Optimal scheduling method for electric vehicle charging and discharging via Q-learning-based particle swarm optimization

Large-scale electric vehicle (EV) access leads to grid fluctuations and reduces operational reliability. User charging demand unpredictability further increases the complexity of scheduling models. Additionally, the unstable output of distributed energy in relation to EV carbon emissions also poses new challenges. This paper proposes an optimal scheduling method for EV charging and discharging. First, an optimization model for grid load fluctuations and EV user cost was constructed considering time-of-use electricity price, EV access to the network, distributed energy generation, EV carbon quota, and charging and discharging load response characteristics. Second, a Q-learning-based particle swarm optimization (QPSO) algorithm was designed. The Q-learning algorithm was used to dynamically adjust the inertial parameters and learning factors to improve the QPSO algorithm search efficiency. An orthogonal experiment was conducted to determine the QPSO algorithm parameters, which were validated via simulations. The superior QPSO algorithm performance in solving this problem was demonstrated via multi-factor variance analysis. The grid load fluctuation and user charging cost before and after scheduling as well as the user carbon quota and grid load fluctuation under different carbon prices were analyzed, and the feasibility of the scheduling scheme was demonstrated.