Research on the Prediction Method of Power Battery SOC Based on Deep Learning

Precise prediction of power battery SOC plays a key role in the accurate navigation of electric vehicles that are running on the way. Since traditional models adopt methods of manual intervention to extract characteristic factors that affect the battery, they fail to find the correlations between these factors, which leads to the unsatisfactory prediction precision of SOC. This article proposes a power battery SOC prediction model based on DBN-SVM through analyzing the nonlinear characteristics of power battery in an innovative way. By using deep belief networks, this model automatically learns high nonlinear combination characteristics of the power battery from basic data including the physical parameters of the power battery and the environmental factors. Afterwards, this model improves the precision via classifier. Test results show that the prediction model based on deep learning achieves a higher precision, which approves to be a reliable prediction method of battery power SOC.