Resilient Distribution Networks by Microgrid Formation Using Deep Reinforcement Learning

Resilience becomes vital for power grids facing the increasingly frequent extreme weather events. Microgrid formation is a promising way to achieve resilient distribution networks (RDN) when utility unavailable. This paper proposes RDN-oriented microgrid (RoMF) method based on deep reinforcement learning (DRL) technique, which integrates OpenDSS as an interaction object and searches optimal control policies in model-free fashion. Specifically, we formulate problem Markov decision process, taking into account complex factors such unbalanced three-phase flow operation constraints. Next, simulator-based RoMF environment constructed integrated OpenAI Gym, providing standard agent-environment interface applying DRL algorithms. Then, Q-network used search strategies, offline-training online-application framework of DRL-based given. Finally, extensive numerical results validate effectiveness our proposed method.