This article presents an approach to estimate state-of-health (SoH) of the Li-ion batteries using parts of the frequency data from the whole electrochemical impedance spectroscopy (EIS) dataset. Two commercialized 18650 cylindrical cells have been cycled, and both the cycling data and the EIS data have been collected after each cycle. A long-short-term memory (LSTM) based model has been developed to train with different impedance data subsets at different cycle numbers for extracting sequential features. Later, those features are fed to two fully connected layers to achieve the estimated SoH value for a particular cell and cell condition. This approach will eliminate the need for specialized devices and sophisticated control algorithms required in both extremely high and low frequency EIS measurement for SoH estimation purposes.
