Bayesian Optimization of wind farm power generation via wake steering using ADM

Abstract

As new wind farms are built, the best locations are becoming fewer and fewer. Therefore, the debate is currently focused on possible measures to make better use of the resource. While the major reason for the drop in efficiency of a wind farm is wind speed variability, aerodynamic losses in large turbine arrays can be significant, potentially leading to a drop in annual energy production of up to 20%. There are several methods under study on how to improve the performance of wind farms, among which are the variation of yaw angle or axial-induction-based control, which is based on varying the pitch of the wind turbine blades or varying the torque. This study provides a novel approach to the optimization of wake steering in wind farms via Bayesian Optimization. This way, if the upwind turbines are misaligned, their wakes may not directly affect the downstream turbines. Hence, increasing their power output and the output of the wind farm at the cost of lower power output in the upwind turbines. Therefore, this study chooses to develop a model to provide yaw angle variated optimizations. The model to be used is based on the Actuator Disk Model and Large Eddy Simulations. To this end, a validation of the model used is first carried out by comparing the results obtained with those described in Jiménez et al. (2009). Once the code has been validated, a study of the possible optimisation of an array of turbines is carried out.