Study of EKF and VPA for parameter estimation in a custom-designed battery system

The aim of this work is to investigate onboard parameter identification methods for battery equivalent circuit models (ECM). The ECM is often used as a basis of battery state detection algorithms to calculate state-of-charge (SOC), state-of-health (SOH) and state-of-function (SOF, maximal available power). Especially the latter requires that the parameters of the ECM describe at every moment precisely the battery impedance. This is not a trivial task, as the battery impedances change with SOC, temperature and over the battery lifetime due to aging. Therefore, the parameters of the ECM have to be updated continuously. In this work, starting from different vehicle speed profiles, current profiles are generated by means of a vehicle model. Current profiles thus generated are fed into custom-designed battery systems (composed by one to three time constants) to get correspondent voltage profiles. Extended Kalman Filter (EKF) and Varied-Parameters Approach (VPA) are employed, for on-line estimation of impedance parameters of the battery ECM, having as input the current and voltage profiles previously obtained. Those approaches are successful in fitting an ECM to a lithium-ion cell dataset to within a maximum absolute error of 0.5 % on the battery output voltage.