Power Dispatching Strategy of Electric Vehicle Charging Station Based on Reinforcement Learning and Heuristic Priority

Electric vehicles (EVs) are developing more and more rapidly because of their low carbon, environmental protection, and many other advantages. However, the charging behavior of a lot of EVs will bring shocks to the grid. For electric vehicle charging stations (EVCS), it is beneficial to formulate a reasonable power dispatching strategy to improve economic efficiency and reduce power grid loss. This paper proposes a two-stage efficient power-dispatching strategy. In the first stage, the charging station develops a real-time power purchase strategy based on the deep Q-learning network (DQN) to maximize profit. In the second stage, the charging station distributes power to each electric vehicle (EV) through a heuristic priority-based power distribution method to minimize cost. Simulation results show that reinforcement learning plays a significant role in the power dispatching of EVCS. The net profit of charging stations has also improved significantly. At the same time, the application of the heuristic power distribution mechanism further highlights the superiority of reinforcement learning.

I. Introduction

Smart grid is one of the important ways to deal with the shortage of human resources in the 21st century, which can meet sustainable development and realize the optimal allocation of power in a wide range. Due to environmental pollution and other issues, EVs have gradually replaced traditional fuel vehicles in recent years. Therefore, as an essential part of the smart grid, EVCS also face certain challenges in power distribution and scheduling [1]. For example, dispatching has low economic benefits, and resource allocation is unreasonable.