Abstract: [Objectives] To simultaneously suppress voltage fluctuations at the point of common coupling of doubly-fed induction generator wind farms and enhance their reactive power margin, thereby strengthening their transient voltage support capability, a reactive power margin optimization method for doubly-fed induction generator wind farms based on an improved proximal policy optimization algorithm is proposed. [Methods] A mathematical model for optimizing the reactive power margin of a doubly-fed induction generator wind farm was established, with doubly-fed induction generator wind turbines and static reactive power generators as the main reactive power regulation equipment, and transformed into a Markov decision process model; A proximal policy optimization algorithm based on variable clipping coefficients is proposed to address the problem of difficulty for agents to balance exploration and exploitation caused by improper clipping coefficient settings in the original proximal policy optimization algorithm. [Results] Based on the improved IEEE 39 node system, the correctness and effectiveness of the proposed algorithm were verified. Simulation results showed that the proposed algorithm can improve the convergence speed of the agent, and is superior to traditional proximal policy optimization algorithms and heuristic algorithms in decision speed and reactive power margin optimization. [Conclusions] This algorithm effectively improves the reactive power margin of doubly-fed induction generator wind farms, enhances their voltage support capability under transient conditions, and is significant for better solving reactive power optimization problems in power systems with uncertain sources and loads.

Key words: double-fed induction generator wind farm, reactive power margin optimization, deep reinforcement learning, proximal policy optimization, clipping coefficient