Multi-state joint estimation of series battery pack based on multi-model fusion

Aiming at the problem of on-line real-time estimation of the state of charge (SOC) and capacity of the battery pack, a new mean-difference model (MDM) for series battery pack is proposed. The model realizes the fusion of data-driven Autoregression (AR) model and equivalent circuit model (ECM) to improve the dynamic characteristics of the battery pack model under complex operating conditions. For the model parameters of the new MDM, the differentiated updating strategy is proposed. The differentiated updating strategy proposed can timely update the new MDM parameters online according to their changing characteristics and changing rate. Based on the new MDM, new mean-difference space equations expressing the coupling relationship between battery pack SOC and capacity, as well as a multi-timescale estimation framework based on mean state / difference state changing speed, are proposed. Finally, under this multi-timescale estimation framework, the multi-timescale H infinity filter (Mts-HIF) is used to realize the joint estimation of SOC and capacity of series battery packs. Based on 3 operating conditions, we have verified the proposed new MDM under noiseless and colored noise environment and compared it with the MDM based on n-order RC-ECM (nRC-ECM). The Akaike information criterion (AIC) indicates that the 3-order AR mean model (AR-MM) has the optimal tradeoff between accuracy and complexity under noiseless environment, while the 4-order AR-MM has the optimal tradeoff under colored noise environment and its root mean square error (RMSE) is less than 14.5 mV. Compared with the multi-timescale extended Kalman filter (Mts-EKF) joint estimation algorithm, the Mts-HIF proposed has better robustness under colored noise environment. For the capacity estimation of each cell of battery pack, the RMSE of the MtsHIF does not exceed 2%, while the RMSE of the Mts-EKF exceeds 4%. For the SOC estimation, the RMSE of the Mts-HIF does not exceed 1.2%, while the RMSE of the Mts-EKF is close to 3%.

Automated summary

A new mean-difference model (MDM) is proposed for online real-time estimation of the state of charge (SOC) and capacity of the battery pack. The model combines data-driven Autoregression (AR) model and equivalent circuit model (ECM) to improve the dynamic characteristics of the battery pack model. A differentiated updating strategy is proposed to update the model parameters online based on their changing characteristics and changing rate. Based on the new MDM, new mean-difference space equations and a multi-timescale estimation framework are proposed for the coupling relationship between battery pack SOC and capacity. The multi-timescale H infinity filter (Mts-HIF) is used for joint estimation of SOC and capacity. The proposed MDM is verified under different conditions and compared with another model, showing its effectiveness and accuracy. The evaluation of MDM and estimation algorithm gives a rating of 8 out of 10.