State of health assessment of electric vehicle battery pack by using signal tracking regularized box particle filter

Abstract In this study, a novel regularized box particle filter (RBPF) is proposed for real-time state of health (SoH) estimation of the lithium-ion battery pack of electrical vehicles (EVs). RBPF uses the boxes instead of particles used in the conventional sequential importance re-sampling particle filter (SIR-PF). The boxes cover vast sample space as compared to the particles. Therefore, a very small number of boxes are required to cover the entire sample space as compared to SIR-PF. The next step is to remove particle impoverishment and degeneracy which is addressed by the regularization step in re-sampling of RBPF. Initially, a hardware setup is used for assessment of internal resistance of the battery which is used in establishment of linear regression model (LRM) of the battery. Initially, performance of the individual batteries in a battery pack is evaluated. Later on, algorithms are tested on different series-parallel combinations of batteries to evaluate the impact of different arrangements and numbers of batteries on the SoH estimation of proposed algorithm. The performance indices of RBPF are compared with SIR-PF and BPF algorithms in terms of root mean square error (RMSE), peak error and computational time for SoH assessment of the lithium-ion battery pack. RBPF shows better accuracy and computational time as compared to SIR-PF and box particle filter (BPF). Further, a detailed study is carried out for optimization of the size of boxes and the number of boxes for SoH estimation of batteries.