State of Health Estimation of Lithium Batteries through Vision Transformer Analysis of Electrochemical Impedance Spectroscopy

The rapid advancement and widespread deployment of energy storage systems, particularly lithium-ion batteries, underscore the necessity for precise and reliable state of health (SOH) estimation techniques. This paper develops a novel approach to battery SOH estimation by leveraging Vision Transformers (ViTs), a class of transformer models excelling in handling visual data, to analyze Electrochemical Impedance Spectroscopy (EIS) data. Unlike traditional methods that rely heavily on handcrafted features or classical impedance models, our methodology treats EIS Nyquist plots as images and utilizes ViTs for direct SOH classification. We use a comprehensive real dataset comprising thousands of Nyquist plots, labeled with their corresponding SOH and augmented through various colormap transformations to enhance ViT robustness. Our experiments demonstrate that ViTs can capture subtle, yet critical, impedance characteristics indicative of battery health, achieving superior classification accuracy compared to conventional machine learning techniques. The broader impact of this research lies in its potential to enhance the reliability and longevity of battery systems across various applications, from consumer electronics to electric vehicles and grid storage, ultimately contributing to more sustainable and efficient energy management. © 2024 Computers and Industrial Engineering. All rights reserved.