Dynamic Scheduling of Mobile Energy Storage for Post-Disaster Power Recovery from Storm Tides: A Multi-Agent Reinforcement Learning Framework

Storm tide disasters may lead to extensive power outage in distribution networks. The usage of energy storage resources is necessary to ensure the power critical loads. Previous research has been designed to address typhoon disasters without considering the impact of storm tides, which reduces its effectiveness. This paper proposes a multi-agent deep reinforcement learning framework to address the issues, based on the integration of power and transportation networks, facing dynamic scheduling decisions of mobile energy storage for power supply recovery after storm tide. Firstly, to simulate the dynamic allocation of power demand and energy storage resources in various regions during storm tide disasters, a coupled transportation and power grid digital twin is constructed, ensuring that Mobile Energy Storage Systems (MESS) can efficiently transfer and distribute power resources between regions affected by storm tide. Then, the reinforcement learning method based on Multi-Agent Depth Q-Network (MADQN) is designed, and can optimize the real-time scheduling strategy of mobile energy storage resources, maximize power recovery efficiency, and minimize recovery time. Finally, a hazard assessment framework for storm tides in coastal cities is proposed. The experimental results show that this method has significantly improved recovery speed and power supply security after storm-tide disasters.