Online estimation of lithium-ion battery equivalent circuit model parameters and state of charge using time-domain assisted decoupled recursive least squares technique

摘要

To ensure the battery management system (BMS) operates effectively, it is quite important to accurately determine the model parameters of the equivalent circuit model (ECM) and the state of charge (SOC). Decoupled recursive least squares (DRLS) technique which separately identifies the parameters of the battery's fast and slow dynamic response impedance can improve the real-time estimation accuracy, and time-domain parameter extraction (TDPE) methods can obtain model parameters from experimental voltage data without optimizing the parameters iteratively; hence it is simple and time-effective. To have both the advantages of fast identification speed of TDPE and high accuracy of DRLS at the same time, this study proposes a parameter identification technique that combines the two methods mentioned above. Integrating the TDPE approach and DRLS technique can effectively simplify the complexity of the DRLS estimation method and improve its estimation accuracy. In addition, a direct current resistance (DCR) compensation term considering fast and slow dynamics parameters is also proposed to further improve the identification accuracy of DCR, which further enhances the SOC estimation precision. Compared with the original DRLS technique, the mean absolute percentage error (MAPE) of the modelling error obtained by the proposed method can be improved by 78.7 /73.5 under the federal urban driving schedule (FUDS) and the dynamic stress test (DST) test patterns for simulated results, and enhanced by 9.0 /5.3 under FUDS and DST test profiles for experimental results. In addition, the MAPE of the SOC error is reduced by 72.3 /7.1 under FUDS and DST test patterns for simulated results, and lowered by 18.8 /8.3 under FUDS and DST test profiles for experimental results. Those results validate the effectiveness and cor-rectness of the proposed method.