Analytical Models for the Sizing Optimization of Fuel Cell Hybrid Electric Vehicle Powertrains 2023-24-0133

Improving the development of electrified vehicles requires finding efficient methods for the component sizing of complex powertrains, since they may require a control optimization (for the energy management) which, when added to the sizing optimization, significantly increases the design space. A methodology to estimate the fuel consumption with a closed-form expression is found in the literature, which can be used to reduce the control/plant co-optimization to a static optimization problem. This approach can be used by either estimating the consumption of an existing powertrain: the descriptive level; or by predicting how the consumption will vary with the sizing parameters of the powertrain components: the predictive level. In previous works, the descriptive level was applied to the Toyota Mirai, a Fuel Cell Hybrid Electric Vehicle, showing that it can be extended to vehicles with a fuel cell system. In the present work, the models required for the predictive level are presented, which allow the actual sizing optimization to be performed. The model coefficients are fitted with data from components of different sizing: the battery pack model is fitted with experimental data from roller test benches, the electric machine model is fitted with numerically generated efficiency maps, and the power electronic models are derived from datasheets of discrete components, which are then integrated into the model of the boost converters and the inverter. Finally, the fuel cell system of the Toyota Mirai 2 is used as a reference system for the model. The validation was done by evaluating the errors introduced by each component model. A MATLAB routine was used to calculate the vehicle consumption on driving cycles using the Equivalent Consumption Minimization Strategy for the energy management. The results show that the proposed models adequately estimates the vehicle consumption through the considered cycles, with all errors remaining below three percent.