A Fast Plant-Controller Optimization Process for Mild Hybrid Vehicles

Hybrid vehicles are an important technology for reducing oil use and transportation-related emissions, and there has been recent renewed interest in mild hybrid powertrains due to their ability to provide moderate fuel savings at a relatively low cost. Simulation plays a major role in the design of hybrid vehicles, but slow simulation run times can sometimes be a limiting factor in the optimization process. This paper proposes a fast script-based optimization process that speeds up optimization iterations by 130 times compared to running a full Simulink model in rapid accelerator mode. This increase in speed can allow larger amounts of real-world data to be used in the design process. To investigate the use of real-world data in the design process, 5400 km of pick-truck driving data is used to optimize one plant and one controller parameter in a mild hybrid powertrain, and the results are compared to the optimal parameters found using three standard drive cycles. It was found that when testing on a 500-km validation dataset, the optimal designs from the UDDS, HWFET, and a created combination cycle led to 2.1%-3.8% higher fuel consumption than the optimal design from the large real-world dataset.