Behavior-driven capacity configuration for highway electric vehicle charging stations via soft actor-critic reinforcement learning

With the rapid expansion of the electric vehicle (EV) industry and surging intercity travel demands, highway charging stations face severe challenges of queuing congestion and supply-demand imbalance. Previous charging infrastructure planning studies often ignore the spatiotemporal heterogeneity of dynamic traffic flow and the interactive behaviors between driving and charging. Accordingly, this study proposes a Dynamic EV Driver Behavior-based Highway Charging Planning (DEB-HCP) model. First, by integrating macroscopic traffic trajectories with microscopic charging data via the SUMO simulator, the model captures stochastic charging demands and charging behaviors driven by range anxiety and battery state of charge. Next, utilizing the Soft Actor-Critic algorithm, this study designs three reward functions (Static Cost-Benefit Reward, State-Aware Adaptive Reward, and Target Utilization Guiding Reward) to reveal the regulatory mechanisms of different incentive schemes. Finally, using real-world data from four consecutive service areas along the G50 Expressway in Hubei, China, this study conducts multiple experiments across an annual average scenario and two high-load holiday scenarios. Results indicate that, under annual average scenario, DEB-HCP reduces construction and maintenance costs by 25.6% and 27.1%, while increasing peak-hour utilization by 26.5% and 24.6% compared to existing configurations and multi-objective optimization scheme, respectively. During holidays, it alleviates congestion and stabilizes facilities utilization at 60% to 70%. Furthermore, directly transferring the trained model between distinct holiday scenarios yields stable metrics, confirming the generalization ability and robustness of the model. This study provides a high-fidelity simulation and evaluation framework for the capacity sizing of chargers in complex traffic environments.