Improvement of The Battery State of Charge Estimation Using Recursive ‎ Least Square Based Adaptive Extended Kalman Filter ‎

Abstract

Battery Management System (BMS) including measurements errors that causes decrease in ‎the quality of ‎calculated State of the Charge (SOC). It will limit the accurate estimation of ‎the SOC that is a critical challenge in ‎some of the engineering fields such as medical science, ‎robotics, navigation and industrial applications. These ‎facts implies on the significance of ‎SOC estimation from battery measurements that is the matter of the literature ‎through the ‎recent years. Due to the dependency of the EKF to the system model, the change in the ‎battery ‎parameters and noise information cause losing performance in the SOC estimation ‎over the time. In this paper, we ‎assume that the battery parameters including internal ‎resistance and capacitor and also the noise information are ‎varying over the time. To solve ‎that, two separate on-line identification algorithms for parameters and noise ‎information are ‎introduced. In more details, a Recursive Least Square (RLS) algorithm is used to identify ‎the ‎resistance and capacitor values. Moreover, the process and measurement noise covariance are ‎estimated based ‎on iterative noise information identification algorithm. Then all of the ‎updated values are used in the EKF ‎algorithm. This paper aims to address the issue of uncertainty in SOC estimation by proposing two algorithms. ‎The first algorithm focuses on identifying deterministic uncertainty, which refers to uncertainty in model ‎parameters. To address the challenge of uncertain model parameters, RLS is introduced.

Keywords

• battery charge
• state of ‎the charge (soc)
• ‎kalman filter
• estimation