An analysis of wake loss and atmospheric stability on a wind farm situated on complex terrain

Abstract

In order to analyse the wake loss and atmospheric stability of a wind farm located on a complex terrain, Computational Fluid Dynamics (CFD) simulations were performed with the WindModeller software, which is a module for wind farm simulation developed by ANSYS. The wake was modelled using an actuator disc model approach which is based on the wind turbine thrust coefficient and wind speed. A WindModeller simulation was carried out for DBK wind farm located on Jeju Island, Korea. The nacelle wind speed data from 15 Hanjin 2MW turbines were collected through the Supervisory Control And Data Acquisition (SCADA) system. The wind data was measured from an 80 m tall met mast near the wind farm, which was used as a reference. The WindModeller module simulated the wind speed and the turbulence intensity on a complex terrain, which were compared with the wind data from the SCADA system. Then, the wake effect was analysed with the distance between the wind turbines and the surrounding natural obstacles such as hills and quarries. Atmospheric stability was analysed on different atmospheric conditions by the turbulence intensity and the shear exponent factor. As a result, the wake effect predicted by the WindModeller simulation was greater than the actual wake effect. The actual wind speed ratio decreased by 22% and 35% when the turbines were separated with the distances of around 3~6 times rotor diameters, respectively. Also, the results of the wind speed deficit from the neighbouring hills 1 and 2 were 10% and 8% from the WindModeller simulation. The actual wind speed deficits from the measurement data were 14% and 9%, respectively. In addition, quarry 1 resulted in a wind speed deficit of 3% from the SCADA data analysis while the actual wind speed deficit was 4% from the WindModeller simulation. Lastly, the atmospheric stability had a significant effect on the wind shear and the turbulence intensity in the wind farm. The trend from the actual wind data analysis was the same as that from the WindModeller analysis.