Distributed Predictive Control Approach for Fuel Efficient Gear Shifting in Hybrid Electric Vehicles

Energy management strategies (EMS) employed in hybrid electric vehicles determine the toque split between propulsion machines in such a way that the driver torque demand is satisfied. Therefore, a certain cost function is minimized over a driving cycle subject to constraints on the powertrain. Apart from optimizing the torque distribution the EMS can consider the optimization of clutch engagement and gear choice to further improve the overall fuel economy of the hybrid system. This paper presents a distributed predictive control architecture in which one control module computes the fuel optimal sequence of torque distributions while the other control entity determines the fuel optimal sequence of gear shift commands. To formulate the distributed control problem, the original mixed-integer optimal control problem is reformulated using partial outer convexification and relaxation. The fuel performance and computational complexity are evaluated in comparison to a centralized mixed-integer model predictive controller. Simulation results on a real-world driving cycle show that while nearly the same fuel economy is achieved, significant computational asset is gained with the distributed predictive controller being able to be executed on a rapid control prototyping platform.