Reduced order model for optimization of power production from a wind farm

This paper focuses on the optimization of power production from a column of wind turbines aligned with the mean wind direction. Optimal settings of the wind turbines are estimated through dynamic programming. The optimization is conducted using a reduced order model that estimates the downstream evolution of wind turbine wakes with high level of accuracy and low computational costs. The model, which is the main contribution of this paper, consists in a data-driven Reynolds-averaged Navier-Stokes (RANS) algorithm calibrated through Large Eddy Simulation (LES) data. The RANS model is first assessed against LES data for a single turbine operating with different tip speed ratios. Then, the optimal settings of a column of four wind turbines are estimated from this model through dynamic programming. Finally, the cumulative power curve of the wind turbine column is obtained.