A Deep Reinforcement Learning Voltage Control Method for Distribution Network

Aiming at the distributed voltage control optimally in real-time for distribution network with high penetration of renewable energy sources, a deep reinforcement learning voltage control (DRLVC) method is proposed in this paper. Based on the network partition, the voltage control method with coordinated optimizing the reactive power output of PV inverter and active power output of energy storage system (ESS) is formulated as the Markov decision process for each sub-network. Then, with information communication, the distributed voltage control model is formulated as the multi-agent deep reinforcement learning model. And the improved multi-agent depth determination strategy gradient (MADDPG) algorithm is employed to solve the constructed model efficiently. Finally, numerical simulations are conducted on the 322-bus test network to show the effectiveness of the proposed DRLVC method.