Comparison of compact reformer configurations for on-board fuel processing

international journal of hydrogen energy 35 (2010) 2305–2316

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/he

Comparison of compact reformer configurations for on-board fuel processing Mustafa Karakaya, Ahmet K. Avci* Department of Chemical Engineering, Bogazici University, Bebek 34342, Istanbul, Turkey

article info

abstract

Article history:

Two compact reformer configurations in the context of production of hydrogen in a fuel

Received 9 October 2009

processing system for use in a Proton Exchange Membrane Fuel Cell (PEMFC) based

Received in revised form

auxiliary power unit in the 2–3 kW range are compared using computer-based modeling

4 January 2010

techniques. Hydrogen is produced via catalytic steam reforming of n-heptane, the surro-

Accepted 5 January 2010

gate for petroleum naphtha. Heat required for this endothermic reaction is supplied via

Available online 25 January 2010

catalytic combustion of methane, the model compound for natural gas. The combination of steam reforming and catalytic combustion is modeled for a microchannel reactor config-

Keywords:

uration in which reactions and heat transfer take place in parallel, micro-sized flow paths

Microchannel reactor

with wall-coated catalysts and for a cascade reactor configuration in which reactions occur

Cascade reactor

in a series of adiabatic packed-beds, heat exchange in interconnecting microchannel heat

Methane combustion

exchangers being used to maintain the desired temperature. Size and efficiency of the fuel

Naphtha steam reforming

processor consisting of the reformer, hydrogen clean-up units and heat exchange

Auxiliary power unit

peripherals are estimated for either case of using a microchannel and a cascade configu-

Computational fluid dynamics

ration in the reforming step. The respective sizes of fuel processors with microchannel and cascade configurations are 1.53  103 and 1.71  103 m3. The overall efficiency of the fuel processor, defined as the ratio of the lower heating value of the hydrogen produced to the lower heating value of the fuel consumed, is 68.2% with the microchannel reactor and 73.5% with the cascade reactor mainly due to 30% lower consumption of n-heptane in the latter. The cascade system also offers advanced temperature control over the reactions and ease of catalyst replacement. ª 2010 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved.

1.

Introduction

Recent years have seen a growing interest in process intensification and microsystem technology due to the fact that chemical manufacturing systems that can be used in small scale and in distributed production are proved to have significant advantages such as reduced capital cost, compactness and potentially easier transportation [1]. These advantages can be particularly valuable when considering deployment of fuel-cell-power-driven vehicles in line with

stringent environmental regulations or integration of auxiliary power units (APU) into vehicles to provide power for comfort features such as climatization and lighting which would otherwise be generated by the drive engine. Miniaturization, down to the submillimeter scale, of confinements and/or repeated units within a process equipment is the crucial strategy towards process intensification. Such microstructured equipments have internal characteristic dimensions like channel diameter or gap height within the micrometer range [2]. Compared with conventional

* Corresponding author. Tel.: þ90 212 359 7785; fax: þ90 212 287 2460. E-mail address: [email protected] (A.K. Avci). 0360-3199/$ – see front matter ª 2010 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.ijhydene.2010.01.010

2306

international journal of hydrogen energy 35 (2010) 2305–2316

equipment comprised of macrosized elements, the common feature of microstructured devices is their strictly betterdefined operating condition range, higher flexibility and availability for internal and external numbering-up to increase the throughput [3]. One of the basic building blocks of process intensification has been the development of microchannel reactors that have parallel, identical channels with characteristic dimensions varying between w106 and 103 m and laminar flow conditions. Heterogeneous solid–gas, solid–liquid or solid–liquid– gas-phase reactions can be made to run over thin catalyst layers washcoated on the microchannel walls that are usually made of metallic substrates (Fig. 1a). The high surface area-tovolume ratio, resulting from these features of the microchannel reactors, is often exploited so as to overcome heat and mass transfer limitations frequently encountered in conventional reactors. The specific surface area in microchannel reactors vary in the range 10,000–50,000 m2 m3 whereas it seldom exceeds 1000 m2 m3 in their conventional counterparts [4,5]. In such microfluidic geometries, owing to the several orders-of-magnitude increase both in the heat exchange surface per unit volume and the heat transfer coefficients, very high overall heat transfer rates can be realized across the metallic channel walls and coated catalyst layers, which allows effective use of the catalyst and is also important for coupling exothermic and endothermic reactions in various configurations within a microchannel reactor stack (Fig. 1a) [4–7]. Petrachi et al. [8] investigated, using a 2-D mathematical model, the coupling dynamics of iso-octane steam reforming and catalytic combustion in a multifunctional reactor consisting of one combustion and one reforming channel separated by a steel wall. They also considered the flow of flue gas from an external burner as the heat source for steam reforming and reported that the latter configuration

Fig. 1 – Description of the parallel microchannel reactor configuration (a) and the characteristic unit cell (b).

was more advantageous for use in on-board applications due to shorter start-up times. The recent work by Stefanidis and Vlachos [9] involves the parametric study of coupled methanol/methane combustion and reforming in catalytic plate microreactors modeled in 2-D in order to determine the operating conditions under which complete conversion in either side of the reformer can be achieved without leading to formation of temperature hot spots. Methanol combustion and reforming coupling in various microreactor configurations using computational fluid dynamics techniques was studied by Arzamendi et al. [10]. They have carried out 3-D simulations of two different microreactor conceptual designs: one that involves 4 square parallel channels and one that is formed by two superposed sheets consisting of 10 channels each. The latter design was also classified into three types: cocurrent, counter-current and cross-flow of combustible and reforming streams. In each of the configurations complete conversion of both streams have been achieved with near isothermicity of the reactor blocks except for the 2-sheet configuration with counter-current flow of streams. Apart from complete miniaturization of the process equipment, highly intensified and integrated processes can be devised by promoting intensification in heat transfer while decoupling it from the reaction zones, as demonstrated by Seris et al. [11,12] via a pilot plant for the production of synthesis gas (H2 þ CO) by methane steam reforming. In this configuration, called multiple adiabatic bed arrangement or the cascade reactor system, reactions compartments, which are of adiabatic packed-beds based on conventional design, are interconnected via microchannel heat exchangers as shown in the flow diagram given in Fig. 2. The catalysts can be placed in slots within a heat exchange block, and as a result, there is no need to redistribute the reactants after each heat exchange stage. This system offers some operational advantages; it is beneficial for reactions demanding strict temperature control as the bed sizes can be arranged together with the amount and distribution of reactant flows to obtain the desired temperature profiles. Moreover, the removal of deactivated catalyst is much easier than in the case of microchannel configuration (Fig. 1) in which either the coated catalyst should be disintegrated from the reactor or the whole unit should be replaced. Desired conversion levels can be obtained by increasing the number of reactor-heat exchanger stages, which may be limited by pressure drop considerations.

Fig. 2 – Description of the cascade reactor configuration.

international journal of hydrogen energy 35 (2010) 2305–2316

The concept of using catalytic reactor beds with interstage heat exchange to achieve higher conversions is a classic approach and has industrially established examples such as SO2 oxidation. However, the cascade reactor system is novel since it employs microchannel technology for heat exchange purposes (e.g. via the printed circuit heat exchangers [13]) and is therefore an intensified version of an existing concept. Avci et al. [14] have recently performed a computer-based study for the designs and comparison of cascade and microreactor systems involving methane combustion and methane steam reforming/ethane dehydrogenation. The purpose of the present work is to explore the size, efficiency and fuel requirements of the microchannel reactor system compared to the cascade configuration. This is done in the context of the production of fuel-cell-grade hydrogen via steam reforming of n-heptane, a surrogate of petroleum naphtha [15,16] in an integrated fuel processor in order to drive a PEMFC-based APU in the 2–3 kW range. The heat needed to initiate and sustain endothermic steam reforming is supplied by the catalytic oxidation of methane, the model compound for natural gas. The microchannel configuration is simulated by a finite element based CFD technique that accounts for the flow fields, heat and mass transport phenomena and reaction in the channels and heat transport in the metallic walls at the steady-state. The simulation of the cascade configuration involves the use of the steady-state one-dimensional pseudohomogeneous fixed-bed reactor model for the packed-beds and CFD technique for the interconnecting microchannel heat exchangers.

2.

2307

Description of the process

The hydrocarbon fuel, n-heptane – surrogate for petroleum naphtha – can be converted to high purity, fuel-cell-grade hydrogen in a series of catalytic steps involved in the fuel processor/PEM fuel cell system shown in Fig. 3. In this system, n-heptane conversion takes place in the reformer unit via the steam reforming (SR) reaction which is considered to run over a Ni/Mg–Al2O4 catalyst [16]: C7 H16 þ 7H2 O ¼ 7CO þ 15H2 ;

DH0298 ¼ þ1108 kJ=mol

(1)

The major side reactions accompanying steam reforming are the methanation of synthesis gas (Reaction (2)), methane steam reforming (Reaction (3)) and the water–gas shift (Reaction (4)): DH0298 ¼ 206 kJ=mol

CO þ 3H2 ¼ CH4 þ H2 O; CH4 þ 2H2 O ¼ CO2 þ 4H2 ; CO þ H2 O ¼ CO2 þ H2 ;

DH0298 ¼ þ165 kJ=mol

DH0298 ¼ 41 kJ=mol

(2) (3) (4)

The heat required to drive the endothermic steam reforming is provided by catalytic combustion of methane, the model compound for natural gas, over a Pt/d–Al2O3 [17] catalyst. Effective thermal coupling of these reactions can be achieved in two intensified reactor configurations, the microchannel (Fig. 1) and the cascade (Fig. 2) systems, which are introduced in Section 1 and will further be described in Sections 3.1 and

Fig. 3 – Process flow diagram of the fuel processor/PEM fuel cell assembly.

2308

international journal of hydrogen energy 35 (2010) 2305–2316

3.2, respectively. Methane combustion can be represented by its total oxidation: CH4 þ 2O2 ¼ CO2 þ 2H2 O;

DH0298 ¼ 802 kJ=mol

(5)

The hydrogen-rich stream leaving the reformer contains carbon monoxide whose concentration must be reduced to below 10 ppm since it is a poison for the Pt-based catalysts of the PEM fuel cell [18–20]. The clean-up operation of the hydrogen-rich stream is considered to be undertaken in three stages. The first two stages involve the water–gas shift (Reaction (4)) run in fixed-bed adiabatic reactors at high- and low-temperature modes, respectively. The high-temperature water–gas shift (HTS) reaction takes place over a Fe-based catalyst and reduces the CO concentration to ca. 10% by mole [21]. The low-temperature water–gas shift (LTS) reaction runs over a Cu-based catalyst and reduces the CO concentration to less than 1% [22]. Preferential CO oxidation (Reaction (6)) as the final stage of cleaning-up has been reported to achieve CO removal at the desired level [23,24]: CO þ 1=2O2 ¼ CO2 ;

DH0298 ¼ 283 kJ=mol

(6)

The reformer feed, consisting of inputs for steam reforming (liquid n-heptane and water) and for combustion (methane and air) needs to be evaporated and heated from room temperature to the relevant reaction temperatures, and this can be achieved in the context of heat integration depicted in Fig. 3. Unburned methane from the combustion reaction, unconverted n-heptane from the reforming reactions and purged hydrogen from the fuel cell, which corresponds to ca. 25% of the hydrogen fed to the PEMFC [25], are catalytically burned in an afterburner, whose exhaust is then used to preheat, and if necessary, evaporate the feed streams through several heat exchange stages (HEX 1, 2 and 3 in Fig. 3). Furthermore, the product stream that leaves the reformer at temperatures around 763 K has to be cooled down to 673 K before being fed into the high-temperature water–gas shift (HTS) reactor, so another exchange takes place between the stream mentioned and the reforming feed stream (HEX-4). Even though the HTS reactor alone reduces the CO content substantially, the CO removal unit accepts streams with CO concentration less than 1%, which calls for another intermediate operation, namely the low-temperature water–gas shift (LTS) [26]. However, the LTS operation runs at temperatures between 473 K and 600 K [22], therefore, an exchanger (HEX-5) is installed between the high- and low-temperature water–gas shift reactors, which facilitates heat transfer from the SR/HTS product stream to the SR feed stream. The final heat exchange stage (HEX-6) applied down the process line is required to cool the outlet stream from the LTS reactor from 600 K to 353 K, the operating temperature of the CO oxidation reactor [27], and to preheat the SR feed stream. This configuration enables, at steady-state, the autothermal operation of the fuel processor/ PEMFC assembly and eliminates the need for external heat supply. Process intensification can be achieved through the compact design of all the reactor and heat exchanger units described above. However, as reported in previous studies [14,28,29], coupling scheme of the exothermic and endothermic reactions in the reforming unit makes the major

contribution to the size and efficiency of the fuel processor (reformer þ hydrogen clean-up units þ heat exchangers)/fuel cell assembly shown in Fig. 3. Therefore, the comparison is based on the design of the reformer that establishes itself in the coupling configuration of the exothermic catalytic combustion and endothermic steam reforming. The remaining parts of the fuel processor are considered to be of the same design, if not the same size, for both reactor configurations. The basis flow rate used to determine the reformer size is set as the amount of hydrogen required to drive a 2-kW PEM fuel cell, which corresponds to 2.2  102 mol H2 s1 [30,31]. However, when assessing the effectiveness of the fuel processing unit in terms of productivity, the amount of hydrogen that enters the fuel cell is used as the basis: in either reformer configuration, whether of microchannel or of cascade, hydrogen is also produced in the subsequent water–gas shift stages, so that the overall amount of hydrogen produced can drive a PEMFC-based APU in the 2–3 kW range, as desired.

3.

Mathematical modeling

3.1.

Microchannel reactor configuration

The microchannel reactor model geometry shown in Fig. 1 involves a repeating unit that consists of two parallel channels, called the unit cell, in which the exothermic (methane combustion) and the endothermic (n-heptane steam reforming) reaction streams flow in the counter-current mode, and heat is transferred through the stainless steel wall between the channels. Any given channel is part of a horizontal array of channels, each of which has the same reaction occurring within, as shown in the frontal view of the microchannel array on the y–z plane in Fig. 1a. Considering the shaded combustion channel in the figure, heat released due to the reaction is transferred in all directions. However, for two reasons the direction of net heat flow is towards the endothermic reforming channels, as indicated by the solid vertical arrows: (i) lateral heat flow in the z-direction (shown by the dashed horizontal arrows) out of the shaded channel is counterbalanced by heat flow from the respective combustion channels to its left and right, and (ii) the lateral temperature gradients column-wise are much smaller than those gradients row-wise due to the fact that the reactions taking place in a given row are identical (either combustion or reforming). Another implication of the alternating and symmetrical arrangement of the exothermic and endothermic channel groups is that inbound and outbound heat fluxes (solid vertical arrows) essentially cancel each other out, hence that the unit cell (Fig. 1b) is adiabatic. Moreover, the microchannel array is thought to be ideally insulated on the sides, thus heat loss to the surroundings is negligible. Convenience gained by the symmetric arrangement of the channels gives way to using a 2-D mathematical model for analyzing the coupling of combustion and reforming reactions in microchannel or catalytic plate reactors, as has been demonstrated widely in the literature [9,10,14,28,32–37]. Arzamendi et al. [10] have performed a three-dimensional simulation of coupled methanol combustion and reforming in a similar microchannel system, and their graphical results indicate that lateral

international journal of hydrogen energy 35 (2010) 2305–2316

variations of temperature and species concentrations are not significant. Therefore, a two-dimensional domain that excludes any variation in velocity, pressure, concentration or temperature along the channel width (in the z-direction) suffices to model the microchannel reactor. The square crosssectional channels are taken as 0.1 m long and considered to have a side length of 400 mm. The wall thickness is set as 200 mm. The porous, catalytic layers are assumed to be washcoated on the channel walls and their thickness are taken as 60 mm. Simulation of such a system requires the simultaneous solution of equations for conservation of momentum, energy and mass in the fluid passages and the wall (Table 1). The conservation of momentum is described by the two-dimensional Navier–Stokes equations in the Cartesian coordinates for an incompressible Newtonian fluid. Brinkman-type equations are used to model flow in the porous catalytic washcoat phase. Axial and lateral heat conduction effects in the metallic wall are also significant [38], so conjugate heat transfer between the wall and the fluids is accounted for by the equation of energy in the wall. The rates and heats of reactions are introduced to the mass and heat transport equations through the source terms. Only those reactions taking place in the catalyst washcoats are considered; homogeneous gasphase reactions are discarded. This is a sound assumption to make on the combustion side of the reformer since it takes longer times for the homogeneous reactions to ignite [39]. As for the reforming side, at high temperatures and pressures gas-phase homogeneous reactions can become important as de Smet et al. [40] reported in their work which involved the catalytic partial oxidation of methane on a Ni-based catalyst at 773 K and 1 bar. At these operating conditions, residence time required for heterogeneous conversion of reactants was

2309

found to be almost one third of that required for homogeneous conversion, therefore, gas-phase reactions were seen not to play a significant role. The equations given in Table 1 are solved subject to the boundary conditions given in Table 2. Species concentrations and stream temperatures are specified at the channel inlets and plug-flow velocities are assumed. At the outlets, diffusive flux is taken to be negligible, thus only convective mass and heat flow are considered, and atmospheric pressure is specified. Heat transfer through the wall is considered by means of heat flux continuity at the fluid–solid interfaces. For all types of transport phenomena, axial symmetry about the channel centerlines are imposed. The model equations together with the associated boundary conditions are solved using the finite element method in the Comsol Multiphysics CFD environment run over an HP xw8400 workstation equipped with 4  2.00-GHz Xeon processors. The solution domain shown in Fig. 1b is meshed using triangular elements. The number of elements comprising the unstructured grid has been increased progressively from 2529 to 6555 to seek for the mesh-independent solution which is obtained with 4411 elements. The multicomponent gas mixtures are composed of methane, air (oxygen þ nitrogen), steam and carbon dioxide in the combustion channel, and of n-heptane, steam, carbon monoxide, hydrogen, carbon dioxide and methane in the steam reforming channel. The reactions are carried out at atmospheric pressure, so ideal gas behavior is assumed. The flows are strictly in the laminar regime, i.e. the Reynolds number in both channels is less than or equal to 250. Variations of the physical properties with temperature and composition have been tested in trial runs involving catalytic methane combustion in a single microchannel. The improvement of the results, however, was minor compared to

Table 1 – Two-dimensional mathematical model used to simulate transport and reaction in the catalytic microchannels. Fluid phase Equation of continuity Equation of motion Equation of species continuity Equation of energy

vvxj vxj

vvyj vyj

¼0     vv vv vp v2 v v2 v vv vv vp v2 v v2 v rfj ðvxj vxxjj þ vyj vyxjj Þ ¼ vxjj þ mj vx2xj þ vy2xj rfj ðvxj vxyjj þ vyj vyyjj Þ ¼ vyjj þ mj vx2yj þ vy2yj j j j j   vc vc v2 c v2 c vxj vxijj þ vyj vyijj ¼ DAB vx2ij þ vy2ij j j   vT vT v2 T v2 T rfj Cpf ;j ðvxj vxjj þ vyj vyjj Þ ¼ lfj vx2j þ vy2j þ

j

Washcoat phase Equation of continuity Equation of motion Equation of species continuity Equation of energy

vvxj vxj

þ

vvyj vyj

j

¼0

  m m vp v2 v v2 v ð kj Þvyj ¼ vyjj þ ð3pj Þ vx2yj þ vy2yj j j   vc vc v2 c v2 c vxj vxijj þ vyj vyijj ¼ DAB;eff vx2ij þ vy2ij  rs Rij ðcij ; Tj Þ j j   P vT vT v2 T v2 T rsj Cps;j ðvxj vxjj þ vyj vyjj Þ ¼ lj;eff vx2j þ vy2j þ rsj N k¼1 ðDHk Þðrk ðcij ; Tj ÞÞ m ð kj Þvxj

vp

m

¼ vxjj þ ð3pj Þ



v2 vxj vx2j

þ

v2 vxj vy2j



j

Solid phase Equation of energy j – Channel or washcoat 1. Combustion 2. Steam reforming i – Species i( j¼1): CH4, O2, H2O, CO2, N2 i( j¼2): C7H16, H2O, CO, CO2, H2, CH4

 lw

v2 Tw vx2

 2 þ vvyT2w ¼ 0 w

j

2310

international journal of hydrogen energy 35 (2010) 2305–2316

Table 2 – Boundary conditions associated with the model equations given in Table 1. Boundary conditions 1. Channel entrance: xj ¼ L ðfor COMBÞ; xj ¼ 0 ðfor SRÞ; cyj Uj ¼ Uin j cij ¼ cin ij Tj ¼ Tin j 2. Symmetry at the centerline: cxj ; yj ¼ 0 ðfor SRÞ; yj ¼ H þ Lw ðfor COMBÞ n,vj ¼ 0 n,ðDAB Vcij þ vj cij Þ ¼ 0 n,(lfVTj þ vjrfjCpf,jTj) ¼ 0 3. Along the fluid–solid wall interface: cxj ; yj ¼ H=2 þ ds ðfor SRÞ; yj ¼ H=2 þ ds þ Lw ðfor COMBÞ n,vj ¼ 0 n,ðDAB Vcij þ vj cij Þ ¼ 0 n,(lwVTw) ¼ n,(lfVTj þ vjrfjCpf,jTj) 4. Channel exit: xj ¼ 0 ðfor COMBÞ; xj ¼ L ðfor SRÞ; cyj pj ¼ pout j n,ðDAB Vcij Þ ¼ 0 n,ðlf VTj Þ ¼ 0 5. Solid boundaries: x ¼ 0 and x ¼ L; cyw

then assigned, which also dictate the amount of interstage heat exchange. The procedure is repeated until the number of bed pairs sufficient to produce hydrogen at the basis flow rate of 2.2  102 mol s1. If the assumed cascade configuration violates any of the process constraints given below or the desired reactant conversion cannot be achieved in an acceptable number of stages, then the process is reiterated by varying the assumed bed sizes and the amount of heat exchange. The upper limit of allowable methane conversion in Reaction (5) (w20%) is the most restrictive of the abovementioned constraints, because exceeding it may lead to light-off and cause difficulties in temperature control. Another but less stringent constraint is the aspect ratio of the packed-beds which forces the length-to-diameter ratio to be in a range so as not to cause process infeasibilities due to high pressure drop. Thus, the volume of the packed-beds in the combustion array, dictated by the amount of catalyst used in each bed, should be such that combustion conversion per bed is limited to 20%. Furthermore, pressure drop across each bed in the combustion and reforming arrays should not exceed 3%. A one-dimensional pseudohomogeneous reactor model is used for the simulation of the adiabatic fixed-bed operation that constitute a set of ordinary differential equations (ODE) solved in the MATLAB environment using a stiff solver:

n,(lwVTw) ¼ 0

dFij ¼ Rij dWj the immense increase in the computational and memory requirements. The number of degrees of freedom solved for was seen to increase an order-of-magnitude which made comparison-oriented simulation of the system impractical. Therefore, constant values for the thermal conductivities and viscosities, evaluated at the inlet conditions, have been used [41]. Gas-phase and washcoat effective diffusivities are taken as 1.8  105 and 5.35  107 m2 s1, respectively [35]. Effective thermal conductivities in the combustion and steam reforming washcoat layers are taken as 4.2 and 4.5 times the respective gas-phase conductivities [14]. Permeability and porosity of the washcoats are taken as 1  108 m2 and 4  101, respectively. Sizing of the microchannel system is based on the principle of numbering-up of the catalytic microchannels. This is handled by dividing the basis hydrogen flow rate (see Section 2) by the corresponding single steam reforming channel exit hydrogen flow rate, which gives the number of reforming channels required. The total number of channels (combustion and reforming) is found by doubling this value, which is then used together with the channel dimensions and wall thickness to calculate the size of the microchannel block.

3.2.

Cascade reactor system

Sizing of the cascade reactor system shown in Fig. 2 involves a trial-and-error procedure. Starting with assumed sizes of adiabatic packed-beds in the exothermic and endothermic reaction arrays, the outlet temperatures from each bed pair are determined via solving the reactor model given in Equations (7)–(9). The inlet temperatures to the subsequent pair are

dTj ¼ dWj

(7)

P ðDHk Þðrk Þ k P Fi Cpi

(8)

i

dpj b pj0 Tj FTj ¼ 0 dWj Acj rbj pj Tj0 FTj0 Fij ¼ Fij0 ; Tj ¼ Tj0 ; pj ¼ pj0 at Wj ¼ 0

(9)

(10)

The design of microchannel heat exchangers connecting the catalytic beds differs slightly from the microchannel reactor design in that only the fluid-phase equations given in Table 1 are considered due to the absence of catalytic porous media. Comsol Multiphysics CFD package is used to solve the equations. The representative unit cell of the heat exchanger is similar to the microchannel configuration shown in Fig. 1 and involves two adjacent rectangular channels in which the hot and cold streams flow co-currently, and the separating steel wall. The variables available for adjustment are the lengths and cross-sectional area of the channels. Inasmuch as heat loads at the intermediate exchange stages are variable, design is suited to the highest one. Sizing of the microchannel heat exchangers in the cascade system is done by dividing the total flow leaving each catalytic bed by flow through a single pair and by calculating the number of channels that meets the actual load. Owing to the interstage temperatures that are determined during the phase of sizing the beds, the heat load is known prior to design. Flow through a single pair of microchannels or equivalently, the number of microchannels is assumed initially. With the channel dimensions set, the transport equations are solved as explained above. The dimensions, i.e. length and cross-sectional area, are varied until the exit criteria

international journal of hydrogen energy 35 (2010) 2305–2316

(inlet temperatures to the subsequent pair of beds) are met. If the acceptable sets of variables cannot meet the criteria, the number of channels is changed and the procedure is started all over again. For either reactor system, rates of Reactions (1)–(4) describing steam reforming and of Reaction (5) describing combustion are given by Langmuir–Hinshelwood–type expressions presented in Table 3. The intrinsic kinetics of methane oxidation over a Pt/d–Al2O3 catalyst [17] is adopted for the methane combustion reaction. The intrinsic rate expression proposed by Tottrup [16] is used to describe steam reforming kinetics of n-heptane over a Ni/Mg–Al2O4 catalyst. The kinetics of the side reactions (2)–(4) on this nickel-based catalyst are evaluated using the rate expressions reported by Xu and Froment [42]. The bulk densities of the Pt/d–Al2O3 and Ni/Mg–Al2O4 catalysts are taken as 1109.5 [17] and 1205.8 kg m3 [42], respectively. In the microreactor configuration, methane and heptane feeds are set at 1.5  108 and 1.5  107 mol s1 per channel, respectively. Feed temperatures of the combustion and steam reforming channels are taken as 850 K and 750 K, respectively. In the cascade reactor system, methane feed to the combustion array and n-heptane feed to the reforming array are set at 1.8  103 mol s1 and 1.5  103 mol s1, respectively, and feed temperatures of both streams are taken as 750 K. Carbon-tooxygen ratio at the combustion inlet (C:O2), defined as the ratio of number of moles of carbon atoms in the hydrocarbon molecule to the number of moles of molecular oxygen is set to be at the stoichiometric value of 0.5 in the microchannel system and substoichiometric value of 0.25 in the cascade system. In both configurations, the steam-to-carbon ratio at the reforming inlet (S:C), defined as the number of moles of steam to the number of moles of carbon atoms in the hydrocarbon molecule is set as 3 for ensuring the minimization of coke formation over Ni-based catalysts [43]. The inlet conditions for both reformer configurations are summarized in Table 4.

3.3.

CO clean-up and heat exchange units

The water–gas shift and CO oxidation reactors, assumed to be of packed-bed type, are required to clean up the reformate

2311

stream leaving the microchannel or the cascade reactor (Fig. 3). Their contributions to the overall fuel processor volume in either configuration are determined by reactor simulations using a one-dimensional fixed-bed reactor model (Eqs. (7)-(9)): the catalyst weight in each of the reactors is calculated via a trial-and-error procedure until the effluent stream meets the pertinent criteria which are set as the limits of CO exit concentrations in the pertinent units and their operating temperature ranges (see Section 2). The high- and low-temperature fixed-bed reactors are assumed to run over Fe- and Cu-based catalysts, respectively. The preferential CO oxidation is considered to run over a Cu-based catalyst. The rate expressions for the high- and low-temperature water–gas shift reactions are adopted from [21,22], respectively, and the kinetics of CO oxidation in a hydrogen-rich stream is described using the rate expression proposed by Sedmak et al. [27]. These kinetic expressions are given in Table 3. The heat exchangers (HEX 1–6) integrated into the fuel processor/PEMFC assembly (Fig. 3) are considered in terms of the heat load they are to undertake, instead of their sizes. The percentage of heat recovery is taken to be 50% for all of the exchangers, which is a reasonable assumption if they are to be comprised of parallel microchannels [44]. The unconverted reactants n-heptane and methane and purged hydrogen are assumed to be fully oxidized over the Pt-based catalyst in the afterburner unit, so it is also treated as a heat exchanger working with 50% heat recovery.

4.

Results and discussion

The results of the simulations for the microchannel reformer configuration are presented in Fig. 4 and Table 4. It can be observed that the combustion and reforming channel temperatures equilibrate at a small distance down the channel length (w0.01 m) at around 730 K. This is the direct result of the high rate of heat flow between the reactive flow zones, which is one of the superior characteristics of microchannel reactors compared to their conventional counterparts: the overall heat transfer coefficient between the combustion and reforming channels are calculated to be

Table 3 – Rate expressions used to describe catalytic reactions involved in the fuel processor. 1 Rate (mol kg1 cat s )

Reaction 1

Catalyst Ni/Mg–Al2O4

[16]

Ni/Mg–Al2O4

[42]

pCH4 p2H

Ni/Mg–Al2O4

r1 ¼ ½1þK

C7 pC7 ðpH2 =pH2 O ÞþKH2 O ðpH2 O =pH2 Þ

2 3 4

H2

1þKCO pCO þKH2 pH2 þKCH4 pCH4 þKH2 O pH2 O =pH2

r3 ¼ pk3:53

1þKCO pCO þKH2 pH2 þKCH4 pCH4 þKH2 O pH2 O =pH2

r4 ¼ pkH4

pCO pH2 O pH2 pCO2 =Keq;4 1þKCO pCO þKH2 pH2 þKCH4 pCH4 þKH2 O pH2 O =pH2

2O

2

4 (HTS) 4 (LTS) 5

r4ðHTSÞ ¼ k4

p4H pCO2 =Keq;3

p0:9 p0:31 CO H2 O ðHTSÞ p0:156 p0:05 CO2 H2

2

1 ð1  Keq;4

r4ðLTSÞ ¼ k4ðLTSÞ ðpCO pH2 O  1=2 k5 KCH4 KO pCH4 pO2 2

r5 ¼ ½1þK

pffiffiffiffiffiffiffiffiffiffiffiffi 2

CH4 pCH4 þ

6

2

r2 ¼ pk2:52 H2

KCO KO2 pCO p0:2 O

r6 ¼ 0:5K

Ref.

pCH4 pH2 O p3H pCO =Keq;2 2

k1 pC7

pCO2 pH2 pCO pH2 O Þ

pCO2 pH2 Keq:ð4Þ Þ

Ni/Mg–Al2O4 Fe2O3/Cr2O3/CuO

[21]

Cu/ZnO/Al2O3

[22]

Pt/d–Al2O3

[17]

Cu0.1Ce0.9O2y

[27]

KO2 pO2 

2 0:2 CO pCO þKO2 pO 2

2312

international journal of hydrogen energy 35 (2010) 2305–2316

Table 4 – Comparison of the results of microchannel and cascade systems for the coupled methane combustion–heptane steam reforming reactions. Inleta 1

CH4 (mol s ) 8

1

Outleta 1

C7H16 (mol s )

C:O2

S:C

T (K)

xHC (%)

H2 (mol s )

T (K)

– 1.5  107

0.5 –

– 3

850 750

43 87

– 1.52  106

734 764

Microchannel (per channel)

COMB SR

1.5  10 –

Microchannel (28,900 channels)

COMB SR

2.17  104 –

– 2.17  103

0.5 –

– 3

Cascade

COMB SR

1.8  103 –

– 1.5  103

0.25 –

– 3

Total volume (m3)

1.04  103

– 2.2  102 750 750

86 88

– 2.2  102

496 764

1.39  103

a Inlet and Outlet refer to the reformer in Fig. 3.

9270 W m2 K1 (Nu ¼ 59 based on channel side length), whereas coupling of exothermic and endothermic reactions arranged in a similar fashion (adjacent reaction compartments, counter-current flow) but with catalysts packed in channels instead of being washcoated on the walls, results in overall heat transfer coefficients around 50 W m2 K1 [45]. Owing to the mean channel temperatures of w740 K, 87% steam reforming conversion of n-heptane (Fig. 4) and 43% combustion conversion of methane are obtained. The stream leaving the reforming channel contains 1.52  106 mol s1 H2 (w33% by mole), 7.74  107 mol s1 CO (w17%), trace amounts of CO2 and CH4, unconverted n-heptane and excess water (w50%). The molar H2:CO ratio of 1.96 is lower than the stoichiometric ratio dictated by Reaction (1) and implies the effect of reverse water–gas shift reaction (CO2 þ H2 ¼ CO þ H2O, DH0298 ¼ þ41 kJ/mol) at elevated temperatures. Formation of trace quantities of methane can also be related to the temperature effect, which does not favor the thermodynamics of exothermic methanation (Reaction (2)). Pressure drop along the microchannel unit is found to be negligible and turned out to be 700 Pa in the reforming channel and 100 Pa in the combustion channel. Coupling of catalytic combustion and steam reforming is also demonstrated using the cascade system shown in Fig. 2. Heat generated in a catalytic combustion bed is transferred to

the steam reforming gas stream through a microchannel heat exchanger. The cooled stream is then fed to the next combustion bed to generate heat for the subsequent reforming bed. Upon heat exchange, each designated by the vertical lines in Fig. 5, the heated reformer stream is fed to the next catalytic bed in which further reforming conversion takes place adiabatically (Fig. 2). Such heating–cooling cycles lead to saw-tooth temperature profiles along the combustion and reforming arrays. The resulting saw-tooth pattern and evolution of the cumulative n-heptane conversion are shown in Fig. 5. Even though conversion in a combustion bed is limited to 20%, the heat generated is enough to drive the endothermic reactions in the corresponding reforming bed, as indicated by 88% conversion of n-heptane and 86% conversion of methane (Table 4). The reformer stream leaving the cascade is comprised of w39% H2 (2.2  102 mol s1), 18% CO (9.4  103 mol s1), trace CO2 and CH4 (<1%), unconverted n-heptane and excess water (w42%). The H2:CO ratio at the exit is higher than that in the microchannel reactor configuration because lower temperatures are encountered in the last three reforming beds, a situation favored by the forward water–gas shift (Reaction (4)). By the same token of lower bed temperatures, Reaction (2) is more favored due to its exothermic nature and, therefore, the exit methane concentration, however small, is 12 times that for the microchannel

Fig. 4 – Simulation results of the methane combustion–heptane steam reforming coupling in the microchannel configuration.

international journal of hydrogen energy 35 (2010) 2305–2316

2313

Fig. 5 – Simulation results of the methane combustion–heptane steam reforming coupling in the cascade configuration.

configuration. The total pressure drop across the combustion and reforming arrays are respectively 21% and 10% of the inlet pressure. Pressure drop values are observed to come from the packed-beds of the individual arrays. The contribution of the microchannel heat exchanger sections to the overall pressure drop is found to be negligible. As explained in Section 2, sizing of both reformer configurations has been carried out on the basis of producing hydrogen at a rate of 2.2  102 mol s1. In the microchannel configuration, sizing is based on the H2 exit flow rate from a single steam reforming channel and turned out to be 14,450 microchannels for producing 2.2  102 mol H2 s1. The total number of microchannels (combustion plus reforming) is twice this quantity, i.e. 28,900. For the given channel length of 0.1 m, a nearly cubical arrangement can be made by a 170  170 array of microchannels. The dimensions of this arrangement, excluding the manifolds, are 0.1 m (length)  0.102 m (height)  0.1022 m (width), and the corresponding volume is 1.04  103 m3. In the cascade configuration, pairs of combustion and reforming catalyst beds are designed for side-by-side placement in single slots, therefore, they are assumed to be of the same length, differing in size only in diameter. Based on this constraint and the iterative algorithm explained in Section 3.2, the hydraulic diameter and length of each of the combustion beds are found to be 2.64  102 m and 5.28  102 m, respectively. The Pt-based catalyst loading per bed is 0.032 kg. Each of the steam reforming beds has 5.48  102 m diameter and 5.28  102 m length, and contains 0.15 kg of Ni-based catalyst. Controlled temperatures and desired conversion values are found to be achieved by the use of 9 beds per gas stream, summing up to a total of 18 catalytic beds. The use of a lower number of beds violates the combustion conversion constraint so that the operation ceases to be feasible. On the other hand, using more beds brings only incremental improvements in conversion. Dimensions of the interconnecting microchannel heat exchangers, based on 270 W, the largest amount of interstage heat exchange, are 7.5  103 m (length)  6.4 5  103 m (height)  2.84  102 m (width). Each heat exchanger houses a total of 1024 channels: the hot stream flows in channels of

360 mm depth while the cold stream channels have a depth of 240 mm. Both the hot and cold channels are 240 mm wide, are separated by 200-mm thick walls, and they are arranged into an array of 64  16. The cascade configuration consisting of 18 catalytic beds and 9 microchannel heat exchangers, occupies a block volume of 1.39  103 m3, 81% of which comes from the reforming beds alone. Size of the alternate reformer designs has been used in estimating the total volume of the fuel processing system by adding up the volumes of the CO clean-up units (see Section 3.3). When the microchannel reactor configuration is used in the primary hydrogen generation step via steam reforming of n-heptane, the total volume of the reactors including the highand low-temperature water–gas shift converters and the CO oxidation reactor is calculated to be 1.53  103 m3. In the case of using the cascade reactor configuration, that volume is found to be 1.71  103 m3, 1.12 times that for the microchannel reactor case even though the reactors involved in the cleaning steps are substantially smaller for the cascade configuration because of lower CO content in the reformate (1.12  102 versus 9.4  103 mol s1). Other comparison criteria between the reaction systems are the total size of the heat exchangers installed (Fig. 3) and the overall efficiency. Since the exchangers are thought to be of the same design, comparison based on heat load, calculated by multiplication of the mass flow rates, average specific heats and temperature differences across the heat exchangers, should suffice. The total heat exchange carried out across the six heat exchangers amounts to 4100 W and 3480 W for the microchannel and cascade reactor systems, respectively. The efficiency, on the other hand, is defined as the ratio of the lower heating values (LHV) of the product hydrogen and the fuels natural gas and n-heptane: h ¼ 100 

nH2  LHVH2 nCH4  LHVCH4 þ nC7 H16  LHVC7 H16

(11)

The lower heating values of natural gas, hydrogen and petroleum naphtha are 47.1, 120.2 and 44.9 MJ kg1, respectively [46]. nH2 refers to hydrogen fed to the fuel cell while nCH4 and nC7 H16 are amounts of methane and n-heptane fed to the

2314

international journal of hydrogen energy 35 (2010) 2305–2316

Table 5 – Comparison of the microchannel and cascade systems based on the amount of fuel consumed, overall efficiency, size and total heat load. Fuels fed (mol s1) CH4 Microchannel Cascade

C7H16 4

2.17  10 1.8  103

2.17  103 1.5  103

H2 produced (mol s1) 3.26  102 2.86  102

fuel processor. The amount of hydrogen at the downstream processing is different for the microchannel and cascade reformer configurations. Hydrogen that is rejected by the PEM fuel cell, corresponding to ca. 25% of the hydrogen fed to the PEMFC [25], is sent to the afterburner unit as part of the heat integration (Fig. 3). Based on this configuration, the efficiency of the cascade reactor system is found to be higher than that of the microchannel reactor, i.e. 73.5% versus 68.2%. The difference can be attributed to the amount of n-heptane consumed in the cascade system, which is w30% less than that used in the microchannel configuration. Data relevant to comparison criteria and the results are summarized in Table 5. Even though n-heptane conversions in both systems are nearly identical (Table 4), the number of moles of hydrogen produced per mole of heptane fed is 15 for the microchannel reactor and 19 for the cascade configuration (Table 5). Extra hydrogen is the product of the exothermic water–gas shift reaction that is favored by the distinctly lower average temperatures in the cascade (Figs. 4 and 5). Higher reforming temperatures, attainable by adjusting the inlet temperatures of the combustion beds, are automatically restricted by the process limitations imposed on the combustion reactors (allowable methane conversion of 20% per bed). In the microchannel configuration, on the other hand, heat from the combustion channel is immediately drawn into the steam reforming channel that constitutes a large sink due to the very high heat of reaction (1108 kJ/mol) and presence of excess steam (21 moles of steam per mole of n-heptane, S:C of 3). At lower inlet temperatures than the specified ones (850 K for combustion, 750 K for reforming), the reduction in the combustion stream temperature at the very upstream of the channel becomes so drastic that the reaction extinguishes before reaching the light-off value of conversion (20% of methane fed). Even after combustion initiates, continuous heat removal along the microreactor is so fast that the rate of methane conversion remains low only to give 43% conversion at the outlet although the mixture is stoichiometric. Therefore, the option of operating the microreactor at lower temperatures to take advantage of higher water–gas shift reaction rates is ruled out. Increased hydrogen production per unit amount of fuel in the cascade configuration, however, comes with the cost of using larger reforming reactors and feeding methane that is 8.3 times that used in the microchannel reactor. Because of the presence of interstage heat transfer and cyclic nature of the cascade configuration, heat generated in a combustion bed upon 20% conversion at a maximum (light-off restriction of the combustion beds, see Section 3.2) is transferred to the next reforming bed in the sequence. Therefore, each successive combustion step experiences low-temperature inlet conditions (Fig. 5). In order to compensate for the interrupted

Efficiency h

68.2% 73.5%

Reactor volume (m3)

Total volume of reactors (m3)

1.04  103 1.39  103

1.53  103 1.71  103

Total heat load (W) 4100 3480

heat generation and to sustain the adiabatic reforming reactions in successive adiabatic beds, more methane must be combusted. The combustion and reforming processes in the microreactor, however, run in single compartments (channels) and proceed uninterruptedly. Therefore less methane will be required to drive the endothermic reactions in the microreactor configuration. As explained in Section 2, the unconverted portion of methane is further used to complete the heat integration of the fuel processor/fuel cell assembly. The comparable results presented in Table 5 show that the microchannel reactor offers an advantage in so far a strict size restriction is concerned. However, in addition to its higher efficiency, the benefit of using the cascase reactor configuration lies in easy replacement of deactivated catalyst and decoupling of reaction and heat transfer for much better temperature control.

5.

Conclusions

Two intensified catalytic reactor configurations, parallel microchannels and cascades, are compared, using computerbased modeling techniques, in the context of production of fuel-cell-grade hydrogen to drive a 2–3 kW PEMFC-based APU via steam reforming of surrogate naphtha (n-heptane). Both of the processes involve coupling of endothermic n-heptane steam reforming, the hydrogen producing reaction, with an exothermic reaction, which is considered to be the catalytic combustion of methane. It is shown that fast heat transfer between the two reactions can lead to n-heptane conversions in excess of 87% in either reactor system. Comparison within the proposed designs of the reaction systems, based on the amount of hydrogen produced at the reforming step, reveals the benefit of microchannel configuration due to the lesser size (1.53  103 versus 1.71  103 m3) of the fuel processor comprised of the reformer and subsequent clean-up units. However, based on the ratio of the lower heating value of the hydrogen produced to the lower heating value of the fuel (methane þ n-heptane) consumed, a fuel processor with the cascade system is more efficient (73.5% versus 68.2%). Higher efficiency is mainly associated with n-heptane consumption which is ca. 30% lower than that of the microchannel system. The presence of catalyst in packed bed form allows its facile replacement in case of any deactivation and, therefore makes the cascade configuration the choice in terms of ease of operability. Nomenclature cross-sectional area of the fixed-bed reactors in Acj cascade array j, m2 cij concentration of species i in channel j, mol m3 Cpf,j specific heat of fluid in channel j, J kg1 K1

international journal of hydrogen energy 35 (2010) 2305–2316

Cpi Cps,j DAB DAB;eff H Fij DHk i j k kk Ki L Lw n N Nu pj pi rk Rij

Tj Tw Uj vxj,vyj v Wj xj yj yw

specific heat of species i, J mol1 specific heat of washcoat j, J kg1 K1 species diffusivity (A into stagnant B), m2 s1 effective species diffusivity in the washcoat layer, m2 s1 depth of microchannel, m molar flow rate of species i in cascade array j, mol s1 heat of reaction k, J mol1 species index channel or cascade array index reaction index rate constant for reaction k adsorption constant for species i microchannel length, m wall thickness, m normal unit vector number of reactions Nusselt number total pressure in channel j or cascade array j, bar partial pressure of species i, bar 1 rate of reaction k, mol kg1 cat s total rate of generation/depletion of species i in P channel j or cascade array j ð¼ nijk ðrk ÞÞ, k 1 mol kg1 cat s temperature in channel j or cascade array j, K wall temperature, K plug-flow velocity into channel j, m s1 x- and y-components of fluid velocity in channel j (Cartesian coordinates), m s1 fluid velocity vector in channel j, m s1 weight of catalyst in each reactor in exo- or endothermic array j, kg axial coordinate in channel j, m direction normal to the x-axis, m direction normal to the x-axis (inside the microchannel wall), m

Greek letters constant in Eq. (9), Pa m1 b0 ds thickness of catalytic washcoat/membrane, m void fraction of washcoat layer 3p k permeability of waschoat layer, m2 lfj fluid thermal conductivity in channel j, W m1 K1 lj,eff effective thermal conductivity of the catalytic washcoat j, W m1 K1 lw thermal conductivity of microchannel wall, W m1 K1 rbj bulk density of catalyst in each reactor in cascade array j, kg m3 rfj fluid density in channel j, kg m3 rsj catalytic washcoat solid density, kg m3 mj viscosity of fluid in channel j, kg m1 s1 nijk stoichiometric coefficient of species i in channel/ cascade array j for reaction k Subscripts and Superscripts 0 inlet of the fixed-bed reactor eff effective diffusivity or conductivity f fluid in inlet value of the variable

out s w

2315

outlet value of the variable washcoat wall

Abbreviations APU auxiliary power unit COMB combustion HTS high-temperature water–gas shift LHV lower heating value LTS low-temperature water–gas shift PEMFC proton exchange membrane fuel cell SR steam reforming

Acknowledgements Financial support is provided by TUBITAK through project MAG-108M509 and by Bogazici University Research Fund through project BAP-09HA507D. Ahmet K. Avci acknowledges TUBA-GEBIP program.

references

[1] Benson RS, Ponton JW. Process miniaturization – a route to total environmental acceptability. Chem Eng Res Des 1993; 71:160–8. [2] Kockmann N. Transport processes and exchange equipment. In: Brand O, Fedder GK, Hierold C, Korvink JG, Tabata O, editors. Advanced micro & nanosystems. Micro process engineering, vol. 5. Weinheim: Wiley-VCH; 2006. p. 71–113. [3] Schenk R, Hessel V, Hoffmann C, Kiss J, Lo¨we H, Ziogas A. Numbering-up of micro devices: a first liquid-flow splitting unit. Chem Eng J 2004;101:421–9. [4] Ehrfeld W, Hessel V, Lo¨we H. Microreactors: new technology for modern chemistry. Mainz: Wiley-VCH; 2000. [5] Kiwi-Minsker L, Renken A. Microstructured reactors for catalytic reactions. Catal Today 2005;110:2–14. [6] Kolb G, Hessel V. Micro-structured reactors for gas phase reactions. Chem Eng J 2004;98:1–38. [7] Tonkovich AY, Yang B, Perry ST, Fitzgerald SP, Wang Y. From seconds to milliseconds to microseconds through tailored microchannel reactor design of a steam methane reformer. Catal Today 2007;120:21–9. [8] Petrachi GA, Negro G, Specchia S, Saracco G, Maffettone PL, Specchia V. Combining catalytic combustion and steam reforming in a novel multifunctional reactor for on-board hydrogen production from middle distillates. Ind Eng Chem Res 2005;44:9422–30. [9] Stefanidis GD, Vlachos DG. High vs. low temperature reforming for hydrogen production via microtechnology. Chem Eng Sci 2009;64:4856–65. [10] Arzamendi G, Die´guez PM, Montes M, Centeno MA, Odriozola JA, Gandı´a LM. Integration of methanol steam reforming and combustion in a microchannel reactor for H2 production: a CFD simulation study. Catal Today 2009;143: 25–31. [11] Seris ELC, Abramowitz G, Johnston AM, Haynes BS. Demonstration plant for distributed production of hydrogen from steam reforming of methane. Chem Eng Res Des 2005; 83:619–25. [12] Seris ELC, Abramowitz G, Johnston AM, Haynes BS. Scaleable, microstructured plant for steam reforming of methane. Chem Eng J 2008;135S:S9–16.

2316

international journal of hydrogen energy 35 (2010) 2305–2316

[13] Tsuzuki N, Kato Y, Ishiduka T. High performance printed circuit heat exchanger. Appl Therm Eng 2007;27:1702–7. [14] Avci AK, Trimm DL, Karakaya M. Microreactor catalytic combustion for chemicals processing. Catal Today; 2009;. doi:10.1016/j.cattod.2009.01.046. [15] Chen Z, Ya Y, Elnashaie SSEH. Modeling and optimization of a novel membrane reformer for higher hydrocarbons. AIChE J 2003;49:1250–65. [16] Tottrup PB. Evaluation of intrinsic steam reforming kineticparameters from rate measurements on full particle size. Appl Catal 1982;4:377–89. [17] Ma L, Trimm DL, Jiang C. The design and testing of an autothermal reactor for the conversion of light hydrocarbons to hydrogen I. The kinetics of the catalytic oxidation of light hydrocarbons. Appl Catal A: Gen 1996;138:275–83. [18] Brown LF. A comparative study of fuels for on-board hydrogen production for fuel-cell-powered automobiles. Int J Hydrogen Energy 2001;26:381–97. [19] Semelsberger TA, Brown LF, Borup RL, Inbody MA. Equilibrium products from autothermal processes for generating hydrogen-rich fuel-cell feeds. Int J Hydrogen Energy 2004;29:1047–64. [20] Tingelo¨f T, Hedstro¨m L, Holmstro¨m N, Alvfors P, Lindbergh G. The influence of CO2, CO and air bleed on the current distribution of a polymer electrolyte fuel cell. Int J Hydrogen Energy 2008;33:2064–72. [21] Hla SS, Park D, Duffy GJ, Edwards JH, Roberts DG, Ilyushechkin A, et al. Kinetics of high-temperature water–gas shift reaction over two iron-based commercial catalysts using simulated coal-derived syngases. Chem Eng J 2009;146: 148–54. [22] Choi Y, Stenger HG. Water gas shift reaction kinetics and reactor modeling for fuel cell grade hydrogen. J Power Sources 2003;124:432–9. [23] Ahmed S, Krumpelt M. Hydrogen from hydrocarbon fuels for fuel cells. Int J Hydrogen Energy 2001;26:291–301. [24] Trimm DL, Onsan ZI. On-board fuel conversion for hydrogen-fuel-cell-driven vehicles. Catal Rev Sci Eng 2001;43: 31–84. [25] Golunski S. HotSpot fuel processor: advancing the case for fuel cell powered cars. Platinum Metals Rev 1998;42:2–7. [26] Song CS. Fuel processing for low-temperature and hightemperature fuel cells – challenges, and opportunities for sustainable development in the 21st century. Catal Today 2002;77:17–49. [27] Sedmak G, Hocevar S, Levec J. Kinetics of selective CO oxidation in excess of H2 over the nanostructured Cu0.1Ce0. 9O2y catalyst. J Catal 2003;213:135–50. [28] Karakaya M, Avci AK. Simulation of on-board fuel conversion in catalytic microchannel reactor-heat exchanger systems. Top Catal 2009;52:2112–6. [29] Avci AK, Trimm DL, Onsan ZI. Quantitative investigation of catalytic natural gas conversion for hydrogen fuel cell applications. Chem Eng J 2002;90:77–87.

[30] Zalc JM, Lo¨ffler DG. Fuel processing for PEM fuel cells: transport and kinetic issues of system design. J Power Sources 2002;111:58–64. [31] Tan O, Masalaci E, Onsan ZI, Avci AK. Design of a methane processing system producing high-purity hydrogen. Int J Hydrogen Energy 2008;33:5516–26. [32] Zanfir M, Gavriilidis A. Modelling of a catalytic plate reactor for dehydrogenation–combustion coupling. Chem Eng Sci 2001;56:2671–83. [33] Zanfir M, Gavriilidis A. Catalytic combustion assisted methane steam reforming in a catalytic plate reactor. Chem Eng Sci 2003;58:3947–60. [34] Zanfir M, Gavriilidis A. Influence of flow arrangement in catalytic plate reactors for methane steam reforming. Chem Eng Res Des 2004;82:252–8. [35] Cao C, Wang Y, Rozmiarek RT. Heterogeneous reactor model for steam reforming of methane in a microchannel reactor with microstructured catalysts. Catal Today 2005; 110:92–7. [36] Deshmukh SR, Vlachos DG. Effect of flow configuration on the operation of coupled combustor/reformer microdevices for hydrogen production. Chem Eng Sci 2005; 60:5718–28. [37] Stefanidis GD, Vlachos DG. Millisecond methane steam reforming via process and catalyst intensification. Chem Eng Technol 2008;31:1201–9. [38] Hardt S, Ehrfeld W, Hessel V, Vanden Bussche KM. Strategies for size reduction of microreactors by heat transfer enhancement effects. Chem Eng Commun 2003;190:540–59. [39] Veser G, Frauhammer J. Modelling steady state and ignition during catalytic methane oxidation in a monolith reactor. Chem Eng Sci 2000;55:2271–86. [40] de Smet CRH, de Croon MHJM, Berger RJ, Marin GB, Schouten JC. Design of adiabatic fixed-bed reactors for the partial oxidation of methane to synthesis gas. Application to production of methanol and hydrogen-for-fuel-cells. Chem Eng Sci 2001;56:4849–61. [41] Rase HF. Fixed-bed reactor design and diagnostics. Massachusetts: Butterworths; 1990. [42] Xu J, Froment GF. Methane steam reforming, methanation and water–gas shift: I. Intrinsic kinetics. AIChE J 1989;35: 88–96. [43] Rostrup-Nielsen JR. Catalytic steam reforming. In: Anderson JR, Boudart M, editors. Catalysis, science and technology, vol. 5. Berlin: Springer; 1984. p. 1–117. [44] Brandner JJ, Anurjew E, Bohn L, Hansjosten E, Henning T, Schygulla U, et al. Concepts and realization of microstructure heat exchangers for enhanced heat transfer. Exp Therm Fluid Sci 2006;30:801–9. [45] Kolios G, Frauhammer J, Eigenberger G. Efficient reactor concepts for coupling of endothermic and exothermic reactions. Chem Eng Sci 2002;57:1505–10. [46] Kolb G. Fuel processing: for fuel cells. Weinheim: Wiley-VCH; 2008.