Getting benefits from CFD. in the wind energy industry

Transcription

1 Getting benefits from CFD in the wind energy industry Didier DELAUNAY - METEODYN (In Turkey:

2 Meteodyn software suite Meteodyn WT Meteodyn Forecast TopoWind UrbaWind SimulaWind

3 Meteodyn in the world

4 Getting benefits from CFD in the wind energy industry Wind Resource assessment: Meteodyn WT and validations Analysis of wind parks performances and Optimization Corrections of LIDAR measurements in complex terrain Downscaling for Production Forecast

5 Reynolds Averaged Navier Stokes equations - Steady flow Closure of the system (turbulence modeling): The Equations of Meteodyn WT 0 i i x u 0 ' ' i j i i j j i j i j i j F u u x u x u x x P x u u i j j i t j i x u x u u u ' ' T L T k 2 1/ where

6 Ground Boundary Conditions No canopy F S C U S t U Surface sink term introduced in the momentum equations at the ground cells C S = 0.5*(k / ln(z 1 /z 0 ) ) 2 and C S (ζ)/c s = 1 Ψ M (ζ)/ln( z z 0 ) 2 Inside the canopy F V C U U d Volume sink term introduced in the momentum equations at the canopy cells C d depends on the canopy density

7 Height, m Height, m Calibration of Canopy model Eurocode profiles for wind design z 0 = 1 m TopoWind 160 EC Normalized wind velocity TopoWind 160 EC Turbulence Intensity

8 Height of the Internal Boundary Layer Validation: Roughness change

9 Calibration of drag forces (stability effects) Turbulence intensity Log linear profiles on homogeneous terrain

10 Validation on a single hill: The Askervein test case vitesse normalisée LINE A-A - Direction 210 deg. 1,8 1,6 1,4 1,2 1 0,8 dx = 15 m ; dz = 4 m, dx = 25 m ; dz = 4 m dx = 50 m ; dz = 4 m dx = 15 m ; dz = 2 m Mesures dx = 15 m ; dz = 6 m 0,6 0, distance par rapport au point CP Askervein Hill Normalised velocity 2 1,8 1,6 1,4 1,2 1 0,8 0,6 0,4 0,2 0 Meteodyn_WT measurements Ligne A Distance from HT

11 142 WT Puissance Validation in a very complex terrain: Andhra Lake Wind Park

12 Validation in a very complex terrain Andhra Lake Wind Park Mast 3018 Mast 3027 Mast 3037 Mast 3045 Mast met masts measurement periods > 1 year

13 Validation in a very complex terrain: Andhra Lake Wind Park Domain size: 30 km x 30 km Horizontal mesh resolution: 20 m Vertical mesh resolution: 4 m Orography map 60 Million cells 36 directions (every 10 deg) Computational time to convergence: 1 direction (1 processor) = 8 h 36 directions (8 processors) = 72 h Roughness map

14 Validation in a very complex terrain: Andhra Lake Wind Park Dir = 270 deg Dir = 90 deg Speed-up coefficients

15 Validation in a very complex terrain: Andhra Lake Wind Park Dir = 270 deg Dir = 90 deg Turbulence Intensity

16

17 Wind Speed Uncertainty (%) Net Energy Output Uncertainty (%) Anemometer accuracy Variability of reference period Long term extrapolation Extrapolation (wind flow modelling) Overall Wind Speed Estimation 18.1

18 Validation in a complex terrain: CREYAP BLIND TEST Comparative Resource & Energy Yield Assessment Procedures Wind Park comissionned in 2007 Installed capacity1.3 x 22 = 28.6 MW

19 Copyright Walter Baxter, Jim Barton, Poljes and Panoramio. Validation in a complex terrain: CREYAP BLIND TEST

20 Validation in a complex terrain: CREYAP BLIND TEST Meteodyn WT Computation Horizontal grid resolution: 25m, Vertical resolution: 4m 24 direction sectors (15 deg wide) Number of cells: 5 Millions Convergence ration:99%-100% 4 processors version: 24 directions computed in 8 hours!

21 Validation in a complex terrain: CREYAP BLIND TEST

22 Validation in a complex terrain: CREYAP BLIND TEST

23 Mean wind speed (m/s) Validation in a complex terrain: CREYAP BLIND TEST T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 Mean wind speed Turbulence intensity

24 Production (GWh) Validation in a complex terrain: CREYAP BLIND TEST Net energy P50 (Meteodyn) Observation T1 T2 T3 T4 T5 T6 T7 T8 T9 T10T11T12T13T14T15T16T17T18T19T20T21T22 Total Production: Observed 76.4 GWh - Computed (Meteodyn ) 76.2 GWh Individual turbines: Maximum error 0.2 GWh (7%) ; rms error : 3%

25 Validation: Combined meso-scale and CFD

26 Validation: Combined meso-scale and CFD RSM error = 5.2%

27 ANALYSIS OF WIND PARK PERFORMANCES DIAGNOSIS AND OPTIMIZATION

28 Diagnosis and Optimization On-site Cp and Power curves: Wind speed at hub height deduced from met mast and Meteodyn WT coefficients

29 Wind park analysis example 1

30 Wind park analysis Example 2

31 LIDAR MEASUREMENTS: CFD CORRECTION

32 Lidar measurements Corrections r V i h V i i i i i r V i h V i LIDAR conical scan angle Deviation from the mean wind direction Vertical angle of the wind vector Radial wind along the LIDAR beam Horizontal component along the mean wind direction LIDAR algorithm: Hypothesis of flow homogeneity over the disk: i V i h and V i i r / cos i cos are assumed constant CFD correction: computation of i and i at each scan point h correction of the LIDAR estimation V i i

33 U (Lidar) / Uref Lidar measurements Corrections New Zealand site (ZephIR Lidar) Correction factors computed by Meteodyn WT Horizontal resolution : 5 m Vertical Resolution : 2 m Site n 2 - Mast n 2- H=80 m y = 1.00x R² = 0.96 y = 0.95x R² = U (Mast) / Uref

34 Correction factor Lidar measurements Corrections Spanish site (Windcube LIDAR) Mesh 5 x 5 Mesh 2 x Wind direction ( )

35 Lidar measurements Corrections H Bias rms error Windcube (Leosphere) Spanish site 70 m No correction % 0.89 m/s WT correction % 0.78 m/s ZephIR (Natural Power) NZ site 80 m No correction % 0.75 m/s WT correction % 0.65 m/s

36 WindCube Correction German Site Lidar measurements Corrections

37 METEODYN FORECAST

38 Meteodyn Forecast 4.6 GW by Meteodyn Forecast 38

39 Meteodyn Forecast Average NMAE for 12 hours forecast horizon vs RIX Source: Best Practice in Short-Term Forecasting. A Users Guide Gregor Giebel(Risø National Laboratory, DTU), George Kariniotakis( Ecole des Mines de Paris)

40 Meteodyn Forecast NWP Models global scale ( km ) NCEP/NCAR/ERA/ CFS/MERRA/GFS/ JRA/ECMWF meso scale (2-30 km) Meso- Models WRF MM5 NWP inputs local scale (25m-50m) Local scale MASS.. WASP.. METEODYN WT Meso-scale inputs

41 (17) Meteodyn Forecast: Downscaling The mesoscale points are transferred to each wind turbine by using the local «speed coefficients» obtained by the CFD At each time step, The coefficients are depending on: The wind direction The thermal stability Calibration takes into account seasonal variations (snow, foliage density, ) C site site h meso Vh V i 100 C100 Several mesoscale points can be used.

42 Meteodyn Forecast: Wake effects Thanks to downscaling, individual wake effects can be considered Windspeed deficit and turbulence intensity increase (Modified PARK Jensen or Eddy Viscosity Model) The wake effect is computed at each time step Better consideration of out of order wind turbines in the production forecast

43 Meteodyn Forecast: Statistical Correction NWP Real Production History of Forecats Self-learning Error Modelling Physical Forecast ANN approach Wind Park Corrected Forecast

44 Meteodyn Forecast 4 runs a day,horizon: 22h-46h Park 1 Park 2 Park 3 Park 4 Park 5 Park 6 Nominal Power [kw] Normalized RMSE [%] Annual Capacity Factor [%]

45 CFD is now largely used and validated for Wind Resource Assessment in complex terrain But CFD can gives much more : LIDAR measurements corrections Site suitability analysis Turbulence intensity at hub height (fatigue) Vertical shear of the wind speed Vertical wind speed Extreme wind speeds at hub height (wind design) All IEC requirements Wind Parks in Operation Downscaling for forecast Analysis and Optimization CONCLUSION

46 Thank you for your attention! For more information on Meteodyn WT: Ucyel Enerji