SoC Estimation of Lithium Battery Based on Improved BP Neural Network

Abstract

Lithium iron phosphate battery as the research object, in view of the traditional battery state of charge (SoC) estimate methodological shortcomings and deficiencies, combined with battery charge and discharge characteristics. This paper presents an improved BP neural network algorithm which can significantly reduce the SoC prediction error of the lithium iron phosphate battery. Through a lot of experiments, the improved training function is put forward in this paper. Comparing the SoC value of the lithium battery and the actual SoC value estimated by the neural network, the effectiveness of the algorithm is proved. Research results show that the consumption of the algorithm to the real-time measurement of voltage, current and temperature value and the internal resistance value as input to learn lithium battery power remaining, with estimated speed, wide application range and prediction error is small, the engineering easy to achieve, can effectively solve the prediction problem of the state of charge (SoC) of the lithium batteries for electric vehicles.