Electrochemical impedance spectroscopy (EIS) is a power technique in battery research that includes kinetics characteristic analysis, degradation mechanism, and state estimation and prediction. The impedance model is usually employed to analyze battery EIS quantitatively, and parameter identification for the impedance model is of vital importance. However, the current parameter identification for the impedance model is mainly achieved through heuristic optimization and nonlinear least squares methods, which may be time consuming when used to process a large amount of EIS data at once with low convergence speed. To deal with these issues, this study proposes an innovative parameter identification method for the battery impedance model using the Neural Network approach. First, the relationship between cyclic aging and model parameter variation was investigated. To enhance the networks' adaptiveness and effectiveness, 32000 synthetic EISs were generated to assist the offline training process. Then four Neural Network models with EIS data within different frequency ranges as input were employed to identify the model parameters. The experimental results indicate that the proposed method can achieve a relative mean absolute error of 2.30% compared with software RelaxIS, and a mean error impedance 3.64E–4 Ohm compared to experimental EIS.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    A deep learning-based parameter identification approach of electrochemical impedance spectrum equivalent model for lithium-ion batteries


    Contributors:
    Wang, Jiangwei (author) / Jiang, Bo (author) / Ou, Yuxin (author) / Wang, Xueyuan (author) / Wei, Xuezhe (author) / Dai, Haifeng (author)


    Publication date :

    2024-10-10


    Size :

    492720 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English