为了提高锂电池健康状态(State of health,SOH)的预测准确率,该文将支持向量回归(Support vector regression,SVR)算法与集成学习理论相结合,提出一种基于局部信息融合的支持向量回归集成(Local information fusion with ensemble support vector regression,LIF-ESVR)算法.该算法的核心思想是利用数据的局部信息融合替代原有全局信息,并将信息层融合问题转化为决策层融合问题.首先将原始的训练集划分为若干个子训练集,每个子训练集都包含了原始训练集中的部分重要信息;然后,在每个子训练集上训练一个对应的SVR模型;最后,利用集成学习算法将已训练好的多个SVR模型进行融合.在美国国家航空航天局蓄电池数据上的实验结果表明,所提方法的性能优于现有的锂电池SOH预测方法,具有广泛的应用价值.
Prediction for state of health of lithium-ion batteries by local information fusion with ensemble support vector regression
To improve the prediction accuracy of state of health(SOH)for lithium-ion batteries,this paper developes a local information fusion with ensemble support vector regression(LIF-ESVR)method,which is implemented by combining support vector regression(SVR)algorithm with ensemble learning theory. The basic idea of LIF-ESVR is to use local information fusion to replace the global information and switch the information fusion problem to the decision fusion problem.Firstly the orig-inal training dataset is divided into multiple subsets,each of which contains the important local infor-mation;then,for each subset,the corresponding SVR is trained on it;finally,the ensemble learning technology is adopted to incorporate multiple trained SVRs. The experimental results on batteries datasets of the National Aeronautics and Space Administration(NASA)of USA have demonstrated that the LIF-ESVR outperforms the existing methods for predicting lithium-ion batteries SOH and can be used practically and extensively.
lithium-ion batteriesstate of healthsupport vector regressionensemble learninginformation fusion
国家科技期刊平台
NSTL
万方数据
维普
