1. Improving WAsP predictions in (too) complex terrain
Niels G. Mortensen and Ioannis Antoniou
Risø National Laboratory
Anthony J. Bowen
University of Canterbury
2006 EWEC
2 March 2006

2. Outline Case study in northern Portugal RIX and RIX concepts
RIX configuration
Correction procedure
Improving WAsP predictions in (too) complex terrain?
Wind farm verification
Conclusions

3. Case study in northern Portugal

4. Cross-correlation of wind speeds
(From Bowen and Mortensen, 1996 EWEC conference)

5. Effect of a steep hill – flow separation
The flow behaves – to some extent – as if moving over a virtual hill with less steep slopes than the actual hill =>
actual speed-up is smaller than calculated by WAsP
N. Wood (1995). “The onset of flow separation in neutral, turbulent flow over hills”, Boundary-Layer Meteorology 76,

6. Complex terrain analysis
Ruggedness index, RIX
fraction of terrain surface which is steeper than a critical slope c
Calculation radius ~ 3.5 km
Critical slope c ~ 0.3
Onset of flow separation
Performance envelope for WAsP is when RIX = 0
Performance indicator, RIX
ΔRIX = RIXWTG – RIXMET
ΔRIX < 0  under-prediction
ΔRIX > 0  over-prediction
Terrain steeper than c is indicated by the thick red (radial) lines

7. Prediction error vs. RIX difference
“This performance indicator provides encouraging results…” (Bowen and Mortensen, 1996 EWEC conference)

8. The Ruggedness Index revisited
Reanalyses of Portuguese data sets
Larger and more detailed maps (SRTM 3)
Improved RIX calculation
Calculation implemented in WAsP and ME
More calculation radii: 72 rather than 12
RIX configuration corresponds to BZ-model grid
Improved predicted wind climate and power production
Emergent wind speed distribution
Both the prediction errors and RIX change

9. Maps for RIX calculation and modelling
Hand-digitised map
8 by 8 km2
50- and 10-m cont.
SRTM-derived map 
20 km diameter
50-, 10- and 5-m height contours + spot heights

10. Wind speed prediction error vs. RIX

11. ln(Up/Um) vs. RIX Up = Um exp( RIX) where  = 1.5
R = 3500 m and c = 0.3

12. Influence of radius and critical slope
R [m]
Critical slope c
0.30
0.35
0.40
0.45
3000
0.960
0.967
0.978
0.973
3500
0.972
0.974
0.984
0.986
4000
0.971
0.982
0.979
5000
0.969
0.977
Wind speed prediction error is (almost) fixed…
Number of sectors
Modelling parameters
RIX configuration can be varied easily
Original configuration somewhat arbitrary…
Different calculation radii: 3, 3.5, 4, and 5 km
Different critical slopes: 0.30, 0.35, 0.40, 0.45
Matrix of R2 (coefficient of determination) for different configurations
Weighting RIX with wind rose frequencies
R2 for different values of the calculation radius and critical slope.

13. Influence of calculation height
Vertical wind profile in complex terrain with RIX = 16%
40-m anemometer used as predictor
Vertical profile is predicted well because of similarity in ruggedness index: ΔRIX = 0

14. Improving WAsP predictions in complex terrain
Analysis procedure:
Observed Wind Climate
+ sheltering obstacles
+ roughness map
+ elevation map
Regional Wind Climate
Application procedure:
Predicted Wind Climates
+ power and thrust curves
Predicted wind farm AEP
Post-processing:
Insert WTG at met. stations
Make hhub
Plot ln(Pp/Pm) vs. RIX
Linear fit Pp = Pm exp( RIX)
Slope of trend line 
Correct production estimates:
Apply correction factor
Pm = Pp/exp( RIX)
Corrected gross AEP
Apply wake model results
Corrected net AEP

15. ln(AEPp/AEPm) vs. RIX @ 50 m a.g.l.

16. Step 1-2: AEP [GWh] = F(WAsP)

17. Step 3-4: AEP [GWh] = F(WAsP, RIX)

18. Case study summary
WAsP predictions in (too) complex terrain were improved
69% on average for five sites with 10% < RIX < 33%
88% on average for sites with ∆RIX > 10%
In addition, we have found
SRTM 3 data can be applied for wind resource assessment
optimal configuration values for ruggedness index calculation
an empirical relation between WAsP prediction error and ∆RIX
Can this be verified elsewhere?

19. Wind farm in complex terrain
Elevation map w/ 20-m contours
23-MW wind farm w/ 38 turbines
Two reference met. stations ()
RIX coloured map, range 0-18%
Turbine site RIX range: 4-14%
Met. station RIX range: 4-5%

20. Prediction of power production
Measured power productions, wind speeds and directions over one year available for analyses
Measured wind farm power production overestimated by 13% using standard WAsP procedure
Correction procedure applied:
Correction based on Portuguese data set (similarity)
Percentage applied to each turbine site
Corrected wind farm power curve applied
After correction, the power production is overestimated by 3%
Prediction of actual AEP improved by 70%
Site is also partly forested…

21. Conclusions Ruggedness index RIX and performance indicator RIX
Concepts supported by new data and procedures
Optimum radius and slope for RIX determined
(RIX, U) relation not very sensitive to calculation radius R, critical slope c, or prediction height h
Relation between WAsP prediction errors and RIX
Linear relation between ln(Up/Um) or ln(Pp/Pm) and RIX
RIX weighted with the wind rose does not improve the relation between ln(Up/Um) and RIX
Correction procedure outside WAsP operational envelope
Percentage can be applied to each turbine site
Note, that all this is purely empirical…
Similarity of sites: ridges, escarpments and mountain tops
constant  should be determined for site and height
Extension of WAsP procedures outside operational envelope
Requires two or more (non-similar) met. stations
Linear relation between ln(Pp/Pm) and RIX
Case study predictions improve significantly
Linear fit after extended procedure
AEPP = 1.01 AEPM (R2 = 0.92)
Procedure can be applied with (2…n) met. stations
Procedure is purely empirical and should therefore be tested with other data sets…