Global sensitivity analysis towards non-invasive parameterization of the electrochemical-thermal model for lithium-ion batteries

Abstract

High-fidelity electrochemical-thermal models are essential for performance improvement, charge/discharge strategy optimization, and the safe operation of lithium-ion batteries. However, model performance significantly relies on the accuracy of parameters, whose measurement is limited by laboratory conditions. Non-invasive methods based on relatively accessible current, voltage, and temperature data combined with artificial intelligence are promising for rapid parameterization of battery models. However, the model’s complexity and the data’s poor quality increase the difficulty of applying the methodology. To design a reasonable identification framework and obtain reliable data, the identifiability of model parameters must be analyzed under different operating conditions. This paper develops an identifiability analysis framework to investigate the impact of model parameters on voltage and temperature outputs and the impact of key operating variables, i.e., current rate and ambient temperature. By adjusting operating conditions, the sensitivity of specific parameters can be improved by two orders of magnitude. The results are discussed in detail concerning the model modeling mechanism and the physical meaning of the parameters, with a focus on improving non-invasive parameterization in terms of experimental design and identification strategy.