Prediction of wind farm energy yield using NWP considering within-cell wake losses

Numerical Weather Prediction (NWP) models such as Weather Research and Forecasting (WRF) are widely used for prediction of wind resource at potential wind farm sites and, increasingly, for energy yield prediction. Such models solve a reduced form of the Navier-Stokes equations with typical resolution of 20-1000 m in the vertical axis and 1-2 km in the horizontal axes. Sub-grid models have previously been developed to represent wind farms including by modification of momentum sink and turbulence kinetic energy source terms within cells occupied by turbines. Here, semi-empirical wake models are employed to assess the extent of losses between turbines within a small group such as within a single WRF. Variation of thrust and power with wind speed and direction were obtained using the modified PARK and Eddy Viscosity methods in OpenWind. The influence of wake-losses on yield was evaluated through WRF simulations of resource only and with standard and modified turbine parameterizations. Annual energy yield of the Horns Rev wind farm was considered for reference with predictions obtained for annual data and a short duration subset of wind data with comparable occurrence. Predictions with an Eddy Viscosity model and ERA-Interim data for 2007 were approximately 13% higher than from WRF simulations and than measured yield, approximately 18%. Consideration of within cell losses reduced WRF predictions by approximately 2.5%. The subset of wind data used for WRF simulations was in close agreement with annual occurrence data but further assessment is required of yield over longer samples and directional dependence of losses.