Sensitivity analysis of Kalman filter based estimation methods of lithium-ion batteries under non-stationary conditions

Abstract

This paper addresses the sensitivity analysis of an extended and dual Kalman filters used for state estimation (state of charge (SoC), inner resistance) of lithium-ion batteries mainly for traction applications. The Kalman filters in different configurations are now widely applied in modern battery management systems. It is well known that the quality of estimation is dependent from the accuracy of model used. In this case a compromise between model’s complexity and quality of estimation is one of main critical factors while developing a Kalman filter based estimation solution. In case of state estimation of lithium-ion batteries this point becomes important due to the following reasons: transformation of chemical effects to electrical model causes the loss of model accuracy; model parameters are not constant due to ageing and temperature rice; parameters of derived discrete time models are sensitive to measurement noise. In this paper authors analyse the influence of varying inner parameters of a lithium-ion battery and the choice of time discrete model onto accuracy of SoC estimation. An extended and dual discrete Kalman filter using zero-order hold method and Tustin transform are tested using new European driving cycle.