基于局部离群点检测和标准差方法的锂离子电池组早期故障诊断  

Early fault diagnosis of lithium-ion battery packs based on improved local outlier detection and standard deviation method

作　　者：李纪伟 刘睿涵 吕桃林[2,3,4,5] 潘隆 马常军 李清波 赵芝芸 杨文 解晶莹[2,3,4] LI Jiwei;LIU Ruihan;LU Taolin;PAN Long;MA Changjun;LI Qingbo;ZHAO Zhiyun;YANG Wen;XIE Jingying(East China University of Science and Technology,Shanghai 200237,China;Shanghai Institute of Space Power-Sources,Shanghai 200245,China;Shanghai Aerospace Power Technology Co,Ltd.,Shanghai 201112,China;Shanghai Power&Energy Storage Battery System Engineering Tech Co,Ltd.,Shanghai 200245,China;Shanghai Jiuxing Energy Technology Co,Ltd.,Shanghai 200801,China)

机构地区：[1]华东理工大学,上海200237 [2]上海空间电源研究所,上海200245 [3]上海航天电源技术有限责任公司,上海201112 [4]上海动力储能电池系统工程技术有限公司,上海200245 [5]上海玖行能源科技有限公司,上海200801

出　　处：《储能科学与技术》2023年第9期2917-2926,共10页Energy Storage Science and Technology

基　　金：国家重点研发计划项目(2022YFB2404304);上海市科学技术委员会“科技创新行动计划”资助项目(22DZ1206800);上海市科学技术委员会“科技创新行动计划”资助项目(22DZ1200600);上海市产业协同创新(科技)项目(XTCX-KJ-2022-2-09);国家电网有限公司总部科技项目(4000-202317096A-1-1-ZN);上海市2021年度“科技创新行动计划”启明星项目(21QB1401400)。

摘　　要：锂离子电池由于具有能量密度高、使用寿命长等优点,被广泛应用。为了更加准确地检测出电池组中存在安全隐患的故障电池,本文提出了一种故障检测方法。首先,根据单体故障引起电池组一致性差异,使用引入滑动窗的局部离群点检测算法,检测电池组中不一致单体,同时捕捉单体不一致特性的演化性,根据演化性区分仅存在不一致的单体和具有隐患的故障单体,并对单体的不一致程度划分等级;其次,利用改进标准差算法对检测出的单体诊断故障类型,提取包含故障类型信息的特征作为输入,引入“故障系数”,结合不同故障的判定标准和阈值,实现放大故障特征的同时,区分不同故障类型,有效诊断出早期内短路故障;最后,通过电池组真实运行数据,对所提方法进行验证,分析结果证明本工作所提出算法的有效性、可靠性。Lithium-ion batteries are commonly used for their high energy density and long service life.This paper suggests a fault detection method for actual battery pack operation data to more accurately detect faulty batteries with safety hazards in the battery pack.First,considering the differences in battery pack consistency caused by faulty cells,the local outlier detection algorithm is used with sliding windows to diagnose faulty cells more accurately,while detecting the inconsistency of battery cells and capturing the evolutionary characteristics of cell inconsistencies.Cells only with inconsistencies and faulty cells with hidden dangers are distinguished based on evolution;meanwhile,the inconsistency degree of cells is graded.Second,the fault types of detected cells are diagnosed by an improved standard deviation algorithm.Features containing fault type information are extracted as input,and a malfunction coefficient is introduced to realize the amplification of fault features,distinguishing different fault types by combining judgment criteria and thresholds for different faults.Early internal short circuits can be diagnosed effectively.Finally,the proposed method is verified through real operating data of the battery pack.The analytical findings demonstrate the effectiveness and reliability of the algorithm proposed in this study.

关 键 词：锂离子电池 局部离群检测 标准差 故障检测 

分 类 号：TM911.3[电气工程—电力电子与电力传动]