Intelligent Topology Control of Distribution Power Network for Increased Clean Energy Absorption Using Reinforcement Learning

With the increased penetration of distributed energy resources (DERs), traditional model-driven grid control strategies face limitations in adaptability, real-time decisionmaking, and predictive capability. This paper presents a reinforcement learning (RL)-based intelligent topology control method to address the challenges of local absorption of clean energy in modern distribution networks. By creating the Markov Decision Process (MDP) model, the state space consists of line loading, reverse power flow, switch statuses, and other grid information; the action space primarily considers topology changes via breaker switching. Through careful reward function design, the RL agent learns the optimal control policies through iterative interactions with the grid simulator to minimize the reverse power flow of the substation while ensuring security constraints. The proposed method is validated on a real-world regional distribution power grid in Shaoxing, China, demonstrating up to 30 % reduction in reverse power flow and improved clean energy absorption. The framework enhances grid flexibility, supports renewable energy integration, and provides a scalable solution for future high-renewable power systems.