Reinforcement Learning-Based Optimization for EV Charging Scheduling and Resource Allocation

Abstract

Electric vehicle charging stations must manage uncertain vehicle arrivals, limited charger availability, variable demand, and grid-side constraints within the same operational setting. Conventional scheduling methods often perform poorly because they separate charging-time control from charger allocation, which reduces adaptability under congestion. This study proposes a reinforcement learning-based framework for EV charging scheduling and resource allocation, where the station is modeled as a sequential decision environment using queue status, charger occupancy, charging demand, and load conditions. The methodology integrates state design, action formulation, reward modeling, and constraint handling into one learning architecture. Simulation results show that the learned policy reduces charging cost, smooths grid load, improves charger utilization, lowers waiting time, and increases allocation efficiency compared with conventional baseline methods. The strongest gains appear under moderate and peak arrival scenarios, where static rules become less effective. The study shows that reinforcement learning is a practical and effective approach for coordinated EV charging control in smart charging infrastructure.