Physics-based electrochemical modelling of Li-ion batteries

Physics-based battery models with the Doyle-Fuller-Newman (DFN) model have been regarding as the very powerful model to simulate the Li-ion batteries behaviour with good accuracy. However, the high computational demand and parameterization are the two main challenges have been the main obstacle for the model have limited it use in applications. In this thesis, a comparison of various simplifications of the DFN model has been conducted to compared to improve the simulation speed while maintaining good accuracy, the thesis also propose a guideline on how the select the optimum model simplification. Furthermore, in this thesis, the multiple-particle DFN model is introduced, which incorporates particle size distributions for improving simulation accuracy while maintaining low computational demand. A full parameterization framework have been developed, including full cell teardown. In addition, comparison and analysis are introduced on how to determine the solid-phase diffusion coefficient and the reaction-rate constant. The result shows that the combination of the galvanostatic intermittent titration techniques combined with DFN model is the most accurate approach. Furthermore, a parameterization of the cell without cell tear-down is performed by collecting data from the electric vehicles, while charging and driving. The combination of sensitivity analysis and optimization shows that RMSE below 8 mV is achieved. Finally, simulation are conducted to improve the charging speed in the EV, compared with conventional constant current-constant voltage (CC-CV) protocol. A control-based charging protocol that employs physics-based models proposed in the thesis, demonstrates a 32 % faster charging speed compared to the standard CC-CV protocol from a state of charge of 5 % to 80 %, while keeping the anode potential within safe limits to prevent Li-plating.