基于卡尔曼滤波和特征指数化的电动汽车电池故障诊断方法研究

Battery Fault Diagnosis for Electric Vehicle Based on the Kalman Filter and Feature-Exponential-Function Method

针对电池组内的热失控、内短路等故障问题,提出了一种基于卡尔曼滤波和特征指数化的电池故障在线诊断方法。首先基于历史数据和卡尔曼滤波方法对电压数据进行降噪,可有效去除异常点,并提出一种特征指数化方法以提取和放大电池组单体之间的电压特征。最后,为了减少电池组不一致性导致的单体电池故障误报,提出一种基于余弦相似度的故障值计算方法以提高算法诊断精确度,并实现故障电池的在线自动检测和定位。云端车辆数据验证结果表明,所提出的基于卡尔曼滤波和特征指数化的电池故障诊断算法能够有效地检测故障电池并进行预警。

A battery fault online diagnosis method based on Kalman filtering and feature indexing was proposed for the faults such as thermal runaway and internal short circuit in the battery pack.Firstly,data noise reduction was performed based on historical data and Kalman filtering method to effectively remove voltage anomalies and a feature indexing method was proposed to extract and amplify the voltage characteristics between battery pack cells.Finally,a fault value calculation method based on cosine similarity was proposed in order to reduce the false alarm due to battery pack inconsistency and to automatically detect and locate the faulty battery online.Verification in cloud-based vehicle data shows that the proposed battery fault diagnosis algorithm based on Kalman filtering and feature indexing can effectively detect faulty batteries and provide early warning.