Self-Learning Control Systems for Autonomous EV Charging Operations

Abstract

Electric vehicle charging systems are moving toward autonomous operation as charging demand, charger occupancy, and grid constraints become more dynamic across large service networks. Existing studies on intelligent EV charging show the value of reinforcement learning and adaptive scheduling, but self-learning control architectures that continuously improve operational behavior remain limited. This gap matters because fixed control policies often perform poorly under fluctuating arrival intensity, variable demand, and changing power availability, leading to higher waiting time, inefficient allocation, and unstable control. This article presents a self-learning control system for autonomous EV charging operations based on state observation, adaptive action selection, reward-driven policy updating, and safety-constrained charger control. The results show that the proposed controller reduces charging cost and waiting time, improves allocation efficiency, and maintains strong control stability and adaptive response under varying EV arrival and grid constraint conditions. The study shows that self-learning control can support efficient, stable, and autonomous EV charging operations.