Lithium Ion Batteries Impedance Approximation and Model Parameterization Using Artificial Neural Networks

The accurate determination of the condition of a battery is crucial for an optimized operation in the context of several loading and ambient conditions. A tool to achieve this goal is the electrochemical impedance spectroscopy (EIS), which is a non-destructive measuring method to analyze the internal processes in a cell and results in a complex valued impedance as a function of the frequency. The measurements require high efforts for the experimental setup and the measurement time. This paper presents a method to approximate the EIS with an artificial neural network (ANN) and an equivalent circuit model (ECM). Only four frequencies are used to estimate a whole spectrum. The lowest total distance between the measured and the estimated points, which is used as evaluation criterion, is at a value of 28. 10
^-5
. This as well as the visual results show the accuracy of the method and the potential to dramatically reduce the measurement time.