Kalman Filter Based on SVM Innovation Update for Predicting State-of Health of VRLA Batteries

Abstract

This paper describes the dynamic model of valve regulated lead-acid (VRLA) battery and presents application of state-estimation techniques for online prediction of the state-of-health (SOH) and state-of-charge(SOC) of VRLA batteries. Specifically, approach of Kalman filter (KF) based on the innovation obtained through support vector machine (SVM) is presented. In the proposed approach, factors related with the capacity of a VRLA battery have been taken into account as elements of the multi-dimensional input vector while time series is trained by SVM to predict future innovation. Then step by step prognosis is implemented through Kalman predictor with innovation obtained through SVM. Measurements using real-time data are used to verify the capability of the proposed method for estimating SOH and SOC of VRLA batteries. Experiment results show that our proposed methodologies are suitable for monitoring the SOC and SOH of the battery, with accuracy in determining the SOH and its trend of deterioration within 4%. Moreover, by accounting for the nonlinearities present within the dynamic cell models, the application of KF prediction based on SVM innovation update is shown to provide verifiable indications of SoH.