11 PART: OPERATION, PERFORMANCE CHECK &
POST CONSTRUCTION
11.1 Introduction ............................................................................................................................ 3
11.2 Chapter 1: Performance Check & Post Construction Method .......................................... 4
11.1.1 Workflow .................................................................................................................. 5
11.1.2 File Type Overview .................................................................................................. 5
11.3 Data Preparation .................................................................................................................... 6
11.4 Data ......................................................................................................................................... 7
11.1.3 Start Performance Check and Concept Choice ....................................................... 7
11.1.4 Data – Data Source ................................................................................................. 8
11.1.5 Data – Import Setup ................................................................................................. 9
11.1.6 Data – Pair and Load ............................................................................................. 12
11.1.7 Data – Time Series ................................................................................................ 16
11.1.8 Data – Wind Energy Index Database .................................................................... 18
11.1.9 Data – Time Shift ................................................................................................... 24
11.5 Setup Error Codes ............................................................................................................... 25
11.1.10 Error Code in Time Series ..................................................................................... 26
11.1.11 Manage Error Codes ............................................................................................. 26
11.1.12 Separate Error Code File(s) ................................................................................... 26
11.1.13 Import error codes Set-up ...................................................................................... 27
11.1.14 Pairing to WTG object ............................................................................................ 28
11.1.15 Manage Error Codes ............................................................................................. 28
11.1.16 Import Rules ........................................................................................................... 33
11.1.17 Load/Review .......................................................................................................... 34
11.1.18 No Error Code Available ........................................................................................ 36
11.6 Sensor Calibration ............................................................................................................... 41
11.1.19 Wind speed substitution ......................................................................................... 43
11.7 Create Basic PARK Calculation for losses ....................................................................... 45
11.8 Statistics ............................................................................................................................... 49
11.9 Analysis ................................................................................................................................ 52
11.1.20 Analysis – Time series ........................................................................................... 52
11.1.21 Analysis – Error Loss Statistics ............................................................................. 54
11.1.22 Analysis - Error Loss Matrix ................................................................................... 55
11.1.23 Analysis – Power Curve ......................................................................................... 56
11.1.24 Analysis – XY Graph .............................................................................................. 58
11.10 Monthly Gross ...................................................................................................................... 60
11.1.25 Analysis – Wind Energy Index ............................................................................... 62
11.11 Report ................................................................................................................................... 64
11.1.26 Main section ........................................................................................................... 64
11.1.27 Historic Power Curve ............................................................................................. 65
11.1.28 Actual Losses......................................................................................................... 66
11.1.29 Normalized Production .......................................................................................... 67
11.1.30 NET Production...................................................................................................... 68
11.1.31 Export ..................................................................................................................... 69
11.12 Chapter 2: Monthly production to long-term yield ........................................................... 70
Post Construction - Introduction 2
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.32 Wind Energy Index Correction ............................................................................... 70
11.1.33 Analysis – Monthly time resolution ........................................................................ 70
11.1.34 WCP – loss form .................................................................................................... 72
11.1.35 WCP - reports ........................................................................................................ 73
11.13 Chapter 3 Model calibration & Performance Check ......................................................... 76
11.1.36 Model Calibration, loss estimate and long-term expectation. ................................ 76
11.1.37 Time Varying PARK Calculation for Performance Check ...................................... 77
11.1.38 Wind data. .............................................................................................................. 78
11.1.39 Calculation setup traditional PARK calculation ...................................................... 78
11.1.40 Data preparation .................................................................................................... 80
11.1.41 Performance Check – Model calibration start ........................................................ 81
11.1.42 Detailed 10-min / hourly data vs model ................................................................. 82
11.1.43 Data – Pair and load .............................................................................................. 82
11.1.44 Analysis of detailed time resolution ....................................................................... 82
11.1.45 Time series ............................................................................................................ 82
11.1.46 Power curve ........................................................................................................... 83
11.1.47 Analyzer, detailed data .......................................................................................... 85
Post Construction - Introduction 3
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1 Introduction
Operating turbines should frequently be checked if they perform as expected. There can be many
reasons for poorer performance than expected when the project was designed. A major reason is that
the model calculation had poor quality or wrong data/assumptions, or the short-comings of the model
itself But also, operational issues like yaw errors, pitching problems, and unforeseen curtailments (e.g.
grid), icing loss, etc. could be the reason.
With the windPRO module PERFORMANCE CHECK, it is possible to analyse the performance in an
efficient way and find reasons for under-/over-performance or fine-tune a wind model to be used for
new projects in the area.
Also, monthly production and availability figures can be efficiently analysed with the tool.
Since windPRO 3.4, PERFORMANCE CHECK has been divided into three work paths.
• Post-Construction:
o Primary focus, detailed investigation into historic losses, and calculation of Potential
Production, Long Term Correction, and Future Net Production reporting.
o Methodology is based on historic measured SCADA power curve, for when the turbine
is in normal operation.
o Based on 10min SCADA data, and turbine error/event logs for defining losses and their
causes.
• PERFORMANCE CHECK:
o Monthly production and long-term correction.
o Primary focus on wind model validation, to existing wind farm production data.
o Long term correction of production data, 10min, hourly or monthly data.
• TR10: Dedicated manual: https://help.emd.dk/knowledgebase User manual Chapter 13.
o Primary focus, German wind farms built under FGW17, must evaluate Goodness factor,
according the TR10 standard. Model is certified by TÜV SÜD.
This manual is focused on the Post-Construction path, explaining in detail every Tab in the module
and the workflow. Multiple operations are identical if one uses Performance Check for quantification of
error losses or for model validation.
• 1
st
Chapter is for Post-Construction.
• 2
nd
Chapter is for having only monthly production data and wanting to evaluate long-term
expected yield.
• 3
rd
Chapter is for model validation path, where a time-varying PARK calculation is paired to
actual production to calibrate a wind model to reality, such that the model can be used for re-
powering a wind farm or designing future neighbouring projects.
Post Construction - Chapter 1: Performance Check & Post Construction Method 4
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.2 Chapter 1: Performance Check & Post Construction Method
The Performance Check module in windPRO allows the user to analyse 10-minute SCADA production
data together with wind turbine error/event logs to evaluate the lost production corresponding to
individual errors and predict the future long-term corrected production.
The calculation is based on the actual 10-minute SCADA data of the wind turbines, combined with the
error/event logs to identify each 10-minute period stop or non-optimal performance, identify the cause
of each event, and categorise and assign responsibilities. If error codes are logged in separate files
with start/stop times, the information can be synchronised with the 10-minute SCADA time series. This
results in a 10-minute time, series of production, wind data, and relevant error codes and their
description.
Once all events with errors and/or non-optimal performance have been filtered out, the remaining
records describe normal operation which is used to establish the historic power curve for each turbine
(typically for 0.5 m/s wind speed bins). For periods with faults or stops the nacelle wind speed and the
historic power curves of the individual WTG are used to derive the potential production. The loss
categories follow IEC 61400 26-1. In this process, the logged actual production during error (and non-
optimal) periods is being replaced with the production found by applying the historic power curve to the
nacelle wind speed at the relevant time steps. The potential production values will tell, what the
turbines would have produced on average under normal operation, if it was not affected by an error
event or partial performance. It represents 100% availability. The actual production from the SCADA
data and potential production are then compared in periods of irregular operation. The difference
results in the lost production for each time step.
Please note, that wake losses (internal and from neighbouring wind farms) are inherent in the SCADA
data and are not quantified separately. Electrical losses i.e. within the cables from the turbine to the
metering point) are handled separately as they appear after the SCADA data are logged.
This potential production is then long-term corrected using a wind energy index based on mesoscale
data or other sources and results in the normalised production. Subsequently, expected future long-
term losses are deducted from the normalised production to assess the expected future net
production. Future losses are either assumed to be identical with previous losses or defined by the
user, if for example, changes in the losses are expected. Additionally, electrical losses and losses due
to degradation can be included.
This method is primarily used for evaluation of operating wind turbines/farms, as the uncertainty of the
net production is typically half of the traditional pre-construction yield estimates based on wind
measurement and models. It can also be used for a detailed investigation into the causes of lost
production and to evaluate if actions should/could be taken to address the causes of lost production.
Post Construction - Chapter 1: Performance Check & Post Construction Method 5
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.1 Workflow
The most typical application and workflow for post-construction analysis is described in the Quick
Guide.
However, since only a few error code logs follow the IEC 61400-26-1 structure, the user can be
confronted with various formats. In case of cascading errors and alarms it might be difficult to identify
the relevant event. Therefore, iterations in the work process might be necessary. For example, the
user might find that some events with a specific fault show no poor production. This could be due to
the fact that the time stamp in the SCADA data belongs to the end of the 10-minute period and not the
start. Consequently, the user has to revisit the loading of the 10-minute SCADA data, change the
setting, load the data again and pair the production data with the right WTGs.
Apart from the pure analysis of SCADA data, the Performance Check module can be operated
together with Performance Check track/work path, such the user can fine-tune and validate a chosen
model set-up by comparing real production with modelled production. Specifically, for assets with poor
availability, or poor data recovery of the SCADA data, it might be advisable to use Performance Check
track rather than Post-Construction to predict the future production.
There are interfaces between the two methodologies. Please note, that both tracks require running a
time-varying PARK calculation. In case of Post-Construction, the nacelle wind speed and historic
power curve of the individual WTG will be used to calculate what the WTG should have produced for
each time step. The time-varying PARK calculation does not require any flow modelling in this case. It
can be started from within the Performance Check module.
In the case of the Performance Check track, a full flow model must be set-up. It is possible to use local
mast data as well as mesoscale data for wind speed information. The model can be queried, for
example to find the most suitable scaling factor of the mesoscale data, or to find the best wake
parametrisation which reflects the observed wakes best. The actual production data can be prepared
in the Post construction module and filtered for events with error codes or sub-optimal performance.
One of the key parameters to identify the success of a specific model set-up is the Goodness factor
which describes how well the model reproduces the measured production.
11.1.2 File Type Overview
By default, all windPRO files are located in the folder \windPRO Data\Projects\…
What
Project file extension
wind base file
*.wbf
Performance Check Error
Settings file
*.pfc
Performance Check Import
Setup
*.pci
Post Construction - Data Preparation 6
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.3 Data Preparation
Typically, 10-minute SCADA data are used for Post-Construction analysis. The data have to be in form
of an ASCII text file, e.g., comma or tab delimited. If they are in MS Excel files (.xls/.xlsx), these must
be saved as .txt or .csv or other text file formats. Tab separated text data are recommended.
For the import of the files, it is convenient if the turbine identification in the import files is identical to
the ID in the existing WTG object (description or user label) in the windPRO project. It is important that
the files contain data organized in rows, each row starting with the time stamp, preferably something
like dd-mm-yyyy hh:mm. Please note that midnight (00:00) will be empty in Excel, however, there is no
reason for concern as this empty time stamp is handled correctly in the data importer in windPRO.
The temporal resolution of SCADA data is typically 10 minutes. windPRO is currently able to handle
data with a granularity of 1 minute.
As a minimum, the files must contain production data and wind speed. Both production data in kW and
kWh can be handled, as well as accumulated production. Additionally, wind direction and any
operational parameter like rpm, pitch angle, etc. can be imported. The direct use of error codes is also
possible as long as the error codes come in 10-minute steps. If the error codes are available in a
separate file with start/stop times they can be converted to 10-minute steps later in the process (see
Section 11.1.12).
Table 1 shows an example for a data structure. Each signal contains the WTG ID. Please note that the
data structure is shown transposed! For importing the file structure has to be in rows per time stamp.
Table 1: Example SCADA data structure – please note for display reasons the table has been transposed!
Post Construction - Data 7
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.4 Data
11.1.3 Start Performance Check and Concept Choice
The Performance Check module can be found under Loads & Operation tab in the main ribbon (Figure
1) under the SCADA Data Analysis group select Performance check. Start with creating a new
“session” or open a previously created session.
Figure 1: Start Performance Check Module
Performance Check offers different workflows. In this section we explain the workflow and
functionalities for Post construction evaluation (Figure 2). The process follows the IEC 61400-26-1
structure. The use of this functionality requires a license for Post construction.
Post Construction - Data 8
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 2: Choice of workflow in Performance Check
11.1.4 Data – Data Source
After having chosen the concept (post construction evaluation) the user must define what type of data
is imported (Figure 3).
For SCADA data analysis typically 10-minute data are used. This category also handles other
temporal resolutions like 1-minute or hourly data.
In case the existing WTG object already contains production data, the user can decide to use the
already imported production data.
Starting a new session requires the user to decide if new data are to be imported into the existing
WTG objects, or already imported shall be used. In the latter case, the import process is skipped.
Post Construction - Data 9
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 3: Define type of import data.
11.1.5 Data – Import Setup
The Importer for SCADA data is very similar to the importer in the Meteo object. The user has to either
directly select text files or a folder which contains the text files.
The time zone has to be specified.
The Auto detect button will suggest a structure. This might identify some of the signals, but typically
some manual adjustment will be needed. The user has the option to modify the suggested structure
and specify the line with the header, the first line with data and field separators. At any step the user
can preview the file structure and adjust file formats if required. It is also important to define if the time
stamps refer to the start of a 10-minute period or to the end of a 10-minute period. SCADA systems
often log the data at the end of a 10-minute period, however this is not a universal/standard practice.
The impact of this setting can become important later in the process when the production data is
coupled with error codes.
Post Construction - Data 10
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 4: Importer for 10-minute data
For post construction analysis, users are required to have the production and the wind speed as bare
minimum. Following additional pre-defined signals are available:
• Wind direction
• Availability
• Error Code
• Temperature
• Pressure
• Pitch angle
• RPM
For some signals it is necessary to define the sub-type: Is the wind speed signal representing the
mean, max, min, or std? Is the production signal accumulated production? Once the sub-type is
defined, and the units are set, the data are converted and appears in the right-most column. Please
note that you also must specify the unit for the time stamp. Specific care must be taken in setting the
unit of the power signal correctly. It is, however, at any time possible to return and change the unit.
Please note that all negative production is set to zero. For any selected period, the sum is logged and
documented as a total for the selected turbines.
Multi-editing of the signal is possible by marking the signals you want (by dragging the mouse or by
“Shift-/control-click”), specify the type and sub-type and finally pressing “Apply”.
Further it is possible to change text to numbers. E.g. if your date stamp contains text like DEC, you
can translate this into 12. Or if invalid data is marked NAN, you can change it to -999.
It is possible to save the structure as a “.*pci” file and use it for other projects. It is also possible to use
several import filters in case the SCADA data has been provided in different formats.
Post Construction - Data 11
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
In the next step of the process, it is necessary to assign the production data to a specific WTG. There
are two options: In case you do not know where to find the WTG-ID, a WTG-ID Guide will help you
with the necessary specifications, or alternatively you can directly enter where to find the WTG-ID in
the file (Figure 5).
Figure 5: WTG-ID Guide
In first step of the WTG-ID Guide you have to specify the number of WTGs you want to analyse. Then,
you have to define whether the data are in individual files per WTG, if the WTG-ID is inside each file,
or if the WTG-ID is part of the file name. As shown in Figure 5, selecting underscore as delimiter after
WTG-ID, windPRO will identify the individual names of each of the turbines.
If the WTG-ID is inside the file, you have to specify if the WTG-ID is part of the column header (e.g.
WTG02_Spd as column header) or in a specific column.
In the next step you have to define the first, second WTG-ID, as well as the last WTG-ID by clicking at
the top of the columns. This assists the tool to identify the remaining WTGs automatically. You have to
specify the delimiter, which separates the WTG-ID from the signal name. If the delimiter is an
underscore “_”. Sometimes the column name might include several underscores (e.g.
WTG03_Spd_1). Therefore, it is possible to define which sub-string includes the WTG-ID, in our
example it is the first sub-string that contains the WTG-ID.
Post Construction - Data 12
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 6: WTG-ID selection GUI Example.
Figure 7: Successful selection of WTG-ID
If the selection has been successful, the WTG-ID will now show up (Figure 7). You now have to pair
the production data with existing WTG objects.
11.1.6 Data – Pair and Load
The user has to choose which of the existing WTGs to add to the analysis (Figure 8).
In the case where the naming convention of the existing WTGs is identical with the WTG-ID used in
the SCADA files, windPRO will automatically pair the object with the production data. In the case
where the automatic pairing does not work properly, use the drop-down in upper right corner to match
the ID in the import filter to the objects.
***IMPORTANT NOTE: the exiting WTG object works as a database of SCADA data loaded. If you
load new data into the same layer containing the same existing turbine objects in another Performance
Check session, the existing data with all filters and settings will be overwritten. Thus, it’s important to
clone the object layer containing the existing WTG objects if multiple investigations on different
parameters are required.
Once correctly paired, click and the data will be imported into the WTG objects
(Figure 9).
It is possible to re-visit at a later time. This might be the case, if you have
received further data: For example, you started out by only having monthly production data but now
Post Construction - Data 13
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
you want to add data with a different temporal resolution e.g. 10-minute production data. Depending
on the application you might want to load all data, or you might want to load only the new additional
data into the existing WTG objects, and consequently maintain data that was earlier loaded. Or you
want to load production data, which you had loaded directly into the existing WTG object into the post-
construction calculation.
Figure 8: Chose existing WTGs for pairing.
Figure 9: Load measured data
Post Construction - Data 14
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
On the tab Pair and Load a lot of other functions can be found. Some of them will be treated in a
different manual section (as annotated in Figure 10).
Figure 10: Additional functions on Pair-and-Load tab
Setup error codes is described in Section 11.5. It deals with how to combine the error logs, error
descriptions and categories with 10-minute SCADA data.
Sensor calibration is described in Section 11.6. It deals with checking the nacelle wind speed for
consistency by comparing with references, and for substituting filtered and/or missing wind speeds.
Create time-varying PARK calculation is described in Section 11.7. It allows the user to run a time-
varying PARK calculation.
Additionally, to these larger functionalities following possibilities can be accessed:
Create meteo objects: This button will convert the wind speed signal of each WTG into a Meteo
object.
Apply colours to WTG objects will give the WTG symbols the same colours as used in the graphics
within Performance Check. This is very convenient when evaluating results WTG by WTG.
Setup: Here a number of boundary conditions can be defined (Figure 11). In the first section the user
can define which signal should be used for further analysis, e.g. turbulence intensity is only considered
for wind speeds above 4 m/s. Turbulence intensities at wind speed lower than this value are
automatically disabled. How many samples must be present when aggregating data from e.g. 10-
minute data to hourly data, can also be defined.
Post Construction - Data 15
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 11: Setup settings
Multi Signal setup defines the handling of several signals of the same type, e.g. there might be two
anemometers installed on a nacelle. If both signals are loaded, it can be decided to use the average,
maximum, minimum, or sum of the two signals.
Figure 12 shows an example with two wind speeds being imported. The two signals are combined by
averaging. The user can further define which of the signals is used in future graphical displays.
Figure 12: Handling of multiple signals
Post Construction - Data 16
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.7 Data – Time Series
Before moving on, it is a good idea to quality control all choices by visiting the time series tab. Here
you can inspect your data and perform some basic filtering. Please refer to Meteo manual for details
as the functions are identical to those in the Meteo object.
The data can be sorted by clicking on the top of the column you want to sort by. If data are out of
range or there are duplicates, etc., the small arrows will allow you to jump to the specific time step.
The power signal can be investigated for e.g. by sorting the power column values. The maximum
power [kW] should be identical to the rating of the WTG. If the maximum power is wrong by a factor of
6, the reason could be the choice of units in the import filter: [kWh] or [kW]. If the maximum power is
wrong by a factor of 1000, check if [Wh] or [kWh] should be used. Simply change the unit in the import
filter (see Section 11.1.5) and reload the data.
Figure 13: Example resulting time series.
From windPRO 4.0 a new data cleaning tool within the “Trim and Clean” button is introduced.
This is capable of working on all individual turbines in one iteration, where previously the trimming and
managing of the loaded SCADA data needed to be dealt with one turbine at a time.
Post Construction - Data 17
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 14 window for Trim and Clean feature, all turbines in one work step.
By selecting the Trim and Clean button, a new window appears, here you can select if you want to
work on all loaded SCADA data or a sub section in time.
Further if you want to remove some turbines from the trim and cleaning task, you can de-select the
turbines individually.
Hitting the trim outside period button, deletes measurements outside the selected time interval for
selected turbines after a confirmation notification. All data can be re-loaded under Pair and Load tab.
Trim option is disabled, as long as entire period is selected.
The Clean inside period button, will start scanning through the selected turbines and time period, for
identical wind speed samples following each other of minimum set sample length, for all samples
below minimum wind speed threshold, and disable these samples. These are presumed to be faulty
data that are “frozen” & trended wind speed signals that should not be used to calculate potential
production.
Note for trends of a frozen wind speed signal the first and last sample is disabled for the duration of
found samples. For linear/sloped trends, the first and last sample is not disabled. This is due to the
assumption that such events represent a logger system where it’s a fixed linear inserted wind speed
range, between a gap in the data from last known value and first known value after the gap.
Later these faulty and disabled or missing signals for wind speed readings can be replaced with a
calculated windspeed signal following the TR10 “German Standard” methodology, see section 11.6 for
further details.
Post Construction - Data 18
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
The trend detector is based on a gradually increasing, moving window which considers the difference
between each sample. Example:
1) Moving window starts at 05:30. Time to next sample is 10 minutes, so we continue
2) Time to next sample is 10 minutes, so we continue.
3) The sample at 05:50 does not match the trend of the previous two samples. The moving
window size is once again decreased to one sample, and we assign the base of the window to
the next sample at 05:40.
4) The window is gradually increased again and since the number of samples in the window is
greater than or equal to the threshold, we mark all samples in the window to be cleaned -
except for the first and last sample as highlighted below:
Figure 15 Example of sloped trend detection
The last button “Enable all windspeeds in period” is to re-load all wind speeds which have been
disabled by the cleaning function.
Note, some 10-min SCADA wind speed signals only have one value beyond the decimal point, hence
it becomes statistically more and more realistic that at random, three samples following each other will
have a 100% linear trend/constant offset for 3 x consecutive time steps. Thus it’s recommended to
investigate the time series post running the “Clean inside” method and evaluate if the threshold for
your type of SCADA data require a larger threshold for the automatic filter setting for constant or
trended data.
11.1.8 Data – Wind Energy Index Database
In order to calculate the long-term expected energy (WCP - wind corrected power), it is necessary to
load a minimum of one wind energy index. windPRO has its own file format for storing indices: *.wbf
file (wind base file).
Post Construction - Data 19
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 16: Different choices for energy indices
There are several options (Figure 16):
Option 1: You can load an existing *.wbf file (wind base file) which you have created earlier.
Option 2: Import/update online: For some countries, there are “official” wind energy indices. So
far, this online data includes only the DK-index. While the German BDB index is not included in
windPRO and therefore must be added manually, windPRO assists in identifying the relevant
region (Figure 17).
1.
2.
3.
4.
5.
Post Construction - Data 20
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 17: Identification of relevant region BDB index
Option 3: Add from Meteo object: You can create an index based on a Meteo object and save it
either in a local or in a global folder (for later use). Typically, reanalysis data like MERRA-2 or
ERA5, or mesoscale data like EMD-WRF Europe+ are used. To convert wind speed to energy, the
user can define a power curve and the expected long-term wind speed at hub height. The wind
speed from the long-term data is scaled to match the expected long-term wind speed on site. This
way the dynamic behaviour of the index reflects the production variations as best as possible, and
it becomes possible to use also long-term wind speed data, which may be very different from the
site conditions.
In the next step, the relevant Meteo object is selected, and the reference period defined.
Finally, the calculated wind energy index can be viewed graphically.
This process can be repeated for different sources and/or settings. All results are written into a
Wind Index Database (Figure 19).
Option 4: Add user index base: This function allows to create a user index either by manually
editing row-by-row or by pasting from the clipboard.
Post Construction - Data 23
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Option 5: Since WindPRO 4.1 Create Turbine Indices.
Note this option requires that error codes have been loaded and a loss calculation has been
performed.
Turbine indices are auto created pr turbine, it uses a solver to fit the reference long term wind speeds
such that it on average can predict what was measured. Thus it requires loaded production data to be
used for the solver.
The user can choose a Turbine Index, which follows strictly the same process as defined in
www.emdenergyindex.com for turbine indices, user can here set the training period and the reference
“Index100% period” windPRO will then automatically download all required ERA5t data nodes, and
initiate the solver pr wtg and create a unique Turbine Index pr turbine.
Alternative the user can select a meteo object, Note it requires minimum data since January 2004 and
windspeed, direction and temperature at 10 and 100m elevation, like ERA5 data points or EMD WRF
Mesoscale data time series.
Figure 20 Create turbine Indices
Note with many years of production data for the training and many turbines, the solver and creation of
the unique turbine indices can take +30min. “about 2min pr turbine”.
Once the solver is finished, a new monthly energy index is saved pr turbine, and under Wind energy
Index tab, the Index will now change if a new turbine is selected as they are fixed together and not as
the other types of Indices which are representative to the area excluding wakes and micro site-specific
speedups.
Typically, the turbine index, should see a higher fit to the actual production variation over time, as the
solver have forced it to do so, eg training the index to know each individual turbines micro site-specific
conditions. According to our findings validating the turbine index for a period after the selected training
data period, we seen a +40% increased correlation or reduced uncertainty by using the trained turbine
Index vs a traditional energy Index.
Post Construction - Data 24
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.9 Data – Time Shift
In the time shift tab (Figure 21) time shifts can be applied, either to individual or to all WTGs. Such a
time shift might become necessary when you want to compare data with, for example, mesoscale data
or production generated from a flow model. The user can add as many lines as necessary. The status
column keeps track if the time shift has been performed. By pressing Load measured data the time
shift is applied.
Figure 21: Use of time shift function
Post Construction - Setup Error Codes 25
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.5 Setup Error Codes
The setup of error codes can be accessed from the Pair-and-Load tab. In Section 11.1.4 it is explained
how SCADA data are loaded. Some SCADA data might include an error code in the 10-minute time
series, but it might not include a description of what that error code means. Also, a category has to be
added. This allows the allocation of lost production to different categories. Other SCADA systems do
not include error codes at all in the 10-minute time series. Consequently, the SCADA data has to be
linked with a separate error log. Besides the error code number (e.g. 309) you might also want to add
a description to that specific number and introduce sub-categories based on the description. The user
is required to define and add these categories.
In this part of the performance check module, you learn how you can tie all the information together,
so in the end you have production data, error code, description of error code, and the category tied
together in a 10-minute time series.
The process of setting up error codes and their respective categories or sub-categories can be time
intensive and might require a number of iterations. However, the resulting settings can be saved and
can be used for other projects with the specific error code structure (*.pfc file).
After loading 10-minute SCADA data you have three concepts to choose from (Figure 22) depending
on which information is available and therefore how you want to load and set up the error codes.
Depending on which option is chosen the user must run through various steps to load error logs,
connect error code numbers with descriptions, categorize the errors, convert them into 10-minute (if
they are provided as start/stop time), and connect with the existing WTG objects. You can also define
whether any lost production related to a specific error code is compensated for or not.
• Error codes in the time series: Here the error codes are already part of the production data
and are consequently already imported along with the production data. The next step is to
manage error codes to assign descriptions and categories.
• Separate error code file(s): Error codes are available in a separate error log file with
start/stop times. This file can be added to the production time series using an advanced
interpreter.
• No error codes available: In case error codes are not available, it is possible to define user
error codes based on the scatter plot of the power curve.
Figure 22: Setup error codes: Concept choice
Post Construction - Setup Error Codes 26
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.10 Error Code in Time Series
Some SCADA systems include an error code in the 10-minute time series, e.g. the manufacturer logs
the first error code within each time stamp. In such a case the error code can be imported together
with the production data (Figure 23). Information can be found in Section 11.1.5 Data – Import Setup.
Figure 23: Loading error codes in Data – Import Setup
It is important to note that only ONE error code column per WTG can be loaded. If there are more
columns (e.g. a main and a secondary error code), these must be merged into ONE error code column
before loading. This can be done, for example, by multiplying the first error code with 10000 and then
adding to the other. This way a unique error code for each sub-category is created.
If an error code is a text string, it must be converted to a number (e.g. in Excel) before loading.
Merging two error code numbers is possible in the importer based on a log file, see Section 11.1.13.
Even though the 10-minute time series includes the error code number, you might want to add a
description to the error code. Additionally, it can be defined if losses due to curtailments are
compensated for or not. Also, the loaded error codes have to be categorised such that the user is able
to differentiate between compensated and non-compensated losses. Pressing the top button (Error
Codes in Time Series) in Figure 22 brings you to Manage Error Codes.
11.1.11 Manage Error Codes
The Load Error Code Translator allows you to assign categories and descriptions to the already
loaded error codes. It is also possible to define curtailments, which can add significant value to the
analysis: If a turbine is curtailed, e.g. for flicker or bats, this is an upfront known loss included in the
pre-construction AEP expectations. Therefore, this would often be given a special treatment, which is
possible to do so by marking this. An example for compensated curtailment could be grid curtailment,
where the operator is reimbursed for shutdown due to grid limitations.
Setup is the same as if error logs were from files, see Section 11.1.15
It’s now possible under Load/review to ignore pre-loaded error codes, e.g. if they are covering mostly
normal operation modes, and are just an alarm that’s has been set in the 10-min production time
series. You can now choose the individual error codes and ignore them. Note this will change the error
code and set it to a 0 value, and it is then considered as normal operation for the 10-min samples it
covered.
11.1.12 Separate Error Code File(s)
In this section you learn how you can merge start/stop error logs from separate files with the 10-minute
production time series.
Several manufacturers do not store the error codes as part of the 10-minute time series in the SCADA
system. The error codes might be available in a separate error log files with the time stamp for start
and stop of individual error codes and/or the duration of the error. Often multiple error codes can be
present simultaneously, meaning that there are several error codes active within a 10-minute period.
The tool presented here gives the user different options for the interpretation of such data and the
possibility to evaluate the consequences of the chosen interpretation.
Post Construction - Setup Error Codes 27
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
The error log file has to be in ASCII format. It is advisable to align the naming of the WTG such that it
is the same in the SCADA data and the error logs. The error code importer can load two columns of
error signals (primary and secondary level). Consequently, it is not necessary to merge primary and
secondary error codes before loading. Only numerical values can be imported, any error codes in form
of text have to be converted to numbers.
The use of error logs in separate files requires several steps, starting with the import. After importing,
the error log(s) must be paired with the WTGs. The next step involves managing the error codes.
Import rules can be set and finally the resulting statistics can be loaded and reviewed.
11.1.13 Import error codes Set-up
The error log files (in ASCII format) are selected. The format has to be defined, including the line
which the header is in, the first line with error code and the delimiter (Figure 24).
Figure 24: Importing error log file(s)
The concept is similar to the error code translator loader described in Section 11.1.11. However, since
error logs normally do not come as 10-minute time series, not only the error code and its description
but also the start and stop time/duration have to be defined.
The user has to set the correct format for date and time. If both stop time and duration is available,
only one of these can be chosen. Some SCADA systems have neither a duration nor end time in the
log. Instead they might have a time stamp with a text, e.g. “incoming” means that the event starts,
“reset” or “phasing out” means the end of the event. In such a case the type has to be set to “start/stop
text”. An extra box will appear (Figure 25) where you can specify the text which stands for activating
the event (e.g. “incoming”) and which text deactivates the event (e.g. “reset” or “phasing out”). It is
possible to specify several words, which must be separated with a semicolon. The setup is case
sensitive.
Post Construction - Setup Error Codes 28
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 25: Set-up for start/stop text.
It is essential to assign the error codes to the specific WTGs. In our example the WTG-ID can be
found in a column (Figure 26).
Figure 26: WTG-ID in error log
11.1.14 Pairing to WTG object
After the import of the file the error log has to be paired with the existing WTG object similar to Section
11.1.6. If the WTG-ID in the error log is identical with the name of the existing WTG, the tool will
automatically pair the object with the error log. In case the automatic pairing does not work properly,
use the drop-down in upper right corner to match the ID in the import filter to the objects.
11.1.15 Manage Error Codes
Since error logs have various forms, it depends very much on the individual error log how much
information is included. The aim is to not only have the error code but also the error description and
category connected to the 10-minute time series.
In this tab the error codes are given both descriptions and categories. There are five options on how to
connect error codes with their description (if required) and categories.
It is also possible to define curtailments, which can add significant value to the analysis: If a turbine is
curtailed, e.g. for flicker or bats, this is an upfront known loss included in the AEP expectations.
Therefore, this would often be given a special treatment, which is possible by marking this. An
example for compensated curtailment could be grid curtailment, where the operator is reimbursed for
shutdown due to grid limitations.
Post Construction - Setup Error Codes 29
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 27: Five options to load error code translator
Option 1: From loaded time series – this transfers the error codes which you have loaded before
(Section 11.1.5) into a table. The user now has the possibility to manually assign descriptions and
categories to each error code. As modern turbines can have several thousands of error codes, this
can become a time-consuming process.
Option 2: From attached error log file(s) – in contrast to Option 1, here the error log file is used to
connect the error code numbers with their descriptions. Often categories must be assigned manually.
Option 3: From file – From previous experience you might have built up a directory which connects
error codes with their descriptions for a specific SCADA system. The main advantage compared to
Option 2 is that you can benefit from your own error code library, which can also include categories.
The user has to specify how the ASCII file is structured (Figure 28), e.g. in which line (row) the header
is and what the delimiter is. Via a drop-down menu the user can define the type of data. Normally just
the error code and description are needed.
Sometimes files containing error codes, and their descriptions are part of the contract with OEM. The
contractual agreement might refer to the error codes the manufacturer is responsible for and thereby is
included in the calculated availability guarantees.
Figure 28: Define types in error code file.
As a special feature it is possible to define both primary and secondary error codes (Figure 29). In
such a case an extra item appears: By defining a multiplying factor for the primary (main) or secondary
error code it is possible to create a unique error code. It is recommended that you assign a multiplier
Post Construction - Setup Error Codes 30
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
which exceeds the highest value of the secondary error code to ensure unique codes. In our example
the primary error code 309 will be multiplied with 10000 resulting in 3090000 and 201 from the
secondary error code will be added. The new, unique error code is 3090201.
Figure 29: Secondary error codes
Option 4: From Clipboard is like Option 3 but used when the data is in the clipboard e.g. after
copying from Excel.
Option 5: From another Performance Check session – This option is only available when you
create a new Performance Check session within the same windPRO project. To get access to a list
from another project, use the Export – Import option.
In general (independent of the chosen option) multi-editing is possible by marking a group of error
codes (Figure 30). The list is auto sorted by the most frequent occurring error codes.
Post Construction - Setup Error Codes 31
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 30: Multi-editing of error code categories
A problem can arise if a unique error code has more meanings. An example is shown in Table 2. In
our example error code 309 appear in more variants.
Table 2: Example error code with various meanings
If windPRO identifies an error code with varying descriptions, a green menu “Separate error codes”
bar will appear. A pop-up window lists all observed error codes, which have varying descriptions. For
each description a new error code is suggested. Update error code list all events under error code 309
will be split into sub-events with a new, unique error code and a specific description.
Post Construction - Setup Error Codes 32
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 31: Separate error codes with multiple meanings
After setting up error codes with description and category, the list could look like Figure 32. Some
errors are curtailments, of which some are compensated for. In our example “Remote shut down” is a
grid curtailment, compensated by the utility.
Post Construction - Setup Error Codes 33
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 32: Example of categorized error codes
11.1.16 Import Rules
The tab Import rules give different options for handling the import. In most cases it will be ok to choose
the default option.
In some cases, the control system uses an error code called “alarm chain activated” independent of
the reason. Consequently, all errors will have the same code, which of course will prohibit any
meaningful analysis. Therefore, more choices are available.
Figure 33: Import rules
Post Construction - Setup Error Codes 34
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.17 Load/Review
Here the error codes are loaded from the error log file(s) into the existing WTG objects by pressing the
“Load” button. There are two tabs on the left-hand side: WTG time series and Error code frequency.
In WTG time series the resulting 10-minute time series is displayed (Figure 34). Please note that only
10-minute periods without any error will fall in the category “Normal operation”. If within the 10-minute
period, there is one milli-second subject to an error event, the whole 10-minute period will be marked
with that specific error code, unless otherwise specified on the Import Rules tab under “Drop error
codes with duration less than” field.
Clicking on individual 10-minute time stamps with an error code allows to see in the bottom of the
window all error codes that were active in that time stamp. It is possible to manually select (button on
the right) another error code within the 10-minute time stamp other than the one selected by the import
rules setup.
Figure 34: 10-minute time series from error log
On the tab WTG time series, the user again has access to user error codes (see Section 11.1.18), as
there might be time stamps remaining, where the WTG clearly does not operate optimally but has no
error code. Using user error codes, it is possible to manually assign an error code to these events.
On the second tab to the left Error Code Frequency, the main statistics are shown for each error
code together with the description: MTTR (mean time to repair), the period count (how many time
steps has this event been logged for), and frequency (how often has the event been logged).
A plausibility check can be performed by pressing the “Preview” button. The specific error code can be
reviewed. In Figure 35 events for yaw untwist are shown. Luckily the control system untwists the WTG
during low wind speeds.
Post Construction - Setup Error Codes 35
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
As a second example, the events which belong to error code 309 are shown in Figure 36. Not all
events are below the power curve as one might expect, since the error codes description is “remote
shutdown”. However, as explained earlier, a 10-minute time stamp is marked with an error code even
if the error has been active for a millisecond. This might have been the case in this example, where
the shutdown has only been active for a very small part of the time stamp.
However, it could also be the case that the assumptions as to which period the SCADA data refers to
were wrong. In the initial data import of the SCADA data, the user had to choose if the time-stamp
refers to beginning or the end of the 10-minute period (see Section 11.1.5). If the wrong choice has
been made, the 10-minute error table and 10-minute production table do not refer to the same time but
are shifted by 10 minutes. In such a case, the user must exit the error code set-up and go back to the
start. The data import setup must be revisited (see Section 11.1.5), the time stamp setting must be
changed, and pair & load must be repeated. The Setup Error Codes must be re-done. After reloading
the error logs files and converting them to 10-minute periods, the stop events should now concur with
zero production. It can consequently be assumed that the production and error logs are now
synchronized.
Finally, in case the user has merged error codes – the merged code can be checked for plausibility.
Figure 35: Error code statistics and plausibility check
Post Construction - Setup Error Codes 36
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 36: Error code 309 – remote shutdown: Likely wrong import setting for time stamp.
11.1.18 No Error Code Available
This feature can be used when error codes have not been provided. It can also be of use, if the error
codes have been provided, but not all events are assigned an error code. E.g. the WTG de-rates, but
the operational status shows normal operation. Please note that user error codes will NOT overwrite
imported error codes. Figure 37 it is demonstrated, as to how to use this feature where no error codes
are available, but it will work similarly if error codes are already loaded.
Figure 37: No error codes available
Post Construction - Setup Error Codes 37
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
User error codes can be defined in the Load/review tab: The button “User error codes” brings you to
the following form (Figure 38).
Figure 38: Form for defining user error codes.
Four different types of user-defined error codes are setup. Please note that you can specify from
which WTG you will see the scatter plot. The user-defined error codes however will apply to all WTGs.
Type 1 allows to identify events with stops without error code: A combination of wind speed and power
is used to find these events. The algorithm is based on that when wind speed is bigger than x m/s
(default 5 m/s) and the power is smaller than x% of rated power, one can specify a new error code
number for these events. The Preview button brings you to the scatter plot and shows the impact of
the chosen settings (Figure 39). In the scatter plot the filtered events are shown in a colour. The new,
user defined error code becomes effective after you have closed the preview and pressed Run. The
new error code will then be added to the 10-minute production time series.
Post Construction - Setup Error Codes 38
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 39: Preview: Stop without error code.
Type 2 addresses power curtailments and outliers without error codes. Two sets of parametrisations
are used: one for the steep part of the power curve below rated power and one for the flat part of the
power curve above rated wind speed.
For the steep part of the power curve events below a fictive, shifted power curve are filtered. The user
can specify if the measured, binned SCADA power curve or the contracted power curve from the WTG
catalogue is used, which is assigned to the existing WTG object. The user can also specify how much
the power curve is shifted to the right. Obviously the further the power curve is shifted to the right, the
less aggressively it is filtered. Please note that the filtering affects the measured, binned power curve.
Consequently, running the same filter once more will pinpoint a few, additional time stamps.
For the flat part of the power curve, events with x % of the rated power can be filtered.
It is possible to assign several user-defined error codes e.g. first you might want to isolate the rating at
approximately 2300 kW. Once you have found appropriate settings you can run the tool, which will add
the error code to the 10-minute time series. You can then re-visit the scatter plot and define new,
different error codes.
In the example in Figure 40 the filtering might have been too aggressive, as too many data are filtered
out - especially in the knee area of the power curve. In such a case, it is possible to go back and
adjust the settings, e.g. a bigger shift of the power curve.
Post Construction - Setup Error Codes 39
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 40: Preview: Filtering for curtailments and non-optimal performance
Type 3 is to establish an error code for operation above cut-out wind speed. The purpose of this
option is mainly to avoid these samples disturbing the comparison between measured and modelled
production from a time-varying PARK calculation.
Type 4 can be used to catch a specific operating mode which has no error code, by selecting a box
parametrised by a wind speed and a power range. Pressing preview allows you to drag & drop a box
to mark an area for filter.
The history of the user-defined error codes can be re-visited. The user can see how many samples
have been affected by which setting. The user-defined error codes can be deleted.
As mentioned above, the user defined error codes can also be used even though you have the error
code in the 10-minute SCADA data. You might want to use this option e.g. if the WTG de-rates, but
this de-rating is treated as normal operation in the SCADA system. In the bottom of the form (Figure
38) you have to confirm if you want to replace the existing error code with the user-defined error code.
With the user-error codes established, a rough loss evaluation for each WTG can now be performed
(Figure 41). The WTG can be changed in the top-right corner.
Post Construction - Setup Error Codes 40
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 41: Loss evaluation of user-defined error codes: example user code 10000
As a check, it is possible to see the events which have been assigned to a user error code by pressing
the preview button.
The user can choose to ignore the error code. Alternatively, the error can be treated as normal
operation. Both options require loading the data again (see Section 11.1.6).
Post Construction - Sensor Calibration 41
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.6 Sensor Calibration
Since the methodology behind the post construction analysis assumes consistent nacelle wind
speeds, the wind speed signal should be checked for any inconsistencies. Inconsistencies could occur
due to anemometer exchange or changes of the scaling factor of the wind speed signal in the turbine
controller and should be corrected before undertaking loss evaluations.
In Figure 42, the daily average wind speed signal of a single “Selected WTG” is compared to a daily
average reference wind speed signal, obtained from the average of all available signals for each day
of the chosen reference turbines and/or other measurement. The reference wind speed signal can
come from another single WTG or from a reference wind speed signal (e.g. mesoscale data). The
upper graph shows the two wind speed signals to be compared. The lower graph displays the selected
time series minus the reference time series with a set averaging time. The averaging time can be
annual, monthly, or daily.
Please note, that by default only data is shown where the WTG operates optimally by switching filters
on. This ensures that the comparison is not affected by a changing behaviour of the nacelle
anemometer when the WTG is switched off, de-rated or similar.
The analysis is supported by a heat map (Figure 43) and the scatter plot of power curve.
Figure 42: Sensor calibration
The heat map indicates that there is an offset between the average of all chosen reference data sets
and the turbine under investigation. It is strongly recommended to also visit the scatter plot of the
SCADA power curve for a plausibility check before concluding which signal is trustworthy and which
one is not.
Post Construction - Sensor Calibration 42
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 43: Heat map of reference versus individual wind speed
In this case, a persistent offset is found in the data. The affected time period can be marked by
pressing the right mouse button and drag to the end period. Let go of mouse right button and a pop-up
window informs about the difference in wind speed between the two signals during the marked period
(Figure 44). The user can now decide if the offset should be applied or not.
Figure 44: Example of inconsistent nacelle wind speed
Under the condition that the filters are switched on, the graph will update and display the original as
well as the resulting corrected signal. It is recommended to plausibility-check the impact on the power
curve. E.g. if you chose to compare two individual WTGs rather than using an average as reference,
using the heat map and power curve will help in understanding which of the two signals is the correct
one.
In the bottom right of the window all applied calibrations can be revisited and deleted again.
Post Construction - Sensor Calibration 43
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.19 Wind speed substitution
Wind speed substitution follows the TR10 standard method 3 for finding the best correlating reference
wind speed signal, available in a minimum 3 month running window.
It requires error codes to have been preloaded and a reliable wind direction signal.
The aim is to substitute partly disabled wind speed signals from the imported nacelle/SCADA wind
speeds or if there’s gaps in the provided data.
This method establishes a monthly correlation to all other windspeed signals loaded and in 12 x 30°
sectors, with a resulting scale and offset between each data signal loaded.
As an example, with 10 turbines and 3 years of Scada data loaded, plus one Mesoscale reference
data set; the TR10 method automatically calculates 11 data sets x 36 months x 12 sectors = a TR10
matrix consisting of 4752 correlations along with the Scale and offset between all loaded datasets for
these.
Then for any given 10min period where a wind speed signal is disabled or missing, it will select the
wind direction for this period, and find the reference which has enabled windspeed data and the
highest correlation. It will then use the derived scale and offset on the reference data to substitute over
the disabled or missing windspeed signal.
If no wind direction data is available, it can be imported as Mesoscale or possible on-site measured
wind data. This is done in the Reference database where loaded METEO objects in the windPRO
project are available, and it’s further possible to directly import online reference data sets.
Under “Setup wind direction master,” it’s possible to use one master wind direction dataset for the
TR10 evaluation on correlation matrix between all loaded datasets of nacelle wind speeds, reference
wind speeds, and direction data are based.
Once reference data and wind direction setup are chosen, pressing calculate will start windPRO’s
TR10 calculation, depending on the number of turbines and length of loaded timeseries this can take
several minutes.
During the calculation, which is parallelized, thus uses multiple cores if available, the top left window
gets updated as each sector and month is completed. When it is done, the top right table is written,
listing the combined matrix of all calculated correlations and number of months used to fulfill the TR10
minimum requirements of bins filling per month and sector. See the TR10 manual for further details on
the model and its requirements.
The combined TR10 matrix can be exported for documentation purposes.
Once the TR10 method has been completed, pressing “Substitute disabled/missing windspeed
signals”, will begin reading all loaded nacelle wind speeds for every sample where it is disabled, or a
time stamp is missing a wind speed reading in between first and last known date & time per turbine. It
will follow the TR10 method, find the highest correlating reference data which has data in the said
period, and apply the scale and offset found. This will overwrite the disabled signal or insert into
missing periods.
Note the original wind speed signal will be overwritten if disabled!!!
Post Construction - Sensor Calibration 44
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 45: Sensor calibration – Wind speed substitution : TR10 methodology
During the process of substituting the TR10 derived wind speeds, the bottom window gets updated as
the software progresses through the loaded timeseries, and graphically indicated with Green original
un-touched data. If the data have been substituted by concurrent neighbouring nacelle wind speeds or
if no turbines in the park had data, that reference windspeed data has been used instead.
If hourly mesoscale or era-interim data is used as reference, linear fit between hourly windspeed
readings are used to find the 10min windspeed substituted result.
Note!! Depending on the number of turbines and length of data set, along with the CPUs of user’s
computer, this may take several minutes to complete.
Also, if the users regret the choice of substituting the faulty/disabled or missing wind speeds, it’s only
possible to reload data from the import filters in order to revert to the original import settings. This is
done under the main “data” tab, Pair & Load, and re-selecting “Load measured data”. Here it’s
possible recover existing error codes after the fresh reloading of the original SCADA data by selecting
“Recover existing error codes after load”.
Post Construction - Create Basic PARK Calculation for losses 45
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.7 Create Basic PARK Calculation for losses
To calculate the lost production a PARK calculation has to be performed (Figure 46). It calculates the
potential production based on the historic power curves of each WTG and the nacelle wind speeds.
The nacelle wind speeds of each WTG should have been checked for consistency by comparison with
a reference wind speed (see Section 11.6). The basic time-varying PARK calculation does not require
a detailed flow model set-up. Terrain and roughness information are not required.
Prior to starting the PARK calculation, the user has the possibility to perform a more detailed
plausibility check of the wind speed signal (Figure 47). While the Sensor Calibration addresses
consistency in time, the Wind Speed Correction allows the user to check consistency for different
operational conditions: It is known that the flow around the nacelle of the WTG is different during
stand-still than during operation. Consequently, the wind speed measured on top of the nacelle might
not be comparable for these two conditions. Since the nacelle wind speed is used for calculating the
potential energy, a correction of the wind speed signal during non-optimal performance might be
required. It is possible to enter a correction for all WTGs. Alternatively, it is possible to determine a
correction factor for each individual WTG.
Figure 46: Create time-varying PARK calculation.
For Post Construction path, the PARK calculation is set up under the Data tab, selecting the Pair &
load on the left side under the sub-tab window. Here the green button “Create” allows one to initiate
the PARK calculation setup.
From windPRO 4.0, it’s possible to include Seasonal and Day & Night power curves. These historic
power curves are all binned on “normal operation data” and losses will be calculated relative to time of
Post Construction - Create Basic PARK Calculation for losses 46
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
day and period of year. The individual Power Curve (PC) can now be exported to a file or copied for
use in excel. Additionally, Air density correction can be applied to the nacelle wind speed for samples
with error codes. This is only recommended if proper concurrent measured temperature and pressure
are available.
Figure 47: Wind speed correction
The function Wind Speed Correction helps the user to make a qualified decision and allows the
correction of the wind speed. For each WTG, the nacelle wind speeds are plotted against a reference
WTG of the user’s choice as shown in Figure 48 (right plot - after selecting calculate for an individual
WTG). Typically, two neighbouring WTGs are compared. The trend line for concurrent time steps with
both WTGs operating optimally is shown (red scatter with black trend line). Additionally, another trend
line shows the relationship of the nacelle wind speeds for time steps, where the WTG under
investigation is operating non-/sub-optimally, but the reference WTG is fully operational (pink scatter
with pink trend line). As a plausibility check the scatter plot of the WTG under investigation is shown
on the left-hand side. In our example, all pink dots indeed represent non-/sub-optimal operation or
stand-still. The difference of the slope between the two trend lines is suggested as a correction factor
for the nacelle wind speed and will be applied for each time step with non-optimal performance.
The user can define specific directions to ensure that the two WTGs that are compared are not wake-
affected.
Post Construction - Create Basic PARK Calculation for losses 47
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 48: Evaluation of impact of operational condition on the nacelle wind speed
Since windPRO 4.1 two methods for wind speed correction is possible a simple scale factor, or the
new OTF “Operational transfer function” this allows for selected windspeed bins an individual scale
factor. Should the windspeed ratio to the reference indicate a clear non-linear trend, this would be the
preferred method for correcting the windspeed for stopped conditions. If this method is chosen, the
resulting OTF can be seen under view, or after accepting the WTG list is updated with the notation
“Transfer Function” is selected the individual OTF pops-up and can be copied for documentation if
needed.
Post Construction - Create Basic PARK Calculation for losses 48
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
After correction of the nacelle wind speeds, the user must pay attention that the measured power
curve is selected. If the filters are switched on (which they are by default), all time steps with non-
optimal performance are disabled. The remaining data forms the historic power curve. The button
displays the historic power curve (Figure 49). By pressing the right mouse button, it
is possible to copy the historic power curve data from the table.
The binned measured power curve as well as the contracted power curve from the existing WTG
object can be displayed together with the seasonal power curves if chosen. NB the table to left is
always the binned normal power curve for all data, the seasonal, and/or time-varying power curves
can be exported to file. It is also possible to use alternative set-ups for displaying the data (Figure 50).
The user can either display the scatter curve of a different WTG of the wind farm, choose another wind
speed signal or limit the period.
Finally, the user has to ensure that the second tick is enabled to “Replace calculated production with
measured for non-error code samples”.
Post Construction - Statistics 49
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 49: Historic power curve
Figure 50: Alternative set-up for historic power curve display
After the setup of the calculation is done, or if changes are made in filters or periods, a run of the
chosen setup needs to be initiated by the Run button.
11.8 Statistics
After having loaded the 10-minute time series, having combined it with error codes, description of error
codes and their categorisation, the recovery rate and some basic statistics can be found in the
Statistics tab.
Post Construction - Statistics 50
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
For each signal and for each WTG, the count, recovery rate, mean, standard deviation, minimum, and
maximum are shown (Figure 51).
For each signal for each WTG the monthly means can be visited (Figure 52).
Finally, the recovery rate per day can be visited for each WTG (Figure 53). It is possible to mark
specific samples in the table and change the status from disabled to enabled or vice versa.
Figure 51: Main statistics
Post Construction - Statistics 51
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 52: Monthly mean
Figure 53: Daily recovery rate
Note: if the new TR10 methodology for substitution of faulty or missing wind speeds have been used,
the wind speed availability reflects the current availability of the wind speeds with no distinction
between original or substituted wind speeds.
Post Construction - Analysis 52
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.9 Analysis
In the Analysis tab it is now possible to visualise the data.
11.1.20 Analysis – Time series
All loaded signals from each WTG can be displayed in various ways, either as time series in the
temporal resolution they have been loaded in, as monthly values, as a diurnal plot or as radar plot.
The functionality is very similar to the tools in the Meteo object.
In Figure 54 the time series of wind direction and wind speeds are shown together with the production
from two WTGs. The user can adjust how many days are displayed in the window. By clicking and
holding the left mouse button it is possible to zoom in a specific area. It is also possible to display an
averaged timeline.
Further it is possible to manually enable/disable data. Specifically, the wind speed should be checked
as the potential production is later determined using the nacelle anemometer. For example, the
nacelle anemometer could be suffering from icing and shows zero or a constant small value, this data
should be corrected or disabled. The methodology assumes consistent nacelle wind speeds, thus
jumps occurring from anemometer exchange or changes of the scaling factor of the wind speed signal
in the turbine controller should be corrected before undertaking loss evaluations (see Section 11.7).
A number of extra functions become accessible when the user has prepared a time-varying PARK
calculation. It is possible to compare the measured production data with modelled production data. As
preparation, the user must run a time-varying PARK calculation. There are two concepts available:
In case of post construction, the nacelle wind speed and historic power curve of the individual turbines
will be used to calculate for each time step what the WTG should have produced. The time-varying
PARK calculation does not require any flow modelling in this case. It can be started from within the
Performance Check module, see Section 11.7.
In case of Performance Check, a full flow model has to be set-up. As wind speeds, it is possible to use
local masts as well as mesoscale data, see page 76.
If a time-varying PARK calculation has been performed, the thin line series represents the modelled
signal for periods where there has been an error code. If the filters (top right corner) are switched off,
also the corresponding measured values are visible as thick line.
Post Construction - Analysis 53
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 54: Display time series
By default, the filters are switched on and all error codes other than zero are filtered out (Figure 55).
This is necessary to calculate the potential production later in the process. It is possible to define
additional filters. For example, the user might want to investigate a specific direction and/or a specific
wind direction range. “Inside Range” means that the data inside the range will be filtered out. If a
PARK calculation has been loaded it is also possible to add filters to the modelled signals.
Please note, that if you use the “Excel” button on the left-hand side of the graph all data including the
filtered data will be copied.
If sensors have extra calibration, they can be applied here (Figure 55). Please note that sensor
calibrations are explained in Section 11.6.
It is possible to define the minimum number of WTGs that must be operating simultaneously.
It is also possible to filter events out depending on how much the measured production deviates from
the modelled production.
Figure 55: Application of filters
Post Construction - Analysis 54
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.21 Analysis – Error Loss Statistics
As explained in Section 11.5 the user can either combine the 10-minute SCADA time series with error
codes and their description or as a fall-back plan set up user-defined errors in case error codes are not
available. Depending on which one of these two options is selected, the tab Error Loss Statistics
contains different levels of details.
Measured negative production is not accounted as a loss, but the amount is informed in the top right
corner of the window. In addition, the following are shown:
MTBF (mean time between failures): the average time between the error code’s appearance and
disappearance.
MTTR (mean time to repair): the average length of a specific error code.
Count: the number of periods with a specific error code (the count is independent of the length of the
individual event; the count is 1 no matter if the event is one second or one weeklong)
If the user wants to change the categorisation if losses are compensated or not, it is possible to re-visit
Setup Error Codes. It can be decided to exclude each of these in the sum.
It is possible to limit the period to see the losses within e.g. a specific month or year. Note: it’s
recommended to set the time to cover complete years for later use in long-term correction, to avoid
potential seasonal bias.
It is also possible just to look a specific turbine or a specific category or sub-category.
With detailed error codes loaded, the above-described information becomes available for each
individual error code (Figure 56).
Figure 56: Error loss statistics – detailed error codes
Post Construction - Analysis 55
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.22 Analysis - Error Loss Matrix
In this tab (Figure 57) functionalities similar to a Pivot table can be found. Different data fields have
been defined, which can be dragged and dropped freely in the table section to aggregate losses. The
losses can be presented as absolute numbers or percentage (radio button selection on top). Following
data fields are available:
• Year
• Month
• WTG
• Category
With the two last listed data fields (WTG and Category), it is possible to deselect specific WTGs or
categories and thereby only show a subset of the data by clicking on the arrow at the data field (Figure
58).
Figure 57: Error loss matrix
Figure 58: Selecting WTGs and categories.
Post Construction - Analysis 56
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Further, when for example dragging additionally Month into table set-up, the Expand All and Collapse
All button facilitate to reveal analysis such as monthly production per WTG quickly and easily.
The Reset Layout to Default button brings the default pivot set-up back.
The graphical display is fully customisable. The user can change from bar charts to pie charts, etc.
11.1.23 Analysis – Power Curve
Before explaining the functionalities on this tab, it must be stressed that visualised power curve is not
necessarily identical with the contracted power curve provided by the manufacturer. The contracted
power curve refers to the free wind speed, while the power curve displayed here is based on the
nacelle wind speed and does consequently not follow the IEC 61400-12-1. The IEC 61400-12-1 allows
the validation of the power curve through use of the nacelle anemometer. However, this standard
requires a very strict and precise measurement setup which is very rare for a commercial operating
wind farm.
Figure 59: Power Curve tab
In the top left corner (Figure 59), the AEP table gives a quick overview of the expected production for a
Rayleigh distribution for a user-defined wind speed. The measured power curve (normally based on
the nacelle wind speed) can be compared with the contracted power curve. Of course, this only makes
sense if the used wind speed signal is identical with free wind speed, which is not the case if the
nacelle wind speed is used. Please note, that not all high wind speed bins might contain enough data
to derive the full measured power curve. In such a case the power curve is extrapolated to the cut-out
wind speed as specified at the power curve in the WTG object.
Following display options are available (starting from the top):
It is possible to use alternative set-ups for displaying the data (Figure 50). The user can either display
additionally the scatter curve of a different WTG of the wind farm, choose another wind speed signal or
limit the period. This can be useful for example for comparing two binned power curves from two
Post Construction - Analysis 57
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
different WTGs. If two WTGs are chosen here, the AEP table will update accordingly and display both
AEPs for the defined Rayleigh distribution. The user can also compare two periods with each other.
The measured binned power curve as well as the contracted power curve from the existing WTG
object can be displayed. Please note, that the contracted power curve is valid for one, fixed air-
density. Also, the calculated production can be displayed. If the user has run the time-varying PARK
calculation as part of the post construction process, the calculated production corresponds to the
potential production. If the user has run a full flow model-based time-varying PARK calculation (see
Section 11.4), the calculated production can optionally include site-specific correction for turbulence,
shear, air-density etc.
The wind speed signal which is used for the scatter plot can be altered. By default, it is using the
calculated wind speed. The difference between calculated wind speed and the wind speed in the WTG
object consists of any correction the user might have applied to the nacelle wind speed, like described
in Section 11.6 or wind speed correction as described in Section 11.7. If the user has run a full flow
model-based time-varying PARK calculation (see Section 11.4), the calculated wind speed comes
from the flow model. Additionally, the user can load data from any Meteo object. Of course, all data
can be exported to Excel for further analysis e.g. advanced comparison for different temperatures or
turbulence intensities.
Post Construction - Analysis 58
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.24 Analysis – XY Graph
With this tool, any measured signal can be plotted against each other, either as scatter plot or binned.
The x- and y-values can be from the same or different turbines, and any signal imported can be
chosen.
The period can be reduced to see if a specific phenomenon belongs to a specific period.
Averaging in time can give a better overview than having a large “cloud of points”.
Finally, a trend line can be shown (limited to linear) – the formula will appear in the legend.
Figure 60 shows an example how the direction measurements from the one WTG differ from the other
WTG.
An example of operational strategy visualization based on the XY graph tool is given in Figure 61,
where RPM is plotted against pitch angle with wind speed as colour scale.
Figure 60: Example XY graph - nacelle directions
Post Construction - Monthly Gross 60
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.10 Monthly Gross
The tab Monthly Gross aggregates the production data to a monthly time series, which is the basis for
long-term correcting the production data. Depending on the available input data, different methods can
be chosen to compensate for incomplete time series e.g. missing periods or unreliable nacelle wind
speed. The output of this process is the GROSS production “Monthly Potential production”, which is
fed into the WCP tool (wind corrected production = long-term corrected production) for long-term
energy index correction, see Section 11.1.25. The resulting time series of monthly production will be
saved in the Existing WTG object with the label Month Gross.
The user has to choose a time-varying PARK calculation, which is based on the nacelle wind speed
(see Section 11.7).
Four different methods can be chosen to compensate for missing data. Note the recovery rate is
based on the enabled pairs of nacelle wind speed and production.
Divide with recovery rate: Recommended for data recovery rate > 95%. The missing data is
assumed to be caused by data loss. During the period of data loss, the WTG is assumed to produce
as observed on average for the available data in each month.
Ignore recovery rate: Recommended for data recovery rate > 99% and when TR10 data substitution
have been used. The missing data is assumed to represent periods with lost production, which is also
representative for the future.
Skip month if recovery rate is too low: If this option is chosen a field is activated where the user can
define the minimum recovery rate, which only affects the long-term correction.
Take from alternative calculation: This is not to be used for post construction methodology.
Figure 62: Monthly gross tab
Further, two methods become available if monthly data has been read directly into the Existing WTG
object. These options are greyed out in the post construction process if none of the required monthly
data has been loaded into the Existing WTG object.
Post Construction - Monthly Gross 61
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Method 2: Month NET + calculated loss: If the monthly NET (sold production) is loaded into the
Existing WTG object, it is possible to evaluate the grid losses. Please note that 100% availability in the
other methods does not include the grid loss.
Method 3: Monthly RAW divided with month availability: If the monthly RAW is loaded into the
Existing WTG object and includes monthly availability figures, the production can be availability-
corrected by division with the availability. This very simple method assumes there is a column with
availability information in the RAW data. The RAW data can both be with or without grid loss, this must
be handled in the later loss settings when creating a report.
Pressing aggregates the 10-minute data to monthly Gross and saves the
monthly data to the Existing WTG objects. A graph appears on the right-hand side showing the actual
production (called “measurement”) and the actual loss (called “substituted”).
The user can jump directly to the Wind Energy Index tab.
Post Construction - Monthly Gross 62
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.25 Analysis – Wind Energy Index
Note: For a new project setup, the user must first aggregate Monthly Gross production, before this tab
is available( see Section 11.10).
This tab allows the user to analyse the long-term correction of the potential production for each WTG.
There are many different options available, but the basic concept is simple: Based on monthly
production and monthly wind energy indexes, the Wind Index Corrected Production (WCP) is
calculated.
There are two basic concepts for performing this calculation:
The sum method: The sum of production is divided with the sum of index for the enabled months.
The regression method: A linear regression line is established for the monthly potential production
versus monthly wind energy index, and the production for an index value of 100% is found.
The two methods deliver very similar results. Only for shorter periods and/or poor correlation with the
index, will the results from the two methods differ.
Figure 63: Wind Energy Index tab
The user can choose and edit the data base of wind indices (see Section 11.1.8). The indices which
are part of the chosen database can be selected individually. This could be e.g. several re-analysis
and mesoscale indices. The correlation (R
2
) of the different indices with the monthly potential gross
production can be seen in the AEP table in the top left corner. Next to the correlation, the WCP (Wind
Index Corrected Production), the annual production per MW installed, the capacity factor, the length of
the available data, and the usage of that data are shown. Please note, that all production at this point
reflect 100% availability as all times steps with any form of non-optimal performance have been
identified and corrected for.
Following graphical displays can be chosen from.
Energy Production vs wind index plots the monthly potential gross production against a chosen
index (see ).
Post Construction - Monthly Gross 63
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Time series RAW plots the monthly potential gross production per month. Additionally, the monthly
index is shown referring to the secondary y-axis. Ideally the two curves should follow each other.
Time series WCP shows the index-corrected production per month. The index-corrected production
resembles a combination of the previous graph showing the monthly production and the index. This
line should be ideally horizontal. Any deviations from the horizontal could point towards
inconsistencies during time. Any slope could point towards degradation in performance or other
issues. A step change could indicate that a neighbouring wind farm has been commissioned.
Time series WCP/MW shows the index-corrected production per MW per month.
WCP vs wind index shows the index-corrected production against the wind energy index. Ideally the
values should all exhibit the same production. Any slope indicates that the correlation for low and high
wind speeds differs. This is typically the case when the index has been developed based on the wrong
wind speed assumption (see Section 11.1.8 and Figure 18). The precise set-up of the wind energy
index is more important when the production data only covers a short period.
The following option show results for all WTGs:
All WTGs, AEP Avg + WCP shows the measured AEP, and the wind energy index corrected
production for each WTG. If the production data spans a different period than an integer number of
years, the measured production is corrected by dividing the production with the number of months and
multiplying with 12 months.
All WTGs, AEP Avg + WCP/MW shows the measured AEP per MW and the wind energy index
corrected production per MW for each WTG.
All WTGs, Cap. Factor shows the measured capacity factor, and the wind energy index corrected
capacity factor for each WTG.
All WTGs, Correlation shows the correlation (R
2
) between monthly gross production per WTG and
wind energy index. Ideally this should be 1.
WF GEO shows the map of the WTGs under investigation. This graph is most meaningful if a full flow-
model-based time-varying PARK calculation has been performed (see Section 11.4)
For some of the graphical displays further options are available like increasing the averaging from 1
month to several or showing the accumulated average WCP.
For the regression plot, a simple tool can help to identify outliers by defining a maximum deviation
from the WCP in %. For example, 90% means that the actual monthly production should not be more
than 90% different than the WCP of that very month.
Further, it is possible to limit the period, e.g. to filter out the first months of operation, because they
could have some poorer performance.
A scaling option for the data is available.
The WCP loss form is not relevant in the context of post construction analysis.
Post Construction - Report 64
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.11 Report
The post construction report highlights the actual production with actual losses as covered by the 10-
minute SCADA data. The potential production represents 100% availability. The actual losses are
ranked, the lost kWh, MTBF (mean time between failure) and MTTR (mean time to repair) are given.
Long-term correction of the potential power results in the normalised production. Including future
losses leads to the future expected Net production. The process follows the IEC 61400-26-1 structure
using the historic power curve and nacelle wind speed.
The report consists of following sections:
• Main
• Historic Power Curve
• Actual Losses
• Normalized Production
• NET Production
• Comments
Figure 64: Post construction report
11.1.26 Main section
The Main section starts with an explanation of calculation method with the glossary. A table with the
main results follows (Figure 65). The table includes the potential production of the analysed period
(which could be different from one year) as well as an annual basis simply by correcting for the
number of months. Please note, that having, for example 1½ years of data with two summers and one
winter can make the annual production look out of proportion due to seasonal bias. It is recommended
to use complete years to avoid any misleading seasonal bias.
Post Construction - Report 65
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 65: Main results
11.1.27 Historic Power Curve
The table with the results is followed by the scatter plot and the historic power curve (Figure 66). The
filtered events are marked in a lighter colour.
The user can choose in the tab Turbine selection – historic power curve which WTG to display (Figure
67).
Figure 66: Example historic power curve based on the nacelle wind speed.
Post Construction - Report 66
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 67: WTG selection historic power curve
11.1.28 Actual Losses
The next pages show the actual losses aggregated in different ways: by WTG, by category, by error
code, and by month (Figure 68 and Figure 69). Combination of actual production and actual losses
leads to the potential production which is summarized in a table (Figure 70).
Figure 68: Actual losses by category and WTG
Post Construction - Report 67
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 69: Actual losses by month and by error code
Figure 70: Potential Production
11.1.29 Normalized Production
This section of the report shows the normalized production. At the top of the page the normalized
production per year and per MW/y is given. The normalized production is assumed to have 100%
availability.
The monthly potential production versus the wind energy index, the time series of the production, and
the chosen wind energy index as well as the timeline are displayed (Figure 71).
Post Construction - Report 68
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 71: Normalized production - wind energy index correction
11.1.30 NET Production
The NET production requires the quantification of future losses (Figure 72). The actual losses are by
default transferred as future losses; however, the user can modify these values, either manually one-
by-one or through definition of the loss of a category. For the second option the Apply button has to be
pressed to actively change the settings. An electrical loss default of 1% is assumed, with degradation
0.5% which leans towards the FGW TR6.
The final table (Figure 73) summarises the NET production and lists all assumptions.
Figure 72: Quantification of future losses
Post Construction - Report 69
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 73: NET Production
11.1.31 Export
The user can export either the key figures or the time series to the clipboard or to a file (Figure 74).
Data that have been disabled or filtered can be included upon request.
Figure 74: Export options
Post Construction - Chapter 2: Monthly production to long-term yield 70
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.12 Chapter 2: Monthly production to long-term yield
The simplest and very robust approach within Performance check module is to wind index correct
measured monthly production data. It is although important to notice that a wind energy index can be
biased and thereby could lead to wrong results!
11.1.32 Wind Energy Index Correction
Data setups see: 11.3
Load PERFORMANCE CHECK module from the main Tab, Loads & Operation, and
Performance Check.
Set up import filter for measured production data if data are not already loaded in existing WTG
objects. See Section 11.1.5
Decide which existing WTG objects to include in analysis (assumed created in project).
Pair WTG ID’s from the import setup to existing turbine objects and load (import) the data.
Go to “Analysis” > choose “Wind index WCP”
Select a wind index (see section: 11.1.8)
Check the correlation; utilize the filters, availability settings etc. to establish a trustworthy Wind
Index Corrected Production (WCP) = long term production expectation for each turbine in
analysis.
Use the different graph views to extract results to Excel or print the reports for documentation,
includes the long-term expectations and also a coarse loss evaluation and prognoses with P50,
P90 etc. (WCP report)
11.1.33 Analysis – Monthly time resolution
Very often the data basis which the evaluation of the performance of operating wind farms is done is
just on monthly basis. This is also quite good, if the data period is reasonable in length and there are
well established methods for performing performance check on monthly basis, like the wind energy
index method. Higher time resolution data can also be loaded as described in previous chapters.
Monthly data can be with or without grid losses. This is important to know when the calculation model
is fine-tuned.
With monthly data it is possible to “just” calculate expected long term production (wind energy index
corrected). But it is also possible to compare with model calculations based on time-varying PARK
calculations where results are aggregated on monthly level.
Figure 75 Monthly data work path
Post Construction - Chapter 2: Monthly production to long-term yield 71
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 76 Normalized monthly production data.
With monthly production data imported with or without availability, the monthly production is paired to
the wind energy index, and the WCP is calculated. This gives the expected long-term yield based on
correlation between wind energy index and the observed production. In Figure 76, in blue is measured
monthly production, green is the wind energy index from the Danish index model, and in red is the
accumulated average wind corrected production, showing that this turbine has a very stable yield over
the 8 years production period.
The WCP (Wind index Calculated Production = Long term expected production based on wind index
correction method) calculator has many refinements. But the basic concept is simple: Based on
monthly production and monthly wind energy indexes, the Wind Index Corrected production is
calculated. There are two basic concepts for performing this calculation:
The sum method: The sum of productions is divided with the sum of index’s (for non-filtered months).
The regression method: A linear regression line is established for production vs index, and the
production for index 100% is found.
We have performed numerous analyses with the two different methods, and we cannot in general say
which is the best – they both have advantages. If data quality is good and a long period of data is
available, the results will be almost identical. It is when having shorter data periods and/or poor
correlation, the two methods can differ.
NB the WCP is the expected P50 long term yield of the evaluated turbine. It’s important to know if it’s
the raw power from SCADA or from the Grid metering point; if it’s including availability or not to assess
possible losses imposed on the WCP value that is calculated.
It’s also possible to load a time-varying PARK calculation based on monthly aggregated results), This
PARK calculation shall include the existing turbines under investigation. Then it’s possible to evaluate
if the wind model is over- or under-estimating the actual yield. In the Scaler it’s then possible to make
changes to the mean wind speed, and/or the seasonal, possible wake settings, and iterate the
comparison of the wind model to the actual production observed.
Post Construction - Chapter 2: Monthly production to long-term yield 72
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.34 WCP – loss form
The loss form is a small tool for making assessments of losses for a specific period. This can be used
e.g. to assess if the availability information from the control system is in the right size/order; or to
assess how large the loss was in a period with specific problems, e.g. power curtailment. To base loss
assessments on wind index is a coarse method. Much more refined methods are described using the
analyser (require more detailed production data).
Figure 77: WCP Loss calculation
Three different “basic settings” that can be used:
1. NO filter on WCP calculation. All months are used, but with full availability correction (if
availability data are imported or an availability figure for all periods is entered in result table)
2. With filter on WCP calculation. Months not meeting the filter criteria are filtered out in the WCP
calculation. For the remaining months, availability-based correction is performed as described
above.
3. With filter on WCP calculation AND filters are used within the period in investigation.
Here only months not disabled by filter criteria are included in the loss evaluation form.
Based on the calculated WCP from all non-filtered data, and the wind index for the months with
available production data, the expected production for the months in investigation are calculated. This
gives measured production and the expected loss as the difference WCP – measured for the same
months.
Post Construction - Chapter 2: Monthly production to long-term yield 73
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Due to the wind index is describing the production expectation relatively coarse, the calculated loss
will also be somewhat coarse. Due to this, it can be negative for some periods. The longer the period,
the more reliable.
With availability information for each month, the calculated availability-related loss for the months in
investigation is shown separately. If this reflects the found loss precisely, the field “other loss” then
becomes 0. Therefore, it will be 0 if all data used for calculating the WCP is used. This is “self-
fulfilment”, while the calculation method assumes losses equal to the informed availability loss taking
all data.
Whether the data does or does not include grid losses, it can be set in the drop-down box. This is just
an info field. If grid losses are included (data from sales metering used), the WCP is the expected
annual sale. If data from turbine controller is used, the WCP will typically be higher and assumed grid
losses must be subtracted from WCP to get expected annual energy sale.
11.1.35 WCP - reports
When having established the settings giving “comfortable results”, reports can be printed.
There are 3 report pages from the WCP tool:
1. Main – assumptions and results including graphics.
2. Losses – Loss tables and graphics giving “rough” loss estimates.
3. Prognosis – Expected monthly productions incl. probability of exceedance based on user
specifications.
Post Construction - Chapter 2: Monthly production to long-term yield 74
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 78 WCP report page 1
The main page is where the results from the WCP tool are presented. Note that even though there is a
very good correlation, this is no guarantee the result really represents the long-term production. This
requires that the wind energy index is correctly calibrated.
Post Construction - Chapter 2: Monthly production to long-term yield 75
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 79 WCP report losses & prognosis
The loss page gives a year-by-year table, where the measured production subtracted from the (wind
index x WCP) based production gives the loss. For recent 12 months, a similar table is visible. As
seen in the table of recent 12 months loss calculation, negative losses appear which illustrates the
limitation of the wind index accuracy. But it shows how approximately 300 MWh were lost December-
2013. Especially 2010 and 2013 were years with loss above what should be; and 2014 does not look
too good either. 2011-2012 were seen better with low losses and good wind. It can also be seen that
low winds were worse than the losses during 2010 and 2013.
The prognosis page show what to expect for a future year – based on a guess on the wind, where a
100% wind index normally is the most reasonable guess. Also, the loss is a guess, with a value of 3 %
here.
With these two guesses, the presentation shows like P90, which is based on wind index variation
calculations. There is 90% probability that at least 2602 MWh will be the outcome in 2015. But most
likely (P50) 2997 MWh will be the outcome after loss deduction. The expectations are shown month by
month, again based on historical wind index variations.
The “trumpet graph” shows how well the wind index correction works at this specific turbine. Having
only 6 months of data, up to +/-10% error could be seen when having the worst period seen during the
almost 5 year of operation. With a full year of data, the maximum error that could be seen is round +/-
4%. This illustrates how well an index corrected production can be trusted.
Post Construction - Chapter 3 Model calibration & Performance Check 76
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.13 Chapter 3 Model calibration & Performance Check
The basic concept behind the Model validation / Calibration / Performance check path is:
Evaluate/calibrate PARK calculation against measured production based on time step PARK
calculation.
Input can be external mast(s) or nacelle wind speed + direction loaded in Meteo object. The wind
speed can be wake cleaned in Meteo object OR data from more masts/WTGs can be merged, so only
wake free wind is used.
Based on the comprehensive analyser in Performance Check, it can be found which adjustments to
the PARK model setup that will be needed to get a good match, e.g. adjust the Wake Decay Constant.
The gains by this method:
• Wake losses can be quantified quite precisely.
• Model calibration experience can be used for calculation of EYA on new wind farms.
• Loss evaluation does not require all WTGs to have good wind speed signals and the problem
with biased wind speeds when turbines are stopped can be handled.
The calibrated model can be reused with mesoscale modelled winds (or initially based on); then the
modelled winds can be calibrated using the Scaler for the period with production data. With such a
setup, long term expectation can be calculated very precisely using long term mesoscale data (10-
20y), where e.g. direction distribution is long term representative.
All variants can be run for the same park, but in separate “sessions”. When starting the tool, the
session name must be entered by giving it a name e.g. Southern windfarm. More sessions of each
type can be setup for e.g. When different data filtering is allocated to each session or different periods
with different number of turbines operating. It should although be mentioned that a PARK calculation
can utilize Start/Stop dates within each WTG object and thereby a performance check can be
performed as ONE session on wind farms where different turbines are in operation during the period
with data. PARK calculation also handles like day/night operation modes (noise reduction) or other
curtailments like sector management.
The wind energy index database must be mentioned as an important part of the basic concept. A
comprehensive wind energy index database concept is developed. This handles as well “official” wind
energy index is as it has its own index calculator based on wind data. The wind data can be
downloaded in METEO objects from the ONLINE data service, where datasets such as MERRA2 data
or the ERA5 data can be used for creating site- and turbine-specific wind energy indices.
11.1.36 Model Calibration, loss estimate and long-term expectation.
The more refined use of Performance Check is to use it in combination with a PARK calculation. The
measured performance is compared to the calculated-on time step basis, and thereby comparisons
can be performed by different aggregations, like by direction. Problems in the calculation model or
model setup or scaling of mesoscale wind data can be identified and corrected.
Set up and run a time-varying PARK calculation (must be based on “existing WTG objects.”).
Note that the aggregation level in the PARK setup should match the available production data
(like monthly or hourly or 10 min.). The wind data for the PARK calculation can be from one or
more measurement masts at site or from nacelle measurements or from Meso/Model data. BUT
note the wind data must be “wake free,” which might require some preprocessing, like merging
data from more sources by direction or utilizing the wake cleaning option in Meteo object.
Post Construction - Chapter 3 Model calibration & Performance Check 77
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Arrange the measured production data etc. in ASCII text files ready for import.
Load PERFORMANCE CHECK module from main Tab, Loads & Operation.
Setup import filter for measured production data if data are not already loaded in existing WTG
objects.
Decide which existing WTG objects to include in analysis.
PAIR WTG ID’s from the import setup to existing turbine objects and load (import) the measured
data.
Load PARK calculation – now concurrent calculated and measured data are available by time
step.
Load error codes if available, either as part of the production data files or separate from error
log, or if no error code data available, use the “user error code” to assign error codes to events
with stop at wind > x m/s or production below shifted power curve (sub-optimal
performance/power curtailment).
Utilize the analyzer and other tools to get an idea if some model setup must be revised or data
cleaned further. E.g., modify the PARK calculation setup and run this again. Use analyzer to see
if modifications worked. Continue this process until results are satisfying.
Extract results from graphs, like loss, to Excel, to build up the documentation (no reporting yet
available within this part of Performance Check)
If e.g. a Meso data-based calibration is performed, now run a new PARK calculation with found
calibrations, based on for e.g. 20 years of mesoscale data to calculate the expected long-term
AEP, wake losses etc.
11.1.37 Time Varying PARK Calculation for Performance Check
The time-varying PARK calculation is essential for the performance check analysis. There are two
ways to go:
• A traditional PARK time step calculation, where the models calculate free and wake reduced
wind speeds at each WTG position from which the power is calculated with optional corrections
(air density, Turbulence, shear etc.).
• Nacelle wind-based calculation (no models involved, see Post construction path manual).
Setting up a traditional time-varying calculation model where it is included to offer the user the
possibility to check if the model calculation performs well aggregated e.g. by wind direction, wind
speed, season, etc.
The basic concept is to scale the wind data for each time stamp to each turbine position. The transfer
function (by direction) between the wind data source (measurement(s) or Meso scale data) and turbine
position is calculated and used for each time stamp in the used wind data source.
The transfer function is calculated by performing a WAsP or WAsP-CFD or FLOWRES based
calculation on a generic wind distribution at each turbine position in hub height. The ratios between the
calculated mean wind speeds (or Weibull A parameters) are the scaling factors.
Refined options like RIX correction or displacement heights can be part of the calculation. Also,
turbulence intensity (TI) data are transferred to WTG positions by assuming a constant st.dev. by
height, the change in wind speed thereby leads to change in turbulence intensity at each WTG, or by
using a model calculated TI. The turbulence can be used for each time step for the power curve
correction or as an input for advanced wake loss calculation. Similarly, the air density can be
calculated at each time stamp based on temperature and/or pressure data for correction of the power
curve. Finally, also shear and veer correction of power curves are optional.
Using mesoscale data gives access to a refined downscaling (taking out meso terrain, applying micro
terrain)., Another advantage of mesoscale data is that the shear variations in time are included in the
calculation. See more details on calculation method in Chapter 3 ENERGY.
Post Construction - Chapter 3 Model calibration & Performance Check 78
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.38 Wind data.
As input for the time-varying PARK calculation, wind speed and direction for one or more specific
positions in METEO object(s) can be used. This can be data from measurement masts,
LIDAR/SODAR, nacelle measurements, or modelled data. It must although be data that are
concurrent to the turbine operation in time.
If wind data is from a local mast within the vicinity of another wind farm or is from WTG nacelles, wake
reductions are included, whereas the calculation model assumes the free wind speed. This can be
handled by a quite comprehensive data cleaning where the wake reductions are “removed”.
An alternative is to combine data from more sources by direction, where only wake free
measurements are used. Finally, the scaling by direction can be a simple way to compensate for the
wake influence if the data are waked, and similarly this could also be used for removing issues such
as mast shadow. For air density correction, it is often convenient to take temperature and turbulence
from mesoscale data. Even though these are at hourly resolution, they can be used with 10 min. data,
while an interpolation is automatically performed.
Although, turbulence data from modelled datasets can be tricky to rely on. Having local measured data
is of course the best, but there is no simple way to correct for wake influence of TI. Only merging data
from free measurements will give reliable data.
They can be taken from mesoscale data, but there a quality check that should be performed. They
could be very wrong, while the mesoscale data do not capture the mechanical turbulence reasonably -
due to the coarse resolution of the terrain data.
11.1.39 Calculation setup traditional PARK calculation
The time-varying calculation is selected in the PARK module using the following highlighted option as
shown below:
Figure 80 Park module start time varying calculation.
The two right most icons allow to run a time-varying calculation.
Post Construction - Chapter 3 Model calibration & Performance Check 79
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 81 Time varying park calculation setup
At the SETUP tab (Figure 81), the output for Performance check is defined within the shown red
square.
Within the green square is seen a number of more advanced options, like:
• Use time dependent power curves, like night noise-reduced power curve
• Use curtailments based on settings within the WTG objects (can be edited from PARK
calculaiton setup too)
By default all turbines (existing) in the calculation are included in the output, and the data are
aggregated to monthly values. By monthly aggregation, the result file (and thereby the project file size)
is heavily reduced. But if 10-min or hourly production data are available for doing Performance check,
the resolution shall be set to match the measurements, normally by setting aggregation to “none”.
Post Construction - Chapter 3 Model calibration & Performance Check 80
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 82 Scaling of time varying park setup
At the SCALING tab it is decided which wind data source(s) is to be used for time varying calculation.
A very important component is the SCALER.
The SCALER decides how the transfer function between wind data position and turbine positions are
established. This can be based on the WAsP model or the WAsP-CFD model or other CFD modelled
results.
It holds advanced features like mesoscale data downscaling, RIX correction, displacement height
calculation, and the post-calibration scaling option, which makes it possible to adjust the transfer
function by direction, month, hour, and/or wind speed. But most importantly the “main scaling”, which
is the place where the general level of the wind speeds is adjusted to make the calculation match
measurements. The reason for this need is mainly when using mesoscale data, which typically will be
biased some relative value to the real wind. For more information on the scaler, see Chapter 3
ENERGY.
11.1.40 Data preparation
See section: 11.3
Post Construction - Chapter 3 Model calibration & Performance Check 81
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
11.1.41 Performance Check – Model calibration start
Activate the Performance Check module from the main Tab “Loads and Operation”.
Start with creating a new “session” or open a previously created.
Choose the desired work path, (TR10 requires separate license to be activated)
Figure 83 Create new Performance Check session.
Figure 84 Model validation & calibration
Post Construction - Chapter 3 Model calibration & Performance Check 82
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Starting a new session requires for the user to decide if new data are to be imported into the existing
WTG objects, or already imported shall be used. Note: for hourly data or other time resolution, select
the 10 min Scada data option.
11.1.42 Detailed 10-min / hourly data vs model
A performance check session based on detailed data gives access to the power curve tool and the
analyser tool, where a time-varying PARK calculation must be performed with non-aggregated data –
or aggregated to the level where measured data are available.
For most of the settings and work flow it’s identical to the Post Construction methodology.
Data Import See section: 11.1.5
11.1.43 Data – Pair and load
Load PARK calc. – load the results from the PARK calculation shown next to the “auto pair” button
into the Performance check session. Later on while doing the analyses, an alternative PARK
calculation can be selected and can be loaded and replace currently loaded PARK calculation. (Note
that PARK calculation can also be created within Performance check tool, edited, and rerun).
• Setup Error codes see: 11.5
• Sensor Calibration see: 11.6
• Data Time series: See 11.1.7
• Data time shift see: 0
• Wind Energy Index database see: 11.1.8
• Statistic Tab see: 11.8
11.1.44 Analysis of detailed time resolution
Detailed time resolution is referred to as being more detailed than 1 month. It would typically be daily,
hourly, or 10-min data. These data typically will come from Scada systems or turbine controllers. This
normally means that grid loss is not included in data. These data are especially well-suited for
checking details in the model calculations, partly while data can be aggregated by wind direction, and
for power curve validation, where data should have a preferable resolution of 10 minutes. They can
also be used for detailed Post-Construction analysis to quantify losses from the error event logs, and
to derive Potential production (100% availability), and based on historic performance estimate future
NET Production.
11.1.45 Time series
Inspecting time series gives the possibility to screen the data quality on a very detailed level. It is
especially interesting if a time-varying PARK calculation is included. It is also interesting for Post-
construction, through the use of the historic power curve derived for each turbine’s SCADA data, when
the turbine is only in normal operation mode.
Post Construction - Chapter 3 Model calibration & Performance Check 83
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 85 Time Varying Park vs Measurements
Figure 85 shows an example of time variation graphics. Both measured and calculated data are
shown. Thick line for measured, thin for calculated. A checkbox can be used to only use calculated
direction as well. This is normally recommended, especially to make the radar graph make sense. Just
a small turn between measured and calculated makes the radar graph non-comparable. By checking
“use only calculated”, concurrent time stamps are compared.
Data evaluation based on different aggregations can be performed like in Meteo object.
ONE difference from using Meteo object is that we Do NOT disable “poor performance” data here. An
important part of the analysis is to identify losses due to poor performance. How to handle this by the
FILTERS is explained later.
Only erroneous data should be disabled. Erroneous data are data that do not represent how the
turbine did produce. An example is when the Scada data show exact same production from many time
stamps in a row. This is also called “frozen signal” – the computer does not get new input but just
continue to store the most recent value it got for the new time stamps.
11.1.46 Power curve
It is of high importance to notice that validation of the power curve and whether it is within the
specifications as provided by manufacturer cannot be expected, unless a very precise measurement
setup is arranged (see IEC 61400-12-1 standard). This is very rare to find within a commercial
operating wind farm.
The power curve evaluation tool is more for finding changes in time, e.g. before and after a
modification of the turbine and more such analyses.
Post Construction - Chapter 3 Model calibration & Performance Check 84
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 86 Power curve
The basis selections are: Wind speed signal: choosing the WTG object data, the wind speeds
loaded into the turbine object is used. This will typically be the nacelle anemometer data, which are
influenced by the nacelle and rotor blades quite a bit. But it is also possible to choose calculated wind
speed (if a time-varying PARK calculation model is loaded), or data from any Meteo object. So, there
is freedom in selection and the user must assess what can be used for which specific analysis.
Turbine: power output from any of the turbines in the Performance check session can be selected.
[ ] Measured – show the power vs selected wind speed signal – filtered data will be shown with light
colour.
[ ] Measured binned – as above but binned by choice of bin size.
[ ] Contract PC – show the power curve selected at the WTG object “as is”, meaning the values in the
table without any corrections. If the site differs from standard air density corrections, the site-specific
power curve at site air density should be selected to make it comparable to measured.
[ ] Calculation output – show the power curve from the calculated time series, which includes power
corrections as air density, turbulence, shear, and veer corrections, if chosen in calculation.
A special feature is the AEP window, where the power curves seen is converted to an annual energy
production at a given mean wind speed (Rayleigh distributed). This tells very exact how much the
measured power curve deviates from e.g. the contract power curve or another time period’s power
curve. The ratio will also be shown. The calculation is based on a binned power curve with 0.5 m/s bin
size, and the use of windPRO’s “normal” power curve extrapolation features. This means that the
power curve is extrapolated to the Cut-out wind speed specified within the power curve in the WTG
object. If there are too few data points for creating this part, then an extrapolation to cut-out wind
speed and the corresponding power value is performed.
Post Construction - Chapter 3 Model calibration & Performance Check 85
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
View alt. setup - this button opens the window below (Figure 87):
Here, comparisons to other turbines can be made, with an alternative wind speed signal or an
alternative period can be chosen.
Figure 87 Two periods binned power curves plot.
An example where two periods are compared is shown in Figure 87, along with the calculated output.
For the two calculation periods, exact same power curve performance is seen (as expected).
For power curve evaluations, the filtering shall be handled carefully. If e.g. all power <=0 is filtered out,
the curve at the low wind speeds are “lifted”. Best option is therefore to add USER error codes and
filter those. Filters and selections as described previously are also available in the power curve
analysis.
For power curve evaluations based on measurement mast in front of a turbine, use the filter by
direction to filter out data in disturbed wind directions. Use the IEC 61400-12-1 tool in windPRO to find
the disturbed directions, which also can find the directions disturbed by terrain complexity.
11.1.47 Analyzer, detailed data
The analyser is the place where the most comprehensive comparisons between measured and
calculated data can be performed, and wherever needed, inputs to the calculation revisions can be
identified.
Notice: For Post-Construction work path, the analyser tab is removed as this simple calculation is
based on turbine SCADA data and not a wind model, which is analysed here to identify potential
changes in performance, investigate wake model settings, etc.
Post Construction - Chapter 3 Model calibration & Performance Check 86
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 88 XY plot Calculated power vs Measured power.
Here, hourly measured versus calculated data are seen. When using mesoscale data, quite a huge
scatter will often be seen because these data can often be shifted in time. But by aggregating the data
to a different resolution, quite good correlations can be seen.
Another important thing is that applying filters will take out the availability events of the analysis and
the comparison is then made only on the good data, where turbines are in normal operation. With
filters applied, reasonably precise production losses due to availability problems can be identified.
Filters can be applied to measured and calculated data.
Most common is to filter data with error codes, assuming these are established. Even if no error codes
are available, it is possible to establish USER error codes and thereby utilize error code filtering. See
section 11.1.15
A very simple filter is measured must be a certain percent of calculated. Here it is important to be
aware of the time resolution.
Having monthly data, 75% seems like a reasonable value to filter out months with larger availability
problems.
Having daily data, 25% seems like a reasonable to filter out days with larger availability problems.
Having hourly data, 1% seems like a reasonable value; thereby just the “non-performance” hours
typically are taken out.
But having a large spread in the data, it is important not just to filter out the low performing and keep
the high performing, which would introduce a skewness in the filtering.
While using performance Check on a larger wind farm, it can be an issue that several turbines are not
running for some periods. This can lead to wrong conclusions regarding wake losses. Therefore, it is
possible to set that a certain minimum number of turbines to be in operation to include the time stamp.
For the measured or calculated, the lists are automatically build based on signals available.
It is possible to apply filter inside or outside an interval.
Taking negative plus 0 values out, set start to -9999 and end to 0, apply to “inside interval”
Post Construction - Chapter 3 Model calibration & Performance Check 87
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Taking just negative, set start to 0 and end to 9999 and apply to “outside interval”.
With the filter settings, also special analyses can be made. Filter e.g. on temperature or rpm.
Figure 89 Setup of different graph views for x/y plots.
Several different graph views are available. These have the purpose to analyse how the model
calculation works compared to the measurements.
The XY-graph (scatter) has been presented. In the following the other views are presented.
Figure 90 Binned calculated power vs measured for model validation.
XY Graph (binned by x) Figure 90. The binned version of the measured vs. calculated show good
correlation. When using mesoscale model data, this is often not the case. It will typically be seen that
at lower production, the calculation underestimates and at higher production it overestimates. This
typically will require an offset at the wind speed post-calibration level to solve the issue.
We also see the wind rose here, showing a good match between measured and calculated by
direction.
Post Construction - Chapter 3 Model calibration & Performance Check 88
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 91 WF graph view.
The WF graph (Wind Farm graph) shows results for all turbines in the same graph. This is where it
can be observed if some turbines are over- or under-predicted. If so, the hunt for identifying model
problems starts. There are many possible issues to consider, like wake model parameters, terrain, etc.
Figure 92 Difference between measured and calculated, by sector using calculated windspeed.
The WF difference graph showing the difference OR ratios between measured and calculated for all
individual turbines aggregated by sector, season etc. is a very strong tool for evaluation of model
problems. If all turbines perform poor from one direction sector, the problem is how the model handles
this direction sector. It could be a bias in mesoscale wind, or a mast speed up (due to sensor mounting
details), a roughness issue, forest issue, or hill speedup problem. Below are a few explanations:
Showing the difference, the over- or under-performance in total are shown. The deviation in one
sector is divided with the total calculated production.
Showing the ratios, each e.g. direction sector is evaluated not depending on how large part of the
production that comes from this sector.
Post Construction - Chapter 3 Model calibration & Performance Check 89
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
It is thereby possible with the use of the difference graph to focus on what really matters, while the
ratio graph makes an equal evaluation of each sector, season, etc.
The WF difference graph allows to evaluate by sector, by season, diurnal and by wind speed interval.
The next WF loss can be seen as a “historical” feature, while in the versions of windPRO (from 3.2
onwards) there are more comprehensive tools for loss evaluation, see the following paragraph.
The WF loss graph show the calculated “loss” by turbine. With loss what is meant is calculated –
measured for a specific selection of the data. The “loss” calculation selection can be based on:
Figure 93 Filter settings for loss selection.
The filtered will typically represent the values where turbines have been out of operation and thereby
account for the availability loss. But also, poor performance data can be included. It is the filter settings
that decide what is included, and it shall also be mentioned (again) that the concept is: Calibrate
model to reproduce “normal operation” as best as possible. It is a reasonable assumption that the
difference calculated – measured for the “non-normal” operation will reflect losses, i.e. potential
production, that has not been captured.
Figure 94 WF Geo view.
WF geo gives the result geographically. The NE turbine is calculated to perform 4% better than the
two other relativ.
Post Construction - Chapter 3 Model calibration & Performance Check 90
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
NB its an enginnering analysis to investigate the directional and wake models output VS the actual
performance of individual turbines in a park, and evaluate if the wake model needs adjustment, or its
possibly related to forrest model, or a seasonal bias in the wind data and re-adjust the Park calculatoin
and re-validate the model output vs the real yield of the existing turbines.
With iterations it possible to calibrate the windmodel to quite accurately be able to predict the real yield
of existing wind turbines, thereby have confidence that the moddel can be used to predict expansion,
or nearby new project or repowering of the existing site with low uncertainty.
Post Construction - Chapter 3 Model calibration & Performance Check 91
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 1: Start Performance Check Module ............................................................................................ 7
Figure 2: Choice of workflow in Performance Check .............................................................................. 8
Figure 3: Define type of import data. ....................................................................................................... 9
Figure 4: Importer for 10-minute data .................................................................................................... 10
Figure 5: WTG-ID Guide ........................................................................................................................ 11
Figure 6: WTG-ID selection GUI Example. ........................................................................................... 12
Figure 7: Successful selection of WTG-ID............................................................................................. 12
Figure 8: Chose existing WTGs for pairing............................................................................................ 13
Figure 9: Load measured data .............................................................................................................. 13
Figure 10: Additional functions on Pair-and-Load tab ........................................................................... 14
Figure 11: Setup settings ....................................................................................................................... 15
Figure 12: Handling of multiple signals.................................................................................................. 15
Figure 13: Example resulting time series. ............................................................................................. 16
Figure 14 window for Trim and Clean feature, all turbines in one work step. ....................................... 17
Figure 15 Example of sloped trend detection ........................................................................................ 18
Figure 16: Different choices for energy indices ..................................................................................... 19
Figure 17: Identification of relevant region BDB index .......................................................................... 20
Figure 18: From wind speed to energy index ........................................................................................ 21
Figure 19: Example Wind Index Database ............................................................................................ 22
Figure 20 Create turbine Indices ........................................................................................................... 23
Figure 21: Use of time shift function ...................................................................................................... 24
Figure 22: Setup error codes: Concept choice ...................................................................................... 25
Figure 23: Loading error codes in Data – Import Setup ........................................................................ 26
Figure 24: Importing error log file(s) ...................................................................................................... 27
Figure 25: Set-up for start/stop text. ...................................................................................................... 28
Figure 26: WTG-ID in error log .............................................................................................................. 28
Figure 27: Five options to load error code translator ............................................................................. 29
Figure 28: Define types in error code file............................................................................................... 29
Figure 29: Secondary error codes ......................................................................................................... 30
Figure 30: Multi-editing of error code categories ................................................................................... 31
Figure 31: Separate error codes with multiple meanings ...................................................................... 32
Figure 32: Example of categorized error codes .................................................................................... 33
Figure 33: Import rules........................................................................................................................... 33
Figure 34: 10-minute time series from error log .................................................................................... 34
Figure 35: Error code statistics and plausibility check ........................................................................... 35
Figure 36: Error code 309 – remote shutdown: Likely wrong import setting for time stamp. ................ 36
Figure 37: No error codes available ...................................................................................................... 36
Figure 38: Form for defining user error codes. ...................................................................................... 37
Figure 39: Preview: Stop without error code. ........................................................................................ 38
Figure 40: Preview: Filtering for curtailments and non-optimal performance ........................................ 39
Figure 41: Loss evaluation of user-defined error codes: example user code 10000 ............................ 40
Figure 42: Sensor calibration ................................................................................................................. 41
Figure 43: Heat map of reference versus individual wind speed .......................................................... 42
Figure 44: Example of inconsistent nacelle wind speed ........................................................................ 42
Figure 45: Sensor calibration – Wind speed substitution : TR10 methodology..................................... 44
Figure 46: Create time-varying PARK calculation. ................................................................................ 45
Figure 47: Wind speed correction .......................................................................................................... 46
Figure 48: Evaluation of impact of operational condition on the nacelle wind speed ............................ 47
Figure 49: Historic power curve ............................................................................................................. 49
Figure 50: Alternative set-up for historic power curve display ............................................................... 49
Figure 51: Main statistics ....................................................................................................................... 50
Figure 52: Monthly mean ....................................................................................................................... 51
Figure 53: Daily recovery rate ............................................................................................................... 51
Figure 54: Display time series ............................................................................................................... 53
Figure 55: Application of filters .............................................................................................................. 53
Figure 56: Error loss statistics – detailed error codes ........................................................................... 54
Figure 57: Error loss matrix ................................................................................................................... 55
Post Construction - Chapter 3 Model calibration & Performance Check 92
EMD International A/S • www.emd.dk • windPRO 4.1 • September 2024
Figure 58: Selecting WTGs and categories. .......................................................................................... 55
Figure 59: Power Curve tab ................................................................................................................... 56
Figure 60: Example XY graph - nacelle directions ................................................................................ 58
Figure 61: Example XY graph ............................................................................................................... 59
Figure 62: Monthly gross tab ................................................................................................................. 60
Figure 63: Wind Energy Index tab ......................................................................................................... 62
Figure 64: Post construction report ....................................................................................................... 64
Figure 65: Main results .......................................................................................................................... 65
Figure 66: Example historic power curve based on the nacelle wind speed......................................... 65
Figure 67: WTG selection historic power curve ..................................................................................... 66
Figure 68: Actual losses by category and WTG .................................................................................... 66
Figure 69: Actual losses by month and by error code ........................................................................... 67
Figure 70: Potential Production ............................................................................................................. 67
Figure 71: Normalized production - wind energy index correction ........................................................ 68
Figure 72: Quantification of future losses .............................................................................................. 68
Figure 73: NET Production .................................................................................................................... 69
Figure 74: Export options ...................................................................................................................... 69
Figure 75 Monthly data work path ......................................................................................................... 70
Figure 76 Normalized monthly production data. .................................................................................... 71
Figure 77: WCP Loss calculation .......................................................................................................... 72
Figure 78 WCP report page 1 ................................................................................................................ 74
Figure 79 WCP report losses & prognosis ............................................................................................ 75
Figure 80 Park module start time varying calculation. ........................................................................... 78
Figure 81 Time varying park calculation setup ...................................................................................... 79
Figure 82 Scaling of time varying park setup ........................................................................................ 80
Figure 83 Create new Performance Check session. ............................................................................. 81
Figure 84 Model validation & calibration ............................................................................................... 81
Figure 85 Time Varying Park vs Measurements ................................................................................... 83
Figure 86 Power curve........................................................................................................................... 84
Figure 87 Two periods binned power curves plot. ................................................................................. 85
Figure 88 XY plot Calculated power vs Measured power. .................................................................... 86
Figure 89 Setup of different graph views for x/y plots. .......................................................................... 87
Figure 90 Binned calculated power vs measured for model validation. ................................................ 87
Figure 91 WF graph view. ..................................................................................................................... 88
Figure 92 Difference between measured and calculated, by sector using calculated windspeed. ....... 88
Figure 93 Filter settings for loss selection. ............................................................................................ 89
Figure 94 WF Geo view. ........................................................................................................................ 89
