基于GRA-BP神经网络的锂电池剩余容量估计方法  

Estimation method of lithium battery residual capacity based on GRA-BP neural network

作　　者：韦雨亭 刘欣伟 孙金磊[1] 景含笑 温珂镌 WEI Yuting;LIU Xinwei;SUN Jinlei;JING Hanxiao;WEN Kejuan(School of Automation,Nanjing University of Science and Technology,Nanjing Jiangsu 210094,China)

机构地区：[1]南京理工大学自动化学院,江苏南京210094

出　　处：《电源技术》2023年第10期1308-1312,共5页Chinese Journal of Power Sources

基　　金：国家自然科学基金项目(52007085)。

摘　　要：针对锂电池老化过程中特征不明显、对容量波动点追踪不准确、模型长期使用后精度下降等问题,提出了一种基于灰色关联分析(GRA)-反向传播(BP)神经网络的锂电池剩余容量估计方法。通过GRA筛选出能够表征电池老化的特征量,利用计算机辅助寿命周期工程中心(CALCE)公开的锂电池充放电数据集训练BP神经网络模型,并实现电池剩余容量估计。结果表明,对于同一电池,训练集占80%时,容量衰减的估计误差为2.28%,在训练集仅占20%的情况下,估计误差为5.99%。A method based on the GRA-BP neural network for estimating the residual capacity of lithium batteries was proposed to address the problems of inconspicuous features in the aging process of lithium batteries,inaccurate tracking of capacity fluctuation points,and degradation of the accuracy of the model after long-term use.After filtering the feature vectors that can characterize the battery aging through grey relation analysis(GRA),the back propagation(BP)neural network model was trained by using the battery charging and discharging dataset published by the center for advanced life cycle engineering(CALCE)to estimate the battery residual capacity.The results show that the estimation error of capacity decay is 2.28%for the same battery with 80%of the training set and 5.99%with only 20%of the training set.

关 键 词：锂电池 老化特征量 BP神经网络 电池剩余容量 

分 类 号：TM912[电气工程—电力电子与电力传动]