Research on Optimization Configuration of Urban Substation Charging Station Based on Deep Reinforcement Learning Algorithm

With the deepening of the “dual carbon” policy, the electric vehicle industry, as an important way to replace fossil fuels, has developed rapidly in recent years and become a focus of attention in multiple industries such as electricity, transportation, and chemical engineering. Electric vehicle fast charging stations occupy an important position in the public charging market due to their efficient and fast charging characteristics. However, the high randomness and rapidly increasing charging load brought by fast charging stations pose serious challenges to the safe and stable operation of the power grid. This study proposes an optimized operational configuration method for typical fast charging stations for electric vehicles based on DQN (Deep Q-Network) algorithm, and specifically introduces the content of substation charging configuration. This study aims to optimize the capacity of each charging station based on economic and operational efficiency by analyzing the output characteristics of distributed power sources within the station and the charging behavior patterns of electric vehicles, and to construct a typical fast charging station optimization operation configuration model. This model is not only constrained by key factors such as power balance within the station and distributed power output, but also considers optimization of substation charging configuration to ensure efficient operation of charging stations and stability of the power grid. By using DQN algorithm instead of traditional genetic optimization algorithm to solve the optimal solution of the model, this study achieved refined management of charging station configuration. Finally, through empirical analysis of various configuration examples, the effectiveness and feasibility of the proposed method were verified, providing a new technical path and decision support for the optimized operation of urban fast charging stations.