Reforming reactors
Vasco H. Manaçasa, Henrique A. Matosa, Štěpán Špatenkab
a
Chemical Engineering Department, Instituto Superior Técnico, Lisbon, Portugal
b
Process Systems Enterprise Ltd., London, United Kingdom
December, 2013
Abstract
Natural gas reforming (NGR) is an important sector in the chemical and petrochemical industry. It
is a source of hydrogen for the refining industry and the main source of synthesis gas for the
manufacture of chemicals such as ammonia and methanol. Production of synthetic fuels via
Fischer-Tropsch has great potential to become another major application of NGR. Demand for
NGR products will be increasing in the near future so it is important to develop accurate and
predictive modelling tools for decision support in reactor and process design and operation.
This dissertation presents initial developments on NGR modelling in gPROMS®. The two main
reformers, the steam methane reformer (SMR) and the autothermal reformer (ATR) were
modelled. The gPROMS® Advanced Model Library for Fixed-Bed Reactors (AML:FBCR) was used
for detailed modelling of the catalyst beds considering rigorous intra-particle mass and heat
transport. Simplified models were developed for the furnace of the SMR and for the combustion
chamber of the ATR.
Two industrial cases from the literature were studied. The one for SMR was simulated with a
fixed wall temperature profile and afterwards with heat fluxes calculated from a radiation fired
heater model. Both approaches revealed good agreement with plant data. The one for ATR was
simulated with the simplified combustion chamber and predictions had only minor discrepancies
from plant outlet data.
A set of operational conditions, of the SMR case study, were optimised for maximum hydrogen
throughput, at the expense of lower energy efficiency (more 0.9 kJ/mol H2 spent).
Keywords: Natural gas reforming, steam methane reforming, autothermal reforming, gPROMS,
modelling, optimisation.
1
mechanisms controlled by the intra-particle diffusion 2.2 Chemical Routes
rates and with radial gradients of temperature and In a natural gas reforming process stable molecules,
composition on the catalyst tube side; complex need to be broken into less stable, highly reactive
combustion or radiation phenomena takes place on molecules. To accomplish this there are two main
the heating side of the reactor. routes: steam reforming and partial oxidation.
Detailed and accurate modelling of natural gas Methane steam reforming produces a syngas with a
reformers would give better insight into process H2/CO ratio of 3.0. Higher hydrocarbons react in a
design and operation [15] and would become a very similar way to steam reforming reaction (SR).
useful tool for managers and engineers to make (1)
informed process innovation, design and operation
decisions. (2)
The objective of this work is to start developing Along with the SR, water gas shift reaction (WGS)
modelling solutions for natural gas reforming with also occurs.
the purpose of, in the future, creating a (3)
comprehensive accurate and predictive modelling These reactions are all reversible with exception for
pack for synthesis gas manufacture in gPROMS®. steam reforming of higher hydrocarbons and all of
them require a catalyst, typically supported nickel.
2 Natural Gas Reforming Some natural gas reforming processes take
NGR processes and technology were reviewed. A advantage of partial oxidation of hydrocarbons (POX)
process example is showed in Figure 1 and only the which is an irreversible process. The reactions may,
main reformer will be described hereinafter. For a or may not require a catalyst.
more in depth review please consult [15–17].
Figure 1 Illustration of hydrogen production process flow diagram with SMR [17].
2
developed catalysts minimize this issue. Reaction (5) operating pressures of 10-40 bar and outlet
is known as the Boudouard reaction and it is temperatures of 800-950 °C. Inlet temperatures to
responsible for metal dusting corrosion [17]. the reformer tubes are limited by thermal cracking of
Carbon formation can also occur via thermal cracking higher hydrocarbons so the typical range is 450-550
of higher hydrocarbons at temperatures above 550 °C [16]. The upper limit rises up to 650 °C when a pre-
°C. The introduction of pre-reforming minimizes this reformer is installed.
issue by removing higher hydrocarbons from the Synthesis gases produced by SMR from natural gas
main reformer [16]. have H2/CO ratios between 2.8 and 4.8 [16].
Other sources of undesirable reactions are sulphur
2.3.2 Autothermal reformer – ATR
poisoning and sintering of the catalyst [15,16].
An autothermal reformer (ATR) is a refractory lined
2.3 Reactors vessel composed by a burner, a combustion chamber
The main, and most wide-spread, reactor designs for and a catalyst fixed-bed. As the name suggests, the
NGR are steam methane reforming (SMR) and reformer is thermally self-sufficient and operates
autothermal reforming (ATR). While these two will be adiabatically with exothermic POX reactions
described hereinafter, other important reforming providing the necessary heat for the endothermic SR
technologies include catalytic and non-catalytic reactions.
partial oxidation and, more recently developed, heat In the combustion chamber flame temperatures over
exchange reformers (HER) and oxygen membrane 3000 °C can be achieved. As homogeneous SR
reformers, and are addressed in [15]. reactions occur, gas temperatures at catalyst bed
Two-step or combined reforming reactors are also inlet drop down to 1100-1300 °C [15]. Outlet
important reactor configurations. Two of the referred reformer temperatures are usually around 900-1100
reactors, usually ATR, SMR or HER, are arranged in °C [15]. Typical operating pressures are higher than
series or parallel. for SMR and can reach 30-50 bar. Oxygen is fed at
sub-stoichiometric conditions, typically at oxygen to
2.3.1 Steam methane reformer – SMR
methane ratios (O2/C ratios) of 0.5-0.6 [2]. Carbon
As the name suggests, the steam methane reformer
free operation requires a small amount of steam
(SMR) follows the SR reactions. This reformer is often
corresponding to S/C ratio lower than 1.0.
called tubular fired reformer because of the reactor
Hydrocarbon feed can be natural gas or partially
design.
reformed gas (from a primary reformer). The latter is
In SMR the reaction proceeds inside a number of
usually called secondary reformer. The concept is the
reformer tubes while heat is being transferred from a
same but operating conditions are different and
furnace radiant chamber. This radiant chamber has
require different a different reactor design [17]
burners installed on the walls where the fuel is
The oxidant used can be pure oxygen or (enriched)
burned releasing heat. Only 50% of this heat is
air and both have advantages and draw backs. Air-
transferred to the reformer tubes, the remainder is
blown ATR doesn’t need an air separation unit but
recovered on the convection section of the furnace
requires larger equipment and compression costs
for pre-heating duties and steam generation. Flue gas
and so it is only used when nitrogen is needed in the
temperatures are brought down close to 200 °C and
syngas such as in ammonia production
overall thermal efficiencies of 95% can be attained
ATR reforming of natural gas can provide lower
[17].
H2/CO outlet ratios (1.8-3.8) than SMR due to the
Reactor operation is limited by tube material
introduction of oxygen instead of steam [16].
maximum allowed temperature, as increasing this
temperature will decline expected tube lifetime. 3 Modelling
Recent metallurgical developments allow maximum
3.1 gPROMS Libraries
tube wall temperatures of 1050 °C [17].
The gPROMS Advanced Model Library for Fixed-Bed
Since most reformers operate near or at equilibrium
Catalytic Reactors (AML:FBCR) was used to model
outlet conditions, the key variables that define outlet
reactor sections with fixed catalytic beds.
gas composition are: hydrocarbon feed composition,
The two main reactor models used from this library
inlet S/C ratio, outlet temperature and pressure.
were one and two dimensional catalyst beds with
Typical conditions are inlet S/C ratios of 2-5,
3
rigorous mass and heat transport inside the catalyst
∫ ( ) ⁄ (12)
pores: Catalyst_Pellets_Section and
Catalyst_Pellets_Section_1D_Adiabatic.
The library uses a set of user-specified sub-models, ∏
(13)
where kinetic equations or other correlations can be
defined. These customisations are discussed in the 3.4 Bed properties
following sections. Bed porosity was predicted according to Reichelt and
A full description of these models’ equations won’t Blass.
be presented because it wasn’t a model developed Radial heat transfer parameters correlations were
during the course of this work and because of customised to the ones used in [4] and also can be
intellectual property rights. found in [18]. These parameters are used in
It is important to understand how the radial heat equations (9) and (8), and have static and dynamic
transfer is modelled in the two-dimensional model. contributions. The static contribution relates to heat
Both convective and diffusive heat transfer are transfer in the hypothetic situation of zero flow and
considered. The diffusive heat transfer uses an the dynamic relates accounts for hydrodynamics.
effective radial conductivity, λe,r, and wall heat
(14)
transfer coefficient, αw.
(8) (15)
The dynamic correlations for in AML:FBCR include an
adjustable parameter that allows the user to adjust
[ ] (9)
the correlations to their own experimental data.
The gPROMS Process Model Library for Steady State
(PML:SS) conversion reactor and signal models were 3.5 Fired heater
used to model the combustion chamber of the ATR. Detailed modelling the heat transfer from the firebox
to the reformer tubes is a challenge. Modelling
3.2 Physical properties
radiative heat transfer requires intensive
General physical properties were predicted using
calculations, taking into account geometry view
GERG 2008 corresponding states model which is
factors, gas emissivities and burner geometry. In this
specific to natural gas applications.
work a simplified approach to the furnace was used
3.3 Reaction kinetics and a typical radiation heat flux equation was
The reaction kinetics model used was the one considered. Equation (16) is used to calculate heat
developed by Xu and Froment [3]. They established a flux or tube wall temperature and applies two
Langmuir Hinshelwood model that has been effective parameters: an effective emissivity and an
extensively used throughout the literature. effective temperature.
The kinetic model considers three global reactions This approach assumes radiation is the only form of
(10): steam reforming of methane to carbon heat transfer, neglecting convection contributions.
monoxide, SR(I), and to carbon dioxide, SR(II) and The effective emissivity lumps geometry effects
water gas shift, WGS. It was assumed the natural gas (number and location of burners, flame size, firebox
does not contain higher hydrocarbons so no reaction size and number of burners); and emissivities from
kinetics are needed for them. the different sources present (furnace, reformer
tube, fluid and flame). While the effective
(10)
temperature reflects the heating intensity (fuel
The reaction rate expressions and coefficients can be flowrate, flame temperature).
found in the original paper. ( ) (16)
Equilibrium was calculated from: 3.6 Combustion chamber
∑ For natural gas, it is common to approximate partial
(11)
combustion reactions in ATR as a one molecular
Besides kinetic rate equations some key performance reaction. Since a secondary reformer may receive
indicators were also calculated: the effectiveness
factor and the approach to equilibrium:
4
partially reformed gas with some hydrogen content, one catalytic reactor bed: one connected to a fired
hydrogen combustion was also considered [9]. heater and the other working in adiabatic mode.
(17) The methods used in this adiabatic one dimensional
catalyst section model from AML:FBCR were
(18) configured to the same corresponding options as in
The combustion chamber was modelled using the the SMR two-dimensional model.
conversion reactor from PML:SS The model was set Some key performance indicators are automatically
adiabatically with reaction stoichiometry. calculated by the model, like inlet O/C ratio, outlet
Conversion needs to be set for both reactions and methane conversion and outlet H2/CO ratio.
the signal library from PML:SS was used to that
purpose. Complete oxygen conversion is considered, 4 Simulation Results
and selectivity is set between methane partial 4.1 Steam Methane Reforming
oxidation and hydrogen combustion reactions. The example used for this case study comes from the
3.7 Steam methane reformer work by Xu, Froment and Plehiers [4,6] where an
A possible configuration for the SMR is showed in industrial steam reformer was modelled. Their
Figure 2. There are two material stream sources, one reactor tube model consists of a one dimensional bed
for natural gas and another for steam; a junction model with one dimensional pellet model accounting
(mixer); two catalytic reactor beds, one connected to for intra-particle diffusion. The firebox is modelled
a fired heater and the other working in adiabatic using a detailed zone method model.
mode. 4.1.1 Reactor configuration
The methods used in catalyst section model from Reactor and catalyst dimensions were taken from the
AML:FBCR need to be configured. The radial heat original papers, as well as the major properties.
transfer was considered through fluid. Real mixture Missing parameters, like conductivities or emissivities
partial enthalpy calculations were. Intra-particle were taken from other literature examples.
diffusion is calculated using the Dusty-gas model as
4.1.2 Thermodynamic results
Knudsen diffusivity is used in [4]. Finally, solid-fluid
Methane equilibrium conversion is shown in Figure 4
heat and mass transport uses the correlations from
at different temperatures, pressures and S/C ratios.
Hougen (1961), the default in AML:FBCR.
These results clearly depict equilibrium constraints
Some key performance indicators are automatically
on steam reforming reactors. These are important
calculated by the model, like outlet methane
restrictions which explain why steam reforming
conversion and outlet H2/CO ratio.
reactors need to operate at such high temperatures.
3.8 Autothermal reformer
The ATR model uses the combustion chamber model
in PML:SS and the SMR catalyst bed model in
AML:FBCR. A “gML to PML” connector model is
required to change the reference state from pure
elements (PML:SS) to compounds (AML:FBCR).
A possible configuration for the SMR is showed in
Figure 3. There are two material stream sources, one
for the feed (natural gas or reformed process gas)
and another for oxygen or enriched air; one mixer;
Figure 2 SMR model topology using a fired heater. Figure 3 ATR model topology.
5
100 According to these results the model suggests plant
Methane conversion (%)
90 data conversion is better than equilibrium
80 conversion, which is impossible. The ideal gas
70 assumption for reaction rates and reaction
60
equilibrium may be invalid, and vapour fugacity
50
40 rather than vapour concentration should be used. Yet
30 it is also possible the reported plant measurements
20 are inaccurate as a result of temperature or pressure
10
drops in the reactor outlet system or from non-
0
800 1000 1200 1400 1600 uniform operation between different catalyst tubes
Temperature (K) inside the reformer.
1200
20 bar, S/C 4 30 bar, S/C 4 20 bar, S/C 2
Temperature (K)
Figure 4 Methane equilibrium conversion as a function of
1000
temperature at different pressures (20 and 30 bar) and at
different S/C ratios (1, 2 and 4).
900
4.1.3 Case I: Fixed temperature profile
Here the reactor from this case study was modelled 800
heat transfer correlations were reduced to 48.9% and Fluid radial average Fluid at wall
6
active layer is used. These results agree with other 4.2 Autothermal Reforming
literature findings [15]. The example used for this case study comes from the
0,3
work by Al-Dhfeery et. al. [9] where an industrial
secondary reformer was modelled.
In [9], their reactor model consists of a global mass
Reaction rate (mol/kg-cat s)
7
4.2.3 Case I be adjusted by changing pellet tortuosity, porosity or
Results shown in Table 3 reveal good prediction in mean pore radius from AML:FBCR default values to
terms of outlet composition and only slight increase internal diffusion limitations. Without plant
deviations in temperature, pressure and conversion. measurements inside the reactor it’s not possible to
Outlet pressure difference relates to combustion confirm how the reactor bed behaves.
zone and burner pressure drops, as well as inert Another possible explanation is that ATR catalysts
catalyst sections that weren’t considered. Outlet may have lower activity than SMR catalysts, so the
temperature deviations between this model and Al- kinetic parameters would need to be adjusted.
Dhfeery and co-workers’ model can relate to Moreover, it is unlikely the reactor operates with two
different thermodynamic models, but the deviation thirds of inactive bed. Other possible explanations
to plant data might relate to heat losses. ATR could reflect that the catalyst bed is essential to
reformer vessels are considered to be adiabatic, but stabilize the reaction mixture after the combustion
they operate at such high temperatures that a water zone – the catalyst bed can be slightly over-
cooling jacket may be used to stabilize external wall dimensioned.
temperature. In fact, Aasberg [15] refers that outer
metal skin temperatures are reduced to 100-200 °C
5 SMR Optimisation
The SMR model used in section 4.1.4 with a fired
so it is possible that minor heat losses are present.
Table 3 ATR Case I combustion and catalyst zone outlet heater model operational conditions were optimised.
results. Those results were starting point for optimisation.
Plant Present Al-Dhfeery, unit 5.1 Problem formulation
data work 12
s The reactor system was optimised for maximum
Combustion zone
hydrogen production at a fixed natural gas feed
Outlet T - 1361.8 1369.8 K flowrate. The objective function is:
Outlet P - 32 32 bar ( ) (19)
Catalyst The control variables included in are:
zone
Outlet T 1242.2 1264.8 1282.8 K { ⁄ } (20)
Outlet P 31.6 31.9 31.9 bar These correspond to four key design variables of the
Conversion 95.6% 97.4% 95.5% reactor: inlet temperature to the reactor tube, ;
H2/CO 4.39 4.45 4.44
inlet pressure to the reactor tube, ; fired heater
effective temperature, ; and inlet ⁄ .
H2/N2 2.51 2.50 2.51
The problem was subject to the following constraints:
(H2+CO)/N2 3.08 3.06 3.08
(21)
Molar composition at combustion and catalyst zone
(22)
outlets shows almost no difference with plant data or
Maximum temperature constraint reflects a material
the paper’s predictions. Although these differences
constraint and maximum pressure drop was assumed
are very small, Predicted conversion deviates from
to be 120% of the base case scenario.
plant data due to a small absolute difference in
-4
The region where an optimal solution was searched
methane outlet composition, 8x10 , which
for is defined on each design variable by:
corresponds to a significant relative error of 69%.
(23)
Predicted axial profiles suggest that almost only one
(24)
third of the reactor is being used for reaction (see
Figure 7) and are quite different from the original (25)
paper, where the whole catalyst bed is active. ⁄ (26)
Al-Dhfeery and his co-workers used effectiveness The upper bound on inlet temperature to the
factors of around 0.003, predicted from the Thiele reformer is a real material constraint to avoid
modulus for spherical pellets, while the present thermal cracking of hydrocarbons. The fired heater
model calculates effectiveness factors around 0.03, effective temperature is an indirect model parameter
from detailed diffusion inside a hollow cylindrical for heating intensity. In real operation the fuel
pellet. This difference can explain the different axial flowrate and air excess would be adjusted to get
profiles. The profiles predicted with this model could required process gas outlet temperature.
8
The optimisation problem had 109837 equations and Since the WGSR will convert more carbon monoxide
111261 variables. A total of 1420 variables were to hydrogen more energy can be recovered. This
assigned and the remainder 4 degrees of freedom should compensate the SMR energy efficiency
correspond to the four design variables considered. decrease towards hydrogen production. The
The optimal solution was found in 499 seconds using downside is that this heat will be recovered in the
an Intel® Core™ i7-3770S CPU with 16.0 GB of RAM. form of steam and this may be undesired.
9
6.1 Achievements Shift: I. Intrinsic Kinetics, AIChE J. 35 (1989)
The objective of this work was to initiate the 88–96.
development of models for NGR reactors with the [4] J. Xu, G.F. Froment, Methane Steam
purpose of creating accurate and predictive Reforming : ll. Diffusional Limitations and
Reactor Simulation, 35 (1989) 97–103.
modelling solutions for hydrogen and synthesis gas
[5] J.H. Ghouse, T.A. Adams Ii, A multi-scale
manufacture in gPROMS®. This was achieved with the dynamic two-dimensional heterogeneous
SMR and ATR reactor case studies. model for catalytic steam methane reforming
The simplified approaches to furnace and combustion reactors, Int. J. Hydrogen Energy. 38 (2013)
chamber modelling proved to reasonably represent 9984–9999.
real profiles. [6] P.M. Plehiers, G.F. Froment, Coupled
Simulation of Heat Transfer and Reaction in a
Finally, the operating conditions of the SMR industrial
Steam Reforming Furnace, 12 (1989) 20–26.
case were optimised using the detailed SMR reactor [7] S.S.E.H. Elnashaie, A.M. Adris, M.A. Soliman,
model with fired heater. This demonstrated that A.S. Al-Ubaid, Digital Simulation of Industrial
optimisation of complex accurate models is possible, Steam Reformers for Natural Gas Using
within the same modelling environment. Heterogeneous Models, Can. J. Chem. Eng. 70
(1992) 786–793.
6.2 Future work [8] Y.-H. Yu, Simulation of Secondary Reformer in
In order to further complete this work towards a Industrial Ammonia Plant, Chem. Eng.
comprehensive accurate and predictive modelling Technol. 25 (2002) 307–314.
pack for synthesis gas manufacture there are still [9] A. AL-Dhfeery, A. Jassem, Modeling and
several steps to be taken. simulation of an industrial secondary
Kinetics modelling can be improved by tackling the reformer reactor in the fertilizer plants, Int. J.
Ind. Chem. 3 (2012) 14.
ideal gas assumption and implementing fugacity
[10] M.H. Sosna, Y.H. Yu, Modeling for Industrial
calculations. Steam reforming of higher hydrocarbons Heat Exchanger Type Steam Reformer,
also needs to be considered. The undesired coking Korean J. Chem. Eng. 18 (2001) 127–132.
reactions can also be modelled. [11] P. Stehlik, Radiative Component in Thermal
The SMR furnace model needs to be improved, Calculation of Tubular Heat Exchangers, Heat
ideally accounting for detailed radiation and Transf. Eng. 16 (1995) 19–28.
[12] M.H. Wesenberg, Gas Heated Steam
convection heat transfer.
Reformer Modelling, Trodheim, Norway,
Modelling of NGR related equipment such as the pre- 2006.
reformer and the WGSR will allow modelling of the [13] J.K. Rajesh, S.K. Gupta, G.P. Rangaiah, A.K.
complete NGR process. Developing models for other Ray, Multi-objective optimization of industrial
type of reactors like the heat exchanging reactors will hydrogen plants, Chem. Eng. Sci. 56 (2001)
provide different options for process design 999–1010.
optimisation. [14] R.C. Baliban, J.A. Elia, C.A. Floudas, Novel
Natural Gas to Liquids Processes : Process
Finally, complete process modelling and
Synthesis and Global Optimization Strategies,
optimisation, considering reactor or process design, AIChE J. 59 (2013) 505–531.
with detailed high-fidelity models would be a proof of [15] K. Aasberg-Petersen, I. Dybkjær, C.V. Ovesen,
the power and capabilities of advanced process N.C. Schjødt, J. Sehested, S.G. Thomsen,
modelling to provide decision support and help Natural gas to synthesis gas – Catalysts and
achieving the best process innovation, design and catalytic processes, J. Nat. Gas Sci. Eng. 3
operation decisions. (2011) 423–459.
[16] J.R. Rostrup-Nielsen, L.J. Christiansen,
Bibliography Concepts in Syngas Manufacture, Imperial
[1] International Energy Agency, Gas Medium- College Press, 2011.
Term Market Report, 2012. [17] I. Dybkjær, Tubular reforming and
[2] J.R. Rostrup-Nielsen, Syngas in perspective, autothermal reforming of natural gas - an
Catal. Today. 71 (2002) 243–247. overview of available processes, Fuel Process.
[3] J. Xu, G.F. Froment, Methane Steam Technol. 42 (1995) 85–107.
Reforming, Methanation and Water-Gas [18] G.F. Froment, K.B. Bischoff, J. De Wilde,
Chemical Reactor Analysis and Design, Wiley,
2010.
10