Application of Kalman prediction algorithm combined with SVM in monitoring states of VRLA battery

It is difficult to obtain the accurate system state model of a valve-regulated lead acid (VRLA) battery with performance degradation. In order to solve this problem, firstly, using the equivalent circuit model of a VRLA battery and linear dynamic state-space mode, a non-linear mode well suited for the deteriorative battery is deduced. Furthermore, based on the deduced non-liner mode, a Kalman prediction algorithm combined with support vector machine (SVM) method (SVM-KF) is proposed. In the proposed approach, SVM is employed to iterative correct information error during Kalman prediction, so the prediction algorithm is provided with correction ability while a battery is in the degradation. All the obtained results show that the proposed algorithm can accurately predict the remaining capability of the battery and identify the nonlinear deterioration tendency of the battery.