A Novel Deep Reinforcement Learning Based Method for Real-time Pricing and Scheduling in Electric Vehicle Charging Stations

This study introduces an innovative approach using deep reinforcement learning (DRL) algorithms to optimize real-time pricing and scheduling control in electric vehicle (EV) charging stations. Given a set of charging speed options, the objective is to minimize the total charging cost for all vehicles while ensuring each vehicle reaches its required target power. The core problem is formulated as a Markov Decision Process (MDP), where algorithms like Double Deep Q-Network (DDQN) are developed and tested to explore and refine charging strategies, enabling the agent to select the most efficient policy based on the current state. Through a simulated environment, the study evaluates various charging strategies by focusing on metrics such as charging efficiency, cost-effectiveness, user experience, and grid impact. A DRL-based simulation environment and agent were implemented, with parameters like learning rate and exploration rate optimized through iterative training. Results demonstrate that the agent effectively balances exploration and exploitation, achieving stable and high cumulative rewards throughout training. This research significantly contributes to EV charging station operations by enhancing efficiency and profitability while supporting grid stability.