An Attention-Based Reinforcement Learning Approach For Electric Vehicle Charging Control

Efficient electric vehicle (EV) charging management is crucial for sustainable energy utilization and smart grid operation. To overcome charging inefficiency and system load imbalance, we propose a deep-reinforcement-learning-based approach to learn the optimal charging control strategy. The conundrum is expounded as a Markov decision process characterized by obscure state transition functions. The principal objective encompasses the determination of the supreme sequential charging stratagem capable of harmonizing the dual imperatives of charging efficacy and charging station expenditure. We provide an overview of EV charging management and related work and describe the principles of the Proximal Policy Optimization algorithm and its application in charging management. By integrating Long Short-Term Memory, the algorithm use the past state information to predict future trends. Meanwhile, Squeeze-and-Excitation Network dynamically adjusts channel weights to capture important information, improving charging efficiency and load balancing. Finally, the simulation results verify that this method can significantly improve charging efficiency and load balance compared with traditional method.