DTU Wind Energy, Technical University of Denmark
Power curve measurement using ๐๐โ estimates from nacelle lidars
R. WagnerN. DimitrovN. TroldborgA.R. Meyer Forsting
Project: UniTTewww.unitte.dk
5-beam Avent demonstratorZephIR Dual-Mode
Wind Europe 2017, 30 Nov. 2017, Amsterdam
DTU Wind Energy, Technical University of Denmark
Power performance testing: where are we?
โข New standards: IEC 61400-12-1:ed2 (2017)โข Whatโs new?
โ mast and/or RSD e.g. ground-based lidarโ hub height spd + shear measurementor rotor equivalent wind speedโ (somewhat) more thorough power curve uncertainty assessment
โขBut STILLโ no nacelle lidar(coming in IEC 61400-50-3)
โ measurements between 2Drot and 4Drot from the turbine
2
DTU Wind Energy, Technical University of Denmark
1)โGeneric methodology for field calibration of nacelle-based wind lidarsโ (link)
2)โWind field reconstruction from nacelle-mounted lidarshort-range measurementโ(link to WES)
3)Uncertainty propagation in WFR models (using Monte Carlo methods)
4)Applied to power perf.
In my PhD โฆ the story
3
Link to thesis on DTUโs site
DTU Wind Energy, Technical University of Denmark
Searching for free stream wind speed
4
๐ฝ๐ฝโ ??๐ฝ๐ฝ๐๐.๐๐๐๐
Modern turbines: 2.5D ~ 200-400m
โข Decorrelation WSpeed / powerโข Hhub speed sufficient?
โข 2.5D not really free wind โฆ
DTU Wind Energy, Technical University of Denmark
Model-fitting wind field reconstruction for power performance testing
โข Several possibilities for lidar measurements:
1) 2.5D distancefitting wind speed + direction + shear to lidar-measured LOS velocities
2) Multiple distancesclose to rotorinduction integrated in wind field reconstruction
๐๐ ๐ฅ๐ฅ๐๐โ
= 1 โ ๐๐ 1 + ๐๐1+๐๐2
๐๐ = 12
1 โ 1 โ ๐ถ๐ถ๐ก๐ก5
Extrapolation of โtrueโ free stream wind speed
DTU Wind Energy, Technical University of Denmark
Lidar measurements @ multi-dist (near flow)Mast comparison, Nรธrrekรฆr Enge campaign, 7 months
5B-Demo: use the 5 LOS ZDM: use 6 pts@[0.5; 0.75 ; 1.0 ; 1.15] Drot @[0.3 ; 1.0 ; 1.25] Drot
HWS estimated @hub height and @2.5D distance
6
๐ต๐ต๐๐๐๐๐๐๐๐๐๐๐๐ = ๐๐๐๐๐๐๐๐
๐๐ = ๐๐.๐๐๐๐๐๐๐๐๐๐ + ๐๐.๐๐๐๐๐๐๐๐ ๐น๐น๐๐ = ๐๐.๐๐๐๐๐๐๐๐ ๐๐ = ๐๐.๐๐๐๐๐๐๐๐๐๐ โ ๐๐.๐๐๐๐๐๐๐๐ ๐น๐น๐๐ = ๐๐.๐๐๐๐๐๐๐๐
๐๐ = ๐๐.๐๐๐๐๐๐๐๐๐๐ ๐น๐น๐๐ = ๐๐.๐๐๐๐๐๐๐๐ ๐๐ = ๐๐.๐๐๐๐๐๐๐๐๐๐ ๐น๐น๐๐ = ๐๐.๐๐๐๐๐๐๐๐
DTU Wind Energy, Technical University of Denmark
Measured power curves โ 10-min data
7
ZDMusing fitted ๐๐โ
Mast, cup @Hhub
5B-Demousing fitted ๐๐โ
Using nacelle lidar measurements close to turbine rotor!
Accurate
& low scatter
power curves
can bemeasured ..
DTU Wind Energy, Technical University of Denmark
Measured power curves โ binned data
8
ZDMusing fitted ๐๐โ
5B-Demousing fitted ๐๐โ
Mast, cup @Hhub
DTU Wind Energy, Technical University of Denmark
Uncertainty quantification in WFR models
9
โข Models are always wrong framework for UQ
โข The model-fitting WFR approach is too sophisticated for using analytical uncertainty propagation (โGUMโ)
โข Instead we can use numerical techniques: e.g. Monte Carlo methods
DTU Wind Energy, Technical University of Denmark
Monte Carlo UQ results for combinedwind-induction WFR model
10
vs spd ๐๐โ
vs relative dir. ๐๐๐ซ๐ซ
vs shear exponent ๐๐๐๐๐๐๐๐
vs inductionfactor ๐๐๐ข๐ข๐ข๐ข๐ข๐ข
๐๐ ๐๐โ ๐๐ ๐๐๐ซ๐ซ ๐๐ ๐๐๐๐๐๐๐๐ ๐๐ ๐๐๐ข๐ข๐ข๐ข๐ข๐ข
DTU Wind Energy, Technical University of Denmark
Monte Carlo UQ results for combinedwind-induction WFR model
11
Conclusionthe model uncertainty on Vโestimated by the nacelle lidars is negligibly different from the wind speed uncertainty of the reference anemometer used during the LOS velocity calibration campaign
DTU Wind Energy, Technical University of Denmark
Power curve uncertainty assessment (1/4)
12
โขThe procedure is based on the new standards IEC 61400-12-1:ed2 (2017)with some deviations: no โmethodโ uncertainty considered (related to REWS, and shear, veer, TI normalisation, etc)
โขMethod to estimate the cat. B wind speed unc. for the lidars combines the model uncertainty (Monte Carlo) with fitting residuals (inadequacy)
โขThe โflow distortion uncertaintyโ2% for the cup (no site cal, default IEC for 2.5D dist)1% for the lidars: fair enough since measurements taken close to the rotor (about 1Drot)
DTU Wind Energy, Technical University of Denmark
Power curve uncertainty assessment (2/4)cat. B wind speed uncertainty
13
โขThe reduction of combined wind speed uncertainty is โartificialโ since due to the different flow distortion uncertainty value
need for finer quantification of this component in standards
โขFitting residuals slightly higher for ZDM than 5B-Demo explain the difference
DTU Wind Energy, Technical University of Denmark
Power curve uncertainty assessment (3/4)cat. A power uncertainty
14
Lower scatter for the measured power curves with the lidars lower cat. A uncertainty on power output
DTU Wind Energy, Technical University of Denmark
Power curve uncertainty assessment (4/4)combined power curve uncertainty (k=1)
15
DTU Wind Energy, Technical University of Denmark
Take-aways๐ฝ๐ฝโ is found! โข The solution: measurements close to rotor, within the induction zone,
at multiple distances, e.g. with nacelle lidarsno need for more powerful lasers!
โข Wind Field Reconstruction algo. provide estimates comparable to classic mast instrumentation (< 1% difference)
โข Power curves in flat terrain verified accurately, reduced scatter (as usual with nacelle lidars)extinction of the โmet. mast speciesโ is comingโฆ the dinosaurs of
wind measurementsnext generation of IEC standards work ongoing (-50-3)some studies on PCurve uncertainty assessment desirable
Coming soon in UniTTe:โ Complex terrain: demonstration of nacelle lidar short-range
measurement technique in two campaigns Hill Of Towie, Scotland (RES), ZDM & 4-beam Wind Iris Ogorje, Croatia (Akuo Energy), with a 4-beam Wind Iris
16
DTU Wind Energy, Technical University of Denmark
Thanks for your attention!
More info: website www.unitte.dk contact [email protected]
17
DTU Wind Energy, Technical University of Denmark
Acknowledgements
This work was performed within the UniTTeproject (www.unitte.dk) which is financed
by Innovation Fund Denmark.
18
DTU Wind Energy, Technical University of Denmark
Model-fitting Wind Field Reconstruction for power performance testing
โข Several possibilities for lidar measurements:
1) 2.5D distancefitting wind speed + direction + shear to lidar-measured LOS velocities
19
DTU Wind Energy, Technical University of Denmark
Wind speed evolution within the inductionProcess: 1) lidar-estimated Hhub speed @each distance adimensionned
by lidar-estimated ๐๐โ (for each 10min period)2) Averaging of non-dimensional spd by ๐๐โ bins of 0.5 ms-1
20
ZDM5B-Demo
DTU Wind Energy, Technical University of Denmark
A simple induction model
โข Derived from the Biot-Savart lawโsee The upstream flow of a wind turbine: blockage effect
โtwo parameters: induction factor ๐๐, free wind speed ๐๐โ๐๐๐๐โ
= 1 โ ๐๐ 1 + ๐๐1+๐๐2
, with ๐๐ = ๐ฅ๐ฅ๐๐๐ ๐ ๐๐๐๐๐๐
โข What does the induction looks like in NKE?
Black: theoretical, ๐๐ = 0.3Colored lines: different 10min periods
Fitting ๐๐ and ๐๐โ should be possible
21
DTU Wind Energy, Technical University of Denmark
AEP resultsโข IEC -12-1 methodologyโข extrapolated AEPs โข 0.5 m/s bin widthโข Relative difference in % of cup-based AEPโข Rayleigh avg speed = 8 m/s
AEP estimations as good with the โmulti-distancesโ method as with the 2.5D (<1.5% difference)
22
Lidar measurements
@2D (5B-Demo)@2.5 D ZDM) (case 1)
multiple distances
@ โ (case 2)
Avent 5-Beam demonstrator lidar
Wspeed difference: +0.59% Wspeed difference: +0.52%
-0.8% -0.9%
Zephir Dual Mode lidar
Wspeed difference: +0.32% Wspeed difference: -0.27%
-0.3% +0.5%
DTU Wind Energy, Technical University of Denmark
AEP results
23
DTU Wind Energy, Technical University of Denmark
Model-based wind field reconstruction
โข Doppler wind LiDaRs do notโฆโฆmeasure wind speed, wind direction, shear, โฆsee Hardesty, 1987 (wonderful description of lidar principles)
โข They:โonly measure LOS velocitiesโestimate/reconstruct wind field characteristics (WFC)
24
DTU Wind Energy, Technical University of Denmark
Does this make it any easier?
25
Perdigรฃo.credit: N. Vasiljevic
โข In complex terrain: โany โfree streamโ wind speed idea?โsite calibration? Maybe
โข Offshore:โmast expensive โfree wind may not be measurable due to wakes
DTU Wind Energy, Technical University of Denmark
Monte Carlo methods in brief(dummy example)
26
INPUTS DISTRIB. OF ERRORSOUTPUTS DISTRIB. OF ERRORS
DTU Wind Energy, Technical University of Denmark
Power curve in complex terrainwind-induction model @4 distscatter
27
โข Mast: top cup wind spd, corrected with SC (experimental)โข Lidar: free stream wind spd ๐๐โ, no correction
Clear scatter reduction in lidar PCurve
Lidar (4B WI) Mast cup
DTU Wind Energy, Technical University of Denmark
Power curve in complex terrainwind-induction model @4 distbinned
28
โข Mast: top cup wind spd, corrected with SC (experimental)โข Lidar: free stream wind spd ๐๐โ, no correction