Predicting batteries second-life state-of-health with first-life data and on-board voltage measurements using support vector regression

Electric vehicle (EV) batteries experience significant degradation during their primary use. While reaching End-of-Life (EOL) for EVs, these batteries hold the potential for a “Second-life” in less demanding applications. However, accurate estimation for State-of-Health (SoH) remains a challenging t...

Full description

Saved in:

Bibliographic Details
Published in:	Journal of energy storage Vol. 104; p. 114554
Main Authors:	Jameel, Shymaa Mohammed, Altmemi, J.M., Oglah, Ahmed A., Abbas, Mohammad A., Sabry, Ahmad H.
Format:	Journal Article
Language:	English
Published:	Elsevier Ltd 20-12-2024
Subjects:	Battery second-life Data prediction Lithium-ion Remaining useful life (RUL) State-of-health (SoH) Support vector regression (SVR) Data prediction Battery second-life Remaining useful life (RUL) Support vector regression (SVR) State-of-health (SoH) Lithium-ion
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Electric vehicle (EV) batteries experience significant degradation during their primary use. While reaching End-of-Life (EOL) for EVs, these batteries hold the potential for a “Second-life” in less demanding applications. However, accurate estimation for State-of-Health (SoH) remains a challenging task as it requires extensive monitoring communications in Second-life settings. This study proposes a novel data-efficient approach to predicting Second-life SoH with minimal Second-life measurements and readily available first-life data. This work introduces a Support Vector Regression (SVR) model trained on first-life features to estimate discharge capacity in the Second-life. Only terminal voltage measurements (TIECVD and TIEDVD) during Second-life operation are utilized to predict SoH. Unlike existing methods involving broad Second-life monitoring, this approach focuses on energy delivery as an indicator of the battery's ability to power continuous operation, reducing complexity and data acquisition costs. To validate the proposed technique, we conducted experiments using Lithium-ion batteries with NASA's dataset including three different battery models. The results of using the SVR model achieved a Root Mean Square Error (RMSE) between actual and predicted SoH data ranging from 0.0012 to 0.0158, signifying its effectiveness over various battery types. This innovative SoH prediction method using first-life data and minimal Second-life measurements clears the way for better predicting the Remaining Useful Life (RUL) in Second-life EV batteries. •1st-life data predicts 2nd-life SoH (less monitoring).•Estimates 2nd-life SoH from voltage data.•Leverages 1st-life discharge capacity features.•SVR model predicts 2nd-life capacity (1st-life trained).•Low RMSE on NASA Li-ion battery dataset.
ISSN:	2352-152X
DOI:	10.1016/j.est.2024.114554