Dissertation Defense: Optimally-Personalized Hybrid Electric Vehicle Powertrain Control

Xiangrui Zeng, PhD Candidate, Mechanical Engineering

All dates for this event occur in the past.

E525 Scott Lab
E525 Scott Lab
201 W. 19th Ave
Columbus, OH 43210
United States

Committee

  • Professor Junmin Wang
  • Assistant Professor Ryan Harne
  • Professor Chia-Hsiang Menq
  • Associate Professor Haijun Su

Abstract

One of the main goals of hybrid electric vehicle technology is to improve the energy efficiency. In industry and most of academic research, the powertrain and aftertreatment control is designed and evaluated under standard driving cycles. However, the situations that a vehicle may encounter in the real world could be quite different from the standard cycles. Studies show that the human drivers have a great influence on the vehicle energy consumptions and emissions. The actual operating conditions that a vehicle faces are not only dependent on the roads and traffic, but also dependent on the drivers. A standard driving cycle can only represent the typical and averaged driving style under the typical driving scenarios, therefore the control strategies designed based on a standard driving cycle may not perform well for all different driving styles. This motivates the idea to design optimally-personalized hybrid electric vehicle control methods that can be adaptive to individual human driving styles and their driving routes. Human-subject experiments are conducted on a driving simulator to study the driving behaviors. A stochastic driver pedal model that can learn individual driver’s driving style is developed first. Then a theoretic investigation on worst-case relative cost optimal control problems, which is closely related to vehicle powertrain optimal control under real-world uncertain driving scenarios, is presented. A two-level control structure for hybrid electric vehicles is proposed, where the parameters in the lower-level controller can be on-line adjusted via optimization using historical driving data. The methods to optimize these parameters are designed for fixed-route driving first, and then extended to multi-routes driving using the idea similar to the worst-case relative cost optimal control. The performances of the two proposed methods are shown through simulations using human driving data and stochastic driver model data respectively. The energy consumption results in both situations are close to the posteriori optimal result and outperform other existing methods, which show the effectiveness of applying optimally-personalized energy management strategy on hybrid electric vehicles. Finally, a route-based global energy-optimal speed planning method is also proposed. This off-line method provides a useful tool to evaluate the potential of other speed planning methods, for either eco-driving guidance applications or future automated vehicle controls.