LAUSR

Accurate modeling is of great importance for describing battery dynamics under different working conditions. To achieve high accuracy and obtain insights into battery operations, it is critical to bridge the time-domain equivalent circuit model and frequency-domain electrochemical impedance spectroscopy. However, the practical battery running data are recorded at a very low sampling frequency, it is thus challenging to identify the frequency-domain impedance spectrum using time-domain response data. To solve this problem, this article proposes a systematic battery time-domain modeling approach considering impedance spectrum identification. First, a probabilistic approach for global sensitivity analysis is proposed to evaluate the impact of different parameters on model performance. The parameter sensitivity of both integer-order and fractional-order models are evaluated using the Sobol indices obtained by Monte Carlo simulations. Then, a Bayesian optimization algorithm is proposed to identify model parameters to reduce the evaluation times of the battery model. Finally, the experimental data conducted on LiFePO$_{4}$ cells are employed to verify the proposed method. The results indicate that the total Sobol indices show a miscellaneous sensitivity of parameters on impedance spectra, which consolidates the understanding of impedance characteristics, and the higher order integer-order models with well-initialized low sensitive parameters are more suitable for predicting frequency-domain characteristics.