Journal of Natural Gas Science and Engineering 17 (2014) 119e130

Contents lists available at ScienceDirect

Journal of Natural Gas Science and Engineering

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

Correlating the additional amine sweetening cost to acid gases load in

natural gas using Aspen Hysys
Ayyaz Muhammad a, b, Yasser GadelHak a, *
a
Chemical and Materials Engineering Department, Faculty of Engineering, King Abdulaziz University, P.O. Box 80204, Jeddah 21589, Saudi Arabia
b
Institute of Chemical Engineering and Technology, University of the Punjab, Lahore 54590, Pakistan

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

Article history: Natural gas with high acid gas contents represents a challenge in process engineering and energy in-
Received 28 December 2013 dustry. Exploring the alternative solutions requires correlating additional costs accompanying the pro-
Received in revised form cess scale-up to treat high acid-gas contents. This work estimates this relation and validates it against
14 January 2014
various conditions in amine sweetening plants. An amine sweetening process was simulated using Aspen
Accepted 20 January 2014
Available online 11 February 2014
Hysys to treat a natural gas (25 MMSCFD, 1.7 mol% H2S and 4.13 mol% CO2). Amine circulation rate, lean
amine temperature, re-boiler temperature and amine concentration were chosen as the main input
variables to optimize the process total cost using the central composite experimental design model. The
Keywords:
Amine sweetening
process was scaled-up to handle a sour gas up to (25 mol% CO2 and 3 mol% H2S) by scaling up the amine
Aspen Hysys circulation rate. The capital and operating costs showed a linear relationship with the increase of CO2
High acid gas percentage in natural gas. Moreover the utility requirements, regenerator column diameter and the
Sour natural gas surface area of the re-boiler, the lean-rich amine heat exchanger and the lean amine cooler showed a
Process optimization strong linear increase correlation. There was no statistical difference using either LieMather or Kent
eEisenberg thermodynamic models. Low feed gas pressure yielded a similar linear relationship irre-
spective of the costs for purchasing and operating the inlet compressor and air cooler. No statistical
difference found between cost regressions lines of different tray efficiency of the distillation tower. For
various partial pressures of CO2 and H2S, assuming no other sulphur compounds, the economics scaled
up as a second order polynomial.
Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction appropriate sweetening process based on the partial pressure of

acid gases in the feed and product stream (Manning and
The growing needs of the world for natural gas imply Thompson, 1995). These charts show that amines, potassium
improving the existing processes for production, treatment, en- carbonate and physical solvents are generally suitable till a
ergy harvesting and pollution control. One of the most serious maximum of about 8% acid gas concentration in the feed gas to
problems facing natural gas processing technology is the presence meet the pipeline specification (2% CO2 and 4 ppm H2S). To
of high contents of acid gases such as CO2 and H2S. Acid gases are handle sour gases with more contamination only physical sol-
known to contaminate about 40% of the known gas reserves (van vents and membranes are more suitable. One drawback of these
Kemenade et al., 2013). About 13.46% of the world’s natural or guidelines is that they are approximate and general because many
associated gas reserves have H2S over 10% and about 26.9% have other factors should be considered including operating condi-
more than 10% CO2 (Lallemand et al., 2011). Several methods have tions, the gas sales contract and detailed gas analysis (Manning
been developed to reduce the content of acid gases in natural gas and Thompson, 1995; Kidnay and Parrish, 2006). In order to
streams including absorption by physical and chemical solvents, compare different alternatives the design engineer needs to
adsorption, cryogenic or low-temperature separation and mem- perform an economic analysis for the additional costs per unit
brane separation (Berstad et al., 2012; Olajire, Jun. 2010). Manning increase in the acid gas contents. The capital cost increase due to
and Thompson reported charts to aid the designer select the the scale up of the processing equipment has to be analysed and
correlated to the acid gas content in the feed gas. Also the need for
purchasing and installing new equipment e.g. compressors to
* Corresponding author. Tel.: þ966 569264619; fax: þ966 26952257. recompress the feed gas to the high pressures required by the
E-mail address: yasser191919@gmail.com (Y. GadelHak). physical absorbent processes or the membrane module has to be

1875-5100/$ e see front matter Ó 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jngse.2014.01.008
120 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130

included in the capital cost analysis. Additional operating costs Table 1

due to the additional energy requirements for solvent regenera- Composition of sour natural gas fed to the DEA amine sweetening simulated case
(temp. ¼ 25  C and pressure ¼ 1000 psia).
tion, lean solvent pumping and acid gas separation has to be also
evaluated. This paper focuses on correlating additional fixed Component Mol Component Mol Component Mol
capital and operating costs with acid gas increase for amine fraction fraction fraction

sweetening processes. The current work is divided into four parts. N2 0.0016 i-C4 0.0026 C2 0.0393
The first part focuses on simulating an amine sweetening process H2S 0.0172 n-C4 0.0029 C3 0.0093
CO2 0.0413 i-C5 0.0014 i-C4 0.0026
to treat acid gases (1.7 mol% H2S and 4.13 mol% CO2) in a sour
C1 0.8692 n-C5 0.0012 C6 0.0018
natural gas feed. The second part is concerned with identifying H2O 0.0122 DEAmine 0
the key variable(s) and the effect of their interactions on the
estimated total cost of the process. This is done by extending the
concept of experimental design models for laboratory setups to
parameters. This data also saves time required to converge the
the numerical output of the simulation. The third part deals with
towers and the recycle operation. For this simulation case Kente
scaling up the process to treat higher content-sour gas (up to 3%
Eisenberg along with the non-ideal vapour model were selected
H2S and 25% CO2). The key variable(s) and the total cost are then
based on the guidelines in the package manual (A. HYSYS, 2004).
correlated to the percentage of acid gases in the feed. Finally, the
The sour gas (25 MMSCFD, composition shown in Table 1) initially
obtained relation is tested against different operating conditions
enters a Knock-out drum to separate any free water carried with the
to assure the validity of the results for a wide range of conditions
gas.
that reflect situations in typical industrial amine sweetening
The gas is then sent to the contactor (absorber) fed with 28wt%
plants.
DEAmine solution (@ 35  C and 995 psia) flowing at a standard
ideal liquid volumetric flow rate of 43 m3/h. Operating conditions
2. Materials and methods
and temperature profile estimates of the absorber column are
summarized in Table 2.
Aspen Hysys v8.0 was used to model the process flow sheet. This
Rich DEA amine is directed to a control valve to reduce its
software was chosen for two reasons. First, the software offers a
pressure to 620 KPa (close to the regenerator operating pressure).
powerful thermodynamic package called the “Amine package”
After pressure reduction the exit stream is directed to a flash
which is a special package designed for modelling amine sweet-
separator to flash off gases from the stream. The stream that exits
ening units. This package includes experimental solubility and re-
the flash tank enters a shell and tube heat exchanger (tube side)
action kinetics data over wide range of conditions (A. HYSYS, 2004).
to be heated to 95  C before being sent to the regenerator. The
Incorporating this package in the simulation, generally, reduce er-
bottom product from the regenerator (lean regenerated amine) is
rors arising from using general thermodynamic equations of state
the shell side stream that will be further processed to be recycled
or activity models. The second reason is the powerful cost analysis
back to the absorber. Pressure drop for both tube and shell sides
tool integrated in the software which was used to estimate the fixed
was set to 70 KPa. The simple weighted design was chosen as
and operating costs of the process under different conditions. This
design basis for the software to calculate the heat exchanger size
tool saves the time and effort needed through the tedious paper-
and duty.
and-pencil calculations to size and estimate the cost of various
The regenerator (distillation column) consists of 18 stages
equipment. It provides can provide summarized and highly
(trays). It should be noted that Aspen Hysys considers the re-boiler
detailed estimated information depending on the user’s
automatically as the nineteenth stage. The details used to simulate
requirements.
the regenerator are summarized in Table 3.
Since the program provides numerical results as an output after
After being cooled in the heat exchanger the lean amine solution
modifying the tested non-modifiable variable(s), the optimization
is mixed with pure make-up water (temp. ¼ 25  C). The mixer
procedure and the statistical analysis was done using Statistica v7.0
parameters were set to equalize all pressures and the combined
software. Statistica provides powerful industrial experimental
stream was assigned a standard ideal liquid volumetric flow rate of
models to relate a numerical output from an experiment to inputs
43 m3/h. A cooler is further needed (pressure drop ¼ 35 KPa) to
through experimental design techniques. This procedure reveals
further cool the lean amine solution which is then pumped through
the relative effect of the variables being studied and their interac-
a centrifugal pump to be recycled to the tower. The exit stream from
tion on the output. The program provides experimental models for
the pump was assigned a temperature of 35  C and a set operation
multi-input experiments which is more relevant to analyse and
was used adjust its pressure with respect to the Gas to contactor
optimize an industrial process.
stream pressure with a multiplier of 1 and an offset of 35 KPa (i.e.
pressure of lean amine is always less than the entering gas pressure
3. Results and discussion
by 35 KPa). This stream is recycled through a recycle operation to
the tower.
The typical amine sweetening plant consists of an absorber
The simulated flow sheet is shown in Fig. 1.
column where the sour gas feed is contacted with amine solution in
a counter-current fashion where trays or packing are used to
enhance mass transfer. Acid gases are absorbed into the solvent Table 2
that is then regenerated in a regeneration distillation tower to Operating conditions and temperature profile estimates used to simulate the amine-
recover the amine solution which is cooled and recycled back to the sour gas absorber column.
absorber tower. Item Value Item Value
As a starting point, inputs and equipment parameters data from
Number of stages 20 Bottom stage pressure 6900 KPa
an available tutorial file accompanying the software documentation DEA to contactor Top stage Estimates for top stage 40  C
(2004 version) were used to simulate the main case. Such data was stream fed at temperature
important as a starting point to set the software solvers computa- Sour gas feed Bottom stage Estimates for bottom 70  C
tional parameters like the solution algorithm for equipment like the stage temperature
Top stage pressure 6850 KPa
distillation tower and the heat exchangers with the optimum solver
 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130 121

Table 3 Table 4
Operating conditions, calculation method coefficients and temperature profile es- Variables and their change domains chosen to optimize the total cost of the amine
timates of the rich amine distillation column. sweetening process through experimental design models.

Item Value Item Value Variables Variables/coded values 2 1 0 1 2

symbol
Feed tray Fourth tray Tray one temp. 100  C
Condenser set Full reflux Re-boiler 125  C A Amine circulation rate (m3/h) 46 47 48 49 50
temperature B Lean amine to contactor 80 90 100 110 120
Solution method Modified HYSIM CO2 efficiency 0.15 temperature ( F)
Inside-Out C Re-boiler temperature ( F) 260 265 270 275 280
Fixed damping factor 0.4 H2S efficiency 0.8 D Amine concentration (wt%) 25 27.5 30 32.5 35
of the solver
Condenser pressure 190 KPa Overhead rate 1.5 MMSCFD
Re-boiler pressure 220 KPa Reflux ratio 1.5
Condenser pressure 15 KPa plotted in Fig. 2 (right). As shown the observed variables vary lin-
drop early with the predicted ones and the residuals follow a smooth
normal distribution. This supports the use of the model to relate
inputs and outputs in a similar manner to that used for laboratory
3.1. Optimizing the main case experiments design and supports further statistical interferences of
the values of model coefficients and their level of significance on
The first goal was to optimize the main case and determine the the output.
main variables affecting the total cost. The total cost is defined here As shown in Table 6, the high values of the significance level
as the sum of the capital cost and the operating cost estimated by show strong evidence towards accepting the null hypothesis. Also it
the software for 365 days of operation. is observed that the confidence interval includes zero for all con-
Four variables were chosen to be manipulated and a central stants of the model which supports the conclusion that none of
composite four-variable, five-level experimental design model was these variables shows statistical evidence for affecting the output.
generated by Statistica software which was used to statistically This suggests that the domain for all variables considered in this
determine a model relating these variables to the total cost. The model represent an optimum minimum with no significant change
variables and their levels are shown in Table 4. in the output with any change in any variable.
These variables were chosen as a balance between a represen- A deeper optimization procedure was needed in order to
tative model and computational requirements for flow sheet investigate the effect of more variables and their interaction on the
convergence and economics estimation. The levels of the last three total cost so that a limiting variable would be identified for further
variables were chosen based on the recommended range for process scale-up or modification. One of the most important vari-
running the amine package by the software manual. The first var- ables is the temperature profile of the absorption column which has
iable was changed in the shown range based on experience with a crucial effect on the process (Lunsford and McIntyre, 1999). As
flow sheet convergence at different conditions that are required by seen in Fig. 3 the temperature distribution of the absorber column
the statistical software to model the phenomenon. While the trays is not uniform. The values for the first and last trays were
minimum flow rate was found to be 43 m3/h, corresponding to the provided to the simulator as an estimates but still as seen, no need
maximum acid gas loading, that flow range did not allow flow sheet for the high temperature provided (70  C) at the bottom tray. Also
convergence at different operating conditions. The experimental the first tray can accept a lower temperature input.
design output is shown in Tables 5 and 6 respectively. In order to include such variable, an experimental design model
The predicted versus observed values for this model (R2 ¼ 0.92) was used with 10 variables at two levels each. The added variables
are plotted in Fig. 2 (left). The histogram of residuals distribution is included the number of stages in both the regenerator and the

Fig. 1. DEA amine sweetening plant flow sheet simulated by Aspen Hysys.
122 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130

Table 5 important variables affecting the flow sheet are adjusting the
Aspen Hysys fixed and operating costs estimation results corresponding to various absorber temperature (through cooling the lean amine) and the
operating conditions.
solvent flow rate.
Run Var. A Var. B Var. C Var. D Fixed cost Op. Cost
(US $) (US $) 3.2. Scaling up the amine package
1 1 1 1 1 5,735,300 2,884,940
2 1 1 1 1 5,717,190 2,890,450 To scale up the process only the amine flow rate was the main
3 1 1 1 1 5,754,780 2,893,400
variable to be scaled up. This is because investigating the effect of
4 1 1 1 1 5,714,950 2,895,840
5 1 1 1 1 5,627,810 2,889,170 decreasing the absorber column temperature would require
6 1 1 1 1 5,639,590 2,899,380 considering column side operations and pinch analysis techniques
7 1 1 1 1 5,627,920 2,889,510 to integrate the heat utilization in the process. This type of analysis
8 1 1 1 1 5,627,280 2,899,690
is beyond the scope of this paper and therefore shall not be
9 1 1 1 1 5,787,050 2,923,400
10 1 1 1 1 5,744,870 2,922,280
considered for the current case. The importance of considering the
11 1 1 1 1 5,763,880 2,922,570 amine circulation rate can be further supported by Chan and
12 1 1 1 1 5,726,170 2,932,710 Tontiwachwuthikul (1996) who reported that about 70% of the
13 1 1 1 1 5,648,660 2,916,400 plant investment is directly associated with the solvent recircula-
14 1 1 1 1 5,627,370 2,916,280
tion rate since it has a direct impact of the size of the absorber
15 1 1 1 1 5,642,520 2,919,120
16 1 1 1 1 5,628,470 2,926,820 tower, the piping, the circulation pump and the size of the regen-
17 2 0 0 0 5,698,510 2,885,150 erator facilities (Chan and Tontiwachwuthikul, 1996).
18 2 0 0 0 5,727,360 2,935,140 To scale up acid gas content, first the mole percent of H2S was
19 0 2 0 0 5,707,440 2,946,140 increased to 3%. Hydrogen disulfide was not increased further
20 0 2 0 0 5,638,600 2,910,070
21 0 0 2 0 5,711,230 2,907,750
because in nature if the inlet concentration of H2S exceeds 2e3%,
22 0 0 2 0 5,693,770 2,907,750 other sulphur species e.g. carbonyl sulphide (COS), carbon disulfide
23 0 0 0 2 5,695,000 2,902,960 (CS2), elemental sulphur, and mercaptans may be present in the
24 0 0 0 2 5,651,620 2,907,510 feed gas (Kidnay and Parrish, 2006). This factor requires further
25 0 0 0 0 5,699,480 2,905,810
investigation and therefore was not part of this work. Such increase
26 0 0 0 0 5,693,770 2,906,330
in the feed gas content of H2S caused the fixed cost to increase from
5,703,180 $ to 5,755,590 $ and the operating cost from 2832660 $ to
2952090 $. The main factor affecting the increase of fixed capital
absorber, the regenerator feed temperature, the DEA recycle tem-
cost was the increase of the diameter of the absorber from 2.5 ft to
perature and the temperatures of the first and last stages in each
3 ft and the re-boiler size that increased its cost by 9.1%. As for the
tower.
operating costs, the re-boiler duty increased by 10.38% where the
The design model consisted of 16 runs. About half of these runs
steam cost increased by an equivalent percentage. Also a more
could converge but the remaining couldn’t converge. Even with
remarkable change was the increase of the use of cooling water by
several trials to change the domain of each variable, converging the
69.1% leading to an increase in its cost by 67.82%.
flow sheet was so challenging especially for the cases with a long
The percentage of carbon dioxide was further changed from
absorber and short regenerator. The original design with the out-
4.13% to 25% which was the maximum percentage allowed by the
puts is presented in Tables 7 and 8 respectively.
amine package at the feed gas operating conditions. For such
Based on trial and error and experience with this design, a
changes the specifications of the distillation column needed to be
simpler design had to be used through fixing the numbers of stages
changed to allow more acid gases to be removed so the overhead
of both towers and fixing the regenerator feed temperature at
vapour rate spec. was de-activated and instead the column was set
100  C and the values of the new design are presented in Table 9.
to allow full recovery of the DEA amine. This would set the flow rate
The Pareto chart of effects for this design is shown in Fig. 4. And
of acid gases (overhead vapour) free to change according to process
as seen the effect of variable D; the amine flow rate and variable C;
conditions to meet the target specifications of the sweet gas. All
the amine temperature had the greatest effect irrespective of their
changes in the mole percent of acid gases was balanced by
narrow range of change. This shows that in general the two most
decreasing the mole percent of methane gas.
The relation between the amine circulation rate and the mole
fraction of carbon dioxide in the feed gas is presented in Fig. 5. The
Table 6 linear relation was so strong with a coefficient of correlation of
Model coefficients and the statistical interference of the effect of single variables and
0.9988. It worth to mention that during scaling up the process the
their interactions on the mean value of the estimated total cost of the process.
acid gases partial pressure in the feed to the distillation column was
Factor Coefficient P 95% 95% outside the range of specifications allowed by the amine package so
Var. A linear 1183700 0.510486 5.014331Eþ06 2.646930Eþ06 the pressure drop across the valve VLV-100 was increased to meet
Var. A quadratic 3886 0.493541 8.189723Eþ03 1.596077Eþ04 these specifications. The values of mole fractions and the corre-
Var. B linear 5011829 0.799328 4.736535Eþ07 3.734170Eþ07 sponding minimum pressure drop are showed in Fig. 6.
Var. B quadratic 160323 0.775553 1.367847Eþ06 1.047202Eþ06
Var. C linear 1040935 0.946541 3.235782Eþ07 3.443969Eþ07
The capital and operating cost for different values of CO2 mol
Var. C quadratic 16951 0.903873 2.849306Eþ05 3.188316Eþ05 fraction are shown in Fig. 7. The capital and operating costs showed
Var. D linear 33561 0.993858 9.413787Eþ06 9.346664Eþ06 a strong linear relationship (coefficient of correlation ¼ 0.98 for
Var. D quadratic 45301 0.213273 1.207717Eþ05 3.016884Eþ04 both) with the increase of CO2 mole percent. As shown the oper-
A*B 69244 0.252233 1.953656Eþ05 5.687806Eþ04
ating costs increased with a higher slope being even greater than
A*C 13622 0.643776 7.668278Eþ04 4.943903Eþ04
A*D 12527 0.400580 4.405702Eþ04 1.900389Eþ04 the fixed cost of the plant at values 25 mol% acid gases (22 mol%
B*C 6719 0.981711 6.373278Eþ05 6.238903Eþ05 CO2 and 3 mol% H2S). The noticeable increase of operating costs
B*D 194734 0.201252 1.205702Eþ05 5.100389Eþ05 suggests that process modifications should be more concerned by
C*D 5039 0.945177 1.626913Eþ05 1.526132Eþ05 decreasing such huge increase. Alie et al. (2006) cited a references
Constant 19064190 0.968531 1.058769Eþ09 1.020641Eþ09
that reached a similar conclusion, empathising that the operating
 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130 123

Fig. 2. Predicted versus observed values (left) and histogram of residuals (right).

costs are more important than capital costs for amine sweetening was to treat a feed gas with a maximum of 60e100 psi acid gas
plants that treat flue gases from coal-fired power plants (Alie et al., partial pressure, corresponding roughly to 2e5.5 mol% of CO2 in the
2006). feed stream (with H2S being set to 3% already) (Kidnay and Parrish,
Moreover not only the total costs increased linearly but also the 2006). Another recommendation was to accept a feed having acid
key variables of the processes also increased linearly with CO2 mol gases with 200 psi partial pressure which will correspond to
percent increase. As shown in Fig. 8, electricity, cooling water and 13.6 mol% of CO2 (Kidnay and Parrish, 2006). Although these values
steam requirements, increased linearly (R2 ¼ 0.98, 0.99 and 0.97 were based on equilibrium conditions and another process vari-
respectively) with the increase of CO2 contamination. This result ables should be considered during process selection such as
supports the claims that the energy utilization in general is pro- detailed gas analysis, operating conditions, well location and the
portional to the intensity of acid gases in the feed gas for solvent sales contract, still they represent a general guide line for process
absorption techniques (van Kemenade et al., 2013; Kidnay and selection (Olajire, Jun. 2010; Bergel and Tierno, 2009). Based on that
Parrish, 2006). Fig. 7 is seen to represent the intended reference line for the cost
Not only did the utility requirement change linearly but also penalty accompanying process scale-up and can be used as basis of
the main assets of the process were scaled up linearly. For comparison for other processes or modifications aiming at
instance, as shown in Fig. 9, the surface area requirement of the re- decreasing the cost of acid gas sweetening processes.
boiler, the lean-rich amine heat exchanger and the cooler E-100
increased linearly. Also the regenerator diameter (as shown in 3.3. Conditions in typical industrial plants
Fig. 10) increased linearly (R2 ¼ 0.98) with the increase of acid gas
content. In order to extend the analysis to study conditions that may be
The recommendation for maximum acid gases content treated encountered in typical industrial plants, several variables were
by amine package e as reported by Kidnay and Parrish (2006) e changed and the relation between estimated costs and acid gases
increase was re-estimated. These variables will be discussed in the
next section in more details.

3.3.1. The thermodynamic model

Several models were chosen to study the effect of the used
thermodynamic model to calculate the kinetic, solubility and

Table 7
Variables and their change domains chosen to further optimize the total cost of the
amine sweetening process through experimental design models.

Var. code Variable Units Lower level Higher level

Absorber number of stages tray 13 18


Var. A Absorber top stage temp. C 15 35

Var. B Absorber bottom stage temp. C 25 50

Var. C DEA recycle temp. F 90 110
Var. D DEA flow rate m3/h 46 49
Var. E DEA concentration Wt% 25 35

Regenerator feed temp. C 80 110
Regenerator number of stages tray 13 18

Var. F Regenerator top stage temp. C 60 90

Var. G Regenerator boiler temp. F 250 290
Fig. 3. Absorber column temperature profile.
124 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130

Table 8
Aspen Hysys fixed and operating costs estimation results corresponding to various equipment sizes and operating conditions.

Run A B C D E F G H I J Ya F.C.b O.C.c

1 1 1 1 1 1 1 1 1 1 1 0.003 5,288,180 3,045,260

2 1 1 1 1 1 1 1 1 1 1 0.001 5,024,870 2,591,990
3 1 1 1 1 1 1 1 1 1 1 0.0027 5,383,690 3,112,960
4 1 1 1 1 1 1 1 1 1 1 0.0011 8,324,460 2,646,390
5 1 1 1 1 1 1 1 1 1 1 0.0024 8,576,840 2,626,220
6 1 1 1 1 1 1 1 1 1 1 Not converged
7 1 1 1 1 1 1 1 1 1 1 0.0035 4,997,830 2,523,850
8 1 1 1 1 1 1 1 1 1 1 Not converged
9 1 1 1 1 1 1 1 1 1 1 Not converged
10 1 1 1 1 1 1 1 1 1 1 Not converged
11 1 1 1 1 1 1 1 1 1 1 0.002 11,116,500 2,617,250
12 1 1 1 1 1 1 1 1 1 1 Not converged
13 1 1 1 1 1 1 1 1 1 1 0.002 5,364,750 3,114,180
14 1 1 1 1 1 1 1 1 1 1 Not converged
15 1 1 1 1 1 1 1 1 1 1 0.0015 5.226.950 3,090,180
16 1 1 1 1 1 1 1 1 1 1 0.0006 6,857,700 2,658,350
a
Y ¼ volume fraction of carbon dioxide in the sweet gas.
b
F.C. ¼ fixed cost.
c
O.C. ¼ operating cost.

Table 9 A different result was obtained from co-relating the operational

Aspen Hysys total cost estimation results corresponding to various operating costs. As shown in Fig. 12, there is statistical evidence that both co-
conditions.
relations vary significantly with LieMather model always yielding
Run A B C D E F G Total cost higher estimated operational cost. This was explained based on the
1 1 1 1 1 1 1 1 8,473,650 sweetening degree estimated by each model. Using LieMather
2 1 1 1 1 1 1 1 8,435,610 model always resulted in a lower carbon dioxide volume fraction in
3 1 1 1 1 1 1 1 8,422,260 the sweet gas for a given minimum amine circulation rate. This
4 1 1 1 1 1 1 1 8,461,650
would result in higher cooling requirements in the distillation
5 1 1 1 1 1 1 1 8,426,521
6 1 1 1 1 1 1 1 8,305,680 condenser which increases the operational costs. To avoid such a
7 1 1 1 1 1 1 1 8,350,060 lurking variable the scale-up calculation was repeated for both
8 1 1 1 1 1 1 1 8,433,720 models but for a constant carbon dioxide volume fraction in the
sweet gas. The required minimum Amine circulation rate is sown in
Fig. 13. Using KenteEisenberg model always required a higher
even thermodynamic properties of chemical species. First a Amine circulation rate to achieve the same sweetening degree
comparison between LieMather and KenteEisenberg models was (carbon dioxide volume fraction in the sweet gas).
done. Comparing the capital costs still didn’t show any statistical dif-
As shown in Fig. 11, using LieMather model resulted in a slight ference as shown in Fig. 14.
increase in the estimated capital cost. Statistically this increase was As for the operating costs, as shown in Fig. 15 using both models
insignificant as deduced from the trajectory of both regression lines didn’t yield any statistical difference in the co-relation regression
fall within the 95% confidence band of each other. This indicates line. This validates the use of any of both models for a required
that there is no statistical difference between both the co-relation degree of sweetening interchangeably.
resulting from both regression lines which indicated that there is Other general models like NRTL (both the General and the
no statistical difference between using either of the models in the extended) and Wilson package were used to simulate the process.
calculation of the capital cost.

Fig. 4. Pareto chart of effects of the operating conditions on the total estimated cost of Fig. 5. Minimum amine flow rate required to meet the specification of sweet gas
the process. versus carbon dioxide mole fraction in the sour gas feed.
 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130 125

Fig. 6. Values of pressure drop in the control valve VLV-100 Vs CO2 mole fractions in
Fig. 8. Steam (squares), electricity (circles) and cooling water (triangles) requirements
the feed gas.
increase Vs CO2 mol fraction increase in the sour gas feed.

These models didn’t yield any satisfactory results compared to the

amine package. For example using 25 mol% CO2 sour gas all models
estimated the exit volume fraction to be around 0.22, which is so
overestimated compared to the amine package that yielded a vol-
ume fraction estimate of 0.015. Increasing the circulation rate, more
than one fold, to 400 m3/h showed only a slight decrease in the
outlet volume fraction reaching 0.2146. These results were ob-
tained irrespective of the used gas model implemented in the
package i.e. ideal model, PR or SRK. These overestimates were
explained due to the lack of experimental coefficients in the model
for some components which may have affected the final result.
Based on that no further usage of general models was used and only
the amine package was selected to further investigate the effect of
other parameters.

3.3.2. The feed gas pressure

In order to extend the analysis to include lower sour gas feed
pressures, a compressor (K-100) followed by an air cooler (AC-100)
was added in the simulation as shown in Fig. 16. The inlet pressure Fig. 9. Surface area of re-boiler (circles), lean-rich heat exchanger (diamond) and E-
to the absorber column was fixed to 35 bar. The inlet sour gas 100 cooler (squares) Vs CO2 mol fraction in the sour gas feed.

pressure was set to 10, 6 and 2 bar and the process was scaled-up to
investigate the change in the co-relation between total costs and
As shown in Fig. 17, the capital cost shown a strong linear co-
the acid gas load after this modification.
relation (with R2 > 0.97 for all cases). Also the operating costs
were still linearly increasing with the acid load (with R2 > 0.99 for
all cases) even when including the compressor and the air cooler
operational costs.

Fig. 7. Capital (circles) and operating (squares) costs estimated by Aspen Hysys versus
CO2 mole fraction in the sour gas feed. Fig. 10. Regenerator diameter Vs CO2 mol fraction in the sour gas feed.
126 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130

Fig. 11. Capital cost Vs CO2 conc. using KenteEisenberg (95% confidence interval (left)) and LieMather models (95% confidence interval (right)).

Fig. 12. Operational cost Vs CO2 conc. using KenteEisenberg (95% confidence interval (left)) and LieMather models (95% confidence interval (right)).

3.3.3. The regenerator tray efficiency The CO2 stage efficiency was set to 10, 30, 70 and 100% to study the
The stage efficiency in the amine package for the distillation effect of this parameter on cost estimates.
tower is a function of the physico-chemical properties of the amine As seen in Fig. 18, the trajectory of the four regression lines of
solution, the pressure, temperature and the mechanical tray design the capital cost at different stage efficiency fell in the 95% con-
variables such as tray diameter, weir height and weir length (A. fidence interval of one another. Consequently it was deduced
HYSYS, 2004). This factor is important and encompasses various that no statistical difference was found between all regression
operational and design parameters found in typical amine plants. lines.
A similar result was obtained for the operational costs (Fig. 19),
which clearly reflect the dominating effect of the re-boiler heating
duty and amine circulation rate on total costs as compared to the
stage efficiency value.

3.3.4. Partial pressures of CO2 and H2S

The effect of various partial pressures of CO2 and H2S e
assuming the absence of any other sulphur compounds e on the
minimum lean amine flow rate required was studied. For a total
pressure of 70 bars, the molar concentrations of both acid gases
were changed to cover a wide range of partial pressures. The total
composition was balanced by methane for a given concentration of
acid gases in the sour gas feed. As shown in Fig. 2, the minimum
DEA amine circulation rate can be fitted to a second order poly-
nomial function of both acid gases partial pressures (Equation (1)).
If the concentrations of both acid gases are close to each other, in
general a specific flow rate will remove both simultaneously. On
the other hand if one of the acid gases is present in a relatively
larger quantity, additional lean amine flow rate is required to treat
Fig. 13. Amine circulation rate Vs CO2 conc. using KenteEisenberg and LieMather this acid gas which should be studied on a case by case basis.
models. Fig. 20
 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130 127

Fig. 14. Capital cost Vs CO2 conc. using KenteEisenberg (95% confidence interval (left)) and LieMather models (95% confidence interval (right)).

Fig. 15. Operating cost Vs CO2 conc. using KenteEisenberg (95% confidence interval (left)) and LieMather models (95% confidence interval (right)).

The capital and operating costs of the process are shown in and the values of the regression coefficients along with the value of
Figs. 21 and 22 respectively. As shown any increase in either gas (R2) are shown in Table 10.
significantly affects the process economics. In general the output Y, where Y being either the min. flow rate,
In order to quantify the effect of the partial pressure of acid capital or operating costs, can be related to the partial pressures of
gases, all three surfaces were fitted to a second order polynomial both gases using Equation (1).

Fig. 16. DEA amine plant flow sheet after installing the compressor and the air cooler.
128 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130

Fig. 17. Capital (left) and operating (right) cost Vs CO2 conc. for a feed gas of 10, 6 and 2 bar.

dY
2
Y ¼ a1 þ a2 PCO2 þ a3 PCO 2
þ a4 PH2 S þ a5 PH (1) ¼ a4 þ 2*a5 *PH2 S (3)
2 2S dPH2 S
These relations show that the correlation found in Section 3.2
Therefore, differentiating Y with respect to the partial pressure can be considered as a special case of the general correlation rep-
of carbon dioxide for instance yields a linear equation as shown in resented by Equation (1) which shows that amine sweetening
Equation (2). economics scale up linearly with the increase of either acid gases
and increase according to a second order polynomial function in
dY terms of both gases.
¼ a2 þ 2*a3 *PCO2 (2)
dPCO2
4. Conclusion
Similarly differentiating Y with respect to the partial pressure of
hydrogen sulphide yields a linear equation as shown in Equation An amine sweetening process was simulated using Aspen
(3). Hysys v.8.0 in order to estimate the fixed and operating cost

Fig. 18. Capital cost Vs CO2 conc. for a tray efficiency of 10% (95% conf. int. (top left)), 30% (95% conf. int. (top right)), 70% (95% conf. int. (bottom left)) and 100% (95% confidence
interval (bottom right)).
 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130 129

Fig. 19. Operating cost Vs CO2 conc. for a tray efficiency of 10% (95% conf. int. (top left)), 30% (95% conf. int. (top right)), 70% (95% conf. int. (bottom left)) and 100% (95% confidence
interval (bottom right)).

accompanying scaling up of the process to treat high acid gas to scale-up the process and the percentage of carbon dioxide was
contents in sour natural gas. The process was optimized using the changed from 4.13% to 25%. The capital and operating costs showed
central composite design with four process variables at five levels. a strong linear relationship with the increase of CO2 mole percent.
Results showed that none of the variables have a significant effect Also all utility requirements, heat exchangers surface area and the
on the total cost. A deeper optimization study showed that in regenerator vessel increased linearly with the increase of acid gases
general the two most important variables affecting the flow sheet content. Such relation is considered as a useful reference line for
are adjusting the absorber temperature (through cooling the lean the cost penalty accompanying process scale-up and can be used as
amine) and the solvent flow rate. The amine flow rate was chosen basis of comparison for other processes or modifications aiming at
decreasing the cost of acid gas sweetening processes. The linear

Fig. 20. Min. lean amine volumetric flow rate at various CO2 and H2S partial pressures
(KPa). Fig. 21. Estimated capital cost at various CO2 and H2S partial pressures (KPa).
130 A. Muhammad, Y. GadelHak / Journal of Natural Gas Science and Engineering 17 (2014) 119e130

irrespective of the costs for purchasing and operating the inlet

compressor and air cooler. No statistical difference found between
cost regression lines of different tray efficiency of the distillation
tower reflecting the dominating effect of the re-boiler heating duty
and amine circulation rate on total costs. For various acid gas partial
pressures the capital and operating costs were fitted to a second
order polynomial function with R2 > 0.97.

Acknowledgement

The second author is grateful for Eng. Abdel Rahman Habib, Eng.
Ahmed Deyab and Eng. Reda Rashwan for their technical guidance
and support.

References

A. HYSYS, 2004. Release 2004, Simulation Basis. Aspen Technol., Cambridge, MA.
Alie, C., Douglas, P., Croiset, E., 2006. Simulation and optimization of a coal-fired
power plant with integrated CO2 capture using MEA scrubbing. In: GHGT-8
Conference.
Bergel, M., Tierno, I., 2009. Sweetening technologies e a look at the whole picture.
In: 24th World Gas Conf. Argentina, pp. 1e17.
Berstad, D., Nekså, P., Anantharaman, R., 2012. Low-temperature CO2 removal from
Fig. 22. Estimated operating cost at various CO2 and H2S partial pressures (KPa). natural Gas. Energy Procedia 26, 41e48.
Chan, C.W., Tontiwachwuthikul, P., 1996. A decision support system for solvent
selection of CO2 separation processes. Energy Conversat. Manag. 37 (6), 941e
Table 10
946.
The fitted regression coefficients and R2 values for the min. lean amine flow, capital Kidnay, A.J., Parrish, W.R., 2006. Fundamentals of Natural Gas Processing, vol. 200.
and operating costs. CRC Press.
Lallemand, F., Perdu, G., Maretto, C., Weiss, C., Magne-Drisch, J., Lucquin, A.-C., 2011.
R2 a1 a2 a3 a4 a5
Solutions for the Treatment of Highly Sour Gases. Gastech.
Min. flow 0.998126 10.69591 0.08218 0.00001 0.04844 0.00001 Lunsford, K., McIntyre, G., 1999. Decreasing contactor temperature could increase
Capital cost 0.974532 4887440 1727 0 1949 1 performance. In: Proc. Annu. Conv, pp. 1e23.
Operating cost 0.997291 1618287 2329 0 4153 1 Manning, F.S., Thompson, R.E., 1995. Oilfield Processing of Petroleum: Crude Oil, vol.
2. Pennwell Books.
Olajire, A. a., Jun. 2010. CO2 capture and separation technologies for end-of-pipe
relation was preserved with LieMather model which didn’t show applications e a review. Energy 35 (6), 2610e2628.
van Kemenade, H.P., Brouwers, J.J.H., de Rijke, S.J.M., 2013. Comparing the volume
any statistical difference compared to KenteEisenberg model. Low and energy consumption of sour-gas cleaning by condensed rotational sepa-
feed gas pressure yielded also a similar linear relationship ration and amine treatments. Energy Technol. 1 (7), 392e394.