State of charge estimation for the Vanadium Redox Flow Battery based on Extended Kalman filter using modified parameter identification

Vanadium Redox Flow Battery (VRFB) is widely utilized in energy storage due to its excellent characteristics. Credible knowledge of the state of charge (SOC) is a pre-condition for the effective health management of batteries. The SOC estimation depends on the second-order RC equivalent circuit model (ECM) parameters identified by the forgetting factor recursive least squares (FF-RLS). Considering the time-varying characteristics of the model parameters, an adaptive forgetting factor recursive least squares (AFF-RLS) based on the gradient descent method is proposed to identify the EMC parameters. The forgetting factor can be obtained based on the error between the terminal voltage measurement and terminal voltage estimation. The proposed joint estimator has been verified by performing charge and discharge experiments for VRFB single cell. The mean error, maximum estimated error and root mean square error of SOC under 6 A pulse discharging current are 1.38 x 10(-3), 3.76 x 10(-5), and 1.38 x 10(-5). The result indicates that AFF-RLS is robust against outliers of model parameters and improve the anti-interference of EKF. In addition, this paper finds that the different value of learning rate can affect the sensitive and anti-interference of EKF.