Deep reinforcement learning method for energy management in fast charging station

Electric vehicle (EV) fast charging stations often experience high peak charging loads and strong fluctuations. Therefore, developing an energy management strategy for the energy storage system is crucial to improve load characteristics and reduce peak power. Addressing these challenges, this paper proposes a fast-charging station energy management strategy based on deep reinforcement learning. Firstly, the load characteristics of the charging station are analyzed. Subsequently, a mathematical optimization model is established, aiming to minimize the daily electricity purchase cost of the fast charging station while considering peak power reduction as a constraint. The control strategy is designed using a deep deterministic policy gradient algorithm. To validate the efficacy of the proposed deep reinforcement learning control strategy, a case study is conducted. The results demonstrate the significant reduction in peak load power, which validates the efficacy of the strategy.

I. Introduction

With the depletion of fossil energy and the escalating environmental concerns, energy transition has become a critical issue. Electric vehicles (EVs), renowned for their "zero emission and low noise" attributes, are playing an increasingly important role [1]. Consequently, the rapid expansion of EVs has led to rapid developments in their associated charging infrastructure. However, EV fast charging stations, as high and volatile electrical loads, can negatively impact and disrupt the reliable operation of the power grid [2]. It is imperative to regulate the charging load and enhance the load characteristics of fast charging stations. Common control methods include guiding users through time-of-use tariffs, direct power control, and implementing energy storage system(ESS).