Robust model predictive control of a catalytic autothermal methane reformer for fuel cell applications

Abstract

Model uncertainty is an inseparable part of a methane reforming process since many of its main parameters describe microscale behavior. A robust control approach for a catalytic fuel reformer is presented to guarantee performance constraints are met even in the presence of model uncertainty. The control strategy is based on a recently developed sampling-based robust model predictive control (RMPC) with a new approach to capturing the uncertainty. Particular features of the control strategy is the use of a metric for coking susceptibility as a constraint on the inlet gas composition, a cost based metric for evaluating decisions on controller commands, and the ability to meet constraints on reactor temperature, outlet gas energy content, and hydrogen to carbon monoxide ratio. Simulation results carried out in MATLAB on random plants show the constraints are met regardless the model uncertainty and measurement noise.