基于高阶扩展卡尔曼滤波的锂电池剩余使用寿命自适应预测  

Adaptive prediction of remaining useful life for lithium battery based on high-order extended Kalman filter

作　　者：余伟 周云刚 朱文博 黎海兵 张忠波 YU Wei;ZHOU Yungang;ZHU Wenbo;LI Haibing;ZHANG Zhongbo(School of Mechatronic Engineering and Automation,Foshan University,Foshan 528225,China)

机构地区：[1]佛山大学机电工程与自动化学院,广东佛山528225

出　　处：《佛山科学技术学院学报(自然科学版)》2025年第1期19-26,共8页Journal of Foshan University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61803086,62106048);广东省基础与应用基础研究基金资助项目(2022A1515110187,2022A1515140042)。

摘　　要：锂电池的剩余使用寿命(Remaining Useful Life,RUL)预测对于电池健康管理系统的预防性维护至关重要,基于随机过程的退化模型在此领域中发挥了关键作用,特别是采用双隐含状态非线性维纳过程,能够更加灵活具体地描述锂电池退化趋势。在锂电池RUL自适应预测中,退化模型参数在线更新较多采用扩展卡尔曼滤波(Extended Kalman Filter,EKF)方法,它是近似精度只有一阶的非线性高斯滤波器,对强非线性系统滤波精度偏低。基于锂电池双隐含状态非线性维纳退化模型,设计高阶扩展卡尔曼滤波器(High-order Extended Kalman Filter,HEKF),利用高阶项信息减小截断误差,得到最优在线参数估计值,推导出RUL的概率密度函数,实现锂电池RUL自适应预测。通过NASA的锂电池退化数据进行实例验证,相较其他两种方法,所建立方法RUL预测结果的RMSE值、MAE值分别降低53.3%、61.6%和65.7%、66.1%。Remaining Useful Life(RUL)prediction of lithium batteries is crucial for preventive maintenance of battery health management systems,and degradation models based on stochastic processes play a key role in this field,especially the use of two-implied-state nonlinear Wiener process,which can describe the degradation trend of lithium batteries more flexibly and specifically.In the lithium batteries RUL adaptive prediction,the degradation model parameters are often updated online using the Extended Kalman Filter(EKF)method,which is a nonlinear Gaussian filter with an approximate accuracy of only one order,and the filtering accuracy is low for strongly nonlinear systems.This paper is based on the nonlinear Wiener degradation model of lithium batteries with double hidden states.It designs a High-order Extended Kalman Filter(HEKF)to reduce truncation error by utilizing information from high-order terms and obtains optimal online parameter estimates.It further derives the probability density function of RUL and achieves adaptive prediction of RUL for lithium batteries.Finally,the example validation by NASA’s lithium battery degradation data shows that the accuracy of this paper’s method is improved by 83.9%and 53.3%,respectively,compared with the prediction results of the other two methods.

关 键 词：锂电池 剩余使用寿命 高阶扩展卡尔曼滤波 非线性维纳过程 双隐含状态 

分 类 号：U462.3[机械工程—车辆工程]