车用锂离子电池剩余使用寿命的混沌时序非线性组合预测模型研究  

Study on Chaotic Sequence Nonlinear Combination Prediction Model of RUL for Vehicle Lithium-Ion Batteries

作　　者：徐东辉 徐新爱[1] 刘海峰 郑萍[1] Xu Donghui;Xu Xin’ai;Liu Haifeng;Zheng Ping(Nanchang Normal University,Nanchang 330032;The Multi-Function to Multiply Vehicle Research Center of Jiangxi Province,Jiangling Motors Company Limited,Nanchang 330052)

机构地区：[1]南昌师范学院,南昌330032 [2]江铃汽车股份有限公司,江西省多功能乘用车工程研究中心,南昌330052

出　　处：《汽车技术》2021年第10期30-36,共7页Automobile Technology

基　　金：国家自然科学基金项目(51176014);江西省科技支撑计划项目(20151BBE50108);江西省科技厅科技项目(20202BBEL53019);江西省重点研发计划项目(20192BBHL80002);江西省教育厅科学技术研究项目(GJJ191129,GJJ202609,GJJ202610);江西省教学改革一般项目(JXJG-18-23-15)。

摘　　要：针对用单一方法难以准确预测锂离子电池剩余使用寿命(RUL)的问题,提出了非线性组合预测模型。对电池系统进行混沌判断,得到与实际容量相关性高的间接参数,利用重构后的数据对Elman网络及非线性自回归(NARX)网络进行训练和预测,得到表征电池性能退化的特征量;然后,用最小二乘支持向量机进行非线性组合,获得RUL的预测值。试验结果表明:非线性组合预测模型的预测精度优于Elman和NARX,具有更强的非线性预测能力。Aiming at the problem that it is difficult to accurately predict the Remaining Useful Life(RUL)of lithiumion batteries by a single method,a nonlinear combined prediction model is proposed.By means of the chaotic judgment of the battery system,the indirect parameters which are high correlative with actual capacity are obtained.Moreover,the Elman network and Nonlinear AutoRegressive with eXogenous input(NARX)network are trained and predicted with the reconstructed data so that the characteristic quantities that represented the battery performance degradation are acquired.Then,the Least Square Support Vector Machine(LS-SVM)is used for nonlinear combination to gain the predicted value of RUL.The experimental results show that the prediction accuracy of the nonlinear combination prediction model is better than that of Elman and NARX.In addition,it has stronger nonlinear prediction ability.

关 键 词：锂离子电池 非线性组合 寿命预测 混沌时序 

分 类 号：TM912.9[电气工程—电力电子与电力传动]