Deep Reinforcement Learning Control to Maximize Output Energy for a Wave Energy Converter

Abstract

This paper presents a deep reinforcement learning control method to maximize output energy for a point absorber type wave energy converter (WEC) with a linear generator. Conventional control methods require the dynamic model of the WEC. Modeling errors of the dynamic model, however, make energy absorption smaller and cause incorrect control. The proposed method, which is a model- free control method learns the optimal damping and stiffness coefficients based on experiences. In the proposed control method, damping and stiffness coefficients are able to vary in time-domain depending on the incident waves by deep reinforcement learning. The performance of the proposed control method is investigated through numerical simulation in both regular and irregular waves. Compared with the conventional control method, averaged output power increased, and the power fluctuation decreased without the dynamic model. It is understood that the proposed method is more effective than the conventional control method.