Intelligent Layout and Optimization of EV Charging Stations: Initial Configuration via Enhanced K-Means and Subsequent Refinement through Integrated GCN

Abstract

This paper proposes an optimization model for the layout of EV charging stations, aiming to ensure a wide and efficient service area to meet the increasing demand for charging. Through an in-depth study of the deployment optimization of EV charging stations, a layout algorithm based on K-Means and simulated annealing is first introduced to determine the optimal locations for new charging stations. Building on this, a layout optimization algorithm utilizing a Residual Attention Graph Convolutional Network (RAGCN) is proposed, which leverages the efficient learning capability of Graph Convolutional Networks (GCN) on graph-structured data to learn and obtain the best layout for charging stations. Finally, the effectiveness of the model is validated in Nanjing, Jiangsu Province. The results show that the optimized layout of charging stations, which added 493 new stations in high-demand areas such as business districts and corporate enterprises, significantly enhances the convenience and utilization rate of charging for EV users. Additionally, sensitivity analysis and ablation experiments based on Points of Interest (POI) data are conducted to evaluate the impact of various POI features on the layout of charging stations and to explore the contribution of different model components to classification performance.