Power enhancement of a tuned inerter-equipped point-absorber wave energy converter using deep reinforcement learning

This study investigates deep reinforcement learning (DRL) control of a point-absorber wave energy converter (WEC) equipped with a tuned inerter (TI) mechanism. The inerter generates an inertial force proportional to the relative acceleration between its terminals, enabling a large effective mass effect, referred to as inertance, with a relatively small physical mass. In the proposed configuration, the inertance is tuned to the dominant wave frequency so that the internal motion of the mechanism is amplified, enhancing energy absorption without requiring a large floater mass. The control objective is to maximize the net electrical power harvested under irregular wave conditions while considering practical generator power limits. A time-domain WEC model is developed, and the generator damping is controlled by a DRL agent. Both semi-active control, where the damping coefficient is restricted to non-negative values, and active control with bidirectional power exchange are examined. Numerical simulations are performed for a conventional WEC and TI-equipped WECs. The results show that the TI mechanism significantly improves passive energy harvesting. Furthermore, under semi-active control, the TI-equipped WEC achieves substantially higher energy conversion ratios than the conventional WEC, with the maximum improvement exceeding 100% under certain wave conditions.