Techniques for hybrid electric vehicle controller synthesis
Control strategies have previously been developed for Hybrid Electric Vehicles (HEVs) that minimize fuel consumption while satisfying a charge sustaining constraint. Since one of the components of an HEV is the internal combustion engine, tailpipe emissions must also be considered. Additionally, the powertrain's behavior needs to be controlled in such a manner that the driver has a favorable opinion of its drive quality. In this dissertation, model-based control synthesis methods are presented as a way to develop feedback control policies for fuel consumption, tailpipe emissions, and drive quality in HEVs.
Shortest-path stochastic dynamic programming is used to design supervisory controllers that minimize a weighted sum of total expected fuel consumption and tailpipe emissions over drive cycles modeled by a Markov process. The use of shortest-path stochastic dynamic programming allows the charge sustaining constraint to be satisfied by a penalty applied when the vehicle is turned off. To synthesize a controller, an approximate shortest-path stochastic dynamic program is solved using a combination of linear programming (LP) and barycentric interpolation. Solution of the LP is accelerated by various techniques. To control drive quality, an optimal fuel and emissions controller is blended with a controller that considers instantaneous drive-quality.
To simplify the design of the supervisory controller, a technique for designing a static low-level controller is introduced.
The methods developed in this thesis are applied to a parallel hybrid truck to synthesize a charge sustaining controller that minimizes fuel consumption and engine-out emissions. This controller performs up to five percent better than a controller formulated using an infinite horizon, discounted cost, stochastic dynamic program. In order to illustrate the methods of the thesis on the important problem of minimizing tailpipe emissions, an HEV based on a dual-mode electrically variable transmission and a thermally transient catalyst is considered. The optimal controller maps four states to their optimal actions and is synthesized in three hours on a desktop computer. The controller reduces tailpipe emissions by more than fifty percent compared to a popular baseline controller. Drive quality is improved in a systematic manner by modifying the value function associated with this optimal controller.
0544: Electrical engineering