Fast and reliable calibration of thermal-physical model of lithium-ion battery: a sensitivity-based method

Physical simulation of lithium-ion battery is crucial to consolidate the understanding of its operating mechanisms and, potentially, its state of health; nevertheless, a reliable model calibration is complex due to the large number of physical parameters involved. Here, a thorough sensitivity analysis is performed on the simulation of discharge, relaxation and impedance spectroscopy tests, to highlight the response of the Doyle-Fuller-Newman model output, implemented with a thermal model to compute heat transfer effects, to a variation of 28 model parameters as a function of similar to 160 combinations of temperature, battery state of charge and C-rate. The analysis highlights how up to 14 parameters can be regarded as insensitive, reasonably excludable from model calibra-tion, while other parameters show a strongly miscellaneous response, possible to be maximized adopting specific conditions. Therefore, an innovative method is proposed by experimentally exploiting two temperature levels and combining the three techniques, demonstrated to be highly complementary for a fast and reliable model calibration. As a case study, it is applied on a commercial battery sample, enabling a repeatable and physically sound calibration of the model parameters, as successfully demonstrated over a set of full discharges in 12 combinations of temperatures and C-rate. The comparison with a standard discharge-based calibration process highlights the strength of the proposed protocol.

Automated summary

A comprehensive sensitivity analysis is performed on the physical simulation of a lithium-ion battery, revealing the response of the Doyle-Fuller-Newman model to variations in 28 model parameters under different temperature, state of charge, and C-rate conditions. The analysis shows that up to 14 parameters can be considered insensitive and excluded from model calibration, while other parameters exhibit a strongly diverse response that can be maximized by adopting specific conditions. An innovative method is proposed, which experimentally exploits two temperature levels and combines three techniques for a fast and reliable model calibration. The proposed protocol is successfully applied to a commercial battery sample, enabling repeatable and physically sound calibration of the model parameters.