Self-tuning Yaw Control Strategy of a Horizontal AxisWind Turbine Based on Machine Learning

Abstract

The gradual depletion of fossil fuels and the atmospheric pollution caused by their combustion have led to significant growth in renewable energy systems, particularly wind energy. Numerous studies show positive trends in wind energy, with Denmark producing 40% of its power from wind in 2015, while Spain reached 17%, up from 10.4% in 2007. Recent reports by WindEurope also highlight substantial increases in wind energy capacity in countries such as Germany, the United Kingdom, France, and Sweden in 2018. This growth in wind energy generation is closely tied to research focused on reducing the Levelized Cost of Energy (LCOE) of wind turbines, which encourages investment in the sector. One key area of research is the development of advanced control strategies to optimize turbine performance. This chapter introduces the design of a yaw control system for Horizontal Axis Wind Turbines (HAWTs) based on Machine Learning (ML). The goal of this ML-based control strategy is to enable fully autonomous operation of the yaw system, learning from its experience either through offline training (when the turbine is paused) or online training (during operation). The chapter proposes an offline training process but notes that continuous online learning could also be implemented. A combination of Reinforcement Learning (RL) and Artificial Neural Networks (ANN), known as Deep Reinforcement Learning, is used to optimize the system. The approach aims to increase power generation and reduce mechanical loads from yaw rotations. The chapter is structured as follows: Section 2 outlines the objectives and applications of the proposed yaw control strategy; Section 3 discusses the theoretical foundations of the AI techniques used; Section 4 presents the design procedure for the ML-based yaw control system; and Section 5 provides the conclusions.