Journal of Cleaner Production 171 (2018) 795e810

Contents lists available at ScienceDirect

Journal of Cleaner Production

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

Synthesis and mixed integer programming based optimization of

cryogenic packed bed pipeline network for purication of natural gas
Abulhassan Ali a, Khuram Maqsood b, *, Lam Pei Shin a, Varsheta Sellappah a, Sahil Garg a,
Azmi B.M. Shariff a, Saibal Ganguly c, **
a
Chemical Engineering Department, Universiti Teknologi PETRONAS, Bandar Sri Iskandar, 31750, Perak, Malaysia
b
Chemical Engineering Department, NFC Institute of Engineering and Fertilizer Research, Faisalabad, Pakistan
c
Chemical Engineering Department, BITS-PILANI, GOA Campus, 403726, India

a r t i c l e i n f o a b s t r a c t

Article history: The variation of temperature, pressure and composition during cryogenic purication of natural gas
Received 15 March 2017 using cryogenic packed bed network signicantly affects separation. The present work aims to optimize
Received in revised form cryogenic packed bed network for maximum product purity with minimum hydrocarbon losses. The
5 October 2017
separation concept between components of natural gas is based on the difference in freezing points of
Accepted 6 October 2017
Available online 7 October 2017
individual components and their composition in the mixture. The overall process of cryogenic packed
bed pipeline network consists of the initial cooling of each bed, the capture of individual component and
regeneration of each bed. In previous studies, the process concept of cryogenic packed bed network was
Keywords:
Cryogenic CO2 capture
validated using experimentation but the optimization study was not carried out. The simulation results
Cryogenic packed bed network and thermodynamic behavior of the individual component in the mixture are included to further explain
Process intensication the process concept. The sensitivity study was carried out to investigate the effect of the temperature and
the pressure of each packed bed. For sensitivity study, node-edge schemes were developed and discussed
in details by changing operating conditions of packed beds. Based on the node-edge scheme analysis, the
optimum values for both temperature and pressures were selected for maximum separation and mini-
mum methane loss. The energy requirements for both dehydration and CO2 removal were calculated
experimentally and compared with simulation results. For the simulation results, Aspen Hysys simulator
was used along with Peng-Robinson uid package. The energy requirements for both simulation and
experimental results were calculated at atmospheric and high pressures. Both local and global optima
were calculated using golden section search and depth-rst strategy. The initial local optima result
showed that the product purity can reach up to 77% methane, further optimization using depth-rst
strategy showed the product purity up to 94% with reduction of hydrocarbon losses from 39 to 16%.
2017 Elsevier Ltd. All rights reserved.

1. Introduction
Natural gas is a fossil fuel, which has emerged as one of the
cleanest, safest and most useful sources of energy in the world. The
worldwide consumption of natural gas is expected to increase 1.6%
annually till 2030. Raw natural gas consists largely of methane
along with impurities such as water, carbon dioxide, and other
higher hydrocarbons. Though several impurities are present in
* Corresponding author. Chemical Engineering Department, NFC Institute of
Engineering and Fertilizer Research, Faisalabad, Pakistan.
** Corresponding author.
E-mail addresses: khuram.maqsood@gmail.com (K. Maqsood),
gangulysaibal2011@gmail.com (S. Ganguly).
Malaysian natural gas, the major issue is the high content of CO2 (up
to 80%) (Darman and Harum, 2006).
The technologies of CO2 separation mainly comprise of ab-
sorption, adsorption, membranes separation and cryogenic sepa-
ration (Shimekit and Mukhtar, 2012). Chemical solvents based
absorption is presently the most developed method used by in-
dustry. A comparative study of separation technologies for pro-
cessing carbon dioxide rich natural gas in ultra-deepwater oil elds
has been presented by Arajo et al. (2017). This work evaluates
carbon dioxide separation alternatives for the challenging scenario
in terms of technical, economic and environmental aspects. Reis
et al. (2017) presented a hybrid processing scheme of optimized
membrane skid with chemical absorption for the high content
carbon dioxide natural gas. However, these processes are yet to be

https://doi.org/10.1016/j.jclepro.2017.10.060
0959-6526/ 2017 Elsevier Ltd. All rights reserved.
796 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810
developed and tested to handle high CO2 concentration in natural
gas.
Cryogenic separation technology for CO2 capture has been
researched for several decades. However, for the past few years,
this technology was not extensively studied due to the high ex-
pected cooling cost. In the cryogenic separation the operating
temperature is usually very low which is needed to be optimized
very sensibly to minimize the energy requirements and to make it
good competitor with other separation technologies. In contrast,
cryogenic separation does not require absorbents, and solid or
liquid CO2 can be produced directly at high pressure. With these
advantages, the research on cryogenic CO2 capture technology has
made signicant progress.
A detailed cryogenic processes classication based on desu-
blimation and liquication has been presented in previous work
(Maqsood et al., 2014b). The cryogenic processes were classied as
conventional, non-conventional and hybrid cryogenic technologies.
Northrup and Valencia (Northrop and Valencia, 2009) developed
CFZ technology by modifying internal design of existing cryogenic
distillation columns. Feed streams containing CO2 from 8 to 71%
have been processed below the targeted 2% pipeline quality stan-
dard. However, presence of small amount of water can cause
plugging due to ice formation was the main disadvantage. In
addition, the accumulation of solid CO2 or ice on to heat exchanging
surfaces causes a decrease in the heat transfer rate and overall
process efciency. Additional molecular sieve dehydration unit
adds substantial cost and volume to the platform based operation.
Cryocell process was proposed and tested by Hart et al. (Hart and
Gnanendran, 2009). The operating conditions were maintained
such that the vapor phase becomes rich in CH4 whereas the liquid
phase becomes rich in CO2. The available literature unfortunately
does not provide any details on modeling, freezing kinetics or other
fundamental specications due to commercial reasons. A process
concept was developed on the basis of semi-cryogenic distillation
along with mechanical centrifugation by Willems et al. (2010).
Operation difculties were present in the work regarding the for-
mation of droplets of required sizes.
The non-conventional desublimation based packed bed was rst
studied by Tuinier et al. (2010). They developed a novel cryogenic
process using dynamically operated packed beds to capture CO2
and water from ue gases containing 10 vol % CO2 and 1 vol % H2O
(Tuinier et al., 2011). The limitations of this process are that it was
yet to be tested on natural gas at high CO2 concentrations and
pressure.
The cryogenic packed bed process concept for high pressure and
CO2 concentration was further experimentally validated by Abul-
hassan et al. (Ali et al., 2014a, b). It was concluded in this study that
cryogenic packed bed is promising for high CO2 content naturel gas
due to low energy requirements. Energy requirement of cryogenic
packed bed was reported to be signicantly lower than the cryo-
genic distillation process, as the former requires 660e810 kJ/kg CO2
energy while the later need 1472 kJ/kg CO2 to purify natural gas
with 69% of the CO2 content. However, both cryogenic packed bed
and cryogenic distillation process showed similar energy re-
quirements with a feed with a lower concentration of CO2. The
desublimation process had been proven to be efcient for high CO2
natural gas (Maqsood et al., 2014a). Furthermore, a detailed simu-
lation study with experimental validation was reported using 1-
dimensional pseudo homogeneous model (Ali et al., 2013) and
nucleation study (Ali et al., 2014a, b).
Recently, a novel concept of desublimation based multiple
packed bed pipeline network for purication of natural gas was
presented by Abulhassan et al. (Ali et al., 2016) for possible offshore
applications. A series of packed beds were used for simultaneous
dehydration and CO2 removal from natural gas. A novel strategy
using multiple equilibrium temperature (MET) concept was also
introduced for the simulation studies. The advantages of this
cryogenic process are clean, efcient pipeline based separation
with low capital cost. The detailed optimization study of multiple
packed bed pipeline network was not included in the previous
work.
Optimization is one of the most important engineering tools for
addressing the issues related to increased cost of energy, global
competition in product purity and increasingly stiff environmental
regulations (Edgar and Himmelblau, 1988). The maximum utiliza-
tion of the available resources for maximum protability is essen-
tial for successful industrials operations (Basak et al., 2002).
The present study attempts to bridge the gap in existing
research and literature on cryogenic packed beds by investigating
the optimal operating conditions of cryogenic packed bed pipeline
network to purify high CO2 content natural gas at high pressures.
The study also covers the separation of heavy hydrocarbon fractions
with minimum methane losses. The separation behavior of
methane, higher hydrocarbon and carbon dioxide at high pressure
by using multiple equilibrium model (MET) is presented in this
study during optimization.
2. Optimization study for cryogenic packed bed pipeline
network
The optimization study comprises of the problem formulation
and complete strategy to optimize the multiple packed bed pipeline
network.
2.1. General schematics and process concept of cryogenic packed
bed pipeline network
The basic separation principle is based on the difference of the
desublimation and freezing point of each component in the
mixture. The rst step in the purication of the feed gas is dehy-
dration, in which water contents were removed by using a series of
beds at different initial temperatures. The cryogenic packed beds
used for dehydration are the initial part of the complete schematics
presented in Fig. 1. The temperature range of the beds varies from
0
C to 30
C. The complete study on dehydration and CO2
removal using dual bed have been reported by Abulhassan et al. (Ali
et al., 2016).
The feed continuously passes through a series of dehydration
beds in order to achieve the desired pipeline specication i.e.
60e110 mg of water/MMSCF. The feed gas enters in rst dehydra-
tion bed at comparatively high initial bed temperature as compared
to the second or third bed. At very low initial bed temperature and
high pressure, the hydrocarbon losses and energy requirements
were found to be higher based on the simulation results. If the feed
gas through rst bed could not achieve the allowable water con-
tents, then the second dehydration bed is inevitable which has a
comparatively low initial bed temperature. For example, if the rst
bed is operating at 10
C then the second bed would operate
at 30
C or below. The feed gas pressure also drops in the rst bed,
so the gas entering in the second bed has a lower pressure. This
decrease in pressure also allows the operation of the next bed at a
lower temperature because of the shift of freezing point of the
natural gas components. The lower initial bed temperature allows
removing the maximum remaining quantity of the water contents.
After removal of the water contents, the dry gas enters in the next
bed where CO2 removes from the gas.
The beds used for CO2 removal operates on the same principle as
discussed earlier. The temperature range of CO2 removal varies
from 50
C to 110
C based on the feed pressure in order to
minimize hydrocarbon loss and maximize the separation. The
 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810 797

Fig. 1. General cryogenic packed bed pipeline network arrangement.

freezing point of CO2 at atmospheric pressure is 78
C which in- Table 1

crease further by increasing the pressure. The feed gas which comes Natural gas compositions.

from dehydration unit is at higher pressure, so in order to minimize Components Feed 1 Feed 2
the hydrocarbon loss, the pressure of feed gas need to be reduced Mole fraction Mole fraction
because at higher pressure the hydrocarbons liquefy and mix with
CH4 0.4400 0.2900
CO2.
C2H6 0.0500 0.0054
The multiple packed bed pipeline network is the addition of C3H8 0.0270 0.0040
several packed bed in order to purify the multi-component gas feed i-C4H10 0.0100 0.0010
(natural gas). The schematic of multiple packed bed pipeline n-C4H10 0.0100 0.0010
network is given in Fig. 1. PB-1 represents the rst packed bed i-C5H12 0.0100 0.0004
n-C5H12 0.0010 0.0002
which serves as a dehydration unit. At high pressure, the feed splits C6H14 0.0010 0.0001
into vapor and liquid which further goes to PB-2 and PB-3. These C7H16 0.0005 0.0001
packed bed can be used either as a dehydration unit or CO2 removal C8H18 0.0005 e
unit, which depends on the product specication. For example, if H2O 0.0400 e
CO2 0.4000 0.6900
the water contents are not in the desired range, the packed bed will
N2 0.0100 0.0078
be used as a dehydration unit, otherwise CO2 removal unit. The
natural gas compositions used for this study are presented in
Table 1.
Appendix A The schematics of the experimental setup used to methodology has been previously presented by Abulhassan et al.
validate the simulation study is illustrated in Fig. 2. The main (Ali et al., 2016) for multiple cryogenic packed bed network.
component of the experimental setup consists of a jacketed packed
bed, gas feeding system, and refrigerant supply system. The packed 2.2. Optimization strategy of packed bed network
beds were constructed with stainless steel and glass balls were
used as packing material. Several thermocouples were used to Optimization is the use of specic numerical techniques to
investigate the temperature changes during cooling and capture determine the most cost-effective and efcient solution to a
cycles. The sampling device was attached to the system to deter- problem. The systematic determination of optimal solutions leads
mine the percentage removal of CO2 with the help of gas chroma- to a large family of methods and algorithm. Optimization problems
tography analysis (SHIMADZU GC, 2010). The detailed experimental that arise in chemical engineering can be classied in terms of
798 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810
Fig. 2. Experimental setup for measurement of CO2 removal.
continuous and discrete variables. When represented in algebraic
form, the formulation of discrete/continuous optimization prob-
lems can be written as mixed integer optimization problems.
The optimization of the cryogenic process needs to be dealt with
great care because of the involvement of the extreme cryogenic
operating conditions. The optimizations strategy involves three
steps.
1 Development of Node-edge diagram
2 Optimization of operating conditions of each node locally using
Golden search method
3 Optimization of overall network
2.2.1. Development of Node-edge diagram
The cryogenic packed bed network is represented as the node-
edge diagram in Fig. 3. Packed bed 1, as shown in the node-edge
diagram, serves as a root node of the solution tree. As the path
going down, two terminal nodes are created for each node through
the branching steps.
The numbers are shown on the node-edge diagram symbolizes
the packed bed number, however with a few exceptions. For
example, digit 0 refers to raw natural gas storage tank number 16
symbolizes water storage tank, number 17 is being referred as CO2
storage tank, number 18 is denoted as a storage tank for methane
with high purity, storage tank number 19 is used to store methane
with small amount of CO2 and number 20 symbolizes storage tank
for methane with low purity.
In the node-edge diagram, it can also be observed that natural
gas feed coming from the feed storage tank is introduced into the
packed bed 1, where separation takes place and the solid ice is
recovered and sent to the water storage tank numbered 16. In this
case packed bed 1, is considered as the dehydration bed. Mean-
while, the heavy hydrocarbons that have condensed into the liquid
state will be sent to the packed bed 3, while vapor product of
packed bed 1 will undergo another stage of cryogenic separation at
packed bed 2. If the complete water removal does not take place in
packed bed 1, dehydration process will again be carried out in
packed bed 2.
However, if water is completely removed from packed bed 1, the
next packed beds will be considered as CO2 removal and recovered
CO2 will be sent to CO2 storage tank number 7. The vapor and liquid
product will then become feed for the next two packed beds. The
highly pure methane product will be sent to a storage tank
numbered 18 whereas the impure methane product will be sent to
storage tank numbered 20.
2.2.2. Optimization of operating conditions of single node locally
using golden section search method
The rst step in the optimization of the node-edge diagram is to
nd out the optimal condition of each node for maximum separa-
tion at minimum energy cost and minimum hydrocarbon losses.
For this purpose, golden section search method was used ((Moler,
2011). The method works as:
 By dening an interval with a single answer inside the given
range
 Dividing interval into 3 sections by adding two internal points
between ends using golden ratio
 Evaluating the function at the two internal points X1 and X2

if f(X1) > f(X2) the minimum is between X1 and Xmx, redene
range Xmn X1, Xmx Xmx

if f(x1) < f(x2) the minimum is between x2 and xmn, redene
range xmn xmn, xmx x2
2.2.3. Depth-rst strategy
The construction of the node-edge diagram that represents the
multiple packed bed network was performed using the depth-rst
search branching strategy. Depth-rst search is an algorithm used
to perform branching on the most recently created node, where it
goes through a vertical branching until the terminal node is
reached. The operating parameter of the initial node is optimized
using the Golden Section Search technique. Upon obtaining the
optimal composition of the feed product stream at their respective
optimal condition, the initial node was expanded into two main
branches namely vapor, and liquid.
As mentioned earlier, throughout this node-edge diagram
 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810
Fig. 3. Node-edge diagram of multiple packed beds for dehydration and CO2 removal from natural gas.
synthesis, optimization was performed on the vapor and liquid
stream that was channeled to the next node. Therefore, two sub-
nodes were generated from both these product stream. One of
this sub-node is chosen and optimization was performed. This node
was further expanded. The nodes in between the initial and ter-
minal node are known as the intermediate nodes.
When there were no nodes to be expanded, then backtrack to
the base node and if it does not split further into sub-node the
search was terminated. This process repeats until a terminal node
was reached. Every path from the initial node to the terminal nodes
dene a complete solution.
After completing the 1st iteration, the prot objective function
was computed. The parameters and steps involved to perform the
prot objective function have been described earlier. Further iter-
ation on the synthesis of multiple packed beds needs to be
continued in order to perform a comparative analysis of the prot
objective values.
In continuation to perform the 2nd iteration, the terminal nodes
with their respective product stream into product tank 19 and 20
were analyzed on their suitability for further optimization. Product
tank 19 holds methane (CH4), and carbon dioxide (CO2) which
could be recovered and channeled to product tank 18 and 17
respectively. Meanwhile, in product tank 20, there was a high
amount of hydrocarbon losses which were further reduced.
The optimization of the terminal node as a single node was
executed using golden section search. It is noteworthy that, certain
terminal nodes may not have their sub-nodes due to the effect of
the operating parameters on the compositions and energy
799
required. In each iteration, only one terminal node was chosen for
optimization. When there were no further terminal nodes to be
explored, comparative study on the prot objective value from each
analysis was carried out.
Finally, the most optimal scheme was chosen based on the
highest prot value achieved using depth-rst strategy. The
advantage of depth-rst search method is that it has the tendency
of nding the optimal solution in the enumeration procedure. The
ow chart for optimization of complete pipeline network is given in
Fig. 4.
3. Optimization methodology for cryogenic multiple bed
network
A detailed research was conducted on the synthesis and
formulation of a multiple cryogenic packed bed network to purify
natural gas. Optimization based on golden section local search and
depth-rst strategy has been employed. Initially, the problem
synthesis and formulation were performed, then the Node-edge
diagram was constructed. The solution of each node of Node-edge
diagram represent the local optima which was obtained using
golden section search while overall solution of Node-edge diagram
represents the globe optima which was obtained using depth-rst
strategy. The simulation results were obtained for sensitivity study.
The optimal scheme was then obtained for the multiple cryogenic
packed beds for products with minimum energy utilization and
minimum hydrocarbon losses.
800 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810

Fig. 4. Algorithm for single node (Golden section) and whole network (Depth-rst strategy).
 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810 801

3.1. Problem formulation

By choosing a range of different suitable pressure-temperature
combinations obtained from simulation and the process analysis
table for each packed bed, many schemes can be formed. In order to
determine which scheme provides the optimum pressure-
temperature combination for the packed beds, an objective func-
tion that serves as a comparison value was required. The detail of
the formulation of the objective function is detailed in the next
section.
3.2. Objective function for single node optimal operating conditions
Each node in the node-edge diagram represents a packed bed
which was optimized for maximum separation with minimum
energy and hydrocarbon loss. The optimization variables are pres-
sure and temperature. The objective function can be represented as
a cost of the undesired component in the nal product.
Minimize j P1 P2  P3
j 14 and i 111
j is the nal storage facility for high purity CH4, low purity CH4,
natural gas liquids and CO2 products respectively, whereas, i is the
component number for CH4, C2H6, C3H8, C4H10, iC4H10, C5H12,
iC5H12, C6H14, C7H16, C8H18, CO2 respectively.
 Model equation used for single equilibrium stage
The model equations used for single equilibrium stage are given
below
X zi Ki  1
0 (10)
i
1 bKi  1
where:
zi is the mole fraction of component i in the liquid feed
b is the fraction of feed that is vaporized
(1) Ki is the equilibrium constant of component i

Where yi
Ki (11)

 xi
$
P1 cost of undesired component CO2 in the final product
h where:
(2)

 xi is the mole fraction of component i in the liquid phase;
$
P2 cost of energy for separation (3) yi is the mole fraction of component i in the gaseous phase.
h
Once the equation has been solved for b, the compositions xi and

 yi can be immediately calculated as:
$
P3 cost of desired component like methanein final product
h zi
xi (12)
(4) 1 bKi  1

yi Ki xi (13)
3.3. Prot objective for packed bed network
One additional constraint which is related to high-quality
product is
The objective function for the overall prot of cryogenic
network in its nal is represented in Equation (5). Producthigh purity CH4  0:9 (14)
Maximize F C1  C2  C3 (5)

 4. Single node optimization results using golden section
where F $
cycle
is the prot based objective function and the time search method
of one cycle has been taken 10 years.
The product drawn rate was taken as (kg/h) while the price of The rst step in the optimization was to develop node-edge
the product was taken as $/kg. diagram and once the diagram was developed, the optimization
was carried out for each node. During the optimization of each


$ node, the effect of temperature and pressure on separation and
C1 Product revenue per cycle (6) energy requirement was investigated. The feed at 80 bar and 25
C
Cycle
entered in the rst bed where most of the associated water was

 removed. The temperature range of the bed was selected from 0
C
$
C2 Cost of energy required per cycle (7) to 30
C. Below 30
C, the whole feed convert into liquid.
Cycle

 4.1. Optimal operating conditions for Node-1
$
C3 Equipment cost (8)
Cycle
The optimum conditions for Node-1 are explored in this section.
The maximization of the prot objective function is required, The simulation data for all possible conditions was generated
the detail of product selling price is given in Equation (9). (sample data is given in Supplementary Information). The simula-

 XX

 tion results of Node-1 for vapor phase are presented in Table 2. It is
$ kg $ clear from the results that by decreasing the temperature the
C1 Prouductj  Pricei (9)
Cycle j i
cycle kg amount of water removed and energy requirements increases. On
the other hand, the amount of methane in the vapor phase is
decreased by decreasing bed temperature. The main objective of
802 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810

Table 2
Vapor phase simulation results at 80 bar for Node-1.

Feed Conditions Product mass ow in vapor phase (kg/h)

P(bar) T(
C) Components Mass Flow (kg/h) 5
C 0
C 5
C 10
C 15
C 20
C 30
C

80 25 CH4 7058.876 6691.255 6190.972 5331.38 4058.43 2337.28 31.76 0.000

C2H6 1503.495 1351.082 1182.957 940.61 646.87 331.73 3.91 0.000
C3H8 1190.619 999.952 827.772 614.86 393.51 188.77 2.20 0.000
i-C4H10 581.240 454.847 358.807 253.38 154.81 71.55 0.58 0.000
n-C4H10 581.240 438.819 338.840 234.38 140.68 64.15 0.58 0.000
i-C5H12 721.510 489.983 358.747 236.99 137.24 61.07 0.72 0.000
n-C5H12 72.151 46.896 33.690 21.92 12.54 5.54 0.00 0.000
C6H14 86.178 46.134 31.096 19.35 10.75 4.66 0.00 0.000
C7H16 50.103 20.945 13.382 8.07 4.40 1.89 0.00 0.000
C8H18 57.116 17.669 10.869 6.44 3.48 1.49 0.00 0.000
H2O 720.604 5.930 4.069 2.50 1.31 0.50 0.00 0.000
CO2 17603.880 16035.2 14146.6 11301.87 7783.83 3988.06 47.53 0.000
N2 280.130 270.554 255.892 228.32 183.05 113.15 1.68 0.000
Total (kg/h) 30507.14 26869.278 23753.781 19200.07 13530.9 7169.84 88.97 0.000
Energy kJ/h 2029384.8 2746901.3 3547614.2 4396604.7 5259024.8 6116358.2 7324914
MW 0.5637 0.7630 0.9854 1.2213 1.4608 1.6990 2.0347

the Node-1 is the dehydration of the feed gas so that the resultant
stream should contain a minimum amount of water. The other
objectives are minimum energy requirements and the maximum
amount of methane in the vapor stream.
In order to run the algorithm for Node-1, the rst step was to get
the raw simulation data and the second step was to obtain the cost
curves for each quantity which were needed to be optimized. It can
be observed from the cost curves shown in Fig. 5 that, by decreasing
the temperature not only the cost of the unwanted component in
the product (water in the case of Node-1) is decreased, but the cost
of methane also decreases. At low temperature more amount of
feed converts into the liquid phase resulting in the methane loss
which should be avoided.
The third step was to run the iterative procedure using golden
section algorithm for optimum bed temperature. The results for the
optimum conditions are presented in Table 3 and Fig. 6. It is shown
from the results that the optimum bed temperature for minimum
water and maximum methane in the product is 4.7
C.
4.2. Optimal operating conditions for Node-2
The vapor product for Node-1 was taken as the feed for the
Node-2. The feed condition for this node was 80 bar and 4.7
C.
The Node-2 can be used either for dehydration or CO2 removal
which depends on the specication of the product stream for
Node-1. If the product stream from Node-1 meets the pipeline
specication of natural gas (65e110 mg/m3 water contents) then
the Node-2 will serve as CO2 removal unit otherwise dehydration
unit.
The simulation results for Node-2 are presented in Table 4. The
objective function for Node-2 was still same as Node-1, so it was
used for dehydration of the natural gas. Hence, most of the water
was removed in Node-1 so in order to remove the remaining
water, this packed bed should be operated at a lower temperature
than the previous bed. The simulation results show that at the
lower temperature, most of the associated water was removed in
this bed.
The resultant curves from simulation data are presented in

Fig. 5. Effect of temperature on the cost of important targets Node-1.

 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810 803

Table 3
Optimum temperature for Node-1.

Iteration X1 P1 ($/h) P3 ($/h) P2 ($/h) FX1 X2 P1 ($/h) P3 ($/h) P2 ($/h) FX2

1 11.4 302.75 790.47 964.42 476.69 18.5 32.26 166.96 1222.35 1087.65
2 7.08 572.28 1076.80 812.06 307.54 11.4 302.76 790.50 964.41 476.68
3 4.38 777.96 1215.92 720.60 282.63 7.08 572.29 1076.81 812.05 307.54
4 2.71 920.01 1287.45 665.10 297.66 4.38 777.97 1215.92 720.59 282.63
5 4.38 777.96 1215.92 720.59 282.63 5.41 695.88 1166.20 755.29 284.97
6 3.74 830.87 1244.55 699.30 285.63 4.38 777.96 1215.92 720.59 282.63
7 4.38 777.96 1215.92 720.59 282.63 4.77 746.10 1197.43 733.81 282.48
8 4.77 746.10 1197.43 733.81 282.48 5.02 726.72 1185.69 742.00 283.03
9 4.62 758.19 1204.56 728.76 282.39 4.77 746.10 1197.43 733.81 282.48
10 4.53 765.72 1208.93 725.64 282.42 4.62 758.19 1204.56 728.76 282.39
11 4.62 758.19 1204.56 728.76 282.39 4.68 753.56 1201.85 730.69 282.40
12 4.68 753.56 1201.85 730.69 282.40 4.71 750.71 1200.16 731.88 282.42

Fig. 6. Optimum temperature for Node-1.

Table 4
Simulation results at 80 bar for Node-2.

Feed Conditions Product mass ow in vapor phase (kg/h)

P(bar) T(
C) Components Mass Flow(kg/h) 10 15 20 25 30

80 5 CH4 5331.38 5012.71 4563.99 3933.41 3132.38 2104.19

C2H6 940.61 816.64 675.15 514.54 356.93 206.58
C3H8 614.86 477.85 355.26 242.72 152.58 81.58
i-C4H10 253.38 174.20 118.09 74.60 44.17 22.67
n-C4H10 234.38 151.22 98.39 60.20 34.87 17.44
i-C5H12 236.99 126.64 74.85 42.94 23.81 11.54
n-C5H12 21.92 10.82 6.18 3.47 2.16 0.72
C6H14 19.35 6.81 3.51 1.86 0.86 0.86
C7H16 8.07 1.87 0.89 0.46 0 0
C8H18 6.44 0.92 0.42 0.21 0 0
H2O 2.50 1.42 0.86 0.48 0.18 0.18
CO2 11301.87 9934.51 8223.40 6234.54 4293.59 2477.31
N2 228.32 219.85 206.71 186.62 158.27 115.41
Total (kg/h) 19200.07 16935.46 14327.71 11296.04 8199.80 5038.50
Energy kJ/h 309.0 559424.91 1161714 1766737 2362528
MW 0.0001 0.1554 0.3227 0.4908 0.6563

Fig. 7 and the optimal temperature for Node-2 is illustrated in Fig. 9. These results show that for Node-2 the optimal temperature
Fig. 8 and Table 5. The optimal temperature for Node-2 was and pressure were 20
C and 70 bar.
observed around 20
C. The second step was to nd the optimal A similar strategy was used to optimize each node in the node-
pressure of the bed, the results for optimal pressure are given in edge diagram for optimal conditions of the complete network.
804 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810

Fig. 7. Effect of temperature on the cost of decisive parameter at Node-2.

Fig. 8. Optimum temperature for Node-2.

Table 5
Optimum temperature for Node-2.

Iteration X1 P1 ($/h) P3 ($/h) P2 ($/h) FX1 X2 P1 ($/h) P3 ($/h) P2 ($/h) FX2

1 20.73 141.25 235.18 104.30 10.38 24.27 83.36 199.61 139.80 23.55
2 18.54 187.05 254.11 82.29 15.24 20.73 141.26 235.18 104.30 10.38
3 20.73 141.25 235.18 104.30 10.38 22.08 116.77 222.31 117.87 12.34
4 19.89 157.84 242.68 95.90 11.06 20.73 141.25 235.18 104.30 10.38
5 20.73 141.25 235.18 104.30 10.38 21.25 131.56 230.37 109.49 10.68
6 20.41 147.46 238.08 101.09 10.47 20.73 141.25 235.18 104.30 10.38
7 20.73 141.25 235.18 104.30 10.38 20.93 137.50 233.36 106.28 10.43
 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810 805

Fig. 9. Optimum pressure for Node-2.

4.3. Results of golden section on each node (base case)
The rst node-edge diagram illustrated by Fig. 10 was ob-
tained from golden section search method. The results from
Fig. 10 shows that the required purication of methane product
in storage tank 18 is not according to desired purity range. The
methane losses at this temperature are also higher, which is 39%
(product tank 20).
When associated water has been separated from the natural gas
in solid form, the vapor phase product is directed into the packed
bed 2 while the water-free condensed liquid phase product is
channeled into the packed bed 3. Due to some small amount of
water vapor remains in the vapor phase stream in packed bed 1, the
natural gas undergoes dehydration process again in packed bed 2.
Meanwhile, the water-free liquid product that has been directed
into packed bed 3. As the operating pressure of packed bed 3

Fig. 10. The results of optimization in node-edge diagram starting from feed at 80 bar and 25
C.
806 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810
reduces, some of the liquid feed vaporizes and this phenomenon
separates some of the heavy hydrocarbons present in the feed. Both
the vapor and liquid product of packed bed 3 are then fed into
packed beds 6 and 7 respectively, to undergo CO2 removal.
The third level packed beds, namely packed bed, 4, 5, 6 and 7 are
for both CO2 removal and CH4 purication. The other nodes like 8, 9,
10, 11, 13, 14 and 15 are not taking part in the separation because the
maximum separation is already achieved.
4.4. Optimization of multiple packed bed network using depth-rst
strategy
The second step in the optimization of multiple packed bed
network was done by using depth-rst strategy. In this strategy, the
whole network was optimized for prot objective, whereas, the
selected variables were product purity, energy and equipment cost.
The optimal condition for minimum methane losses, maximum
separation and minimum energy requirements is explored in this
section. The temperature and pressure conditions obtained from
golden section search provided the starting point to optimize the
full network. The depth rst strategy result with keeping pressure
decision variable is given in Fig. 11.
Each node in the Node-edge diagram was also subject to the
sensitivity analysis. The sensitivity analysis shows that the
maximum separation is achieved with minimum energy require-
ment when packed bed 1 is operated at 70 bar and 10
C (Fig. 12).
Next, by setting the operation conditions for packed bed 1 to be
at 70 bar and 10
C, the sensitivity analysis is conducted on packed
bed 2, which further dehydrates the feed gas in order to meet the
pipeline specications. As shown in Figs. 13 and 14, keeping the
same pressure and temperature combinations for all packed beds
except packed bed 2, the effect on separation was signicant. It was
Fig. 11. Depth rst strategy with pressure decision variable (Y&N).
observed that the change in operating temperature affects the
purity as well as the amount of the products being produced. For
instance, increasing the operating temperature from 60
C
to 30
C not only increases the purity of methane by 1% but also
increases the amount of high purity methane. This increment in
terms of purity and ow rate is desirable as the main objective of
this separation process is to have maximum separation with min-
imum methane losses.
After the complete removal of moisture contents from the nat-
ural gas, the next packed beds were used to remove CO2. In order to
achieve the separation, the bed temperature must be below the
desublimation temperature of CO2.
Similar to the sensitivity analysis that has been done on level 1
and level 2 of the node-edge diagram, by having packed bed 1 and
packed bed 2 operated at optimal conditions, 70 bar and 10
C
and 40 bar and 10
C respectively, different pressure and tem-
perature combinations of the third level packed beds, namely
packed bed 4, 5, 6 and 7 were investigated. As the purity of nal
methane gas is an important objective of this study, (packed bed 4
and packed bed 5) where the methane content is higher were
investigated primarily. The simulation results show that the
optimal temperature for packed bed 4 was 90
C for maximum
methane recovery (Fig. 15).
The analysis of NS-5 shows the most optimal condition for
dehydration and CO2 removal from natural gas. It can be observed
from Fig. 16 that by changing the condition in node 9 and node 3,
the overall prot objective is greatly increased as compared to NS-
4. This further proves that the sensitivity of each node is greatly
affected by changing the operating conditions. The comparative
analysis for all schemes is given in Table 6. It is evident from Table 6
that the overall prot in NS-5 is higher than the other schemes and
the purity criteria for the primary product is also met.
A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810 807

Fig. 12. Node-edge diagram (NS-1).

Fig. 13. Node-edge diagram (NS-2).

808 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810

Fig. 14. Node-edge diagram (NS-3).

Fig. 15. Node-edge diagram (NS-4).

 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810 809

Fig. 16. Node-edge diagram (NS-5).

Table 6
Comparison of different node-edge schemes.

Scheme number NS-1 NS-2 NS-3 NS-4 NS-5

Total product Cost CO2 Storage Mass ow(kg/h) 17534 17534 17515 17525 17534
Cost($/h) 701.36 701.36 700.60 701.00 701.36
CH4 Storage Mass ow(kg/h) 5877 5361 5020 5303 5877
Cost($/h) 1160.33 1160.33 1074.32 1117.55 1242.11
CO2, CH4 and H h/c Storage Mass ow(kg/h) 4978 5494.5 5323.00 5228.50 4978.5
Cost($/h) 593.48 593.48 570.38 557.46 572.50
CO2, and H h/c Storage Mass ow(kg/h) 4978 5494.5 5323.00 5228.50 4978.5
Cost($/h) 593.48 593.48 570.38 557.46 572.50
C1, Product Cost ($/cycle) 233420000 233420000 233320000 232730000 238670000
Energy (MW) 3 3.18 3.09 3.07 3.09
C2, Energy Cost ($/cycle) 28550000 30240000 29390000 29170000 29360000
C3, Equipment ($/cycle) 498000 1090000 955000 976000 1390000
F, Prots ($/cycle) 204373204 20208818 202983828 202591191 207930738

4.5. Effect of feed composition on energy requirement
The CO2 concentration in the feed has a signicant effect on both
separation and energy requirements. The study was carried out
using two different feed composition as provided in Table 1. For
both concentrations, the optimal bed condition found was
different. The comparison of each feed conditions is given in Table 7.
The amount of CO2 removed in 40% and 69% CO2 feed gas was
17534 kg/h and 30324 kg/h respectively, and the respective energy
requirements were 3.09 and 3.41 MW. Although, the energy re-
quirements in 69% is higher, however, a huge amount of CO2 is
being removed and the energy requirements per kg of CO2 is lower
than the 40%. The prot ($/cycle) in the 40% feed gas is higher as
compared to the 69% feed gas. This higher prot is due to the
presence of high CH4 amount in the feed gas, which ultimately
gives the higher revenue. These results show that in high CO2 feed
gas concentrations the energy requirements per kg of CO2 removal
are low. The number of packed bed required for 69% feed gas are
less which further decreases the equipment cost. The lesser num-
ber of packed bed required for high CO2 concentration resulted in
low footprints which show that proposed process is suitable for
offshore applications. The effect of CO2 concentration on energy
requirements at atmospheric pressure also showed the same re-
sults in previous work with the experimental ndings (Maqsood
et al., 2014a). For example, at atmospheric pressure, the energy
requirement for 13% and 100% was found 1800 kJ and 680 kJ
respectively.
5. Simulation and experimental results comparison for
multiple packed bed pipeline network
The experimentation was conducted using binary CH4eCO2 gas
mixture and compared with the simulation results. In previous
study, the simulation results for natural gas and the binary
810 A. Ali et al. / Journal of Cleaner Production 171 (2018) 795e810

Table 7
Comparison of different feed composition.

CO2 concentration in feed 69% 40%

Total product cost CO2 Storage Mass ow(kg/h) 30324 17534

Cost($/h) 1212.96 701.36
CH4 Storage Mass ow(kg/h) 37877 5877
Cost($/h) 554.09 1242.11
CO2, CH4 and H h/c Storage Mass ow(kg/h) 3294.2 4978.5
Cost($/h) 399.05 572.50
F1 ; F2 ; F3 , Product Cost ($/cycle) 198049000 238674000
Energy (MW) 3.41 3.09
F4 , Energy Cost ($/cycle) 32410000 29360000
F6 , Equipment ($/cycle) 1220000 1390000
Prots ($/cycle) 164418459 207930738

Table 8 Appendix A. Supplementary data

Comparison between experimental and simulation results.


Node # Temperature ( C) Pressure (bar) Percentage Separation (%) Supplementary data related to this article can be found at
Experimental Simulation
https://doi.org/10.1016/j.jclepro.2017.10.060.

3 80 20 98.3 99.5

4 90 25 97.9
6 100 20 71.1
7 90 10 78.5
8 100 25 96.6
CH4eCO2 gas mixture was compared and it was observed that the
exit vapor phase composition of CO2 in the product is similar in
both cases (Ali et al., 2016). The comparison between experimental
and simulation results were based on the exit product composition
which eventually gave the percentage separation. The experimen-
tation was conducted on some selected nodes due to the limitation
of experimental setup. The experimental and simulation results
comparison is given in Table 8 which shows promising results.
6. Conclusion
The present study attempts to optimize a multiple cryogenic
desublimation based pipeline network for removal of water and
CO2 from natural gas. Optimization studies were conducted using
Golden section search and Branch & Bound strategy. The results
were promising for natural gas applications at high pressure.
Cryogenic packed bed network was able to handle the full range of
natural gas at high pressure with lower energy consumption
without any chemical addition. Initially the network was optimized
using golden search method which provided the base case. The
product purity in that case was calculated 77% of CH4 which is not
compatible with pipeline specication of natural gas, further opti-
mization using depth rst strategy resulted the product purity up to
94% with reduction of hydrocarbon losses from 39 to 16%. The
cryogenic packed bed pipeline network showed promising poten-
tial for future commercial exploitation through minimization of
energy, and minimum hydrocarbon losses for purication of nat-
ural gas.
Acknowledgment
The authors gratefully acknowledge the nancial and infra-
structural help from Department of Chemical Engineering, Uni-
versiti Teknologi PETRONAS.
99.3 References
73.6
79.4 Ali, A., Ganguly, S., Shariff, A.B.M., 2013. Simulation of Cryogenic Packed Bed Using
98.6 1-dimensional Pseudo Homogeneous Model. International Oil and Gas Sym-
posium and Exhibition (IOGSE 2013), Sabah, Malaysia.
Ali, A., Maqsood, K., Redza, A., Hii, K., Shariff, A.B., Ganguly, S., 2016. Performance
enhancement using multiple cryogenic desublimation based pipeline network
during dehydration and carbon capture from natural gas. Chem. Eng. Res. Des.
109, 519e531.
Ali, A., Maqsood, K., Syahera, N., Shariff, A.B.M., Ganguly, S., 2014a. Energy mini-
mization in cryogenic packed beds during purication of natural gas with high
CO2 content. Chem. Eng. Technol. 37 (10), 1675e1685.
Ali, A.H., Ashraf, A.M., Shariff, A.B.M., Ganguly, S., 2014b. Study of simultaneous
mass transfer and nucleation in cryogenic packed bed for the removal of CO2
from natural gas. Appl. Mech. Mater. Trans Tech Publ 229e232.
Arajo, O.d.Q.F., Reis, A.d.C., de Medeiros, J.L., Nascimento, J.F.d., Grava, W.M.,
Musse, A.P.S., 2017. Comparative analysis of separation technologies for pro-
cessing carbon dioxide rich natural gas in ultra-deepwater oil elds. J. Clean.
Prod. 155 (Part 1), 12e22.
Basak, K., Abhilash, K., Ganguly, S., Saraf, D., 2002. On-line optimization of a crude
distillation unit with constraints on product properties. Ind. Eng. Chem. Res. 41
(6), 1557e1568.
Darman, N.H., Harum, A.R.B., 2006. Technical Challenges and Solutions on Natural
Gas Development in Malaysia, the Petroleum Policy and Management Project,
4th Workshop of the China-Sichuan Basin Study (Beijing, China).
Edgar, T.F., Himmelblau, D.M., 1988. Optimization of Chemical Processes. McGraw-
Hill, New York.
Hart, A., Gnanendran, N., 2009. Cryogenic CO2 capture in natural gas. Energy Pro-
cedia 1 (1), 697e706.
Maqsood, K., Ali, A., Shariff, A.B., Ganguly, S., 2014a. Synthesis of conventional and
hybrid cryogenic distillation sequence for purification of natural gas. J. Appl. Sci.
14 (21), 2722e2729.
Maqsood, K., Mullick, A., Ali, A., Kargupta, K., Ganguly, S., 2014b. Cryogenic carbon
dioxide separation from natural gas: a review based on conventional and novel
emerging technologies. Rev. Chem. Eng. 30 (5), 453e477.
Moler, C.B., 2011. Numerical Computing with MATLAB. Society for Industrial &
Applied Mathematics, Philadelphia, Pa.
Northrop, P.S., Valencia, J.A., 2009. The CFZ process: a cryogenic method for
handling high- CO2 and H2S gas reserves and facilitating geosequestration of
CO2 and acid gases. Energy Procedia 1 (1), 171e177.
Reis, A.d.C., de Medeiros, J.L., Nunes, G.C., Arajo, O.d.Q.F., 2017. Upgrading of natural
gas ultra-rich in carbon dioxide: optimal arrangement of membrane skids and
polishing with chemical absorption. J. Clean. Prod. 165 (Suppl. C), 1013e1024.
Shimekit, B., Mukhtar, H., 2012. Natural gas purication technologies - major ad-
vances for CO2 separation and future directions. In: Al-Megren, H. (Ed.), Ad-
vances in Natural Gas Technology. InTech, Rijeka, Croatia.
Tuinier, M., van Sint Annaland, M., Kuipers, J., 2011. A novel process for cryogenic
CO2 capture using dynamically operated packed bedsdan experimental and
numerical study. Int. J. Greenh. Gas. Control 5 (4), 694e701.
Tuinier, M.J., van Sint Annaland, M., Kramer, G.J., Kuipers, J.A.M., 2010. Cryogenic
CO2 capture using dynamically operated packed beds. Chem. Eng. Sci. 65 (1),
114e119.
Willems, G.P., Golombok, M., Tesselaar, G., Brouwers, J.J.H., 2010. Condensed rota-
tional separation of CO2 from natural gas. AlChE J. 56 (1), 150e159.