Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Theoretical Analysis of A Pure Hydrogen Production Separation Plant For Fuel Cells Dynamical Applications

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

International Journal of Hydrogen Energy 31 (2006) 10791090

www.elsevier.com/locate/ijhydene
Theoretical analysis of a pure hydrogen production separation plant
for fuel cells dynamical applications
L. Capobianco
a,
, Z. Del Prete
a
, P. Schiavetti
a
, V. Violante
b
a
Department of Mechanics and Aeronautics, University of Rome La Sapienza, Via Eudossiana 18, 00184 Rome, Italy
b
ENEA, Frascati Research Centre, Via Enrico Fermi 45, 00044 Frascati, Rome, Italy
Received 30 November 2005; received in revised form 17 January 2006; accepted 10 February 2006
Available online 30 March 2006
Abstract
This article proposes a mathematical model and develops the numerical simulation of a single stage hydrogen
productionseparation process during transient behaviour, suited for proton exchange membrane (PEM) fuel cell (FC) applica-
tions. Methanol reforming process is performed in a commercial catalytic membrane reactor (CMR), lled with a commercial
ZnOCuO, alumina supported catalyst. The permeate hydrogen is accumulated in a reservoir volume (buffer) connected to the
permeate side. This conguration was studied in order to avoid the feed back control of the reactor feeding, even when transient
power loads to the cell are applied. By numerical simulation, we veried that the system comprised by the CMR and the PEM,
with an appropriate constant reactor feeding ow, is always self-sustaining so that the hydrogen demand by the FC can be sat-
ised at all power regimes. The achievement of this goal was obtained by redistribution of the hydrogen produced in the reactor
between the buffer and the exhaust tail gases. Only the control of two independent variables of the system, such as reactor
temperature and pressure, are needed, therefore, the conguration proposed here results in a simplied approach to the control
strategy for the entire system. We apply the theoretical analysis to a pilot plant designed and assembled at the University of Rome
La Sapienza, in order to verify its functional parameters and the theoretical performance of the system before its real operation.
2006 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved.
Keywords: Fuel cells; Catalytic membrane reactor; Buffer; Methanol steam reforming
0. Introduction
In the recent years, an extensive work has been
done concerning the use of proton exchange membrane
(PEM) fuel cells (FCs) to clean energy and automo-
tive applications. For the latter one, the main problem
is still the safe and inexpensive storage of hydrogen
on board. Therefore, many authors [17], instead of

Corresponding author. Tel.: +39 06 44585585;


fax: +39 06 4881759.
E-mail address: luigi.capobianco@uniroma1.it (L. Capobianco).
0360-3199/$30.00 2006 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2006.02.013
storage, focused on efcient on board production
methods, by means of liquid and gaseous fuels reactions
as: water gas shift of carbon monoxide, steam reform-
ing of methane, steam reforming coupled by partial
oxidation of a liquid fuel, thermal cracking of hydrocar-
bons. Moreover, the use of PEM FCs implies a number
of problems, like purication and humidication of hy-
drogen feeding and control of gases pressure [812]. In
this study, we examine by numerical simulations, the
performance of an H
2
feeding system constituted by a
catalytic membrane reactor (CMR) [10,1315,33,35],
by which the production and purication steps
are integrated. CMRs are high efciency devices
1080 L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090
Nomenclature
A
m
total permeating membrane area, m
2
c concentration in gas phase, mol/m
3
c
0
reference concentration, (=P
0
/(RT
0
))
C
T
SI
, C
T
SIa
catalyst surface concentration of site
types I or Ia (I = 1, 2), mol/m
2
D reactor diameter, m
D
m
membrane diameter, m
E activation energy, kJ/(mol K)
F molar ow rate, mol/s
F
0
reference molar ow rate,
(=(v
0
c
0
V
b
)/L)

F dimensionless molar ow rate, (=F/F


0
)
J molar ux, mol/m
2
s
J
0
reference molar ux, (=F
0
/A
m
)

J dimensionless molar ux, (=J/J


0
)
k
D
rate constant for methanol
decomposition, m
2
/s mol
k
R
rate constant for methanol
reforming, m
2
/s mol
k
W
rate constant for water gas shift reaction,
m
2
/s mol
K
D
equilibrium constant for methanol
decomposition, 1/bar
2
K
R
equilibrium constant for methanol
reforming, 1/bar
2
K
W
equilibrium constant for water gas shift
K

adsorption constant, 1/bar


0.5
L reactor total length, m
L
m
membranes total length, m
N
m
number of membranes
p
i
partial pressure of species i, bar
p
(rs)
H
2
retentate side partial pressure of hydro-
gen, Pa
p
(ps)
H
2
permeate side partial pressure of hydro-
gen, Pa
P total pressure, bar
P
0
reference pressure, bar
P
c
cell power load, W
Q/o permeability, mol/ms Pa
0.5
r
i
overall rate of consumption of species i,
mol/m
3
cat
s
r
M
rate of reaction M, bar/m
3
cat
s
R gas constant, bar m
3
/mol K
R

gas constant in permeability expression,


J/mol K
R
j
rate of reaction j, mol/m
3
cat
s
S surface area, m
2
/gr
T temperature, K
T
0
reference temperature, K
v molar mean velocity, m/s
v
0
reference velocity, m/s
V dimensionless velocity, (=v/v
0
)
V
r
reactor volume, m
3
V
b
buffer volume, m
3
X dimensionless length, (z/L)
y
i
molar fraction of species i
Y dimensionless molar fraction,
(=y
i
/y
i0
)
z reactors axial coordinate, m
Greek letters
: membrane to reactor length ratio
(=L
m
/L)
[ permeating zone surface to volume
ratio (=(4N
m
D
m
)/(D
2
N
m
D
2
m
))
hydrogen utilization factor
(=F
(perm)
H
2
/(F
(prem)
H
2
+ F
(out)
H
2
))
o membrane thickness, m
H
R
heat of reaction, kJ/mol
H
c
high heating value of hydrogen, kJ/mol
c bed void fraction, m
3
void
/m
3
reactor
p
c
cell efciency
v stoichiometric coefcient
H dimensionless buffer pressure
(=P
b
/P
0
)

i
(=((1 c)/c)r
i
U(z L
1
)([/c)J
H
2
)

t
(=L

N
c
i

i
)
1
i
(=L
i
/(c
0
v
0
y
i0
))
1
t
(=L
t
/(c
0
v
0
))
j
cat
catalyst density, kg/m
3
cat
t dimensionless time (=v
0
t /L)
c frequency, Hz
Superscripts
c cell
in reactor inlet conditions
out reactor outlet conditions
perm permeating
Subscripts
a CH
3
OH
b buffer
d decomposition
m membrane
cat catalyst
0 reference value
r CO
2
, reforming
s H
2
w CO, water gas shift
L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090 1081
because the hydrogen produced by dehydrogenation
reactions permeates through metallic membranes, and
the reaction is moved towards the products beyond the
thermodynamic equilibrium theoretically attainable in
the corresponding closed system [32]. The CMR is cou-
pled with a limited reservoir volume (buffer) designed
to accumulate the permeate hydrogen used to feed the
FC anode. We consider here a new commercial CMR,
inside which methanol steam reforming takes place on
a commercial ZnOCuO, alumina supported catalyst.
In most studies on CMRs, vacuum or sweep gas is uti-
lized to maintain a trans membrane hydrogen concen-
tration gradient, while in other cases, it is assumed that
the permeate hydrogen is maintained at atmospheric or
xed pressure [13]. Those are not really feasible condi-
tions for dynamic applications, when the permeate side
is connected to a FC with variable power load. Since
FC hydrogen consumption during fast load transients is
strongly variable, FC system requires rapid hydrogen
owadjustments. In these cases, the reactor should meet
the FC demand adjusting the reactors fuel feeding (i.e.
by a greater or smaller hydrogen production). The CMR
dynamic transient can be quite slower than that required
for fast load commutation, and, if feasible, needs for a
much more complicated control strategy. We observed
that, by means of an accumulation device (buffer) con-
nected to the permeate side and a constant reactors
feeding ow, this problem can be overlapped. At sta-
tionary conditions, with a xed hydrogen production
rate in the reaction side coupled with the permeation
process, the system automatically set the difference
between the average partial pressure of hydrogen in the
reactor lumen and the pressure in the permeate side
(driving force) in order to accomplish the required
hydrogen ow acrossthe membrane. The use of the
buffer allows to control only the reactors temperature
and pressure, resulting in a simpler and more efcient
control strategy.
1. Experimental plant layout and design
Many authors already focused on the study of opti-
mal layout and design of an on board fuel processor
system coupled with a FC. Although PEM FC seems to
be so far the best solution for automotive power gen-
eration systems, due to low operation temperature and
exibility, various problems arise regarding hydrogen
productionpurication systems and their primary fuel
conversion [812]. Energy efciency is another fun-
damental task to pursue. Therefore, various plant lay-
outs with different energy recovery strategy have been
proposed. In this study, Methanol steam reforming
was considered in order to realize a feasible on board
hydrogen source.
The advantages of using methanol, as primary fuel,
are the following:
easier to transport on board than hydrogen,
no desulphuring is needed compared to traditional
gasolines,
high energy density,
low reaction temperature 200300

C.
A novel strategy to feed FC has been studied here. Us-
ing a membrane reactor, no ne regulation of hydrogen
partial pressure in the retentate side is possible so, in
order to maintain the trans membrane hydrogen con-
centration gradient, a different system conguration in
the permeate side has been realized. Classical methods
foresee the use of a vacuum pump, a sweep gas or back
pressure regulators. A simple buffer accumulation de-
vice on the permeate side has been indeed proposed
here, without any control of its inside pressure.
The basic idea is that, providing an appropriated vol-
ume for the reservoir connected to the permeate side
of the membrane, we completely avoid the necessity
of controlling the reactor feeding, even when transient
power loads are applied to FC. In fact, with a constant
fuel feeding to the reactor, a xed overall hydrogen pro-
duction rate is guaranteed. When FC requests no power,
the buffer can be considered as a closed volume con-
nected to the permeate side of the reactor. The buffer
will be lled by the permeate hydrogen until the pres-
sure inside it equals the average hydrogen partial pres-
sure in the retentate side. When FC is asked to supply
electrical power, a constant hydrogen owis drawn from
the buffer, the pressure drop inside it (with respect to
the retentate side) produces the necessary pressure gra-
dient (driving force) across the membranes to accom-
plish the required hydrogen ow. At the same time, the
outgoing of the permeated hydrogen causes also an hy-
drogen partial pressure drop in the reactor lumen. The
system evolves towards a new stationary state, where
the average pressure difference between the retentate
side and the permeate side is sufcient to accomplish
the required hydrogen ow. The hydrogen produced in
the reactor which exceeds the amount strictly required
by FC, exits the system together with the reactors tail
gases.veri
The buffer is also necessary to feed FC during the
start-up phases or in acceleration phases, when the
needed hydrogen ow would not pass instantaneously
through the membranes. Fig. 1 shows the layout of
1082 L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090
Fig. 1. Plant layout of a pilot power station designed and assembled at Romes University La Sapienza.
a small scale (P = 150 W) pilot plant for on line
hydrogen production, feeding a self-breathing and
self-humidied PEM FC (BCS Fuel Cells, Inc. Model
50-24), building up at the Mechanical Measurements
Laboratory of Romes La Sapienza University. The
reactor consists of a 13 mm inner diameter and an
about 254 mm length stainless steel tube, hosting seven
Pd coated tubular membranes with a high hydrogen
permeability, of about 203 mm length and 1.98 mm
outer diameter (REB Research and Consulting). The
fuel is a methanolwater mixture, stored at atmospheric
temperature and pressure inside a polyethylene tank,
and pumped inside the plant by a positive displace-
ment, high head, alternative pump (MILTON ROY,
LMI HH743-N8P). A boiler, made by a 2 m length
1
8
in
copper tube wrapped around a 315 W cartridge heater
(WATLOW, Firerod

), is used to vaporize and over


heat the reaction mixture before entering the reactor.
External heat needs to be input to the reactor to sustain
the energy consumption by endothermic reactions and
to maintain the temperature of the system near to the
optimal working value. This has to be done in order to
both optimize the reaction kinetics, and maintain a suf-
ciently high membrane permeability. Therefore, a 90 cm
exible and insulated 125 W heater (OMEGA, FGR-
030 Rope Heater) has been wrapped around the stainless
steel tube. Both heaters are provided with their own PID
controllers (WEST, Model 6500 and Model 6100). The
system was operated for maximum power load condi-
tions. It was calculated that for a 150 W power PEM FC,
a hydrogen molar ow rate F
(c)
H
2
= 1.32 10
3
mol/s
would be needed, considering a 40% efciency. The
reactor operating conditions were chosen with the aid
of a test plug ow model computer code, resulting in:
T = 573 K, P = 10 bar, methanol feeding: F
(in)
CH
3
OH
=
6.0 10
4
mol/s.WaterMethanol molar ratio was
chosen to be 1.1 , so we have F
(in)
H
2
O
=6.6 10
4
mol/s
and F
(in)
= 1.26 10
3
mol/s. A commercial DN 40
stainless steel (38.1 mm o.d.) tube with CONFLAT

ange ends with an internal volume of 170 cm


3
(PFEIFFER) as buffer volume is used. A back pressure
L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090 1083
regulator (PARKER VERIFLO, APR1) in the exhaust
line of the reactor is used to control reactors pressure,
and a check valve (PARKER, 1 psig CO Check Valves)
between the permeate side and the buffer is used in or-
der to avoid hydrogen back ow during shut down. Be-
fore actually starting up the system in the lab, our aim
was to verify the system performances and effective-
ness with the hydrogen buffer, by means of numerical
simulation.
2. Mathematical model of the reactorbuffer
system
As already mentioned, membrane reactor combines
the reaction and purication processes within one inte-
grated device, hence we need analytical expressions for
the reaction kinetics and the permeation law across the
membranes to describe its pattern. Those expressions
are presented at the end of this section.
2.1. Mathematical model
In order to develop a mathematical model to anal-
yse the theoretical dynamic behaviour of the system,
we refer to the scheme reported in Fig. 2, and con-
sider a pseudo-homogeneous model [16,17]. The fol-
lowing approximations are introduced for the reactors
modelling:
the reactor temperature is kept uniform, constant and
controlled at T = 573 K,
the ideal gas law applies to the gas phase,
the pressure gradient across the packed bed is ne-
glected,
axial dispersion of gaseous species along the reactor
is neglected,
diffusion limitations across catalyst pellets are ne-
glected.
Uniform and constant temperature prole is not a real-
istic situation in the presence of a strong endothermic
reaction. Nonetheless, we examine here an ideal case
study. In the real case, the presence of a thermal gradi-
ent along the reactor axis involves the decrease of the
reactions kinetic in the points where the temperature is
below the uniform value considered in this work. Our
assumptions imply the examination of the best possi-
ble performances the reactor may reach. Furthermore,
despite the ideal gas law is not obeyed by non-ideal
compounds like methanol and steam at high pressure,
we assume this law is true for the entire gas phase in-
Fig. 2. Schematic view of the reactor + buffer module, with the
coordinate system used for the modelling equations. Membranes are
constituted by hollow tubes with a blank oating head end pointing
the reactors inlet; other ends are brazed together to a single tube
sheet, at the reactor outlet, opened to a common manifold which
collects the permeate hydrogen. Catalyst lls overall reactor internal
volume.
side the reactor in order to simplify the mathematical
modelling, considering that in a wide part of the reac-
tor there is high methanol and water conversion, so the
overall gas phase volumetric behaviour is close to the
ideal gas.
The following notation is used:
r
i
=
N
R

j
v
i,j
R
j
,

i
=
(1 c)
c
r
i
U(z)
[
c
J
i
,

t
=
N
c

i
, (1)
(J
i
, the permeating ux term, is present only if i =
H
2
) where N
c
is the number of components, N
R
the
number of reactions, c is the packed bed void fraction,
v
i,j
is stoichiometric factor of component i in reaction
j, r
i
is the overall production rate of the component
i by all reactions, R
j
the rate of the single reaction,
[ is the permeating area over volume ratio (i.e. [ =
4N
m
D
m
/(D
2
N
m
D
2
m
)) in the permeating zone, and
U(z) is the unit step function dened as
U(z) =
_
0, 0z <L L
m
,
1, L L
m
z <L.
(2)
1084 L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090
The function U(z) is introduced to take into account
that in the rst part of the reactor there is only cat-
alytic bed; in fact, the oating head of the tubular mem-
branes do not reach the reactors inlet. As can be seen in
Fig. 2, the membranes oating head start at z=LL
m
,
while the reactor inlet is at z = 0.
With the above assumptions, the material balances
(on molar basis) of single components lead to
jc
i
jt
=
j(c
i
v)
jz
+
i
, i = 1, . . . , N
c
,
c
(in)
i
(t ) = c
i
(t, 0), (3)
while the total balance leads to
jc
jt
=
j(cv)
jz
+
t
,
J
(in)
(t ) = c
(in)
(t, 0)v
(in)
, (4)
where J
(in)
(t )=F
(in)
/(cS) is the specic molar owrate
of reaction mixture, S =(D
2
)/4 is the cross-sectional
area of reactor inlet, c
(in)
= P
(in)
/RT
(in)
.
In the buffer, considering a perfect mixing, we have
the simple balance equation
V
b
RT
0
dP
b
dt
= N
m
D
m
_
L
0
U(z)J
H
2
dz F
(c)
H
2
(t ), (5)
where F
(c)
H
2
(t ) is the molar ow required by the cell,
1
that can be approximately described by the equation
F
(c)
H
2
(t ) =
P
(c)
(t )
p
c
H
c
, (6)
where P
(c)
(t ) [W] is the time variable cell power load,
H
c
= 284 kJ/mol is the hydrogen high heating value,
and p
c
40% [18,19] is the cell efciency.
Eqs. (3)(5) joined by equation of state
P = cRT (7)
and the necessary kinetics and permeation pro-
cess information, completely solve the problem.
As can be seen, due to hydrogen permeation and
non-equimolar reactions, the continuity equation is
not obeyed, so that supercial uid velocity varies
along the reactor and also not all of Eqs. (3) are
independent of each other, due to the total mo-
lar balance. Combining Eqs. (4) and (7) we obtain
a time-independent differential equation describing
1
We assume here that for FC, a variable load required from
the electrodes instantaneously converts to a variation of hydrogen
consumption.
supercial velocity prole along the reactor axis at all
time instants.
Introducing the following dimension-less variables:
H =
P
b
P
, Y
i
=
y
i
y
i0
, X =
z
L
,
V =
v
v
0
, t =
t v
0
L
, (8)
and the following parameters:
c
0
=
P
0
RT
0
, F
0
=
v
0
c
0
V
b
L
, J
0
=
F
0
A
m
,

J =
J
H
2
J
0
,

F
H
2
=
F
(c)
H
2
F
0
,

O =
T
T
0
, (9)
and by use of following functions:
1
t
=
L
c
0
v
0

t
, 1
i
=
L
c
0
v
0
y
i0

i
, (10)
Eqs. (3) can be put in dimensionless form
jY
i
jt
= V
jY
i
jX

OY
i
1
t
+

O1
i
, i = 1, . . . , N
c
,
Y
(in)
i
(t) = Y
i
(t, 0), (11)
with
jV
jX
=

O1
t
,
V(t, 0) = 1, (12)
and Eq. (5) in the form
dH
dt
=
1
:
_
1
0
U(X)

J dX

F
(c)
H
2
(t), (13)
U(X) =
_
0, 0X<1 :,
1, 1 :X<1
(14)
with : = L
m
/L.
This is a set of partial integral differential equations,
that describes the time evolution of the system variables
during power load transients.
2.2. Reactions rate and hydrogen permeating ux
expressions
It was well studied by several authors [9,12,13,20,21]
that methanol steam reforming proceeds rapidly on
L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090 1085
alumina supported ZnOCuO(i.e. commercial methanol
synthesis catalyst) by the following reactions scheme:
R
r
: CH
3
OH + H
2
OCO
2
+ 3H
2
,
H
R
= +49.51kJ/mol,
R
d
: CH
3
OHCO + 2H
2
, H
R
= +90.70 kJ/mol,
R
w
: CO + H
2
OCO
2
+H
2
, H
R
=41.19kJ/mol.
Reactions R
r
, R
d
are both reversible and endothermic
and proceed under volume increase, so the highest
methanol conversions are obtained at high temperature
and low pressures. Reaction R
w
, the water gas shift
reaction, takes place simultaneously with methanol
steam reforming: it is an exothermic reaction that pro-
ceeds without volume change, so low temperatures are
preferable.
Peppley et al. [20] developed a comprehensive ki-
netic model for a commercial BASF K3-110 CuO/ZnO
catalyst, taking into account heterogeneity of catalysts
sites and thermodynamic equilibrium limitations, and
we use this kinetics model in simulations presented
thereafter. The analytical expressions are reported in the
Appendix.
Hydrogen selective metal membrane are widely
studied. Most works are focused on Pd-based ones
[9,10,1315,21,22,34,36], but refractory metals mem-
branes (vanadium, niobium, tantalum alloys) seem to
be even better, due to high hydrogen permeability,
provided that supercial dissociation limitations are
avoided [2329]. Permeation across dense membranes
is an activation process. If diffusion in the bulk is the
rate limiting step, hydrogen dissociates fast on the
surfaces, so the permeating ux can be expressed as
J
H
2
=
Q
0
o
exp
_

E
R

T
__
_
p
(rs)
H
2

_
p
(ps)
H
2
_
, (15)
where the superscripts (rs) (ps) refer to retentate side
and permeate side, Q(T ) = Q
0
exp(E/R

T ) is the
membrane permeability, and o is the membrane thick-
ness. To relate this expression to the balance equations
we consider p
(rs)
H
2
= p
H
2
and p
(ps)
H
2
= P
b
. For the mem-
branes used, a conservative permeability value (a value
three times greater than the best PalladiumSilver alloy
membrane is declared for REBResearch and Consulting
membranes [30]) of Q/o = 1.5 10
3
mol/sm
2
Pa
0.5
is assumed at T = 573 K.
2.3. Numerical technique
Eqs. (12) were discretized with a backward nite dif-
ference scheme in space, and the resulting set of ODEs,
CO
H
2
V
L, reactor length [m]
V
,

d
i
m
e
n
s
i
o
n
l
e
s
s

s
u
p
e
r
f
i
c
i
a
l

v
e
l
o
c
i
t
y
Y
,

m
o
l
a
r

f
r
a
c
t
i
o
n
2.0
1.6
1.2
0.8
0.4
0.25 0.20 0.15 0.10 0.05 0.00
1.0
0.8
0.6
0.4
0.2
CH
3
OH
H
2
O
CO
2
Fig. 3. Stationary gas phase molar fractions proles along reactor
axis with P
b
=1.2 bar. The second-order discontinuity in the graph
at z = L L
m
is due to the beginning of the permeation process
(oating membrane heads).
were integrated in time with a explicit Euler method by
mean of a FORTRAN code. At each time step, velocity
in every point of the space domain was calculated by
quadrature of Eq. (13) by trapezoidal rule. For the sake
of brevity, analytical details of numerical procedure are
reported in the Appendix.
3. Results and discussion
To analyse the dynamical response of the system,
we consider the system behaviour in three kinds of
situations.
3.1. Constant buffer pressure
In the rst examined case, the hydrogen buffer pres-
sure is kept at a constant value of P
b
= 1.2 bar and
Eq. (15) is not considered. The situation of con-
stant pressure in the permeate side is considered fre-
quently by some authors [13] and, as we will see
later, its representative, for our system, of a con-
stant power regime for FC.
2
Fig. 3 shows stationary
concentration and dimensionless velocity proles of
the gas phase inside the reactor. It is clear that the
rst part of the reactor is strongly effective in the
methanol conversion process by reforming and de-
composition of methanol, while, after methanol is
totally converted, only water gas shift and permeation
processes take place. Moreover, observing the slope
of hydrogen, carbon dioxide and carbon monoxide
2
In this situation, asymptotically, the difference between the
average hydrogen pressure in the reactor side and that in the buffer
is such as to guarantee FC required ow.
1086 L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090
proles, we may conclude that water gas shift proceeds
at a constant rate, sufciently high to reach equilibrium
conditions in all points of the reactor. As a matter of fact,
while hydrogen content reduces along the reactor axis
because of continuum subtraction of hydrogen due to
permeation, carbon monoxide and carbon dioxide mo-
lar fraction increases, while the methanol one decreases
as a consequence of the equilibrium shifting effect that
is characteristic for a CMR [12,15,22]. Moreover, as
expected, after an increase due to hydrogen production
in the rst part of reactor, supercial velocity (i.e. the
volumetric ow rate) decreases following roughly the
hydrogen concentration prole. Hydrogen production is
estimated by integrating hydrogen ux across the mem-
brane along the axis
F
(perm)
H
2
= N
m
D
m
_
L
0
U(z)J
H
2
dz. (16)
In this situation, we obtain a permeating hydrogen molar
owrate of F
(perm)
H
2
=1.3810
3
mol/s, while hydrogen
leaving the reactor with the tail gases is F
(out)
H
2
= 1.2
10
4
mol/s, so if we dene a hydrogen utilization factor
as
=
F
(perm)
H
2
F
(perm)
H
2
+ F
(out)
H
2
(17)
(hydrogen consumed vs. hydrogen produced), we have
in this case = 91%.
3.2. Variable buffer pressure with no power load
An opposite situation than before arises when buffer
loading dynamic is considered with zero power load ap-
plied. In this case, the hydrogen demand by FC is ab-
sent (i.e. F
(c)
H
2
(t ) = 0.0) so we only observe the buffer
lling by the permeating hydrogen. Fig. 4 shows sta-
tionary concentration and velocity proles along reactor
axis in this situation. At stationary conditions, since the
average hydrogen partial pressure above the membrane
surface inside the reactor equals the pressure reached
in the buffer, the permeating hydrogen ux vanishes.
This situation represents the asymptotical result of the
system start up when the buffer is lled without hydro-
gen demand by FC. It can be compared to the situation
arising when the reactor acts like a simple fuel proces-
sor for hydrogen production. In this case no equilibrium
shift effect can exist. Moreover, almost total conversion
of methanol is realized in the rst part of reactor, while,
from this standpoint, only water gas shift proceeds, as
CO
V
V
,
d
i
m
e
n
s
i
o
n
l
e
s
s

s
u
p
e
r
f
i
c
i
a
l

v
e
l
o
c
i
t
y
Y
,

m
o
l
a
r

f
r
a
c
t
i
o
n
2.0
1.6
1.2
0.8
0.4
0.25 0.20 0.15 0.10 0.05 0.00
1.0
0.8
0.6
0.4
0.2
CH
3
OH
CO
2
H
2
H
2
O
L, reactor length [m]
Fig. 4. Stationary gas phase molar fractions proles along reactor
axis taking into account buffer dynamics with F
(c)
H
2
= 0.0 mol/s.
Asymptotic value P
b
= 6.34 bar. The second order discontinuity in
the graph at z =LL
m
is due to the beginning of the permeation
process (oating membrane heads).
F
(c)
, hydrogen flow consumed by the cell [mol/s]

,

u
t
i
l
i
z
a
t
i
o
n

f
a
c
t
o
r

f
o
r

h
y
d
r
o
g
e
n
0.0012 0.0010 0.0008 0.0006 0.0004 0.0002
1.0
0.8
0.6
0.4
0.2
0.0
H
2
Fig. 5. Utilization factor as a variable of the hydrogen ow
consumed by the FC.
can be inferred observing the CO content prole. Sta-
bilized pressure in the buffer is P
b
= 6.34 bar. As be-
fore, supercial velocity has approximately the same
slope of hydrogen concentration prole, meaning that
the change in volumetric ow rate is now strongly in-
uenced by the moles produced by the reactions. We
conclude that for all possible stationary power regimes
for the cell ranging from zero to maximum power load
(i.e. P
(c)
= 150 W), we have a buffer hydrogen pres-
sure ranging from P
b
= 6.34 bar to a value that cannot
be less than P
b
=1.2 bar, because during operation, the
maximum ow rate required by the cell F
(c)
H
2
= 1.32
10
3
mol/s is less than what the reactor is able to pro-
duce with P
b
=1.2 bar. From this standpoint, even when
FC requires maximum ow of hydrogen, the residual
buffer pressure is sufcient to win the losses across the
L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090 1087
direction of increasing time
L, reactor length [m]
Y
H
2
,

h
y
d
r
o
g
e
n

m
o
l
a
r

f
r
a
c
t
i
o
n
0.25 0.20 0.15 0.10 0.05 0.00
1.0
0.8
0.6
0.4
0.2
Fig. 6. Time evolution of hydrogen molar fraction prole along
reactor axis, when step power load equivalent to a molar ow rate
F
(c)
H
2
= 1.32 10
3
mol/s from F
(c)
H
2
= 0.0 mol/s is required from
the cell from the buffer. The second-order discontinuity in the graph
at z = L L
m
is due to the beginning of the permeation process.
t, time [s]
P
b
,

B
u
f
f
e
r

p
r
e
s
s
u
r
e

[
b
a
r
]
365.4 313.2 261.0 208.8 156.6 104.4 52.2 0.0
8.0
6.0
4.0
2.0
P
f
= 1.84 bar
P
i
= 6.34 bar
Fig. 7. Time evolution of hydrogen shell pressure P
b
when step
power load equivalent to a molar ow rate F
(c)
H
2
=1.3210
3
mol/s
from F
(c)
H
2
= 0.0 mol/s is required from the buffer. The bold lines
represent the pressure limits in the buffer at maximum and minimum
power load.
pressure regulator at the cell inlet, avoiding a possible
problem of starvation of the fuel. It seems clear that at
different power regimes, we have different utilization
factor ranging from =0% to 91% as can be seen in
Fig. 5.
3.3. Buffer dynamic forced by power load applied to
the cell
In Figs. 6 and 7 we can observe the transient be-
haviour from the latter to the rst of the two limit situ-
ations previously described, respectively, for hydrogen
concentration prole inside the reactor, and hydrogen
P
b
P
b
,

B
u
f
f
e
r

p
r
e
s
s
u
r
e

[
b
a
r
]
1.0
0.8
0.6
0.4
0.2
261.0 208.8 156.6 104.4 52.2
6.5
6.0
5.5
5.0
4.5
4.0
3.5
F
(c)
/F
o
t, time [s]
F
(
c
)
/
F
o


,

d
i
m
e
n
s
i
o
n
l
e
s
s
H
2
H
2
H
2
H
2
c
e
l
l

f
e
e
d

f
l
o
w

r
a
t
e
Fig. 8. Dynamic oscillations induced in the hydrogen shell pres-
sure P
b
when transient sinusoidal molar ow rate of frequency
c = 0.1 (v
0
/L) is required from the cell. The dotted line rep-
resents normalized molar ow rate F
(c)
H
2
(t )/F
0
H
2
oscillations, with
F
0
H
2
= 1.32 10
3
mol/s.
P
b
t, time [s]
P
b
,

B
u
f
f
e
r

p
r
e
s
s
u
r
e

[
b
a
r
]
1.0
0.8
0.6
0.4
0.2
261.0 208.8 156.6 104.4 52.2
7.0
6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0
F
(c)
/F
o
F
c


/

F
o

,

d
i
m
e
n
s
i
o
n
l
e
s
s
H
2
H
2
c
e
l
l

f
e
e
d

f
l
o
w

r
a
t
e
H
2
H
2
Fig. 9. Dynamic oscillations induced in the hydrogen shell
pressure P
b
when transient sinusoidal molar ow rate of fre-
quency c = 0.2 (v
0
/L) is required. The dotted line repre-
sents dimensionless molar ow rate F
(c)
H
2
(t )/F
0
H
2
oscillations, with
F
o
H
2
= 1.32 10
3
mol/s.
buffer pressure. This evolution represents the dynamical
response of the systemwhen, after a start up in which the
buffer is lled (without FC hydrogen demand), reach-
ing the maximum equilibrium pressure P
b
= 6.34 bar
a fast commutation to maximum power load is super-
imposed starting from t = 10.44 s. As can be seen in
the gure, the system evolves towards the minimum
buffer pressure and the maximum utilization factor for
hydrogen (i.e. the ratio between hydrogen utilized from
1088 L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090
the cell and hydrogen produced), leaving, also in this
situation, a quantity of residual hydrogen in the exhaust
gases ( 12%). Finally, we examine the dynamical be-
haviour of the system subjected to a typical sinusoidal
power load cycle. As a consequence, a sinusoidal vary-
ing hydrogen supply to the cell is required. We choose
for our simulation an expression for the consumed mo-
lar ow rate that attains the form
F
(c)
H
2
(t ) =
F
0
H
2
2
+
F
0
H
2
2.5
cos(ct ), (18)
with F
0
H
2
= 1.32 10
3
mol/s. This is a cycle with a
mean value of 50% and an amplitude of 80% of the
ow at the nominal maximum power load (i.e. P
(c)
=
150 W). Fig. 8 shows the time evolution of hydrogen
buffer pressure compared to normalized molar ow rate
F
(c)
H
2
(t )/F
0
H
2
. As can be seen, buffer pressure follows the
dynamical evolution of the power load with a phase lag
and a pressure mean value that, in general, depends on
the applied frequency. Fig. 9 reveals the effect of varying
frequency. The highest is the frequency, the lower is the
phase time lag and correspondingly, the highest is the
pressure mean value. These results are consistent with
those expected and previously described. It seems clear
from the simulations that the system is dynamically sta-
ble under a wide range of forced conditions without
the control of reactor feeding, due to its natural self
equilibrium.
4. Conclusions
Computer simulation evidenced that, theoretically, in-
serting a mass ow capacitance element between the
hydrogen source (i.e. the reactor) and FC is a valid
strategy to improve the dynamical quality of the on
board hydrogen production system. In point of fact a
PEM FC follows almost immediately a load commuta-
tion [18,19] so that, if to produce enough hydrogen, the
reactors fuel ow rate is adjusted, it is its dynamical
transient which controls the overall time response of the
system.
By the utilization of the hydrogen buffer, this dynam-
ical transient can be avoided, at the cost of a variable
utilization factor for the produced hydrogen. The uncon-
sumed hydrogen leaves the system with the reactors tail
gases. The more is the power load applied to the cell,
the smaller is the amount of the unused hydrogen. The
hydrogen waste with the tail gases might seem a draw-
back of the proposed strategy, because at low power
regimes we have a low fuel efciency. Nonetheless, in
view of a necessary energy recovery and efciency op-
timization, we must consider that the sensitive heat and
the heating value of the reactors tail gases can be par-
tially recovered.
It is the case to point out that methanol reform-
ing reaction is thermodynamically enhanced also
at low temperature with high conversions. The ad-
vantages of the equilibrium shifting effects are less
evident than for other reforming processes. For
this particular process, the choice of using a CMR
is useful only to obtain ultra-pure hydrogen by
means of one single device. The steam reforming
of methanol was used in this numerical simulation
as a model for a generic endothermic process in
view of the on board hydrogen production to take
place in our pilot plant, however. Looking forward
to a cleaner energy system, we have to think to a
similar process with fuels that will avoid the high
thermal loss due to vaporization (gaseous fuels),
and will produce lower or, if possible, zero CO
x
emissions.
Appendix A
A.1. Numerical procedure
Numerical simulations presented in Section 2
have been obtained by adopting a backward nite
difference approach for solving the transport equa-
tions inside the reactor. The spatial domain has
been discretized into N = 100 internal elementary
cells (N = 100 internal grid points plus one to en-
forcing boundary conditions). Let us indicate with
Y
(j)
i
the dimensional molar fraction of component
i (i = 1, . . . , 5) in the space position X
(j)
= jX
with X = 1/N, at a time t. The unit step func-
tion U(X (1 :)) was discretized obtaining
in the grid points U(X
(j)
) = 0, j = 1, . . . , N
K
,
and U(X
(j)
) = 1, j = N
K+1
, . . . , N, were N
K
=
(1 :)/X.
From Eqs. (11) we obtain the following set of 5N
ODE:
dY
(j)
i
dt
=V
(j)
Y
(j1)
i
Y
(j)
i
X

OY
(j)
i
1
(j)
i
+

O1
(j)
t
, (A.1)
coupled with Eq. (13). Eqs. (13), (A.1) were integrated
using Euler method in time with time step t = 1.0
10
4
. At a xed time t, velocity prole were derived
by quadrature of Eq. (12) by mean of trapezoidal rule
L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090 1089
obtaining the recursive formulae
_
X
(j)
X
(j1)
1
t
dX = (1
(j)
t
+ 1
(j)
t
)
X
2
,
V
(j)
= V
(0)
+

O
_
X
(j)
0
1
t
dX = V
(0)
+

O
j

K=1
_
_
X
(k)
X
(k1)
1
t
dX
_
, V
(0)
= 1. (A.2)
Integral term in Eq. (13) is obtained at each time step
by the following quadrature formulae:
_
1
0
U(X)

J dX
=
N

j=1
[U(X
(j)
)

J
(j)
+ U(X
(j1)
)

J
(j1)
]
X
2
. (A.3)
A.2. Kinetic expressions and simulations parameters
The following reaction rate expressions are used
[20,3]:
R
R
=
k
R
K

CH
3
O(1)
(p
a
/p
0.5
s
)(1 p
3
s
p
r
/K
R
p
a
p
d
)C
T
S1
C
T
S1a
S
cat
j
cat
[1 + K

CH
3
O(1)
(p
a
/

p
s
) + K

HCOO(1)
p
r

p
s
+ K

OH(1)
(p
b
/

p
s
)][1 + K
0.5
H
2
(1
a
)
_
p
s
]
, (A.4)
R
D
=
k
D
K

CH
3
O(2)
(p
a
/

p
s
)(1 p
2
s
p
w
/K
D
p
a
)C
T
S2
C
T
S2a
S
cat
j
cat
[1 + K

CH
3
O(2)
(p
a
/

p
s
) + K

OH(2)
(p
w
/

p
s
)][1 + K
0.5
H
2
(2
a
)

p
s
]
, (A.5)
R
W
=
k
W
K

OH
3
(1)
(p
w
p
d
/

p
s
)(1 p
s
p
r
/K
W
p
w
p
b
)(C
T
S1
)
2
S
cat
j
cat
[1 + K

CH
3
O(1)
(p
a
/

p
s
) + K

HCOO(1)
p
r

p
s
+ K

OH(1)
(p
b
/

p
s
)]
2
. (A.6)
Table 1
Rate and equilibrium constants values used in the base case simu-
lations
Rate or equilibrium constants
k
R
= 3.11 10
4
m
2
/mol s K

CH
3
O(2)
= 2.46 10
3
bar
0.5
K

CH
3
O(1)
= 0.44 bar
0.5
K
H
2
(2a)
= 139.92 bar
0.5
K

OH(1)
= 0.31 bar
0.5
k
W
= 6.04 10
5
m
2
/mol s
K
H
2
(1a)
= 0.20 bar
0.5
K
R
= 5.49 10
4
bar
2
K

HCOO(1)
= 1.76 bar
1.5
K
D
= 1.61 10
3
bar
2
k
D
= 1.18 10
5
m
2
/mol/ s K
W
= 34.07
K

CH
3
O(2)
= 2.46 10
3
bar
0.5
Table 2
Catalyst properties values used in the base case simulations
Catalyst properties
C
T
S1
= 7.5 10
6
mol/m
2
C
T
S1a
= 1.5 10
6
mol/m
2
C
T
S2
= 7.5 10
6
mol/m
2
C
T
S2a
= 1.5 10
6
mol/m
2
S
cat
= 50 m
2
/g j
cat
= 1300 kg/m
3
Table 3
Parameters values used in the base case simulations
Parameter Unit Value
D m 0.013
D
m
m 0.00198
F
(in)
CH
3
OH
mol/s 6.0 10
4
F
(in)
H
2
O
mol/s 6.6 10
4
L m 0.254
L
1
m 0.203
N
m
7
P
0
bar 10.0
Q
0
/o mol/s m
2
Pa
0.5
1.5 10
3
R bar m
3
/mol K 8.05 10
5
R

J/mol K 8.31
T K 573.0
T
0
K 298.0
V
b
m
3
17 10
4
c 0.45
The detailed description of temperature laws of kinetic
constants can be found in [20,31]. For the sake of
brevity, we report in Tables 1 and 2, the kinetics con-
stant values and the catalyst properties at T =573 K that
we used in simulations. Process characteristics used in
the base case simulations, are reported in Table 3.
References
[1] Basile A, Paturzo L, Lagan F. Catal Today 2001;67:65.
[2] Brown LF. Int J Hydrogen Energy 2001;26:381.
1090 L. Capobianco et al. / International Journal of Hydrogen Energy 31 (2006) 10791090
[3] Melo F, Morlanes N. Catal Today 2005;107108:45866.
[4] Silberova B, Venvik HJ, Walmsley JC, Holmen A. Catal Today
2005;100:457.
[5] Suna J, Qiub X, Wua F, Zhub W, Wanga W, Haoa S. Int
J Hydrogen Energy 2004;29:1075.
[6] Semelsberger TA, Brown LF, Borup RL, Inbody MA. Int
J Hydrogen Energy 2004;29:1047.
[7] Wang L, Murata K, Inaba M. J Power Sources 2005;145:707.
[8] Amphlett JC, Mann RF, Peppley BA. Int J Hydrogen Energy
1996;21:673.
[9] Han J, Kim IS, Choi KS. Int J Hydrogen Energy 2002;27:1043.
[10] Lin YM, Rei MH. Int J Hydrogen Energy 2000;25:211.
[11] Sridhar P, Perumal R, Rajalakshmi N, Raja M, Dhathathreyan
KS. J Power Sources 2001;101:72.
[12] Takeda K, Baba A, Hishinuma Y, Chikahisa T. JSAE Rev
2002;23:183.
[13] Itoh N, Kaneko Y, Igarashi A. I&EC Res 2002;41:4702.
[14] Lattner JR, Harold MP. Int J Hydrogen Energy 2004;29:393.
[15] Tosti S, Basile A, Chiappetta G, Rizzello C, Violante V. Chem
Eng J 2003;93:23.
[16] Capobianco L, Violante V. In: Proceedings of the HYSY DAYS,
rst world congress of young scientists on hydrogen, Turin,
Italy; 1820 May 2005.
[17] Froment GF, Bischoff KB. Chem reactors anal des. New York:
Wiley; 1979.
[18] Del Prete Z, Schiavetti P, Capobianco L. In: Proceedings of
the VI Italian national congress on mechanical and thermal
mesurements, Desenzano Del Garda, Italy, 1214 September
2005.
[19] Del Prete Z, Schiavetti P. In: Proceedings of the HYSY DAYS,
rst world congress of young scientists on hydrogen, Turin,
Italy, 1820 May 2005.
[20] Peppley BA, Amphlett JC, Kearns LM, Ronald FM. Appl Catal
A: General 1999;179:31.
[21] Wiese W, Emonts B, Peters R. J Power Sources 1999;84:187.
[22] Wieland IS, Melin IT, Lamm IA. Chem Eng Sci 2002;57:1571.
[23] Buxbaum RE, Lie H. J Power Sources 2003;123:43.
[24] Buxbaum RE, Kinney AB. I&EC Res 1996;35:530.
[25] Buxbaum RE. Sep Sci Technol 1999;34:2113.
[26] Buxbaum RE, Marker TL. J Membr Sci 1993;85:29.
[27] Zhang Y, Ozaki T, Komaki M, Nishimura C. J Membr Sci
2003;224:81.
[28] Ozaki T, Zhang Y, Komaki M, Nishimura C. Int J Hydrogen
Energy 2003;28:297.
[29] Nishimura C, Komaki M, Hwang S, Amano M. J Alloys Compd
2002;330332:902.
[30] Buxbaum RE. Palladium-coated metal membranes for ultra
high purity hydrogen and future applications. http://www.
rebresearch.com/membs/MetMems.html.
[31] Harold MP, Nair B, Kolios G. Chem Eng Sci 2003;58:2551.
[32] Drioli E, Violante V. Reattori Catalitici a Membrana. ENEA
Report, Italy; 1994.
[33] Elnashaie SSEH, Moustafa T, Alsoudani T, Elshishini SS.
Comput Chem Eng 2000;24:1293.
[34] Tosti S, Violante V. ENEA ITER Task TEP 1998;2..
[35] Moustafa T, Elnashaie SSEH. J Membr Sci 2000;178:171.
[36] Shu J, Grandejean PA, Van Neste A, Kaliaguine S. Can J Chem
Eng 1991;69:1036.

You might also like