Wind Turbine Load Mitigation Using MPC with Gaussian Wind Speed Prediction

Abstract

An important control challenge for large wind turbines is to maximize the power capture whilst at the same time mitigating against potential fatigue damage. It is now quite well understood that this challenge may be beyond the capability of classical controllers, even when individual pitch actuation is considered. Recent research shows that preview controllers requiring future wind speed knowledge will offer an enhanced combination of good load mitigation and power capture performance. In this context, some preview control schemes use future wind speed prediction data generated by Light Detection and Ranging (LiDAR) systems. However, LiDAR devices tend to be expensive and may not always be available for individual wind turbines. This paper shows how accurate short-time wind speed data can be predicted from past measurements using a Gaussian Process (GP) model based on Matern class kernel. The short-time wind speed prediction is combined with model predictive control (MPC) and detailed simulations are carried out using the FAST NREL 5MW wind turbine model.