Parameter Decoupling and Sensitivity of Different Discretisation Modes for Lithium-Ion Battery Models

Abstract

The online parameter estimation of a lithium-ion battery model is a key part of the state estimation of a lithium-ion battery management system (BMS). To improve the speed of the identification algorithm and the processing of abnormal data, a BMS must discretise the data. In this study, as shown in graphical abstract, the effects of backward difference and bilinear transformation on the accuracy and stability of identification parameters were studied comprehensively. The stability and sensitivity of the numerical model were analysed via discrete sampling step size, current sensor accuracy and voltage sensor accuracy. Simultaneously, three types of data under low dynamic, high dynamic and real complex conditions were used to evaluate the influences of different factors on the model parameters. The adaptive forgetting factor recursive least squares method was used for parameter estimation. Finally, a meaningful conclusion was drawn by comparing the simulation results with the experimental verification. For a given cell model, the influences of sampling step size, current and voltage accuracy on the parameter estimation of backward difference and bilinear transformation cannot be disregarded. This study reveals the influences of measurement uncertainty on model parameter estimation and strategies for reducing algorithm sensitivity, providing scientific guidance for subsequent engineering research.

Keywords: Backward difference, Bilinear transformation, AFFRLS, Measurement uncertainty, Algorithm sensitivity