5 LOADS SITE COMPLIANCE & LOAD RESPONSE 
5.1  Introduction, definitions, and step-by-step guide ................................................................................. 2 
  Requirements of IEC 61400-1 standards: ed. 4 (2019), ed. 3 (2010) and ed. 2 (1999) ........... 2 
  Typical modes of use ................................................................................................................ 3 
  Step-by-step guide .................................................................................................................... 5 
5.2  SITE COMPLIANCE ................................................................................................................................... 5 
  Setting up a SITE COMPLIANCE calculation ........................................................................... 5 
  IEC Checks - Main checks ...................................................................................................... 21 
  IEC Checks - Other checks ..................................................................................................... 55 
  (Re)calculate all ....................................................................................................................... 64 
5.3  LOAD RESPONSE ................................................................................................................................... 66 
  Setting up and running the fatigue load calculation ................................................................ 69 
  Results – the fatigue load estimates ....................................................................................... 73 
  Description of the generic wind turbine models ...................................................................... 75 
  How-to add a new turbine load response model (manufacturers only) .................................. 76 
5.4  Exports and Result-to-file ...................................................................................................................... 81 
5.5  Reports and printing  .............................................................................................................................. 83 
5.6  References ............................................................................................................................................... 91 
Appendix I     -   Gumbel’s Theory of Extremes and more ............................................................................ 93 
Appendix II    -   Frandsen Effective turbulence model ................................................................................. 98 
Appendix III   -   Critical, Caution & OK limits in: IEC ed. 2 / ed. 3 / ed. 4. ................................................. 101 
Appendix IV   -   Theory of LOAD RESPONSE and Fatigue ........................................................................ 105 
Appendix V    -   Curtailment........................................................................................................................... 110 
Appendix VI   -   IEC 61400-1 ed. 2 (1999)...................................................................................................... 112 
Appendix VII  -   IEC 61400-1 ed. 4 (2019)...................................................................................................... 114 
Appendix VIII -  Lifetime extension DNVGL-ST-0262 (2016) ........................................................................ 119 
Appendix IX   -  Siteres ambient climate files (from Resource, GASP, etc.) ............................................. 121 
Appendix X    -  Tropical Cyclone Analysis ................................................................................................... 122 
Appendix XI   -  Spectral Correction Method / Downscaling ....................................................................... 122 
 
   

Introduction, definitions, and step-by-step guide 2

5 LOADS SITE COMPLIANCE & LOAD RESPONSE

5.1 Introduction, definitions, and step-by-step guide

Establishing a proper layout and selecting a suitable turbine model are important steps in developing a wind

energy project. Wind turbines are typically designed for 20 years lifetime and according to a number of standard

climatic design classes, e.g., the IEC classes IA or IIIC. The Roman numeral defines the wind speed class I, II

or III, and the letter defines the turbulence class A, B or C. Class IA is the strongest standard class, and the least

strong class is IIIC.

Installing a turbine of the incorrect class on a site could result in premature structural failure and ruin a project.

On the other hand, installing a turbine from a class above that required could add unnecessary extra costs,

making a project financially unviable.

The windPRO modules SITE COMPLIANCE and LOAD REPONSE help the user determine which particular

turbine class is suitable for a site. The modules also help the user identify critical risks of a project.

The requirements of the wind turbine design classes are defined in the international standard:

IEC 61400-1 ed. 3 (2010) “Wind turbines Part 1 - Design requirements” [1, 2]

Most sections concern design requirements of the standard wind turbine classes. However, Section 11 describes

“assessment of a wind turbine for site-specific conditions”, i.e., the assessment of whether a wind turbine class

complies with the conditions in a particular site and layout or, in other words, “site compliance”.

Requirements of IEC 61400-1 standards: ed. 4 (2019), ed. 3 (2010) and ed. 2 (1999)

Table 1 in the IEC 61400-1 ed. 3 defines the fundamental design parameters of the standard wind turbine climate

classes described above.

Wind turbine class

III

Vref [m/s]

50.0

42.5

37.5

Values

Iref [-]

0.16

Specified

Iref [-]

0.14

by the

Iref [-]

0.12

designer

Table 1. IEC 61400-1 ed. 3 “Table 1”.

Section 11 of the IEC 61400-1 ed. 3 (2010) standard (henceforth referred to as the “IEC standard”) describes

seven main parameters used to classify the site. Firstly, there is a parameter describing the terrain complexity

then six parameters relating to the site’s wind climate. Together, these seven “Main IEC checks” are:

• Terrain complexity

• Extreme wind

• Effective turbulence

• Wind speed distribution

• Wind shear

• Flow inclination

• Air density

The standard also lists a number of “other environmental conditions” to be assessed for a site. Of these additional

parameters, we have selected three parameters which are more likely to be critical and which can be estimated

with an acceptable accuracy. These parameters referred to as “Other checks” are:

• Earthquake hazard

• Lightning rate

• Extreme and normal temperature range

Introduction, definitions, and step-by-step guide 3

The omitted parameters are “Icing, hail and snow”, “Humidity”, Salinity”, “Solar radiation” and “Chemically active

substances”.

Section 11.1 of the IEC standard ([1], p 52) describes how to compare a site specific wind climate to the design

climate for the relevant IEC class, e.g., IIB, to prove that the site conditions do not violate the turbine class.

“…It shall be shown that the site-specific conditions do not compromise the structural

integrity. The demonstration requires an assessment of the site complexity, see 11.2, and an

assessment of the wind conditions at the site, see 11.3. For assessment of structural integrity

two approaches may be used:

a) a demonstration that all these conditions are no more severe than those assumed for

the design of the wind turbine, see 11.9;

b) a demonstration of the structural integrity for conditions, each equal to or more severe

than those at the site, see 11.10.

If any conditions are more severe than those assumed in the design, the structural and

electrical compatibility shall be demonstrated using the second approach.”

In short, ‘approach a’ means that, if all the wind climate checks in SITE COMPLIANCE are OK, it can be

concluded that a turbine class is suitable. But, if one or more climate checks are exceeded, ‘approach b’ should

be used – which corresponds doing the load calculation in LOAD RESPONSE based on the results from SITE

COMPLIANCE.

IEC 61400-1 ed. 2 (1999)

As of windPRO version 3.1 SITE COMPLIANCE and LOAD RESPONSE also support the now obsolete second

edition of IEC 61400-1 released in 1999 [23]. The reason to support this older version is that some conservative

manufacturers are still selling turbines certified according to this second edition (ed. 2). In addition, new wind

farm extensions can influence existing turbines certified to ed. 2, and their suitability must be evaluated according

to their original design standard.

The design parameters partly have different definitions in ed. 2 and ed. 3. The reference wind speed Vref is

defined in the same way, as the extreme wind speed with 50 years return period. However, the “reference”

turbulence intensity is defined differently. In ed. 3, “Iref” is the mean turbulence at 15m/s. In ed. 2, the turbulence

parameter is called “I15” and is the 84

percentile of turbulence at 15m/s.

Wind turbine class

III

Vref [m/s]

50.0

42.5

37.5

30.0

Values

Specified

by the

designer

Vave [m/s]

10.0

8.5

7.5

6.0

I15 [-]

0.18

a [-]

I15 [-]

0.16

a [-]

Table 2. IEC 61400-1 ed. 2 “Table 1”.

The implementation of ed. 2 in SITE COMPLIANCE is based as closely as possible on the ed. 3 implementation,

without violating the intention of ed. 2. The reason for this approach is that ed. 3 was created to handle several

problematic or too open issues in ed. 2. Appendix VI summarizes the requirements in the IEC 61400-1 ed. 2

(1999).

IEC 61400-1 ed. 4 (2019)

Appendix VII summarizes the requirements in the IEC 61400-1 ed. 4 (2019).

Typical modes of use

SITE COMPLIANCE can be used with various qualities of input data and external software licenses. The

following five main modes are available:

Introduction, definitions, and step-by-step guide 4

1. Mast data & flow model(s)

2. Mast data only

3. No mast data

4. 3

party WTG results (*.xml)

5. Ambient site result

From windPRO 4.0 SITE COMPLIANCE offers the option to modify each main modes as an Offshore site. The

implication of this modification is described in appendix XI.

Full functionality is obtained for projects with site mast(s) with multiple measuring heights and valid external

licenses for both WAsP

and WAsP Engineering (WEng 3)

and pre-run WAsP-CFD

flow results available

(alternatively pre-run Flowres results, from any CFD model). When using WAsP, the module is considered to be

operating in its main mode. In this mode, a WAsP Engineering license is not a requirement, but it adds additional

calculation options which improve the quality of the results.

Minimum data level is the second mode Mast data only. This mode requires only an on-site mast with multiple

measuring heights and no external software licenses to calculate all main checks.

Mode three, No mast data, is for projects with no on-site mast, as is typically the case in mature markets with

many installed WTGs like Denmark or Germany. This mode requires valid external licenses for WAsP and WEng

3 as well as regional wind statistics (wind atlas / lib file) to complete all seven main IEC checks.

In Mode one, there are two sub-modes for the WAsP setup: Long term corrected wind statistics and Mast directly.

The first sub-mode covers the situation where a long term corrected wind statistics has been generated via MCP

for each on-site mast. This mode ensures a WAsP calculation which is consistent with any PARK calculations

based on these wind statistics. The second sub-mode (Mast directly) integrates the STATGEN calculation for

each mast and so includes both steps of the two-step WAsP procedure, taking the on-site mast data directly as

input. This sub-mode opens the possibility to perform a simple long-term correction within SITE COMPLIANCE,

as this is required in the IEC standard if on-site data is not representative of the long-term.

Mode four provides the possibility to load results of the main IEC checks provided by a 3

party. These results

can be used in LOAD RESPONSE but cannot be combined with any calculation of the IEC checks in SITE

COMPLIANCE. The required file format is an xml-file, either with all WTGs in one file or one file for each WTG.

This xml-format is also available as a result-to-file option for standard SITE COMPLIANCE calculations.

http://www.wasp.dk/Products/WAsP.aspx

http://www.wasp.dk/Products/WEng.aspx

http://www.wasp.dk/Software/WAsP-CFD

Note: that WAsP ≥11 is required to use WAsP-CFD results in WAsP

SITE COMPLIANCE 5

Step-by-step guide

For a quick step-by-step getting-started introduction to SITE COMPLIANCE and LOAD RESPONSE, the user is

recommended to use windPRO Quick Guides:

SITE COMPLIANCE Quick Guide

LOAD RESPONSE Quick Guide

Lifetime Analysis Quick Guide

5.2 SITE COMPLIANCE

Setting up a SITE COMPLIANCE calculation

Subsections:

5.2.1.1 Main

5.2.1.2 Mast data

5.2.1.3 WTGs

5.2.1.4 Mast-WTG

5.2.1.5 Long term correction

5.2.1.6 WAsP

5.2.1.7 WEng

5.2.1.8 WAsP-CFD

5.2.1.9 Flowres

5.2.1.10 “3rd party results” mode

5.2.1.11 Curtailment

Before starting the SITE COMPLIANCE calculation make sure you have a licensed version of SITE

COMPLIANCE and that your windPRO project contains the following data/licenses:

1. A layout of WTG objects

2A. A digital elevation model: line object or elevation grid object (TIN)

3A. A site mast with multiple heights (all carefully checked and cut to a number of full years)

And/Or

3B. A site data object (with a wind statistic)

A valid WAsP license

A valid WEng license

And/Or

3C. A valid WAsP 11 license

And WAsP-CFD result files

And/Or

3D. A Flowres result file(s)

(export from any external CFD model)

3E. A predefined xml-file

with “3

party results” for the main IEC checks

The following section 5.2.1 describes the important setup steps of a SITE COMPLIANCE calculation tab by tab.

Note that not all tabs described in the following will be available in each of the main modes.

See Flowres format in: http://www.emd.dk/files/flow/EMD_technote_Generalized_Flow_Request_Result.pdf

See xml format specification in: http://www.emd.dk/files/windpro/windPRO_third_party_Load_Response.pdf

SITE COMPLIANCE 6

Main

On the Main tab of SITE COMPLIANCE, the general mode of use (described previously) is selected under Site

and layout check using:. The appropriate choice depends on the availability of on-site measurements and

licenses for the flow models WAsP, WAsP Engineering (WEng) or WAsP CFD. Alternatively, external (CFD) flow

results in Flowres files or external 3

party IEC check results in xml files can be used.

Figure 1. Main tab in SITE COMPLIANCE setup, showing all sub tabs.

The Flow models with a valid license should be ticked. WEng 3 and WAsP / WAsP-CFD are only selectable if

valid licenses are available. Note that selection of WEng requires that WEng version 3 or newer is installed and

licensed and that the PC has internet access. The license system of WEng and non-dongle based WAsP

versions (from version 11) will check license status regularly via the internet.

If a valid license is available for LOAD RESPONSE, the Include LOAD RESPONSE option can be checked to

include load calculations based on the IEC main check results for each wind turbine.

If curtailment criteria have been defined on the individual WTG objects to shut down in certain wind speed and

direction intervals, these curtailment criteria may be included by checking the option Apply sector curtailment.

The curtailment will influence the result of the Effective turbulence check due to reduced wake effects from shut

down turbines. Curtailment also affects results in LOAD RESPONSE because a shutdown turbine will experience

less fatigue loads.

The Design standard must be selected between the options: IEC61400-1 ed. 4 (2019), IEC61400-1 ed. 3 (2010)

or IEC61400-1 ed. 2 (1999).

SITE COMPLIANCE 7

If a Design class is not set individually in each of the relevant WTG objects, Overrule option may be selected to

select the Design class. The basic design parameters are summarized for each design class in the table below

the overrule selection. If Class S is selected, the empty fields for class S in the summary table becomes editable

to fill in the class design values. The checkbox “Wind speed dependent TI90” enables a table to explicitly define

custom TI design values for each wind speed bin, instead of just filing the Iref class S field.

Mast data

On the Mast data tab, the relevant Site masts must be selected by checking them. Once a mast is selected, it

expands, and the Main height must be selected. This is the height used as basis in all IEC calculations and in

WAsP calculations if “run in Mast directly mode” is used. If multiple heights are available, the heights to be used

in calculation of vertical shear (i.e., wind speed variation with height) must be checked (including the main

height).

Note that if any fields such as Sample rate, Duration or Recovery are shaded red, this indicates potential

problems. The IEC standard requires 10-minute data to be used for on-site measurements, and that data should

not be seasonally biased, i.e., duration should be an integer number of years.

If available, one or more Long term reference series may be selected as well. This is done by checking it and

setting the Purpose to Long term reference. A third Purpose available is Climate, which may be used in case no

temperature series has been measured on the site mast.

Figure 2. Mast data tab showing a typical setup with a Site mast, a Long term reference series and a Climate

mast (typically the nearest 10m Synop station for long-term temperature data).

WTGs

On the Layout tab, the WTGs of the project are selected. In SITE COMPLIANCE, the IEC checks are performed

only for WTGs classified in the project as New WTGs. Existing WTGs which are selected are included in relevant

calculations (e.g., wake effects), but no individual results appear for such WTGs. Check the layer(s) with the

SITE COMPLIANCE 8

New WTGs and adjust individual selection at the bottom of the tab if only some WTGs in the layer are to be

included.

Note that at the bottom of the page you may select whether you want to have the object description or the user

label in the site tables, graphs and reports.

WTGs in noise reduced modes should generally be avoided as the wind speed at rated power is extracted from

the power curve as part of a main IEC check. This may not work well for power curves in reduced mode.

Figure 3. Layout tab, where the relevant project WTGs (New WTGs) are selected.

Displacement height (Common for mast and WTGs)

There are three options for displacement height (see Figure 3, left side): “No displacement height”,

“Displacement height from objects” and “Displacement height calculator”. The last two options are relevant if

omnidirectional displacement heights have been set on mast or WTG objects. The “calculator” (last option) is

the most complete option as it works with sector wise displacement heights and a buffer zone around forests

with fractional ramp down of displacement heights for nearby forests.

Mast-WTG

This tab is only available in modes one and two where an on-site mast is available. Here, the user decides which

mast to use for which WTG. The default setup for Mast-WTG link choses the nearest mast for each WTG.

SITE COMPLIANCE 9

However, this may be adjusted manually where there are multiple choices for mast by selecting the Manual

mast-WTG matrix.

Figure 4. Mast-WTG tab – the appropriate site mast should be chosen for each WTG.

Long term correction

This tab is only available in mode one and two and when WAsP mode is set to Mast directly. In these cases, it

is important to evaluate whether the measurement period of the mast(s) is long-term representative. If not,

a long-term correction may be appropriate and can be applied from within SITE COMPLIANCE.

Note that this tab offers a simple alternative to the typical procedure of making a long term corrected wind

statistics for each site mast via MCP and using the WAsP option Long-term corrected wind statistics on the Main

tab.

The long-term correction available within SITE COMPLIANCE supplements the MCP methods in the MCP

module as it is a wind speed index correction, where the index correction in MCP is based on an energy index.

Another reason for this difference is that, in SITE COMPLIANCE, the focus is loads and not energy.

Where the selection is No correction - data are representative, no further action is required. If a long term series

was setup on the Mast data tab and it fully overlaps with the Site mast data, then the option Wind speed index

correction may be selected. By selection of this option, the Calculate corrections button turns yellow and must

SITE COMPLIANCE 10

be clicked. Evaluate the results for each site mast (if more than one) in terms of Index and R

(correlation

coefficient). The plot is only shown for the one selected.

Figure 5. Long-term correction tab.

An index of 100% means that the mast period is representative of the long-term. A value either above or below

100% means that the mean wind speed measured by the site mast must be corrected by the appropriate amount

(inverse index) to represent the long term climate.

The index is calculated from the reference series as the ratio of the mean wind speed of the concurrent period

divided by the mean of the full length of the series. Correlations coefficients are based on concurrent monthly

mean wind speeds, which are also plotted on the graph where they are normalized to an average of 100.

WAsP

The WAsP tab is available if WAsP was checked on the main tab. Options on this tab depend on the selected

WAsP mode. In WAsP mode, Mast directly a single Site data object must be selected to define the terrain and

roughness maps to send to WAsP (see below).

SITE COMPLIANCE 11

Figure 6. WAsP tab in WAsP mode Mast directly.

Running SITE COMPLICANCE in WAsP mode Long-term corrected wind statistics, a Site data object with a

wind statistics must be selected for each mast (only one in the case illustrated). Please ensure that the selected

wind statistics in each site data object, in fact, holds a long-term corrected wind statics for the mast.

Figure 7. WAsP tab in WAsP mode Long-term corrected wind statistics.

In No mast mode, the WAsP tab requires the user to check the Site data objects to be used in the calculation.

The nearest site data objects will then be used for each WTG.

Figure 8. WAsP tab in No mast mode.

When the Site data object(s) has been selected, the WAsP calculation is run by clicking the green Run WAsP

Calculation button. WAsP parameters may be adjusted by clicking the Edit WAsP parameters button.

Figure 9. WAsP tab – ready to run the calculation or change WAsP parameters.

Figure 10. WAsP tab, before and after successful calculation.

When the WAsP calculation is complete the calculation button turns white and the red stop sign on the WAsP

tab becomes a green tick mark (see Figure 10).

Displacement height

See section Displacement height.

SITE COMPLIANCE 12

WEng

SITE COMPLIANCE offers an easy-to-use integration of WAsP Engineering 3 (WEng 3) and later versions. We

recommend using WEng 4 for stability (available in windPRO 3.1 SP1). The integration requires full external

installation and licensing of WEng, and turns windPRO into the GUI of the WEng flow model. This has enabled

a significant simplification of setting up and running WEng. Contrary to WAsP, WEng is a grid-based model that

calculates the flow parameters (except turbulence) for each grid point of an entire rectangular calculation domain

defined by the user.

First, a Site data object must be selected to define the terrain and roughness maps passed to WEng.

Then a Buffer around all masts/WTGs must be chosen. This buffer distance defines the extension of the

calculation domain. Default is 5km, which is a compromise between accuracy and calculation time. In cases with

prominent roughness changes just beyond 5km or large scale terrain features, the buffer should be extended as

required.

The grid resolution defaults to 50m which, typically, is acceptable. In sites with rapid variations in terrain, such

as a narrow ridge, a finer resolution should be chosen.

Notice that if the Buffer or Grid resolution is changed the Number of grid points will update accordingly. The

calculation time of WEng is optimized if the Number of grid points stays just below 170, 340 or 680 …etc.,

in both directions. This is due to internal zero-padding inside WEng.

Figure 11. WEng tab.

SITE COMPLIANCE 13

Once appropriate settings are made, the calculation can be performed by clicking the green Run WEng (WAsP

Engineering) calculation button. A WEng calculation may take several minutes. A main reason is that turbulence

is predicted for each WTG and mast position.

Figure 12. WEng tab before and after successful calculation.

Note that for larger domains and fine resolutions, in particular WEng 3.x may crash. WEng does not inform

windPRO that it has crashed but windPRO’s connection to WEng simply times-out. In such cases it is possible

to have windPRO automatically dump a WEng project with the WEng calculation setup (wind direction and

domain) which caused the crash. This file will be named “WEngCrashProject.wep” and is situated in the

windPRO project root folder. This WEng project file may be opened and tested directly in WEng and if the crash

is reproduced, the project should be passed directly to the WAsP/WEng support team at DTU for diagnostics

and further help. To enable this option see WEng Advanced setup.

Displacement height

See section Displacement height.

If displacement heights have been set, either for the Meteo or WTG objects, in the WEng calculation, the

typical effect is to decrease predicted wind speeds at the object in question, whereas predicted turbulence and

wind shear will normally increase.

Advanced setup

Prior to running the WEng calculation, the Advanced setup options may be reviewed and adjusted.

Figure 13. WEng tab, Advanced setup.

This setup illustrates how the flow modelling actually works inside WEng via the Setup of reduced geostrophic

wind that allows adjustment of Wind speed, Height, Roughness length and the number of Sectors.

As WEng is a linearized model, the results of the WEng flow modelling in terms of relative speed-up factors and

turbulence intensity will not depend on this setup. Only in the special case of offshore or semi offshore conditions

is extra caution needed as, here, the linearity does not hold due to increasing sea roughness with wind speed.

In such special cases, it is advised to run WEng and close SITE COMPLIANCE and export the WEng flow results

via right-clicking on the calculation and choosing Result-to-file. The wind speeds predicted for each WTG in the

result should match approximately the expected extreme wind speed for the WTGs. If the result is too low or

high, the reduced geostrophic wind speed can be adjusted accordingly in the WEng advanced setup to properly

model the flow conditions during the on-site extreme wind conditions.

Turbulence calculation lets the user deselect the calculation of turbulence via the option None or choose the

alternative model Scanlan. However, it is recommended to use the default choice of including turbulence with

the model Kaimal.

SITE COMPLIANCE 14

Note that dump of a WEng project file and use of obstacles is activated in these advanced settings. The obstacle

option is only active for WEng version 4.X and newer.

WAsP-CFD

From windPRO 3.0, SITE COMPLIANCE integrates flow results from WAsP-CFD. The integration requires a

full external installation and licensing of at least WAsP 11 and that a WAsP-CFD calculation (the cluster

calculation which establishes the raw CFD flow results) has been performed using appropriate CFD tiles which

cover the relevant masts and turbine positions.

First step is to add the ‘raw’ CFD results by pressing the button Add WAsP-CFD file(s) or Add all from calculation

and then, in the file browsing window which pops up, select the relevant cfdres files. The cfdres files are generally

situated in a project folder named OnlineCFDResults.

Figure 14. Pop-up menus to select WAsP-CFD result files. Left: menu for selecting individual files. Right: menu

for selecting all results of an entire calculation.

SITE COMPLIANCE 15

Figure 15. WAsP-CFD tab in Mast directly mode.

If SITE COMPLIANCE is run in the modes Long-term corrected wind statistics or No mast data, it is also

necessary to select the appropriate wind statistics. In the Long-term corrected wind statistics mode, a wind

statistics must be selected explicitly for each on-site mast.

Figure 16. Additional part of WAsP-CFD tab which is shown in the mode: long-term corrected wind statistics.

Once appropriate selection of CFD result files and, if required, also wind statistics has been done, the calculation

can be performed by clicking the green Run WAsP-CFD & extract raw CFD results button. The calculation may

take several minutes and will first call WAsP with the CFD results and wind data or wind statistics to predict the

sector-wise Weibull distributions and frequencies at each WTG position, at hub height, and hub height +/- ½

rotor diameter. The last option is default and may be deactivated. After that, SITE COMPLIANCE will extract the

additionally required flow parameters like vertical inflow angle, speed-up and veer directly from the raw CFD

flow.

Figure 17. WAsP-CFD tab before and after successful calculation.

Displacement height

SITE COMPLIANCE 16

See section Displacement height.

Advanced setup

On the WAsP-CFD tab, the tick box Advanced settings (Raw CFD results) will give access to additional options

for using the raw CFD flow results. The options relate to the how the information in all 36 directions in the flow

results may be used to smoothen the flow results. The reason is that each of the 36 directions is run as an

independent numerical simulation. When comparing the results of all sectors, “numerical noise” from the

solutions is obvious as a scattering or high frequency variation in the results, such as speed-up. To limit this

“noise” and improve the robustness of, e.g., propagation models in extreme wind, applying a smoothing filter is

good solution. Per default, a smoothing kernel of [0.25, 0.5, 0.25] is applied to the raw CFD parameters veer, TI

and speed-up. This means that each directional result is replaced by the weighted average of itself and its two

neighbouring simulations using the weights of the chosen “kernel”. Inside WAsP, a smoothing kernel of [1/3, 1/3,

1/3] is used for predicting Weibull distributions when using 12 sectors of 30 degrees. Setting the smoothing

kernel to [0,1,0] corresponds to no smoothing.

Figure 18. WAsP-CFD tab, Advanced settings (Raw CFD results).

Flowres

From windPRO 3.1, it is possible to load and utilize the results of any external flow model, typically CFD models.

The required data format is called “Flowres” and is defined here

. This format contains a more complete set of

the basic flow results from the “raw” simulation output than, e.g., the wrg/rsf (resource) format. Figure 19 shows

how the Flowres option is selected on the main tab of SITE COMPLIANCE as an alternative to WAsP, WEng

and WASP-CFD. In many regards, Flowres is very similar to the underlying flow data format of WAsP-CFD

(called .cfdres). Both formats are zip files, which contain an xml file defining the overall simulation setup. File

names of the result data files are also included in the zip file as a number of grid files, one for each simulation

direction and for each result parameter, such as turbulence intensity for direction 170° or speed-up for 20°.

Figure 19. Selection of Flowres on the main tab, which activates the Flowres tab.

Figure 20 shows the Flowres tab, where the first step is to select the relevant Flowres files by pressing the button

“Add Flowres file(s)” and selecting the relevant files in the pop-up menu. Once the Flowres files are selected,

the “Run calculation” button will turn green and when clicked it starts the “calculation” and the processing of the

files. Depending on the file size and number of turbines, this may take several minutes.

Flowres format: http://www.emd.dk/files/flow/EMD_technote_Generalized_Flow_Request_Result.pdf

SITE COMPLIANCE 17

Figure 20. Flowres setup and calculation tab.

Once the loading, calculation and processing of the Flowres files is completed, the “Run calculation“ button turns

grey, and the red stop sign on the tab becomes a green tick mark (see Figure 21).

Figure 21. Flowres tab before and after completed loading/calculation/processing.

Advanced setup

By checking “Advanced settings”, the advanced option shown in Figure 22 becomes available. These

parameters relate to the smoothing of the CFD results across direction to reduce the numerical noise. See

Advanced setup for WAsP-CFD for further details.

Figure 22. Advanced setup options for Flowres.

Displacement height

SITE COMPLIANCE 18

See section Displacement height.

“3rd party results” mode

From windPRO 3.1, it is possible to load pre-calculated results of the main IEC checks to allow use of LOAD

RESPONSE by users with in-house calculation procedures for the main IEC checks. The 3rd party file format

is xml and is also used as result-to-file export option for ordinary SITE COMPLIANCE calculations. It is

optional whether to have results for several turbines in one 3rd party xml file or to have one xml file per turbine.

Figure 23. Selection of 3rd party on the SITE COMPLIANCE Main tab.

Figure 24. 3rd party tab with xml files selected and loaded.

Note the summary column named “Resolution” of the loaded IEC main check results in the table in Figure 24,

which indicates whether the loaded results for each IEC check are omnidirectional or sectorwise. For Effective

3rd party format description: http://www.emd.dk/files/windpro/windPRO_third_party_Load_Response.pdf

SITE COMPLIANCE 19

turbulence, there is also a third option called “Total turbulence” which represents turbulence results at a

resolution of 1 degree, prior to integration using the Wöhler exponent (m). The resolution of the loaded results

will determine which calculations options are available in a subsequent LOAD RESPONSE calculation.

In third party mode, all calculation options and settings are disabled in the IEC checks since the result is loaded

from the xml files. Only abs with result plots and tables are shown in 3rd party mode. In addition, to close the

window of an opened IEC check after viewing the results is only possible by pressing “Cancel” – this is to indicate

that no calculations have been performed.

“Ambient Site results” mode

From windPRO 3.5, it is possible to load pre-calculated resource and site parameter files as basis for the main

IEC checks. The required format is called .siteres and may also be used in PARK calculations for AEP

assessments. For most checks the use of siteres overrides all calculation options as it defines the ambient

climate. However, for the turbulence checks the wake effects are calculated on top of the ambient climate and

wake related calculation options are available. See Appendix IX - Siteres ambient climate files (from

Resource, GASP, etc.) for further details of the siteres format.

Figure 25. Selection of siteres-file on the SITE COMPLIANCE main tab.

Curtailment

The Curtailment tab is available if “Apply Sector curtailment” is checked on the SITE COMPLIANCE main tab

(see Figure 26).

Figure 26. Selection of Curtailment on the main tab.

This tab simply provides an overview of the curtailment settings defined on each individual WTG, and the

possibility to change these settings. If curtailment settings on WTG objects are adjusted from within this tab, it

is important to be aware that this might influence other calculations which use the objects. The details of how

the curtailment setting influences the results is described in Appendix V - Curtailment. However, it is important

to understand that only results in the Effective turbulence check and in LOAD RESPONSE are affected by the

curtailment.

SITE COMPLIANCE 20

Figure 27. The Curtailment tab, with one curtailment line added for one WTG.

SITE COMPLIANCE 21

IEC Checks - Main checks

Subsections:

5.2.2.1 Terrain complexity

5.2.2.2 Extreme wind

5.2.2.3 Effective turbulence

5.2.2.4 Wind distribution

5.2.2.5 Flow inclination

5.2.2.6 Wind shear

5.2.2.7 Air density

Once all the setup steps have been completed and chosen flow models have been run, the actual IEC

calculations can be performed on the IEC checks tab. To start the calculation of the Main IEC checks, mark the

relevant checkboxes in the Include column. It is recommended to include all Main checks. If any steps in the

setup are incomplete or missing this is indicated with a Missing:…….. text in the Setup/Calculate of the affected

checks.

Some checks will be shown as “Missing” until connected calculations have been completed. Effective turbulence

always requires the Terrain complexity to be calculated first. If no WEng calculation result is available, the Flow

inclination also requires the Terrain complexity to be completed first.

Note also the Result legend as this is used extensively as a clear visual evaluation of all calculation results.

Figure 28. The result legend from the Calculations tab.

The overall wind farm evaluation is shown in the Result column when then calculation of a check is completed,

and it is determined by the evaluation of the worst WTG of the wind farm. Inside each check, the evaluation is

presented for each WTG.

Figure 29. The IEC checks tab with the Main checks expanded and checked, ready for calculation.

To start the calculation of each Main IEC check, click the Edit button. It is recommended first to calculate the

Terrain complexity check.

Note: A single or a few CRITICAL exceedances for a project may not always mean that a wind turbine

class can be fully excluded as suitable. CRITICAL issues should always be addressed with the relevant

turbine manufacturer. A turbine manufacturer knows the load margins and detailed aeroelastic model of

their WTGs and can perform detailed load calculations for the particular model to evaluate if the buffer

present for the OK checks or in the design can balance the identified CRITICAL issues.

SITE COMPLIANCE 22

Terrain complexity

Description and limit

The IEC standard describes, in quite elaborate detail, the Terrain complexity check. The check evaluates the

terrain steepness and variability in the vicinity of each WTG position. A number of circular and “pizza slice”

planes must be fitted to the terrain and the inclination and terrain variation from each plane must be evaluated.

These requirements are described in detail below.

The Terrain complexity check is not a binary “GO” / “NO GO” check that may kill a project in itself. Instead, the

check serves to identify complex sites / WTG positions, mainly to enable an appropriate correction referred to

as Turbulence structure correction to the measured turbulence. This correction is to compensate for the fact that

cup-anemometers only measure the horizontal component of turbulence, whereas, in complex terrain, a

significant fraction of the turbulent kinetic energy may be transferred to the vertical component of turbulence.

Setup, Calculation and Results

Setup and calculation of the IEC Terrain complexity check is very simple in SITE COMPLIANCE. The module

uses the active elevation model (i.e. the Digital Elevation Model) defined for the project. This model can be either

a contour file in a Line object, referred to as the “TIN”, or can be defined by an Elevation grid data object.

Figure 30. Setup and calculation of Terrain complexity check.

The only user option is the Grid resolution, which defines the resolution of a suitable quadratic grid of point

elevations extracted from the elevation model to be used in the required terrain fits. The IEC states that the

resolution may not be greater than the smaller of 100m or 1.5xHH. The default resolution automatically fulfils

this criterion. However, often the quality of the available elevation data does not grant the use of a finer resolution.

Click the green Calculate button to perform the Terrain complexity check. Calculation may take longer if the full

TIN has not already been calculated by the user. After successful calculation of the check, several new tabs

emerge presenting the results. The module jumps directly to the tab Results (Graphics) giving the main result

overview. On this tab, the so-called Terrain complexity index, as defined in the IEC standard, is illustrated for

each WTG position. A WTG position is complex if the index equals 1 and not complex if it equals 0. An index

between 0 and 1 indicates a partly complex WTG position.

Note also that, in the case shown, a green square has emerged beside the calculation name in the “calculation

list” to the upper right. This indicates that the result of this check is OK, and that no problems or risks are expected

relating to the Terrain complexity.

SITE COMPLIANCE 23

Figure 31. Results (Graphics) tab presents the Terrain complexity results as a simple overview.

SITE COMPLIANCE allows the user to Add new additional calculations with a different setup (in this case, grid

resolution). This is done by clicking the Add new button selecting the required new setup and clicking the green

Calculate button again. In this way, several supplementary calculations may be added to check how strongly the

results depend on the setup/assumptions. It is recommended to adjust the names of each calculation on its

Setup tab – this may also be done after calculation. The setup and result tabs of a calculation may be viewed by

clicking on that calculation in the calculation list. The shown calculation is highlighted in blue.

Before finalising the check by clicking the OK button, one of the performed calculations must be Selected as the

final result.

Figure 32. Left: The “calculation list” with multiple calculation setups and the upper one selected (marked) but

the lower one is viewed (blue highlight). Right: the “view legend” defining shown and selected calculations in the

calculation list.

The Results (Table) tab summarizes the results in terms of the Terrain complexity index for each WTG. A colour

code highlights complexity indices in green if they are 0 (i.e., OK) and in yellow (i.e., CAUTION) if an index

exceeds zero.

Expand the results of a WTG by clicking the “>”, this reveals the three sub-levels: “R=5xHH”, “R=10xHH” and

“R=20xHH”. Expanding one these levels reveal the results of the fit(s) with that radius in terms of the slope and

its direction, and the energy that is available in the directional span represented by that fit. Further details of how

these fits are performed are shown in the following.

SITE COMPLIANCE 24

Figure 33. Results (Table) presents the many details of the fits performed in the Terrain complexity check.

IEC requirement – Terrain complexity

For the Terrain complexity check, the IEC standard describes how a number of planes of varying size and shape

must be fitted to the terrain around each WTG position. The following text describes these requirements as

defined in [2]. In total 25 planes must be fitted with the following characteristics, all passing the WTG base

position:

• A circular plane with radius 5xHH centred on the WTG

• 12 “pizza-slice” fits with a radius of 10xHH, one fit in each sector

• 12 “pizza-slice” fits with a radius of 20xHH, one fit in each sector

The fits are illustrated for a contour file below showing the points used in some of the fits. For the “pizza-slice”

fits, all the points are shown, but only the points used within the northern sector are highlighted. The standard

requires that the grid resolution must not exceed the smaller of 100m or 1.5xHH.

Figure 34. Illustration of the fits required in the Terrain complexity check. Left: outline of the 5xHH omni-

directional fit and the included grid points. Middle: the grid points in a 10xHH sectorial fit, with points in the north

sector highlighted (5xHH circle shown as scale). Right: the points in a 20xHH sectorial fit with the northern sector

highlighted (5xHH circle shown as scale).

For each fit the slope is estimated. For the 5xHH fit, the slope is in the direction of the gradient (i.e. max slope);

for the 2x12 “pizza-slice” fits, (10 and 20xHH) the slope is estimated along the sector median. If the estimated

slope exceeds +/-10°, the fit is considered ”failed”.

In addition to the slope, a “check area” must be estimated for each of the 25 fitted planes. This check is the area

that has a vertical deviation from the fitted plane exceeding a certain threshold. The vertical thresholds for each

radius as well as the slope demand as defined in the standard are summarized in the table below:

SITE COMPLIANCE 25

Figure 35. Summary of the checks applied to each of the fits required in the Terrain complexity check [2].

The standard defines a Complexity Index, I

, which is calculated based on the outcome of the fits. For each fit,

the relative energy fraction is estimated based on the relevant sector-wise Weibull parameters. The omni-

directional fit (5xHH) represents 100% of the energy. If less than 5% of the energy is in the sectors which fail

either the slope or the vertical deviation check, the WTG position is regarded not complex, and the complexity

index I

=0. If more than 15% of the energy is in sectors which fail either of the checks, the WTG position is

regarded complex and I

=1. In the interval when 5%-15% energy is in sectors which fail, the complexity index is

interpolated linearly between 0 and 1.

If the complexity index is larger than 0, extra caution is required. The standard requires that a correction is

applied in the turbulence calculation. This is to compensate that cup anemometers only measure the horizontal

component of turbulence and not the vertical. In complex terrain, part of the turbulent kinetic energy is transferred

from the horizontal component to the vertical component and, hence, not accounted for by the measurements

made by a cup anemometer. The standard defines a Turbulence Structure Correction Parameter, C

, which

may be calculated from the complexity index. This correction parameter must be applied to the measured

turbulence when the complexity index is larger than zero. This information is automatically transferred to the

Effective turbulence check.

As a foot note, the standard mentions that the inclination angle of the plane from the omni-directional fit with a

radius of 5xHH may be used as the inflow angle estimate. We allow this option in SITE COMPLIANCE module

and the relevant inflow angles are automatically transferred from the Terrain complexity check to the Flow

inclination check.

Note: If both WAsP and WAsP-CFD have been run in the setup of SITE COMPLIANCE, the WAsP-CFD results

will always take preference to define sector energy fractions used in the terrain complexity check.

Extreme wind

Description and limit

The extreme wind speed check is one of the most important and potentially most critical checks in a SITE

COMPLIANCE calculation. The reasons are the very high uncertainty associated with most extreme wind

estimates and that the check represents an extreme load, which may not be compensated by a buffer in the

results of other checks relating to fatigue.

In the context of the IEC standard, extreme wind refers to a 10-minute averaged wind speed event with a

recurrence period of 50 years. Recurrence period (T) is a statistical term derived from the more stringent

statistical term “annual risk of exceedance” (R) via the simple relation . So, by definition, a 50-year wind

speed estimate has an annual risk of being exceeded of 2%.

The IEC standard’s design assumption for extreme wind is V

ref

defined for each wind speed class in Table 1 of

the standard (see Table 1 of this manual). So, e.g., a Class I WTG is designed for extreme wind speeds of up to

50m/s at standard air density of 1.225kg/m

Most methods of extreme wind estimation rely in one way or the other on the theory developed by E. Gumbel in

[3]. Appendix I describes the fundamentals of this theory and details of the theory relating to the models available

in this module.

Setup, Calculation and Results

The following text describes the workflow in setting up, calculating and reviewing the results of a typical Extreme

wind calculation. The setup of this check is split in three groups: Statistical model, Propagation model and

Fit radius

Count

Fit type

Max slope

Max vertical

deviation

Max allowed area

exceeding deviation

5xHH

Omni. (360°)

10°

0.3xHH

5xHH

10xHH

Sector (30°)

10°

0.6xHH

5xHH

20xHH

Sector (30°)

10°

1.2xHH

5xHH

SITE COMPLIANCE 26

Additional model settings. Details of the options within each of these groups are described in more detail after

this description of the workflow. Figure 36 below shows the Setup of the Extreme wind check. There are a

number of input parameters with abbreviated names for the different methods.

Figure 36. The Setup tab of the Extreme wind check. Note that the upper most option “Annual Maximum &

Gumbel” is greyed out and not available in the case shown because not all site masts hold ≥5 years data.

For POT-N & Gumbel;

• N is the number of extreme events to extract from the time series

• Δt is the minimum time separation between two extreme events to secure independence

For Weibull parent (EWTS/Bergström):

• N is the assumed number of independent 10-minute wind speed samples per year.

For Eurocode EN1991-1-4:

Input ‘base values’:

• Wind speed is the design wind speed at standard conditions from the relevant national annex to the

Eurocode (usually available from national standards authorities).

• Height and Roughness are the standard conditions for the base wind speed, typically 10m and 0.05m.

• Direction is needed for the base wind speed to select speed-up factors, if known or given in the national

annex this should be input directly.

• Auto sector let’s SITE COMPLIANCE calculate the most likely extreme wind sector based on the Weibull

parameters.

• Max sector takes the speed-up from the (worst) sector with the highest speed-up factor.

In Additional model settings:

• ρ is the air density at high wind speeds.

• K

is a normalized gust factor, where the default value 3.0 represents a 3 seconds gust.

Note that, for the default setup, no Additional model settings are activated. To include any of these additional

settings, they must always be activated manually with the appropriate input parameter. If any additional modelling

options are not selectable and greyed out, the reason is that a required input source is missing in the general

setup as e.g. a Long-term reference mast for the Index correct POT-N & Gumbel option.

Click the highlighted Calculate button to the lower right to run the Extreme wind calculation using the default

setup. When the calculation is complete, several new results tabs emerge:

The Extracted data (Table) tab shows a table of the extracted extreme samples from the mast time series.

SITE COMPLIANCE 27

Figure 37. The Extracted data (Table) tab.

On the Extracted data (Graphics) tab you see a plot of each of the mast time series with the extracted extreme

samples highlighted. Scroll through the masts (if more than one) using the Next and Previous buttons to the

lower left.

Figure 38. The Extracted data (Graphics) tab.

The Gumbel fit (Graphics) tab shows you the statistical fit for each WTG and the extrapolation to the required

return periods (i.e. risk levels), typically 50 years. Click next (lower left) to scan through the WTGs.

SITE COMPLIANCE 28

Figure 39. The Gumbel fit (Graphics) tab.

The Results (Table) tab gives you the final 50-year extreme wind speed result for each WTG as well as the IEC

limit for the WTG class (V

ref

for the relevant class). In the case shown, all WTGs are OK (green), and hence the

overall result for the park (upper right corner) is also green (OK).

Figure 40. The Results (Table) tab.

The Results (Graphics) tab gives a graphical overview of the results relative to the IEC limit. In the shown case,

the results are within the IEC limit.

SITE COMPLIANCE 29

Figure 41. The Results (Graphics) tab.

You may click the Add new button (lower right corner) to add another calculation setup for extreme wind using

a different method or different parameters for the same method. E.g., select Weibull parent (EWTS/Bergström)

and press Calculate. Now, you have two result options you can compare. If you have run a WEng calculation,

you may also try the Risø NCEP/NCAR extreme wind atlas which covers most of Europe and US. Each

calculation adds a line in the right pane. Click on a calculation in the pane to highlight it and view its setup and

results.

IMPORTANT: The results of the calculation which is marked with the will be used in the final evaluation.

Statistical model

The setup in this group defines the statistical model used to extract the extreme wind data and how this data is

fitted to the Gumbel statistical model, if required in the method. The different models are described in the

following text.

Figure 42. The Setup of the Statistical model in the Extreme wind check. Note that the uppermost option is only

available with at least 5 years of measurements for all site masts.

Annual maximum & Gumbel (requires ≥5 years for all site masts)

This method is only available if all masts have at least 5 years of data, because only the single most extreme

wind speed measurement is extracted from each year from the time series. With less than five years and hence

less than five data points, the linear fit to the Gumbel model becomes very vulnerable to the individual data

points. General recommendations suggest at least seven years of data for the annual maximum (AM) method

to obtain reliable results for this method. The AM method is the classical method proposed by Gumbel and is

still considered the most accurate, provided that sufficient data is available, which, unfortunately, is rarely the

case in wind energy project development.

Once the annual maxima data points are extracted they are processed according to the classical Gumbel method

of extremes, resulting in the so-called Gumbel plot. The fit of the Gumbel asymptote is performed using the

Probability Weighted Moments (PWM). (See [4] or Appendix I for further details of the theory).

POT-N & Gumbel

This method is available when at least one site mast with time series data was selected in the general setup.

The name abbreviates Peak-Over-Threshold and the N signifies the modification that the threshold is not defined

SITE COMPLIANCE 30

in terms of wind speed as is usually the case, but, instead, in terms of a fixed number (N) of samples. Often, this

method is also referred to as “Method of Independent Storms (MIS)” (see [5] and Appendix I) although the original

MIS method [5] uses a slightly different approach for data extraction.

There is no lower limit to the length of the time series needed for this method, but it is strongly recommended to

use an integer number of full years, i.e., at least one full year. N defines the total number of storms to be extracted

(default is 20) and Δt the minimum time separation required for storms to be independent (default is 4 days). The

default time separation is chosen to match the typical time scale of synoptic storm events.

Once extracted, the samples are processed according to Gumbels method with the additional step of correcting

for the annual storm rate (see [5] or Appendix I for further details). The model fit to the Gumbel asymptote is

performed based on the classical plotting positions (see [6]) and a traditional least squares fit with wind speed

as the independent variable.

Weibull parent (EWTS/Bergström)

The European Wind Turbine Standard (EWTS) [7] describes an extreme wind method based on the tail

characteristics of the Weibull mean wind speed distribution (parent distribution). A fundamental principle is that

sites which exhibit wind distributions with a low Weibull k-factor have heavier tails and, thus, a higher likelihood

of extreme events and, hence, higher extreme wind speeds.

This methodology relies on two main assumptions:

1) The wind distribution is a Weibull

2) The number of statistically independent wind samples per year is known (and equal for all sites)

The EWTS publication uses the number 23037 10-min independent events per year with a reference to a 1992

publication by Bergström [8]. However, there is typographical error in the EWTS publication and the number is

incorrect. In Bergström’s original paper [8], the correct number is 2302.

For the Weibull distribution shape parameter, we use the so-called “combined Weibull” [18] which results from

combining the sector-wise Weibull distributions, typically resulting from the WAsP calculation, to a resulting omni-

directional Weibull with the same first (mean wind speed) and third (energy) moments.

Risø NCEP/NCAR extreme wind atlas

The Risø NCEP/NCAR extreme wind atlas method uses a database of Regional Extreme Wind Climate (REWC)

files established for WEng 2 as part of a research project at Risø [9]. These atlases are based on the

NCEP/NCAR global reanalysis data with some ad hoc corrections established as part of the project to

compensate for the coarse temporal and spatial nature of the NCEP/NCAR data.

This method requires that a WEng calculation has been performed in SITE COMPLIANCE. The user may select

one of the nearest four REWC files. Each REWC file contains 12x30 reduced geostrophic wind speeds, one for

each of twelve direction sectors in each of thirty reference years. The wind speed samples are adjusted for each

WTG position using the predicted sector-wise flow corrections established by WEng to obtain the annual

maximum samples for each reference year. From this point, the method progresses identically to the Annual

Maximum method described above.

Eurocode EN1991-1-4

During the past decade or so, all national building codes within the European Community have been

standardized into the Euro Codes (EN). The Euro Code EN1991-1-4 treats wind loads and defines the guidelines

for handling of design extreme wind speeds. Each country has a national annex to EN1991-1-4 that describes

the Base wind speed to be used in different regions of the country as well as appropriate corrections at high

altitude and other exceptions or corrections. Thus, national codes like DIN (Germany), BS6399 (UK), DS

(Denmark), SS (Sweden), PN (Poland), SFS (Finland) are now referred to as e.g. DS-EN1991-1-4.

The input Base wind speed, height and roughness is to be found in the relevant national annex. The wind speed

is defined as a 50-year extreme wind speed at reduced geostrophic conditions. The typical standard conditions

are: 10m agl. (flat terrain) and a uniform roughness of z

= 0.05m (roughness class 2,).

A simple example is Denmark where the 50-year base wind speed is defined as 24m/s for all inland positions,

except in a 25km wide belt along the Danish west coast. In this belt, the base wind increases linearly from 24m/s

to 27m/s at the west coast. In countries with mountainous areas such as Poland or Germany, the national annex

SITE COMPLIANCE 31

specifies base wind speeds but also adjustment factors to these base wind speeds for altitudes above a certain

threshold.

In the calculation, the base wind speed is propagated from the standardized reduced geostrophic conditions

(10m, flat terrain and uniform roughness) to the specific terrain and roughness conditions of each WTG using

the flow results of WEng, WAsP or WAsP-CFD.

Several countries outside Europe have also adopted or are in the process of adopting the Euro Codes (or

selected parts of them) as design standards. At the time of writing (2012), these countries include Singapore,

South Africa, New Zeeland and Australia.

Propagation models

Once the extreme samples have been extracted from the measurement series, either using the Annual Maximum

method or the Peak-Over-Threshold method (POT-N), they are scaled to better represent the conditions at the

WTG positions. This scaling using a Propagation model may be based on the methods summarized in the figure

below.

Figure 43. Setup of the Propagation model in the Extreme wind check.

For the option WAsP-CFD / Flowres, an advanced scaling method is used which accounts for the potentially

strong turning (veer) of the wind from a mast to the WTG in complex terrain. For WEng and WAsP, the scaling

is performed using the sector-wise mast-to-WTG speed-up factors predicted in the WEng and WAsP calculations

respectively. In the Shear option, the measured sector-wise shear is used to calculate the sector-wise speed-up

factors from mast height to WTG hub height. For the choice No model, no scaling is performed of the measured

extremes.

The Gumbel fitting is always performed after the chosen scaling is applied to the extracted extreme samples.

Additional model settings

The Extreme wind check offers several Additional model settings as shown in the figure below. These are

described in the following.

Figure 44. Setup of Additional model settings for the Extreme wind check.

Index correct POT-N & Gumbel

This long-term index correction method for extreme wind estimates is only available for the POT-N method. It is

based on work published at the EWEA conference 2010 [10]. The fundamental basis of this method is the

observation that two main sources of error dominate in extreme wind estimation:

1) Too short a time-series (resulting in a statistical uncertainty)

2) Bias introduced by the method of the Gumbel model fit.

SITE COMPLIANCE 32

In addition to this, we know that the Annual maximum method with the PWM fit is virtually free of error source

two: bias from the Gumbel fit [11].

The idea of this index correction is based on calculating an extreme wind index for the period of the site

measurements by using an overlapping long-term reference series which is representative in regards to the

storm events. The index is estimated by applying the same POT-N setup to the concurrent part of the reference

series and the Annual maximum (AM) method and PWM fit to the full length of the reference series. The index

is then defined as the ratio of the concurrent POT-N estimate to the full AM estimate for the reference. The

appropriate correction, which is applied to the extreme wind estimate using the site mast, equals the inverse of

the extreme wind index.

As a validation, an index is calculated for each year of the site series and for the reference series. This allows a

visual validation to check if the extreme wind climate is properly accounted for by the reference series. If this is

the case, the index curves will have the same shape.

Note: The Extreme wind index correction should be used with great caution if only one year of site/reference

overlap is available. This prevents evaluation of the mast and reference index trends that can be prevalent in

such short periods, e.g., if an evaluation based on several years overlap reveals very different trends than using

only one year.

Air density at high wind speed

The aerodynamic force is proportional to the square of the wind speed and to the air density. Thus, a decrease

of air density results in a decreased thrust. The IEC limit for extreme wind is defined at standard air density of

1.225kg/m

. Extreme wind estimates at other air densities may be corrected to standard air density assuming

that the thrust force is unchanged but that the site air density is replaced by the standard value.

The air density inserted by the user should represent the air density expected at high speeds. Often the expected

mean air density is used as an approximation.

Include 3s gust estimate

Gust values may be estimated from measurements using the max of each 10-minute interval which is often

logged. However, the averaging period of such estimates is unknown. The IEC standard requires 3-second

averages to be used for gust estimates. Instead of maximum 10-minute 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 a normalized peak factor equal to 3.0 for t=3s according to Cook (1990) in [12], TI is the expected

turbulence intensity (10-minute averaging) which is calculated as the mean turbulence intensity for the extracted

extreme samples. No adjustment is made to the turbulence; hence the gust factor estimated at each mast is

used directly for the relevant WTG positions.

k-factor pre-conditioning

The Gumbel model is an asymptotic model - it assumes that the number of independent events per year is

infinite. In [5], Cook demonstrates that the error due to this assumption depends on the Weibull k-factor of the

wind speed distribution. If the k-factor is one (also called the exponential distribution), the error is zero no matter

the real sample size. For k larger than 1, the error increases with k. The error introduced by this assumption

makes the Gumbel plot of the extracted extreme samples curve slightly down. The curvature introduces an

overestimation (conservatism) in the Gumbel fitting.

To reduce this error, the wind speeds may be transformed prior to Gumbel fitting to turn the distribution into a

Weibull with a k-factor of one. This is done by raising the wind speeds to the power of k. After the fit has been

performed and extrapolated to obtain the 50-year estimate, the wind speeds are transformed back taking the k

root. Often a factor of two is used as a general estimate of the k-factor. The default value in this setup is the

mean “combined” k-factor for WTGs [18]. The typical effect of the k-factor pre-conditioning is to reduce the

extreme wind estimates by 5-10%.

SITE COMPLIANCE 33

Thus, the use of k-factor preconditioning rests on a solid statistical argument. See the Extreme wind appendix

for further details.

Safety factor correction for COV>0.15

In the ed. 4 of the IEC standard, it was identified that further safety is needed for particular distribution parameters

of the extreme wind Gumbel distribution, in the case when the coefficient of variation of the distribution exceeds

0.15. See further details in Appendix VII.

Effective turbulence

Description and design limit

The Effective turbulence check is together with the Extreme wind check - one of the most important IEC checks.

Where Extreme wind represents the extreme loads, the Effective turbulence mainly represents the fatigue loads,

a more long-term degradation of structural integrity of the turbine. Calculation of the Effective turbulence is

described in a revised version in the 2010 Amendment to IEC61400-1 ed. 3 [2]. The Effective turbulence model

is based on the publication [13] by the late Sten Frandsen from Risø/DTU, hence the model is also known as

the “Frandsen model”.

The design limit for Effective turbulence is called the “Normal turbulence model” in the IEC standard and is

calculated from the basic parameter I

ref

in Table 1 (see section 5.1.1). I

ref

has the values 0.12, 0.14 and 0.16 for

the standard turbulence classes A, B and C, respectively. The calculated Effective turbulence is based on the

90th percentile of measured ambient turbulence and must be compared against the Normal turbulence model

(the design limit) for a range of wind speeds. When the power curve is known, the range is from 60% of the wind

speed at rated power to the cut out wind speed. In SITE COMPLIANCE, a WTG object, and hence a power

curve, is always defined prior to calculating Effective turbulence. Appendix II describes further theoretical details

of the Effective turbulence model.

Effective turbulence is calculated as a function of wind speed only. This is done by integrating the directional

variation of turbulence over all directions for each wind speed bin. However, effective turbulence is NOT a

measurable quantity as it combines the directional contributions with a special weighting that accounts for

material fatigue via use of the material parameter, the Wöhler exponent, hence the name Effective turbulence.

Prior to the integration over all directions for each wind speed, the estimated wake added contribution is

combined with the 90th percentile of the ambient turbulence at each WTG. The normal turbulence model is

illustrated below for each of the three turbulence classes.

Figure 45. Illustration of the IEC “Normal turbulence model” which is the IEC design limit for the Effective

turbulence check. Left: plotted as standard deviation of wind speed. Right: plotted as turbulence intensity.

Setup, Calculation and Results

Note:

Please be aware that the Terrain complexity check must always be completed before the Effective

turbulence calculation can be initiated. The reason is that the results of the Terrain complexity check are

used in the Effective turbulence calculation via a correction factor called turbulence structure correction

factor which is required by the IEC standard.

0,1

0,2

0,3

0,4

0,5

0 10 20 30

Normal turbulence

model,

/u [-]

Wind speed [m/s]

Class A

Class B

Class C

0 10 20 30

Normal turbulence

model,

[m/s]

Wind speed [m/s]

Class A

Class B

Class C

SITE COMPLIANCE 34

The following text describes the workflow in setting up, calculating and reviewing the results of a typical effective

turbulence calculation. The setup of this check is split in five groups: Turbulence data, Propagation model,

Turbulence structure correction, Frandsen model and Sector management. Details of the options within each of

these groups are described after this description of the workflow and various tabs. The figure below shows the

Setup tab of the Effective turbulence check.

Figure 46. Setup tab of the Effective turbulence check.

Click the highlighted Calculate button to the lower right to run the Effective turbulence calculation using the

default setup. When calculation is done, several new results tabs emerge, which are described in the following.

When measured turbulence data from a Site mast is used in the calculation, this data must be analysed. A model

must be fitted to the mean turbulence and to its standard deviation to fill in gaps in the data and to extrapolate

them to the required wind speed bins without data. The Data fit (Table) summarizes the omnidirectional and

sector-wise outcome of the procedure in a tabular form.

Note that at the bottom of the tab it is possible to toggle between viewing the table as standard deviation in m/s

and as turbulence intensity. The standard deviation is recommended throughout as this is the measured quantity,

and because the fundamental assumption in the Frandsen model is that loads are proportinal to the standard

deviation. Using turbulence intensity tends to move focus to low wind speeds where loads are not as significant.

Viewing turbulence data as TI also tends to overemphasize small insignificant deviations of the fits at low wind

speeds.

SITE COMPLIANCE 35

Figure 47. Data fit (Table) tab of the Effective turbulence check. Note that only when WAsP-CFD is the

propagation model the fit is also shown for each WTG in addition to each mast.

The Data fit (Graphics) presents the turbulence measurements and fits them into a graphical form. Use the

buttons in the lower left corner to change to another mast (if available) or to skip through the sectors. It is worth

noting that in the plots the…:

- asterisk illustrates the measurements

- open circles show the chosen value

- red lines are the fitted data model

If the asterisk and the open circles coincide, it means that the measurements in the bin are accepted as the bin

fullfills the selection criteria (minimum number of samples in each bin) defined on the Setup tab. If the asterisks

and open circles deviate for a bin, the open circle will always fall on the redline, indicating that the fitted value

was selected because the bin failed the selection criteria.

Note that for the default setup, the standard deviation of turbulence (σ

) is the same for all sectors. This is due

to the fact that this quantity (the standard deviation of the standard deviation) is ususally not well determined

sector-wise with only one year of data. The IEC standard allows this use the weighted mean σ

instead of sector-

wise values to stabilize the estimation.This is introduced in a footnote in [2].

If the data selection or the fits are not satisfactory, press the Add button to the lower right to try another setup

where the data selection criteria have been changed. This setup is described in more detail below.

SITE COMPLIANCE 36

Figure 48. Data fit (Graphics) tab of the Effective turbulence check. Note the three highest samples on the left

plot. These are measurements with too few samples in their bin to be selected for use. Thus, the asterisks

(measurements) do not coincide with the circles (selected value) for those four bins. Instead those circles fall on

the red line (the fit) as the fit is used for bins with too few samples. Sample thresholds are defined in the setup.

Results of the Effective turbulence calculation are summarized on the tab Results (Table) showing the results

for each wind speed bin within the required check interval by the IEC standard. This check interval is described

above (in Description and limit) and is relative to the wind speed at rated power, which is estimated from the

power curve defined for each particular WTG object in the calculation.

Any bins exceeding the IEC limit are highlighted in the table, so that a WTG with no highlighting is fully within

the IEC limit for all required wind speeds. Yellow highlighting is used for Caution, i.e. when the exceedances for

the particular WTG are not considered critical and, thus, are compensated by the buffer at other wind speeds.

Red highlighting is used when overall the exceedances for the WTG are expected to accumulate to a critical

exceedance.

For each WTG, the first line in the table shows the result of the Effective turbulence calculation for each wind

speed bin. The second line shows the relevant IEC limit for the same bins. In the column Equivalent, the

equivalent accumulated effective turbulence normalized to the IEC design climate is shown. Thus, the value of

Equivalent is always one for the row IEC demand (second line). The threshold between Caution (yellow) and

Critical (red) is reached at value of Equivalent exceeding 1. Third line shows the omnidirectional mean

turbulence, and the lines below summarize the sector-wise part results and scaling factors used for each WTG.

SITE COMPLIANCE 37

Figure 49. Results (Table) tab of the Effective turbulence check. Note that the Effective turbulence is exceeded

for some WTGs at some wind speeds in the table. Those bins are highlighted in yellow as the exceedance is not

expected to be critical.

The tab Results (Graphics) illustrates the Effective turbulence results as a function of wind speed for each WTG.

Use the buttons to the lower left to shift to next WTG. Any exceedance within the relevant IEC check interval is

highlighted by light red shading of the wind speed bin.

Note that also the Effective ambient turbulence is shown as the dashed curve, which illustrates the result of the

Effective turbulence calculation disregarding all wake-added turbulence contributions. The difference between

the full and the dashed curve clearly illustrates the wake contribution of the wind farm to the effective turbulence.

The difference is usually highest up to 8-12 m/s, because at higher wind speeds, the thrust coefficient decreases

quickly for most WTGs and lower thrust results in less wake turbulence.

SITE COMPLIANCE 38

Figure 50. Results (Graphics) tab of the Effective turbulence check.

The rightmost tab Detailed results (Graphics) illustrates all the contributions and part results in the calculation of

effective turbulence as a function of wind speed. Each contribution may be de-activated in the menu below the

graph. The wind speed is adjusted using the slide bar. To scroll through the graphs for the WTGs use the

Previous/Next buttons to the lower left. If the box All is ticked, all WTGs are plotted together with their relative

positions. Use the mouse-over-effect to see the label of each WTG. If all contributions are unchecked except

Wake turbulence, this graph is particularly useful to identifying which neighbour WTG is the cause of a too high

Effective turbulence level.

The graphics on this tab are also very useful to help better understand the calculation of Frandsen’s Effective

turbulence but also to analyse and help understand the cause of turbulence-related problems of a wind farm

design. Is it the ambient turbulence or is it the wake-added turbulence that is causing the problem? And from

which sector/directions are the problems caused?

Figure 51. Detailed results (Graph) tab of the Effective turbulence check.

The following sub sections describe the setup options of the Effective turbulence check in more detail.

Turbulence data (ambient)

The starting point of an Effective turbulence calculation is the ambient turbulence on the site.

Figure 52. Setup of ambient turbulence data input and handling.

Ambient turbulence from mast measurements

When the ambient turbulence is measured, it must be treated properly to ensure a robust outcome of the

calculation as real data always have outliers, holes and other problems. This is handled by applying selection

SITE COMPLIANCE 39

criteria to the data and by fitting a model to the ambient turbulence measurements and using this for problematic

or missing bins.

First setup choice for the mean turbulence (Mean σ) is the selection of Sector-wise or Weighted mean. The first

option is default and highly recommended. The latter choice is only to be used experimentally as a very robust

estimate for very problematic data, e.g., with a very short time span.

Second choice is the number of samples (N) in a bin for the bin-average turbulence to be used. Bins with less

than N samples will be disregarded; they will not contribute to the fit, and the turbulence of those bins will be

replaced by the fit. The default value of N has been chosen by testing the methodology on a number of site

masts. The fit is performed as a standard least squares fit to the bin values with equal weights. This disregards

the fact that there are many more samples in bins at low wind speeds compared to bins at higher wind speeds,

thus, effectively increasing the relative weight of bins at higher wind speeds where loads are more critical. If a fit

weighting each bin by the number of samples is used, any upward curvature of the σ(u) curve at higher wind

speeds, as in forests and off/near shore, tends to be ignored, which can lead to falsely low estimates of

turbulence at high wind speeds as illustrated in the figure below.

Figure 53. Plot showing the difference between a weighted fit and the fit employed in SITE COMPLIANCE. Note

how the weighted fit severely underestimates values of accepted bins between 15 and 20 m/s.

Next choice is activating the Use fit for all bins option. If this is activated the fit is used for all bins no matter the

number of samples. This is only recommended as a robust approach if the data seems very noisy and oscillatory.

The last choice is whether or not Charnock effect should be included by adding a second order term to the

turbulence fit. This option should only be activated for offshore sites where the surface roughness increases with

wind speed due to the build-up of the wave height. If Charnock effect is included Site Compliance has a built-in

extreme wind speed limit of 35m/s for which it is assumed that waves start breaking and the drag coefficient

saturates (i.e. the second order growth of turbulence ceases).

Setup for handling of measured standard deviation of turbulence (i.e., σ

) has the same three choices as for the

mean turbulence (σ), but the defaults are different. By default the weighted mean is used for σ

as this is allowed

in the standard [2] and ensures a much more stable estimation of this otherwise quite unstable parameter. In

addition, a higher default bin acceptance limit of 50 samples is used, as well as using the fit for all bins. These

defaults ensure a quite robust estimation of σ

. However, our analyses have shown that still, for some sites with

only a year of data, an even more robust estimation is required. Therefore, an extra fit criterion is included which

controls the type fitting with the options Auto, Linear or Robust. The default setting is Auto. In this mode, an

ordinary linear fit is used if the R

of the fit is above 0.8. When R

is below 0.8, a very robust estimator is used,

namely the median sample which is very insensitive to outliers. The user may also force the fit to always be a

Linear fit or a Robust fit (i.e. median) no matter the value of R

using the last to options.

0 5 10 15 20 25 30

0.5

1.5

2.5

u [m/s]



(u) [m/s]

data

fit

weighted fit

Selected

SITE COMPLIANCE 40

Ambient turbulence from model (no mast data)

If no measurements of turbulence are available the turbulence must be predicted using a model, which means

either WEng or WAsP-CFD or a pre-run flow result in flowres format. In addition, the flow models only predict

the mechanically generated contribution to the mean turbulence intensity and not the thermal contribution or the

standard deviation of the turbulence intensity. There is just one sub-option ‘COV’, with default value COV=0.3,

which means that the turbulence intensity from flow model will be used as the mean turbulence, whereas the

standard deviation is assumed to equal 30% of the mean, i.e., that the mean has a coefficient of variation (COV)

of 0.3. This assumption is chosen to match observations and includes a deliberately slight conservative bias on

the 90 percent quantiles of turbulence on average for many masts in simple terrain

. A COV of 0.2 yields on

average an unbiased result, but has a variation which might lead to underestimation for some sites, which is

effectively compensated by using COV=0.3 according to the results in

Figure 54. Comparison of COV=0.2 assumption with the assumptions in the IEC standards.

Propagation model

Turbulence measured on a mast only represents the mast position and measuring height. Conditions might be

significantly different at other heights and at other locations across a site. To compensate for this SITE

COMPLIANCE allows a number of scaling options to adjust measured turbulence to become more

representative for each WTG position. This scaling is setup in the Propagation model group which also allows

the option No scaling, which is recommended only for sites where the mast is fully representative for WTG

conditions.

Figure 55. Setup of propagation model, which defines the scaling method of measured turbulence from mast

position and height to each WTG position and hub height.

The scaling can be based on predicted turbulence and speed-up from WAsP-CFD / Flowres or WEng or on just

sector-wise speed-up factors predicted by WAsP. Additionally, the actual scaling may be done using three

different basic assumptions, abbreviated as:

1) Asymptotic

2) Constant σ-error

3) Uniform

For WAsP-CFD / Flowres, the scaling performed for all individual samples so that the complete time series of

wind speed and turbulence is scaled out to each WTG position. The tabulation/binning and fitting is then

performed for each WTG. This time series approach allows the scaling to properly account for potentially very

strong turning (veer) of the wind from mast to WTG, which is likely in complex terrain. With the WAsP-CFD time

For details see: http://help.emd.dk/knowledgebase/content/WindEuropeSummit2016_Paper_L_Svenningsen_annex.pdf

SITE COMPLIANCE 41

series approach only the scaling option Uniform is possible. Scaling based on WEng turbulence is applied to the

binned turbulence table at the mast and allows all three scaling assumptions, which are described in the

following.

Asymptotic scaling using WEng turbulence

This method relies on two assumptions:

a) That WEng, as a neutral model, captures the variation of mechanically generated turbulence well. Hence,

WEng turbulence ratios must approach the ratios of site turbulence in the high wind speed limit.



























 























for 



 

b) That thermally-generated turbulence dominates at wind speeds approaching zero and is a general feature

across the site (i.e. a constant).



























  for 



 

Where index t indicates “true” which for the mast is the measured turbulence and for the WTG the quantity we

want to estimate. Index p abbreviates “predicted” and represents WEng predictions.

The resulting scaling factor is a function of direction sector and wind speed.

Constant σ-error scaling using WEng

In this method it is assumed that the error of the WEng predicted turbulence (σ) is constant across the site,

related to, e.g., thermal or larger scale turbulence contributions not captured by the WEng micro-scale model.













 























 











Speed-up across the site is accounted for via the predicted WEng mast-to-WTG speed-up factor









(in the above equation, u is used for u

WTG

). Indices on sigmas are used as in the previous scaling

method and the resulting scaling factor, in this case, is also a function of direction sector and wind speed.

Uniform scaling using WEng

The uniform scaling relies on the very simplifying assumption that the “true” ratio of mast-to-WTG turbulence

equals the ratio predicted by WEng, which is constant for all wind speeds within a sector. The scaling factor

depends on the sector.









 











 











 











 



No speed-up is included and the scaling factor does not vary with wind speed. This scaling assumption is coarser

than the other WEng based methods, but on the other hand the simplicity of the underlying assumptions also

makes this method quite robust.

Constant σ-error scaling using WAsP

This scaling approach is related to the Constant σ-error scaling using WEng (see above) but, in a simplified form

as no predicted turbulence values are available. If the predicted turbulence values are set to zero in the

equivalent WEng equation, the corresponding WAsP version is obtained.









 











 



The fundamental assumption is that, for a certain flow condition (wind speed and direction) at the mast, the

turbulence in terms of σ is constant across the site. But, as wind speed varies across the site due to roughness

and terrain speed-ups, the turbulence intensity will vary.

Uniform scaling using WAsP

The uniform scaling based on WAsP relies on a similar simple assumption as the uniform WEng scaling is based

on. The scale factor is assumed constant for all wind speeds and simply equal to the inverse of the mast-to-

WTG speed-up factor predicted by WAsP.









 











 



SITE COMPLIANCE 42

No scaling

This last option is to assume that the mast measurements are representative for all WTG positions, implying no

scaling at all. The underlying assumption is rather trivial.

























This assumption is only recommended for simple and small sites with little variation in terrain or from mast to

WTG or if the scaling options, for some reason, perform poorly. For offshore sites this setting may also be most

appropriate depending on the mast setup.

Turbulence structure correction

The IEC standard specifies that if a WTG position is estimated to be complex or partly complex in the terrain

complexity check, i.e., the complexity index is I

>0, then a correction should be applied to turbulence values.

The argument is that in complex terrain turbulent kinetic energy is transferred from the horizontal component to

the vertical component. Cup anemometers measure only the horizontal component. To compensate for the

transferred turbulence, the IEC standard introduces a correction factor C

, the Turbulence structure correction

parameter. This correction has its own setup group with several options as shown below.

Figure 56. Setup of the Turbulence structure correction requirement of the IEC standard.

The setup allows the following three options.

Complexity check

The IEC standard describes how C

may be calculated from the terrain complexity index (I

) for each WTG

position using the following equation.





  



The IEC standard [2] states that this equation is to be used when no measurements or modelling is available for

the three components of turbulence.

So far, experience from several complex sites has shown that this Complexity check option is quite conservative

when compared to modelled components of turbulence. This is the case even when the modelling of these

components is using a linearized model, such as WEng, expected to over-estimate inflow angles and, hence,

over-estimate transfer of horizontal turbulent kinetic to the vertical/transverse component.

WEng turbulence components

When measurements or modelling of the three components of turbulence are available, the IEC standard states

that the following equation for the turbulence structure correction parameter (C

) may be used.









  















 

















Where σ

, σ

and σ

are the three components of turbulence (longitudinal, vertical and transverse). These three

components are predicted for each WTG position in a WAsP Engineering calculation and are used when the

setting WEng turbulence components is chosen. This setting is default when a WEng calculation is available.

The current release of SITE COMPLIANCE does not support the use of three-component measurements.

No correction

This option should only be selected if terrain complexity indices for all WTGs equal zero, i.e., for not complex

sites. The only exception is when ambient turbulence is predicted using WAsP Engineering. In this case, No

correction is the default setting, because WEng models the longitudinal component and not the horizontal

component as measured by a cup anemometer. This decision is in-line with the comparable use of WEng

calculation results in Risø/DTU’s own software tools e.g. WAT.

SITE COMPLIANCE 43

Frandsen model

The Frandsen Effective turbulence model plays a central role in the IEC context and the model is described in

detail in Annex D of the Standard [1, 2]. The theoretical details of the model are summarized in Appendix II of

this manual.

In very brief terms, the Frandsen model consists of several sub components. The first component is a model

that predicts the wake-generated turbulence downstream of a WTG at a certain wind speed and for a given

thrust coefficient. The second component is a model for combining wake-generated turbulence with ambient

turbulence into a “total” turbulence for each direction and each wind speed bin. The third component is a special

integral over all directions that accounts for frequency of each direction but also includes a non-linear weighting

relating to the accumulation of material fatigue in different materials. Hence, a material parameter called the

Wöhler exponent is needed as input. The integration is performed for each wind speed bin and the resulting

integral is the Effective turbulence as a function of wind speed.

Figure 57. Setup of the Frandsen model also referred to as the Effective turbulence model.

Wöhler exponent

The general material parameter which is used in weighting according to fatigue accumulation, as described

above for the Frandsen Effective turbulence model. Usually, the value m=10 is assumed as it represents glass

fibre and thus the WTG blades. A m value of 3-5 represents welded steel and e.g. the tower. Generally, using a

high value e.g. m=10 will be a conservative assumption for materials with a lower Wöhler exponent.

Wake width

From Frandsens original publication [13] uses so-called ‘view angles’ to quantify the directions with wake effects.

In the proposed model for the standard Frandsen used constant view angles of 21.6° independent of the spacing

between turbines - this is the default option in SITE COMPLIANCE. However, Frandsen also proposes an

alternative view angle estimate ‘Variable’, which depends on spacing, x, in RD via: 



 

 

 .

Between typical spacings of 4 and 5 RD spacing the two options produce very similar results. For spacings less

than 3 RD the variable option results in larger view angles (wake widths), and thus in higher loads, which is likely

more realistic.

Large wind farm correction

A fourth and additional component in the Frandsen model is a large wind farm correction which accounts for

extra added turbulence levels inside very large wind farms. The original specification of the correction assumes

a regular layout not generally applicable to the real world, but this has been generalized in [15]. Our

implementation is in line with this generalization; It assess the need for large wind farm correction for each sector

and adjusts ambient turbulence levels as requested in the standard. Only the number of neighbour WTGs in the

sector decides if a sector is large wind farm or not.

Sector management

There are two options for sector management: the 1) Simple sector curtailment and 2) Advanced sector

curtailment. Option one is described below and can be selected in the calculation without any prerequisites.

Option two, described in section 5.2.1.11 Curtailment, requires curtailment rules to be defined directly on the

relevant WTG objects and that Curtailment has been selected on the SITE COMPLIANCE main tab.

The simple curtailment option allows definition of a distance threshold in rotor diameters (RD), e.g., 3 RD. In the

calculation, all wake effects from neighbour WTGs within the set distance for a particular direction will be

excluded, assuming, therefore, that those WTGs have been shut down.

This implementation makes it very easy to assess the effect of most manufacturers’ standard requirement of

shutting WTGs down in directions where distances are below 3 RD.

SITE COMPLIANCE 44

Figure 58. Selection of sector management. Note that Advanced sector curtailment is not selectable, as this

option is selected on the SITE COMPLIANCE main tab.

Wind distribution

Description and limit

The Wind distribution check evaluates the frequency of occurrence at different wind speeds for each WTG by

comparing them to the frequency of occurrence assumed in the IEC design limit.

The IEC design limit for the Wind distribution check is a Weibull distribution with a shape factor of k=2. The mean

wind speed is defined as 20% of the basic design parameter V

ref

from Table 1 (see section 5.1.1) which is 10m/s,

8.5m/s and 7.5m/s for the wind speed classes I, II and III, respectively. A range of wind speeds ranging from

20% to 40% of V

ref

must be checked, i.e. from the mean wind speed to twice the mean wind speed of each WTG

class.

In the IEC standard, it is required that the wind distribution estimated for each WTG is long-term representative.

Hence, an evaluation of the long-term level and possibly a long-term correction is required. This may be handled

in three ways in SITE COMPLIANCE.

1) The first and typical way is by basing the WAsP calculation on a long-term corrected wind statistics saved

from windPRO’s MCP module by choosing this WAsP-option in the Main setup tab of SITE COMPLIANCE.

2) The second option is to use the mast data directly (either in a WAsP calculation or without) and utilize the

simple (MCP supplement) long-term index correction for wind speeds available in SITE COMPLIANCE (see

section 5.2.4).

3) The third option is that the data are evaluated to be long-term representative in their own right, and no long-

term correction is applied. The last option is not uncommon with around three or more years of site data.

Setup, Calculation and Results

The following text describes the workflow in setting up, calculating and reviewing the results of a typical Wind

distribution calculation. The setup of this check is quite simple with few options as illustrated in the figure below

showing the Setup tab.

Figure 59. Setup tab of the Wind distribution check.

SITE COMPLIANCE 45

In most cases, a WAsP calculation has been run and using these results predicted at each WTG will then be the

default option. In the shown case, a WAsP-CFD calculation has also been performed which then takes

preference as default.

The Results (Table) summarizes the results for all relevant wind speed bins for each WTG. Bins are highlighted

if the frequency within that bin exceeds the IEC limit. If the overall exceedance for the WTG is evaluated to be

critical the highlighting is red or if it is only evaluated to be Caution highlighting is yellow.

First row for each WTG shows the frequencies of occurrence in each bin predicted for the WTG. Second row

shows the IEC limit, i.e., the frequencies of the relevant IEC class. If the sub-levels are expanded, the sector-

wise frequencies for the predicted WTG climate are presented.

Figure 60. Results (Table) tab of the Wind distribution check.

On the Results (Graphics) tab the WTG probability density function (frequencies of occurrence in each wind

speed bin) is graphically compared to that for the IEC class. The IEC wind speed check interval is highlighted

by the blue vertical lines, and any exceedance is highlighted by a light red box covering the exceeded interval.

Results for Next or Previous WTG and are viewed using the WTG selection buttons to the lower left.

Figure 61. Results (Graphics) tab for the Wind distribution check.

Calculation options

There are only a few calculation options for the Wind distribution check with the most common being the WAsP-

CFD sector Weibulls or WAsP sector Weibulls. The two other options should only be used where the mast

position is truly representative for the WTG positions.

It is well known that the WAsP model has limitations in very steep or complex terrain where it may over-estimate

the terrain speed-up factor considerably. If the mast position is representative of the WTGs, this should not pose

a major problem.

SITE COMPLIANCE 46

WAsP-CFD sector Weibulls

With this setup, the sector-wise predicted Weibull distributions from WAsP-CFD are used. The so-called

Emergent wind speed distribution (WAsP nomenclature) is used in the actual IEC check. It is simply the sector-

wise Weibull distributions weighted by their sector frequency and summed to form the omni-directional

distribution of wind speeds. The resulting distribution is generally not a Weibull itself and hence may exhibit bi-

or multi-modality (i.e., multiple maxima). Finally, the Emergent distribution is compared against the IEC design

limit for the wind speed bins required by the IEC standard.

Flowres sector Weibulls

With this setup, the sector-wise Weibulls distribution is calculated by transferring the measured time series from

each mast and main height to the relevant WTG positions at hub height using the Flowres flow simulation results.

Once transferred, the time series are binned and sectorwise Weibull distributions are fitted. As the Flowres

results typically represent neutral stability, it is recommended to use this option with measurements close to hub

height. If measurements are much lower than hub height it is recommended first to use the shear extrapolation

option “Add synthesized height” in the meteo object Data/Data setup tab to establish a time series at or close to

hub height at the mast position.

WAsP sector Weibulls

As described above but using the standard WAsP flow model.

Mast sector Weibulls and sector shears (mast shear required)

Basing the vertical extrapolation on shear measured on the mast relies on the assumption that wind shear is

constant with height, which is certainly not the case in areas with strong terrain speed-up effects. It also relies

on the assumption that horizontal variations of the wind distribution are negligible.

If no WAsP calculation/license is available, this calculation option may be used if the mast position is considered

to be representative of the WTG positions. The calculation will use the fitted sector-wise Weibull parameters for

the main height of each mast. The Weibull A parameter for each sector is then scaled from mast main height to

WTG hub height using the corresponding sector shear exponents estimated for the mast shear heights. Finally,

the Emergent distribution is formed by summing the scaled sector Weibull parameters, weighting them by the

sector frequencies.

Mast Weibulls directly

In this option, the Emergent distribution is calculated directly from the mast sector-wise Weibull parameters. This

is only advisable if the mast is representative for the WTG positions, both in terms horizontal variations and in

terms elevation above ground (i.e. mast main height ≈ hub height).

Flow inclination

The Flow inclination check predicts the flow inclination in each sector and identifies the sector with the highest

absolute (positive or negative) flow inclination for each WTG. The resulting flow inclinations are compared

against the IEC design limit for each WTG.

Description and limit

The IEC design limit for the Flow inclination check are flow inclinations of +8° and -8° for the worst direction.

Thus, flow inclinations above +8° or below -8° exceed this limit.

The IEC standard mentions that the slope the smallest of the terrain fits (5xHH) required in the Terrain complexity

check may be used as an estimate of the flow inclination. We allow this option in SITE COMPLIANCE, and note

that, generally, the resulting estimates are not too different from more advanced estimates obtained using flow

models.

Setup, Calculation and Results

The following text describes the workflow in setting up, calculating and reviewing the results of a typical Flow

inclination calculation. Setup of this check is quite simple with only two options, as illustrated in the figure below

showing the Setup tab.

SITE COMPLIANCE 47

Figure 62. Setup tab of the Flow Inclination check.

In cases where WAsP-CFD flow results are available and has been run, the WAsP-CFD flow inclination estimate

for each WTG is the default option. Alternatively, WEng flow results or the terrain complexity check may be used.

The Results (Table) summarizes the result, i.e., the maximum flow inclination for each WTG and the direction

(sector median) in which it occurs. If the table is expanded, the flow inclination for each model sector is shown.

The expansion with sector-values is not available for the calculation Terrain fit (5xHH disc), from complexity

check, as sector-wise results are not available for this calculation.

Figure 63. Results (Table) tab of the Flow inclination check.

On the Results (Graphics) tab the results (i.e., max flow inclination) are summarized for all WTGs. IEC design

limits at ±8° are shown as the horizontal blue lines.

SITE COMPLIANCE 48

Figure 64. Results (Graphics) tab for the Flow inclination check.

Calculation Options

The following describes the calculation options for the Flow inclination check.

WAsP-CFD / Flowres, flow model

Using this option, flow inclination is extracted directly from the raw WAsP-CFD flow results, which is always run

for 36 model directions. WAsP-CFD is the most accurate method for complex terrain sites.

WEng, flow model

WEng is a broadly accepted and used flow model suited for engineering calculations like flow inclination. The

advanced setup of the WEng calculation determines the number of sectors in the WEng calculation with a default

of 12 sectors. A finer resolution of the inflow angle estimate may be obtained by increasing the number of sectors

in the WEng calculation. As a linearized flow model WEng has a tendency to over-estimate inflow angles in steep

and complex terrain.

Terrain fitting (5xHH disc) from complexity check

See the section 5.3.1 describing the Terrain complexity check for further details of the terrain fit. Again, it is

important to stress that this option is mentioned as acceptable in the IEC standard.

Wind shear

Description and limit

The Wind shear check evaluates the vertical wind shear or, in other words, the vertical variation of wind speed

across the rotor for each WTG position. Wind shear is quantified by means of the wind shear power law exponent,

α. The IEC design limits for wind shear is an average wind shear above 0 but less than 0.2 for all design classes.

The IEC standard is not clear regarding “the average shear” as it may refer to the shear exponent of the omni-

directional mean wind profile or to the average of the sector-wise shear exponents. Since the shear calculation

is non-linear (it is based on a log-log fit), these two estimates generally differ - but not drastically. In Section 11.3

of the Standard [1] it is mentioned that the shear must be estimated for at least 30° direction sectors. In SITE

COMPLIANCE the shear average is taken to be the weighted average shear of the sector-wise shear exponents.

A reason for this is that it enables more detailed adjustments to the sector-wise shear estimates. This

interpretation of “average shear” is consistent with the approach adopted in the Risø/DTU WAT tool [15].

SITE COMPLIANCE 49

Setup, Calculation and Results

There are seven options to select between for the Wind shear check, as illustrated in the Setup tab below. The

two upper options seek to combine mast measurements of shear with flow predictions to improve the estimation.

The next three options are based purely on flow modelling results and the last option is based purely on

measurements. From version 3.2 there is also a general option for the methods based on measured shear, to

based the shear estimates on the Frequency tables in the meteo object (original option) or to base it on

Concurrent samples (new option), which is more robust for errors in directions across measuring heights.

Figure 65. Setup tab of the Wind shear check.

The Results (Table) tab for Wind shear shows the IEC high and low design limits and the wind shear estimate

for each WTG. If the table is expanded for a particular WTG, the sector-wise shear estimates and their frequency

are presented.

Figure 66. Results (Table) tab of the Wind shear check.

SITE COMPLIANCE 50

Finally, the Results (Graphics) presents the overview of all the WTG results with the IEC limits highlighted as

blue lines.

Figure 67. Results (Graphics) tab for the Wind shear check.

Calculation Options

The following text describes the detailed calculation options of the Wind shear check and their main assumptions.

The figure below shows the options.

Figure 68. Calculation options in the Wind shear check.

Flowres propagation of mast shear heights

In this option, the selected Shear heights on each mast are transferred as time series to the same heights on

the relevant WTG positions, and, then, the sectorwise wind shears are estimated based on these heights. This

approach ensures that the stability information in the profile measurements is retained and only the effects of

roughness and terrain variations on the profile are accounted for. It is recommended to use this option with

measurements close to hub height where the shear heights cover a significant part of the rotor to ensure that

the shear estimates are represent the rotor area.

WEng shear adjusted using WEng error on mast shear

The sector-wise shear estimates based directly on WEng results are usually quite biased to the low side as

WEng assumes neutral stability. This bias may be estimated for each sector of the mast and applied as a

correction to the WEng shear estimated for each sector at the WTG positions. After applying the bias-correction,

SITE COMPLIANCE 51

the corrected sector-wise shear exponents are averaged using the sector frequencies for the turbine position to

obtain the final average shear estimate.

This method seeks to combine the best from measurements with the best from the flow modelling.

Measurements accurately represent the shear at a single position, namely the mast, but may not be

representative for the site as such. Flow modelling has its strength in resolving the relative variation of shear

across the site, although generally with a constant bias, which may be considerable.

The method relies on the assumption that wind does not turn significantly across the site or vertically, i.e,. flow

conditions in one sector at the mast are assumed not to have turned significantly when they reach the WTG

positions. The larger the sector width the better this assumption is. Turning across a site and vertically is usually

much less than the 30° of standard sectors, however, in very complex terrain turning may be very significant. It

is also important to validate that all direction sensors used on a mast are consistent, because such errors might

mess up this calculation option considerably.

WAsP shear adjusted using WAsP error on mast shear

This method is similar to the adjusted WEng method described above. The only difference is that the relative

variation of shear is extracted from the results of WAsP flow modelling. The WAsP profile is quite different from

the WEng profile as WAsP has a built-in stability model that has a very strong effect on the predicted profiles.

WAsP-CFD +/-½RD shear

As for WAsP, the default setup of WAsP-CFD includes modelling of wind conditions at HH and at HH+½RD and

HH-½RD, which allows estimation of the shear across the rotor. The WAsP-CFD model combines the “raw” CFD

based (Ellipsys model) flow perturbations with the WAsP modelling setup and stability model which only affects

the wind profile. The WAsP stability is rather simplified and has a very strong influence of the wind shear. With

default WAsP stability settings, WASP-CFD results are likely to result in conservative estimates of the wind

shear.

The quality of this calculation option depends on how well the WAsP stability parameters have been adapted to

match the observed profile at the site. The quality may be high if WAsP stability settings have been adjusted

appropriately and the WAsP-CFD profile matches well with the observed profile at the mast position.

WAsP +/-½RD shear

The default setup of the WAsP calculation includes modelling of wind conditions at HH (hub height), but also at

HH+½RD and HH-½RD, where RD is the rotor diameter. Having the flow conditions modelled at these three

heights allows the estimation of the wind shear across the rotor for each sector and, hence, definition of the

“average shear”.

The quality of this calculation option depends on how well the WAsP profile has been validated and adapted for

the site. The quality may be high if WAsP stability settings have been adjusted appropriately and the WAsP

profile matches well with the observed profile at the mast position.

In tropical and mid-latitudes, the WAsP profile using default stability settings tends to over-estimate wind shear.

The reason is that WAsP defaults are calibrated using the measurements from Northern Europe.

WEng (WAsP engineering) shear

Flow results from WAsP Engineering include the derivatives of the flow field and bases the calculation of wind

shear on these derivatives. Hence, the estimated shear represents the hub height exactly. However, as WEng

assumes neutral stability, the wind shear is usually under-estimated significantly using WEng results. This

method is not generally recommended. It is included to allow comparison with the more suitable estimation

methods as it is not common to see that WEng shear is used directly.

Mast shear assumed representative for all WTGs

This calculation is only recommended if the mast is representative for all the WTG positions and for the WTG

hub height, both in terms of terrain and roughness conditions.

Mast shear based on..

This is a general option for all the methods using measured shear at the mast(s). Shear estimates can be

based either on the Frequency tables in the meteo object (original option) or be based on Concurrent samples

(new option), which is more robust for errors in directions across measuring heights which relatively common.

SITE COMPLIANCE 52

Ignore all displacement heights in shear estimates.

This option is activated per default and included from windPRO version 3.3. The purpose is to prevent

underestimating wind shear when displacement heights are used to predict the wind speed distributions. It is

important to stress that this option will not affect the actual wind speeds predicted using the displacement

heights, the option simply uses the un-displaced heights together with the predicted wind speeds (including

any displacement effects) to get the “real” wind shear exponents relative to the true surface. If the

displacement heights are used when calculating the actual shear exponents they will typically be significantly

smaller as now the profile is described using two parameters. If such reduced shear exponents are used for

load simulations in an aero-elastic model, first step is to simulate turbulent and sheared wind fields. If this

model setup does not include displacement heights, but the input shear exponents are calculated using

displacement heights the wind shear is significantly underestimated and most likely also the resulting aero-

elastic loads.

Air density

Description and limit

The IEC standard requires an assessment of the air density at hub height. The IEC design assumption for air

density is 1.225 kg/m

. Hence, air densities lower than this limit results in less loading on the WTG and vice

versa.

In principle, the IEC standard requires the air density to be averaged only for wind speeds greater than the wind

speed at rated power of the WTG. This adds a significant complexity to calculation of this check. However, a

number of test cases have shown that the difference between the annual mean air density and the average for

wind speeds above U

rated

are usually within 1-2%. Considering the other uncertainties that affect air density

estimates, this small difference is not considered significant. In addition, it is common practice among turbine

manufacturers to use the annual mean air density instead. This practice has also been adopted in the SITE

COMPLIANCE module.

The model adjustments of temperature (and pressure) from mast height amsl (above mean sea level) to WTG

HH amsl are based on the ISO standard atmosphere as mentioned in the IEC standard.

Relative humidity is not an input as it only has a minor effect on air density. This is illustrated in the plot below.

For relative humidity to have an effect on air density exceeding 2%, the mean air temperature must exceed 35°C

and the mean humidity be close to 100% - it is worth noting that the highest annual mean temperature worldwide

is 34°C (in Dalol, Ethiopia

). Hence, in most typical locations, the effect of relative humidity is well below 1%. In

addition, including the effect of humidity decreases air density and, hence, the aerodynamic loads. Thus,

excluding relative humidity is a slightly conservative assumption in regards to wind turbine loads.

http://www.weatherexplained.com/Vol-1/Record-Setting-Weather.html#b

0.77

0.78

0.79

0.8

0.81

0.82

0.83

0.84

0.85

0.86

0.87

0.88

0.89

0.9

0.91

0.92

0.93

0.94

0.95

0.96

0.97

0.98

0.99

Temperature [

Relative humidity

Ratio of mixed to dry air density

-10 0 10 20 30 40 50 60 70 80 90

100

0.75

0.8

0.85

0.9

0.95

SITE COMPLIANCE 53

Figure 69. Plot showing the relative effect on air density of varying mean temperature and relative humidity.

Setup, Calculation and Results

The Setup tab for the Air density check requires only the choice of the input data. If a site mast is available with

temperature (and pressure, but not required), this option can be chosen. If there are more site masts with

temperature, just one has to be chosen. Alternatively, mast, e.g., SYNOP or METAR, data may also be

downloaded using the online data option in a Meteo Data Object. The mast must have temperature (and

pressure) and have been defined as a Climate mast in the general SITE COMPLIANCE setup.

If there are no on-site measurements of temperature (and pressure), climate statistics from the Global Historical

Climatological Network (GHCN) may be chosen. By default, this option will choose the nearest station.

Sometimes a more distant station may be more appropriate if its elevation above sea level is more representative

for the considered site. To review and choose other stations click the Climate database button.

Figure 70. Setup tab of the Air density check.

Once the calculation is run by clicking the green “Calculate” button, the results tabs are generated. The tab Base

data summarizes the selected masts annual mean data and the measuring height above sea level. Below that,

all the relevant atmospheric model parameters are listed and, at the end, the air density calculated directly for

the mast data is shown.

SITE COMPLIANCE 54

Figure 71. Base data tab of the Air density check.

The Results (Table) tab shows the resulting annual mean air density for each WTG. A column shows the height

difference to the mast used and another coulumn the predicted annual mean temperature and pressure.

Figure 72. Results (Table) tab of the Air density check.

Finally, the tab Results (Graphics) provides the overview plot of all the WTG results and the IEC design limit.

SITE COMPLIANCE 55

Figure 73. Results (Graphics) tab of the Air density check.

Calculation Options

For both calculation options, the same extrapolation model is used. It corrects the mean temperature using the

standard temperature lapse rate of -0.0065 K/m from measuring height to WTG hub height. If pressure

measurements are available too, these are corrected using the temperature difference as input and the

hydrostatic equation. If no pressure data is available, the sea level pressure is corrected to WTG HH using the

standard atmosphere model.

Site or climate mast with Temperature (and Pressure)

If temperature (and pressure) data are available for a site mast, the air density calculation may be based on

these - as described above.

GHCN Climate database

GHCN is a database

maintained by the US institution NOAA (National Oceanic and Atmospheric

Administration). The database contains historic climate statistics for a large number of stations worldwide.

IEC Checks - Other checks

The IEC standard [1] lists a number of Other environmental conditions to be assessed and compared with the

design assumptions for the WTG. In addition, Earthquake conditions must be evaluated to assess if further

actions in this regard are needed for the site. The list below summarizes these additional site parameters

mentioned in the IEC standard.

• Earthquake conditions (seismic hazard)

• Normal and extreme temperature ranges

• Lightning

• Icing, hail and snow

• Humidity

• Solar radiation

• Chemically-active substances

• Salinity

In SITE COMPLIANCE, we have only included the additional parameters highlighted in bold above. The selection

was based on a combination the possibility of allowing a meaningful estimation method and data source and the

risk that a specific parameter is critical for a site.

http://www.ncdc.noaa.gov/ghcnm/

SITE COMPLIANCE 56

Seismic hazard

Description and limit

The IEC standard requires a site estimate of the peak ground acceleration (PGA) with a recurrence period of

475 years for the site area. This recurrence level is equivalent to an annual risk of exceedance of 0.2%. There

is no explicit IEC limit for PGA, hence, the general hazard levels of GSHAP have been adopted in SITE

COMPLIANCE. “Low hazard” results in OK, “Moderate hazard” in Caution, and “High” and “Very high hazard”

results in Critical, emphacizing the need for futher and detailed investigation of seismic loads.

For the PGA site estimate, SITE COMPLIANCE uses a database which results from the UN funded Global

Seismic Hazard Assessment program (GSHAP). GSHAP database is compiled from several regional and

national sub-projects and summarizes the PGA in m/s

at a resolution of 0.1° by 0.1°. The GSHAP data is

available via the ETHZ or GeoForschungsZentrum Potsdam homepages. The figure below illustrates the global

GSHAP seismic hazard map.

The PGA mapped in GSHAP refers to standard soil conditions which is rock (rock/firm soil in US/Canada). As a

consequence very special local conditions which differ significantly from this like a deep and soft sedimentary

setting may amplify ground motions significantly, making conditions more severe than mapped in GSHAP.

Nevertheless, GSHAP is considered a reliable source of seismic hazard (in terms of PGA) as it compiles

numerous local and regional studies made by local experts and authorities.

Figure 74. GSHAP seismic hazard map showing global variation of PGA. From [16].

Setup, Calculation and Results

No inputs are required in the Seismic hazard calculation as only the GSHAP database is available.

SITE COMPLIANCE 57

Figure 75. Setup tab of the Seismic hazard check.

The Results (Table) tab summarizes the site estimated seismic hazard in terms of the PGA level in m/s². It also

gives the expected hazard level.

Figure 76. Results (Table) tab of the Seismic hazard check.

The last tab in the Seismic hazard check provides a map of the seismic hazard in an area of approximately

1000km by 1000km around the site.

Note that since GSHAP is compiled of several national and regional projects, some variations in PGA may not

allways be 100% smooth across national borders, as is seen for the U.S. - Canada border (not shown).

SITE COMPLIANCE 58

Figure 77. Results (Graphics) tab of the Seismic hazard check.

Temperature range

Description and limit

Wind turbines are designed for operation within a certain range of temperatures: the operational or Normal range.

A wind turbine may survive temperatures beyond the operational range, but will cease operation in order to

survive. If temperatures exceed the so-called Extreme range or survival range, components within the wind

turbine may suffer permanent damage.

The Normal range for the IEC standard class WTGs is: -10°C to +40°C.

The Extreme range for the IEC standard class WTGs is: -20°C to +50°C.

Most manufacturers produce high or low temperature versions of their WTG models with extended normal and

extreme ranges for specific markets. Such models are categorized as Class S WTGs. The temperature range is

typically extended by around 10°C at the high or low end of the ranges noted above in Class S versions.

Calculations for customized Class S temperature ranges are supported by SITE COMPLIANCE and described

below.

Setup, Calculation and Results

On the Setup tab of the Temperature range check, there are two setup groups; Select data and fit and

Temperature design limits. In the former group, a mast with temperature measurements must be selected. In

the latter group, the fit option must be selected as either Full Gaussian or Tail Gaussian, deciding whether to fit

the full the range of measurements or just the high and low tails, respectively. If the Tail Gaussian option is

selected, the tail fraction may be adjusted. By default, the highest and lowest 10% of the samples are fitted in

this option.

SITE COMPLIANCE 59

Figure 78. Setup tab of the Temperature range check.

Once the calculation has been performed by clicking the green “Calculate” button, the results tabs appear. The

Base data tab summarizes the chosen temperature measurements and the mast details as well as the mean

WTG HH above sea level, which is used to estimate a single result for the entire site.

Figure 79. Base data tab of the Temperature range check.

The Results (Table) lists the chosen limits for the normal and extreme temperature ranges. In the table, the

column labelled “hours < Tmin [h/year]” shows the estimated hours below the normal and extreme low

temperature limits, respectively. The column labelled “hours > Tmax [h/year]” shows the hours above the normal

and extreme high temperature limits, respectively. The total result is the estimated sum of hours outside the two

ranges, shown in the rightmost column.

SITE COMPLIANCE 60

Figure 80. Results (Table) tab of the Temperature range check.

On the Results (Graphics) tab, the measurements and the fits are presented. In the upper (larger) plot, the

cumulative sample distribution is shown with the normal and extreme ranges indicated using the vertical red

lines. If the Full Gaussian fit was chosen in the setup, this fit is also illustrated in the plot as the blue curve.

The two lower plots illustrate the upper and lower quantiles (25%) of the samples, respectively. The frequency

of occurrence axis (y-axis) has been transformed to illustrate, directly, the hours below (left graph) or above

(right graph) the temperatures on the x-axis. The fit(s) are also illustrated in both these plots. The hours below

or above the normal and extreme limits may be read directly at the y-axis besides the crossing points where the

fit(s) cross the vertical red lines.

Figure 81. Results (Graphics) tab of the Temperature check.

SITE COMPLIANCE 61

Calculation options

The calculation options of the Temperature range check are described in the following.

Select data and fit

A Site or Climate mast with temperature data may be chosen in the setup as illustrated below. The SYNOP

stations are often a good alternative if no on-site data is available. Users should make sure that the mast

elevation above sea level is properly defined in the properties of the Meteo object as this may strongly influence

the results. Usually, the mast evelation may be read from the header lines of the data files in the Meteo object.

Figure 82. Setup details of data source and fit for Temperature range check.

Full Gaussian fit

Selecting this option fits the full sample distribution using a Gaussian relationship, which has the same mean

and standard deviation as the samples (method of moments). This method is very robust but for many sites. The

fit is not always satisfactory which is clearly visible from the graphs.

Tail Gaussian fit

This option fits separate Gaussian distributions to the low tail and to the high tail ranges with the default “tail

range” set to fit the lowest 10% and the highest 10% of the samples. The fits are performed using the so-called

Normal probability plots of the ordered samples against the theoretical Gaussian quantiles, but only fitting the

relevant tail range.

Temperature design limits

The figure below illustrates the setup of the Temperature design limits.

Figure 83. Setup of Temperature design limits in the Temperature range check.

Standard limits

The default option Standard limits is used for standard class WTGs or in initial screenings of the Temperature

range check. Note that some manufacturers, such as Vestas, have extended the temperature range on all their

standard WTG models to, e.g., -20°C for the Normal range minimum temperature. Such models should be

calculated using Class S limits for temperature range although the WTG model is not a special high- or low-

temperature version.

Class S limits

The Class S limits option should be selected and Normal and Extreme temperature ranges adjusted

appropriately if the WTG model in question is known to be a low- or high-temperature version. This option is also

appropriate if, perhaps, a Temperature range calculation was initially performed for the standard limits, but

resulted in a Critical exceedance. In such cases, it is necessary to explore which temperature limits would then

be needed/acceptable. Usually, the special high- or low-temperature versions have Normal and Extreme limits

which are extended by 10°C or 20°C, to, e.g., -30°C and -40°C for a low-temperature version.

SITE COMPLIANCE 62

Lightning rate

Description and limit

The estimation of Lightning rate in SITE COMPLIANCE is based on a database established by the NASA Global

Hydrology and Climate Center (GHCC) [17]. Data were collected using two kinds of satellite detectors LIS

(Lightning Imaging Sensor) and OTD (Optical Transient Detector).

The IEC standard does not specify a specific limit for the lightning rate.

Setup, Calculation and Results

No alternatives are available in the setup of the Lightning rate check.

Figure 84. Setup tab of the Lightning rate check.

The Results (Table) provides the expected lightning rate for the site.

SITE COMPLIANCE 63

Figure 85. Results (Table) tab of the Lightning rate check.

The last tab Results (Graphics) illustrates the lightning rate in a ca. 1000km by 1000km square, centred on the

site.

Figure 86. Results (Graphics) tab of the Lightning rate check.

SITE COMPLIANCE 64

(Re)calculate all

Once a complete SITE COMPLIANCE calculation has been setup and each included calculation is performed,

it is time to assess the overall results. In the case shown below, design class IIIB is checked for a site. Most

checks came out as OK, three checks got the result Caution, and two checks (Extreme wind and Wind

distribution) ended in the category Critical.

To check another design class for the site is very easy and fast in SITE COMPLIANCE. Just go back to the

Main tab and change the design class to, e.g., a higher wind speed class, e.g., class IB or IIB instead of IIIB.

Figure 87. Part of the Main tab where design class is set.

Back on the Calculations tab, all results will be cleared as one of the basic settings has been changed. To

recalculate all included checks simply click the (Re)calculate all button. This will recalculate all calculations

included for each check and update the visual results accordingly.

Class IIIB Class IIA

Figure 88. Left: Calculation and overall result for the initial design class IIIB. Right: Re-calculation for the design

class IIA.

A suitable design class maybe found by iteratively increasing the selected design class for wind speed (e.g., III

to II) and turbulence (e.g., B to A) until all main checks are green or, perhaps, yellow.

However, there is a more accurate and less conservative alternative to just iteratively increasing the selected

design class. This alternative is the module LOAD RESPONSE, which allows the user to perform a fatigue

load assessment for each WTG position in the layout. LOAD RESPONSE is an independent module but is fully

integrated in SITE COMPLIANCE and may be activated on the Main tab of SITE COMPLIANCE, if a valid

license is available.

The next section of this manual describes LOAD RESPONSE.

SITE COMPLIANCE 65

LOAD RESPONSE 66

5.3 LOAD RESPONSE

From windPRO version 3.0, the module LOAD RESPONSE has been integrated in SITE COMPLIANCE for

estimation of fatigue loads for each WTG in a layout.

LOAD RESPONSE is developed as part of the publicly funded research project: “Optimized Integration of Load

Calculations in Development and Design of Wind Farms” led by Post Doc. Henrik Stensgaard Toft. The project

is a collaboration between EMD, Aalborg University (Department of Civil Engineering)

and Innovation Fund

Denmark

. This setup has ensured a very strong scientific basis for LOAD REPONSE. In addition, the generic

wind turbines provided with LOAD RESPONSE have been thoroughly validated in collaboration with a leading

wind turbine manufacturer to ensure their representability and accuracy. Finally, the response surface

methodology employed in LOAD RESPONSE to establish fatigue loads has been previously certified by TÜV-

SÜD for consistency with the ÌEC61400-1 ed. 3 standard.

LOAD RESPONSE, in its current implementation (windPRO 4.0), is focussed on fatigue loads during power

production, which in IEC 61400-1 e. 3 (2010) is denoted “design load case 1.2”. This is typically the most

significant fatigue design load case, because it covers most of the wind turbines design lifetime and depends

strongly on the site specific wind climate. However, extreme loads which are not covered in LOAD

RESPONSE can also be design driving for wind turbines, but these are typically handled directly via, e.g.,

evaluation of the Extreme wind speed check in SITE COMPLIANCE.

Justification

Many SITE COMPLIANCE calculations are not fully conclusive in the sense that one or more of the main IEC

checks relating to fatigue loads are partly (yellow) or fully (red) exceeded, whereas other main checks are well

within IEC limits. The yellow and red main checks are often due to exceedances of the wind climate

parameters for specific wind speeds as shown in the figure below. It is, therefore, possible that the exceedance

of some wind climate parameters at specific wind speeds can be compensated by lower values at other wind

speeds or in other wind climate parameters. This evaluation is performed in LOAD RESPONSE based on a

response surface methodology.

Figure 89. Exceedance of wind speed distribution (left) and turbulence (right) at specific wind speeds.

In the SITE COMPLIANCE calculation shown below, a case is illustrated where the main checks Effective

turbulence, Wind distribution, Wind shear and Air density are all partly exceeded, whereas Flow inclination is

well within the limits.

http://www.civil.aau.dk/

http://innovationsfonden.dk/en

LOAD RESPONSE 67

Figure 90. A SITE COMPLIANCE calculation illustrating the need for LOAD RESPONSE.

Based on the above SITE COMPLIANCE result, it is not possible to directly make an accurate decision if the

turbine model/design class is suitable for the site and layout or not. A fatigue load assessment is required to

make an appropriate conclusion. In fact, the IEC standard requires a fatigue load calculation in these cases by

the statement in Section 11.1 ([1], p 52):

“…It shall be shown that the site-specific conditions do not compromise the structural

integrity. The demonstration requires an assessment of the site complexity, see 11.2, and an

assessment of the wind conditions at the site, see 11.3. For assessment of structural integrity

two approaches may be used:

a) a demonstration that all these conditions are no more severe than those assumed for

the design of the wind turbine, see 11.9;

b) a demonstration of the structural integrity for conditions, each equal to or more severe

than those at the site, see 11.10.

If any conditions are more severe than those assumed in the design, the structural and

electrical compatibility shall be demonstrated using the second approach.”

In other words, approach b) is required if one or more of the IEC checks are exceeded. LOAD RESPONSE is

an implementation of the requirement in approach b), as it allows the user to compare the site specific fatigue

loads with the design loads assumed in the IEC standard for all the important turbine components.

Only the manufacturers have access to the complete aero-elastic model of a wind turbine model, which is

needed for “approach b” (see above). For this reason, LOAD REPONSE comes with two generic wind turbine

models, which allows the user to estimate fatigue loads without access to the original aero-elastic model.

Figure 91 below illustrates the overall result of a LOAD RESPONSE calculation for the problematic SITE

COMPLIANCE result above in Figure 90. The results show that site specific loads for all WTG positions are

OK for all the main components, and that the WTG with the highest loads is “WTG 27”. The column Load Index

shows the WTG loads for the site/layout normalized with the IEC design loads. A load index of <100% means

that WTG loads are less than the design loads and, hence, OK.

Figure 91. The overall LOAD REPONSE result for the SITE COMPLIANCE result in Figure 90.

Practicalities

LOAD RESPONSE is a separate module and requires a separate license, but it is fully integrated in SITE

COMPLIANCE. In the module list, this is illustrated by the different shapes of the green markers of the two

modules (see below).

LOAD RESPONSE 68

Figure 92. Left: module list with the Loads group expanded. Note that LOAD REPONSE has a green circle and

SITE COMPLIANCE a green triangle. Right: the message if one tries to start LOAD RESPONSE on its own from

the module list.

LOAD REPONSE is activated from the main tab of SITE COMPLIANCE under the headline Load calculation

by checking the box Include LOAD RESPONSE (see figure below). When this box is checked, a new tab

named LOAD RESPONSE appears. Initially, the tab has a red stop mark to indicate that no load results are

available yet.

Figure 93. Activation of LOAD RESPONSE from the main tab of SITE COMPLIANCE. Note the yellow indication

on the tab “IEC checks” and the stop sign on the tab LOAD RESPONSE illustrating the overall result of SITE

COMPLIANCE and that the LOAD RESPONSE calculation is not yet run.

Wind turbine manufacturers can easily, and without any data leaving the company, add their own turbine

models for in-house use or to share with, e.g., developers. When implemented, manufacturer specific models

are protected by a high level of security via both encrypted file format and license control (see section 5.3.4 for

further details).

What is a response model?

LOAD RESPONSE uses a response surface methodology to calculate fatigue loads for a particular set of WTG

wind climate parameters. The response model is simply a set of pre-run aero-elastic simulation results (fatigue

loads) for certain combinations of wind climate parameters combined with a surrogate model trained on these

pre-run data.

LOAD RESPONSE support two different types of surrogate models, see Appendix IV - Theory of LOAD

RESPONSE and Fatigue for more information.

A response model is required for each component or cross section in the wind turbine model – referred to as

“sensors”. Typical sensors are, e.g., “Tower bottom for-aft bending moment” or “Blade out-of-plane bending

moment”. Figure 94 shows an example of the aero-elastic response for a sensor as function of wind speed and

wind shear, where the surrogate model is trained on a subset of points located on this surface.

LOAD RESPONSE 69

Figure 94. Aero-elastic response for the sensor: “Blade out-of-plane bending moment” as a function of wind

speed and wind shear. The vertical axes show normalized Damage Equivalent Loads (DEL).

Setting up and running the fatigue load calculation

First, make sure that LOAD RESPONSE is activated by checking it on the main tab of SITE COMPLIANCE.

The next step is to select the response model for use in the load calculation. This is done on the LOAD

RESPONSE tab via the dropdown menu Select WTG response file for all turbines at the top of the tab. LOAD

RESPONSE comes with several generic response models, as shown below: a response model for rotor

diameters of 90m and up and a model for rotors below 90m.

LOAD RESPONSE 70

Figure 95. The LOAD RESPONSE tab in SITE COMPLIANCE.

Figure 96. Selection of generic response model.

This choice of response model will apply to all WTGs in the layout, which is the typical situation when using the

generic turbine models unless the layout consists of turbines with very different rotor diameters. If the layout

consists of turbines with different IEC design classes which have been setup in each individual WTG object, this

is handled automatically when the generic turbine models are used.

The alternative option, named Select WTG response file (individual turbines), just below the default option,

activates the Select button to access selection of individual response models for each WTG in the layout. This

option is only relevant if the generic model is used for layouts with different rotor diameters or when manufacturer

specific responses are used for layouts which include different turbine models or different IEC design classes.

Once the response model has been selected, the next step is to select Fatigue loads by checking the Include

box beside it. Once included, an Edit button appears, which gives access to the fatigue load setup and

calculation.

The Fatigue loads window is shown below. To start the calculation, press the green Calculate button at the

bottom of the window. Before doing that it is recommended to review the main assumptions and calculation

setup summarized in the window. The following text describes these assumptions and options.

LOAD RESPONSE 71

Figure 97. The Fatigue Loads setup and calculation window in LOAD RESPONSE.

IEC design load case (DLC)

The fatigue load estimation in LOAD RESPONSE includes the IEC design load case DLC 1.2 “Power production”

which covers fatigue during normal operation and there is an option to include ‘DLC other’. The latter represents

the combined contribution of the additonal fatigue related load cases 3.1, 4.1 and 6.4 covering ‘start-up’, ‘normal

shut-down’ and ‘parked’ situations. For further details, see the appendix on the theory behind LOAD

RESPONSE.

The option “Directional tower loads” may be activated to include the effect of wind direction on fatigue

accumulation in the tower. It works by dividing the tower circumference into 36 equidistant points, and then the

fatigue load is calculated at each point. Subsequently, the critical point with the highest fatigue load is chosen to

be representative of the strength requirement of the tower. Note that this option combines the tower “fore-aft”

and “side-side” loads into a single combined load which is simply named “TwbMax” for tower bottom max across

all evaluated points. For more information on the applied method please refer to [33].

Wind climate parameters

Only some of the IEC main checks contribute to the estimation of fatigue loads, these checks are listed in this

field and include: Effective turbulence (including the turbulence structure correction parameter for complex

terrain), Wind distribution, Flow inclination, Wind shear and Air density.

WTG Information

This field descripes the selected wind turbine model which forms the basis of the response model - in the case

shown, the model chosen is the generic turbine model for typical/intermediate rotors (≥90m) which comes with

LOAD RESPONSE. The design lifetime is part of the turbine specification and is typically defined as 20 years.

The bottom entry in the field specifies the mathematical response model named Central composite

approximation – this is the response surface method previously certified by TÜV-SÜD (see Appendix IV for

further details).

WTG Components

This field lists the WTG components included in the selected response model. The generic response models

include the components: Blade, Tower, Nacelle and Shaft. In the actual fatigue load calculations, each

component is represented by a number of cross sections and material parameters referred to as Sensors and

Wöhler exponents, respectively. Manufacturer specific turbine models may include other components and

sensors than those included for the generic models.

LOAD RESPONSE 72

Directional resolution

LOAD RESPONSE provides three options for the directional resolution of the fatigue load assessment. All

options are in accordance with the requirements in the IEC standard, however, the omnidirectional case is the

default case in the IEC standard and also in LOAD REPONSE.

Omnidirectional (IEC61400-1 ed. 3, 2010)

With this option, omnidirectional values are used for all the wind climate parameter in the main IEC checks.

Sectorwise

In this option, the fatigue load estimation is performed sectorwise using the sectorial values of the wind climate

parameters except for air density. Using a sectorwise wind climate partly relieves the dependence on the coarse

assumptions, e.g., in the Frandsen model of effective turbulence and using the sector-weighted average of the

wind shear. Thus, overall sectorwise load assessments are expected to provide more accurate results than

omnidirectional assessments.

Full resolution (no effective TI integration)

This option uses the wind climate parameters wind shear, flow inclination, and total turbulence (wake+ambient)

for each single degree (centred on half degrees, e.g., 0.5°, 1.5°, 2.5° etc.) with the response model. In most

cases, the climate parameters do not vary within each 30 degree sector, except for turbulence, where the wake

added turbulence does not align with sectors and may inflict significant variation of turbulence within sectors.

The use of the “Full resolution” options avoids the need to make a Frandsen integration over all directions or

over each sector to approximate the effect of varying turbulence on fatigue accumulation. In this option, the

fatigue accumulation is done explicitly and directly using the response models of LOAD RESPONSE.

Note: This option has been added after the original certification by TÜV SÜD. However, it is very similar to the

sectorwise option, and will, in most cases, provide close to identical results. From a theoretical perspective, this

option is the most stringent one, as it relives the need for the much simplifying and disputed approximation in

the Frandsen integration, which mixes the climatic parameter turbulence with the material parameter (Wöhler

exponent).

Effective turbulence

For application of the Effective turbulence, LOAD RESPONSE provides two options, variable or fixed. Both

options are in accordance with the IEC standard, but can result in quite different load levels mainly for the tower

or other steel parts represented by low Wöhler exponents. Hence, for manufacturer specific models, it is

important to use the same approach as used in the type certification of the turbine model.

Variable Wöhler exponent (IEC61400-1 ed. 3, 2010)

In this option, the Effective turbulence is calculated for all the relevant Wöhler exponents covered by the different

components/sensors. For each sensor/component, the appropriate Wöhler-dependent Effective turbulence is

used.

Fixed Wöhler exponent (m=XX from Site Compliance)

In this approach, the Effective turbulence is calculated only for the particular Wöhler exponent chosen by the

user in the SITE COMPLIANCE calculation (the value XX, default is m=10). This Wöhler-independent Effective

turbulence is used for all components/sensors regardless if their material is represented by a different Wöhler

exponent.

Normalization loads

Normalization load is synonymous with ‘design load’ or ‘reference load’ which represent the denominator when

calculating load indices. For generic turbine models no loads margins/reserves are known and, hence, the

normalization loads are always calculated using the response model (Use response model) for the design wind

climate for the design class in question. For manufacturer specific models the design loads are defined directly

in the load response file (From response file), and these will be used in calculation using the turbines correct

design class. However, in the special case where the user wishes to calculate for another design class than the

turbine is certified for, the included design loads are no longer valid and the normalization loads must be

calculated using with the Use response model option.

LOAD RESPONSE 73

Results – the fatigue load estimates

Once the calculation has been run (takes a few seconds), a number of new result tabs will appear. The following

text describes the content of these result tabs and serves as guidance on how to interpret the results.

Figure 98. The result tabs, which emerge when the calculation is completed.

Results (Table)

The Result table provides the most comprehensive and complete presentation of the fatigue load results. For

each WTG position, the table presents the Load Index for each sensor grouped according to the component it

represents. For each sensor the data columns describe the following: the Sensor description in human language,

the Wöhler exponent, the Load index and finally, a column called Fatigue life. For manufacturer specific turbine

models, it is also possible to see the estimated fatigue loads.

Load index is simply defined as: “WTG loads” / “IEC design loads”, where IEC design loads means the loads

obtained by using the wind climate parameters (limits) of the IEC design class in question, e.g., IIIA.

Fatigue life is calculated from the load index and Wöhler exponent as described in the appendix. Note that a

Load index of 100% will result in a Fatigue life equal to the design lifetime - 20 years for the generic WTGs.

Figure 99. The Result (Table) tab.

It is important keep in mind that the fatigue life does not account for other material degradation phenomena,

such as abrasion or corrosion. Fatigue life only accounts for fatigue, hence the name. Nevertheless, fatigue life

estimates can be very useful to quantify the effect of extending a wind farm in terms of the accelerated material

fatigue on the turbine components, e.g., blades, due to the increase in wake turbulence from installing additional

turbines. Such an assessment can be done rather simply by performing two LOAD RESPONSE calculations:

one without the additional turbines and one with the additional turbines. The difference in fatigue life for the

original layout obtained in the two calculations is a meaningful quantification of the reduction in component

(fatigue) lifetime. In the case that the existing wind farm has already been in operation for several years, this

should be taken into account in the assessment.

LOAD RESPONSE 74

Results (Graphics)

The graph in the Results (Graphics) tab shows, per default, the Load Index for the worst component/sensor for

each WTG. This load index can represent different components for the different WTGs, to see which

component/sensor represents the worst load index, move the cursor/mouse to the data point of interest. Use the

dropdown menus below the graph to show load indices for a particular component/sensor.

Figure 100. The Results (Graphics) tab.

Results (Map)

The final tab shows an overview map of the results with the relative turbine positions and a colour coding to

indicate if the Load index is OK or not (i.e., below or above 100%). Again, the default view shows the results for

the worst component and can be changed to any component/sensor using the dropdown menus just below the

graph. The Color scale can be changed from binary OK / Critical (green / red) to a full color scale indicating the

variation in the load across the WTG layout. This view gives the user a fast overview of the most critical wind

turbines in the layout.

LOAD RESPONSE 75

Figure 101. The Results (Map) tab.

Visualize Damage Matrix via ‘Results (Table)’

On the Results (Table) tab for each WTG and sensor result there is a button named Visualize in the left most

column named Visualize damage matrix. By clicking this button the fatigue damage in each bin of wind speed

and direction is visualized.

Figure 102. Visualization of a damage matrix for a sectorwise calculation. This plot shows that most damage is

accumulated for operation between 70-100° & 7-12m/s and 190-290° & 10-16m/s (deep orange areas).

Description of the generic wind turbine models

LOAD RESPONSE comes with several generic response models:

• a model for turbines with small rotors (<90m),

• a model for turbines with typical/intermediate rotors (≥90m)

• a model for turbines with very large rotors (≥160m).

• a direct drive (DD) version of the model for typical/intermediate rotors (≥90m)

LOAD RESPONSE 76

Each of these models are available with and without carbon in the blades. All these response models are based

on a modern standard wind turbine design with the main characteristics being the following:

Generic turbine configuration:

• 3 blades

• Pitch control (collective)

• Standard PID controller

• Gearbox (except for the DD version)

• Steel tower

• Onshore

The underlying aero-elastic simulations for the generic response models have been run for a selection of sensors

to represent the main components of the wind turbine as listed below.

Components and sensors included for the generic response models:

• Blade:

o Root out-of-plane (DEL)

o In-plane moment (DEL)

• Tower:

o Bottom for-aft (DEL)

o Side-to-side moment (DEL)

• Nacelle:

o Yaw bearing tilt (DEL)

o Yaw moment (DEL)

• Shaft:

o Low-speed-shaft torque (DEL)

o Low-speed-shaft torque (LDD)

Figure 103. Left: list of sensors for the generic response models. Right: turbine sketch with the sensor hot spots

indicated.

The abbreviation DEL indicates Damage Equivalent Load and LDD indicate Load Duration Distribution. The

concept “damage equivalent load” refers to a common way of expressing cumulative fatigue loads.

Mathematically, the DEL results from accumulating the fatigue contributions at different cycles (extracted by

rainflow-counting) using Miner’s rule to get the cumulative fatigue damage. This damage is combined with a

linear S-N curve to get an equivalent stress range (load) for a chosen number of cycles or cycle frequency. The

resulting equivalent stress range is referred to as the Damage Equivalent Load. The linear S-N curve (defined

by the Wöhler exponent m) relates stress range to number of cycles to failure. Load duration distribution is a

very similar concept for load durations which better represents damage accumulation in mechanical

components, e.g., gearboxes (See Appendix IV for further details).

Wind turbine manufactures who wish to implement their own turbines in LOAD RESPONSE can freely choose

their preferred combination of sensors and component groups independent of the setup EMD has chosen for

the generic turbine models. The main principles of adding manufacturer specific response models are

described in the following section.

How-to add a new turbine load response model (manufacturers only)

Wind turbine manufacturers are recommended to implement their own turbine models for in-house use in LOAD

RESPONSE. This will enable them to perform very fast and accurate fatigue load assessments to speed-up the

internal project turnaround time by optimizing the interaction between the wind & site department and the load

department.

The pre-requisite to implement a wind turbine is to run a number of aero-elastic simulations using the particular

turbine model. A document specifying these required runs and the format in which to save them is available from

EMD upon request

. It is important to note that all aero-elastic simulations and post-processing are done in-

house by the manufacturer and using the manufacturers own in-house tools. See the figure below:

Please contact: Troels Juul Pedersen, tjp@emd.dk

LOAD RESPONSE 77

Figure 104. Schematic representation of the steps in order to implement wind turbines LOAD RESPONSE.

Once the required aero-elastic simulations and post-processing is completed, the results need to be saved in a

standard binary file format, “mat”-format also used in MATLAB. The formatting is described in the EMD reference

document for turbine implementation

. In addition, an “xml”-file is required, which describes the general name

and configuration of the wind turbine model. Once the mat-file and xml-file have been prepared, the next step is

to wrap and encrypt the files into windPRO’s internal “.loadresponse” file format. In this process, it is also possible

to define which windPRO license number will be allowed to use the file from LOAD RESPONSE in windPRO

(given the.loadresponse file is available). In summary, the LOAD RESPONSE files implemented by turbine

manufacturers for in-house use have the following three layers of security:

• Encrypted file format - the aero-elastics data a stored in a binary format with encryption

• License control - the manufacturer decides exactly which windPRO licenses can access the file

• File sharing - only users which have received the specific load response file from the manufacturer and

copied it to the PC can use it (if their license allows it, see above)

As an additional fourth layer of security, the manufacturer can choose to normalize all the aero-elastic load input

data for the response model with an arbitrary constant only known to the manufacturer. This will not influence

the main result - the load indices - as they are the ratio of WTG loads to IEC design loads which will both contain

the constant.

Zip and Encrypt tool

The “Zip and encrypt Load Response” tool is accessed via the Loads & Operation menu in windPRO 4.0 as

shown in the figure below. Note that this tool can only be used with a valid license for LOAD RESPONSE.

Figure 105. Top: How to open the “Zip and Encrypt Load Response” tool via Tools menu. Bottom: The “Zip and

Encrypt Load Response” tool.

LOAD RESPONSE 78

Once opened the Zip and Encrypt tool provides an interface to:

• Select with type of model type.

• Select the required .XML file describing the turbine and the aero-elastic runs

• Select the required .MAT file with the fatigue loads from the aero-elastic simulations

• Define which user license IDs are allowed to access the response file from windPRO

Note: If the field with user IDs is left blank, there will be no restriction in terms of which license IDs can access

the file in windPRO – of course, given that the file has been copied to the user’s PC.

Once all the required inputs have been defined, press the button Check, Pack and save loadsponse file. The file

should be saved in the LoadResponse folder in windPRO Data: “…./windPRO Data/LoadResponse”. Once

completed, a window will pop-up (see Figure 106) with a summary of the consistency checks of the xml and mat

files.

Figure 106. Summary of the consistency checks for the selected xml and mat file.

If allowed by the license check, the saved specific wind turbine model will now appear as a selectable response

file in LOAD RESPONSE.

Accuracy tab and Certification requirements

For manufacturer specific wind turbine models implemented in LOAD RESPONSE, certain restrictions have

been agreed between EMD and TÜV-SÜD as part of the original certification of LOAD RESPONSE. These

restrictions ensure that the accuracy of the response model is acceptable for the combination of the manufacturer

specific model and WTG climate parameters for the site in question.

Note that since these requirements were based on a surrogate model developed using the Central Composite

Design (CCD), these tabs are not available when using a model based on Gaussian Process Regression (GPR).

These certification restrictions apply if the results are used in official contexts as, e.g., in a building permit. For

this reason, an additional result tab named Accuracy will appear in LOAD RESPONSE when used with

manufacturer specific in-house turbines. The top of the tab summarizes the check against the requirements. The

following text describes the details of the two check criteria on the Accuracy tab.

Figure 107. Accuracy tab (top part only) with the summary of the accuracy checks.

LOAD RESPONSE 79

Deviation of WTG climate parameters (Accuracy tab)

First sub-tab on the Accuracy tab is an analysis of the of the WTG wind climate parameters relative to the

reference conditions in the response model and the limit agreed in the original certification. The reason for this

check is that the response model is based on a number of aero-elastic simulations covering the typical region of

variation of the four WTG wind climate parameters: Effective turbulence (I

ref

), Wind shear (α), Inflow angle (φ)

and Air density (ρ). However, if the WTG wind climate parameters are significantly outside the region covered

by the response model, the error on the response model will increase and it’s accuracy decrease. To estimate

how far outside the area covered by the response model a WTG wind climate is, a normalized “radius” is

caculated for each WTG. This radius is simply the euclidian distance from the reference point (I

ref,0

, α

, φ

, ρ

)

of the response model (typically the IEC design climate of the relevant class) to the WTG wind climate in the

four dimensional (I

ref

, α, φ, ρ) space. However, each of the four dimensions (axes) are normalized to a

characteristic step length used in the response model (ΔI

ref

, Δα, Δφ, Δρ) as is evident from the equation:















 











 

 









 

 









 

  











As an example, let us take a case where a WTG has all wind climate parameters equal to the IEC class except

for the Air density, which is one step length (Δρ) higher than the reference. In this case, the normalized radius

becomes equal to R

norm

=1.

Figure 108. Accuracy tab showing the deviation of site climate parameters from the reference climate

(manufacturer specific response files only).

Accuracy of response model (Accuracy tab)

The response model in LOAD RESPONSE is based on a fit to 25 combinations of wind climate parameters at

each wind speed bin. Each sensor has its own response model. For each sensor and each of the 25 “fit points”,

it is possible to calculate the error or residual of the fit. The sub-tab Accuracy of response model summarizes

the errors for these 25 “fit points” in terms of the maximum absolute relative error.

In addition to the 25 fit points, 8 additional “control points” are required to fullfill the requirements of the original

certification. These 8 control points are not used in the fit, and thus, allow for an extra independent check of the

LOAD RESPONSE 80

accuracy of the model. The errors for the control points are also summarized in the sub-tab Accuracy of response

model.

Figure 109. Accuracy tab showing the check of the response surface accuracy for the fitted 25 points and for the

8 supplementary accuracy control points. In the case shown, the error/deviation for all points are within the

acceptable limit.

Fatigue Loads in OPTIMIZE

Users with an active LOAD RESPONSE license can also enable fatigue lifetime constraints in the layout

optimization within the OPTIMIZE module.

Both specific and generic turbine models can be used in the OPTIMIZE module. However, only specific turbine

models based on the Central Composite Design (CCD) are available.

Exports and Result-to-file 81

5.4 Exports and Result-to-file

All result tables inside each of the SITE COMPLIANCE check calculations and LOAD RESPONSE results may

be copied and pasted to, e.g., Excel (see the figure below).

Figure 110. Example of copying data from the results table of a check. Notice the options to copy only the

selected part or to copy all of the table including all expandable levels.

Another way of exporting results and data from the module is available once a calculation is completed and

closed by clicking OK. The user may right-click on the SITE COMPLIANCE calculation and select the Result-to-

file option. This action results in a rather long list of export options available (see below).

Figure 111. Result-to-file menu accessed via right-clicking on the calculation.

Exports and Result-to-file 82

The following outlines each export option in SITE COMPLIANCE along the given requirement before the option

is available.

• WAsP data for each site

o Description: Includes position, sector-wise Weibull parameters, sector frequencies, and shear

exponent with and without displacement height for selected mast(s) and WTG(s). These

values are inputs to the SITE COMPLIANCE propagation model and may differ from values in

result tables.

o Requirements: WAsP selected in flow model(s) and calculation completed.

• WAsP-CFD weibull data for each site or Flowres weibull data for each site

o Description: Contains similar information as the WAsP data for each site export.

o Requirements: WAsP-CFD selected in flow model(s) or using Flowres CFD results and

calculation completed.

• WEng data for each Site

o Description: Includes position, sector-wise Weibull parameters, sector frequencies,

turbulence components, turbulence intensity, turbulence correction factor, and shear exponent

with and without displacement height for selected mast(s) and WTG(s). These values are

inputs to the SITE COMPLIANCE propagation model and may differ from values in result

tables.

o Requirements: WEng selected in flow model(s) and calculation completed.

• CFD data for each site or Flowres data for each site

o Description: Includes height, sector-wise wind speed, wind direction, speed up, mesoscale

roughness, flow inclination, and turbulence intensity for selected mast(s) and WTG(s).

o Requirements: WAsP-CFD selected in flow model(s) or using Flowres CFD results and

calculation completed.

• Statistics for site mast (all in one file and in separate files)

o Description: Includes position, measurement duration, recovery rate, wind climate

parameters (e.g., Weibull parameters, mean wind speed, turbulence intensity, wind shear, air

density, inflow angle), extreme wind conditions (e.g., 50-year wind), and environmental

conditions (e.g., temperature, lighting rate, seismic hazard).

o Requirements: Mast(s) data used in SITE COMPLIANCE calculation.

• Summary of WTG results

o Description: Provides all relevant information from the SITE COMPLIANCE calculation,

including summaries of mast(s), flow model(s), methods used for IEC checks, and siting

parameters summary for each selected WTG(s).

o Requirements: IEC checks completed in SITE COMPLIANCE.

• XML export to load assessment (3rd party result format)

o Description: Includes position, wind turbine characteristics (e.g., rated power, rotor diameter),

extreme and normal wind conditions for selected WTG(s). Can be used in the LOAD

RESPONSE module for fatigue load assessment.

o Requirements: IEC checks completed in SITE COMPLIANCE.

• IEC 61400-15-1 Standard Export (JSON)

o Description: Contains input parameters for site suitability analyses as per the preliminary IEC

standard 61400-15-1. The export is formatted as a JSON file, subject to future modifications

as the standard evolves.

o Requirements: IEC checks completed in SITE COMPLIANCE.

Reports and printing 83

5.5 Reports and printing

Main result

Figure 112. The main page of the report highlights the conclusion of the site assessment and gives an overview

of the results on a park level in the table.

Reports and printing 84

WTG results

Figure 113. Second report page summarizing the results of all the main checks (columns) for each WTG (rows).

Reports and printing 85

Details and Assumptions

Figure 114. The Details and assumptions page which summarizes the most important of the general assumptions

in the setup of SITE COMPLIANCE.

Reports and printing 86

Detailed results and setup (available for each check)

For each of the checks included and calculated, a separate detailed report is available. The detailed report for

each check summarizes the selected method, the available methods for that check, and the results in a graphical

and tabular format. Below is shown the detailed report (two pages) for the Extreme wind check.

Figure 115. First page of the Extreme wind check, detailed report.

Reports and printing 87

Figure 116. Second page of the Extreme wind check, detailed report.

The remaining IEC checks have similar “detailed reports” which are not shown in this manual.

Reports and printing 88

LOAD RESPONSE

Figure 117. Main overview report page of the LOAD RESPONSE report.

Reports and printing 89

Figure 118. Second page of the LOAD RESPONSE report.

Reports and printing 90

Figure 119. Third page of the LOAD RESPONSE report.

References 91

5.6 References

[1] IEC 61400-1 ed. 3, 2005, Wind turbines – Part 1: Design requirements.

[2] IEC 61400-1 ed. 3, 2010, Amendment 1.

[3] Gumbel, E., 1958, Statistics of Extremes, Columbia University Press.

[4] Abild, J., Andersen, E. Y. and Rosbjerg, D., 1992, The Climate of Extreme Winds at the Great Belt, Denmark,

Journal of Wind Engineering and Industrial Aerodynamics, vol. 41-4, p. 521-532.

[5] Cook, N., 1982, Towards better estimation of extreme events, Journal of Wind Engineering and Industrial

Aerodynamics, vol. 9, p. 295-323.

[6] Makkonen, L., 2007, Problems in the extreme value analysis, Structural Safety, vol. 30, p. 405-419.

[7] Winklaar, D. (ed.), 1998, European Wind Turbine Standards II, part I: Load Spectra and Extreme Wind

Conditions.

[8] Bergström, H., 1992, DISTRIBUTION OF EXTREME WIND SPEED, Wind Energy Report WE 92:3,

Department of Meteorology, Uppsala University.

[9] Larsén, X. G. and Mann, J. 2009, Extreme winds from the NCEP/NCAR reanalysis data. Wind Energy, vol

12, p556-573. DOI: 10.1002/we.318.

[10] Svenningsen, L. et al., 2010, An Index-like correction to reduce uncertainty of extreme wind estimates from

short-term data, Proceedings of the European Wind Energy Conference.

[11] Unpublished Monte Carlo study of extreme wind estimation presented at Vindkraftnet meeting at Risø/DTU

2010.

[12] Cook, N. J., 1990, The Designer's Guide to Wind Loading of Building Structures, Butterworths. (Book)

[13] Frandsen, S. T., 2007, Turbulence and turbulence generated loading in wind turbine clusters, Risø report

R-1188.

[14] Nielsen, M., WAT tool Help file, 2011, Risø/DTU.

[15] Nielsen, M., Jørgensen, H. E. and Frandsen, S. T., 2009, Wind and wake models for IEC 61400-1 site

assessment, Proceedings of the European Wind Energy Conference (EWEC 2009).l

[16] Giardini, D., Grünthal, G., Shedlock, K. M. and Zhang, P., 2003, The GSHAP Global Seismic Hazard Map.

In: Lee, W., Kanamori, H., Jennings, P. and Kisslinger, C. (eds.): International Handbook of Earthquake &

Engineering Seismology, International Geophysics Series 81 B, Academic Press, Amsterdam, 1233-1239.

[17] NASA, Global Hydrology and Climate Center. http://thunder.nsstc.nasa.gov/data/index.html.

[18] Troen, I. and Petersen, E. L., 1989, European Wind Atlas, Risø National Laboratory. (Book)

[19] ASTM E1049-85 (2011): ASTM Standard, Standard Practice for Cycle Counting in Fatigue Analysis,

ASTM E1049-85, 2011.

[20] Miner, M. A. (1945): Cumulative Damage in Fatigue, Journal of Applied Mechanics-Transactions of the

ASME 1945, vol. 12(3), p. A159-A164.

[21] DIN 3990-6 (1994): Deutsche Norm: Tragfähigkeitsberechnung von Stirnrädern, Teil 6: Betriebsfestigkeits-

rechnung, DIN 3990-6, 1994.

[22] Box & Wilson (1951): On the Experimental Attainment of Optimum Conditions, Journal of the Royal

Statistical Society, Series B (methodological), vol. 13, p. 1-45, 1951.

[23] IEC 61400-1 ed. 2, 1999, Wind turbine generator systems – Part 1: Safety requirements.

[24] IEC 61400-1 ed. 4, 2019, Wind turbine generator systems – Part 1: Safety requirements.

[25] Justus, C. G., et al., 1978, Methods for Estimating Wind Speed Frequency Ditributions, Journal of Applied

Meterology, vol. 17.

[26] Winterstein, S. R., et al., 1993, Environmental Parameters For Extreme Response: Inverse FORM With

Omission Factors, ICOSSARS Proceedings, Innsbruck, Austria.

[27] DNVGL-ST-0262, 2016, Lifetime extension of wind turbines, DNV-GL A/S, URL:

https://rules.dnv.com/docs/pdf/DNV/ST/2016-03/DNVGL-ST-0262.pdf

[28] Joint Committee for Guides In Metrology, “Evaluation of measurement data - Guide to the expression of

uncertainty in measurement,” Int. Organ. Stand. Geneva ISBN, vol. 50, no. September, p. 134, 2008.

[29] Jonkman, J. et al., 2009, Definition of a 5-MW reference wind turbine for offshore system development,

NREL/TP-500-38060.

[30] Rinker, J. and Dykes, K., 2018, WindPACT Reference Wind Turbines, National Renewable Energy

Laboratory.

[31] Slot, R. M. M. S. et al., 2018, Consistent direct-drive version of the NREL 5MW turbine, WindEurope

Conference Proceedings 2018.

[32] Bak, C. et al, 2013, Description of the DTU 10 MW Reference Wind Turbine, DTU Wind Energy Report-I-

0092.

[33] Slot, R. M. M. S. et al, 2018, Directional fatigue accumulation in wind turbine steel towers, Journal of

Physics: Conference Series (2018) 1102 012017.

[34] Knapp, K. R. et al., 2018, International Best Track Archive for Climate Stewardship (IBTrACS) Project,

Version 4, NOAA National Centers for Environmental Information, doi:10.2592182ty-9e16/

References 92

[35] World Meteorological Organization (WMO): Harper, B. A., Kepert, J. D. & Ginger, J. D., Guidelines for

converting between various wind averaging periods in tropical cyclone conditions, WMO/TD No. 155.

[36] Larsen, X. G., 2012, Recipes for correcting the impact of effective mesoscale resolution on the estimation

of extreme winds, Journal of Applied Meteorology and Climatology, vol. 51, p. 521-532.

[37] Slot, R. M. M., Sørensen, J. D., Sudret, B., Svenningsen, L., & Thøgersen, M. L. (2020). Surrogate model

uncertainty in wind turbine reliability assessment. Renewable Energy, 151, 1150–1162.

https://doi.org/10.1016/J.RENENE.2019.11.101.

[38] C. E. Rasmussen and C. K. I. Williams, Gaussian Processes for Machine Learning. MIT Press, 2006

Appendix I - Gumbel’s Theory of Extremes and more 93

Appendix I - Gumbel’s Theory of Extremes and more

Emil J. Gumbel’s model of extremes published in “Statistics of Extremes” [3] in 1958 is the classical standard

model for describing the statistics of extreme events. The model is also called Fisher-Tippett Type 1 asymptote

or Generalized Extreme Value model (GEV) type 1. The Gumbel model describes the distribution function of

annual extremes, i.e., the cumulative probability, G(u), that a yearly maximum wind speed u is not exceeded and

takes the form:















Where the parameters β and α, are called mode and dispersion or, sometimes, location and scale.

There is a simple relation between the Gumbel distribution of annual extremes and the cumulative mean

distribution of all wind speeds samples, F(u). The mean distribution, also called the parent distribution, is typically

assumed to be Weibull. The simple relation states that the probability that a given extreme wind speed is the

largest among N samples is given as the cumulative parent distribution multiplied by itself N times. For large N,

the exact distribution of annual extremes converges asymptotically to the Gumbel distribution:

























The error of using the Gumbel asymptote is related to the number of independent samples in a year as well as

the k-parameter of the Weibull parent distribution (i.e., the tail behaviour). For lower k-parameters the rate of

convergence (i.e., more accurate already for small N) is much faster than for higher k-factors. It is also important

to note that all N samples are assumed to be independent, i.e., not correlated. In real life, this is not the case for

all 10-minute or hourly wind speed samples in a year. The demand for independence leads to considerable

complexity. Thus, usually the Gumbel distribution is not estimated from the parent distribution, but rather, directly

from extracted extreme samples of the time series and then combining these with theoretical estimates of the

cumulative probability of non-exceedance (P

≈G(u

)) called “plotting positions”. A Gumbel distribution is then

fitted to (u

, P

) to obtain the Gumbel distribution parameters, α and β.

In the original Gumbel approach, the annual maximum samples (u

) are extracted for each year of an N year

time series. These extreme samples are then ranked (i) smallest to largest (i equals 1 to N) and attributed the

plotting positions, P

, the theoretical estimates of the cumulative distribution function, approximating the

probability that the annual maximum wind speed u

is not exceeded. Several formulas for plotting positions have

been suggested. The original Gumbel formula is:

















Which, in fact, was introduced by Weibull (Makkonen, 2004) [6].

An alternative plotting position is due to Hazen (Makkonen, 2004) [6], which is used in the extreme wind plots

in, for example, most Risø/DTU software (e.g., WAsP Engineering):







 



The extreme wind speed samples, u

, are then plotted on the x-axis versus a transform of the chosen P-values,

called the reduced variate, y:















With this transform the Gumbel asymptote takes a very convenient linear form:





















Thus, a linear fit to a plot of (u

, y

) provides the parameters of the Gumbel distribution.

The IEC design criterion for extreme wind speed is a 50-year event, where “50” is referred to as the return period.

In other words the design criterion is the wind speed that is expected to occur only once in 50 years. The return

period is related to the annual risk of exceedance (R) via:











  















Thus, for T=50 years the annual risk of exceedance is R=2%.

Appendix I - Gumbel’s Theory of Extremes and more 94

Once the Gumbel parameters, α and β, have been obtained from a linear fit to (u

, y

) the extreme wind estimate

for T=50 years (i.e. R=2%) may be obtained from:

















  





  

Where y(T=50) equals 3.9. Thus, to obtain the estimate of the 50-year wind speed, the linear Gumbel fit of (u

) must be extrapolated to y=3.9.

Implicit assumptions in the choice of “plotting positions”

The plotting position associated to each of the extracted extremes is a theoretical estimate of the annual

probability that this wind speed is not exceeded. As such, this probability also implies an assumption of the return

period of the highest of the extracted wind speeds.

For the classical Gumbel plotting position, the max wind speed extracted is assumed to have a return period of:









  







 





  

Using the Hazen plotting positions, the same max wind speed is assumed to have the return period:









  







 







As an example with 10 years of data (i.e., N=10), employing the classical Gumbel plotting positions relies on the

assumption that the overall maximum wind speed recording has a return period of 11 years. For the Hazen

plotting positions, the assumed return period is 20 years. A more extreme example is a time series of 25 years.

Using the Hazen estimates assumes the max recording to have a return period of 50 years. Thus, it is obvious

that the Hazen plotting positions are much less conservative than those of the classical Gumbel method.

Extreme wind speed estimate at return period 1 year

The IEC standard also mentions that the extreme wind speed for T=1 year must be estimated although it is not

used directly in the extreme wind check. However, the exact expression above for y(T) is not defined for T=1

year. Instead, the most likely extreme to encounter in any given year is usually chosen as the most appropriate

estimate; this value equals the mode of the Gumbel distribution, i.e., the parameter β from the linear fit (which

does not exactly equal the mean as the distribution is not symmetric). So, we employ the definition of the extreme

wind estimate with return period of 1 year as:











This definition is consistent with the equation derived using a Poisson process (see for example, [4]).

Fitting the Gumbel asymptote

The linear fit to (u

, y

) described above represents the basis of Gumbel’s asymptotic model of extremes.

However, this linear fit may be performed in various ways. Firstly, it is worth noting that the fit is performed on

, y

), i.e., with the reduced variate as the dependent variable. The reason for this is the implicit assumption in

the standard least-squares fitting routine that the dependent variable (here, y) has much higher uncertainty than

the independent one (u). The argument is that the wind speeds are measured using high quality equipment,

whereas y (reduced variate) is a transform of a theoretical estimate of the annual probability of each wind speed

not being exceeded, which is associated with considerable uncertainty.

The standard fit is performed using the least-squares method. Monte Carlo simulations (not published) have

shown that typically this fit introduces a slight conservative bias.

An alternative fit is done using the Probability Weighted Moments, PWM (Abild, 1992) [4], which only takes the

ranked wind speeds as input, and, hence, does not utilize the reduced variate. In this way, the PWM fit avoids

the main source of method-induced bias. The PWM expressions for the fit parameters to the Gumbel asymptote,

scale (α) and location (β) are:



















and 



 

With estimates of the sample probability weighted moments given as:

Appendix I - Gumbel’s Theory of Extremes and more 95





















and 























Monte Carlo simulations (not published) have shown that the PWM fit to the Gumbel asymptote does not

introduce a bias in the Gumbel fit. Unfortunately, the PWM fit does not work equally well with all the ways of

extracting the extreme samples. It seems that PWM only is bias-free for the traditional Gumbel approach where

only the annual extremes are extracted.

Annual maximum method (AM)

The traditional Gumbel method only extracts the most extreme sample of each year, or from, alternatively, the

most extreme sample of each period of fixed length sub-dividing the time series. Hence, the method is referred

to as the Annual Maximum method (AM) or Periodical Maximum method.

The drawback of the AM method is the requirement of relatively long time series for the fit to the Gumbel

asymptote to be meaningful. Typically, at least 5-10 years is recommended to constrain the fit parameters

reasonably well.

In SITE COMPLIANCE, at least 5 years of data are required for the AM method to be available.

Fit

The PWM fit is used with the AM method as it guarantees the least bias in the fitting. Since the PWM fit does

not require plotting positions, no Gumbel plot is needed. But is simply used for visual presentation.

Peak-Over-Threshold method (POT)

In some applications, this method is also referred to as method independent storms. In most applications, 5-10

years of on-site measurements are rarely available, and, within each year, there may be more than one

significant storm event. Hence, a group of extreme wind methods have been developed which utilize more than

a single storm from each year. These methods are referred to as Peak-Over-Threshold methods. Storms are

typically extracted by defining a high threshold to select only high wind events which exceed this threshold. To

ensure that the storm events are statistically independent events, a minimum time difference is required between

the extracted events, typically a few days. The extracted extreme samples may then be analysed in a way very

similar to that of the standard AM Gumbel approach.

Normally, the recommendation for POT methods is given as the number of events to be extracted as 20-50

extremes. This makes the selection of a proper threshold an iterative procedure. As a more efficient way of

extracting the extreme samples in SITE COMPLIANCE, we have introduced a variation of the POT method which

we call POT-N. Instead of defining a threshold, the wished number of extremes is defined directly, and the

program, then, internally selects the proper threshold to obtain this number of extremes.

As in the AM method, the extracted extreme samples are ranked and the “plotting position” (P

) is attributed to

each of the extracted extremes, i.e., the theoretical estimate of the probability of not being exceeded. For POT-

N we have decided to use the classical Gumbel plotting positions in SITE COMPLIANCE.

Instead of a “storm rate” of just one storm/year, as in the AM method, the storm rate is λ storms/year in a POT-

estimation. Thus, a direct Gumbel fit to the extracted extremes would not yield the distribution of annual

extremes, but simply the distribution of the extracted storms. To compensate for this, the plotting positions, P

may be raised to the λth power, provides an estimate of the PDF of the annual extremes (see Cook, 1982 [5]).

This transformation is equivalent to a simple shift on the y-axis, i.e., the standard reduced variates are shifted

by ln(λ):









 

After this transform, the POT Gumbel plot is fully equivalent to the AM plot, with 



 for T=50 years.

Fit

Our studies have shown that the PWM fit does not work well for the POT method as for the AM, unfortunately.

Instead, a linear least squares fit to the (u,y) is used. The classical Gumbel plotting positions are used as the

implicit assumption of return period of the max wind recording seems more sensible than for the Hazen plotting

positions.

Appendix I - Gumbel’s Theory of Extremes and more 96

Weibull parent (EWTS/Bergström) method

The occurrence of high extreme events is closely linked to the tail behaviour of the wind speed distribution. The

heavier the tail the more likely are high extreme events to occur. For Weibull distributions commonly adopted in

wind energy, the shape of the tail is determined by the Weibull shape or k-parameter. A lower k-parameter

means a heavier tail and that extreme events are more likely.

This effect has been quantified in the European Wind Turbine Standard (EWTS), which includes a method for

extreme wind estimation based on the “Parent”-distribution - in this case, the Weibull distribution. The method

simply assumes a universal number of independent extremes per year (N). The so-called “exact distribution” of

the annual maximum is then obtained by raising the Weibull cumulative distribution function to the power of this

number, N.

There is an error in the EWTS publication in the number of independent samples which they set to 23,037 per

year with reference to Bergström (1992) [8]. However, in Bergström (1992) [8], the correct number for 10-minute

data is n=2,302 independent samples per year, or around every 20

10-minute sample. For hourly-averaged

data, the number is 883 or approximately every 10

hourly sample. The error arises due to an exponent of

effective frequency which is incorrectly transferred by a factor of 10 in EWTS.

The slope and offset of the Gumbel asymptote (for high n) to the “extracted distribution” of annual extremes are

given as (Bergström, 1992 [8], EWTS, 1998 [7]):

























The difference between the “exact” and Gumbel asymptote is not significant, and working with the Gumbel

asymptote allows a fully consistent plotting with the other extreme wind estimation methods.

Omni-directional or sector-wise

The EWTS/Bergström method requires omni-directional Weibull parameters. In the WAsP context, Weibull

parameters are sector-wise, which is much more realistic and allows for multimodal omni-directional total

distribution (several peaks). However, an omni-directional Weibull distribution called “Combined” may be

estimated from the sector-wise Weibulls according to the method in the European Wind Atlas [18].

Fit - is the WAsP Weibull fit appropriate for extreme wind estimation?

The WAsP-type Weibull fit, fits exactly the third moment (energy) and frequency above the mean speed of the

table data (no power curve or truncation is applied). Thus, the WAsP fit has a very strong emphasis on the tail

behaviour. This is in contrast to ordinary least-squares or maximum-likelihood fits, that fit the wind speeds (and

not the energy). These fits tend to fit well around the mean where the highest frequencies of occurrence are, at

the cost of reproducing the tail behaviour less well. In conclusion, the WAsP Weibull-fit is in fact better than most

other fits at reproducing the right tail behaviour, which is of main importance in extreme wind estimation.

Preconditioning

The Gumbel distribution is an asymptotic distribution. As the number of independent (i.e., not correlated) samples

in the pool from which the extremes are extracted, e.g., 1 year, approaches infinity, the Gumbel asymptote

becomes exact. The accuracy of the asymptotic assumption depends on the number of independent samples

but also on the shape of the parent distribution, i.e., the Weibull distribution. For a k-parameter of 1, the

convergence is extremely fast and the asymptote practically exact for just few samples. For higher k-factors the

convergence is much slower (see Cook, 1982 [8]).

The deviation of the true annual extreme distribution from the Gumbel asymptote is a slight curvature of the

extreme samples when plotting the reduced variate, y, on the y-axis versus wind speed on the x-axis. This

curvature will be curved downwards (i.e., concave) and generally results in a conservative fit (over-estimation)

which is further exaggerated upon extrapolation to high return periods like 50 years (y=3.9) and higher.

A possible solution is to precondition the data before fitting the slope and offset. The wind speeds are

transformed so that the parent distribution becomes a Weibull with a k-parameter of 1 for which the convergence

is extremely fast, and, thus, the Gumbel approximation is always very good (Cook, 1982 [8]). To achieve this,

the wind speeds of the extreme samples are simply raised to the power of k, where k is the parent Weibull

distribution. Often k=2 is used as a common assumption in wind energy. In addition, using k=2 makes the

transformed wind speeds proportional to the dynamic pressure, related to the thrust exerted by the wind.

However, the real argument for preconditioning is purely statistical and is illustrated in the graphs below.

Appendix I - Gumbel’s Theory of Extremes and more 97

Figure 1. Illustration of the asymptotic nature of the Gumbel model. In both graphs, blue curves show the exact

distribution for an annual number of independent samples of N=10

to 10

in steps of 10. Red curves show the

Gumbel asymptote assuming N is infinite (hidden behind the blue curves on the right graph).Note that as N

increases the blue curves converges to the red. Left graph illustrates the situation for k=2 and the right graph

for k=1, which is equivalent to using preconditioning.

0 5 10 15 20 25 30 35 40 45 50

-2

-1

k=2 (no preconditioning)

u [m/s]

y=-log(-log(G(u))))

0 5 10 15 20 25 30 35 40 45 50

-3

-2

-1

k=1 (pre-conditioned)

u [m/s]

y=-log(-log(G(u))))

Appendix II - Frandsen Effective turbulence model 98

Appendix II - Frandsen Effective turbulence model

The following text describes the main assumptions and steps in the implementation of the Frandsen model or

Effective turbulence model. Most assumptions are directly specified in the IEC standard [1, 2] or in Frandsens

original publication [13].

The Frandsen model [1, 2, 13] defines the so-called effective turbulence as a combination of ambient and

wake generated turbulence integrated overall directions in a way that accounts for accumulation of fatigue

using material properties. In the edition specified in the IEC61400-1 ed. 3 2010 amendment [2], the effective

turbulence is calculated using the 90th percentile of ambient turbulence.

The illustration below gives a simplified overview of the calculation steps of effective turbulence for a WTG.

Ambient σ (P90) Wake added σ “Total” σ

“Total” σ Effective σ

Figure 1. Simplified illustration of the main calculation steps in the Frandsen effective turbulence model. m is

the Wöhler exponent.

For each WTG position in the calculation the Frandsen model needs the following input:

1.  and 



 - Ambient turbulence (mean & st.dev. functions of direction and speed)

2. 











 - Sector-wise frequencies and Weibull distributions

3. 



(turbine thrust curve) and park geometry

4.  - Relevant material fatigue property “Wöhler exponent

Input 1 is used to calculate the ambient characteristic turbulence, i.e., the 90th percentile.

Input 2 is used to calculate the directional wind sped distribution conditioned on wind speed.

Input 3 is used to calculate the wake generated contribution to turbulence.

Input 4 is used in the fatigue weighted combination model of single directions to obtain an omnidirectional

effective turbulence as a function of wind speed only.

A main decision in SITE COMPLIANCE regarding the implementation of the Frandsen has been working

directly with standard deviations of wind speed (σ) instead of turbulence intensity (TI), which is more common.

There are several arguments supporting this decision. Firstly, Frandsen’s original publication [13, p. 84] states

directly that “The model is expressed in terms of standard deviation of wind speed fluctuations rather than

turbulence intensity”. Secondly, a fundamental assumption in Frandsen’s model is that loads are proportional

to σ, the standard deviation of wind speed. Thirdly, the variation of σ with wind speed assumed in the IEC

design limit (“Normal turbulence model”) is linear for (u,σ), but not for (u,TI).

dθ

∫

0°

360°

Appendix II - Frandsen Effective turbulence model 99

The implementation of Frandsen’s model as specified in [1, 2] includes on a number of assumptions and

calculation steps. The main assumptions in the SITE COMPLIANCE implementation are (most are directly given

in the standard [1, 2]):

• “No reduction of mean wind speed inside the wind farm shall be assumed” (no wake deficit!) [1, 2]

• Only wake from nearest neighbour WTG considered when multiple wakes overlap [1, 2]

• Turbulence structure correction parameter is applied to both σ and σ

• Wakes have a fixed angular width of 22° independent of distance [13, 15]

The main steps in the calculation of effective turbulence (as illustrated in figure 1) are:

• “Total” turbulence (σ

) is calculated in each direction combining measured 90th percentile of ambient

turbulence (σ

) and calculated wake added turbulence (σ

wake

) [1, 2] - “^” indicates measured data:















 





where





  



(90th percentile of ambient turbulence

)

and

















 

 



 

 



















• “Effective turbulence” is calculated from “Total turbulence” raised to the power of m (Wöhler exponent)

and integrated (numerically) over all directions weighted by its relative frequency (f):





































 



The Frandsen model requires a “large wind farm correction” when certain conditions are met, but [1, 2] only

considers the special case of a regular rectangular layout. In SITE COMPLIANCE, a more general sector-wise

version of this large wind farm correction has been implemented. For each sector, the following evaluation is

performed:

• If >5 neighbour WTGs in a sector, the sector is a “large wind farm sector”:

• In direct wake directions σ

is calculated as above (no correction of ambient level) [1, 2]

• In non-wake directions (>10RD), ambient σ

is adjusted using [1, 2]:





















 



   



where















































The latter fraction on the right in the above equation was proposed in [15] as a more generally applicable

version of the fraction on the left, which is given explicitly in the standard, covering only regular rectangular

layouts. The square root argument represents a “thrust versus area”. In the fraction on the right, the

rectangular expression is replaced with an angular expression with the area of an “pie slice” 





sec

is the number of WTGs within the “slice” or sector, and d

max

the radius of the slice and, thus, distance to

the furthest neighbour WTG in the sector.

Calculation of Equivalent effective turbulence (not part of the IEC standard)

This calculation is not part of the IEC standard but is based on considerations presented in Frandsen’s original

paper [13]. The Equivalent effective turbulence is used to decide when the calculated Effective turbulence

For complex terrain 



is multiplied by the Turbulence Structure Correction Parameter, C

Appendix II - Frandsen Effective turbulence model 100

exceeds the IEC design limit to evaluate if this exceedance is critical or not. The main assumption is the same

as in the Frandsen model, but an extra calculation step is added: integrating the effective turbulence over all

relevant wind speeds, whereas Frandsens model only integrates over directions. This second integration over

wind speeds assumes that the sensitivity of the WTG to wind speed fluctuations is constant for the relevant wind

speed range, which is a significant extension to the approximation of the Frandsen model. The Equivalent

effective turbulence is calculated as following:





 





















Where m is the Wöhler exponent, f(u)

total frequency of a wind speed bin (omnidirectional) and σ

eff.

(u) is

effective turbulence as a function of wind speed bin, i.e., the result from the Frandsen models effective turbulence

calculation.

The effective turbulence result for a particular WTG is compared to the IEC design limit for the WTG class by

calculating the equivalent effective turbulence for the relevant IEC design class (turbulence and frequency

values) and for the actual WTG results. These results for the actual WTG result are normalized by the result

for the IEC class. If the normalized result (ratio) exceeds 1, the IEC exceedance is considered critical. The

integration limits are set to match the IEC check interval.

Appendix III - Critical, Caution & OK limits in: IEC ed. 2 / ed. 3 / ed. 4. 101

Appendix III - Critical, Caution & OK limits in: IEC ed. 2 / ed. 3 / ed. 4.

The following values are used in the windPRO SITE COMPLIANCE module to help the user evaluate if an

obtained check value is critical or not. They are based on experience from a large number of projects but further

validation and improvements are on-going.

General decisions:

For all checks, the result of a calculation is green/“Ok” if the result is fully within the IEC limits. If the result for a

check exceeds the IEC limit, the result will be yellow/“Caution” or red/“Critical”, depending on the degree of

exceedance. Caution is used when the exceedance is not considered critical.

At WTG level:

• A WTG is set to Critical if just one check is Critical, Caution if just one check is Caution, and only Ok if

all checks are Ok.

At Park level:

• A check is set to Critical for the park if just one WTG is Critical for the check, it is Caution if just one

WTG is Caution and only Ok if all WTGs are Ok.

Assessment of Critical, Caution and Ok for each Site Compliance check (ed. 3):

Terrain complexity (I

IEC limit: (none/see details of terrain check)

• Critical: Never

• Caution: If I

> 0

• Ok: if I

= 0

Wind shear (α):

IEC limits: 0≤α≤0.2

• Critical: if α>0.3 or α<0

• Caution: if 0.3≥α>0.2

• Ok: if 0≤α≤0.2

Air density (ρ):

IEC limits: 1.225 kg/m

is assumed in design

• Critical: Never

• Caution: ρ>1.225 kg/m

• Ok: ρ≤1.225 kg/m

Inflow angle (φ

max

IEC limits: φ≤8° & φ≥-8°

• Critical: φ>12° or φ<-12°

• Caution: 12°≥φ>8° or -12°≤φ<-8°

• Ok: φ≤8° & φ≥-8°

Extreme wind (u

50y

IEC limits: V

ref

≥u

50y

• Critical: u

50y

>Vref or gust >1.4*Vref

• Caution: Never

• Ok: V

ref

≥ U

50y

or 1.4*Vref ≥ gust

Effective turbulence (σ

eff.

(u)):

IEC limits: 

(u) > 

eff. WTG

(u) for all u in check interval

• Critical: 

eq, WTG

> 

eq, IEC

• Caution: 

eq, WTG

< 

eq, IEC

• Ok: 

eff, IEC

(u) > 

eff. WTG

(u) for all u

Appendix III - Critical, Caution & OK limits in: IEC ed. 2 / ed. 3 / ed. 4. 102

eq, XXX

is the Equivalent effective turbulence for WTG for IEC design class. See Appendix II for a description of

the calculation of this quantity.

Wind distribution (pdf(u)):

IEC limits: f

IEC

(u) > f

WTG

(u) for all u in check interval

• Critical: 



 or (



 



)

• Caution: 



 & (



 



)

• Ok: f

IEC

(u) > f

WTG

(u) for all u

Where:





 









 























 









 





















The main point here is that exceedances at the lower half of the check

interval is not as severe as those in the upper half at higher wind speeds

where loads are expected to be higher.

Seismic hazard (PGA):

• Critical: PGA > 2.4

• Caution: 2.4 ≥ PGA > 0.8

• Ok: 0.8 ≥ PGA

Lightning rate:

• Critical: rate > 20

• Caution: 20 ≥ rate > 10

• Ok: 10 ≥ rate

Temperature range (T):

Normal range

• Critical: h outside > 240h

• Caution: h outside > 24h

• Ok: 24 h ≥ h outside

Extreme range

• Critical: h outside > 1h

• Caution: 1h ≥ h outside > 0.0 h

• Ok: 0.0 h outside

Assessment of Critical, Caution and Ok – differences for IEC ed. 2:

Only the Effective turbulence check and the Wind distribution check have different OK, Caution and Critical

limits in IEC ed. 2 compared to IEC ed. 3.

Effective turbulence (σ

eff.

(u)):

IEC limits: 

(u) > 

eff. WTG

(u) for u=15m/s

• Critical: 

eq, WTG

> 

eq, IEC

• Caution: 

eq, WTG

< 

eq, IEC

• Ok: 

eff, IEC

(15m/s) > 

eff. WTG

(15m/s)

eq, XXX

is the Equivalent effective turbulence for WTG for IEC design class (See Appendix II for a description of

the calculation of this quantity).

Appendix III - Critical, Caution & OK limits in: IEC ed. 2 / ed. 3 / ed. 4. 103

Wind distribution (pdf(u), with Weibull A, k and mean V

ave

IEC limit: not explicit, but Compare design and WTG distributions

• Critical: f

IEC

(15m/s) < f

WTG

(15m/s)

• Caution: f

IEC

(15m/s) > f

WTG

(15m/s)

• Ok: V

ave, IEC

> V

ave, WTG

& k

IEC

< k

WTG

The main point here is that the design climate assumes a Weibull distribution with k

IEC

=2 and average wind

speed defined for each design class. Higher average wind speeds will lead to increased loads as will a lower

Weibull k parameters. As the wind speed 15m/s plays a special role in the ed. 2, the limit from caution to

critical is defined by the frequency of occurrence at 15m/s.

Assessment of Critical, Caution and Ok – differences for IEC ed. 4:

The IEC ed. 4 is mostly consistent with the ed. 3, hence, only the differences for ed.4 check limits are

summarized below.

Terrain complexity (I

IEC limit: (none/see details of terrain check)

• Critical: Never

• Caution: If C

> 0 (complexity category: H, M, L)

• Ok: If C

= 0 (complexity category: N)

Wind shear (α):

IEC limits: 0.05≤α≤0.25

• Critical: If α>0.3 or α<0

• Caution: If OK not fulfilled

• Ok: if 0.05≤α≤0.25

Air density (ρ):

IEC limits: ρ≤1.225 kg/m

• Critical: Never

• Caution: If OK not fulfilled

• Ok: ρ≤1.225 kg/m













 ≤ 1.225 kg/m

when ρ>1.225 kg/m

Wind distribution (pdf(u)):

IEC limits: f

IEC

(u) > f

WTG

(u) for u 









• Critical: 



 or (



 



)

• Caution: If OK not fulfilled.

• Ok: f

IEC

(u) > f

WTG

(u) for u interval &  









 



 









 

Where:





 









 























 









 





















The main point here is that exceedances at the lower half of the check

interval is not as severe as those in the upper half at higher wind speeds

where loads are expected to be higher.

Ambient 90% Turbulence [NTM]:

IEC limits: 

(u) > 

omni,90%

(u) for all u in check interval

• Critical: If OK not fulfilled

• Caution: Never (ultimate loads)

Appendix III - Critical, Caution & OK limits in: IEC ed. 2 / ed. 3 / ed. 4. 104

• Ok: 

















for   



  





Ambient Extreme Turbulence (AET) [ETM]:

IEC limits: 

1,ETM

(u) > 

AET

(u) for all u in operation

• Critical: If OK not fulfilled

• Caution: Never (ultimate loads)

• Ok: 

















for 









Max Centre-wake 90% Turbulence (MCWT) [ETM]:

IEC limits: 

1,ETM

(u) > 

MCWT

(u) for all u in operation

• Critical: If OK not fulfilled

• Caution: Never (ultimate loads)

• Ok: 

















for 









Appendix IV - Theory of LOAD RESPONSE and Fatigue 105

Appendix IV - Theory of LOAD RESPONSE and Fatigue

This appendix describes the fatigue design load cases, which should be considered according to the wind

turbine standard IEC 61400-1 ed. 3 (2010) and ed. 4 (2019) along with the workflow for determining fatigue

loads.

Design load cases

The design load cases in IEC 61400-1 ed. 3 (2010) are developed in order to represent all significant design

situations during a wind turbines design lifetime in both ultimate (extreme) and fatigue loading.

1. Power production

2. Power production plus occurrence of faults

3. Start up

4. Normal shut down

5. Emergency shut down

6. Parked (standing still or idling)

7. Parked and fault conditions

8. Transport, assembly, maintenance and repair

The design load cases which should be considered for fatigue loading are listed in the table below (see IEC

61400-1 ed. 3 (2010) for abbreviations). In LOAD RESPONSE, only design load case 1.2 is considered

explicitly in the response model, since this is normally is the most dominant design load case and highly

dependent on the site specific wind climate parameters. However, the minor contributions from the other DLCs

may be included via the ‘DLC Other’ option from windPRO 3.4. This contribution should represent the

expected combined fatigue contribution within the design lifetime of all the fatigue DLCs ‘other’ than DLC 1.2.

other IEC61400-1 ed. 4 provides guidelines for the number of expected start-ups and shut-downs.

Design situation

DLC

Wind model

Wind speeds

1) Power production

1.2

NTM

< V

hub

< V

out

2) Power production plus occurrence of fault

2.4

NTM

< V

hub

< V

out

3) Start up

3.1

NWP

< V

hub

< V

out

4) Normal shut down

4.1

NWP

< V

hub

< V

out

6) Parked (standing still or idling)

6.4

NTM

hub

< 0.7V

ref

Table 1. Fatigue design load cases in IEC 61400-1 ed. 3 (2010).

Aero-elastic simulations

Wind turbine loads are normally determined based on aero-elastic simulations of the wind turbine during the

different design situations / load cases. The aero-elastic model needs, in general, input concerning:

• Site specific wind climate parameters

o Mean wind speeds

o Turbulence

o Wind shear

o Air density

o Inflow angle

• Structural properties

o Blade properties (length, aero-dynamics, weight, stiffness, etc.)

o Nacelle properties (gearbox, generator, shaft, etc.)

o Tower properties (height, weight, stiffness, etc.)

o Foundation properties (size, stiffness, etc.)

• Control system

o Power curve

o Pitch system

o Actuators

o Control algorithms

Appendix IV - Theory of LOAD RESPONSE and Fatigue 106

In the aero-elastic model, a turbulence generator is used to generate a random turbulence box containing the

3-dimensional wind speed vector at different locations in the rotor plane at different time steps. The turbulence

box is normally generated based on the Kaimal or Mann turbulence model (see [1, 2 ]). From the turbulence

box, the aerodynamic force on the wind turbine at each time step is determined. This is used to determine

time-series for the forces, moments and deflections in different parts of the structure, taking the structural

dynamics and the control system into account.

Fatigue – Structural components

The fatigue loads for structural components (blades, tower, shaft, etc.) are normally determined by rainflow

counting of the simulated time-series (see [19]). The purpose of rainflow counting is to extract the stress

ranges from the time-series, which is the cause of the damage of the structure. In rainflow counting, these

stress ranges are extracted in a similar way as raindrops falling from a roof, hence the name rainflow counting.

Compared to other methods, rainflow counting focuses on extracting the largest stress ranges in the time-

series.

The number of cycles to failure is normally determined based on a SN-curve, which, depending on the

material, can be more or less complex. However, in order to secure an easy interaction between the aero-

elastic model used to determine the overall structural loading on all wind turbine components and the finite-

element models used to determine the local stresses for specific structural details, a linear SN-curve is

normally used. The linear SN-curve describes the relationship between the stress range  and the number of

cycles to failure N, where, also, the Wöhler exponent m is introduced corresponding to the negative slope of

the SN-curve. Note that the term “linear” SN-curve refers to linear in a log-log plot (see the equation and figure

below).

log log log

N N K m



=  = − 



Figure 1. Linear SN-curve in log-log coordinate system.

The total damage D from the individual stress cycles extracted by rainflow-counting is normally determined by

Miner’s rule for linear damage accumulation proposed by MA Miner [20]:

( )







where n

is the number of cycles with stress range 

extracted by rainflow counting and N() is the number of

cycles to failure for stress range 

determined from the SN-curve. If the damage is <1, the applied fatigue

loading will not lead to structural failure, whereas a damage >1 will result in structural failure.

In LOAD RESPONSE, the fatigue load for structural components is expressed as Damage Equivalent Load

(DEL). The DEL is determined as the fatigue load which for an equivalent number of cycles n

will lead to the

same damage as the actual load time-series for the wind turbines lifetime.



























Appendix IV - Theory of LOAD RESPONSE and Fatigue 107

Fatigue – Mechanical components

For mechanical components, the fatigue stresses normally occur due to the interaction between different

gears. This fatigue load is often characterized by the Load Duration Distribution (LDD), which describes how

long time the stresses are at a given level (see, e.g., [21]). The LDD is also dependent on the sampling

frequency f and the number of time steps in the simulations with a given stress level. It is noted that the term

Wöhler exponent also is used for the Load Duration Distribution, even though the same relationship to the

linear SN-curve not is present.

























Surrogate modelling

Fatigue loads in windPRO are approximated using surrogate models. Surrogate modelling involves creating

simplified and computationally efficient models to approximate the complex relationship between input variables,

such as wind conditions, and the resulting fatigue loads on the different turbine components. These models

enable much quicker predictions and assessments compared to full aero-elastic simulations, facilitating a more

efficient design process.

LOAD RESPONSE supports two different types of surrogate models: Central Composite Design (CCD) and

Gaussian Process Regression (GPR). All generic turbine models in LOAD RESPONSE are based on the CCD,

whereas for specific turbine models these can be created for either type.

Central Composite Design

The Central Composite Design, first proposed by Box & Wilson (1951) [22], was initially developed to assess

second order effects in a response where interaction between different variables is present. Therefore, it

constitutes a good candidate for a surrogate model, as is often the case for fatigue loads on turbines.

For each sensor in LOAD RESPONSE, a response surface is fitted at each wind speed bin in order to assess

the fatigue loads accurately. In design load case 1.2, the following four wind climate parameters (variables)

influence the fatigue loads:

• Effective turbulence standard deviation 

(90% quantile – mean wind speed / sector dependent)

• Wind shear  (mean value – sector dependent)

• Inflow angle  (maximum value – sector dependent)

• Air density  (mean value)

The Central Composite Design used in LOAD RESPONSE is, therefore, four dimensional. The figure below

shows the principle in establishing the response surface for a two dimensional case using central composite

design. The black point refers to the reference point (normally, the wind conditions for the IEC-class), the blue

points refer to variations of single wind climate parameters (variables) in order to estimate second order

effects, and the red points refer to variations of several wind climate parameters, simultaneously, in order to

estimate their interaction.

Figure 2. Central Composite Design for two variables.

-2 -1 0 1 2

-2

-1

Appendix IV - Theory of LOAD RESPONSE and Fatigue 108

In order to setup the four dimensional response surface in LOAD RESPONSE, aero-elastic simulations for 25

combinations of the four wind climate parameters are needed at each wind speed bin. The response surface

central composite design has been selected for LOAD RESPONSE, because it provides a reasonable balance

between accuracy and the required number of simulations.

The regression model 







is fitted to the response at the individual wind speed bins using linear least squares

regression, where x denotes the wind climate parameters and



and



are regression parameters.













 











 















 





Gaussian Process Regression

From windPRO 4.1 a more advanced surrogate model using Gaussian Process Regression (GPR) is available

for specific turbine models.

Gaussian process regression is a probabilistic surrogate model which assumes the model output 







to be a

realization of a deterministic mean defined by a regression model 







and a correlated stochastic process 

























 







The first terms 







to is similar to the Central Composite Design a second order regression model fitted using

linear least squares. The second term 







is interpolating the known residuals at the experimental design by a

stationary zero mean Gaussian process fully described by its covariance.



























The overall process variance 



is assumed constant and  models the correlation between 







and 









their inter-distance and the hyperparameters 



. Regression coefficients , process variance  and

hyperparameters  is estimated by optimization, maximizing the likelihood of observing the response at the

experimental design.

The GPR model in windPRO includes mean wind speed at hub height as an additional input variable compared

to the Central Composite Design. Furthermore, the model can utilize random sampling of design points. This

enables much greater flexibility in terms of model accuracy, as a one can increase the sampling rate at different

wind conditions to increase accuracy.

For more information on GPR as a surrogate model for wind turbine fatigue assessment, see [37]. Additionally,

for a more fundamental introduction to the topic of Gaussian Process Regression, see [38].

Results

The damage equivalent load 



 or load duration distribution (



) is determined at each mean wind

speed bin using the given surrogate model. The damage equivalent loads or load duration distribution for the

full wind speed distribution (DEL / LDD) can then be determined by weighting the fatigue loads from the

individual wind speed bins with the wind speed distribution:

( ) ( )

( )

, , , ,

out

hub in

hub hub hub

DEL w V DEL V V

   



( ) ( )

( )

, , , ,

out

hub in

hub hub hub

LDD w V LDD V V

   



where w(V

hub

) is weight factor determined from the wind speed distribution.

Appendix IV - Theory of LOAD RESPONSE and Fatigue 109

The estimated fatigue loads are presented in the following three ways in load response:

• Load (specific wind turbines)

• Load index (generic / specific wind turbines)

• Fatigue lifetime (generic / specific wind turbines)

The value of the damage equivalent load or load duration distribution is shown directly in LOAD RESPONSE

for each sensor when calculations are performed for specific wind turbines implemented by manufactures.

The load index is the main result, both for specific and generic wind turbines. The load index for each sensor is

defined as the ratio between the fatigue load (DEL / LDD), estimated based on the site specific wind

conditions, and the fatigue load (DEL

ref

/ LDD

ref

), estimated based on the wind conditions for the considered

IEC class.











 











 

A load index <100% shows that the site specific fatigue loads are less severe than the fatigue loads for the IEC

wind climate conditions. If ‘DLC Other’ is included in the calculation it will be added as a constant contribution

to both numerator and denominator for the DEL or LDD (shown only for DEL below).















 















 







 

The fatigue lifetime is also determined for each wind turbine component relative to the assumed design

lifetime, which, for the generic wind turbines, is assumed to T

turbine

= 20 years. The fatigue lifetime will,

therefore, be larger than 20 years for sensors with a load index below 100% (below shown without DEL Other

contribution).

ref

lifetime DEL turbine

ref

lifetime LDD turbine

DEL

LDD



=







=





It is noted that the fatigue lifetime does not consider other material degradation phenomena such as abrasion

or corrosion.

Appendix V - Curtailment 110

Appendix V - Curtailment

This appendix describes the curtailment or “wind sector management” calculations in SITE COMPLIANCE and

LOAD RESPONSE. The detailed curtailment calculations require curtailment settings to be defined on the

WTG objects individually and that the Curtailment option has been checked on the main setup tab of SITE

COMPLIANCE.

Curtailment in SITE COMPLIANCE

In SITE COMPLIANCE, the curtailment settings will NOT influence the calculation of the ambient wind climate

parameters. The ambient wind shear or turbulence is the same no matter if the turbine is in operation or not.

However, the wake induced turbulence behind a wind turbine depends strongly on, whether or not the turbine

is operating or shut down (or running in a reduced mode). Hence, in SITE COMPLIANCE, the curtailment will

only influence the turbulence checks which combine both the ambient and wake contributions to turbulence,

‘Effective turbulence’ and ‘Max centre-wake 90% turbulence’.

In the Effective turbulence check, the effect of curtailment is quite simple. For each wake situation between two

turbines, it is considered whether the wake generating turbine is curtailed or not for each direction wind speed

bin. In case the wake generating turbine is curtailed, its wake contribution is omitted for the speed and

direction in question. The wake of any further upstream turbines might become important and contribute to the

turbulence as the Frandsen wake model only considers the wake turbulence of the nearest upstream turbine

(in operation).

The wake signature, sometimes called “view angles”, typically has a fixed width of 22 degrees in the Frandsen

model (see Appendix II - Frandsen Effective turbulence model). Hence, a simplifying assumption is needed

to handle partial curtailment within the direction interval generating the wake. A basic assumption is adopted

saying that each degree where the wake generating turbine is curtailed within the 22 degree directional

interval, leads to one degree less wake at the receiving WTG.

Curtailment in LOAD RESPONSE

The effect of curtailment on the load calculation for a wind turbine is more intuitive and directly influences the

turbine itself in contrast to the effect of curtailment on the wake effects. For directions and wind speeds where

the turbine is shut down, it will experience much reduced fatigue compared to being in operation. On the other

hand, there is a minor contribution to fatigue from the additional starts and stops required from the

implementation of the curtailment rules.

LOAD RESPONSE implements the main fatigue design load case DLC1.2 “Normal operation” using a

response model. Design load cases DLC3.1 “Start up” and DLC4.1 “Normal shut down” are handled via the

accumulated “DLC other” contribution. The effect of additional starts and stops enforced by the curtailment

strategy would require a detailed time series calculation to estimate the needed number of additional stops and

starts. This is not inline with the current ‘statistics’ version of the IEC standard and followingly of SITE

COMPLIANCE and LOAD RESPONSE, but it is a clear goal to move towards time series in the future

versions.

To compensate for the above simplification of not directly accounting for the fatigue of the additional starts and

stops, a simple assumption has been made. Whenever a turbine is shut down, the fatigue load (DEL)

contribution of wind speed and direction bin is not decreased to zero (or nearly zero), but reduced by a fixed

fraction. If the fatigue of the starts and stops was explicitly accounted for, the fraction would be very close to

zero, but since they are not accounted for, the fraction is set to 0.5.









This may seem as a very conservative assumption, but, since the DEL is used in the very non-linear fatigue

calculation (see Appendix IV - Theory of LOAD RESPONSE and Fatigue, page 101) using the Wöhler

exponent, this is not the case. For a Wöhler exponent of m=10, the fatigue damage contribution of a curtailed

wind speed and direction bin is reduced to 0.1% and for m=5 it is reduced to 3%.

The integration of loads in LOAD RESPONSE follows a fixed discretization of 1 degree (centred on half

degrees) and 0.5 m/s centred on X.25 m/s and X.75 m/s. Hence, curtailment settings defined with finer

Appendix V - Curtailment 111

resolution than this discretization e.g. X.1 m/s will be rounded off to the actual discretization, however these

round-off effects will generally be negligible.

Appendix VI - IEC 61400-1 ed. 2 (1999) 112

Appendix VI - IEC 61400-1 ed. 2 (1999)

From windPRO version 3.1, SITE COMPLIANCE supports the second edition of IEC 61400-1 (IEC ed. 2),

which was released in 1999. The second edition was replaced by the third edition in 2005, which was updated

in 2010 (IEC ed. 3). Hence, IEC ed. 2 is, in principle, obsolete, but, nevertheless, it continues to be used by

conservative turbine manufacturers or when new park extensions influence existing ed. 2 turbines, and they

need to be assessed for, e.g., life reduction due to increased wake turbulence.

The general IEC ed. 2 requirements for assessment of a wind turbine using site-specific conditions is defined

in section 11.2 of the standard in the part inserted below.

From [23] IEC 61400-1 ed. 2 (1999).

For many of the wind climate requirements in IEC ed. 2, no explicit or direct calculation guideline is given. The

below text summarizes the decisions made for calculation and evaluation of the IEC ed. 2 checks – the governing

principle has been to re-use as much as possible from the current IEC ed. 3 version.

Terrain complexity

There is no explicit description of how to assess terrain complexity in IEC ed. 2, only that it is required.

We have decided to fully adopt the terrain complexity calculations of IEC ed. 3, to fulfill the IEC ed. 2

requirements.

Extreme wind

The design value and, hence, upper limit of the 50 year extreme wind is Vref for the design class in question as

in IEC ed. 3.

Wind distribution

There is no specific limits or check ranges defined for wind distribution in IEC ed. 2. Since the design climate is

defined by mean wind speed for the wind turbine class and k=2, these values are used as the check limits. (see

Appendix III - Critical, Caution & OK limits in: ).

Appendix VI - IEC 61400-1 ed. 2 (1999) 113

(Effective) Turbulence

The IEC ed. 3 Effective turbulence calculation to combine ambient and wake turbulence has been adopted to

fulfill the IEC ed. 2 requirements. However, it is with the important difference that the 84

percentile of the

ambient turbulence is used for IEC ed. 2 as opposed to the 90

in the IEC ed. 3. The IEC ed. 2 standard only

requires the check performed at 15 m/s with a design value and, hence, upper limit of I15 for the relevant deign

class.

Wind shear

The design value and upper limit of wind shear is set to 0.2 in IEC ed. 2. We have chosen to supplement this by

the additional lower limit at 0.0 defined in IEC ed.3, as very low wind shears are equally damaging as high wind

shears.

Flow inclination

The upper and lower limits of flow inclination are defined as +/-8° from horizontal as in the IEC ed. 3.

Air density

A design air density and, hence, upper limit of 1.225 kg/m

is defined in IEC ed. 2, as in IEC ed. 3.

Appendix VII - IEC 61400-1 ed. 4 (2019) 114

Appendix VII - IEC 61400-1 ed. 4 (2019)

From windPRO version 3.3, SITE COMPLIANCE supports the fourth edition of IEC 61400-1 [24] (hereafter IEC

ed. 4), released in February 2019. Fourth edition is consistent with the third edition in several regards, but at

the same time, it introduces new turbine design classes and new checks for the extreme wind climate. The

terrain complexity check is replaced by a new methodology, but maintains its purpose of deciding the degree

to which ‘turbulence structure correction’ is required for the turbine positions. Another improvement in ed. 4 is a

more explicit split of the wind climate checks in the checks that contribute to fatigue and those from extreme

events that represent ultimate loads. The following text summarizes these main differences in IEC ed. 4

compared to IEC ed. 3, which are relevant in the context of SITE COMPLIANCE and LOAD RESPONSE.

Design classes

The IEC ed. 4 allows the same standard design classes as IEC ed. 3, the wind speed classes I to III and

turbulence classes A to C, with unchanged design parameters Iref, Vref and Vave. In addition, the IEC ed. 4

introduces a new extra high turbulence class “A+”; A new “Tropical” option is also introduced for all design

classes, which replaces the design extreme wind speed “Vref” with the increased “Vref,T” of 57m/s, suitable for

regions prone to tropical cyclones/typhoons. These classes are summarized in Table 1 of the ed. 4 standard,

inserted below.

From [24] IEC 61400-1 ed. 4 (2019): “Table 1 – Basic parameters for wind turbine classes”.

Section 11.2 Terrain Complexity

The updated ed. 4 version of the terrain complexity check changes the way terrain complexity is calculated and

categorized. Sectorial fits for the radius measured in hub heights of 5xHH has been included, with an option to

extend the sectors by 2xHH downwind of the WTG position. In total 1+3x12 = 37 terrain fits must be performed

for the following radii (in HH):

• 5xHH: 12 sectorial fits & 1 omnidirectional fit

• 10xHH: 12 sectorial fits

• 20xHH: 12 sectorial fits

Contrary to the IEC ed. 3, the terrain fits are not constrained to go through the WTG tower base, and the

resolution of the utilized terrain grid is required to be ≤50m.

For each of the 37 fits the slope (θ) along the sector center-line (or gradient for the omni-directional fit) is

estimated as well as the standard deviation of the difference between the fitted plane and the original terrain

heights of the grid points (D

). The slopes and standard deviations are used to calculate the “Terrain Slope

Index” (TSI) and “Terrain Variation Index” (TVI) for the omni-directional fit and across the sectorial fits for each

of the radii, as summarized in the equation (34) in the standard (see below).

Appendix VII - IEC 61400-1 ed. 4 (2019) 115

Equation (34) from [24] IEC 61400-1 ed. 4 (2019), expressions for the TSI and TVI based on sectorial (with

subscript 30) and omni-directional (with subscript 360) terrain fits. f

Energy

is the energy fraction of each sector at

the WTG position based on the sector Weibulls. R is the radius of the fit e.g. 5xHH and k

=5/3 and k

=3 are

constants.

Based the resulting TSI and TVI for each WTG position the complexity category can be found using Table 5 in

IEC ed. 4 (see below). The worst TSI or TVI for a WTG position determines the category and each category is

defined by its lower TSI and TVI thresholds. The categories “L”, “M” and “H” represent low, medium and high

terrain complexity, respectively. We use the label “No” complexity for the category below the “low” category.

Contrary to IEC ed. 3 no complexity index is calculated in the IEC ed. 4.

From [24] IEC 61400-1 ed. 4 (2019): “Table 5 - Threshold values for terrain complexity categories L, M and H”.

Finally, the turbulence structure correction parameter (C

) can be inferred directly from the complexity category

using Table 7 of the IEC ed. 4 standard (see below). In SITE COMPLIANCE this parameter is directly transferred

to the relevant turbulence checks, where it may be replaced by more detailed alternative values if WEng

modelling results are available.

From [24] IEC 61400-1 ed. 4 (2019): “Table 7 – Values of the turbulence structure correction parameter

depending on terrain complexity category L, M and H”.

Section 11.9.2 Fatigue load assessment

The IEC ed. 4 wind climate checks are split into a list of checks for fatigue loads and a list for ultimate loads.

The below list of five checks, a) to e), represent fatigue loads or “normal climate”, were also in IEC ed. 3.

a) Wind Distribution

b) Effective turbulence

c) Flow Inclination

d) Wind Shear

e) Air Density

In these five checks, there are smaller changes to the following checks: Wind Distribution, Flow Inclination, Wind

Shear and Air density, which are briefly described below.

Check (a) Wind Distribution introduces an additional check criterion on the Weibull k-factor, which is summarized

in the standard’s equation (35) and figure 12, shown below. This extra k-factor check penalizes deviations of k

outside a range, which depends on the ratio of the mean wind speed of the WTG and the design class. If this

ratio is 1 or above the criterion is violated. For ratios below 1 the acceptable k-range centered on the design

k=2.0, gradually grows with decreasing mean wind speed ratio. k-factors below 1.4 will also lead to a violation

Appendix VII - IEC 61400-1 ed. 4 (2019) 116

no matter the ratio of mean wind speeds. This additional check makes the Wind Distribution check more

restrictive in IEC ed. 4 as compared to IEC ed. 3.

Equation (35) and figure 12 from [24] IEC 61400-1 ed. 4 (2019), which describe the additional check introduced

on the Weibull k-factor.

Check (b) Effective turbulence has to be checked in the wind speed interval 



to 



. Previously in IEC ed.

3 the interval was from 60% of the wind speed at turbine rated power to the cut out wind speed.

Check (c) Flow Inclination is adjusted to use the energy frequency weighted average inflow angle across sectors

as opposed to the max sector inflow angle used in IEC ed. 3. To prevent the non-conservatism from averaging

high positive and negative inflow angles to zero, in SITE COMPLIANCE we allow the option of using absolute

averages of inflow angles, which also appears more consistent with equation 43 in the terrain complexity check

(see above). The sector energy frequencies are calculated from the sector Weibulls. IEC ed. 4 allows the use of

slopes of the sectorial 5xHH fits as estimates of the sector flow inclinations in ed. 3 the slope of the omni-

directional fit was used for this purpose.

Check (d) Wind Shear is adjusted similarly to the Flow Inclination check to use the energy frequency weighted

average wind shear across sectors instead of the simple average shear as in IEC ed. 3. In addition, the accepted

range of wind shears is adjusted to 0.05 to 0.25 in IEC ed. 4 compared to 0.0 to 0.2 in IEC ed. 3.

Check (e) Air Density is adjusted to allow a correction based on the WTG and design mean wind speeds for

sites where the design air density of 1.225kg/m

is exceeded. This adjustment effectively reduces the WTG air

density by the squared ratio on WTG mean wind speed and mean wind speed of the design class. Hence, a

margin of 10% on the WTG mean wind speed (a ratio of 0.9), translates into a reduction of 20% of the WTG site

air density in the check against the design limit.

Equation (37) from [24] IEC 61400-1 ed. 4 (2019), adjustment criterion for the air density when the design air

density is exceeded.

Section 11.9.3 Ultimate load assessment

The list of ultimate load checks for wind climate is extended to four explicit checks in IEC ed. 4, from only one

explicit check on extreme wind speeds in IEC ed. 3. The three new ultimate load checks all refer to different

aspects of extreme turbulence as seen from the list below.

a) Ambient 90% Turbulence [NTM]

b) Extreme Wind

c) Ambient Extreme Turbulence [ETM]

d) Max Centre-wake 90% Turbulence [ETM]

Appendix VII - IEC 61400-1 ed. 4 (2019) 117

Check (b) Extreme Wind is adjusted slightly compared to IEC ed. 3. Firstly, the WTG extreme wind speed

estimates must be corrected by an additional “safety factor” when the coefficient of variation of the WTG extreme

wind Gumbel distribution exceeds 15%. Secondly, the WTG 50year extreme wind speed estimates may be air

density corrected according to equation (39) in the standard (inserted below). This effectively reduces the WTG

extreme wind speed by the square root of the ratio of the WTG air density and design air density of 1.225 kg/m

Equation (39) from [24] IEC 61400-1 ed. 4 (2019), allowed air density correction criterion for extreme wind.

The new check (a) Ambient 90% Turbulence [NTM], checks that the WTG ambient 90% of turbulence including

turbulence structure correction (C

) is within the design Normal Turbulence Model (NTM) for wind speeds

between 0.6 and 1.6 times the rated wind speed. One may wonder why this check of a non-extreme turbulence

quantile against a non-extreme design turbulence is a part of the ultimate load checks. The reason is this checks

link to a design load case, DLC1.1 where the resulting loads are extrapolated to extreme quantiles/return-

periods. If this check fails, DLC1.1 and load extrapolation must be performed.

The new check (c) Ambient Extreme Turbulence [ETM], checks the WTG ambient 50y extreme turbulence

against the design Extreme Turbulence Model (ETM). No turbulence structure correction is required. An explicit

wind speed check interval is not defined for this check, but the check is related to DLC1.3 in “Power production”

group, defined for all wind speeds between cut-in and cut-out, hence this is the check range we have adopted.

The new check (d) Max Centre-wake 90% Turbulence [ETM] checks the WTG 90% turbulence including wake

turbulence for the worst direction against the design Extreme Turbulence Model (ETM). This check bears a

strong resemblance to the Effective turbulence check for fatigue loads, with the differences that the maximum

centre-wake turbulence across all directions is used instead of the effective integration across directions for the

effective turbulence. Another, difference is that this check (d) does not require turbulence structure correction

) to be applied as is the case for the Effective turbulence. An explicit wind speed check range is not defined,

but with similar arguments as in check (c), the full power production range from cut-in to cut-out is adopted.

The two new checks (a) Ambient 90% Turbulence [NTM] and (c) Ambient Extreme Turbulence [ETM] both refer

to the ambient turbulence. Since no wake affects are included, these checks are handled omni-directionally as

this makes the statistical handling across wind speeds bins and the propagation to the WTG position much more

robust. The omni-directional propagation of the turbulence is performed using the frequency weighted average

wind speed and turbulence for the mast and WTG. An exception is for CFD/Flowres propagation where the full

time series is propagated from the mast to each WTG and then binned omni-directionally there.

Effect of curtailment on fatigue and ultimate checks

If curtailment is applied to some WTG in a layout some of fatigue and ultimate checks may be affected. The

curtailment does not change the ambient wind climate parameters, but will only affect the wake contribution to

turbulence. The control of the WTG itself e.g. shut down or derating shall be handled as a part of the load

simulation. Hence, only the checks which include wake effects will be affected by curtailment, this is the fatigue

check (b) Effective turbulence and ultimate check (d) Max-centre wake 90% turbulence.

Connection to design load cases

The assessment of structural integrity based on the wind climate parameters is closely linked to particular design

load cases (DLC), which depend on site specific wind climate parameters. For the group of checks related to

fatigue loads and the ‘normal’ wind climate parameters, the primary relevant design load case is DLC1.2 ‘Power

production’. Minor contributions to fatigue loads also occurs from DLC3.1 ‘Start-up ‘, DLC4.1 ‘Shut-down’ and

DLC6.4 ‘Parked’, however, much less than the Power production part as summarized in the table below.

Fatigue Load Check

Design Load Cases (DLC)

(a) Wind Distribution

DLC1.2 + (DLC3.1,DLC4.1, DLC6.4)

(b) Effective turbulence

DLC1.2 + (DLC3.1,DLC4.1, DLC6.4)

(d) Wind Shear

DLC1.2 + (DLC3.1,DLC4.1, DLC6.4)

(e) Air Density

DLC1.2 + (DLC3.1,DLC4.1, DLC6.4)

Connections between ‘normal’ wind climate checks and fatigue load DLCs.

Appendix VII - IEC 61400-1 ed. 4 (2019) 118

For the checks relating to the ‘extreme’ wind climate parameters and ultimate loads, the relations to DLCs are

more complex as summarized in the table below:

Ultimate Load Check

Design Load Case (DLC)

(a) Ambient 90% Turbulence [NTM]

DLC1.1 + Load extrapolation

(b) Extreme Wind

DLC6.1 + DLC6.2

DLC1.3

(d) Max Centre-wake 90% Turbulence [ETM]

DLC1.3 alternatively DLC1.6

Connections between extreme wind climate checks and ultimate load DLCs.

DLC1.1 is ‘Power production’ using the normal wind climate via the normal turbulence model (NTM), but for the

purpose of extrapolating the resulting normal operation loads to extreme loads. DLC1.3 is also ‘Power

production’ but with the extreme turbulence model (ETM), where the alternative DLC1.6 also considers wake

effects but in a more narrow wind speed interval. Finally, DLC6.1 and DLC6.2 are both ‘Parked’ during the

extreme 50 year wind speed event, the latter with loss of grid connection.

Design loads

For a generic turbine model there are no explicitly defined design loads from a type certification, hence, the

design loads are estimated as the loads which result from applying the load model (aero-elastic model or

response model) to the design wind climate. IEC61400-1 ed. 3 explicitly mentions that the design load should

represent the worst loading in regards to vertical inflow angles in span from -8° to +8°. Hence, for calculating

design loads in load response the highest load of -8°, 0°, and +8° is always evaluated and the highest taken as

the design load. In IEC61400-1 ed. 4, this requirement is no longer explicitly mentioned. For consistency, we

interpret the intention to be same in ed. 4 in regards to inflow angles influence on deign loads although not

explicitly stated.

Robust Weibull fit for turbulence

In the IEC61400-1 ed. 4 the design distribution of turbulence (wind speed standard deviation) conditioned on

wind speed is described by a two-parameter Weibull, similarly to the annual wind speed distribution. In ed. 3,

turbulence was assumed to follow a log-normal distribution. However, in all calculations relating to the site

specific 90%-quantile of turbulence both ed. 3 and 4 uses an expression for the 90%-quantile, which assumes

a normal distribution, i.e. 







 



. In ed. 4 the Weibull parameters of the design turbulence (Normal

Turbulence Model) are calibrated such that the Weibull based 90%-quantile matches the NTM 90% design

turbulence. To be in-line with the ed. 4 standard we also allow the estimated site 90%-quantile to be based on

fitted Weibull distribution of turbulence conditioned on wind speed. However, as the turbulence data are also

binned on direction very few samples can be available for these fits. To ensure on the one hand a robust fit and

on the other hand consistency with the standard equation based on the mean and standard deviation we have

chosen a method of moments fit [25]. This solves two equations with two unknowns (i.e. the Weibull parameters)

to find the Weibull with same mean and standard deviation as the data. This makes the fit quite robust even in

bins with few samples.

Ambient extreme turbulence and the IFORM method

The IEC61400-1 ed. 4 introduces the Ambient Extreme Turbulence check and presents two alternative methods

to estimate the 50 year extreme turbulence at each wind speed. One method is using a parameterized

expression given in the standard based on peak factors (k

) to be used similarly as for the estimation of the 90%-

quantile, 







 







at each wind speed (



depends on wind speed). This method is said to be due to

Leo Thesbjerg from Vestas in the IEC working group TC88 (no known publication). The alternative method

proposed is the Inverse First Order Reliability Method, or IFORM [26], which progresses in a number of steps to

estimate the 50 year turbulence conditioned on wind speed. Here the concept of IFORM is briefly sketched. The

IFORM method performs the following steps for each wind speed:

1. Calculate the standard normal quantile (x

) for the cumulative frequency of the wind speed

2. Calculate the target quantile (beta) for the target probability of exceedance of 50 years

3. Calculate the standard quantile of the turbulence (x

) from x

and beta in the 2D standard normal space

4. Calculate the target 50y turbulence from its CDF such that it has the same cumulative probability as x

The IFORM is very sensitive to the estimated standard deviations of turbulence and is considerably less robust

than the alternative method. Hence, IFORM is only recommended when data duration and quality is high so that

turbulence distributions are well defined at most wind speeds and only little influenced by the fitting and

extrapolation to higher wind speeds. The IFORM method is deactivated until industry guidelines to improve its

robustness is established/published.

Appendix VIII - Lifetime extension DNVGL-ST-0262 (2016) 119

Appendix VIII - Lifetime extension DNVGL-ST-0262 (2016)

Many onshore wind turbines in Europe will be approaching the end of their design lifetime, traditionally 20

years. The operational lifetime may in many cases be extended beyond the original design lifetime – but this is

typically subject to different national requirements mainly driven by inspection protocols and other such

schemes. There is also an on-going effort to establish an international IEC standard 61400-28 for “Through life

management and life extension of wind power assets”, however, currently only two official standards on

lifetime extension have been released by the commercial entities, DNV-GL and UL. The former entity has

released the standard ”DNVGL-ST-0262 Lifetime extension of wind turbines”, which is a focused, concise and

freely available standard [27] covering both the theoretical fatigue calculations of lifetime and the practical

guidelines for inspection.

From windPRO version 3.5 lifetime calculations in LOAD RESPONSE have been aligned with the DNVGL-ST-

0262 standard. The purpose is providing a very fast, easy and cost-effective tool for assessing the structural

potential for lifetime extension prior to any further detailed and costly investigations and inspections. Physical

inspections are always required for lifetime extensions and cannot be replaced by fatigue load calculations,

only supplemented.

Methods in DNVGL-ST-0262

DNVGL-ST-0262 splits the lifetime assessment into an ‘Analytical’ calculation part (§2.2) and a ‘Practical’

inspection part (§2.3). Our focus is the Analytical part which is further split in to three alternative methods:

‘Simplified’ (§2.2.3), ‘Detailed’ (§2.2.4) and ‘Probabilistic’ (§2.2.5). The former two approaches are deterministic

and covered by windPRO LOAD RESPONSE and described in the following.

‘Simplified approach’ (=generic response models)

This approach is used when the actual turbine design is not available and use of a generic model is necessary.

Although this is a deterministic approach where uncertainties in general are not accounted for, the Simplified

approach specifically requires assessment of the uncertainty related to use of the generic load model. Except

for this uncertainty the Simplified approach is consistent with using LOAD RESPONSE with one of the module’s

generic turbine models. Further description of the uncertainty model and how to handle known load margins is

provided in the following subsections.

‘Detailed approach’ (=specific response models)

The Detailed approach is used when the actual turbine design is available in the form of actual load calculations

for the turbine model in question. In LOAD RESPONSE this corresponds to using a specific response model

which also includes the actual design loads (certification loads) including any load margins. No uncertainty is

considered for this approach.

Load margins (Simplified approach only)

Some specific turbine models are stronger than required for their design class, here referred to as ‘load margins’.

Some users may have knowledge of load margins for the components of specific models after several iterations

with a manufacturer comparing load results for a generic model with those of a specific turbine. Such information

may be included via a new option to combine known load margins with the use of generic models. Load margins

as such cannot be handled as uncertainties as they are systematic biases from the design class basis introduced

by each manufacturer for each turbine model.

Uncertainty model (Simplified approach only)

The uncertainty model relies on the IEC/ISO standard for uncertainty assessments [28] to calculate the resulting

uncertainty on the lifetime. However, there is no simple or direct way to calculate the uncertainty due to the

generic model, hence, the assessment is done as an order of magnitude. In general, the uncertainty related to

using a generic reference turbine model depends on its similarity compared to the design of the actual turbine

model for which the lifetime extension is conducted. An overall description of the reference turbines used in

LOAD RESPONSE is found in the following subsection.

Initially, uncertainties are estimated for each component/sensor by analysing the variation in load index response

across several different turbine models with moderate variations in their structural design. This is then simplified

to a single baseline uncertainty across all sensors/components - in line with the largest variations observed for

blades and tower loads – and therefore possibly conservative for other sensors/components. Three uncertainty

categories are then defined via a doubling of the uncertainty when going one category up, as follows:

• High similarity / low uncertainty: 1 x baseline uncertainty

Appendix VIII - Lifetime extension DNVGL-ST-0262 (2016) 120

• Moderate similarity / uncertainty: 2 x baseline uncertainty

• Low similarity / high uncertainty: 4 x baseline uncertainty

The resulting uncertainties on the calculated component load indices are then propagated to uncertainties on

the component lifetimes using the methodology of the ISO/IEC standard for uncertainty [28]. The relation

between load index and lifetime is very non-linear and results in a significant amplification of uncertainties – this

is inherent in the non-linear nature of fatigue damage accumulation and load calculations.

Reference models (Simplified approach only)

All generic models in LOAD RESPONSE are adapted from public reference models like NREL 5MW geared or

direct-drive [29,31], Windpact 1.5MW [30] and DTU 10MW [32]. These all have a traditional standard design with

the following main characteristics:

• Traditional upwind three-bladed design (fibreglass or carbon - choose appropriate model)

o Max blade chord ca. 8% at ca. 25% of length, narrowing towards the root

• Geared or gearless (choose appropriate model)

• Standard “soft-stiff” steel tower

• Standard variable speed, collective pitch control (PI controller)

Special control strategies are not accounted for and is beyond the scope of the assessment as special control

strategies could in principle result in arbitrarily large differences in the turbine load response.

Appendix IX - Siteres ambient climate files (from Resource, GASP, etc.) 121

Appendix IX - Siteres ambient climate files (from Resource, GASP, etc.)

From windPRO 3.5 a new resource data format “.siteres” has been introduced. Contrary to existing resource

formats such as “.rsf” or “.wrg”, siteres files also include other site parameters in addition to the Weibull

resource parameters.

These parameters are statistical parameters characterising ambient site conditions of extreme wind,

turbulence, wind shear, air density, flow inclination and terrain complexity – sufficient to make a full IEC 61400-

1 site assessment in SITE COMPLIANCE. Hence, from windPRO 3.5 siteres-files have been integrated as an

alternative calculation basis in SITE COMPLIANCE supplementing the existing based on measurements and

flow-modelling. The siteres format is described further in this reference: Technical Note - Siteres

Siteres files may be obtained from several sources. The user may generate own siteres files in the windPRO

RESOURCE module in a setup that inherits most calculation functionality from SITE COMPLIANCE (licence for

SITE COMPLIANCE is required).

An alternative option is GASP - the Global Atlas of Siting Parameter - which provides a global atlas of siting

paramers at 250m resolution and three heights 50m, 100m and 150m. GASP data may be downloaded directly

from windPRO through the “Data / Siting Parameter (GASP etc.)” menu. This opens a download window which

allows specification of the area of interest.

The same download window also includes another new concept, Regional Atlas of Siting Parameters (RASP),

which is for atlases of siting parameters with regional coverage and considerably increased accuracy of the siting

parameters compared to GASP. Currently RASP-SWE 0.3 (beta version) is available as a demonstration dataset

for Sweden.

Appendix X - Tropical Cyclone Analysis 122

Appendix X - Tropical Cyclone Analysis

IEC 61400-1 ed. 4 (2019) includes in Annex J a method to quantify tropical cyclone extreme wind speeds.

However, this theoretical method is not very practical, as it utilizes Monte Carlo simulation to predict tropical

cyclone induced extreme wind speeds.

As a more practical alternative, from windPRO 4.0 SITE COMPLIANCE includes a screening tool to analyse

historical cyclone tracks around a site. The Tropical Cyclone Analysis tool integrates the International Best Tack

database (IBTrACS) [34] of historical tropical cyclone tracks with worldwide coverage going back to pre-1900.

This analysis tool gives the user the possibility to analyse the sensitivity to size of region of interest and to the

historical period of interest. The extracted tracks are visualized, and the important statistics are extracted and

summarized. Finally, a Gumbel extreme wind model is fitted to the extracted data to quantify a 50-year extreme

wind estimate at standard conditions. The Tropical cyclone analysis is added to the group Other IEC checks &

analysis.

Tropical cyclones are also known as typhoons or hurricanes and are so-called warm core cyclones which can

only be seeded and maintained when the temperature of the upper parts of the water column exceed 26°C.

Exactly at the equator the Coriolis force is zero and tropical cyclones cannot form and will die out. North of the

equator the Coriolis force make cyclones turn counterclockwise as winds veer to the right. The opposite is the

case south of the equator, where wind veers left and cyclones rotate clockwise.

Tropical cyclones represent the fiercest weather and wind speed events on our globe, and, luckily, they are rare

and relatively localized phenomena. This is also a reason for the difficulty when analysing extreme winds from

tropical cyclones - getting a sufficient long-term record and sufficient spatial coverage to sample the variation in

tracks. Even several years of data for a point location will rarely be enough to properly evaluate the risk of tropical

cyclone winds.

Tropical cyclones are typically classified following the Saffire-Simpson hurricane scale, which uses so-called “1-

minute maximum sustained wind speed” at 10m height. This simply means the highest recorded/modelled 1min

averaged wind speed within the cyclone at a given time. Not all international cyclone agencies report cyclone

data with the same averaging time, hence, IBTrACS contains data with both1min, 3min and 10min data

depending on agency. To circumvent this problematic inconsistency, we use only wind speeds reported through

the USA agencies which also incorporate data from the other agencies and convert them consistently to 1min

data. The table below summarizes the Saffire-Simpson tropical cyclone scale with thresholds converted to m/s.

The rightmost column indicates the typical damage incurred by each cyclone category.

Saffir-Simpson hurricane scale converted to m/s (see e.g. https://www.nhc.noaa.gov/aboutsshws.php).

Setup - extraction of data from IBTrACS database

To extract cyclone events and statistics from the IBTrACS database the user needs to set three parameters, the

start year of the extraction, the radius around the site centre to extract tracks within and finally a threshold wind

speed below which extracted annual maxima are not included in the fitted Gumbel extreme wind model. By

default, this threshold is set to 20 m/s however this should be specified site specific by the user.

Saffir-Simpson tropical cyclone scale

10m, 1min wind speed

Category min [m/s] max [m/s] Damage

5 70 - Catastrophic

4 58 70 Extreme

3 50 58 Extensive

2 43 50 Moderate

1 33 43 Some/Minimal

Tropical storm 17 33 -

Tropical depression - 17 -

Appendix X - Tropical Cyclone Analysis 123

Results (Graphics) – spatial plot of tracks using Saffir-Simpson scale

This track plot summarizes all tracks dating back to the user-defined start year within an area of ca. 1000km by

1000km. Track points are shown in the colour legend of the Saffire-Simpson scale and higher track category

points/segments are always plotted on top to emphasize historical storm severity. The plot also highlights the

user’s selected radius of representativity around of the site centre.

Results (Table) – summary of extracted cyclone events and annual maxima

The first table on this tab summarizes the number of individual cyclone tracks of each Saffir-Simpson category

within the user defined radius. The table below lists the annual maximum storm point/track for each year back

to the user selected start year – within the selected radius. In this table all the IBTrACS 1min wind speeds

reported by the USA agencies have been converted to 10min averages using the guideline provided by WMO

for tropical cyclones [35]. This conversion is done assuming offshore/coastal exposure and will be conservative

for inland locations, however, as tropical cyclones decay rapidly over land, they are mainly a concern for coastal

regions or islands.

Appendix X - Tropical Cyclone Analysis 124

Gumbel fit (Graphics) – extreme statistics and 50-year wind estimate (at 10m)

The annual maximum wind speeds listed on the previous tab are plotted on a Gumbel plot and fitted to a Gumbel

extreme wind model (see Appendix 1). Wind speeds below the user selected threshold are not included in the

Gumbel plot and fit, however, they still contribute in the calculation of the plotting positions on the y-axis (see

Appendix 1). The Gumbel fit allows for the indicative calculation of a 50-year wind speed for tropical cyclones.

This estimate represents 10m height and offshore/coastal conditions, hence, it may be used as the input base

wind speed when using the EN1991-1-4 method in the Extreme wind check. In this case the reference roughness

should be set to offshore roughness

=0.0002m as the IBTrACS wind speeds represent offshore or coastal

conditions.

The roughness will be increased offshore at higher wind speeds, but this ‘charnock’ effect is not modelled in

models like WAsP, so to avoid artificial speed-ups/down, use of WAsP’s offshore roughness length is

recommended.

Appendix X - Tropical Cyclone Analysis 125

Disclaimer and limitations of the tropical cyclone analysis:

- Data provided from IBTrACS are provided as is with the conversions described above

- IBTrACS data before 1965 to 1980 have limited validity varying between regions

- Some wind speeds are coarsely discretized (e.g. 0.5m/s) visible as step-like artefacts in the Gumbel fit

- The tropical cyclone analysis is intended for offshore and coastal sites

o The analysis may not be representative for inland sites and may be overly conservative

- The analysis does not account for other storm mechanisms - only tropical cyclones

o Other extreme wind mechanisms (e.g. thunderstorms or tornados) may contribute significantly

to the extreme wind conditions at a particular site.

Appendix XI - Downscaling, Offshore mode & Spectral Correction 126

Appendix XI - Downscaling, Offshore mode & Spectral Correction

From windPRO 4.0 SITE COMPLIANCE includes the option to base site assessments on downscaling, that is,

the combination of mesoscale data and microscale flow modelling. This functionality is based on the SCALER

which is already well established and documented in windPRO.

The original motivation for the downscaling option was for offshore applications where mesoscale data mostly

have limited bias and perform well, yet, with some limitations, particularly, regarding turbulence and extreme

wind. Therefore, together with the downscaling option a new “Offshore site” option has been added to simplify

setup for offshore sites in general.

Lastly, mesoscale data have a known bias with regards to extreme wind speeds, which may be sought

corrected by applying a spectral correction which is available from windPRO 4.0. Unfortunately, no general

correction model has been proposed for mesoscale TI, hence, this work is ongoing. However, in general

results based on downscaled turbulence will be non-conservative.

The downscaling and offshore site option can be selected at the main tab in the SITE COMPLIANCE setup as

illustrated in the figure below.

These three new options downscaling, offshore mode and spectral correction are briefly documented in the

following.

Downscaling

As mentioned above selecting the downscaling option will allow for using the windPRO SCALER setup together

with mesoscale data and a microscale model. In fact, by deselecting the Terrain scaling option in the SCALER

setup, no microscale modelling is done at all, which can be justified offshore. Downscaling may be selected

together with the main option Mast data & flow model(s) or No mast data. In the former case, two new tabs

named Calibration mast and Downscaling will appear. In the latter case, only the Downscaling tab will appear.

Calibration mast

On this tab the user may select a meteo object holding measurements from a mast within the region or situated

on-site. These measurements may subsequently be employed by the spectral correction method in the extreme

wind check. Unfortunately, no calibration method is so far available for turbulence or other variables.

Downscaling

This tab provides access to the standard setup of the SCALER and selection of mesoscale data to be used by

the selected SCALER setup.

Appendix XI - Downscaling, Offshore mode & Spectral Correction 127

IEC Checks

Each of the IEC Checks has a new option based on downscaling, named e.g. Downscaled weibulls (via

downscaled timeseries). The exceptions are the checks Terrain complexity and Flow inclination which have no

new downscaling related method. The three checks related to extreme turbulence, introduced in IEC61400 ed.

4, are not allowed in downscaling mode as they require a higher quality of turbulence data.

Warning:

- Be careful with assessments of Effective turbulence based on mesoscale turbulence intensity (TI)

- Even if the mean TI is well captured the variation of TI with wind speed may be significantly biased

- Standard deviation of TI within each wind speed bin is often too low, particularly at high wind speeds

- These biases may lead to underestimation of characteristic TI and component loads

Offshore mode

The offshore mode is mainly establish for the sake of better handling of offshore turbulence data and to avoid

the need to calculate terrain complexity. Within the Terrain complexity check an offshore option has been

introduced that avoids the need to actually establish terrain data via a DEM or TIN model.

In addition, the Effective turbulence check has a new option to include the Charnock effect. This introduces a

second order term in the turbulence fits that extremely well captures the dynamic charnock effect, i.e. that

higher wind speeds generate lager waves that in turn increase roughness and turbulence.

Spectral correction

From windPRO 4.0 SITE COMPLIANCE implements the extreme wind spectral correction method of Xiaoli

Larsén et al. from DTU [36]. This method applies to reanalysis data and may be based on either theoretical

assumptions of the spectrum or using an observed spectrum from a representative mast to quantify the relative

damping of the reanalysis data’s spectrum.

Fundamentally, the spectral correction method (hereafter SCM) seeks to compensate for the fact that the time

series of reanalysis data are too smooth or equivalently put that the spectrum falls off too rapidly at high

frequencies. The SCM simply quantifies the expected reduction of extreme wind speeds due to this error in the

spectrum. The result is a simple correction factor that is applied to the extreme wind results based on

reanalysis data as shown in the Results (Table) screen shot below, where the correction is +6%.

128