Dimensioning of Integrated Starter-Generator Mild Hybrid System Using Real World Drive Cycles

Abstract

Hybrid vehicles are an important technology for reducing oil use and transportation-related emissions. It is well-known that hybrid and electric vehicles are often designed and tested using standard cycles such as the Highway Fuel Economy Test (HWY), Urban Dynamometer Driving Schedule (UDDS), and the US06 Supplementary Federal Test Procedure (US06). However, this begs the questions: How does real world driving compare to these cycles? Can a vehicle be designed using real world driving data which saves fuel in the real world compared to a vehicle designed using standard cycles? This thesis investigates this issue using a set of 5000km of real world driving data by light-duty pickup trucks, with the goal to optimize the fuel savings of a mild hybrid truck. The challenge with using a model-based design approach on thousands of kilometers of real driving data is the long model run-time required to iterate through plant and control parameters. Thus, this work develops a novel script which reduces optimization time by 78%. The key is to run the full model of the non-hybrid truck one time on the full driving data set, and then use the resulting vehicle speed, engine efficiency, engine torque, and engine speed, as inputs to the faster script. The script is then used to quickly iterate through the driving data set many times to find optimal control and plant parameters. In this work, exhaustive search is used; however, evolutionary optimization algorithms could also be used and would benefit from the fast script iteration on real world driving cycles. Overall, the use of the real world driving set for design of the mild hybrid truck resulted in a 7.10% decrease in fuel consumption compared to the non-hybrid truck, while the use of standard driving cycles for design resulted in a 5.45% fuel consumption decrease compared to the non-hybrid truck.