Dynamic voltage collaborative control of distributed photovoltaic clusters based on multi-agent deep reinforcement learning

Due to the randomness and volatility of distributed photovoltaic output, large-scale access to the distribution network can easily lead to issues such as voltage exceeding limits and fluctuations. A dynamic voltage collaborative control method for distributed photovoltaic clusters based on multi-agent deep reinforcement learning is proposed. Based on electrical distance spectrum clustering, distributed photovoltaic distribution network nodes are partitioned to form physically closely connected and internally autonomous photovoltaic clusters. Map each cluster as an intelligent agent, use multi-agent deep deterministic policy gradient algorithm, and achieve dynamic collaborative control of reactive power and voltage between clusters through centralized training and distributed execution paradigm. Through case analysis, the superiority of this method in dynamic voltage collaborative control of distributed photovoltaic clusters has been verified, which can provide an effective technical path for voltage governance under high penetration rate photovoltaic access.