Non-Destructive Parameterization of Lithium-Ion Batteries via Machine Learning with Simulated EIS Data

Lithium-ion batteries are ubiquitous in modern energy storage applications, necessitating efficient methods for assessing their state and performance. This study explores a non-destructive approach to extract vital battery parameters using machine learning techniques applied to simulated Electrochemical Impedance Spectroscopy (EIS) data. EIS is a powerful diagnostic tool for batteries and provides a safe and repeatable alternative to the physical intrusion of battery dismantling, which could alter the batteries properties. The research focuses on the design and training of machine learning models for accurate prediction of battery parameters within the widely used P2D model. By leveraging the power of machine learning, this approach aims to accurately characterize the battery parameters using an electrochemical model as opposed to the less accurate equivalent circuit models, contributing to the reliability and longevity of lithium-ion batteries in diverse applications. The second part of this paper incorporates real-life experimental EIS data by utilizing an improved version of an open-source model called “Impedance Analyzer”. Multiple approaches have been explored and discussed to leverage machine learning algorithms to accurately estimate the battery parameters. The findings of this study pave the way for more robust, non-destructive battery assessment methods, crucial for advanced state of health prediction models of lithium-ion batteries.