State of Health Estimation for Lithium-Ion Batteries Based on Transition Frequency’s Impedance and Other Impedance Features with Correlation Analysis
Mohammad K. Al‐Smadi, Jaber A. Abu Qahouq
Abstract
This paper presents data-driven impedance-based state of health (SOH) estimation for commercial lithium-ion batteries across an SOH range of ~96% to ~60%. Battery health indicators at the transition frequency of the battery impedance Nyquist plot are utilized to develop an SOH estimator based on an artificial neural network (ANN). In addition, two more ANN-based SOH estimators utilizing some impedance magnitude and phase values are developed. Spearman correlation analysis is utilized to identify the frequency points at which the impedance magnitude and phase values show strong correlations with SOH values and are thus utilized as SOH indicators. The performance evaluation of the developed SOH estimators shows that the maximum root mean square error (RMSE) is equal to 1.39%, the maximum mean absolute error (MAE) is equal to 1.25%, the maximum mean absolute percentage error (MAPE) is equal to 1.55%, and the minimum coefficient of determination (R2) is equal to 0.983.