State-of-charge estimation using data fusion for vanadium redox flow battery

Abstract

Accurate estimation of the state of charge (SOC) is important for preventing overcharge and overdischarge of vanadium redox flow batteries (VRFBs). In this paper, we propose a data-fusion (DF) method to improve the accuracy of SOC estimation for VRFBs by combining two different SOC estimation results. First, the extended Kalman filter (EKF) and adaptive extended Kalman filter (AEKF) were used to estimate the SOC according to the equivalent circuit model (ECM) of the VRFB. Thereafter, the SOC estimation results of the EKF and AEKF were utilized separately as the basic estimation of the DF method to obtain two different DF results, which can be explicitly divided into DF-EKF and DF-AEKF. The ECM parameters were extracted using the recursive least squares (RLS) algorithm in real time. Subsequently, an experiment involving a hybrid pulse discharge was conducted to validate the proposed method. Here, the root-mean-square error (RMSE) and mean absolute error (MAE) are used as statistical methods for outcome evaluation. Compared with the RMSE and MAE (0.0012, 0.0010, and 0.0010 and 0.0009) of the EKF and AEKF, the RMSE and MAE of the DF-EKF decreased by 33.3%, 20%, and 36.3% and 22.2% to 0.0008 and 0.0007, respectively. The same downward trend also occurs in DF-AEKF. The RMSE of the results of DF-AEKF decreased by 33.3% and 20% to 0.0008, and the MAE decreased by 45.5% and 33.3% to 0.0006, respectively. The results show that the proposed DF method exhibited a high fidelity and accuracy in estimating the SOC of the VRFB.