comparison of reformer.pdf

Comparison of methanol-based fuel processors for
PEM fuel cell systems
James R. Lattner, Michael P. Harold*
Department of Chemical Engineering, University of Houston, 4800 Calhoun Road, Houston, TX 77204, USA
Received 15 March 2004; received in revised form 25 May 2004; accepted 10 June 2004
Available online 19 October 2004
Abstract
The deployment of proton exchange membrane (PEM) fuel cells requires efficient conversion of fuels to hydrogen in distributed facilities.
Methanol is often considered as a fuel source because it is stored as a liquid and can be reformed to hydrogen at relatively mild conditions. We
have compared steam reforming (SR), autothermal reforming (ATR), and ATR membrane reactor based fuel processors using a commercial
CuO/ZnO/Al2O3 catalyst. In our approach, we first optimize the flowsheet and identify reaction conditions that maximize overall system
efficiency. Reaction kinetics and heat transfer are then incorporated into the process efficiency analysis as well as volume requirements for
each fuel processor. We show that the SR and ATR fuel processors coupled with a PEM fuel cell achieve about 50% overall efficiency (LHV
basis), with roughly equal fuel processor volumes of about 29 and 22 liters for 50 kW net power generation, respectively. The ATR fuel
processor requires distributed air injection in order to avoid overheating the copper catalyst and the formation of excessive CO. The ATR
membrane reactor combines the same features with hydrogen removal in the reforming section of the reactor. The main benefit of the ATR
membrane reactor is a reduction of the fuel processor volume to 13 liters, at the expense of a more complex steam system and a small reduction
in overall efficiency.
# 2004 Elsevier B.V. All rights reserved.
Keywords: Fuel processors; Methanol; Membrane reactors; PEM fuel cells; Hydrogen; Reactor engineering; Steam reforming; Autothermal reforming
1. Introduction
Fuel cell powered vehicles offer the potential for high
efficiency and reduced emissions compared to internal
combustion engines. Low temperature proton exchange
membrane (PEM) fuel cells require hydrogen and oxygen
(or air) as reactants. Although direct use of hydrogen has
many advantages, the production, distribution and storage
on a vehicle presents many challenges [1]. Hydrocarbon
fuels such as gasoline or diesel provide much higher storage
densities for hydrogen [2]. Natural gas (methane) is abun-
dant, particularly in remote locations, and is a good source
of hydrogen, but is not easily distributed and stored on a
vehicle [3]. Methane can be converted to methanol, which is
easy to store and ship, making methanol a ‘‘transportable’’
form of methane. In addition, methanol can be converted
to hydrogen at milder conditions than petroleum-based
hydrocarbons, making it an attractive fuel for on-board
hydrogen production.
Options for converting methanol to hydrogen include
steam reforming (SR), catalytic partial oxidation (CPO), and
autothermal reforming (ATR). ATR combines the endother-
mic SR reaction with the exothermic CPO reaction. Of these
options, steam reforming has the advantage of producing the
highest hydrogen concentration. However, a more compli-
cated reactor with external fuel combustion is required to
supply the necessary heat. Partial oxidation and autothermal
reforming provide lower hydrogen concentrations, particu-
larly if air is used as the oxidant. Autothermal reforming of
methanol has the potential to produce reasonable hydrogen
concentrations using a relatively simple reactor design.
The hydrogen produced from each of these reformer types
contains carbon monoxide, which is a poison to the platinum
PEM fuel cell catalyst. Options for removing carbon
monoxide from the reformate include preferential oxidation
www.elsevier.com/locate/apcatb
Applied Catalysis B: Environmental 56 (2005) 149–169
* Corresponding author. Tel.: +1 713 743 4304; fax: +1 713 743 4323.
E-mail address: mharold@uh.edu (M.P. Harold).
0926-3373/$ – see front matter # 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcatb.2004.06.024

(PrOx) and hydrogen separation membranes, such as
palladium-based alloys or proton-conducting ceramic
materials.
With the many options available for hydrogen production
and purification, it is difficult to determine which overall
system is optimal. In a previous study, several fuel processor
configurations for the autothermal reforming of hydrocarbon
fuel (n-tetradecane, a model component for diesel fuel) were
studied [4]. In the present study, we apply a similar approach
to the analysis of fuel processor configurations for the
autothermal reforming of methanol. Reaction kinetics and
heat transfer constraints are factored into the fuel processor,
which is then integrated into an overall fuel processor/fuel
cell system. We consider three basic fuel processor
configurations:
1. Steam reforming of methanol, with heat supplied by
catalytic combustion, followed by CO removal in a PrOx
reactor.
2. Adiabatic autothermal reforming of methanol, followed
by CO removal in a PrOx reactor.
3. Adiabatic autothermal reforming in a palladium mem-
brane reactor with countercurrent steam sweep.
Our objective is to compare overall system efficiencies
and reactor volumes as a function of fuel processor design.
The efficiency of the fuel processor/fuel cell combination
depends heavily on the heat integration scheme employed.
Each process utilizes similar heat recovery schemes, and
practical limits in heat exchanger approach temperatures are
applied uniformly in each case. The water balance is esp-
ecially important for the on-board reformer; each case re-
covers sufficient water from the exhaust stream such that no
additional water is needed. Variables in the design and o-
peration of the fuel cell system include oxygen/carbon ratio,
steam/carbon ratio, and reformer feed temperature. These
variables are optimized in the context of the overall process
model using a HYSYS process simulator to perform the heat
and material balance calculations. After optimization of the
reformer operating conditions, reaction kinetics are applied
consistently to each case to estimate and compare reactor
volumes for the various configurations.
For the present study, literature kinetic models were
utilized for the oxidation and reforming reactions on a CuO/
ZnO/Al2O3 catalyst. The steam reforming kinetics of
Peppley et al. [5] utilize three reactions:
methanol steam reforming ðSRÞ :
CH3OH þ H2O $ CO2 þ 3H2; DH
f ¼ 49:4 kJ mol1
(1)
methanol decomposition ðMDÞ :
CH3OH $ CO þ 2H2; DH
f ¼ 90:5 kJ mol1
(2)
water gas shift ðWGSÞ :
CO þ H2O $ CO2 þ H2; DH
f ¼ 41:1 kJ mol1
(3)
The steam reforming reaction (1) is the desired reaction, as it
produces only CO2 and hydrogen. The methanol decom-
position reaction (2) produces CO, which must be removed
prior to entering the PEM fuel cell. At methanol steam
reforming conditions, the reverse of the water gas shift
reaction (3) must be minimized to limit CO production.
The reforming and decomposition reactions, and the reverse
water gas shift reaction are all endothermic. An external
J.R. Lattner, M.P. Harold / Applied Catalysis B: Environmental 56 (2005) 149–169
150
Nomenclature
A
surf surface:volume ratio of membrane or heat transfer
surface
Co
p molar heat capacity of gas mixture
CSj concentration of surface sites
Ea activation energy
F Faraday’s constant = 96 485 C/mol
DHj heat of reaction for reaction j
kj rate constant for reaction j
Ki adsorption equilibrium constant for component i
Keq,j reaction equilibrium constant for reaction j,
Keq;j ¼ expðDGj=RTÞ
Ni molar flux of component i
Ng molar flux of total gas stream
pi partial pressure of component i
Qi permeability of component i
rj molar rate of reaction j based on mass of catalyst
R gas constant = 8.314 J mol1
K1
Sg surface area of catalyst
T pseudo-homogeneous reaction temperature
Uo overall heat transfer coefficient
V voltage
z axial dimension
Greek letters
d membrane thickness
hfc fuel cell efficiency
hj effectiveness factor for reaction j
ucat fraction of reactor cross-section containing cata-
lyst
nij stoichiometric coefficient for component i in reac-
tion j
rb catalyst bulk density
Subscripts
cm radial mean catalyst property
i component number
j reaction number
p permeate
r retentate
s solid
w wall

source of heat is required to obtain a reasonable conversion
of methanol.
An alternate to supplying heat externally is to co-feed
oxygen (air) to the reforming reactor. This is referred to as
autothermalreforming(ATR),wheretheexothermicoxidation
reactions supply heat to the endothermic reforming reactions.
Reitz et al. found that copper–zinc catalysts in the presence of
oxygen will catalyze the combustion of methanol:
methanol combustion : CH3OH þ 1:5O2 $ CO2 þ 2H2O;
DH
f ¼ 675:4 kJ mol1
(4)
Catalyst in the oxidized state was found ineffective at produ-
cingeitherCOorhydrogen.However,intheabsenceofoxygen
andbysupplyingsufficientheat,thecatalystwillbereducedby
methanol, becoming active for hydrogen production. By co-
feeding oxygen (or air) with methanol and steam to a reactor,
the combustion reaction occurs at the front of the bed and
produces a considerable amount of heat. Beyond the point in
the bed where the oxygen is depleted, the catalyst becomes
reduced and the remainder of the catalyst bed conducts the
steam reforming reactions (1)–(3) [6]. This two-step reaction
mechanism is used to model the ATR reactor cases.
2. Process modeling of fuel cell systems
2.1. PEM fuel cell considerations
In the proton exchange membrane (PEM) fuel cell,
hydrogen molecules ionize on the anode to form protons and
electrons. The electrons pass through an electrical circuit,
and the protons diffuse through an acidic polymer membrane
(Nafion1
or some variant) to the cathode side. At the
cathode, the protons, electrons, and oxygen molecules
combine to form water. At the PEM cell conditions of 80 8C,
1.1 bar pressure, and with excess air as the oxygen source,
the product water is present as a gas. The overall reaction is:
H2 þ 0:5O2 ! H2O ðgÞ ; DH
f ¼ 241:6 kJ mol1
;
DG
f ¼ 226:1 kJ mol1
(5)
The maximum reversible voltage that can be produced from
this reaction is based on the Gibbs energy change, given by
VmaxðGibbsÞ ¼
DG
f
2F
¼ 1:17 V (6a)
For the present study, the fuel cell efficiency is defined as the
electrical energy produced by the fuel cell stack divided by
the total enthalpy (not Gibbs energy) of reaction. The
product water is taken to be in vapor form, so the heat of
reaction is on a lower heating value (LHV) basis. Based on
this definition, the maximum voltage is given by:
Vmax ¼
DH
f
2F
¼ 1:25 V (6b)
The efficiency is thus:
hfc ¼
DHemf
DH
f
¼
Vemf
Vmax
(7)
where DHemf is the electrical energy produced by the cell.
The remaining enthalpy of reaction must be rejected as heat.
We recognize that the voltage Vmax based on enthalpy is not
theoretically attainable, but it allows calculation of the fuel
cell power output from enthalpies of reaction, which are
readily obtained from process simulation software. For the
present study, the actual cell voltage is assumed to be 0.75 V,
for an efficiency (LHV basis) of 60%. The operating tem-
perature of the stack is assumed to be 80 8C.
Not all of the hydrogen fed to the fuel cell anode reacts;
some of the hydrogen will remain unconverted and exit with
the anode exhaust gas. The fraction of hydrogen reacted in
the fuel cell is defined as the hydrogen utilization, and this
value will depend upon the purity of the anode feed
hydrogen stream. Assumed hydrogen utilizations are 85%
for the reformate feed case [7] and 95% for the high purity
hydrogen in the membrane cases [8].
Note that in the overall process model, unconverted
hydrogen in the fuel cell anode exhaust is combusted, and
the heat of combustion is utilized to preheat the reformer
feeds. In the present study, the fuel cell efficiency, as defined,
is not debited by the hydrogen utilization; only the converted
hydrogen is considered in the fuel cell energy balance.
The rate-limiting step in the PEM fuel cell is typically the
reaction kinetics at the cathode [9]. Excess air is required to
maintain a sufficient oxygen concentration at the cathode
exit to ensure reasonable current densities. An air
stoichiometry of 2.0 is assumed for the present study. This
means that twice the stoichiometric requirement of oxygen
is fed to the cathode based on the amount of hydrogen that is
converted (not fed to the anode). Higher air rates will
marginally improve fuel cell performance, but at the expense
of increased air blower energy and water losses from the
exhaust. Lower air rates cause an undesirable decline in fuel
cell voltages due to low oxygen concentrations in the
cathode [10].
2.2. Steam reformer system description
The fuel processor/fuel cell flow scheme for the steam
reforming of methanol is shown in Fig. 1. Liquid methanol is
fed to exchanger E-1, where it is vaporized and preheated
with combustion exhaust gases. The majority of the
methanol vapor is for reformer feed; a small stream is split
off to feed the combustion side of the reformer as
supplemental fuel (if needed). The reformer feed methanol
is mixed with steam and fed to the shell side of the reformer
reactor containing ca. 2 mm diameter pellets of CuO/ZnO/
Al2O3 catalyst. The reaction occurs at 3 bar pressure and an
exit temperature of about 270 8C. The methanol is reformed
with steam into a mixture of hydrogen, CO2, and a small
J.R. Lattner, M.P. Harold / Applied Catalysis B: Environmental 56 (2005) 149–169 151

amount (1%) of CO (lower operating temperatures are
more favorable for the water gas shift equilibrium).
The tube side contains a supported noble metal
combustion catalyst. The combustor feed consists of a
mixture of anode vent gas containing unused hydrogen,
cathode vent gas containing nitrogen, water vapor, unused
oxygen, and a small amount of methanol vapor for
supplemental fuel.
The reformer effluent is mixed with air and fed to the
preferential oxidation (PrOx) reactor. This reactor oxidizes
the small amount of CO to CO2 and also oxidizes the
unconverted methanol from the reformer. There is no
information in the literature regarding the fate of methanol
in a PrOx reactor; for purposes of this study we assume the
methanol undergoes methanol decomposition to CO and
hydrogen, followed by oxidation of the CO. We have
assumed that 50% of the oxygen reacts with CO to form
CO2, and 50% reacts with hydrogen to form water. With this
assumption, the net overall reactions for CO and methanol in
the PrOx reactor are:
CO þ H2 þ O2 ! CO2 þ H2O; DH
f ¼ 524:3 kJ mol1
(8)
CH3OH þ O2 ! CO2 þ H2O þ H2;
DH
f ¼ 433:8 kJ mol1
(9)
The PrOx reactor is sized to achieve a GHSVof 10 000 h1
,
based on work at Argonne National Laboratory on a pelle-
tized precious metal catalyst [11]. The catalyst volume to
achieve this space velocity in a 50 kW fuel cell is ca. 12.3 l.
A tubular reactor with boiling water on the shell side
(exchanger ‘‘E-4’’ in Fig. 1) is utilized to control the reaction
temperature at a constant 200 8C and to remove the heat of
reaction for the oxidation reactions (8) and (9). The pressure
on the steam generator is varied to control the PrOx reaction
temperature. The steam generated in this exchanger-reactor
is used in the reformer. The process effluent from the PrOx
reactor is further cooled in exchanger E-6 by exchanging
heat with the boiler feedwater.
The cooled reformate at about 150 8C exchanges heat with
the fuelcellcathodeairinexchangerE-5.Themainpurpose of
thisexchangeristoequilibratethetemperaturesofthecathode
and anode feed streams feeding the PEM fuel cell, at about
80 8C. The anode feed stream contains about 67% hydrogen,
anditisassumedthat85%ofthishydrogenpermeatesthrough
the fuel cell membraneandreacts with oxygen. Thefuelcellis
assumed to operate at 1.1 bar pressure. The heat generated in
the fuel cell stack is removed by circulating a cooling water
streamthroughthefuel cell stack.Thiscoolingwaterstreamis
cooled in an external air-cooled exchanger E-8.
The anode exhaust gas, containing some unreacted
hydrogen ( 24.8%, dry basis), and the cathode exhaust gas,
containing unreacted oxygen ( 11.7%, dry basis), are
mixed and become the primary reactants for the combustion
side of the reformer reactor. As mentioned previously,
supplemental methanol vapor is also fed to the combustor to
maintain the heat balance. The flow rate of the supplemental
methanol fuel is regulated to control the temperature of the
reforming side of the reactor.
152
Fig. 1. Steam reforming fuel processor/fuel cell system process flow diagram.

The exhaust from the combustion step leaves the
reforming reactor at 280–300 8C, and is then used to
vaporize and superheat the methanol feed in exchanger E-1.
This exhaust stream is then cooled further in air-cooled
exchanger E-7 to about 48 8C. A portion of the water
condenses in this exchanger, is separated from the exhaust
gas, and becomes the boiler feedwater for the steam
generator E-4. Excess water is purged from the system. One
advantage for using methanol as a feedstock compared with
hydrocarbons is that methanol produces more water per
mole of hydrogen than the hydrocarbon feedstocks, and
recovery of sufficient water at near-ambient conditions is not
an issue.
2.3. Autothermal reformer with PrOx system description
The autothermal reformer with PrOx flowsheet is shown
in Fig. 2. The methanol feed is preheated with hot exhaust
gas in exchanger E-1. The preheated methanol is mixed with
steam and is further preheated in exchanger E-2 using hot
exhaust gas from the combustor. The preheated methanol/
steam vapor mixture enters the reformer containing the
CuO/ZnO/Al2O3 catalyst at 3 bar and about 200 8C.
Compressed air is distributed axially along the bed to avoid
excessive local temperatures that would sinter the Cu
catalyst (this feature will be described further in Section
4.2). In the front section of the reformer bed the exothermic
oxidation reaction primarily occurs, although endothermic
reforming also occurs in oxygen-depleted areas, such as the
core of the catalyst pellets. Downstream of the air
distributors the endothermic methanol reforming and
decomposition, and exothermic water gas shift reactions
occur.
The effluent from the reformer is mixed with air and fed
to the PrOx reactor. The design of this system is identical to
the steam reforming system (Section 2.2). The final
reformate contains about 60% hydrogen. This is a lower
concentration than the steam reforming case due to the
presence of nitrogen diluent that is introduced with the
air.
The anode exhaust, containing unused hydrogen
( 17.6%, dry basis), and the cathode exhaust, containing
unused oxygen ( 11.7%, dry basis), are mixed and fed to a
catalytic combustor. During steady state operation, the
maximum temperature generated in the combustor is about
300 8C. This is well below the temperature where nitrogen
oxides are produced. The heat generated from the
combustion of the unused hydrogen is sufficient to provide
preheat for the ATR reactor under steady state conditions.
However, during system startups, or during vehicle
acceleration in an on-board system, additional methanol
may be fed to the catalytic combustor to provide requisite
heat to the ATR reactor.
The combustor exhaust preheats the ATR feed in two
exchangers. The exhaust is first cooled in exchanger E-2,
which preheats the methanol/steam mixture, and is further
cooled in E-1, where methanol feed is preheated. The
exhaust is finally cooled to 55 8C in air-cooled exchanger E-
7. Condensed water is recovered in a separator for recycle as
boiler feedwater, and excess water is purged.
Fig. 2. Adiabatic ATR with PrOx fuel processor/fuel cell system process flow diagram.

2.4. Autothermal reforming membrane reactor system
description
The ATR membrane reactor uses a dense Pd or Pd alloy
membrane to separate hydrogen from the reformate gases
within the ATR reactor, thus eliminating the PrOx reaction
step. The overall process flow diagram is shown in Fig. 3.
Fig. 4 is a schematic of the ATR membrane reactor with
staged air injection and countercurrent steam sweep on the
permeate side of the membrane. In the design of Fig. 4, the
oxidation zone and the membrane zones do not overlap.
While this is important in the autothermal reforming of
hydrocarbons to prevent temperatures at the membrane from
exceeding 900 8C [4], it is not so critical in the methanol
ATR case due to the lower temperatures involved.
A membrane reactor requires a significant hydrogen
partial pressure driving force to achieve reasonable
permeation rates and high hydrogen recovery. The use of
a sweep gas lowers the partial pressure on the permeate side,
and a countercurrent sweep gas provides the lowest
hydrogen partial pressure at the exit of the reformer bed,
where it is most needed to achieve high recovery (low H2
partial pressure at the exit). The choice of reformer pressure
is a tradeoff in reactor design parameters, with a high
reformer pressure allowing for lower membrane surface
area, but with high air compression costs and possible
154
Fig. 3. Adiabatic ATR membrane reactor fuel processor/fuel cell system process flow diagram.
Fig. 4. Schematic of an adiabatic ATR membrane reactor showing both air distribution tubes and membrane tubes with provisions for countercurrent sweep gas.

mechanical design problems. A reformer pressure of 5 bar
was chosen for this study as the minimum pressure that
allowed high hydrogen recovery in the membrane reactor for
maximum system efficiency.
Steam for this system is generated at two pressure levels.
Low pressure steam at 1.1 bar and 102 8C is generated for
use as sweep gas. High pressure steam at 5 bar and 152 8C is
generated for the ATR feed. The two pressure levels allow
for optimal recovery of waste heat at two different
temperature levels, albeit at an increase in complexity of
the fuel processing system.
As shown in Fig. 3, the methanol feed is mixed with the
high pressure steam and preheated with hot combustor
exhaust gas in exchanger E-2. The preheated methanol/
steam is fed to the shell side of the ATR membrane reactor
containing the CuO/ZnO/Al2O3 catalyst. Air for the ATR
reactor must be compressed to 5 bar in a two-stage
compressor with intercooler, C-1 and E-9. The air is
introduced into the ATR reactor using feed distribution
tubes. These tubes may be made of a porous ceramic or
metal material, or may be metallic tubes having discrete
holes along their length for distribution of the air. The
distributed introduction of air avoids overheating and
sintering of the copper catalyst (discussed further in Section
4.2). The exothermic combustion reaction heats the reaction
mixture, and endothermic steam reforming occurs in the
oxygen-depleted zones of the reactor. The membrane tubes
begin after the air distributor tubes, allowing permeation of
hydrogen along the remaining length of the reactor. Steam
reforming, methanol decomposition, and water gas shift
occur simultaneously with membrane permeation. The
removal of hydrogen from the reaction mixture improves the
thermodynamic driving force for the methanol reactions,
whereas the lower hydrogen partial pressure reduces the
driving force for the undesirable reverse water gas shift
reaction. Finally, the membrane, with sufficiently high
permselectivity, provides an essential purification function.
The permeate leaves the membrane reactor at a pressure
of 1.1 bar and a temperature of about 290 8C. It contains
hydrogen at nearly 80% purity, with the balance consisting
primarily of the sweep steam. The hot permeate is cooled by
preheating boiler feedwater in exchanger E-6. The permeate
exchanges heat with the cathode air stream in exchanger E-5
before entering the anode side of the fuel cell stack. The
water present in the permeate is beneficial to the fuel cell
stack, as humidification helps prevent dry-out of the PEM
membrane [12]. Due to the absence of inert impurities in the
anode feed (e.g. N2, CO2), very high utilizations of hydrogen
can be achieved; 95% utilization is assumed.
The retentate leaves the membrane reactor at a pressure
of 5 bar and a temperature of about 190 8C. The retentate
contains small amounts of CO, unconverted methanol, and
unrecovered hydrogen. This stream mixes with the anode
exhaust, containing a small amount of unused hydrogen, and
the cathode exhaust, containing unused oxygen. This
mixture is fed to the catalytic combustor. The maximum
temperature achieved at steady state in the combustor is
about 360 8C, which is sufficiently low to prevent formation
of nitrogen oxides. During startups and transients, supple-
mental methanol can be fed to the combustor to provide
additional preheat to the ATR reformer.
The hot combustor exhaust is cooled in four stages. First,
the hot exhaust preheats the methanol/steam mixture in
exchanger E-2. Next, high pressure steam is generated in
exchanger E-3, followed by generation of low pressure
steam in exchanger E-4. The exhaust is finally cooled to
about 55 8C in air-cooled exchanger E-7. Condensed water
is recovered in a separator for use as boiler feedwater.
Excess water is purged.
2.5. Process optimization methodology
In Sections 2.6 and 2.7 following, a first pass at
optimization of the process parameters is performed. A
heat and material balance process simulator, HYSYS.Pro-
cess1
v.2.1.1 with the Peng–Robinson equation of state is
utilized. The main focus of this exercise is to optimize the
variables of reaction temperature and steam:carbon ratio in
the reforming reactor, and oxygen:carbon and steam:carbon
in the ATR case. No reaction kinetics are incorporated at this
point. The steam reforming reaction is assumed to reach
thermodynamic equilibrium at the reactor outlet tempera-
ture. The extent of the reverse water gas shift reaction is
kinetically limited; this is adjusted to achieve 1.0% CO in
the reformate; a value considered consistent with the data for
CuO/ZnO/Al2O3 catalyst [13,14]. This somewhat arbitrary
CO concentration is utilized only for optimization of process
parameters; the kinetic model is used in the subsequent
detailed simulations for prediction of reformer outlet
compositions.
The optimum reaction temperature is a function of the
heat integration scheme and the approach temperature in the
heat integration exchangers. The heat integration schemes
described in Sections 2.2–2.4 are considered the simplest
configurations that achieve acceptable fuel reformer
efficiencies (80%). The log-mean temperature difference
(LMTD) for the heat integration exchangers was
fixed so that a fair comparison of efficiencies for each case
was obtained. An arbitrary value of 100 8C was selected as
this fixed value. In each of the cases, all of the steam
produced in the steam generators is used within the process;
no ‘‘excess’’ steam was generated and sent directly to a
condenser.
The process variables determined from the HYSYS
modeling is utilized as an input to the more detailed kinetic
and heat transfer modeling performed in Section 3.
2.6. Process optimization—steam reforming case
The steam reforming process described in Section 2.2 is
optimized with respect to the reformer outlet temperature
and the steam:carbon ratio using the HYSYS model. The

effect of reformer pressure was not explicitly studied in this
case. As will be shown, the reforming reactions benefit from
low pressure. The selected pressure of 3 bar was considered
minimum from the standpoint of achieving reasonable
pressure drops and volumetric flow rates.
For this first pass optimization, as mentioned earlier, we
assume that the steam reforming reaction achieves thermo-
dynamic equilibrium at the reactor outlet temperature. The
water gas shift reaction is not allowed to reach equilibrium,
rather, the extent of the reverse water gas shift reaction is
adjusted to achieve an arbitrary 1 mol% CO in the reformer
exit.
The variables to optimize in the steam reforming case
are steam:carbon ratio, reaction temperature, and supple-
mental fuel fired in the combustor. The results are shown in
Fig. 5a–c. The following iteration technique was used:
(i) Select steam:carbon ratio.
(ii) Adjust reaction temperature to meet 100 8C LMTD in
the heat integration exchangers E-1, E-6 and the
reforming reactor (based on extents of reaction
discussed above).
(iii) Adjust supplemental fuel rate to produce required
steam generation rate.
(iv) Iterate (ii) and (iii) until both LMTD constraints and the
selected steam:carbon ratio is achieved.
Fig. 5a shows the equilibrium methanol conversion and
average reformer temperature as a function of the steam:-
carbon ratio. As the steam production rate is increased, the
average reforming temperature increases in order to main-
tain the 100 8C approach temperature. As reforming tem-
perature and steam:carbon are increased, the equilibrium
methanol conversion increases. For steam:carbon ratios
exceeding 1.1, the equilibrium methanol conversion is
100%.
Fig. 5b shows the supplemental fuel fired as a function of
steam:carbon ratio. The values shown are the percentage of
supplemental fuel in the total methanol fed to the process.
The minimum amount of fuel methanol at low steam:carbon
ratios is about 6% of the total methanol feed. As the
steam:carbon ratio exceeds about 0.9, this percentage
increases.
Fig. 5c shows the overall system efficiency (LHV basis)
as a function of the steam:carbon ratio. Based on the selected
constraints, the maximum efficiency occurs at a steam:car-
bon ratio for the steam reforming + PrOx case of about 0.98.
At lower steam:carbon ratios, the methanol conversion is
too low, resulting in low hydrogen generation efficiency. At
higher steam:carbon ratios, the additional fuel methanol
required to generate that steam reduces the overall system
efficiency. At a steam:carbon ratio of 0.98, the average
reformer temperature is 215 8C (200 8C in, 230 8C exit),
the equilibrium methanol conversion is 96.6%, and the
supplemental methanol fuel in the combustor is 8.2% of the
total methanol feed rate.
156
Fig. 5. (a) Process model results for methanol conversion and average
reaction temperature as a function of H2O:C ratio in the steam reforming
fuel processor. (b) Process model results for the supplemental methanol fuel
required (as a percent of total methanol feed) in the steam reforming reactor
combustor as a function of H2O:C ratio. (c) Process model results for the
overall fuel processor/fuel cell system efficiency as a function of the H2O:C
ratio for the steam reforming fuel processor.

2.7. Process optimization—autothermal reforming case
The HYSYS process model was used for a first pass
optimization of the reaction temperature, steam:carbon
ratio, and oxygen:carbon ratio for the ATR fuel processor.
As in the previous case, the steam reforming reaction is at
thermodynamic equilibrium at the adiabatic reactor outlet
temperature, and the extent of the reverse water gas shift
reaction is adjusted to achieve 1.0 mol% CO. The LMTD of
the heat integration exchangers is fixed at 100 8C.
The effect of increasing pressure was tested in the
HYSYS model. The base pressure selected was 3 bar, but the
optimization of parameters was also tested at 5 bar. As will
be shown, the lower reformer pressure results in a higher
overall system efficiency. The lowest pressure in a practical
fuel processor will be dictated by pressure drop considera-
tions, which was beyond the scope of the present study.
The variables to optimize in the ATR reforming case are
oxygen:carbon ratio, steam:carbon ratio and reaction
temperature. The results are shown in Fig. 6a–d. The
following iteration technique was used:
(i) Select oxygen:carbon ratio.
(ii) Assume steam:carbon ratio.
(iii) Adjust reaction temperature to meet 100 8C LMTD in
the heat integration exchangers E-1, E-2, E-4 and
E-6.
(iv) Determine the amount of steam generated from waste
heat in exchanger E-4. This steam rate sets the
steam:carbon ratio for the next iteration.
(v) Iterate (ii)–(iv) until the LMTD constraint is met at the
specified oxygen:carbon ratio, then select another ratio
and repeat all steps.
In the process flow scheme of Fig. 2, waste heat from the
ATR and PrOx reactors are used to generate the dilution
steam that is fed to the reformer. The amount of waste heat
available for steam generation is primarily a function of the
oxygen:carbon ratio in the ATR feed. As the air rate is
increased, more heat is generated in the ATR and thus more
steam is generated from waste heat. Because of this relat-
ionship, the steam:carbon ratio is not an independent var-
iable, but is a function of the oxygen:carbon ratio. This
relationship is shown in Fig. 6a. Fig. 6b shows the resulting
reaction temperatures as a function of the oxygen:carbon
ratio at a reformer pressure of 3 bar. At low O2:C ratios, the
temperature drops across the adiabatic reactor. This is be-
cause the endothermic steam reforming reaction dominates
the exothermic oxidation reaction. At O2:C ratios greater
than about 0.125, the temperature rises across the reactor as
the exothermic oxidation reaction becomes significant. The
reactor inlet temperature drops as the O2:C ratio increases.
This is because the steam dilution increases with increasing
oxygen (as shown in Fig. 6a), and because the steam is a
relatively cool 133 8C. As shown in Fig. 6b, the average of
the reformer inlet and outlet temperatures goes through a
minimum of 220 8C, which occurs at an O2:C ratio of about
0.125.
Fig. 6c shows the methanol conversion as a function of
O2:C ratio. At low levels of oxygen (air) addition, the low
reactor exit temperatures do not allow complete conversion
of the methanol due to thermodynamic constraints. Above
an O2:C ratio of about 0.125, methanol conversion is ess-
entially complete.
The overall fuel processor/fuel cell system efficiency is
shown in Fig. 6d, which shows an optimum system
efficiency at O2:C ratio of about 0.125. Curves are shown
at reformer pressures of 3 and 5 bar; the lower pressure case
achieves a higher efficiency by about 0.5%. The optimum
oxygen:carbon ratio occurs at about the same value for both
pressure levels. At low values of the O2:C ratio, there is
insufficient temperature and steam to achieve high
equilibrium methanol conversion, which results in reduced
system efficiency. At high values of the O2:C ratio, there is
excess combustion of methanol in the reformer, leading to
reduced hydrogen yields and subsequent reduction of system
efficiency.
It is interesting to note that the optimum system
efficiency occurs at the O2:C ratio that gives the minimum
average reformer temperature and also just gives 100%
methanol conversion. It should also be kept in mind that
the preceding studies are based on idealized process
models that account for thermodynamic equilibrium, but
that do not account for kinetic or other equipment sizing
factors.
3. Kinetics and reactor modeling
3.1. Reaction kinetics
Several kinetic models have been proposed for the steam
reforming of methanol over CuO/ZnO/Al2O3 catalysts. The
following model by Peppley et al. [5] is one of the more
comprehensive models, and has been used in this study
methanol steam reforming ðSRÞ :
CH3OH þ H2O $ CO2 þ 3H2
r000
SR ¼
kSRKCH3OðpCH3OH p3
H2
pCO2
=KSRpH2OÞCS1CS1aSg
ðp0:5
H2
þ KCH3OpCH3OH þ KHCOOpCO2
pH2
þ KOHpH2OÞ
ð1 þ K0:5
H p0:5
H2
Þðmol=kgcat sÞ
(10)
methanol decomposition ðMDÞ :
CH3OH $ CO þ 2H2
r000
MD ¼
kMDKCH3Oð2ÞðpCH3OH p2
H2
pCO=KMDÞCS2CS2aSg
ðp0:5
H2
þKCH3Oð2ÞpCH3OHþKOHð2ÞpH2OÞð1þK0:5
Hð2Þ
p0:5
H2
Þ
ðmol=kgcat sÞ (11)
water gas shift ðWGSÞ : CO þ H2O $ CO2 þ H2

r000
WGS ¼
kWGSKCH3Op0:5
H2
ðpCOpH2OpH2
pCO2
=KWGSÞC2
S1Sg
ðp0:5
H2
þKCH3OpCH3OHþKHCOOpCO2
pH2
þKOHpH2OÞ2
Reitz and coworkers found that the copper must be in
reduced form to be active for the steam reforming reactions
[6]. In the presence of oxygen, the CuO/ZnO/Al2O3 catalysts
will only catalyze the complete combustion of methane to
carbon dioxide and water. They found the following rate
expression to apply under oxidizing conditions
methanol oxidation : CH3OH þ 1:5O2 $ CO2 þ 2H2O
r000
OX ¼ kOX
p0:18
CH3OHp0:18
O2
p0:14
H2O
The parameters for Eqs. (10)–(13) are given in Table 1. KSR,
KMD, KWGS are the equilibrium constants for the SR, MD,
and WGS reactions.
The reaction equilibrium constants were calculated as a
function of temperature using the physical property data and
procedures outlined by Reid et al. [15].
3.2. Catalyst effectiveness factors
The intent of the present study was to simplify the
engineering model of the reforming reactor so that
comparison of various reactor types could be made
without excessive computational requirements. On the
other hand, commercial CuO/ZnO/Al2O3 catalyst pellets
commonly operate in the diffusion-limited regime in low-
temperature water gas shift reactors at 200–220 8C [16].
To address these issues, we solved the reaction and
diffusion problem for 2 mm pellets for the methanol
oxidation and steam reforming conditions. (The details of
this study will be reported elsewhere; a brief summary is
described here.) The effectiveness factors were found
to be primarily functions of temperature; the effect of
composition was weak for the mixture feed composi-
tions and trajectories pertinent to the reaction systems in
this study. As such, empirical effectiveness factor
equations were developed as a function of temperature
for reactions in both the reduced catalyst (reac-
tions (1)–(3)) and oxidized catalyst (reaction (4))
regimes.
158
Fig. 6. (a) Process model results for the optimum H2O:C ratio as a function of the O2:C ratio for the adiabatic ATR fuel processor. Curves are shown at
reforming pressures of 3 and 5 bar. (b) Process model results for the ATR temperatures as a function of the O2:C ratio for the adiabatic ATR fuel processor
operating at 3 bar pressure. (c) Process model results for the methanol conversion in the ATR fuel processor as a function of O2:C ratio. (d) Process model results
for the overall fuel processor/fuel cell system efficiency as a function of the O2:C ratio in the ATR fuel processor. Curves are shown at reforming pressures of 3
and 5 bar.

One must be careful when applying such empirical
relations to reactions with equilibrium limitations. This is
because the effectiveness factor can approach either negative
or positive infinity as compositions cross equilibrium [17].
This arises from the fact that the effectiveness factor is
defined as the ratio of the observed rate normalized by the
rate based on bulk conditions; i.e. robs = hrbulk. When the
bulk gas is at thermodynamic equilibrium with respect to,
say, the WGS reaction, the net WGS reaction rate calculated
at bulk gas conditions is zero. The steam reforming reaction
will cause the conditions inside the catalyst particle to be
non-equilibrium with respect to the WGS reaction, resulting
in a small but finite reaction rate. As a result, the
effectiveness calculated at this condition approaches
infinity. This problem has been avoided by noting that only
two of reactions (1)–(3) are independent. In this approach,
the effectiveness factors for two reactions are calculated by
attributing the formation of products to those two selected
independent reactions. Under reducing conditions we found
that at temperatures below 310 8C, the local species fluxes at
the particle surface can be determined by selecting steam
reforming (1) and water gas shift (3) as the independent
reactions. At temperatures above 310 8C, the methanol
decomposition reaction (2) becomes significant. At these
temperatures, the local species fluxes can be determined by
selecting steam reforming (1) and methanol decomposition
(2) as the independent reactions.
Fig. 7 shows the best-fit regression of the effectiveness
factors as a function of temperature under typical oxidizing
conditions, while Fig. 8 is for typical reducing conditions.
Under oxidizing conditions, the combustion reaction (4) is
the dominant reaction because the oxidized form of the
copper catalyst is ineffective at conducting steam reforming
reactions. The small effectiveness of the steam reforming
reaction is due to the fact that methanol is consumed by
oxidation in an outer shell of the catalyst particle (Fig. 7). In
the oxygen-depleted pellet core the catalyst is active for
steam reforming and decomposition, but much of the
methanol has been consumed in the outer shell. We
determined that the water gas shift reaction is insignificant
and can be ignored. Under reducing conditions the steam
reforming effectiveness is much higher (Fig. 8). Both the
reforming and WGS effectiveness show a similar decreasing
dependence with temperature. Interestingly, at temperatures
higher than 310 8C, the WGS effectiveness factor becomes
negative. This means that the direction of the WGS reaction
inside the pellet is opposite of the direction that would be
calculated based on bulk phase conditions.
3.3. Hydrogen permeability through Pd membrane
The hydrogen flux through a dense Pd membrane is
limited by the diffusion of hydrogen atoms through the
membrane film of sufficient thickness, in which case the flux
can be represented by [18]:
NH2
¼
QH2
d
ðp0:5
H2;r p0:5
H2;pÞ (14)
Table 1
Parameters for kinetic rate expressions for methanol reforming, decom-
position, water gas shift, and oxidation reactions (units are consistent with
pressures in bar and overall rate in mol/kgcat s)
Parameter Expression
Sg (m2
kg1
) 1.02 105
CS1 = CS2 (mol m2
) 7.5 106
CS1a = CS2a (mol m2
) 1.5 105
kSR 7.4 1014
exp(102 800/RT)
kWGS 5.9 1013
exp(87 600/RT)
kMD 3.8 1020
exp(170 000/RT)
kOX 3.6 109
exp(115 000/RT)
KCH3O 6.55 103
exp(20 000/RT)
kOX 4.74 103
exp(20 000/RT)
kH 5.43 106
exp(50 000/RT)
kHCOO 2.30 109
exp(100 000/RT)
KCH3Oð2Þ 36.9 exp(20 000/RT)
kH(2) 3.86 103
exp(50 000/RT)
rb (kg m3
) 1300
Fig. 7. Effectiveness factors for 2 mm catalyst spheres in oxidizing con-
ditions at 3 bar pressure.
Fig. 8. Effectiveness factors for 2 mm catalyst spheres in reducing condi-
tions at 3 bar pressure.

where d is the membrane thickness and the driving force is
proportional to the difference in the square roots of the
hydrogen partial pressures on the retentate and permeate
sides of the membrane (represented by the subscripts r and p,
respectively). We consider the resistance of the porous
support to be negligible. The permeability of hydrogen
through palladium, QH2 (mol m1
s1
Pa0.5
), was taken
from Holleck [19]:
QH2
¼ 4:40 107
exp
15 700
RT

(15)
3.4. One-dimensional reactor modeling
The fuel processor reactor sizes were estimated using the
kinetic expressions integrated with one-dimensional species
and energy balances. The species fluxes in the absence of a
membrane are given by:
dNi
dz
¼ ucatrb
X
j
nijrjhj (16)
where the subscript i refers to the species i and j refers to the
reaction j. The volume fraction of catalyst in the reactor
cross-section is represented by ucat. Formally, the effective-
ness factor of species j, hj, is a function of temperature and
reacting species concentration for reaction j; the empirical
treatment described earlier considers the temperature depen-
dence for oxidizing and reducing conditions (Figs. 7 and 8).
An additional term accounts for hydrogen permeation when
the membrane is present:
dNi
dz
¼ ucatrb
X
j
nijrjhj A
surf
Qi
d
ðp0:5
i;r p0:5
i;p Þ (17)
where A
surf is the membrane specific surface (m2
m3
); the
Qi’s are zero for all species except hydrogen for a defect-free
membrane. The energy balance in the absence of heat
transfer surface (adiabatic operation) is given by:
NgCo
p
dT
dz
¼ ucatrb
X
j
ðDHjÞrjhj (18)
where the heat capacity, Co
p and heats of reactions DHj are
both functions of temperature. In the steam reforming
reactor, a constant wall temperature is assumed, and the
energy balance becomes:
NgCo
p
dT
dz
¼ ucatr b
X
j
ðDHjÞrjhj þ A
surfUoðTw TÞ
(19)
where Uo is the overall heat transfer coefficient, and Tw the
temperature of the wall, which is assumed to be constant and
equal to the well-mixed temperature of combustion gases.
The energy balance on the retentate side of the membrane
reactor is similar to Eq. (19):
NrCo
p
dT
dz
¼ ucatrb
X
j
ðDHjÞrjhj þ A
surfUoðTp TrÞ
(20)
The exception is that the permeate temperature, Tp, is not
constant but a function of position z. The one-dimensional
permeate side energy balance must account for the heat
content of the permeating gas, as well as convective heat
transfer across the membrane surface. The following equa-
tion was derived for the permeate side energy balance:
NpCo
p;p
A
surf
dTp
dz
¼ ðNH2
Co
p;H2
þ UoÞðTp TrÞ (21)
where Np is the axial flux of permeate gas and NH2 is the flux
of hydrogen permeating through the membrane per equation
(15).
Typical one-dimensional heat transfer coefficients in
commercial steam reforming furnaces range from 300 to
500 J m2
s1
K1
at a particle Reynolds number of 104
[20]. The particle Reynolds number of the reforming
reactors in this study are 102
to 103
, which call for a lower
heat transfer coefficient; a value of 200 J m2
s1
K1
was
used for the steam reforming reactor. In the membrane
reactor, the heat transfer across the membrane would be
expected to be lower due to the use of a porous ceramic
support. A value of 40 J m2
s1
K1
was used, as this gave
a good approach temperature between retentate and
permeate while avoiding numerical instabilities.
4. Results
4.1. Steam reforming reactor simulations
The process modeling work detailed in Section 2.5
provided a starting point for optimization of the fuel
processor with the addition of reactor kinetics and heat
transfer. The species and energy balance equations were
integrated for the steam reformer with the use of a fourth
order Runge–Kutta algorithm in MATLAB1
. The
MATLAB1
kinetic model was linked to the HYSYS1
process model for simulation of the remainder of the fuel
processor/fuel cell system. After selection of a set of process
variables, the two models were iterated until convergence
was obtained for feed and effluent conditions. We then
adjusted the reactor size to achieve the maximum efficiency
at the particular conditions.
The first reformer design tested utilized 1.0 cm diameter
heat transfer tubes on a 1.1 cm pitch. The temperature of the
gases on the combustion side of the reformer was simulated
as a constant temperature equal to the adiabatic reaction
temperature after accounting for heat transferred to the
process. In a real combustor design, the combustion would
160

be more plug-flow and a non-uniform temperature profile
would result. The use of a constant combustion temperature
greatly simplified the convergence of the reactor model and
is considered reasonable for the objectives of this study.
There are two key differences between the afore-
described idealized process calculations and the kinetic
simulation. In the idealized model the methanol conversion
was always at equilibrium. An H2O:C ratio was selected, and
the supplemental fuel firing rate was adjusted to generate the
required steam rate. In the kinetic simulation, equilibrium
conversion is not achieved. In these simulations, supple-
mental fuel was not added. Rather, the reactor size was
adjusted to change the methanol conversion. The uncon-
verted methanol in the reformer effluent reacts in the PrOx
reactor to generate additional heat, rather than by supple-
mental firing of methanol in the exhaust combustor.
The second key difference is in the prediction of the CO
content in the reformer effluent. In the idealized model, the
extent of the reverse WGS reaction was set to achieve 1%
CO in the reformer effluent, while reaction kinetics dictate
the CO content in the kinetic simulation. As a result of these
differences, the optimum reformer design and conditions are
different than in the idealized process model. The CO
content and the amount of unconverted methanol in the
reformate have an impact on the steam generation rate in the
PrOx oxidation step, which then affects the overall system
efficiency. The overall methanol feed rate is adjusted as
necessary to achieve 50 kW of net power output.
The results of the combined process/kinetic model at
various H2O:C ratios are shown in Figs. 9–11. Fig. 9 shows
that the overall fuel processor/fuel cell efficiency peaks at an
H2O:C ratio between 1.0 and 1.1. By inclusion of reaction
kinetics, the optimum steam rate has increased over that of
the equilibrium process model shown in Fig. 5c. Fig. 9 also
shows the space velocity dependence, which increases
monotonically as the H2O:C ratio increases. Along this locus
the reactor size corresponds to that giving the methanol
conversion needed to meet the heat balance requirements.
The corresponding methanol conversion increases
(decreases) to the left (right) of the maximum (Fig. 10).
The methanol conversion decreases to the right of the
maximum as unconverted fuel is utilized for additional
energy to generate the steam. So while the reactor size
decreases as H2O:C increases, the efficiency decreases due
to the increased energy needs to generate steam. Fig. 10 also
shows the CO content decreases monotonically as the
H2O:C ratio increases, due primarily to the more favorable
water gas shift equilibrium. Fig. 11 shows the reformer
temperatures to be relatively independent of H2O:C, with a
typical reformer inlet temperature of 210 8C, an exit
temperature of 270 8C, and a wall temperature of 279 8C.
The model predicts that production of CO in the methanol
steam reformer is limited by kinetics, not equilibrium; i.e.,
CO content in the reformate depends upon the temperature
and space velocity. High reforming temperatures or low
space velocities result in additional CO formation. At low
steam rates, the CO content is high because the space
velocity is low. The low space velocity is required to achieve
a reasonable methanol conversion at the low steam rate. The
high CO content at low space velocity also suggests that the
CO content will increase as the fuel processor is turned down
to low rates in any methanol fuel processor.
Unlike the results of the ideal process model, the
methanol conversion decreases with increasing H2O:C ratios
above 1.1 (as shown in Fig. 10). The lower conversions are
necessary in order to provide sufficient heat production in
the PrOx reactor to generate the necessary steam require-
ment. Based on this study, it appears that the optimum steam
reformer H2O:C ratio is about 1.1, as this ratio provides the
maximum system efficiency with the smallest reactor size.
The closely spaced small diameter tubes of the first
design provide for a relatively small catalyst requirement in
the shell side of the reformer. It was found, however, that the
Fig. 9. Overall system efficiency and GHSV for the combined process/
kinetic model of the fired steam reformer system as a function of H2O:C
ratio. The steam reformer utilizes 1.0 cm diameter heat transfer tubes on a
1.1 cm triangular pitch. The GHSV is based on the volume occupied by the
catalyst on the shell side of the reactor.
Fig. 10. Carbon monoxide at reformer outlet and methanol conversion for
the combined process/kinetic model of the fired steam reformer system as a
function of H2O:C ratio.

overall reformer size could be reduced with larger diameter
tubes and a greater tube spacing. This design increases the
fraction of catalyst in the reactor (from 0.28 to 0.49) and
decreases the heat transfer surface (from 300 to 91 m2
m3
).
The improved design utilizes 2.54 cm diameter tubes on a
3.18 cm triangular pitch spacing. The detailed reactor
simulation results for this design are shown in Figs. 12 and
13. Fig. 12 shows the temperature profile within the
reformer, with an inlet temperature of 205 8C, an exit
temperature of 272 8C, and a combustion temperature of
279 8C. Fig. 13 shows the partial pressure profiles of each
species along the length of the reformer. Table 2 compares
the results of the two tubular steam reformer designs. The
second design achieves a smaller overall reformer volume
with fewer heat transfer tubes, but requires a larger volume
of catalyst.
The overall system design parameters and results of
the integrated kinetic and process models are shown in
Tables 3–5. The steam reforming fuel processor design with
2.54 cm tubes was utilized in these tables with an H2O:C
ratio of 1.1. The overall system efficiency for this steam
reformer fuel processor/fuel cell system is 50.4% on a LHV
basis.
4.2. Autothermal reformer with preferential oxidation
The process model study described in Section 2.6 showed
that the optimum oxygen:carbon molar ratio is about 0.125,
and that the optimum steam:carbon ratio is based on the
maximum amount of steam that can be generated with waste
heat. This guidance was used for the kinetic modeling of the
adiabatic ATR reactor. In the first pass reactor design, the air
was mixed with the steam and methanol at the entrance to
the ATR reactor. The temperature and partial pressure
profiles for this case at 94% conversion are shown in Fig. 14a
and b. As seen in these figures, there is a ‘‘hot spot’’
generated of about 370 8C from the oxidation reaction,
which is much faster than the steam reforming reactions.
This high temperature is unacceptable for two reasons: (i)
the copper catalyst will sinter at this temperature [13], and
(ii) the reaction kinetics predict a CO content in the
162
Fig. 11. Reformer inlet, outlet, and average temperatures for the combined
process/kinetic model of the fired steam reformer system as a function of
H2O:C ratio.
Fig. 12. Temperature profiles for the steam reforming reactor kinetic
simulation at the ‘‘optimum’’ H2O:C ratio of 1.1. The steam reformer
utilizes 2.43 cm diameter tubes on a 3.18 cm triangular pitch.
Fig. 13. Partial pressure profiles for the steam reforming reactor kinetic
simulation at the ‘‘optimum’’ H2O:C ratio of 1.1. Methanol conversion is
95.9% and the CO concentration at the exit is 0.92 mol%.
Table 2
Comparison of tubular steam reformer design parameters from combined
kinetic/process model
Tube OD (cm)
1.0 2.54
Reactor diameter (cm) 12.0 11.0
Reactor length (m) 1.8 1.8
Number of tubes 108 11
Volume fraction of reactor containing catalyst, ucat 0.25 0.49
Heat transfer surface (m2
m3
) 299.8 91.4
Catalyst volume (l) 5.0 8.4
Total reactor volume (l) 20.1 17.1
GHSV based on catalyst volume (h1
) 4900 2900

reformate of 3.7 mol%, which is much too high for efficient
operation of the fuel processor system.
The ‘‘hot spot’’ problem can be avoided by utilizing
staged air injection into the adiabatic autothermal
reformer. Such a reactor was simulated; the results are
shown in Fig. 15a and b. Air is evenly distributed into the
bed in 12 discrete increments over the first 1.2 m of bed
length. This was accomplished in the model with a series
of finite length adiabatic PFRs with interstage air
addoption. The first two air injections initiate the
exothermic reaction; ignition is achieved on the third air
injection. Once ignition is achieved, the oxygen is quickly
depleted in a steep exotherm. After the oxygen is
consumed, the endothermic reforming reaction occurs
and a slightly less steep endotherm occurs. This pattern
repeats until the last increment of air is injected at 1.2 m
Table 3
Material balance data for 50 kW fuel cell system
Case
1 2 3
Fuel processor description
MeOH steam reforming plus PrOx MeOH ATR plus PrOx MeOH ATR Pd membrane reactor
Flow rates (g mol/h)
MeOH to reformer 525.0 570.0 570.0
MeOH to combustor 41.5 0 0.0
Air to ATR N/A 366.4 366.4
Air to PrOx 184.7 128.7 N/A
Air to fuel cell 6061 6173 6003
Raw reformate 2113 2430 N/A
Clean reformate 2298 2558 N/A
Retentate N/A N/A 1067
Permeate N/A N/A 1677
Dilution steam to reformer 577.5 456.0 456.0
Sweep steam N/A N/A 350.0
Exhaust 6638 7461 7112
Excess water 457.5 51.2 107.3
Stream name Raw reformate Raw reformate Retentate
Temperature (8C) 272.2 265.4 217.7
Stream compositions (mol%)
Methanol 0.93 0.54 2.37
Hydrogen 70.84 62.80 12.36
Water 4.32 1.11 7.08
Nitrogen 0.00 12.16 27.13
Oxygen 0.00 0.00 0.00
Carbon monoxide 0.91 0.57 0.97
Carbon dioxide 23.01 22.82 50.07
Stream name Clean reformate Clean reformate Permeate
Temperature (8C) 209.6 202.1 285.7
Methanol 0.00 0.00 0.00
Hydrogen 63.50 59.63 79.13
Water 5.66 2.11 20.87
Nitrogen 6.35 15.52 0.00
Oxygen 0.00 0.00 0.00
Carbon monoxide 20 ppm 20 ppm 0.00
Stream name Exhaust Exhaust Exhaust
Temperature (8C) 48 55 55
Methanol 0.00 0.00 0.00
Hydrogen 0.00 0.00 0.00
Water 10.18 14.37 14.37
Nitrogen 74.33 70.68 70.77
Oxygen 6.96 7.15 6.85
Carbon monoxide 0.00 0.00 0.00

into the bed. In the remaining 0.8 m of bed length, the
reforming and water gas shift reactions occur. The final
methanol conversion is 97.4% and the CO content of the
reformate is 0.6 mol%. The peak temperature achieved in
this reformer is about 320 8C, but this temperature is very
localized and apparently does not contribute to excessive
CO production in the reformate. Because the average
reforming temperature is lower in the distributed air case, a
33% larger reactor volume is required to achieve the same
methanol conversion.
The staged air injection autothermal reformer kinetic
model was integrated into the process model with an O2:C
ratio of 0.135. The maximum amount of steam generation
within the design constraints achieved an H2O:C ratio of
0.80. The results of the integrated fuel processor/fuel cell
system are shown in Tables 3–5. The overall system
164
Table 4
Design of reactors for 50 kW fuel cell system
Case
1 2 3
Reforming reactor parameters
O2:C ratio in feed N/A 0.135 0.135
H2O:C ratio in feed 1.1 0.8 0.8
Feed temperature (8C) 204.8 200.0 197.0
Maximum temperature (8C) 272.2 325 310
Pressure (bar) 3.0 3.0 5.0
Reactor diameter (cm) 11.0 7.6 8.9
Reactor length (m) 1.80 2.00 2.04
Methanol conversion (%) 96.3 97.7 95.6
Outlet H2 partial pressure (bar) 2.12 1.88 0.62
Reactor volume (l) 17.11 9.1 12.7
GHSV (based on cat volume) (h1
) 2900 4000 9900
Membrane parameters
Tube OD (cm) N/A N/A 1.0
Number tubes N/A N/A 59
Tube length (m) N/A N/A 0.79
Tube pitch (cm) N/A N/A 1.1
Tube layout N/A N/A Triangular
Membrane surface (m2) N/A N/A 1.47
Fraction catalyst in cross-section N/A N/A 0.25
Membrane material N/A N/A Pd
Membrane thickness (mm) N/A N/A 10
Palladium mass (g) N/A N/A 177
Permeate parameters
H2O:permeate H2 ratio N/A N/A 0.26
Sweep inlet temperature (8C) N/A N/A 102.3
Permeate pressure (bar) N/A N/A 1.1
Permeate outlet temperature (8C) N/A N/A 285.7
Flow direction N/A N/A Countercurrent
H2 recovery across membrane N/A N/A 91.0%
PrOx parameters
Air stoichiometry 2.0 2.0 2.0
Selectivity (%) 50 50 50
Pressure (bar) 3.0 3.0 3.0
Temperature (8C) 200 200 200
Fuel cell parameters
Air stoichiometry 2.0 2.0 2.0
Operating pressure (bar) 1.1 1.1 1.1
Cell voltage 0.75 0.75 0.75
LHV efficiency (%) 60 60 60
H2 utilization (%) 85 85 95
Gross power (MJ/h) 186 187.9 182.7
Net power (MJ/h) 183 183.8 178.4
Overall system efficiency (LHV basis) (%) 50.4 50.3 48.8

Table 5
Heat exchanger data for 50 kW fuel cell system
Case
1 2 3
Exchanger E-1
Duty (MJ/h) 29.56 4.83 N/A
Stream Methanol Methanol N/A
Inlet temperature (8C) 25.0 25 N/A
Outlet temperature (8C) 250.0 97 N/A
Stream Exhaust Exhaust N/A
Inlet temperature (8C) 278.6 204 N/A
Outlet temperature (8C) 158.4 184.9 N/A
LMTD (8C) 68.1 131.7 N/A
Exchanger E-2
Duty (MJ/h) 34.17 38.8 29.18
Stream FC exhaust + MeOH Methanol + steam Methanol + steam
Inlet temperature (8C) 278.6 116 131.8
Outlet temperature (8C) 280.3 200 210.0
Stream Reformate Exhaust Exhaust
Inlet temperature (8C) 204.8 299 363.7
Outlet temperature (8C) 272.2 204 252.9
LMTD (8C) 27.6 93.1 136.7
Exchanger E-3
Duty (MJ/h) N/A N/A 16.27
Stream N/A N/A Exhaust
Inlet temperature (8C) N/A N/A 252.9
Outlet temperature (8C) N/A N/A 189.8
Stream N/A N/A Water/stream
Inlet temperature (8C) N/A N/A 151.8
Outlet temperature (8C) N/A N/A 151.8
LMTD (8C) N/A N/A 64.4
Exchanger E-4
Duty (MJ/h) 22.73 17.95 11.97
Stream Water/stream Water/stream Exhaust
Inlet temperature (8C) 133.5 133.5 189.8
Outlet temperature (8C) 133.5 133.5 142.8
Stream Reformate Reformate Water/stream
LMTD (8C) 185.0 96.8 61.0
Exchanger E-6
Duty (MJ/h) 3.9 4.8 8.38
Stream Reformate Reformate Permeate
Outlet temperature (8C) 157.9 168.2 122
Stream Boiler feedwater Boiler feedwater Boiler feedwater
Inlet temperature (8C) 48 55 55
LMTD (8C) 92.0 89.0 96.6
Exchanger E-7
Duty (MJ/h) 71.34 54.17 61.34
Stream Exhaust Exhaust Exhaust
Outlet temperature (8C) 48 55 55
Exchanger E-8
Duty (MJ/h) 123.8 126.0 116.2
Stream FC cooling water FC cooling water FC cooling water
Inlet temperature (8C) 85 85 85
Outlet temperature (8C) 75 75 75

efficiency for this steam reformer fuel processor/fuel cell
system is 50.3% on a lower heating value basis.
4.3. Autothermal reformer membrane reactor
The membrane reactor integrates a dense palladium
membrane into the reformer for selective hydrogen removal
as the reforming reactions proceed. The goal is to improve
the driving force for the steam reforming reaction by
removal of hydrogen. A schematic of the ATR membrane
reactor is shown in Fig. 4. Staged injection of air was utilized
to avoid excessive hot spots and consequent CO production,
just as in the conventional ATR reactor case. Membrane
surface was introduced just downstream of the air injection
tubes. Tube diameters of 1.0 cm on a 1.1 cm triangular pitch
were utilized for the membrane support, with a Pd thickness
of 10 mm. Countercurrent steam sweep is introduced on the
permeate side of the membranes to allow high hydrogen
recovery.
The results of the kinetic model simulation of this
reactor are shown in Fig. 16a–c. The temperature profile on
the retentate side of the reformer shown in Fig. 16a is similar
to the profile in the staged air injection of Fig. 15a. In the
membrane zone, however, heat transfer to the sweep
steam cools the retentate significantly. The exit temperature
of the retentate on the membrane reactor is 218 8C. This low
temperature slows down the reaction kinetics. Thus, the
beneficial aspect of hydrogen removal by the membrane
is at least partially offset by the low reforming temperatures
caused by cooling from steam sweep in the adiabatic
166
Fig. 14. (a) Temperature profile for the ATR reactor kinetic simulation with
an O2:C ratio of 0.135 and an H2O:C ratio of 0.80. All of the air is introduced
at the bed entrance. (b) Partial pressure profiles for the ATR reactor kinetic
simulation with an O2:C ratio of 0.135 and an H2O:C ratio of 0.80. All of the
air is introduced at the bed entrance. Methanol conversion is 94% and the
CO concentration at the exit is 3.7 mol%.
Fig. 15. (a) Temperature profile for the ATR reactor kinetic simulation with
staged air injection, with an overall O2:C ratio of 0.135 and an H2O:C ratio
of 0.80. (b) Partial pressure profiles for the ATR reactor kinetic simulation
with staged air injection with an overall O2:C ratio of 0.135 and an H2O:C
ratio of 0.80. Methanol conversion is 97.4% and the CO concentration at the
exit is 0.7 mol%.

methanol reforming membrane reactor. The final methanol
conversion is 95.6% and the CO content of the
retentate is 1.0 mol%. The hydrogen recovery across
the membrane is 91.0%. The hydrogen purity of the
permeate is 79.1%, with the balance consisting of the sweep
steam.
The staged air injection ATR membrane reactor kinetic
model was integrated into the process model with an O2:C
ratio of 0.135. The maximum amount of high pressure steam
generation within the design constraints achieved an H2O:C
ratio of 0.80. The remainder of the waste heat was utilized to
generate low pressure steam for the sweep gas. This sweep
steam rate is equivalent to an H2O:C ratio of 0.61, making
the total amount of steam generated per mole of carbon
fed equal to 1.41. The detailed results of the integrated
fuel processor/fuel cell system are shown in Tables 3–5.
The overall system efficiency for this ATR membrane
reactor fuel processor/fuel cell system is 48.8% on a lower
heating value basis.
5. Discussion
The overall system efficiency calculation is directly
dependent upon the assumed fuel cell efficiency. It is more
useful to compare fuel processor efficiencies between
different designs. One definition for fuel processor
efficiency that is frequently used is to look at the heating
value of the hydrogen produced in the reformer divided by
the heating value of the feed. This definition is not useful,
however, in systems where the fuel processor is highly
integrated with the fuel cell system. For example, the fuel
Fig. 16. (a) Temperature profile for the ATR membrane reactor kinetic simulation with staged air injection and countercurrent steam sweep in the permeate. The
overall O2:C ratio is 0.135 and the H2O:C ratio is 0.80. (b) Partial pressure profiles on the retentate side of the ATR membrane reactor with staged air injection
and countercurrent steam sweep in the permeate. The methanol conversion is 95.6% and the CO content of the retentate effluent is 1.0 mol%. (c) Partial pressure
profiles on the permeate side of the ATR membrane reactor with staged air injection and countercurrent steam sweep in the permeate. The hydrogen recovery
across the membrane is 91.0% and the hydrogen purity in the permeate is 79.1%.

cell anode exhaust gases are burned in a combustor, and this
heat is then used in the fuel processor. This heat integration
is not accounted for in the common definition. It is more
valuable in this case to use the following definition for fuel
processor efficiency
hfp ¼
hnet
hfc
(22)
where hfp is the efficiency of the fuel processor and hfc is the
efficiency of the fuel cell. The net system efficiency, hnet, is
the net power production divided by the lower heating value
of the feed. This definition for fuel processor efficiency takes
into account all of the heat integration with the fuel cell
exhaust. The results of the three fuel processor/fuel cell
configurations are summarized in Table 6. Shown are the
fuel processor efficiencies and volume requirements.
It is interesting to note that the SR and ATR systems each
had essentially identical efficiencies. The implication here is
that, provided that each system is optimized on a consistent
basis, that one method of hydrogen production is not
inherently more efficient than the other. In both cases, the
heat of reaction for the steam reforming reaction is provided
by combustion of fuel. In the steam reformer, this combus-
tion occurs external to the reactor, while in autothermal
reforming, the combustion is internal. The ATR system has
the advantage of a simpler reactor design, while the SR
system has the advantage of a higher hydrogen concentration
in the reformate.
The ATR membrane reactor has a slightly lower
efficiency than the SR or ATR cases. We believe this is
due to the additional steam generation required for the sweep
gas, compared to the steam generated for the SR or ATR
cases. The main advantage for the membrane reactor is a
reduction in fuel processor volume due to elimination of the
PrOx reaction step. Note that the reactor volume for the ATR
membrane reactor is slightly larger than the straight ATR
reactor. This is due to the volume occupied by the membrane
tubes, as well as due to the lower average temperature in the
membrane reactor. The lower temperature requires a larger
catalyst volume to achieve the needed conversion. The lower
temperature is primarily caused by the cooling effect of the
steam sweep. This could be overcome by superheating
the sweep steam upstream of the membrane reactor, but at
the expense of another heat exchanger.
The ATR membrane reactor accomplishes hydrogen
generation and purification in a single unit. We estimate that
177 g Pd are needed for the 50 kW processor. The Pd cost
obviously has to be factored into an overall comparison of
the two designs.
Each design utilizes a condenser and separator on the
combined exhaust stream to recover water for re-use in the
steam reforming and sweep gas steps. The use of methanol
as a fuel results in an excess of water in the exhaust. This is
in contrast to the use of hydrocarbon fuels, where less water
is produced and the temperature and pressure of the exhaust
condenser must be carefully controlled to ensure adequate
water recovery [4].
One final observation can be made regarding the CO
content of the reformate product. The exit CO levels in the
SR and ATR reactors are below the thermodynamic
equilibrium value. This means that the production of CO
is kinetically limited. Increasing temperature or contact
time with the catalyst will increase the CO content of the
reformate. The temperature can be controlled by air
addition, however the contact time is a function of the fuel
cell load requirement. As the reformer is turned down, the
increased contact time will result in higher CO levels and
lower resulting system efficiency. This, of course, would not
be an issue in the ATR membrane reactor.
6. Conclusions
There are a variety of viable reactor design options
available for the reforming of methanol to produce hydrogen
for PEM fuel cells. A steam reformer design that uses
catalytic combustion of fuel to supply heat for the
endothermic reaction in a tubular shell and tube reactor
can achieve a fuel processor efficiency of about 84%. The
combustion fuel includes the unused hydrogen from the
fuel cell anode exhaust, as well as some methanol as
supplemental fuel. An adiabatic autothermal reformer uses
in situ combustion with air in the fixed bed reformer to
provide heat for the steam reforming reaction. The full
amount of air cannot be introduced with the feed, as the
resulting exotherm will cause sintering of the copper-based
catalyst, and the high temperature will cause unacceptably
high levels of CO in the reformate. Distributed air injection
along the length of the adiabatic bed overcomes this
problem. Unused hydrogen from the fuel cell anode exhaust
is combusted with unused oxygen in the cathode exhaust to
provide preheat for the ATR reactor. This system is also able
168
Table 6
Fuel processor efficiencies and volume requirements for various fuel processor designs of combined fuel processor/fuel cell systems for 50 kW net power
production
Case Overall LHV
efficiency (%)
Fuel processor
efficiency, Eq. (22) (%)
Fuel processor sizes (l)
Reformer PrOx Total
(1) Methanol steam reformer w/PrOx 50.4 83.9 17.1 12.3 29.4
(2) Methanol ATR w/PrOx 50.3 83.8 9.1 12.3 21.4
(3) Methanol ATR membrane reactor 48.8 81.3 12.7 0 12.7

to achieve a fuel processor efficiency of about 84%. In both
the SR and ATR reactors, the effluent contains up to about
1% CO, which is removed by oxidizing to CO2 in a
preferential oxidation reactor.
An adiabatic ATR Pd membrane reactor can provide
nearly pure hydrogen for the fuel cell (a mild methanation
step may be needed to remove CO leakage through the
membrane defects). The use of a countercurrent steam
sweep on the permeate side of the Pd membrane allows for a
hydrogen recovery of 91% at a moderate reformer pressure
of 5 bar. The presence of steam in the hydrogen feed to the
PEM fuel cell is beneficial to the polymer membrane as it
helps keep it hydrated. The lack of CO2 or other impurities in
the hydrogen allows for higher hydrogen utilization in the
fuel cell. Selective removal of hydrogen from the reformer
by the Pd membrane does not reduce the reformer volume
relative to the straight ATR reactor. Although the thermo-
dynamic driving force for the steam reforming reaction is
increased by the presence of the membrane, the benefit is
minimal as this reaction is not limited by thermodynamics.
In addition, the injection of sweep steam provides unwanted
cooling of the reformer. The net result is an increase in the
reformer reactor volume for the Pd membrane reactor
relative to the straight ATR reactor. The Pd membrane
reactor does achieve an overall reduction in fuel processor
volume, however, by elimination of the PrOx reaction step.
The fuel processor efficiency of the ATR Pd membrane
reactor is slightly lower than the SR or straight ATR reactors.
This is attributed to the larger amount of steam required to
supply both the reforming reactor and the permeate sweep
steam.
The reduction in fuel processor volume achieved with
the Pd membrane reactor may have attractive implications
with dynamic performance (e.g. startup time). On the other
hand, the cost and stability of Pd-based membranes must
obviously be considered.
Acknowledgement
We acknowledge the partial support of this research by
ACS Petroleum Research Fund (ACS-PRF #37053-AC9).
References
[1] C.E. Thomas, G.D. James, F.D. Lomax, J. Ira, F. Kuhn, Int. J.
Hydrogen Energy 25 (2000) 551–567.
[2] P.J. Berlowitz, C.P. Darnell, in: Proceedings of the SAE Meeting,
2000, pp. 8–18.
[3] J.R. Rostrup-Nielsen, Phys. Chem. Chem. Phys. 3 (2001) 283–288.
[4] J.R. Lattner, M.P. Harold, Int. J. Hydrogen Energy 29 (2004) 393–417.
[5] B.A. Peppley, J.C. Amphlett, L.M. Kearns, R.F. Mann, Appl. Catal. A:
General 179 (1999) 31–49.
[6] T.L. Reitz, S. Ahmed, M. Krumpelt, R. Kumar, H.H. Kung, J. Mol.
Catal. A: Chemical 162 (2000) 275–285.
[7] E.D. Doss, R. Kumar, R.K. Ahluwalia, M. Krumpelt, J. Power Sources
102 (2001) 1–15.
[8] J. Larminie, A. Dicks, Fuel Cell Systems Explained, Wiley, Chiche-
ster, 2000.
[9] J.C. Amphlett, R.F. Mann, B.A. Peppley, Int. J. Hydrogen Energy 21
(1996) 673–678.
[10] J.H. Hirschenhofer, D.B. Stauffer, R.R. Engleman, M.G. Klett, Fuel
Cell Handbook, 4th ed. B/T Books, 1993.
[11] R.K. Ahluwalia, E.D. Doss, R. Kumar, J. Power Sources 117 (2003)
45–60.
[12] F. Barbir, PEM fuel cell stack design considerations, in: Proceedings
of the AIChE Spring Meeting, New Orleans, 2002.
[13] J. Agrell, H. Birgersson, M. Boutonnet, J. Power Sources 106 (2002)
249–257.
[14] K. Geissler, E. Newson, F. Vogel, T.-B. Truong, P. Hottinger, A.
Wokaun, Phys. Chem. Chem. Phys. 3 (2001) 289–293.
[15] R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases and
Liquids, 4th ed. McGraw-Hill, New York, 1987.
[16] G. Petrini, P. Schneider, Chem. Eng. Sci. 39 (1984) 637–641.
[17] A.M. De Groote, G.F. Froment, Rev. Chem. Eng. 11 (1995) 145–183.
[18] K. Jarosch, H.I. de Lasa, Ind. Eng. Chem. Res. 40 (2001) 5391–5397.
[19] G.L. Holleck, J. Phys. Chem. 74 (1970) 503–511.
[20] J.R. Rostrup-Nielsen, Catalytic steam reforming, in: Catalysis:
Science and Technology, Springer-Verlag, Berlin, 1984.

comparison of reformer.pdf

Recommended

Recommended

More Related Content

Similar to comparison of reformer.pdf

Similar to comparison of reformer.pdf (20)

More from ZeenathulFaridaAbdul1

More from ZeenathulFaridaAbdul1 (11)

Recently uploaded

Recently uploaded (20)

comparison of reformer.pdf