Model based state-of-energy estimation for LiFePO4 batteries using unscented particle filter

Abstract

Lithium-ion batteries have been treated as the most efficient energy storage candidates in electric vehicles. However, range anxiety hinders its popularity in daily transportation. When compared with the extensively discussed state-of-charge, state-of-energy (SoE) is a more appropriate indicator of remaining driving mileage by considering the voltage degradation in the discharging process. In this paper, an equivalent circuit model (ECM) based SoE estimator is proposed where a pseudo power definition is utilized to exclude the internal loss influence, and an unscented particle filter is employed to address system nonlinearities and noises. Additionally, to accommodate battery time-variant characteristics, the ECM parameters and variables are decoupled in complex-frequency domain, upon which the recursive least square algorithm is utilized for the on-line identification of parameters. Finally, a series of validation experiments on a LiFePO4 battery verify that the proposed methodology can track the real SoE closely with an error of less than 1.8%.