基于IGA-MRVR的锂离子电池剩余使用寿命预测  

Prediction of Lithium-ion Battery's Remaining Useful Life Based on IGA-RVR Algorithm

作　　者：高栋[1] 黄妙华[1] GAO Dong HUANG Miao-hua(Hubei Key Laboratory of Advanced Technology for Automotive Components ,School of Automotive Engineering,Wuhan University of Technology,Wuhan 430070,Chin)

机构地区：[1]武汉理工大学汽车工程学院现代汽车零部件技术湖北省重点实验室,武汉430070

出　　处：《自动化与仪表》2017年第7期10-15,共6页Automation & Instrumentation

基　　金：国家科技支撑计划项目(2015BAG08B02)

摘　　要：锂离子电池是电动汽车的关键部分,科学准确地预测其剩余使用寿命RUL是决定电动汽车性能的重要因素。该文针对传统的单核相关向量回归RVR算法存在预测精度低且泛化能力弱的问题,提出了基于多项式核和径向基核函数的多核MRVR;应用改进遗传算法IGA对多核核参数进行联合寻优;使用美国国家航空航天局NASA电池数据进行验证。验证结果表明,改进遗传算法优化的MRVR算法有更好的预测精度和泛化性。The lithium-ion battery is a key part of electric vehicles（EVs）. Scientific and accurate prediction of re- maining useful life（RUL） of lithium-ion batteries is an important factor to determine the performance of EVs. Aiming at the problem that the traditional single kernel relevance vector regression（RVR） has low accuracy and poor generalization ability,a novel multi-kernel RVR（MRVR） based on polynomial kernel and radial basis kernel function is proposed. Moreover, the improved genetic algorithm （IGA） is used to search the kernel parameters of the MRVR model. The NASA battery data is used to verify the model. The results show that the improved MRVR algorithm has better prediction accuracy and better generalization.

关 键 词：电动汽车 锂电池 剩余使用寿命 多核相关向量回归算法 改进遗传算法优化 预测 

分 类 号：U469.722[机械工程—车辆工程] TM912.9[交通运输工程—载运工具运用工程]