A Robust Extended Kalman Filtering Approach for State of Charge Estimation in Batteries

State of charge (SOC) estimation is a crucial challenge faced by battery management systems. Extended Kalman filter (EKF)-based approaches have been widely explored in literature for SOC estimation. The challenge with the EKF approach to SOC estimation is that the parameters of the underlying state-space model (SSM) are not perfectly known. Such uncertainty in the SSM may arise at both the model order and the model parameter levels. The SSM of the SOC estimation problem at the most reduced model order form is defined using close to ten parameters, all of which suffer from uncertainties. A disadvantage of the EKF approach is that when the SSM parameters deviate from reality, the filter starts to produce incorrect SOC estimates unbeknown to the user. This article presents a novel EKF approach to SOC estimation that is robust against model parameter uncertainties. The proposed robust EKF employs additional states to absorb the uncertainties in the model parameters and employs two metrics, both computed based on filter innovations, to detect model order uncertainty. The proposed approach is demonstrated using a battery simulator.