Research on Optimization Method of Power Grid Recovery Path Based on Reinforcement Learning

After a power outage occurs, the power grid recovery can be accelerated according to the reasonable recovery path. This paper proposes a recovery path optimization method based on reinforcement learning. This method can solve complex problems in a model less way and improve the efficiency of the method. The goal is to restore maximum power to the grid. The constraints include over voltage, power flow, frequency, and self-excitation. Through continuous interactive learning between the agent and the power grid during the execution of the recovery path, the Q-value function of the power grid state and the recovery path was obtained. Based on IEEE system data simulation, the effectiveness and rationality of the proposed method are verified.