Energy & Fuels 2007, 21, 2977-2983 2977

Dynamic Optimization Strategies of a Heterogeneous Reactor for

CO2 Conversion to Methanol
G. Zahedi,† A. Elkamel,*,‡ and A. Lohi§
Simulation and AI research center, Department of Chemical Engineering, UniVersity of Razi, Kermanshah,
Iran, Department of Chemical Engineering, UniVersity of Waterloo, Waterloo, ON, Canada, N2L 3G1, and
Department of Chemical Engineering, UniVersity of Ryerson, Toronto, ON, Canada M5B 2K3

ReceiVed June 25, 2007

The inherent sustainability and low carbon dioxide (CO2) emissions of renewable energy technologies provide
the necessary features of future energy policy goals. To secure the contribution of renewables in the future
energy supply, practical dynamic optimization strategies for CO2 conversion into methanol in a heterogeneous
reactor are proposed. The conversion provides a solution to the problem of carbon dioxide emissions reduction
and at the same time produces a readily storable and transportable fuel. The dynamic optimization is carried
out on an industrial scale methanol reactor and considers shell temperature and inlet hydrogen mole fraction,
separately and simultaneously, as optimization variables. The optimizations have been carried out in both
steady- and unsteady-state modes. A heterogeneous model of the reactor has been used to obtain accurate
optimization results. In the dynamic mode, a stagewise optimization for 4 years of reactor operation has been
considered. The methanol production rate (MPR) has been selected as the objective function to find optimal
profiles. In the steady-state case, the optimal input mole fractions of CO, CO2, and H2 have been investigated.
The calculated concentration profiles are realistic and appropriate for the application of an industrial methanol
reactor. The staged optimization suggests a new guideline for operating the reactor. For the optimization of
input mole fractions, it was indirectly concluded that CO2 concentration adversely affects the overall process
as compared to catalyst poisoning due to the presence of CO.

1. Introduction considered to be a future renewable energy.3 The current paper

attempts to suggest practical methods to enhance methanol
The global warming caused by increasing concentrations of production from current industrial methanol reactors.
greenhouse gases in the atmosphere is a serious problem Methanol is one of the worldwide consumed chemicals which
concerning the fundamental conditions of life on Earth. CO2 is is usually used as fuel, as solvent, and as a building block to
considered to be the major cause of global warming. The net produce chemical intermediates.4 The following reactions can
increase of 13 000 million tones of carbon dioxide estimated be assumed to occur in methanol synthesis:
to be added to the atmosphere annually is alerting.1 The rising
level of CO2 is already affecting the atmosphere, sea level, and CO + 2H2 S CH3OH, ∆H 0 ) -90.55 kJ/mol (1)
ecological systems. The global sea level rose 10-20 cm over
the past century, and in this century, it may rise by as much as CO2 + 3H2 S CH3OH + H2O, ∆H 0 ) -49.43 kJ/mol (2)
88 cm. The current atmospheric concentration of carbon dioxide
at 380 ppm is 30% higher than the pre-industrial level and is
expected to increase to somewhere between 800 and 1000 ppm CO2 + H2 S CO + H2O, ∆H 0 ) +41.12 kJ/mol (3)
by the year 2010. The third session of the conferences of parties
to the United Nations framework convention on climate change Equation 2 is dependent on the other two equations and can be
held in Kyoto in December 1997 produced an agreement known obtained by the addition of eqs 1 and 2.5 In this study, eqs 1
as the Kyoto Protocol. The Kyoto Protocol calls for the majority and 3 have been used to represent the reaction system.
of industrialized countries to limit their emissions of greenhouse Independent reactions 1 and 3 are exothermic, and thus, the
gases over the period from 2008 to 2012.2 As a CO2 reduction formation of methanol is preferred with low temperature and
measure, the current study considers the conversion of CO2 to high pressure. Synthesis gas consists of H2, CO2, CO, and some
methanol. The latter is considered to be a valuable product inert components like methane and nitrogen. Figure 1 illustrates
because of its use as a feed stock to produce chemical a typical methanol plant.6 Table 17 represents specifications of
intermediates. Methanol is also easily transported and is the plant.

(3) Mullins, F.; Baron, R. Policies and Measures for Common Action;
* Corresponding author. E-mail: [email protected]. Working Paper 9, International GHG emission trading, OECD, Annex I,
† University of Razi. Expert Group on UN FCCC, 1997.
‡ University of Waterloo. (4) Methanex. www.methanex.com (accessed 2006).
§ University of Ryerson. (5) Skrzypek, J.; Lachowska, M.; Grzesik, M.; Sloczynski, J.; Nowak,
(1) Creutz, C.; Fujita, E. Carbon Dioxide as a Feedstock, Carbon P. Thermodynamics and kinetics of low-pressure methanol synthesis. Chem.
Management; The National Academy Press: Washington, D.C., 2000. Eng. J. 1995, 58, 101-108.
(2) Hileman, B. An urgent plea on global warming. Chem. Eng. News. (6) Lurgi. Integrated low pressure methanol process; technical report,
2004, 82 (26), 44. Lurgi Oil and Gas: Germany, 1995.

10.1021/ef700361m CCC: $37.00 © 2007 American Chemical Society

Published on Web 08/08/2007
2978 Energy & Fuels, Vol. 21, No. 5, 2007 Zahedi et al.

methanol reactor performance. The simultaneous effect of inlet

hydrogen and coolant temperature as optimization variables has
not, however, yet been studied in the literature. Furthermore,
the “stagewise” optimization for methanol reactors with a narrow
time increment (i.e., for each 100-day increment of operation)
has not been considered previously.
The present paper, therefore, deals with the simultaneous
optimization of inlet hydrogen and shell temperature regarding
reversible reactions with an industrial methanol reactor model.
In addition, inlet hydrogen and temperature are optimized
individually. Both effects of equilibrium and catalyst deactiva-
tion on reactor behavior are analyzed. Furthermore, dynamic
optimization and steady-state optimization are presented in this
research in order to provide a complete picture of the optimiza-
Figure 1. Schematic of methanol production plant. tion of the reactor.

Table 1. Specifications of SPC Methanol Reactor7 2. Process Modeling

feed component mole fraction
To get a mathematical model for the methanol reactor, it is
hydrogen 0.659 essential to write mass and energy balances. A heterogeneous
carbon monoxide 0.046
carbon dioxide 0.094 model is considered in this paper.13 The application of the
methane 0.1026 conservative principles to the bulk gas stream and film of gas
water 0.0004 on the surface of the catalyst leads to the following equations
nitrogen 0.093 for the reactor:
methanol 0.005
Continuity equations
Because of catalyst deactivation, the usual practice is to ∂yi Ft ∂yi
continuously increase the working temperature or decrease the Bct )- + aVctkgi(yi - ysi ) i ) 1, 2, ..., N - 1 (4)
feed rate of the reactants. A temperature increase has a secondary ∂t Ac ∂z
side effect that accelerates the deactivation rate, while a decrease
in the feed rate decreases the reactor production rate. So ∂ysi
sct ) kgi(yi - ysi ) + riFBa i ) 1, 2, ..., N - 1 (5)
considering these facts, an optimum operating temperature and ∂t
feed rate should be found in order to have sustainable methanol
production. Energy equations
The optimization of a methanol reactor has been the subject
of few studies. Larsson et. al8 considered the unconstraint ∂T Ft ∂T πDi
Bctcpg )- + aVhf(T s - T) + U (T - T)
optimization of an isothermal methanol reactor at steady state. ∂t Ac ∂z Ac shell shell
The reactor was modeled as a series of isothermal continuously (6)
stirred tank reactors (CSTRs), with a kinetic model from
dT s N

∑
Vanden, Bussche, and Froment.9 Lovik et al.10 studied dynamic
optimization of a methanol reactor. They reported the optimal FBcps ) aVhf(T - T ) + FBa
s
ri(-∆Hf,i) (7)
dt i)1
shell temperature as time goes on. Jahanmiri and Eslamloueyan11
investigated the optimal temperature profile in a steady-state Deactivation rate14
methanol reactor. Their model finds the optimal temperature
profile inside the reactor, which is impractical to implement in
industrial units. Recently, Kordabadi and Jahanmiri12 attempted
to find the optimal temperature of a methanol reactor using a
da
dt
) -Kact exp ( (
-Eact 1
-
R T Tref
1
a5 )) (8)

genetic algorithm. They also investigated the optimal division

Boundary and initial conditions
of a bed into two sections as two reactors. Their work focused
on combining steady state and multi-objective optimization
techniques to estimate optimal conditions. yi(0,t) ) y0i
Only Lovik et al.10 and Kordabadi and Jahanmiri12 used
T(0,t) ) T 0 (9)
dynamic industrial heterogeneous models to optimize the
yi(z,0) ) yi0
(7) SPC. Methanol reactor long sheet; Shiraz petrochemical company:
Shiraz, Iran, 2000-2003. ysi (z,0) ) ysi 0
(8) Larsson, T.; Skogestad, S.; Cheng-Ching, Yu. Presented at the AIChE
annual meeting, Dallas, TX, 1999.
(9) Vanden Bussche, K. M.; Froment, G. F. A steady-state kinetic model T(z,0) ) T0
for methanol synthesis and the water gas shift reaction on a commercial
Cu)ZnO)Al2O3 catalyst. J. Catal. 1996, 161 (1), 10. Ts(z,0) ) T s0
(10) Lovik, I. Modeling, Estimation and Optimization of the Methanol
Synthesis with Catalyst Deactivation. Ph.D. Thesis, Department of Chemical
Engineering, Norwegian University of Science and Technology, Norway, a(z,0) ) a0 (10)
2001.
(11) Jahanmiri, A.; Eslamloueyan, A. Optimal Temperature Profile in where yssi is the steady-state profiles of the component i mole
Methanol Synthesis Reactor. Chem. Eng. Commun. 2002, 189, 713-741.
(12) Kordabadi, H.; Jahanmiri, A. Optimization of methanol synthesis fraction and Tss is the steady-state temperature profile along the
reactor using genetic algorithms. Chem. Eng. J. 2005, 108, 249-255. reactor. The values of Tss and yss
i are determined from a steady-
Optimization of CO2 ConVersion to Methanol Energy & Fuels, Vol. 21, No. 5, 2007 2979

Figure 2. Steady-state optimum shell temperature. Figure 3. Steady-state optimum inlet hydrogen mole fraction.

state simulation. In the steady-state simulation, the left-hand 4.2. Optimization of Inlet Hydrogen Mole Fraction. In
sides of eqs 4-7 are set to zero. contrast to the CO16 and CO217 roles in catalyst deactivation,
H2 does not deactivate the catalyst directly. Noting eqs 1-3,
3. Process Optimization an increase in H2 can cause more methanol production but at
the same time leads to an increase in CO and CO2 production.
As was mentioned earlier, both dynamic and steady-state This fact motivates us to find an optimum mole fraction for
optimization are carried out. The objective function, J, for inlet hydrogen. In order to implement such a policy, the inlet
dynamic optimization is the methanol production rate (MPR), hydrogen mole fraction and consequently other components’
which was described in our earlier work15 and is presented in inlet mole fractions should vary. These are changed in this
eq 11 below. In steady-state optimization, the effluent methanol research in a way that the following equations are satisfied:
mole fraction is considered as the objective function, J. Different
optimization strategies for both the steady and the dynamic xH2O ) x H0 2O(1 - xH2) (15)
modes were implemented and will be described in detail in the
following sections.
The dynamic optimization problem is formulated as follows: xN2 ) x N0 2(1 - xH2) (16)

max (J ) MPR ) FMw,methanol ∫0txmethanol|z)1 dt) (11) xCO ) x 0CO(1 - xH2) (17)

which is subject to the process model (eqs 4-10) and the xCO2 ) x CO
0
2
(1 - xH2) (18)
following additional constraints

490 e TShell e 540 (12) xMeOH ) x 0MeOH(1 - xH2) (19)

0 e T e 543 (13) xCH4 ) 1 - (xH2 + xH2O + xN2 + xCO + xCO2 + xMeOH) (20)

0 e yH2 e 114 (14) x H2 ) x H2 (21)

For steady-state optimization, the deactivation is assumed to The mole fractions indicated with the superscript 0 are Shiraz
be unity, and J ) ymethanol|z)1. Taking into account the length Petrochemical Complex (SPC) plant data that are listed in Table
of the reactor and time simultaneously in optimization is another 1. The philosophy behind this is to change other components’
challenge considered in the current study. Matlab 7 was used mole fractions in a way consistent with their initial mole
to implement all optimization strategies. fractions. Methane and nitrogen are inert components, and
changing the methane proportion in the feed is easily achieved
4. Steady-State Optimization in the SPC methanol reactor because of the available natural
4.1. Optimization of Shell Temperature (Tshell). The gas resources. The optimal hydrogen mole fraction was found
optimization variable shell temperature is often controlled by a to be 0.6. Figure 3 illustrates the variation of the methanol outlet
proportional and integral (PI) controller in a real plant.7 mole fraction with inlet hydrogen.
Therefore, manipulating Tshell is of practical importance. The 4.3. Simultaneous Optimization of xH2 and Tshell. Figure 4
Levenberg-Marquart technique was employed to find the shows the variation of produced methanol mole fraction as a
optimum Tshell. Figure 2 shows the variation of ymethanol with function of inlet hydrogen fraction (xH2) and shell temperature
Tshell. For steady state, a Tshell value of 518 K was found as an
(16) Rahimpour, M. R.; Fathikalajahi, J.; Jahanmiri, A. Selective kinetic
optimum temperature. deactivation model for methanol synthesis from simultaneous reaction of
CO2 and CO with H2 on a commercial copper/zinc oxide catalyst. Can. J.
(13) Rezaie, N.; Jahanmiri, A.; Moghtaderi, B.; Rahimpour, M. R. A Chem. Eng. 1998, 76, 1-9.
comparison between homogeneous and heterogeneous modeling in Lurgi- (17) Zahedi, G.; Jahnmiri, A.; Rahimpor, M. R. Prediction of the CuO-
type methanol reactor. Chem. Eng. Process. 2005, 44, 911-921. ZnO-Al2O3 catalyst deactivation. Int. J. Chem. Reactor Eng. 2005, 3,
(14) Hanken, L. Optimization of methanol reactor. Master’s Thesis, The Article A8.
Norwegian University of Science and Technology, Norway, 1995. (18) Goh, C. J.; Teo, K. L. Control parameterization: A unified approach
(15) Zahedi, G.; Elkamel, A.; Lohi, A.; Jahanmiri, A., Rahimpor, M. R. to optimal control problems with general constraints. Automatica 1988, 24,
Hybrid model formulation for the unsteady state simulation of a packed 3-18.
bed reactor for CO2 hydrogenation to methanol. Chem. Eng. J. 2005, 115 (19) Ramirez, W. F. Process Control and Identification; Academic Press:
(1-2), 113-120. Boston, 1994.
2980 Energy & Fuels, Vol. 21, No. 5, 2007 Zahedi et al.

Figure 4. Simultaneous steady-state optimization of xH2 and Tshell.

Figure 5. Unsteady-state optimization of Tshell during 4 years of operation.

the MPR integral (eq 22) as a function of shell temperature

and hydrogen inlet mole fraction.

max (MPR integral ) ∫0txmethanol|z)1 dt) (22)

5.2. Dynamic Staged Optimization. 5.2.1. Tshell Dynamic

Staged Optimization. The time interval for this section has been
set as 100 days which is the narrowest increment in methanol
optimization. At each 100 days of operation, the optimum values
of Tshell are shown in Table 2. As time goes on, the optimal
Tshell increases. This seems reasonable as the raised temperature
Figure 6. Unsteady-state optimization of xH2 during 4 years of is for compensating the deactivation effect on MPR.
operation. 5.2.2. Simultaneous xH2 and Tshell Dynamic Staged Optimiza-
tion. This section is the most computationally intensive part of
the optimization. The optimization is a kind of an optimal control
(Tshell). The optimal values were found to be xH2 ) 0.62 and problem. Broadly speaking, the necessary conditions that will
Tshell ) 508 K. determine an optimal policy u(t) that causes a system to follow
a path x(t) such that a performance functional J ) ∫tt0f F(x,u,t)
5. Dynamic Optimization dt is maximized will be obtained. In order to apply these
circumstances, the Hamiltonian or adjoint eq 19 is often
5.1. Dynamic Optimization of xH2 and Tshell during 4-year introduced as follows:
Operation. Tshell was found to be 519 K during 4 years of
plant operation. The optimization code for xH2 as a single
variable provided a value of 0.6. Choosing xH2 and Tshell H ) F(x,u,t) + λTf(x,u,t) (23)
simultaneously as optimization variables led to xH2 ) 0.61
and Tshell ) 512 K. Figures 5-7 show the variations of where λ is a vector containing dynamic Lagrange multipliers.
Optimization of CO2 ConVersion to Methanol Energy & Fuels, Vol. 21, No. 5, 2007 2981

Figure 7. Unsteady-state optimization of xH2 and Tshell during 4 years of operation.

Table 2. Optimal Values of Tshell Every 100 days of Reactor The inlet hydrogen mole fraction and the shell cooling
Operation temperature have been set out as the optimization variables.
day Tshell (K) Implementing and changing their values in the real plant is
0-100 498 practical. Equations 12-14 are used as inequality constraints,
100-200 504 and the process model (eqs 4-10) serves as equality constraints.
200-300 511 The solution techniques for optimal control problems can be
300-400 513 categorized into two groups: indirect and direct methods. The
400-500 514
500-600 515 indirect methods solve eqs 23-26 and 28 and thus indirectly
600-700 518 solve the optimal control problem. The most well-known
700-800 520 methods in this category are boundary condition iteration (BCI)
800-900 522 and control vector iteration (CVI).21 However, these methods
900-1000 526
1000-1100 529
can exhibit numerical instabilities or slow convergence rates
1100-1200 534 for many problems.
There are two possible strategies for the direct methods:
The necessary conditions according to the fundamental theorem sequential and simultaneous. The sequential strategy is often
of the calculus of variation are the following:20 called control vector parametrization (CVP).21 Here, piecewise-
Euler-Lagrange equation constant control profiles are used. Hence, an infinite-dimensional
optimization problem in continuous control variables is trans-
∂H formed into a finite-dimensional nonlinear programming (NLP)
λ̇ ) - (24)
∂x problem that can be solved by any gradient-based method (e.g.,
sequential quadratic programming (SQP)). The performance
Process model index evaluation is carried out by solving an initial value
problem (IVP) of the original ordinary differential equation
∂H (ODE) system and gradients of the performance index as well
x̆ ) - (25)
∂λ as the constraints. Moreover, this method is a feasible-type
method, i.e., the solution is improved in every iteration. In the
Optimal control simultaneous strategy, both the control and state variables are
discretized using polynomials (e.g., Lagrange polynomials)
∂H whose coefficients become the decision variables in a much
)0 (26)
∂u larger NLP problem.23,24 Unlike the CVP method, the simul-
taneous method does not require the solution of IVPs at every
Transversality iteration of the NLP. The method is, however, of the infeasible-
type (i.e., the solution is available only once the iterative process
-λT(tf)δxf + H(tf)δtf ) 0 (27) has converged).
(21) Ray, W. H.; Szekely, J. Process Optimization with Applications in
u, λ, x, and H all are vectors of dimension N - 1. As the Metallurgy and Chemical Engineering; John Wiley and Sons: New York,
operating time is constant (4 years), eq 27 reduces to the 1973.
following: (22) Goh, C. J.; Teo, K. L. Control parameterization: A unified approach
to optimal control problems with general constraints. Automatica 1988, 24,
3-18.
λT(tf) ) 0 (28) (23) Biegler, L. T. Solution of dynamic optimization problems by
successive quadratic problems and orthogonal collocation. Comp. Chem.
Eng. 1984, 8, 243-248.
(20) Pinch, E. R. Optimal Control and the Calculus of Variations; Oxford (24) Cuthrell, J. E.; Biegler, L. T. On the optimization of differential-
University Press: Oxford, 1993. algebraic process systems. AIChE J. 1987, 33, 1257-1270.
2982 Energy & Fuels, Vol. 21, No. 5, 2007 Zahedi et al.

Table 3. Comparison of Optimal MPR with Reference Case

ref MPR (heterogeneous optimal MPR
time (day) model (ton/day)) (ton/day)
0-100 305.66 311.2
100-200 293.28 296.2
200-300 285.30 288
300-400 279.36 282.5
400-500 274.57 277.7
500-600 270.53 273.15
600-700 267.03 268.6
700-800 263.94 265.1
800-900 261.16 262.2
900-1000 258.64 260.2
1000-1100 256.32 257.2
1100-1200 254.19 255.3
Figure 8. Schematic of simultaneous xH2 and Tshell dynamic staged
optimization.
u(t) was approximated with 12 constant-piecewise profiles over
1200 days of operation. The profiles were taken to be of the
same length, but this restriction can be easily waived in our
numerical implementation scheme. The optimization problem
is therefore constrained by a set of partial differential equations
(PDEs).
Assuming that a constant control vector at each time
increment (100 days) accelerates the solution of Euler-Lagrange
equations. Compared to other indirect methods, the convergence
of the current method (which is similar to CVI) is fast. A
schematic representation of this algorithm is shown in Figure
8. The integration of the model differential equations and
because of their stiff nature was performed by the modified
Rosenbrock25 formula of order 2. The integrals in the model
and optimization algorithm were numerically evaluated using
Figure 9. Time varying reactor shell temperature for simultaneous a recursive adaptive Simpson quadrature. Matlab 7 has been
optimal control of Tshell and yH2. used too in this part of the study to implement all of the routines
and optimization procedures.
Figure 9 shows the cumulative effect of deactivation kinetics
and chemical equilibrium on the optimal evaluation of shell
temperature. The first characteristic noticed is the continuous
increase of the shell temperature after day 200. This is to
compensate for the effect of catalyst deactivation on the MPR.
The second characteristic effect is the temperature drop after
day 100. According to ref 13, it is obvious that there is a great
amount of H2O formation at the beginning of the operation and
then its production trend becomes slow. Noting this fact and
noting that eqs 1 and 3 are the governing equations of the
Figure 10. Trajectory of inlet hydrogen mole fraction for simultaneous
control of Tshell and yH2. reactor, one can conclude that, at the beginning, endothermic
eq 3 is the rate-controlling step and, in order to increase the
In order to solve the optimal control problem for the methanol MPR at the beginning, temperature should increase. By the
reactor, the control variables, Tshell(t) and xH2(t), were discredited passage of time, the exothermic reaction (eq 1) becomes the
as piecewise-constant profiles (Figure 8) and the control vector rate-controlling step so the temperature should decrease. The

Figure 11. Unsteady optimization based on the presence of only CO, CO2, and H2 in the feed.
Optimization of CO2 ConVersion to Methanol Energy & Fuels, Vol. 21, No. 5, 2007 2983

latter is required in order to move the reaction mixture away with xH2 at each 100 days of operation. As a final effort and in
from the equilibrium state and therefore to increase reactant the dynamic mode of operation, the optimal feed mole fractions
conversion. when the feed contains only CO, CO2, and H2 have also been
Figure 10 illustrates the hydrogen mole fraction variation with obtained. This optimization emphasized our previous finding
time. On the basis of the above arguments for days 0-100, eq that the CO2 effect on catalyst deactivation is more important
3 is the controlling step. After day 100, eq 1 is controlling, so than that of CO.
hydrogen demand increases according to the stoichiometric
coefficient of hydrogen in eq 1. After day 200, by the passage Acknowledgment. The authors would like to acknowledge the
of time, the deactivation mechanism will be the dominant financial support of Razi University and the National Science and
controlling mechanism and hydrogen demand decreases to Engineering Research Council (NSERC) of Canada.
compensate for this effect. Table 3 provides a comparison of
MPR based on the proposed optimal method and a reference Nomenclature
case. This reference case is MPR based on SPC operating
Ac ) cross area of each tube (m2)
conditions tabulated in Table 1. It is obvious from Table 3 that aV ) external particle surface area per unit of reactor volume
the simultaneous strategy provides better MPR. Comparing with (m2‚m-3)
the reference case, MPR increases by 1.4045% over 4 years a ) activity of catalyst (-)
and this is equivalent to a profit of about $1 400 000 USD. a0 ) initial activity of catalyst (-)
For more details on variations of other component mole cpg ) specific heat of the gas at constant pressure (J‚mol-1)
fractions affected by our optimal control policy and singularity cps ) specific heat of the solid at constant pressure (J‚mol-1)
of control, see ref 27. ct ) total concentration (mol‚m-3)
5.3. Optimization of Inlet Components. For this section, Di ) tube inside diameter (m)
an ideal case was considered. Inlet syngas was assumed to Eact ) activation energy used in the deactivation model (J‚mol-1)
contain only CO, CO2, and H2. The xCO/xCO2 and xH2/xCO2 ratios Ft ) total molar flow per tube (mol‚s-1)
were chosen as optimization variables. Figure 11 represents the f ) function denotes process model
H ) Hamiltonian
results for this optimization. It was found that xCO ) 0.147 29,
J ) objective function (kg)
xCO2 ) 0.077 52, and xH2 ) 0.775 19 to be the optimum inlet Kact ) deactivation model parameter constant (s-1)
mole fractions during 4 years of reactor operation. It is kgi ) mass transfer coefficient for component i (m‚s-1)
interesting to note that the mole fraction of CO2 has decreased Mw ) molecular weight (kg‚kg-mol-1)
compared to CO. This can be another proof for our previous R ) universal gas constant (J‚mol-1‚K-1)
result that the role of CO2 in deactivation is important.17 This ri ) reaction rate of component i (mol‚kg-1‚s-1)
case is an ideal case which can be implemented in a laboratory- T ) bulk gas-phase temperature (K)
scale setting. In an industrial-scale rector, however, syngas has Ts ) temperature of gas on the solid surface (K)
other components like methane, nitrogen, and water and their Tshell ) temperature of boiling water in the shell side (K)
removal is not economical. Tref ) reference temperature (K)
t ) time (s)
tf ) final operating time (s)
6. Conclusion u(t) ) vector of optimization variables
The optimization of an industrial methanol reactor in order Ushell ) boiling water-gas overall heat transfer coefficient
to enhance its economical performance in both steady- and (W‚m-2‚K-1)
unsteady-state modes was considered in this paper. In the steady- x ) states of the system
yi ) bulk gas-phase mole fraction for component i (-)
state mode, the optimum values of Tshell and xH2 as separate s
and simultaneous optimization variables have been found. In yi ) mole fraction of the ith component in the solid phase (-)
the dynamic mode, the optimum values of Tshell and xH2 as z ) axial reactor coordinate (m)
separate and simultaneous optimization variables during 1200 Greek Letters
days of operation have also been investigated. Furthermore, the ∆Hf,i ) formation enthalpy of component i (J‚mol-1)
optimization has also considered Tshell and also Tshell combined B ) void fraction of catalytic bed (m3‚m-3)
S ) solid particle void fraction (m3‚m-3)
(25) Shampine, L. F.; Reichelt, M. W. The MATLAB ODE Suite. SIAM FB ) bed density (kg‚m3)
J. Sci. Comput. 1997, 18, 1-12.
(26) Mathwork. http://www.mathworks.com/support/solutions/data/1- λ ) Lagrange multiplier
15NM7.html?solution)1-15NM7 (accessed 2005).
(27) Zahedi, G.; Elkamel, A.; Lohi, A. Enhancing CO2 conversion to
Superscripts and Subscripts
methanol using dynamic optimization, applied on shell temperature and 0 ) inlet or initial conditions
inlet hydrogen during 4 years operation of methanol plant. Energy Sources,
in press. EF700361M