A Kalman Filter SOC Estimation Method for Lithium-ion Batteries Based on Discrete Wavelet Transform Denoising

An extended Kalman filter (EKF) algorithm based on discrete wavelet transform (DWT) denoising is proposed for the problem of the lower accuracy and larger fluctuation of SOC (state of charge) estimation caused by the noise of the discharging/charging voltage (DCV) signals of lithium-ion batteries. The DCV signal with noise is decomposed through the multi-resolution analysis (MRA). Effects of four hard-thresholding-based denoising rules on reducing the noise of DCV signal are compared, and the hard-thresholding-based denoising rule based on Stein's unbiased risk estimation is selected to adjust wavelet coefficients. The parameters of the battery model are identified by the recursive least square method with an adaptive forgetting factor and then the SOC is estimated using EKF. Simulation results show that the selected denoising rule effectively reduces the noise of the DCV signal. The proposed algorithm effectively improves the accuracy of SOC estimation with a strong robustness and the estimation error is less than 3%.