Performance Analysis of Computational Intelligence Techniques for Lithium-Ion Battery Remaining Useful Life Prediction

Authors:
Sara Toumi, S. Rubin Bose, Ajay Kumar Karri, Gui Yinxuan, Watahiki Hayato, Heider Al Mashalah

Addresses:
Department of Hydrocarbon Production, Kasdi Merbah University, Ouargla, Ouargla Province, Algeria. School of Computer Science and Engineering, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India. Department of Mathematics, Texas A&M University, College Station, Texas, United States of America. Department of Computing and Information Systems, Singapore Management University, 81 Victoria Street, Singapore. Department of Information and Communication Engineering, The University of Tokyo, Bunkyo Ward, Tokyo, Japan. Department of Business, McMaster University, Hamilton, Ontario, Canada.

Abstract:

Lithium-ion batteries are of critical importance for electric vehicles, consumer electronics and renewable energy storage. However, their performance deteriorates over time, so predicting the Remaining Useful Life (RUL) is important for safety, reliability and cost-effectiveness. Traditional physics-based and empirical Remaining Useful Life (RUL) estimation models face challenges related to complexity, limited generalisability and high computational requirements. To address these limitations, this study explores a data-driven machine learning approach to predict battery RUL. The proposed work compares three models, Linear Regression, Decision Tree Regressor and Support Vector Machine (SVM), on a simulated battery degradation dataset. Researchers assess each model using standard performance measures, including MAE (mean absolute error), RMSE (root mean square error), and the coefficient of determination (R2). The results show that both Linear Regression and Decision Tree Regressor perform well in terms of predictive accuracy, with R2 scores above 0.95 and low MAE and RMSE scores demonstrating their ability to model degradation trends. Although the SVM model is inaccurate, its consistent predictions suggest hyperparameter optimisation could improve it. Visualising degradation patterns and model predictions with lightweight ML frameworks makes battery prognostics interpretable. Generalising models to real-world datasets and accounting for variations in battery usage remain challenges. This study establishes a solid platform for incorporating machine learning into modern Battery Management Systems (BMS) and enables data-driven maintenance plans. 

Keywords: Remaining Useful Life (RUL); Lithium-Ion Batteries; Machine Learning (ML); Linear Regression; Decision Tree Regressor; Support Vector Machine (SVM); Data-Driven Maintenance.

Received on: 20/10/2024, Revised on: 29/12/2024, Accepted on: 02/03/2025, Published on: 05/10/2026

DOI: 10.69888/FTSES.2026.000795

FMDB Transactions on Sustainable Energy Sequence, 2026 Vol. 4 No. 2, Pages: 84-96

  • Views : 47
  • Downloads : 6
Download PDF