international journal of hydrogen energy 34 (2009) 6221–6230

Available at www.sciencedirect.com

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

Dynamic optimization of a novel radial-flow, spherical-bed

methanol synthesis reactor in the presence of catalyst
deactivation using Differential Evolution (DE) algorithm

M.R. Rahimpour*, P. Parvasi, P. Setoodeh

Chemical Engineering Department, School of Chemical and Petroleum Engineering, Shiraz University, Shiraz 71345, Iran

article info abstract

Article history: In this work, a novel radial-flow spherical-bed methanol synthesis reactor has been
Received 16 February 2009 optimized using Differential Evolution (DE) algorithm. This reactor’s configuration visual-
Received in revised form izes the concentration and temperature distribution inside a radial-flow packed bed with
12 May 2009 a novel design for improving reactor performance with lower pressure drop. The dynamic
Accepted 14 May 2009 simulation of spherical multi-stage reactors has been studied in the presence of long-term
Available online 27 June 2009 catalyst deactivation. A theoretical investigation has been performed in order to evaluate
the optimal operating conditions and enhancement of methanol production in radial-flow
Keywords: spherical-bed methanol synthesis reactor. The simulation results have been shown that
Dynamic optimization there are optimum values of the reactor inlet temperatures, profiles of temperatures along
Differential Evolution algorithm the reactors and reactor radius ratio to maximize the overall methanol production. The
Methanol synthesis optimization methods have enhanced additional yield throughout 4 years of catalyst
Spherical-bed reactor lifetime, respectively.
Catalyst deactivation ª 2009 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights
reserved.

1. Introduction remove the heat by heat exchangers between the reactors.

The flow in SPBRs is radial so that it offers a larger mean cross-
Methanol is an important multipurpose based chemical, sectional area and reduced distance of travel for flow
a simple molecule which can be recovered from many compared to traditional vertical columns. Consequently, the
resources, predominantly natural gas [1]. Tubular packed bed pressure drop in these radial geometry reactors is reduced
reactors (TPBRs) are used extensively in industrial methanol radically. Another characteristic of the spherical radial-flow
synthesis [2]. Some potential drawbacks of this type of reactor (RFR) is that since heat transfer will be dominated by
reactors are the pressure drop across the reactor, high the spherical geometry, a small hot zone should develop
manufacturing costs resulting from a large wall thickness and where reaction occurs. This small reaction zone can provide
low production capacity [3]. One potentially interesting idea several advantages, especially in reversible exothermic
for industrial methanol synthesis is the use of a spherical reactions.
packed bed reactor (SPBR) [4], which is the subject of this work. The factors affecting the production rate in an industrial
The advantages of this reactor are a small pressure drop, low methanol reactor are parameters such as thermodynamic
manufacturing costs as a result of a small wall thickness, and, equilibrium limitations and catalyst deactivation. In the case
if desired, a high production capacity. It is possible to connect of reversible exothermic reactions, such as methanol
different spherical reactors in series in a production plant and synthesis, selection of a relatively low temperature permits

* Corresponding author. Tel.: þ98 711 2303071; fax: þ98 711 6287294.
E-mail address: [email protected] (M.R. Rahimpour).
0360-3199/$ – see front matter ª 2009 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2009.05.068
6222 international journal of hydrogen energy 34 (2009) 6221–6230

Nomenclature r2 rate of reaction for hydrogenation of CO2,

kgmol m3 s1
a activity of catalyst
r3 reaction rate constant for the 3rd rate equation,
cpg specific heat of the gas at constant pressure,
kgmol m3 s1
J kgmol1 K1
ri reaction rate of component j, kgmol m3 s1
Di tube outside diameter, m
R universal gas constant, J kgmol1 K1
Derj diffusion coefficient of component j in the
Ri inner diameter of reactor, m
mixture, m2 s1
Ro outer diameter of reactor, m
Ed activation energy used in the deactivation model
t time, s
J kgmol1
T bulk gas phase temperature, K
fj partial fugacity of component j, bar
TR reference temperature used in the deactivation
Ft total molar flow rate per tube, mol s1
model, K
DH298i enthalpy of reaction i at 298 K
Tshell temperature of coolant stream, K
k1 reaction rate constant for the 1st rate equation,
Ushell overall heat transfer coefficient between coolant
mol kg1 s1 bar1/2
and process streams, Wm1 s1
k2 reaction rate constant for the 2nd rate equation,
ur radial velocity of fluid-phase, m s1
mol kg1 s1 bar1/2
V total volume of reactor, m3
k3 reaction rate constant for the 3rd rate equation,
yj mole fraction of component j in the fluid-phase,
mol kg1 s1 bar1/2
kgmol m3
kjg mass transfer coefficient for component j, m s1
keff conductivity of fluid-phase, Wm1 K1 Greek letters
Kd deactivation model parameter constant s1 3 void fraction of catalytic bed
Kj adsorption equilibrium constant for component j, 3s void fraction of catalyst
bar1 n stoichiometric coefficient
Kpj equilibrium constant based on partial pressure for r density of catalytic bed, kg m3
component j
Superscripts and subscripts
M number of reactions
0 inlet conditions
N number of components
i reaction number index (1, 2 or 3)
P total pressure, bar
j number of components
r radial coordinate m
s at catalyst surface
r1 rate of reaction for hydrogenation of CO,
ss initial conditions (i.e., steady-state condition)
kgmol m3 s1

higher conversion, but this must be balanced against a slower time or the maximum profit per unit of production. When one
rate of reaction, which leads to the requirement of a large design variable is changed, it is often found that some costs
amount of catalyst. Up to the maximum production rate point, increase and others decrease. Under these conditions, the
increasing temperature improves the rate of reaction, which total cost may go through a minimum at one value of the
leads to more methanol production. Nevertheless as the particular design variable, and this value would be considered
temperature increases beyond this point, the deteriorating as an optimum. A number of search algorithm methods for
effect of equilibrium conversion emerges and decreases dealing with optimization problems have been proposed in
methanol production. Therefore one of the important key the last few years in the fields of evolutionary programming
issues in methanol reactor configuration is implementing (EP) [6], evolution strategies (ES) [7], genetic algorithms (GAs)
a higher temperature at the entrance of the reactor for [8] and particle swarm optimization (PSO) [9]. DE algorithm is
a higher reaction rate, and then reducing temperature grad- a stochastic optimization method minimizing an objective
ually towards the exit for increasing thermodynamic equilib- function that can model the problem’s objectives while
rium conversion. incorporating constraints. The algorithm mainly has three
Like in the world of modeling, the field of dynamic opti- advantages; finding the true global minimum regardless of the
mization has its own jargon to address specific characteristics initial parameter values, fast convergence, and using a few
of the problem. Most optimization problems in process control parameters. Being simple, fast, easy to use, very easily
industry can be characterized as non-convex, non-linear, and adaptable for integrand discrete optimization, quite effective
constrained optimization problems [5]. For plant optimization in non-linear constraint optimization including penalty
typical optimization parameters are equipment size, recycle functions and useful for optimizing multi-modal search
flows and operating conditions like temperature, pressure and spaces are the other important features of DE algorithm [10].
concentration. An optimum design is based on the best or In this study, the novel radial-flow spherical-bed methanol
most favorable conditions. In almost every case, these synthesis reactor configuration has been optimized. Optimi-
optimum conditions can ultimately be reduced to a consider- zation tasks have been investigated by novel optimization
ation of costs or profits. Thus an optimum economic design tools, Differential Evolution (DE) algorithm. Optimization of
could be based on conditions giving the least cost per unit of reactor was studied in four approaches. In the first approach,
 international journal of hydrogen energy 34 (2009) 6221–6230 6223

an optimization program has been developed to obtain more 2.1. Mutation

methanol production through optimum inlet temperature of
reactors. In the second approach, the optimal temperature For each target vectorxi;G , a mutant vector is produced by
profiles along the reactors during the period of operation have


been considered to reach maximum methanol production vi;Gþ1 ¼ xi;G þ K  xr1 ;G  xi;G þ F  xr2 ;G  xr3 ;G (1)
rate. In the third approach, the optimum radius ratios of where i; r1 ; r2 ; rr3 ˛f1; 2; .; NPg are randomly chosen and must
reactors were obtained for maximum production rate. Finally, be different from each other. In Eq. (1), F is the scaling factor
the deactivation parameters were optimized using plant data. which has an effect on the difference vector ðxr2 ;G  xr3 ;G Þ, K is
the combination factor [10].

2. Differential Evolution (DE) algorithm 2.2. Crossover

The DE algorithm is a population based algorithm similar to The parent vector is mixed with the mutated vector to
genetic algorithms using similar operators: crossover, muta- produce a trial vector uji;Gþ1
tion, and selection. The main difference in constructing better 

uji;Gþ1 if rndj  CR

or j ¼ rni ;
solutions is that genetic algorithms depend on crossover uji;Gþ1 ¼ (2)
qji;G if rndj > CR and jsrni ;
while DE relies on mutation operation. This main operation is
founded on the differences of randomly sampled pairs of where j ¼ 1; 2; .; D; rj ˛½0; 1 is the random number, CR is
solutions in the population. crossover constant ˛½0; 1, and rni ˛ð1; 2; .; DÞ is the randomly
The algorithm uses mutation operation as a seek mecha- chosen index [10].
nism and selection operation to direct the search toward the
probable regions in the search space. The DE algorithm also 2.3. Selection
uses a non-uniform crossover that can take child vector
parameters from one parent more often than it does from All solutions in the population have the same chance of being
others. Using the components of the existing population selected as parents independent of their fitness value. The
members to build trial vectors, the recombination (crossover) child produced after the mutation and crossover operations is
operator efficiently shuffles information about successful evaluated. Then, the performance of the child vector and its
combinations, enabling the search for a better solution space. parent is compared and the better one is selected. If the parent
An optimization task consisting of D parameters can is still better, it is retained in the population.
be represented by a D-dimensional vector. In DE, initially Fig. 1 shows DE’s process in detail: the difference between
a population of NP solution vectors is randomly created. This two population members (1, 2) is added to a third population
population is successfully improved by applying mutation, member (3). The result (4) is subject to crossover with the
crossover, and selection operators. The main steps of the DE candidate for replacement (5) in order to obtain a proposal (6).
algorithm are given below [10]: The proposal is evaluated and replaces the candidate if it is
found to be better.
Initialization
Evaluation
Repeat 3. Model development
Mutation
Recombination Methanol synthesis is generally performed by passing
Evaluation a synthesis gas comprising hydrogen, carbon oxides, and any
Selection inert gasses at an elevated temperature and pressure through
Until (termination criteria are met) one or more beds of catalyst, which is often a copper–zinc
oxide catalyst. The following three overall equilibrium reac-
tions are relevant in the methanol synthesis [11]:

Feed

Product

Fig. 2 – Schematic diagram of a methanol synthesis single-

Fig. 1 – Obtaining a new proposal in DE [6]. stage spherical reactor.
6224 international journal of hydrogen energy 34 (2009) 6221–6230

Feed

Product

Fig. 3 – Schematic diagram of a methanol synthesis three-stage spherical reactor.

CO þ 2H2 4CH3 OH (3) [13], making the corresponding changes for a spherical
geometry. In this study, homogeneous one-dimensional
models have been considered. The basic structure of this
model is composed of heat and mass balance conservation
CO2 þ 3H2 4CH3 OH þ H2 O (4)
equations coupled through thermodynamic and kinetic rela-
tions, as well as, auxiliary correlations for predicting physical
CO þ H2 O4CO2 þ H2 (5) properties.
In this simple model we assume that gradients of
temperature and concentrations between catalyst and gas
A schematic sketch of the spherical reactor is presented phases can be ignored and the equations for the two phases
in Fig. 2. The catalyst is situated in the dome and between can be combined [14]. The general fluid-phase balance is
two perforated spherical shells. The synthesis gas enters a model with the balances typically account for accumulation,
the reactor between the catalyst bed and the pressure- convection, and reaction. In the current work, axial dispersion
resistant reactor wall. It flows steadily from outside of heat is neglected and the heat loss by a coolant is consid-
through the catalyst bed into the inner sphere. The gas is ered as we study a realistic reactor. The energy and mass
removed from the inner sphere through a tube to the balances can be written as [15]:
outlet [12].
n
Due to small pressure drop and low manufacturing costs vyj 1 v 2  X
3 ¼ 2 r ur yj þ 3s ð1  3Þa yij r
multi-stage, spherical-bed reactors could be utilized instead of vt r vr i¼1
single ones in order to achieve production improvement. The
three-stage configuration is illustrated in Fig. 3. j ¼ 1; 2; .; N and i ¼ 1; 2; .; M (6)

n
vT 1 v X
3.1. Reactor model rur r2 Cp T  Tref þ 3s ð1  3Þa



ð1  3ÞrCpc ¼ 2 DHi r (7)
vt r vr i¼1

The mathematical model corresponding to the spherical where T and yj are, respectively, the temperature and
packed bed flow reactor is derived starting from the dynamic concentration of component j in the fluid-phase and a is the
model developed by Rahimpour et al. for tubular flow reactors activity of catalyst.

Table 1 – Catalyst and reactor specifications [15].

Parameter Value Table 2 – Input data for first reactor [15].
3
rs (kg m ) 1770 Feed conditions Value
dp (mm) 5.47
Composition (mol%)
Cps (kJ kg1 K1) 5.0
CH3OH 0.5
3 0.5
CO2 9.4
3s 0.4
CO 4.6
aap (m2 m3) 626.98
H2O 0.04
Cpg (kJ kg1 K1) 2.98
H2 65.9
V (m3) 80
N2 9.33
Spherical One-stage Two-stage Three-stage CH4 10.26
reactor (m) (m) (m)
Total molar flow rate (kmol s1) 2
Inner radius 1.0 1.0 1.0 Inlet temperature (K) 503
Outer radius 2.72 2.19 1.94 Pressure (bar) 76.98
 international journal of hydrogen energy 34 (2009) 6221–6230 6225

Table 3 – Specification data for reactor feed [12].

Feed (kmol s1) P (bar) T (K) H2 (mol%) CO (mol%) CH3OH (mol%) CO2 (mol%) H2O (mol%) N2 (mol%) CH4 (mol%)

17.7892 81.95 313 77.53 3.9695 0.4114 2.5197 0.0806 3.4285 12.05

The boundary conditions are as follows: 3.3. Deactivation model

vyj vT
r ¼ Ri ; ¼ 0; ¼0 Catalyst deactivation model for the commercial methanol
vr vr (8)
r ¼ Ro ; yj ¼ yj0 ; T ¼ T0 synthesis catalyst was adopted from Hanken [17].

da Ed 1 1
  
The initial conditions are: ¼ Kd exp  a5 (16)
dt R T TR
t ¼ 0; yi ¼ yss
i ;
ss
T¼T ; a¼1 (9)
where TR, Ed and Kd are the reference temperature, activation
energy, and deactivation constant of the catalyst, respectively.
The chosen numerical values for these parameters are:
3.2. Reaction kinetics TR ¼ 513 K, Ed ¼ 91270 J mol1, and Kd ¼ 0.00439 h1.

Reactions (3)–(5) are not independent and therefore one is Steady State Condition
a linear combination of the other ones. Kinetics of the low- 0.1573
pressure methanol synthesis over commercial CuO/ZnO/
0.1573
Al2O3 catalysts has been widely investigated. In the current
work, the rate expressions have been selected from Graaf et al. Objective Function 0.1573
The corresponding rate expressions due to the hydrogenation
0.1572
of CO, CO2, and the reversed water–gas shift reactions are
given in Appendix A [16]. 0.1571
The reaction rate constants, adsorption equilibrium
0.1571
constants, and reaction equilibrium constants, which occur
in the formulation of kinetic expressions, are tabulated in 0.1571
Appendix A, respectively.
0.157

0.1569

0.1569
0 10 20 30 40 50 60 70 80 90 100
Table 4 – Comparison of simulation results with Hartig Iteration
et al. [12] data.
Fig. 4 – Objective function values for steady-state
Outlet composition Homogenous Hartig Relative optimization.
(mol%) model et al. data error (%)

First reactor
H2 78.006 76.550 1.902
CO 3.113 3.321 6.281
CH3OH 1.740 1.613 7.826
Optimum Reactor inlet Temperature
CO2 2.009 2.131 5.705 520
H 2O 0.558 0.532 4.964
N2 3.693 3.510 5.195 515
CH4 12.978 12.340 5.172
510
Second reactor
505
Temperature (K)

H2 77.088 75.640 1.915

CO 2.396 2.612 8.249 500
CH3OH 3.043 2.806 8.479
495
CO2 1.694 1.811 6.471
H 2O 0.960 0.913 5.152 490
N2 3.800 3.591 5.809
CH4 13.355 12.620 5.828 485

Third reactor 480

First Reactor
H2 76.580 75.010 2.093 Second Reactor
475
CO 1.801 2.017 10.709 Third Reactor
CH3OH 4.212 3.705 13.700 470
0 200 400 600 800 1000 1200
CO2 1.476 1.630 9.442
H 2O 1.233 1.141 8.146 Time (day)
N2 3.915 3.652 7.182
Fig. 5 – Dynamic optimal inlet temperatures for (a) first
CH4 14.009 12.840 9.104
stage (b) second stage and (c) third stage.
6226 international journal of hydrogen energy 34 (2009) 6221–6230

The partial differential equations (PDEs) of the one-

Table 5 – Optimal values of reactors’ inlet temperatures
(K). dimensional model of spherical reactor were solved by means
of the orthogonal collocation numerical method [18,19].
Time (day) First reactor Second reactor Third reactor
Numerical solution was performed by Rahimpour et al. [20].
0 471.71 503.74 502.07 For more accuracy of homogeneous model, the results of
100 476.53 505.84 503.70 steady-state predictions for spherical reactors are compared
200 484.64 509.62 507.92
with Hartig et al.’s data [12] and a good agreement was ach-
300 489.44 511.51 509.88
ieved. Tables 3 and 4 show these results.
400 493.03 512.97 511.04
500 495.62 514.11 512.14 Differential Evolution algorithm is applied to determine the
600 497.35 514.82 513.08 optimal reactor operating conditions for methanol production
700 499.16 515.45 513.64 process. The goal of this work is to maximize the methanol
800 500.60 515.98 514.17 production during 1200 days of operation.
900 501.87 516.50 514.66 In this study, the optimization of reactor was investigated
1000 502.97 517.03 515.19
using four approaches: optimal reactor inlet temperature
1100 503.89 517.40 515.55
approach, optimal temperature profile approach, optimal
1200 504.82 517.79 515.89
catalyst deactivation parameter approach, and optimal radius
ratio of reactors.

4. Optimization and results

4.1. Optimization of the reactor inlet temperatures
The technical design data of the catalyst pellet and input data
of the reactor have been summarized in Tables 1 and 2, For this optimization study, inlet temperature of reactor is
respectively. variable because the design conversion of reversible

a Optimum mole fraction from First reactor outlet b Optimum mole fraction from Second reactor outlet
0.044 0.062
With Optimization With Optimization
0.042 Without Optimization 0.06 Without Optimization
0.058
0.04
Methanol mole fraction

Methanol mole fraction

0.056
0.038
0.054
0.036
0.052
0.034
0.05
0.032
0.048
0.03 0.046

0.028 0.044

0.026 0.042
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Time (day) Time (day)

c Optimum mole fraction from Third reactor outlet

0.08
With Optimization
Without Optimization
0.075
Methanol mole fraction

0.07

0.065

0.06

0.055

0.05
0 200 400 600 800 1000 1200
Time (day)

Fig. 6 – Optimal methanol mole fraction for (a) first stage (b) second stage and (c) third stage.
 international journal of hydrogen energy 34 (2009) 6221–6230 6227

of mole fraction of methanol component at the reactor outlet.

Table 6 – Additional yield achieved for output methanol
mole fractions with optimal temperature profiles. Fig. 4 shows the objective function values for steady-state
optimization.
Reference Optimum Additional yield (%)
Optimal inlet temperature profiles through catalyst life-
Activity ¼ 0.9 0.0673 0.0698 3.71 time for three-stage reactors are shown in Fig. 5.
Activity ¼ 0.7 0.0668 0.0694 3.89 Values of optimal inlet temperatures for some days are
Activity ¼ 0.5 0.0663 0.0688 3.83
reported in Table 5.
Optimal reactor inlet temperatures should be increased
during the catalyst lifetime to compensate for the reduction of
exothermic reactions often closely approach the upper
production rate due to catalyst deactivation. For instance, this
boundary imposed by thermodynamics [21].
optimization approach enhanced a 30% additional yield for final
In the case of reversible exothermic reactions, selection of
product at first reactor. Fig. 6 shows the difference between
a relatively low temperature permits higher conversion, but this
methanol mole fractions at reactor outlet for this activity.
must be balanced by the slower rate of reaction resulting in the
need for a large amount of catalyst. Hence, the reactor inlet
temperatures in three-stage reactors can be adjusted (at optimal 4.2. Optimal temperature profile approach
temperatures) to maximize methanol mole fraction at reactor
outlet. Therefore, according to deactivation rate, dynamic From a theoretical point of view, there is an optimal
optimal temperatures are obtained through catalyst lifetime. temperature profile along the methanol synthesis reactor,
In this study, the objective function used to determine which maximizes methanol production rate as reported in
optimized inlet temperature of reactors is defined as the sum the literature for exothermic reactors. The optimal profile

a Optimum Temperatue Profile - First Reactor b Optimum Temperatue Profile - Second Reactor
(Activity = 0.9) (Activity = 0.9)
600 550

590 545

580 540

570
Temperatue (K)

Temperatue (K)

535
560
530
550
525
540
520
530
515
520
510
510
505
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
Radius (m) Radius (m)

c Optimum Temperatue Profile - Third Reactor

(Activity = 0.9)
540

535

530
Temperatue (K)

525

520

515

510

505
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
Radius (m)

Fig. 7 – Optimal temperature profiles at a [ 0.9 for (a) first stage (b) second stage and (c) third stage.
6228 international journal of hydrogen energy 34 (2009) 6221–6230

Average of Square Absolute Error in Methanol Production vs. Time

Methanol Production 310
54 Plant Data
Hanken Deactivation Model
300 Optimized Deactivation Model
52

Methanol Production (ton)

50 290
Objective Function

48
280
46
270
44

42 260

40
250
0 200 400 600 800 1000 1200
38 Time (day)
0 10 20 30 40 50 60
Iteration Fig. 9 – Comparison of Hanken and optimized deactivation
Fig. 8 – Objective function values for optimization of models with observed methanol production rate (ton/day).
deactivation parameters.

m
da CO Ed 1 1
   
changes during operation because of catalyst deactivation ¼ Kd exp  an (17)
dt CO2 R T TR
and therefore, it is not unique over different time intervals
[22]. In this study pseudo-dynamic optimization was used The optimization of parameters (Ed, Kd, n and m) was
instead dynamic optimization. According to deactivation performed by DE algorithm and results of new model for
rate, three activity levels (a ¼ 0.9, 0.7 and 0.5) were chosen to one-stage spherical reactor for feed flow rate 2.22 kgmol s1
study optimal inlet temperatures. These values reflect were compared with Hanken’s model and plant data of
dynamic properties of reactor operation and give some reactor. In this study, the objective function, which is used
information about variation of optimal temperatures to determine optimized deactivation parameters, is defined
through catalyst lifetime [23]. In this approach, the objective as the average of square absolute error in methanol
function is similar to the objective function of the previous production between plant data and model results through
section (Table 6). catalyst lifetime. Fig. 8 shows the objective function values
Fig. 7 shows optimal temperature profile along the meth- for this optimization.
anol synthesis reactors for activity level of 0.9. Values of optimal deactivation parameters are reported in
Table 7.
Fig. 9 illustrates the results of methanol production, in
which new catalyst deactivation model is compared with
4.3. Optimization of catalyst deactivation Hanken’s model.
rate parameters

The deactivation of low-pressure methanol synthesis catalyst

has been investigated by Kuechen and Hoffmann [24]. In their
experiments, the proportion of hydrogen was kept constant Table 8 – Values of methanol production predicted by
whilst the CO:CO2 ratio was varied in order to make a clear optimal deactivation model compared with Hanken’s
model.
distinction between completely different behavior of the
catalyst due to variations in the CO:CO2 ratio. Therefore, Time Plant data Hanken’s Percent New Percent
(day) (ton/day) model error with model error with
deactivation kinetics is a function of temperature and the
(ton/day) plant data (ton/day) plant data
fugacity ratio of carbon monoxide to carbon dioxide as illus-
trated in the following equation: 100 296.50 293.69 0.95 290.26 2.10
200 302.60 291.47 3.68 286.29 5.39
300 284.33 289.95 1.98 283.43 0.32
400 277.90 288.78 3.92 281.12 1.16
500 278.20 287.80 3.45 279.09 0.32
Table 7 – Optimal values of deactivation model.
600 278.00 286.95 3.22 277.31 0.25
Ed 111542.07 J mol1 700 274.00 286.19 4.45 275.73 0.63
Kd 0.4301156 h1 800 268.10 285.51 6.50 274.22 2.28
n 4.000021 900 275.50 284.89 3.41 272.73 1.00
m 0.000960 1000 274.58 284.31 3.54 271.33 1.18
 international journal of hydrogen energy 34 (2009) 6221–6230 6229

Methanol Production vs. Time

310 Appendix A
300
Auxiliary correlations
Methanol Production (ton)

290 To complete the simulation, auxiliary correlations should be

280 added to the model.

270

260 A.1. Reaction kinetics

250 h  i
Raduis Ratio = 2 k1 KCO fCO fH1:5
2
 fCH3 OH = fH0:5
2
KP1
240
Raduis Ratio = 3
R1 ¼  h   i (A.1)
1 þ KCO fCO þ KCO2 fCO2 fH0:5 þ KH2 O =K0:5 H2 fH2 O
230 Raduis Ratio = 4 2

Raduis Ratio = 5
220 h  i
0 200 400 600 800 1000 1200 k2 KCO2 fCO2 fH1:5
2
 fCH3 OH fH2 O = fH1:52 KP2
Time (day) R2 ¼  h   i (A.2)
1 þ KCO fCO þ KCO2 fCO2 fH0:5 2
þ KH2 O =K0:5 H2 fH2 O
Fig. 10 – The effects of reactor radius ratio on methanol
production rate (ton/day). h i
k3 KCO2 fCO2 fH2  fH2 O fCO =KP3
R3 ¼  h   i (A.3)
1 þ KCO fCO þ KCO2 fCO2 fH0:52
þ KH2 O =KH 0:5
2
fH2 O

The methanol production of the one-stage reactor

achieved by optimized deactivation model and Hanken’s A.2. Reaction constants
model is compared with plant data [25] in Table 8.

4.4. Optimization of reactor radius ratio Table A.1 – Reaction rate constants.
B
k ¼ A expðRT Þ A B
In this section, inner and outer radius ratio of one-stage 7
k1 (4.89  0.29)  10 113,000  300
reactor is variable. We try to find the optimum radius ratio, k2 (1.09  0.07)  105 87,500  300
which maximizes the final product. Results show that k3 (9.64  7.30)  1011 152,900  11800
increasing the radius ratio increases methanol production and
compensates catalyst deactivation effect (Fig. 10). However, it
is not clear if there is a unique optimum value for radius ratio.
In Fig. 10, the radius ratio increases from 2 to 5 to restore
methanol production. The objective function was similar to Table A.2 – Adsorption equilibrium constants.
the objective function of the first optimization subsection, and B
k ¼ A expðRT Þ A B
the chosen value for the inner radius was 1 m.
KCO (2.16  0.44)  105 46,800  800
KCO2 (7.05  1.39)  107 61,700  800
0:5
ðKH2 O =KH 2
Þ (6.37  2.88)  109 84,000  1400

5. Conclusion

In this study, a radial-flow spherical-bed methanol synthesis

reactor has been optimized to maximize methanol production Table A.3 – Reaction equilibrium constants.
yield. The optimization problem includes four approaches: A
Kp ¼ 10ð t BÞ A B
optimal reactor inlet temperature approach, optimal temper-
K p1 5139 12.621
ature profile approach, optimal catalyst deactivation param-
K p2 3066 10.592
eter approach, and optimal radius ratio of reactors. Results of K p3 2073 2.029
optimization procedures yield high additional methanol
production during the operating period. A homogeneous
model was used in the optimization investigation. Dynamic
references
optimization is the method of choice for solving the
constrained non-linear problem under study. Dynamic opti-
mization was implemented using Differential Evolution (DE)
[1] Khademi MH, Jahanmiri A, Rahimpour MR. A novel
algorithm, which is a powerful optimization technique with
configuration for hydrogen production from coupling of
a reasonably low computational complexity. Also, the methanol and benzene synthesis in a hydrogen-
proposed method provides guidelines to perform similar permselective membrane reactor. Int J Hydrogen Energy,
designs for methanol synthesis. 2009;34:5091–107.
6230 international journal of hydrogen energy 34 (2009) 6221–6230

[2] Rahimpour MR, Alizadehhesari K. Enhancement of carbon dynamic models for industrial methanol reactors in the
dioxide removal in a hydrogen-permselective methanol presence of catalyst deactivation. Chem Eng Process
synthesis reactor. Int J Hydrogen Energy 2009;34:1349–62. 2005;44(8):911–21.
[3] Rahimpour MR. Enhancement of hydrogen production in [15] Rahimpour MR, Elekaei H. Enhancement of methanol
a novel fluidized-bed membrane reactor for naphtha production in a novel fluidized-bed hydrogen
reforming. Int J Hydrogen Energy 2009;34:2235–51. -permselective membrane reactor in the presence of
[4] Viecco Guillermo A, Caram Hugo S. The spherical reverse catalyst deactivation. Int J Hydrogen Energy 2009;34:
flow reactor. Chem Eng Sci 2002;57:4005–25. 2208–23.
[5] Parvasi P, Khosravanipour Mostafazadeh A, Rahimpour MR. [16] Askari Fatemeh, Rahimpour Mohammad Reza. Dynamic
Dynamic modeling and optimization of a novel methanol simulation and optimization of a dual-type methanol reactor
synthesis loop with hydrogen-permselective membrane using genetic algorithms. Chem Eng Technol 2008;31(No. 4):
reactor. Int J Hydrogen Energy 2009;34:3717–33. 513–24.
[6] Fogel LJ, Owens AJ, Walsh MJ. Artificial Intelligence through [17] Hanken L. Optimization of methanol reacto. Master’s thesis,
simulated evolution. New York: John Wiley & Sons; 1966. The Norwegian University of Science and Technology; 1995.
[7] Rechenberg I. Evolution strategy: optimization of technical [18] Balakotaiah V, Luae D. Effect of flow direction on conversion
systems by means of biological evolution. Stuttgart: in isothermal radial flow fixed-bed reactors. AIChE J 1981;
Fromman-Holzboog; 1973. 27(3):442.
[8] Huang Sy-Ruen, Lin Chiu-Yue, Wu Chueh-Cheng, Liu Shun- [19] Hlavkek V, Kubicek M. Modeling of chemical reactors-XXV,
Tsai, Lin Liang-Chieh, Yun Shin-Joe. Optimizing circuit cylindrical and spherical reactor with radial flow. Chem Eng
parameters of the biohydrogen real-time power generating Sci 1972;27:177.
system by genetic algorithm. Int J Hydrogen Energy 2008;33: [20] Rahimpour MR, Abbasloo A, Sayyad Amin J. A novel radial-
1598–606. flow, spherical-bed reactor concept for methanol synthesis
[9] Kennedy J, Eberhart R. Particle swarm optimization. In: in the presence of catalyst deactivation. Chem Eng Technol
Proceedings of the IEEE international conference on neural 2008;31(No. 11):1615–29.
networks; 1995. p. 1942–8. [21] Rahimpour MR, Ghader S. Incorporation of flexibility in the
[10] Storn R. Differential evolution – a simple and efficient design of a methanol synthesis loop in the presence of
heuristic strategy for global optimization over continuous catalyst deactivation. Chem Eng Technol 2003;26(No. 6).
spaces. J Global Optimization 1997;11:341–59. Dordrecht. [22] Rahimpour MR, Elekaei Behjati H. Dynamic optimization of
[11] Parvasi P, Rahimpour MR, Jahanmiri A. Incorporation of membrane dual-type methanol reactor in the presence of
dynamic flexibility in the design of a methanol synthesis catalyst deactivation using genetic algorithm. Fuel Process
loop in the presence of catalyst deactivation. Chem Eng Technol 2009;90(2):279–91.
Technol 2008;31(No. 1):116–32. [23] Kordabadi H, Jahanmiri A. Optimization of methanol
[12] Hartig F, Keil FJ. Large-scale spherical fixed bed reactors. J Ind synthesis reactor using genetic algorithms. Chem Eng J 2005;
Eng Chem Res 1993;32:424–37. 108(3):249–55.
[13] Rahimpour MR, Moghtaderi B, Jahanmiri A, Rezaie N. [24] Kuechen C, Hoffmann U. Investigation of simultaneous
Operability of an industrial methanol synthesis reactor with reaction of carbon monoxide and carbon dioxide with
mixtures of fresh and partially deactivated catalyst. Chem hydrogen on a commercial copper/zinc oxide catalyst. Chem
Eng Technol 2005;28(2):226–34. Eng Sci 1993;48(No. 22):3767–76.
[14] Rezaie N, Jahanmiri A, Moghtaderi B, Rahimpour MR. [25] Operating data sheets of methanol plant. Domestic
A comparison of homogeneous and heterogeneous Petrochemical Plant; 2000–2003.