基于真实数据的电动汽车锂电池故障检测方法

当电池的异常特征不明显时，传统的电动汽车电池系统故障检测方法很难进行早期故障检测。当前大多方法都是基于实验室条件下的测试数据进行研究，利用电动汽车实际运行数据的研究较少。为解决上述问题，提出一种基于真实数据的电动汽车电池系统内短路故障在线检测方法，通过经验模态分解提取分解后的电压残差值作为故障特征，结合香农熵权重法，以每个采样点的香农熵的冗余度作为权重，对串联电池系统中各电芯单体进行评分，结合改进Z-分数，实现对串联电池组的故障检测和定位。利用真实车辆数据进行验证并与阈值法和相关系数法进行比较，验证了该方法的有效性。结果表明，所提出的方法计算成本低、可靠性高，能够在线应用，不需要精确的模型且无需针对不同型号车辆的得分阈值进行试验，降低了试验成本。

Fault Detection Method for Electric Vehicle Lithium Battery Using Real-world Data

Conventional electric vehicle battery system fault detection methods are tough to detect early faults when the abnormal characteristics of the battery are not obvious. Most of the current methods are based on the test data under laboratory conditions, and there are few studies using the real-world data of electric vehicles. To solve the above problems, an online internal short circuit fault detection method for electric vehicle battery system based on real data was proposed, which extracted the decomposed voltage residual values as fault features by empirical modal decomposition, combined the Shannon entropy weighting method with the redundancy of the Shannon entropy of each sampling point as the weight to score each cell in the series battery system, and combined the modified Z-score to achieve the fault detection and localization of series-connected battery pack. The validity of the algorithm was verified by utilizing real-world electric vehicle data and compared with the threshold method and the Pearson correlation coefficient method. The results show that the proposed method has low calculation cost and high reliability. It can be applied online, without the need for accurate models, and without the need to test for the score threshold of different models of vehicles, which reduces the test cost.