基于机器学习的电动汽车电池系统的风险预警

Risk early warning of electric vehicle battery system based on machine learning

为提高动力电池在实车工况下安全预警的及时性和准确性,将电池系统安全预警问题提炼为关键状态预测及基于预测状态的预警分类2大科学问题,根据实车运行中的电池状态数据,选择电池的单体电压最高值、单体电压极差等作为关键预测对象;利用费舍尔计分和最大信息系数(MIC)进行特征选择,采用样本卷积和交互网络模型(SCINet)实现关键状态预测;基于预测的状态,建立多分类随机森林(RF)模型,对动力电池的安全风险进行分级预警。研究结果表明:该模型对电池多个参数具有很强的预测能力,如预测1 min后单体电压最高值的均方根误差(RMSE)为0.0271,温度最高值为0.0540;对电池系统1 min后安全风险等级预测的查准率为84%,宏平均f 1分数为74%。

In order to improve the timeliness and accuracy of safety early warning of power battery under real vehicle conditions,the safe early warning of the battery system was refined into two scientific problems:key state prediction and early warning classification based on predicted state.According to the battery state data in the real vehicle operation,the maximum value of single cell voltage and the range of the cell voltage were selected as the key prediction objects.Fisher scoring and Maximum Information Coefficient(MIC)were used to realize key feature selection,and Sample Convolution and Interaction Network model(SCINet)were used for key state prediction.Then,based on the predicted state,a multi-classification RF model was established to classify the safety risks of power batteries.The results show that the proposed model has a strong predictive ability for multiple parametrs of the battery.For example,the root mean square error(RMSE)of the highest cell voltage is 0.0271 and the highest temperature is 0.0540 after 1 min of prediction.The prediction accuracy of the safety risk level of the battery system after 1 min is 84%,and the macro-average f 1 score is 74%.