Frequency-Based Parameterization of Semi-Empirical Models for State-of-Health Estimation in Lithium-Ion Batteries

The accurate estimation of state-of-health (SoH) in lithium-ion batteries is crucial for enabling safe and improved battery usage. Yet the complexity of electrochemical aging processes within the cell and their dependence on differing operating conditions make this task challenging. This paper proposes a novel SoH estimation method using equivalent circuit model (ECM) parameters identified based on electrochemical impedance spectroscopy and regression analysis. Six different ECM architectures are calibrated and compared based on quantitative metrics, including root mean square error, Akaike information criterion, and Bayesian information criterion. Based on this analysis, the second-order constant phase element model is chosen, and its parameters are used to perform three supervised tree-based regressions and one supervised linear regression. The random forest regression model is found to be the most accurate in estimating capacity-based SoH, and a rigorous correlation analysis is conducted to examine the relationship between model parameters and SoH. This study utilizes data collected from an aging campaign of 22 5Ah nickel-manganese-cobalt lithium-ion cells and demonstrates the effectiveness of ECMs and regression models for SoH estimation.