Battery health and remaining useful life prediction using lightweight machine learning models
Author Identifier (ORCID)
Abstract
Battery degradation has become a crucial challenge to environmental safety and energy management systems due to the limited lifespan of lithium-ion batteries and their improper handling. This study aims to develop lightweight machine learning models to estimate the remaining useful life of lithium-ion batteries that align with SDG 7 and 12 to support responsible battery usage. Five regression models Gradient Boosting, XGBoost, Random Forest, Decision Tree, and Polynomial Regression were assessed for their predictive performance. Using the training dataset from Kaggle titled Battery Remaining Useful Life (RUL) on 2022, the Gradient Boosting model demonstrates the optimal predictive capability with cross-validation performance showing an R-squared value of 0.994853 ± 0.001025 and a Mean Absolute Error (MAE) score of 11.320829 ± 0.185441, maintaining proper performance hierarchy. Therefore, this study contributes to improving better battery management through the deploy model, which positively reduces environmental risks.
Keywords
lithium-ion batteries, computational modeling, battery management systems, predictive models
Document Type
Conference Proceeding
Date of Publication
12-4-2025
Publication Title
2025 International Conference on Informatics, Multimedia, Cyber and Information System (ICIMCIS)
Publisher
IEEE Explorer
School
School of Business and Law
RAS ID
88791
ISBN
9798331579302
Copyright
subscription content
Recommended Citation
Wime, D. S., Wijaya, J. S., Santoso Gunawan, A. A., Setiawan, K. E., Moniaga, J., & Jie, F. (2025). Battery health and remaining useful life prediction using lightweight machine learning models. In 2025 International Conference on Informatics, Multimedia, Cyber and Information System (ICIMCIS) (pp. 925-930). https://doi.org/10.1109/ICIMCIS68501.2025.11327063.