Voltage Reactive Control Strategy in Distribution Network Based on Multi-agent Deep Reinforcement Learning
In order to meet the demand of voltage control for distribution networks with high proportion access of distributed power supply,a distribution network voltage reactive control strategy based on multi-agent deep reinforcement learning was proposed.Firstly,the mathematical model was constructed with the objective of minimizing the voltage offset of distribution network nodes,and each distributed solar inverter was modeled as an agent;then,by partitioning the distribution network,the inverter collaborative control problem was established as a decentralized and observable Markov decision process for each sub region;the multi-agent twin delayed deep deterministic policy gradient algorithm was used to solve the real-time optimization control strategy;finally,and the simulation testing was conducted on an IEEE 33 node system.The results indicate that the proposed method is effective in voltage reactive controlling in distribution networks.
distribution networkvoltage reactive controldistributed generationmulti-agent deep reinforcement learningMarkov decision process
万方数据
维普
