Robust battery state-of-charge estimation with improved convergence rate based on applying Busse's adaptive rule to extended Kalman filters

The extended Kalman filter (EKF) has been widely used to estimate the state-of-charge (SoC) of batteries over the past decade. Battery SoC estimation with the EKF is initialized without knowing the true value of the SoC. Thus, it requires a fast convergence rate to provide users with an accurate SoC value in the shortest time. Applying an adaptive rule into the EKF is an unfussy way to improve both the accuracy and convergence rate of SoC estimation. However, an adaptive rule requires additional calculations and consumes additional memory space to store the learning history. This paper applies Busse's adaptive rule to improve the accuracy and convergence rate of EKF battery SoC estimation. Experimental data from a lithium titanate battery is applied to examine the battery SoC estimation with EKF, covariance-matching adaptive EKF (CM-AEKF), and Busse's adaptive EKF (Busse-AEKF) algorithms. The findings showed that the Busse-AEKF method has the shortest convergence time with an accuracy that is comparable to that of the CM-AEKF method. After the SoC value is converged, the algorithm gives estimation accuracy of a 1.42% root-mean-square error (RMSE) and a 3.15% of maximum error. In addition, Busse's AEKF does not require a large memory space to operate. Thus, it is a promising solution for battery SoC estimation.