Professional Documents
Culture Documents
Optimization of Demethanizer Distillation Column Using Genetic Algorithm
Optimization of Demethanizer Distillation Column Using Genetic Algorithm
Abstract
In this study the demethanizer distillation column located in PC-1 ethylene plant
in Basrah was subjected to simulation and optimization to find the optimum
operating conditions using genetic algorithm and Aspen plus.
A steady state simulation model was developed to study the behavior of multi-
component mixture in the demethanizer distillation column. The column model is
composed of the MESH (Material balance, Equilibrium, Sum of mole fractions and
Heat balance) equations. The simulation and optimization were conducted by using
Matlab version (2010.a) and Aspen plus. The simulation model was subjected to
validity test by comparing the simulation results with the actual data collected from
PC-1 in Basra. The validity test shows a great reliability for both simulation models
comparing with actual unit results.
The operating conditions were optimized using genetic algorithm and Aspen
plus. The effect of the decision variables (reflux ratio, reboiler heat duty, condenser
cooling duty and the feed temperatures) on the production rate of methane, ethane
and ethylene and hence on total profit were studied. Genetic algorithm results
showed that the multi-objectives can be achieved by manipulating the decision
variables to increase the profit to approximately (82) US dollar more than actual
profit.
Keywords
Optimization, Simulation, Demethanizer, Distillation Column, Genetic
Algorithm, Aspen plus.
The designing and selecting of operating conditions for the distillation column
is a very challenging task because there are many variables must be taken into
consideration in the designer calculations such as; dimensions of the column,
number of plats, type of the plat, feed location, side products location, ... etc.
Distillation is usually modeled with equilibrium MESH model equations. In
addition to MESH equations, thermodynamic correlations for individual
components are required in distillation modeling (1).
The main objective and the driving force for the optimization is the economic
consideration which is done by two methods: minimizing the energy and maximize
the profit. The economic optimization of a distillation column involves the
selection of the optimum design specification and operating conditions such as
number of trays, feed location, side product location and the energy used in the
separation and to minimize the total investment and operation cost (2, 3).
Optimization of the distillation column is a very hard issue because there are
many variables that must be taken in calculations which give very wide range of
variables to choose from, before the computers the selection of the number of the
variables was done by hand and take a lot of time, now with the availability of the
modern computers that task is much easier. There are many stochastic optimization
methods and programs for the optimization research (3).
2
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
fittest (10). Genetic Algorithms has been widely studied, experimented and applied
in many fields in engineering worlds (11).
Mathematical Model
3
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Model Assumptions
The proposed model include the following assumptions:-
X
i 1
i, j 1.0 0 …….. (4)
Heat balance or energy balance for i components on stage j are:
Lj-1 hj-1-(Vj + SVj) Hj - (Lj + SLj) hj + Fj hFj + Vj+1 Hj+1 - Qj = 0 ….…. (5)
The total material balance on each stage is:
Lj-1 -(Vj + SVj) - (Lj + SLj) + Fj + Vj+1 = 0 ….…. (6)
The vapor and liquid on each stage is in equilibtuim so the liquid at each stage
is at its bubble point and the vapor on that stage is at its dew point. Newton-
Raphson's numerical method was used for estimating bubble points.
Equations (1 to 4) were used to calculate the composition of each component at
each tray within the distillation column. The set of equations for each component at
each tray are solved by Gaussian elimination algorithm. The total mateial and heat
balance equations represned in equations (5) and (6) are solved by 2N * 2N
Jacobian matrix to estimate the vapor and liquid flowrates at each stage (1).
4
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
5
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Simulation techniques
6
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
In the (PC-1) complex there are three sources of feed to the complex. The
classification of these feeds was based on the ethylene content in each feed with
(12%, 16% and 19%) ethane. The demethanizer designed to handle all of these
feeds, all of the simulation and optimization in this study was based on the (16%)
ethane feed stocks.
Before starting the optimization procedure, the simulation program was
subjected to validity test to verify the model validity and its reliability to
demonstrate the actual demethanizer tower in PC-1 complex.
The results obtained from the simulation program were compared with the
existing data collected from the PC-1 data sheets (Lummus manual, 1978), for the
demethanizer feed stock of 16% ethane content to verify the validity of the
simulation model. The difference between the actual data and results estimated
from Matlab simulating model are shown in table (2).
Table (2): Difference between the actual data and estimation results (ethane
content 16%)
Top product Bottom product
Comp. Actual Estimated Actual Estimated
Abs. Error Abs. Error
(kmol/hr) (kmol/hr) (kmol/hr) (kmol/hr)
H2 29.34 29.34 0 0 0 0
CO 0.39 0.39 0 0 0 0
CH4 78.23 78.2273 0.0027 0.16 0.1527 0.0073
C2H4 3.51 3.4755 0.0345 712.38 712.3594 0.0206
C2H6 0.17 0.184 0.014 515.78 515.7804 0.0004
C2H2 0.04 0.00784 0.039992 3.83 3.8606 0.0306
C3H6 0 6.22E-11 6.22E-11 11.92 11.8913 0.0287
C3H8 0 1.34E-12 1.34E-12 2.39 2.3843 0.0057
C3H4 0 0 0 0.25 0.2494 0.0006
i-C4H10 0 0 0 0.22 0.2195 0.0005
C4H8 0 0 0 1.55 1.5463 0.0037
C4H6 0 0 0 8.33 8.32 0.01
n-C4H10 0 0 0 1.82 1.8157 0.0043
i-C5H12 0 0 0 1.72 1.7159 0.0041
n-C5H12 0 0 0 0.24 0.2395 0.0005
Total
111.6800 111.625 0.05536 1260.59 1260.54 0.055
kmol/hr
The estimated result from the Aspen plus was also compared with the existing
data of the demethanizer (16%) ethane feed stock to verify the validity of the
Aspen plus simulation model. The difference between actual and estimated data
from Aspen plus 11.1 is shown in table (3).
7
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Table (3): The difference between actual and estimated data from Aspen plus
11.1
Top product Bottom product
Comp. Actual Estimated Actual Estimated
Abs. Error Abs. Error
( kmol/hr) (kmol/hr) ( kmol/hr) ( kmol/hr)
H2 29.3400 29.3400 0 0 0 0
CO 0.3900 0.3900 0 0 0 0
CH4 78.2300 78.2900 0.06 0.1600 0.1000 0.06
C2H4 3.5100 3.5000 0.01 712.3800 712.3900 0.01
C2H6 0.1700 0.1600 0.01 515.7800 515.7700 0.01
C2H2 0.0400 0.0380 0.002 3.8300 3.8300 0.00
C3H6 0 0 0 11.9200 11.9200 0
C3H8 0 0 0 2.3900 2.3900 0
C3H4 0 0 0 0.2500 0.2500 0
i-C4H10 0 0 0 0.2200 0.2200 0
C4H8 0 0 0 1.5500 1.5500 0
C4H6 0 0 0 8.3300 8.3400 0.01
n-C4H10 0 0 0 1.8200 1.8200 0
i-C5H12 0 0 0 1.7200 1.7200 0
n-C5H12 0 0 0 0.2400 0.2400 0
Total
111.68 111.772 0.082 1260.59 1260.54 0.05
kmol/hr
From the above two table’s one can conclude that both two programs are able to
simulate the demethanizer distillation column very well.
Optimization
Finding the optimum operating conditions for the distillation column is very
difficult task, the objective of the distillation is different from one process to
another, due to the difference in the objective function of the distillation column
there are different decision variables.
The calculation of the economic objective function for the distillation column
can be written as follows (15) :
Objective function = (D PD+B PB –F PF)- (R PR + S PS) ……… (8)
Equation (8) was modified by replacing the top products by methane production
rate and the bottom by ethane and ethylene production rates to insure the separation
factor in addition to maximizing the profit.
The modified objective function equation will be written as follow:
Profit = (Methane * unit price $ + Ethane * unit price $ + Ethylene * unit price $ -
Feed(s) * unit price $) - (Qreb * unit price $+ Qcond * unit price $) …….. (9)
Where: Methane, ethane, ethylene and the feed is measured in k-mole per hour
and Q-reb and Q-cond are measured in Kj/hr.
The prices of the (PC-1) complex for the variables of equation (9) are shown in
table (4).
8
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Table (4): unit price for the feed, products and the energy consumed
Variable Measuring Unit Unit Price $
Feed price kmol / hr 4.1587-3
Methane (from Top) kmol / hr 3.7658
Ethane (from Bottom) kmol / hr 6.6139
Ethylene (from Bottom) kmol / hr 23.1485
Q- reboiler kJ / hr 2.8435-5
Q- condenser kJ / hr 2.8435-5
Table (5): Actual operating conditions for (16%) ethane content feed stock
Discretion Value Units
Reflux ratio 1.197 ----
Reboiler heat duty 6.52x106 kj / hr
Condenser heat duty 1.05x106 kj / hr
Feed (F1) temperature -98.1 ºC
Feed (F2) temperature -70 ºC
Feed (F3) temperature -34.9 ºC
For performing optimization on demethanizer and finding the optimum
operating conditions; equation (9) was used to calculate the profit with the aim of
maximizing the profit.
The actual profit based on the values of table (5) are (19958.93) US dollar per
hour. The actual profit will be used as a reference value to be compared with the
optimum profit estimated from Aspen plus programs.
9
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Tables (6) shows that, the profit increasing at each generation and maximum
profit was found at generation (5710) at which the profit equal to (20041.13) US
dollar per hr.
Genetic algorithm optimization results have the flexibility of selection the local
optimum operating conditions due to the changes in objective function. As the
10
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
demands of the marks change or the requirements of the PC-1 complex changes,
the local optimum operating conditions can be selected from the genetic algorithm
optimization results to satisfy the new demands.
The local optimums for the demethanizer from table (6) are:
At generation of (998); the methane separation factor and production reach
the maximum value, where methane separation factor is equal to
(0.999974) and the production of methane is (78.388) kmole / hr. The
methane production increases about (0.2%) compared with actual methane
production at generation (2358).
At generation of (4671); the ethane separation factor and production reach
the maximum value, where ethane separation factor is equal to (0.999983)
and the production of ethane is (515.9302) kmole / hr. The ethane
production increases about (0.04%) compared with actual ethane
production at generation of (2358).
At generation of (4912); the ethylene separation factor and production
reach the maximum value, where ethylene separation factor is equal to
(0.995993) and the production of ethylene is (714.521) kmole / hr. The
ethylene production increases about (0.3%) compared with actual ethylene
production at generation of (2358).
The maximum profit can be found at the last generation of (5710); where
the profit equals (20041.13) US dollar per hour. The increase in the profit
from the actual profit at generation (2358) is equal to (0.42 %) which
represents (82 US dollar per hour).
The interaction between the decisions variables make the searching for the
global optimum operating conditions is necessary. Global optimum can be found at
generation (4912) at which all the decision variables increased.
Table (7) shows the differences between the values of the decision variables and
the profit as well as methane, ethane and ethylene production rates at actual
operating conditions (at generation 2358) and the optimum operating conditions (at
generation 4912).
11
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Table (7): The difference between actual and the optimum values at
generation (4912)
Changing
Actual Optimum Changing
Variables Percentage
Operating Operating Sign
%
Reflux Ratio 1.197 1.201 0.004 Increasing
Q-reboiler KJ/hr 7090000 7091690 1690 Decreasing
Q-condenser KJ/hr -1050000 -1091850 41850 Decreasing
Profit US dollar per hr 19958.93 19999.91 40.98 Increasing
Methane kmole/hr 78.2318 78.385 0.1532 Increasing
Methane separation factor 0.997982 0.999939 0.001957 Increasing
Ethane kmole/hr 515.7804 515.93 0.1496 Increasing
Ethane separation factor 0.999691 0.999981 0.00029 Increasing
Ethylene kmole/hr 712.364 714.521 2.157 Increasing
Ethylene separation factor 0.995089 0.995993 0.000904 Increasing
12
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Table (8) shows that, the profit increasing at each iteration and reach the
maximum profit at iteration (11597) at which the profit equal to (20038.13) US
dollar per hour. The percentage increasing in profit base on the profit at actual
operating conditions (at iteration 9124) is equal to (79) US dollar per hour.
As the same for genetic algorithm aspen plus optimization results have the
flexible of selection the locale optimum operating conditions due to the changes in
objective function. As the demands of the markets change or the requirements of
the (PC-1 complex) changes the optimum operating conditions can be selected
form the aspen plus optimization results to satisfy the new demands.
13
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
The local optimums for the demethanizer from table (8) are:
At iteration (8741); the methane separation factor and production reach the
maximum value, where methane separation factor is equal to (0.999542)
and methane production is (78.354) kmole / hr. The methane production
increases about (0.16%) from the actual methane production at iteration
(9124).
At iteration (6948); the ethane separation factor and production reach the
maximum value, where ethane separation factor is equal to (0.999981) and
ethane production is (515.931) kmole / hr. The ethane production increases
about (0.03%) from the actual ethane production at iteration (9124).
At iteration (11047); the ethylene separation factor and production reach
the maximum, where ethylene separation factor is equal to (0.997849) and
ethylene production is (714.350) kmole / hr. The ethylene production
increases about (0.28%).
The maximum profit can be found at iteration (11597), where the profit is
equal to (20038.13) US dollar per hr. The profit increases from the actual
profit at iteration (9124) by (0.4%) which represents (79 US dollar per
hour).
The optimum results can be found at iteration (11345), where all the
decision variables increase based on the actual operating conditions at
iteration (9124).
Table (9) shows the differences between the values of the decision variables and
the profit as well as methane, ethane and ethylene production at actual operating
conditions (at iteration 9124) and the optimum operation conditions (at iteration
11345).
14
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Table (9): The difference between actual and the optimum values at Iter.
(11345)
Changing
Actual Optimum Changing
Variables Percentage
Operating Operating Sign
%
Reflux Ratio 1.197 1.0883 0.1087 Decreasing
Q-reboiler KJ/hr 7090000 7073300 16700 Decreasing
Q-condenser KJ/hr -1050000 -1041000 9000 Decreasing
Profit US dollar per hr 19958.93 19979.61 20.68 Increasing
Methane kmole/hr 78.2318 78.35312 0.12132 Increasing
Methane separation factor 0.997982 0.999497 0.001515 Increasing
Ethane kmole/hr 515.7804 515.928 0.1476 Increasing
Ethane separation factor 0.999691 0.999976 0.000285 Increasing
Ethylene kmole/hr 712.364 713.164 0.8 Increasing
Ethylene separation factor 0.995089 0.996206 0.001117 Increasing
Observers of the above two tables (7) and (9) can be clearly noted that the
improvement in the final profit is very few and it is due to the fact that the tower
actually designed to work at the best conditions.
The optimization by aspen plus is faster than by genetic algorithm. By GA the
assumptions of the decision variables and creating the population as well as
evaluating of the objective function takes about (0.1) sec. The time for each
simulating of the demethanizer is about (1.01) min. The long simulation time
occurs due to the 2211 equations that need to be solved to simulate demethanizer
distillation column.
As for time consumption; the simulation by aspen plus are solved by (10-13)
sec. and the assumption with the solving of the profit equation about (4) sec. but
the number of iteration needed to find the optimum value is more the number of
genes in GA optimization program. Fig (3) shows the profit change with number of
trays for genetic algorithm and aspen plus optimization programs.
15
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Conclusions
Demethanizer (DA-1501) which is one of the distillation columns in the
petrochemical complex PC-1 in Basra was subjected to optimization. Two
programs were used; Genetic algorithm and Aspen plus to performer the
optimization.
In this study the following conclusions were found:
1- Proposed simulation model is reliable to simulate the demethanizer
distillation column.
2- From the simulation and optimization results may be to note that the results
of mathematical model tremendously apply to practical results taken at the same
conditions.
3- Proposed mathematical model can easily be applied to any distillation
tower in the other process.
4- Simulation by genetic algorithm is slower than optimization by Aspen
plus.
5- The improvement in the final profit is few and it is due to the fact that the
tower actually designed to work at the best conditions.
16
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
Nomenclature
Symbol Definition Units
ASPEN Advanced System for Process Engineering -
D Distillate molar flow rate kmol/hr
GA Genetic Algorithm -
H Vapor molar enthalpy kJ/kmol
h Liquid molar enthalpy kJ/kmol
i Component identification number -
j Stage identification number -
K Equilibrium constant -
L Liquid molar load on each stage kmole/hr
MESH (Material balance, Equilibrium, Summation of mole, -
Heat balance)
N Number of stages -
ODE Ordinary Differential Equation -
P Price $ dollar
PC Petrochemical complex -
PSO Particle Swarm Optimization -
Q Rate of heat transfer kJ/hr
R Reflux ratio -
S Steam flow rate in reboiler kg/hr
SA Simulated Annealing -
SC Simultaneous Convergence -
SL Liquid side stream molar rate -
SV Vapor side stream molar rate -
T Temperature K
TSA Simulated Annealing Temperature -
V Vapor molar load on each tray kmole/hr
Xij Liquid mole fraction -
Yij Vapor mole fraction -
Zij Feed mole fraction -
17
Optimization of Demethanizer Distillation
Column using Genetic Algorithm
References
1) Perry, "Perry’s chemical engineers’ handbook". 8th ed. New York, McGraw-
Hill, (2008).
2) Mauhar S. M., Barjaktarovi C B. G., and Sovilj M. N., " Optimization of
Propylene—Propane Distillation Process", Chem. Pap. 58(6), p386-390,
(2004).
3) Vázquez – Castillo, J.A.; Venegas – Sánchez, J.A.; Segovia - Hernández, J.G.;
Hernández – Escoto, H.; Hernández, S.; Gutiérrez - Antonio, C.; & Briones –
Ramírez A., "Design and Optimization, using Genetic Algorithms, of
Intensified Distillation Systems for a Class of Quaternary Mixtures", Chem.
Eng., 33, 11, p1841-1850, 0098-1354, (2009).
4) Popov A., "Genetic Algorithms for Optimization", Version 1.0 User Manual, ,
Hamburg, (2005).
5) Melanie Mitchell, An Introduction to Genetic Algorithms, Book, Fifth printing,
(1999).
6) Andradottir S., “A Review of Simulation Optimization Techniques”, In
Proceedings of the 1998Winter Simulation Conference, edited by D. J.
Medeiros, E. F. Watson, J. S. Carson, and M. S. Manivannan, p151–158,
(1998).
7) Wei-Zhong A., and Xi-Gang, Y., "A simulated Annealing-Based Approach to
the Optimal Synthesis of Heat-Integrated Distillation Sequences", Comput.
Chem. Eng., 33, 1, p199-212, 0098-1354. (2009).
8) Zheng Jianchao, Jie Jing, Cui Zhihua., "Particle swam optimization Algorithm
[M]", Science Publishing Company of Beijing, (2004).
9) Wayburn T. L. & Seader J. D., "Homotopy continuation methods for
computer-aided process design", Computers & Chemical Engineering 11(1),
p7–25, (1987).
10) Holland J. H., "Adaptation in natural and artificial systems: an introductory
analysis with applications to biology, control, and artificial intelligence",
(1975).
11) Abhijit Tarafder, Ajay K. Ray and Santosh K. Gupta, " Applications of Genetic
Algorithm for Solving Multi-Objective Optimization Problems in Chemical
Engineering", Computers & Chem. Eng., 32, p49-56, (2008).
12) Kister, H. Z., "Distillation Design" (McGraw-Hill), (1992).
13) MATLAB Software version 2010a, www.mathworks.com
14) Aspen plus 11.1, user guide, Aspen Technology, Inc. All rights reserved,
(2001), www.aspentech.com
15) Cecil L. Smith, "Distillation Control", An Engineering Perspective, Published
by John Wiley & Sons, Inc., Hoboken, New Jersey, Published simultaneously
in Canada, chapter 6, p252, (2012).
18