New energy station AGC and primary frequency modulation dynamic coordinated control based on multi-agent collaborative reinforcement learning

Due to the dispersed frequency regulation resources and significant differences in response characteristics of existing methods, this study investigates the dynamic coordination control of AGC and primary frequency regulation in new energy stations based on multi-agent collaborative reinforcement learning. Build a comprehensive dynamic model that covers multiple types of new energy generation units, energy storage systems, and grid frequency response. Model each frequency regulation resource as an independent agent and use a multi-agent deep deterministic policy gradient (MADDPG) algorithm with centralized training and decentralized execution. Through the collaborative mechanism between agents, optimize and coordinate AGC and primary frequency regulation. The controller dynamically adjusts the participation level and power allocation scheme of each frequency regulation resource based on the magnitude and trend of the grid frequency deviation. The simulation results show that in both step disturbance and continuous fluctuation scenarios, the experimental group can achieve rapid recovery of system frequency, and the AGC regional control deviation is maintained within a small range, effectively improving the ability of new energy stations to participate in grid frequency regulation.