Sei sulla pagina 1di 15

Applied Catalysis A: General 250 (2003) 161175

Reaction kinetics and reactor modeling for fuel processing of liquid hydrocarbons to produce hydrogen: isooctane reforming
Manuel Pacheco a, , Jorge Sira b , John Kopasz c
a

Department of Rening and Petrochemicals, Center for Research and Development of the Venezuelan Oil Industry (PDVSA-Intevep), Sector el Tambor, P.O. Box 76343, Los Teques, Edo Miranda, Venezuela b Department of Mechanical Engineering, Universidad de los Andes, Mrida, Venezuela c US Department of Energy, Chemical Technology Division, Argonne National Laboratory, Argonne, IL 60439, USA Received 22 October 2002; received in revised form 21 March 2003; accepted 28 March 2003

Abstract A mathematical model was developed in the framework of the process simulator Aspen Plus in order to describe the reaction kinetics and performance of a fuel processor used for autothermal reforming of liquid hydrocarbons. Experimental results obtained in the facilities of Argonne National Laboratories (ANL) when reforming isooctane using a ceria-oxide catalyst impregnated with platinum were used in order to validate the reactor model. The reaction kinetics and reaction schemes were taken from published literature and most of the chemical reactions were modeled using the LangmuirHinshelwoodHougenWatson (LHHW) formulation to account for the effect of adsorption of reactants and products on the active sites of the catalyst. The water-gas-shift (WGS) reactor used to reduce the concentration of CO in the reformate was also modeled. Both reactor models use a simplied formulation for estimating the effectiveness factor of each chemical reaction in order to account for the effect of intraparticle mass transfer limitations on the reactor performance. Since the data in the literature on kinetics of autothermal reforming of liquid hydrocarbons using CeO2 -Pt catalyst is scarce, the proposed kinetic model for the reaction network was coupled to the sequential quadratic programming (SQP) algorithm implemented in Aspen Plus in order to regress the kinetic constants for the different reactions. The model describes the trend of the experimental data in terms of hydrogen yield and distribution of products with a relative deviation of 15% for reforming temperatures between 600 and 800 C and reactor space velocities between 15 000 and 150 000 h1 . 2003 Elsevier Science B.V. All rights reserved.
Keywords: Kinetic modeling; Reactor modeling; Reforming; Gasoline; Hydrogen

1. Introduction There is consensus in the scientic international community representing both the energy and transportation sectors that hydrogen could nd a more extended use as a fuel or an energy carrier in the future. Hydrogen is the fuel for polymer electrolyte mem Corresponding author. E-mail address: manuelpachecog@yahoo.com (M. Pacheco).

brane fuel cells (PEMFC) which are considered as the most technologically mature of the different types of fuel cells. In these fuel cells pure hydrogen or a reformate gas containing hydrogen can be fed to the anode, while oxygen is fed to the cathode and a electrochemical reaction takes place producing electricity plus heat and water as byproducts. When analyzing the commercial viability of the fuel cell technology, the more relevant issue is the fuel to feed the fuel cell rather than the fuel cell itself. Hydrogen can be produced

0926-860X/$ see front matter 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0926-860X(03)00291-6

162

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

Nomenclature Cp dp Deff,j Dm,j F(y) G h jD jH k kc kj Ki KH2 O L NPr NRe NSc Pi rj rj gas-phase heat capacity diameter of catalyst particle effective diffusivity of species j molecular diffusivity of species j objective function to be minimized during the parameter regression mass velocity of the gas per unit cross section of the reactor heat transfer coefcient at the boundary layer around the catalyst 2/3 dimensionless number (kc NSc /G) 2/3 dimensionless number (hNPr /Cp G) gas-phase thermal conductivity mass transfer coefcient at the boundary later around the catalyst reaction rate constant for reaction j adsorption constant for species i (1/atm) dissociative adsorption constant for water vapor reactor length (cm) Prandtl number (Cp /?) Reynolds number (dp G/) Schmidt number (/Dm ) partial pressure of species i reaction rate for reaction j pseudo-rst order reaction rate expression for reaction j used for the effectiveness factor calculation (rj = rj ) catalyst particle radius calculated mole fraction for the ith experimental measurement measured mole fraction for the ith experimental measurement axial coordinate along the reactor bed (cm)

Thiele modulus gas-phase thermal conductivity catalyst thermal conductivity gas-phase viscosity gas-phase density standard deviation for the ith experimental measurement

rp yi,cal yi,m z

Greek letters H parameter to quantify intraparticle temperature gradients (dened in Eq. (16)) WGS reversibility factor for the water-gasshift reaction (dened in Eq. (11)) H heat of reaction

from fossil fuels through a process of fuel reforming using natural gas, naphthas, gasolines or even heavier fuels like Diesel [1], gasoil, etc. Other pathways to hydrogen have been proposed based on water electrolysis and solar energy, reforming of biofuels or other renewable sources of fuels and energy. Regardless of the hydrocarbon feedstock being used for hydrogen production via fuel reforming, a tool is needed in order to analyze the relationship between fuel formulation and fuel reformability, between fuel reformability and nature of the catalytic system, and to determine the reaction kinetics and assess the importance of heat and mass transfer issues on reactor performance. This work is aimed at developing and validating a preliminary mathematical model to describe these phenomena. This model was validated using experimental data collected in a xed-bed microreactor packed with a noble-metal based catalyst proprietary to the US Department of Energy [2]. Since naphtha cuts are very complex mixtures of hydrocarbons, this work was initiated with the study of reformability of isooctane as a representative parafn in a naphtha cut. The reaction scheme that has been proposed in the literature for partial oxidation and steam reforming of hydrocarbons has been implemented and the unknown kinetic parameters were determined through a data regression algorithm. A detailed description of this reaction scheme will be given in the present contribution. A simplied mathematical model was implemented to account for the intraparticle diffusional limitations due to the high reaction temperatures and high reaction rates. This intraparticle diffusional limitation was accounted for by estimating effectiveness factors for each of the reactions considered in the kinetic mechanism. The low and high temperature water-gas-shift (WGS) reactors used downstream of the fuel reformer were also modeled. The coupling of thermolysis reactions that heavier

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

163

hydrocarbons undergo to produce lighter hydrocarbon molecules, which are also reformed to produce hydrogen, will be the subject of following publications.

2. Reaction kinetics for partial oxidation and steam reforming of liquid hydrocarbons Reforming of liquid hydrocarbons in order to produce hydrogen (or synthesis gas) has not been studied as deeply as that for lighter hydrocarbons like methane or natural gas. Published kinetic data for the partial oxidation or steam reforming reactions of hydrocarbons C5 to C8 is scarce. In the present contribution an attempt is made to develop a kinetic and reactor model to describe autothermal reforming of isooctane using a ceria oxide catalyst impregnated with platinum. Isooctane is used as a model molecule for a C8 naphtha or parafnic gasoline. Due to the fact that hydrocarbons undergo a thermal cracking or thermolysis [3] when in contact with a high temperature media, the hydrocarbon molecules that are actually reformed in the reactor are a spectrum of lighter hydrocarbons formed after the thermolysis process. In order to simplify the analysis, in the present contribution, it is assumed that the molecule that is actually reformed is the actual fuel fed to the reactor, i.e. isooctane. Pant and Kunzru [4,5] show that the conversion of n-heptane and methylcyclohexane due to the thermolysis reactions at temperatures from 680 to 800 C and residence time from 0.1 to 0.6 s can be as low as 10% and as high as 80% for the most severe reaction conditions. The reactor residence time for the hydrocarbon reforming reactions studied in the present contributions was between 0.02 and 0.2 s. A detailed modeling of naphtha pyrolysis coupled to the reforming reactions is underway and will be the subject of a future publication. Direct partial oxidation of isooctane can be represented as C8 H18 + 4O2 8CO + 9H2 (1)

though these reaction schemes have been proposed primarily for methane partial oxidation, the combustion route was assumed in the present contribution because of the relatively high amount of CO2 present in the reformate in the early stages of the reaction. This reaction scheme can be described as C8 H18 + 16O2 8CO2 + 9H2 O C8 H18 + 8H2 O 8CO + 17H2 C8 H18 + 8CO2 16CO + 9H2 (2) (3) (4)

Reaction 2 corresponds to the catalytic combustion of the hydrocarbon, while reactions 3 and 4 correspond to the steam and CO2 reforming reactions, respectively. Xu and Froment [9,10] when modeling steam reforming of methane on a Ni-Al2 O3 catalyst indicated that the hydrocarbon can be steam reformed to CO2 and H2 as well, that is C8 H18 + 16H2 O 8CO2 + 25H2 (5)

Besides the reactions described earlier, it is clear that the presence of CO and water in the system implies that the WGS reaction should play a role CO + H2 O CO2 + H2 (6)

In the present work reactions 2 through 6 were used to describe the partial oxidation, steam reforming and CO2 reforming that take place in the fuel processing reactor for hydrogen production. Rostrup-Nielsen [11] proposed a reaction mechanism for steam reforming of hydrocarbons where it is assumed that hydrocarbons are chemisorbed on a catalyst dual site followed by successive -scissions of the carboncarbon bonds. The resulting C1 -species react with chemisorbed steam. Using Langmuir equation for describing the surface concentration of hydrocarbon (Cn Hm ), steam and hydrogen, the following expression is obtained for the reaction rate r= kA PCn Hm (1 + (kA /kr Kw )(PH2 /PH2 O )PCn Hm +KH2 O (PH2 O /PH2 ) + KH2 PH2 )2 (7)

Two reaction schemes have been proposed to explain the partial oxidation of a hydrocarbon to CO and H2 [68]. Some authors consider that the hydrocarbon undergoes combustion followed by both steam and carbon dioxide reforming, the other mechanism is based on the catalytic pyrolisis of the hydrocarbon followed by hydrogen desorption and carbon oxidation. Even

where Pi is the partial pressure of species i, KH2 O the dissociative adsorption constant of water, KH2 the adsorption constant for hydrogen, kA the reaction rate constant for the hydrocarbon chemisorption, and kr

164

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

the reaction rate constant for the reaction between adsorbed C1 -species and chemisorbed oxygen produced from water dissociation. In the mechanism proposed by Rostrup-Nielsen [11] it is implicit that the catalyst support and the promoters enhance adsorption of steam, which is then transported or spilled over to the active metal surface H2 O + support H2 Osupport H2 Osupport+ O + H2 (8) (9)

Due to the possible strong support-metal interaction and/or to the direct activation of steam by the active metal (H2 O+ O + H2 ), the surface of the active metal may be covered by water. In order to develop Eq. (7), it was assumed that activated oxygen (O ) was the most abundant reaction intermediate and that the surface concentration of heavier hydrocarbons chemisorbed on dual sites (Cn Hz 2 ) was negligible. The basic assumption proposed by Rostrup-Nielsen [11], regarding successive -scissions of the carboncarbon bonds in the heavier hydrocarbon resulting only in C1 -species, is perhaps an important simplication of the actual mechanism because, as Savage [3] describes, there is a family of hydrocarbon pyrolysis reactions that includes homolytic dissociation of the CC bonds, radical recombination (reverse reaction), -scission, radical disproportionation, isomerization, hydrogen abstraction, among others. The magnitude of the CH and CC bond dissociation energies of the different bonds present in the hydrocarbon molecule is what determines the kinetics and product distribution in the hydrocarbon pyrolysis reactions. Another assumption implicit in Eq. (7) is that chemical reversibility is negligible. Despite these simplications in the Rostrup-Nielsen mechanism for steam reforming of higher hydrocarbons, Eq. (7) has been the basis for building reaction kinetic models for steam reforming of hydrocarbons. Using a combination of mass spectrometry and gas cromatography, Kopasz et al. [12] have shown that during autothermal reforming of isooctane and other liquid hydrocarbons using a Pt-CeO2 catalyst, C1 to C4 hydrocarbons are the dominant species that result from the cleavage of the original fuel molecules during the process of thermolysis. This indicates that a spectrum of lighter hydrocarbons are produced and

then reformed. However, the introduction of all these lighter hydrocarbons in the kinetic model would make it much harder to handle especially when modeling reforming of a complex hydrocarbon blend like gasoline or naphtha. Simplications are needed for computational calculations. Most published work on steam reforming of light and heavier hydrocarbons use rate expressions similar to Eq. (7) to explain the experimental observations. Avci et al. [13] used rate expressions similar to Eq. (7) to model steam reforming of methane, propane and octane; however, they did not consider reversibility in the kinetic expressions used for steam reforming of hydrocarbons. Only recently [14] experimental measurements on steam reforming and autothermal reforming of liquid hydrocarbon fuels using a proprietary Pt-Rh-Al2 O3 catalyst have been published; however, no kinetic or reactor modeling was presented in this publication. Due to the lack of published kinetic data specically for the Pt-CeO2 catalyst under study in the present contribution, it was necessary to use the general form of the rate expressions developed for other catalytic systems and regress the kinetic parameters in order to adjust the model predictions to the experimental data. Table 1 shows the rate expressions used in the present work for the kinetic modeling of the reforming reactions, the catalyst used and the respective reference. It has to be emphasized that the kinetic expressions described in Table 1 have been developed during studies of partial oxidation and steam reforming of methane and the general functionality of these expressions has been adopted in this work for the equivalent reactions involving isooctane. As it was mentioned above, Rostrup-Nielsen [11] showed that the experimental data for light and heavy hydrocarbons can be explained by using a Langmuir-type mechanism to describe the adsorption and reaction of reactants and product on the catalyst active sites. The kinetics model used by Xu and Froment [9] implemented in this contribution is consistent with this mechanism.

3. Reaction kinetics for low and high temperature water-gas-shift Since the goal of this work was to develop a mathematical model for a fuel processor for hydrogen

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175 Table 1 Rate expressions and references used in the reformer modeling Reaction C8 H18 + 16O2 8CO2 + 9H2 O C8 H18 + 8H2 O 8CO + 17H2 C8 H18 + 8CO2 16CO + 9H2 C8 H18 + 16H2 O 8CO2 + 25H2 CO + H2 O CO2 + H2 Expression for ri r1 = k1 PiC8 PO2 r2 = k2
2.5 P H2 3 PiC8 PH2 O PH2 PCO /K1

165

T ( C) 800900

Catalyst Ni/Al2 O3

Reference [6]

(1 + KCO PCO + KH2 PH2 + KiC8 PiC8 + KH2 O PH2 O /PH2 )2


2 2 PCO PH2

500750

Ni/Mg Al2 O3

[9,10]

r3 = k3 PiC8 PCO2 1 r4 = r5 = k4
3.5 P H2

K3 PiC8 PCO2
2 4 PiC8 PH2 O PH2 PCO2 /K4

800900

Ni/Al2 O3

[6]

(1 + KCO PCO + KH2 PH2 + KiC8 PiC8 + KH2 O PH2 O /PH2 )2 PCO PH2 O PH2 PCO2 /K5 (1 + KCO PCO + KH2 PH2 + KiC8 PiC8 + KH2 O PH2 O /PH2 )2

500750

Ni/Mg Al2 O3

[9,10]

k5 PH2

500750

Ni/Mg Al2 O3

[9,10]

production with emphasis on fuel cell applications, water-gas-shift (WGS) reactors have to be used downstream of the fuel reformer in order to increase the overall efciency of hydrogen production and to lower the CO content in the reformate which could irreversibly deactivate the catalyst used in the anode of polymer electrolyte membrane (PEM) fuel cells if the CO concentration is higher than about 100 ppm [15]. In this contribution two water-gas-shift reactors (also called shifters), a high temperature shift (HTS) and a low temperature shift reactor (LTS) were modeled downstream from the reformer. The reaction kinetics for the WGS reaction under high and low temperature conditions and for a wide variety of catalyst has been studied by Keiski and coworkers [16,17]. A commercial iron-oxide/chromium-oxide catalyst as well as a copper-based catalyst were studied by these researchers. By varying the reaction temperature, H2 O-to-dry gas ratio, space velocity, and CO, CO2 , and/or H2 concentration in the gas fed to the shifters, the following power-law type of rate expression was developed by Keiski and coworkers:
n m rWGS = kWGS CCO CH2 O CCO2 CH2 (1 WGS ) p q

Depending on the temperature range and catalyst used the kinetic parameters and reaction orders of reactants and products used in Eq. (10) change. These parameters and orders were implemented in the present work for the modeling of the high temperature and low temperature shifters. This kinetic model was initially coupled through a user-supplied subroutine to the Aspen Plus plug-ow reactor model in order to simulate the high and low temperature water-gas-shift reactors and validate the models using the experimental data [16,17].

4. Mass and heat transfer issues Due to the high temperature at which the reforming and WGS reactions take place, the important heat effects, and large reaction rates; it is expected that diffusional limitations and heat transfer resistance inside the catalyst and at the boundary layer around the catalyst play important roles in the reactor performance. The relationship between the importance of mass transfer resistance through the boundary layer at the surface of the catalyst and temperature difference between catalyst surface and bulk uid may be easily derived for steady-state conditions [18]. kc (Co Cs)( Hr ) = h(T s T o) (12)

(10)

where Ci is the molar concentration of species i, WGS is the reversibility factor that accounts for the approach to chemical equilibrium. WGS = CCO2 CH2 KT CCO CH2 O (11)

and KT is the equilibrium constant.

where kc is the mass transfer coefcient at the boundary layer, Co and Cs are the reactant concentrations at

166

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

the bulk uid and catalyst surface; respectively, Hr is the heat of reaction, h is the heat transfer coefcient at the boundary layer around the catalyst, and Ts and To are the temperature at the catalyst surface and bulk uid, respectively. Using the correlations commonly accepted for estimating the heat and mass transfer coefcients in packed beds, Eq. (12) can be written in terms of the Schmidt and Prandtl numbers as follows: (Ts To ) = jD jH NPr NSc
2/3

(Co Cs )

Hr Cp (13)

From Eq. (13) it can be seen that the maximum temperature difference between the catalyst surface and the bulk uid can be estimated by considering a completely mass-transfer controlled reaction (Co Cs ) due to the external mass transfer resistance. Therefore, the extent of the heat transfer resistance at the boundary layer around the catalyst can be estimated by the term (Ts To )max = jD jH NPr NSc
2/3

H r Co Cp

(14)

Table 2 shows the range of values of thermophysical properties like gas-phase (reformate) density, viscosity, heat capacity, and estimated values for heat and mass transfer coefcients at the boundary layer. This range of values is for temperatures between 600 and 800 C and for a reactor pressure of 6 psig. ConsidTable 2 Thermophysical and transport properties of the reformate and estimated heat and mass transfer coefcients at the boundary layer 600 C Gas density (kmol/m3 ) Gas viscosity (g/cm2 s) Gas heat capacity (cal/gmol K) Gas thermal conductivity (cal cm/(s cm2 K)) Schmidt number (using isoctane diffusivity) Prandtl number Reynolds number Nusselt number for heat transfer Gassolid heat transfer coefcient (cal/(cm2 s K)) Mass transfer coefcient using isooctane diffusivity (cm/s) 0.01434 3.56E4 7.647 5.63E4 1.72.1 0.45 0.751.5 2.5 2.5E2 120 800 C 0.01167 4.09E4 7.312 7.07E4

ering that the key component to make these calculations is isooctane and considering that the overall heat of reaction for autothermal reforming is of the order of 5% of the heat of reaction for partial oxidation (40 kcal/mol); the maximum temperature gradient through the boundary layer is about 19 C at the entrance of the reformer. This maximum temperature gradient at the boundary layer rapidly drops to about 6 C at half the bed length and 3 C at the bed exit. This indicates that under the conditions of fuel reforming studied in this work, temperature gradients (and therefore heat transfer resistance) could become important only close to the region where the fuel is fed to the reformer. Also, it is important to mention that Eq. (13) emphasizes that if the heat of reaction is large, mass transfer limitations at the boundary layer on the catalyst surface may be small, yet heat transfer can still be important. This leads to the conclusion that for most of the reactor bed, both heat and mass transfer resistance at the boundary layer could be neglected for modeling purposes under the conditions analyzed in the present study. This assumption may not be correct close to the reactor inlet. It is important to mention that these calculations of heat transfer resistance were performed considering the worst case where there is a complete mass transfer resistance at the boundary layer. It is likely that the actual mass transfer resistance will be shared between the external boundary layer and the intraparticle mass transfer; therefore the external temperature gradient will be lower than that estimated earlier. The other important working hypothesis in this assessment of the heat/mass transfer effect is the overall heat of reaction. A fraction of the heat of reaction for the partial oxidation is considered because this work is concerned with the autothermal reforming of isooctane where the heat released by the partial oxidation reactions is used for the endothermic steam reforming reactions. Only a slightly higher O2 /C ratio is used to maintain the system just slightly exothermic. With respect to the intraparticle heat transfer resistance it can be shown that under steady-state conditions the diffusion ux of reactants across a boundary layer surrounding some fraction of the catalyst porous structure equals the rate of reaction within the surface. Also, the heat released or consumed by the reaction must all be transferred across the same boundary. Sattereld [18] mathematically represents these transport

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

167

phenomena in the following form (T Ts ) = Deff (Cs C) Hr (15)

5. Effectiveness factor model Due to the complexity of the kinetic expressions described in Table 1 the integration of the mass transfer equations that describe the intraparticle mass transfer resistance represents a challenge especially because the solution of the intraparticle problem has to be coupled with the interparticle description in order to model the reactor. In this work a simplied approach was used for estimating the effectiveness factor for each reaction throughout the reactor bed. This approach is based on the use of reversible pseudo-rst order expressions for the reaction rates instead of the LangmuirHinshelwoodHougenWatson (LHHW) formulation only for the purpose of estimating the effectiveness factor. In this approach the LHHW expressions are rewritten in a more simplied form that has an analytical solution for the effectiveness factor. This approach can be explained with an example. The expression for describing the reaction rate for steam reforming of isooctane to produce CO and hydrogen (Eq. (3)) described in Table 1 is
3 PiC8 PH2 O PH2 PCO /K1

where T is the temperature at any point within the catalyst particle, C the molar concentration of the reacting species at the same point, Hr the enthalpy change on reaction, and the thermal conductivity of the porous catalyst. Ts and Cs are the boundary values at the outer surface of the catalyst. From Eq. (15) a dimensionless parameter can be derived that quanties the effect of intraparticle heat transfer resistance and the maximum temperature gradient within the catalyst. Sattereld [18] represents this parameter as H = Hr Cs Deff Ts (16)

Sattereld [18] summarizes data of thermal conductivity of a wide variety of catalysts in powder form k2
2.5 PH2

r2 =

(1 + KCO PCO + KH2 PH2 + KiC8 PiC8 + (KH2 O PH2 O /PH2 ))2

(17)

and shows that it varies between 3E4 and 1E3 cal/(s cm C). If a thermal conductivity for the porous catalyst of 5E4 cal/(s cm C) is used, the value of parameter H is of the order of 104 to 103 for this system and under the conditions studied in the present contribution. This corresponds to a maximum temperature gradient within the catalyst of the order of 0.5 C at the reformer entrance and about 0.04 C at the end of the catalyst bed. For these calculations the effective diffusivity of isooctane was calculated considering both molecular and Knudsen diffusion. A tortuosity factor of 2 was assumed [18], a mean pore radius of 6.3E10 m was calculated given the measured pore size distribution and a catalyst void fraction of 0.050.19 was estimated based on the measured total pore volume (0.010.05 cm3 /g). This assessment indicates that intraparticle heat transfer resistance is unimportant under the conditions studied in the present work; and therefore the intrinsic reaction rates can be evaluated at the bulk uid temperature and corrected for the effect of intraparticle mass transfer resistance by using the effectiveness factor of each reaction.

This expression can be rewritten in reversible pseudo-rst order form: r2 = r2 = k2 PiC8 PCO Keq2 (18)

where k2 is the pseudo-rst order reaction constant and Keq2 is a parameter that includes the original equilibrium constant. k2 = k2 PH2 O 2.5 (1 + K P PH2 CO CO + KH2 PH2 (19)

+KiC8 PiC8 + (KH2 O PH2 O /PH2 ))2

and Keq2 = K1 PH2 O


3 PH2

(20)

Sattereld [18], Froment and Bischoff [19] indicate that the Thiele modulus for a rst order reversible ith reaction taking place in a spheric catalyst of radius rp and using the jth species as the key component is

168

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175 Table 3 Reaction conditions for autothermal reforming of isooctane

dened as i = rp 3 ki (Keqi + 1) Deff,j Keqi (21)

and that the effectiveness factor can be estimated by i = 1 (3i )coth(3i ) 1 i 3i (22)

Reactor temperature ( C) Reactor pressure (psig) Space velocity (h1 ) Catalyst size (mm) Catalyst pore radius (m) H2 O/C (mol/mol in the feed) O2 /C (mol/mol in the feed)

600, 700 and 800 57 15000150000 0.510.76 5.8E102.8E9 1.43 0.42

In this formulation the effective diffusivity of jth species is estimated considering both Knudsen and molecular diffusion 1 1 t 1 (23) = + Dk,j Dm,j Deff,j where is the tortuosity factor and is the void fraction of the catalyst, respectively. Based on the range of values for the tortuosity factor in different catalytic systems given by Sattereld [18], is assumed to be equal to 2. Using the pore size distribution of the catalyst was estimated to be equal to 0.19. It is important to notice that the intrinsic reaction rates calculated by the reversible pseudo-rst order expressions (rj ) are identical to the intrinsic reaction rates calculated using the original formulation (rj ) taken from the reaction mechanism for the partial oxidation and steam reforming reactions described earlier in the present contribution. The functionality of the intrinsic reaction rate expressions is modied only to be able to estimate the effectiveness factor from an analytical solution of the mass transfer equations that describe the intraparticle mass transfer-reaction process.

6. Model validation and data regression: autothermal reforming of isooctane Autothermal reforming of isooctane performed at the facilities of Argonne National Laboratory (ANL) was chosen as the system used for the regression of the kinetic parameters and model validation. The ANL proprietary Pt-CeO2 catalyst was used for fuel reforming in a micro-reactor of 0.42 cm of internal diameter and 5 cm long which is placed inside a furnace for temperature control. About 2 g of catalyst is crushed and packed in the micro-reactor for each experimental run. A detailed description of the experimental setup is given by Kopasz et al. [12]. Table 3 shows the reaction conditions for these experiments.

Performing an energy balance calculation that considers the exothermic heat of reaction of the partial oxidation reactions and the endothermic effect of the steam reforming reactions, it can be shown that a molar steam-to-carbon ratio of 1.43 and oxygen-to-carbon ratio of 0.42 can be used to make the system almost autothermal allowing only for a slightly higher O2 /C ratio to allow for some exothermicity in the system. The kinetic models described in Table 1 for the steam reforming, partial oxidation and CO2 reforming were implemented in a user-supplied FORTRAN subroutine of the process simulator Aspen Plus . This subroutine coupled to the rigorous plug-ow reactor model of Aspen Plus (RPLUG) was also designed to perform mass transfer calculations involving estimation of the effective diffusivity of the different species at the catalyst side, Thiele modulus and effectiveness factor for the different reactions. This kinetic subroutine was also coupled to a data-t algorithm built-in Aspen Plus which uses a version of the sequential quadratic programming (SQP) formulation in order to regress the kinetic parameters needed to adjust the model predictions to the experimental data. This regression was performed by minimizing a objective function, F(x), that quanties the difference between predicted and measured reformate compositions weighted by the inverse of the standard deviation of the measured composition.
N

F(y) = Min
i=1

yi,m yi,calc i yi,m

(24)

where yi,m and yi,calc are the measured and calculated values of the ith data point; respectively, and i is the standard deviation for the experimental measurement. After performing a sensitivity analysis on the effect of kinetic parameters, six parameters were chosen for the regression. These parameters were the ve preexponential factors of the Arrhenius-type expressions for

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

169

the reaction rate constants k1 through k5 described in Table 2, and the pre-exponential factor for the dissociative adsorption constant for water vapor at the catalyst active sites (KH2 O ). KH2 O was the only adsorption constant regressed because the sensitivity analysis performed on the model indicated that the major contribution to the adsorption term in the kinetic expressions was due to the dissociative adsorption of water. As an example, the dissociative adsorption term for water (PH2 O KH2 O /PH2 ) in the reaction rate expressions described in Table 2 accounts for about 9298% of the denominator at 800 C and GHSV = 15 000 h1 . Since only the pre-exponential factors are regressed, with this approach it is assumed that the activation energies for the different reactions and the enthalpy of adsorption of water vapor are the same in this catalytic system with respect to the values reported in the literature. It is important to mention that despite the high temperature, the inhibition effect of reactant and product adsorption on the catalyst active sites is signicant. This inhibition effect due to reactants and products adsorption is illustrated by a denominator in the reaction rate expressions that varies along the reactor bed from 12 to 150 at 800 C and 15 000 h1 . The regression of these kinetic parameters was performed simultaneously for three different temperatures and four different space velocities studied experimentally involving more than 50 experimental measurements. Reaction rate coefcients and a water adsorption dissociative coefcient greater than zero were considered as physical constraint during the data regression process. Table 4 shows the results of the data regression process when the pre-exponential factors of the reaction rate coefcients k1 through k5 described in Table 2,

Table 4 Results of the kinetic parameter regression Parameter Pre-exponential factor Activation energy and heat of adsorption of water (kJ/mol) 166 240.1 23.7 243.9 67.1 HH2 O = 88.7

k1 (mol/(gcat s bar2 )) k2 (mol bar0.5 /(gcat s)) k3 (mol/(gcat s bar2 )) k4 (mol bar0.5 /(gcat s)) k5 (mol/(gcat s bar)) KH2 O (dimensionless)

2.58E+08 2.61E+09 2.78E05 1.52E+07 1.55E+01 1.57E+04

The parameters and their respective use are dened in Table 1.

as well as, the pre-exponential factor of the water dissociate adsorption coefcient were regressed. An Arrhenius-type of expression was considered for the dependency of these parameters with temperature. The activation energies of these reactions and the heat of adsorption of water were taken from references [6] through [10] and are reported on Table 4. The minimum converged value of the objective function (Eq. (24)) was 1.27 which is equivalent to an average relative error between the measured and calculated values of about 15%, considering a standard deviation of 10% for the reformate composition measurements. The adsorption coefcients for hydrogen, CO and isooctane were considered to be the same as those reported in the cited references ([6] through [10]) because, as it was indicated before, the competitive adsorption effect of these three species accounts for a minor fraction to the whole inhibition effect with respect to the contribution of water dissociative adsorption. Figs. 1 and 2 compare the product distribution measured experimentally in the reformate stream and the model predictions for a space velocity of 15 000 and 150 000 h1 , respectively. Since in the model developed in the present work, the thermolysis reactions of heavier hydrocarbons to produce lighter hydrocarbons is not accounted for, the comparison of hydrocarbons in the reformate is based on the lumped hydrocarbons measured experimentally (C1 through C4 ) and unconverted isooctane predicted by the model. Figs. 1 and 2 show that the model is capable of quantifying the distribution of products in the reformate stream under a wide range of conditions; although the relative error in the prediction of hydrogen concentration increased from about 13% for a space velocity of 15 000 h1 to 23% at 150 000 h1 . Fig. 3 depicts a parity plot that indicates the quality of the t for the whole set of experimental data. From this comparison it was determined that the average relative error in the model prediction was 15.5% if only H2 , CO and CO2 are considered in the calculation of the deviation between model prediction and experimental data. When unconverted hydrocarbons are included, the average relative deviation between experimental data and model predictions increased to 21.9%. This shows the importance of incorporating the thermolysis reactions in order to quantify the effect of thermal cracking of the heavier hydrocarbons on the overall product distribution in the reformate.

170

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

Fig. 1. Comparison between experimentally measured reformate composition and model predictions. GHSV = 15 000 h1 and 700 C.

In order to illustrate the model performance and capabilities, Fig. 4 shows the concentration prole predicted by the reactor model at 15 000 h1 and 800 C. This gure indicates that water vapor concentration peaks early in the reactor because of the competitive contribution of consumption and production as shown in the proposed mechanism. Also, oxygen is quickly consumed in the partial oxidation reactions, and both CO and CO2 reach a concentration plateau also due to

the relative contribution of the reactions where these species are consumed and produced. At this low space velocity the model indicates that the concentration of hydrogen along the bed increases steadily. Fig. 5 describes the typical order of magnitude for the effectiveness factors of the reaction network used to describe the reformer. Close to the section where the fuel is fed to the reactor effectiveness factors converge to values of the order of 104 to 103 . However,

Fig. 2. Comparison between experimentally measured reformate composition and model predictions. GHSV = 150 000 h1 and 800 C.

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

171

Fig. 3. Parity plot of reformate composition. GHSV = 15 000150 000 h1 and 600 to 800 C.

as oxygen, water and fuel get depleted along the bed, the order of magnitude of the reaction rates is significantly reduced and the effectiveness factor increase. Particularly for the partial oxidation reaction the maxi-

mum value of the effectiveness factor of 1 is reached only at about 0.1% of the catalyst bed starting from 5 104 because of the rapid depletion of oxygen. The increase in the order of magnitude of is

Fig. 4. Predicted composition prole along the reformer bed at 15 000 h1 and 800 C.

172

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

Fig. 5. Predicted effectiveness factors for the different reactions taking place in the reformer.

more gradual for the other reactions considered in the present work. It is important to notice that the dimensionless axial coordinate in Fig. 5 had to be represented in logarithmic scale to be able to capture the details and rapid change of the effectiveness factors calculated near the feed point of the reactor where the concentration of reactants is highest. The order of magnitude of the estimated effectiveness factors indicates that signicant improvement in the reactor efciency could be gained by reducing the diffusion path to the catalyst active sites.

7. Model validation and data regression: high and low temperature water-gas-shift (WGS) A similar methodology used for validating the kinetic model for the autothermal reformer (ATR) was also used for validating the kinetic model for water-gas-shift at low and high temperatures. Eqs. (10) and (11) for describing the low and high temperature WGS reaction kinetics were implemented in a FORTRAN-based kinetic subroutine that was coupled to the rigorous plug-ow reactor model of Aspen Plus . The approximate methodology described above

for estimating the Thiele modulus and effectiveness factor was also implemented in this subroutine to account for intraparticle mass transfer resistance. The kinetic model implemented has 3 parameters (kWGS , n and m with p = q = 0 in Eq. (10)), because Keiski et al. [16,17] show that it gives better accuracy. Fig. 6 compares the experimental data of CO conversion reported by Keiski et al. at high and low temperatures with the CO conversions predicted by the model implemented in the present work. This gure shows that for the minimum and maximum reaction temperatures studied by these researchers the kinetic model agrees fairly well with the experimental data. However, for the intermediate temperature studied the deviation of the model predictions are larger. Keiski showed a similar trend with respect to the performance of his model.

8. Case study: autothermal reformer and shifter integration Once the autothermal reformer model for processing isooctane as a model molecule and the kinetic model for the water-gas-shift reaction kinetics were

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

173

Fig. 6. CO conversion along the shifter reactor used by Keiski and coworkers. CO/H2 O = 6.1, GHSV = 3260 h1 . Indicated temperatures correspond to the reactor inlet. Symbols correspond to experimental data and lines to model predictions.

validated against experimental data, an integrated model of the two types of reactors was implemented. A reformer, a high temperature shifter and low temperature shifter were integrated in a single model. A H2 O/CO molar ratio of 4 was used at the feed of the high temperature shift and no aditional water was suplied at the inlet of the LTS reactor. Table 5 describes the condition of the different reactors in this case study using conditions for the high and low temperature shifters similar to those reported by Myers et al. [20].
Table 5 Conditions used for simulating the integrated reformer and watergas-shift reactors ATR Space velocity (GHSV) h1 Temperature ( C) Molar H2 O/C in the feed Molar O2 /C in the feed Reactor volume (cm3 ) HTS Temperature ( C) Molar H2 O/CO in the feed LTS Temperature ( C) Molar H2 O/CO in the feed 15000 700 1.43 0.42 0.7 400 4 200 Outlet concentration from the HTS

Fig. 7 depicts the concentration of CO along the reformer and CO conversion reactors. It can be seen that using this conguration and operating conditions, the model predicts a reduction in the CO concentration from about 5.5% molar in the reformate to 0.17% molar in the product of the low temperature shifter. It is important to notice also the chemical equilibrium reached in the second half of the high temperature

Fig. 7. CO prole along autothermal reformer (ATR), high temperature shift (HTS) and low temperature shift (LTS) reactors predicted by the model.

174

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175

shifter responsible for the plateau in the CO concentration. Only the change in operating temperature from the HTS to the LTS can reduce the CO concentration further.

perimental data. Both the autothermal reforming and water-gas-shift reactor models were coupled to study an integrated case including the high and low temperature water-gas-shift reactors.

9. Conclusions A pseudo-homogeneous model that includes description of the reaction kinetics and intraparticle mass transfer resistance was developed and validated for autothermal reforming of a liquid hydrocarbon and the water-gas-shift reaction to be used in the process of hydrogen production via catalytic reforming of a hydrocarbon-base fuel. For the description of the reformer, the model uses the LangmuirHinshelwood HougenWatson formulation as described in the literature for representing most of the reaction network, while for the CO converters the kinetic model is based in a semi-empirical formulation also previously published. Kinetic parameters for the reformer reaction network were regressed using experimental data under a wide range of conditions for isooctane autothermal reforming using a Pt-CeO2 catalyst. Kinetic parameters for the water-gas-shift model were used as published in the literature. The model implemented for describing the performance of the reformer is able to predict the available experimental data for the distribution of H2 , CO and CO2 in the reformate with a deviation of about 15%. The effectiveness factors for the different reactions, estimated using a reversible pseudo-rst order approximation for the more complex kinetics, was of the order of 104 to 103 in the rst stages of the reformer adjacent to the reactor feed point. Larger effectiveness factors were estimated closer to the reactor exit due to the decrease of the reaction rates when the hydrocarbon and other reactants are depleted along the catalyst bed. These low effectiveness factors encountered along the reformer bed despite the small size of the catalyst used in the present work indicates that significant improvements in the reactor efciency could be achieved by using structured catalyst like monoliths or other structures with a suitable and sufciently thin catalytic coating. The model for describing the water-gas-shift reaction was also coupled to an intraparticle mass transfer resistance model and validated against published ex-

Acknowledgements The authors thanks PDVSA-Intevep and the US Department of Energy for allowing the publication of the present contribution. References
[1] C. Pereira, R. Wilkenhoener, S. Ahmed, M. Krumpelt, Catalytic reforming of gasoline and diesel fuel, in: Proceedings of the AIChE Spring 2000 Meeting, Atlanta, GA, 59 March 2000. [2] M. Krumpelt, S. Ahmed, R. Kumar, R. Doshi, Partial Oxidation Catalyst, University of Chicago, US Patent #6,110,861 (29 August 2000). [3] P. Savage, Mechanisms and kinetics models for hydrocarbon pyrolysis, J. Anal. Appl. Pyrolysis 54 (2000) 109126. [4] K. Pant, D. Kunzru, Pyrolysis of n-heptane: kinetics and modeling, J. Anal. Appl. Pyrolysis 36 (1996) 103120. [5] K. Pant, D. Kunzru, Pyrolysis of methylcyclohexane: kinetics and modeling, Chem. Eng. J. 67 (1997) 123129. [6] W. Jin, X. Gu, S. Li, P. Huang, N. Xu, J. Shi, Experimental and simulation study on a catalyst packed tubular dense membrane reactor for partial oxidation of methane to syngas, Chem. Eng. Sci. 55 (2000) 26172625. [7] D.A. Hickman, L.D. Schmidt, J. Catal. 138 (1992) 267. [8] D.A. Hickman, L.D. Schmidt, Steps in CH4 oxidation on Pt and Rh surfaces: high-temperature reactor simulations, AIChE J. 39 (1993) 11641177. [9] J. Xu, G. Froment, Methane steam reforming, methanation and water-gas-shift. Part I. Intrinsic kinetics, AIChE J. 35 (1989) 88. [10] J. Xu, G. Froment, Methane steam reforming. Part II. Difusional limitations and reactor simulation, AIChE J. 35 (1989) 97. [11] Rostrup-Nielsen, Catalytic steam reforming, in: John Anderson, Michel Boudart (Ed.), Catalysis Science and Technology, Springer, 1984 (Chapter 1). [12] J. Kopasz, L. Miller, S. Ahmet, P. Devlin, M. Pacheco, Reforming Petroleum Based Fuels for Fuel Cell Vehicles: Composition-Performance Relationships, 2002 Future Car Congress, Virginia, 35 June 2002. [13] A. Avci, Z. Ilsen nsan, D. Trimm, On-board fuel conversion for hydrogen fuel cells: comparison of different fuels by computer simulations, Appl. Catal. A 216 (2001) 243. [14] S. Springmann, G. Friedrich, M. Himmen, M. Sommer, G. Eigenberger, Isothermal kinetic measurements for hydrogen production from hydrocarbon fuels using a novel kinetic reactor concept, Appl. Catal. A 235 (2002) 101.

M. Pacheco et al. / Applied Catalysis A: General 250 (2003) 161175 [15] Fuel Cell Handbook, Fifth ed., National Energy Technology Laboratory, 2000. [16] R. Keiski, T. Salmi, V. Pohjola, Development and verication of a simulation model for a non-isothermal water-gas-shift reactor, Chem. Eng. J. 48 (1992) 17. [17] R. Keiski, O. Desponds, Y. Chang, G. Somorjai, Kinetics of the water-gas shift reaction over several alkane activation and water-gas shift catalyst, Appl. Catal. A 101 (1993) 317.

175

[18] C.N. Sattereld, in: Robert E. (Ed.), Mass Transfer in Heterogeneous Catalysis, Krieger Publishing Company, USA, 1981. [19] G. Froment, K. Bischoff, Chemical Reactor Analysis and Design, Wiley Series in Chemical Engineering, 1990. [20] D. Myers, T. Krause, J. Bae, C. Pereira, Reducing the Volume/Weight of the Fuel Post Processor for PEFC Power Systems, in: Proceedings of the 2000 Fuel Cell Seminar, 30 October2 November 2000, Portland, USA.

Potrebbero piacerti anche