Uneven Usage Battery State of Health Estimation via Fractional-Order Equivalent Circuit Model and AutoML Fusion

Abstract

To accurately predict the State of Health (SOH) of lithium-ion batteries under the continuously changing charging and discharging conditions in practical applications, this study proposes a hybrid modeling approach that integrates a Fractional Order Equivalent Circuit Model (F-ECM) with the AutoGluon automatic machine learning framework. By leveraging Electrochemical Impedance Spectroscopy (EIS) to capture battery frequency response characteristics, F-ECM accurately fits EIS data to extract detailed internal state parameters. The integration of AutoGluon automates the machine learning process, enhancing the precision of SOH predictions. Through testing and analysis on real battery datasets, this method has demonstrated superior prediction precision and computational efficiency compared to existing mainstream modeling approaches. Specifically, the hybrid method achieved a Root Mean Square Error (RMSE) of 2.12% and a Mean Absolute Error (MAE) of 1.67%. This study presents a highly accurate, interpretable, and adaptable predictive framework for lithium-ion battery health assessment, offering valuable insights for battery health management system development.