基于二重故障区间定位和随机森林算法的锂电池组突发性故障诊断

Sudden Fault Diagnosis of Lithium Battery Pack Based on Double Fault Interval Location and Random Forest Algorithm

锂电池的突发性故障往往会在短时间内造成极大的危害,制约其进一步的发展。因此,本文对突发性故障进行研究,提出了一个完整的锂电池组突发性故障诊断框架。首先,利用经验模态分解(EMD)对电压数据序列进行分解,将子序列构建新的数据序列并导入二重故障区间定位模型中,利用二分法思想的相关系数和基于时间窗的相关系数诊断突发性故障的故障区间。然后,利用核主成分分析法提取故障特征。最后,使用随机森林算法对故障进行训练和分类,并且与其余方法进行比较。此外,建立故障注入平台,以物理方式触发串联4节电池组上的外部短路和接触故障。试验结果表明,该方法对突发性故障的诊断准确可靠。

Sudden faults of lithium batteries often cause great harm in a short period of time,restricting their further development.Therefore,this paper investigates sudden faults and proposes a complete framework for diagnosing sudden faults in lithium battery packs.First,the voltage data sequence is decomposed using empirical modal decomposition,and the subsequence is constructed into a new data sequence and imported into the binary fault interval localization model,and the correlation coefficient of dichotomous ideas and the correlation coefficient based on the time window are used to diagnose the fault interval of the sudden fault.Then,the fault features are extracted using kernel principal component analysis.Finally,the faults are trained and classified using the random forest algorithm and compared with the remaining methods.In addition,a fault injection platform is established to physically trigger external short circuit faults and contact faults on a series-connected four-cell battery pack.The experimental results show that the method is accurate and reliable for the diagnosis of sudden faults.