3 Energy calculation
 
3  ENERGY CALCULATION ..................................................................................................................... 3 
  Introduction, step by step guide ............................................................................................................. 3 
  Having local wind measurements: .......................................................................................... 3 
  Not having local wind measurements ..................................................................................... 3 
  Coordinate system auto selected for WAsP and PARK calculations ..................................... 4 
  Different concepts: wind statistics or time step calculation .................................................................... 4 
  Overview of the different concepts ......................................................................................... 4 
  Why calculate in the time domain? ......................................................................................... 6 
  How is it calculated in the time domain? ................................................................................ 6 
  Resource map, supports the design of the wind farm ............................................................ 8 
  Basic requirements, data & objects ....................................................................................................... 8 
  Wind measurements ............................................................................................................... 8 
  Wind statistics based energy calculations .............................................................................. 9 
  Mesoscale wind data ............................................................................................................ 18 
  Elevation data ....................................................................................................................... 18 
  Roughness data .................................................................................................................... 18 
  Obstacles – displacement heights (forest handling) ............................................................. 23 
  Site data object: the terrain (and wind) data container ......................................................... 25 
  Turbine data .......................................................................................................................... 25 
  Tools (data/MODEL validation/calibration) .......................................................................................... 26 
  Overview of model and wind data validation tools ............................................................... 26 
  SCALER - downscaling/post scaling/interpolation ............................................................... 26 
  Displacement height calculator ............................................................................................. 36 
  ORA (Optimized Roughness Approach) ............................................................................... 37 
  Rix correction ........................................................................................................................ 39 
  Result layer ........................................................................................................................... 43 
  Performance Check module ................................................................................................. 43 
  T-RIX tool .............................................................................................................................. 43 
  MCP (Measure-Correlate- Predict) - Long term correction module ..................................................... 44 
  MCP ...................................................................................................................................... 46 
  Sessions in MCP .................................................................................................................. 66 
  Reports from MCP ................................................................................................................ 66 
  Models used by MCP ............................................................................................................ 68 
  Models/modules for initial calculations ................................................................................................. 77 
  METEO ................................................................................................................................. 77 
  ATLAS................................................................................................................................... 83 
  WAsP interface ..................................................................................................................... 87 
  WAsP-CFD ........................................................................................................................... 90 
  Resource maps ..................................................................................................................... 91 
  STATGEN ............................................................................................................................. 98 
  Flow request export (FLOWREQUEST - FLOWRES format) ............................................ 100 
  PARK calculation ............................................................................................................................... 102 
  The wake loss (PARK) models ........................................................................................... 103 
  Curtailment in PARK calculations ....................................................................................... 111 
  Common settings for all PARK calculation variants ........................................................... 116 
  Common settings for Wind statistics based (standard) PARK calculation ......................... 120 
  Standard PARK calculation with WAsP .............................................................................. 127 
  Standard PARK calculation with WAsP-CFD ..................................................................... 128 
  Standard PARK calculation with resource file .................................................................... 129 
  Common settings for Time series-based PARK calculation ............................................... 129 
  Time varying calculation based on mesoscale data ........................................................... 138 
  Time varying calculation based on measured data ............................................................ 138 
  Other PARK calculations .................................................................................................... 139 
  Output from PARK calculations .......................................................................................... 139 
  PARK with WakeBlaster, external wake model .................................................................. 147 
  Project Cost and LCOE calculation .................................................................................... 160 
  Loss & Uncertainty ............................................................................................................................. 163 
  Introduction, definitions and step-by-step guide ................................................................. 163 
  Step-by-step guide .............................................................................................................. 168 

Introduction, step by step guide   2 
 
© EMD International A/S   •    www.emd.dk   •   windPRO 4.0  •   April 2024 
  Basic data for calculations .................................................................................................. 168 
  Bias ..................................................................................................................................... 171 
  Loss .................................................................................................................................... 173 
  Uncertainty .......................................................................................................................... 183 
  Results ................................................................................................................................ 188 
  Calculation and print ........................................................................................................... 190 
  Appendix: Wake model tests and validations .................................................................................... 193 
  Test of calculated wake loss on varying wind farm sizes ................................................... 193 
  Single row versus multiple row wind farms......................................................................... 195 
  Horns Rev area, Danish Offshore project ........................................................................... 196 
  Lillgrund, Sweden offshore project ..................................................................................... 207 
  Wake calculation validation for large Egypt wind farm ....................................................... 210 
  Large UK offshore wind farm complex ............................................................................... 211 
  Very large Egypt wind farm complex .................................................................................. 217 
  Conclusions on wake modeling .......................................................................................... 218 
  Validation examples and model problem issues ................................................................................ 220 
  MCP validation .................................................................................................................... 220 
  Mesoscale data long term consistency ............................................................................... 221 
  WAsP versions modifications ............................................................................................. 223 
  Displacement height calculation ......................................................................................... 227 
  Elevation model pitfalls ....................................................................................................... 228 
  Checking the Power Curve ................................................................................................. 230 
  Test of Turbulence scaling.................................................................................................. 232 
  Appendices: From windPRO 2.9 manual, not included in this manual: ............................................. 234 
  MCP2005 – Measure/Correlate/Predict (long term correction) .......................................... 234 
  Table of figures and tables ................................................................................................................. 234 
 
 
 
   

Introduction, step by step guide 3

3 ENERGY CALCULATION

Introduction, step by step guide

Calculating the expected long-term Annual Energy Production (AEP) for a wind turbine or wind farm is one of

the most important tasks when developing a wind project. The poorest wind sites (commercial turbines) have

round 5 m/s in annual average wind speed. The best above 10 m/s. Since the energy content in the wind

increases with power of 3, a 10 m/s wind site will provide 8 times more wind energy than a 5 m/s site. Due to the

design of turbines, going for the best in cost efficiency, there will be limitations in utilizing the highest wind

speeds. Therefore, the typical gain in production will “just” increase with a power of 2. This, however, still gives

a factor of 4 increase in AEP for the 10 m/s relative to a 5 m/s site. And the increase in AEP just going from 6

m/s to 7 m/s is around 35%. Therefore, it is very important to know the wind speeds and thereby predict the AEP

very precisely. The AEP, together with the price earned per MWh, are the two main deciding factors for the value

of the project and if it is feasible. Project costs, maintenance costs, interest rates etc. are other deciding factors,

but these vary typically not so much from site to site (except onshore vs offshore). Thereby, the wind speed and

the MWh tariff mainly decide if the specific site is feasible to utilize.

To do a proper AEP estimate, there are several ways to go. In windPRO, the order of the calculation modules is

lined up like this:

MCP – long term corrects the local wind measurements

PARK – do the AEP calculation based on the long-term corrected local wind data, including wake loss calculation

and optional sector management or other curtailments

LOSS & Uncertainty – evaluate/calculate expected losses and uncertainties

OR, the “time varying energy calculation concept” based on mesoscale model data with time step calculation:

Meteo Analyzer OR Performance Check + PARK – Post calibrate the mesoscale data to match local

measurements or turbine production for the period with data

PARK – do the AEP calculation based on recent 10-20-year mesoscale data (post scaled based on above),

including wake loss calculation and optional curtailments

LOSS & Uncertainty – evaluate/calculate expected losses and uncertainties

Before the calculations can be performed, there are several steps to be taken:

Having local wind measurements:

1. Import the logger data in the windPRO METEO Object, do the data screening, and clean for bad data.

Set up a terrain model for WAsP or WAsP-CFD and do analyses, like how does the measured shear

compare to model calculated shear. And if more masts are on the site, how does the model cross

predict – if poor, what could be the explanations and which possible model or data adjustments can

improve the model setup. For instance, fine tuning roughness data and/or include displacement height

(near forest) could be solutions.

2. Get long-term data, import in METEO object, analyze the quality, and compare different sources. Many

long-term data sources are available for ON-LINE download.

3. Create long-term corrected wind statistics or long-term local time series with the MCP module OR use

the new SCALER concept to calibrate long-term data using the local measurements.

Not having local wind measurements

a. Use mesoscale model data. This requires calibration since the mesoscale model data for most sites is

not accurate enough for direct use. Calibration can be based on regional measurements or turbine

production from operating turbines.

b. OR use one or more wind statistics (WAsP format) e.g., purchased from a meteorological department.

The quality of these are often insufficient and they need to be validated e.g., by re-calculating already

operating turbines near the new site.

c. OR use a wind resource map provided by an external source (WAsP format or .siteres format). Here

again, it must be mentioned that quality assurance is the key.

Different concepts: wind statistics or time step calculation 4

Option A is very popular due to constantly improved mesoscale model data, which is fully integrated in

windPRO’s time varying calculation concept, described in Section 7.

Other aspects of wind farm development to consider

First step will typically be to establish the potential wind farm area in the WTG-area object, including restriction

zones, buffers etc. Later, the SITE COMPLIANCE and LOAD RESPONSE modules can be important to find out

which turbines are strong enough to survive at the site based on the wind conditions and the wind farm layout.

Also, modules like METEO or WAsP interface can be useful for calculation of AEP for many different turbine

variants and hub heights at “one spot” at the site to see which turbine comes out with the best AEP/Cost ratio

for the site. Finally, the RESOURCE module is very helpful to get the spatial variations of the wind resource for

the site and, thereby, a basis for designing a wind farm, eventually with help from the OPTIMIZE module.

Many other constraints can dominate the wind farm layout design. Noise restrictions (DECIBEL module) very

often decide. Visual constraints (PHOTOMONTAGE and ZVI modules), shadow flickering (SHADOW module)

and environmental constraints like bird corridors, just to mention a few others.

Set up a calculation model, which requires terrain data (elevation and roughness). With the terrain model and

a flow model (WAsP, WAsP-CFD or flow result files from other model providers delivered in the generalized

FLOWRES file format), vertical and horizontal extrapolations can be performed so the wind expectations at each

turbine position and height can be calculated. windPRO offers full WAsP and WAsP-CFD integration and several

terrain data sources for ON-LINE download.

Finally, yet importantly, it should be mentioned that windPRO has a large turbine database with the most

important specifications for energy calculations. In addition, windPRO has numerous tools for wind farm design

(eg. the OPTIMIZE module and the PARK design object), data validation, check etc. Specifically, the

PERFORMANCE CHECK module shall be mentioned here, since it allows the user to set up detailed analyses

comparing calculated and measured production for one or more wind farm(s) on a very detailed level. Comparing

the measured and modelled calculated energy (PERFORMANCE CHECK module) on many operating wind

farms, teaches the user much about how well the model calculations perform relative to real life operation. This

gives unique feedback on setting up a calculation model for future projects.

Coordinate system auto selected for WAsP and PARK calculations

From windPRO 3.2 all WAsP and PARK calculations are performed in UTM WGS84 for the zone in which the

Site Center is placed. This solves a problem when using Geographic coordinates (which cannot not be handled

by WAsP), and problems arising from coordinate systems with large rotations relative to Geographical North. In

former versions such systems could give doubtful results due to the rotation. By always using UTM WGS84,

rotations are marginal.

Different concepts: wind statistics or time step calculation

Overview of the different concepts

Historically, the majority of AEP calculations have been calculated based on wind statistics. This is the native

WAsP concept (Wind Atlas Method), where wind data is represented by a 3-dimensional matrix with Weibull

distributions and frequency by height, direction sector and roughness. This concept is considered robust and

fast – and it still is.

Calculations in the time domain do not really differ that much, but it gives many advantages. Getting a temporal

dimension makes it possible to match up production with e.g. electricity prices and detailed curtailments. Another

major reason for calculating in the time domain is that it is now possible to access good quality long term

mesoscale wind data all over the world. But the method is not limited to usage of mesoscale data, as local

measurements also can be used. An added benefit of the time varying concept includes the possibility for

advanced interpolation of data from multiple masts - a feature which is well-sought after in the industry.

Different concepts: wind statistics or time step calculation 5

Table 1 Calculation/correction options: Wind statistics vs time domain

Calculation options

Wind statistics

Time step

calculation

Power curve corrections:

Different noise modes Day-Evening-

Night

YES

Air density

YES, by annual

average

YES, by time

step

Turbulence

YES

Shear and veer

YES

Model corrections:

Displacement height by dir. sector

YES

RIX correction

YES

Curtailment calculations in PARK

Sector management

YES

Other curtailments

YES *6)

Resource map calculation:

YES

Interpolation between more

masts/mesoscale points:

Partly *1)

YES

Wake model options:

Alternative models to N.O. Jensen

(WAsP, PARK1 and 2) and

NO2005

YES

Special output like calculated

turbulence and PPV model

YES

Partly *2)

Advanced wake loss settings, like

change WDC by upwind WTGs

YES

Turbulence controlled WDC

Partly, by direction

sector average

YES, by time

step

Blockage

YES

Hub height dependent TI/WDC by

terrain type selection

YES

Utilization of WAsP CFD

YES

Utilization of other CFD models

Partly *3)

Partly *4)

Utilization of mesoscale data from

EMD

YES

Utilization of mesoscale data from

other providers

Partly *5)

*1) Within a wind resource map calculation, interpolation between multiple wind statistics are possible. Afterwards the

resource map can be used as input for the calculation.

*2) The calculated time series including e.g., calculated turbulence, can be copied to e.g., Excel and any aggregation

can be made. Placing a small turbine at the mast position and including this in the calculation will give the output as

calculated wind speed etc. at the mast position.

*3) Delivering output as resource map (WAsP format) can be used as input for the calculation.

*4) In the 2.9 "simple" variant of time step calculation handles resource map data as input. The new FLOWRES file

format allows all model providers to generate output to be used from windPRO. The format also support turbulence,

inflow angles etc. like it can handle different stability classes.

*5) So far, the advanced downscaling removing the Mesoscale terrain/applying the micro terrain and using the shear

from the mesoscale data, is only an option for EMD mesoscale data. However, the datasets from other providers can

be used as an "artificial mast" or with a simplified downscaling method.

*6) For Park1, Park2 and Eddy Viscosity 1988.

Different concepts: wind statistics or time step calculation 6

Below are some arguments and explanations related to the time step calculation concept, also called “the new

energy calculation concept”. But remember, the old wind statistics based concept is still fully supported and also

expanded with new features.

Why calculate in the time domain?

3.2.2.1 More accurate calculation

The validation of model calculation setup is much more precise. Instead of having only a single average value

to check against, it is possible to do multiple checks by e.g direction, wind speed, time on day and month. Also,

Weibull fit problems are avoided.

Corrections of power curves: Air density, turbulence, shear and veer are time step specific and higher accuracy

can be obtained compared to average corrections. It is similar for wake models. It is known how the wake losses

highly depend on turbulence, which varies in time.

Curtailment losses, like noise, flicker, birds, bats or load based (sector management) can be calculated very

precise having the time series productions. It is possible to calculate based on different Day-Evening-Night noise

reduction modes in one process. PARK can natively include most curtailments, which can dynamically influence

the wake losses.

3.2.2.2 Improve the model setup

More detailed feedback recalculating operating turbines - learn about model problems, get better at setting up

models for new projects. Get close to where the model calculation problems are e.g., by filtering data in high/low

turbulence periods etc.

3.2.2.3 Value of turbine production

With volatile power markets causing electricity prices to fluctuate over time, it is key to know the production

pattern to make sound investment decisions.

Output for time varying tariffs can be calculated. Similar if consumption pattern is known, own consumption

covered from turbines can be calculated.

Grid systems and other production units can be designed better if the time varying production is known, e.g.,

how will the spatial distribution of wind farms affect the wind production variations in time.

3.2.2.4 Utilize Mesoscale wind data

Mesoscale wind data is not always precise - calibration with time domain data makes it possible to correct for

e.g., directional and seasonal bias in the mesoscale model data.

Mesoscale data is complete, with no gaps or frozen equipment like with measurements. All relevant signals are

available, like temperature, pressure, turbulence, shear, veer, humidity, solar radiation etc.

Mesoscale data opens up the ability for advanced corrections, like icing loss and heating requirements.

3.2.2.5 Huge future development potential

Manufactures are working towards more intelligent turbines that e.g., adjust to load conditions, meaning that the

power curves will change by climate (load) conditions. MODEL can adapt these features.

How is it calculated in the time domain?

A transfer function between measurement points and calculation point is established by direction sector. The

transfer function is used on each time stamp wind speed to extrapolate it to the calculation point.

Using Mesoscale model data, the calculation height is interpolated in the mesoscale data, thereby there is NO

vertical model extrapolation, and the mesoscale data shear is used for each time step. The mesoscale model

data is downscaled by an advanced methodology, see 3.4.2.

More mesoscale points or measurement masts can be used with advanced horizontal interpolation on the

geostrophic wind level, which means that the mesoscale data is normalized (mesoscale terrain lifted off), before

interpolation. Then, the interpolated, normalized values get the micro scale terrain applied at the interpolation

point (calculation point).

Different concepts: wind statistics or time step calculation 7

3.2.3.1 The transfer function calculator (the SCALER) features

▪ Pre-procsessing mesoscale data, also called downscaling. Mesoscale terrain is “lifted out” and

microscale terrain is applied. Utilization of traditional WAsP (ONLY ver. 11+), WAsP-CFD or the

FLOWRES (open flow model output format) for the transfer function.

▪ Mast to turbine position scaling can be based on WAsP A-parameter ratio calculation OR the raw WAsP

flow ratios (no stability correction). Since windPRO 3.2 the WAsP A-parameter ratio calculation is

made wind speed dependent and stability correction is thereby handled more correct. This can

in case of large height extrapolations change calculation result from windPRO 3.0/3.1 to 3.2 or

later significant.

▪ RIX correction is optional.

▪ Displacement height calculation can be performed, based on an object’s displacement height OR the

sector wise auto displacement height calculator.

▪ Turbulence can be moved from mast to turbine or calculated if the used model (e.g., CFD) supports this.

▪ Post calibration can be performed with wind speed scaling and/or offset + individual sector wise,

seasonal, diurnal or wind speed scaling factors. The post calibration corrects for e.g., bias in mesoscale

data.

3.2.3.2 The SCALER can be used from:

▪ PARK, time varying calculation

▪ METEO analyzer, for e.g., establishment of a calibrated SCALER for mesoscale data if measurement

mast data is available.

▪ METEO object, “Add scaled time series” for e.g., downscaling mesoscale data within the METEO object

to evaluate how the downscaling changes the raw mesoscale data.

▪ METEO object, Graphic/Profile, for scaling wind data within the METEO object (calculating the wind

profile) for testing different SCALER settings, like displacement height.

o Note: the aggregated calculated profile is directly comparable to measured when aggregation

e.g., by season, while the filtering works on calculated as well as measured. A traditional WAsP

calculated profile will always be based on all data.

3.2.3.3 The time varying concept features

Calculation in the time domain adds following features:

▪ Use the shear in the mesoscale data by time step

▪ Air density correction on time step basis.

▪ Turbulence correct the power curve by time step.

▪ Shear and veer corrected power curves by time step.

▪ Use time step turbulence to adjust wake decay constant in the wake model on a time step basis.

▪ Use different operation modes by time step.

Table 2 Establishment of calibrated long-term data time series

Wind used in calculations:

Local measurements (short term data)

Mesoscale data (long term data)

Tool to use:

Long term correct with:

Calibrate with:

Tools to use:

MCP module

Mesoscale data

Local measurements

Meteo analyzer + Scaler

Regional measurements

Meteo analyzer + Scaler

Regional Turbine wind

Meteo analyzer or Perf.Check module

Regional Turbine production

Performance Check module + Scaler

Added value: (partly for future utilizations, like Icing loss calculation)

More weather parameters:

Shear

Used by time step, optional Power

Curve correction

Veer

Optional for Power Curve correction

Temperature

Optional for Power Curve correction

Pressure

Optional for Power Curve correction

Turbulence

Optional for Power Curve correction

and wake correction

Solar radiations

Humidity

Cloud cover

etc.

Alternative

options

Calibration can include direction and season

scaling to compensate for bias in mesoscale data.

This bias is dependent on where in the world and

the local terrain features, like near shore, forest etc.

This "left side option" seem best if there are

directional turn or nonlinear wind speed ratios, as

these are handled well by MCP (Matrix method).

Basic requirements, data & objects 8

Having both mesoscale long-term data AND local measurements (blue arrow), there are two ways to go:

▪ Use MCP to create a long term local data series as input for the time varying calculation for getting the

expected long-term AEP. This method has the advantage that it can utilize the more refined options in

MCP to e.g., turn the mesoscale wind data to match the local measurements better.

▪ Use Meteo Analyzer to calibrate the mesoscale data with a SCALER to reproduce the local

measurements accurately in the concurrent time period, and then calculate with the full mesoscale data

period and the calibrated SCALER for getting the long-term AEP. This method has the advantage that

the shear from the mesoscale data is used for the entire period, and it makes it possible to make power

curve corrections on time step basis, based on the additional information in the mesoscale data. The

present wind statistics based calculation concepts (WAsP) do not utilize the shear from local

measurements, although it can be used in, for example, METEO object to calibrate the WAsP parameter

settings, like heat flux.

Resource map, supports the design of the wind farm

A wind resource map is a calculation of the wind distributions (Weibull A and k for each direction sector) in a

grid. The area to be calculated must be defined (e.g., by a WTG-area object) and the resolution and hub height(s)

must be specified. A wind resource map (.rsf, .wrg or .siteres format) can be used as input for a PARK calculation

or as input for an OPTIMIZATION.

Wind resource maps can also be calculated as well based on wind statistic(s) or based on the SCALER concept

with calculation in time domain, where the output although won’t include the time dimension. Both methods

include interpolation between more wind datasets (masts or mesoscale data points).

Since 3.5 it is possible to rescale resource maps based on one or more measurement points within the resource

grid area.

windPRO offers free access to the GASP dataset which is a global wind resource map. Documentation can be

found here: GASP Global. The data format is .siteres, but can from result layer menu be exported to .rsf or

.WRG.

Basic requirements, data & objects

A very large amount of data, as well a detailed description of the terrain, and numerous long wind data series,

results in a huge data amount. windPRO is specially designed and developed over decades to handle larger and

larger data amounts in a structured way and in a way where the user can keep track of the data, check them for

errors and correct them. Here, the different data types and objects (data containers) for energy calculations are

described. Note the more detailed description of some objects mentioned given in the BASIS chapter.

The fundamentals in an AEP calculation are:

1. Wind measurements/data (can be pre-processed data as wind statistics or model wind data).

2. Model extrapolations of the wind to each turbine position at hub height

a. Vertical

b. Horizontal

c. In time (Long term correction)

3. Turbine data (Hub height, Power curve and Ct curve for wake loss calculation)

4. Losses

a. Wake (depends on PARK layout, wind and terrain, and requires turbine positions)

b. Electrical (can be calculated by eGRID module)

c. Availability

d. Other: like shut down due to environmental issues (curtailment) etc.

More “details”, like temperature data for air density correction etc. are also relevant.

In the following, the fundamental data structures for the needed input are reviewed, then followed by a chapter

describing the models and validation tools.

Wind measurements

Wind measurements are the typical basis for calculations of expected energy production from wind turbines. Of

high importance is partly seasonal variations, partly annual variations. Therefore, measurements should normally

Basic requirements, data & objects 9

be performed for at least one year to capture not only seasonal variations in wind speeds and directions but also

like variations during the seasons. The annual variations should be handled by long-term correction based on

long-term data for at least 10 years, preferably 20 years. Measured wind data and optional additional data like

temperature and pressure can be imported in the METEO object where data can be screened and cleaned for

errors or other problems like icing events etc.

Measuring at multiple heights makes it possible to evaluate how the shear is handled by the model and if model

adjustments are needed. Measuring at the expected hub height for the turbines to be installed takes out an

important uncertainty component - the vertical extrapolation. Measuring with more masts (or other measuring

devices, like Lidar), reduces the horizontal extrapolation uncertainty and gives, in general, an important feedback

on how well the model handles this. The Cross Predictor in Meteo Analyzer is a tool for checking both the vertical

and horizontal extrapolation accuracy of the model. More measurement points can also be handled in windPRO,

where the new time step-based calculation concept with SCALER allows for interpolations between more

measurement points. This is an expanded feature not included in the native WAsP software.

Wind statistics based energy calculations

Historically, the majority of AEP calculations have been calculated based on wind statistics. Even though the

new time-varying energy calculations has many benefits, the wind statistics based energy calculation is still

supported by windPRO.

A wind statistics is a structured collection of wind measurements processed with terrain and holds, thereby, the

information of the wind speed distributions as a function of roughness, direction and height.

Figure 1 Wind Atlas Method

Basic requirements, data & objects 10

The left graphic illustrates the wind atlas method: “cleaning” the wind measurements of local terrain effects and

establishment of a generalized wind climate, the wind statistics. This is then used with the local terrain at the

calculation point to establish the expected “terrain corrected” wind climate. To the right, different wind

distributions (frequency versus wind speed) are shown partly as measured (grey) and as weibull fitted

representations (black line)[ Source: European wind atlas, Risø/DTU].

A wind statistics is a file with a multidimensional matrix with the following dimensions:

• Height (typical: 10, 25, 50, 100, 200 m)

• Wind direction (typical 12 sectors of 30 degrees)

• Roughness (typical 5 classes; 0, 1, 2, 3, 4, corresponding to lengths; 0.0002, 0.03, 0.1, 0.4 and 1.5)

For each dimension, a Weibull distribution describes the wind distribution. This is represented by an A and k

parameter. For each roughness, a frequency by direction sector gives how often the wind comes from this

direction.

A wind statistics can be generated via WAsP from windPRO modules STATGEN or MCP. It is based on a Weibull

distribution, a histogram (table data) or a time series in METEO object with wind speed and direction for each

time stamp (ONLY 1 height) AND a terrain description (elevation, roughness, local obstacles and, eventually,

displacement height). The WAsP model, popularly spoken, lifts off the terrain influence on the wind and creates

“neutral” wind distributions. When generating a wind statistic from MCP, the local measurements are long term

extrapolated based on a long-term time series (see more in MCP section).

The native format for wind statistics in WAsP is called a Library file (*.LIB). However, since WAsP 11, it is called

“New Generalized Wind Climate (*.gwc) file format replaces the Wind Atlas (*.lib) file format. Vertical wind profile model

and parameter settings are now part of the Generalized Wind Climate file; A number of model parameters, which in

WAsP 10 and earlier versions were saved with the WAsP project, are now saved with the Generalized Wind Climate

file.” (from www.wasp.dk )

In windPRO, the file extension for a wind statistics is .wws. This file contains a number of additional parameters,

like coordinates, information of the time series used, the WAsP parameters, WAsP version etc. The number of

parameters has expanded over time by different windPRO versions. From an energy calculation, a report page

“wind statistics information” is available, showing which data and parameters are included in the .wws file. Note

windPRO supports both .lib and .gwc files while .lib files can be generated directly and both types can be used

in calculations from windPRO.

An important part of a wind statistics is the shear: how does the wind speed change with height. This is partly

based on the roughness, but also on an advanced stability model, which is part of the WAsP model. The stability

model can be invoked by changing the heat flux parameters in WAsP. This will typical be needed in very warm

(dessert) regions or very cold climates. With measurements at more heights, the METEO object can be used to

calibrate the stability settings.

3.3.2.1 View/evaluate/edit a wind statistic

It is possible to view wind statistics with the Wind statistic viewer which can be found in the Climate tab. Below

an example:

Basic requirements, data & objects 11

Figure 2 Wind statistic view.

The window shows some basic info as well as some evaluation figures. These, based on Danish average site,

are:

Energy – the “raw” energy as a percentage of 3300 kWh/m

WTG energy – the power curve filtered energy relative to 1025 kWh/m

, where the calculation is based on a

WTG with specific power: 0.45 kW/m

The energy levels arbitrary chosen to correspond to an average Danish site with round 6.5 m/s at 50m a.g.l.

For both variants, a hub height of 50m and roughness class 1 is used for the key value calculation.

The energy key values are very useful for checking if the wind statistic seem to be “reasonable”. Comparing with

other wind statistics in the region, only smaller differences should be seen, otherwise it is most likely that the

data quality is too poor. It can although be related to mesoscale wind differences in e.g., mountainous or coastal

regions.

In addition to the energy key values, it is possible to show direction graphs and maps with wind statistics, see

example below:

Figure 3 View of wind statistics on map and as rose graphs.

As seen above, multiple wind statistics are selected in the list, and to compare e.g., the directional energy level.

The selected are marked yellow on the map.

Edit metadata button give access to modify different information’s, like coordinates.

Basic requirements, data & objects 12

Figure 4 Edit wind statistic metadata

It also gives access to print report with all relevant information’s on the wind statistic.

The “View additional info” shows the following:

Figure 5 View additional wind statistic info.

Modify energy level is a special feature allowing for manipulation of the data within the wind statistic, in case it

is known that it provides too high or too low calculation results, e.g., if used on an operating project with good

long-term expectation knowledge.

Basic requirements, data & objects 13

Figure 6 Modify energy level for wind statistic.

For a given:

▪ Power curve

▪ Height

▪ Roughness

The energy correction factor can be entered.

For each sector, an optimizer finds the A and k parameter adjustment (scaling factor) needed for adjusting the

calculated energy production that respects the desired energy correction factor. These scaling factors are then

used on ALL A and k parameters within the wind statistic.

From windPRO 3.6 a measure has been introduced to stop the optimizer from finding an increased A-parameter,

when the energy is scaled down. If the power curve ramps down at higher wind speeds, this ramp down is

removed in the power curve used in scaling. In practice, the energy correction optimizer uses the highest power

value for all wind speeds up to 100 m/s. This avoids the A-parameter being scaled in the opposite direction of

the energy correction. In addition, there is a limit to how much the k value is allowed to be modified, so this

always will stay within the range 1.0 to 4.0. However, if the k in the windstatistic to be scaled is outside this

interval, then the limit is expanded to this value.

For a site with terrain features (different speed ups by direction at different locations etc.) the calculation result

may not change exactly as expected. If the goal is to match a certain value, then one must try out different

correction factors experimentally.

With the modification of the correction factor, the wind statistic would serve as a better input for the calculation

of new wind turbines in the near region, compared to using a wind statistic that has a known bias.

An illustration of the energy scaling compared to wind speed scaling, and why it is important to not just scale

energy, is shown below.

Basic requirements, data & objects 14

Figure 7 Illustration of energy vs wind speed scaling.

With energy scaling, the correction on AEP in any calculation will be as shown with the “fixpoint” line, here

approximately 22%. With wind speed scaling, the AEP correction will be the same for the fixpoint, which was

basis for the creation of the wind speed scaling, here 25m hub, roughness length 0.03m (Rcl. 1) and a Bonus

150 kW turbine. With wind speed scaling as example for a 100 m hub height, same roughness length 0.03m

(Rcl. 1) and the same power curve, the AEP correction will now be approximately 15%.

The methodology is used in the Danish DK’07 wind statistic, which is based on regional correction curves created

back in 1992 mainly based on 25 m hub height turbines. This wind statistic is still much used for calculations in

Denmark 25+ year later. The methodology will although not work well in complex terrain regions, where the wind

can change radically just within few km.

Extrapolate roughness button gives access to add extra roughness in the wind statistic. This is needed for later

WAsP versions, that do not allow for roughness extrapolations. Especially older wind statistics will often have an

upper roughness length of 0.4 (class 3.0), which then not will work on sites with roughness values above this.

Figure 8 Extrapolate roughness.

Here a new upper roughness value of 1.6 (class 4) is added, and the extrapolate button creates the new data.

Entering “Edit meta data” gives access to print a wind statistic report, in which the added data can be seen.

BeldringeDenmark1972-80 Scaling factor (A):

1,1

100

150

200

250

0% 10% 20% 30% 40% 50%

AEP increase: Bonus 150 kW [MWh/y]

Height a.g.l. (m)

0,03

0,1

0,4

Fixpoint:

Basic requirements, data & objects 15

Figure 9 Results with extrapolated roughness.

3.3.2.2 Global Wind Atlas wind statistics, download and use.

At the web page: https://globalwindatlas.info/ there is access to wind statistics worldwide based on World Bank

sponsored project.

Figure 10 Global Wind Atlas web page.

Zooming into a specific location and placing a marker, gives access to downloading a .LIB file (GWC file):

Figure 11 Place a marker at the specific location for the wanted GWC file.

After placing the marker, a download button appears in the right pane:

Basic requirements, data & objects 16

Then select GWC:

Figure 12 Download the .GWC file for selected location.

Save the file like here:

Figure 13 Save GWC file in folder below "WindPRO Data\Windstatistics" or in project folder.

If your search path setting for wind statistics includes this path, the file is ready to use in windPRO and can be

found by the wind statistic browser:

Figure 14 The GWC file is seen in the wind statistic browser.

To see the GWC file requires the file location is included in the search path setup. The naming is taken from the

file, and will always be the same, but also coordinate information is included, making it possible to distinguish

between more downloaded files.

Figure 15 Comparing GWC file (left) with EU-WindAtlas file (right).

By “Edit meta data” in the wind statistic browser, there are acces to printing a wind statistic report. Above the

mean wind speeds from the report. As seen, the GWC file has essential higher wind speeds than the “old” EU-

Wind Atlas data for the same location.

This would also be seen in the browser, where the WTG energy level, a very useful check, is seen. In this case

the GWC file located at same spot as the “Danmark” wind statistic (Beldringe from EU-WindAtlas) has an WTG

energy level ~15% higher than the “Danmark” files, which has proven to calculate quite precise. This mean that

at this location, the GWC file probably will calculate round 15% too high energy, so the results will not be more

precise than this. Below a small test is performed with the GWC data.

Basic requirements, data & objects 17

Figure 16 The locations of the 13 WTGs in test calculation.

Note the forest in main wind direction for WTG 11-13, this is not handled in the calculation setup, where

displacement height or better the ORA tool should be used.

A calculation based on standard DK roughness maps and elevation is set up with default wake model.

The results of the calculations are below compared to actual wind index corrected production based on 15+

operation years, where months with larger availability problems are filtered.

Figure 17 Testing GWA (left) versus EU-Wind atlas wind statistics (Beldringe), rightmost.

The GWA data overpredict by 17-26%, the EU-WindAtlas data overpredict by 6-14%. Worth to note that “normal”

availability losses and possible noise reduction modes (not known here) justifies an overprediction of 5-10%.

The most overpredicted WTGs 11-13 has heavy forest in main wind direction, SW, which not is handled in the

calculation setup. Conclusion is that at this location, the EU-WindAtlas works quite well, with a few percent

overprediction if all corrections were made, whereas the GWA data overpredicts by at least 10%. The reason is

probably a known bias from mesoscale modeling where for Denmark it is in the order of 3-5% on wind speed,

Basic requirements, data & objects 18

which converts to 10-15% on energy. This bias is seen much worse, in e.g., southern Germany, up to 20% on

wind speed is seen. Offshore in North sea, the bias is close to zero.

Mesoscale wind data

Mesoscale wind data are time series output from a mesoscale model, like WRF. The concept is that data from

global climate simulation models (ie, at a 100 km resolution) is used as the input together with terrain (elevation

and roughness) to create a grid node resolution of ie. 3 km. The mesoscale model thereby adds the regional

terrain effects not included in the global climate simulation models.

Mesoscale wind data can be downloaded from a database with pre-run datasets for selected areas using

windPRO by establishing a METEO object. It is also possible to run the WRF mesoscale model from windPRO

for any place in the world (see further details in Chapter 4). A special feature added in the windPRO mesoscale

data is the used mesoscale terrain, which is stored in the METEO object. This is used when downscaling the

mesoscale data (lifting off the mesoscale terrain and applying the micro terrain). But even when downscaling the

mesoscale wind data, there will be a need for calibration. Mesoscale data does not have a fully accurate wind

speed level all over the world, and the bias will be different depending on where the point is. Therefore, a post

scaling/calibration is needed. In windPRO, there are comprehensive tools for performing this post calibration.

VERY IMPORTANT: The use of downscaling near a coastline is problematic. The reason is partly the

coarse resolution of the mesoscale model runs (including coarse roughness data resolution) and partly

the different stability conditions between land and water. Lifting off mesoscale terrain (downscaling)

from two data points near the coastline - one onshore, another offshore - can, thereby, give large

differences in wind speeds when calculating e.g., right at the coastline. And which one tells the truth? It

is so far recommended NOT to use mesoscale data within +/- 3 km from a coastline. Instead, take the

next point further offshore or further inland, depending on where the calculation shall be performed (on-

or offshore).

Elevation data

Elevation data describes the terrain elevation. In windPRO, elevation can be handled as grid data or as

contour data with conversion options between the two formats. There are more free datasets available for

download that can be directly accessed via windPRO ON-LINE data services, making it easy to establish the

elevation data. Also, comprehensive tools are available within the Line object for manually digitizing the elevation

data based on contour lines on background maps. (See more details in Chapter 2 Basis)

For a wind farm site, elevation should be established for a distance of around 7 km in all directions from the edge

of the site (can also depend on turbine height). For WAsP-CFD calculations, the data should cover a 20 km

radius. The needed resolution of the elevation data depends on the terrain. For WAsP calculations, 5 m

resolution contour lines are normally considered sufficient.

A special issue to consider is when using surface data (top of forest etc.), like SRTM data. These data can create

large errors, especially if turbines are placed inside or near forests and the calculation model, thereby, places

the turbines on the top of the forest instead of on the ground. The only way to proceed in this case is to digitize

the wind farm area manually based on background maps with the correct terrain elevation. Visit the knowledge

base for online data to read more about the available datasets.

Roughness data

The surface roughness is very important for describing how the wind profile is dragged. It describes the friction

created by the landscape elements like houses, trees, and the surface of the ground. There are more free

datasets available for download. These can be directly accessed via windPRO ON-LINE data services, making

it easy to establish the roughness data. Also, comprehensive tools are available within the Area Object for

manually digitizing the roughness data based on polygons on background maps. The Area Object exports the

polygons to Roughness Lines, which is the format required by the WAsP model.

The line object can import roughness data from these file types, where the .tif is new from windPRO 3.4:

Basic requirements, data & objects 19

Figure 18 Import file formats for roughness from line object.

The Area object can import polygons from as well as gridded data from

Using older WAsP versions (< 10.2), or the simple ATLAS model, roughness can be handled as roughness

roses, which can be established within site data objects (see windPRO 2.9 manual for details). But todays

standard is to use roughness line maps. One reason for this is that newer WAsP versions (from 10.2) use the

roughness maps for calculation of the stability corrections on a more refined way than previous WAsP versions.

Therefore, the use of roughness roses is no longer supported.

3.3.5.1 Using Online data for roughness line maps

Roughness lines can be downloaded direct from EMD’s On-line service. It is only recommended for initial

calculations or, for example, the calculation of resource maps for larger regions for finding areas of interest.

When a specific site shall be calculated, the fine-tuning of the roughness data is very important. This is very

difficult based on roughness lines, and, thereby, it is recommended to perform this within the Area Object.

The roughness data sources are in a rapid development, more new datasets appear every year, with different

resolution and quality. Below is described a procedure for manual edition as ONE example based on Data4Wind

roughness data. Other datasets might not require this process, but it is still good to know the features available

for editing.

▪ Import the data into an area object.

▪ Delete the “open farmland” areas. While these normally cover the most, and they will be treated as

background roughness. In non-farmland regions, like deserts, heavy forested regions, or water, it can

be more convenient to let these dominant area types be associated with the background roughness.

▪ In a dataset like “Data4Wind”, the open farmland is the class range of 1.8-1.9 .

▪ Set the background roughness to the dominating roughness of “open farmland” (or whichever class is

dominate), which can be from class 1 up to class 3 if there are many wind breaks (see figure later). If

the farmland areas varies much in surface roughness (seen, e.g., by looking at the areas on google

maps, preferably combined with a site visit), manual digitization of parts of the open farmland areas

might be required to establish a reasonably accurate roughness map.

▪ Take the roughness areas given as > class 3.2, and merge these into the class 3.2 layer. It can be

discussed if there are areas with higher class than 3.2, but typically not for a standard farmland site

(although, mountains could be). Forest areas are recommended not to be set higher than class 3.2

but, instead, use the displacement height calculator for compensating for the effect of the forest (see

chapter related to this).

▪ Finally, export the processed roughness areas to lines. A line object with the data can automatically be

created.

3.3.5.2 General information about roughness

For a wind farm site, roughness should be defined for a distance of around 20 km in all directions from the edge

of the site (can also depend on turbine height). The needed resolution of the roughness data depends on the

terrain. It is important that surface roughness shall be seen as a property of a larger portion of terrain. E.g., a

farmland with many wind breaks should not be digitized wind break by wind break, but as the entire farmland

region where the distance between and heights of the wind breaks decides the roughness (see figure later in

this paragraph).

Roughness can be handled by either roughness classes or roughness lengths (see next table).

Table 3 Roughness definitions

Basic requirements, data & objects 20

Roughness

class

Roughness

length

Relative

energy %

Description

0,0002

100

Water areas.

0,5

0,0024

Mixed water and land area or very smooth land.

0,03

Open farmland with no crossing hedges and with scattered buildings.

Only smooth hills.

1,5

0,055

Farmland with some buildings and crossing hedges of 8 m height and

about 1250 m apart.

0,10

Farmland with some buildings and crossing hedges of 8 m and 800 m

apart.

2,5

0,20

Farmland with closed appearance and dense vegetation – crossing

hedges of 8 m height and 250 m apart.

0,40

Villages, small towns, very closed farmland with many or heigh hedges,

forrest, many abrupt orographic changes, etc

3,5

0,80

Large town, cities with extended build-up areas.

1,6

Large cities with build-up areas and high buildings.

Figure 19 Production variation by roughness and specific power

The deciding factors for AEP in flat terrain are the roughness, the specific power and hub height. Above is

illustrated how the first two mentioned invoke AEP for Danish wind conditions- more than a factor of 2.

40%

50%

60%

70%

80%

90%

100%

110%

250 300 350 400 450 500

Specific power (W/m^2)

Production by roughness and specific power for

~100m hub height

Class 0 Class 1 Class 2 Class 3

Basic requirements, data & objects 21

Figure 20 AEP change vs distance to coastline and hub height

Illustrated in the figure above is the very large influence the roughness has on the AEP, even at high hub heights.

A 100m high turbine loses 20% in production, when moving 15 km from shore inland – even with a relative low

inland roughness of class 2. An only 15m high turbine loses 50%, so larger hub heights do compensate a fair

for roughness influence but are still heavily influenced themselves.

In areas with hedges (windbreakers), the graphs in the following figure, based on formulas from “Danish wind

atlas”, Risø/DTU 1979, can be used to estimate the roughness class or length. Notice the non-linear impact of

the hedge height on the roughness class. Normal farmland is assumed to lie between the hedges. This is

incorporated into the figure by adding 0.03 m to all specified roughness lengths. A porosity of 0.33 is assumed.

Basic requirements, data & objects 22

Figure 21 Roughness in farmland with windbreaks

In an area with many buildings, the roughness length can be estimated by using the following equation from

“Danish wind atlas” Risø 1979:

= 0.5 * h

* b * n / A

Where:

h = height of building

b = width of building

n = number of buildings

A = total area within which the n buildings are situated

NOTE: The roughness length of the area between the buildings must be added to the roughness length, which

has been determined based on the above equation, e.g., add 0.03 m to the calculated roughness length for

normal farmland.

The relation between roughness classes and roughness lengths is shown in the next figure.

Basic requirements, data & objects 23

Figure 22 Roughness class vs length

By applying roughness sector wise (roughness roses), it may be necessary to decide on one roughness value

when, that sector consists of more than one roughness class. It is then recommended to make a simple weighting

of the roughness classes in that sector. For example, if the area consists of 1/2 Roughness Class 2, 1/4

Roughness Class 1 and 1/4 Roughness Class 3, the resulting Roughness Class becomes: (2*2 + 1*1 + 1*3)/4

= 2. It is, again, important to note that the use of Roughness Roses is a legacy practice/option that is only viable

with WAsP versions before 10.2.

It’s important that a roughness classification covers an entire roughness area (belt), i.e. a roughness belt with a

width 1000 m with one crossing hedge of 10 m heights should be assessed to the roughness class 2. It’s often

seen that such an area has been classified as roughness class 1 all the way to the hedge, with a shift to

roughness class 3 for a few meters along the width of the hedge, and with a final roughness class of 1 just after

the hedge. This is incorrect! The “European Wind Atlas” recommends doubling the width of every new roughness

belt when moving outwards from the WTG.

Another important rule: Even if an area is located at a ground level, which is lower than the turbine site, the

roughness classification is not affected by this fact. The differences in terrain heights are included in the elevation

model.

In practice, it’s important to visit the site and take preliminary notes regarding the roughness and the distances

between the roughness changes. Furthermore, notes should be taken regarding local obstacles and their

dimensions. Having completed the site visit, the exact distances between the roughness changes and the final

design of the roughness classifications can be determined at your desk by using the map and the above-

mentioned tools. However, much of the measuring work can be avoided by using digital background maps in

windPRO. An efficient option is the Google synchronization view, where the objects are shown on top of Google

Earth maps.

Obstacles – displacement heights (forest handling)

Single obstacles, e.g., buildings, hedges etc. (of more than ¼ of the hub height) near the WTG (within

approximately 1000 m from the turbines) should be included as local obstacles OR incorporated by

displacement height.

A displacement height calculator was introduced in windPRO 3.0, because experiences showed the obstacle

model does not handle the combination of forest and larger turbines sufficiently. It is, therefore, recommended

to use the displacement height calculator when having forest areas near (within approx. 1 km) the turbines when

they have a 40 m hub height or higher. For smaller turbines (<40 m), the obstacle model seems to handle the

Basic requirements, data & objects 24

reductions sufficiently, although there are problems when the obstacles are very close to the turbines (within the

shaded region shown in the figure below).

If the obstacle is lower than ¼ of the hub height or farther than 1000 m away, it should not be included in the

calculation as an obstacle as it will have little or no influence on the calculation result. However, it must be

included as a roughness element in the roughness classification. In addition to the dimensions of the obstacle,

its porosity must also be estimated. The following figure shows how the WAsP model handles obstacles.

Figure 23 Obstacle model

The displacement height calculator works like this:

Figure 24 Displacement height calculation

With the displacement height calculator, a height is subtracted from the hub height in each direction sector based

on the height of the forest and the distance to the forest. Various input data to describe the forest can be used

(see more detailed documentation later).

Basic requirements, data & objects 25

Site data object: the terrain (and wind) data container

The site data object is a multi-purpose object for definition of how the terrain shall be handled. It can, e.g., merge

more elevation or roughness data files, or be used to point out a specific combination of files, which makes

experimental calculations more structured. When the purpose is set for WAsP calculation or resource map

generation, a wind statistic can also be selected within this object. Thereby, when setting up, e.g., a PARK

calculation, all the input data is defined just by a link to a site data object.

Depending of the use, the object will change color on the map. The different colors are:

• Black – for Atlas

• Blue – for WAsP *)

• Orange – for STATGEN *)

• Green – for RESGEN *)

• Purple – For WAsP-CFD

*) Can be used for terrain input for SCALER, Site Compliance etc.

Turbine data

New WTG

Existing WTG

Wind turbines can be established as new or existing. Both hold links to the turbine catalogue with information on

power curve, Ct curve, noise data, visualization data etc. User defined parameters such as coordinates, hub

heights etc. can be input as well. New from 3.2 is handling of paired Noise and power curves. (More details are

in Chapter 2 of the Basis module).

Existing WTGs can be marked “Treat as PARK WTGs”:

The feature basically identifies if the existing WTG is part of the project, or just a reference turbine that might

give wake and noise additions, but is not part of the project that is calculated. The PARK calculation will then

calculate the production with and without new WTGs. Grid curtailment calculations ignores non-“Treat as Park”

WTGs.

Start-stop dates can be set on turbine objects. This is a very convenient option in time varying calculations, as

one long time series can be calculated with turbines with different (de)commissioning dates, which is respected

in the PARK calculation.

For existing turbine objects, information on actual production can be included, partly as the long-term expected

production (statistic tab) and/or as time series (see more details in chapter 11, Operation – Performance Check).

Since 3.1, the noise reduced power curves can be utilized based on Day-Evening-Night settings and thereby a

PARK calculation can handle different noise modes by time of day.

From 3.2 there is an option to choose “Shut down” in power curve choice, if e.g., Night mode is to stop the

turbine. It can also be used for taking out a turbine of a re-calculation, if this e.g., is an experiment, then it is not

needed to open the PARK calculation to deselect the turbine.

3.3.8.1 Handling of power curve cut-out

If a power curve is set to cut out at a specific wind speed, production is included up to the cut-out wind speed.

Decimal cut-outs (e.g., 22.8 or 22.9 m/s) are even handled when calculating in e.g., 1 m/s bins. Here, the end

will be calculated a little higher in the 23 m/s bin covering production from 22.5-23.5 m/s. For time step

calculations it is simply the time step wind speed which decides if it is below cut-out.

Tools (data/MODEL validation/calibration) 26

Tools (data/MODEL validation/calibration)

Models have uncertainties; data have uncertainties – and even errors.

One of the most important disciplines within energy calculations is to justify that both the data and the models

used are reasonably correct. This means that the calculated wind distribution at each turbine location is as close

as possible to what the project will show. There are plenty of possible mistakes. The more comprehensive the

evaluation of the data and the models, the higher the safety that no major mistakes are made. The

validation/calibration tools therefore, always have had a high focus in windPRO.

Overview of model and wind data validation tools

The data tools are described in separate paragraphs in this chapter, with METEO object and Meteo analyzer as

exception. These are described in Chapter 12. Also, the Performance Check module is described in separate

Chapter 11. Here an overview is given.

Table 4 Model and data validation tools

Extrapolation of the wind data, e.g., coming from a measurement mast, will, in most cases, be needed, as

measurements will not be available at each planned turbine position, are often not in hub height, and not long

enough to be sure data is long-term representative. This is where the models come in, like the WAsP flow model

for horizontal and vertical extrapolation, and MCP models for long-term extrapolation.

SCALER - downscaling/post scaling/interpolation

A SCALER is used for “moving” wind (and other climate parameters) from one location

(mast) to another (turbine) in the time domain. It is a very comprehensive interpolation and book keeping tool

Model and wind data validation tools in windPRO

Meteo object:

Compare more measurement heights, disable (flag) erroneous data, merge "free directions"(remove mast shadow if more

anemometers in same height), calculate shear 24-12 matrix and based on this create synthesized data in other height and many

more options. (See Chapter 12).

Graphic tab, profile:

Validates the shear by showing measured shear together with model-calculated shear from ONE height. Can be shown by

direction sector.

The model calculated shear can be based on:

A) WAsP

OR B) WAsP-CFD for aggregated data (wind statistics based). Thereby, annual average calculated shear is shown which by season

or day/night aggregation of measurements not give a real comparison.

OR C) SCALER. Here it’s possible to do real comparisons by, e.g., season or day/night. Although the model "behind" (WAsP) does

not yet have the option to run with different day/night heat flux settings, then only when using Meteo object with EMD mesoscale

data, the day/night shear differences be compared between measured and calculated.

Meteo analyzer:

General: Compare more masts/mesoscale data on time series level or aggregated with concurrency optional. Disable functionality

available - written back to Meteo objects. (See Chapter 12).

Cross predictor:

Validates the horizontal (and vertical) model extrapolations between more masts and heights based on concurrent data

SCALER:

Creates a SCALER based calculation (e.g., from mesoscale data or other mast) in an existing mast (Meteo object) for comparison to

measured data – then optional: Change SCALER setup (Post calibration) and recalculate to get a match.

MCP:

Creates long-term corrected wind data based on transfer model set up from concurrent local measurements and long-term data.

The module holds comprehensive tools for comparing the modelled and measured data in the concurrent data period and gives

quality indicators.

Performance Check:

The most comprehensive tool (module) for validating the “full way” from wind data to energy output from the wind turbines

(when in operation). Checking performance turbine by turbine, direction by direction, month by month, power curve by power

curve etc. can give the full understanding of where data or models fail or need tunings. (See Chapter 11).

Tools (data/MODEL validation/calibration) 27

utilizing different flow models, that makes it capable to use multiple masts with multiple heights, even with

direction dependent displacement heights.

Besides being able to transfer wind data at each time step sample, the SCALER also includes functionality to

transfer data from one terrain to another. This unique feature of the SCALER means that it can be used for

downscaling meso-scale data. So, in other words, the SCALER can take in to account the terrain used in

mesoscale models and transform wind to the micro scale terrain, illustrated below:

Figure 25 Mesoscale terrain used for Standardization of Mesoscale wind data.

Tools (data/MODEL validation/calibration) 28

Figure 26 From Standardized wind to micro scale wind.

As a windPRO user, think of the SCALER as a component that can “scale” wind from one position to another

using results from a flow model (orography, roughness, obstacles, stability, etc.). So the SCALER “moves” wind

from source data to turbine positions on a site for each time step sample, but, on top of that, it is also able to

calibrate each sample according to user defined post-scaling parameters. This is especially important when

down-scaling meso-scale data, because such data will often include a bias that has to be calibrated against real

measurements from masts or existing turbines. Similarly, e.g., nacelle wind measurements often need a post

calibration.

A SCALER incorporates other important features, such as RIX correction where the complexity of the terrain can

be used to compensate the WAsP IBZ linear flow model problems. Another important feature is that the SCALER

is able to scale the results from the flow-model based on displacement heights for each sector. This feature is

relevant in forest areas where sector-wise displacement heights can be used to simulate the forest “lifting” the

wind profile. For windPRO 3.0, it is important to note that sector-wise displacement heights are only used at the

calculation points and not at the mast positions (points where data are scaled from). For meso-scale data

downscaling calculations, this is not important since forests will not affect the meso-scale data like it will for real

measurement masts.

Starting from windPRO 3.1, displacement heights at mast positions is included in the SCALER, and masts near

forests, will be handled with this feature (unlike in 3.0), which might call for recalculations in specific projects.

Tools (data/MODEL validation/calibration) 29

Figure 27 SCALER setup

The SCALER setup can be reached from the relevant parts in the software, like a PARK calculation, METEO

object, Meteo Analyzer etc.

The tabs to be explained are:

• Terrain scaling

• Rix setup

• Displacement height

• Turbulence

• Post calibration

All 5 correction types are optional. The first 4 are all of the category “terrain scaling” and can be unchecked at

the [ ] Terrain scaling checkbox. The values behind are kept but not used when unchecked. The Post calibration

can, likewise, be unchecked. Rix correction and displacement height can individually be checked/unchecked

within the definition tab.

3.4.2.1 Terrain scaling

Terrain scaling is divided into these main types:

• Meso-scale Data Downscaling

• Measured data scaling

• The “user defined” allows for any mix and to change predefined choices

Meso-scale Data Downscaling assumes the mesoscale model terrain data is part of the METEO object that

holds the mesoscale wind data. Therefore, the “standard” downscaling method only can be used with EMD

mesoscale data in the present version. For the more specific downscaling predefined choices, press “user

defined”.

Tools (data/MODEL validation/calibration) 30

Figure 28 Micro terrain in SCALER

When using EMD mesoscale data as a wind data source, the mesoscale terrain is included in the METEO object

and automatically used as the terrain for the source data.

The only necessary user input is, therefore, the Micro terrain. The options are:

• Site data object – use WAsP 11+ for terrain scaling

• WAsP-CFD results – use WAsP-CFD result file(s) for terrain scaling.

• FLOWRES files – use external e.g., CFD models that support the .flowres file format

• Resource files (.rsf, .WRG, .siteres) since in 3.6

Using WAsP, the number of direction sectors can be specified (default 12, max. 36).

Measured data Scaling is a “simple” variant of the above, where only the Micro terrain is used. Just as specified

above, where instead of downscaled mesoscale data, the data comes from a mast (or an “artificial mast” coming

from other Mesoscale data providers). Here are two variants:

WAsP Stability /A-parameter: The speed up’s are calculated as Weibull A parameter ratios between mast

position(s) and the calculation points.

Important: From windPRO 3.2 the ratios are calculated for more wind speeds and the ratios becomes wind speed

dependent. This can change calculation results

Neutral stability / Raw flow: The speed up’s are calculated based on the raw WAsP speed up output (roughness,

orography and obstacles), including the turns of the wind directions, but NOT including stability correction, while

this is not a part of the WAsP speed up output. This method is therefore ONLY recommended if measurements

are close to hub height, and it is also checked by the software that the height differences are less than 20m. But

due to the more correct handling of the speed up’s and turns, this method is recommended if the stability

correction not is needed.

User defined terrain scaling explains partly the terrain scaling principles and partly gives access to

experimental settings.

Figure 29 terrain scaling in SCALER

The SCALER includes different methods intended for different user cases.

•A: Geostrophic wind up/down: This method uses the geostrophic wind laws to generalize each sample to

geostrophic wind from the generalization terrain and then applies micro-scale terrain effects using flow

perturbations from the micro-scale terrain. The method does not include modelled stability corrections in vertical

scaling, but relies on interpolation in the input heights to do the vertical interpolation.

Tools (data/MODEL validation/calibration) 31

•B: Simple (no mesoscale terrain effects): This method does not take in to account the terrain effects from

the generalization terrain (normally meso-terrain), but only corrects the source data by orography effects in the

micro-terrain. The eventual roughness differences in mesoscale and micro terrain is, thereby, not considered.

This assumption will be unproblematic offshore. It is similar in mountain regions, where the dominating effect is

the orography. But in terrain where the varying roughness dominates the terrain effects, this method will not be

a good choice.

•C: WAsP A-Parameter Scaling: This method scales wind from one location to another by the ratio between

the A-parameters from a WAsP calculation at the two locations. This method accepts more source height for

interpolation if relevant (if there are measurements both below and above calculation height) but uses the WAsP

model to do the vertical extrapolation if calculation height is outside measurement height range. From 3.2 the

WAsP A-parameter ratio calculation is made wind speed dependent and stability correction is thereby

handled more correct. This can in case of large height extrapolations change calculation result from

windPRO 3.0/3.1 to 3.2 significant.

As seen, the SCALER includes different methods intended for different user-cases, but, essentially, they all have

the same purpose of “moving” wind data from one location to another. As a windPRO user, it is normally not

needed to decide since method A is chosen for EMD Mesoscale data, by default, and method C for measured

data. Only for experimental purposes or if using mesoscale data from other sources, the choice of method can

be relevant. A special case could be where mesoscale data is used, but the user wants to control the stability

correction by WAsP heat flux parameters instead of taking the shear from the mesoscale data. In this case,

method C must be used together with the mesoscale data from one height.

[ ] Remove roughness in mesoscale terrain is an option when selecting method A: in user defined mode,

explained here:

The first step in downscaling is to remove the terrain effects from the mesoscale terrain. In order to do that we

use WAsP to get the speedups from the meso-terrain. WAsP will return an orography speedup and a

roughness speedup. So the first step is to divide the wind speed with the orography speedup (always done)

and if the “Remove roughness in mesoscale terrain” is enabled it will then divide by the roughness speedup as

well. So why is this option disabled in the default downscaler? Because tests have shown that the way WAsP

sees roughness (as changes) is not how the meso-scale model sees it. During downscaling the effects of

roughness is still taken into account since the equations going to and from geostrophic wind includes a z0

(meso-scale roughness or reference roughness) which comes from the meso-terrain when going up and from

the micro terrain when going down.

Standardize using mesoscale terrain in scaler

As mentioned above the unique feature of the SCALER is that it is able to differentiate between the terrain in

which the source data is located and the terrain at the calculation points (WTGs/masts). Therefore, it is important

for a SCALER to know where the source data is located.

A SCALER accepts, in general, two types of terrain for generalization, namely:

• Meso-terrain: This terrain type represents the terrain used in the model, the source data

originates from. So, for meso-scale data the terrain would be the terrain used in the meso-

scale model. In windPRO, the meso-scale terrain is associated to the METEO object holding

the meso-scale data. For the SCALER setup, you do not point directly to a meso-scale terrain,

but simply indicate that the SCALER should use the terrain from the METEO object(s) used.

• Micro-terrain: If the SCALER is setup to use the micro-scale terrain as the generalization

terrain, it simply means that the source data is also located in the micro-scale terrain. This is

true for, e.g., masts, LIDARS, or modelled data that are downscaled to micro terrain (artificial

masts).

Micro terrain in scaler

The micro terrain in a SCALER can be defined in following ways:

• Site Data object when using the classical WAsP IBZ linear flow model. The SCALER only

uses the terrain from the Site Data object so any wind statistics in the Site Data object will

not be used in the SCALER calculations.

Tools (data/MODEL validation/calibration) 32

• WAsP CFD result files that have been calculated for relevant area(s). When using WAsP

CFD result files, all flow perturbations used in the SCALER come from the CFD results

instead of the WAsP IBZ model.

• FLOWRES file, from external model calculations.

3.4.2.2 Rix setup

Figure 30 RIX correction in SCALER

The RIX correction is explained in Section 3.4.2.2. Within the SCALER, the RIX values are calculated by WAsP

11+ default settings:

3.4.2.3 Displacement height

Figure 31 Displacement height in SCALER

The displacement height calculation is explained in Section 4.9. Note that the displacement height calculation

using calculator is NOT used on the mast position in windPRO 3.0, but will be used from windPRO 3.1. Updating

calculation using this feature thereby will result in changes in results if mast is near a forest and the “calculator”

used.

3.4.2.4 Turbulence in scaler

Figure 32 Turbulence scaling.

The ambient turbulence can come from the meteo data series (if loaded). If so, this measured turbulence can

be moved to each turbine position by two different propagation methods:

Tools (data/MODEL validation/calibration) 33

Based on wind speed-up

The St.dev. is assumed constant at all turbine positions. Then the mean wind speed calculated at each turbine

position defines the turbulence by formula: TI = St.Dev(u)/u, where u is the wind speed at the given time step,

and the St.Dev(u) is found at the measure point by same formula.

This method is also used if the turbulence is taken direct from a meteo object (not from scaler). The difference

is that when using the feature within the scaler, interpolation from more meteo objects is supported along with

the features described below.

Based on modelled turbulence

If modelled turbulence exists for the site, e.g., from WAsP CFD result file or by other model by use of FLOWRES

files (can come from WENG or other models, e.g., external CFD models supporting the FLOWRES file format),

the turbulence is propagated from the measurement point to the turbine position by the model turbulence ratio.

Use turbulence from model

The model calculated turbulence is used at each calculation point found by interpolation in the model data. This

require the used model include turbulence calculation, like WAsP-CFD or FLOWRES files from external model

with turbulence calculation.

When using the feature WDC controlled by turbulence in time step calculation, each turbine will get individual

WDC for each time step. This WDC is then used to decide the wake expansion for each turbine. Similar when

using turbulence corrected power curve.

3.4.2.5 Post calibration in scaler

Figure 33 Post calibration in SCALER

Post calibration is a very important part of the SCALER. Using Mesoscale wind data for a calculation cannot be

assumed accurate. The post calibration is required to unbias the mesoscale wind data. For measurements, post

calibration can be used for correction of tower shadow or, in the case of nacelle measurements, for correction

Tools (data/MODEL validation/calibration) 34

of calibration problems. Calibration of measurements should typically only be performed if production data is

available to calibrate against using the Performance Check module. It must, although, be mentioned that a

calibration, in this case, could be biased due to a wrong power curve or a wrong wake model calculation (see

details in the performance check chapter).

Having Mesoscale data, the Post calibration can also be based on local wind measurements (Meteo Analyzer)

as well as turbine production (Performance check).

The proposed calibration process is:

1. Adjust the main scale (see Section 4.6.2)

2. Include, if needed, a main offset (and readjust the main scale)

3. Calibrate by direction sector

4. Calibrate by month

5. Redo calibration by sector

6. If needed, include diurnal correction

Ad. 1 – The main scale is adjusted to get the data within a reasonable range. Up to 25% adjustment is needed

at some locations. This is simply a problem in the mesoscale modelling. Offshore regions seem, in general, very

precisely modelled, but, the further inland, the higher the over prediction by the mesoscale models. It is important

to get the data within a reasonable range before starting the more detailed calibration. Comparing the scaled

mesoscale data against measurements, or calculated turbine production based on the mesoscale data, against

real turbine performance often shows that low wind speeds are too low and high wind speeds too high in the

mesoscale data. This bias can be corrected by an offset.

Ad. 2 - Offset corrects for skewness in mesoscale data. For most examples tested, an offset of around +1 m/s

has given the best match. This will then require the main scale to be decreased.

Ad. 3 - Having a reasonable match between scaled mesoscale data and measurements (or turbine production),

the fine tuning can be established. It is described in detail in the Meteo Analyzer section (Section 4.5) and in

the Performance Check chapter on how to perform calibrations by variables like direction and month (**). The

direction calibration is important since a direction bias can result in wrong long-term estimates, even when the

average between measured and scaled match for a specific period (e.g., one year, if this year has a non-typical

direction distribution).

Ad. 4 - Calibration by month can be used to take out seasonal bias coming from the terrain (roughness). The

surface roughness changes by season, in some places a lot (especially where tall crops are grown, like maize)

It can also occur in regions where e.g., the winter season has snow. The mesoscale model does, partly,

compensate for this, but the micro terrain is fixed by present calculation methods. The difference in how the

mesoscale model compensates and the fact the micro terrain does not change in time in the model can create

a seasonal bias when comparing model data to measurements. This can be unbiased by the Post calibration.

Ad. 5 - By calibrating seasonally, the direction calibration will be invoked, as some directions are more frequent

than others are in specific seasons. While the direction calibration is assumed very important, it is recommended

to remake the direction calibration after a seasonal calibration. It is important to notice here that the new found

direction calibration factors must be MULTIPLIED with the previously found values before putting the calibration

factors into the SCALER. It is a multiple of the used calibrations and the new deviations that gives the final

calibration.

Ad. 6 - A diurnal calibration can be performed as the last step. This is especially important if the diurnal variation

is used for reasons like PPA negotiations, having tariffs that vary by hour or for dimensioning of storage systems,

etc.

Calibration by speed is tricky and should be handled carefully. An example of where this can give unforeseen

effects is if the same factor is multiplied on,e.g., 8 m/s, no matter if the 8 m/s is in 30m height or in 80m height.

Thereby, the shear will be modified, which might not be a desired effect of the calibration. But in a situation like

dealing with nacelle measurements, a scaling by speed could be the method that corrects best for the needs of

a non-linear calibration of the measurements.

3.4.2.6 Selection of wind data - Interpolation

With the SCALER it is possible to use data from multiple masts, assuming a gradual change of the wind speeds

between the masts. This is utilized directly by the PARK calculation, instead of having to calculate a wind

Tools (data/MODEL validation/calibration) 35

resource map up front (as in previous versions before 2016). Similarly, when using mesoscale wind data, multiple

mesoscale data points can be used and interpolated.

In PARK, the data selection for data to be used with a SCALER looks like this:

Figure 34 Wind data selection in SCALER

A list with the available METEO objects is seen. Each “mast” can be expanded to see the different heights and

signals. The following rules apply to the “use in scaling” column:

Using mesoscale data, there must be at least a height above and below the calculation height, where

calculation height could be displaced. (Mesoscale data wind profile/shear is used, no vertical model

extrapolation).

Using measurements, multiple heights from each mast can be selected for the calculation and interpolation is

done. If calculation height is outside the mast height interval, the WAsP model does the vertical extrapolation.

Shear heights column does not affect the calculation of the wind speeds for the energy calculation. It is used

for calculating shear for other purposes, like power curve correction based on shear or curtailment based on

shear. When selecting shear heights top mounted anemometers should not be used with side mounted, while

this would distort the shear calculation in the mast direction. If there are very few mast shaded data or these are

corrected, it will although be ok.

For as well mesoscale data as measurement data, MORE horizontal data points (mesoscale points or masts)

can be used. Horizontal interpolation will be performed. Either by taking the nearest (if one mast is, e.g., in the

valley and one on the ridge, this would be best as it adheres to the similarity principle) or by distance weighting

at geostrophic wind level. The latter one means that wind data has the terrain influence “lifted off” before the

Tools (data/MODEL validation/calibration) 36

interpolation, then the micro terrain is applied at the calculation spot to the interpolated value. The distance

weight is based on the inverse squared distance.

When more masts are selected, there MUST be concurrent data with wind speed and wind direction. Up to 50%

deviation of the time step (e.g., 5 minutes if having 10 min. data) is allowed in the concurrency check, so the

clock does not need to be completely synchronized at the different masts.

Displacement height calculator

The Displacement Height Calculator can be found inside the SCALER and as a separate tab in the PARK and

RESOURCE calculations. It can also be accessed from the Definitions tab:

Forests near turbines is a known issue. Experience shows that handling the forests with a displacement height

works reasonably well. A tool that calculates sector wise displacement heights is incorporated since windPRO

3.0.

Theoretically, the displacement height is an offset in the wind profile lifting the perceived ground level off the

ground and treating the forest top as the ground. The offset effectively cuts the bottom part off the turbine and/or

met mast towers, thus reducing their heights in the calculations. An illustration of the effect can be found in the

METEO object in a situation where WAsP may be unable to predict the observed wind profile above forests.

However, if the mast is reduced in height with something close to tree height, the profile may fall into place.

Figure 35 Displacement height input data

The input form for the displacement height calculator is shown above. The ”input forest as” section is a setup

form on how to input the forest information.

The very simple approach is to use the roughness map for defining the forest areas. The input choices are:

• Use the roughness map as used in the energy calculation (selected in site data object), OR

• Point out a specific line object with roughness lines, OR

Tools (data/MODEL validation/calibration) 37

• Point out an Area object with roughness areas

In each case, it must be decided which roughness class (or length) that defines to the program what is a forest

area and what the typical forest height is.

The more refined approach makes it possible to work with varying forest heights:

• Area object with digitized height elements (the height property must be used in the area object)

• Line object with tree heights

• Elevation grid with tree heights

NOTE: For all 3 options, it MUST be the height of the forest, NOT the elevation + height (some data sources

provides the surface elevation. These cannot directly be used but must first be pre-processed, subtracting the

terrain elevation).

It is possible to tune the displacement height calculation with the following:

• Factor on tree height – general experience seem to shows that using the actual forest height works best,

meaning this factor should be kept to 1.0, although for some test cases a factor of 1.5 is seen to work best.

• Distance factor to end of displacement up wind – e.g., having a forest height of 15m and a factor of 50

means that forest in a distance up to 15 x 50 = 750 m in up wind direction from the turbine will create a

displacement height, linearly decreasing from 15 just at the forest edge to 0 at 750 m distance.

• Distance factor to end of displacement down wind – same as above, but where the wind hits the turbine

before it hits the forest.

Considering the two last ones, the result returned is conservative, taking the largest displacement height between

the two.

The calculation is based on 3-degree “beams”. For each beam a displacement height is calculated, and the

average from all beams in a sector is taken. Thereby a forest that only covers part of the sector will give a

reduced displacement height. If the beam hits more forest parts (at different distances with different heights), the

largest displacement height is used.

Several scientific papers regarding this issue are available, and several test cases have been set up by EMD

showing that the concept works well. It is, of course, not a very precise calculation, since issues like density and

size of the forest, etc. not are handled, but is still a better evolved process than was available previously.

See also the validation chapter; 8.4 with tests of the Displacement Height Calculator.

ORA (Optimized Roughness Approach)

Forests in energy calculations is complicated. Therefore several approaches are seen. One of the more

promising, intensively tested is the Optimised Roughness Approach (ORA).

This approach consists in defining the roughness value of forested areas according to the height and type of

trees. Higher trees will give higher roughness value, and, for the same height, coniferous forests will have a

higher roughness than deciduous forests.

The ORA model is implemented in windPRO as an ”add on” tool to the already implemented displacement

calculation tool in windPRO:

Figure 36 Displacement height calculator, part of ORA

The displacement height calculator based on different input sources (line, area or elevation grid objects)

calculates the displacement height at any position in the terrain in any direction sector. The base concept is that

the distance to the forest and forest height decide the displacement height by a linear decrease by distance.

Inside the forest, the displacement height is given by the forest height with a multiplier. New in windPRO 3.2 is

that the factor can be selected based on forest type:

Tools (data/MODEL validation/calibration) 38

• Coniferous: 0.66

• Deciduous: 0.7

These factors will be proposed modified in the displacement height calculator to these defaults from the ORA

tool. Thereby the combined increased roughness values and factors on tree heights in displacement height

calculations are aligned to match the experimentally found best combinations based on several sites. For specific

sites other combinations might work better, but this requires quite good data, e.g., more measurement masts at

site, to be able to find best working combinations.

The ORA tool is started from the Definitions tab:

The ORA tool allows to:

1) Calculate the roughness value of forest areas according to the tree height and type

2) Combine these new forest roughness areas to the original roughness map so that it can be used for energy

calculation.

Figure 37 ORA setup

The ORA tool setup.

Besides choosing/defining a displacement height calculator setup, the ORA tool creates a roughness map based

on the shown settings by pressing the “Create ORA line object”. The line object will be created in the active layer.

The conversion from forest heights to roughness values can be edited manually. The defaults values depend on

the selected forest type Coniferous/Deciduous. These values are experimentally found by Peter Enevoldsen,

(PHd, Aarhus University), based on several test cases. The values of roughness are found together with the

previous mentioned factors on forest height. Therefore, windPRO suggest changing these factors in the

displacement height calculation setup, if the forest type is changed.

An important input is the selection of the original roughness source (Non forest roughness), where forest are

included based on “traditional” assignment of roughness to forest, e.g., coming from On-line roughness datasets

as Globecover. This is the starting point for the ORA modified roughness map, where the forest data overrule

the roughness in the forested areas based on tree heights. The basic concept is that the roughness data are

Tools (data/MODEL validation/calibration) 39

converted to a roughness grid, where the grid cell values with forest are replaced based on the ORA values.

Then the grid is converted back to a line map. The resolution for the temporary grid must be given as input:

If the resolution is set to a too low value (high resolution) in combination with the size of the roughness map, a

warning is shown, and the creation of roughness map is not possible. If the resolution value is too high to capture

the forest areas reasonably well (seen by visual inspection of the created line map), the workaround is to limit

the size of the initial roughness map, eventually dividing it into more map files and running the process more

times. In this case, it shall although be payed attention to the borders between the line maps. The best will be to

combine these into one line object (by “add” function) and thereby inspect and correct inconsistencies along

borders. Inconsistent roughness maps do violate WAsP calculations and must be taken very seriously, while

WAsP do not produce any error message, it just creates unrealistic results near the inconsistent lines.

After creation of ORA roughness map, the method to use it is described in the yellow box in the ORA setup form.

The validation results have been presented at the Resource Assessment Workshop in Edinburgh 2017 (A

Uniform Approach for Wind Simulations in Forested Areas: Examining the Importance of Data Inputs

(Enevoldsen, 2017)), where it was demonstrated that 3 different industrial CFD forest simulations had an average

error in wind speed cross prediction of ~6.5% where the ORA/WAsP model error were as low as 2%.

The Optimized Roughness Approach was developed by Peter Enevoldsen. Further info can be found in

Managing the Risks of Wind Farms in Forested Areas: Design Principles for Northern Europe (Enevoldsen,

2017)

Rix correction

From windPRO 3.0, it is possible to include RIX correction in the AEP calculation. Previous versions had the

possibility to perform a RIX calculation, but no correction.

Rix correction is a separate tab in the setup of the PARK and RESOURCE calculations based on wind statistics.

Using SCALER, the setup is performed within the SCALER setup. Here RIX correction is treated as a model

feature, meaning the WAsP model handles the RIX correction based on the WAsP parameter settings.

RIX is the Ruggedness Index, defined as the percentage of the area around an object that has a steepness

above a given threshold value. At 30% steepness, flow separation typically starts, which means that the WAsP

model assumptions are no longer valid. Experiments show that the RIX values can be used to fix WAsP model

calculation problems in steep terrain – or at least reduce the error introduced by flow separation. It is even seen

in research papers that RIX correction can also improve CFD calculation results. But The RIX correction is,

however, not currently offered as a correction option using WAsP-CFD results for a calculation.

The Latest research by DTU/Risø/ shows that the threshold in a RIX calculation typically works best with 40%

steepness (new default), and that, with a delta Rix within +/- 5%, corrections should not be performed. Cross

predictions based on more masts can fine tune the threshold (see Cross predictor tool in windPRO Meteo

Analyser). In the windPRO LOSS & UNCERTAINTY module, RIX correction can be calculated automatically as

a bias based on the most recent recommended correction formulas, which can be found in EWEC2006 & 08

papers on Rix from DTU/Risø, see extract below:

Tools (data/MODEL validation/calibration) 40

Figure 38 RIX correction EWEC 2006 results

The main conclusion, based on the use of the RIX method, is that if both reference site (measurement mast)

and predicted site (WTG) are equally rugged (Delta RIX < 5%), very small calculation errors are expected.

If the reference site (measurement mast) is very rugged (e.g., RIX = 0.2) and the predicted site (WTG) is less

rugged (e.g., RIX = 0), Delta RIX will be -0.2 and, according to the graph, a 30% too low wind speed prediction

could be expected at WTG site. This could lead to around 60%

under-prediction of the calculated energy

production.

If the reference site is less rugged (e.g., RIX = 0) and the predicted site (WTG) is very rugged (e.g., RIX = 0.2),

Delta RIX will be +0.2, and, according to the graph, a 30% too high wind speed prediction could be expected at

WTG site. This could lead to around 60%

over-prediction of the calculated energy production.*

Doubling of the energy prediction error based on the mean wind speed error is a rough conversion, which holds

for wind speeds around 8 m/s. At 6-7 m/s, tripling the value is more accurate and only 1.5 factors should be used

for 9 m/s (see graph above based on a typical WTG).

Tools (data/MODEL validation/calibration) 41

Figure 39 Change in AEP% per change in wind speed %

Source: EWEC06 paper:

Figure 40 Input data for RIX correction in statistical based calculations.

The form above shows the RIX setup when the traditional wind statistic based PARK calculation is selected.

IMPROVING WAsP PREDICTIONS IN (TOO) COMPLEX TERRAIN

Niels G. Mortensen1, Anthony J. Bowen2 and Ioannis Antoniou1

1Wind Energy Department, Risø National Laboratory

P.O. Box 49, VEA-118, 4000 Roskilde, Denmark

T (+45) 46 77 50 27, F (+45) 46 77 59 70

E-mail niels.g.mortensen@risoe.dk

2Mechanical Engineering Department

University of Canterbury

Christchurch, New Zealand

Tools (data/MODEL validation/calibration) 42

The Steepness threshold is, in some papers, recommended to start at 30%, while others say at 40%. It depends

on the site characteristics. In the Meteo Analyser, it is possible to do experiments, if more masts are available,

to judge which method works best. Here, it also can be judged if the RIX correction should be weighted by

direction (Frequency) or omnidirectional (Equally distributed).

Figure 41 RIX correction input in SCALER

Within the SCALER, a simpler approach is used since the SCALER uses the internal WAsP RIX calculation

where the settings are defined within the WAsP parameter file. This does not allow for frequency weighting, but

steepness threshold and radius can be set.

In the Result to file output, the RIX correction is included. Note the sector wise result to file gives the sector

corrections, where the standard result to file just gives the aggregated results.

The RIX report page in the PARK calculation, statistical based, is shown below.

Figure 42 Report output with RIX correction

In the report from a wind statistic based calculation, full RIX correction documentation is available. But note, this

is the total. If sector wise correction is used, the result shown is the sum of the sector corrections and, thereby,

there can be a RIX correction even if the average delta Rix is within +/-5% and this interval is set to no correction.

This is due to the fact that some sectors have a delta Rix outside the +/-5% intervals.

For the Scaler based RIX correction report, only the RIX values for each WTG is shown, while the correction is

an integrated part of the model calculation.

Tools (data/MODEL validation/calibration) 43

Result layer

The result layer offers a presentation of the spatial distribution of calculated wind data, production etc. It even

allows for presentation of differences by the Compare feature, where, e.g., a CFD calculation can be compared

to a WAsP calculation by ratio or difference or a formula. This gives unique options for checking different models

against each other – or the same model with different data assumptions. The use of the result layer is described

as part of the BASIS chapter.

Performance Check module

The performance check module is an important module for calibration of the SCALER. Having turbine production

monthly, hourly or 10 min. based, gives unique possibilities to calibrate the wind data so the calculated

production, time step by time step, matches the measured production (filtered for downtime), and thereby

establishing a wind data time series reflecting the “real wind” at the site. Also, having turbine production data for

more or many turbines at a site makes it possible to calibrate/test wake model settings, roughness, forest model,

Rix correction, power curve corrections etc.

The unique thing about using turbine production for model and data calibration is that there, typically, are many

more calibration points than having measurement masts, and the transformation from wind to power can be

included in the calibration.

The Performance Check module is described in a separate chapter (11), where the interaction with the PARK

calculation also is described (11.9).

T-RIX tool

From windPRO 4.0 it is possible to run a terrain complexity check according to the German Technishe Rechtlinie

6 (TR6) revision 12. The tool can be found in the Tools tab:

T-RIX is the representativity measure for the Technische Richtlinie 6, a German guideline, that standardizes the

parameter settings for the RIX calculations. For the RIX calculation it provides parameter settings for the radius,

critical slope, number of radii (wind sectors), the spatial resolutions (horizontal and vertical) for the digital

elevation model and the contour line interval. Since RIX values are affected by the chosen parameter values,

the standardized approach allows for a better comparison and correction of RIX values. It also guides with the

choice of parameters. In addition to the standard RIX calculation, T-RIX considers the height distance between

the wind data measurement position and the wind turbine in addition to planar distances.

MCP (Measure-Correlate- Predict) - Long term correction module 44

The method provides two horizontal distance specifications A and B between the wind data measurement

position and the wind turbine in kilometers; A specifies the recommended distance (should-criterion) and B

specifies the distance that must not be exceeded (must-criterion). Correspondingly, it also specifies a

recommended (should-criterion) and maximum vertical distance (must-criterion). Moreover, uncertainty

quantification is defined for AEP-predictions for horizontal and vertical distances between the should and the

must criteria.

It is possible to calculate the complexity check using New WTGs and Existing WTGs and/or Meteo objects.

The T-RIX module produces a main report and a detailed report. The main report lists the input data and

parameters, and it can be checked whether the input data and parameters are coherent with the TR6

guidelines. The detailed report shows the T-RIX values per WTG, distances A and B, and lastly the quantified

uncertainties obtained using the T-RIX specifications. Finally, the complete report can be exported to a text file

or saved to clipboard via Result to File.

MCP (Measure-Correlate- Predict) - Long term correction module

MCP is an advanced module that enable users to calculate long-term corrected wind data in windPRO.

The MCP module can be found in the top Ribbon under Climate tab:

MCP (Measure-Correlate- Predict) - Long term correction module   45 
 
© EMD International A/S   •    www.emd.dk   •   windPRO 4.0  •   April 2024 
 
Figure 43 Start MCP calculation. 
 
Starting with windPRO 3.2 a fully new MCP module is available. This will coexist with the former MCP module, 
now renamed to MCP2005, from its release year 2005. 
The manual focuses on the MCP module, the previous version being described in Appendix: MCP2005. 
 
In general wind measurements are undertaken for a limited period of time, which may represent a different wind 
speed level compared to the long term expectations. If these short term measurements are further used in an 
AEP calculation, the energy estimates may result in errors or may not be regarded as long term representative.  
The purpose of the MCP module is to provide the tools and methods necessary to calculate the potential long-
term wind climate at the site. 
 
The MCP module consists of the following methods: 
•  Simple Mean Wind Speed Scaling 
•  Linear Regression 
•  Matrix 
•  Neural Network 
•  Solver based (for Scaler) 
•  Scaling Local 
 
The documentation for the different MCP methods can be found in the MCP Reference Document as well as in 
section: 3.5.4 Models used by MCP. 
 
In addition, several features are available for conducting an evaluation: of the long term datasets to be used, the 
adequate long term period, and adjustments to be applied on the datasets involved in the MCP process. Some 
of these tools are briefly introduced below, while a comprehensive description is available Chapter 7 of  MCP 
Facilities in windPRO.   
•  Evaluate reference: Alternative long term datasets can be compared to a selected long term reference. 
These datasets can be downloaded within MCP for a 20 years’ time frame. In the download process the 
nearest point to the selected reference is considered. The alternative datasets can be downloaded from 
several sources, such as: the EMD’s comprehensive database ; via METEO object, by using users own 
sources; or from Performance Check database, where e.g., turbine based indexes can be imported.  
The “Compare to other references (LT Bias)” function gives fast and precise answers on how much 
would the alternative datasets predict higher/lower than the selected reference, where the local data 
measurement period is used as part of the evaluation. 
•  Auto filter time offset: In the Adjustments tab long term datasets can be adjusted for time offset based 
on best correlation between local and reference time series wind speeds. This tool can be part of the 
selection process of a long term dataset. 
•  Auto Veer: A feature in the Adjustments tab that automatically calculates the veer between the local 
data and the reference time series. The user can decide to apply this veer correction on any of the two 
datasets, local or reference. 
•  Interpolate reference:  A feature in the  Adjustments  tab. A typical reference has 1-hour resolution, 
while the local measurements typical has 10-min. By resampling the reference to 10-min (by simple 
interpolation)  the  amount  of  data  for  training  the  model  increases  essentially,  and  better  prediction 
results can be obtained. 
•  Concept Choice: This tab provides the option to choose between: 
o  a classic long-term correction, where the output is a corrected long-term time series or a wind 
statistic. 
o  a scaling approach where the output is a short term dataset (local measurements) scaled to be 
long term representative, but keeping the dynamics and direction distribution of local data. 
•  Slicing: In the Model LT tab a more comprehensive evaluation of the methods performance can be 
done by giving the user the option to split the dataset into several concurrent sets. These sets will be 
used for training and testing the models/ methods. Sic different alternative period slicing are available. 

MCP (Measure-Correlate- Predict) - Long term correction module 46

• Uncertainty: As part of the Concept Choice tab, an MCP uncertainty method is available. The

uncertainty calculation is based on the Klintø model (Klintø, April 2015), and considers four main

parameters as uncertainty contributors: the wind index, the correlation (Pearson factor), the inter-annual

variability and the number of concurrent years.

• Session concept: In the Main tab of the MCP calculation the user has the option to create sessions

independent of each other, which allow to combine various long term datasets with the local

measurements and various methods. This window allows the user to structure the work and decide on

a specific combination of reference data and method to be used for the final evaluation.

MCP

3.5.1.1 Measure

On the Measure tab the user will select the input data to be used in the long term correction process.

Figure 44 In the Measure tab the input data is loaded and can be analysed.

In the drop-down menu of Meteo object and height, the user will select the local measurement and the long

term timeseries to be used in the analysis. As a minimum requirement, the two loaded datasets shall be

concurrent for the analysis to happen. In general, one full year of concurrency is recommended between the two

datasets.

The local measurements should cover a full year of data while the long term dataset vary in length from 10 to 30

years.

The top table shows details of the two loaded datasets, such as time period used, the time resolution, the mean

wind speed for the entire dataset and the mean wind speed during concurrent time between the two datasets.

The time series graph shows the two datasets, by default, as monthly averages.

The colour code seen in the graphs represent the colour coded datasets defined through the text color in the top

table.

An overview of the correlation parameters for the concurrent period between the two datasets is presented

underneath the time series graph. The correlation figures are calculated on the averaged datasets, as indicated

in the averaging cell underneath the time series graph.

MCP (Measure-Correlate- Predict) - Long term correction module 47

Figure 45 Correlation of the two datasets averaged to monthly level.

Note that the correlation will typically be poorer, the higher time resolution. Still, a more realistic correlation is

achieved on raw data, the scatter and dynamics of the timeseries being preserved.

The wind energy correlation is calculated using the power curve defined under Session Setup tab. If no power

curve is defined a standard power curve based on u^2 truncated at 13 m/s is used.

On the right side are shown the wind rose and diurnal profile between the two datasets. These can optionally be

set to show e.g., only concurrent data or all data.

The rose view shows parameters such as: Wind direction frequency, Mean wind speed or the Wind energy

distribution.

The number of sectors can be adjusted in steps up to 36 sectors or to a maximum of 360 sectors. By default, 12

sectors are shown.

Figure 46 Wind direction frequency distribution on concurrent data between local measurements (blue)

and concurrent long term reference (red).

Figure 46 gives a good overview of the concurrency between the two datasets during concurrent period. It is

also a visual indication for possible directional bias. Any bias can be corrected in the Adjustments tab.

Reference ST/LT shows the rose between the long term reference dataset and the short term reference dataset

(concurrent with local measurements).

MCP (Measure-Correlate- Predict) - Long term correction module 48

Figure 47 Frequency distribution between the long term dataset (pink) and the long term dataset during

concurrent time with local data (red).

Figure above indicates if the wind direction observed during the measurement year is the same as the long term

observed wind direction. The data from the same time series is compared during different time periods, the

measurement period, and long term one. If the distributions match it can be concluded that the wind direction

during measurement period is long term representative. This analysis indicates whether the wind direction should

be corrected and is relevant in the selection of the long term method.

Figure 48 Wind energy distribution between local measurement and long term reference data (top) and

the diurnal wind speed profile between the local measurements and long term reference data (bottom).

Showing the short term vs. long term wind energy distribution can tell if the period with local measurements is

long term representative. In the diurnal graph example, the short term mean wind speed (blue) shows a stronger

variation day/night compared to the long term data.

These comparisons are relevant in the process of selecting a long term correction method.

MCP (Measure-Correlate- Predict) - Long term correction module 49

Figure 49 Scaled reference data available for creating reference data from more model points.

In the Long term reference dropdown menu the user has the option to scale a reference dataset, by selecting

Scaled reference data.

Scaling the reference data is an option, meaning that more reference datasets can be merged within MCP. This

can as well be used for downscaling more mesoscale points to the local mast position, as to average more model

data points. Or, if having more long-term reference masts near the site these can be scaled to local mast position

and averaged by distance weight. Note: If several reanalysis datapoints selected at some distance to the site,

are to be used as distance averaged points by the Scaler, then the scaler should be setup without terrain scaling.

To preform terrain scaling a terrain model should be available in the dataset, and this is available only for the

mesoscale data.

MCP (Measure-Correlate- Predict) - Long term correction module 50

By downscaling the mesoscale dataset to micro terrain level, it is expected to get closer to the level of the local

mast data as both are represented by the micro terrain effects. Thereby there will be better conditions for the

MCP models to give a more accurate prediction. More details on the Scaler are available in section: 3.4.2

SCALER - downscaling/post scaling/interpolation.

Data statistics button shows statistical information, sectors wise, for each of the two datasets, local

measurements and long-term reference. On the first tab, Table data, the sector wise correlation is presented

together with information on the sector wise mean and standard deviation of the wind speed, the ratio of wind

speeds and the wind veer for the selected datasets.

Figure 50 Data statistics window.

In the Graphics tab all concurrent data are extracted with optional outlier filtering. By default, samples with wind

speeds below 4 m/s and wind direction differences above 90 degrees are excluded both in the local and

reference dataset.

Sectors are defined by the local direction measurements.

The table is based on concurrent time stamps for the ratios between short and long term, and on all data for the

averages of individual datasets.

MCP (Measure-Correlate- Predict) - Long term correction module 51

Figure 51 Data statistic graph

The Graph tab provides a visual representation of the mean wind speed relation between short and long term

datasets.

Limit periods button allows the user to limit the period used in the analysis on any of the two datasets selected.

Figure 52 Limit period for local and/or reference data.

MCP (Measure-Correlate- Predict) - Long term correction module 52

Limit period tool is typically used in cases where one of the datasets needs to be limited to a shorter period or

for sensitivity analysis situations.

For instance, if the reference dataset seem less trustworthy in the older part of the data, then it might be a better

choice just to use the recent 10 years than a longer period of time. In the Evaluate reference paragraph is

explained how to evaluate such a choice.

Another example could be related to testing using one year of local measurements of the full two years period.

The option “Use disabled data” is convenient, if disabling’s performed in the Meteo object were done with other

purposes than removing bad data.

Evaluate Reference is a comprehensive part of the MCP process, which can be opened from the Evaluate

reference button in the Measure tab.

Figure 53 Evaluate reference times series window.

In the Reference series table, the following series are available by default:

Reference, user power curve – session setup. A wind index calculated with the entire reference data series

having 100% index in average. Is based on the index calculation setup chosen in Session setup tab (see 3.5.1.2

Session setup). Note that this series is visible only after a user power curve is selected in the Session setup

window.

Reference, generic power curve – Simple index. As above, but based on the default session setup, with a

power curve based on U^2 truncated at 13 m/s. This is the method for the index’s downloaded with the only

difference that the downloaded always has 1993-2012 as reference period (100% index).

MCP (Measure-Correlate- Predict) - Long term correction module 53

Figure 54 Add on-line wind energy index -downloading alternative reference wind energy indices.

Add On-line wind energy index button, gives the option to download different predefined indices. Depending

on the location of your project on the globe, different indices are available, such as EMD Global Wind Data

(based on ERA Interim), ERA5 or Merra-2.

A number of mesoscale datasets are available in regions where EMD has pre-run the WRF mesoscale model

for a predefined domain.

More information on the pre-run mesoscale datasets and reanalysis data is available here.

NOTE: All indexes have a 1993-2012 (20years) timeline, as long term reference period, giving 100% in average

for this period. The timeline was selected in order to have a fixed reference, evaluated by EMD to be long term

representative for Northern Europe. It is based on around 40 years’ experience with turbine operation data and

also validated against very long-term sources like NAO index. See documentation on www.vindstat.dk. However,

this period might not be the best reference for other places on earth. For other places in the world, issues like

poor data quality or insufficient sources for historical data, could limit the time line to the recent 10 or 15 years

as long-term reference.

It is also possible to create own wind energy index based on own settings, like reference period, dataset and

power curve. This can be done in the Performance Check module.

Figure 55 Wind energy indices can be browsed and loaded using different source. Detailed information

is provided for each series loaded.

MCP (Measure-Correlate- Predict) - Long term correction module 54

Having created an index database in Performance Check it is possible to browse the database and import the

indices.

Add from Meteo object (Figure 55) makes it possible to load a reference dataset which has previously been

downloaded in a meteo object. The energy index will use the settings established in the Session setup window,

namely the power curve choice and wind speed scaling, and will by default treat all data in the meteo object as

the reference period (100% index). To reduce the reference period, the data in the meteo object must be

disabled.

The Reference series (Figure 55) table provides information on the start-end date, the distance to local data

point/ and distance to long-term reference, the concurrent number of days between the reference and the local

measurements. Several parameters are calculated as follows:

• Variability on wind speed and energy is calculated as standard deviation on complete 12 months

average. As a curiosity it can be mentioned that this value is sensitive to the start month (which can be

user defined), probably related to a certain coupling on high/low wind months around new year.

• MK trend test (Mann Keldall) is a quantitative figure that identifies if there is a significant trend in the

reference data. The result will approach 1 if perfectly un-trended and 0 if perfectly trended. A trend in

the dataset can indicate a climatic event or could appear due to a data artefact. The trend should not

necessarily disqualify the data, but rather raise awareness on the suitability of the dataset. A classic

example in the ground stations, are the trees growing near a measuring mast used as reference. The

impact of the trees during the years would results in a trend making the data unusable.

• Sen’s slope: It is only calculated if there is a trend. The slope, or linear rate of change and the intercept

are calculated.

• LT bias: Show how much an alternative reference will adjust the measurements relative to the selected

reference based on wind energy index ratios.

In the bottom of the Evaluate Reference Time Series window the wind energy index and the cumulated wind

energy index are presented. The wind energy index graph can be shown as running average based on different

period lengths. By default, 1 year is used.

Figure 56 Wind energy index graph (top) and the cumulated wind energy index (bottom).

By increasing the running average to e.g., 5 years the relative drift between different data sources will be easier

to spot.

Looking at the accumulated wind energy index graph, for recent 120 months/10y (marked with red line), it shows

a rather constant offset between the different sources, all being very close to index 100 accumulating 10y back.

This would help conclude that recent 10 y seem a good choice, while these sums up to ~100% and the further

back in history discrepancies will be taken out.

MCP (Measure-Correlate- Predict) - Long term correction module 55

One could also consider the e.g. 6 years period as a good long term selection, being the point where all lines

cross first at the 100% index, however it should also be considered if the wind direction is long term

representative during these recent 6 years. The shorter the period, the higher the risk of not having long term

representative direction distribution.

There will later, when predicting, be an adjustment option for the “very long-term wind”, in case it is decided to

use a period, which do not reach 100%.

The Show period dropdown menu (Figure 56) gives the option to show the concurrent datasets. This has an

impact on the datasets which have additional months at the end period, as the accumulated analysis starts from

the latest date.

Compare to other references function gives an overview on how much would the alternative datasets predict

higher/lower than the selected reference, where the local data measurement period is used as part of the

evaluation.

Figure 57 Comparison to other references (LT Bias)

This tool establishes wind energy indexes for the alternative reference and the selected reference.

Then by comparing the index for the local measurement period to long term, it tells how much each reference

will gear up/down the local data. The ratios between the gearings tell how different the prediction with alternative

datasets is compared to the selected reference.

For each tested alternative reference there are:

1. Common reference periods (upper)

The alternative uses the same reference period in this evaluation as the user selected reference. If the

alternative has a shorter period than the selected reference, the evaluation is based on the concurrent

reference period.

2. Individual reference periods (lower)

The alternative dataset uses the period it was defined during the selection process, either defined by the user

(the reference was added from external sources Figure 55), or a predefined period for the wind energy indices

available within windPRO (Figure 54). Note the user can shorten the reference period by using the Limit period

feature (Figure 52). This can make sense in countries with poor data quality back in history, that also affects

MCP (Measure-Correlate- Predict) - Long term correction module 56

the model data like Merra and ERA. In such regions it thereby often will be recommended to use a shorter

reference period.

By mousing over the underlined figures, the periods behind the data used are graphically marked in the coloured

table.

Figure 58 Period used for different alternative and reference datasets.

The text explains what would be calculated by the alternative reference compared to the selected reference.

The decision on the reference source and period to be used is probably the most important choice in an MCP

session. The tool will provide the answer, but it offers the above-mentioned tools to support the choice.

3.5.1.2 Session setup

Figure 59 Session setup window. Default setup changed.

In session setup it is decided how wind speeds are converted to energy, in all places in MCP where wind energy

is used as measure.

There are two options available:

MCP (Measure-Correlate- Predict) - Long term correction module 57

• Use actual power curve of a selected WTG- this option allows you to select a power curve from the

turbine catalogue.

• Use simple power curve approximation with squared wind speeds. This is also the default setup and

uses a simple power curve, established as squared wind speeds, truncated at 13m/s and cut out at

25m/s, to convert the wind speed time series to energy.

In terms of data availability, the default selection uses periods with at least 60% data available to avoid

calculating e.g., weekly or monthly wind energy index with too few data.

The Scale wind speed option can be used if the local and reference wind speeds differ too much. The reference

might be at a level around 5 m/s while the local is around 7 m/s. This will when by looking up in the power curve,

create a relative bias between the local and reference, that will make the energy correlation very poor. Scaling

both series to same level makes the energy indexes comparable.

Going back to “Measure” will right away update the Energy based evaluations.

3.5.1.3 Adjustments

Figure 60 Adjustments available on the loaded time series.

Following adjustments can be set:

• Time offset - calculates which time offset gives the best correlation between local time series and long-

term data on the wind speeds level. This is tested by running a number of correlation calculations (up to

+/- 24 hours). Often the local data has a 10-minute time domain while the reference data comes with an

hourly time resolution. An auto detection method solves the challenge of matching the time stamps

between the two datasets with different time resolution. This also solves the issue of datasets being in

different time zones.

• Wind Veer: Calculates the veer between local and reference data for concurrent samples at wind

speeds above threshold (min 4 m/s) set in Data statistics (Figure 50). The output is the veer correction

which results from the average of all included samples.

Show data points check box will show the entire amount of data points used to create the whisker plot

graph.

MCP (Measure-Correlate- Predict) - Long term correction module 58

There can be several reasons for seeing a veer between the two datasets, such as:

1. poor calibrated wind vane at the local mast. Here the local data should be veered.

2. local turns in wind direction affected by micro terrain effects (steep slopes) not captured by meso/model

data. in this case the reference data should be veer corrected. For other reasons it must be judged which

to veer. Note: Veering will affect the final output!

• Averaging - it averages the local data, and it is possible, for instance, to create hourly averaged values

based on 10-minute data series. The averaging is a moving average, and it is made on the preceding

period up to the time stamp indicated.

• Interpolate reference - A typical reference dataset has a minimum time resolution of 1-hour, while the

local measurements typical have 10-min time resolution. By resampling the reference timeseries to 10-

min (by simple interpolation) the amount of data for training the model increases, and better prediction

results can be obtained.

3.5.1.4 Model input data

Figure 61 Input data after all applied adjustments.

In the Model Input data, the two datasets, local and reference, are presented, this time, considering all

adjustments applied by the user in the Adjustments tab.

The data to be used by the model(s) can now be inspected. The relations between the two datasets and the

features available in this window are similar to the ones in Measure tab.

It is recommended to conduct a final inspection on the datasets in this window, before proceeding to the selection

of methods for long term correction. Any suspicion or erroneous data, e.g., a week with large deviation between

reference and local should be addressed by going back to the Meteo object for a second review in the data

screening, or by reconsidering the adjustments applied in previous tabs.

MCP (Measure-Correlate- Predict) - Long term correction module 59

3.5.1.5 Concept choice and uncertainty

Figure 62 Choice of two different concepts for MCP

As the texts explain, there are two different concepts to choose between:

• Create a long term data series – several methods are available and their performance can be evaluated

by using a slicing method. The user will determine which method best handles the specific data series

by evaluating the performance of this various methods. The resulted long term time series will have the

same length as the reference dataset selected in the Measure tab.

• Scale local data to be long term representative – a single method is available which does a scaling

based on the ratio of wind speed and standard deviation between the local and reference data. The

resulted time series has the length of the local measurements.

More details on each method are provided in 3.5.4 Models used by MCP.

The uncertainty based on selected data and filtering is shown. This is based on the Klintø model (Klintø, April

2015). The model is based on analyses of many datasets, where it is identified which parameters drive the

prediction errors, and in which size order. The results of the analyses are then converted to a formula setup, that

takes the most important parameters into account and creates an uncertainty figure that match the analyses

results.

The parameters considered in the uncertainty evaluation are the wind energy index, the Pearson coefficient

(correlation), the inter annual variability and concurrent period. The formula used is presented in Figure 63.

MCP (Measure-Correlate- Predict) - Long term correction module 60

Figure 63 Uncertainty calculation in MCP

It has so far been decided not to bring uncertainty results based on less than 1 year of concurrent data to the

reporting. The value is however calculated. But having much less than 1 year of data, it is not considered reliable

to give a qualified uncertainty.

The most comprehensive part of the MCP module is the modelling of a long term data series which is being

processed in the Model LT and/or Scaling Local tab, presented below:

3.5.1.6 Model LT data – Model Choice

Here the transfer functions between the two concurrent datasets, local and reference, are generated using five

main methods, as shown below.

Figure 64 Model LT tab

MCP (Measure-Correlate- Predict) - Long term correction module 61

Five different methods are available:

Simple Speed scaling – it is set as a default method. This simple method calculates the ratio between local

measurements and reference mean wind speed during concurrent time. The ratio is further applied on each

reference wind speed time step. Only the wind speed is being scaled in the process.

Regression – It is the most common used model through decades. It makes a regression analysis on the

concurrent datasets and establishes the regression lines per direction sector. Adjustments are available for this

method. Detailed explanations are provided in the model paragraph 3.5.4.2. Both the wind speed and wind

direction can be long term corrected in this process.

Matrix – This method models the changes in the wind speed and wind direction through a joint distribution fitted

on the wind speed and wind direction bins. Both the wind speed and wind direction can be long term corrected

in this process.

Neural network – This model is a machine learning algorithm which trains a neural network to detect a pattern

between the local measured and reference wind conditions during concurrent time.

Solver based – This method finds the best scale and offset parameters for each sector, season and diurnal

index to be applied on the reference wind speed. The method uses the linear regression to determine these

factors. Additionally, the found factors can be further exported to the Scaler and used directly in a Post calibration

process.

Features available in the Model LT data:

Train & test model– provides a basis for the user to assess the performance of the four methods, by alternating

slicing the time series into several intervals, and checking how well the predicted data matches the observations

(measured).

Datasets are divided in concurrent time periods used for training the model and testing it.

For example, if the 5-hr alternating slices is used, the first 5 hours are used for training, next 5 hours for testing,

following 5 hours for training and so on. Further it is possible to evaluate how well the model works based on

having only half of the data, by looking at the other half. Thereby different models can be intercompared on a

fair basis.

The alternating slicing option is selected from the drop-down menu next to the Train & Test button.

Figure 65 Drop down menu with the alternating slicing options and the Tran & Test button.

Note, in the final prediction, after a method is selected and the long term dataset is saved, the entire period is

used in training, no matter the alternating slicing selected. The purpose is to give the method the option to train

on as much data as possible and ideally result in the best possible trained model.

Residual resampling - is optional for all models, except simple scaling. It can be selected from the radio button

marked in Figure 64 Model LT tab.

Residuals add ‘noise’to the predicted time series, which gets the wind speed distribution closer to real

measurements. But it also reduces the correlation and increases the errors shown in the prediction table.

Therefore, a more fair measure for errors and correlations is seen without residuals. Default means that residuals

are only included in evaluations where these normally improve the results. In section 8.3.1 of the MCP Reference

Manual more details are given.

MCP (Measure-Correlate- Predict) - Long term correction module 62

Figure 66 The Compare view in Model LT

Upon running all prediction models with the Train & Test button, results are displayed in a table next to the

selected model, as well as graphically under the Compare tab.

The table values and graphs can be shown based on aggregation on different time resolution (hour-day-week-

month), and as well on wind speeds or energy. These selections are made from the drop down menu next to

Statistics averaging and Parameter.

The table offers an overview of the performance of different selected methods. The first four parameters: MBE

(mean bias error), MAE (mean absolute error), RMSE (root mean square error) and Correlation, provide

information on how well the wind speed is predicted by the model across the time series and the correlation

between the measured and predicted datasets is given.

Measured and Predicted concurrent wind speed – These two parameters are not entirely comparable,

depending on the slicing method selected. If the slicing test ‘Entire concurrent period’ is used, than the two are

comparable. Measured Wind Speed Concurrent represents the mean wind speed of the part of the local

measurements concurrent with the long term reference. The Predicted concurrent wind speed represents the

mean wind speed of the predicted timeseries using a certain method. If any other slicing method is used, the

predicted wind speed will only cover half of the dataset, the test period.

Long term wind speed – is the predicted mean wind speed for the entire reference period, corresponding to

the trained model applied on the full reference dataset.

KS (Kolmogorov- Smirnov) Statistics – This tests if the two datasets, both measured and predicted, follow

the same distribution, and calculates the size of the difference between the two. In MCP it is worth to notice that

it is not just the KS-test hypothesis, which finds if the largest error is below a given threshold, but the statistics,

where we test the modelled vs measured accumulated distributions and find the largest error. This is reported

as the deviation in number of accumulated samples where the deviation is largest, divided with the total number

of samples. Thereby it is not just a yes or no, but a value that tell how well the two distributions match.

All parameters mentioned above are further described in section 8 of the MCP Reference Manual.

Compare graphs – it provides a graphical representation of the parameters shown in the table. The color code

is defined in the text of each method.

The Correlation graph is shown as (1 – result) to get a better resolution and still have 0 y-value at x-axis. Note

that all models are shown with “Default residuals” in figure above. This means all without residuals, except for

KS statistics. For all bar graphs: The smaller numerical value the better!

MCP (Measure-Correlate- Predict) - Long term correction module 63

Figure 67 Evaluation of Selected model with residuals

Selected model tab shows a series of graphs with information on the correlation between measured and

predicted wind speed and wind directions, as well as the frequency, diurnal and seasonal distributions.

Figure 68 The predicted data shown along with the local measurements.

MCP (Measure-Correlate- Predict) - Long term correction module 64

Time series tab – shows the predicted and measured datasets with different time resolutions, selected from the

Statistics averaging drop down menu. At this evaluation it is possible to see performance between the two

datasets, predicted and measured. If the performance is poor a double check of the measured data and

reference data is recommended. In the Meteo object additional cleaning or adjustments can be done.

Predict

Each method has a predict button. The method that performs best should be used for prediction. Click the predict

button next to the model.

Figure 69 Prediction of long term dataset

On the right part of the form, the used model and data, local measured and long term reference, is shown. In

bottom part, it is possible to correct the long-term series if there is a known bias which doesn’t make it long term

representative.

Figure 70 Correction for non long term representative reference

MCP (Measure-Correlate- Predict) - Long term correction module 65

The bias correction can be made on wind speed or wind index by adjusting the mean wind speed.

In the left part of the Predict form, there are two output options:

• Output as time series- The long term corrected data is saves as a time series in a new meteo object or

added to an existing meteo object depending on the user selection.

• Output as wind statistic – The long term corrected data is saved as a wind statistics. To generate a wind

statistic, a site data object must be chosen in the Setup botton. This holds the information on the terrain

(roughness and orography) as well as the handling of eventual obstacles. On the setup page handling

of displacement height can also be chosen.

3.5.1.7 Model Choice: Scale Local

Figure 71 Scaling Local data to LT level

This is often a preferred method, if longer local measurements are available, if the site is in a wind climate that

is very much the same year by year, e.g., where temperature is the main driver, or if the correlation between

local and reference data is very poor and a traditional long term correction with transfer function is not

recommended.

The advantage is as previously mentioned that the local measured direction distribution is used, which on

locations with much turn e.g., near mountains often will be a better approach while the reference data might not

reflect this. Also preserving the dynamics of the local measured wind speeds can improve accuracy, especially

when calculating by time step.

The implemented method is the variance ratio method, 3.5.4.6.Scaling Local, where a Factor and Offset

parameters are used to long term correct the local data.

Based on the Factor and Offset, the scaled calculated time series can be written to a Meteo object or a wind

statistic can be generated based on the scaled time series and the terrain from the site data object.

MCP (Measure-Correlate- Predict) - Long term correction module 66

If the local data includes a period non concurrent to reference, this is included and an extra “post factor” is used

to make the entire local period long term representative.

Sessions in MCP

Figure 72 Example of several sessions compared.

Multiple sessions can be created in MCP. Each session can differ by:

• Choice of reference data

• Choice of local data

• Model selection and parameters

• Different periods for as well reference as local

• Different adjustments (time offset and veer)

Thereby the changes in results based on different choices can be seen fast and easy and confidence can be

gained, or outliers detected. Key parameters as predicted long term mean wind speed and uncertainty within

each session is seen. The figures will be available in the reporting part.

Reports from MCP

Reporting consists of two parts:

Session overview - The different session results are compared based on selected model within each session.

The uncertainties based on selected data for each session is a part of the report, where each parameter included

in the uncertainty calculation is shown.

Session details – For each session, the tested models are reported for comparison.

MCP (Measure-Correlate- Predict) - Long term correction module 67

Figure 73 Part of the report for session overview

Figure 74 Uncertainty calculation

In the uncertainty calculation, the input figures and the constants are shown. In this case the Wind energy index

is the major difference driving the uncertainty up where the index differs most from 100%.

MCP (Measure-Correlate- Predict) - Long term correction module 68

Figure 75 Example of MCP session details report

In session details, it can among many other information’s be seen how the use of residual resampling improves

the wind speed frequency distribution, while it is shown in direct comparison with the result without residuals.

Models used by MCP

In the MCP reference document more information on the math behind the methods is found. In this chapter

detailed explanation is given to the different input options for each of the models.

3.5.4.1 Model LT; Simple Speed scaling

A simple method is used which consists of simply dividing local concurrent mean wind speed with reference

concurrent wind speed. The resulted ratio is further multiplied with the wind speed series of the reference dataset,

resulting so the long-term corrected mean wind speed.

MCP (Measure-Correlate- Predict) - Long term correction module 69

3.5.4.2 Model LT; Regression

Figure 76 Settings for Regression opened by pressing the Edit button.

Binning is by default by direction sector. Optional binning can be set by season, where the EMD Season setup

is used:

MCP (Measure-Correlate- Predict) - Long term correction module 70

Figure 77 Season setup

Setup of seasons and day/night – This setup can divide in time of the year as well as time of the day. It is e.g.,

possible just to divide in day/night by reducing the number of seasons to 1. It is not recommended to divide into

many seasons, this would give too few data for the analyses, only day/night and/or summer/winter would

probably be the best recommended resolution.

Transfer function type - Sector handling

The transfer functions can be made for a small number of sectors (usually 12) or for 360 directions, meaning

360 transfer functions. Division in a number of sectors (typically 12) is the way for “homemade” Excel

applications. This should be chosen for comparison with such calculations. But letting the software tool do the

hard work, making 360 transfer functions based on a specific window size will be more accurate and is here

chosen as default.

Sector window - Each transfer function will be made on the basis of data from a range of directions centred on

the direction in question. The directions refer to the reference directions and 30 degrees is default. If 360 are

chosen only one transfer function is made based on all data.

Skip angle difference larger than - Particularly at low wind speeds the angle between matching reference and

local data may deviate significantly and cause a lot of noise. By rejecting point with large difference in direction

this noise can be reduced. However, this means discarding information that could be important. By default, all

data are included.

Skip wind speeds less than - As very low wind speed contributes with a lot of noise and often deviate from the

linear relation seen at higher wind speeds it is useful to simply discard them from the transformation function.

This does not mean that they are discarded from the full reference dataset being transformed to. Depending on

the amount of noise and actual wind speeds at the reference this limit can be set freely but 2 m/s is default.

Regression model (Wind Speeds)

MCP (Measure-Correlate- Predict) - Long term correction module 71

Figure 78 Wind Speeds- Regression Model.

Here the type of regression on the wind speed is selected. Linear regression means first order and a two-

component linear regression will usually be preferable to a regression through origin (0,0) as this provides a

better fit in the wind speed range relevant to production. This is also default. An alternative is a 2

order

regression, which will fit a parabolic curve to data. This may create a better fit but allow deviating extreme wind

speeds to influence the fit at high wind speeds.

Regression model (Wind Directions);

Figure 79 Wind directions - Regression Model.

Usually, the direction change is independent from wind speed and so a 0

order regression should be used.

Residual Model (for wind speed and/or direction):

Figure 80 Residual Model selection for wind speed or wind direction

Gaussian (normal) distributed residuals are mainly included for backward compatibility. It proved to be a too

simplified approach.

The “Advanced Gaussian” of order: 1 is the recommended model. This method is a function of the reference

wind speed so that the regression formula is y = ax+b+e(x).

The reference wind speed range is divided into a number of intervals. Within each interval the observed scatter

is found as a standard deviation of the scatter together with the bias on the observations. Both are then applied

when transferring the reference data to the site.

The observed standard deviation is modelled as either a first or a second order polynomial. The result is a much

more dynamic fit than a standard linear regression. On well correlating data, internal tests have shown a

significant improvement on the precision of the long-term prediction. The second order residual resampling do

have the problem that few very scattered points at high wind speeds can exaggerate the resampled scatter

MCP (Measure-Correlate- Predict) - Long term correction module 72

unreasonably. It is therefore recommended to use it carefully and if in doubt stay with the default first order

advanced Gaussian residual resampling.

3.5.4.3 Model LT; Matrix

Figure 81 Settings for Matrix method.

The binning is done the same as described in the regression model, by default by direction sector, alternatively

a seasonal binning can be defined.

Sector window - Each transfer function will be made on the basis of data from a range of directions centred on

the direction in question. The directions refer to the reference directions and 30 degrees is default. If 360 are

chosen only one transfer function is made based on all data.

Skip angle difference larger than - Particularly at low wind speeds the angle between matching reference and

local data may deviate significantly and cause a lot of noise. By rejecting points with large difference in direction,

this noise can be reduced. However, this means discarding information that could be important. By default, all

data is included.

Wind speed window - Since the Matrix builds a list of possible outcomes for the transformation for each degree

and each 1 m/s wind speed bin, the user can decide to include also neighbour wind speed bins for each bin in

calculation. Default only the wind speeds within the bin in question is used.

MCP (Measure-Correlate- Predict) - Long term correction module 73

Minimum records behind to accept polynomial fitting - To avoid bins based on very few data records to

influence the model, these can be filtered by choosing a minimum number of records. By default, the minimum

is set to 5 records.

Direction window size for polynomial fitting -Transfer function regression will be based on all transfer function

within a specific window width. By default, the window is set to 10 degrees.

The observed extreme minimum and maximum values of mean and standard deviation on direction and wind

speed change are listed. Extreme values indicate highly uncertain transformations. If this shall be changed, you

must go back to the setup of the parameters for calculating the transfer functions.

Polynomial order - For each of the parameters a polynomial regression can be fitted to describe the

transformation function. The order can be freely chosen and can even be different from parameter to parameter.

The default choice for mean wind speeds is a 1

order fit as this is less sensitive to deviating extreme values.

For mean change of wind direction, a 0

order polynomial is default (wind speed independent direction change).

For the standard deviations, 1

order higher is recommended for wind veer and 2

order for wind speed.

3.5.4.4 Model LT; Neural Network

Figure 82 Settings for Neural Network - MCP model

The Artificial Neural Network (ANN) - establishes transfer functions from reference data to measurement data

by using ANN networks trained with reference data as input and measurement data as output. Once the networks

are trained (with backpropagation algorithm) they can be used to predict wind speed and direction based on the

reference data.

From a user perspective the ANN method only has three settings:

Diurnal variation - The hour of the day is used for each sample as input to the network. It will often give a

better diurnal match when this is enabled.

Monthly variation - Similar to diurnal variation this will add the month of the year as input to the network. In

cases where the difference between reference- and measured data seems to have a monthly variation this

option can be enabled to possibly get better results. This option should be used with caution if the measured

data doesn’t include a full year since the network would then eventually be used with input data that it hasn’t

been trained with.

MCP (Measure-Correlate- Predict) - Long term correction module 74

Add wind speed residuals - Like regression methods an ANN approach tends to “remove” some of the

variability in the predicted data. Therefore, residuals can be added as a post-processing to re-establish the

variability seen in the measured data. The ANN method uses the 1

order Advanced Gaussian method that is

also used in the Regression MCP method. The residual model is calculated based on the predicted data vs.

the measured data, and then used to add a residual on each predicted sample.

3.5.4.5 Solver based (For Scaler)

The goal of the Solver based MCP method is to establish a method that has training parameters identical to the

ones found in the Scaler Post Scaling options, so that the parameters can be exported from MCP to a Scaler

and used in e.g RESOURCE and PARK.

These parameters are:

1. Overall offset that is added to all samples (



)

2. Overall scale that is multiplied to all samples (



)

3. Scale factor based on wind direction sector (





4. Scale factor based on season of year, also called month factors 









5. Scale factor based on diurnal index, which can be seasonal dependent 













Since all these parameters will correlate it is not possible to establish a deterministic method that will set the

parameters, so an iterative solver must be used.

The solver works best if the overall offset and scale is determined first. Those factors are determined by linear

regression on all concurrent samples. So, the basic expression of determining a sample would be:









 



 



Once the overall offset and scale parameters are determined each sample in the concurrent time series is

marked with an index of how they belong to each of the other post-scaling parameters (points 3 to 5 above).

This means that the combined expression for a predicted wind speed would be:









 



 



  









 









 













The goal of the Solver MCP method is then to minimize the objective:





 











In the current implementation we offer three different solver algorithms, namely:

• L-BFGS (default)

• BLEIC

• Levenberg-Marquardt

Tests have so far shown that L-BFGS gives the best results and most users would not need to change that.

Currently the sector, month and diurnal variables are initialized to 1.0 before starting the solver. It could later be

investigated whether results could be improved by finding better start guesses before starting the solver.

The solver uses steps of 0.01 and has a default stop-criteria of 50.000.000 iterations or when the goal function

is below 0.1 m/s.

The setup window for the Solver model looks like this with default settings:

MCP (Measure-Correlate- Predict) - Long term correction module 75

In this setup the overall Scale and Offset are calculated before the solver runs for the other parameters. The

sector factors are set to 12 sectors by default.

For both Month and Diurnal factors, the user chooses a windPRO season setup (see Figure 77 Season setup).

This season setup defines how many seasons a year should be divided into for the Month factors and for the

Diurnal factors it defines how a day should be split into periods and whether different diurnal factors should be

used for different seasons of the year.

If the MCP session is set up to use a Scaler the user can also choose to use the Sector, Month and Diurnal

factors from the original Scaler. So, if the user has already found e.g. the Sector factors, then these can be used

directly and will not be part of the Solver algorithm.

If a Scaler is used in the MCP session there is a new option in the Predict window to output the predicted scaler:

This button creates a copy of the Scaler used in the MCP session and then adds the trained parameters to the

Post Calibration of the copied Scaler. Here is an example of a Post Calibration from a MCP Solver model:

MCP (Measure-Correlate- Predict) - Long term correction module 76

3.5.4.6 Scaling Local

The purpose is to adjust the short-term site time series of wind speed using the long-term reference to ensure

that the wind speeds of the local series match the long-term level.

This requires the local data and reference series to fully overlap in the concurrent period.

With the concept of index correction, we do not establish a direct site-to-reference relationship as in standard

MCP methods. Instead, we base the correction solely on the reference series. As a result, the correction reflects

the relationship between concurrent and full concurrent-to-full for the reference series.

Figure 83 Graphical representation of dataset used and their lengths.

In the Scaling local method, the transfer function is allowed to have both a slope and an offset, so it is possible

to correct both the wind speed mean and stdev from concurrent-to-long term conditions.

Model:







 

Two equations are needed to constrain both parameters, α and β, one for the mean and one for the stdev, which

both follow directly from the definitions of mean and stdev. and the assumed linear relationship/transfer-function.

Constraints:



















 

Models/modules for initial calculations 77









Which together yield the following expressions for α and β.









































A minor problem is that the correction model may lead to negative wind speeds for some of the corrected wind

speed samples. This is pragmatically handled by truncating these samples to zero.

Thus, the full correction equation for u

site

is.

















































The found scale and offset is shown on the form:

Figure 84 Found scale and offset together with predicted long-term mean wind speed.

If the local data include a period non concurrent to reference, this is included and an extra “post factor” is used

to make the entire local period long term representative. The post factor is simply the average wind speed for

concurrent local data divided with the average wind speed for all local data. This is multiplied on all wind speeds

in the scaled local series. Example including post factor is shown below:

Figure 85 Found scale, offset and post-factor together with predicted long-term mean wind speed.

Note the post factor here change the mean wind speed 2.6%. This is unusual high, and the reason is we have

found a period with an extreme high month and excluded this just to test the model.

Models/modules for initial calculations

Most AEP calculations will be performed with the PARK module. But, for providing data to the PARK module or

for doing initial tests calculations or for calculating single turbines, there is a number of auxiliary modules. As an

example of a possible use is the comparison of AEP for many different turbine types at a specific site.

In the following descriptions, selected features are documented within each paragraph, but not repeated for all

modules since the input screens are the same for several other modules.

METEO

The METEO module performs a calculation of the wind distribution and the AEP on a single spot, where the

measurement mast is located. The calculation assumes the wind data are long-term representative. Having,

e.g., ½ year or 1½ year of data, thereby, will give a biased AEP result as all seasons are not equally represented.

AEP results will always be “scaled” to a full year, no matter the length of data period or data recovery.

Input:

• METEO object with wind data.

• Turbine types (from turbine catalogue or user defined wtg file) and hub heights

• Air density correction (world climate database, available within software)

Output:

Models/modules for initial calculations 78

• Calculated AEP (Annual Energy Production) for each turbine and hub height at the position of the

METEO object (calculation point) as the main report.

• Production analyses, Power curve analyses, Wind data analyses, Wind profile (optional) and Map as

detailed report pages.

• Data files for spreadsheet analyses with turbine specifications, calculation results etc. If time series is

used in the calculation, a 24-12 (hours-months) distribution will be available.

The input data must come from a METEO object, where the measured data is imported, or data from EMD On-

line data services is downloaded. No horizontal model extrapolations are available, but, optionally, vertical

extrapolations based on shear input are available.

The METEO calculation is started from the Energy tab and selecting METEO from the Single Point AEP button:

Figure 86 Main page input for METEO calculation

At the main page, it can be decided to include a “simple reduction” to compensate for expected losses and,

eventually, uncertainty deduction. This “primitive” approach can be overruled by different choices, as seen in the

form above. The hub height for key results can be set. This is in order to have comparable results for different

calculations, where the hub heights might differ. It is optional to calculate wind profiles that will be reported as

an average for the site and by direction sector also.

Models/modules for initial calculations 79

Figure 87 Wind distribution and WTGs input

The wind data selection offers data from all METEO objects in the project (filtering based on “purpose” is

possible), where, as well as the time series, the “MEASURE” (the table or histogram data) or the Weibull fitted

distribution can be chosen. Thereby, it is easy to compare the difference in results using, e.g., the measurements

direct with the Weibull fitted data. In the input form, more turbines and hub heights can be selected. Finally, the

power curve handling, especially the air density correction, can be setup. See Section 3.7.4 Common settings

for Wind statistics based (standard) PARK calculation.

Figure 88 Input of shear in METEO calculation

When the METEO calculation starts, an input screen for shear appears. These are the options:

• Use data from METEO objects shear tab, where more sets of shear values can be established.

• User define shear by typing shear values as power law exponents by each direction sector.

NOTE: Shear values in non-flat terrain can, when established from more measurement heights, partly include

hill speedup. This is also similar near forests. It can, thereby, give a highly uncertain calculation result.

The k-parameter correction is a little complicated as it highly depends on the height scaled from and to. Until

windPRO 3.5 the default was 0.008 increase per meter, based on rather old empirical values. From windPRO

3.6 no value is recommended. Taller measurement and hub heights has turned the old default irrelevant. If

Models/modules for initial calculations 80

needed, the below information might support making a good choice. A better choice would although be to create

a synthesized time series in meteo object in for calculation height. Then no extrapolation of k is needed.

Figure 89 k Weibull parameter change with height, left onshore, right offshore wind statistic for DK.

The figures above show how WAsP assumes the k-parameter to change for the 4 roughness classes, 0 – 3, for

a given wind statistics. On the left is the Danish “basis” wind statistics, Beldringe. On the right is a Danish offshore

based wind statistics. The general picture is the same, only the offshore based values are higher, which is partly

due to higher mean wind speeds. It can be seen that onshore, from 50m to 100m, the increase is around 0.2 for

50m, meaning that the increase is 0.004 per m. Going higher than 100m or offshore, the increase becomes

negative. -0.001 seems to be a reasonable value if your measurements are from 100m and you want to scale to

larger heights, or if you go up from 50m offshore.

Figure 90 Measured Weibull k change by height – offshore example

Above is an example of real, measured k-values on a tall offshore mast in Scandinavian waters with long distance

(>30km) to shore in all directions.

Models/modules for initial calculations 81

Figure 91 Output example from a METEO calculation

Above is an example of the main output from a METEO calculation. The results can be taken out and put into a

spreadsheet for further processing by using the “result to file” output, shown here:

Table 5 Result to file output from a METEO calculation

Models/modules for initial calculations 82

Figure 92 Production analyse output

For each turbine (power curve) and hub height, detailed presentations are available.

Models/modules for initial calculations 83

Figure 93 Wind data analyse output

More result pages give a detailed presentation of the results.

ATLAS

Atlas is a simple model based on the “windatlas for Denmark” from 1979 by Risø. The model has been refined

by EMD during the 80’ties, but not further developed since. The model is, although, still an affordable, simple to

Models/modules for initial calculations 84

use and reasonably accurate model. It requires a wind statistics, describing the regional wind climate. These are

available for several countries but of varying quality. Upcoming in 2015 is a world wind atlas from Risø/DTU. To

establish a wind statistics from measurements the WAsP model will be needed.

Input:

• Site data object with wind statistics, roughness rose, hill length and height by sector and obstacles by

distance, height and porosity by sector.

• Turbine types (from turbine catalogue or a user defined wtg file) and hub heights

• Air density correction (world climate database available within software)

Output:

• Calculated AEP (Annual Energy Production) for each turbine and hub height at position for site data

object (calculation point) as main report.

• Production analyses, Power curve analyses, Terrain, Wind data analyses, Wind profile (optional),

Wind statistics info, Map as detailed report pages.

• Data files for spreadsheet analyses with turbine specifications, calc. results etc.

The site data object is the “complicated” part of the calculation setup. First, a site data object is inserted in the

position on the map where the calculation shall be performed.

The purpose is set to “ATLAS” calculation.

A wind statistic is chosen:

Figure 94 Wind statistics selection form. More statistics can be selected by <Ctrl>.

The wind statistics can be selected in the site data object, and there will be graphical features for evaluation of

the wind statistics. An important feature is that the “energy levels” can be compared, which can give an idea of

how realistic a given wind statistics is by comparing to other wind statistics in the region.

Models/modules for initial calculations 85

Figure 95 When more statistics are selected, individual weight can be given.

The weights given to each of more wind statistics can be decided manually. It can also be decided given by

reciprocal distance weight, which can be auto updated when moving the site data object on map. For resource

map calculation, the auto update can be performed for each calculation point.

Figure 96 Input form for roughness rose data

Models/modules for initial calculations 86

Figure 97 Input form for ATLAS Hill/Obstacles

The terrain (roughness, hills and obstacles) can be entered as values for each sector in the form, but more

conveniently, choose OK to site data object and make the input in graphic mode:

Figure 98 Graphic roughness rose establishment.

By right clicking roughness, changes can be established and the change distance can be dragged. Also, the

roughness value can be set by right clicking and choosing “Roughness”. Similarly, hills and obstacles can be

edited graphically by choosing the menu line “ATLAS hills and obstacles” in the right click menu as seen above.

Models/modules for initial calculations 87

WAsP interface

WAsP is the comprehensive wind atlas model calculation engine from Risø/DTU, first released in 1989, still in

development and considered as the Industry standard for wind energy calculations. The WAsP model is fully

controlled from windPRO and the user does not need to use the separate user interface for the WAsP model.

The installed WAsP versions can be seen by opening the Show Data and Models window from the Settings &

Help tab:

Figure 99 Models available/licensed

Green means “license activated”. It is important which WAsP version is used. In the validation Section, 8.4, an

overview of the modifications in recent versions is analysed. It is of high importance to notice that, from WAsP

ver. 10.2, roughness roses are no longer supported. Roughness data must be available as roughness line maps.

This is the major change to WAsP 10.2, but this does not affect the traditional WAsP model. WAsP 11/12 includes

CFD model handling,

Input:

• Site data object with wind statistics, roughness data)

and orography (link to files/line objects/grid

objects)

• ALTERNATIVE to Site data object is WAsP-CFD result file(s) + wind statistics(s)

• Obstacles digitized on the map.

• Turbine types (from turbine catalogue or from a user defined wtg file) and hub heights

• Air density correction (world climate database available within software)

• Displacement height, optional calculated from sector wise displacement height calculator

Output:

• Calculated AEP (Annual Energy Production) for each turbine and hub height at position for site data

object (calculation point) as main report.

• Production analyses, Power curve analyses, Terrain, Wind data analyses, Wind profile (optional),

Wind statistics info, Map as detailed report pages.

• Data files for spreadsheet analyses with turbine specifications, calc. results etc.

Roughness data can be roughness roses (WAsP 9 or previous) or roughness line maps (all WAsP versions).

Models/modules for initial calculations 88

Figure 100 Selection of roughness and elevation data in Site data object

Above is shown the dialog box for input of elevation (orography) and roughness data. Note, more line objects,

map files and elevation grid objects can be linked to the site data object. All the checked ones will be used in the

calculation. Be careful when using multiple datasets as this can create inconsistency, especially having two

different roughness line files overlapping can create huge calculation errors. Thus, it is very important to

be aware of which files are linked. The files linked are controlled by the settings on the line/grid objects, which

have this box checked:

When checked, the data are automatically linked, when creating a site data object. But NEW created line/grid

objects with these setting will not be linked in a previously created site data object.

When using Elevation grid objects for orography, the gridded data will, during the calculation, be converted to

contour lines as specified at the WAsP tab in the elevation grid object. Elevation grid can have properties like

“terrain”, “surface”, “sea depths”, which will decide it the object are relevant as input for energy calculations. By

“surface” option, it can be doubtful, but it is accepted. The reason is that e.g., SRTM data are surface data, but

very popular for energy calculations, although they can give misleading input in forest regions. The turbines will,

if care not are taken, be handled as they are placed on the top of the forest.

Figure 101 WAsP interface input form

Models/modules for initial calculations 89

The structure of input form is similar to ATLAS, but it has the extra tab for displacement height, which in

combination with the wind profile calculation makes it an interesting analyse tool. The effects of as well

Orography, Obstacles and Displacement height can be seen direction sector by sector vs. height in the reports.

Figure 102 Set up displacement height in WAsP interface

3.6.3.1 WAsP parameters

Figure 103 Edit WAsP parameters from windPRO

As an example, a few modifications are made to illustrate where changing parameters can be relevant:

Std. Height 5 is changed from 200 to 120m. If your hub height is 110 m, you might be better off using 100m and

120 m than 100m and 200m. The interpolation distance simply becomes smaller.

Std. Roughness #5 is changed from 1 m to 2 m. This could be relevant in a very high roughness region, where

you expect above 1 m length (class 3.7), and thereby will avoid extrapolation, which is more uncertain than

interpolation. (In the most recent, WAsP 11, extrapolation is not allowed).

Finally, the heat flux parameters could be relevant to change in very hot areas (deserts), or if you want to force

water stability conditions, you can simply set the water parameters for land.

Models/modules for initial calculations 90

Figure 104 Edit WAsP parameters from windPRO

Remark: if WAsP parameters are changed from defaults, the button becomes yellow. When you are in other

parts of the software, like in a METEO object, profile viewer, where you also have access to WAsP parameters,

the button will also be yellow. This reminds you that you are working with modified WAsP parameters.

It is, of course, important that the modified WAsP parameters BOTH are used when generating the wind

statisticss AND in the AEP calculation. (Going up/down with WAsP with same parameters). It will typically be in

the METEO object or the Meteo Analyser experiments where “fixing” the WAsP parameters will be performed.

From WAsP 11, the WAsP parameters, when generating a wind statistics, will be saved as a part of the

wind statistics and USED when calculating with this wind statistics. Thereby, it is not possible to go

up/down with different parameters.

3.6.3.2 Geostrophic shear modelling

From WAsP 12 there are added two new parameters, which also are available from windPRO:

The vertical and horizontal extrapolation models in WAsP are modified to take into account large-scale horizontal

temperature gradients (‘baroclinic’ effects), which induce geostrophic wind shear. The method is implemented

by extracting average geostrophic-scale wind shear from global CFSR reanalysis data, with values from the

nearest grid points automatically used to provide more accurate AEP predictions. (Explanation from WAsP Help)

It must although be noted that the modifications in general are small, and only relevant when extrapolating data

in height.

WAsP-CFD

WAsP-CFD is part of the latest WAsP ver.11. The base concept is the wind atlas method, but with the Ellipsys

(DTU) CFD model as the flow modelling engine. The WAsP-CFD model has a very high resolution and can, so

far, only run as a remote Cluster calculation, hosted by EMD. The calculation procedures are fully controlled

from windPRO. Calculation credits must be purchased to run WAsP-CFD calculations. See further details in

Chapter 4, where also the input data for WAsP-CFD is described as well as links to validation.

Input:

• Site data object with roughness data and orography (link to files/line objects/grid objects) + Defined 2 x

2 km calculation areas.

Output:

• Job to send to Cluster for calculation

• When Job is done, an email notification is sent

• When reopening the calculation, CFD result files can be downloaded

• WAsP-CFD result files can be used in PARK calculations, WAsP and Resource calculations, and also

from the Meteo Analyzer and METEO objects.

Models/modules for initial calculations 91

WAsP compared to WAsP-CFD

Figure 105 Example of WAsP result compared to WAsP-CFD result.

The main reason for using WAsP-CFD is to have a higher calculation accuracy in complex terrain. It is a “known

issue” that WAsP tends to over predict low elevated turbines relative to higher elevated, when the terrain is

relative smooth (not steep). The figure above shows the WAsP minus the WAsP CFD calculated wind speeds.

In the low valleys, WAsP calculates up to 1.5 m/s higher wind speeds than WAsP-CFD. When the terrain gets

complex (steepness > 30%), the differences can be even larger.

Resource maps

This section describes a wind resource map calculation. The tool can calculate a wind resource map based on

• WAsP model.

• WAsP-CFD result files.

• the Scaler, which can use WAsP, WAsP-CFD and .flowres files from third-party CFD flow models.

Resource maps can also be

• rescaled based on one or more measurements within the resource area.

• be downloaded from menu “Data”, here .siteres files from the GASP project are available globally.

The module can also produce a resource map report based on external files, e.g., from other CFD calculation

tools, delivered as .RSF or .WRG files.

3.6.5.1 Resource map calculation based on wind statistic(s)

The resource map calculation can utilize multiple cores on the PC and even get “support” from other computers

connected to the network and, thereby, process large resource maps with high resolution within a reasonable

time.

Input:

• Site data object with wind statistics, roughness data and orography (link to files/line objects/grid

objects)

• ALTERNATIVE to Site data object is WAsP-CFD result file(s) + wind statistics(s)

Optional:

• Obstacles digitized on the map

• Displacement height calculation setup

Models/modules for initial calculations 92

• RIX calculation and RIX correction setup

Output:

• Resource map report

• Resource map file (can be used in PARK and OPTIMIZE module as input for AEP calculation and

visualized by RESULT layer on map). The format is .RSF, the native WAsP format OR the EMD

defined .siteres format.

Figure 106 Purpose in Site data object set for Resource map calculation

Within the site data object, set for purpose “RESGEN”, there will be a few differences from previous

described input to site data object:

Figure 107 Select wind statistics(s) for resource map calculation

When choosing more wind statistics (selected by holding <CRTL> down), these can be distance weighted

within each calculation point. A maximum weight can be entered, e.g., 80% and, thereby, non-logical

"high/low spots” just at the wind statistics point can be avoided.

Distance weights can also be given manually (fixed at any calculation point) by unchecking “Update…”.The

terrain specification follows previous descriptions. For a resource map calculation, roughness roses are NOT

possible no matter which WAsP version is used.

Figure 108 Resource map calculation area

Models/modules for initial calculations 93

The resource calculation area can be defined by a rectangle at the site data object by manually entering the

coordinates, drawn out on the map, or by use of a WTG-area object.

Having the site data object defined, the calculation module can then be activated:

Figure 109 Resource map calculation options

The calculation variants are basically to either perform the WAsP based calculation or to utilize already calculated

data from WAsP-CFD or other providers, that delivers resource map files in WAsP format. The siteres format

developed by EMD can in addition to wind resource parameters also include parameters with relevance for e.g.,

load calculations, like TI, inflow angle etc. This requires the Site Compliance module to be licensed. Please note

that calculations can take a long time to complete if this option is selected as Site Compliance tests have to be

run for all grid points. The larger the area and the finer the grid, the longer this will take. We recommend starting

with a coarse grid to test the speed of the calculation and refine it as required.

Models/modules for initial calculations 94

Figure 110 Resource calculation input options

The figure above shows how the resource specification of hub heights (multiple possible) and resolution are

defined.

The calculation area can be defined in the site data object (square) or it can be defined by one or more a WTG-

area objects (freely digitized). The last option requires the box “Override area…” to be checked. The Parameter

to show on map and/or in a result layer can be defined along with formatting the legend before the calculation.

These choices can always be changed after calculation.

Models/modules for initial calculations 95

Figure 111 Input of RIX correction handling in resource map calculation

RIX correction can also be included (see Section 4.10 on “RIX correction” for details). When RIX

calculation/correction is chosen, there will be (optional) 4 result layers:

Figure 112 Result layer output from Resource map calculation with RIX

The uncorrected resource map as well as the RIX and Delta Rix maps can be seen along with the RIX corrected

resource map. By establishing a compare layer, the difference between the Rix corrected and uncorrected

resource maps can be presented.

Figure 113 Displacement height input by resource map calculation

Finally, displacement height corrections can be included in the resource map calculations (see details in Section

4.9, “Displacement height calculator”).

Models/modules for initial calculations 96

Figure 114 Example of resource map difference output with displacement height

Above, a test calculation with two forest pieces (hatched) with different heights are shown in a resource

calculation. Then, a compare layer is established showing the wind speed without and with a displacement height

calculation. It works as expected, with a gradual reduction as the distance to the forest increases.

3.6.5.2 Resource map calculation based on SCALER

A resource map can be based on the SCALER concept, meaning that each time step wind speed is scaled to

each point in the resource grid. The data are collected in TAB files at each grid point, which are Weibull fitted

and then saved in .RSF file.

Input:

• WTG-area object for defining the calculation area

• Height(s) and grid resolution

• SCALER, including wind data selection (more wind masts/mesoscale data points can be used with

interpolation), terrain/model and eventually post scaling.

Optional:

• Obstacles digitized on the map

• Displacement height calculation setup

• RIX calculation and RIX correction setup

Output:

• Resource map report

• Resource map file (can be used in PARK and OPTIMIZE module as input for AEP calculation and

visualized by RESULT layer on map). The format is .RSF, the native WAsP format OR the EMD

defined .siteres format.

3.6.5.3 Resource map rescaling

Since windPRO 3.5 is has been possible to rescaling an existing resource map.

Models/modules for initial calculations 97

This can be convenient for e.g., large offshore sites with a mesoscale wind speed gradient or if an external flow

modeling is available as a resource map, but based on a preliminary wind speed assumption. Then the wind

speed level can be adjusted without having to rerun the flow model after local measurements have become

available, for example.

Figure 115 Input data for rescaling a resource map.

First a resource map is selected, in either a .rsf, .wrg or .siteres format.

Next, the meteo object(s) are selected. Since the rescaling feature uses the frequency table, the meteo object

must have the same number of sectors as the resource map. The number of sectors can be changed in the

meteo object. Only one height can be used in a rescaling, but multiple meteo objects with different positions can

be used. When selecting one height all non-matching heights are greyed out.

If multiple meteo objects are selected, a spatial (horizontal) interpolation method can be activated. A special

option is to include elevation difference in distance weighting. Thereby masts elevated closer to the calculation

point are given more weight:

As an additional option, the elevation can be used for further adjustment of the resulting resource map. This can

compensate for flow models tending to underestimate wind speeds at higher elevated areas compared to lower

elevated areas of the site. Note that if the site has a steep terrain causing flow separation, then this can go in

the other direction, that the flow model overestimates the highly elevated spots. For such sites, this option is not

recommended. But in normal hilly terrain multiple cases show 0.3% flow model bias, so the default is rather

conservative. An upper limit can be set, just remember to increase this to get the full effect of the adjustment, if

it seems relevant to use with more measurements on a site. See validation chapter 3.10.5 Elevation model

pitfalls.

The Neighbor sector option is default set to 1, meaning that when the model and observations are used to find

the adjustments for a given direction sector, it looks in one neighbor sector to each side to come up with the

revised frequency for the sector. Then a turn of the wind will be smoothed. At sites with much turn and a high

sector resolution, like 10-degree sectors, it might be relevant to expand using two neighbor sectors, but normally

one neighbor sector will be the best choice.

Models/modules for initial calculations 98

With the choices selected, the recalibration is performed with <OK>, and a rescaled/calibrated resource map will

appear as result layer. This can be used as input in a PARK calculation.

STATGEN

The STATGEN module generates a wind statistics based on wind data in a METEO object and terrain data with

the WAsP model or WAsP CFD result files as the generator. Also, mesoscale data can be used. Here although

the mesoscale model terrain should be used. This is possible when using EMD-WRF data, while these includes

mesoscale model terrain files in the meteo object with mesoscale wind data.

Input:

• METEO object with wind data

• Site data object with roughness data and orography (link to files/line objects/grid objects)

• ALTERNATIVE_1 to Site data object is WAsP-CFD result file(s)

• ALTERNATIVE_2 to Site data object is mesoscale terrain from EMD-WRF data in meteo objects

• Data period to be used

Optional:

• Obstacles digitized on the map (Can be turned on/off in site data object)

• Displacement height based on object data OR advanced calculator (sector wise)

Output:

• Wind statistics (.LIB(WAsP original format) or .WWS (windPRO format) file)

Figure 116 Main input for STATGEN calculation

In the main form, the “terrain source” is chosen. In a site data object, the traditional WAsP input of roughness

and orography + info on inclusion of obstacles is defined. WAsP-CFD result files include, indirectly, the terrain

(results are processed with terrain). Making a wind statistics directly from mesoscale modeled wind data

(currently, only from EMD) offers the possibility to use the mesoscale terrain when generating the wind statistics.

Models/modules for initial calculations 99

Figure 117 Statgen input form

On the “statgen” tab, the site data or CFD results or mesoscale data are selected as well as the wind data to be

used.

The WAsP parameters can be modified (see Section 6.3.1 for details).

Saving the results as a .wws file, all relevant information will be included, like length of wind data series, WAsP

parameters etc. This information can be printed as a part of all WAsP based calculation reports (see example

below):

A new 3.1 feature is the selection of a specific period. Thereby measurements can be truncated to full years, or

for mesoscale data it can be decided to use like last 20 or 10 years. In addition, the inclusion of the general

Displacement height calculation tool, optional sector wise by calculator is new in 3.1.

Models/modules for initial calculations 100

Figure 118 Wind statistics info report output generated from MCP

A wind statistic can also be generated from MCP, which normally are recommended, while this includes the long

term correction.

See further information in Section 3.2, “wind statistics,” for more details on the structure of the wind statistics.

Flow request export (FLOWREQUEST - FLOWRES format)

A new windPRO 3.1 feature is a flexible data exchange with external models.

It consist partly of creation of a flow request file, which basically is an export of the terrain data lined up in the

windPRO project, partly of the capability to use flow model results from external model providers. The flow result

file format FLOWRES is detailed described so any model providers can establish this format as output from their

models. Then windPRO can use the flow result files in relevant calculations, like Resource map calculation,

PARK calculation etc.

Models/modules for initial calculations 101

Figure 119 Flow request export

Figure 120 Terrain setup for flow request files.

The terrain (roughness and orography) need to be specified in a Site data object.

Forest data can be included based on roughness map files. The Roughness lengths interval that shall generate

forest data output must be specified. Then forest height and forest density can be specified.

Figure 121 Definition of the Result volume.

Here the volume to be calculated is specified. Partly by horizontal limits, given either by a WTG area object or

by object layer. Last option simply draws an area around the objects at selected layer. Partly by extraction

heights. Depending on which model that uses the flow request file, there might be limitations in calculation

heights. And some models might not read this information but require manual input by the model run.

PARK calculation 102

Figure 122 Definition of direction sectors to be simulated.

Finally the number of, and which direction sectors that model calculations are wanted for can be defined.

When all is defined, the “Export to flow request” button is pressed. Then a name and where to save it must be

specified.

If OK is pressed, an assumption report is generated for documentation of the settings.

Next step will then be to start the external software, read the file and start a calculation.

When calculation is done, a FLOWRES file can be exported from the external software.

This can then be used from windPRO similar to where like WASP-CFD result files can be used, e.g., from a

PARK calculation.

PARK calculation

The PARK calculation is the “center” within an energy calculation. It partly models the wind distribution at each

turbine position in hub height and partly the wake losses, thereby creating the calculated AEP for each turbine.

Further refinement of the AEP expectations can be given by taking a PARK calculation result into the Loss &

Uncertainty module.

Figure 123 PARK calculation, selection of method

Above is the start dialog box of the PARK module. 6 main groups of setting up a PARK calculation are available.

External wake models and Blockage are described in document available from link. From 3.4 the WakeBlaster

model is available as external model, only within the time series concepts so far.

PARK calculation 103

Additional options are available under Other PARK calculations, such as:

Figure 124 2.9 compatible PARK model choices

The windPRO 2.9 options are kept for backward compatibility - but can also still be relevant (though these options

are rarer, nowadays). These will not be documented in this manual, but documentation can be found in the

windPRO 2.9 manual.

The wake loss (PARK) models

The PARK module calculates the wake losses due to the shadowing effects between WTGs sited close to each

other (wind farms or clusters). PARK offers more different models for calculating wake losses, including Blockage

as option, which reduces wind speeds upwind.

Wake model overview

Statistical

Time

step

Blockage

Note

Original N.O. Jensen (PARK1)

Long time usage

Improved N.O. Jensen (PARK2)

Recommended

EMD variant: NO2005

With special tuning for e.g., single

row wind farms

Ainslie 1988 with DAC

New in windPRO 3.5

WakeBlaster (external model)

Advanced flow modelling

TurbOPark

Offshore

Outdated: (to be removed)

EWTS II (Larsen) 1999

Not good for large wind farms

EWTS II (Larsen) 2008

Not good for large wind farms

Ainslie 1986

Outdated implementation

From windPRO 3.6 the TI input to the wake models (and thereby also WDC for N.O. Jensen models), can now

be hub height dependent, when letting the terrain type decide TI. This can be a large advantage for e.g.,

calculating repowering projects, where hub heights can differ much for the turbines in a calculations, and thereby

also the TI.

Since windPRO 3.6 is also possible to use two time step variants for WakeBlaster: One, where a number of

scenarios is calculated and used for a lookup of the result each time step. Another, where the model calculates

for each time step, and thereby can take the individual turbulence by time step into account, see section 3.7.13

PARK with WakeBlaster, external wake model.

From windPRO 3.5 a fully new Ainslie 1988 with Deep Array Correction (DAC) has been implemented. All non-

outdated models run as well based on statistical wind distributions as time series wind. For all models except

the external WakeBlaster model, a Blockage (induction) model can be chosen to run along with the wake models.

The present scientific state of the art blockage models is implemented (Forsting and Branlard). However, these

models do not take turbulence or stability into account, and we must admit that they do not bring any significant

improvements to the wake modelling. The models calculate typically an added loss of 0.5% for larger wind farms.

This is probably in a right size order seen over a year, but at low TI or stable wind, the actual blockage is probably

higher. But these conditions typically only appear in a smaller fraction of a year.

In relation to testing the new Ainslie/DAC model a more comprehensive wake modelling test has been performed.

This gives good confidence that even large offshore areas can be modelled quite precisely. See Validation

PARK calculation 104

Chapter 3.9. The new Ainslie1988 can be run without the DAC model, but this will not work for larger wind farms

as the Ainslie model itself only captures near wakes.

The Turbulence Intensity (TI) which correlates well with stability, is a very deciding parameter in wake modeling.

This is used as direct input for Ainslie and WakeBlaster, and indirectly for N.O. Jensen models via the Wake

Decay Constant (WDC). Over many years, EMD has tested wake models on operating wind farms with high

focus on finding a relation between WDC and TI that work best.

Due to the high importance of TI, an option for scaling the TI is made available in PARK. This should e.g., be

used for: EMD-WRF Europe+ and similar datasets, which is seen to have a factor 1.41 too low TI – at least

offshore, but also seen for several onshore sites. This is not the case for the former EMD ConWx mesoscale

dataset or the present EMD-WRF On Demand data. For onshore sites, the new EMD-WRF Europe+ dataset

has shown to perform better regarding TI than the previous setup based on tests against more than 200 masts

onshore. Therefore, TI is not corrected with√2=1,41, as done in EMD ConWx data. However, onshore TI is in

general very uncertain in mesoscale model data.

It is therefore preferred to use the measured TI on site or similar site to find a scaling factor for the mesoscale

TI. This scaling factor can then be applied when running an onshore PARK calculation.

Alternatively use the DTU recommendations in table below, which performs well on average, but do not capture

deviations for less typical sites.

Table 6 Recommended settings for N.O. Jensen PARK models

DTU recommendations:

EMD recommendations from windPRO 3.6 :

N.O. Jensen (PARK1)

PARK2

PARK1

PARK2

Offshore and low TI onshore*)

0.05

0.06

WDC = TI x 0.67

WDC = TI x 0.8

Onshore

0.075

0.09

WDC = TI x 0.5

WDC = TI x 0.6

Advanced offshore low TI

WDC = 2 x TI -0.07

It should be noted that DTU previously recommended 0.04 for offshore sites with PARK1.

From windPRO 3.6 the recommendations on the factor on onshore TI has been increased from 0.4 to 0.5

for PARK1 and from 0.48 to 0.6 for PARK2! This partly removes the previous inconsistencies in the transition

zone between on- and offshore, partly it is observed in more validation studies, that the new values has a better

match. It is important to mention that there not is one true value for all sites. The above is based on what we

have experienced work best for most sites tested.

The Offshore TI is illustrated below by examples:

Figure 125 Offshore TI, formulas, and examples of measurements

PARK calculation 105

EMD has collected TI measurements (5-15 m/s where wake loss appears), from multiple different offshore sites

to illustrate which “band” of TI can be expected offshore.

The formula, that give a rough idea about TI, is illustrated in the figure above. This is not that useful, when it

comes to offshore locations, as it is extreme sensitive to the roughness length, and what is this offshore? Zo =

0.0002 is normally used, while this can be extrapolated from shear measurements. But also Zo= 0 is seen used

in other contexts. If 0 is entered in the formula, a division by 0 is seen, and this will not work. But lowering to Zo=

0.000001, shows that the TI is lowered much. The sensitivity to roughness lengths offshore in the TI formula is

probably the reason why the fully formula-based concept with WDC = 0.4 x TI (see TI formula below); introduced

in windPRO 3.2, does not work well for offshore locations. Stability although also plays a role.

The measurements shown in the graphic above, indicates partly the “band” TI is within, but also how it changes

with hub height.

For new 8+ MW offshore turbines hub heights 100-120 m will be seen. Here the TI could probably vary from

5.5% to 7.5%, depending on specific location.

Our recommendation for a simplified approach will be to go with 6% or 7.5% TI for low-high TI offshore sites and

thereby the mentioned WDC values for the two PARK variants shown in the table below:

The recommendations for offshore leads to these examples for comparison to DTU recommendations:

PARK-1

PARK-2

PARK-1

PARK-2

High TI = 7.50%

Low TI = 6.00%

Factor

0.67

0.8

0.67

0.8

WDC

0.050

0.060

0.040

0.048

The DTU recommendations were previously 0.04 for PARK1, today 0.05 for PARK1 – these correspond to the

two low and high TI examples in table above. For PARK2 the DTU recommendation 0.06 correspond to the high

TI site.

e.g.,

TI will depend on the site. For rough estimates is used:

TI = A*k/ln(h/z

)

Where;

A = 2.5

k=0.4

h = calculation height

= roughness length

The chosen constants are primary based on this paper:

http://orbit.dtu.dk/files/122284235/On_the_application_of_the_Jensen_wake_model.pdf

Citation (APA):

Pena Diaz, A., Réthoré, P-E., & van der Laan, P. (2016). On the application of the Jensen wake model using a

turbulence-dependent wake decay coefficient: the Sexbierum case. Wind Energy, 19, 763–776. DOI:

10.1002/we.1863

From this part of the conclusion:

Figure 126 N.O. Jensen model better performing than more advanced models.

Table 7 Basic assumptions for hub height dependent WDC with examples for PARK2

Basic input for WDC

Calc. height:(m)

100

Terrain type

Rou.Class

Rou. Length

WDC

PARK calculation 106

Very stable

-1,4

0,0000002

0,051

0,041

0,050

0,040

Offshore (lower

TI)

0,00001

0,065

0,052

0,062

0,050

Offshore

0,0002

0,080

0,064

0,076

0,061

Offshore (higher

TI)

0,5

0,0024

0,101

0,081

0,094

0,075

Very open

1,0

0,03

0,13

0,081

0,12

0,074

Open

1,5

0,06

0,15

0,088

0,13

0,080

Mixed farmland

2,0

0,11

0,16

0,098

0,15

0,088

Closed

2,5

0,20

0,18

0,109

0,16

0,097

Very closed

3,0

0,39

0,21

0,123

0,18

0,108

Dense forest

3,50

0,74

0,24

0,142

0,20

0,122

The table illustrates how the formula-based TI is calculated for two different hub heights, 50 and 100 m. The

corresponding WDC is simply TI x 0.8 for offshore, TI x 0,6 for onshore.

Conversion from roughness class to length is derived from the simple table below as simple linear relations in a

logarithmic plot. Note there are two linear relations, one below class 1 and one above.

Table 8 Roughness class and length relations

Class

Length

0,0002

0,03

0,1

0,4

Figure 127 The WDC recommendations by roughness class for 100m hub height.

5,0%

6,2%

7,6%

9,4%

12,3%

13,3%

14,6%

16,1%

18,0%

0,020

0,040

0,060

0,080

0,100

0,120

0,140

0,160

0,180

0,200

-1,4 0,0 0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5

TI WDC

Rou.Class

PARK2 WDCs and TI @ 100m HH

TI WDC 100m DTU recommend

PARK calculation 107

Note the recommendations are related to PARK-2. For N.O. Jensen original model. the WDC must be divided

by 1.2.

A new feature from windPRO 3.6 is to allow the ambient turbulence based on terrain to directly control the WDC

by hub height. Since windPRO 3.3, it has been possible in time step calculations to make the WDC hub height

dependent by using a time series TI signal.

For the ones who like the more core science behind the new WDC recommendations based on TI, look into this:

Figure 128 WDC calculated from TI based on theory.

Note the TI to WDC converter here are found to ~0.4, PARK1 related (0.48 for PARK2). Based on numerous

validation cases this constant is found to be slightly too small.

PARK calculation 108

3.7.1.1 The PARK 1 & 2 implementations

PARK1 or “Original N.O. Jensen model” has been part of windPRO from the very beginning, is regularly checked

against the similar implementation in WAsP. This is often referred to as the “industry standard”. But when DTU

made PARK2, this should replace PARK1 as it is seen to perform slightly better and is based on more correct

physics.

PARK2 model has some formula revision, see;

http://orbit.dtu.dk/files/139682596/PARK2Validation_Poster.pdf

The main issue is that the combination model is changed from Root Sum Square wind speed deficit summation

to linear summation, while mirror wakes have been removed, but also other formula revisions. Due to these

changes, a higher WDC must be used to compensate for this.

The Wake Decay Constant (WDC) is the main parameter for the PARK 1 & 2 models, already comprehensively

described in previous chapters.

3.7.1.2 The Ainslie/DAC implementation

Ainslie 1986 was implemented in windPRO in 2005, but it was seen that this model did not perform well for larger

wind farms and thereby there was little focus on this model. In 2021, based on user requests, the model was re-

implemented, now based on the Ainslie 1988 improved model descriptions. But this still does not handle large

wind farms particularly well, with calculated wake losses far too low.

Other software packages have therefore introduced deep array correction models to be used along with Ainslie

model. These models basically increase the roughness to compensate for the too low calculated wake losses

when the windfarms are larger (more than ~20 WTGs).

EMD has implemented its own Deep Array correction (DAC) model based on available scientific papers. This is

not fully like other deep array correction models, but comprehensively tested and found to work well for a number

of wind farm configurations, see validation chapter 3.9.

The new Ainslie 1988 with DAC calculates wake loss in same size order to the N.O. Jensen and WakeBlaster

concepts. However, it is less sensitive to TI changes as it does not increase the calculated wake losses enough

when the TI is low. Additionally, the model is too conservative for the new 8MW+ turbine generation (it calculates

too high wake losses).

For Ainslie/DAC, there are several parameters:

The main input for Ainslie model is the TI, selected similar to the WDC for N.O. Jensen models. In addition, there

are several parameters, where defaults for onshore and offshore is established as simple selections. The most

PARK calculation   109 
 
© EMD International A/S   •    www.emd.dk   •   windPRO 4.0  •   April 2024 
important parameter is the DAC roughness settings. These have been comprehensively tested for offshore. For 
onshore  projects,  only  Roughness  Class  1  parameter  suggestions  are  given  so  far.  With  user  defined  it  is 
possible to define freely the parameter settings. See validation chapter 3.9 for more details. 
 
 
3.7.1.3  The NO2005 implementation 
EMD has implemented the N.O. Jensen model in a variant named NO2005. Comprehensive tests performed 
2016 has revealed a difference in the NO2005 implementation compared to Original N.O. Jensen model. It has 
been known from the first tests, that NO2005 calculated slightly lower wake loses. see e.g., 
https://www.emd.dk/files/PSO%20projekt%205899.pdf 
 
But with increasing wind farm sizes, the deviation became larger, see 
https://help.emd.dk/knowledgebase/content/TechNotes/TechNote_5_Park%20model%20revision.pdf 
The solution pinpointed is to use 35% linear weight in the NO2005. 
 
From windPRO 3.2 the Original N.O. Jensen model (PARK1), as well as PARK2 can be used in the time step 
based calculation method. And these can be used with combined Linear and RSS weight. 
 
The NO2005 model is kept as an alternative, while this has the experimental tuning: Change WDC by number 
of upwind turbines. This feature is still relevant for post construction analyses, where there is good data to fine 
tune wake model settings. 
 
3.7.1.4  The WakeBlaster implementation 
WakeBlaster as external model is an interesting alternative, while the calculation method differs from as well the 
N.O. Jensen as the Ainslie concept. This model does although calculate wake losses in similar size order as the 
N.O.  Jensen  models  but  can  have some  deviations  based  on  the  wind farm  layout.  See  3.7.13  PARK  with 
WakeBlaster, external wake model for a detailed walk through a WakeBlaster calculation setup. 
 
From windPRO 3.6 it is possible to use WakeBlaster in real time step calculations, where the TI input vary by 
time step. 
 
 
3.7.1.5  The TurbOPark implementation 
TurbOPark is a modified Park wake model developed by Ørsted with the source code publicly available here: 
https://github.com/OrstedRD/TurbOPark. This model has been implemented natively in windPRO 4.0. 
 
EMD has verified that the implementation in windPRO yield the same result as provided in the above github 
repository by recalculating their examples. The example includes a varying background wind speed and a regular 
grid of 16 turbines of two different turbine types. 
See also: https://help.emd.dk/knowledgebase/content/ReferenceManual/Wake_Model.pdf  
 
 
3.7.1.6  Validation reports on deep array modelling and recommendations 
Below conclusions from 
http://iopscience.iop.org/article/10.1088/1742-6596/753/3/032020/pdf 
 

PARK calculation 110

In general, the predictions of the simple wake model we have tested are in good agreement with the

observations. However, the usefulness of the model for large offshore wind farms has been put into question by

prior assertions that the model systematically underestimates the wake losses inside large wind farms. The

existence of such a ‘deep array effect’ would imply that the model was insufficient or needed corrections. In this

study, we find no evidence of a systematic deep array effect, despite comparing the model with observations

along a row of 26 turbines! This matches the conclusion of previous research on other large offshore arrays [5].

When comparing the Nysted wake losses before and after Rødsand II, we find that the additional wake loss from

the neighboring wind farm is roughly confined to the first few rows in the downstream wind farm.

See also: http://iopscience.iop.org/article/10.1088/1742-6596/524/1/012162/pdf

As seen, some of the probably most comprehensive studies on wake losses show that the original N.O. Jensen

model handles wake loss calculation well, also for large wind farms.

In the last reference, it is mentioned that in one section tested in the London Array windfarm, the N.O. Jensen

model under predict wake losses for the back rows. But here is also noted that the Turbulence is very low for

the sample data. This is what EMD has seen for an Egypt wind farm, that when turbulence is low, wake losses

increase radically. This can be captured in model calculation by decreasing the WDC.

EMD thereby recommend the (N.O. Jensen model) as preferred for pre-construction calculations. PARK2

seem slightly better than PARK1 (org. N.O. Jensen), and PARK2 should be the preferred choice. Pay

attention to the turbulence and make the WDC choice based on this as previous described – this is the

key parameter to make the model calculate correct. The best way to include turbulence is in the time

step-based calculation with WDC as a function of TI on time step basis. Also pay attention to the Ct

curves. It can be hard to judge if these are correct, but comparisons to other turbine models might give

a hint if they seem realistic.

For single row projects, pay specific attention, here the wake loss is calculated too high with normal wake models,

while they do not consider the added fresh wind from both sides of the single row. See validation chapter 3.9.

3.7.1.7 Park rotation by coordinate system

PARK and WAsP calculations requires rectangular coordinate systems. This mean that the park will be slightly

rotated the more the given rectangular coordinate system is rotated relative to geographic north. Some

coordinate systems have quite large rotations. It is therefore decided that before running the calculation,

windPRO convert all data to UTM WGS84 system, and chose the UTM zone in which the Site centre is located.

Thereby rotation is minimized to ~+/- 3 degrees in worst case, which appear when the wind farm is located just

at the border between two UTM zones. What can be wrong by the rotation is that wind data typically are aligned,

so 0 degree is geographic north. Then the wind and the array orientations do not match fully. It is a marginal

problem with the chosen solution in windPRO. The user can although compensate the problem by adding an

offset when importing the wind data. While the accuracy of the wind data directional alignment rarely justifies

this, and the effect on the result is marginal, it is not a general recommendation. The used coordinate system

and zone as well as the resulting rotation is shown on the PARK result main page. Thereby it is documented

how slightly different positions of site center (in different UTM zones), is the reason for slightly different results.

PARK calculation 111

Curtailment in PARK calculations

From windPRO 3.3 most curtailment calculations are included in PARK. Previously, only sector management

was available in PARK with shutdown. In windPRO 3.3 prioritized curtailment rules can be defined shutting down

a turbine or changing operation mode. The advantage of handling curtailments in PARK instead of LOSS &

UNCERTAINTY is that the changes in wake losses will be handled. This can be of some importance especially

with sector management, where e.g., every second turbine is stopped, when the wind comes along the row, just

the situation where the wake losses appear.

All curtailments are defined on the individual WTG objects in a project. Thus, the curtailments applied to a PARK

calculation is dependent on the configuration of the WTGs included in the calculation.

To activate curtailments in a PARK calculation simply check the Use curtailment checkbox in the Setup tab:

Each WTG can contain multiple curtailments like noise, bat and bird curtailments. In case two or more curtailment

rules are valid at the same time, PARK will evaluate the rule with the highest priority first and attribute any

production loss to this rule. Up until windPRO 3.6, only the first valid rule would be executed and attributed the

loss for the entire timestep. In windPRO 4.0, multiple rules can be valid at the same time. Now, the next priority

rule will then be evaluated, and if there is a further subsequent loss caused by rule number two, then the

additional lost production is attributed to rule number two.

To use the old way of only allowing one curtailment rule to be activated per time step, simply check the checkbox

“Allow only on curtailment per time step”

Any PARK calculations including curtailments created before windPRO 4.0 have this setting enabled by default.

Any curtailment rules define on the WTG objects included in the PARK calculation will appear in the Curtailment

tab. Although curtailment rules are defined on the individual WTG objects they can also be edited directly from

PARK by checking the “[x] Allow editing” box. This turns the “View” button into “Edit”.

Time-based PARK can utilize all the curtailment options in the WTG object, whereas statistics-based PARK can

only use curtailment rules defined by wind speed and wind direction.

PARK calculation 112

Figure 129 Curtailment settings in PARK

Curtailment rules can also be defined on multiple turbines when multi-editing multiple turbines. Only identical

curtailment rules shared by all selected turbines can be edited in batch. New identical curtailment rules can

always be added to the selected turbines in multi-edit:

It is possible to view all rules for the selected turbines by selecting “Show for all WTGs”, though the rules cannot

be multi-edited from here:

PARK calculation 113

Here, curtailment rules can be imported and exported to and from the clipboard with the “Import” and “Export”

buttons:

The “Paste” requires that the content in clipboard follow the input data format. As a help, a template can be

copied. The template shows the import format and the available signal abbreviations.

WTG1

Wind sector management

Mode3

Wdir

165

185

WTG2

Wind sector management

Mode3

Wdir

345

Data in Excel for paste into the “Import data for curtailment”

PARK calculation 114

Figure 130 Import curtailment data to WTG objects

Here a WTG_ID is required in the first data column. This can be coordinates, Description or User label.

When importing from clipboard, all existing rules are cleared before the new rules are imported. The priority of

the rules for each turbine is determined by the order they are imported. The priority number dictates in which

order curtailment rules are evaluated. Once the conditions of a rule are met, the curtailment is applied, and no

further rules are applied in the time step.

Always remember that changing the settings on the WTG objects can influence other calculations using the

same objects.

The result of a calculation with curtailment will partly be presented with curtailment loss in PARK report, partly

as a new column in Result to file. And it will be added in Loss&Uncertainty, where it will get the same status as

Wake losses. Where losses normally are combined as (1-Loss1) x (1-loss2), this will not be so for Wake and

curtailment losses, while these in the PARK calculation already are combined. So these will simply be added

before combined with other losses.

Note the added column in result to file might require update of like Excel templates and external software tools

that utilize Result to file. The curtailment column will always be included.

3.7.2.1 Temperature derating

Some turbines located in high altitude and/or high temperature settings may need to be derated depending on

their cooling equipment. Manufactures typically issue a dataset specifying at which points the turbine’s maximum

power output is limited from rated power. In windPRO 4.0 it is possible enter these temperature curves into the

wind turbine catalogue. See BASIS manual section 2.6.4.6 on how to enter the data.

windPRO will interpolate between specified temperatures, but elevation is considered discrete. The elevation

used to determine if a turbine is within the specified elevation envelope is TIN elevation + hub height.

Any turbine model which has a temperature curve defined in the turbine catalogue can be temperature derating

in a time-varying PARK calculation. If a turbine model does not have a temperature curve defined, it will not be

derated. To activate temperature derating in PARK, enable curtailments and go to the Curtailments tab and

Enable Temperature Derating:

PARK calculation 115

Once activated, all WTG objects included in the PARK calculation which has a temperature curve defined in the

turbine catalogue will be derated once the turbine is operating within the specified temperature/elevation ranges.

The loss is applied to the turbine before any other curtailment losses.

Similar to all other curtailments, you need to select which signal should trigger the curtailment:

The losses from temperature derating can be seen in the Production Analysis report and as an aggregate

curtailment loss in the Main report. In the Curtailment Assumptions report it is possible to see how many turbines

have been affected by the temperature derating.

3.7.2.2 Grid curtailment

A special curtailment is grid limits, as this is not applied on specific WTGs but on the windfarm as a whole.

Therefore, this is treated separately and entered in setup form. The grid curtailment is only available for time

varying calculations due to the nature of the loss.

Figure 131 Entering grid curtailment settings.

The calculated production in PARK occurs before any losses in the collector system. Conversely, the limitation

point is typically defined after the loss in the collector system. Therefore, peak grid loss can be added to the limit,

enabling WTGs to compensate for the loss in collector system.

For any given time step the total park production (after adjustment for the electrical loss at peak load) is compared

to the defined grid limitation. Any exceeding production will be subtracted proportionally from all “New WTGs”

and all “Existing WTGs” which are “Treat as Park WTG” .

Existing WTGs which are not treated as Park WTGs will not be part of the grid limit calculation, as these are

assumed to have separate grid access point.

As such, it is possible to control which WTGs shall be included in the grid curtailment calculation. Note the multi-

edit tool can activate/deactivate the “Treat as park WTG” for existing WTGs.

The grid curtailment is reported as a separate, independent loss after the calculation of wakes and other WTG

specific curtailments.

PARK calculation 116

Common settings for all PARK calculation variants

3.7.3.1 Hub height dependent TI based on terrain selection

Figure 132 New 3.6 method for automatic TI calculation by hub height

To the left, a calculation made before ver. 3.6 can be converted to the 3.6 method (right). This has two

advantages:

1) The drop-down list no longer has an entry for each hub height and terrain class. Just the terrain class.

This makes selection easier.

2) The calculation automatically uses the hub height for each WTG to calculate the TI based on terrain

class. Thus, it uses a turbine specific TI (instead of an average based on the selection in the dropdown

list). This means that a project which was first calculated with 100m WTGs, then cloned and calculated

with 159m hub height will automatically change the TI input. Previously, this had to be manually

reselected.

If the calculation contains WTGs with different hub heights, each WTG use the TI as input for this WTG,

which again decides how much wake this WTG gives to its neighbours.

Figure 133 Input of sector defined TI also utilizes the new concept.

When the input is entered by sector, the hub height dependent TI values will also be used if terrain is selected

under “Terrain class”. If “User defined” is selected, then the input values (roughness length, TI or WDC) will be

used “as is”, as the automatic TI calculation only works when based on a terrain class.

3.7.3.2 Use Curtailment

Curtailments can be included in PARK calculation. This mean that if a turbine is reduced or stopped by specific

wind speeds and/or directions (sector management), this will be considered in the wake loss calculation. For

more refined curtailments, like BAT stop a specific period, this can only be handled in time step-based

calculations. The sector management variant is included in all PARK variants (except the 2.9 variants and

external models) from 3.2. See Section 3.7.2 Curtailment in PARK calculations for more details.

PARK calculation 117

3.7.3.3 Use Blockage

Blockage, or induction, calculates the wind speed reduction caused by the wind farm upwind. There are two

scientific models implemented:

Figure 134 Forsting blockage model.

Figure 135 Branlard blockage model.

Parameter adjustments are for expert users that know the model background and wish to test model settings or

have special knowledge.

Blockage reduction will be part of the calculated wake loss. To see the blockage reduction, it is necessary to run

two calculations with and without and subtract results. Due to the very small impact, it is not judged worth to

perform a double calculation to be able to include the blockage loss separately in the reporting.

PARK calculation 118

3.7.3.4 WTGs

Figure 136 Selection of WTGs for PARK calculation

At the WTGs tab, the turbines to be calculated are selected. By default, the turbines from visible layers appear,

but other layers can be activated and, by unchecking “use all objects from selected layers”, turbines can be

individually selected.

Checking the “Use explicit link to Wind distribution for each WTG” adds another tab “Link site data to WTG”:

PARK calculation 119

Figure 137 Explicit link site data – WTG, manual control

Before activating this option, wind distributions must be chosen. By default, each WTG “chooses” the nearest

wind distribution object. In the matrix above, checkmarks can be moved so the user has full control as to which

wind distribution object shall be used for which WTG.

For Scaler calculations it is from windPRO 3.3 possible to link WTGs to specific masts.

When selecting the rightmost option, a new tab “link WTG to mast” appear.

Here it is even possible to link a WTG to more masts, and the full flexibility to use more masts for some WTGs

and only one for other WTGs exist. The Scaler basically scale the selected masts to the given WTG and distance

weight the result at each WTG position. The distance weight is based on the inverse squared distance.

PARK calculation 120

Common settings for Wind statistics based (standard) PARK calculation

3.7.4.1 Setup

Figure 138 Model parameters and report features in “standard” PARK

Model parameters:

“Wake decay constants…” defines how much the wake expands behind the rotor. The example figure 0.09

means that, per meter behind the rotor, the wake expands 9 cm. In “non-advanced mode” the choices are:

Further details can be found in “The wake loss (PARK) model” (Section 7.1), where it is emphasized that, at

larger hub heights, the WDC should be reduced due to lower turbulence.

A lower expansion (offshore) gives a higher reduction of the wind speed behind the rotor and, thereby, larger

calculated wake losses. It is recommended to enable “advanced”, which gives more flexibility.

Report features:

Hub height for key results gives calculated mean wind speed, energy and the equivalent roughness class for

this height at the location where the Site Data Object is placed.

WTG area(s) on map gives the option to include selected WTG Area’s on the map as part of the report.

Handling of losses and uncertainties:

Here, it is decided how the user wants to integrate these variables into the reporting. It is recommended to use

the Loss & Uncertainty module to give this important part of the AEP calculation the right focus. For “simpler” or

preliminary calculations, the other options can be used. The “Add simple reduction” to compensate for bias, loss

and uncertainty reductions will add an extra column in the report with the reduced AEP result. You can define

the text written at the printing stage. The default is “Result – xx%”

PARK calculation 121

Enable advanced options:

Figure 139 Advanced setup options in “standard” PARK

Terrain types available in advanced mode:

‘

Here the terrain type is selected, where the corresponding roughness class and roughness length is shown.

Based on terrain class, the TI is calculated by hub height for each WTG in the calculation. In the three N.O.

Jensen models the TI is converted to a Wake Decay Constant (WDC), while the other wake models use the TI

directly.

Sector wise Wake decay constant/turbulence parameters

PARK calculation 122

Figure 140 TI or Wake decay constant (WDC) by direction sector

Sector wise TI or WDC can be entered or loaded from METEO data. The input options:

• PARK hub height

• Terrain class

• Roughness length

• TI, measured

• Height for measured TI

• WDC

Are fully tied together with formulas. (see further details in 3.7.1 The wake loss (PARK) models ).

New in windPRO 3.5-3.6 is partly:

Hub height dependent TI/WDC:

If the sectors are specified by a terrain type, the TI/WDC values can be recalculated to actual hub heights.

Therefore, the hub height specified in top is of Figure 140 now is called “Example Park hub height”. If the terrain

class is set to “User defined”, scaling to individual hub heights is not an option and input values will be used as

is.

[x] Offshore and onshore with low TI ~<10%

It is seen by many validations that for low TI sites, the factor 0.67 (PARK1) or 0.8 (PARK2) is best choice, while

0.5 (PARK1) and 0.6 (PARK2) is the better choice for onshore (updated in 3.6).

TI scale factor on Meteo TI

This is included to be able to calibrate e.g., mesoscale model data TI to local site measurements. Mesoscale TI

is very uncertain for onshore sites as the mesoscale models work on mesoscale level and thereby does not

capture the micro-level TI. Having TI measurements for a shorter period (1 year) will often be sufficient to

calibrate the mesoscale TI to the real site TI level, and thereby get access to long term TI data. Especially for

the EMD-WRF Europe+ and similar datasets, it is seen that the TI level offshore is 1.41 (√2) too low, and

multiplying with this is recommended. For EMD ConWx mesoscale data or EMD-WRF On Demand data this is

not recommended. Here the TI is as expected offshore and should not be scaled.

When loading TI from Meteo, this is converted to example hub height TI, to give the correct input for the WDC

calculation or the right level for models using TI direct (WakeBlaster and Ainslie).

NOTE: The factor from TI to WDC, above shown as 0.6, will, if PARK1 is chosen before this form is opened, be

0.50. It is thereby dependent on the wake model as described previously. If the wake model is later changed

from PARK2 to another model, this will not change the WDC unless the default selected “(o) WDC” is

changed to “(o) TI” before leaving the form. The default functionality of the form is that it is the WDC that is

kept when leaving the form and the other parameters are recalculated based on the chosen wake model.

PARK calculation 123

Figure 141 Import TI from Meteo object

Defaults are from 5 to 15 m/s, the wind speed interval where the wake loss appear, and therefore the TI within

this interval is the relevant.

Figure 142 Example of conversion of TI by height

Above is seen how TI now, when loading from Meteo, is shown for measurement height along with for expected

hub height, which decide the WDC.

Figure 143 Simple turbulence calculator for PARK input

The calculator next to TI input selection, gives the formulas behind the Roughness to TI conversion. The

calculator can be used to calculate TI based on the user’s own assumptions.

Access to advanced calculations:

PARK calculation 124

These Advanced calculations require you to choose another wake model than the default N.O. Jensen model.

The reduced wind speeds inside a wind farm create a matrix that can be used to lookup the calculated reduction

at a specific point, for any direction and any free wind speed. A way to use this is to place a very small turbine

(0.1 m rotor diameter) at the position where a measurement mast is located inside the wind farm, and, thereby,

get the reduction in wind speed measurements due to surrounding turbines. Then the expected free wind speed

can be established, which can be useful for performance check calculations.

Park Power curve based on a PPV model gives the calculated Park power output related to a specific

measurement mast position outside the park. This can be useful for situations like prognosis systems, where the

wind prognoses are given at a specific point (measurement mast position). With the table, the transfer function

from wind speed and direction to Park output is given.

Turbulence calculations give the ambient + wake added turbulences for each turbine. NOTE this calculation is

now transferred to the Site Compliance module (all versions after 2.8).

In “report features” there is an additional checkbox. Here, you can add calculation results for +/- ½ rotor diameter

in the result to file. This makes it easy to evaluate the variation of the calculation results over large rotors. With

the present AEP calculation tools, it is always assumed the hub height represents the average for the entire rotor

diameter. With a large increase in rotor diameters, this might require changes in future calculation methods.

Wake model selection:

Figure 144 Alternative wake models in "standard" PARK (Wind statistic based)

Regarding wake models, EMD recommends using the N.O. Jensen (RISØ/EMD) PARK2 model. If you need

turbulence calculations, the N.O. Jensen (EMD) 2005 is recommended. The three models mentioned below

N.O.Jensen : 2005 are intended for experimental use. Tests so far indicate that these models do not reduce

calculation results enough when using the standard parameters. The parameters will be tuned based upon

ongoing research in order to better correlate the predicted values to actual performance. New in windPRO 3.4

is that the Larsen model now can be used with linear combination model, but tests show that this do not make

this model perform well. The performance of the model is too dependent on the wind farm size and is only

recommended for experimental use. The Eddy Viscosity Model (J.F.Ainslie) : 1988 is a new implementation

included from windPRO 3.5, which includes a DAC (Deep Array Correction) model, which is a requirement for

making this model work well.

Turbulence model selection:

Figure 145 Alternative turbulence models in "standard” PARK

PARK calculation 125

For turbulence calculations, the S. Frandsen 1999 or the TNO model are recommended as the most-accepted

ones at this time. NOTE: Turbulence calculations will be taken out of the PARK calculation in near future. The

Site Compliance module has a much more refined and precise calculation of the wake added turbulence. The

wake and turbulence models and other advanced functions of the PARK calculation are described in detail in

the appendix on Wake and Turbulence Models.

Finally access to WAsP parameter changes are available here as well (see Section 6.3.1) But note that, from

WAsP 11, the wasp parameters set when generating a wind statistic (when generated with WAsP 11), are the

ones used in the PARK calculation.

3.7.4.2 Power Correction

By default, this option is always activated.

Figure 146 Power curve input (air density correction) for a wind statistic-based calculation

Power curves in the windPRO WTG Catalogue are only available for standard air density (1.225 kg/m

) and are

then recalculated by WindPRO to the site air density. Alternatively, the user can enter the local site air density-

specific power curve in the WTG Catalogue and then deactivate the correction (obviously, if the default correction

is activated with site air density available in Catalogue and in the air density setup, no correction will be made).

By default, windPRO will estimate the site air density from the closest weather station present in its embedded

Climate Database. Click the Edit button to see the details, and/or to change this choice.

PARK calculation 126

Figure 147 Air density setup form

In the air density window, it is possible to choose the Standard air density, or a manual input.

The default concept is to calculate air density at hub height for individual turbines. The crucial parameters are of

course air temperature and pressure. Temperature (by default taken from nearest climate station) is converted

to site using the temperature lapse ratio, applied on the climate station elevation and WTG hub height above

sea level. Air pressure is by default taken from elevation, assuming Standard Atmosphere.

Alternatively, both parameters can be manually entered, together with the respective measurement elevations.

As a minor influence, air humidity can also be entered.

Finally, it is possible to test (see section Example at the bottom) which results for air density the setup will give.

This is based on any value of elevation and hub height. This allows to test the sensitivity of the estimate to the

choices made.

Back to the Power correction main window, three different correction methodologies can be chosen.

Old windPRO method: based on the simple scaling of power values according to the ratio of site-to-standard

air density, typically used for stall-regulated wind turbines. This scaling approach would change the level of the

rated power if applied at all wind speeds. To avoid this, an empirical solution is made, such that the scaling stops

shortly before the scaled rated power is reached, and instead a smooth empirical transition to the real rated

power is made.

The drawback of this approach is that for large corrections (usually to low air densities) the empirical transition

from the steep part of the power curve to the rated power is not smooth enough and the shape does not

accurately mimic the behaviour of pitch-regulated wind turbines. Overprediction of AEP as for the IEC method

can be the result for large corrections to low air densities.

IEC 61400-12 method: it is a two-step procedure. The power output at all wind speeds is assumed to have the

standard ρu

dependence. Instead of scaling the power values at standard air density (P

std

) by the ratio of site

air density-to-standard air density (which would alter the rated power), step 1 scales the wind speeds of the

standard power curve according to:

site

std

(ρ

std

⁄ρ

site

)

1/3

The resulting corrected power curve (u

site

, P

std

) is, however, sampled at the new wind speed values, u

site

. To

obtain the air density corrected power curve at the original wind speeds (u

std

), step 2 interpolates the new

corrected power values at the original wind speeds, i.e. (u

std

, P

site

), from the curve (u

site

, P

std

PARK calculation 127

New WindPRO method: an improved air density correction, based on a simple adjustment of the IEC 61400-

12 method. The assumption implicit in the IEC correction is that the efficiency of the turbine is constant at all

wind speeds, which we know from the C

curve is not fulfilled. As a result, the IEC correction does not perform

well around rated power. Therefore, the IEC 61400-12 correction should only be used for small air density

corrections! The New WindPRO method uses the two-step approach identical of the IEC 61400-12 method,

with the simple but important difference that the exponent in equation above is not constant at 1/3 for all wind

speeds. Instead, the exponent is made a function of wind speed. A full theoretical description of this and the

other methods can be found in Reference documents - Power Curve Options.

Finally, negative values in the power curve can also be handled. Most often, the power value in the interval from

0 to around 4 m/s is 0 W. But some power curves do exhibit negative values. If these are included, they can be

handled in two different ways:

1. Default: the power consumed at low wind speeds is included in the AEP calculation, and the AEP result

thereby reduced.

2. Ignore the negative values in the AEP calculation, but then report the calculated power consumed

separately. This makes sense if a very accurate financial calculation is required, and the power purchase

cost is higher than the price paid for delivered power.

3.7.4.3 PowerMatrix

PowerMatrix is a new format for power curves developed by EMD. This experimental format allows for multiple

power curves in one structured file format. The idea is that manufactures instead of handing out multiple pages

with all different power curve variants (Air density, turbulence, noise, load modes etc.), can write all variants into

one file, in which windPRO can extract/interpolate to the one matching the specific situation. Especially for the

time step-based calculation concept, this will add several benefits, while the relevant power curve for each time

step can be chosen.

Standard PARK calculation with WAsP

3.7.5.1 Wind distribution

Figure 148 Wind distribution selection in standard PARK with WAsP

The site data object(s) to be used are checked – the default is that the nearest will be used at any turbine position.

At the WTGs tab, it can be decided, manually, to link a site data object to each turbine. This can be usefully if,

e.g., one site data object is based on measurements on a hill and another downhill. Then turbines on the hill

and downhill should be linked to the most representative site data object.

3.7.5.2 Displacement height

Figure 149 Displacement height setup in standard PARK

It is possible to use:

PARK calculation 128

• No displacement height – the “original” hub heights for each turbine is used.

• Displacement heights from objects – the hub heights are reduced with the displacement height defined

in the WTG object.

• Displacement height Calculator – Sector wise displacement heights from the calculator, object

displacement heights are ignored (see Section 8.4 for a description of the Displacement height

calculator.

RIX setup

Figure 150 RIX input setup in standard PARK

The RIX settings are explained in Section 3.4.5 Rix correction.

Standard PARK calculation with WAsP-CFD

The calculation setup follows the same as “Standard PARK with WAsP” except for the wind distribution tab,

which is replaced with two tabs: CFD result files and Wind statistics.

Figure 151 Wind distribution selection in standard PARK with WAsP-CFD

PARK calculation 129

The CFD result files are added, typically “all from calculation”, and the calculation “behind” the CFD results is

pointed out. Individual CFD result files can also be pointed out, if coming from another project or a mix from more

calculations.

Figure 152 Wind statistics selection in standard PARK with WAsP-CFD

More wind statistics can be selected. Note: When adding more wind statistics to a WAsP-CFD based

calculation, these additional should be generated based on CFD result files – if they come from a traditional

WAsP Statgen, this can introduce a bias. More wind statistics can be linked individually to turbines.

Standard PARK calculation with resource file

The calculation setup follows the same as “Standard PARK with WAsP” except for the wind distribution tab,

which is replaced with a “Resource files” tab:

Figure 153 Resource files selection in PARK based on resource files

Resource map files (.rsf or .wrg or .siteres) are added and used as the wind data in the calculation. In overlapping

resource files, these will be interpolated in the calculation points.

Note the new 3.5 feature: 3.6.5.3 Resource map rescaling

Common settings for Time series-based PARK calculation

Calculation in the time domain gives many new output options and more refined correction features, like seasonal

correction of wind data, power curve corrections for each time stamp based on variables like temperature, shear,

turbulence, etc., also, for the wake models, more advanced correction options are available.

PARK calculation 130

3.7.8.1 Setup, time step based calculation

Figure 154 Setup input for PARK based on time series

The output can be:

(o) AEP – the average hourly-calculated production based on all used wind data is multiplied with 8766 (hours

in an average year)

Include seasonal correction: Seasons can be defined, e.g., as months or 4-seasons (winter, spring, etc.).

Within each season, the average hourly production is calculated. This is then multiplied with the relative time in

each season and scaled to a full year with 8766 (hours in an average year). Thereby, each season is given the

same weight (corrected for length differences in seasons), and, thus, compensation is made for, e.g., an

overweight of winter data. If a season has less than 1% of all data, calculation is not possible while the season

correction is judged to be random – and if a season has no data, the season correction calculation is not possible.

Include a long-term factor – if the wind data period is known to be biased relative to long term, it is possible to

adjust the results by a factor.

(o) Time period energy – simply, calculated production in each time step and all values added. The result can,

thereby, be for, e.g., ½ a year or 3 years, depending on the data period. This can be convenient for follow up on

a specific period or to find the production that has been lost if a turbine has been out of order for a specific period.

PARK calculation 131

Include data recovery correction – if some time records are missing during the data period (defined by first

and last time stamp in the calculation period), the decision can be made to compensate for this, simply by finding

the data recovery rate and dividing this into the calculated summed production.

Use START-STOP time from WTG objects – is a new 3.2 feature that makes it possible to calculate a wind

farm where the operating turbines change in time – in ONE calculation. At the turbine object there is a new

setting: Operation, where the date for START and/or STOP can be entered. This feature makes it very

convenient when e.g., performing analyses in Performance Check, while the entire data period can be analysed

in one process. It is even possible to utilise if a turbine has changed operation mode during time, simply by

creating two WTG objects, having the different operation modes and set STOP for the one equal to START for

the other. Then although the operation data imported in Performance Check must be divided into two files with

different WTG ID. It is also possible to simulate future added wind farms based on this feature in combination

with an offset on the timeseries used in calculation. Add e.g., 20y to past 20y data and the turbines put in

operation during next 20y will be included from planned installation date set in the WTG object.

Use time of day depending power curves. This makes it possible to calculate like noise reduction modes that

vary during the day. The setup looks like this:

Figure 155 Time of day depending power curve setup.

Note the different power curves must be defined at the WTG object.

Use curtailment, see common settings for all PARK variants 3.7.3.

Use blockage, see common settings for all PARK variants, 3.7.3.3 Use Blockage

Limit park output to grid capacity, see 3.7.2.2 Grid curtailment

Output to PERFORMANCE CHECK and/or result to file, HYBRID etc.

The time step-based PARK calculation can deliver output files time step by time step, partly to be loaded into

Performance Check module for comparison with measured data, partly to a file that e.g., can be further analysed

in spreadsheet tools. Also the output for e.g., HYBRID module is controlled by selections made here. The data

quantity can become VERY large. Therefore, some limitations can be set, like only having detailed data for

selected turbines or only having the sum for all turbines. Finally, the time series can be aggregated to hourly,

monthly etc. data records. By default, aggregation to monthly values is set to avoid storage of large data amounts

that may not have any purpose. For a detailed Performance Check, 10 min. or hourly data is preferred, (assuming

this data is available as measured production). For HYBRID output at least hourly resolution will be needed.

Note the calculated data will be stored in the windPRO project file. Running many calculations with high

resolution/long data period in non-aggregated mode will make the windPRO project file very large.

Report features is as described for the wind statistics based concept.

Report pages for time varying results. Here, it can be decided to include new and/or existing turbines. A

special variant is to include existing ONLY if they have the property “Park WTG” checked in their object setup.

Thereby, a calculation having both existing in the windfarm to be expanded + reference turbines nearby, can be

“split”, so the time varying reports only show the result for a wind farm with some new and some existing, without

considering other existing turbines in the report.

The results presented are in the following graphs:

• 24-12 table with MWh

• 24-12 table with MW

• Graph with monthly production

PARK calculation 132

• Graph with diurnal production

• Table and graph with duration data

Result to file, the time step results are explained in 3.7.12.2 Output from PARK calculations, Result to file

3.7.8.2 Wake settings, time step based

Figure 156 Wake setup in PARK based on time series

Figure 157 Wake models for time step calculations.

The recommended model is as for statistical based calculations the PARK2 model. For performance Check

(model validation/tuning) although the NO2005 offers some more “handles” that can bring the model closer in

alignment with real performance. Especially for single row projects this can be tuned to handle those correct,

where the other models suffer from not being able to separate the “nature” of single row versus multiple row

performance.

An interesting new option since 3.4 is the external WakeBlaster model, see 3.7.13.

The standard input options are as described for the wind statistics-based PARK are shown above. The

interesting part is the new options - The wake model settings have this “advanced” expansion box for all models:

[x] Advanced – gives access to a time step turbulence-based correction.

Time step turbulence based requires a turbulence signal in a METEO object or when using “data from scaling”,

more options are available, see 3.4.8.4 Turbulence in scaler. NOTE: The dataset must exist concurrent to the

entire wind data period in the calculation! Based on the calculated wind speed at each turbine position, the

turbulence is then found from the St.dev. The turbulence at each turbine position then decides the wake decay

constant for the wake loss calculation at the given turbine. The translation from turbulence to WDC is based on

a simple linear relation, where the parameters can be edited. The default is like this:

PARK calculation 133

Figure 158 WDC versus turbulence, PARK2, for time step correction, recommendations.

Above, the relation between turbulence and WDC for PARK2. The DTU recommendations are limited to one

value for onshore, one for offshore. It is although documented in numerous DTU papers, that the TI affects the

wake losses significantly and thereby the WDC choice should depend on the site TI.

The latest EMD recommendations were shown in Table 6 Recommended settings for N.O. Jensen PARK

models. In Figure 158 it is illustrated graphically. There is a grey area between on- and offshore, with TI round

10%.

By analyzing data from more offshore wind farms, the relation WDC = 2 x TI -0.07 is found to give the best

match, when subdividing data into TI bins. This is although problematic as a generalized formula, as the average

TI level impacts this formula. Thus, it will only be working well for sites with average TI round 6-7%.

The large advantage making the WDC controlled by TI is that it takes out the guessing required on an average

WDC, and it does the weighting of the data against turbulence (see the test example in validation, Section 8).

With the fully formula-based TI to WDC concept from windPRO 3.2, the WDC for N.O.Jensen models can be

fully set by the TI or indirect by terrain class.

As seen in the input form, it is possible to select the WTG that controls the wind direction for the wake loss

calculation. It is so, when calculating wake loss in a time step, the wind direction must be defined. This can differ

by turbine and a choice must be made. The default choice is the first turbine in the list.

The reason for this input option is: when comparing two calculations, e.g., one with an existing wind farm and

one with a new + existing wind farm, the existing wind farm might get slightly different results in the two

calculations (apart from the influence from the new). Meanwhile, in the second, a new turbine comes in and

controls the wind direction, because new turbines are first in the list of turbines to be calculated. For such a

comparison calculation, it is then possible to manually select the same turbine to control the wind direction. In

very complex terrain with large turns of the wind direction, it can also be convenient to be able to decide which

turbine that decides the direction for the wake loss calculation, e.g., one in the middle of the site instead of the

outmost.

PARK calculation 134

The choice does also influence the calculated free wind speed aggregated by sectors. The reason for this is that

the sectors are defined by the wake calculation direction and thereby by the turbine chosen for controlling the

direction.

3.7.8.3 Large wind farm advanced features (Deep array losses or single row)

[x] Enable large wind farm advanced features – gives access to modifying the wake combination model with two

different variants for NO2005, and one for PARK1.

The here described features are for POST construction evaluations, where there is a wish to tune the Wake

model to reproduce the measurements. The found needed tunings can of cause be used for PRE construction,

if a similar wind farm is planned in similar environments.

In the original N.O. Jensen wake model, the wind speed deficits when having wakes from more turbines are

summarized by Root Sum Square (RSS) method. An alternative is to summarize the deficits, are the Linear

method (simple sum). Tests show this method reduces the wind speeds far too much, but, on the other hand,

the RSS method might to reduce too little when there are many arrays. Combining the two methods might give

a better reproduction. It is now possible to control the weighting of the two methods with full flexibility. For PARK2

the summarizing is fully linear, but based on revised formula set, so not direct comparable.

Figure 159 Linear and RSS weight in PARK combination model

The simple modification is not to give the RSS wind speed deficit calculation full weight, but to give some weight

to a linear sum of the deficits. An experimentally found reasonable weight for NO2005 is 35% linear weight,

given as the default (see further details in the Park model validation chapter, Section 8.2. The method is, e.g.,

supported by the work RES: OFFSHORE WAKE MODELLING, Presentation at Renewable UK Offshore Wind

2011, by Gerd Habenicht, Senior Technical Manager, 29th of June 2011.

But, this combination tends to give too high of reductions inside the wind farm. An additional correction proposed

in “Evaluation and Benchmarking of Wind Turbine Wake Models” by Mathieu Gaumond, DTU, June 30th 2012,

is to decrease the WDC by number of upwind turbines.

Some experiments have been performed and resulted in 3 different versions of the way the change in WDC by

upwind turbines are calculated. The version 3 is the recommended, ver. 1 & 2 are only seen as relevant for

backward combability. This option is ONLY available for the NO2005 model variant.

Version 1: Semi aggregated reduction by number of upwind turbines.

Version 2: Full aggregated reduction by number of up wind turbines starting from WTG 2.

Version 3: Full aggregated reduction by number of up wind turbines starting from WTG 1.

Figure 160 reduction of WDC by number of upwind turbines, Version 2 & 3.

This more advanced/experimental added modification of decreasing the WDC for the more upwind turbines is

found to improve the result, especially handle row by row more accurate, where the average only changes slightly

by the use of the shown defaults (see examples in validation, Section 8).

We recommend stopping the reduction after row 5.

PARK calculation 135

How version 3 works, including recommended input data, is illustrated below. Set 1 is recommended as initial

input. If the results (seen in Performance Check module) show that row by row not is reproduced correct, set 2

(less aggressive) or set 3 (more aggressive) can be brought in.

Table 9 Decreasing WDC by upwind turbines.

Multiple rows: decrease WDC by upwind turbines

set1

set2

set3

-0,3

-0,2

-0,5

1,4

1,3

1,5

Figure 161 Decreasing WDC by upwind turbines.

For multiple row wind farms, there will typically be seen a need for decreasing the WDC by number of upwind

turbines. This can be done more or less “aggressive” illustrated by the three sets of constants, resulting in the

three shown graphs where it can be seen how a “base WDC” of 0.05 converts to WDC values used for WTGs

with one or more upwind WTGs.

Table 10 Increasing WDC by upwind turbines.

Single rows: increase WDC by upwind turbines

set1

set2

set3

0,6

0,3

0,9

0,6

0,8

0,4

Figure 162 Increasing WDC by upwind turbines.

0,02

0,04

0,06

0,08

1 2 3 4

WDC

Upwind turbines

Changing WDC by upwind turbines, multiple rows

WDC_1 WDC_2 WDC_3

0,02

0,04

0,06

0,08

0,10

1 2 3 4

WDC

Upwind turbines

Changing WDC by upwind turbines, single row

WDC_1 WDC_2 WDC_3

PARK calculation 136

For single row wind farms, there will typically be seen a need for increasing the WDC by number of upwind

turbines. This can be done more or less “aggressive” illustrated by the three sets of constants, resulting in the

three shown graphs where it can be seen how a “base WDC” of 0.05 converts to WDC values used for WTGs

with one or more upwind WTGs.

IMPORTANT to notice is: A correct wake loss calculation is a balance of the chosen model (e.g., the N.O. Jensen

2005 implementation), the WDC and evt. Deep array settings (Linear weight and WDC decrease by upwind

turbines). Ongoing experiments show that a 35% Linear weight in combination with turbulence based WDC gives

the best reproduction of the measured wake loss when using NO2005. If Linear weight of 100% is used, the

increase of WDC should be 0.03 (from 0.04 to 0.07 offshore). But then the wake loss along the rows is over

predicted, but the 360° result is fine for the test case HR1. See validation chapter.

3.7.8.4 Power curve corrections by time series PARK calculation

Power curves are traditionally used as an “annual average” power curve in AEP calculations. With the time step

calculation method, it is possible to introduce corrections for power curves based on the meteorological

parameters present at each time step. This will, typically, not change much in the calculated AEP, but, for sites

with, e.g. very low or high turbulence, it can create improvements on the AEP accuracy. And, when doing

Performance Check analyses, the corrections can explain periods with high/low performance based on the

influence of the meteorological parameters on the power curves.

Figure 163 Power curve correction options in time domain

The possible corrections are:

• Air density (by temperature and/or pressure)

Selecting the “according to IEC 61400-12-1 ed.2” OR using PowerMatrix, gives following extra:

• Turbulence (only pitch regulated turbines)

• Shear (based on shear heights selected in Scaler)

PARK calculation 137

• Veer (based on Veer signal in Meteo object or mesoscale shear)

The Air density correction follows the formulas described in the IEC standard 61400-12-1 ed.2, and differs slightly

from the “Adjusted IEC method” described in Section 3.7.4.2. A full theoretical description of the two methods

can be found in https://help.emd.dk/knowledgebase/content/ReferenceManual/PowerCurveOptions.pdf

It should be noted that the IEC standard describes how to correct a measured power curve to standard conditions

based on the meteorological parameters. Here, the methods are “reversed” so they correct the standard power

curves to modified power curves based on the parameters. To make the concept work smoothly, some decisions

have to be taken upon implementation.

By default, windPRO will estimate the site air density from the closest weather station present in its embedded

Climate Database. However, in a timeseries-based scenario, it is certainly wiser to extract temperature and

pressure from the METEO object with the wind data selected in the SCALER tab. (Use data from Scaling). Then,

it will be required that ALL METEO objects (if multiple) have the required signals present at EACH time stamp.

For example, wind speed AND direction must be present in all METEO objects for the given time stamp. The

calculation, however, allows for a time offset of up to 50% of the sample time, if more METEO objects are used.

Thereby, non-syncronised time stamps in different METEO objects can be used.

If temperature or pressure are taken from ONE METEO object, the demands on data recovery are less. Here,

up to 13 time steps (equal to 2 hours+ having 10min wind data) is allowed to be missing or disabled and a linear

interpolation is used between two nearest time stamps to find the value. This means that calculations based on

10 min. wind data can use, e.g., mesoscala data with hourly temperature and pressure values as calculation

basis for the air density correction. In that case, up to 2 hour values can be missing, and the calculation will still

be performed for each 10 min. value based on the interpolated signals.

Same rule for turbulence, where the SCALER from 3.1 partly can scale the turbulence based on more source

data(more masts, heights or mesoscale points). Partly take the turbulence from model calculations, if e.g., WAsP

CFD result files or FLOWRES files are used in the SCALER. Turbulence Intensity (TI) is scaled to each

calculation position assuming a constant Standard deviation by height. From the source data, the standard

deviation is calculated from TI and wind speed. Then, based on the calculation point wind speed, the TI is

calculated. This is a quite simple approach, not taking into account if there are parts of the site with large

turbulence deviations. If the turbulence information is missing in a time step (outside the interpolation frame), no

turbulence correction is performed, but the data point is still calculated. Before a calculation is started, it is tested

that at least 50% of the time steps has turbulence information, otherwise an error is reported, since it does not

make sense, in that situation, to calculate with turbulence correction.

Turbulence can be taken from Mesoscale data (see Chapter 12 regarding METEO objects and how to establish

turbulence data from mesoscale data in a METEO object).

Figure 164 Reference turbulence for TI correction of power curve

The turbulence correction needs an assumption, for the turbulence at which the power curve used, that is

representative. This value can be user defined.

Shear can be calculated by time step based on selected shear heights in scaler. For veer this is not an option,

here the user must establish a veer signal in a meteo object. The reason is that having two wind vanes normally

not will give a reliable veer signal for all time steps, and non valid results will appear. Therefore the user must

PARK calculation 138

import and screen the data before use. Eventually substitute part of the time series based on mesoscale data

veer – or solely base veer signal on mesoscale data.

In the PERFORMANCE CHECK manual, examples on calculated air density and turbulence corrections are

compared to measurements.

For details on the correction methods, please see the IEC standard. The corrections applied can be seen in

details in the result to file with a time step calculation. Here as well, the signals used at each time step as the

calculated corrections can be seen.

Time varying calculation based on mesoscale data

Figure 165 Selection of wind data and SCALER in PARK time series based

Basing the time step PARK calculation on Mesoscale data from EMD (The EMD-WRF model or the EMD-ConWx

dataset), all heights of the mesoscale dataset are selected by default. Heights below the lowest calculation height

and above highest calculation height MUST be included – no extrapolations are permitted in this calculation

concept. The mesoscale data defines the shear and veer. It is, although, possible to also choose the

“measurement SCALER” with Mesoscale data (e.g., mesoscale data from other providers), and let the WAsP

model do the vertical scaling, then only one data height is used.

More mesoscale data points can be used, and horizontal interpolation can be chosen – the alternative is to “take

nearest”.

Time varying calculation based on measured data

This option is like time varying based on mesoscale data apart from:

• Instead of using mesoscale data, measured data is used (or an “artificial mast”).

PARK calculation 139

• Model is used for extrapolation in height, can be WAsP, WAsP-CFD, Flowres format or resource file.

Only local terrain is used.

• The default scaling is based on the WAsP A parameter ratio.

• Heights for wind speed calculation and shear calculation can be individually selected.

Other PARK calculations

The “other” PARK calculations are included for backward compatibility to previous windPRO versions. Please

refer to windPRO 2.9 manual for details on these options.

Figure 166 The 2.9 compatible PARK methods

Output from PARK calculations

Partly the report pages and partly the result to file options will be described here that give an output for taking

the results to spreadsheets or other external software tools for further processing. It will also be explained how

the output options will be different from a wind statistics-based calculation and a SCALER (time step) calculation.

One of the differences is that the SCALER based calculations will be able to provide the wake reduced wind

speeds as well as the non-wake reduced, since the wind statistics-based calculation only provides the non-wake

reduced wind speeds.

Another difference is that the wind statistics-based calculations deliver the modifications due to, e.g., hills and

obstacles as percentages on AEP relative to flat terrain and no obstacles. This calculation is very fast, and extra

calculations are automatically performed with/without e.g., hills. The SCALER based calculations are only one

run, but, as an output, there will be the speed up factors (flow perturbations) from WAsP used in the SCALER.

This means that the time step calculations deliver the speed-ups on wind speed due to hills, obstacles etc. as

an alternative to the percent increase in AEP. The speed-ups are relative to flat terrain/no obstacles. If the speed

up relative to a measurement mast position is wanted, this can be found as the ratio between the speed up at

turbine position and mast position based on a calculation where both positions are included, where a turbine is

included at the measurement mast position.

3.7.12.1 Output from PARK calculations, Reports

A feature from windPRO 3.3 that not many users are aware or is the option to copy graphs and tables to clipboard

from the report preview:

PARK calculation 140

Figure 167 Right click on preview graph to copy to clipboard.

Figure 168 Right click on preview table to copy to clipboard.

When tables are copied, it is not “just” a picture, but the cells, meaning that pasting e.g., into Excel, the texts and

values will be available for further processing.

PARK calculation 141

Figure 169 The PARK main page report results based on wind statistics

The report page shown above is with “all inclusive”, meaning that existing PARK WTG’s as well as new WTG’s

are included in the calculation. This includes extra information compared to “just” calculating a new wind farm.

The status of “PARK WTG” can be set on the WTG object and means that the existing WTG is represented on

the main page. Alternatively, is that the existing WTG’s are treated only as reference turbines and will only

appear on the report page dealing with reference turbines.

PARK calculation 142

Figure 170 Part of the PARK main result from time step calculation

As seen above, free and wake reduced wind speeds are also included as a mean across hub heights. The main

results from a time step calculation show, by default, the long term expected AEP based on the wind data used

and is expected to be long term representative (see further details in chapter 7.1.1: Setup of time step-based

calculation).

3.7.12.2 Output from PARK calculations, Result to file

Result to file is the way to get structured output for further processing in e.g., Excel of by scripts.

Figure 171 Result to file output options from PARK, left wst based, right time step based.

PARK calculation 143

Above, the left form shows the output options from a wind statistics based calculation while the right form shows

the time step-based calculation.

As seen, there are more options for time step calculations, like the “Scaler result dump”, where the calculated

speed ups can be seen for each calculation point.

The PARK result to file (upmost choice) differences by the two different calculation concepts is shown here:

Figure 172 Result to file output comparison (shown transposed)

For the first 26 columns, the output is identical. Therefore, comparisons are easy to make. The only difference

is in column 3, where, in the wind statistics based calculation, the wind statistics name is shown and, in the

SCALER based calculation, the scaler name + the METEO object name is shown.

A new column from windPRO 3.2 is curtailment loss, inserted as column 26! In sector wise output

curtailment loss by sector is included as well as a new block (rightmost).

From column 27, the output is different. The wind statistics based calculation shows the Weibull A, k and f by

direction sector. The SCALER based calculation shows the Free and Wake reduced wind speeds and the

frequency by sector.

Figure 173 Result to file output comparison (shown transposed), rightmost columns

In the rightmost columns in the result to file, different values will appear depending on which corrections are

included in the calculations. Changes due to hill/obstacles are shown as percentages for each WTG in a wind

statistics based calculation, but NOT in a time step-based calculation. Note that the changes are on AEP without

Column: Wind statistic Time step Comment

Header Column text column unit 1. data line Header Column text column unit 1. data line

1 Name: Label Time: 6 Name: Label Time: 6 No. of WTG

2 Std. PARK lt meso data ############ New

Lt Meso calibrated with RIX and DH 5 WTGs

############ New New/Existing wtg

3 LIB file

C:\Users\per.EMD\Documents\WindPRO Data\Projects\Ireland\Cronalagth\IE Old met mast,Gillespie - A Scaled 30,00 m.wws

Scaler

RIX calibrated Meso Scaler+DH-test

EmdConwx_N55.070_W008.230 (21)

Wind data input

4 ITM X(East) 585.990 ITM X(East) 585.990

5 Y(North) 924.153 Y(North) 924.153

6 Z [m] 290,8 Z [m] 290,8

7 WTG type Valid No WTG type Valid No

8 Manufact. VESTAS Manufact. VESTAS

9 Type-generator V47-660 Type-generator V47-660

10 Power, rated [kW] 660 Power, rated [kW] 660

11 Rotor diameter [m] 47 Rotor diameter [m] 47

12 Hub height [m] 40 Hub height [m] 40

13 Row data/Description

VESTAS V47 660 47.0 !O! hub: 40,0 m (TOT: 63,5 m) (142)

Row data/Description VESTAS V47 660 47.0 !O! hub: 40,0 m (TOT: 63,5 m) (142)

Power curve

Creator EMD

Power curve

Creator EMD

15 Name Level 0 - calculated - - 07-2001 Name Level 0 - calculated - - 07-2001

16 User label User label

Annual Energy

Result [MWh] 2.678,70

Annual Energy

Result [MWh] 2.697,90

18 Park Efficiency [%] 97,79 Park Efficiency [%] 98,04

19 Regional Correction Factor 1 Regional Correction Factor

20 Equivalent roughness 0,5 Equivalent roughness NA for time step

21 Mean wind speed [m/s] 8,96 Mean wind speed [m/s] 8,93 Wake free

22 HP-value [%] 99 HP-value [%] 99

Calculated prod. without new WTGs

[MWh] 0

Calculated prod. without new WTGs

[MWh] 0 Only existing WTG

Actual wind corrected energy

[MWh] 0

Actual wind corrected energy

[MWh] 0 Only existing WTG

25 Goodness Factor [%] Goodness Factor [%] Only existing WTG

26 Curtailment loss [%] Curtailment loss [%]

27 A (Sum) [m/s] 10,12

Wake reduced mean wind speed

[m/s]

28 k (Sum) 2,05 Free WS (0) [m/s] 8,09

29 A (0) [m/s] 9,64 Red WS (0) [m/s] 7,81

30 k (0) 1,936 f (0) [%] 5,6

31 f (0) [%] 5,5 Free WS (1) [m/s] 7,1

Column: Wind statistic Time step Comment

Header Column text column unit 1. data line Header Column text column unit 1. data line

65 Air density [kg/m³] 1,208 Air density [kg/m³] 1,213 comparable

66 Displacement height [m] Sector wise Displacement height [m] 0 comparable

67 Park

Decrease due to obstacles

[%] 0,38 Park

Decrease due to obstacles

[%] Sector wise NA in time step

68 Park Increase due to hills [%] 25,86 Park Increase due to hills [%] Sector wise NA in time step

69 Sensitivity

[dAEP/dWS %]

1,29 Sensitivity

[dAEP/dWS %]

0,36 comparable

70 Reference site RIX [%] 0,5

71 WTG RIX [%] 1,3 WTG RIX [%] 0,0106

Delta RIX (WTG site - Reference site)

[%] 0,8 Calc. period Start 26-09-1999

73 RIX correction [MWh] -26,38 End 03-06-2000

More RIX info in

wst based. In

time step start-

stop dates

included

PARK calculation 144

Wake influence. Sensitivity is “comparable” although calculated differently. In a wind statistics based calculation

wind statistics, it is calculated by running an extra calculation with 1% decrease of wind speed (Weibull A-par).

In the time step calculation, similar 1% decrease is used, but for each time step and aggregated.

If the displacement height is non-sector wise, the displacement height will be seen.

The sector wise reports will give the details on displacement height, obstacles, hills and Rix .

Figure 174 Result to file; Sector wise; output comparison

For the sector wise output, there will be similar results from the two calculation types, apart from effect from the

obstacles, hills and Rix. The wind statistics based calculation shows the percent change, where the time step

based calculation shows the flow perturbations (speed-ups = factors on wind speeds from model). Rix correction

is shown as MWh change in a wind statistics based calculation, while the time step based calculation shows the

Delta Rix value. The reason for the difference is as previously mentioned that a wind statistics based calculation

is very fast, and, thereby, auto run more times with/without corrections, allowing the change in AEP to be found

directly. The Time step calculation requires more time and, therefore, it is not calculated several times. Instead,

the parameters calculated by the SCALER/WAsP is presented. If there is a need for identifying the change in

AEP by different corrections more precisely, the only way is to run the calculation more times without the

corrections, one by one, and then compare results.

Figure 175 Result to file; Park results, WAsP 11

There is an option to obtain the native WAsP results. These are shown and commented on above. The

roughness speedups are relative to a reference roughness set by WAsP, seen in the output as the Mesoscale

roughness (as roughness lengths).

Time series output

The “Park time variation” output results for each WTG (or selected) for each time step – or aggregated time steps

- depending on setup:

Wind statistic Time step Comment

One column for each direction sector + 26 first columns from previous

Annual Energy (MWh) Annual Energy (MWh) comparable

Park efficiency (%) Park efficiency (%) comparable

Decrease due to obstacles (%) Obstacle speedup different

Increase due to hills (%) Orographic speedup different

RIX correction (MWh) Delta rix (%) different

Displacement height (m) Displacement height (m) comparable

PARK calculation 145

Figure 176 Setup can decide which WTGs and time resolution for Result to file output.

PARK calculation 146

In the time varying Result to file, the following information can be found:

Figure 177 Result to file output from PARK based on time series

The results can be copied to the clipboard and pasted into Excel.

Table 11 Output from PARK based on time series to spreadsheet

Above is a part of the result, where results are aggregated to annual values, are copied to Excel. To the left is

the inserted calculations, where “Time” column is multiplied with “Power” column to get the production. In this

way the annual variations in expected production can be visualized and annual production indexes calculated.

More detailed description of the result columns below.

PARK calculation 147

Table 12 Output from PARK based on time series to spreadsheet – column documentation

The columns in result to file output are shown above. Here, they are shown transposed with an explanation for

each line (column). After the columns with WT-1, it continues with WT-2 etc. (or selected WT’s).

The format is fixed for the first 19 columns per WTG, independent from which corrections applied. If curtailment

is included, there can be added more columns per WTG, depending on which curtailment signals used.

NOTE: Signals can be taken from Scaler or METEO objects.

If from Scaler, data can be model signals (from e.g., FlowRes or WAsP-CFD result files) or Meteo signals scaled

by the models, depending on settings in Scaler.

If from METEO objects, signals are scaled from Meteo object position and height to WTG position and height,

where models for this scaling are available (apart from wind speed/direction, temperature, pressure and

turbulence always will be scaled based on simple models).

PARK with WakeBlaster, external wake model

WakeBlaster is the first external wake model windPRO offer aces to, but probably more will come. The use is

limited to time series calculations. Before ver. 3.6 based on precalculated scenarios, from 3.6 also time step by

time step calculation based on like different TI or by time step. This is especially relevant for validation

calculations, where measured data per time step is available. Then it can be validated how well the wake loss

by varying TI is reproduced by the model.

Header row 1 Header row 2 Header row 3 Header row 4 1. data line Explanation

LocalWind-PARK2 Wake_Curtailment & all corrections

18-02-2020 10:51 Time stamp Date-time 19-01-2018 14:20

Scaler: Total Power [kW] 1044,8 Sum of power for the wind farm in time step

Def.MeasureScaler Time [h] 0,1667 Hours in the time stamp, multiply with power to get production

Meteo data:

For reference WTG: [1] 570715000001306876: 3075 kW Vestas - Ørndrup Hovedgå- HH-mod.

Free wind speed [m/s] 4,4

WT1&4 Merge-period calibrated - dir -5 Reduced wind speed [m/s] 4,4

Wind direction [°] 162,5

Bolded are variables Temperature [°C] 1

Pressure [hPa] 987

Air density [kg/m3] 1,254

WDC Turbulence [] 0,071

WDC [] 0,057

Turbulence [] 0,071

Shear [] -0,6

Veer [°] 0,1

Curtailment index [] 0

Ref power [kW] 296,1

All corrections [-] 1,44

Air density correction [-] 1,02

Turbulence correction [-] 0,97

Shear correction [-] 1,45

Veer correction [-] 1

Temperature [Deg C]

Inflow angle [Degrees]

< Dynamic curtailment signals> Below for WTG 1, similar to find for each WTG

1 Power [kW] 296,1 Average power in time step for WTG 1

1 Free wind speed [m/s] 4,4 Free wind speed (before wake reduction)

1 Reduced wind speed [m/s] 4,4 Wake reduced wind speed

1 Wind direction [°] 162,5 Below signal has only values if used in time step

1 Temperature [°C] 1 Used in Air density calculation

1 Pressure [hPa] 987 Used in Air density calculation

1 Air density [kg/m3] 1,254 Calculated for hub height

1 WDC Turbulence [] 0,071 Turbulence value (ambient) used if WDC (TI) used per time step

1 WDC [] 0,057 Wake Decay Constant (can be different by WTG) ONLY if wake

1 Turbulence [] 0,071 Turbulence value (ambient) used if power curve correction by TI used

1 Shear [] -0,6 Shear as power law exponent (only shown if shear correction used)

1 Veer [°] 0,1 Degrees difference over the rotor diameter (shown only if used)

1 Curtailment index [] 0 See report page "Curtailment assumptions" which rule this refer to

1 Ref power [kW] 296,1 The power if no curtailment is present

1 All corrections [-] 1,44 All power curve corrections combined, set to 1 if curtailed

1 Air density correction [-] 1,02 Correction factor on power due to Air density

1 Turbulence correction [-] 0,97 Correction factor on power due to TI

1 Shear correction [-] 1,45 Correction factor on power due to Shear

1 Veer correction [-] 1 Correction factor on power due to Veer

1 Temperature [Deg C] Curtailment signals added IF USED, might be different from the

1 Inflow angle [Degrees] earlier shown signals, while they can be taken from another source

Note: The values in first block are for the reference WTG,

default the first in calculation list. This can be changed in PARK

setup. The

reference WTC control the inflow direction in PARK

calculation and the used TI if TI used for Wake calculation.

The shown signals are explained in the block below for WTG_1

NOTE1: Wake added TI is NOT included in any TI signal at

present.

NOTE2: The number of columns are dynamic in case

Curtailment is used in calculation. FIRST 19 columns for each

WTG although fixed, no matter calculation settings.

PARK calculation 148

3.7.13.1 About WakeBlaster in windPRO

WakeBlaster is a fast 3D wake model, developed by ProPlanEn

. It uses a custom solver of the 3D Reynolds-

Averaged Navier Stokes Equations.

Figure 178 Example of a WakeBlaster CFD simulation.

A flow case for Anholt offshore wind farm. Colour-coded as wind speed deficit, relative to free upstream wind

speed.

WakeBlaster is implementing a RANS solver that models a complete wind farm, not single turbines. Resolving

3D effects, like wake expansion and wake superposition, results in the replacement of empirical approximations

and the reduction of uncertainties. For more information, visit https://proplanen.info/wakeblaster.

WakeBlaster runs on a remote server and requires an API key (see the paragraph below, regarding purchasing

and costs).

WakeBlaster users benefit from the capability to run WakeBlaster from windPRO, gaining access to windPRO’s

advanced handling of meteorological data (by Scaler) and power curve correction options, as well as its reporting

features. This makes it easy to compare with other models, as all output is similarly formatted.

WindPRO users benefit from an almost seamless integration of a new state of the art wake model into their

existing workflow. It is also possible to use WakeBlaster calculations in, e.g., a performance check module for

validation against SCADA data.

3.7.13.2 Setting up the request calculation in windPRO

The first step is to define the layout to be calculated, and to process this in order to generate the necessary files

for the WakeBlaster calculation. This is done within the PARK module.

Start PARK module

Wolfgang Schlez, “Virtual Wind Farm Simulation – A Closer Look at the WakeBlaster Project”, WindTECH International,

Volume 13, No 6, 2017 (https://proplanen.info/wakeblaster).

Philip Bradstock and Wolfgang Schlez, “Theory and Verification of a new 3D RANS model”, Wind Energ. Sci. Discuss.,

https://doi.org/10.5194/wes-2020-33, in review, 2020.

PARK calculation 149

Only the time series-based variants support WakeBlaster calculations in this version.

Figure 179 Selection of WakeBlaster model.

At “Setup”, the output specifications can be decided for later use. E.g., by default, result output to file as time

series will be aggregated monthly. This can be changed to aggregation level “None”, to get full output resolution

for, e.g., use within a performance check module with 10-min resolution SCADA data. At the “Wake” tab, select

“WakeBlaster” and a new tab will appear.

Choose WTGs “as usual”.

Note: when running a WakeBlaster calculation on a remote server, the returned results are “locked” to the chosen

layout and turbine choice. Any changes in WTGs will require a new WakeBlaster calculation on a remote server.

If you later want to calculate with the WakeBlaster result file and SAME layout/turbine type, but different wind

data, this is possible if you clone the first calculation, then change the wind data selection in the “Scaling” tab,

see below.

Figure 180 Chosing wind data.

Behind the scenes, the WakeBlaster model is currently operated in a statistical mode, but we will endeavour to change this

in a forthcoming release cycle.

PARK calculation 150

Chose the “Scaling” settings, as usual. At least one Meteo dataset must be selected, but more can be selected

to reflect how the wind varies over the site. Later, a WakeBlaster calculation based on results from the remote

sever calculation can be run by different Meteo selections, without running a new WakeBlaster calculation on

the remote server. This is done by cloning your first WakeBlaster calculation, then editing it in the

“Scaling” tab settings. You will not be able to modify your WTG selections in the cloned calculation, because

the WakeBlaster run is for the specific wind farm.

Figure 181 Preparing WakeBlaster calculation.

Choose between scenarios and time step calculation. The time step calculation has the advantage that individual

TI by time step can be used and thereby a more accurate calculation is performed. But for long time series it will

be an expensive calculation, and thereby mainly recommended for validation calculations, where the focus is to

test how well the model handles low versus high TI situations. Typical number of flow cases by scenario

calculation is 2700, which correspond to round 4 months of 1-hour values or less than 1 month of 10-min. values.

Preparation for Scenarios:

Set up the flow cases. The minimum span is 2-16 m/s, which will cover all wake loss situations, although this

can be expanded in special cases. The recommended step is 1 m/s.

The number of sectors determines how precise a calculation will be. A 2-degree step is high accuracy and means

180 sectors. Less than 72 sectors (5 degrees) cannot be chosen, because it does not make sense to use

WakeBlaster for lower wake calculation direction resolution.

You can “Generate Flow Plane image” (costing 100 flow cases) for a specific wind speed, direction, and hub

height. This will provide a plot in the report and a picture for viewing/copying - see later example.

PARK calculation 151

Figure 182 TI setup for WakeBlaster.

Finally, the turbulence must be entered. This can be loaded from a Meteo object, if quality TI data are available.

However, please be aware that (like the mesoscale model) TI might not be sufficiently accurate. (The EMD-WRF

Europe+ dataset is known to have TI values that are too low, which leads to calculation of wake losses that are

too high – but, these TI values can be recalibrated to compensate for this effect). Press “OK” to revert to the

“WakeBlaster” tab.

Preparation for Time step calculation:

Figure 183 Input for WakeBlaster time step calculation.

The time series containing wind speed and direction data is selected in the Scaling tab. If these data include

turbulence, the choice of Turbulence should be “from scaler”. It is also possible to select turbulence data from a

different meteo object. The TI data can be scaled. For instance, if the TI data is biased. Finally, a fixed TI can

be experimentally chosen, but this voids the whole point of using the time step calculation is to have the TI by

time step.

After preparation

Press the “OK” button, and the WakeBlaster input preparation calculation will run. This can take some minutes,

depending on the size of the wind farm and the number of meteo positions.

PARK calculation 152

Figure 184 WakeBlaster status appear when ready to start on remote server.

When the request files have been calculated, the “WakeBlaster status” window appears.

3.7.13.3 Starting the WakeBlaster calculation on a remote server

Figure 185 Enter URL and API.

Enter the URL and API key which you received by email when you purchased the WakeBlaster packages

. Your

monthly inclusive credits will show up in the status when you select: “Get account status”. Additional credit

packages will be assigned directly to your account and will show up here as soon as they are activated.

If enough are credits available (see below):

Here, 2,700 flow cases are needed, so it is ok to continue.

Press ”OK”.

Please keep your key safe; it is valid for use within your company, and provides access to your data and flow cases. Flow

case credits are non-transferable, and no refund is available for flow case credits that are used without authorisation.

PARK calculation 153

Accept, and the initiated job will be sent to the remote server.

3.7.13.4 Waiting for processing at a remote server

Now wait to receive a confirmation email, stating that the calculation is done:

Figure 186 Email notification when result file is ready.

The email will look like this in your mailbox. This process will take from around 10 minutes to more hours

depending on the (cost-optimized) server load.

Figure 187 Check calculation status on remote server.

If a calculation takes longer than usual to return a result, you can check the job status. Open the calculation

(right click on “calculation” and choose “properties”) and select “Get job status”. You will receive an update, and

potential error messages are displayed. If you cannot identify the cause of an error yourself, please send a copy

of the error message to user support. Select “Cancel” to leave the window – if you select “OK”, a new

calculation request will be activated for no reason.

The “Load old result file” gives the opportunity to read a previous downloaded WakeBlaster result file saved

locally. To use this for a new WakeBlaster result based file, possible with other wind data, it is important to notice

that the WTG configuration must be exact the same as when the WakeBlaster calculation were requested. There

will be a check when starting a wake blaster calculation that as well WTG Ct curves as coordinates etc. match.

If not, the calculation cannot be performed. The user will be informed about the missing matches.

WakeBlaster is designed, and the available resources are configured, so that the calculations are fast enough to use

WakeBlaster in your daily work. If you require a reduced latency and/or faster response-time for a specific application case,

please contact us with the details, and we can provide a quote.

PARK calculation 154

Figure 188 Ready to download WakeBlaster results.

When the calculation is done (you have received the email), open the calculation and download the result file.

You can “Show flow plane”, if it is included in the calculation: (here shown for the Anholt project, not similar to

the calculation setup shown in all the previous screen shots)

PARK calculation 155

Figure 189 Downloaded WakeBlaster results, ready to calculate.

When downloaded, the highlighted fields changes, and you are ready for performing the WakeBlaster based

PARK calculation by pressing OK. If added, you can choose to show Flow plane or Calculation info. Last one

should be copied in case there seem to be something wrong with the calculation, and hotline support is needed,

then the WakeBlaster team can identify which calculation and thereby hopefully the problem.

3.7.13.5 PARK calculation based on WakeBlaster result file

Now press “OK” and the calculation based on the WakeBlaster result file will run. Note: Meteo data in “Scaling”,

“Power corrections”, and “Setup” (e.g., selection of the aggregation resolution for the time series output), can all

be modified before “OK” is selected.

PARK calculation 156

When calculation is done, the reports can be shown/printed and “result to file” data extracted.

A new page option, the “flow plane”, will show the above graph in the report, if it is selected at the point of

calculation.

3.7.13.6 Evaluation of WakeBlaster results, compared to PARK2

Here, WakeBlaster ver. 2.1.3 is being used and, in this case, free wind for each WTG position. Power curve

corrections are available, as for other wake models in windPRO. Here, air density correction adds 0.4% to the

calculated AEP and turbulence correction adds 0.7%, compared to using the standard power curve.

The main results are a 6.4% wake loss, calculated for the Anholt offshore wind farm. This is slightly higher than

in PARK2, which shows a 6.3% wake loss, with default offshore WDC settings of 0.06 - see similar results,

below:

With EMD’s recommended approach for PARK2 (using WDC = 0.8 x TI for offshore), the result is:

Here, there is calculated 7.2% wake loss, somewhat higher than in the two previous examples. The sensitivity

to the accuracy of the TI data is high.

The actual production (normalised to 1 year) for the calculation period is 1,763 GWh, which gives the following

evaluation of the 3 calculations:

PARK calculation 157

Table 13 WakeBlaster and PARK2 results.

Measured

PARK2 TI

PARK2 DTU

WakeBlaster

Production [GWh]

1.763

1.853

1.872

1.868

Ratio

0,95

0,94

The real production at this low-resolution follow up level cannot tell which model performs best. An over-

prediction of around 5% can easily be explained by losses and downtime (technical and market regulation), wind

data bias, etc.

Figure 190 Wake loss by turbine.

A result comparison, by turbine. There is no big difference between the calculated wake losses here, but other

detailed tests against measurements show that the WakeBlaster model does make it better than the N.O. Jensen

based models in some cases, but not all. In any case, such comparisons are very sensitive to turbine operation

issues, and the accuracy of Ct curves, etc. In the OWA (Offshore Wind Accelerator) project, several wake models

were compared against operation data for six Danish and UK offshore wind farms. Here, WakeBlaster (and

PARK2) also performs well (it is not outperformed by any other wake models). A project result presentation is

available. It cannot be concluded that any one wake model is always the best but having multiple opinions will

always be better than having just one.

Plotting the calculated wake loss based on the time step calculation, with a high directional resolution, illustrates

the model differences:

PARK calculation 158

Figure 191 Wake loss by direction, WakeBlaster and PARK2.

Looking at the layout in next figure, when the wind comes along the long rows, PARK2 calculates higher wake

losses than WakeBlaster. It is difficult to validate what is most correct while precise direction data are rare.

However, from the test for Horns Rev1 (for example), there are indications that PARK2 calculates wake losses

that are too high when the wind comes along the rows, compared to when wind comes skew to the rows. All in

all, both models end up producing similar overall results.

PARK calculation 159

Figure 192 The windfarm layout (Anholt) in the calculation test.

The numbers correspond to those shown in the previous graph.

Figure 193 The calculated wake loss by wind speed, WakeBlaster and PARK2.

100

101

102

103

104

105

106

107

108

109

110

111

6260000

6265000

6270000

6275000

6280000

6285000

6290000

630000 632000 634000 636000 638000 640000 642000 644000

Anholt layout

PARK calculation 160

The calculated wake loss by wind speed. PARK2 calculates more loss from 5 to 10 m/s, whereas WakeBlaster

calculates slightly more than PARK2, with DTU default WDC at the higher wind speeds, where there is the most

production.

3.7.13.7 Purchasing WakeBlaster credits

Visit https://www.emd-international.com/windpro/cfd/wakeblaster/ for actual info on purchasing WakeBlaster

credits. The costs are in size order 20 – 50 € per WakeBlaster calculation.

Project Cost and LCOE calculation

From windPRO 3.6 it is possible to calculate projects costs and Levelized Cost Of Energy (LCOE) for the

designed wind farm based on cost functions. The cost functions transform the physical properties of the wind

farm to costs using cost formulas and specific costs, typically per MW. The grid and road lengths are

automatically calculated based on shortest distance between WTGs within the wind farm.

See detailed description of the cost functions in BASIS manual, chapter 2.18.

NOTE: Only New WTGs can be used in the cost calculator.

Figure 194 Cost calculator in PARK.

Select “use cost calculator”, then select a cost model in drop down:

Then the “Setup” button gives the option to define other cost models.

PARK calculation 161

Figure 195 Create new cost models, choose currency, and calibrate costs.

If you already know some of the costs of a component, like the turbine cost, the cost model can be calibrated.

See Chapter 2.18.1.2

In the Cost Model Setup, several actions can be taken:

Insert a new cost model based on EMD defaults. Here 3 variants, low, mid, high for onshore and offshore are

available for typical projects. These should give realistic values based on the designed wind farm, although

variations from project to project can be high.

For the AEP, which is very deciding for e.g., LCOE, two important notes:

1) When using the cost function BEFORE a PARK calculation is performed, a rough guess is used

(Capacity Factor of 35%). Reopening after a calculation, the calculated values are imported into the

“example data” panel in the right-hand side of the window. This will be using the last calculated AEP. So

if the wind farm has changed since the last calculation, these changes will not be reflected in the AEP

shown in the example. Updating the PARK calculation will also update the example data.

2) Loss deductions of AEP should be included. These are taken from setup tab:

From windPRO 4.0 it is possible to attach cost maps to add known spatial costs depending on soil, ownership,

water depths etc. See the BASIS manual section 2.18 for more information about attaching a cost map.

In a PARK calculation the spatial costs are calculated based on the position of New WTGs. Values outside the

cost map are assumed to be zero. So, if a WTG’s position is not within the cost map, it will have no spatial cost.

PARK will warn the user if a turbine is located outside the cost map area.

Results: At the cost function form, tables with costs, specific costs COE and LCOE can be seen. In the report

similar results is seen.

PARK calculation 162

Loss & Uncertainty 163

Loss & Uncertainty

The PARK calculation traditionally provides the results assuming all turbines are running “full time” and it

provides the AEP results measured at the turbine, which is usually the output before the step-up transformer.

Although, in some cases, power curves can represent the output after the step-up transformer. Power curves

will also typically include noise reduced operation modes (selected by user). Since windPRO 3.3, curtailment

losses and wake losses can be included in the calculated AEP by PARK. This has the advantage that the saved

wake losses due to curtailment also are correct calculated. In loss & Uncertainty module, the already included

curtailment losses are transferred from PARK similar to wake losses, as a fixed reduction that cannot be edited

in L&U. They are also handled different from other losses, described in info box:

To get the expected AEP sold to grid, as well as grid losses as losses due to turbine availability, curtailments,

etc., must be subtracted. This can be done in a PARK calculation as a “lump sum” - a simple percentage

reduction. But a far more comprehensive evaluation can be done by importing the PARK results into the LOSS

AND UNCERTAINTY evaluation module in windPRO. This partly makes sure all possible losses are judged and

partly that there will be specific calculation options helping to get the losses more precisely calculated. An

example is the high wind hysteresis, which can be calculated based on the wind distribution and entered

controller settings. High wind hysteresis is about how low the wind speed shall go and for how long of a time

before a turbine restarts after a high wind, cut-out event.

The uncertainty can also partly be judged, partly calculated, by help from the LOSS AND UNCERTAINTY

module. Here are more refined calculation options, like calculation of the uncertainty based on the distance

between measurement mast and each turbine.

Introduction, definitions and step-by-step guide

After calculating the expected AEP (Annual Energy Production) with the windPRO PARK module, the next step

to bring a wind farm project to a “Bankable” level is to estimate losses and uncertainties. Losses have the recent

years become a more and more important part of the AEP estimate, partly because the losses typically are

higher for modern wind farm projects, partly because the margin in AEP estimates has been lowered due to

larger project sizes, and more tight budgets for wind farm projects. While wind farm investments have increased

heavily, the need of knowing the uncertainties similarly has become of huge importance to get the projects

financed.

With the windPRO LOSS & UNCERTAINTY module the estimation of expected losses and uncertainties can be

performed on a structured basis, with numerous tools for quantifying the individual components quite accurately.

Besides losses and uncertainties, the module also offers a Bias correction part. A Bias is a “known issue”, like

model problems (e.g., RIX correction) or wind speed measurement bias, which have not been corrected in the

calculation basis.

3.8.1.1 Basic definitions

The basic concept behind the module is:

Calculated GROSS AEP

+/- BIAS correction

- LOSSES

= NET AEP (expected sold energy production) = P50

The expected NET AEP is also named the P50 value, which is the expected outcome of the project. There is a

probability of 50% that the outcome will be more than P50 and a probability of 50% that the outcome will be less.

This can also be named the “central estimate”. The uncertainty must be judged/calculated to find out how

accurate the estimate is, and thereby the risk of getting a lower outcome than expected.

Loss & Uncertainty 164

Including the uncertainty, the AEP estimate is assumed to follow a normal distribution. All uncertainty

components are assumed independent and, thus, combined as standard deviations, i.e. the square root of

summed squares of individual contributions. The individual uncertainty components, judged or calculated, shall

be given as 1 std dev (Standard Deviation or simply σ).

If the std dev (hereafter σ) is 10%, this means that the production at a given AEP exceedance level (PXX) for a

calculated result can be calculated using the inverse normal distribution as:

P84 = P50 – 1 x Uncertainty (= P50 - 10%, for σ=10% as above)

P90 = P50 – 1,28 x Uncertainty (= P50 - 12,8%, for σ=10% as above)

Below are listed additional coverage factors for other typical exceedance levels (e.g., 75%), all based on the

normal distribution.

Prob. (%)

Coverage factor

(0,00)

0,67

84 *)

1,00

1,28

1,64

2,33

*) For P84,00 the coverage factor is not exactly 1,00 but 0,99. The coverage factor 1,00 corresponds to P84,13

which we round off to P84 here for convenience.

A special component in the uncertainty evaluation is the year-to-year variability of the wind, which can be

included in the calculations. The variability describes how much the annual average wind speed varies from

year-to-year for the region. This figure can be calculated in the MCP module based on long term data series, or

it can be found in different research projects.

The expected probability of exceedance is calculated for 1, 5 10, 20 years with the variability for the time span

in question included in the uncertainty. Contrary to the other uncertainties the variability depends on how many

years the forecast covers, referred to as “expected lifetime”. This can be of importance for judgment of the risk

of the investment.

3.8.1.2 Understanding the uncertainty concept (Probability of exceedance)

The uncertainty concept is well illustrated by the figure below.

Loss & Uncertainty 165

Figure 1 Based on calculations of 1806 wind turbines in Denmark, the count of goodness factor

(Actual/calculated AEP corrected with wind energy index) for each turbine shows that the actual results

are close to a normal distribution with a σ of 8,1%. In other words the uncertainty for these calculations

is 8,1%.

Figure 2 Illustration of the Normal distribution

The normal distribution is defined so that roughly 2/3 (more precisely 68,3%) of all events will be within +/- 1σ

and around 32% is outside. In the one tail (e.g., below -1σ), there is around 16%, so there is 16% probability that

100

150

200

250

0,7 0,8 0,9 1 1,1 1,2 1,3

Goodness for: kW: 600 - 2500 Goodness: 0,7 - 1,3 Count: 1806

Average: 0,99 σ: 8,1%

Count Normal dist.

70% 75% 80% 85% 90% 95% 100% 105% 110% 115% 120% 125% 130%

Number out of 100

AEP relative to P50

Normal distribution with σ= 10%

P84

16%

P95

70% 75% 80% 85% 90% 95% 100% 105% 110% 115% 120% 125% 130%

Number out of 100

AEP relative to P50

Normal distribution with σ= 5%

P84

16%

P95

Loss & Uncertainty 166

the estimate will be below 1σ subtracted from P50, or 84% probability that it will be above (exceed). In other

words, the P84 is the value where 84 out of 100 realizations will result in an outcome better than P84. For P95,

there is only 5% probability to get an outcome poorer than this exceedance level which is found by subtracting

the σ multiplied by 1,64 from the P50. So for σ=10%, the P95 value in the left graph is found where 5% is in the

shaded area (P95). This would be found 16,4% below 100%, i.e. at 83,6% on the x-axis.

Similarly, if σ=5%, 5% x 1,64 = 8,2%, so P95 is found at AEP of 100%-8,2% = 91,8% of the P50 on the x-axis.

Figure 3 The probability of exceedance graph

The probability of exceedance will normally be shown as a cumulative graph showing the probability of

exceedance on the y-axis and the corresponding AEP PXX values on the x-axis.

3.8.1.3 What is included in GROSS value?

The module follows the DNV (Det Norske Veritas) definition as presented at AWEA 2008:

Included in GROSS calculation:

- roughness effects

- topographic effects

- obstacle effects

- air density correction

- (long term correction)

- (wind data correction)

Last two should be included, but it is up to the user to decide what is included. If e.g., a post calibration show

that the wind data has been offset, it can be decided to redo PARK calculations with updated wind data or it can

be decided to include the offset as a Bias correction of the GROSS.

NOT included in GROSS calculation:

- Wake losses (The PARK result includes wake losses, but these are “taken out” in the loss module so

the “real Gross” based on the DNV definition is used as basis for all loss reductions.

- Other losses like availability, grid losses etc., see complete list below.

- Model issues like RIX correction or known power curve bias, will should be included as Bias, not as

Losses, because these are considered “known issues” and should thereby be treated as corrections to

the calculation results applied before the loss evaluation.

- Curtailments included in PARK calculation are treated similar to wake losses.

The structure of the module set demands to the user keeping track of what has already been compensated in

the PARK AEP calculation, and what should be added in the loss, bias and uncertainty evaluation. The only

“automized” issue is that the wake losses are taken out of the windPRO PARK AEP calculation (the size of the

wake loss is automatically filled in), so the LOSS & UNCERTAINTY module starts from the non-wake loss added

wind farm AEP calculation result.

Loss & Uncertainty 167

The module has these features:

1. All Bias, loss and uncertainty components can be judged by the user and entered manually.

2. Some of the components can be calculated by the software based on different data sources, typically

wind data time series.

The wind data time series are used to divide the expected AEP in time steps, to enable calculation of time, wind

speed or wind direction dependent losses. But also links to other windPRO calculations like SHADOW can be

used to give an accurate estimate of AEP loss due to flicker stop, or a PARK RIX calculation can be used to

perform a RIX bias correction.

3.8.1.4 Loss definitions

The loss definitions in the module follow the below definitions (in italic the EMD modifications). Note we have

switched group 1 and 2 relative to the original paper so Wake effects occur first and availability second.

Paper, AWEA 2008: Standard Loss Definitions for Wind Resource / Energy Assessments

Prepared by Steve Jones of Global Energy Concepts (DNV)

Standard

Loss Category

Recommended

Subcategories

Comments

1. Wake Effects

Wake effects, all

WTGs

Losses within the turbines which are the subject of the energy assessment.

Helimax currently includes wake losses in the gross yield. Losses on the

turbines which are the subject of the energy assessment, from identified

turbines that are not the subject of the energy assessment, which either

already operate or which are expected to operate the entire useful life of the

facility being studied.

If the PARK calculation includes existing turbines (which it should), the wake

losses from as well internal as external wake effects are included in the wake

loss calculation, therefore the EMD has brought the two groups from original

document into one.

1. Wake Effects

Future wake

effects

Losses due to additional development in the vicinity of the turbines being

studied, but which would occur after commissioning of the turbines being

studied.

2. Availability

Turbine

GEC further divides this into routine maintenance, faults, minor components,

and major components. AWS Truewind uses a separate factor (Availability

Correlation with High Wind Events) that could be buried into this number or

categorized with “7. Other” below.

2. Availability

Balance of plant

(Substation)

Losses due to downtime in components between the turbine main breaker to

and including project substation transformer and project-specific transmission

line.

2. Availability

Grid

Losses due to downtime of power grid external to the wind power facility.

2. Availability

Other

Other availability losses not accounted for above or in other categories below.

3. Turbine

performance

Power curve

(can be part of

Bias)

Losses due to the turbine not producing to its reference power curve (even

with new blades and wind flow within test specifications).

3. Turbine

performance

High wind

hysteresis

Losses due to shutdown between high-wind cutout and subsequent cut back

in.

3. Turbine

performance

Wind flow

Losses due to turbulence, off-yaw axis winds, inclined flow, high shear, etc.

These represent losses due to differences between turbine power curve test

conditions and actual conditions at the site.

3. Turbine

performance

Other

Other turbine performance losses not accounted for above.

4. Electrical

Electrical losses

Losses to the point of revenue metering, including, as applicable, transformers,

collection wiring, substation, transmission.

4. Electrical

Facility

consumption

Losses due to parasitic consumption (heaters, transformer no-load losses,

etc.) within the facility. This factor is not intended to cover facility power

purchase costs, but does include the reduction of sold energy due to

consumption “behind the meter.”

5. Environmental

Performance

degradation not

due to icing

Losses due to blade degradation over time (which typically gets worse over

time, but may be repaired from time to time), and blade soiling (which may be

mitigated from time to time with precipitation or blade cleaning).

5. Environmental

Performance

degradation due to

icing

Losses due to temporary ice accumulation on blades, reducing their

aerodynamic performance.

Loss & Uncertainty 168

5. Environmental

Shutdown due to

icing, lightning,

hail, etc.

Losses due to turbine shutdowns (whether by the turbine controller, SCADA

system, or by an operator) due to ice accumulation on blades, lightning, hail,

and other similar events,

5. Environmental

High and low

temperature

Losses due to ambient temperatures outside the turbine’s operating range.

(Faults due to overheating of components that occur when ambient conditions

are within the turbine design envelope would be covered under turbine

availability category above.)

5. Environmental

Site access and

other force

majeure events

Losses due to difficult site access due to, for example, snow, ice, or remote

project location. Note that this environmental loss and some other

environmental losses may be covered under the availability definition, above.

However, these “environmental” losses are intended to cover factors outside

the control of turbine manufacturers.

5. Environmental

Tree growth or

felling

Losses due to growth of trees in the facility vicinity. This loss may be a gain in

certain cases where trees are expected to be felled.

6. Curtailment

Wind sector

management

Losses due to commanded shutdown of closely spaced turbines to reduce

physical loads on the turbines.

6. Curtailment

Grid curtailment

and ramp-rate

Losses due to limitations on the grid external to the wind power facility, both

due to limitations on the amount of power delivered at a given time, as well as

limitations on the rate of change of power deliveries.

6. Curtailment

Power purchase

agreement

curtailment

Losses due to the power purchaser electing to not take power generated by

the facility.

6. Curtailment

Environmental,

Noise

Losses due to shutdowns or altered operations to reduce noise and shadow

impacts, and for bird or bat mitigation. This would include use of a low-noise

power curve vs. a standard one from time to time.

For Noise and flicker, there are in windPRO access to detailed calculation

options. Therefore EMD has expanded this with more groups, same for Birds

and Bats, which can be set based on “free of choice” parameters like date

interval, hour interval etc.

6. Curtailment

Environmental,

Flicker

6. Curtailment

Environmental,

Birds

6. Curtailment

Environmental,

Bats

6. Curtailment

Temperature

Derating

Losses due to the turbine needing to throttle at rated power at certain

temperature ranges and elevations. Calculated in time-varying PARK.

6. Curtailment

Other curtailment

7. Other

This would cover anything that doesn’t fit into the above six main categories.

Step-by-step guide

❑ Establish a PARK calculation (see Energy, Section 3.3.5), BUT note the following:

If more site data objects or turbine types are used, group these in separate layers before calculation.

If RIX Bias correction should be included, make the RIX calculation in PARK

If calculation of time dependent losses etc. makes sure you have a proper time series with required data.

The WTI generator in Meteo analyzer tool may be used to establish this. Include temperature if high/low

temperature shut down is expected. Include turbulence or gust if high wind hysteresis loss is expected.

❑ Start LOSS & UNCERTAINTY module

❑ Load PARK calculation

❑ You may attach a wind data time series from Meteo object or WTI file

❑ Input needed parameters in Bias, Loss and Uncertainty tab sheets

❑ Where “Edit” buttons are available, start detailed calculations – you might need to go back to “Main” to

reselect the wind data to more a or less detailed series.

❑ When all inputs are established, review at “result” tab and start calculation for generating report by OK.

Basic data for calculations

A PARK calculation is the basis. From this all relevant data on AEP for each turbine, wake loss, elevation, hub

height etc. are loaded. In addition, sensitivity is calculated. The sensitivity defines the transfer from changes in

wind speed to changes in AEP for each turbine by recalculation of the PARK with a small change in wind speed.

It is worth to notice that if a RIX bias calculation is wanted, the PARK calculation loaded must hold a RIX

Loss & Uncertainty 169

calculation. Similarly, if a flicker stop loss calculation is wanted, there must be a shadow calculation for exactly

the same wind farm configuration as in the loaded park calculation. For Noise loss calculation it must be noted

that if the PARK calculation already includes turbines in noise reduced mode, no additional Noise loss should

be entered. If it is a wish to present the loss due to noise in the loss calculation, the PARK calculation must be

based on no noise reduced turbines, and the noise reduced modes must then be implemented in the loss

module. We are aware of this is somewhat “tricky” and the handling of noise loss calculation will be improved

further in the future. If for example the noise reduced mode only occurs during e.g., night hours, the

Loss&Uncertainty module is very convenient to use, as the calculation setup allows limiting the noise reduced

mode to specific hours or wind directions.

In addition, following data can be used:

Climate data as time series: Either by link to a Meteo object time series or to a .WTI file (Wind TIme variation file

that can be established from the Meteo analyzer or selected from the windPRO Data\Standards\ folder). From

windPRO 3.3 selection of datasets for different loss calculation parts can be selected individual for each loss

calculation type.

Power curve uncertainty can be specified detailed in the WTG catalogue and used from uncertainty module, but

also simpler approaches for this calculation are available if no detailed data are available for the turbine.

Figure 4 The Main tab where PARK calculation is loaded.

When loading a PARK calculation, it can be decided if existing turbines included in PARK shall be included in

the loss & uncertainty evaluation. Further it is possible only to include the existing turbines if these are flagged

“treat as PARK WTG” (property on existing WTG objects).

This tab shows the main results from the PARK calculation and the calculated sensitivity for propagation of

changes in wind speed to changes in AEP (AEP%/ws%).

Loss & Uncertainty 170

Figure 196 Decide how to handle existing WTGs.

Checking the “Use advanced loss calculation tools…” gives access to add time varying data or to assume

constant power. The last option is used if no time varying data are available, but the user still wants to calculate

flicker or temperature loss assuming constant AEP in each time step.

With the time step PARK calculation concept, it is obvious to use a time step PARK calculation as input. This

must although in the Setup be set to non-averaged output or hourly averaged. Default is monthly. The time step

PARK calculation should be long term representative, meaning being calculated based on e.g., 1 or 3 y LT scaled

data or e.g., 10 or 20y calibrated mesoscale data.

The expected lifetime only influences the uncertainty contribution from the variability of the wind. All other

calculations are based on annual averages. The uncertainty component coming from the year to year variability

decreases with the number of years and will thereby be lower the longer the lifetime (part of the variability is

averaged out).

3.8.3.1 Climate data

Several loss calculations are based on climate data including also temperature data. Of high importance is that

the climate data represent a typical year. A time series averaged over several years will not hold the information

of the dynamic behavior that is of high importance for the expected shut down situations of the wind turbines.

However, calculations can be performed based on more than 1 year of data in a Meteo object.

In a Meteo object, a dataset of one or more years of data can be established from more years of data by disabling

data, so that one or more representative year(s) is enabled. If the data series merely holds a ½ year or 1½ year,

the calculation will be seasonally biased. This should be avoided. But no matter how long (or short) period of

data the Meteo object used hold, it is important to remark that the calculations always will assume these data as

long term representative and scale the calculations to annual values.

A specific way to establish exact 1 year of data prepared for such analysis is found in the Meteo analyzer. Here

you can generate exactly 1 year of data with a specific temporal resolution (data can be down or up sampled)

based on one or more time series in Meteo objects. See further details in the chapter on time varying data from

Meteo analyzer.

Climate data can be selected individually for each loss calculation type. It the best (most precise) wind variation

time series e.g., not have temperature information, a temperature loss calculation (shut down at low/high

temperature) can be calculated based on e.g., Merra data which includes both wind and temperature. It is

although important that wind data must exist in the meteo object used in each specific calculation, while it is

concurrent wind and e.g., temperature, that decides the loss.

3.8.3.2 Model results

Figure 5 At the Model results calculated Gross is shown.

Loss & Uncertainty 171

Figure 197 For "Standard" PARK also model modifications shown.

Depending on which PARK variant is loaded, there can be more details shown on Model results, e.g., the effect

of the layout. This can give an idea of how large model corrections are applied. The software calculates the AEP

if all turbines are positioned at the position(s) of the site data object(s) – if more than one site data objects, the

turbines “belonging” to a specific site data object is moved to this object and calculated. This is compared to the

actual calculation with the turbines at their “real positions” (the Layout). Thereby it can be seen how much the

model transforms data based on roughness, orography and local obstacles. The higher the effect of the layout,

the higher the risks are of errors in calculations if the model does not perform accurately. In other words, the

more measurement masts the calculation is based on, the lower the effect of the layout, or the lower complexity

of terrain/roughness, the lower the effect of the layout.

Note the results here are EXCLUDING wake losses and curtailment losses included in PARK; these are simply

taken out of the calculation and transferred automatically to the LOSS sheet where belong according to the DNV

standard definitions.

3.8.3.3 General concept for input of data in bias, loss and uncertainty sheets

In general, there can only be entered one value for a loss/bias/uncertainty to represent the entire wind farm. But

if a calculation module is available (i.e. an “Edit” tickbox), values can be entered in a more flexible way:

• Individually for each turbine

• For all turbines on a specific layer (in Maps&Objects)

• For all turbines

This means that if there is a need for specific data on half of the turbines and other values on the other half, it

would be a very good idea to place these two groups in different layers in the project setup. A lot of individual

input can then be avoided. An example could be if a wind farm is established with 2 or 3 different wind turbine

types or if e.g., one group is more exposed (on a ridge) than another group and therefore needs a lower cut out

wind speed value.

Input of data for an individual turbine or for all turbines in a layer is simply selected by clicking with the mouse

on the individual turbine or on the specific layer. The input will then be assigned to the selected turbine or group.

Bias

Bias is a correction for “known issues”, like e.g., the RIX (Ruggednes IndeX) modifications of wind speeds in

complex terrain introduced by RISØ, or e.g., power curve correction, where those are known to be too pessimistic

or optimistic based on experience or evaluation by the HP method. Also wind measurements can have a known

bias. For example, specific anemometers are known to have a systematic error, or post calibration could show

an error, in both cases it is more convenient to include these corrections as biases than by reanalysing all the

wind data behind the calculations. It is important is that bias corrections only are included once, either in the data

basis of the PARK calculation or as a bias in the LOSS & UNCERTAINTY module. An advantage by having bias

corrections in the LOSS & UNCERTAINTY module is that it will be clearly documented, and easy to change if

new information appears at a later stage.

Loss & Uncertainty 172

A bias can be entered as a simple correction in percent either on wind speed or in percent on AEP. If entered

as wind speed percentage, this quantity is converted to percent on AEP using the sensitivity AEP%/WS%

(WS=Wind Speed). The AEP percentage is then multiplied with calculated GROSS and added (or subtracted)

to GROSS before loss subtraction. Remember that a bias can have a positive or negative value - so do remember

the sign.

Figure 6 The input form for Bias.

As seen above five different predefined bias input lines are available. If PARK calculation includes a RIX

calculation, but NO correction, there will be a “calculate” tick box option for this (see next chapter for details). If

PARK calculation includes RIX correction, this is not treated as a bias in Loss&Uncertainty. However, there will

be an information in the comment box how much RIX correction was included from the PARK calculation.

Wind speed correction

If the wind data is known to have a bias, which has not already been corrected in the wind data used for

the PARK calculation, the correction should be included here.

Wind data bias can have many reasons and is probably the most frequent reason for biased calculation results.

But it can be very difficult to discover such a wind speed bias. The best method to avoid wind bias is to have

more wind data sources for the site/region. Existing turbines with available production figures present near the

site is also a valuable source of validation of the wind data level.

If local wind measurement equipment is used, the wind data correction can simply be due to known offset related

to the equipment used. Often this will be corrected for at previous stage in calculations, if so it SHALL NOT be

entered as a bias, the correction would then be double. But it is a good idea to write a comment if corrections

are performed before PARK calculation, or if any validation of the wind speed level is made.

The correction can be entered as a modification on wind speed or AEP, remember to include the sign (- if it is a

reduction) of the bias, because corrections can go in either directions.

RIX correction

For details, see 12.3.2. If a RIX correction proposal is made via another tool than windPRO, or it just is a rough

user estimate, the correction can be entered here. But it will then ONLY be as a common correction that will be

the same for all turbines. So if RIX correction is issues always make a PARK calculation with RIX and use the

correction calculation tool described in “RIX correction calculation”.

Model problems for very large wind farms

This topic is less relevant today, where comprehensive tests on large wind farms show the wake models handles

these reasonable well, see e.g., http://iopscience.iop.org/article/10.1088/1742-6596/524/1/012162/pdf. It seems

the problem issue previous seen more are related to low turbulence, which require use of lower WDC in PARK

calculation. The right choice of WDC seem to solve this. But if it is assumed that the PARK calculation not

includes the full expected wake losses, a bias expectation can be entered.

Power curve correction

If it is known that a power curve is too optimistic or pessimistic, a simple correction should be entered here.

Other

Any other issues that the user knows is a bias in the calculation should be compensated here.

Loss & Uncertainty 173

3.8.4.2 RIX correction calculation

For the RIX correction a calculation module is established. The main source for the implementation is:

EWEC06 paper:

IMPROVING WAsP PREDICTIONS IN (TOO) COMPLEX TERRAIN

By Niels G. Mortensen, Anthony J. Bowen and Ioannis Antoniou

Wind Energy Department, Risø National Laboratory

This paper describes why complex terrain with steepness > 30-40% violates the WAsP model calculation

method, and how calculation accuracy can be improved by applying the RIX correction method.

Figure 7 Input of details for the RIX correction calculation and main results.

The basic formula is: U

= U

x exp(-α x ∆RIX), where U

is the predicted wind speed using WAsP and U

(measured) is the corrected wind speed. The parameter α is found empirically (e.g., via cross prediction tool in

Meteo analyzer in windPRO) and ∆RIX is calculated by windPRO in a park calculation based on the elevation data

at the site. The key issue is to estimate the α value and to decide the radius and slope threshold for the ∆RIX

calculation. Given these the RIX correction is simple math. The calculation tool finds the appropriate (given α

and ∆RIX) correction of the wind speed at each WTG position and converts this to an AEP modification based

on the AEP%/ws% sensitivity for each WTG position. The calculated modification will be stored individually on

each WTG.

Loss

Loss is the AEP that should be produced based on the available wind and the turbine power curve, but never

reach the “sales metering”. Partly due to physical losses such as grid losses, partly due to wake losses, where

turbines take wind from each other and partly due to reductions in turbine operation, e.g., due to shut down at

low temperatures or availability losses when out of order.

The seven loss main groups defined by DNV are listed in the Intro chapter. Here is how the general calculation

runs.

For each turbine a given loss component is converted to efficiency, i.e. a 3% loss is converted to 100%-3% =

97% efficiency. This is done turbine by turbine. The efficiencies from each component are then multiplied and a

resulting efficiency found. This is multiplied with the GROSS AEP after Bias correction, if any. Then the NET

AEP = P50 is the result of the loss reduction.

Loss & Uncertainty 174

Pre-calculated losses, Wake and Curtailment, included in PARK calculation, that already account for their

interaction, are summed prior to combination.

Figure 8 The loss input screen.

The loss input screen holds seven main groups that can be expanded for input of the relevant loss estimates.

Some input lines hold a “Calculate” option. When checked, the edit button opens a form for detailed calculation

of the loss due to the specific component.

For the losses that can be calculated by the module, a more detailed description of calculation method follows.

For all components a comment can be added. This is an important part of the loss evaluation. In the report all

lines with comments will be shown, so the user can see the background for the evaluation even if no loss is

assumed due to the specific component.

Besides what can be calculated, it is of high importance to emphasize that two loss components always should

be included:

1. Turbine availability, typically 2-5%, depending on service arrangement and turbine quality.

2. Grid losses (can be calculated with eGRID module), will typically be 1-3% depending on distance to

meter point, and if e.g., staff house consumption should be included. Note that the power curves used

in PARK-calculation are measured at the low voltage side of the turbine transformer, so the turbine

transformer losses should always be included; this alone is typically between 0.5 - 1%.

Loss & Uncertainty 175

3.8.5.1 High wind hysteresis

Figure 9 High wind hysteresis loss calculation.

Note the feature for adding a separate time series for this specific calculation. The data used for distributing the

production in time (e.g., a time varying PARK calculation), might not hold the needed data for this specific

calculation type. For the High wind hysteresis, the turbulence is of importance for redistribution of 10-min data

to minute and gust data. Therefore, an alternative time series might be needed.

High wind hysteresis loss is where the turbine is stopped below the cut-out wind speed. All stop time above cut

out wind speed (defined in power curve) are already corrected for in AEP calculation. But high wind speed

shutdown events can cause significant fatigue loading. Therefore, to prevent repeated start up and shut down

of the turbine when winds are close to the shutdown threshold, hysteresis is commonly introduced into the turbine

control algorithm. Thereby losses in relation to AEP calculation are introduced. The setup of the stop/start

procedure must be confirmed by the turbine manufacturer. This is individual from turbine type to turbine type but

is sometimes also set individually from site to site.

The threshold for stop and restart of the turbine can be defined on different time resolution parameters: “Gust”,

“Minute” or “10 minutes values”. The latter is the standard mean wind speed as defined in the wind data selected

on the main tab (a meteo object or a wti file). The “Gust” values may be estimated from measurements using

the max of each 10-minute interval (often logged as “maximum mean wind speed”). However, the averaging

period of such estimates is unknown. The IEC61400-1 standard for turbine design requires 3-second averages

to be used for gust estimates. So instead of maximum 10-measurements 3-second gust estimates may be based

on a simple model originally introduced by Davenport. The method uses the formula below to estimate the gust

at averaging time t.









  





Where K

is the peak factor dependent on the averaging time t for which the conversion is required (defined in

the article BELJAARS, The Influence of Sampling and Filtering on Measured Wind Gusts, 1987)

The same equation is used for the “Minute” parameter, allowing to define the stop and restart threshold on an

averaging period of 1 or 3 minutes for example.

Once the time resolution is selected the values of the threshold for stop wind speed and restart wind speed (with

their respective averaging periods must be defined).

“Restart delay after u<urestart” is a safety margin in time to prevent the turbine restarting too rapidly if the wind

shortly after increases again.

“Restart delay after u>ustop” ensures that the wind has decreased steadily below the restart speed before it

really starts again. This prevents the turbine to to restart too shortly (and may be unnecessarily) again.

Loss & Uncertainty 176

Note that if all three “Time resolution” parameters are selected, they must all be fulfilled for a WTG to “restart” in

windPRO’s calculation of the hysteresis loss. But a WTG will stop if any of the stop criteria is fulfilled.

Calculation:

windPRO runs through the time series, find the events that triggers cut-out and logs the period the turbines would

have to stop until restart criteria are fulfilled.

The loss is calculated and normalized to a full year.

3.8.5.2 Degradation losses due to Icing

This loss is not yet established as a calculation feature, but there are made a quite comprehensive work in

relation to this. Based on meso scale modeling with focus on Icing calculations, the needed output can be

established. A comprehensive validation of the results has been made, see EMD-WRF On-Demand ICING -

Wiki-WindPRO for further details. It is e.g., possible to calculate an icing loss map with expected AEP loss for a

specific site, see example below.

Figure 198 Example of calculated icing loss with EMD WrfOnDemand.

See also manual Chapter 4 Cluster Services.

3.8.5.3 High and low temperature

AEP is calculated for each time step in the climate data time series based on a scaling of the wind speed to the

calculated average wind speed for each turbine. The AEP results based on this method is then scaled so the

annual sum equals the main calculation result.

Based on entered shut down threshold temperatures, the AEP calculated for each time step is summed for all

time steps outside the temperature threshold values. The AEP loss sum is then converted to a loss percentage

that is saved for each WTG.

Loss & Uncertainty 177

Figure 10 The loss due to high/low temperature is calculated.

In the example above it is seen that the time series with temperature vary in the range from -3.6 to 26.7 degrees.

The setting for temperature shut down is below -0 and above +35 deg. C (this is set low just to illustrate the

calculation). The loss is calculated to 0.8% based on an AEP calculation for each time step, where the time steps

with temperatures outside operation range is summed and included as a loss. The loss calculation for each

turbine is shown, in this example it is almost same for all turbines while only one time series can be handled.

Note the temperature is taken from mesoscale model data in this example, see bottom of form.

3.8.5.4 Wind sector management

Wind sector management is stop of turbines when the wind comes from specific directions within specific wind

speed intervals, to prevent damage of neighboring turbines due to wake added turbulence due to dense spacing.

From windPRO 3.2 it is possible to perform this calculation within the PARK calculation. This will then be loaded

into Loss & Uncertainty and treated like wake losses, as pre-calculated loss. It will although still be possible to

input sector management settings and calculate losses direct in L&U module if not calculated in PARK.

This is quite complicated to input, while it is individual from turbine to turbine. Below is seen an example where

all turbines in the East group have the same settings. But to input this realistically, there must be an individual

input for each turbine based on e.g., a Site Compliance calculation. By mouse click at one specific turbine

(highlighting this), the settings in the field above will only relate to this specific turbine. For a large wind farm this

work is quite troublesome. Therefore, it is possible partly to import from a file. These data could have been

established within the Site Compliance analyses.

The” right” way from windPRO 3.2 onwards will be to include sector management in the PARK

calculation. Then curtailment settings can be taken from WTG objects. The curtailments will decrease

wake losses, and the resulting curtailment and wake loss are transferred and handled correctly in L&U

module.

Loss & Uncertainty 178

Figure 11 Wind sector management, flexible input options.

For wind sector management there are two variants:

• Based on wind statistic calculation

• Based on time step calculation

Only one of these should be used.

Figure 199 Statistic based wind sector management.

For the statistic-based sector management there is an option to set a hysteresis. This works so: In a 2-degree

hysteresis, there are calculated loss 2 degrees outside the specified interval at each “side”. But only the half of

this loss is included. In that way is simulated if e.g., the wind direction gradually changes from 178 degrees to

180 degrees there is no stop before 180 degrees is reached. But after stop at 180 degrees, and the wind direction

rotates back, the turbine is first back in operation when the direction is below 178 degrees. This feature is not

included in time step based, but the user can just manually set the angle according to above example to 179

degrees, and almost same result will be seen.

3.8.5.5 Grid curtailment and ramp-rate

In case of included grid curtailment loss in PARK calculation, this will just appear in the list as information. The

Gross result transferred from PARK to Loss & Uncertainty module will be without the grid curtailment and applied

as a loss afterwards. This is due to the grid curtailment being calculated independently from wakes and WTG

curtailments. If grid curtailment is not included in PARK calculation, a value can be entered. This can aside from

limitations in grid capacity also hold losses due to restrictions on how fast the power delivery are allowed to

change, ramp-rate.

3.8.5.6 Power purchase agreement curtailment

If the power purchaser not always can take all production. This can be a quite complex calculation with several

market regulation mechanisms.

Loss & Uncertainty 179

3.8.5.7 Noise

Figure 12 Input for noise loss.

Some turbines might run in noise reduced mode, maybe only within specific time of day, maybe only at certain

wind directions (or combinations). Besides time and direction interval, the noise reduced power curve (or no

power curve meaning full stop), can be selected. A tricky issue here is if the PARK calculation already is

calculated with noise reduced power curves. In this case, there shall not be entered noise loss. So to include the

noise reduced mode loss correctly, the PARK calculation must be without noise reduction, and the noise reduced

modes selected here. In future version, there will be an option to treat noise reduction the same way as wake

reduction, meaning the software first takes out the effect of noise reduction, and transfers the loss, where it is

automatically set up. This will also include L

den

calculations, where different settings for day, evening and night

will be required. Note in the power curve selection field, there will be information telling with which power curve

each turbine has been calculated.

3.8.5.8 Flicker

Figure 13 Setup of input for shadow flicker stops loss calculation.

Loss due to stop caused by flicker at neighbors is simple to performed. A Shadow calculation for exact the same

wind farm layout as PARK calculation is based on, must be loaded. Then time step by time step it is checked if

Loss & Uncertainty 180

there is flicker at a neighbor and the loss due to stop within flicker time is calculated. The calculation is based

on the turbine running in “calendar mode”, meaning all possible events of giving flicker is included (worst case

calculation). If a more advanced flicker reduction mode is implemented, the stops will be less, and a simple

reduction due to this can be entered – typically around 50%.

3.8.5.9 Bat

Figure 200 Setup of Bat stop loss calculation.

A special feature for calculating Bat curtailment, sunrise and sunset can be included. But note that two separate

lines must be included, while conditions in one line is AND, and therefore like 2 hours before and after sunrise/set

would not work if entered in the same line. The AND would eliminate the two options.

The shown setup above illustrates how a curtailment calculation is set up based on:

From 15-04-yyyy to 15-10-yyyy, 2 hours before and after as well sun set as sun rise, the turbine will be stopped

if wind speed is below 6 m/s.

Note the calculation is performed based on the time varying dataset that is chosen in the setup at the main page.

This should always include ONE FULL YEAR. So it is if e.g., a .wti file is generated from the Meteo analyzer.

But if more years, it will still work, although note the year will be ignored, it only looks at the dates and assume

same stop period each year.

Note also the wind speed checked against is SCALED to the calculated wind speed at the turbines in hub height.

If the regulations refer to e.g., 10m height wind speed, the user must modify the limits, e.g., if stop demanded

below 4 m/s at 10 m height, a stop below 6 m/s in turbine hub height (e.g., 80m), must be used. Example: (look

in manual for details, search for “Shear”)

Shear

0,2

Height

Wmean calculated

4,0

4,5

4,9

5,2

5,5

5,7

5,8

Loss & Uncertainty 181

6,0

6,1

100

6,3

110

6,4

120

6,5

The table show the wind speeds at different heights, by a typical shear of 0.2. Here having 6 m/s in 80m height.

An often-seen problem although is that the authorities that set the criteria’s, forget to mention for which height.

Then most often it is just assumed it is for hub height.

Note if a field is entered with no value having “-“ in the field, it mean “all included”. If eg. Sunrise (from hours

before) is set to “-“ it takes all the hours before sunrise into the calculation. (From midnight).

The data for when sunrise – sunset appears is calculated based on geometry from the site center position (same

formulas as used in the Shadow flicker calculation without a horizon angle threshold.)

3.8.5.10 Other

Figure 14 “Other” gives large flexibility for loss calculation depending on any parameter combination.

Using “Other” any parameters available in Meteo object or .WTI file can be used for setting up any parameter

combination. In the example above, stop for the one group of turbines is every Sunday between 10:00 – 11:00

if wind direction is between 160 and 200 degrees. This could be when wind blowing towards the Church within

church time.

Loss & Uncertainty 182

3.8.5.11 Manual entering losses by turbine

Figure 201 Checking "Manually", individual loss per turbine is entered.

For the loss types where calculation features are available, it is possible to enter losses by turbine manually.

This is relevant if calculated by other tools, e.g., Icing loss calculation.

Loss & Uncertainty 183

Uncertainty

Uncertainties are grouped in 5 groups,

A. Wind data

B. Wind MODEL

C. Power Conversion

D. Bias

E. Loss

Each of those groups must be judged, and as for bias and loss, some groups have calculation features which

will be described in separate chapters. In later versions, more calculation features will be implemented.

Before going to the calculation features, the Wind data group will be explained, while this is one of the more

important ones.

Figure 15 The five uncertainty groups A-E.

For category E all lines with a loss entry is shown.

3.8.6.1 Wind data uncertainty

Wind data can be used in the PARK calculation in different ways:

• Measurements on site, typically along with a long-term correction.

• A wind statistic for the region, possibly verified/calibrated based on performance from existing turbines

in the region.

• A wind resource map, based on a model, like mesoscale model, CFD model or WAsP model – behind

the wind resource map there will be wind data, that can be based several different sources.

Loss & Uncertainty 184

To judge the quality of the wind data is probably the most essential part of the uncertainty evaluation. If turbines

with longer operation period (>1y) exists in the region, a test calculation with the used wind data is one of the

best ways to reduce uncertainty of the wind data basis. It is essential that the production from these turbines is

properly long term corrected and cleaned for availability problems. If the actual cleaned production from those

can be reproduced accurately, the uncertainty on the wind data can be assumed small.

If only local measurements are available, the uncertainty depends much on measurement equipment, mast

configuration, sensor calibration and quality. Long term correction is normally a must, but here additional

uncertainties are introduced, while the long term sources often are of poor quality, and might even be trended,

if e.g., trees has grown up around the reference mast of if it is modeled data there might be trends due to changes

in the data basis for the model. Such trends should NOT be considered just as an uncertainty, but should be

corrected for up front or included as bias correction.

Even with high quality data, the wind measurement uncertainty should not be assumed lower than 2% on wind

speed - an “upper limit” is difficult to give. If it is a low wind site, the wind speed uncertainty converted to AEP

uncertainty can be as high as 3 times the wind speed uncertainty, while it at a high wind site only will be 1,5

times the wind speed uncertainty.

A specific source of uncertainty is the position on the measurement mast. If the mast location is in a hilly

environment, it is crucial that the position is correct, and that the elevation information around the mast is

accurate. It is often seen that measurement masts are placed on a small hill top. If the elevation data are rough,

the little hilltop is not included in the data and an error is introduced when cleaning the data based on orography.

This is not an uncertainty but an error that must be handled by establishing the elevation data round the mast in

a correct way. Photomontage tool should be used to verify that the elevation data round the mast is correctly

established. If the mast position is uncertain, this should be included in uncertainty, for instance in “Other wind

related”.

Long term expectations might be the component within the wind data group with highest uncertainty. It is

therefore important to understand how the composition of this part should be established.

In the input forms, there are 3 different input fields related to this topic:

• Long term correction

• Year–to-year variability

• Future climate

3.8.6.2 Long term correction

Here the uncertainty based on the facts used in the long-term correction, typically performed with the MCP

module shall be entered. From 3.2 there are implemented an uncertainty calculation, the “Klintø model” which is

based on studies of many data series, finding which parameters decides the uncertainty. See details in MCP

chapter.

3.8.6.3 Year-to-year variability

The figure entered here is decides how the 1,5,10, 20 year uncertainty is calculated. It tells how much the wind

varies from year to year in the specific region. A typical value is around 6% on wind speed, but several sources

are available at the Internet giving more specific regional variations. In the MCP module, the variability is

calculated based on the long term reference used. The variability entered is used for the 1-year calculated

uncertainty, while the 5 year then is the σ

/sqrt(5) etc. So the 20y variability uncertainty is the σ

/sqrt(20). E.g.,

for σ

= 6%: 6%/sqrt(20) = 1,3% (on wind speed, which converts to AEP% depending on wind speed level). It is

important to be aware of that the variability tells about the fluctuations within few years, not the very long term

variations seen in e.g., Northern Europe described by the NAO index (North Atlantic Oscillations). This is handled

separately in the “future climate” input field.

3.8.6.4 Future climate

E.g., in Northern Europe, we have seen large variations during the 30 year 1980-2009 of modern turbine

operation in Denmark. While 1986-95 (10y) were 8% above long term average measured in AEP, 1996-2006

(11y) were 5% below long term average. This illustrates well that 10 year for sure is too short a period to use as

long term, and that there are climate variations that no one can predict. So far it seems that the variations in

wind climate are not related direct to global warming etc. The slow variations have been seen for 150 years (e.g.,

Loss & Uncertainty 185

by the North Atlantic Oscillation); going up and down, but not trending towards more or less wind. Prediction the

future 20y wind is a hard task that no one really can do. So to assume an uncertainty around 1-3% on wind

speed due to future climatic variations seems appropriate – at least for Northern Europe – other parts in the

world have similar variations, some has not. This should be studied region by region.

3.8.6.5 Reference WTGs

Option to account for calibration of wind based on reference WTGs.

3.8.6.6 Model uncertainty

Vertical extrapolation

Figure 16 Input for vertical extrapolation uncertainty calculation.

The vertical extrapolation uncertainty is divided into the uncertainty due to elevation (above sea level) difference

and difference between mast height and turbine height (above ground level).

The proposals given for the uncertainty is based on different studies, but can be very site dependent. The best

way to get a reasonable basis for the judgment is if there are more masts at the site, the cross prediction accuracy

can give an idea on the uncertainty.

Loss & Uncertainty 186

Figure 17 A large number of calculations in Denmark suggest a linear relationship between uncertainty

and increased elevation in non-complex terrain.

For the DK example above, it is important to emphasize that it is actually the absolute elevation that is shown.

But in the Danish landscape the elevation difference is linked to the absolute elevation as the typical data basis

is based at low elevation. Therefore, the figure indicates an increased uncertainty with increased elevation

difference. Several other studies come up with similar findings. If the terrain is very complex, a RIX correction

might have been performed. In this case the elevation difference uncertainty will be lowered.

The recommendations written in the input fields are intended to give an idea for input of uncertainty, but as

terrain types vary very much from site to site, also the uncertainties vary similarly much. The best way is always

to have more measurement mast at site and use the cross prediction tool in Meteo Analyzer to help give more

precise indications of the uncertainty of the wind model.

8,5

9,5

10,5

0 10 20 30 40 50 60 70 80 90

std dev (%)

Elevation(m)

σ on AEP with elevation, 2200 WTGs in DK

Loss & Uncertainty 187

Horizontal extrapolation

Figure 18 Input for horizontal extrapolation uncertainty calculation.

This calculation is similar to the vertical extrapolation. The critical issue is to judge the distance dependency of

the uncertainty. An upper threshold value will normally be reasonable to use, while the uncertainty does not just

continue to increase with distance. As for the vertical uncertainty, cross prediction based on more masts will be

the best way to establish a basis for the judgments of uncertainty versus distance.

3.8.6.7 Uncertainty of terrain data

Makes it possible to account for partly the complexity of the terrain, partly the quality of the terrain data.

3.8.6.8 Power conversion uncertainty

Power curve uncertainty will often be found in the reports from power curve measurements. But please note that

these will typically give very high uncertainty estimates, which might not be fair. Often power curves are

measured on more turbines of the same type at different locations and the manufactures then perform their best

judgments/averaging of more measurements to reduce their risk. This will reduce the uncertainty. A simple input

can be given as illustrated above like a detailed input requiring input in the WTG Catalogue can be used. It is

our hope to get the uncertainty included for the most power curves in future, but it will probably take some time

before these are available – at least the structures are ready now.

Loss & Uncertainty 188

Power curve uncertainty

Figure 19 Input for power curve uncertainty calculation.

3.8.6.9 Bias uncertainty

For each bias component the user attributes a value, the uncertainty on this value can (and should) be set as

well. Note that the entered uncertainty estimate is multiplied by the bias value, so if e.g., a bias is set to 5% with

an uncertainty of 10% on that value, the resulting uncertainty is 0,5% on AEP resulting from that component.

3.8.6.10 Loss uncertainty

For each loss component included with a value, the uncertainty should also be set by the user. Note that the

estimate is multiplied with the loss value-. A loss of e.g., 5% with an uncertainty of 10%, results in an uncertainty

on AEP of 0,5% due to that loss component.

Results

Figure 20 Evaluation of results.

Loss & Uncertainty 189

On the Results sheet to the lower right presents results for 1, 5, 10 and 20 years of averaging (i.e. life time) and

at several probabilities of exceedance values (50%, 75%, 84%, 90% and 95%).

3.8.7.1 Sum of WTG P-values differ from PARK P-values

If and only if the total loss (in percent) of each WTG and the total park loss are identical the WTG P50s will sum

up to the PARK P50.

However, if some WTGs have significantly different losses (usually due to wake effects) the sum of the WTG

P50s will NOT add up to the PARK P50.

But why is it necessary to have both methods then?

Because we need consistency across P50, P90 etc. – and the WTG P90s will never add up to the PARK P90

as they are the result of a non-linear statistical model. Hence, either the calculation of P50, P90 etc. is done on

a park level OR on WTG level – what is correct depends on if the project is sold/developed as a whole or

individual WTGs.

Loss & Uncertainty 190

Calculation and print

3.8.8.1 Main result

Figure 21 The printout of main result gives an overview of all results as a table and as graphics.

Loss & Uncertainty 191

3.8.8.2 Assumptions and results

Figure 22 Second report page collect all input data on 1-2 pages. Also the detailed result matrix will be

shown at the bottom of this page (not shown here).

Loss & Uncertainty 192

3.8.8.3 WTG results

Figure 23 This page show results turbine by turbine, thereby the needed details for projects sold turbine

by turbine are available.

3.8.8.4 Detailed results

For each “calculation option” a separate report is available. These describe the calculation setup and partly the

data basis. Only one sample is given below. These reports can be quite detailed.

Figure 24 Flicker stop loss is shown with losses for for each turbine can be seen.

Appendix: Wake model tests and validations 193

Appendix: Wake model tests and validations

Test of calculated wake loss on varying wind farm sizes

A good way to compare wake models (verification) is to see how the calculated wake loss changes by wind farm

size.

Here we show a calculation setup, with a square layout with 7 RD spacing based on a large WTG, the Vestas 8

MW V164 with 100m hub height offshore. Then from 3 x 3 up to 31 x 31 rows (961 WTGs) are tested. This will

reveal how the different models compare.

Figure 202. The development of calculated wake loss by wind farm size. NO2005 test.

Using a logarithmic x-axis, the calculated wake loss increases almost linearly with wind farm size

when spacing is kept constant.

For small wind farms (< 20 turbines), the three variants calculate almost identical results. The

deviations increase as project size increases.

By the time the project reaches 100 turbines, the NO2005 calculates around 2% more AEP than

the original N.O. Jensen model. At 250 turbines, this increases to 3%, showing some issues with

this model variant for large wind farms.

Using a linear weight of 35% in the NO2005 combination model, brings the result closer to the

original N.O. Jensen model and AEP deviations are less than +/- 1%. Mirror wake is used in NO2005

in the figure above.

3.9.1.1 Extended test including PARK2, WakeBlaster and Ainslie DAC.

In earlier windPRO versions (< 3.0) the NO2005 was the only wake model for time step calculations. Now multiple

wake models can be used for time step calculations.

Appendix: Wake model tests and validations 194

Figure 203 Park 1&2 and WakeBlaster test by wind farm size.

Increasing wind farm size from 3 to 961 WTGs and calculating wake losses for the same layout, is illustrated

above for the N.O. Jensen variants + WakeBlaster. A very good agreement is seen. The impact of the WDC

choice is seen for PARK2, where the WDC 0.06 (DTU recommendation for offshore) is compared to the lower

0.048 (low TI site). The low TI site shows around 1 percentage point higher calculated wake losses for medium

size and 2-3 percentage points higher wake losses for large wind farm sizes. The original N.O. Jensen (PARK1)

does have a slight “saturation” with very large wind farm sizes. This has been seen as a problem, which did

require some deep array correction for very large wind farms, e.g., the Zafarana wind park in Egypt with 700

WTGs, but this is also seen at e.g., Horns Rev area, where PARK2 handles the wake loss calculation better than

PARK1. PARK2 and WakeBlaster almost fully agree in the test. Here WakeBlaster is run with slightly higher TI

which explains why it calculates slightly lower wake loss than PARK2.

PARK1&2 almost agree for up to 100 WTGs, which is good as PARK1 has been the most used and recognized

wake model in the most recent 30+ years.

Figure 204 Ainslie in Open Wind and windPRO test by wind farm size.

Above, the new windPRO Ainslie 1988 and DAC implementation is compared to Openwind from UL similar

models. This leaves no doubt that the Ainslie as a “stand alone” won’t work, even just for 6 x 6 row wind farm; it

needs a deep array correction model.

Appendix: Wake model tests and validations 195

The windPRO and Openwind implementations agree well, although a little higher wake loss is calculated by

windPRO. However, there are many parameter options both in windPRO and in Openwind, so the differences

are just a matter of default choices. Later validation examples for Horns Rev wind farms and other show that the

higher calculated wake loss by Ainslie 1988 is related to the large WTG size (8MW) used in this example. For

smaller turbines 2-3 MW, there are better agreements also with PARK2.

For the very large wind farms, the recovery zone settings make a difference in the windPRO Ainslie 1988

implementation. In the chart above there are two series of Ainslie with DAC using different recovery zone

settings: “Ainslie DAC-60/80” and “Ainslie DAC-80/120”. It can be observed that the flattening of wake losses as

the wind farm size increases gets delayed when increasing the recovery zone from 60/80 RD to 80/120 RD.

Where and if this flattening of wake loss shall occur, no recommendations are given.

Some of the tests presented in this chapter will illustrate how the deep array settings perform. This manual

chapter concludes that deep array effects are actually more a question of using the correct turbulence. But still,

for tuning wake models for post construction evaluations, the so-called deep array effects still can be relevant.

Especially if using the Ainslie model as a “stand alone” there is a high underestimation of the wake reductions

in the “deep” arrays.

Single row versus multiple row wind farms

A special problem to pay attention to is the single row vs multiple row projects. For single row projects, the wake

models tend to estimate too high wake losses for backrow turbines. The simple explanation is that undisturbed

wind is fed in from the sides along the row, reducing the actual wake loss. In actuality, it is typically seen that the

backrow turbines have similar losses to the second-row turbines, while the wake models instead increase the

wake losses with the number of upwind turbines. An example is illustrated below:

Figure 205 Measured left, calculated reductions by PARK1 (org. N.O. Jensen) right. (by direction).

WTG 6 is the upwind turbine, with just one wake turbine in front (WTG7). X-axis shows direction (degrees). Note

the measurements show almost the same reductions (Power%) for all six wake-affected turbines.

Calculations show essentially larger decreases in centre angles and by a number of upwind turbines.

0,00

0,20

0,40

0,60

0,80

1,00

244 248 252 256 260 264 268 272 276

Reduction, measured

WT1 WT2 WT3

WT4 WT5 WT6

0,20

0,40

0,60

0,80

1,00

238 242 246 250 254 258 262 266 270

PARK1_WDC 0.04 power_2015

WT1 WT2 WT3

WT4 WT5 WT6

Appendix: Wake model tests and validations 196

Figure 206 Calculated reductions by PARK2 left and by NO2005 with DA, right.

With PARK2 it looks better, but the only real way to solve the single row calculation problem in this example is

to increase WDC by number of upwind turbines using the deep array settings in NO2005.

In the calibrated version above to the right, the WDC change for upwind turbines are set as follows:

This result in the following:

Figure 207 Factor on WDC and thereby increased WDC by number of upwind turbines.

As seen, the WDC with “base” set to 0.04 will convert to 0.03 for the first wake turbine, gradually increasing to

0.5, which is the upper limit after four upwind turbines. This work for single row projects (offshore for the

mentioned values), where numerous other single row projects have been tested.

Horns Rev area, Danish Offshore project

The Horns Rev area at the Danish West coast is a good test case as there are three large offshore projects with

many operational years for the first two areas and with large 8 MW turbines for the last project, HR3. This makes

it possible to test for long-term operation, wind farm area interaction and large turbines.

0,00

0,20

0,40

0,60

0,80

1,00

238 242 246 250 254 258 262 266 270

PARK2_WDC 0.06 power_2015

WT1 WT2 WT3

WT4 WT5 WT6

0,00

0,20

0,40

0,60

0,80

1,00

238 242 246 250 254 258 262 266 270

NO2005_WDC 0.04 DA power_2015

WT1 WT2 WT3

WT4 WT5 WT6

0,03

0,04

0,05

0,05 0,05 0,05

0,01

0,02

0,03

0,04

0,05

0,06

0,20

0,40

0,60

0,80

1,00

1,20

1,40

1,60

0 1 2 3 4 5 6 7

WDC

Factor on WDC

Upwind turbines

Single row wind farm expirience

Factor on WDC

WDC: 0,04

Appendix: Wake model tests and validations 197

3.9.3.1 Verification for Horns Rev area, wind statistic based, new Ainslie DAC focus

For the statistical based calculations, the focus is on comparing models (verification), not validation.

Figure 208 Map of Horns Rev area, HR1 south, HR2 mid and HR3 north.

At Horns Rev there are three wind farms making this area suitable for testing single wind farm calculations with

different turbine sizes and different layout configurations.

The Turbulence Intensity (TI) in this area is around 7.5%, which is relatively high for an offshore site. The TI at

the site is found from multiple measurement masts in the area combined with mesoscale data and is valid for

the wind speed interval of 5-15 m/s. This is the wind speed interval where wakes dominate.

Several validations show that PARK2 handles this site very well with the DTU default Wake Decay Constant,

WDC = 0.06. This corresponds to the EMD recommendations to use WDC = 0.8 x TI for offshore sites. (0.8 x

7.5% = 0.06)

For the Ainslie model, where TI is the main input parameter, this is set to 7.5% in the following calculations. The

Deep Array Correction model (DAC) has more parameters as shown in Section 3.7.1.2 The Ainslie/DAC

implementation. The added roughness for offshore areas is by default 0.02, used in the tests along with 0.01, to

illustrate the sensitivity of the parameter.

The below figures show calculated wake losses with long term (20y) mesoscale model data for the area using

the EMD-WRF Europe+ dataset. This is found to reproduce the wind speeds seen very precisely for this area.

For comparison, the Openwind Eddy Viscosity model (Ainslie) and PARK2 model are calculated. While we do

not believe the formula revisions for PARK2 compared to PARK1 has been made in Openwind, the N.O. Jensen

model can run with linear combination model and excluding mirror wakes, which is close to PARK2. And then

the Ainslie with DAC is run with default settings for offshore. For full disclaimer, there are several tuning

parameters in Openwind that have not been tested here.

Appendix: Wake model tests and validations 198

Calculating HR1 and HR2 as “stand alone” and combined.

Appendix: Wake model tests and validations 199

Figure 209 Plots showing calculated long-term wake losses for different combinations.

The new windPRO Ainslie 1988 with DAC default settings, returns almost exactly the same wake loss as

windPRO PARK2 (N.O. Jensen) for HR1 alone, HR2 alone and when both windfarms are calculated together.

It can be seen how sensitive the DAC model is to the roughness increase by testing 0.01 as alternative to the

default 0.02. The result shows a decrease in calculated wake loss of just under 1 percentage point.

The Openwind calculations show some higher calculated wake loss with their PARK2 model and a little lower

with Ainslie-DAC.

Which one of the windPRO Ainslie-DAC or PARK2 models performs best is not possible to say based on the

real operational data, as the difference is too small to detect. However, based on detailed tests with 10-min

operational data, we know that the size order of the calculation results for windPRO PARK2 are very accurate.

Figure 210 Impact of neighbour wind farm in calculations.

Looking at the impact by inclusion of another wind farm 15 km away, the impact is relatively small. The biggest

impact shall be expected on HR1 from HR2 due to the wind direction distribution. This is also demonstrated with

PARK2, but only by 0.2 percentage points. This is as expected, because when looking at the data for HR1 before

and after commissioning of HR2 there is practically no difference to see. Ainslie DAC has a higher increased

wake loss, although also small.

Figure 211 Turbine by turbine calculated wake loss.

Appendix: Wake model tests and validations 200

In Figure 211 there are three calculation variants with calculated wake losses WTG by WTG. The first 80 WTGs

are HR1, the next 91 are HR2. There are small differences, but quite small and very similar patterns with PARK2

and Ainslie DAC with defaults.

Conclusion for HR1&2: The windPRO Ainslie-DAC (1988) performs almost as PARK2 with default settings

– and PARK2 has been validated in several studies to perform very accurately for these wind farms.

Calculating HR2 and HR3 as “stand alone” and combined.

In this example, there is an added challenge compared to HR 1&2, with HR3 having significantly larger turbines,

RD 164 against HR2’s 93 m. And the two wind farms are located much closer together.

Figure 212 Calculated wake losses for HR3 with 8 MW WTGs.

Above, when calculating HR3 with much larger WTGs, there are much larger deviations. Ainslie-DAC with default

settings calculates ~2 percentage points more wake loss than PARK2. Even with a lower increase of roughness,

it calculates 1.5 percentage points more wake loss than PARK2. In this case, Openwind PARK2 and windPRO

PARK2 are in better agreement and the two different models get more similar results than with the smaller wind

turbines.

Figure 213 details on wake loss calculation for HR3.

Figure 213 shows how deviations are seen for all turbines and directions. Here, it is hard to tell what is correct

as HR3 has not been operating without HR2 impact. Combining with HR2 we might get an idea.

Appendix: Wake model tests and validations 201

Figure 214 HR2 & 3 wake loss calculation with different combinations.

Here is seen how much impact the different models calculate by including the neighbor wind farm.

Especially interesting is the increase in calculated wake loss for HR2 by the presence of HR3 as this might be

detected in the operation data. Albeit, difficult due to market regulation and operational issues.

All model variants agree well on the impact of including the neighbor wind farm.

Conclusion for HR2&3: There are differences in PARK2 and Ainslie/DAC in calculated wake loss when the

turbine size gets larger, here 8 MW, where Ainslie/DAC calculates higher wake losses. This might lead to revised

default recommendations for either one or both models when the turbines are larger, but it is too difficult to say

based on a single example. All models agree on the size order of increased wakes from the neighbor wind farm.

3.9.3.2 Tests for Horns Rev area, time step based.

First, a detailed validation of both the mesoscale data and the wake loss calculation setup is made based on

HR1. This detailed validation is possible due to access to 10-min data. Next all models are tested on all three

wind farms in the HR area based on monthly measured production data.

For the monthly data it is worth keeping in mind where these are measured. In Denmark, and assumed similar

in the UK, the monthly data is measured at the low voltage side at the substation. This means that the

measurements are reduced with 1.5 – 2% electrical losses (step up transformer and internal grid each account

for typical 0.7-1%). In addition, internal consumption during stops can add 0.1-0.2% by reasonable normal

operation. For sites with heavy market regulation shut down, this can be significantly higher.

Using the 10-min SCADAdata, this is typically measured after the step-up transformer, which includes approx.

1% loss. Thus, applying a ratio of 0.99 measured/calculated for all normal 10-min SCADAdata samples yields

the perfect result from the wake model

For monthly data the ratio shall be 0.98 if there are no other losses than electrical. In other words, if less than

2% loss in a month, the calculation model is biased, which could be the wake calculation or a wind data bias.

Horns Rev 1 detail validation of wake model

In a pre-analysis the ratio between back and front row power is found by TI bin, which then is paired with similar

ratios for calculations with different WDC values. This gives the following relation between WDC and TI for

PARK2:

Appendix: Wake model tests and validations 202

Figure 215 Advanced modelling of WDC by TI based on detailed Power by TI analyses.

The good thing is that a similar analysis on another UK offshore wind farm with access to detailed 10-min data

gives a similar result. Therefore, the advanced option for WDC(TI) is added as an alternative to the general

recommendation WDC = 0.8 x TI for offshore.

Here, it is seen how the calculation based on mesoscale data and a WDC = 2 x TI – 0.07, handles the wake loss

calculation to almost perfection, where a filter criterion is that minimum 79 of the 80 turbines must be running.

This leaves 27.000 10-minute approved samples, which is 26% of available samples for the two years 2008 and

2012 with data available.

Figure 216 HR1 calculation for 2008 & 2012 compared to measurements.

Figure 217 HR1 measured and calculated at TI>6%.

Figure 218 HR1 measured and calculated at TI<6%.

Appendix: Wake model tests and validations 203

It is worth saying, that the ratio measured/calculated turbine by turbine is good for both higher and lower TI:

Figure 219 Measured/calculated for lower and higher TI for HR1.

There are no signs of “curtains” or east-west bias meaning the wake model handles the site to near perfection.

The bias for low TI for all WTGs is caused by mesoscale data having too high wind speeds at low TI. This could

be a blockage issue.

But it is good to see how the much larger variation in production between the middle and end row wtg’s at low

TI is captured well by the wake model with the advanced settings.

Figure 220 Long term calculation of HR1 compared to measured.

With the advanced wake model settings, close to 20-year operation can be calculated and compared to monthly

measurements. A very fine agreement is seen, but also that there are some months with quite high losses. From

2018 we know that one turbine has been taken out of operation permanently due to lightning damage.

Figure 221 The calculated wake loss vs loss on top of wake loss and binned loss.

Appendix: Wake model tests and validations 204

The left-side graph above shows no correlation between loss after wake loss and calculated wake loss, which is

good. The red square in the right-side graph shows where there should not be any data, apart from mesoscale

bias related. The loss/mesoscale bias is nicely normally distributed as should be expected with this very long

dataset.

Apart from "extraordinary"

losses mean values are:

Average if <

10%

if < 15%

All

Extraordinary

HR1

6,6%

8,3%

17,2%

8,9%

The table above shows “normal” operational losses in the size order of 7-8%, of which ~1% is grid loss and the

remaining is availability, sub-optimal performance AND market regulation. The extraordinary loss by major

downtime and possible market regulation is ~9% seen over 20 years.

Horns Rev 2&3 detail validation of wake model

The wind data used is the EMD-WRF Europe+ dataset. While there can be a bias, which can violate the wake

model validation, it has the great advantage that it can be compared to actual production from the Danish

Stamdataregister. This is done month by month, which results in a much better validation basis than for statistical

calculations in spite of missing 10-minute data as we had for HR1.

Figure 222 Comparing calculated wake losses with all losses including wake loss, HR2 left, HR3 right.

First, a look at the production data compared to the calculation without a wake model. Here the calculated wake

losses by PARK2 (DTU default WDC 0.06) is plotted against the losses by the wind model calculation, without

wake loss calculation. Left is HR2, right is HR3. It is seen on the trend how a significant part of the loss is related

to wake loss. This is very clear for HR2, less for HR3 due to short operation period, including start up period.

It is to expected that there are no data above the blue line, as the calculated wake losses then would be too high

(or the wind data biased). A few points above the line are ok due to the lack of precision of the mesoscale data

and the fact that there are no stability or TI correction month by month.

It is seen that the calculated wake losses by PARK2 vary by month from 5-18% for HR2 while they vary from 2-

9% for HR3. But the total losses based on the wind model varies much more, and this is a problem for the

validation, that we cannot separate wake and other losses precisely.

Below the calculation with wake loss is compared to the measurements:

Appendix: Wake model tests and validations 205

Figure 223 Monthly calculated and measured production for HR2.

Viewing the monthly figures gives a good idea if the wind model seems reasonable. And it does. The grey dots

show the loss after wake loss and the yellow the calculated wake loss.

Figure 224 Calculated wake losses and seen "other loss" for HR2.

Above to the left is seen for HR2 that there is no systematic trend that losses are higher/lower where the

calculated wake losses are high/low. This is a good indication that the wake loss calculation does not seem

biased.

To the right the loss distribution after wake loss reduction in calculations. This looks ok, although we do not know

the monthly operational losses. Looking at averages, where extremes are taken out:

Apart from "extraordinary"

losses mean values are:

Average if < 10%

if < 15%

All

HR2

6,4%

8,1%

11,4%

HR3

5,4%

7,2%

18,9%

Probably acceptable loss figures for the “normal operation”, where some grid losses must be assumed for the

internal cabling (substation and sea-land cable losses are not subtracted in production figures. Measurements

are at the substation on the WTG side). But the availability and sub-optimal WTG operation losses are dominant

together with possible market regulation.

It thereby seems that PARK2 (WDC 0.06) and the wind data basis handles the calculation well.

Then we have a reference for the Ainslie/DAC test.

Appendix: Wake model tests and validations 206

Figure 225 Calculated and measured + losses by Ainslie DAC calculation.

The Ainslie DAC-based calculation performs like PARK2 for HR2.

A similar loss table as for PARK2 justifies that the calculation works well, although the HR3 losses seem too low

(too high calculated wake loss for this farm).

Apart from "extraordinary"

losses mean values are:

Average if < 10%

if < 15%

HR2

5,4%

7,5%

HR3

3,2%

4,6%

Appendix: Wake model tests and validations 207

Figure 226 Timeseries calculated and measured by month for HR3.

Comparing Ainslie/DAC with PARK2 – both with default settings. That the losses get negative with Ainslie DAC

but not with PARK2 indicates that Ainslie/DAC calculates too high a wake loss for this wind farm.

HR2 results

Calculated wake loss

HR2-3_PARK2_WDC 0.06

10,4%

HR2-3_Ainslie_

11,1%

HR2-3_PARK2_WDC = 2 x TI - 0.07

10,2%

The main results of three calculation variants for HR2 as time step calculations for the full HR2 operation period

are shown above. Note the PARK2 with WDC(TI) is based on the EMD-WRF Europe+ TI, which is scaled with

factor 1,41. The advantage of making the wake loss TI depending per time step is that the wake losses are

calculated more precisely. This way sites with higher or lower TI will be handled more precisely.

Figure 227 Monthly loss distribution by loss bin for HR2 and different calculation variants.

Showing months by loss bin for eleven operational years. We cannot know for sure how much is related to

mesoscale wind bias, but we know from HR1 validations with 10-min operational data for two years that the

losses on top of wake losses are in a realistic size order and the wake model calculation is in the right size order.

Lillgrund, Sweden offshore project

This project is special due to the dense spacing - around 3.2x the rotor diameter (RD). The main wind direction

is from WSW, along the row orientation.

Appendix: Wake model tests and validations 208

Figure 228 Measured and calculated from Performance Check, month data (includes “other” losses).

Shown above is the output from the PERFORMANCE CHECK module, where both measured and calculated

turbine by turbine can be seen. Data is filtered by taking out larger down times in both measured and calculated

values. Monthly production data for each turbine for five years (from December 2008) is used. The calculation

is based on mesoscale data. In general, a very good match, using PARK2 WDC = 2 x TI -0.05. Thereby

subtraction of 0.07 is replaced with 0.05, which is a fine-tuning handle, that in this case change the calculated

wake loss from 28% to 26%. The tuned version calculates all WTGs within +/- 2% of measured with respect to

differences in wake loss. Other loss is assumed the same for all WTGs after outlier filtering.

Figure 229 The calibration tool: goodness vs calculated, PARK2 adv. default (left) and tuned (right).

Showing the goodness (measured/calculated for concurrent outlier filtered data) is a good way of finetuning the

wake loss model. If there is a down or up trend, the wake model is incorrectly calibrated as seen to the left,

Appendix: Wake model tests and validations 209

where the highest calculated (free in main wind direction) has the lowest goodness. Thereby too high wake

losses are calculated. Here mesoscale data are scaled 0.98 which probably are a reasonable size order giving

round 6% avg. loss on top of wake loss for the outlier filtered data.

Figure 230 Wind farm layout and ratios measured/calculated P2 tuned, Lillgrund offshore.

The calculated wake losses are for three variants:

PARK2, WDC = 2 x TI – 0.07 (advanced default): 28%

PARK2, WDC = 2 x TI – 0.05 (best WTG by WTG reproduction): 26%

Ainslie default settings: 27%

Thereby both PARK2 and Ainslie by default settings are close to the fine-tuned most trustworthy result for this

very high wake loss site.

Figure 231 Ainslie with WTG by WTG goodness slightly poorer than PARK2.

Appendix: Wake model tests and validations 210

Wake calculation validation for large Egypt wind farm

This project is checked each half year for five years, where wake modeling is tested. So here we know very

precisely what the wake losses are. Below a comparison for a specific period, 1½ year, with good mast

measurements in front of park including Turbulence measurements.

The in-row distance is ~3 RD, between rows ~14 RD. The turbine model is the Gamesa G80 with 60m hub

height.

Figure 232 Layout of the Egypt wind farm.

The special thing for this site is the uniform wind direction, NW, but with the specialty that in WNW the TI is very

low, around 5%, while in NNW it is more “normal” for onshore, close to 9%:

Figure 233 TI by direction sector.

Therefore, the Ainslie model is tested particularly for sectors 29 and 34, to compare to PARK2 wake modeling

at low-high TI conditions. The results:

0,02

0,04

0,06

0,08

0,1

28 29 30 31 32 33 34 35

TI by 10 degree sector

Appendix: Wake model tests and validations 211

Figure 234 Egypt large wind farm wake loss calculations.

Ainslie is used with onshore background roughness 0.03, but added roughness as for offshore, 0.02. With these

settings Ainslie compares well to PARK2. Both calculate 9-10% wake loss, which is the correct value based on

5+ years of measurements. Ainslie does calculate a lower wake loss in the low TI sector 29 compared to PARK2

and the reverse in high TI sector 34. This is as expected from the offshore tests, where it was seen that Ainslie

model does react less to changes in TI compared to PARK2.

There is no complete agreement WTG by WTG, but all in all good agreement. In addition, the test of new offshore

recommendations for PARK2 is included: WDC = 2 x TI – 0.07. This is seen to work very similarly to the PARK2

calculation with WDC by direction based on WDC = 0.8 x TI.

Setting the added roughness for Ainslie to 0.1 (onshore recommendations) calculates some higher wake

losses, although just a few percentage points. This illustrates how sensitive the Ainslie/DAC is to added

roughness for a wind farm covering a large area. The site although must be considered “offshore -like” where

also the PARK2 model needs offshore settings to behave well. The combination of low TI, low roughness and

high wind speeds makes it offshore similar, and this probably explains the need of handling this site with offshore

settings.

Large UK offshore wind farm complex

How does the wake modeling work for very large wind farm complex, not just a single or two wind farms?

Appendix: Wake model tests and validations 212

This we will try to answer by the Irish sea wind farm complex with Walney etc. see below.

Figure 235 Large wind farm complex covering 45 km east-west.

Monthly data from each wind farm are available. The Walney extension consists of two types of turbines,

Siemens Gamesa 7 MW (green) and Vestas 8MW (light blue).

Period:

From

Wind farms

Months

Years

WTGs

acc.

WTG type

Jan-08

Jan-

Barrow

3.1

V90 3MW

Jun-12

Jan-

+W1, W2,

Ormonde

1.7

162

SWT 3.6 (107+120) Repower 5M

Oct-14

Nov-

+West of

Duddon Sands

3.2

270

SWT 3.6 120

Aug-

Dec-

+Walney

extension

2.4

357

V164 8.25 + SWT 154 7MW

An interesting feature of this wind farm area (and in windPRO) is that data can be grouped in 4 periods with

different wind farms in operation. Thereby it can be checked if the wake models capture the increase in losses

due to new neighbours.

A calculation model is setup based on multiple EMD-WRF Europe+ mesoscale data points to include the

horizontal variation of the wind climate in the region. Calculation is by hour, only including operating turbines in

relevant months. The months with partly operating wind farms are not used.

While we do not have information on losses apart from the difference between calculated and measured, we

evaluate the performance of the wake modelling based on the loss difference from period to period.

Appendix: Wake model tests and validations 213

Four periods, where different number of wind farms are operating, are analyzed. Note that there are months with

extremely high losses, where some wind farms have been out of operation most of a month, which influences

the data too much. Therefore, we take out months with more than 15% losses to make conclusions based on

“normal operation”:

Figure 236 Losses on top of wake losses for months with < 15% loss (74% of month data).

The “trade off” between

wake losses and operational

losses / mesoscale bias is

the difficult part. What is

what? This is of importance

to separate to be able to

make better pre-construction

calculations.

Availability information do

not give the full answer.

Many analyses have shown

there are more operational

losses than the availability

information shows.

Losses on top of wake

loss apart from extremes

Period:

Appendix: Wake model tests and validations 214

Losses on top of wake losses for months with < 15% loss (74%) for the four periods with a different number of

wind farms in operation.

The markers show:

Red – the losses increase with new wind farm that must be expected to have an impact.

Green – the losses decrease with new wind farm that must be expected to have an impact.

Yellow – change in loss is not expected related to new wind farms but change in operation.

There are examples where the wake model does not seem to capture the effects of new wind farms well enough.

Conversely, there are also examples where the wake model calculates too much loss increase. This leaves no

clear conclusion.

As an alternative approach, advanced wake tuning is applied:

Figure 237 Calculation with advanced WDC(TI) for offshore.

Almost same picture, 71% of 535 months data used. But some higher losses than previous calculation with fixed

WDC = 0.05.

The red-green differences from the previous chart are overlayed. This shows how Walney 2 previously had

decreasing losses, but now has increased losses. The other examples also show slightly smaller loss increases

as neighbor farms are built, compared to the losses using WDC=0.05.

By further adjusting the offset from -0.07 to -0.09 returns more months fulfilling the 15% threshold, 77%.

Losses on top of

wake loss apart

from extremes

Losses on top of

wake loss apart

from extremes

Period:

Appendix: Wake model tests and validations 215

Compared to using 2xTI-0.07, the losses are now around 2 percentage points smaller, due to higher calculated

wake losses. And this also reduces the change of losses as neighbor farms are built. Some of the change partly

disappears, some partly get reduced.

Very importantly, the losses are now comparable to the losses observed at the detailed 10-min calibrated sites,

which is around 4-7% for most wind farms and periods. The reason why West of DS (Duddon Sands) has higher

losses can be related to power curve or Ct curve issues. Similarly, the bias in the other direction for Ormonde

can have multiple causes (mesoscale wind, Ct curve, power curve), but the increase in losses does not seem

wake related, but related to operational problems:

Figure 238 Ormonde measured and calculated by month.

Above, observe the “heavy” drops, first in Oct. 2016 then several months in 2017-18 and again in 2019 and 20.

This does not seem wake loss related but caused by operational issues. This explains the loss increase by

period, which could be even higher if the model setup for this windfarm was tuned further.

Figure 239 West of Duddon Sands measured and calculated by month.

West of Duddons Sands is performing in a quite stable manner month by month, but with quite high losses every

month. This is unlikely a wake modeling issue, but more likely a turbine data issue. The calculated production is

a little too high every month probably due to mesoscale bias or power curve bias.

Appendix: Wake model tests and validations 216

Figure 240 Assumed all time losses in addition to wake losses for the 6 wind farms.

As mentioned, Ormonde and West of Duddon Sands probably skew 1-2% in each direction, maybe simply due

to using too few mesoscale wind data points – only 3 points were used in the calculations. While 3-5% normal

operation loss seems reasonable, the serious problem is the months with > 15% loss, which might perhaps

include market regulation?

Figure 241 Final calculated wake losses per period per wind farm.

With the “most plausible” calculation setup based on losses on top of wake losses, these are the calculated wake

losses for the four periods for the six wind farms. The reason why Barrow and Ormonde have lower calculated

wake losses in period 4 than 3 is due to the wind speed and direction distribution. Walney 1&2 have higher

calculated wake losses in period 4 due to the impact from the Walney extension.

Finally, a comparison to Ainslie DAC handling of this large area:

Appendix: Wake model tests and validations 217

Figure 242 Calculation for 20y all wind farms running full time, compare PARK2 and Ainslie.

As seen, there are differences between PARK2 and Ainslie DAC. In general, these are small, typically below 1

percentage point. But most importantly: the two wake models seem to handle even these large wind farm

complexes very well. There is no indication of serious problems.

Very large Egypt wind farm complex

With 700 WTGs, Zafarana is probably the largest onshore wind farm complex measured in number of turbines.

Figure 243 Zafarana wind farm.

Appendix: Wake model tests and validations 218

Wind is always from northern directions. TI is round 9%. Using PARK 2 with 0.8 x TI gives a good match

calculated vs measured:

Figure 244 Row by row calculated and measured production with ratio meas/calc.

In 2011 EMD calibrated the wake model for this site. The result was a need of increased roughness to

compensate partly for:

1) The original N.O.Jensen model (PARK1) did not reduce calculation result “enough” for the down wind

part of the site.

2) There were meso scale effects which the traditional WAsP setup did not include.

In 2022 using PARK2 and mesoscale model data, this site is calculated quite accurately without any needs for

added roughness to compensate for model and data inaccuracies. This shows how continuous model and data

improvements within windPRO have brought the calculation model accuracy very far within the past 10 years.

Figure 245 2011 calculations, increased roughness as model compensation very deciding.

With a decent model calculation setup in 2011, the predicted wake loss would be calculated to 13,5%, where

real operation data showed 20%. The 2011 “solution” for calibration of the model setup was to increase the

roughness. In 2022 the combined mesoscale model data and PARK2 are capable of handling this site just as

well without roughness calibration.

Conclusions on wake modeling

WindPRO offers three alternative wake model concepts:

1. N.O. Jensen (PARK1, PARK2, NO2005). EMD recommends PARK2

2. Ainslie with DAC (Deep array correction)

Appendix: Wake model tests and validations 219

3. WakeBlaster (external model from ProPlanEn)

For the two first wake models, blockage can be calculated as part of the wake calculation. Although typically this

will only reduce production around 0.5% for 100 WTGs and is not TI dependent. Therefore, there is no significant

change in calculated losses by present blockage implementations based on state-of-the-art scientific methods.

All three model variants calculate wake losses with reasonable accuracy and similar magnitudes. This has been

validated based on several wind farms with access to detailed 10-minute operation data per turbine and verified

by comparisons of the different wake models on real wind farms and on simulated wind farms with tests of

calculated wake losses vs. the size of wind farm.

The key input for concepts of Ainslie with DAC and WakeBlaster is the Turbulence Intensity (TI), which

determines the level of calculated wake loss. For N.O. Jensen it is the Wake Decay Constant (WDC), which

EMD recommends being adjusted based on TI.

Detailed data analyses show a trend towards wake losses being even more TI-dependent than the models

calculate. This can be handled by N.O. Jensen by making the WDC(TI) more “aggressive” than so far

recommended, illustrated below:

Figure 246 WDC(TI) for different configurations with PARK2.

The updated recommendation for PARK2 onshore is to use WDC = 0.6 x TI. For offshore and low TI onshore

sites EMD recommends using a WDC = 0.8 x TI. However, we do see in data/validation examples that the more

aggressive relation WDC=2x TI -0.07 for offshore/low TI site works better when subdividing data in TI bins.

The subtracted 0.07 can be used as tuning parameter, where for some sites like Lillgrund offshore with dense

spacing 0.05 works best and for very large wind farm complexes like the Irish Sea Complex 0.09 works best.

Most likely it is a question of wind farm size. For windfarms with around 100 WTGs the 0.07 seems to work best.

The reason for more or less reduction can be that bigger wind farms has a greater impact on the wind regime of

the site, which then can be compensated by a lower WDC. A lower limit on WDC of 0.01 and higher limit of 0.2

is recommended.

An important piece of information is that for offshore, the EMD-WRF Europe+ and similar pre-run mesoscale

datasets have too low a value of TI offshore. EMD recommends scaling this TI with √2 (=1.41). This experience-

based adjuster is built-in for the discontinued EMD ConWx mesoscale dataset and the EMD-WRF On Demand

data. But while onshore validations show better agreement on TI onshore without this adjuster, it is not included

in EMD-WRF Europe+ and similar datasets.

Calculating for the new generation of offshore turbines from 7 MW and up, there seem to be a slight

overestimation of the wake loss with Ainslie DAC model. This is based on a few tests so far.

Validation examples and model problem issues 220

Validation examples and model problem issues

MCP validation

The validation is based on Høvsøre 40m data. Here there are 14 years of measurements, so the long-term wind

speed is well-known. Data for one year at a time are taken as local data and EMD ConWx mesoscale data are

used as reference. Here the years 2015 (high wind), 2016 (low wind) and 2017 (normal wind) are used to see if

the long-term expectations are as the long term measured based on these different local years.

Figure 247 More MCP sessions can show more combinations for comparisons

The session overview page with the examples is shown above.

Below is a table with the different LT predicted mean wind speeds.

Two variants of the 2015 and 2017 based predictions are shown (by cloning the one and changing the selected

model).

Below results of the experiments shown in table:

Table 14 Testing MCP with Høvsøre data

LT truth

7,93

Input

Method

Predicted

LT m/s

Deviation

Deviation%

Uncertainty

2015

Regression

7,96

0,03

0,4%

6,3%

2016

Regression

7,85

-0,08

-1,0%

3,5%

2017

Regression

7,87

-0,06

-0,8%

N/A

2017

Simple scaling

7,93

0,0%

4,30%

2015

Scale local

7,93

0,0%

6,30%

As seen the largest deviation to the true long-term wind speed (7.93m/s over 14 y) is 0.08 m/s (1,0%). The AEP

uncertainty will depend on the turbine technology and wind speed level. For this wind regime, the AEP

uncertainty will be around a factor of 2 on any wind speed uncertainty The largest error is therefore well inside

the estimated uncertainty. It can thereby be concluded that the tool seems to do a good job and that the EMD

ConWx data as reference work almost perfect at this site.

Validation examples and model problem issues 221

For the first entry labelled “2017 Regression”, one year of data is used, therefore the uncertainty is not presented

(as windPRO only calculates uncertainty when there is more than twelve months of overlapping data). Expanding

2017 slightly with some data from 2016, gives the uncertainty shown by “Simple scaling” method.

Mesoscale data long term consistency

Figure 248 Long-term consistency using mesoscale data

This screen shot shows how well the mesoscale data based calculation matches measured production on a time

scale of 18 years. The Tunø Knob offshore wind farm has been operating very well with few problems and a

high availability during all 18 years. Therefore, it is a good validation case. It is seen that year-by-year model

errors are within +/-5% and month by month within +/- 10%, apart from few months where the problems probably

were related to availability issues.

This example alone cannot validate the long-term consistency.

Another test case is Høvsøre wind measurement mast versus MERRA-2 and EMD ConWx:

Validation examples and model problem issues 222

Figure 249 Wind speed ratios Model data/measurements 2004-18

It can be seen as well that the EMD ConWx mesoscale dataset and the MERRA-2 data has a trend relative to

measurements.

It is in size order 1-2% per 10 years as an average for this site. What could be the explanations?

• Roughness increase, more trees, buildings and turbines affect the measurements, but not the model

data.

• Measurement equipment changes in time, although it is assumed this is recalibrated regularly (?)

• Measurements are never “perfect”.

What is certaintly seen is different behaviour in low and high wind years. See graph below:

Figure 250 MERRA-2/measurements for Høvsøre 100m (MERRA-2 scaled to Høvsøre average)

y = 0,0023x + 0,9925

y = 0,0012x + 1,0057

0,94

0,96

0,98

1,02

1,04

1,06

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018

Model/measurement ratios (100m)

EMDConW/Høvs 100 Merra2/Høvs 100

Linear (EMDConW/Høvs 100) Linear (Merra2/Høvs 100)

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

0,975

0,98

0,985

0,99

0,995

1,005

1,01

1,015

1,02

1,025

8,6 8,8 9 9,2 9,4 9,6 9,8 10

Ratio model/measurement

Annual wind speed (m/s)

Merra2/Høvs 100

Validation examples and model problem issues 223

There is a clear trend, that in low wind years, the model data are too high and in high wind years they are too

low. In other words, model data varies less by high/low wind years than do the measurements. This problem can

be handled by post processing the model data, something EMD is working on finding good methods to solve.

Later tests have revealed that the seen trend is higher when looking ONLY at eastern directions, where

roughness increases in time, and there is basically no trend with wind from west, where the roughness

remains constant due to the met mast is close to the sea in west direction. Therefore the observed bias

is identified mainly to be increased roughness in time, and is not reflected in the model data.

WAsP versions modifications

The WAsP model itself will not be explained in detail here (see Risø/DTU WAsP manual), only the changes of

high importance for the user in the more recent versions. From the very first versions, until and including ver. 9,

the model itself has only changed marginally, and the calculation results, thereby, also. The major improvement

during the earlier WAsP model changes is the capability to handle more map file points. However, from version

10.0, model modifications have been made. These mainly relate to stability correction handling, especially for

offshore and near shore, but, also, the roughness map interpretation has been improved. The corrections were

partly included in ver. 10.0, but first fully implemented in ver. 10.1 and 10.2. We, therefore, do not recommend

using ver. 10.0 in offshore or coastal regions. The corrections relate to the default heat flux parameters - the way

roughness in a coastal zone is interpreted and to formula modifications. The result of the corrections is a

smoother change between on shore and offshore stability correction. The wind statistics for offshore is different

if it is generated from WAsP 9 or WAsP 10.2+.

This means that an offshore or near shore wind statistics made from WAsP 9 SHOULD NOT be used in

any version from WAsP 10.2 or vice versa.

This can best be illustrated by an example: 4 wind statistics are generated from the same time series data:

Top row is using WAsP 9 and bottom row is using WAsP 10.2. Left column is assuming a class 0 site and the

right column is assuming a class 1 site

In the left column, the class 0 data is almost the same using the two WAsP versions. The highest level is at

200m, although the result is slightly lower with WAsP 10.2 – a part of the formula modification. BUT, for the

onshore classes, the wind speeds are essentially higher with WAsP 10.2 - around 0.2 m/s.

This means that a WAsP 10.2 calculation could calculate up to around 10% higher AEP than WAsP 9 at

an onshore site, if the data basis were offshore.

In the right column, the onshore site, the onshore class data 1, 2 and 3 are almost identical from the two versions.

Again 200m has slightly lower results with WAsP 10.2 – a part of the formula modification. BUT the class 0 data

is around 0.3 m/s lower when calculated with WAsP 10.2.

Figure 251 Four windstatistic results for Lillgrund offshore, different WAsP’s.

Validation examples and model problem issues 224

This means that a WAsP 10.2 calculation would calculate up to around 10% lower AEP at an offshore

site, if the data basis were onshore.

The changes are, of course, not serious as long as a consistent approach is used in the sourcing of the raw data

(e.g., use offshore data for an offshore site). But, it should be mentioned that, in the coastal region, where there

is part water and part land, quite unpredictable changes can be seen. Here, it will be especially important to use

the same WAsP version for generation of wind statistics.

From the tables, it is also seen that, with large hub heights (>100m), WAsP 10.2 calculates slightly lower wind

speeds than WAsP 9 – this is so both on and offshore have a general modification of the stability correction

model. Based on quality test data in an offshore environment and from tall masts, this correction appears to give

a better reproduction of the measurements.

Figure 252 Test case calculations showing the WAsP stability model shift.

The example above illustrates the improvement regarding smoothing the stability model shift. Two turbines just

200m apart, both around 4 km offshore, are calculated. In Sector 5, there is a roughness change from class 0 to

class 0.4 at a 10 km distance. The graphs show the ratio of the calculated AEP between WAsP 9/WAsP 10.2.

Looking at the rose, it is seen that, in Sector 5, what happens for the two turbines is very different. This is where

the distance to the roughness change in Sector 5 is just around 10 km. A change from 0 to a higher class decides

that WASP 9 shall change between on- and offshore stability in that direction. The parameter: “Width of coastal

zone” decides this and can be changed, but the default is 10 km. It’s an obvious inconsistence, which is in WAsP

9, but not in WAsP 10.2. But, the graph also shows how the calculation results are smaller in general with WAsP

9 when the wind comes from land, while the WAsP 9 and WAsP 10.2 results are the same when wind comes

from the open sea.

Validation examples and model problem issues 225

Figure 253 Wind profiles, measured and calculated

A test on reproducing measured profiles in large heights is shown above. It is seen that WAsP 10.2 predicts the

vertical profile slightly better than WAsP 9. The 160m points (purple squares) are from another mast nearby and,

therefore, not fully comparable to the measured (red triangles). Based on both masts measured at 100m, the

160m point is scaled (yellow triangle), and it is observed that the WAsP calculated profile matches

measurements well up to 160m, with a small advantage for WAsP 10.2 relative to WAsP 9.

3.10.3.1 Tests of calculation in a coastal region with different hub heights.

The calculation setup can be seen in the figure on the next page:

Figure 254 Model setup for test of WAsP model

100

120

140

160

180

200

4 6 8 10 12

Høvsøre - W sector

Wasp 10.2

Measured

Measured-scaled from

OMS mast

Measured - OMS mast

100

120

140

160

180

200

3 5 7 9 11

Høvsøre - E sector

Wasp 10.2

Measured

Measured-scaled from

OMS mast

Measured - OMS mast

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75

Elevation (m)

Elevation for the 75 test turbine pos.

Validation examples and model problem issues 226

Based on a met mast 1800m from the west coast, a row of turbines are calculated based on the two different

WAsP versions, where the wind statistics are generated with same version as is used in the calculation. The

turbine row starts 13.5 km offshore and ends 23.5 km onshore.

Figure 255 Map details for test setup

The map gives an idea of the surface roughness around the test row. The row crosses a large forest area along

the onshore part.

Figure 256 Results of WAsP 10.2 vs WAsP 9 calculations

For a 15 m hub height there are two variants: one calculated without orography and one with. For all other

heights, orography is included, and for all calculations, digital roughness maps are included.

It is seen that there are quite some differences between the WAsP 10.2 and WAsP 9 results, but it is not that

easy to come to a clear conclusion as to why. Starting with the 150m hub height, WAsP 10.2 results are around

1% lower the first 3 km offshore, but then rise up to 3% higher around 10 km offshore. Onshore, the results are

lower the first 15 km, up to 3%, but get then slightly higher. A “reversed” pattern is seen for very low hub heights,

but with up to 8% higher results onshore. There are two dramatic peaks in the graph. The leftmost peak is due

to differences in roughness interpretation in the two models. The rightmost is due to different orography

interpretation. This just illustrates that besides the major model changes, some minor bug fixes also contribute

to the differences and leads to a recommendation to always use the latest version.

In conclusion, there are obvious improvements, and especially offshore and in coastal regions, we recommend

using WAsP 10.2 due to better reproduction of the measured shear. For fully onshore sites (when measurements

also are fully on shore) there seems only to be differences when hub heights are > 100m. However, if

97%

99%

101%

103%

105%

107%

109%

-13.500

-12.500

-11.500

-10.500

-9.500

-8.500

-7.500

-6.500

-5.500

-4.500

-3.500

-2.500

-1.500

-500

500

1.500

2.500

3.500

4.500

5.500

6.500

7.500

8.500

9.500

10.500

11.500

12.500

13.500

14.500

15.500

16.500

17.500

18.500

19.500

20.500

21.500

22.500

23.500

Distance to shoreline (negative is offshore)

WAsP 10.2/WAsP 9 based on near shore windstatistics

15 hub height, FLAT terrain

15m hub height

40m hub height

90m hub height

150m hub height

Met mast

40m

Validation examples and model problem issues 227

measurements are in coastal regions, the changes in calculation results can be quite large. We cannot yet say,

based on actual turbines, if one or the other models performs better. But the most important conclusion is: Do

not use different WAsP versions in coastal regions to generate and use wind statistics!

Displacement height calculation

Figure 257 Test wind farm for displacement height

Above is an example of a large wind farm: 67 x 225 kW turbines (31m hub, 29m RD), with a 15 m forest just to

the west (main wind direction). This is an ideal test case, except for the production data logging not being very

good.

Figure 258 Results of calculation with and without displacement height

Validation examples and model problem issues 228

For a shorter period (one quarter), reasonably good production data are collected for the turbines marked on the

map in Figure 167 and shown in this graphic. The standard calculation is a calculation based on mesoscale wind

data for the same period as the production data. Note especially the turbine row T10-T18. Here, the west most

(T10) is calculated as having the highest values due to low wake losses. But the actual production from these

turbines is lowest and the production increases towards east. The obvious reason is the influence from the forest.

Calculation with the forest model (displacement height calculator) captures the decrease in production very well

for the turbines nearer to the forest.

Figure 259 Test of displacement height used for 4000 DK turbines

In another example, an analysis was performed for 4000 operating DK wind turbines where calculated results

were compared with measured production. The forest concept was tested simply by giving a roughness class of

3 and higher a “forest status”. On the x-axis is how much this corrects the individual wind farms, on average: up

to 18% reduction. And, also shown is how the goodness (measured/calculated) is lifted from 80% to almost

100% for most of the influenced turbines. It is obvious that the forest model performs well in improving the

goodness of the more forest- influenced turbines, and that it brings the trend closer in line with that of the least

influenced turbines.

Elevation model pitfalls

Based on many post construction evaluations, some trends are seen when performing a calculation in elevated

terrain. The two major issues are:

• In less steep terrain, higher elevated turbines are under predicted by the WAsP model relative to lower

elevated turbines. With WAsP-CFD, similar results are seen, although there is some improvement.

• In steep terrain (>30% slopes) higher elevated turbines are over predicted by the WAsP model relative

to lower elevated turbines. The RIX correction can partially repair this problem, however, the WAsP-

CFD, or other CFD models, would be a better choice than the WAsP model in this scenario.

Validation examples and model problem issues 229

Figure 260 Ratio measured/calculated for a site in Germany with elevation differences.

Above is an example from a German site with turbines in elevation from 133m to 157m. A very clear trend is

seen in that the turbines with lower elevation are over predicted relative to the turbines with higher elevation -

almost 7% per 10m elevation difference. Based on more internal calculation examples, the value varies from 3-

7% per 10m. This has high importance when having reference turbines with different elevations relative

to the new project to be calculated.

Figure 261 Example as in previous figure but including WAsP CFD calculation.

Above is another example, where WAsP-CFD is tested as an alternative to WAsP. For WAsP with Rix, a ratio of

3.5% per 10m is shown. This ratio is lowered to 3.4% per 10m when based on WAsP-CFD – almost no

improvement.

Validation examples and model problem issues 230

For complex terrain (large steepness), see the WAsP-CFD validation paper.

Checking the Power Curve

3.10.6.1 The C

value

A key parameter when checking the power curve is the non-dimensional C

curve, which can be used for direct

comparison of different power curves. If the maximum C

value exceeds 0.5, the power curve must be assumed

incorrect. C

is the electrical efficiency, meaning that losses in energy conversion of rotor and drive train (gearbox

and generator) are included. If these losses were zero, the C

could have a maximum of 0.593, the “Betz-limit”,

where it, mathematically, can be proven that a turbine rotor cannot take more than 59.3% of energy out of the

wind. Theoretically, if the rotor took 100% of the energy out of the wind, the wind speed behind the rotor would

be 0 m/s, and the air would stop behind the rotor. While the rotors are not “ideal”, there will be losses in the drive

train, therefore, a C

of 0.5 is considered an upper limit.

Figure 262 Example of power, C

and C

curves from windPRO

Above is an example of power, C

and C

curves of a 2 MW turbine with a 90 m rotor diameter. Max. C

is 0.45

in this example. The example is a turbine developed around 2000. Since then, the standard C

has increased

towards 0.5. This is due to better blade design, more efficient control systems and drive train solutions, and the

use of permanent magnets in the generators. The most important issue in turbine design, however, is still the

Cost Of Energy (COE), and the highest C

might not mean the lowest COE. So, even if it was possible to go

above a C

of 0.5, it probably would not bring down COE – at least not with present technology options.

3.10.6.2 HP/PN-check

Another possible method for checking the power curve is the HP/PN method, where HP is the abbreviation for

Helge Petersen, first manager of Risø test station for wind turbines, and PN is for the Manager of EMD, Per

Nielsen.

Helge Petersen did a comprehensive study around 2000 where a large number of power curves were compared.

The conclusion was that grouping the turbines by design/control strategy (pitch/stall and 1-generator/2-

generator), and normalizing by specific power, all turbines had identical power curves. This resulted in a set of

tables, giving the output per square meter rotor area for the different designs/control strategies as a function of

mean wind speed with specific power as a parameter.

Per Nielsen has later taken over the work of Helge Petersen and updated the study with turbines that are more

recent and added some experience-based adjustments (see below). The tables are implemented in windPRO,

and a comparison tool for checking how a given power curve compares to the standard tables is available.

Figure 263 Example of HP check of power curve

Validation examples and model problem issues 231

Above is an example of the HP/PN comparison. As seen in this example, the HP value at 5-6 m/s average wind

speed matches exactly the supplied power curve value. At 7 m/s, the HP value is 2% higher than the power

curve value. At 9 m/s, it is 1% lower. If the deviation is more than +/-3% at a given wind speed, there should be

concerns about the reality of the power curve. Note that the comparison is based on Rayleigh distributed wind

(Weibull k=2) at standard air density (1.225kg/m³). If turbines are running in noise-reduced mode, the HP check

value tells the losses due to noise reduction. If the non-noise reduced mode has a check value of 1 and the

noise reduced a value of 7, the noise reduced operation can be considered to cause a loss of around 6% (see

example below).

Figure 264 Example of HP check of power curves with noise reduction

From windPRO version 2.4, the HP values have been improved in that WTGs smaller than 150 kW are given

reduced HP values based on experience. This means that:

• 50-149 kW: power is reduced to 90% for 50 kW, and, in a linear manner, reduced up to 100% for 150

kW.

• WTGs below 50 kW are reduced to 90%.

• 1 generator, pitch-regulated WTGs are reduced by 4% for mean wind speeds below 7 m/s.

With these corrections, the HP values will be more accurate for all WTG sizes.

From windPRO version 2.8, further improvements are made. There are now 5 check tables:

Table 15 HP/PN grouping of turbine types for power curve check

NOTE: When using the HP check it is VERY important that the field in the turbine catalogue called “Generator

type” is correctly filled. From windPRO version 2.8, the 2-generator and variable speed means two different

check tables. The “variable speed” version reflects more the 2000-12 generation of modern turbines with

inverters to control the rotor speed, while the “2-generator” version is more relevant for older turbines.

The updates are based on comprehensive analyses by Per Nielsen, which was based on 23 reasonable, new

power curves (2000-12 turbines). These are divided into Specific power classes, analysed part by part (foot, leg,

shoulder and arm), and are the basis for establishment of “artificial power curves.” These are then the basis for

calculation of the new “HP/PN-table values” included in most recent windPRO versions.

Stall – 1 generator

Stall – 2 generator/variable

Pitch – 1 generator

Pitch – 2 generator (updated 2012)

Pitch – variable speed (NEW 2012)

Validation examples and model problem issues 232

Figure 265 Example of production vs wind speed for generic turbine

Above are the results of the analyses for the Pitch-variable speed, modern 2000-12 turbine. The parameter is

specific power in W/m

rotor area.

Test of Turbulence scaling

With the SCALER and time step PARK calculation, it is possible to establish the turbulence for each time step

at each WTG position in hub height, see Section 4.8.4. This can partly be used for turbulence correction of the

power curve (only for pitch regulated turbines) and partly to control the WDC in wake loss calculations.

500

1.000

1.500

2.000

2.500

3.000

3 4 5 6 7 8 9 10 11

kwh/m2/y

Mean wind speed (m/s)

New HP/PN production figures per Square m

rotorarea at different specific power

200

250

300

350

400

450

500

550

600

Validation examples and model problem issues 233

Figure 266 The site mast with measurements, turbulence in 40 m and 50 m.

Below the different variants of establishing the turbulence by SCALER calculations at a specific turbine position

is tested.

Figure 267 Turbulence calculated at WTG-1 at 47 m hub height from different sources.

As seen the calculated turbulence based on 40 m as well as 50 m measurements is exactly the same. This

shows the model performs the transformation correct based on different heights.

Appendices: From windPRO 2.9 manual, not included in this manual:   234 
 
© EMD International A/S   •    www.emd.dk   •   windPRO 4.0  •   April 2024 
The Model calculated TI is based on WAsP CFD model results. This comes out with slightly higher turbulence 
than measured. Finally, the TI taken from mesoscale data (EMD ConWx) is shown. This is again somewhat 
higher than the measurements. Overall this means there are not very large deviations. 
 
To test the calculation concept more comprehensive, the calculated TI at two different turbines is tested. At WT-
10 the calculated (free) wind speed  is 5% higher than at  WT-1, therefore round 5%  lower calculated TI  are 
expected at WT-10. 
 
 
Figure 268 The ratio of TI at WT-1 and WT-10 with different calculation settings. 
 
As seen, the result comes out reasonable as expected. The calculated TI is 3-5% higher at WT-1 than at WT-
10. In addition, it works similar based on all four described calculation settings. 
 
It is hereby shown that the SCALER calculates the TI at different positions as expected. The accuracy of the 
calculated turbulence will although not be more precise than the data and models behind. In some terrain, the 
real turbulence might differ more than the model calculations show. 
 
 
 
 
  Appendices: From windPRO 2.9 manual, not included in this manual: 
  MCP2005 – Measure/Correlate/Predict (long term correction) 
The previous version of MCP, first release in 2005, will still be available for backward compatibility. 
 
  Table of figures and tables 
Figure 1 Wind Atlas Method .................................................................................................................................. 9 
Figure 2 Wind statistic view. ................................................................................................................................ 11 
Figure 3 View of wind statistics on map and as rose graphs. ............................................................................. 11 
Figure 4 Edit wind statistic metadata ................................................................................................................... 12 
Figure 5 View additional wind statistic info. ......................................................................................................... 12 
Figure 6 Modify energy level for wind statistic..................................................................................................... 13 
Figure 7 Illustration of energy vs wind speed scaling. ......................................................................................... 14 
Figure 8 Extrapolate roughness. ......................................................................................................................... 14 
Figure 9 Results with extrapolated roughness. ................................................................................................... 15 
Figure 10 Global Wind Atlas web page. .............................................................................................................. 15 

Table of figures and tables   235 
 
© EMD International A/S   •    www.emd.dk   •   windPRO 4.0  •   April 2024 
Figure 11 Place a marker at the specific location for the wanted GWC file. ....................................................... 15 
Figure 12 Download the .GWC file for selected location. .................................................................................... 16 
Figure 13 Explanation of the GWC file from web page. ....................................... Error! Bookmark not defined. 
Figure 14 Save GWC file in folder below "WindPRO Data\Windstatistics" or in project folder. .......................... 16 
Figure 15 The GWC file is seen in the wind statistic browser. ............................................................................ 16 
Figure 16 Comparing GWC file (left) with EU-WindAtlas file (right). ................................................................... 16 
Figure 17 The locations of the 13 WTGs in test calculation. ............................................................................... 17 
Figure 18 Testing GWA (left) versus EU-Wind atlas wind statistics (Beldringe), rightmost. ............................... 17 
Figure 19 Import file formats for roughness from line object. .............................................................................. 19 
Figure 20 Production variation by roughness and specific power ....................................................................... 20 
Figure 21 AEP change vs distance to coastline and hub height ......................................................................... 21 
Figure 22 Roughness in farmland with windbreaks ............................................................................................ 22 
Figure 23 Roughness class vs length ................................................................................................................. 23 
Figure 24 Obstacle model ................................................................................................................................... 24 
Figure 25 Displacement height calculation .......................................................................................................... 24 
Figure 26 Mesoscale terrain used for Standardization of Mesoscale wind data. ................................................ 27 
Figure 27 From Standardized wind to micro scale wind. .................................................................................... 28 
Figure 28 SCALER setup .................................................................................................................................... 29 
Figure 29 Micro terrain in SCALER ..................................................................................................................... 30 
Figure 30 terrain scaling in SCALER ................................................................................................................... 30 
Figure 31 RIX correction in SCALER .................................................................................................................. 32 
Figure 32 Displacement height in SCALER ........................................................................................................ 32 
Figure 33 Turbulence scaling. ............................................................................................................................. 32 
Figure 34 Post calibration in SCALER ................................................................................................................ 33 
Figure 35 Wind data selection in SCALER .......................................................................................................... 35 
Figure 36 Displacement height input data ........................................................................................................... 36 
Figure 37 Displacement height calculator, part of ORA ...................................................................................... 37 
Figure 38 ORA setup ........................................................................................................................................... 38 
Figure 39 RIX correction EWEC 2006 results ..................................................................................................... 40 
Figure 40 Change in AEP% per change in wind speed % .................................................................................. 41 
Figure 41 Input data for RIX correction in statistical based calculations. ............................................................ 41 
Figure 42 RIX correction input in SCALER ......................................................................................................... 42 
Figure 43 Report output with RIX correction ....................................................................................................... 42 
Figure 44 Load measured data and long-term reference .................................................................................... 46 
Figure 45 View of concurrent Energy in Rose graph. .......................................... Error! Bookmark not defined. 
Figure 46 Short Term (dark red) to Long Term (light red) ratios. ......................... Error! Bookmark not defined. 
Figure 47 Full scaler available in MCP for creating reference data from more model points. ............................ 49 
Figure 48 Data statistics table ............................................................................................................................. 50 
Figure 49 Data statistic graph ............................................................................................................................. 51 
Figure 50 Limit period for local and/or reference data. ....................................................................................... 51 
Figure 51 Evaluate reference data ...................................................................................................................... 52 
Figure 52 downloading alternative reference wind energy index's ..................................................................... 53 
Figure 53 Wind index comparison between Merra2 and EMD ConWx datasets. Error! Bookmark not defined. 
Figure 54 Cumulated wind energy index for Merra2 and EMD ConWx. .............. Error! Bookmark not defined. 
Figure 55 Compare to other references (LT Bias) .............................................................................................. 55 
Figure 56 Session setup, how to convert wind speed to energy ......................................................................... 56 
Figure 57 Adjusting the data................................................................................................................................ 57 
Figure 58 The time limited and adjusted concurrent data ................................................................................... 58 
Figure 59 Choice of two different concepts for MCP ........................................................................................... 59 
Figure 60 Uncertainty calculation in MCP ........................................................................................................... 60 
Figure 61 Neural network setup ........................................................................... Error! Bookmark not defined. 
Figure 62 The Compare view in Model LT shown with residuals except for KS Statistics (Default) ................... 62 
Figure 63 Evaluation of Selected model without residuals ................................... Error! Bookmark not defined. 
Figure 64 Evaluation of Selected model with residuals ....................................................................................... 63 
Figure 65 Prediction of long term dataset ........................................................................................................... 64 
Figure 66 Correction for non-long-term representative reference ....................................................................... 64 
Figure 67 The modelled data shown along with the measurements ................................................................... 63 
Figure 68 scaling Local data to LT level .............................................................................................................. 65 
Figure 69 Example of several sessions compared. ............................................................................................ 66 
Figure 70 Part of the report for session overview ............................................................................................... 67 
Figure 71 Uncertainty calculation ........................................................................................................................ 67 
Figure 72 Example of MCP session details report .............................................................................................. 68 
Figure 73 Settings for regression MCP model .................................................................................................... 69 

Table of figures and tables   236 
 
© EMD International A/S   •    www.emd.dk   •   windPRO 4.0  •   April 2024 
Figure 74 Season setup ...................................................................................................................................... 70 
Figure 75 Settings for Matrix MCP model ........................................................................................................... 72 
Figure 76 Settings for Neural Network MCP model ............................................................................................ 73 
Figure 77 Main page input for METEO calculation ............................................................................................. 78 
Figure 78 Wind distribution and WTGs input ....................................................................................................... 79 
Figure 79 Input of shear in METEO calculation................................................................................................... 79 
Figure 80 k Weibull parameter change with height ............................................................................................. 80 
Figure 81 Measured Weibull k change by height – offshore example ................................................................ 80 
Figure 82 Output example from a METEO calculation ........................................................................................ 81 
Figure 83 Production analyse output ................................................................................................................... 82 
Figure 84 Wind data analyse output .................................................................................................................... 83 
Figure 85 Wind statistics selection form. More statistics can be selected by <Ctrl>........................................... 84 
Figure 86 When more statistics are selected, individual weight can be given. ................................................... 85 
Figure 87 Input form for roughness rose data ..................................................................................................... 85 
Figure 88 Input form for ATLAS Hill/Obstacles ................................................................................................... 86 
Figure 89 Graphic roughness rose establishment. ............................................................................................. 86 
Figure 90 Models available/licensed ................................................................................................................... 87 
Figure 91 Selection of roughness and elevation data in Site data object ........................................................... 88 
Figure 92 WAsP interface input form .................................................................................................................. 88 
Figure 93 Set up displacement height in WAsP interface ................................................................................... 89 
Figure 94 Edit WAsP parameters from windPRO ............................................................................................... 89 
Figure 95 Edit WAsP parameters from windPRO ............................................................................................... 90 
Figure 96 Example of WAsP result compared to WAsP-CFD result. .................................................................. 91 
Figure 97 Purpose in Site data object set for Resource map calculation ........................................................... 92 
Figure 98 Select wind statistics(s) for resource map calculation ........................................................................ 92 
Figure 99 Resource map calculation area ........................................................................................................... 92 
Figure 100 Resource map calculation options .................................................................................................... 93 
Figure 101 Resource calculation input options ................................................................................................... 94 
Figure 102 Input of RIX correction handling in resource map calculation ........................................................... 95 
Figure 103 Result layer output from Resource map calculation with RIX ........................................................... 95 
Figure 104 Displacement height input by resource map calculation ................................................................... 95 
Figure 105 Example of resource map difference output with displacement height............................................. 96 
Figure 106 Input data for rescaling a resource map. .......................................................................................... 97 
Figure 107 Main input for STATGEN calculation ................................................................................................ 98 
Figure 108 Statgen input form ............................................................................................................................. 99 
Figure 109 Wind statistics info report output generated from MCP .................................................................. 100 
Figure 110 Flow request export ......................................................................................................................... 101 
Figure 111 Terrain setup for flow request files. ................................................................................................. 101 
Figure 112 Definition of the Result volume. ...................................................................................................... 101 
Figure 113 Definition of direction sectors to be simulated. ............................................................................... 102 
Figure 114 PARK calculation, selection of method ........................................................................................... 102 
Figure 115 2.9 compatible PARK model choices .............................................................................................. 103 
Figure 116 Offshore TI ...................................................................................................................................... 104 
Figure 117 N.O. Jensen model better performing than more advanced models. ............................................. 105 
Figure 118 The WDC recommendations by terrain type and hub height. ......................................................... 106 
Figure 119 WDC calculated from TI based on theory. ...................................................................................... 107 
Figure 120 Curtailment settings in PARK .......................................................................................................... 112 
Figure 121 Import curtailment data to WTG objects ......................................................................................... 114 
Figure 122 Entering grid curtailment settings. ................................................................................................... 115 
Figure 123 Forsting blockage model. ................................................................................................................ 117 
Figure 124 Branlard blockage model. ............................................................................................................... 117 
Figure 125 Selection of WTGs for PARK calculation ........................................................................................ 118 
Figure 126 Explicit link site data – WTG, manual control .................................................................................. 119 
Figure 127 Model parameters and report features in “standard” PARK ........................................................... 120 
Figure 128 Advanced setup options in “standard” PARK (N.O. Jensen) .......................................................... 121 
Figure 129 Wake decay constant (WDC) or TI by direction sector ................................................................... 122 
Figure 130 Import TI from Meteo object ............................................................................................................ 123 
Figure 131 Example of conversion of TI by height ............................................................................................ 123 
Figure 132 Simple turbulence calculator for PARK input .................................................................................. 123 
Figure 133 Alternative wake models in "standard" PARK (Wind statistic based) ............................................. 124 
Figure 134 Alternative turbulence models in "standard” PARK ........................................................................ 124 
Figure 135 Power curve input (air density correction), non-time step based .................................................... 125 
Figure 136 Air density setup form ..................................................................................................................... 126 

Table of figures and tables   237 
 
© EMD International A/S   •    www.emd.dk   •   windPRO 4.0  •   April 2024 
Figure 137 Wind distribution selection in standard PARK with WAsP .............................................................. 127 
Figure 138 Displacement height setup in standard PARK ................................................................................ 127 
Figure 139 RIX input setup in standard PARK .................................................................................................. 128 
Figure 140 Wind distribution selection in standard PARK with WAsP-CFD ..................................................... 128 
Figure 141 Wind statistics selection in standard PARK with WAsP-CFD ......................................................... 129 
Figure 142 Resource files selection in PARK based on resource files ............................................................. 129 
Figure 143 Setup input for PARK based on time series .................................................................................... 130 
Figure 144 Time of day depending power curve setup. .................................................................................... 131 
Figure 145 Wake setup in PARK based on time series (only NO2005 offer all features) ................................. 132 
Figure 146 Wake models for time step calculations. ......................................................................................... 132 
Figure 147 WDC versus turbulence for time step correction, example. ............................................................ 133 
Figure 148 Linear and RSS weight in PARK combination model ..................................................................... 134 
Figure 149 reduction of WDC by number of upwind turbines, Version 2 & 3. ................................................... 134 
Figure 150 Decreasing WDC by upwind turbines. ............................................................................................ 135 
Figure 151 Increasing WDC by upwind turbines. .............................................................................................. 135 
Figure 152 Power curve correction options in time domain .............................................................................. 136 
Figure 153 Reference turbulence for TI correction of power curve ................................................................... 137 
Figure 154 Selection of wind data and SCALER in PARK time series based .................................................. 138 
Figure 155 The 2.9 compatible PARK methods ................................................................................................ 139 
Figure 156 Right click on preview graph to copy to clipboard. .......................................................................... 140 
Figure 157 Right click on preview table to copy to clipboard. ........................................................................... 140 
Figure 158 The PARK main page report results based on wind statistics ........................................................ 141 
Figure 159 Part of the PARK main result from time step calculation ................................................................ 142 
Figure 160 Result to file output options from PARK, left wst based, right time step based. ............................. 142 
Figure 161 Result to file output comparison (shown transposed) ..................................................................... 143 
Figure 162 Result to file output comparison (shown transposed), rightmost columns...................................... 143 
Figure 163 Result to file; Sector wise; output comparison ................................................................................ 144 
Figure 164 Result to file; Park results, WAsP 11 .............................................................................................. 144 
Figure 165 Setup can decide which WTGs and time resolution for Result to file output. ................................. 145 
Figure 166 Result to file output from PARK based on time series .................................................................... 146 
Figure 167 Example of a WakeBlaster CFD simulation. ................................................................................... 148 
Figure 168 Selection of WakeBlaster model. .................................................................................................... 149 
Figure 169 Chosing wind data. .......................................................................................................................... 149 
Figure 170 Preparing WakeBlaster calculation. ................................................................................................ 150 
Figure 171 TI setup for WakeBlaster. ................................................................................................................ 151 
Figure 172 WakeBlaster status appear when ready to start on remote server. ................................................ 152 
Figure 173 Enter URL and API. ......................................................................................................................... 152 
Figure 174 Email notification when result file is ready. ..................................................................................... 153 
Figure 175 Check calculation status on remote server. .................................................................................... 153 
Figure 176 Ready to download WakeBlaster results. ....................................................................................... 154 
Figure 177 Downloaded WakeBlaster results, ready to calculate. .................................................................... 155 
Figure 178 Wake loss by turbine. ...................................................................................................................... 157 
Figure 179 Wake loss by direction, WakeBlaster and PARK2. ......................................................................... 158 
Figure 180 The windfarm layout (Anholt) in the calculation test. ...................................................................... 159 
Figure 181 The calculated wake loss by wind speed, WakeBlaster and PARK2. ............................................ 159 
Figure 182 Decide how to handle existing WTGs. ............................................................................................ 170 
Figure 183 For "Standard" PARK also model modifications shown. ................................................................. 171 
Figure 184 Statistic based wind sector management. ...................................................................................... 178 
Figure 185 Setup of Bat stop loss calculation. .................................................................................................. 180 
Figure 186 Checking "Manually", individual loss per turbine is entered. .......................................................... 182 
Figure 187. The development of calculated wake loss by wind farm size. NO2005 test. ................................. 193 
Figure 188 Park 1&2 and WakeBlaster test by wind farm size. ........................................................................ 194 
Figure 189 Ainslie in Open Wind and windPRO test by wind farm size. ........................................................... 194 
Figure 190 Measured left, calculated reductions by PARK1 (org. N.O. Jensen) right. (by direction). .............. 195 
Figure 191 Calculated reductions by PARK2 left and by NO2005 with DA, right. ............................................ 196 
Figure 192 Factor on WDC and thereby increased WDC by number of upwind turbines. ............................... 196 
Figure 193 Map of Horns Rev area, HR1 south, HR2 mid and HR3 north. ...................................................... 197 
Figure 194 Plots showing calculated long-term wake losses for different combinations. ................................. 199 
Figure 195 Impact of neighbour wind farm in calculations. ............................................................................... 199 
Figure 196 Turbine by turbine calculated wake loss. ........................................................................................ 199 
Figure 197 Calculated wake losses for HR3 with 8 MW WTGs. ....................................................................... 200 
Figure 198 details on wake loss calculation for HR3. ....................................................................................... 200 
Figure 199 HR2 & 3 wake loss calculation with different combinations. ........................................................... 201 

Table of figures and tables   238 
 
© EMD International A/S   •    www.emd.dk   •   windPRO 4.0  •   April 2024 
Figure 200 Advanced modelling of WDC by TI based on detailed Power by TI analyses. ............................... 202 
Figure 201 HR1 calculation for 2008 & 2012 compared to measurements. ..................................................... 202 
Figure 202 HR1 measured and calculated at TI>6%. ....................................................................................... 202 
Figure 203 HR1 measured and calculated at TI<6%. ....................................................................................... 202 
Figure 204 Measured/calculated for lower and higher TI for HR1. ................................................................... 203 
Figure 205 Long term calculation of HR1 compared to measured. .................................................................. 203 
Figure 206 The calculated wake loss vs loss on top of wake loss and binned loss.......................................... 203 
Figure 207 Comparing calculated wake losses with all losses including wake loss, HR2 left, HR3 right. ........ 204 
Figure 208 Monthly calculated and measured production for HR2. .................................................................. 205 
Figure 209 Calculated wake losses and seen "other loss" for HR2. ................................................................. 205 
Figure 210 Calculated and measured + losses by Ainslie DAC calculation. .................................................... 206 
Figure 211 Timeseries calculated and measured by month for HR3. ............................................................... 207 
Figure 212 Monthly loss distribution by loss bin for HR2 and different calculation variants. ............................ 207 
Figure 213 Measured and calculated from Performance Check, month data (includes “other” losses). .......... 208 
Figure 214 The calibration tool: goodness vs calculated, PARK2 adv. default (left) and tuned (right). ............ 208 
Figure 215 Wind farm layout and ratios measured/calculated P2 tuned, Lillgrund offshore. ............................ 209 
Figure 216 Ainslie with WTG by WTG goodness slightly poorer than PARK2. ................................................ 209 
Figure 217 Layout of the Egypt wind farm. ........................................................................................................ 210 
Figure 218 TI by direction sector. ...................................................................................................................... 210 
Figure 219 Egypt large wind farm wake loss calculations. ................................................................................ 211 
Figure 220 Large wind farm complex covering 45 km east-west. ..................................................................... 212 
Figure 221 Losses on top of wake losses for months with < 15% loss (74% of month data). .......................... 213 
Figure 222 Calculation with advanced WDC(TI) for offshore. ........................................................................... 214 
Figure 223 Ormonde measured and calculated by month. ............................................................................... 215 
Figure 224 West of Duddon Sands measured and calculated by month. ......................................................... 215 
Figure 225 Assumed all time losses in addition to wake losses for the 6 wind farms....................................... 216 
Figure 226 Final calculated wake losses per period per wind farm. ................................................................. 216 
Figure 227 Calculation for 20y all wind farms running full time, compare PARK2 and Ainslie. ........................ 217 
Figure 228 WDC(TI) for different configurations with PARK2. .......................................................................... 219 
Figure 229 More MCP sessions can show more combinations for comparisons ............................................. 220 
Figure 230 Long-term consistency using mesoscale data ................................................................................ 221 
Figure 231 Wind speed ratios Model data/measurements 2004-18 ................................................................. 222 
Figure 232 Merra-2/measurements for Høvsøre 100m (Merra-2 scaled to Høvsøre average) ........................ 222 
Figure 233 Four windstatistic results for Lillgrund offshore, different WAsP’s. ................................................. 223 
Figure 234 Test case calculations showing the WAsP stability model shift. ..................................................... 224 
Figure 235 Wind profiles, measured and calculated ......................................................................................... 225 
Figure 236 Model setup for test of WAsP model ............................................................................................... 225 
Figure 237 Map details for test setup ................................................................................................................ 226 
Figure 238 Results of WAsP 10.2 vs 9 calculations .......................................................................................... 226 
Figure 239 Test wind farm for displacement height .......................................................................................... 227 
Figure 240 Results of calculation with and without displacement height .......................................................... 227 
Figure 241 Test of displacement height used for 4000 DK turbines ................................................................. 228 
Figure 242 Ratio measured/calculated for a site in Germany with elevation differences. ................................ 229 
Figure 243 Example as in previous figure but including WAsP CFD calculation. ............................................. 229 
Figure 244 Example of power, C
e
 and C
t
 curves from windPRO ...................................................................... 230 
Figure 245 Example of HP check of power curve ............................................................................................. 230 
Figure 246 Example of HP check of power curves with noise reduction .......................................................... 231 
Figure 247 Example of production vs wind speed for generic turbine .............................................................. 232 
Figure 248 The site mast with measurements, turbulence in 40 m and 50 m. ................................................. 233 
Figure 249 Turbulence calculated at WTG-1 at 47 m hub height from different sources. ................................ 233 
Figure 250 The ratio of TI at WT-1 and WT-10 with different calculation settings. ........................................... 234 
 
Table 1 Calculation/correction options: Wind statistics vs time domain ................................................................ 5 
Table 2 Establishment of calibrated long-term data time series ........................................................................... 7 
Table 3 Roughness definitions ............................................................................................................................ 19 
Table 4 Model and data validation tools .............................................................................................................. 26 
Table 5 Result to file output from a METEO calculation ..................................................................................... 81 
Table 6 Recommended settings for N.O. Jensen PARK models ...................................................................... 104 
Table 7 Basic assumptions for hub height dependent WDC with examples (onshore) .................................... 105 
Table 8 Roughness class and length relations.................................................................................................. 106 
Table 9 Decreasing WDC by upwind turbines................................................................................................... 135 
Table 10 Increasing WDC by upwind turbines. ................................................................................................. 135 
Table 11 Output from PARK based on time series to spreadsheet .................................................................. 146 

Table of figures and tables 239

Table 12 Output from PARK based on time series to spreadsheet – column documentation .......................... 147

Table 13 WakeBlaster and PARK2 results. ....................................................................................................... 157

Table 14 Testing MCP with Høvsøre data ........................................................................................................ 220

Table 15 HP/PN grouping of turbine types for power curve check ................................................................... 231