Machine learning approach for full impedance spectrum study of Li-ion battery

Electrochemical Impedance Spectroscopy (EIS) has been widely applied for Li-ion battery research because EIS can reflect the physical characteristics. The full impedance spectrum sweep generally takes several minutes. Thus, it is impossible to implement a full spectrum sweep for real-time investigations. In this paper, machine learning approach is proposed to address the issue. The proposed approach is based on multi-sine signal sweep technique, where the impedances at corresponding frequencies are derived with a fast Fourier transform. The full impedance spectrum is obtained via machine learning approach. The results are compared with three alternative techniques namely, the piecewise cubic Hermite interpolation polynomial, modified Akima piecewise cubic Hermite interpolation, and Spline. The results demonstrate that the proposed machine learning approach has the best performance.