A Deep Reinforcement Learning Method for Charging Station Management and Load Balancing

Electric vehicles (EVs) and the corresponding renewable energy-based charging stations (CSs) are rapidly gaining popularity as an effective solution for environmental protection. However, the increasing adoption of EVs and CSs can lead to load fluctuations and pose safety risks to the grid. To address this issue and develop a practical management strategy for CSs, we propose a load-constraint hierarchical deep reinforcement learning approach. Our method aims to achieve two main objectives: 1) maximize the total profit and balance the load for CSs, taking into account the unpredictable behavior of EV users; and 2) separate pricing and charging scheduling actions to enhance training efficiency. We conducted experiments using real data, and the results validate the effectiveness of our algorithm in achieving load balancing and maximizing profits for CSs.