Artificial neural network in estimation of battery state of-charge (SOC) with nonconventional input variables selected by correlation analysis

Abstract

The selection of input variables is important to improve the prediction accuracy of artificial neural networks (ANNs). A three-layer feedforward backpropagation ANN is presented to estimate and predict the battery state-of-charge with nonconventional input variables selected. Initially, a few candidate input variables are derived from three basic input variables: discharging current, discharging time and battery terminal voltage. Then, three techniques of correlation analysis - the linear correlation analysis, nonparametric correlation analysis and partial correlation analysis - are used to select the input variables, and the results obtained are compared. With several nonconventional input variables included in the input sets, high prediction accuracy of the ANN model is obtained.