Deep Learning-Driven SOH Assessment of Batteries from EIS Data Using Markov Transition Fields

The assessment of the State of Health (SOH) of batteries is crucial for optimizing their performance, ensuring safety, and extending their lifespan in various applications. Electrochemical Impedance Spectroscopy (EIS) has been widely used as a diagnostic tool for monitoring battery behavior, but its complex and high-dimensional data present challenges for traditional analysis methods. In this paper, we propose a novel approach for SOH estimation by leveraging Markov Transition Fields (MTF) to transform EIS data into image-like representations, which are then classified using deep learning techniques. Specifically, we evaluate the performance of 5 pre-trained Convolutional Neural Networks (CNNs) in classifying the MTF-transformed EIS data to estimate the SOH of batteries. To assess the robustness and generalization capability of the proposed approach, we incorporate two distinct battery datasets and evaluate the performance across different networks. Our results demonstrate that combining MTF with pre-trained CNNs offers a promising solution for accurate and efficient SOH assessment, providing valuable insights into the potential of deep learning-based methods for battery health monitoring. This work contributes to the growing body of research on using advanced machine learning techniques for battery diagnostics and highlights the advantages of using image transformation methods like MTF for enhanced classification performance.