A fast convergence strategy based on gray wolf optimization algorithm for co-estimation of battery state of charge and capacity

Abstract

Under the same initial noise covariance, the extended Kalman filter (EKF) algorithm and the adaptive extended Kalman filter (AEKF) algorithm show different convergence rates in the battery state estimation experiments. It can be concluded from the experimental results that the convergence rate of the EKF algorithm is usually faster than that of the AEKF algorithm, but the estimation accuracy of the AEKF algorithm is usually better than that of the EKF algorithm. In response to this issue, this article proposes a fast convergence strategy based on the grey Wolf optimization (GWO) algorithm for the co-estimation of battery SOC and capacity. Based on battery parameter identification, the proposed algorithm utilizes the GWO to optimize the initial noise covariance of the EKF algorithm. Then, the EKF algorithm with optimized initial noise covariance is used to quickly pull the SOC estimation results into a stable region and switch to the AEKF algorithm to jointly estimate the SOC and capacity of the battery. Through data validation under different working conditions, it is shown that the proposed target algorithm has a much faster convergence ability than the comparison algorithm, and the proposed algorithm also exhibits excellent robustness under different initialization errors and temperatures.