You are on page 1of 11

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1

Available online at www.sciencedirect.com

ScienceDirect

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

Steam reforming of glycerol for hydrogen


production: Modeling study

Joel M. Silva, M.A. Soria, Luis M. Madeira*


LEPABE, Department of Chemical Engineering, Faculty of Engineering, University of Porto, Rua Dr. Roberto Frias s/n,
4200-465 Porto, Portugal

article info abstract

Article history: A phenomenological model that predicts the operation of a packed bed reactor for glycerol
Received 23 September 2015 steam reforming (GSR), in terms of glycerol conversion and products generation (first time
Received in revised form to our knowledge), especially H2, was built and used to perform a parametric analysis
13 November 2015 (range of conditions: 748e848 K, water-to-glycerol feed ratios (WGFRs) of 3e12 and 1
Accepted 14 November 2015 e5 atm). First, the model was validated in terms of glycerol conversion, products yields and
Available online xxx selectivities, having been observed very good agreement. Regarding the parametric study,
higher temperatures, WGFRs and pressures were found to be more beneficial in terms of
Keywords: glycerol conversion. Moreover, the H2 yield and selectivity are enhanced when higher
Glycerol temperatures and lower WGFRs and pressures are employed. Still, the maximum amount
Steam reforming of 4.93 mol of H2/mole of glycerol fed to the reactor was observed at 848 K, WGFR of 9 and
Hydrogen 2 atm.
Modeling Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights
reserved.

biodiesel production process and converted into hydrogen


Introduction through steam reforming (Eq. (1)).
 1 
Nowadays hydrogen is seen by many authors as the energy C3 H8 O3 þ 3H2 O#7H2 þ 3CO2 DH298K
r ¼ 128 kJ$mol (1)
carrier of the future due to its potential to fulfill many of our
However, the formation of secondary products of a more
society needs without ignoring some of the environmental
complex mechanism than Eq. (1), such as CO and CH4, must be
problems that fossil fuels, the main energy carriers of the
considered as well, according to some works reported in the
present, cannot avoid. Therefore, several distinct processes
literature regarding glycerol steam reforming (GSR) [1e4]. The
for hydrogen production have been researched. While during
use of glycerol as feedstock would prove itself advantageous
the last century the focus was manly on using fossil fuels-
compared to methanol or ethanol since it would decrease its
based feedstock (e.g.: natural gas), in the last decades the in-
disposal and associated costs, thus valorizing biodiesel.
terest has shifted towards cleaner raw materials such as
Even though there has been done a huge amount of work
methanol, ethanol and glycerol, among others. Glycerol, in
regarding the search of GSR catalysts and there are several
particular, can be produced as a by-product during the
thermodynamic assessments that have been done [5],

* Corresponding author. Tel.: þ351 225081519.


E-mail address: mmadeira@fe.up.pt (L.M. Madeira).
http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
0360-3199/Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
2 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1

Notation and glossary si temperature based parameter for component i, K


T bed absolute temperature, K
A kinetics pre-exponential factor, mol/(m2s kPa(bþg))
Tbi normal boiling point of species i, K
Aij dimensionless parameter for the mixture of i and j
Ti,0 reference temperature for species i in Sutherland's
Ametal surface metal surface area of the catalyst, m2/kgcat.
equation, K
C total concentration of species, mol/m3
Tw reactor wall absolute temperature, K
Ci molar concentration of compound i, mol/m3
us superficial velocity of the gas mixture, m/s
Cp,i molar heat capacity of species i, J/(mol K)
VR volume of the reactor bed, m3
Cp,f Mass heat capacity of gas mixture, J/(kg K)
Wcat mass of catalyst in the reactor bed, kgcat.
C1 Sutherland constant, kg/(m s K0.5)
x dimensionless axial coordinate
Dax mass axial dispersion coefficient, m2/s
XG conversion of C3H8O3
DC3 H8 O3 ;j binary diffusivity of C3H8O3 when diffusing into
Yi yield of species i
species j, m2/s
yi molar fraction of species i
DC3 H8 O3 ;mix molecular diffusivity of C3H8O3 in the reaction
mixture, m2/s Greek letters
dp catalyst particles diameter, m b partial reaction order for C3H8O3
Ea activation energy, J/mol DHf,i heat of formation of species i, J/mol
F dimensionless parameter εb void fraction of the catalyst bed
Fi molar flux of species i, mol/(m2s) fij dimensionless parameter for the mixture of i and j
g dimensionless parameter of steam kinetics g partial reaction order for H2O
h convective wall-fluid heat transfer coefficient, W/ lax heat axial dispersion coefficient, W/(m K)
(m2K) li thermal conductivity of species i, W/(m K)
L reactor length, m lf thermal conductivity for the gas mixture, W/(m K)
Mi molar mass of species i, mol/kg rf gas mixture density, kg/m3
P reactor total pressure, Pa mi viscosity of species i, Pa s
pi partial pressure of component i ¼ C3H8O3 and H2O, mf gas mixture viscosity, Pa s
kPa v atomic diffusion volume
R ideal gas constant, J/(K mol) ni unitary constant that indicates if i is consumed or
Ri consumption/formation rate of component i, mol/ produced
(kgcat.s)
List of acronyms
R0 reactor inner radius, m
GSR glycerol steam reforming
S Sutherland temperature, K
WGFR water-to-glycerol feed ratio
Si selectivity of species i

phenomenological models that allow describing the GSR re-


action system in packed bed reactors, for example, are still too Phenomenological model
few. To the best of our knowledge there are only a couple of
works where such models are reported [6e8]. Even though all Model and governing equations
these are 2D models that allow consequently to do an inter-
esting analysis of the several variables profiles (e.g. contours The pseudo-homogeneous model here proposed for
over the 2D domain [8]), they only predict the consumption of describing the GSR reaction in a packed bed reactor considers
glycerol. There is one work in which the generation of prod- the following assumptions:
ucts (and by-products) is predicted; however, it is for the
sorption-enhanced GSR process and assumes theoretical  Steady state;
yields [9], which is not realistic for conditions far from those  Axially dispersed plug flow;
employed by the authors. With the aim of covering up this  Non-isothermal operation;
hole, a phenomenological model capable of predicting accu-  Pressure drop along the bed described by the Ergun
rately the real operation of a GSR packed bed reactor in terms equation;
of consumption of glycerol and production of the main prod-  Non-constant velocity along the bed;
ucts (H2 and CO2) and by-products (CO and CH4), based on  Ideal gas behavior.
actual individual products generation kinetics, has been
developed and will be target of analysis here. First a critical The governing equations used to simulate such system are
comparison between the model's predictions and the experi- the following, being composed by 7 differential equations (one
mental results obtained by Cheng et al. [10] for a Co-Ni/Al2O3 for each species - partial mass balances), 1 differential equa-
catalyst is done, being then followed by a parametric analysis tion for the total mass balance (required due to the non-
considering different temperatures (748e848 K), water-to- constant velocity along the reactor bed), as well as 2 other
glycerol feed ratios (WGFRs) (3e12) and pressures (1e5 atm). differential equations to obtain temperature and total pres-
sure profiles:

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1 3

Partial mass balance


1  yC3 H8 O3
d2 Ci L dðus Ci Þ Wcat L2 DC3 H8 O3 ;mix ¼ Pn yj (10)
Dax  þ Ri ¼ 0 (2)
dx2 εb dx VR εb j¼1 DC
3 H8 O3 ;j

Total mass balance being jsC3H8O3 and DC3 H8 O3 ;j the binary diffusivity of glycerol
when diffusing into species j. In order to calculate this last
dðus CÞ Wcat L X parameter the Fuller method [14] was used:
 Ri ¼ 0 (3)
dx VR i
 0:5
Energy balance 0:1013T1:75 1
MC3 H8 O3
þ M1
j
DC3 H8 O3 ;j ¼ " #2 (11)
d T us rf Cp;f L dT Wcat L X
2 2   hL2 P 1=3 P 1=3
 þ DHf ;i Ri  2 ðT  Tw Þ ¼ 0 P ð vÞi þ ð vÞj
dx2 lax dx VR lax i R0 lax
(4) P
where v is the sum of the atomic diffusion volumes for each
Momentum Balance component (taken from Ref. [14]) and Mi is the molar mass of
species i (in this case expressed in g/mol).
dP ð1  εb Þ2 Lmf ð1  εb ÞLrf 2 In terms of heat axial dispersion coefficient, it was calcu-
þ 150 us þ 1:75 us ¼ 0 (5)
dx ε3b d2p ε3b dp lated using the Wakao and Funazkri correlation [15]:
Since both partial mass balance and energy balance are lax ¼ 7lf þ 0:5rf us dp Cp;f (12)
second order differential equations, then two boundary con-
ditions are required for each. The following boundary condi- where lf is the thermal conductivity of the gas mixture,
tions were used: calculated as detailed below. Both mass and heat axial
dispersion coefficients were calculated at the inlet conditions
x¼0: Ci ¼ Ci;in ; T ¼ Tin ; us ¼ us;in ; P ¼ Pin (6) in order to simplify the calculations and since no significant
variation was observed along the bed.
dCi dT The gas mixture density was calculated locally through Eq.
x¼1: ¼ 0; ¼0 (7)
dx dx (13):
where Ci is the molar concentration of compound i (i ¼ C3H8O3,
P Xn
H2O, H2, CO2, CO, CH4 and Ar), x is the dimensionless axial rf ¼ yi Mi (13)
RT i¼1
coordinate, Dax is the mass axial dispersion coefficient, L is the
reactor length, us is the superficial velocity, Wcat is the mass of where R is the ideal gases constant and yi is the molar fraction
catalyst in the reactor bed, VR is the volume of the reactor bed of species i. Similarly, the viscosity of the gas mixture was
(Wcat/VR is the bulk density, rb), εb is the void fraction of the calculated locally with the Wilke method [13]:
catalyst bed and Ri is the consumption/formation rate of
component i. C is the total concentration of species, T is the X
n
yi m
mf ¼ Pn i (14)
bed temperature, rf is the gas mixture density, Cp,f is the gas i¼1 j¼1 yj fij
mixture heat capacity, lax is the heat axial dispersion coeffi-
being mi the viscosity of species i that is calculated using a
cient, DHf,i is the heat of formation of species i, h is the
polynomial equation (Eq. (15)) [16], except for Ar whose vis-
convective wall-fluid heat transfer coefficient, R0 is the reactor
cosity was calculated through Sutherland's equation (Eq. (16))
inner radius, Tw is the reactor wall temperature, P is the total
[17]:
reactor pressure, mf is the gas mixture viscosity and dp is the
catalyst particles diameter. mi ¼ A þ BT þ CT2 (15)
The void fraction of the catalyst bed was estimated with Eq.
(8) [11]: C1 T3=2
mi ¼ (16)
A TþS
εb ¼   n þ B (8)
2R0 dp where A, B and C are constants that present specific values for
each gas (taken from Ref. [16]), while C1 and S are the
where A ¼ 1.5, B ¼ 0.35 and n ¼ 1 assuming that the catalyst
Sutherlands constants for gas i (C1 ¼ 2.08  106 kg/
particles are lumps of irregular shape.
(m s K0.5) and S ¼ 191 K for Ar). Parameter fij is calculated with
Regarding the mass axial dispersion, it was calculated
Eq. (17) [13]:
using the Edwards and Richardson correlation [12]:
2 !0:5 !0:25 32
  0:5
0:5us dp Mi 41 þ mi Mi 5
Dax ¼ 0:73DC3 H8 O3 ;mix þ (9) fij ¼ 8 1 þ (17)
1 þ 9:49
DC
3 H8 O3 ;mix Mj mj mj
us dp

where DC3 H8 O3 ;mix is the molecular diffusivity of glycerol in the where jsi. Similarly to the axial dispersion coefficients, the
reaction mixture, which can be estimated using the Wilke mass heat capacity of the gas mixture was calculated at feed
method [13]: conditions for the same reasons:

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
4 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1

Pn the consumption of glycerol but also for the generation of H2,


yi Cp;i
Cp;f ¼ Pi¼1
n (18) CO2, CO and CH4 was used:
i¼1 yi Mi

where Cp,i is the molar heat capacity of species i and can be Ri ¼ ðni ÞAeEa =RT pbC3 H8 O3 pgH2 O Ametal surface (26)
calculated with Eq. (19) [16]:
being that ni takes the values of ±1, depending on if species i is
Cp;i ¼ A þ BT þ CT2 þ DT3 þ ET4 (19) being consumed (1) or produced (þ1). A is the pre-
exponential factor, Ea stands for the activation energy, b and
being A, B, C, D and E constants that present specific values for g are the partial reaction orders for glycerol and steam,
each gas (taken from Ref. [16]). respectively, and Ametal surface is the catalyst's metal surface
The convective wall-fluid heat transfer coefficient was area. The values for the kinetic parameters estimated by
calculated using Leva's correlation [18]: Cheng et al. [10] are presented in Table 1.
  !0:9 The metal surface area of the catalyst is 740 m2/kgcat.. Since
lf dp rf us
h ¼ 0:813 e6dp =ð2R0 Þ (20) no kinetic parameters were presented for steam, its kinetics
2R0 mf was estimated so that the error associated to the mass balance
lf was calculated using the Wassiljewa method [13]: is very small (3%).

X
n
yi li Numerical solution
lf ¼ Pn (21)
i¼1 j¼1 yj Aij

The set of ordinary differential and algebraic equations pre-


where li is the thermal conductivity of species i and was ob-
viously described was solved using the bvp4c function of
tained through the correlation proposed by Eucken [13]:
MATLAB software, a finite difference code which implements
 
R Cp;i the three-stage Lobatto IIIa formula. The solutions were ob-
li ¼ mi 1:25 þ (22)
Mi Mi tained at 200 equally spaced points along the column length
and a value of 105 was considered for both absolute and
The parameter Aij was obtained by the method of Lindsay
relative tolerances.
and Bromley [14]:
( "    #1=2 )2  
3=4
1 m Mj T þ si T þ sij
Aij ¼ 1þ i (23) Results and discussion
4 mj Mi T þ sj T þ si

where sij is obtained from Eq. (24): Reaction metrics


pffiffiffiffiffiffiffi
sij ¼ F si sj (24)
The reactor performance was evaluated in terms of glycerol
where F is a constant that takes the value of 1 except when conversion (Eq. (27)), products yields (Eq. (28)) and selectivities
either of both components i and j are very polar (F ¼ 0.73). si,j is (Eq. (29)):
a component temperature based parameter that takes the
FC3 H8 O3 ;in  FC3 H8 O3 ;out
value of 79 K for H2 and is expressed by Eq. (25) for the other XG ¼  100 (27)
FC3 H8 O3 ;in
components:

si;j ¼ 1:5Tbi;j (25) Fi;out


Yi ¼ (28)
FC3 H8 O3 ;in  FC3 H8 O3 ;out
being that Tbi,j is the normal boiling temperature of compo-
nent i or j (taken from Refs. [19,20]). Fi;out
Si ¼  100 (29)
FH2 ;out þ FCO2 ;out þ FCO;out þ FCH4 ;out
Kinetics
where Fi is the molar flux of species i; in stands for the reactor
inlet while out stands for the reactor outlet. The experimental
An extensive literature review on GSR kinetics has been done.
conversions obtained by Cheng et al. [10] were calculated
Most of the found kinetics are of the power-law type
based on the conversion into gaseous products. On the other
[3,4,6,21e25] and refer to the glycerol consumption rather
hand, no further kinetics besides those shown in Table 1 were
than to each of the reactions that normally take part in the
used. Therefore, neither liquid nor solid - e.g. coke - products
GSR process (glycerol thermal decomposition, water-gas shift
and methanation of carbon monoxide) or to each of the
products that are formed (H2, CO2, CO and CH4). More complex
kinetics that take into consideration adsorption constants for Table 1 e Kinetic parameters for the species
glycerol and steam (of the LangmuireHinshelwood type) or consumption/formation [10].
several kinetic constants corresponding to several elementary Species A (mol m2 s1 kPa(bþg)) Ea (kJ mol1) b g
steps of the GSR process were found as well [3,10,23]. How- C3H8O3 0.036 63.30 0.253 0.358
ever, and once again, they refer only to glycerol consumption. H2 0.468 67.28 0.253 0.274
Therefore, in order to describe the GSR process, the power- CO2 0.074 64.06 0.281 0.403
law-type kinetics (Eq. (26)) proposed by Cheng et al. [10] for a CO 0.062 61.73 0.308 0.065
CH4 0.555 100.90 0.601 0.393
Co-Ni/Al2O3 catalyst with estimated parameters not only for

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1 5

formation is considered herein. Even though other definitions


Table 2 e Parameters used during the simulations.
of selectivity are often used in the literature, the one of Eq. (29)
is here adopted in order to match the one used by Cheng et al. Parameter Value Units
[10]. It should be stressed that, according to this definition, L 0.12 m
even when no by-products are formed, CO2 is always pro- R0 5  103 m
duced along with H2. Thus, the maximum selectivities for the dp 1.40  104 m
Wcat 5  104 kg
desired products are 70% and 30% for H2 and CO2, respectively.
Gas hourly space velocity 9000 LN/(kgcath)
yC3 H8 O3; in a 0.05
Model validation yH2 O; in a 0.15e0.60
a
Balanced with Argon.
Before being used to perform a parametric analysis, the model
must be validated so that we are sure that its outputs in terms
of glycerol conversion and products yields and selectivities are Profiles along the reactor bed
trustable. In Fig. 1 it is presented the comparison between the
model predictions and the experimental results obtained by For the parametric study, a packed bed reactor with the
Cheng et al. [10] (results shown refer to different operating characteristics presented in Table 2 was used (most of them
conditions, in the ranges 773e823 K, WGFRs between 3.4 and similar to those used elsewhere [10]).
11.9 and atmospheric pressure) for (a) the conversion of Before going into the parametric analysis, it is important to
glycerol, (b) yields and (c) selectivities towards the different understand how the species molar fractions and temperature
species. vary along the packed bed length. For this the profiles ob-
The parity plots show that there is in fact very good tained at 798 K, WGFR of 9 and 1 atm are presented in Fig. 2.
agreement between the model predictions and the experi- Regarding the concentration profiles in Fig. 2(a), it can be
mental results, particularly for selectivites. It is also worth to observed the existence of 2 distinct regions. In the first
notice that relatively high H2 (and also CO2 yields) were 10e15% of the reactor a rapid decrease of the glycerol con-
reached, although below the maximum theoretical value of 7 centration as well as a fast enhancement of the concentra-
(cf. Eq. (1)). Moreover, very low CO and CH4 yields and selec- tions of particularly H2 and CO2 are observed. This happens
tivites are observed, which are desirable. because of the initial high glycerol and steam partial pressures

Fig. 1 e Parity plots for (a) glycerol conversion, (b) products yields and (c) selectivities for temperatures between 773 and
823 K, WGFRs between 3.4 and 11.9 and atmospheric pressure. Experimental yields and selectivities taken from Ref. [10],
while conversion values were kindly provided by Doctor Chin Kui Cheng.

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
6 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1

Parametric analysis

In order to understand how the performance of a catalytic


packed bed reactor for GSR can be enhanced, the effect of
several operating conditions like temperature, WGFR and
pressure was analyzed. In the parametric analysis, the ranges
considered for temperature and partial pressures of the re-
actants in the reactor feed were slightly increased as
compared to those used in the model validation. It was veri-
fied for all the range of simulated conditions that the glycerol
conversion in the thermodynamic equilibrium (calculated
with Aspen Plus software) was always higher or equal to the
actual conversion obtained with the phenomenological
model.
The first parameters to be target of analysis were temper-
ature (748e848 K) and WGFR (3e12). The effect of both vari-
ables on the conversion of glycerol is presented in Fig. 3.
Regarding temperature, it is clear that its increase at con-
stant WGFR benefits the conversion of glycerol. This happens
due to the enhancement of the reaction kinetics according to
Eq. (26). It has also been verified that the increment in terms of
glycerol conversion caused by the increase of temperature
increases as the WGFR becomes higher, achieving a maximum
absolute increment of approximately 44% when the temper-
ature is increased from 748 K up to 848 K at WGFR of 12.
Regarding the WGFR, its increase at constant temperature also
results in an enhancement of the glycerol conversion. This
happens because by increasing the amount of steam in the
reaction mixture the glycerol consumption kinetics is
Fig. 2 e Simulated profiles of (a) molar fractions of C3H8O3, enhanced since g > 0 in Eq. (26). Also, it has been observed that
H2, CO2, CO and CH4 (dry basis and without carrier gas) and the increment in terms of glycerol conversion caused by the
(b) temperature along the dimensionless length of the increase of the amount of steam at constant temperature
reactor bed at 798 K, WGFR of 9 and 1 atm. slightly increases as the WGFR gets close to 9. On the other
hand, when it is varied from 9 to 12, this increment decreases
slightly. This happens because the reaction rate becomes
and high temperature that result in a fast GSR kinetics. In the approximately zero-order for water due to its excessive
second region the concentrations of glycerol, H2 and CO2 vary amount in the reaction media.
much more slowly and start stabilizing due to the much
smaller partial pressures of glycerol and steam and the
already diminished temperature. For CO and CH4 this effect is
much smaller since both present very small concentrations,
especially CH4 due to the low methanation selectivity of the
Co-Ni/Al2O3 catalyst.
In terms of temperature profile, it can be seen that its shape
is very similar to that normally observed in adiabatic reactors
[18,26]. However, the model here described takes into
consideration the existence of heat transfer from the reactor
wall (considered to be at the same temperature as the oven
which on its side is considered to be equal to the inlet tem-
perature) to the reaction mixture, which is at a lower tem-
perature than the wall due to the endothermicity of the GSR
reaction. Therefore, a non-adiabatic profile would be expect-
able. Nevertheless, it has been observed that the wall-fluid
heat transfer coefficient obtained from Eq. (20) is very small
and so heat transfer from the reactor wall to the reaction
mixture only occurs up to a very limited extent.
Regarding total pressure, it presented a linear decrease
along the bed length, as expected, being that an average Fig. 3 e Effect of both temperature and WGFR on the
pressure drop of approximately 12 kPa (~12%) was obtained. conversion of glycerol at 1 atm.

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1 7

In terms of the products, their generation was evaluated in CO formation. However, at lower WGFRs a more pronounced
terms of yield and selectivity. The effect of both temperature effect of temperature is observed, being that the yield of CO
and WGFR on the yields of H2, CO2, CO and CH4 is presented in increases with temperature, contrarily to what was expected
Fig. 4. based on the slightly lower activation energy of CO. In fact it
By analyzing Fig. 4 (a) it can be seen that the yield of H2 is was noticed that the decrease of both glycerol and steam
positively influenced by the increase of temperature at con- partial pressures along the packed bed have a decisive role
stant WGFR. This happens because of the highly accelerated here. For a WGFR of 3, it was observed that the individual ef-
kinetics of H2 formation at higher temperatures, according to fect of increasing the inlet temperature from 748 K up to 848 K
Eq. (26). In other words, when the temperature is increased (assuming constant the glycerol and water partial pressures,
from 748 K up to 848 K, the rate of formation of H2 increases and equal to the feed ones) led to an enhancement of the
more than the rate of consumption of glycerol and so higher instantaneous CO yield (assessed based on reaction rates) of
H2 yields are observed. This is exclusively related to the dif- only 3%. On the other hand, when the individual effect of the
ference between the values of activation energy presented in reactants pressures decrease along the reactor bed for 748 K
Table 1 for glycerol and H2. Regarding the WGFR, its increase and 848 K (Supplementary data, Fig. S1) was evaluated
leads to a decrease of the H2 yield. This means that when the (assuming constant temperature, and equal to the feed one),
steam content in the reaction mixture is increased the for- an increase of the instantaneous CO yield from the inlet to the
mation rate of H2 increases less than the consumption rate of outlet reactor conditions of 53% and 22% was observed at
glycerol. This happens because the partial order of steam for 848 K and 748 K, respectively. This happens because at 848 K
the consumption of glycerol is higher than that for the pro- the decrease of both glycerol and steam pressures is more
duction of H2, as can be seen in Table 1. Regarding CO2, it is pronounced, being that steam has an even more important
observed that temperature has almost no effect on its yield. role than glycerol here since its partial order for both glycerol
This is most probably due to the very close activation energy and CO kinetics are very different (Table 1). The difference in
values presented in Table 1 for glycerol and CO2. In terms of terms of instantaneous CO yield enhancement is responsible
WGFR, a much more pronounced effect is observed as for its overall increase when temperature is raised from 748 K
increasing WGFRs lead to the enhancement of the CO2 yield. up to 848 K (Fig. 4 (c)). Moreover, at a WGFR of 12 the same
This is explained by the higher partial order of steam for the evaluation (Supporting information, Fig. S1) led to a much
rate of CO2 formation (0.403) than for the rate of glycerol smaller difference between 748 K and 848 K in terms of
consumption (0.358). instantaneous CO yield enhancement along the reactor (7.9%
Concerning CO, Fig. 4 (c) shows that temperature has and 8.7% at 848 K and 748 K, respectively). This much smaller
almost no effect on its yield at higher WGFRs. This is due to difference is due to the much higher glycerol partial pressure
the very close activation energies of glycerol consumption and relative reduction along the bed comparatively to the steam

Fig. 4 e Effect of both temperature and WGFR on the yield of (a) H2, (b) CO2, (c) CO and (d) CH4 at 1 atm.

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
8 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1

partial pressure reduction, which leads to a more emphasized CH4 selectivities increase with the steam content, the selec-
negative effect of glycerol partial pressure decrease and a less tivities of H2 and CO show the opposite trend. This is due to
important positive effect of steam partial pressure decline on the higher steam partial orders for the formation of both CO2
the CO kinetics. Consequently there is almost no variation of and CH4 or, for the case of CO, due to the negative steam
the CO yield with temperature at WGFR of 12 (Fig. 4 (c)). partial order, which results in a highly pronounced relative
Regarding the effect of the WGFR, it is clear that the decrease decrease of the CO selectivity (42%). On the other hand, the
of the CO yield caused by the increase of the steam content selectivity of H2 only suffers a relative decrease of 1.2%. Also,
shown in Fig. 4 (c) is related to the negative partial order of since the steam partial order for CO2 formation is higher than
steam for the rate of CO formation. that for CH4 formation, its selectivity enhancement caused by
Finally, in terms of CH4 yield, the increase of temperature the increase of the steam content is also higher (21% and 8%,
led to its big enhancement, as can be seen in Fig. 4 (d). This is respectively). It is finally worth highlighting the completely
due to the much higher activation energy for CH4 formation. different selectivities, with average values of 64.6% for H2,
As for the WGFR, it presented almost no effect on the CH4 yield 26.7% for CO2, and far lower for CO (7.9%) and particularly CH4
(variation of 0.005 at 848 K), because g is very similar for (0.8%), which were obtained.
methane and glycerol. By analyzing Fig. 6 it is possible to conclude that one ben-
The effect of both temperature and WGFR on the products efits from the use of higher temperatures and WGFRs when
selectivities was analyzed as well and is presented in Fig. 5 (a) intending to achieve higher H2 production (FH2 ;out =FGlycerol;in ),
and (b), respectively. As can be seen in Fig. 5 (a), while both H2 being that at atmospheric pressure a maximum production of
and CH4 selectivities increase with temperature, the selec- 4.66 mol of H2/mole of glycerol fed to the reactor was observed
tivities of CO2 and CO decrease. This is related to the respec- at 848 K and WGFR of 12.
tive activation energies for the rate of formation of each The last parameter to be analyzed was the operating
component, being that H2 and CH4 present higher activation pressure. The effect of varying the operating pressure be-
energies than CO2 and CO. Also, since the activation energy for tween 1 and 5 atm on the glycerol conversion at different
CH4 formation is much higher than that for H2, its selectivity temperatures and WGFR of 9 is presented in Fig. 7.
suffers a much more pronounced relative increase (55%), It is possible to observe that higher pressures enhance the
while H2 selectivity only suffers a slight relative increase conversion of glycerol. This is explained by the positive values
(2.4%). On the other hand, since the activation energies for CO of the partial orders of glycerol and steam in its consumption
and CO2 are very similar, both suffer very similar variations. rate (b and g are both > 0). Moreover, even though increasing
Regarding the effect of WGFR (Fig. 5 (b)), while both CO2 and pressures have a more pronounced effect on glycerol con-
version at lower temperatures (748e773 K), only at higher
temperatures the glycerol conversion gets close and even
reaches 100%. At 848 K, a pressure of 2 atm was observed to be
already enough to allow glycerol to be completely converted.
A similar analysis was done for the products yields. Such
effects are presented in Fig. 8.
It can be seen in Fig. 8 (a) that as the pressure increases, at
constant temperature, the yield of H2 becomes lower. This

Fig. 5 e Effect of (a) temperature (WGFR ¼ 9 and 1 atm) and Fig. 6 e Effect of both temperature and WGFR on the
(b) WGFR (798 K and 1 atm) on the selectivities of H2, CO2, amount of hydrogen produced per glycerol fed to the
CO and CH4. reactor at 1 atm.

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1 9

has reached completion, this variation stops, as shown in


Fig. S2 (a), leading to a higher average temperature in the
reactor. This benefits slightly more the H2 than the glycerol
kinetics due to the higher H2 formation activation energy, and
that is why the H2 yield does not decrease so much at higher
temperatures. At lower temperatures this is not important
because glycerol conversion never reaches completion, and so
average reactor temperature is smaller for higher pressures
(Fig. S2 (b)).
Regarding CO2, higher pressures enhance its yield because
of the higher glycerol and steam partial orders of CO2 forma-
tion compared to glycerol kinetics. Also, it must be remarked
that even though an apparent increase of the CO2 yield with
temperature at 5 atm is observed, it is almost negligible <
0.03. As for CO, its yield is decreased when higher pressures
are employed, as can be seen in Fig. 8 (c), due to the negative
partial order of steam for CO kinetics. Regarding CH4, higher
Fig. 7 e Effect of pressure on the conversion of glycerol at pressures exercise a positive effect on its yield, especially at
different temperatures and WGFR of 9. higher temperatures. This once again occurs because of the
higher partial orders of CH4 formation compared to those of
glycerol consumption. Also, the methane kinetics benefits
much more than glycerol kinetics from the already mentioned
happens because the same increase of pressure increases the progressively lower decrease of temperature in the reactor
glycerol consumption rate more than it increases the H2 pro- after the glycerol conversion is complete (higher temperatures
duction rate, according to the respective partial orders pre- and pressures) (Supporting information, Fig. S2), due to its
sented in Table 1. Also, at higher temperatures such negative much higher activation energy. Therefore, CH4 yield presents
effect of the total pressure on hydrogen yield is less pro- a considerably higher increase with pressure at higher
nounced. It has also been observed that the initial tempera- temperatures.
ture decrease along the reactor is more pronounced with A similar analysis was also done for the products selec-
increasing total pressure/glycerol conversions (Supporting tivities as well, being that their variation with pressure at
information, Fig. S2). However, after conversion of glycerol 798 K and WGFR of 9 is presented in Fig. 9.

Fig. 8 e Effect of pressure on the yield of H2 at different temperatures and WGFR of 9.

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
10 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1

reactor was observed at 848 K, WGFR of 9 and 2 atm. In a


recent work, Ebshish et al. [27] used response surface meth-
odology to optimize process conditions through a non-
phenomenological statistical approach (multiple regression
analysis). In such study the ranges of conditions employed
were similar to those reported by us. Ebshish et al. [27] have
got to maximum experimental conversion of glycerol and H2
production of 75% and 4.03 mol of H2/mole of glycerol fed to
the reactor, respectively, at 873 K, WGFR of 12 and atmo-
spheric pressure while using a Ni/Al2O3 catalyst.

Conclusions

The developed phenomenological model that describes the


GSR reaction in a packed bed reactor, developed for the first
Fig. 9 e Effect of pressure on the selectivities of H2, CO2, CO
time for predicting both glycerol consumption and products/
and CH4 at 798 K and WGFR of 9.
by-products formation, presented quite good adherence with
the experimental results. A parametric analisys was also
performed, being that it was ultimately concluded that higher
Regarding H2 and CO, their selectivities decrease when temperatures, higher WGFRs and higher pressures are more
higher pressures are employed due to their lower partial or- advantageous in terms of glycerol conversion and H2 pro-
ders with respect to glycerol and steam compared to those for duction (moles of H2/moles of glycerol fed to the reactor). Even
CO2 and CH4 (Table 1); this is particularly evident for CO due to though higher WGFRs and pressures lead to lower H2 yields
its negative partial order for steam. Also, CH4 selectivity in- and selectivities, its absolute production (produced flow rate)
creases much more than CO2 selectivity due to its much is increased until the total glycerol conversion reaches
higher glycerol partial order. completion. After that, subsequent increments of these vari-
Finally, in terms of H2 production, it can be seen in Fig. 10 ables will decrease the absolute H2 production due to their less
that pressure positively affects the amount of hydrogen positive effect on the H2 kinetics compared to the other ki-
produced per glycerol fed to the reactor up to a certain point, netics (except CO). Therefore, maximum H2 production of
being that afterwards it has a negative impact (mainly at 4.93 mol of H2/mole of glycerol fed to the reactor was observed
higher temperatures). This happens since after glycerol con- at 848 K, WGFR of 9 and 2 atm. It is nonetheless necessary to
version reaches 100% (Fig. 7), the selectivity towards H2 pro- remark that higher temperatures, WGFRs and pressures mean
duction keeps decreasing. Ultimately, a maximum higher operation costs and so, alternatives that would allow
production of 4.93 mol of H2/mole of glycerol fed to the the use of milder conditions without losing process efficiency,
like membrane reactors or sorption-enhanced reactors, would
be most welcome.

Acknowledgments

Joel M. Silva acknowledges the Doctoral Program in Chemi-


cal and Biological Engineering from the Faculty of Engi-
neering of University of Porto and the Portuguese
Foundation for Science and Technology (FCT) for the PhD
grant (PD/BD/52625/2014). M.A. Soria is grateful to the FCT
for his postdoctoral grant (SFRH/BPD/88444/2012), with
financing from the European Social Fund (ESF) and the
Human Potential Operational Program (POPH). The authors
especially acknowledge Doctor Chin Kui Cheng for providing
the experimental data necessary for the validation of the
phenomenological model.

Appendix A. Supplementary data


Fig. 10 e Effect of pressure on the amount of hydrogen
produced per glycerol fed to the reactor at different Supplementary data related to this article can be found at
temperatures and WGFR of 9. http://dx.doi.org/10.1016/j.ijhydene.2015.11.055.

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 5 ) 1 e1 1 11

references [14] Perry RH, Green DW. Perry's chemical engineer's handbook. 7
ed. , New York: McGraw Hill; 1999.
[15] Wakao N, Funazkri T. Effect of fluid dispersion coefficients
on particle-to-fluid mass transfer coefficients in packed
[1] Silva JM, Soria MA, Madeira LM. Thermodynamic analysis of
beds: correlation of sherwood numbers. Chem Eng Sci
glycerol steam reforming for hydrogen production with in
1978;33(10):1375e84.
situ hydrogen and carbon dioxide separation. J Power
[16] Appendix C - physical properties of liquids and gases, in
Sources 2015;273(0):423e30.
Ludwig's applied process design for chemical and
[2] Dieuzeide ML, Amadeo N. Thermodynamic analysis of
petrochemical plants. In: Coker AK, editor. 4th ed. Boston:
glycerol steam reforming. Chem Eng Technol
Gulf Professional Publishing; 2010. p. 757e92.
2010;33(1):89e96.
[17] Sutherland W. The viscosity of gases and molecular force.
[3] Sundari R, Vaidya PD. Reaction kinetics of glycerol steam
Phylosophical Mag 1893;36:507e31.
reforming using a Ru/Al2O3 catalyst. Energy & Fuels
[18] Froment GF, Bischoff KB, De Wilde J. Chemical reactor
2012;26(7):4195e204.
analysis and design. 3rd ed. New York: John Wiley & Sons;
[4] Pant KK, Jain R, Jain S. Renewable hydrogen production by
2011.
steam reforming of glycerol over Ni/CeO 2 catalyst prepared
[19] Smith JM, Van Ness HC, Abbott MM. Introduction to chemical
by precipitation deposition method. Korean J Chem Eng
engineering thermodynamics. 7 ed. , Singapore: McGraw-
2011;28(9):1859e66.
Hill; 2005.
[5] Silva JM, Soria MA, Madeira LM. Challenges and strategies for
[20] Glycerine - Boiling and Freezing Points. [cited 2015 18/02];
optimization of glycerol steam reforming process. Renew
Available from: http://www.engineeringtoolbox.com/
Sustain Energy Rev 2015;42(0):1187e213.
glycerine-boiling-freezing-points-d_1590.html.
[6] Adhikari S, Fernando SD, Haryanto A. Kinetics and reactor
[21] Dave CD, Pant KK. Renewable hydrogen generation by steam
modeling of hydrogen production from glycerol via steam
reforming of glycerol over zirconia promoted ceria supported
reforming process over Ni/CeO2 catalysts. Chem Eng Technol
catalyst. Renew Energy 2011;36(11):3195e202.
2009;32(4):541e7.
[22] Wang C, Dou B, Chen H, Song Y, Xu Y, Du X, et al. Hydrogen
[7] Caglar OY, Demirhan CD, Avci AK. Modeling and design of a
production from steam reforming of glycerol by NieMgeAl
microchannel reformer for efficient conversion of glycerol to
based catalysts in a fixed-bed reactor. Chem Eng J
hydrogen. Int J Hydrogen Energy 2015;40:7579e85.
2013;220:133e42.
[8] Bakhtiari M, Zahid MA, Ibrahim H, Khan A, Sengupta P,
[23] Cheng CK, Foo SY, Adesina AA. Steam reforming of glycerol
Idem R. Oxygenated hydrocarbons steam reforming over Ni/
over Ni/Al2O3 catalyst. Catal Today 2011;178(1):25e33.
CeZrGdO2 catalyst: kinetics and reactor modeling. Chem Eng
[24] Cheng CK, Foo SY, Adesina AA. H2-rich synthesis gas
Sci 2015;138:363e74.
production over Co/Al2O3 catalyst via glycerol steam
[9] Iliuta I, Radfarnia HR, Iliuta MC. Hydrogen production by
reforming. Catal Commun 2010;12(4):292e8.
sorption-enhanced steam glycerol reforming: sorption
[25] Sutar PN, Vaidya PD, Rodrigues AE. Glycerol-reforming
kinetics and reactor simulation. AIChE J 2013;59(6):2105e18.
kinetics using a Pt/C catalyst. Chem Eng Technol
[10] Cheng CK, Foo SY, Adesina AA. Glycerol steam reforming
2010;33(10):1645e9.
over bimetallic CoNi/Al2O3. Industrial Eng Chem Res
[26] Smith JM. Chemical engineering kinetics. 3 ed. , Singapore:
2010;49(21):10804e17.
McGraw-Hill; 1981.
[11] Pushnov AS. Calculation of average bed porosity. Chem
[27] Ebshish A, Yaakob Z, Taufiq-Yap Y, Bshish A. Investigation of
Petroleum Eng 2006;42(1e2):14e7.
the process conditions for hydrogen production by steam
[12] Edwards MF, Richardson JF. Gas dispersion in packed beds.
reforming of glycerol over Ni/Al2O3 catalyst using response
Chem Eng Sci 1968;23(2):109e23.
surface methodology (RSM). Materials 2014;7(3):2257.
[13] Poling BE, Prausnitz JM, O'Connell JP. The properties of gases
and liquids. 5 ed. , New York: McGraw-Hill; 2001.

Please cite this article in press as: Silva JM, et al., Steam reforming of glycerol for hydrogen production: Modeling study, International
Journal of Hydrogen Energy (2015), http://dx.doi.org/10.1016/j.ijhydene.2015.11.055