Multi-time scale identification of key kinetic processes for lithium-ion batteries considering variable characteristic frequency

The electrification of vehicles puts forward higher requirements for the power management efficiency of integrated battery management systems as the primary or sole energy supply. In this paper, an efficient adaptive multi-time scale identification strategy is proposed to achieve high-fidelity modeling of complex kinetic processes inside the battery. More specifically, a second-order equivalent circuit model network considering variable characteristic frequency is constructed based on the high-frequency, medium-high-frequency, and low-frequency characteristics of the key kinetic processes. Then, two coupled sub-filters are developed based on forgetting factor recursive least squares and extended Kalman filtering methods and decoupled by the corresponding time-scale information. The coupled iterative calculation of the two sub-filter modules at different time scales is realized by the voltage response of the kinetic diffusion process. In addition, the driver of the low-frequency subalgorithm with the state of charge variation amount as the kernel is designed to realize the adaptive identification of the kinetic diffusion process parameters. Finally, the concept of dynamical parameter entropy is introduced and advocated to verify the physical meaning of the kinetic parameters. The experimental results under three operating conditions show that the mean absolute error and root-mean-square error metrics of the proposed strategy for voltage tracking can be limited to 13 and 16 mV, respectively. Additionally, from the entropy calculation results, the proposed method can reduce the dispersion of parameter identification results by a maximum of 40.72% and 70.05%, respectively, compared with the traditional fixed characteristic frequency algorithms. The proposed method paves the way for the subsequent development of adaptive state estimators and efficient embedded applications.