摘要
电动汽车作为电网中的重要负荷,具有较高的需求响应潜力.为降低电动汽车集群用能成本,缓解电网峰值负荷压力,文中首先分析了V2G模式下电动汽车的用电特性,构建了电动汽车集群充放电调度模型,通过成本分析为充放电调度提供决策依据.然后应用多智能体深度确定性策略梯度(Multi-Agent Deep Deterministic Policy Gradient,MADDPG)算法对电动汽车集群进行充放电行为的实时优化,利用用户的历史用电数据完成学习过程,并依据当前的用电信息进行调度决策.算例分析表明,该方法可以进行电动汽车集群充放电行为的实时在线优化决策,在保证用户用电需求的前提下,提高用户用电经济性,实现峰值负荷的转移.
As an important load in power grid, electric vehicles have high demand response potential. To reduce the energy cost of electric vehicles and peak load pressure on the grid,this paper first analyzes the electricity characteristics of electric vehicles in V2G mode,constructs the electric vehicle cluster charging and discharging scheduling model,and provides the basis of decision making for charging and discharging scheduling through cost analysis. Then apply the multi-agent deep deterministic policy gradient(MADDPG) algorithm to optimize the charging and discharging behavior of electric vehicle cluster in real time,complete the learning process through users’ historical electricity consumption data,and make scheduling decisions through the current electricity consumption information. Example analysis shows that this method can optimize the charging and discharging behavior of electric vehicle cluster online in real time to improve the users’ power consu-mption economy and transfer the peak load under the premise of ensuring the users’ demand for electricity.
作者
戴武昌
刘艾冬
申鑫
马鸿君
张虹
Dai Wuchang;Liu Aidong;Shen Xin;Ma Hongjun;Zhang Hong(Key Laboratory of Modern Power System Simulation and Control&Renewable Energy Technology(Northeast Electric Power University),Ministry of Education,Jilin Jilin 132012;Fushun Power Supply Company,State Grid Liaoning Electric Power Co.Ltd.,Fushun Liaoning 113000)
出处
《东北电力大学学报》
2021年第5期80-89,共10页
Journal of Northeast Electric Power University
基金
吉林省科技计划重点研发项目(20180201010GX)。
关键词
电动汽车充电调度
需求响应
多智能体深度强化学习
在线优化
Electric vehicle charging scheduling
Demand response
Multi-agent deep reinforce ment learning
Online optimization