基于经验模态分解的门控循环单元神经网络的锂离子电池荷电状态估计  

State-of-Charge Estimation of Lithium-Ion Battery Based on Gated Recurrent Unit Using Empirical Mode Decomposition

作　　者：李宁[1] 何复兴 马文涛[1] 蒋林[2] 张小平[3] Li Ning;He Fuxing;Ma Wentao;Jiang Lin;Zhang Xiaoping(School of Electrical Engineering Xi'an University of Technology,Xi’an 710048,China;Department of Electrical Engineering and Electronics University of Liverpool Liverpool,L693GJ,United Kingdom;Department of Electronics,Electrical and Systems Engineering University of Birmingham Birmingham,B152TT,United Kingdom)

机构地区：[1]西安理工大学电气工程学院,西安710048 [2]利物浦大学电气工程和电子系利物浦,英国L693GJ [3]伯明翰大学电子、电气和系统工程系伯明翰,英国B152TT

出　　处：《电工技术学报》2022年第17期4528-4536,共9页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(51507140);国家留学基金委国际清洁能源拔尖人才项目([2018]5046,[2019]157);江苏省配电网智能技术与装备协同创新中心开放基金项目(XTCX202007)资助。

摘　　要：锂离子电池荷电状态(SOC)估计技术是电动汽车电池监测系统(BMS)设计的重要组成部分。该文提出一种基于经验模态分解(EMD)的门控循环单元(GRU)神经网络的锂离子电池荷电状态估计方法,在GRU估计SOC的基础上,引入EMD算法分解放电电流,不仅提高GRU模型对长时间电流信号保持长期信息的能力,而且提高锂离子电池荷电状态估计精度。仿真实验表明,与传统的循环神经网络和长短期记忆网络相比,该文所提基于EMD-GRU方法的锂离子电池SOC估计平均绝对误差为1.5093%,同比降低了20.7924%。State of charge(SOC)estimation technology of lithium-ion battery is an important part of the battery management system(BMS)design of electric vehicles.In this paper,an SOC estimation method for lithium-ion batteries based on gated recurrent unit(GRU)using empirical mode decomposition(EMD)is proposed.The EMD algorithm is introduced to decompose the discharge current,based on the GRU estimation of the SOC,which not only improves the ability of the GRU model to maintain long-term information for long-term current signals,but also betters the accuracy of SOC estimation of lithium-ion battery.Simulation experiments show that,compared with the traditional recurrent neural network and long-term and short-term memory network,the EMD-GRU method proposed in this paper displays the average absolute error of the lithium-ion battery SOC estimation is 1.5093%,a year-on-year decrease of 20.7924%.

关 键 词：锂离子电池 荷电状态估计 门控循环单元 经验模态分解 

分 类 号：TM912[电气工程—电力电子与电力传动]