Load Shedding Control Strategy in Power Grid Emergency State Based on Deep Reinforcement Learning

In viewing the power grid for large-scale new energy integration and power electrification of power grid equipment, the impact of power system faults is increased, and the ability of anti-disturbance is decreased, which makes the power system fault clearance more difficult. In this paper, a load shedding control strategy based on artificial intelligence is proposed, this action strategy of load shedding, which is selected by deep reinforcement learning, can support autonomous voltage control. First, the power system operation data is used as the basic data to construct the network training dataset, and then a novel reward function for voltage is established. This value function, which conforms to the power grid operation characteristics, will act as the reward value for deep reinforcement learning, and the Deep Deterministic Policy Gradient algorithm (DDPG) algorithm, with the continuous action strategy, will be adopted. Finally, the deep reinforcement learning network is continuously trained, and the load shedding strategy concerning the grid voltage control problem will be obtained in the power system emergency control situation, and this strategy action is input into the Pypower module for simulation verification, thereby realizing the joint drive of data and model. According to the numerical simulation analysis, it shows that this method can effectively determine the accurate action selection of load shedding, and improve the stable operational ability of the power system.