Real-time model for wave attenuation using active plate breakwater based on deep reinforcement learning

Abstract

This paper presents a numerical study of the interaction between irregular waves and actively controlled plate breakwater based on deep reinforcement learning (DRL), in which wave attenuation performance is mainly discussed. A numerical wave flume (NWF) is built and an in-house solver is developed to integrate CFD and DRL. Validations on wave generation and interaction between incident waves and the plate are carried out. An NWF-DRL coupled model is proposed to simulate the actively controlled plate moving to attenuate waves, where the motion of the plate is controlled by DRL. It is shown that under the regular wave and trichromatic wave group, both the trained agents can obviously reduce the wave height compared to that under fixed plate. Moreover, the further trained agents can reduce the transmitted wave heights to a significantly low level in the three more complex and unknown irregular wave cases. Meanwhile, the model does a great job of preventing the development of focused wave height in an irregular wave train. This research first examines the wave attenuation of a DRL-controlled breakwater in irregular wave environments, which may offer fresh perspectives on wave attenuation and even the design of intelligent marine equipment.