State‐of‐health prediction for lithium‐ion batteries via electrochemical impedance spectroscopy and artificial neural networks

Abstract A detailed and in‐depth prediction of the state‐of‐health of lithium ion batteries (LIB) remains a major challenge. Meanwhile, the dynamic changes in the thermal and electrochemical characteristics of the interphases are important for the determination of battery health. Herein, we performed electrochemical impedance spectroscopy (EIS) measurements and equivalent circuit analysis on Panasonic NCR 18650B batteries under different states of charge (SOC), overcharge, and overdischarge cycling conditions. Three indicators of the comprehensive state of health (CSOH) of the batteries were summarized based on the values of the resistances obtained from equivalent circuit analysis from the EIS measurements. CSOH represents the dynamic electrochemical characteristics of the LIBs. By evaluating the CSOH indicators, we have developed an artificial neural network (ANN) model which can provide an effective prediction of the future CSOH of the LIBs. Percent error estimations have been done by implementing the Tanh activation function, and the estimation results of predicted equivalent series resistance, charge‐transfer resistance, and solid‐electrolyte interphase resistance are 5%, 1.5%, and 1%, respectively. The ANN model based on EIS analysis is a straightforward and effective approach that can provide novel routes for CSOH prediction of LIBs.