Voltage Control Strategy of Distribution Networks with Distributed Photovoltaic Based on Multi-agent Deep Reinforcement Learning

Abstract

It is of great significance to control the voltage fluctuation and network loss increase caused by the random fluctuation of photovoltaic equipment output for the stable operation of distribution network. In order to solve the distribution network, voltage fluctuation problem. Firstly, a model-free multi-agent reinforcement learning framework based on depth deterministic strategy gradient algorithm is proposed. The method of centralized training and decentralized execution is adopted to solve the voltage fluctuation problem. Then, the reward function of the algorithm is adjusted to reduce reactive power loss under the premise of controlling voltage fluctuation. The adjustment can better meet the voltage control requirements of the distribution network. The deep reinforcement learning algorithm does not require accurate power flow modeling, nor does it depend on the prediction of the data before the day, so it is suitable for some observation distribution networks with weak communication capability. Finally, an example is given to verify that the algorithm has strong voltage control ability.