Estimation of Battery State of Charge Based on Electrochemical Impedance Spectroscopy and Machine Learning

Abstract

With the increasing application of lithium-ion batteries in energy storage (ES) and electric vehicles (EV), the operation state and the safety early warning are particularly important. Due to the complexity of lithium-ion batteries, predicting the state of charge (SOC) of batteries is still a great challenge. In this paper, an electrochemical impedance spectroscopy (EIS) based on the SOC estimation method was proposed. The EIS of different SOC under 25℃ was measured, and the variation characteristics of EIS were summarized from the aspects of amplitude and phase by using the full band information. Finally, the linear system and BP neural network model were constructed by selecting the characteristic parameters at a specific frequency. The results demonstrate that the prediction effect of the BP neural network is better, and the error is found to be less than 3%.