Predicting the degradation of li-ion battery using advanced machine learning techniques

Yi Ru Li, Kuan Jung Chung

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper, advanced machine learning techniques called ensemble learning process (ELP) are proposed to predict the performance degradation of Li-ion batteries. The ELP is to use a variety of machine learning, Artificial Intelligence, and statistical modeling techniques and approaches to achieve the best results. Prediction quality and accuracy are quantified by the Pearson Correlation and Root-Mean-Square-Error (RMSE) respectively. A commercial software named ThingWorx Analytics (PTC) was applied to the study of predicting the capacity loss of Li-ion batteries using data from a dual dynamic stress accelerated degradation test (D2SADT). The results show that the proposed algorithms are able to precisely predicting the capacity loss of batteries in terms of a low root-mean-square error (RMSE = 0.01) and a high Pearson Correlation (0.99) for one step ahead prediction. Furthermore, the ensemble learning process confirms the good performance in the two steps ahead (long-term) prediction of degradation resulting with the only little increase of RMSE (from 0.01 to 0.02) and maintaining the same Pearson Correlation (0.99).

Original languageEnglish
Title of host publicationIMPACT 2017 - 12th International Microsystems, Packaging, Assembly and Circuits Technology Conference, Proceedings
PublisherIEEE Computer Society
Pages258-262
Number of pages5
ISBN (Electronic)9781538647196
DOIs
Publication statusPublished - 2017 Jul 1
Event12th International Microsystems, Packaging, Assembly and Circuits Technology Conference, IMPACT 2017 - Taipei, Taiwan
Duration: 2017 Oct 252017 Oct 27

Publication series

NameProceedings of Technical Papers - International Microsystems, Packaging, Assembly, and Circuits Technology Conference, IMPACT
Volume2017-October
ISSN (Print)2150-5934
ISSN (Electronic)2150-5942

Other

Other12th International Microsystems, Packaging, Assembly and Circuits Technology Conference, IMPACT 2017
CountryTaiwan
CityTaipei
Period17-10-2517-10-27

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture
  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Predicting the degradation of li-ion battery using advanced machine learning techniques'. Together they form a unique fingerprint.

Cite this