Energy optimal dispatching of ship's integrated power system based on deep reinforcement learning

Abstract

With the pressure mounting on environmental protection, all-electric ships(AES)are becoming increasingly popular for maritime transportation. The optimal dispatch of energy is of great significance to achieving AES' safe and economic operation. This paper proposes a deep reinforcement learning (DRL) based energy optimization scheduling method for the ship power system, and the generator and the energy storage system (ESS) are directly driven by the original measurement data of the ship's power system. This paper first describes optimal energy scheduling in the ship power system mathematically and then expresses the scheduling decision problem as a reinforcement-learning framework. Next, it introduces a deep Q network algorithm to optimize the end-to-end control strategy between the measurement data of the ship power system and the action instructions of the generator and ESS. The method proposed in this paper does not need to model the complex system and can realize the dynamic optimization scheduling decision of energy with the goal of economy. Finally, two case studies are analyzed based on the historical data of the ship power system, and the simulation verifies the effectiveness and superiority of the proposed energy optimization scheduling method based on DRL.