Two-stage optimization framework for electric vehicle automatic charging stations with differential charging strategies in photovoltaic-energy storage system microgrid

Artikel i vetenskaplig tidskrift, 2026

The rapid expansion of the electric vehicle (EV) market has intensified the demand for efficient charging infrastructure, yet conventional fixed-charger paradigms suffer from low utilization rates, pronounced grid peak-valley disparities, and insufficient renewable energy absorption due to the one-vehicle-one-charger binding constraint. This study proposes a two-stage robust optimization framework for Electric Vehicle Automatic Charging Stations integrated with a photovoltaic-energy storage system microgrid: Stage 1 employs a reinforcement-learning-enhanced NSGA-II and TOPSIS to simultaneously optimize economic efficiency, reclassification count, and grid load variance under a budget-of-uncertainty formulation that jointly handles EV arrival/departure time uncertainty and photovoltaic output variability; Stage 2 combines CPLEX-based exact EV grouping with a Wolpertinger-DDPG architecture for dynamic rail-mounted mobile charger (RMC) assignment, enabling a one-charger-serves-multiple-vehicles paradigm through rail-mounted robotic charging arms. A scheduling-flexibility-based EV trichotomy with a formal reclassification safety-net mechanism and a price-inversion-preventing compensation design are embedded to guarantee differentiated yet equitable charging services. Comprehensive validation was conducted across three urban scenarios with 10,000 users. These results confirm that the proposed framework offers a scalable and robust solution for large-scale sustainable EV charging infrastructure deployment.