Journal of Natural Gas Science and Engineering 26 (2015) 470e479

Contents lists available at ScienceDirect

Journal of Natural Gas Science and Engineering

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

APCI- LNG single mixed refrigerant process for natural gas liquefaction
cycle: Analysis and optimization
Peyman Moein a, Mehran Sarmad a, Hadi Ebrahimi a, *, Marziyeh Zare b,
Saeed Pakseresht b, Shahrokh Zandieh Vakili b
a
Gas Research Division, Research Institute of Petroleum Industry (RIPI), Tehran 14665-1998, Iran
b
Research & Technology Directorate of National Iranian Gas Company, Iran

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

Article history: In this paper, a typical single mixed refrigerant (SMR) with low energy consumption is analyzed to
Received 8 April 2015 determine the optimum operating conditions. Genetic Algorithm (GA) tool is used to minimize the total
Received in revised form required work by optimizing eleven variables including the outlet pressures of all compressors and
21 June 2015
throttling valves and the molar flow rate of the MR components. The results of the process simulation
Accepted 22 June 2015
Available online xxx
show that the total work is varied linearly by the compositions. It is observed that to have a minimum
work, when the ambient temperature increases, the compositions of both methane and ethane should be
decreased, while those of nitrogen, propane and n-butane should be increased. Moreover, optimized
Keywords:
LNG
values of the total required work at different ambient temperatures were correlated with a second order
Single mixed refrigerant polynomial. Increasing feed pressure has small effect on the MR compositions, but significant effect on
Optimization the total work.
Genetic algorithm © 2015 Elsevier B.V. All rights reserved.
Ambient temperature
Feed pressure

1. Introduction capital and operational costs or between hot and cold composite
curves (Hatcher et al., 2012). N2 expander processes have several
The demand for Natural Gas (NG) as a relatively clean fossil fuel advantages compared to MR ones, including simplicity, less number
has increased in the past years. Although research and develop- of equipments, no refrigerant phase change, non-flammability and
ment on other fossil fuels mainly oil, coal and their derivatives are non-toxicity of N2. However, higher energy consumption per pro-
going on (Zhang et al., 2014), NG produces approximately less than duced LNG tones particularly for large capacity plants is well-
half the CO2 emissions per unit of generated electricity. Two main known for N2 expander processes. The cascade cycles contain
ways to transport NG from resources to consumers are pipeline and some refrigeration stages that each stage has its own refrigerant
Liquefied Natural Gas (LNG). The pipeline is most commonly used and independent compression system.
due to its lower operating cost; however it is inflexible in the case of NG essentially contains mixed hydrocarbons and so its
transmission capacity, and requires enormous initial construction enthalpy with temperature is nonlinearly varied during the
cost. LNG is a liquid form of natural gas with a temperature liquefaction. Whenever the temperature difference between hot
about
162  C. It has a volume ~1/600 of that of gas at ambient and cold streams in the heat exchangers reaches below a certain
temperature. Although production and transmission of LNG require value, the entropy generation reduces, but the efficiency in-
complex machinery and vehicles, the incremental cost of transport creases. This temperature difference in MR cycles could be
per mile is less than that of pipelines (Cranmore and Stanton, reduced effectively without need for more auxiliary equipment
2000). There are several processes for LNG production such as N2 (Venkatarathnam, 2008). Although the pure refrigerant cycles are
expander (Yuan et al., 2014), cascade and Mixed Refrigerant (MR) simple, a large number of refrigeration stages are required in or-
processes. The difference among these processes is related to their der to reduce the temperature difference (Andress and Watkins,
2004).
Single Mixed Refrigerant (SMR) processes are the simplest MR
cycles for liquefaction of natural gas. Compared to mixed-
* Corresponding author.
refrigerant cycles (MRC) that have been applied for both large
E-mail address: [email protected] (H. Ebrahimi).

http://dx.doi.org/10.1016/j.jngse.2015.06.040
1875-5100/© 2015 Elsevier B.V. All rights reserved.
 P. Moein et al. / Journal of Natural Gas Science and Engineering 26 (2015) 470e479 471

and small-scale systems (Cao et al., 2006), SMR systems have only Because of the nature of nonlinearity and thermodynamic
one series of compressors beside the main cryogenic heat ex- complexity of the LNG processes, optimization of these processes
changers to cover all range of the liquefaction temperatures should be performed by global search methods such as GA, Tabu
(Swenson, 1977). Nevertheless, these MR processes are highly Search (TS), or Simulated Annealing (SA). Due to the fact that GA
energy intensive, so an optimization procedure is required to does not require derivatives and initial points, it was chosen as the
minimize energy consumption. Several MR process optimization optimization method for the current research.
methods are available in the open literature. Lee et al. (2004) In this work a typical APCI (Air Products and Chemicals, Inc.)
applied a graphical targeting method to optimize a multi-stage process (Roberts et al., 2002) as a SMR liquefaction process for LNG
MR system by considering minimization of both crossovers and production was chosen because of its improved efficiency
shaft work requirement. Gong et al. (2000) optimized the PRICO compared to the other SMR processes. Since special design of this
process by BOX method. Energy and exergy analysis of SMR, a LNG production process, it has low energy consumption. Therefore
two-stage expander nitrogen refrigerant and two open-loop this process could be a suitable replacement with current LNG
expander processes have been evaluated by Remeljej and production processes due to simplicity, high efficiency and mini-
Hoadley (2006) for LNG production with small-scale LNG. mum number of existing heat exchangers. The improvements
Aspelund et al. (2010) used combination of Tabu Search (TS) and include the compression of vaporized refrigerant at reduced
Nelder-Mead Downhill Simplex (NMDS) method linked to Aspen- compressor inlet temperatures and the generation of inter-stage
HYSYS simulation software to optimize the PRICO process. liquid refrigerant streams at ambient temperature which could be
Mokarizadeh and Mowla (2010) used coded PengeRobinson used beneficially in the refrigeration cycle. On the other hand this
equation of state and genetic algorithm method to optimize the APCI process can achieve high thermodynamic efficiency by using a
PRICO process. Alabdulkarem et al. (2011) applied genetic algo- minimum number of heat exchangers to reduce both capital and
rithm (GA) for optimization of propane pre-cooled MR (C3MR) operating costs (Roberts et al., 2002). The APCI has not been already
and reported MR power consumption at different pinch temper- investigated with the current GA method.
atures. Mortazavi et al. (2012) enhanced the performance of C3MR Although optimization of several MR components in some
process by implementing two phase expanders, liquid turbines special mixed refrigerant processes were investigated in previous
and gas expanders instead of conventional expansion devices. Xu studies, but the current APCI process has not been analyzed and
et al. (2013) optimized the PRICO process using GA under different optimized in detail yet. Moreover, the pattern of total required
cold box inlet temperatures and correlated linear regression on work variation at different MR compositions when the other pa-
the MR composition. Li et al. (2012) also used GA and exergy ef- rameters are fixed has not been already presented. On the other
ficiency to optimize large scale liquefaction systems. Khan and Lee hand in open literature in which the effect of MR compositions was
(2013) used particle-swarm optimization methodology coded in considered on the total work, the optimization was performed in
MATLAB connected to UniSim simulation software to optimize a different ambient temperatures and at last only the optimized
SMR process. Wahl et al. (2013) applied sequential quadratic values were presented. So there is a lack of study to show how the
programming (SQP) to optimize the PRICO process. Castillo et al. total work varies with changing MR compositions by fixing other
(2013) compared different precooling cycles for LNG production parameters such as ambient temperature particularly in APCI pro-
processes at two different climate conditions, warm (25  C) and cess. In the present study this pattern was obtained and the results
cold (6  C). They investigated both power consumption and heat show that the total work will be varied linearly by MR composi-
transfer coefficient and found that a three stage propane pre- tions. Also the effect of feed (natural gas) pressure on the total work
cooled cycle for warm case and two stage mixed refrigerant cycle of refrigeration cycle was investigated.
for the cold case could be more adequate. Hwang et al. (2013) The present APCI single mixed refrigerant cycle which investi-
optimized a dual mixed refrigerant LNG process for different gated in the present study has some advantages compared to the
number of refrigerant sub-streams and compression stages using other SMR cycles which mentioned above. However, this special
combination of the genetic algorithm and sequential quadratic process has not been investigated yet. So optimization of this cycle
programming. Jacobsen and Skogestad (2013) used both UniSim can be useful in the case of design of mixed refrigerant liquefaction
and MATLAB softwares for modeling and optimization of a SMR processes. In this research, the APCI process was simulated in Aspen
process with disturbances in the flow rate of natural gas and HYSYS as a base case and optimized by connecting GA MATLAB tool
ambient temperature. They expressed that the feasible region of to Aspen HYSYS to minimize the total required work. Then the ef-
operation could be divided into five sub-regions with different fects of ambient temperature and feed pressure on the total
active constraints. Khan et al. (2013) optimized SMR and C3MR required work were investigated. Afterwards, the concentrations of
processes using knowledge based optimization (KBO) method. By the MR components were optimized. In the current paper, not only
considering boiling point difference of the MR components, they effects of temperature and pressure are considered as presented in
improved energy efficiency of the system. Skaugen et al. (2013) several previous studies, but also change of MR compositions are
investigated a concept including a detailed heat exchanger investigated using the proposed correlations.
model for the optimization of a SMR process. Vatani et al. (2013)
considered two mixed refrigerant cycles for design of an LNG 2. Simulation and optimization framework
production process with integrated natural gas liquids (NGL) re-
covery. They formulated an optimization problem with mini- 2.1. APCI process
mizing specific power consumption. Wang et al. (2013) optimized
two types of C3MR processes by investigating different objective The flow sheet of the APCI process modeled by Aspen HYSYS is
functions including shaft work consumption, exergy efficiency shown in Fig. 1.
and operating expenditure. They reported that shaft work con- As shown in Fig. 1, outlet stream of the last compressor
sumption is the best objective function for the optimization of (Compressor-3) is passed through an air cooler (AC-3) divided into
these processes. Cammarata et al. (2001) used GA for optimization two phases (liquid and vapor) in a separator (V-2). Each of them
of a modified Collins cycle which was used for liquefaction of enters the first heat exchanger (HX-1) separately as hot streams.
helium. Austbo et al. (2014) have presented a detailed review The second cooled liquid stream (stream 16) after passing through a
related to the optimization of different LNG processes. throttling valve returns to the heat exchanger as a cold stream and
472 P. Moein et al. / Journal of Natural Gas Science and Engineering 26 (2015) 470e479

Fig. 1. Process flow diagram of APCI LNG liquefaction process.

after heat exchanging, enters the second stage of compression fitness function. On the other hand, all generated data from the
(Compressor-2). The cooled vapor phase goes to a phase separator simulation case of HYSYS are exported to the MATLAB code. There is
(V-3) and the same operation occurs in the second heat exchanger a continuous linking between the MATLAB optimization code and
(HX-2). The outlet cold stream of the third heat exchanger (stream HYSYS simulation case. During data acquisition, the optimization
26) mixes with that of the second exchanger (stream 25) and after by GA is going on until the minimization of the objective function is
heat exchanging enters the first stage of compression (Compressor- performed.
1). The temperature of natural gas after three steps decreases to
about
155  C and after passing through throttling valve (Valve-4),
2.3. Optimization formulation
LNG is produced at
162  C and atmospheric pressure. In this paper
the properties of natural gas and outlet temperature of LNG from
In the current optimization problem, minimization of the total
the last heat exchanger (HX-3) are fixed and only the properties of
work of the plant (for all compressors and a pump) was defined as
MR can vary during the optimization. The simulation assumptions
an objective function expressed as:
are listed in Table 1.
X 
Minimize f ðXÞ ¼ WCompressors þ WPump (1)
2.2. Optimization problem
In the above equation X is an adjusted variable vector including
In order to optimize the process by GA method, MATLAB was MR molar flow rates of components, the outlet pressure of MR
connected to Aspen HYSYS process simulator. In this situation the compressor stages and the MR pressure after each throttling
GA method acts as a controller and the Aspen HYSYS model is a valve. Nitrogen, methane, ethane, propane and n-butane are used
server. A MATLAB program code adapted from a previous work as the components of the mixed refrigerant. PengeRobinson
(Behroozsarand et al., 2009) was prepared to call the Aspen HYSYS equation of state is applied to calculate the properties of both MR
simulation and transfer the data produced by GA. Calculated total and NG.
required work by Aspen HYSYS is sent to MATLAB to be used in GA optimization of this process is limited to three constraints.
First, the minimum approach temperature in the heat exchangers
should be greater than 2  C to satisfy reliability and feasibility of
Table 1 the process. Second, the temperature of MR in the inlet of
Assumptions of the APCI plant simulation. compressors should be greater than the dew temperature of the
fluid in that pressure to prevent formation of liquid in the suction
Parameter Value
part of the compressors. Third, the pressure of streams entering
Adiabatic compressor efficiency 75%
to a mixer should be equal. These constraints are formulated as
Minimum temperature approach of cold box ( C) 2
Pressure drop in heat exchangers (bar) 0.1 follow:
Pressure drop in air coolers (bar) 0.2
Heat leakage of cold box 0% DTmin;HX
i  2  C
Flow rate of natural gas(kg mole/h) 26700
Pressure of natural gas entering cold box (bar) 66.5
Temperature of hot stream at cold box inlet ( C) 32
Ti  Tdew;i
Temperature of natural gas at cold box outlet ( C)
155
Composition of natural gas (% mole)
Pi;in ¼ Pj;in ðfor MixersÞ
N2 0.90
CH4 94.00
C2H6 3.10 where DTmin;HX
i represents the minimum approach temperature
C3H8 1.30 in heat exchanger i, Ti and Tdew;i refer to operating temperature and
n-C4H10 0.40 dew point temperature of the stream i and Pi;in and Pj;in illustrate
i-C4H10 0.30
the mixer pressure of inlet streams (i and j).
 P. Moein et al. / Journal of Natural Gas Science and Engineering 26 (2015) 470e479 473

Table 2 composite curves in the heat exchangers should be greater than

Tuning parameters of GA. 2  C while the temperature of compressors inlet streams should be
Tuning parameters Value greater than the dew point of these streams. This figure is obtained
Population size 300
when the other variables are assumed to be fixed while only the
Selection method Tournament composition of MR can be varied. According to Fig. 2, the total
Tournament size 4 required work extremely depends on the MR composition. It is
Mutation method Adaptive feasible shown that in this process, the relationship between work and
Crossover fraction 0.8
composition is linear. Additionally, the effects for all components
Crossover function Two-point
Stopping criteria do not have the same trend. For instance, rising nitrogen mole
Maximum number of generations 200 fraction causes the required work to be increased, while increase in
Objective function tolerance 10
6 methane mole fraction decrease the work. As illustrated in Fig. 2,
ethane and n-Butane behave similar to methane and the behavior
of propane is similar to nitrogen. The regressed linear equations are
listed below:

8  9
>
> 1:7616 xN2 þ 1:1856  105 0:058 < x < 0:07 >
>
>
>   N2 >
>
>
< 
2:668 xCH4 þ 2:0177  105 0:267 < x < 0:274 >
=
 CH4
WT ¼
1:9899 xC2 H6 þ 1:9512  10 5 j 0:324 < xC2 H6 < 0:334 (3)
>
>  >
>
>
> 2:9399 xC3 H8 þ 0:52886  105 0:258 < xC3 H8 < 0:265 >
>
>
:  >
5 0:0745 < xn
C4 H10 < 0:0775 ;

7:7894 xn
C4 H10 þ 1:8907  10

The last limitation is defined as a linear constraint for GA opti- where WT is total required work in kW and xN2 , xCH4 , xC2 H6 , xC3 H8
mization while the first and second constraints are defined as the and xn
C4 H10 represent the molar composition of nitrogen, methane,
penalties for the objective function as shown below: ethane, propane and n-butane, respectively.
! Therefore, it is important to find the best values for MR com-
P 2 positions in addition to the other variables to minimize the total
Minimize PðX; rÞ ¼ f ðXÞ þ r ½maxf0; gi ðxÞg
i
required work. The base case solution was performed when the
(2)
g1 ðxÞ ¼ 2
DTmin;HX
i inlet hot stream of the cold box was at 32  C. The search conver-
g2 ðxÞ ¼ Tdew;i
Ti gence process is shown in Fig. 3. As shown in Fig. 3 the optimization
was terminated before maximum 200 generations which indicates
where r is the penalty factor, assumed here as 10þ14 . If all the that the average change in the objective function (total required
constraints are satisfied, the second term in the right hand side of work) value is less than 10
6 after about 66 generations according
Eq. (2) would be zero (PðX; rÞ ¼ f ðxÞ). Otherwise the second term to Table 2.
will be a large value that GA modifies the penalty function in the The adjusted variables and the range of each one used in the
next generation. The tuning parameters of GA are listed in optimization solution are presented in Table 3. The lower and upper
Table 2. bounds were selected as 50% less and 50% more than each variable
value in the base case, respectively.
The values of adjusted variables and the total required work in
3. Results and discussion the base and optimized case are presented in Table 4.
As shown in Table 4 after the optimization, the total required
3.1. The base case solution work decreases about 14% compared to the base case. Fig. 4 illus-
trates the hot and cold composite curves of the cold box for the base
Design and control of a MR process is extremely related to the and optimized cases. In fact, the composite curves of all three heat
total required work. This work depends on not only amount of exchangers are collected in a figure.
mixed refrigerant composition, but also the amount of increase or As shown in Fig. 4(a) there is a gap between composite curves
decrease in compositions. that causes low process efficiency due to considerable thermody-
Knowing total required work at different mixed refrigerant namic irreversibility and exergy loss during heat exchange. After
compositions could give a good view about the design of a new MR the optimization temperature difference between hot and cold
process or control of the current processes. Analyzing the results composite curves became smaller than that of the base case
obtained from the simulation could present the relationship be- (Fig. 4(b)). The total required work decreases about 14% due to the
tween total required work and MR composition in APCI and predict decrement in temperature difference. Fig. 5 shows the temperature
the system performance before the optimization. However, only a approach of the cold box in different heat duties.
few researches are found in open literature about it for other MR The approach temperature is based on minimum approach
processes. For considering the effect of compositions, all other temperature of all three heat exchangers. For each heat exchanger a
parameters such as temperature and pressure are assumed hot composite curve is obtained by combination of all hot streams
constant. and at the same way a cold composite curve is obtained by com-
Effect of MR composition on the total required work is shown in bination of all cold streams. The aim of optimization is minimizing
Fig. 2. The hot stream enters to the cold box at a temperature of the difference temperature (approach temperature) between these
32  C. two composite curves (hot and cold streams) for LNG heat
The range of each MR component mole fraction was selected as exchangers.
follow: a minimum approach temperature between hot and cold
474 P. Moein et al. / Journal of Natural Gas Science and Engineering 26 (2015) 470e479

Fig. 2. The variation of total work with Nitrogen, Methane, Ethane, Propane and n-Butane content of MR at 32  C cold box inlet hot stream.

As shown in Fig. 5 the minimum approach temperature in all 3.2. Effect of ambient temperature
heat duties is greater than 2  C that satisfies the feasibility of the
process. In the case of optimized process the approach tem- One of the most important variables that affect the required
peratures along three heat exchangers are smaller than that of work of the LNG plants is ambient temperature. When the ambient
the base case and close to the minimum allowable approach temperature varies, in order to keep the process in an optimum
temperature (2  C) that causes decreasing in energy condition the composition of MR should be changed. In fact, the
consumption. variation of ambient temperature affects basically the outlet
 P. Moein et al. / Journal of Natural Gas Science and Engineering 26 (2015) 470e479 475

xNitrogen ¼ 5  10
4 Ta þ 0:0151

xMethane ¼
4:3  10
3 Ta þ 0:4826
xEthane ¼
4:3  10
3 Ta þ 0:3572 (4)
xPropane ¼ 7:2  10
3 Ta þ 0:1457
xn
Butane ¼ 9  10
4 Ta
0:0006

where Ta is ambient temperature in  C. xNitrogen , xMethane , xEthane ,

xPropane and xn
Butane represent the molar composition of nitrogen,
methane, ethane, propane and n-butane, respectively. In this study
the ambient temperature has been considered equal to cold box
inlet MR temperature. As shown in Fig. 6 when ambient tempera-
ture increases from 20  C to 40  C, the concentration of nitrogen,
propane and n-butane should be increased about 40%, 50% and
103%, respectively. Also the concentration of methane and ethane
should be decreased about 22% and 32%, respectively. Furthermore
as illustrated in Fig. 6 and Eq. (4), the variation slope of propane
Fig. 3. Search convergence process at 32  C cold box inlet hot stream.
concentration versus ambient temperature is greater than that of
other components and the amount of n-butane concentration was
increased more than that of other components (about 2 times).
Table 3 Fig. 7 shows the variation of total required work versus ambient
The adjusted variables and optimization range.
temperature and its second order polynomial regression.
Adjusted Variable Unit Lower bound Upper bound The regressed equation is presented below:
Flow rate of NitrogenðX1 Þ (kmol/h) 2000 9000
Flow rate of Methane ðX2 Þ (kmol/h) 10000 41000
WT ¼ 0:0003Ta2
0:0221Ta þ 1:8772 (5)
Flow rate of Ethane ðX3 Þ (kmol/h) 12000 50000
Flow rate of Propane ðX4 Þ (kmol/h) 8000 34000
Flow rate of n-Butane ðX5 Þ (kmol/h) 3000 13000 where WT is total required work in MW. As shown in Fig. 7 when
Compressor-1 outlet pressure ðX6 Þ (kPa) 650 2000 ambient temperature increases, the total required work decreases
Compressor-2 outlet pressure ðX7 Þ (kPa) 2000 4000 to about 33  C. After 33  C increasing the ambient temperature
Compressor-3 outlet pressure ðX8 Þ (kPa) 5000 7000
causes the total required work increases too. Therefore the base
valve-1 outlet pressure ðX9 Þ (kPa) 600 2000
valve-2 outlet pressure ðX10 Þ (kPa) 150 450 design of this APCI process should be performed at ambient tem-
valve-3 outlet pressure ðX11 Þ (kPa) 150 450 perature of about 33  C and the operating temperature should be
kept at about 33  C to minimize the total required work.
When ambient temperature increases it is required to provide
more refrigeration capacity to liquefy natural gas. In comparison to
Table 4
lighter hydrocarbons such as methane and ethane, heavier hydro-
The values of adjusted variables in the base and optimized process.
carbons with higher boiling points such as propane and n-butane
Case Variables could produce this refrigeration capacity in higher ambient tem-
MR components flow rate (kmol/h) peratures. So the concentration of propane and n-butane in mixed
X1 X2 X3 X4 X5
refrigerant should be increased at higher ambient temperatures.
There are two factors that affect the total required work at different
Base 4388 20557 25470 17223 6478
ambient temperatures. The first one is increment of mixed refrig-
Optimized 4461 21206 25965 19937 6112
erant temperature at the suction part of the compressors. The
Case Variables Obj. Function
compressors work will be increased by increasing fluid tempera-
Pressures (kPa) Total work (kW) ture at the suction part. The second factor is related to the type of
X6 X7 X8 X9 X10 X11 fluid which will be compressed. Since the compressor work varies
by fluid molecular weight inversely (because the outlet pressure of
Base 1061 3197 6040 1051 224 234 149188
Optimized 1325 2719 6021 1315 339 349 128325 all compressors are fixed), heavier hydrocarbons require less work
for compression rather than that of light hydrocarbons. As
mentioned above the amount of propane and n-Butane should be
increased by increasing ambient temperature. So, the required
work for compression of this fluid will be decreased. Fig. 8 shows
temperature of the air coolers. Due to changing ambient temper-
that MR molecular weight varies linearly by ambient temperature.
ature, outlet temperature of air coolers is varied from 20  C to 40  C
Variation of the total required work by MR molecular weight is
with 0.5  C intervals. Hence, the optimization process was per-
shown in Fig. 9.
formed at different ambient temperatures to find the best
As shown in Fig. 9 when MR molecular weight value is equal to
composition of MR. Finally linear regression curves are obtained for
about 31.5 (kg/k mol) which is corresponding to ambient temper-
each component of MR that shows the variation of concentration
ature of about 33  C, the total required work is at minimum value.
with ambient temperature to keep the process in an optimum
Fig. 10 shows variation of total required work by both effective
condition.
parameters: MR molecular weight and ambient temperature. Ac-
Fig. 6 shows the variation of MR composition and its linear
cording to this figure, the ambient temperature and MR molecular
regression to keep the system in the optimized condition at
weight have opposite effects on the total required work. From 20  C
different ambient temperatures.
to about 33  C the effect of fluid type (MR molecular weight) is
The regressed equations are listed below.
more significant and causes the work to be decreased. For
476 P. Moein et al. / Journal of Natural Gas Science and Engineering 26 (2015) 470e479

320 320

300 300
280 280
260 260
Temperature (K)

Temperature (K)
240 240
220 220
Cold Stream
200 200 Cold Stream
Hot Stream
180 180 Hot Stream

160 160
140 140
120 120
100 100
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350
Heat Flow (MW) Heat Flow (MW)
(a) (b)
Fig. 4. Hot and cold composite curves of three heat exchangers in the cold box (a) the base case, (b) the optimized case.

16 12

App.Temp.
14
10
Min. App. Temp.
Approach Temperature (K)
Approach Temperature (K)

12
8
10

8 6

6 App. Temp.
4
4 Min. App. Temp.
2
2

0 0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350
Heat Flow (MW) Heat Flow (MW)
(a) (b)
Fig. 5. Approach temperature curves of three heat exchangers in the cold box (a) the base case, (b) the optimized case.

0.5 159

0.45 158

0.4 157
Nitrogen
Molar Composition

Total Work (MW)

0.35 156
Methane
0.3 155
Ethane
0.25 154
Propane
0.2 153
n-Butane 152
0.15

0.1 151

0.05 150

0 149
18 20 22 24 26 28 30 32 34 36 38 40 42 20 22 24 26 28 30 32 34 36 38 40
Ambient Temperature (°C) Ambient Temperature (°C)

Fig. 6. Optimized MR molar composition in various ambient temperatures. Fig. 7. The total required work in various ambient temperatures.
 P. Moein et al. / Journal of Natural Gas Science and Engineering 26 (2015) 470e479 477

33.5 0.45

33 0.4

32.5 0.35

MR Molar Composition
MR Molecular Weight

32 0.3
31.5 N2
0.25
CH4
31
0.2
C2H6
30.5
0.15 C3H8
30 n-C4H10
0.1
29.5
y = 0.187x + 25.296 0.05
29
0
28.5 5500 5900 6300 6700 7100 7500
20 22 24 26 28 30 32 34 36 38 40 Feed Pressure (kPa)
Ambient Temperature (°C)
Fig. 11. Optimized MR molar composition in various feed pressures.
Fig. 8. Variation of MR molecular weight with ambient temperature.

3.3. Effect of feed pressure

159

158 Another parameter which affects the total required work of the
y = 0.7755x - 49.165x + 930.06
LNG plants is the pressure of natural gas. In this section optimiza-
157
tion of the APCI process was performed by considering the molar
156 flow rates of MR components as adjusted variables and minimi-
Total Work (MW)

155 zation of the total required work as an objective function. Also the
154
other variables used in the optimization of the base case were kept
constant. Feed pressure was changed from 55 bar to 75 bar with a
153
0.5 bar pressure step size. The linear regressions of MR mole frac-
152 tions at various feed pressure levels are indicated in Fig. 11.
151 The regressed equations are listed below:

150

149 xNitrogen ¼
1  10
6 Pf þ 0:034

29 29.5 30 30.5 31 31.5 32 32.5 33
xMethane ¼
7  10
6 Pf þ 0:3995
MR Molecular Weight
xEthane ¼ 1  10
5 Pf þ 0:1311 (6)
Fig. 9. Variation of total required work with MR molecular weight. xPropane ¼
8  10
6 Pf þ 0:4257
xn
Butane ¼ 3  10
6 Pf þ 0:0096

MR Molecular Weight where Pf is the feed pressure in kPa. xNitrogen , xMethane , xEthane ,
28 28.5 29 29.5 30 30.5 31 31.5 32 32.5 33 33.5 34
159 159 xPropane and xn
Butane represent the molar composition of nitrogen,
158 158 methane, ethane, propane and n-butane, respectively. As shown in
Fig. 11 when the feed pressure increases from 55 to 75 bar, the
157 157
concentration of nitrogen, methane and propane should be
156 156
Total Work (MW)

decreased to about 7%, 4% and 4%, respectively. Also the concen-

155 155 tration of ethane and n-butane should be increased to about 11%
154 154 and 23%, respectively. According to Fig. 11 and Eq. (6) the slope of
153 153 ethane concentration versus feed pressure is greater than that of
152 152
other components while the concentration of n-butane concen-
tration was increased more than that of other components. On the
151 151
other hand the feed pressure has not an effective influence on the
150 150
optimization variables including compositions of mixed refrigerant.
149 149 Variation of ambient temperature affects directly the efficiency
20 22 24 26 28 30 32 34 36 38 40
of compressors (at different inlet temperatures) and the mixed
Ambient Temperature (°C)
refrigerant conditions. This is due to changing temperature and
Amb. Temperature Molecular Weight
phase condition (liquid/gas ratio) of outlet streams of air coolers (12
Fig. 10. Variation of total required work by MR molecular weight and ambient and 13) and operation of HX-1, HX-2, and HX-3. So for the opti-
temperature. mization of this new condition, the composition of MR must be
changed clearly to minimize the gap between composite curves of
HXs. Feed stream (natural gas) is independent and its pressure
variation has an indirect effect on mixed refrigerant conditions.
temperatures above 33  C the effect of fluid temperature at Although the feed pressure affects the total required work of mixed
compressor suction is more significant and causes the required refrigerant compression, its amount is less significant than those of
work for compression to be increased. ambient temperature on the mixed refrigerant compositions. Also,
478 P. Moein et al. / Journal of Natural Gas Science and Engineering 26 (2015) 470e479

156 Pf feed pressure (kPa)

T temperature ( C)
154
Ta Ambient temperature ( C)
152 DTmin minimum approach temperature ( C)
Total Work (MW)

150 W work (kW)

WT total required work (kW, MW)
148 x molar composition
146 X adjusted variable vector

144
Subscripts
142 dew dew point
i stream i
140
5500 5900 6300 6700 7100 7500 in inlet stream
Feed Pressure (kPa) j stream j

Fig. 12. The total required work in various feed pressures.

Abbreviations
APCI Air Products and Chemicals, Inc
the variation of Feed pressure has a negligible effect on HX com- GA Genetic Algorithm
posite curves. Fig. 12 shows the variation of total required work LNG Liquefied Natural Gas
versus feed pressure and its linear regression. MR Mixed Refrigerant
The regressed equation is presented below: NG Natural Gas
SMR Single Mixed Refrigerant
WT ¼
0:0047Pf þ 179:26 (7)
References
where WT is the total required work in MW and Pf is the feed
pressure in kPa. As shown in Fig. 12 when the feed pressure in- Alabdulkarem, A., Mortazavi, A., Hwang, Y., Radermacher, R., Rogers, P., 2011.
Optimization of propane pre-cooled mixed refrigerant LNG plant. Appl. Therm.
creases, the total required work decreases continuously. According Eng. 31 (6e7), 1091e1098.
to Eq. (7) by increasing the feed pressure from 55 bar to 75 bar, the Andress, D.L., Watkins, R.J., 2004. Beauty of simplicity: phillips optimized cascade
total required work decreases about 6%. LNG liquefaction process. Adv. Cryogen. Eng. 49, 91e98.
Aspelund, A., Gundersen, T., Myklebust, J., Nowak, M.P., Tomasgard, A., 2010. An
optimizationesimulation model for a simple LNG process. Comput. Chem. Eng.
4. Conclusion 34 (10), 1606e1617.
Austbo, B., Lovseth, S.W., Gundersen, T., 2014. Annotated bibliographyduse of
optimization in LNG process design and operation. Comput. Chem. Eng. 71,
The APCI LNG process was simulated and optimized by coupling 391e414.
of Aspen HYSYS as the process simulation software and MATLAB as Behroozsarand, A., Ebrahimi, H., Zamaniyan, A., 2009. Multiobjective optimization
the optimizing software. GA optimization method was chosen in of industrial autothermal reformer for syngas production using nonsorting
genetic algorithm II. Ind. Eng. Chem. Res. 48, 7239e7529.
order to make sure a global optimum condition would be reached. Cammarata, G., Fichera, A., Guglielmino, D., 2001. Optimization of a liquefaction
The optimized value of each variable was obtained and reported. plant using genetic algo rithms. Appl. Energy 68 (1), 19e29.
The total required work was reduced by 14% compared to the base Cao, W., Lu, X., Lin, W., Gu, A., 2006. Parameter comparison of two small-scale
natural gas liquefaction processes in skid-mounted packages. Appl. Therm.
case. Also, the effect of ambient temperature on the process per- Eng. 26 (8), 898e904.
formance was investigated. The optimization was performed in the Castillo, L., Dahouk, M.M., Di Scipio, S., Dorao, C.A., 2013. Conceptual analysis of the
range of 20  Ce40  C with 0.5  C step sizes by variation of molar precooling stage for LNG processes. Energy Convers. Manag. 66, 41e47.
Cranmore, R.G., Stanton, E., 2000. Natural gas. In: Dawe, R.A. (Ed.), Modern Petro-
flow rate of MR components only. When the ambient temperature leum Technology, Upstream Volume. Institute of Petroleum, John Wiley and
increases, to keep the process in the optimum condition the con- Sons Ltd, Chichester, pp. 337e382.
centration of nitrogen, propane and n-butane should increase and Gong, M.Q., Wu, J.F., Luo, E.C., Zhou, Y., Liang, J.T., Zhang, L., 2000. Optimum
composition calculation for multicomponent cryogenic mixture used in JeT
methane and ethane should decrease. When the ambient temper-
refrigerators. Adv. Cryo. Eng. 45, 283e290.
ature increases, the total required work behavior follows a second Hatcher, P., Khalilpour, R., Abbas, A., 2012. Optimisation of LNG mixed-refrigerant
order polynomial with about 33  C minimum point which is cor- processes considering operation and design objectives. Comput. Chem. Eng.
41, 123e133.
responding to MR molecular weight of about 31.5 kg/kmol. In
Hwang, J.H., Ku, N.K., Roh, M.I., Lee, K.Y., 2013. Optimal design of liquefaction cycles
addition, the optimization was performed in various feed pressures of liquefied natural gas floating, production, storage, and offloading unit
from 55 bar to 75 bar. The results showed that there is not a sig- considering optimal synthesis. Ind. Eng. Chem. Res. 52 (15), 5341e5356.
nificant change in MR compositions in this range of pressure. Jacobsen, M.G., Skogestad, S., 2013. Active constraint regions for a natural gas
liquefaction process. J. Nat. Gas. Sci. Eng. 10, 8e13.
Moreover the total work decreases linearly by increasing the feed Khan, M.S., Lee, M., 2013. Design optimization of single mixed refrigerant natural
pressure. gas liquefaction process using the particle swarm paradigm with nonlinear
constraints. Energy 49, 146e155.
Khan, M.S., Lee, S., Rangaiah, G.P., Lee, M., 2013. Knowledge based decision making
Acknowledgements method for the selection of mixed refrigerant systems for energy efficient LNG
processes. Appl. Energy 111, 1018e1031.
Lee, G.C., Smith, R., Zhu, X.X., 2004. Optimal synthesis of mixed-refrigerant systems
We gratefully acknowledge the support of Research & Tech-
for low temperature process. Ind. Eng. Chem. Res. 41, 5016e5028.
nology Directorate of National Iranian Gas Company. Li, Y., Wang, X., Ding, Y., 2012. An optimal design methodology for large-scale gas
liquefaction. Appl. Energy 99, 484e490.
Mokarizadeh Haghighi Shirazi, M., Mowla, D., 2010. Energy optimization for
Nomenclature liquefaction process of natural gas in peak shaving plant. Energy 35,
2878e2885.
Mortazavi, A., Somers, C., Hwang, Y., Radermacher, R., Rodgers, P., Al-Hashimi, S.,
2012. Performance enhancement of propane pre-cooled mixed refrigerant LNG
Symbols plant. Appl. Energy 93, 125e131.
P pressure (kPa) Remeljej, C.W., Hoadley, A.F.A., 2006. An exergy analysis of small-scale liquefied
 P. Moein et al. / Journal of Natural Gas Science and Engineering 26 (2015) 470e479 479

natural gas (LNG) liquefaction processes. Energy 31 (12), 2005e2019. Wahl, P.E., Løvseth, S.W., Mølnvik, M.J., 2013. Optimization of a simple LNG process
Roberts, M.J., Agrawal, R., Daugherty, T.L., 2002. Single Mixed Refrigerant Gas using sequential quadratic programming. Comput. Chem. Eng. 56, 27e36.
Liquefaction Process. US Patent 6,347,531. Wang, M., Khalilpour, R., Abbas, A., 2013. Operation optimization of propane pre-
Skaugen, G., Walnum, H.T., Hammer, M., Wahl, P.E., Wil-helmsen, Ø., Kolsaker, K., cooled mixed refrigerant processes. J. Nat. Gas. Sci. Eng. 15, 93e105.
2013. Design and optimization of compact heat exchangers in process used for Xu, X., Liu, J., Jiang, C., Cao, L., 2013. The correlation between mixed refrigerant
liquefaction of natural gas. In: 5th International Conference on Applied Energy composition and ambient conditions in the PRICO LNG process. Appl. Energy
ICAE 2013. 2013 July 1e4, Pretoria, South Africa. 102, 1127e1136.
Swenson, L.K., 1977. Single Mixed Refrigerant, Closed Loop Process for Liquefying Yuan, Z., Cui, M., Xie, Y., Li, C., 2014. Design and analysis of a small-scale natural gas
Natural Gas. US Patent 4,033,735. liquefaction process adopting single nitrogen expansion with carbon dioxide
Vatani, A., Mehrpooya, M., Tirandazi, B., 2013. A novel process configuration for co- pre-cooling. Appl. Therm. Eng. 64 (1), 139e146.
production of NGL and LNG with low energy requirement. Chem. Eng. Process: Zhang, X., Myhrvold, N.P., Caldeira, K., 2014. Key factors for assessing climate
Process Intensif. 63, 16e24. benefits of natural gas versus coal electricity generation. Environ. Res. Lett. 9,
Venkatarathnam, G., 2008. Cryogenic Mixed Refrigerant Processes. Springer, New 114022.
York.