Chemical Engineering Journal 281 (2015) 1036–1044

Contents lists available at ScienceDirect

Chemical Engineering Journal

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

Process simulation and optimization of H2 production from ethanol

steam reforming and its use in fuel cells. 2. Process analysis and
optimization
Ilenia Rossetti ⇑, Matteo Compagnoni, Mauro Torli
Dip. Chimica, Università degli Studi di Milano, INSTM Unit Milano-Università and CNR-ISTM, via Golgi 19, 20133 Milano, Italy

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Process layout has been suggested for

H2 production, purification and
feeding to FC.

Heat exchange intensification has
been considered.

Electrical and thermal output has
been considered for diluted water–
ethanol mixtures.

The use of diluted ethanol proved
effective for higher efficiency.
HT-WGS = high temperature water gas shi

LT-WGS = low temperature water gas shi

METH = methanaon reactor

a r t i c l e i n f o a b s t r a c t

Article history: The feasibility of power cogeneration through fuel cells using bioethanol with different concentration has
Available online 22 August 2015 been considered. Data and layout have been inspired by an existing unit Helbio, GH2-BE-5000
(5 kWelectrical + 5 kWthermal) system for combined heat and power generation (CHP). The system is consti-
Keywords: tuted by six reactors connected in series for hydrogen production and purification and by a fuel cell of the
Bio-ethanol steam reforming mentioned capacity.
Process simulation To evaluate process efficiency and the possibility to operate with diluted bioethanol feed, characterized
H2 production
by lower purification cost, different process layouts have been tested. Particular attention is paid to the
Fuel cells
intensification of the heat exchange network, to increase the overall plant efficiency. Heat supply to the
steam reformer has been accomplished by burning part of the reformate, since diluted ethanol is not
suitable to feed the burner as in the experimental process layout.
The water/ethanol feeding ratio has been taken as major parameter for simulation. An increase of this
variable improved H2 yield due to promotion of the water gas shift reaction and lower impact of the
hydrogen-consuming methanation step. However, higher heat input was required by the reformer, imply-
ing the delivery of a higher fraction of the reformate to the burner instead than to the fuel cell. This means
lower electric output and efficiency. However, the presence of a high enthalpy steam exhaust increased the
available thermal output, with consequent increase of the thermal and overall efficiency of the plant.
Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction

In order to find out alternative routes for the co-generation of

heat and power (CHP) from renewable sources, different strategies
⇑ Corresponding author. Fax: +39 02 50314300.
have been proposed. Among these, H2 production from bioethanol,
E-mail address: [email protected] (I. Rossetti).

http://dx.doi.org/10.1016/j.cej.2015.08.045
1385-8947/Ó 2015 Elsevier B.V. All rights reserved.
 I. Rossetti et al. / Chemical Engineering Journal 281 (2015) 1036–1044 1037

Table 1
Results of Model 3 implementation and estimated kinetic and adsorption parameters.

mol s1 g1 J mol1 J mol1

Model 3
k05 1.544E+20 Ea5 302,980 C0 2.926E02 D HC 55,199
k07 1.920E+05 Ea7 41,605 F0 2.412E04 D HF 76,661
k013 7.756E+09 Ea13 187,783 G0 9.940E+01 D HG 13,965
k014 5.044E+05 Ea14 56,252 H0 2.322E+00 D HH 27,945
I0 4.907E02 D HI 67,738
M0 1.369E01 D HM 32,808
N0 1.660E05 D HN 16,489

coupled to fuel cells raised considerable attention in recent years vector [22]. Alternatives for the SRE reactor may be a unit for
[1–3]. In addition, highly innovative solutions for the production ethanol dehydrogenation to acetaldehyde, followed by reform-
of second generation biofuels are becoming available, leading to ing of the latter [23] or autothermal reformers. A multichannel
environmentally, ethically and economically sustainable bioetha- reactor has been also considered in the literature, with detailed
nol. The economical plan proposed by Biochemtex, for instance, modeling [24].
is based on 0.3 euro/L for the production of lignocellulosic anhy-
Water gas shift (WGS) reactors in variable number to ensure
drous bioethanol [4]. proper H2 yield and reformate purification from CO. Typically,
A 250 W system based on autothermal reformer and a fuel cell a first reactor aims at increasing H2 yield and it is usually oper-
stack has been studied [5]. A minimum amount of process controls ated at relatively high temperature (350–400 °C). WGS is an
and little internal heat integration kept system architecture simple, exothermal reaction and under these conditions ca. 90% of the
as required for portable applications, at difference with the pre- CO outflowing from the SRE reactor may be converted to CO2.
sently studied system, in which heat integration was part of the opti- Subsequent reactor(s), working at decreasing temperatures,
mization. Indeed, for stationary applications the increase of abate CO concentration in the reformate to meet the specifica-
efficiency is seen as a predominant factor with respect to tions of the fuel cell. The catalysts are usually based on Fe–Cr
simplification. oxides for the high temperature stage (HT-WGS) and Cu–ZnO
A similar system has been proposed, with reformate purifica- for the low temperature one (LT-WGS) [25].
tion from CO based on preferential oxidation and attention to the
A preferential oxidation (PROX) reactor, or, alternatively, a
control logic and heat integration [6–8]. On a completely different selective methanation (METH) reactor is commonly added if
scale, the technical feasibility of using existing steam reforming fuel cells operating at low temperature are used. The purpose
and hydrogen separation technologies to produce hydrogen from of these units is to reduce the content of CO in the reformate
bioethanol at industrial level (100,000 N m3/h) has been explored below 20 ppm, i.e. the threshold CO tolerance of a Polymer Elec-
[9]. The product distribution in a steam reformer as a function of trolite Membrane FC (PEMFC). However, both options present
water/ethanol feeding ratio, possibly including a carrier gas, has some drawbacks. In the case of PROX [26]
been simulated [10]. Moreover, very recently thermodynamic anal-
ysis and process simulation of a reactor producing reformate by 2CO þ O2 ! 2CO2 ðR1Þ
oxidative reforming of n-butanol has been carried out [11], coupled
O2 demand is usually overstoichiometric leading to some H2
with an experimental study on oxidative reforming of ethanol [12]
depletion. Moreover, inert N2 coming from air further dilutes
and n-hexadecane [13] in microreactors to feed micro-fuel cell sys-
the reformate. By contrast, in the case of METH [27]
tems. Computational fluid dynamics simulation of ethanol steam
reforming in catalytic wall microchannels has been performed on CO þ 3H2 $ CH4 þ H2 O ðR2Þ
a Co3O4–ZnO catalyst [14].
Membrane reactors also attracted attention for similar applica- the consumption of H2 depends on residual CO amount in the
tions [15–17]. Structured membranes can allow significant H2 reformate. In these units the parallel methanation of CO2 may
purification from CO and CO2. also occur. The latter reaction may be keep under control by tun-
More in general, a typical layout of a CHP system is composed ing catalyst selectivity and operating temperature (ca. 200–
of: 215 °C). In addition, the low CO concentration may favor the
reverse WGS with sudden methanation of the CO thus formed

A multi-tubular reactor filled with an ethanol steam reforming [28]. The most commonly used methanation catalysts are based
(SRE) catalyst. The heat of reaction is provided on the outer wall on Ni, Ru or Rh supported over oxides (e.g. alumina). Another
of the tubes by combustion of part of the reactant (C2H5OH with possible alternative may be the physical separation of CO by
high concentration), as described e.g. in [18]. It is alternatively Pressure Swing Adsorption (PSA), which is however indicated
possible to use part of the reformate from the reactor or a por- when pressure gradients are significant and admitted.
tion of the H2 produced [19]. The most innovative designs pro-
A stack of FC: usually PEMFC, also as new HT-PEMFC (High Tem-
vide a catalytic burner, the catalysts for the catalytic perature PEMFC), SOFC (Solid Oxide FC) or MCFC (Molten Car-
combustion of C2H5OH being coated on the outer surface of bonate FC). These cells have widely different operating
the reformer tubes, in very efficient thermal contact with the temperatures and they allow operation with increasing
reforming catalyst which is coated on the internal skin of the amounts of CO (from ppm to percent), in the order: 60–80 °C,
same tubes [20]. Different possible configurations strictly 160–180 °C, 800–1000 °C and 650–800 °C. Therefore, according
depend on the size of the system. Another solution, feasible only to the type of FC, the number of WGS units ranges from 1 to 3
with high temperature fuel cells, such as the solid oxide ones, and PROX/METH reactors may be unnecessary.
proposes the use of the stack effluent to heat up the reforming
reactor. A further possibility is the use of an afterburner for the The main purpose of this work is to quantify the electric power
fuel cell (FC) effluent [21] or the use of molten salts as thermal and thermal energy output, as well as the overall efficiency of a
1038 I. Rossetti et al. / Chemical Engineering Journal 281 (2015) 1036–1044

plant for CHP with residential size (5 kWelectrical + 5 kWthermal). In The expressions for the coefficients A–N appearing in the rate
particular, the system is based on a PEMFC, fed with reformate pro- equations above are reported in the original paper [30]. The rate
duced by steam reforming of bioethanol. Reformate purification expressions for WGS and SRM are much more complex than those
from CO is accomplished by a series of WGS and METH reactors. usually derived for such reactions. WGS and SRM in this case are
Different tools have been used, such as Aspen PlusÓ, MatlabÓ, not occurring alone on catalyst surface, but they are part of a com-
Athena Visual StudioÓ for process simulation and analysis. The plex reaction mechanism. Based on a Langmuir Hinshelwood
operational variables chosen for the simulation are taken by an approach, all the species concurring for adsorption over active sites
actually existing unit GH2-BE-5000 (Helbio SA, Hydrogen and should appear in the denominator of the rate expressions and are
Energy Production Systems), better described in the following included in the overall balance on the active sites, leading to com-
and capable of delivering the required output [18,29]. The thermo- plex rate equations.
dynamic and kinetic input data have been selected as detailed in This model has been applied by us to a full set of experimental
the first part of this work, in particular by using Model 3. data collected for a Ni/Al2O3 sample [31]. This allowed to represent
Among the different operating variables considered, the water/ with good accuracy the experimental data, validating the proposed
ethanol ratio was found particularly relevant to optimize process model also for a different catalytic system, and to provide a reliable
yield and its economic sustainability. estimate of the whole set of kinetic parameters needed for the pre-
sent simulations and for steam reformer reactor sizing, as sum-
2. Models and methods marized in Table 1. Indeed, the required data were not fully
available in the literature.
2.1. Kinetic model

A detailed revision of kinetic models available was necessary, 2.2. Layout of the CHP unit
since process simulation including reactor sizing in the case of the
steam reforming reactor requires a suitable reaction scheme with In the present work, we concentrated on a CHP unit Helbio,
the relative kinetic parameters. Therefore, the kinetic model has GH2-BE-5000 [18]. It is composed of a prereformer, a reforming
been adapted from literature as extensively described in part 1, reactor, two units of HT- and LT-WGS and two selective METH
and labeled as Model 3. Briefly, such model was originally developed reactors, all connected in series. To provide reformate to feed a
for a Rh(1 wt%)MgAl2O4/Al2O3 catalyst [30] and includes 14 elemen- 5 kWel FC, the plant is fed with 142 g/min of a solution with
tary steps, 4 of which were proposed as rate determining ones. H2O/C2H5OH = 5.7 mol/mol. Heat supply to the reformer is ensured
The following set of rate equations have been proposed for by the catalytic combustion of 96 vol% C2H5OH. The reformate is
ethanol decomposition (ED), ethanol steam reforming (SRE), fed to a PEMFC operating at 80 °C and 1.8 bar. A heat recovery sys-
methane steam reforming (SRM) and water gas shift (WGS): tem is present to increase the overall process efficiency.
The process flowsheet used in the following simulations has
CH3 CH2 OH ! CH4 þ CO þ H2 ðEDÞ been modified, being based on a single SRE reactor and one METH,
since during our experimental testing we have seen that the sec-
CH3 CH2 OH þ H2 O ! CO2 þ CH4 þ 2H2 ðSREÞ ond METH is only a guard reactor, because the reformate meets
the specifications for CO concentration already after the first one.
CH4 þ H2 O¡CO þ 3H2 ðSRMÞ Additionally, we have redrawn the heating and heat recovery sys-
tem in order to allow the use of diluted ethanol solutions, unsuit-
CO þ H2 O¡CO2 þ H2 ðWGSÞ
able for the catalytic burner used in the experimental set up.

½pC2 H5 OH  2
k5 N 0:5 ½C T 
½pCH4 ½pH2 
r ED ¼ h i h i h i h i0:5 2 ð1Þ
½pCH4  ½pH2 O  ½pC H OH  ½pC H OH  ½pC2 H5 OH 
1 þ A pC2 H5 OH þ C 0:5 þ F 0:5 þ G pCH4 þ H½pCO  þ I pCO2 þ J pH2 þ L 2 5 0:5 þ M 2p 5 þ N
½pH2  ½pH2  ½pH2  ½ H2  ½pCH4 ½pH2 
0:5

½pC2 H5 OH ½pH2 O 
k7 NF ½C T 2
½pCH4 ½pH2 
r SRE ¼ h i h i h i h i0:5 2 ð2Þ
½pCH4  ½pH2 O  ½pC2 H5 OH  ½pC2 H5 OH  ½pC2 H5 OH 
1 þ A pC2 H5 OH þ C þ F þ G p þ H ½ p  þ I p þ J p þ L þ M þ N
½pH2 
0:5
½pH2 
0:5 CH4 CO CO2 H2
½pH2 
0:5
½pH2  ½pCH4 ½pH2 
0:5

 h i2 
½pCH4 ½pH2 O 
k13 CF  1
½ p 
CO pH2 ½C T 2
½pH2  K SRM
r SRM ¼ h i h i h i h i0:5 2 ð3Þ
½pCH4  ½pH2 O  ½pC H OH  ½pC H OH  ½pC2 H5 OH 
1 þ A pC2 H5 OH þ C 0:5 þ F 0:5 þ G pCH4 þ H½pCO  þ I pCO2 þ J pH2 þ L 2 5 0:5 þ M 2p 5 þ N
½pH2  ½pH2  ½pH2  ½ H2  ½pCH4 ½pH2 
0:5

 h ih i0:5 
½pCO ½pH 
k14 HF 2O
0:5  K WGS
1
pCO2 pH2 ½C T 2
½pH2 
r WGS ¼  h i h i h i h i0:5 2 ð4Þ
½pCH  ½ pH  ½pC H OH  ½pC H OH  ½pC2 H5 OH 
1 þ A pC2 H5 OH þC 4
0:5 þ F
2O
0:5 þ G pCH4 þ H½pCO  þ I pCO2 þ J pH2 þ L 2 5 0:5 þ M 2p 5 þ N
½pH2  ½pH2  ½pH2  ½ H2  ½pCH4 ½pH2 
0:5
 I. Rossetti et al. / Chemical Engineering Journal 281 (2015) 1036–1044 1039

Fig. 1. Schematic representation of process layout, including a basic representation of heat flows, for (a) the experimental GH2-BE-5000 apparatus and (b) the simplified, heat
integrated system proposed for the present simulations.

Therefore, we used reformate to heat up the reformer in the dnH2 O ¼ v ðr ER  r SRM  r WGS Þds ð6Þ
simulated plant. A sketch of the experimentally available layout
and of the modified system used for the present simulations is dnH2 ¼ v ðr ED þ 2r ER þ 3r SRM þ r WGS Þds ð7Þ
reported in Fig. 1.
For the implementation of process simulations it was necessary dnCO2 ¼ v ðr ER þ r WGS Þds ð8Þ
to use an appropriate set of kinetic equations and the relative opti-
mized parameters to describe the above processes. These models dnCO ¼ v ðrED þ r SRM  r WGS Þds ð9Þ
have been recovered in the literature, while the parameters were
obtained by regression of experimental data, as extensively dnCH4 ¼ v ðr ED þ r ER  r SRM Þds ð10Þ
described in part 1 of the present work and summarized in para-
graph 2.1. where s is the contact time, v the volumetric flow rate and ri the
rate of each reaction.
2.3. Sizing/rating of the SRE reactor The following thermal balance has been added:

v qcp dT i ¼ dwQ þ ½rED ðDr HED Þ þ rER ðDr HER Þ

A continuous downflow tubular reactor has been modeled. A
rigorous sizing of the system is out of the scope of the present þ r SRM ðDr HSRM Þ þ r WGS ðDr HWGS ÞdV ð11Þ
work. Therefore, the following approximations have been taken where r is reaction rate, q is density and cp the specific heat at con-
into account. According to the possible reactor configuration, cata- stant pressure of the reacting system at a given axial coordinate; Ti
lyst particle size and type and volumetric flow, external (turbulent the temperature inside the tubes, dwQ heat input to the reactor,
flow and similar performance with different flow rate at constant ðDr Hi Þ the reaction enthalpy of each reaction and dV the elemen-
contact time) and internal diffusional limitations have been tary volume of catalyst bed (correlated to dW, elementary catalyst
neglected (limited porosity, catalyst efficacy ca. 1). Radial and axial mass through the known catalyst density). The thermal input/out-
temperature and concentration gradients have been neglected, put may be calculated as:
assuming a plug-flow approximation.
The kinetic model defined as Model 3 in the first part of this 4dV
dwQ ¼ U i ðT e  T i Þ ð12Þ
work has been selected [30], with the set of kinetic parameters Di
estimated in part 1 and here summarized in Table 1. We recall in where Te is the temperature of the heating fluid at the same coordi-
the following the material balances for each species and the reac- nate, Ui and Di are the global coefficient of thermal exchange and
tion set used. internal pipe diameter, respectively, while 4dV/Di represents the
internal elementary heat exchange surface. Furthermore, the fol-
dnCH3 CH2 OH ¼ v ðr ED  rER Þds ð5Þ lowing thermal balance holds for the external heating medium:
1040 I. Rossetti et al. / Chemical Engineering Journal 281 (2015) 1036–1044

v e qe cpe dT e ¼ dwQ ð13Þ 3. Results and discussion

The pressure (P) profile along the catalyst bed has been taken
3.1. Flowsheet and parameters to be optimized
into account:
 
DPa 4dV Experimental data relative to the existing GH2-BE-5000 unit
dP ¼ ð14Þ
L pD2i have been presented elsewhere [18,29]. Testing allowed to validate
the integration of the six reactors to accomplish reformate produc-
where (DPa/L) represents the uniform pressure drop across the bed, tion and purification with the desired flow rate. Satisfactory H2
which has been calculated according to the Ergun equation [32,33]. purity has been achieved after methanation, CO concentration
Variable GHSV (gas hourly space velocity) values have been being well below the threshold of 20 ppm imposed by the PEMFC
used. Referring to ethanol as limiting agent with constant feed in use. However, interestingly low CO amount was also obtained
and considering the gas volume in normal conditions GHSV was after the LT-WGS stage, i.e. 0.4 vol%. This allows the direct coupling
varied between 314 and 7700 h1. the fuel processor with a HT-PEMFC, working at high temperature
and more tolerant to CO.
The flowsheet of the system is reported in Fig. 2. This conceptu-
2.4. Fuel cell
ally represents the scheme of the experimental system in use,
except for the intensification of the heat exchange network and
A PEMFC has been considered. Its efficiency g has been
the heat supply, in order to improve the overall efficiency, i.e.
calculated as follows:
including the modifications summarized in Fig. 1. Furthermore,
Erev me F Eme F the burner which furnished reaction heat to the reformer is repre-
g ¼ gV gI gmax ¼ gV gI ¼ gI ð15Þ sented in Fig. 2 as external unit operation, whereas it is conceived
Dr H Dr H
as a catalytic burner located in the shell side of the real reformer in
where gmax represents the thermodynamic efficiency, gV the poten- use. Similarly, heat recovery allowed by the heat exchangers EX4,
tial efficiency (taking into account ohmic losses, activation polariza- EX6 and EX8 is truly made in the shell side of the two WGS reactors
tion and concentration polarization), gI the current efficiency. Erev and METH unit. The start up of the unit is allowed by burning an
represents the reversible potential of the cell, me is the number of auxiliary fuel, although transient response is not considered in
equivalents of electrons transferred during the redox reaction, F is the present work.
the Faraday’s constant and DrH the enthalpy variation during the The following chemical species have been taken into account
same reaction. C2H5OH, H2O, H2, CO2, CO, CH4, O2 and N2. The Peng–Robinson
No adequate unit operation was available to model the PEMFC equation of state has been adopted, since it is particularly suited
in AspenPlusÓ, so it was treated as reactor for H2 combustion. to describe light gas mixtures in a wide temperature and pressure
The electrical work and heat were calculated from the enthalpy range. However, negligible differences have been observed by
change across the reactor, by taking into account the efficiency using different thermodynamic packages, except in some cases
parameters reported for similar systems [34]. (vide infra).

Fig. 2. Flowsheet of the simulated CHP unit. Material streams are indicated with continuous lines, heat flow with dotted lines. I = input, O = output of the relative reactor.
 I. Rossetti et al. / Chemical Engineering Journal 281 (2015) 1036–1044 1041

The feed has been set equal to the experimental value, i.e. Table 2
44 g min1 (0.92 mol min1), for C2H5OH, whereas the water molar Sizing details of the optimized reformer reactor used for the simulations.

flow has been widely varied between 5 and 14 times the molar Spatial distribution Triangular network
flow of ethanol. The steam reformer temperature has been set at Tubes number 109
750 °C. Tube length 510 mm
The lower limit is imposed by the need to operate overstoichio- Tube diameter 12 mm
metrically to promote the WGS reaction and to limit coking. As for Interaxial distance 16.16 mm
Shell internal diameter 190 mm
the latter parameter we selected the operating window as for tem- Material AISI 316
perature and water/ethanol ratio in a conservative way. Indeed,
based on our own experience and other literature data, coking
can be considered negligible over the most commonly used cata-
methanation of CO2 at 210 °C with selective commercial catalysts.
lysts when operating at T > 600 °C with stoichiometric feeding
Data for a 5%Ru–Al2O3 catalyst [28], similar to those employed in
ratio. Lower temperature is admissible in overstoichiometric
our case, report the following stoichiometry as worst acceptable
water/ethanol conditions. The upper limit is imposed by the heat
case in the chosen temperature range and this scenario has been
input needed to fully vaporize the feed: above this limit an addi-
considered as the more pessimistic condition in our simulations:
tional heater would be necessary.
Fictitious 108 mol min1 inlet flow of CH4 and H2 is imposed CO þ 0:5CO2 þ 5H2 ! 1:5CH4 þ 2H2 O ðR3Þ
by Model 3 as detailed in Part 1. A variation over two order of mag-
All these reactors have been connected to proper heat-links for
nitudes of this value did not affect the calculated concentration
heat recovery.
profiles, even in the first integration intervals. Air flow to the bur-
ner has been set 110% with respect to the stoichiometric need for
the fuel (ethanol and/or H2). 3.2. Sizing of the SRE unit
Since part of the reformate is used as fuel to heat the steam
reformer, a crucial parameter is the split ratio of the reformate More detailed sizing and analysis was done for the reformer
between the burner and the fuel cell. Indeed, on one hand it would reactor and its burner. A multitubular reactor configuration was
be preferable to feed as much as possible the FC, but this would chosen, with reforming catalyst on tube side and the hot combus-
decrease the reformer temperature with the consequent drop of tion gases on the shell side. The best results were obtained by
H2 yield. Therefore, this parameter was optimized and the air flow adopting a co-current configuration, in order to provide higher
to the burner and the FC was varied accordingly. heat amount in the first catalyst layers, where most reactants are
One pump and one compressor are considered and operated so converted. The optimized catalyst particle size was 1.2 mm, allow-
to allow 1.8 bar pressure at FC inlet. ing to achieve limited intraparticle diffusional limitations and
Different heat exchangers are present, allowing fine heat recov- acceptable pressure drop across the catalytic bed. Accordingly,
ery and decreasing the reformate supply to the burner. This is pos- tubes diameter was set ca. 10 times larger to avoid by-pass phe-
sible because a cascade cooling is compulsory to decrease stream nomena. Reactor length was set sufficiently long to neglect inlet
temperature to ca. 80 °C before feeding the FC. The configuration phenomena, but not too long to increase appreciably pressure
of heat exchangers has been always considered countercurrent drop. The tubes number was varied according to the operating con-
with a minimum difference of temperature between the fluid ditions. A summary of one possible configuration is reported in
streams of 5 °C. Particular attention is needed when dealing with Table 2.
heat exchangers characterized by liquid–vapor equilibrium, in par- A specific calculation has been carried out to determine the glo-
ticular EX2 and EX3. Indeed, the sizing of the system is usually bal heat transfer coefficient (U) across the heat exchanger reactor.
done by respecting these conditions: The resistance to heat transfer across the metallic tubes has been
neglected, so that
Hih:s:  Hoh:s: ¼ Hoc:s:  Hic:s: ð16Þ 1 1 1
¼ þ ð18Þ
U hi he
T ih:s: > T oc:s: [ T oh:s: > T ic:s: ð17Þ
hi and he being the liminar coefficients inside and outside the tubes,
labeling with h.s. and c.s. the hot and cold sides, respectively. In respectively. The former term has been calculated through the fol-
such case, an improper evaluation of the liquid–vapor equilibrium lowing equation, adapt to estimate heat transfer between a solid
may lead to physically unreliable temperature profiles across the wall and a layer of solid particles [32,33]:
heat exchanger. In these cases, an UNIQUAC model has been found k h i
more reliable to predict the behavior of aqueous solutions of etha- hi ¼ 0:203ðRePrÞ1=3 þ 0:220Re0:8 Pr0:4 ð19Þ
Dp
nol and it has been implemented in the simulation. Similar results
have been obtained with the NRTL thermodynamic package imple- The Reynolds and Prandtl numbers are calculated as follows:
menting the Wilson mixing rule.
Dp us q
Both the WGS reactors and the METH unit have been considered Re ¼ ð20Þ
as Gibbs reactors, since the experimental outflowing concentration
l
was always comparable to the equilibrium conversion. The use of
cp l
equilibrium conditions in this section should not be confused with Pr ¼ ð21Þ
k
the use of the WGS reaction rate above described. That was
intended as part of the complex reaction set for the modeling of 4v
the steam reformer reactor. Here a less complex reaction mixture us ¼ ð22Þ
npD2
is present and thermodynamic regime is reached downstream
the SR reactor. where k, q, cp and l are the thermal conductivity, density, heat
Their operating temperatures were set to 350 °C for the HT- capacity and viscosity of the fluid (all calculated at the fluid temper-
WGS, 280 °C for the LT-WGS and 210 °C for the METH reactor, ature), Dp is particle diameter, v the volumetric flow rate, n the
respectively. Literature data represent very limited unselective number of tubes in the selected configuration, D tube diameter
1042 I. Rossetti et al. / Chemical Engineering Journal 281 (2015) 1036–1044

and us the surface velocity (with empty reactor). To calculate the

same parameter for the shell side, an additional term for radiative
heat transfer (hr) has been added to the liminar coefficient (h0 e):
0
he ¼ he þ hr ð23Þ
"   #
0 ke le 0:14
he ¼ 0:25Re0:6 Pr0:33 ð24Þ
D lwe

DGe
Re ¼ ð25Þ
le

4v e qe
Ge ¼ h i ð26Þ
p D2s  nD2
where the pedix e is referred to the external fluid mixture and Ds is Fig. 4. Molar flow rate of different products outflowing the HTWGS reactor at
shell diameter. To calculate the contribution of radiative heat trans- different inlet H2O/ethanol ratio.

fer, the following equations have been used:

ðg T 4e ag T 4w Þ
hr ¼ r
0 ðT e T w Þ ð27Þ

0 ¼ ðw þ 1Þ=2 ð28Þ

r is the Stefan–Boltzmann constant, while  is a correction coeffi-

0

cient to take into account that the surface is not black. The emissiv-
ity of the pipes ðw Þ was calculated by interpolation at wall
temperature Tw [32,33] for AISI316 steel tubes. g has been calcu-
lated from nomograms as a function of CO2 and H2O partial pres-
sure in the gas, gas temperature Te and the average optical path
[32,33]. ag has been calculated similarly, but it also depends on
the Tw/Te temperature ratio. Finally, Tw has been considered con-
stant on the inner and outer skins of the pipe, because we have con-
sidered negligible resistance to heat transfer across the tube
thickness, so that:

he T e þ hi T i Fig. 5. Molar flow rate of different products outflowing the LTWGS reactor at
Tw ¼ ð29Þ different inlet H2O/ethanol ratio.
ðhi þ he Þ
The pressure drop across the catalyst bed has been calculated
according to the Ergun equation [32,33] by considering the bed
porosity e = 0.35.
The GH2-BE-5000 prototype incorporates an ethanol catalytic
combustion unit in the shell side. The rigorous modeling of this
unit has not been considered essential for the purpose of this work,
mainly because it was here substituted by reformate combustion
to allow the use of diluted bioethanol for SRE. Therefore, in the

Fig. 6. Molar flow rate of different products outflowing the COMETH reactor at
different inlet H2O/ethanol ratio.

simulation we considered an external burner for heat supply, mod-

eled as an adiabatic Gibbs reactor.

3.3. Fuel cell unit

Fig. 3. Molar flow rate of different products outflowing the SRE reactor at different The elaboration of the fuel cell system was complicated by the
inlet H2O/ethanol ratio. need of considering the non conventional electrical output of a
 I. Rossetti et al. / Chemical Engineering Journal 281 (2015) 1036–1044 1043

Table 3
Electrical and thermal power output as a result of simulations with variable water/
ethanol feeding ratio. Plant efficiency calculated with respect to the lower heating
value of ethanol.

H2O/C2H5OH inlet (mol/mol) 5 6.5 8.5 11 14

DH of the cell [W] 14,771 14,583 13,745 12,626 10,845
Thermodynamic efficiency 0.434 0.431 0.429 0.427 0.425
Electric power output of the 6417 6288 5895 5386 4609
cell [W]
Power used by PUMP and 613 614.0 614.2 614.6 614.9
compressor [W]
Net electric output [W] 5803 5674 5281 4772 3994
Thermal power output of the 8353 8296 7850 7239 6236
cell [W]
Thermal power absorbed by 436 1315 2564 4099 5996
H2O in COND [W]
Total thermal power [W] 8790 9610 10,413 11,339 12,233
Fig. 7. H2 flowrate trend after different reactors as a function of inlet water/ethanol Power lossa [W] 6452 5761 5351 4935 4818
ratio (increasing in the sense of the arrow). Electric efficiencyb 0.307 0.300 0.279 0.252 0.211
Thermal efficiencyb 0.464 0.508 0.550 0.599 0.646
Total efficiencyb 0.771 0.807 0.829 0.851 0.857
reactor. Therefore we modeled it as a reactor converting the refor- a
Referred to the DH of combustion of C2H5OH at 15 °C, to produce CO2 and H2O
mate isothermally at 80 °C, 1.8 bar. liquid at 15 °C = 1372.5 kJ mol1.
Q and L being heat and electrical work output of the cell, we b
Referred to the lower heating value of pure C2H5OH = 1234.8 kJ mol1.
took into account the specifications of a commercial unit [34],
which accounts for a 0.4 electrical efficiency. L and Q have been cal-
culated from the enthalpy variation (DH) of the reactor unit called consumption of the SRE reactor, a consistent heat amount is
cell, obtained as output of the simulation. needed for that reactor, increasing with the dilution of the feed.
For the present simulations, we have chosen to thermally feed
such reactor by splitting some reformate to the burner. According
to the higher heat input required by SRE in the case of more diluted
3.4. Simulation results ethanol solutions, the reformate withdrawn from the fuel cell feed
and sent to the burner increases, as shown in Fig. 8.
The output flow rates of interesting products estimated with The power output, electrical, thermal and global, together with
variable H2O/C2H5OH ratio are summarized in Figs. 3–6 for each the relative efficiency, are reported in Table 3. According to the
reactor. These figures may be nicely compared with H2 concentra- higher amount of reformate used as fuel to heat up the SRE reactor
tion profiles obtained upon simulation of a similar system [8] and when diluted ethanol is fed to the plant, the fraction used in the
with the experimental ones reported for the GH2-BE-5000 unit fuel cell is lower, with consequent decrease of the electrical output
[18,29]. An increase of H2 productivity is evident with increasing and efficiency. By contrast, the thermal output increases due to a
water feed. H2 yield increased from 84.3% to 97.6% when increasing higher amount of heat made available by excess steam, which
the water/ethanol feeding ratio from 5 to 14 (mol/mol). This can be may be recovered downstream. Therefore, the electrical efficiency
ascribed to the promotion of the WGS reaction as testified by the decreases from 0.307 at H2O/C2H5OH = 5 (mol/mol), to 0.211 at
corresponding decrease of CO and increase of CO2. Therefore, the H2O/C2H5OH = 14 (mol/mol), while the thermal efficiency
use of diluted ethanol is an effective tool to improve hydrogen increases from 0.464 to 0.646. As a consequence, the overall effi-
yield. The decrease of CO concentration also contributes to a lower ciency of the system increases from 0.771 to 0.857, of course at
impact of the methanation reaction, which consumes H2, further the expenses of the most valuable form of energy output, i.e. the
contributing to the improvement of the overall hydrogen yield, electrical one. The overall efficiency here reported is of course
as highlighted by Fig. 7. Form this point of view the use of diluted much higher than that experimentally achieved [18,29] due to sub-
ethanol would be the best choice to increase hydrogen productiv- stantial modification of the layout and more efficient heat
ity. However, in spite of the extensive heat exchange network integration.
introduced in the present flowsheet, aiming at decreasing the fuel The electrical efficiency here achieved was higher than what
reported [35] for a similar system with SRE heating provided by
ethanol combustion. Slightly different system and higher efficiency
has been instead reported elsewhere [19].
At last we may conclude that the use of diluted bioethanol is
technically feasible and this may open the way to a decrease of
the purification costs of this biofuel with respect to its use in inter-
nal combustion engines, for which it should be heavily dehydrated.
Investigations on this point have been recently summarized else-
where [36]. Of course in this way it is compulsory to use the refor-
mate as fuel to heat up the steam reformer, providing a different
fuel or heating system for the start up of the unit. This may be fea-
sible for stationary devices designed for continuous steady state
operation. By contrast, for automotive use it is not the right choice
due to frequent start up and the need of transporting excessively
diluted mixtures.
The use of diluted bioethanol improves the overall efficiency of
Fig. 8. H2 flowrate produced and fed to the fuel cell as a function of inlet water/ the process due to higher heat power available, but decreases the
ethanol ratio. amount of reformate which is possible to valorise in the fuel cell
1044 I. Rossetti et al. / Chemical Engineering Journal 281 (2015) 1036–1044

to produce electrical power. Therefore, these parameters should be [4] <http://betarenewables.neencloud.it/proesa/what-is>.

[5] T. Aicher, J. Full, A. Schaadt, Int. J. Hydrogen Energy 34 (2009) 8006.
taken into account in the economic evaluation of the solution.
[6] S. Biset, L. Nieto Deglioumini, M. Basualdo, V.M. Garcia, M. Serra, J. Power
Sources 192 (2009) 107.
4. Conclusions [7] L. Nieto Degliuomini, D. Zumoffen, M. Basualdo, Int. J. Hydrogen Energy 37
(2012) 14801.
[8] L. Nieto Degliuomini, S. Biset, P. Luppi, M.S. Basualdo, Int. J. Hydrogen Energy
A system was evaluated for the electrical and thermal cogener- 37 (2012) 3108.
ation from bioethanol. The apparatus is constituted by a fuel pro- [9] J.H. Oakley, A.F.A. Hoadley, Int. J. Hydrogen Energy 35 (2010) 8472.
[10] L. Hernández, V. Kafarov, J. Power Sources 192 (2009) 195.
cessor, including a steam reformer, two water gas shift and a [11] U.W. Hartley, S. Amornraksa, P. Kim-Lohsoontorn, N. Laosiripojana, Chem. Eng.
methanation reactors in series, and a PEM fuel cell. The target J. 278 (2015) 2.
power is nominally 5 kWel + 5 kWth, amenable for residential [12] T. Hou, S. Zhang, T. Xu, W. Cai, Chem. Eng. J. 255 (2014) 149.
[13] J. Thormann, P. Pfeifer, U. Kunz, Chem. Eng. J. 191 (2012) 410.
cogeneration.
[14] I. Uriz, G. Arzamendi, E. Lopez, J. Llorca, L.M. Gandìa, Chem. Eng. J. 167 (2011)
In the present work, different layouts have been tested, trying in 603.
particular to optimize the heat exchange network. The possibility [15] Y.S. Cheng, M.A. Pena, K.L. Yeung, J. Taiwan Inst. Chem. Eng. 40 (2009) 281.
[16] K.L. Yeung, R. Aravind, J. Szegner, A. Varma, Stud. Surf. Sci. Catal. 101 (1996)
to operate with diluted ethanol solutions has been checked, open-
1349.
ing the way to lighter purification strategies for bioethanol, with [17] D. Montané, E. Bolshak, S. Abelló, Chem. Eng. J. 175 (2011) 519.
decreasing production cost. Therefore, the water/ethanol feeding [18] I. Rossetti, C. Biffi, G.F. Tantardini, M. Raimondi, E. Vitto, D. Alberti, Int. J.
ratio was the main parameter varied in process simulation, in order Hydrogen Energy 37 (2012) 8499.
[19] P. Giunta, C. Mosquera, N. Amadeo, M. Laborde, J. Power Sources 164 (2007)
to check its effect on the operation of the reactors and on power 336.
output and efficiency. Heat supply to the reformer was accom- [20] E. Verykios, U.S. Patent No. 6,605,376 (2003), to Helbio S.A.
plished by feeding part of the reformate to a burner. [21] W. Jamsak, P.L. Douglas, E. Croiset, R. Suwanwarangkul, N. Laosiripojana, S.
Charojrochkul, S. Assabumrungrat, J. Power Sources 187 (2009) 190.
H2O/C2H5OH was varied between 5 and 14 mol/mol. H2 yield [22] M. De Falco, F. Gallucci, Int. J. Hydrogen Energy 35 (2010) 3463.
increased with increasing this parameter. However, more diluted [23] V.M. García, E. López, M. Serra, J. Llorca, J. Power Sources 192 (2009) 208.
solutions required higher heat input to the SRE reactor, imposing [24] Y.M. Bruschi, E. Lòpez, N.S. Schbib, M.N. Pedernera, D.O. Borio, Int. J. Hydrogen
Energy 37 (2012) 14887.
to withdraw a higher fraction of the reformate from the fuel cell. [25] T.L. LeValley, A.R. Richard, M. Fan, Int. J. Hydrogen Energy 39 (2014) 16983.
As a consequence, the electrical output and efficiency decreased. [26] A. Mishra, R. Prasad, Bull. Chem. React. Eng. Catal. 6 (2011) 1.
However, heat remained available in residual steam, which may [27] D.L. Trimm, Appl. Catal. A: General 296 (2005) 1.
[28] P. Djinovic, C. Galletti, S. Specchia, V. Specchia, Catal. Today 164 (2011) 282.
be recovered effectively and used, increasing the thermal output
[29] I. Rossetti, C. Biffi, L. Forni, G.F. Tantardini, G. Faita, M. Raimondi, E. Vitto, D.
and efficiency, with a global increase of the overall plant efficiency. Alberti, in: Proceedings of the ASME 2010 Eight International Fuel Cell Science,
In conclusion, the use of diluted bioethanol may be envisaged Engineering and Technology Conference, Fuel Cell 2010, June 14–16, 2010,
Brooklyn, New York, USA, vol. 2, 2010, p. 465.
for H2 production, provided that sufficient heat is furnished to
[30] C. Graschinsky, M. Laborde, N. Amadeo, A. Le Valant, N. Bion, F. Epron, D.
the steam reforming reactor by burning part of the reformate. If Duprez, Ind. Eng. Chem. Res. 49 (2010) 12383.
electrical output is the most valuable goal, it is better to operate [31] V. Mas, M.L. Bergamini, G. Baronetti, N. Amadeo, M. Laborde, Top. Catal. 51
with the lowest water/ethanol ratio, whereas at higher dilution (2008) 39.
[32] H. Scott Fogler, Elements of Chemical Reaction Engineering, fourth ed.,
the thermal output and plant efficiency may be maximized. Prentice Hall, 2005.
[33] L. Forni, I. Rossetti (Ed.), Fenomeni di Trasporto, Cortina, 2009.
References [34] <www.horizonfuelcell.com>.
[35] J.A. Francesconi, M.C. Mussati, R.O. Mato, P.A. Aguirre, J. Power Sources 167
(2007) 151.
[1] V.A. Kirillov, V.D. Meshcheryakov, V.A. Sobyanin, V.D. Belyaev, Yu.I. Amosov, N.
[36] I. Rossetti, J. Lasso, M. Compagnoni, G. De Guido, L. Pellegrini, Chem. Eng.
A. Kuzin, A.S. Bobrin, Theor. Found. Chem. Eng. 42 (2008) 1.
Trans. 43 (2015) 229.
[2] F. Díaz Alvarado, F. Gracia, Chem. Eng. J. 165 (2010) 649.
[3] M. Ni, D.Y.C. Leung, M.K.H. Leung, Int. J. Hydrogen Energy 32 (2007) 3238.