Multi-agent Deep Reinforcement Learning Voltage Control Strategy for Active Distribution Networks Considering Privacy Protection
With the implementation of China's"dual carbon"goals and the rapid expansion of renewable energy,voltage control in distribution network faces new challenges.This study proposes a multi-agent deep reinforcement learning algorithm with inter-regional privacy protection to address decentralized voltage control in active distribution network,aiming to prevent privacy leakage from global information sharing during centralized training while improving voltage control performance.Firstly,a privacy-preserving collaborative control framework is developed,leveraging the principles of the multi-agent deep reinforcement learning.Then a decentralized reinforcement learning algorithm integrating local observation with global objective is established to optimize the coordinated control of voltage regulation device.Finally,case studies confirm that the proposed method improves voltage stability and security in the distribution network,achieving the efficient voltage control while preserving the privacy.
voltage controlprivacy protectiondecentralized controlmulti-agent deep reinforcement learning
万方数据
维普
