Machine Learning and AI-Based Approaches for Battery State-of-Health Prediction in Energy Storage Systems: A Comprehensive Review

Document Type : Review Article

Author

Department of chemistry, Faculty of science , Qom university , Qom , Iran

10.22091/jaem.2026.16369.1050

Abstract

This review aims to provide a critical and structured analysis of artificial intelligence (AI)-based approaches for lithium-ion battery health prediction and energy storage management, with a particular focus on their role in enabling decision-centric battery management systems (BMS). A systematic evaluation of 54 recent studies is conducted, covering machine learning (ML), deep learning (DL), and physics-informed hybrid models.
The results demonstrate that deep learning and hybrid approaches consistently outperform traditional model-based methods, with reported improvements of up to 30–50% in prediction accuracy (e.g., reduced RMSE) across benchmark datasets such as NASA and Oxford battery datasets. However, these performance gains are often accompanied by increased computational complexity, data dependency, and reduced interpretability, which significantly limit their real-world deployment in embedded systems.
This review introduces a decision-centric framework that shifts the focus from prediction accuracy to actionable intelligence, emphasizing the integration of prognostic outputs into real-time control, safety management, and lifecycle optimization. The analysis reveals a critical gap between high-performing laboratory models and their practical implementation in industrial battery management systems.
It is concluded that future research should prioritize lightweight, interpretable, and uncertainty-aware AI models, alongside standardized datasets and benchmarking protocols, to enable scalable and reliable deployment in next-generation energy storage infrastructures.

Keywords

Main Subjects

Advances in Energy and Materials Research

Articles in Press, Accepted Manuscript
Available Online from 25 August 2026

Files

Share

How to cite

Statistics