基于孤立森林算法的锂离子电池微内短路故障诊断方法  

Fault diagnosis of micro-internal short circuits in lithium-ion battery using the isolated forest algorithm

作　　者：郭煜 王亦伟[1,3,4] 彭鹏[1,3,4] 王银飞 丘意书[1,3,4] 蒋方明[1,3,4] GUO Yu;WANG Yiwei;PENG Peng;WANG Yinfei;QIU Yishu;JIANG Fangming(Guangzhou Institute of Energy Conversion,Chinese Academy of Sciences,Guangzhou 510640,Guangdong,China;Dongfeng Motor Corporation Research and Development Institute,Wuhan 430056,Hubei,China;CAS Key Laboratory of Renewable Energy,4Guangdong Provincial Key Laboratory of New and Renewable Energy Research and Development,Guangzhou 510640,Guangdong,China;School of Mechanical and Power Engineering,Shenyang University of Chemical Technology,Shenyang 110142,Liaoning,China)

机构地区：[1]中国科学院广州能源研究所,广东广州510640 [2]东风汽车集团有限公司研发总院,湖北武汉430056 [3]中国科学院可再生能源重点实验室 [4]广东省新能源和可再生能源研究开发与应用重点实验室,广东广州510640 [5]沈阳化工大学机械与动力工程学院,辽宁沈阳110142

出　　处：《储能科学与技术》2024年第11期4102-4112,共11页Energy Storage Science and Technology

基　　金：广州市科技计划(202201010418)。

摘　　要：电池系统的内短路故障是造成电池热失控和火灾事故的主要原因之一,因此有必要对电池内短路故障进行诊断对事故做出早期预警。孤立森林算法是一种无监督的异常检测算法,被广泛应用于异常数据识别领域。根据串联电池组中内短路电池的电压会与正常电池发生偏离的特点,本文提出了基于孤立森林算法的锂离子电池微内短路故障诊断方法。为了对方法进行验证,本文构建了串联电池组进行了不同短路电阻和充放电工况的短路实验,并在实际运行工况下对一个锂离子电池储能系统进行了电池短路实验,然后利用孤立森林算法对实验数据进行分析诊断。结果表明,对于循环充放电工况,孤立森林算法对短路电阻为1000Ω的短路故障诊断精准率超过了74%,召回率超过了76%,准确率超过了91%;在模拟电动汽车实际驾驶的动态工况中,算法对300Ω的短路故障诊断精准率和召回率超过了86%,准确率超过了95%;算法对电池储能系统在实际运行工况下25Ω的内短路故障检测召回率大于98%。实验结果表明,孤立森林算法可以在多种工况下对锂离子电池微内短路故障进行有效检测,被检测出的锂离子电池内短路电阻达到了千欧姆数量级。Internal short-circuit(ISC)faults in lithium-ion battery(LIB)systems are major contributors to thermal runaway and fire incidents.Diagnosing ISC faults is crucial for early warning of potential accidents and ensuring the safe operation of LIB systems.The isolated forest algorithm,an unsupervised anomaly detection method,is widely utilized in identifying anomalous data.Leveraging the characteristic voltage deviation of ISC-affected LIBs within a series-connected LIB pack,this study proposes an ISC fault diagnosis method based on the isolated forest algorithm.To validate the proposed method,a series-connected LIB module was constructed to perform ISC experiments under various short-circuit resistance conditions.ISC experiments were also conducted in an echelon-utilized LIB energy storage system(ESS)under real-world operating conditions.The isolated forest algorithm was then applied to analyze the experimental data.Results indicate that,under cyclic charging and discharging conditions,the algorithm achieved an accuracy rate of over 74%,a recall rate exceeding 76%,and a precision rate above 91%for a 1000ΩISC fault.For dynamic driving conditions of electric vehicles,the algorithm demonstrated an accuracy rate above 86%and a recall rate over 95%for a 300ΩISC fault.In the ESS's actual operating conditions,the recall rate for detecting a 25ΩISC fault exceeded 98%.The experimental outcomes confirm that the isolated forest algorithm effectively detects micro-ISC faults in LIBs across various operational scenarios,with detected ISC resistance reaching the magnitude of thousands of ohms.

关 键 词：锂离子电池 孤立森林算法 微内短路故障 电池储能系统 

分 类 号：TM912[电气工程—电力电子与电力传动]