基于多尺度数据融合的锂电池健康状态评估

Lithium battery health evaluation based on multi-scale data fusion

针对锂电池健康状态评估问题,提出一种以人工神经网络为核心多尺度数据融合框架下的锂电池健康状态评估方法.选取内阻、充电电压样本熵和等压降放电时间作为典型特征参数,建立3层分布式人工神经网络对特征参数进行多尺度融合计算,以计算拟合输出结果作为评估健康状态的参考值,并通过美国国家航空航天局(national aeronautics and space administration,NASA)试验数据集进行验证.结果表明:提出的评估方法能够基于锂电池充放电测量数据和解算特征参数,利用多尺度数据融合框架迅速迭代收敛,完成锂电池健康状态评估拟合;该评估方法的计算结果与测试平台实测值相比,平均误差小于3%,评估性能衰退趋势与实际劣化趋势一致.

A method of lithium battery health assessment was proposed based on the framework of multi-scale data fusion with artificial neural network as the core.The internal constant resistance,the sample entropy of charging voltage and the isobaric discharge time were selected as typical characteristic parameters.The three-layer distributed artificial neural network was established for the multi-scale data fusion,and the calculated fitting output was used as reference value for the health assessment.The proposed method was verified by the national aeronautics and space administration(NASA)experimental datasheet.The results show that the method based on the lithium battery typical characteristic parameters and the multi-scale data fusion framework can rapidly iteratively converge to complete the evaluation and fitting of the health status of lithium battery.Comparing the calculation results of the proposed method with the test platform data,the average error is less than 3%,and the evaluation performance degradation trend is consistent with the actual deterioration trend.