Preliminary results of a physical model for extrapolating the wind field over complex terrain

In complex terrain, wind power production depends not only on wind velocity but on wind direction too. Estimating such volatile variables is a difficult but essential task for the installation of a Wind Turbine (WT) and the operation of Wind Farm (WF). Our goal is to create a short term (5-15min) forecasting physical model when terrain complexity excludes measurements using long-range scanning lidars. In this study we will try to extrapolate the wind field over a wind farm with the use of steady state CFD simulations. Steady state RANS CFD simulations are extremely faster than time dependent ones (LES), which makes them fast enough for operational use, as well as for estimating wind flow over a wanted location for many different scenarios.

 

The computer used for the simulations has 24-cores and 128gb ram installed. The software used for the simulations is COMSOL Multiphysics. In this study, each steady state simulation represents 10 min averages and the computational time needed is below 2 min. The simulations were run for an area 2.5x2x0.6 km³ large with a non-uniform mesh and the distance between computational nodes ranging from 22,5 to 75.5m in the area of interest.

 

Wind measurements from a vertical Wind Profile LIDAR installed at CRES WF located at Lavrio (Greece), were used for the validation of the model. The terrain location is complex with a RIX index of 10%. Lidar measurements at 6 different heights (54 m, 78 m, 100 m, 120 m, 140 m, 160 m) were used as inlet conditions after being vertically extrapolated and interpolated, acknowledging the factors that govern lidar accuracy in complex. The inlet plane was positioned 1.3 km away from the LIDAR location. Horizontal homogeneity was assumed at inlet plane. The simulations were run at an expanded area to capture the effects of the terrain on wind movement, because of lack of more measurement locations.

 

Results from 291 simulations that were corrected with linear regression, were compared with data from the three WTs and the LIDAR. Extreme errors were detected when sudden changes in wind speed or direction were observed. The maximum mean absolute error (MAE) of wind velocity with respect to LIDAR measurements at all available heights is 0.28 m/s and the minimum is 0.17 m/s.  In addition, absolute errors are smaller than 0.7 m/s and maximum absolute percentage error is 11% for 95% of the estimations. The accuracy of the model increases with height because of the terrain anomalies and turbulence effects.