基于双DQN和扩展卡尔曼滤波的锂离子电池荷电状态估计  

State of Charge Estimation of Li-ion Batteries Based on Double DQN and Extended Kalman Filters

作　　者：高洪森[1] 王雪 魏宏博 游国栋[2] 侯晓鑫 赵双乐 GAO Hongsen;WANG Xue;WEI Hongbo;YOU Guodong;HOU Xiaoxin;ZHAO Shuangle(Tianjin Lishen Battery Joint-stock Co.,Ltd.,Tianjin 300384,China;College of Electronic Information and Automation,Tianjin University of Science&Technology,Tianjin 300222,China)

机构地区：[1]天津力神电池股份有限公司,天津300384 [2]天津科技大学电子信息与自动化学院,天津300222

出　　处：《天津科技大学学报》2022年第4期49-54,共6页Journal of Tianjin University of Science & Technology

基　　金：天津市科技支撑重点项目(17YFZCNC00230);天津市自然科学基金重点资助项目(13JCZDJC29100)。

摘　　要：为了提高锂离子电池荷电状态(SOC)的估计精度,设计了一种基于双深度Q网络(双DQN)和扩展卡尔曼滤波(EKF)的锂离子电池SOC估计算法.选择锂离子电池二阶RC等效电路为研究对象,采用EKF算法重构了锂离子电池的离散系统数学模型;结合深度强化学习思想,构造了一种深度强化学习扩展卡尔曼滤波算法.该算法设计了双DQN,并对EKF参数进行优化.仿真结果表明,与DQN扩展卡尔曼滤波算法相比,双DQN扩展卡尔曼滤波算法具有更好的收敛性、自适应能力以及估计精度.In order to improve the accuracy of the state of charge(SOC)estimation of the Li-ion battery,a Li-ion batteries state of charge estimation algorithm based on double depth Q network(double DQN)and extended Kalman filters(EKF)is proposed in this article.The second-order RC equivalent circuit model was selected as the research object,state-space model of coefficient of variation of the Li-ion battery was structured by EKF algorithm.Moreover,combined with the idea of deep reinforcement learning,a reinforcement learning extended Kalman filters algorithm was structured,which designed a double DQN and optimized the EKF parameters.Compared with DQN extended Kalman filter algorithm,double DQN extended Kalman filters has better convergence,adaptive ability and better estimation accuracy through simulation.

关 键 词：锂离子电池 双DQN 荷电状态 扩展卡尔曼滤波 

分 类 号：TM912[电气工程—电力电子与电力传动]