摘要
针对大型充电场站内规模化电动汽车的有序充电问题,提出一种基于双深度Q网络(DDQN)深度强化学习方法的电动汽车充电安排策略,能有效计及电动汽车出行模式和充电需求的不确定性,实现充电场站充电成本最小化的目标。首先,对电动汽车泊车时间和充电需求特征进行提取;其次,为解决维数灾问题,提出一种“分箱”方法和充电次序优化策略以控制状态和动作空间的大小,从而建立了一种适用于大规模电动汽车有序充电的马尔可夫决策过程(MDP)模型;然后,应用DDQN的强化学习算法对电动汽车有序充电策略进行求解;最后,通过仿真算例验证了所提方法的有效性,不仅能有效减少充电场站的充电成本,而且能使模型训练难度不受电动汽车规模影响。
Aiming at the coordinated charging problem of large-scale electric vehicles(EVs)within a large charging station, an EV charging scheduling strategy based on double deep Q network(DDQN)of deep reinforcement learning is proposed, which can effectively take into account the uncertainty of traveling pattern and charging demand of EVs and achieve the objective of minimizing the charging cost of the charging station. Firstly, characteristics of the parking time and the charging demand of EVs are extracted. Secondly, to solve the curse of dimensionality problem, a binning method and an optimal charging order strategy are proposed to control the size of the state and action space, and thus a Markov decision process(MDP)model which is applicable to the coordinated charging of large-scale EVs is established. Then, the DDQN of reinforcement learning algorithm is used to calculate the EV coordinated charging strategy. Finally, the validity of the proposed method is verified through a simulation example. It is verified that not only the charging cost of the charging station can be effectively reduced, but also the difficulty of model training becomes unaffected by the scale of EVs.
作者
陈果
王秀丽
原晟淇
帅轩越
周前
CHEN Guo;WANG Xiuli;YUAN Shengqi;SHUAI Xuanyue;ZHOU Qian(School of Electrical Engineering,Xi'an Jiaotong University,Xi'an 710049,China;State Grid Jiangsu Electric Power Co..L.td.Research Institute,Nanjing 210036,China)
出处
《电力系统自动化》
EI
CSCD
北大核心
2023年第2期88-95,共8页
Automation of Electric Power Systems
基金
国家电网公司科技项目(5400-202099508A-0-0-00)。
关键词
电动汽车
充电场站
深度强化学习
有序充电
维数灾
马尔可夫决策过程
electric vehicle
charging station
deep reinforcement learning
coordinated charging
curse of dimensionality
Markov decision process
