The demand for energy storage is increasing massively due to the electrification of transport and the expansion of renewable energies. Current battery technologies cannot satisfy this growing demand as they are difficult to recycle, as the necessary raw materials are mined under precarious conditions, and as the energy density is insufficient. Metal–air batteries offer a high energy density as there is only one active mass inside the cell and the cathodic reaction uses the ambient air. Various metals can be used, but zinc is very promising due to its disposability and non-toxic behavior, and as operation as a secondary cell is possible. Typical characteristics of zinc–air batteries are flat charge and discharge curves. On the one hand, this is an advantage for the subsequent power electronics, which can be optimized for smaller and constant voltage ranges. On the other hand, the state determination of the system becomes more complex, as the voltage level is not sufficient to determine the state of the battery. In this context, electrochemical impedance spectroscopy is a promising candidate as the resulting impedance spectra depend on the state of charge, working point, state of aging, and temperature. Previous approaches require a fixed operating state of the cell while impedance measurements are being performed. In this publication, electrochemical impedance spectroscopy is therefore combined with various machine learning techniques to also determine successfully the state of charge during charging of the cell at non-fixed charging currents.

This work presents a data-driven approach for state-of-charge estimation of rechargeable zinc-air batteries based on electrochemical impedance spectroscopy. Due to the nonlinear electrochemical behavior and flat discharge voltage profile of zinc-air batteries, accurate state-of-charge estimation remains challenging. The investigated cells employ a three-electrode configuration with a dedicated gas diffusion electrode for discharge and a separate electrode for charging. This work focuses exclusively on discharge operation, as the two current paths involve physically distinct electrodes with fundamentally different impedance characteristics. High-dimensional impedance spectra are combined with physically interpretable features derived from a simplified equivalent circuit model and compressed via principal component analysis. A long short-term memory network models the relationship between the resulting feature representation and state-of-charge, with Bayesian hyperparameter tuning applied to optimize architecture and training configuration. Performance is compared against baseline models including multilayer perceptrons. The model is trained on multiple battery cells and evaluated on a completely held-out cell to assess cross-cell generalization. The results show that principal component analysis compression of the combined impedance spectrum and equivalent circuit feature vector is the decisive optimization step, achieving a mean absolute error of 1.04% on an unseen test cell. In contrast, the choice of model architecture has a smaller impact on performance.