Non-Linear Control Strategy for a Two-Body Point Absorber Wave Energy Converter Using Q Actor-Critic Learning

In this paper, a reinforcement learning method is used to tune a non-linear reactive control model parameters of a two-body point absorber Ocean Wave Energy Converter (OWEC). In particular, an Actor-Critic algorithm, as a model-free method is adopted for the maximization of the energy extraction, adaptive to the sea state. Different values of Power Take-Off (PTO) control parameters are applied to the system to observe reward and penalty of the taken action. Reward is determined by the average power over a specific time horizon lasting several wave periods. A two-body point absorber, simulated in WEC-Sim, is developed as the agent in order to validate the control strategy for different wave conditions. Results for the analyzed sea states verifies that the proposed non-linear control law learns the optimal PTO control parameters in specified sea states.