Skip to Main Content
Comprehension of open-circuitvoltage (OCV) function and battery impedance can be used to evaluate the residual energy and dynamic voltage response of lithium batteries. However, noise contamination during data sampling may affect the accuracy of the derived OCV function and battery impedance extraction, resulting in a poor evaluation performance. By aiming at formulating a more representative circuit-based model, this study proposes a hybrid averaging method along with an impedance network observation approach that is embedded within a time-frequency localisation method, anticipating eliminating the influences of sampling noise such that the characterisation of the OCV function and battery impedances can be better ensured. In such an integrated approach, the trimmed relevance vector regression is also added, which is followed by two cascaded R-C networks in order to fit the characteristics of battery impedances. Through the method proposed in this paper, it is found that the average error can be decreased to be less than 20 mV. Both battery residual energy and I-V performance are more effectively comprehended.