Optimization Study of Active Distribution Network Operation Considering “Source-Storage” Coordination

In this paper, a deep reinforcement learning based multi-intelligence cooperative operation optimization method for active distribution network zoning is proposed. The proposed method transforms the “source-storage” coordinated operation optimization model of an active distribution network into a Markov game based on the deep reinforcement learning theory of multiple intelligences, constructs intelligences for distribution network zoning, and employs a multi-intelligence double-delay deep deterministic policy gradient algorithm to achieve centralized training and distributed deployment. The proposed method improves the problem of high percentage of distributed power access and effectively consumes renewable energy.