Wiener Model and Extremum Seeking Control for a CO Preferential Oxidation Reactor with the CuO-CeO 2 catalyst Hyun Chan Lee, Sin Kim, Jae Pil Heo, Dong Hyun Kim, and Jietae Lee* Department of Chemical Engineering, Kyungpook National University, Taegu 702-701, Korea *Corresponding author. Tel:+82-53-950-5620, Fax:+82-53-950-6615, e-mail: jtlee@knu.ac.kr Abstract: Hydrogen rich gas produced by hydrocarbon reforming contains CO up to several thousand ppm which acts as a poison for the proton exchange membrane fuel cells. Preferential oxidation (PROX) is one of the promising methods that reduce its concentration below 10ppm. The PROX reactor with the CuO-CeO 2 catalyst shows optimal reactor temperature and a control system that maintains the PROX reactor temperature at its optimal one under changing environments and catalyst deactivations. Experimental studies to obtain a dynamic model for the CO PROX reactor have been done and the Wiener-type nonlinear model is obtained. Then the extremum seeking control that can track the optimal temperature for the Wiener-type nonlinear system is applied and its performance is verified experimentally. The extremum seeking control is a non-parametric real-time optimization method and can be applied effectively to other CO PROX reactors with different operating conditions. Keywords: PROX Reactor, CO Oxidation, PEM Fuel Cell, Wiener Model, Extremum Seeking Control. 1. INTRODUCTION Hydrocarbon reforming can be used to produce economically hydrogen, a source for the proton exchange membrane fuel cell. The hydrogen-rich gas from hydrocarbon reforming reactors can have CO up to several thousand ppm which poisons the anode catalyst of the fuel cell. The CO concentration need be removed below 10ppm. The preferential oxidation (PROX) is one of the simplest ways for this purpose. It uses the following oxidations: ]) / [ 828 . 241 ( 5 . 0 ]) / [ 984 . 282 ( 5 . 0 298 2 2 2 298 2 2 mol kJ H O H O H mol kJ H CO O CO o o           (1) In reactions of Eq. (1), the CO oxidation should occur preferentially over the H 2 oxidation. The copper-cerium oxide catalyst of CuO-CeO 2 has been reported to have high activity and selectivity for CO oxidation (Dudfield et al. 2001; Kim and Cha, 2003).The CO PROX reactors with CuO-CeO 2 catalyst can remove CO from 1% to less than 10ppm. In the PROX reactors, the H 2 oxidation also occurs, degrading the overall efficiency of fuel cell system with hydrocarbon reforming. To design the CO PROX reactor, kinetics data for the CO and H 2 oxidations under various mixture conditions are necessary. Recently, Lee and Kim (2008) proposed such kinetics for CO and H 2 oxidations over the CuO-CeO 2 catalyst as 62 . 0 37 . 0 91 . 0 10 2 2 ] / [ 4 . 94 exp 10 4 . 3 )] /( [              O H CO CO CO P P P RT mol kJ s kg mol r 69 . 0 48 . 0 13 2 2 2 2 ] / [ 142 exp 10 1 . 6 )] /( [              O H CO H H P P P RT mol kJ s kg mol r (2) Here the partial pressures are in the unit of kPa. Both oxidation rates are independent of the oxygen partial pressure, i.e., zero-order reactions for the oxygen concentration, and CO and H 2 reaction rates are near first- order for CO and H 2 partial pressures, respectively. Rigorous simulations that use the above kinetics with diffusion of the reactants into the catalyst pore structure have been done (Kim et al., 2013), showing effects on the reactor performances for various process conditions such as the feed composition of oxygen and the reactor temperature. One of the results to be noted especially is that, under limited oxygen feed flow rate, there is the reactor temperature window where the CO concentration is below 10ppm. There is the optimal reactor temperature within this temperature interval. The optimal reactor temperature will change as reactor conditions and environments change. To maintain the reactor temperature at its optimal one under changing environments, a control system is required for optimal and longer operations of the reactor. For this, a dynamic model of the CO PROX reactor is studied first. The mathematical model is constructed and, for high thermally conductive reactor, the dynamics between the heater input and the CO conversion is shown to be well approximated by a Wiener-type nonlinear model. Experimental step responses support this Wiener-type nonlinear model. To maintain the exit CO concentration at its minimum is a task different from the traditional control one to regulate the output at a given set point. There have been several approaches to attack this real-time optimization problem such Preprints of the 9th International Symposium on Advanced Control of Chemical Processes The International Federation of Automatic Control June 7-10, 2015, Whistler, British Columbia, Canada TuM2.2 Copyright © 2015 IFAC 575