Li-ion batteries equivalent circuit model parameter estimation from electrochemical impedance spectroscopy measurements using neural networks

Although electrochemical impedance spectroscopy (EIS) is a powerful tool for understanding the internal processes in batteries, it is still not widely utilized in practical implementations. One of the reasons why that is the case is due to complexity of the equivalent circuit model (ECM) and time required for its analysis with the conventional approach, and lack of publicly available datasets for the machine learning approach. In this paper we suggest a new approach of estimating ECM parameters from EIS measurements using neural networks. For training of neural networks we propose using a completely synthetic dataset, bypassing the issue of the lack of the publicly available EIS datasets for Li-Ion batteries and only validating the trained model on experimentally measured dataset.