SAC-based Multi-agent Framework for Continuous Volt-Var Control in Distribution Network with High Penetration of PVs

The penetration of high proportional PVs has greatly changed the structure and operation of the distribution network, which brings great challenges to the stable operation of the distribution network. In this paper, we propose a soft actorcritic based multi-agent reinforcement learning method for Volt-Var control of distribution networks. The proposed method adopts a framework of centralized training and decentralized execution, and the designed agent can learn coordinated control strategies from historical data. Compared with traditional algorithms this paper proposed multi-agent soft actor-critic algorithm has good performance, stable performance, and strong anti-interference capability. Comparative studies on IEEE 33-node test systems demonstrate the performance of the proposed solution.