MARLYC: Multi-Agent Reinforcement Learning Yaw Control

Abstract

Inside wind farms, turbines are subject to physical interactions such as the wake effects. Such phenomena damage the performance of wind turbines, especially offshore. Yaw control consists in rotating a turbine’s nacelle on a horizontal plane and can be used to reduce the detrimental consequences of wake effects by steering them. In this work, a new method called multi-agent reinforcement learning yaw control (MARLYC) is proposed to control the yaw of each turbine in order to improve the total energy production of the farm. MARLYC consists in the centralized training and decentralized execution of multiple reinforcement learning agents, each agent controlling the setting of one turbine’s yaw. Agents are trained together so that collective control strategies can emerge. During execution, agents are completely independent, making their usage simpler. Numerical simulations are conducted on 15 different wind farms whose size ranges from 4 to 151 turbines with real time-varying wind data. For each wind farm, MARLYC increases the total energy production by controlling the yaws of the turbines judiciously with negligible increase of the computation time.