Reduction of high fidelity lithium-ion battery model via data-driven system identification

The battery management system of a hybrid electric vehicle requires a computationally simple yet accurate model of the battery. In this paper a reduced order battery model is developed using a stochastic top-down approach. Firstly a pseudo2D, multi-particle electrochemical model, considered as a surrogate for the real system, is used to obtain the observational data. Then the model structure is inferred directly from the data. The dependencies between the states and the model parameters are analysed, which results in a 5th order piecewise state dependent parameter model which can describe the nonlinear relationship between the current, the voltage and the state of charge of the battery.