Due to the voltage plateau, the state of charge (SOC) estimation based on voltage signals remains challenging for LiFePO4 batteries. Electrochemical impedance spectroscopy (EIS) provides detailed information on the complex electrochemical processes within the battery, facilitating a more comprehensive understanding of the battery's state, thereby improving the accuracy of SOC estimation. Five types of features are extracted from publicly available EIS data, employing light gradient boosting machine ensemble learning techniques for SOC estimation of LiFePO4 batteries. A comprehensive performance comparison is conducted, including features constructed from raw EIS data, equivalent circuit model parameters, real and imaginary parts of impedance, and their combinations. Results indicate that utilizing the imaginary part of impedance achieves excellent SOC estimation performance for LiFePO4 batteries, with the average root mean square error of different batteries within 2.35%. Additionally, considering the frequency range of EIS data (0.01~1000Hz), the imaginary part of impedance shows promise for online acquisition, providing a new approach for data-driven battery management.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Impedance-Based State of Charge Estimation of LiFePO4 Batteries Using Ensemble Learning


    Beteiligte:
    Qian, Guangjun (Autor:in) / Han, Xuebing (Autor:in) / Zheng, Yuejiu (Autor:in) / Sun, Yuedong (Autor:in)


    Erscheinungsdatum :

    10.10.2024


    Format / Umfang :

    369526 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch