The current effort focuses on the Autothermal Reforming (ATR) reactor. Within this reactor it is frequently assumed that three non-elementary reactions take place: Total Oxidation, Steam Reforming, and Water Gas Shift. The ATR reactor consists of a solid catalyst support residing within an adiabatic tubular vessel. Reactant gases are then passed through the support at a high space velocity ~50,000/h. The axial temperature profile within the (tubular) reactor consists of a hot spot near the inlet (from the fast oxidation reaction) followed by a substantial temperature drop (due to endothermic, but slower steam reforming). Dynamically, one can identify two clear time-scales. In the gas phase, the high space velocity indicates that axial species concentrations will respond almost instantaneously to concentration and flow changes at the inlet. However, in spite of potentially large changes in heat generation (due to new reaction rates), the temperature response of the solid will be fairly slow.
The control system objectives of the ATR are as follows. As power demands at the fuel cell change, ATR throughput will need to respond quickly, while maintaining tight specifications on reformat composition and operating temperature. Among the most challenging transient scenario is start-up. Since the first goal of start-up is to arrive at the desired temperature as soon as possible, CPOX operation is usually the first phase (a feed of only fuel and air). However, to achieve the desired low CO concentration, steam must be added (ATR operation). This addition of steam will, however, drop reactor temperature and may even extinguish the reaction, if an increase in air flow does not occur. However, this increase should be such that temperature limits are not exceeded. Thus, the start-up objective is to develop an air flow-rate scheme to achieve ATR operation in minimum time, while continuously satisfying reactor temperature constraints.
In a previous effort, the authors have developed a CDF type dynamic model of the ATR reactor as well as tested classic feedback and feed-forward control schemes, aimed at the start-up problem. In particular, it was concluded that a control scheme based on a lumped parameter linear model was insufficient to achieve the fast start-up required by the on-board fuel processing application. Thus, in the first part of the presentation we will describe our recent efforts to develop a reduced order ATR model capable of capturing nonlinear and spatial aspects, while being computationally efficient enough for real-time applications. In the second part, we will illustrate the application of this model within a predictive control framework.