i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2

Available online at www.sciencedirect.com

ScienceDirect

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

Real time optimization of steam reforming of

methane in an industrial hydrogen plant

M. Taji, M. Farsi*, P. Keshavarz

Department of Chemical Engineering, School of Chemical and Petroleum Engineering, Shiraz University, Shiraz, Iran

article info abstract

Article history: The main goal of this research is modeling and real time optimization of an industrial
Received 7 March 2018 steam methane reformer considering catalyst deactivation. In the first step, the reformer is
Received in revised form heterogeneously modeled based on the mass and energy balance equations considering a
23 April 2018 detailed kinetic model. To prove the accuracy of developed model, the simulation results
Accepted 15 May 2018 are compared with the available plant data at steady state condition. In the second step,
Available online xxx based on the mechanism of catalyst deactivation, a first order decay model is proposed and
the parameters of the model are calculated to minimize the absolute difference between
Keywords: calculated methane conversion and plant data. In the third step, an optimal control
Steam methane reforming problem is formulated to maintain hydrogen production capacity at the desired level.
Process modeling Based on the formulated optimization problem, optimal dynamic trajectories of feed
Catalyst deactivation temperature and steam to methane ratio are calculated considering two strategies. Then,
Optimal control the performance of developed optimization procedure is proved considering furnace
temperature and feed concentration as disturbance. The simulation results show that
operating at the proposed optimal condition increases hydrogen production about 11.6%. In
addition, the process emission performance defined as hydrogen to carbon dioxide ratio in
the product is 6.72 and 7.03 at the conventional and optimized conditions, respectively.
© 2018 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

reforming of hydrocarbons and coal gasification [4]. The

Introduction
Syngas as a mixture comprising of hydrogen and carbon
monoxide is an important feedstock in refinery and petro-
chemical complexes. It is used to produce methanol,
ammonia, ethylene glycol, and various aldehydes [1]. In
addition, syngas is severed as a feedstock to Fischer-Tropsch
Synthesis. Although a variety of technologies have been
developed based on pure hydrogen as an energy carrier, it is
used as feedstock in the hydrogenation and hydrocracking
processes in oil refinery units [2,3]. Industrially, hydrogen is
produced through steam reforming, partial oxidation and dry
composition of produced syngas varies based on the hydrogen
to carbon molecular ratio in the hydrocarbon feed stream and
process type. Typically, methane is widely used in the
reforming process due to higher hydrogen to carbon ratio
compared to other hydrocarbons [5]. In the partial oxidation, a
sub-stoichiometric mixture of methane and oxygen reacts
catalytically and syngas is produced [6,7]. Steam reforming is
an endothermic catalytic reaction that methane and steam
react at high temperature in a coil furnace. Auto thermal
reforming is an integration of steam reforming and partial
oxidation to achieve efficient thermal management [8].

* Corresponding author.
E-mail address: farsi@shirazu.ac.ir (M. Farsi).
https://doi.org/10.1016/j.ijhydene.2018.05.094
0360-3199/© 2018 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
2 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2
Although steam methane reforming is more economical
compared to the other methods due to high efficiency and low
feed cost, finding the optimal condition and catalyst deactiva-
tion by sintering, coking and poisoning are the major chal-
lenges in the steam reforming process. Catalyst deactivation
decreases production capacity gradually. In this regard, many
researchers have focused on the process modeling, optimiza-
tion and intensification of reforming process to improve the
production capacity and reformer performance. Alatiqi et al.
developed a mathematical model to investigate the perfor-
mance of steam methane reforming at steady state condition
[9]. Comparison between simulation results and plant data
proved the accuracy of the developed model and considered
kinetic model. Asleshirin et al. developed a homogeneous
model for steam methane reforming at dynamic condition [10].
They proposed a second order polynomial equation for catalyst
decay and calculated the parameters of the proposed model.
Jong et al. modeled an steam methane reformer and investi-
gated the effects of operation parameters on the hydrogen
production [11]. The results showed that increasing reactor
length, air concentration in the burner fuel and thickness of
insulating shield could improve the performance of reformer.
Pantoleontos at el. developed a one dimensional model for
steam methane reforming and calculated the profile of wall
temperature to maximize hydrogen yield [12]. The results
showed that applying the optimal wall temperature on the coil
tube could improve conversion and hydrogen yield.
For complicated problems, the multi-objective optimiza-
tion satisfies a number of different objectives simultaneously,
and the results are more accurate and real compared to the
single objective approach. In this regard, the multi-objective
optimization is used to find the optimal operating condition
of different process [13]. Sinaei et al. modeled and optimized a
steam methane reformer using design of experiment and
response surface methodology [14]. The hydrogen production
and methane mole fractions were considered as two re-
sponses. Temperature, pressure, steam to methane ratio, and
hydrogen to methane ratio in the feed stream, and tube wall
temperature were selected as the independent factors, and
the optimum values of independent factors were calculated.
Rajesh et al. developed a multi objective optimization problem
to achieve maximum carbon monoxide production and min-
imum methane conversion [15]. They selected feed tempera-
ture, furnace temperature, and steam flow rate as decision
variables. The optimal Pareto front was developed by non-
sorting genetic algorithm and a single optimal solution was
selected from list of alternatives by TOPSIS method. Farsi and
Hosseini developed a heterogeneous model for an industrial
reformer and investigated the effect of feeding type on the
process performance. They developed a multi objective opti-
mization problem to maximize methane conversion and
minimize pressure drop [16]. The results showed that
hydrogen production is improved about 18.1% in the proposed
configuration compared to the conventional process. Rajesh
et al. simulated an industrial hydrogen plant at steady state
condition [17]. The non-dominated sorting genetic algorithm
was employed to optimize the operating condition consid-
ering hydrogen production and inlet steam flow rate as the
objective functions. Then, Pareto optimal front was obtained
at different process conditions.
Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
Currently, membrane separation, thermally coupling and
sorption enhanced technologies have been proposed to couple
with the reforming process to shift the reforming reactions
toward the higher conversion [18]. In this regard, Cruz and
Silva developed a mathematical model to compare the per-
formance of conventional and membrane reactors in the
reforming process [19]. The results showed that methane
conversion and hydrogen yield are significantly enhanced by
hydrogen removal from reaction zone through the Pd-based
membrane. Abbas et al. simulated the sorption enhanced
steam reforming of methane using CaO as CO2 sorbent [20].
The result showed that methane conversion in the proposed
and conventional processes approaches to 65% and 24% at the
same operating condition, respectively. Aboosadi et al.
coupled the steam methane reforming and nitrobenzene hy-
drogenation reactions in an reactor [21]. The simulation re-
sults showed that produced heat through nitrobenzene
hydrogenation in the exothermic side could play role of fire
box in the conventional process.
Generally, optimization of the operating conditions
without considering catalyst deactivation suffers from reli-
ability and results are valid at the start of run when the
catalyst is completely active. In this regard, the main goal of
this research is modeling and real time optimization of an
industrial reformer considering catalyst decay. Based on the
mechanism of catalyst deactivation presented in the litera-
ture, a decay model is proposed to calculate the catalyst ac-
tivity during the process run time. Then, an optimal control
problem is formulated to fix production capacity at the desired
level. Based on the formulated optimization problem the
optimal trajectories of feed temperature and steam to
methane ratio in the feed stream are obtained during the
process run time.
Process description
Steam reforming of methane is widely used to produce
hydrogen in oil refinery and petrochemical complexes. After
desulfurization of natural gas in a catalytic reactor, methane
is preheated, mixed with superheated steam and feeds to the
reformer. The reforming reactions are typically carried out in
a furnace over the nickel catalyst. The furnace consists of a
radiant section including burners and reforming tubes, and a
convection section to recover the waste heat of the flue gases.
In the radiant section, the catalyst is loaded in the reforming
tubes placed in a row along the furnace. The produced heat in
the fire box is transferred to the reforming tubes and supplies
the heat of reforming reactions. It should be noted that the use
of steam, slightly more than the stoichiometric ratio, im-
proves the rate of reforming reactions and prevents coke build
up on the surface of catalyst [15]. The outlet product from
reformer feeds into the shift reactors to convert steam and
carbon monoxide to hydrogen and carbon dioxide. After
steam recovery, the outlet stream enters to the hydrogen
separation unit. Amine absorption and PSA treating are two
conventional processes to separate hydrogen from product. In
the amine absorption process, the aqueous solution of amine
and outlet stream from shift convertor feed the top and bot-
tom sections of absorption column, respectively and carbon
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2 3

dioxide is continuously absorbed. In the PSA units, the im-

Table 1 e The kinetic of steam methane reforming
purities are adsorbed on an absorbent and pure hydrogen is
reactions.
produced. Fig. 1 shows the schematic diagram of industrial
Reaction rate equation (kmol kg1 1
cat h )
hydrogen plant in Iran.
!,
k1 P3H2 PCO
R1 ¼ PCH4 PH2 O  DEN2 (4)
P2:5
H2
ke1
 
Process modeling R2 ¼
k2
PCO PH2 O 
PH2 PCO2
DEN2 (5)
PH2 ke2
!,
Reaction kinetic k3 P4H2 PCO2
R3 ¼ 3:5 PCH4 P2H2 O  DEN2 (6)
PH ke3
2

The reforming reactions are reversible and endothermic, and KH2 O PH2 O
DEN ¼ 1 þ þKCO PCO þKH2 PH2 þKCH4 PCH4 (7)
PH2
according to the Le Chatelier principle the desired conditions
to achieve maximum conversion are high temperature and
low pressure [16]. As well as steam reforming reactions, the
rate and adsorption constants of reactions [16]. It is
water gas shift reaction takes place on the catalyst surface.
mentioned that the kinetic constants are Arrhenius type.
The coke formation through methane and carbon monoxide
decomposition is negligible when steam concentration in the
Coil modeling
feed stream is slightly higher than the stoichiometric ratio
[5,15]. The steam reforming and water gas shift reactions are:
Developing a detailed and accurate model plays a key role in
Kj the process optimization and intensification. In this section, a
CH4 þ H2 O4CO þ 3H2 DH ¼ 206:6 (1)
mol one dimensional heterogeneous model is developed based on
the mass and energy balance equations to simulate an in-
Kj dustrial reformer. The considered assumptions are:
CH4 þ 2H2 O4CO2 þ 4H2 DH ¼ 164:9 (2)
mol
 Pseudo steady state condition due to low rate of catalyst
Kj deactivation
CO þ H2 O4CO2 þ H2 DH ¼ 41:1 (3)
mol  Since compressibility factor is in the range of 0.99e1.0, the
Table 1 shows the kinetics of steam reforming and water ideal gas assumption is acceptable.
gas shift reactions. Table 2 shows the equilibrium, reaction  Plug flow pattern is assumed due to high Reynolds number.

Fig. 1 e Process flow diagram of steam methane reforming.

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
4 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2

Table 2 e Equilibrium, reaction rate and adsorption 1 X 4

¼ Ri (15)
constants. U i¼1

Reaction Equilibrium constant (Ke) k0 Ej The considered heat transfer resistance includes convec-
1 24:58  10 3 4:225  10 15 240,100 tive heat resistance in the coil, conductive resistance in the

4:707  1012 e T tube material, and radiative and convective resistances in the
2 4:91  103 1:955  106 67,130 fire box. Equations (16) and (17) show the mass and energy

2
1:142  10 e T balance equations for solid phase:
3 17:52  103 1:020  15 243,900
    
5:375  1010 e T av kgi Ci  Csi ¼ rB at hj  Rj (16)

  
av hg ðT  Ts Þ ¼ rB at hj Rj DHj (17)
 Since Pe[1 the axial diffusion of mass and heat are
where hj and at are the effectiveness factor and catalyst ac-
ignored [22].
tivity, respectively. Pressure drop in the reactor is calculated
 Since Bi≪0:1 temperature gradient in the catalyst is ignored
from Ergun moment equation [26]:
[23].
 The CH4 decomposition and CO reduction reactions are dP 150 ms ð1  εb Þ2 1:75ð1  εb Þrg 2
negligible due to high steam to methane ratio [5,24]. ¼ us  us (18)
dz 2
dp ε3b dp ε3b

In the catalytic reactions, mass is transferred from the bulk
fluid to the external surface of the porous catalyst [25]. Then,
the reactants diffuse from the external surface into the cata-
lyst, and the reaction takes place over the pores. In this paper,
the mass and heat transfer resistances in the gas phase and
mass transfer resistance inside the catalyst are considered in
the developed model. The heat and mass transfer resistances
in the gas phase are considered in the model by transfer co-
efficients. The effect of intra particle gradient is applied in the
model by effectiveness factor. The effectiveness factor is
defined as the ratio of reaction rate considering mass transfer
resistance in the particle to the reaction rate applying the
surface concentration in the particle. The considered method
to calculate the effectiveness factor could be find in Appendix
A. Typically, the reaction rate of components is:
rH2O ¼ h1 R1  h2 R2  h3 R3
rCH4 ¼ h1 R1  h3 R3
rH2 ¼ 3h1 R1 þ h2 R2 þ 4h3 R3
rCO ¼ h1 R1  h2 R2
The considered auxiliary correlations to estimate the
physical and thermal characteristics of the gas mixture, such
as the viscosity, density, diffusion coefficient, specific heat
capacity, heat of reaction could be find in Appendix B. Table 3
shows furnace, feed and catalyst specifications of the
considered industrial reformer.
Process optimization
In the steam methane reforming process, production capacity
decreases gradually due to catalyst decay and it has an enor-
mous effect on the performance and controllability of plant. In
this regard, a real time optimization problem is formulated to
maintain production capacity at the desired level considering
feed temperature and steam to methane ratio as decision
(8) variables. Typically, there are many paths to maintain
hydrogen production capacity at the desired level. In the re-
(9) gard, two strategies are proposed based on the priority in the
manipulation of feed temperature and steam to methane
(10) ratio. According to pattern implemented in the industrial
process, the first optimization strategy is that manipulation of
(11) the feed temperature is prior to the methane concentration. In

rCO2 ¼ h2 R2 þ h3 R3 (12)
Table 3 e Typical operating condition at the start of run
In the above equations, R1, R2 and R3 are the rate of CO (SOR) conditions.
hydrogenation, water gas shift and CO2 hydrogenation re-
Feed Condition
actions. To develop a mathematical model, an element along Pressure (bar) 24,5
the axial direction of the reactor is selected, and mass and Inlet temperature (K) 818
heat balance equations are written on the gas and solid pha- Outlet temperature (K) 1133
ses. The mass and energy balance equations in the gas phase Methane flow rate (kmol h1) 425
are: Steam flow rate (kmol h1) 1867
Reactor Dimension
  Reactor tube inside diameter (m) 0.11
v
us ðCi Þ  kgi av Ci  Csi ¼ 0 (13) Reactor tube outside diameter (m) 0.14
vz
Number of pipe 40
Reactor tube length (m) 81.6
v  
us CP r ðTÞ  hg av ðT  Ts Þ þ UA T  Tf ¼ 0 (14) Catalyst Data
vz Catalyst shape Raschig ring
where U and Tf are the overall heat transfer coefficient and fire Catalyst porosity 0.52
box temperature, respectively. The overall heat transfer co- Catalyst diameter (mm) 17  6
Catalyst tortuosity 3.5
efficient is explained as:

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2 5

the first strategy, the feed temperature is manipulated and

vy
when the considered upper bound for the feed temperature is ¼ 4ðx:y:uÞ (27)
vt
activated, the steam to methane ratio is changed to maintain
where x represents composition, pressure and temperature
production capacity at the desired level. In the second strat-
states, and y is catalyst activity. Also, xðL:0Þ and xðL:tÞ are the
egy, manipulation of the feed concentration is prior to the feed
produced hydrogen at the start and end of runs, respectively.
temperature in the formulated optimization problem.
Based on the Lagrangian method, the formulated problem
Objective function, decision variable and constraint could be written as:

Ztf (
In the oil refinery complexes, the object of steam methane  2
L¼ xðL:0Þ  xðL:tÞ þ
reforming is hydrogen production to use in the hydrocracking
0
and hydrotreating units. Since steam methane reforming re-
X      )
5
 vFi vT va
actions are reversible and endothermic, increase in the feed  li Þ  fi þ l f m 4 dt (28)
i¼1
vl vl vt
temperature enhances rate of reactions. On the other hand,
the water gas shift reaction is exothermic and reversible, and The optimal dynamic trajectories of decision variables are
increase in the operating temperature decreases the equilib- calculated iteratively by:
rium conversion and results in the lower hydrogen produc-
a
tion. To avoid coke build up on the catalyst surface, ui aþ1 a
ðtÞ ¼ ui ðtÞ þ εli ðtÞ (29)
operational limitations and plant cost, the steam to methane
ratio is set in the range of 3e6 [27e29]. In addition, due to dL
li aðtÞ ¼ (30)
safety and material limitations, the upper bounds of feed and dui
skin temperatures are set at 900 and 1300 K, respectively [15].
where li is calculated by numerical derivative method, and it
The formulated optimization problem could be explained as:
approaches to 0 the optimal condition.
Objective Function ¼ FH2 (19)
Activity model
H2 O
3 6 (20) In this work, a catalyst decay model is proposed to predict
CH4
catalyst activity during the process run time, and the param-
TFeed  900 K (21) eters of the developed model are calculated based on plant
data by genetic algorithm method. The considered activity
TSkin  1300 K (22) model is as follows:

da  Ed

Optimal control ¼  kd eRT a (31)
dt

The optimal control as an extension of the calculus of varia-
tions is a dynamic optimization procedure developed to find
the optimal dynamic trajectory of decision variables [30].
Without no path constraints on the states and the control
variables, the continuous time optimal control problem can be
formulated as:
where kd and Ed are decision variables. To calculate the
optimal parameters of the decay model, the absolute relative
difference between calculated methane conversion and plant
data is considered as objective function. In the genetic algo-
rithm the number of population and generation are set to be
100 and 200, respectively to guarantee the accuracy of results.

Ztf
J ¼ Pðxf Þ þ Fðx:u:tÞ dt (23) Mathematical solution
to

From modeling perspective, process furnace is a complicated

x_ ¼ f ðx:y:tÞ (24) heat and mass integrated system. In this system, the mass
and energy balance equations, activity model and auxiliary
where t0 and tf are the initial and final times in the considered
equations form a set of nonlinear differential equations. In
problem, x, P, F and f are state vector, terminal cost function,
this paper, the numerical methods are considered to solve the
an intermediate cost function, and a vector field, respectively.
developed equations. Initially, the activity model is dis-
In this research, the formulated optimal control problem
cretized at the time domain using finite difference method
could be written as:
and the balance Equations (13)e(18) are solved at each time
Ztf step using fourth order Runge Kutta method. To calculate the
 2
JðxÞ ¼ xðL:0Þ  xðL:tÞ dt (25) rate of heat transferred from fire box to the coil, the skin
to temperature is required. The produced heat in the fire box is
transferred to the outer wall of coil tubes through radiation
Subject to:
and convection mechanisms. Then, heat is conducted along
vx the tube thickness and transferred to the gas inside the tubes
¼ f ðx:y:uÞ (26) by convection mechanism. The following procedure is
vz

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
6 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2

considered to calculate the temperature and concentration

(a)
profiles along the reforming tubes at each time step.

 Estimate a skin temperature for element “k” of tube, Tks

 The rate of heat transferred from fire box to tube wall, Q ko ,
is calculated based on the estimated skin temperature.
 The mass and energy balance Equations (13)e(18) are
solved for element “k” and the rate of heat transferred from
k
to the gas phase, Q i , is calculated.
wall

 If Q ko  Q ki > ε correct Tks based on the ðQ ko  Q ki Þ


 If Q ko  Q ki < ε save Tks , and repeat the procedure for
element k þ 1.

Result and discussion

Model validation

Since the developed mathematical model of the process is (b)

used to optimize the reformer's operating condition, the
model validation is necessary. In this regard, to verify the
accuracy of the developed model and considered assump-
tions, the simulation results are compared with the plant data
at the presented feed condition and design data in Table 3.
Table 4 shows the comparison between calculated outlet
pressure, temperature and methane conversion and plant
data. It is appeared that there is a good agreement between
model result and plant data.

Activity model

In this section, based on the mechanism of catalyst deacti-

vation, a decay model is proposed and the parameters of
model are calculated to minimize the absolute difference be-
tween predicted methane conversion and plant data. Typi-
cally, operating at high temperature sets severe conditions on
the nickel catalyst, and causes sintering [31]. The sintering
results in the loss of catalyst activity and changes the con-
version and selectivity. On the other hand, operating at low Fig. 2 e (a), Methane concentration in the feed stream
steam to methane ratio results in the coke build up on the during the process run time. (b), Feed temperature profile
catalyst surface, and decreases the catalyst active sites. In the during the process run time.
industrial plant, the catalyst poisoning by sulfur impurities is
minimized by hydrotreating and desulfurization of the feed
stream. Fig. 2a and b shows the feed temperature and
methane concentration in the inlet of considered reformer
da  
¼  460:24  eð T Þ a
11346
during the process run time. The process run time in the (32)
dt
considered reforming plant is 57 months.
The developed catalyst activity model based on the avail- Process optimization
able data is as following:
In this section, the results of process optimization are pre-
sented based on the two strategies. The first optimization
strategy is that manipulation of the feed temperature is prior
Table 4 e Comparison between plant data and simulation
results. to the methane concentration. In this regard, the feed tem-
perature is manipulated and when the considered upper
Parameters Simulated Plant Relative
result data error bound for the feed temperature is activated, the steam to
methane ratio is changed to maintain production capacity at
Pressure (bar) 20.99 21 0.04%
desired level. In the second strategy, manipulation of the feed
Outlet temperature (K) 1131 1133 0.17%
CH4 conversion 0.773 0.78 0.9%
concentration is prior to the feed temperature in the

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2 7

results of strategies from product distribution point of view.

Table 5 e Optimization results based on the first and
While in the first strategy lower methane consumption and
second strategies.
carbon monoxide production are the main benefits, the sec-
Parameter Conventional Strategy 1 Strategy 2
ond one presents a lower carbon dioxide production. The
1
Hydrogen (kmol h ) 25.94 28.96 28.96 carbon oxide production as total carbon dioxide and carbon
Carbon monoxide 3.50 4.15 3.98 monoxide in the outlet product is 8.15 and 8.10 kmol h1 based
(kmol h1)
on the first and second strategies, respectively. It is concluded
Carbon dioxide 3.86 4.00 4.12
(kmol h1)
that difference between total carbon emission in the both
Steam (kmol h1) 34.63 33.70 33.62 strategies is negligible. Fig. 3aec shows the calculated optimal
Consumed 7.36 8.15 8.10 trajectories of feed temperature, steam to methane ratio and
Methane (kmol h1) hydrogen production capacity for both strategies. It can be
Sum of carbon 7.36 8.15 8.10 concluded that, hydrogen production decreases from 28.98 to
emission (kmol h1)
23.21 kmol h1 in the conventional process during the process
run time. The hydrogen production capacity is improved
formulated optimization problem. Table 5 shows the optimi- about 3.02 kmol h1 when the optimal trajectories are applied
zation results based on the first and second strategies. on the reformer. Based on the simulation results, hydrogen
Although both strategies are capable to minimize the production increases about 11.6% in the optimized process.
objective function and maintain production capacity at the From theoretical point of view, operating at high temperature
desired level, there is a considerable difference between is a practical solution to overcome the effect of catalyst decay

(a) (b)

(c)

Fig. 3 e (a), Hydrogen production capacity during the process run time. (b), Calculated optimal trajectory of feed
Temperature. (c), Calculated optimal trajectory of steam to methane ratio.

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
8 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2
on the reaction rate, and it increases the kinetic constant. In
addition, operating at low steam to methane ratio in the feed
stream has a considerable effect on the rate of reactions. It
was found that increase the feed temperature from 818 to
900 K and decrease the molar steam to methane ratio from
4.38 to 3.45 are practical methods to maintain production ca-
pacity at the desired level.
Figs. 4 and 5 show the carbon monoxide and carbon dioxide
production in the conventional and optimized reformers,
respectively. It was found that, carbon monoxide production
decreases from 3.99 to 2.98 kmol h1 in the conventional
process during the process run time. Based on the simulation
Fig. 4 e Carbon monoxide production during the process
run time.
Fig. 5 e Carbon dioxide production during the process run
time.
Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
results, the average carbon dioxide production is
3.86 kmol h1 in the conventional process and it increases
about 0.14 and 0.26 kmol h1 when the first and second stra-
tegies are applied on the reformer, respectively. In addition,
carbon monoxide production increases 0.65 and 0.47 kmol h1
when the first and second strategies are applied on the sys-
tem, respectively. In the considered plant, produced carbon
dioxide is separated from gaseous product and vents to the
atmosphere. In this regard a quantitative criterion is defined
as hydrogen per carbon dioxide ratio in the outlet product
from reformer to compare the emission performance of opti-
mized and conventional processes. The calculated emission
performance in the reformer based on the conventional, first
and second strategies are 6.72, 7.24 and 7.03, respectively. The
results show that operating at the proposed condition in-
creases the emission performance.
Fig. 6 shows the coil outlet temperature during process run
time. Generally, the coil is placed in the radiation section of
the furnace and heat is transferred from combustion chamber
to the coil. The transferred energy is consumed through
endothermic reforming reactions and syngas is produced.
Typically, decreasing catalyst activity reduces methane con-
version and the rate of heat consumption in the reformer.
Thus, the rate of heat transferred from fire box to the gas is
dominated over the rate of heat consumption and tempera-
ture increases in the coil. The results show that temperature
of outlet product from coil increases linearly in the conven-
tional process. Typically, increasing methane concentration
in the feed stream increases rate of reforming reactions and
causes higher heat consumption in the coil. Thus, in the part
where the steam to methane ratio is manipulated, lower
temperature rise is developed.
It is concluded from Fig. 6 that operating based on the first
strategy increases average operating temperature in the
reformer and results in the lower catalyst activity. While in
the second strategy, the feed temperature trajectory is applied
on the reformer after the concentration trajectory, and the
Fig. 6 e Calculated outlet coil temperature.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2 9

catalyst experiences higher activity. Based on the simulation
results, the average catalyst activity at the 30th and 57th
months of continous operation is 0.253, 0.108 when the first
strategy is applied, respectivelly. On the other hand, the
average catalyst activity at the 30th and 57th months of
operation is 0.270, 0.112 when the second strategy is applied,
respectivelly. The results show that the lower methane con-
version, lower carbon monoxide production and higher cata-
lyst activity are the main benefits of second strategy, while
lower carbon dioxide is produced when the first strategy is
applied on the reformer. Since the same amount of total car-
bon oxide and hydrogen are produced when the both strate-
gies are applied on the reformer, the second strategy is
proposed as the optimal procedure.
Effect of furnace temperature
In this section the ability of considered optimization proced-
ure to find the optimal trajectories of manipulated variables is
investigated against applied changes in the furnace temper-
ature. Fig. 7aec shows applied changes in the furnace
temperature and calculated trajectories of feed temperature
and steam to methane ratio to maintain hydrogen production
at the desired level. Operating at high furnace temperature
increases rate of reforming reactions and lower feed temper-
ature and higher steam to methane ratio are required to
maintain production capacity at the desired level. Although
furnace temperature could be considered as the decision
variable, as well as hot spot formation, increasing furnace
temperature decreases thermal efficiency of furnace and de-
creases coil life time due to metal carbonization and hot spot
formation.
Effect of feed concentration
Fig. 8aec shows the applied changes in the feed concentration
and the calculated trajectories of feed temperature and steam
to methane ratio to maintain hydrogen production at the
desired level. The simulation results prove the ability of
formulated real time optimization problem to calculate the
optimal trajectories of the feed temperature and steam to
methane ratio during the process run time. Generally,

(a) (b)

(c)

Fig. 7 e (a), Variation of the furnace temperature during the process run time. (b), The optimal trajectory of feed temperature
considering furnace temperature as disturbance. (c), Calculated optimal trajectory of steam to methane ratio considering
furnace temperature as disturbance.

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
10 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2

(a) (b)

(c)

Fig. 8 e (a), Variation of carbon dioxide flow rate in the feed stream during the process run time. (b), Calculated optimal
trajectory of feed temperature considering feed concentration as disturbance. (c), Calculated optimal trajectory of Steam to
methane ratio considering feed concentration as disturbance.

increase carbon dioxide concentration in the feed stream
shifts reforming reaction toward the left side and results in
the lower hydrogen production.
Conclusion
In this paper an industrial steam methane reformer is het-
erogeneously modeled based on the mass and energy balance
equations considering catalyst decay. Based on the mecha-
nism of catalyst deactivation, a first order decay model was
proposed and the parameters of the model were calculated to
minimize the absolute difference between simulation results
and plant data. The results showed that production capacity
decreases gradually due to catalyst deactivation in the in-
dustrial plant. To maintain production capacity at the desired
level, an optimal control problem was formulated considering
feed temperature and steam to methane ratio as decision
variables based on two strategies. In the first strategy, the feed
temperature regulation is prior to feed concentration accord-
ing to the applied pattern in the industrial process. In the
second strategy, manipulation of the feed concentration is
prior to the feed temperature. The simulation results showed
that operating at the optimal condition increases hydrogen
production about 11.6%. The hydrogen per carbon dioxide
ratio in the outlet product from reformer were 6.72, 7.24 and
7.03 based on the conventional, first and second strategies,
respectively. The lower methane conversion, carbon monox-
ide production and higher catalyst activity were the main
benefits of second strategy, while lower carbon dioxide was
produced when the first strategy was applied on the reformer.
Finally based on the simulation results, the second strategy
was selected as the optimal procedure to maintain hydrogen
production at the desired level.

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2 11

Viscosity of species are calculated by:

Appendix A
Ai TBi
mi ¼ (B.8)
In order to calculate catalyst effectiveness factor, a cylindrical 1 þ CTi þ TD2i
shell element is selected in the raschig ring type catalyst and
Specific heat capacity of pure components and mixture are
the mole balance equations are written on the considered
estimated by temperature dependent correlation using:
element.
Based on the considered assumptions temperature ig
Cp;i
gradient in the catalyst is negligible. The balance equations ¼ Ai þ Bi T þ Ci T2 þ Di T3 (B.9)
R
are solved simultaneously to calculate the concentration
profile along the catalyst radius. It is mentioned that effec- X
6

tiveness factors are calculated at each element of reactor. By Cp ¼ Yi Cig

p (B.10)
1
integrating the actual rate of reaction inside the catalyst,
effectiveness factor is calculated [25].
  Nomenclature
1 d dCj
De;j r  rs rj ¼ 0 (A.1)
r dr dr
A Cross section ðm2 Þ
Z Ro av Catalyst specific surface area (m2 kg1 Þ
r  rj dr Ci Concentration of species (mol m3 )
Ri
hj ¼   (A.2) Cp Special heat capacity (kJ kmol1 K1 )
R20  R2i rsj
Cs Concentration of species on surface catalyst (mol
m3 )
Ct Total concentration (mol m3 )
Appendix B
Dp Catalyst particle diameter (m)
D Diffusion coefficient (m2 s1 )
The convective heat transfer coefficients in the coil tube and
dout Outer diameter of the reactor tube (m)
furnace are calculated by the following correlations [32,33]:
din Inner diameter of the reactor tube (m)
0 Ej Activation energy of reaction j (kJ kmol1 )
kB Fi Molar flow rate of specie i (mol s1 )
hg ¼1:6 B2
D@ hg Heat transfer coefficient (W m2 K1 )
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ki Thermal conductivity of specie i (W m1 K1 )
u 0  0:3 2 1ffi1
 12 u
= Kg;i Mass transfer coefficient for species (m s1 )
Re u
Re 3
0:0557 εb Pr CC
1
u1þ B AC Rate constant of reaction j (kmol kgcat 1 h1 )
=
þ0:644Pr 3 @  0:1  2 (B.1) kj
εb t = A
1þ2:443 εb Re
Pr 1 3
L Length of reactor (m)
P Total gas pressure (bar)
 0:8   Pi partial pressure of gas species i (bar)
k dout us rg mcp 0:33 ri Rate of species i (kmol kgcat 1 h1 )
hfurn ¼ 1:45  (B.2)
dout m k R Gas constant (pa m3 mol1 K1 )
The mass transfer and diffusion coefficients are calculated Rj Rate of reaction j (kmol kgcat 1 h1 )
from the following equations [34,35]: Ri Inner diameter of the catalyst (m)
  Ro Outer diameter of the catalyst (m)
Di;m 1 1
Tf Furnace temperature (K)
kg ¼ 0:6 Re2 Pr3 (B.3)
dp Ts Catalyst temperature (K)
Tw Wall temperature (K)
1y U Overall heat transfer coefficient (W m2 K1 )
Di;m ¼ P yi i (B.4)
Di;j us Superficial gas mixture velocity (m s1 )
z Axial direction (m)
2= qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0:001858T 3
MWm
1
þ MW 1
i Greek letters
Di;j ¼ (B.5)
Pt UD s2mi DHj Heat of reaction j (kj kmol1 )
a Catalyst activity
Thermal conductivity of gas mixture is estimated based on
εb Backed bed porosity
the pure gases thermal conductivity. The pure gases thermal
hj Effectiveness factor of reaction j
conductivity is calculated from following correlations [34]:
mg Gas viscosity (pa s1 )
X
6 ms Catalyst viscosity (pa s1 )
k¼ ðYi ki Þ (B.6) rB Bulk density (kg m3 )
1
rg Gas mixture density (kg m3 )
si Collision diameter of species i (Ǻ)
Ai TBi
ki ¼ (B.7) UD;i Collision integral for diffusivity of species i
1 þ CTi þ TD2i

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094
12 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 8 ) 1 e1 2

references [17] Rajesh J, Gupta S, Rangaiah G, Ray A. Multi-objective

optimization of industrial hydrogen plants. Chem Eng Sci
2001;56:999e1010.
[18] Bhat SA, Sadhukhan J. Process intensification aspects for
[1] Manaças VHC. Advanced steady-state modelling and
steam methane reforming: an overview. AIChE J
optimisation of Natural Gas Reforming reactors. Portugal:
2009;55:408e22.
University of Lisbon; 2013.
[19] Cruz BM, da Silva JD. A two-dimensional mathematical
[2] Baykara SZ. Hydrogen: a brief overview on its sources,
model for the catalytic steam reforming of methane in both
production and environmental impact. Int J Hydrogen Energy
conventional fixed-bed and fixed-bed membrane reactors for
2018;43:10605e14.
the Production of hydrogen. Int J Hydrogen Energy
[3] Haseli Y. Maximum conversion efficiency of hydrogen fuel
2017;42:23670e90.
cells. Int J Hydrogen Energy 2018;43:9015e21.
[20] Abbas S, Dupont V, Mahmud T. Modelling of H2 production
[4] Steinberg M, Cheng HC. Modern and prospective
in a packed bed reactor via sorption enhanced steam
technologies for hydrogen production from fossil fuels. Int J
methane reforming process. Int J Hydrogen Energy
Hydrogen Energy 1989;14:797e820.
2017;42:18910e21.
[5] Roh H-S, Eum I-H, Jeong D-W. Low temperature steam
[21] Aboosadi ZA, Rahimpour M, Jahanmiri A. A novel integrated
reforming of methane over NieCe (1x) Zr (x) O 2 catalysts
thermally coupled configuration for methane-steam
under severe conditions. Renew Energy 2012;42:212e6.
reforming and hydrogenation of nitrobenzene to aniline. Int J
[6] Zhu J, Zhang D, King K. Reforming of CH4 by partial
Hydrogen Energy 2011;36:2960e8.
oxidation: thermodynamic and kinetic analyses. Fuel
[22] Alam I, West DH, Balakotaiah V. Transport effects on pattern
2001;80:899e905.
formation and maximum temperature in
[7] Dincer I, Acar C. Review and evaluation of hydrogen
homogeneouseheterogeneous combustion. Chem Eng J
production methods for better sustainability. Int J Hydrogen
2016;288:99e115.
Energy 2015;40:11094e111.
[23] Holman J. Heat transfer. 9th ed. 2002. New York.
[8] Rowshanzamir S, Eikani M. Autothermal reforming of € zinger H, Taglauer E, Ertl G, Kno
€ zinger H, Weitkamp J,
[24] Kno
methane to synthesis gas: modeling and simulation. Int J
editors. Handbook of heterogeneous catalysisvol. 1.
Hydrogen Energy 2009;34:1292e300.
Weinheim: Wiley-VCH; 1997.
[9] Alatiqi I, Meziou A, Gasmelseed G. Modelling, simulation and
[25] Fogler HS. Elements of chemical reaction engineering. 4th ed.
sensitivity analysis of steam-methane reformers. Int J
Edinburgh: Pearson Education Limited; 2014.
Hydrogen Energy 1989;14:241e56.
[26] Alpay E, Kershenbaum L, Kirkby N. Pressure correction in the
[10] Asleshirin S, Bahmani M, Fazlali A, Fadavi O. The static and
interpretation of microreactor data. Chem Eng Sci
dynamic modeling of a steam methane reforming hydrogen
1995;50:1063e7.
plant. Petrol Sci Technol 2012;30:1882e92.
[27] Rostrup-Nielsen JR. Catalytic steam reforming. Catalysis:
[11] De Jong M, Reinders A, Kok J, Westendorp G. Optimizing a
Springer; 1984. p. 1e117.
steam-methane reformer for hydrogen production. Int J
[28] Rostrup-Nielsen JR. Production of synthesis gas. Catal Today
Hydrogen Energy 2009;34:285e92.
1993;18:305e24.
[12] Pantoleontos G, Kikkinides ES, Georgiadis MC. A
[29] Pen M, Gomez J, Fierro JG. New catalytic routes for syngas
heterogeneous dynamic model for the simulation and
and hydrogen production. Appl Catal Gen 1996;144:7e57.
optimisation of the steam methane reforming reactor. Int J
[30] Kirk DE. Optimal control theory: an introduction. Courier
Hydrogen Energy 2012;37:16346e58.
Corporation; 2012.
[13] Isebor OJ, Durlofsky LJ. Biobjective optimization for general
[31] Rostrup-Nielsen JR, Sehested J. Steam reforming for
oil field development. J Petrol Sci Eng 2014;119:123e38.
hydrogen. The process and the mechanism. Fuel Chem Div
[14] Nobandegani MS, Birjandi MRS, Darbandi T, Khalilipour MM,
Preprint 2003;48:218e9.
Shahraki F, Mohebbi-Kalhori D. An industrial Steam
[32] Koning GW. Heat and mass transport in tubular packed bed
Methane Reformer optimization using response surface
reactors at reacting and non-reacting conditions:
methodology. J Nat Gas Sci Eng 2016;36:540e9.
experiments and models. Twente University Press; 2002.
[15] Rajesh J, Gupta SK, Rangaiah Gt, Ray AK. Multiobjective
[33] Grevskott S, Rusten T, Hillestad M, Edwin E, Olsvik O.
optimization of steam reformer performance using genetic
Modelling and simulation of a steam reforming tube with
algorithm. Ind Eng Chem Res 2000;39:706e17.
furnace. Chem Eng Sci 2001;56:597e603.
[16] Farsi M, Shahhosseini H. A modified membrane SMR reactor
[34] Green DW, Perry RH. Perry's chemical engineers' handbook.
to produce large-scale syngas: modeling and multi objective
McGraw-Hill Professional; 1999.
optimization. Chem Eng Process Process Intensification
[35] Stewart WE, Lightfoot EN, Bird RB. Transport phenomena. J.
2015;97:169e79.
Wiley; 1962.

Please cite this article in press as: Taji M, et al., Real time optimization of steam reforming of methane in an industrial hydrogen plant,
International Journal of Hydrogen Energy (2018), https://doi.org/10.1016/j.ijhydene.2018.05.094