State-of-Charge Estimation of VRLA Batteries Using Neural Networks and Extended Kalman Filter

In this paper, the state of charge (SOC) estimation of Valve-Regulated Lead-Acid (VRLA) batteries using Neural Networks (NNs) and Extended Kalman Filter (EKF) is considered. The NN, which is of Radial Function Basis (RBF) type is trained off-line, providing the necessary model for EKF. Proper selection of inputs to the NN can help it to provide better modeling of the battery. The inputs to this NN are the terminal voltage and current, and SOC of the battery at the present sampling time. The output is the predicted terminal voltage at the next sampling time. In order to estimate the SOC during the charging and discharging processes, two NNs are trained, one for the charging process and the other one for the discharging process. This is mainly due to the fact that the behavior of batteries is different during charging and discharging. The estimated states in EKF are the SOC and the terminal voltage of the battery. One of the main advantages of the proposed method is that there is no need to know the initial values of the SOC. The proposed estimator can converge to the actual SOC in relatively short time. Experimental results show good estimation of the SOC of VRLA batteries.