Research on the Collaborative Scheduling Strategy for Power System AVC Based on Reinforcement Learning

With the expansion of power system scale and the transformation of energy structure, traditional scheduling methods are inefficient when facing complex power environments. This paper proposes a collaborative scheduling method for automatic voltage control (AVC) based on reinforcement learning, combining Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) algorithms, aiming to improve the voltage stability and economy of power systems. By constructing a power system scheduling model and defining the state space, action space, and constraints, reinforcement learning is used to achieve autonomous decision optimization. Simulation results show that, compared to traditional methods, the reinforcement learningbased scheduling strategy has significant advantages in voltage fluctuation control, load change response, and system stability. Especially in the multi-agent collaborative scheduling framework, the cooperation optimization between agents further enhances the overall scheduling performance. In addition, the reinforcement learning method demonstrates stronger adaptability and efficiency in large-scale power grids and distributed energy integration environments, showing its potential for future applications in power systems.