基于XGBoost预测和ARIMA残差校正的锂离子电池SOC估算  

Estimation of State of Charge of Lithium-ion Battery Based on XGBoost Prediction and ARIMA Residual Correction

作　　者：费陈 宋树祥[1] 夏海英[1] 冀聪聪 陆利芬 FEI Chen;SONG Shu-xiang;XIA Hai-ying;JI Cong-cong;LU Li-fen(School of Electronic Engineering»Guangxi Normal University,Guilin 541004,China;Guilin Zhishen Information Technology Co Ltd,Guilin 541004,China)

机构地区：[1]广西师范大学电子工程学院,桂林541004 [2]桂林智神信息技术有限公司,桂林541004

出　　处：《武汉理工大学学报》2019年第10期94-101,共8页Journal of Wuhan University of Technology

基　　金：广西科技重大专项项目(桂科AA18118009)。

摘　　要：为提升锂离子电池性能和延长锂离子电池寿命,提出一种基于XGBoost预测和ARIMA(Autoregressive Integrated Moving Average Model)残差校正的锂离子电池电荷状态估算方法。该估算方法首先采用集成学习XGBoost回归算法对锂电子电池SOC进行预测,然后基于预测残差建立ARIMA模型对预测值进行修正,进一步提升SOC估算的准确性,并与线性回归、支持向量机、神经网络、K-邻近算法等4种经典回归预测算法进行了实验对比分析。结果表明,文中方法通过引入ARIMA残差校正,将XGBoost的预测精度在绝对误差,均方根误差和均方根百分比误差3个技术指标上提高了15%~20%,且优于其他5种回归预测算法。In order to improve the performance of lithium-ion battery and extend the life of lithium-ion battery,we proposed a method of SOC estimation of lithium-ion battery based on XGBoost prediction and ARIMA residual correction.This method first predicted the SOC of lithium-ion battery by using XGBoost algorithm.Then,the predicted values were corrected based on the residuals by using ARIMA model to further improve the SOC estimation accuracy.Finally,we compared the method with four classical regression algorithms,including linear regression,SVM,neutral network and KNN.The experimental results demonstrate that our method increases by 15%to 20%in terms of absolute error,RMSE(root mean squared error)and RMSPE(root mean square percent error)by introducing ARIMA residual correction,and outperforms the other five comparators in the three indexes.

关 键 词：锂离子电池SOC XGBoost算法 ARIMA模型 估算 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]