Intelligent charging scheduling algorithm for electric vehicle cluster access in transportation charging facilities

Abstract

Electric vehicle (EV) cluster access is becoming a key control problem for urban intelligent transportation charging facilities, because concentrated arrivals at residential and public chargers can reshape feeder loading, driver waiting time, and voltage security at the same time. This paper proposes an intelligent rolling scheduling algorithm that combines scenario-based demand encoding, reinforcement-learning candidate dispatch, and model-predictive safety projection. The method explicitly considers stochastic arrival time, departure deadline, state of charge, charger rating, time-of-use tariff, feeder voltage, branch loading, and user service priority. A 24-h case study with 300 EVs, 96 control intervals, and repeated Monte Carlo demand samples is used for verification. Compared with uncontrolled charging, tariff-only charging, and deterministic MPC, the proposed scheduler reduces the daily peak load from 742 kW to 580 kW relative to the tariff-only baseline, lowers charging cost by 14.6%, raises the minimum bus voltage from 0.946 p.u. to 0.962 p.u., and maintains a 97.3% on-time completion rate. Ablation and sensitivity tests further show that the safety projection is critical for voltage feasibility, while the priority repair mechanism prevents service degradation under dense evening arrivals.