Capacity Estimation for Lithium-Ion Batteries Based on Impedance Spectral Dynamics and Deep Gaussian Process

The charging and discharging processes of lithium-ion batteries are accompanied by complex electrochemical behaviors that have long-lasting and irreversible effects on the performance and lifetime of the batteries, accurate online capacity estimation of the batteries has been challenging. Considering that electrochemical impedance spectra contain rich frequency domain information, which can fully reflect the dynamics of the battery, a new capacity estimation method based on impedance spectra dynamics is proposed. First, based on the prior knowledge of battery dynamics, the relationship between the Nyquist curve of electrochemical impedance spectra and the parameters of the equivalent circuit model is analyzed, and several characteristic frequency points roughly linearly correlated with the capacity degradation are extracted. Then, a deep Gaussian process regression model with autoregressive recurrent structure is designed to implement capacity estimation, the hyperparameters are recognized using sequential model-based optimization. Finally, the proposed estimator is validated on battery experimental data under different temperature conditions, the results show that the proposed method has good estimation accuracy and general applicability.