Optimization of Electric Vehicle Smart Charging and Discharging Decisions Based on Policy Gradient Reinforcement Learning

Electric vehicles (EVs) are becoming increasingly prevalent, presenting significant challenges for power system operations. As a result, smart charging solutions that maximize cost-effectiveness, grid stability, and environmental sustainability are required. To address this need, this paper employs policy gradient reinforcement learning (RL) to develop a technique for EV optimization. Specifically, the created framework in-corporates time-of-use pricing, EV battery specifications, and power system limits. Within this framework, a multi-objective reward function is developed to balance long-term goals like grid load balancing, including strategies such as discharging during peak times and charging during low times, and short-term cost reduction. Importantly, research has found that the adaptive nature of this method enables it to effectively respond to dynamic changes in electricity pricing and grid conditions, demonstrating its potential for real-world applications.