Modeling and health feature extraction method for lithium-ion batteries state of health estimation by distribution of relaxation times

Abstract

Accurate state of health (SOH) estimation of lithium batteries (LIBs) depends on the accuracy of LIBs modeling and the correlation of extracted health features (HFs). Aiming at the problems of a priori assumptions for existing battery modeling and poor correlation of HFs, this paper proposes a SOH estimation method for LIBs based on distribution of relaxation times (DRT). Firstly, we deduce and establish the developed DRT-based impedance model from battery electrochemical impedance spectroscopy (EIS), and the model parameters are calculated by EIS split-frequency domain method based on the principle of parsimony. Then, we analyze the variation rules of the peak and valley information in the distribution function of relaxation times curve with the capacity decay, and we extract multiple features strongly correlated with SOH as HFs. The extracted HFs are subjected to Weighted Principal Component Analysis (WPCA) to obtain indirect health features (IHFs), which achieve the estimation of SOH for LIBs. Finally, the effectiveness of the proposed modeling and health feature extraction method is validated by SOH estimation using the public dataset from Nature subseries. The results indicated that the EIS fitting accuracy R2 of the developed model is the best and the RMSE of SOH estimation is within 0.873 %.