international journal of hydrogen energy 33 (2008) 7062–7073

Available at www.sciencedirect.com

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

Optimal design of an experimental methanol fuel reformer

Chih-Chiang Chuanga, Yih-Hang Chenb, Jeffrey D. Warda, Cheng-Ching Yua,*,

Yen-Chun Liuc, Chiou-Hwang Leec
a
Department of Chemical Engineering, National Taiwan University, Taipei 106-07, Taiwan
b
Energy and Environment Research Laboratories, Industrial Technology Research Institute, Hsinchu 310, Taiwan
c
Material and Chemical Research Laboratories, Industrial Technology Research Institute, Hsinchu 310, Taiwan

article info abstract

Article history: We report on the steady state modeling of an experimental methanol fuel reformer for fuel
Received 4 July 2008 cell applications. The fuel reformer consists of an AutoThermal Reformer (ATR) followed by
Received in revised form an Oxygen Removal (OR) reactor, Steam Reformer (SR) and Water Gas Shift (WGS) reactor.
19 August 2008 The effluent from the WGS is fed to a series of three Preferential Oxidation (PROX) reactors
Accepted 19 August 2008 that reduce the CO concentration to less than 40 ppm. A mathematical model of the
Available online 21 October 2008 reformer is developed and selected parameters of the model are fit to experimental data
collected from a fuel reformer that was designed, built and operated by the Material and
Chemical Research Laboratories (MCL) of the Industrial Technology Research Institute
(ITRI) in Hsinchu, Taiwan. In order to develop a compact and high-performance fuel
reformer system, the mathematical model is used to design a reformer that has the
minimum possible combined volume of the steam reformer and water gas shift reactor.
The result is that the volume of the optimized reactor units can be reduced by 17.2%
without a significant change in the overall efficiency.
ª 2008 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights
reserved.

1. Introduction review is given by Song [2]. Further background information

is available in Refs. [3,4].
As a method of power generation, fuel cells have the There are a number of considerations in the design of fuel
potential to provide higher efficiency and therefore also reforming systems [1]. Carbon monoxide is a poison for
reduced emissions compared to internal combustion engines polymer electrolyte membrane (PEM) fuel cells, thus if this
[1]. Fuel cells have been designed that can directly oxidize kind of fuel cell is used, the carbon monoxide concentration
fuels such as methane, methanol, dimethylether, etc. must be reduced to a very low level, typically less than
However, the greatest efficiency is achieved when elemental 100 ppm. For transportation applications in particular but for
hydrogen is oxidized directly at the anode. However, as stationary applications as well, it is desirable to make the fuel
a fuel, hydrogen is difficult to store and transport. Therefore, reforming system as small and lightweight as possible.
in recent decades, considerable effort has been made in the Because many fuel reforming reactions take place at high
design of fuel reforming systems that can produce temperatures, it is desirable to employ heat integration in
a hydrogen-rich stream from a fuel such as methanol, order to improve the efficiency of the reforming process.
ethanol, dimethyl ether or hydrocarbon fuels. A detailed Finally, it is desirable that the fuel reforming system start up

* Corresponding author.
E-mail address: ccyu@ntu.edu.tw (C.-C. Yu).
0360-3199/$ – see front matter ª 2008 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2008.08.045
 international journal of hydrogen energy 33 (2008) 7062–7073 7063

Nomenclature Wcat;n catalyst weight of each lump (kg)

yi mole fraction of component i
A reactor cross-section area (m2)
z axial direction in a reactor
CP;g heat capacity of gas (J/(mol K))
CP;S heat capacity of solid (J/(kg K)) Greek letters
F molar flow rate (mole/min) rcat molar density of catalyst (mol/m3 of support)
h heat transfer coefficient (J/(min m2 K)) rg molar density of gas (mol/m3 of gas)
DHrxn;j reaction heat of jth reaction (J/mol) rS density of solid (kg/m3 of support)
kcond thermal conductivity of solid (J/(min m K)) nij stoichiometry coefficient of ith component in jth
Qloss heat loss per unit reactor volume(J/min/m3) reaction
rj jth reaction (mol/kg/min) 3 void fraction of the monolith (m3/m3 of reactor
S geometric surface area per unit reactor volume volume)
(m2/m3)
subscripts
Ta temperature of ambient (K)
i index of component
Tg temperature of gas (K)
j index of reaction
TS temperature of solid (K)
n index of lump in the reactor
U overall heat transfer coefficient (J/(m2 min K))
S solid
Vn reactor volume of each lump (m3)
g gas

and achieve steady state quickly, and respond quickly to gas shift reactor. Finally, in Section 6 conclusions are
changes in hydrogen demand. presented.
Many fuels have been proposed for reforming to produce
hydrogen, including methane, methanol, dimethyl ether,
ethanol and even gasoline. Among these, methanol has the 2. Experimental
advantages that it has a relatively high hydrogen-carbon ratio,
it is a liquid at room temperature, and it can be formed either 2.1. Experimental set-up
from natural gas or from biomass. A number of authors have
studied methanol fuel reforming for fuel cell applications, The experimental fuel processor was designed and installed in
including Refs. [5–15]. the facility of MCL of ITRI. The entire fuel processor setup is
In this contribution, we report on the steady-state composed of several reactors, heat exchangers, and a pre-
modeling and design optimization of a 1.5 kW methanol fuel heating system (Regenerative Thermal Oxidizer (RTO)). A
reformer built and operated by the Material and Chemical process flow diagram for the fuel reformer system is shown in
Research Laboratories (MCL) of the Industrial Technology Fig. 1. The feeds enter the system at room temperature. The
Research Institute (ITRI) in Hsinchu, Taiwan. The fuel methanol, water and air feed flow rates can be adjusted by
reformer consists of feed preheating unit, autothermal pumps and a blower. The methanol and water mixture first
reformer, oxygen removal reactor, steam reformer, high- passes through the RTO which uses a catalyst to burn residual
temperature water gas shift reactor, and preferential oxida- hydrogen in the fuel cell exhaust gas or off-spec reformate gas
tion reactors. The experimental set-up and analysis are to heat the feed stream and vaporize the water and methanol.
similar to work previously reported [16] for a methane fuel The vapor mixture then passes into the autothermal
reformer system. reformer. The reformer is made of stainless steel 304 with
According to a US Department of Energy report [17], the good heat transfer capability. Monolith catalyst carrier is
goals for future PEM fuel cell performance and power gener- placed inside the reformer and the catalyst is coated on
ation per volume of a feed reformer are 75%(LHV) and 2 kW/L a ceramic support. The catalyst used here is a Pt/CeO2–ZrO2
by 2015. Reducing the volume of fuel reforming systems is an catalyst developed by MCL of ITRI. Adding Pt on the catalyst
important step toward meeting these goals. Therefore, in this surface increases the thermal resistance of the catalyst.
work an optimized design with a minimized fuel reformer The hot reformer effluent is cooled down using a heat
volume is also developed. exchanger and direct water injection before entering the
This manuscript is organized as follows: In Section 2, the oxygen removal reactor. The purpose of the oxygen removal
experimental setup is described. In Section 3 each reactor unit reactor is to remove the oxygen from the outlet stream of the
is discussed, and the mathematical model that is used to autothermal reformer, because oxygen will poison the
describe each unit is given, including the chemical reactions commercial Cu–ZnO/Al2O3 (MDC-3) catalyst in the steam
that are assumed to take place and the rate equations that reformer.
describe them. In Section 4, experimental data from the MCL After passing through the oxygen removal reactor, the
fuel reformer is given, and the method used to adjust model reformed fuel gas enters the steam reformer reactor.
parameters and the results of the fitting are described. In Following the steam reformer is the water gas shift reactor.
Section 5, it is shown how the mathematical model was used Water injection between these two reactors is employed to
to design an optimized process that has the minimum cool the fuel gas. The operating temperature of the high-
possible combined volume of the steam reformer and water temperature water gas shift reactor (HTS) is 310  C and the
7064 international journal of hydrogen energy 33 (2008) 7062–7073

Fig. 1 – Detailed process diagram showing all reactor units, pipes, valves and temperature measurements.

catalyst is a Pt/mixed-oxide catalyst developed by MCL of ITRI (3) collect data to develop a mathematical model of the
and coated on honeycomb catalyst carriers. process.
The final step in the fuel reforming process is preferential The experiment is operated in two stages: (1) heat-up and
oxidation (PROX). Oxygen is introduced before the fuel gas (2) autothermal reforming.
enters the PROX reactors. A catalyst containing 3 wt% Pt The start-up procedure of the methanol steam reformer is
prepared by MCL of ITRI catalyzes the PROX reactions. All as follows:
dimensions and catalyst properties of all reactors are shown Methanol and water are mixed in the pre-mixing tank and
in Tables 1 and 2. pumped into the autothermal reformer. In front of the auto-
thermal reformer, an igniter system is used to preheat the
2.2. Experimental runs and results methanol and water by using nickel–chromium filament. The
igniter system starts to heat the nickel–chromium filament in
The objectives of this experimental setup are to (1) test the the first 3 min. Then, methanol solution and air (oxygen) are
reactivity and properties of the catalysts developed by MCL of fed into the autothermal reformer to carry out the combustion
ITRI; (2) obtain operating parameters to achieve the desired reaction. A large quantity of fuel is consumed to heat up the
hydrogen flow rate and carbon monoxide concentration; and catalyst and monolith support of the reformer. During startup,
the reformate gas is sent to the RTO unit and combusted to
release energy. When the temperature of RTO is satisfactory,
Table 1 – Reactor dimensions the igniter system is turned off. The RTO instead of igniter
becomes the feed preheating unit. When the concentration of
ATR Oxygen SR HTS PROX1 PROX2 PROX3
CO in outlet stream of the PROX reactor reaches the ppm level,
remover
the reformate gas is fed into PEM Fuel cell system. Figs. 2 and 3
Height 4 1 3 10 9 9 9 show the results of typical experimental runs, including both
(cm) startup and steady-state reforming operation. Fig. 2 shows the
Outer diameter 8 8 8 11.8 7.6 7.6 7.6
temperatures profiles for different sectors of the fuel processor.
(cm)
Fig. 3 gives the concentration profiles for CO and hydrogen.
 international journal of hydrogen energy 33 (2008) 7062–7073 7065

Table 2 – Catalyst properties

ATR Oxygen SR HTS PROX1 PROX2 PROX3
remover

Type of support Ceramic Ceramic Tablet Ceramic Metal Metal Metal

monolith monolith monolith monolith monolith monolith
400 cell/in2 400 cell/in2 j6.4  3.2 mm 400 cell/in2 300 cell/in2 300 cell/in2 300 cell/in2
Volume of support (ml) 201.1 50.3 151.2 1093.6 364.7 364.7 364.7
Catalyst weight (g) 39.0 16.5 204.6 186.5 74.4 70.4 68.8

Qloss is a term that accounts for loss of thermal energy to the

3. Reactor models environment. Qloss is assumed to be negligible for every unit
except the ATR, that is all units except the ATR are well
A process flow diagram for the fuel reformer system is shown insulated and therefore assumed to be adiabatic. For the ATR,
in Fig. 1. All reactor units are modeled as plug flow reactors Qloss is calculated according to:
(PFRs). At steady state, the models consist of coupled 4U
nonlinear ordinary differential equations that describe the Qloss ¼ ðTS  Ta Þ (4)
D
change in gas phase molar flow rate and reactor temperature
where U is the overall heat transfer coefficient and D is the
down the length of the reactor. All units except the ATR are
outside diameter of the ATR.
well insulated; therefore, it is assumed that they are adiabatic.
Eqs. (1)–(3) apply to every reactor in the fuel processor;
It is further assumed that the ideal gas equation of state is
however, a different set of reactions are assumed to take place
valid, and that the pressure drop across the reactor units is
in each reactor. Table 3 summarizes the reactions that take
negligible. For the gas phase flow rates, a material balance on
place in each reactor, together with the enthalpy of the reac-
each species gives:
tion, the kinetic expression used, and the literature reference
1 dFi X
¼ ð1  3Þ nij rj rcat (1) for the kinetic expression.
A dz j In the autothermal reactor three reactions may take
place: methanol is reversibly converted to carbon monoxide
where A, F, n, 3, r and rcat are the reactor cross-sectional area,
and hydrogen, carbon monoxide and water can convert
species molar flow rate, stoichiometry coefficient, void frac-
reversibly to carbon dioxide and hydrogen (the water gas
tion, reaction rate and catalyst density, respectively. The
shift reaction), and methanol can be oxidized to form
subscripts i and j refer to species and reactions, respectively.
carbon dioxide and water. In the steam reforming reactor,
For the temperature of the gas in the reactor:
! only the water gas shift reaction and the reversible
1 d X   conversion of methanol to carbon monoxide and hydrogen
Fi CP;i Tg ¼ hS TS  Tg (2)
A dz i
can occur. In the WGS reactor only the WGS reaction can
occur. Finally, in the PROX reactor, carbon monoxide is
where h is the heat transfer coefficient between the catalyst
catalytically oxidized to form carbon dioxide, and some
surface and the gas phase, S is a geometric factor equal to the
hydrogen is oxidized as well. Tables 4–6 summarize the
catalyst surface area per unit catalyst volume, and TS and Tg
literature values of the kinetic and absorption parameters
are the temperature of the catalyst surface and the gas in the
for the reactions.
reactor, respectively.
Cooling between reactor units is accomplished either by
For the temperature of the solid catalyst:
heat exchange with cooling water or direct cold water injec-
kcond d2 TS X    tion. Where direct injection of water was employed, the
 ¼ ð1  3Þ DHrxn;j rj rcat  hS TS  Tg  Qloss (3)
A dz2 j

800 10 100
ATR[Inlet] SR[inlet] 700
700 9 SR(CO Conc.) 90
HTS[Intlet] PrOX1[Inlet] 600 PrOX3(CO Conc.)
8 80
600 PrOX2[Intlet] PrOX3[Intlet]
Rxn. Temp.(°°C)

500 7 PrOX3(H2 Conc.) 70

PrOX3[Outlet]
CO Conc.%

500
H2 Conc.%

6 60
400
400 5 50
300 300 4 40

200 3 30
200
2 20
100 100
1 10
0 0 0 0
0:00 0:04 0:08 0:12 0:17 0:21 0:25 0:30 0:34 0:38 0:43 0:47 0:51 0:00 0:04 0:08 0:12 0:17 0:21 0:25 0:30 0:34 0:38 0:43 0:47 0:51
time (hr) time (hr)

Fig. 2 – Temperature versus time at various points in the Fig. 3 – Concentration versus time at various points in the
fuel reformer system. fuel reformer system.
7066 international journal of hydrogen energy 33 (2008) 7062–7073

ATR SR HTS PrOX1 PrOX2 PrOX3

500 4. Experimental data and parameter fitting

The fuel reformer was operated at steady state with the feed
400 conditions given in Table 7. The temperature and composition
of the gas stream exiting each reactor was measured. Table 8
Temp (°C)

shows the steady-state inlet and outlet temperature and

300 outlet composition for each reactor unit. In order to match the
experimental data to the mathematical model, certain model
parameters were adjusted iteratively according to the
200 following procedure:

(1) Initial values are supplied for the parameters U and kcond
100 and the parameter group (hS ).
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 (2) Adjust certain kinetic parameters in order to match the steady-
Length (m) state reactor effluent compositions as closely as possible.
(3) Using the values of the kinetic parameters from (2),
Fig. 4 – Temperature profile indicated by the mathematical
calculate the temperature at the reactor effluent and
model (solid line) and experimental data (open circles).
compare with the experimental results. If there is signifi-
cant deviation, return to (1) and adjust the parameters U
and kcond and the parameter group (hS ). If there is no
following energy balance is used to calculate the outlet significant deviation, proceed to (4).
temperature: (4) Adjust capacity parameters 3rg, 3rgCp,s,(1  3)rsCp,s in order
0 to match the dynamic response of the process. If a match
X Z Tin
B vap is possible, then the process parameters have been
Fi;out CP;i dT ¼ FH2 O;in  @CP;H2 OðlÞ ð373298ÞþDHH2 O
i Tout appropriately identified. If not, return to (1) and adjust U
1 and kcond and the parameter group (hS ).
Z Tin
C
þ CP;H2 OðgÞ ðTÞdTA (5)
T373 Table 9 shows the values of the physical property param-
eters that were identified by this procedure. Tables 10 and 11
where the term on the left hand side represents the temper- show the modified kinetic parameters, together with their
ature change of the hydrogen-rich gas stream and the right literature values.
hand side represents the temperature increase and vapor- Although this paper discusses only a steady-state process
ization of the injected water. model, in step (4) some of the parameters of the steady-state

Table 3 – Reactions and kinetics

Reaction DH298 kJ/mol Kinetics Reference Note

Autothermal reactor (ATR)

Ka;2 ðkD PCH3 OH  kD;B P2H2 PCO ÞCS2 CS2a Sg
CH3 OH%CO þ 2H2 90.5 rMD ¼ [5–7]
ðP0:5 0:5 0:5
H2 þ Ka;2 PCH3 OH þ Kb;2 PH2 O Þð1 þ Kc;2 PH2 Þ
CO þ H2 O%CO2 þ H2 41.2 Kb;1 ðkW PCO PH2 O  kW;B PH2 PCO2 ÞC2S1 Sg [5–7]
rW ¼ 2
0:5 þ K P
ðPH 2
a;1 CH3 OH þ Kd PCO2 PH2 þ Kb;1 PH2 O Þ
a
CH3 OH þ 1:5O2 /CO2 þ 2H2 O 675.4 rOX ¼ kOX PCH3 OH P1:5
O2 –
Oxygen removal
reactor (OR)
a
CH3 OH þ 1:5O2 /CO2 þ 2H2 O 675.4 rOX ¼ kOX PCH3 OH P1:5
O2 –
Steam reforming
reactor (SR)
Ka;2 ðkD PCH3 OH  kD;B P2H2 PCO ÞCS2 CS2a Sg
CH3 OH%CO þ 2H2 90.5 rMD ¼ [5–7]
ðP0:5 0:5 0:5
H2 þ Ka;2 PCH3 OH þ Kb;2 PH2 O Þð1 þ Kc;2 PH2 Þ
CO þ H2 O%CO2 þ H2 41.2 Kb;1 ðkW PCO PH2 O  kW;B PH2 PCO2 ÞC2S1 Sg [5–7]
rW ¼
0:5
ðPH 2
þ Ka;1 PCH3 OH þ Kd PCO2 PH2 þ Kb;1 PH2 O Þ2
Water gas shift (HTS)  
P 2 PH2
CO þ H2 O%CO2 þ H2 41.2 rWGS ¼ kWGS PCO PH2 O  CO
KWGS [18]
Preferential Oxidation (PROX)
kCO PCO PO2
CO þ 12O2 /CO2 283 rCO ¼ [19]
ð1 þ KCO PCO Þ2
1 0:5 b
H2 þ 2O2 /H2 O 242 rH 2 ¼ kH2 PH 2
[20]

a The form of the kinetic rate expression for methanol oxidation was developed by assuming a quasi-elemental reaction. The activation energy
for the reaction is taken from the literature [11]. The pre-exponential factor was determined by regression of experimental data.
b The form of this kinetic rate expression is modified from [20].
 international journal of hydrogen energy 33 (2008) 7062–7073 7067

Table 4 – Literature values of kinetic parameters for the Table 6 – Literature values of kinetic parameters for the
autothermal reactor (ATR), steam reforming reactor, and WGS and PROX reactors
oxygen removal reactor Reaction Pre-exponential factor A0 Activation energy
Kinetic Pre-exponential Activation (rate constant) EA (KJ/mol)
parameter factor a0 energy EA
kWGS 4.933  106 mol/(g min bar2) 47.4
(rate constant) (kJ/mol)
kWGS 1.312  102 (dimensionless) 38.06
kD 2.28  1022 mol/(g min bar) 170 kCO 2.15  1016 mol/(g min bar2) 75.41
kD;B 8.57  1010 mol/(g min bar2) 80 KCO 1.248  103 bar1 3.13
kW 3.54  1015 mol/(g min bar) 87.6 kH2 1.232  103 mol/(g min bar0.5) 18.74
kW;B 5.42  1017 mol/(g min bar) 128.8
kOX – 115

concentration of oxygen rapidly drops almost to zero (Fig. 5).

model were determined by matching the dynamic response of This is necessary because any oxygen present at this point in
the process with a dynamic model. Appendix A gives addi- the gas stream will poison the copper catalyst in the water gas
tional details about the procedure that was used. If a satis- shift reactor.
factory match between the calculated and experimentally In the steam reforming and water gas shift reactors, the
determined trajectories was not achieved, then the numerical temperature rises gradually (Fig. 4) due to the exothermic
values of the physical property parameters were again WGS reaction. The concentrations of hydrogen and carbon
adjusted and the procedure repeated. Table 12 shows the dioxide gradually rise, while the concentration of CO gradually
values of the capacity parameters that were identified using falls, which is also consistent with the occurrence of the water
dynamic data. gas shift reaction (Figs. 5,6).
Figs. 4–6 show the steady state temperature and compo- Before each of the PROX reactors, the gas stream is cooled
sition profiles in the reactor as suggested by the mathematical in a heat exchanger and oxygen is injected into the reactor.
model, along with the experimental data points. Fig. 4 shows The temperature rises in the reactor because the oxidation
the temperature profile, Fig. 5 shows the mole fraction of reactions are exothermic. Because concentrations are
methanol (left axis) and oxygen (right axis) on logarithmic expressed on a mole fraction basis, there are discontinuities in
scales, and Fig. 6 shows the composition of CO2 and H2 (left the mole fraction of all species when air is injected into the
axis, linear scale) and CO (right axis, logarithmic scale). Along system, because the air injection increases the total number
the top of each figure are labels that identify each reactor in of moles of gas in the stream. In each PROX reactor, the
the process. Note that as shown in Fig. 1 there is water concentration of CO decreases and the concentration of CO2
injection after the oxygen removal and steam reforming increases, while there is a very slight decrease in the hydrogen
reactors, and inter-stage cooling and oxygen injection before concentration.
each of the PROX reactors.
In the autothermal reactor, the temperature rises rapidly
as the highly exothermic oxidation reaction (ROX) occurs first, 5. Design optimization
and then decreases as the endothermic molecular decompo-
sition reaction (RMD) and reverse water gas shift reaction (RW) 5.1. Optimization of the fuel reformer volume
proceed. Some thermal energy is also lost to the environment.
In the ATR, the concentration of methanol and oxygen drop One key consideration in the design of fuel reformers,
rapidly (Fig. 5), while the concentration of CO2, H2 and CO particularly for transportation applications, is the weight and
rapidly rise (Fig. 6), consistent with the occurrence of all three volume of the reformer components. Therefore, after
reactions. a mathematical model of the process was developed, a design
In the oxygen removal reactor, the temperature rises due to optimization was conducted, with the objective of minimizing
the highly exothermal oxidation reaction (Fig. 4), while the the combined volume of the fuel reformer reactor units. Table
2 shows the volume of each reactor unit. The largest contri-
butions to the total volume come from the HTS and PROX
reactors (48% and 42% of the total volume, respectively). Our
Table 5 – Literature values of adsorption parameters for experience indicates that it is difficult to reduce PROX reactor
the autothermal reactor (ATR), steam reforming reactor,
and oxygen removal reactor
Parameter Pre-exponential Adsorption DH0
factor A0 (kJ/mol)
(adsorption constant)
Table 7 – Steady state values of feed flow rates into the
Ka,1 6.65  103 bar0.5 20 ATR
Ka,2 3.69  101 bar0.5 20
Feed Feed Feed
Kb,1 4.74  103 bar0.5 20
CH3OH H2O O2
Kb,2 3.69  101 bar0.5 20
(mol/min) (mol/min) (mol/min)
Kc 5.43  106 bar0.5 50
Kd 2.3  109 bar2 100 0.575 0.92 0.115
7068 international journal of hydrogen energy 33 (2008) 7062–7073

Table 8 – Steady state composition data

CO% (dry base) CO2% (dry base) H2% (dry base) CH3OH Conv O2 Conv (%) Tin ( C) Tout ( C)
(%) (dry base)

ATR 6.2 17.04 53.64 94 90 423 451

Pt/Cz75R – – – – 100 – –
SR 3.87 18.8 55.22 – – 365 –
HTS 1.01 20.88 56.44 – – 310 143
PROX3 37 (ppm) 20.91 51.59 – – 143 170

volume without violating the CO constraint; however, it is

possible to reduce the volume of the HTS reactor.
Because the SR and HTS reactors work together to reduce
Table 9 – Physical property parameters the CO concentration in the ATR outlet, we attempted to find
Parameter Value Units

Kcond 1.0 J/(m min K)

ATR SR HTS PrOX1 PrOX2 PrOX3
U 200 J/(min m2 K) 50 10
8
hS 1.5  105 J/(min m3 K) 40
6
4
2
30

20
YCH3OH (%)

YO
2
10

(%)
Table 10 – Modified kinetic parameters for the ATR
Pre-exponential factor Literature Modified
(rate constant)

kD mol/(g min) 2.28  1022 1.37  1018 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
kW mol/(g min bar) 3.54  1015 1.18  1019
Length (m)
kW;B mol/(g min bar) 5.42  1017 5.42  1016
kOX mol/(g min bar2.5) – 6.83  106 Fig. 5 – Composition profile for methanol and oxygen in the
fuel reformer system. Solid lines represent the
mathematical model, and open circles represent
experimental data.

Table 11 – Modified kinetic parameters for the SR, WGS ATR SR HTS PrOX1 PrOX2 PrOX3
60
and PROX reactors
Pre-exponential factor Literature Modified 6.2
50
(rate constant) 3.87
YCO2 (%) & YH2 (%)

SR: kD mol/(g min) 2.28  1022 2.3  1019 40 1.01

SRkW mol/(g min bar) 3.54  1015 2.61  1015
YCO (%)

HTSkWGS mol/(g min bar2) 4.933  106 3.09  106 30

PROXkCO mol/(g min bar2) 2.15  1016 1.89  1016
PROXkH2 mol/(g min bar0.5) 1.232  103 6.16  104 20

10

0 0.0037
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
Length (m)
Table 12 – Mass and energy capacity parameters
Fig. 6 – Composition profile for hydrogen, carbon dioxide,
Parameter Value Units
and carbon monoxide in the fuel reformer system. Lines
3 0.8928 represent the mathematical model, and open circles
rg 300.0 mol/m3 represent experimental data. For CO2 and H2
rS 1550 mol/m3 concentrations, which are read on the left axis, the solid
CP;g 30 J/(mol K)
line represents H2 concentration and the dashed line
CP;S 1000 J/(mol K)
represents CO2 concentration.
 international journal of hydrogen energy 33 (2008) 7062–7073 7069

a b 7
c
375

FCH3OH,SR (mol/min)
6 0.06

YCO,SR (%)
TSR (°C)
5
370 0.05
4
0.04
3

365 2 0.03
0 1 2 0 1 2 0 1 2
Wcat (Kg) Wcat (Kg) Wcat (Kg)

Fig. 7 – Steam reformer properties versus catalyst mass (a) Temperature, (b) CO mole fraction, and (c) methanol flow rate.

the design that minimized the combined volume of these two HTS inlet. The difference is necessary to allow for proper
units, subject to certain constraints. The independent vari- automatic temperature control (by water injection). The third
ables for the optimization were the inlet temperature to those constraint is a maximum reactor temperature constraint due
reactor units. Note that the volumes of the SR and HTS can be to the properties of the HTS Cu–ZnO/Al2O3 catalyst. The fourth
adjusted independently in the optimization. constraint specifies the maximum allowable CO concentra-
The objective can be stated mathematically as: tion at the HTS outlet. When the CO concentration at the exit
Minimize : f ðXÞ ¼ ðVSR þ VHTS Þ (6) of the HTS is below this level, it can be reduced to the PPM
level by the PROX reactors.
  The second constraint is not as intuitive as the others, but
X ¼ TSR;in ; THTS;in (7)
it is necessary in order to ensure that the optimization
8 9 produces a reasonable process design. The reason for
>
>
> TSR;out  THTS;in  10 C >
>
> including the second constraint can be understood in light of
>
> >
>
< dTSR;out = Fig. 7, which shows several steam reformer properties plotted
¼0
Subject to : dW cat (8)
>
> THTS;out < 350 C >
> versus catalyst mass (i.e. position in the plug flow reactor).
>
> >
>
>
: YCO;HTS ¼ 1:01% >
; The methanol concentration (Fig 7(c)), which is already low,
drops continuously throughout the reactor because of the
It is worthwhile to discuss the implications of the constraints methanol decomposition reaction. The reactor temperature
given in Eq. (8) and the reasons that they are included in the (Fig 7(a)) rises at first because the exothermic water gas shift
optimization. The first constraint specifies a minimum reaction dominates, consuming carbon monoxide (Fig 7(b)).
difference in the temperature between the SR outlet and the Farther down the reactor, this reaction becomes less

a 1500
b 1000

800
VHTS (cm3)

1000
VSR (cm3)

600

400
500
200

0 0
280 300 320 340 360 380 280 300 320 340 360 380

c 4
d 380

360
TSR,max (°C)

3
YCO,SR (%)

340

2
320

1 300
280 300 320 340 360 380 280 300 320 340 360 380
TSR,in (°C) TSR,in (°C)

Fig. 8 – System properties as a function of steam reformer inlet temperature calculated using process model. (a) SR volume,
(b) HTS volume, (c) mole fraction of CO in SR outlet, and (d) maximum temperature in SR.
7070 international journal of hydrogen energy 33 (2008) 7062–7073

a 1250 1.33
b

1200 1.32

VSR + VHTS (cm3)

FH2,HTS (mol/min)
1150 1.31

1100 1.3

1050 1.29

1000 1.28
280 300 320 340 360 380 280 300 320 340 360 380
TSR,in (°C) TSR,in (°C)

Fig. 9 – Combined volume of the steam reforming reactor and the HTS reactor (a) and HTS outlet hydrogen flow rate (b)
versus steam reformer inlet temperature.

important and the endothermic methanol decomposition this case. However, the higher temperature is also disadvan-
reaction, which produces carbon monoxide, becomes more tageous from a thermodynamic point of view: the equilibrium
important. As a result, the concentration of carbon monoxide CO concentration is greater at higher temperatures (Fig 8(c)).
in the steam reformer drops quickly initially but then As a result, the mole fraction of CO at the SR outlet increases
plateaus. A practical design will include a steam reformer with increasing temperature, as does the required volume of
sized so that the CO concentration is beginning to plateau at the HTS (Fig 8(b)). The higher SR reactor temperature will
the exit of the reactor. This outcome can be guaranteed by cause a higher CO concentration in the SR outlet stream. This
requiring that the derivative of reactor temperature with means that a larger HTS reactor volume will be needed to
respect to catalyst mass at the exit of the reactor be equal to meet the CO concentration specifications in the HTS outlet
zero. stream. Therefore, there is a tradeoff in the reactor network
The optimization problem was solved by iterating over the design and a minimum in the combined reactor volume.
two optimization variables. For a given value of TSR,in, the Fig. 9(a) shows the combined volume of the steam reformer
value of THTS,in that minimized the combined volume was and HTS reactor versus the inlet temperature of the steam
identified. The procedure was repeated for various values of
TSR,in until the global minimum was identified.
Fig. 8 shows certain properties of the fuel reformer system
Table 14 – Literature values of kinetic parameters for the
versus TSR,in with the optimal (volume-minimizing) value of
methanation reaction
THTS,in. As the SR inlet temperature increases, the SR volume
Component Pre-exponential factor Adsorption
decreases (Fig 8(a)). The maximum temperature of the SR also
A0 (adsorption constant) DH0 ðKJ=molÞ
increases with increasing inlet temperature (Fig 8(d)). The
higher temperature is favorable from a kinetic point of view kf mol/(min bar0.5) 7.02  1016 240.1
and the WGS reaction approaches equilibrium more quickly in kr mol/(min bar1.5) 5.862  103 67.1
CH4 (bar1) 6.65  104 38.3
CO (bar1) 8.23  105 70.7
H2 (bar1) 6.12  109 82.9
Table 13 – Reaction and kinetics for methanation reactor H2O (bar1) 1.77  105 88.7

Reaction Kinetics Source

CH4 þH2 O%CO þ 3H2 kf PCH4 PH2 O  kr PCO P3H2 [21]

rM ¼
P2:5
H2 ðaÞ
2 Table 15 – Modified values of kinetic parameters for the
 methanation reaction
a¼ 1 þ KCO PCO þ KH2 PH2 Pre-exponential factor Literature Modified

PH O (rate constant)
þKCH4 PCH4 þ KH2 O 2
PH2
kf mol/(min bar0.5) 7.02  1016 7.02  1016
kr mol/(min bar1.5) 5.862  103 4.66  104
 international journal of hydrogen energy 33 (2008) 7062–7073 7071

Fig. 10 – Combined volume of the steam reforming reactor and the methanation reactor (a) and HTS outlet hydrogen flow
rate (b) versus steam reformer inlet temperature.

 
reformer. The optimal design is marked with a triangle and X ¼ TSR;in ; TM;in (10)
the experimental (base case) design is marked with a circle. 8 9
The minimum combined volume occurs at a temperature of >
>
> TSR;out  TM;in  10  C >
>
>
>
> >
>
317  C and a combined volume of 1031 cm3, as compared with < dTSR;out =
¼0
Subject to : dW (11)
the initial design at a temperature of 365  C and 1247 cm3, for >
>
cat
TM;out < 500  C >
>
>
> >
>
a reduction in volume of 17.2%. The steam reformer inlet >
: YCO;HTS ¼ 1:01% >
;
temperature in the optimal design is 317  C.
It is important to verify that the reduction in reformer Fig. 10(a) shows the combined volume of the steam
size does not come at the expense of a significant reduction reforming and methanation reactors versus SR inlet temper-
in the hydrogen production rate (i.e. the efficiency of the ature for the optimal value of the methanation inlet temper-
reformer). Fig. 9(b) shows the hydrogen flow rate at the exit ature. In contrast with Fig. 9(a), the combined volume
of the HTS versus the steam reformer inlet temperature. decreases with increasing temperature over the entire
Compared with the base-case design, the hydrogen produc- temperature range 280–370  C. For temperatures beyond
tion rate of the volume-optimized system is reduced by only 370  C it is impossible to satisfy the constraints. The minimum
0.69%. possible combined volume is just 287 cm3, a reduction of 77%
compared to the base-case system.
5.2. Optimization of the reaction pathway However, from Fig. 10(b) this improvement comes at the
expense of a drastic reduction in the hydrogen flow rate. At
Besides optimizing the process as described above, an alter- the design corresponding to the minimum combined volume,
native way to reduce the fuel reformer volume is to use the hydrogen flow rate is reduced by 19%. The reason for this
a different process design in which the HTS reactor is replaced is that the methanation reaction consumes three moles of
with a methanation reactor. The methanation reactor reduces hydrogen for every mole of CO converted. Although the
the CO concentration by hydrogenating CO to form methane methanation reactor is attractive from the point of view of
and water. Table 13 shows the reaction and the kinetic volume reduction, it results in a prohibitive loss of efficiency
expression that was used to model the reaction rate [21]. of the reformer. Therefore, this process design alternative is
Table 14 shows the literature values of the kinetic parameters not considered any further.
taken from [22]. Table 15 shows the modified kinetic param-
eters that were determined by regressing experimental data
for the methanation reaction from MCL of ITRI. 6. Conclusions
For the alternative process design with the methanation
reactor, the optimization problem can be stated as: In this work, we have reported on the modeling and optimi-
zation of an experimental methanol fuel reforming system for
Minimize : f ðXÞ ¼ ðVSR þ VM Þ (9) fuel cell applications developed by MCL of ITRI. Kinetic
7072 international journal of hydrogen energy 33 (2008) 7062–7073

parameters for all of the catalysts, including a Pt/CeO2–ZrO2   dTS;n kcond 

1  3 rS Vn CP;S ¼ ðTS;nþ1  2TS;n þ TS;n1 Þ  hSVn TS;n
catalyst developed by MCL of ITRI are identified by fitting dt Dz2
experimental data. The experimental system can produce  X 
 Tg;n þ DHrxn;j rn;j Wcat;n  Qloss
a hydrogen stream with a CO concentration less than 40 ppm, j
and with an efficiency of 66.1 percent. The method of model
(17)
development, in which kinetic parameters are taken from the
literature and the forward reaction rates are adjusted to The parameters 3rg, 3rgCp,s, and (1  3)rsCp,s were adjusted so
match experimental data, has the advantages that it is simple that the dynamic trajectories predicted by the model
and can be completed rapidly with minimal experimental equations match the experimental data. The results are
data, thus facilitating rapid product design. shown in Table 12.
The model was also used to develop an optimized design in
which the volume of the reactor units is minimized. The result
of the optimization suggests that the combined volume of the
references
SR and HTS reactors can be reduced by 17.2% with a reduction
in the hydrogen flow rate of only 0.69%.
[1] Ahmed S, Krumpelt M. Hydrogen from hydrocarbon fuels for
fuel cells. International Journal of Hydrogen Energy 2001;26:
291–301.
Acknowledgement [2] Song C. Fuel processing for low-temperature and high-
temperature fuel cells: challenges, and opportunities for
sustainable development in the 21st century. Catalysis
This work was supported by the National Science Council of
Today 2002;77:17–49.
Taiwan under grant NSC 96-2628-E-002-022-MY3.
[3] Larminie J, Dicks A. Fuel cell systems explained. 2nd ed. New
York: John Wiley & Sons; 2003.
[4] Fuel cell handbook. 6th ed. EG&G Services Parsons, Inc.
Appendix A
Science Applications International Corporation.
Determining capacity parameters from dynamic Morgantown, WV: US Dept. of Energy, 2000.
responses [5] Asprey SP, Wojciechowski BW, Pepply BA. Kinetic studies
using temperature scanning: the steam-reforming of
Although this contribution deals only with the development methanol. Appl Catal A Gen 1999;179:51–70.
of a steady state model, some of the parameters of the steady [6] Pepply BA, Amphlett JC, Kearns LM, Mann RF. Methanol–
state model were estimated by fitting the dynamic response of steam reforming on Cu/ZnO/Al2O3. Part 1: The reaction
network. Appl Catal A Gen 1999;179:21–9.
the process to a dynamic process model. This appendix briefly
[7] Pepply BA, Amphlett JC, Kearns LM, Mann RF. Methanol–
discusses the development of the dynamic model and the
steam reforming on Cu/ZnO/Al2O3 catalysts. Part 2. A
fitting of these parameters. comprehensive kinetic model. Appl Catal A Gen 1999;179:
Eqs. (1)–(3) in the body of the paper are steady-state equa- 31–49.
tions. The analogous dynamic equations are partial differen- [8] Chan SH, Wang HM. Thermodynamic and kinetic modeling
tial equations: of an autothermal methanol reformer. J Power Sources 2004;
  126:8–15.
vyi 1 v Fyi X
3rg ¼ þ ð1  3Þ nij rj rcat (12) [9] Idem RO, Bakhshi NN. Kinetic modeling of the production of
vt A vz j hydrogen from the methanol-steam reforming process over
Mn-promoted coprecipitated Cu–Al catalyst. Chem Eng Sci
1996;51(14):3697–708.
 [10] Lattner JR, Harold MP. Autothermal reforming of methanol:
vTg 1 v X  
3rg CP;g ¼ FCP;i Tg þ hS TS  Tg (13) experiments and modeling. Catalysis Today 2007;120:
vt A vz i 78–89.
[11] Lattner JR, Harold MP. Compariosn of methanol-based fuel
processors for PEM fuel cell systems. Appl Catal B Environ
  vTS kcond v2 TS X  2005;56:149–69.
1  3 rS CP;S ¼ þ ð1  3Þ DHrxn;j rj rcat
vt A vz2 [12] Lindstrom B, Pettersson LJ. Hydrogen generation by steam
j
  reforming of methanol over copper-based catalysts for fuel
 hS TS  Tg  Qloss (14) cell applications. Int J Hydrogen Energy 2001;26:923–33.
[13] Bichon P, Asheim A, Jordal A, Sperle T, Fathi M, Holmen A,
These equations are converted into a series of coupled ordi-
et al. Hydrogen from methanol steam-reforming over Cu-
nary differential equations by discretizing the length coordi- based catalysts with and without Pd promotion. Int
nate into a number of finite positions (lumping): J Hydrogen Energy 2007;32:1799–805.
dyn;i X [14] Lee JK, Ko JB, Kim DH. Methanol steam reforming over cu/Zn/
3rg Vn ¼ Fn1 yn1;i  Fn yn;i þ nij rn;j Wcat;n (15) Al2O3 catalyst: kinetics and effectiveness factor. Appl Catal A
dt j
Gen 2004;278:25–35.
[15] Reitz TL, Ahmed S, Krumpelt M, Kumar R, Kung HH.
Characterization of CuO/ZnO under oxidizing conditions for
dTg;n X X the oxidative methanol reforming reaction. J Mol Catal A
3rg Vn CP;g ¼ Fn1 yn1;i CP;n1;i Tg;n1  Fn yn;i CP;n;i Tg;n
dt i i
Chem 2000;162:275–85.
  [16] Lin ST, Chen YH, Yu CC, Liu YC, Lee CH. Modeling of an
þ hSVn Ts;n  Tg;n
experimental methane fuel processor. J Power Sources 2005;
(16) 148:43–53.
 international journal of hydrogen energy 33 (2008) 7062–7073 7073

[17] Multi-Year Research, Development and Demonstration Plan: [20] Choi Y, Stenger HG. Kinetics, simulation and insights for CO
Planned Program Activities for 2005-2015, http://www1.eere. selective oxidation in fuel cell applications. J Power Sources
energy.gov/hydrogenandfuelcells/mypp/index.html. 2004;129:246–54.
accessed 5.30.2008. [21] Xu J, Froment GF. Methane steam reforming, methanation
[18] Choi Y, Stenger HG. Water gas shift reaction kinetics and and water-gas shift. I. Intrinsic kinetics. AIChE J 1989;35(1):
reactor modeling for fuel cell grade hydrogen. J Power 88–96.
Sources 2003;124:432–9. [22] de Smet CRR, de Croon MHJM, Berger RJ, Marin GB,
[19] Venderbosch RH, Prins W, van Swaaij WPM. Platinum Schouten JC. Design of adiabatic fixed-bed reactors for the
catalyzed oxidation of carbon monoxide as a model reaction partial oxidation of methane to synthesis gas. Application to
in mass transfer measurements. Chem Eng Sci 1998;53(19): production of methanol and hydrogen-for-fuel-cells. Chem
3355–66. Eng Sci 2001;56:4849–61.