Lithium-ion battery remaining useful life estimation based on ensemble learning with ls-svm algorithm

Yu Peng, Siyuan Lu, Wei Xie, Datong Liu, Haitao Liao

Research output: Chapter in Book/Report/Conference proceedingChapter

5 Scopus citations

Abstract

This chapter proposes the integrated use of parameter-selection-based ensemble learning (PSBEL) and least square support vector machine (LS-SVM) algorithm for lithium-ion battery remaining useful life (RUL) estimation. Technically, given a set of monitoring parameters, some groups of parameters are randomly selected to construct LS-SVM submodels. On the basis of these submodels, ensemble learning is utilized to achieve a final result, which overcomes the difficulty of accurately determining the model parameters and significantly improves the precision and stability of RUL estimation. The proposed approach provides practitioners with confidence intervals and probability distributions of RUL estimates for uncertainty management. The validity and applicability of the PSBEL with LS-SVM for RUL estimation are demonstrated using capacity-changing data of lithium-ion batteries during discharging cycles. To analyze the impact of different submodels on the precision of final RUL prediction and to improve the uncertainty representation, more advanced ensemble learning techniques should be considered.

Original languageEnglish (US)
Title of host publicationAdvances in Battery Manufacturing, Services, and Management Systems
PublisherWiley-IEEE Press
Pages217-232
Number of pages16
ISBN (Electronic)9781119060741
ISBN (Print)9781119056492
DOIs
StatePublished - Oct 3 2016

Keywords

  • Data-driven approaches
  • Discharging cycles
  • Least square support vector machine algorithm
  • Lithium-ion battery remaining useful life estimation
  • Parameter-selection-based ensemble learning
  • Uncertainty management

ASJC Scopus subject areas

  • General Engineering

Fingerprint

Dive into the research topics of 'Lithium-ion battery remaining useful life estimation based on ensemble learning with ls-svm algorithm'. Together they form a unique fingerprint.

Cite this