(721c) Application of Economic MPC to a Hybrid Fuel Cell Vehicle

The notion of a fuel cell powered vehicle has captured the imagination of many as a clean/efficient alternative to the internal combustion engine. However, the capital cost of a fuel cell power unit is significant, and threatens to be a show-stopping hurtle. As such, many have looked to hybridize the fuel cell with lower cost energy storage devices; a rechargeable battery and/or super-capacitor. If given a hybrid vehicle configuration, a fundamental question concerns the coordination of power output from each device. How much power should each device provide in response to a demand? How fast should a storage device be recharged and to what level? If the system contains more than one storage technology - each with unique power/energy density characteristics - then the issue becomes even more complicated. Central to the power coordination question is the physical limits of each device. Clearly, a battery or super-capacitor can hold only a finite amount of energy and cannot output power if the reserve is fully depleted. In addition, the rate of charging and discharging of these devices should be limited to observe heat dissipation related safety concerns. Similarly, a fuel cell will have a maximum power output limit and cannot accept any power. In addition, one may wish to limit the ramp rate of fuel cell power output, in an effort to reduce degradation rates.

Clearly the prominent role of equipment limitations, with respect to energy storage capacity and maximum power, suggests the use of predictive control for constraint enforcement. In this work we investigate the use of Economic MPC for the hybrid vehicle application. In this novel formulation the economic objective is defined as the instantaneous energy loss from the storage devices. Unfortunately, the resulting on-line optimization problem turns out to be a computationally slow quadratically constrained quadratic programming problem. Thus, the main result of the paper is to show that an approximation of the EMPC problem will lead to a computationally efficient quadratic programming implementation with virtually zero loss in performance.