An AVC Optimization Strategy Based on Improved Deep Reinforcement Learning

An AVC optimization strategy based on improved deep reinforcement learning is proposed to address the problem of traditional voltage control methods being unable to meet the voltage regulation requirements of new power systems. Firstly, an AVC optimization model is constructed with the objectives of voltage, main transformer tap changer, and capacitor operation times, including modeling of on load tap changers and reactive power compensation capacitor banks to comprehensively improve voltage stability. Then, an improved Twin Delay Deep Deterministic Policy Gradient (TD3) algorithm is introduced to optimize the Double Deep Q-Network (DDQN) and applied to solve the AVC optimization model to obtain reliable optimization measures. Finally, experimental analysis is conducted on the proposed strategy based on the IEEE-33 node system. The results show that the optimization results of the proposed strategy are the most ideal, with average voltage deviation, tap action times, capacitor action times, and optimization time divided into 0.0015p. u, 5 times, 6 times, and 36ms, which can provide effective control strategies for actual power grid operation.