You are on page 1of 8

Chemical Engineering Science 122 (2015) 291–298

Contents lists available at ScienceDirect

Chemical Engineering Science


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

Comparison of equilibrium and non-equilibrium models of a tray


column for post-combustion CO2 capture using DEA-promoted
potassium carbonate solution
Tohid Nejad Ghaffar Borhani a, Vahid Akbari a, Morteza Afkhamipour b, Mohd. Kamaruddin
Abd Hamid a, Zainuddin Abdul Manan a,n
a
Process Systems Engineering Center (PROSPECT), Faculty of Chemical Engineering, Universiti Teknologi Malaysia, 81310 Johor Bahru, Johor, Malaysia
b
Department of Chemical Engineering, Faculty of Gas and Petrochemical Engineering, Persian Gulf University, P.O. Box 75169-13798, Bushehr, Iran

H I G H L I G H T S

 Potassium carbonate process was simulated using Aspen Plus by equilibrium and non-equilibrium approaches.
 The two process models are compared with each other.
 The effects of mass transfer coefficients and film discretization were examined.
 The thermodynamic model was developed for potassium carbonate solution.

art ic l e i nf o a b s t r a c t

Article history: In this study, the equilibrium and non-equilibrium models for CO2 absorption from natural gas stream by
Received 24 April 2014 DEA-promoted potassium carbonate solution in a tray column were developed and compared with each
Received in revised form other. The Electrolyte NRTL thermodynamic model was utilized to calculate the activity coefficient in the
10 September 2014
liquid phase, and the SRK equation of state was used for the gas phase. The non-equilibrium model was
Accepted 11 September 2014
Available online 22 September 2014
based on the two-film theory, and the effect of film discretization was examined. The equilibrium model
was based on the theoretical number of stages combined with the concept of Murphree efficiency from
Keywords: 0.1 to 0.3 for three stages in the lower section of the column and 0.4 from stage 4 onwards.
Potassium carbonate solution A thermodynamic study was performed to describe the equilibrium behavior of the solvent. All the
CO2 capture
necessary reactions in the liquid phase were considered in all simulations. The models were validated by
Tray column
comparing the obtained results with the published experimental data. Results of absorber column
Equilibrium model
Non-equilibrium model simulation show that the non-equilibrium model gives a better prediction of the temperature and
Murphree efficiency concentration profiles as compared to the equilibrium model.
& 2014 Published by Elsevier Ltd.

1. Introduction post-combustion CO2 capture is the most attractive and common


option especially the post-combustion CO2 capture with chemical
The potentially adverse environmental, social and economic or reactive absorption. During the past decades, there have been
impacts of global warming associated with greenhouse gas (GHG) numerous efforts to develop cost-effective and energy-efficient
emissions have encouraged widespread efforts to reduce CO2 and processes for manufacturing chemical solvent (reactive absorption
other GHG emissions. The industrial sector is among the highest processes) for capturing the CO2 from various gas streams (Wang
contributors of the CO2 emissions. There are various methods et al., 2011). In contrast to the physical absorption, the reactive
for CO2 capture and storage (CCS), which are classified in three absorption does not require the use of a high pressure or high
main groups: pre-combustion CO2 capture, oxyful CO2 capture, capacity solvents. Because the chemical reactions are in the liquid
and post-combustion CO2 capture. It should be noted that the phase and the absorption of components occur simultaneously
(Kale et al., 2013).
The potassium carbonate process is an important chemical
solvent technology to reduce CO2 emissions. This process, also
n
Corresponding author. Tel.: þ60 75535502, þ 60 7 5535609; fax: þ60 7 5536165.
known as Benfield process, was developed for the synthesis of
E-mail addresses: zain@cheme.utm.my, zainmanan@gmail.com (Z.A. Manan). liquid fuel from coal (Kohl and Nielson, 1997). The process has also

http://dx.doi.org/10.1016/j.ces.2014.09.017
0009-2509/& 2014 Published by Elsevier Ltd.
292 T.N.G. Borhani et al. / Chemical Engineering Science 122 (2015) 291–298

been demonstrated to be suitable for partially selective removal of Some equilibrium and non-equilibrium process models have
hydrogen sulfide in the presence of carbon dioxide (Astarita et al., been developed and used to study the CO2 absorption in promoted
1983). This process has been applied in more than 700 plants and un-promoted potassium carbonate solutions. It should be
worldwide for carbon dioxide and hydrogen sulfide removal from noted that in some of these studies, simple thermodynamic
streams like ammonia synthesis gas, crude hydrogen, natural gas, models have been employed (Rahimpour and Kashkooli, 2004;
and town gas (Kohl and Nielson, 1997). Potassium carbonate Sanyal et al., 1988). In some other investigations, rigorous thermo-
solution is low cost, low toxic, easily regenerable, less corrosive, dynamic models such as Electrolyte NRTL model (Al-Ramdhan,
low degradable, and has high chemical solubility of CO2. However, 2001; Mumford et al., 2011; Oexmann et al., 2008; Smith et al.,
the reaction between CO2 and potassium carbonate solution is not 2012) and modified Pitzer model (Thiele et al., 2007) have been
fast that is demonstrating the slow mass transfer in the liquid utilized. Recently, Borhani et al. (in press) studied about the effects
phase (Savage et al., 1984). In addition, fouling in the form of of various promoters on the potassium carbonate solution using
precipitation of the potassium carbonate and accumulation of the rate-based non-equilibrium model.
crystals in unit operations can pose serious problems (Fosbøl et This study focuses on the absorber modeling where its perfor-
al., 2013). mance is much more dependent on accurate modeling of the
In order to improve the potassium carbonate process in terms transfer phenomena and rates. All the necessary experimental data
of reaction with CO2, the promoters (activators) have been used. were taken from the literature. First, a thermodynamic model
The promoters, which have been used for potassium carbonate using the stoichiometric and heterogeneous approaches has been
process, can be divided into two main categories, as organic established. Trustable thermodynamic solubility data of CO2 in
promoters (such as various alkanolamines) and inorganic promo- potassium carbonate solutions have been used to validate the
ters (such as arsenic trioxide, alkali metal salts of selenious or thermodynamic model. Next, the equilibrium and non-equilibrium
tellurous acid, and alkali metal salts of weak inorganic acids such models for capturing the CO2 in DEA-promoted potassium carbo-
as potassium and sodium salts of boric acid, vanadic acid, and nate solution have been implemented in Aspen Plus software
arsenious acid). Utilization of promoters for the potassium carbo- using the Electrolyte NRTL thermodynamic model and appropriate
nate solution leads to substantially lower capital and operating property models. The inlet and outlet experimental data of a real
costs and higher treated gas purity (Kohl and Nielson, 1997). industrial DEA-promoted potassium carbonate process for captur-
Numerous studies have focused on the kinetic reactions of the ing the CO2 from natural gas stream have been used for construc-
potassium carbonate solution in the presence of various promoters tion and validation of the model (Karaj Petrochemical Complex,
(Astarita et al., 1981; Cullinane and Rochelle, 2004; Knuutila et al., 2010). Furthermore, the kinetic and equilibrium reactions, hydro-
2010). dynamics (mass transfer coefficient, interfacial area, liquid holdup,
Equilibrium stage and non-equilibrium stage models are the and pressure drop), film discretization, and flow models have been
two main approaches to model the reactive absorption processes. specified and examined.
Traditionally, modeling and simulation of reactive absorption units
have been based on the well-known equilibrium stage model
(Taylor and Krishna, 1993). More complicated modeling and 2. Models development
simulation of absorption units have been established on the
non-equilibrium (rate-based) model. It should be noted that 2.1. Thermodynamic model
distillation and absorption are non-equilibrium processes. In a
real process, mass and energy are transferred across the fluid In order to reasonably explain the complex phenomena taking
interface at rates, which depend upon the extent to which the place in the CO2 capturing by the DEA-promoted potassium
phases are not in equilibrium with one another. Historically, this carbonate solution, the model must properly apply the thermo-
departure from equilibrium conditions has led to the use of dynamics of DEA–K2CO3–KHCO3–H2O–CO2 aqueous solution
efficiencies (for tray columns) and the Height Equivalent to a (liquid phase speciation and vapor–liquid equilibrium), reaction
Theoretical Plate (HETP) (for packed columns). In the equilibrium kinetics of CO2 with the aqueous solution, and the various
model, vapor and liquid are assumed to enter a tray or cross chemical and physical properties affecting the mass and heat
section of packing in a column, exchange matter and energy, and transfer (Zhang et al., 2009).
leave in equilibrium with each other. Thermodynamic equilibrium The term speciation is used to describe what happens when an
is assumed to exist between the vapor and liquid streams leaving electrolyte solution is dissolved into water. The composition of the
each stage (Seader et al., 2010). In rate-based model, it is assumed liquid solution can be estimated using the equations of speciation
that mechanical, chemical, and thermodynamic equilibrium equilibria. A speciation equilibrium calculation can be implemen-
exist only at the fluid interface (Taylor and Krishna, 1993). In this ted alone or simultaneously with the vapor–liquid equilibrium
approach, in addition to the equations related to equilibrium calculations. After calculating the mole fraction of each species in
modeling, the mass and heat transfer rate equations are solved. the liquid phase using the speciation calculation, the vapor phase
This model totally avoids the approximation of efficiency composition for molecular solute CO2 and solvents H2O and DEA
(Krishnamurthy and Taylor, 1986). There are various rate-based can be obtained from the equilibrium between vapor and liquid. In
models from simple type to complicated (rigorous) one. In the order to calculate the vapor–liquid equilibrium there are two main
simplest non-equilibrium rate-based model, the reaction kinetics approaches: homogeneous approach (Fürst and Renon, 1993) and
in the bulk solution and enhancement factors are considered. In heterogeneous approach (Al-Rashed and Ali, 2012; Barreau et al.,
the most rigorous non-equilibrium rate-based model, the reaction 2006; Cullinane and Rochelle, 2005; Faramarzi et al., 2009). In the
kinetics, film discretization with several segments, and electrolyte homogenous approach, a single equation is used to determine the
thermodynamics are considered (Kenig et al., 2001). In the rate- properties of both liquid and vapor phases. In the heterogeneous
based approach, in addition to the main equations, the mass and approach, which is also well known as gamma–phi approach, an
heat transfer rates, transfer coefficients, hydrodynamics equations, activity coefficient model and an equation of state are utilized to
phase equilibria, and physical properties models must be used. By represent the liquid phase and the vapor phase, respectively. In
using these models, in addition to having reliable results of the this study, the stoichiometric and heterogeneous approaches have
column design, the process mechanism needs to be fully under- been used to perform the speciation and vapor–liquid equilibrium.
stood (Kale et al., 2013). The Electrolyte NRTL thermodynamic model and SRK equation of
T.N.G. Borhani et al. / Chemical Engineering Science 122 (2015) 291–298 293

state have been considered in both thermodynamic and process  The data sets reported for CO2 solubility in terms of CO2 partial
models for the liquid and gas phases, respectively. pressure as a function of fractional conversion (loading) of
In the system under study the liquid phase composes of CO2, K2CO3 to KHCO3 and temperature (Tosh et al., 1959) over the
DEA, H2O, K2CO3, and KHCO3. Therefore, nine species must be KHCO3–H2O–CO2 solution.
considered, which are K þ , HCO3  , CO3 2  , H þ , OH  , CO2, DEA,
DEAH þ , DEACOO  , and H2O. The following chemical reactions
The vapor–liquid equilibrium and liquid phase speciation of the
occur in the DEA–K2CO3–KHCO3–H2O–CO2 aqueous solution and
DEA–K2CO3–KHCO3–H2O–CO2 aqueous solution are calculated
show the effects of potassium carbonate and DEA in absorption of
using a flash model in Aspen Plus software. In the thermodynamic,
CO2. These reactions must be considered for the thermodynamic
equilibrium, and non-equilibrium models presented in this study,
model:
the thermodynamic properties such as fugacity coefficient,
2H2 O⇌OH  þ H3 O þ ð1Þ enthalpy, entropy, free Gibbs energy, and volume were calculated
using appropriate thermodynamic methods. On the other hand,
CO2 þ 2H2 O⇌HCO3  þ H3 O þ ð2Þ the transport properties such as viscosity, thermal conductivity,
diffusion coefficient, and surface tension were calculated using the
HCO3  þH2 O⇌CO3 2  þH3 O þ ð3Þ physical property models (Aspen Technology, 2001a).

DEAH þ þ H2 O⇌DEA þ H3 O þ ð4Þ


2.2. Process model
DEACOO  þ H2 O⇌DEA þ HCO3  ð5Þ
Accurate and detailed process models can contribute to the
Reactions (1)–(3) are related to the potassium carbonate effects,
evaluation of different operating conditions for process optimiza-
and reactions (4) and (5) signify the effect of DEA on the solution.
tion purposes without the operation of the actual plant. These
It should be noted that reactions (1)–(5) are equilibrium reactions.
models can also save time and human resources since process
The equilibrium constants can be achieved using two methods.
models can be relatively quick and easy to perform compared to
The first method is to calculate it from the standard Gibbs free
actual plant tests. Obviously, the appropriate modeling of reactive
energy change, and the second one is to obtain the constants by
absorption column depends on the proper selection of column
regressing them against the experimental data of the related
internals, sufficient knowledge of the process behavior, and details
literature, in the form of equilibrium constant as a function of
about the design of the column. In this study, the Aspen Plus V7.2
temperature (Austgen, 1989). The parameters of equilibrium con-
was used to simulate the process since this process simulator has
stant have been extracted from the literature (Austgen, 1989;
suitable thermodynamic model and unit operations to simulate
Austgen et al., 1991).
the processes related to electrolyte solutions like potassium
The natural gas stream considered in this study contain CO2,
carbonate solution. In this work, RateFrac (the rate-based non-
H2O, CH4, C2H6, C3H8, nC4H10, iC4H10, nC5H12, iC5H12, nC6H14, and
equilibrium model for simulating multistage vapor–liquid fractio-
N2. Therefore, all the components except water are selected as
nation operations in Aspen Plus) was utilized as absorber of the
Henry-components. The Henry's constants were retrieved from
process (Aspen Technology, 2001b). The chemical kinetics, hydro-
the Aspen Plus databanks for these components with water, except
dynamics (mass transfer coefficient, interfacial area, liquid holdup,
for those for nC5H12 and nC6H14 with H2O which are taken from
and pressure drop), column properties, film discretization, and
NIST web database. SRK is an equation of state (EOS) which can
flow models have been specified in the model (Aspen Technology,
calculate fugacity coefficients for the vapor phase (Aspen
2001c).
Technology, 2001a). Electrolyte NRTL is a generalized excess Gibbs
In RateFrac the mass and heat transfer resistances in the liquid
energy model that accounts the interactions between all true
and gas phases have been considered based on two-film theory,
liquid phase species. This model is also well known as an activity
and the liquid film at each stage has been discretized into several
coefficient thermodynamic model, which can calculate activity
non-homogeneous segments with various thicknesses (Qi et al.,
coefficients, enthalpies, equilibrium vapor pressure of CO2 and
2013; Zhang et al., 2009). By using the film discretization, the
H2O, solution speciation, and Gibbs energies. The Electrolyte NRTL
accurate modeling of the chemical reactions and consequently the
thermodynamic model has three contribution terms of excess
more accurate speciation profiles in the liquid film is obtained. It is
Gibbs energy, which are Pitzer–Debye–Hückel term to account
an important model parameter in order to improve the model's
the long-range ion-ion interactions, the born term to consider the
predictive power (Asprion, 2006; Kale et al., 2013). The concept of
effect of mixed solvent i.e., the difference in Gibbs energies
film discretization in the liquid film layer is illustrated in Fig. 1.
between ionic species in a mixed solvent and in water, and the
NRTL term to reflect the local interactions. To ensure an accurate
Vapor Bulk Vapor Film Liquid Film Liquid Bulk
representation of the liquid phase thermodynamics for the DEA-
Y
promoted potassium carbonate solution, the adjustable para- DEACOO
OH
meters of Electrolyte NRTL thermodynamic model were deter-
T H O
mined by the data regression system (DRS) in Aspen Plus using the Y
following binary and ternary VLE data: CO HCO

T CO
 The data sets reported for CO2 solubility (Maddox et al., 1987; H O

Maddox and Elizondo, 1989) over the DEA–H2O–CO2 solution. X H O

 The data sets reported for mean ionic activity coefficient DEA

(Hilliard, 2005), water vapor pressure depression (Aseyev, T CO


1999; Puchkov and Kurochkina, 1970), and heat capacity of K
X
solution (Hilliard, 2005) over the K2CO3–H2O solution.
 The data sets reported for vapor pressure depression (Aseyev, Interface

1999) and heat capacity of solution (Hilliard, 2005) over the Fig. 1. The concept of two-film theory with film discretization and counter-current
KHCO3–H2O solution. flow model.
294 T.N.G. Borhani et al. / Chemical Engineering Science 122 (2015) 291–298

For the process modeling, two sets of reactions are considered: considered in this section. The non-ideal behavior of stages is
one set is the equilibrium reactions (1)–(5) described above, and accounted using the Murphree efficiency. Different values of
the other one is the kinetic reactions between DEA and CO2 and Murphree efficiency from 0.1 to 0.3 have been selected for three
the reactions of CO2 with OH  . Reactions (6) and (7) are the stages in the lower section of the column and 0.4 has been
forward and backward reactions for bicarbonate formation, and considered from stage 4 onwards (Afkhamipour and Mofarahi,
reactions (8) and (9) are the forward and backward reactions of 2013).
DEA carbamate formation.
CO2 þ OH  -HCO3  ð6Þ
3. Results and discussion
HCO3  -CO2 þ OH  ð7Þ
3.1. Thermodynamic model validation and results
DEA þ CO2 þ H2 O-DEACOO  þ H3 O þ ð8Þ
As mentioned before, the thermodynamic model has been
DEACOO  þ H3 O þ -DEA þCO2 þ H2 O ð9Þ implemented using a flash calculation in Aspen Plus. For the
thermodynamic model developed in this research, the deviation
The following power law form is used as the kinetic reactions rate:
 n    N between the experimental data and simulated values was calcu-
T Ej 1 1  α lated by using average absolute relative deviation (%AARD):
r j ¼ kj exp   ∏ xi γ i i ð10Þ
T0 R T T0
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} i ¼ 1 Sim Exp
K j ¼ reaction rate constants 100 N Y i  Y i
%AARD ¼ ∑ ð11Þ
N i Y Exp
i
where r j is the reaction rate of reaction j, kj is the pre-exponential
factor for reaction j, T is the absolute temperature of the system, T0 where N is the number of process variables and Y Sim
i and Y Exp
i are
is the reference temperature, Ej is the activation energy, n is the the simulated values and experimental data for component i,
temperature exponent, and R is the gas law constant. In this respectively.
equation, the concentration basis is Molarity. E and k which are According to our literature review, there is not any trustable
derived from Pinsent et al. (1956) and Rinker et al. (1996) are solubility data for DEA–K2CO3–KHCO3–H2O–CO2 solution. How-
illustrated in Table 1. ever, there are some appropriate solubility data for other pro-
A number of correlations are available to calculate the mass moted and un-promoted potassium carbonate solutions as
transfer coefficients and effective interfacial area in tray columns. follows:
These correlations vary in their accuracies, limitations, and some-
times system specific applicability. The performances of gas and  PZ–K2CO3–KHCO3–H2O–CO2 solution (Cullinane and Rochelle,
liquid mass transfer coefficients presented by Grester et al. (1958) 2005; Hilliard, 2005);
and Scheffe and Weiland (1987) compared to each other. The  PZ–K2CO3–H2O–CO2 and 2MPZ–K2CO3–H2O–CO2 solutions
correlation of Scheffe and Weiland (1987) has been used for (Kim et al., 2012);
interfacial area calculations. The Bennett et al.'s (1983) technique  MEA–PZ–K2CO3–KHCO3–H2O–CO2 solution (Hilliard, 2008);
was employed as the holdup method. The Chilton–Colburn corre-  BA (boric acid)–K2CO3–KHCO3–H2O–CO2 solution and 30 wt%
lation was used to calculate the heat transfer in the absorber of K2CO3–H2O–CO2 solution at 50 and 70 1C (Endo et al., 2011);
(Aspen Technology, 2001b).  K2CO3–H2O–CO2 solutions in 5 and 10 wt% of potassium carbo-
There are four different flow models in the RateFrac, which nate at 25 and 50 1C (Park et al., 1997);
determine the bulk properties required to evaluate the mass and  K2CO39H2O–CO2 solutions (Pérez-Salado Kamps et al., 2007);
energy fluxes and reaction rates. The four flow models are Mixed,  K2CO3–KHCO3–CO2–H2O solutions in 20, 30, and 40 equivalent
Counter-Current, VPlug, and VPlug-Pavg (Aspen Technology, concentration of potassium carbonate at temperatures 70, 90,
2001b). Zhang et al. (2009) illustrated that flow models only have 110, 130 1C (Tosh et al., 1959).
minor impacts on the predicted absorber temperature profiles,
and the mixed flow model has the most reliable results. Thus, the In this study, it is obvious that the mentioned promoted
mixed flow model is chosen. The pressure drop in the packed potassium carbonate solubility data cannot be used because any
columns can be calculated by using various models; but, for tray of the promoters have their effects on the solution. Therefore, only
columns, the user only can fix the pressure at the top or bottom of solubility data of un-promoted potassium carbonate solution have
the column. The film discretization option is set to geometric, and been used to validate the presented thermodynamic model.
the film discretization ration is set to 2.5. The absorber is a tray The predicted CO2 partial pressures and experimental values
column with 4-m diameter and 13 trays. The Weir height is for K2CO3–KHCO3–CO2–H2O solutions are compared in Figs. 2 and
60 mm, and the column pressure is set to 6000 kPa. The tray type 3, respectively. In Fig. 2, the CO2 solubility data in 20 wt%
is Glitsch Ballast. The stage number for liquid stream and gas equivalent concentration of potassium carbonate solution at four
stream iterance numbers are 13 and 1, respectively. different temperatures were used (Tosh et al., 1959). In Fig. 3, the
The equilibrium model is achieved by switching the modeling CO2 solubility data in 30 wt% potassium carbonate solution was
approach option from rate-based (non-equilibrium) to equilibrium utilized (Endo et al., 2011). The average absolute relative deviation
in the RateFrac model. Similar to rate-based model, 13 stages are between the experimental data and the calculated amounts CO2
solubility in Figs. 2 and 3 were 10.7 and 7.4, respectively. The
Table 1 results for Figs. 2 and 3 were satisfactory since the data have been
Parameters k, n, and E in the kinetic reactions rate.
presented by Tosh et al. (1959) and Endo et al. (2011) was
Reaction k E (cal/mol) n Reference trustable, not scatter, and shows appropriate consistency. In
general, there is a good agreement between the model predictions
(6) 4.315  10 þ 13 13,249 0 Pinsent et al. (1956) and the experimental values.
(7) 2.38  10 þ 17 29,451 0 Pinsent et al. (1956) The speciation calculation for the DEA–K2CO3–KHCO3–H2O–
(8) 6.48  10 þ 16 5072 0 Rinker et al. (1996)
(9) 1.43  10 þ 17 11,497 0 Rinker et al. (1996)
CO2 has been done at four different temperatures by using three
various DEA concentrations. The calculated values are summarized
T.N.G. Borhani et al. / Chemical Engineering Science 122 (2015) 291–298 295

1000 6

323 K
343 K
323 K ENRTL
5 343 K ENRTL
100
CO2 Partial Pressure (kPa)

CO2 Partial Pressure


10

3
343 K
363 K
383 K
403 K
1 343 K ENRTL
2
363 K ENRTL
383 K ENRTL
403 K ENRTL

1
0.1
0 0.2 0.4 0.6 0.8 1
Loading (mol CO2/mol K2CO3)

Fig. 2. CO2 partial pressure in 20 wt% equivalent concentration of potassium


carbonate solution at four different temperatures; points: (Tosh et al., 1959); line:
Electrolyte NRTL. 0
0 0.1 0.2 0.3 0.4 0.5
Loading
in Table 2. The concentration unit of all species in this table is
Fig. 3. CO2 partial pressure in 30 wt% potassium carbonate solution at two different
kmol/m3. However, there was not any experimental data for temperatures; points: (Endo et al., 2011); line: Electrolyte NRTL.
speciation of the mentioned solution but the results of speciation
for DEA–K2CO3–KHCO3–H2O–CO2 solution were in relative agree-
Table 2
ment with the results of Thee et al. (2012) for speciation calcula-
The speciation of DEA–K2CO3–KHCO3–H2O–CO2 system predicted by Aspen Plus.
tion of MEA–K2CO3–KHCO3–H2O–CO2 solution.
Temperature ½DEAadded ½DEAfree ½DEAH þ  ½DEACOO   ½HCO3   ½CO3 2  
3.2. Process model validation and results (1C)

50 0.4 0.15112 0.00025 0.21400 0.84500 2.03000


In this study, for the non-equilibrium (rate-based) model, in
0.8 0.33446 0.00045 0.37962 0.65218 1.97415
addition to the equilibrium reactions (1)–(5), the kinetic reactions 1.2 0.60648 0.00060 0.52898 0.46946 1.90897
(6)–(9) have been considered. The equilibrium reactions (1)–(5)
70 0.4 0.20267 0.00036 0.15873 0.88997 2.00343
and the Murphree efficiencies have been employed for equilibrium 0.8 0.42004 0.00066 0.28567 0.73548 1.95043
stage model. The industrial data from Karaj Petrochemical Com- 1.2 0.71248 0.00093 0.40936 0.57826 1.88572
plex were used to construct the model. In this industrial data, the 90 0.4 0.24388 0.00039 0.11270 0.92443 1.97450
natural gas stream contain the carbon dioxide, water, nitrogen, 0.8 0.48857 0.00073 0.20750 0.80191 1.92186
and alkanes from methane to normal hexane and as it mentioned 1.2 0.79975 0.00108 0.30647 0.66932 1.85791
before, the liquid solution stream contain diethanolamine, potas- 110 0.4 0.27290 0.00035 0.07829 0.94695 1.94045
sium carbonate, and water. The pressure and temperature of the 0.8 0.53781 0.00068 0.14760 0.84978 1.88860
process can be found in Table 3. The non-equilibrium and 1.2 0.86472 0.00104 0.22435 0.73933 1.82579
equilibrium models specifications and simulation results relying
on the industrial data (Karaj Petrochemical Complex, 2010) are
summarized in Table 3. was performed in this study. It was found that when the Grester
As can be seen, Table 3 shows the non-equilibrium model with et al. (1958) mass transfer model is used in the study most
Grester et al. (1958) mass transfer coefficient which gives a agreement between the industrial outlet gas stream and simulated
satisfactory CO2 amount in the treated gas stream because the outlet gas stream is obtained. The average absolute deviation
non-equilibrium model considers the mass and heat transfers in between the calculated amount of CO2 in the treated gas using
the process. In this study the inlet and outlet real industrial data of Grester et al. (1958) mass transfer coefficient model and experi-
a tray column has been considered. Therefore, the inlet data was mental values is 0.9268. The AAD% between the experimental data
used to construct the model. The outlet stream obtained by Aspen and calculated values for Scheffe and Weiland (1987) model is
Plus was compared with the industrial outlet gas stream. There- around 8.7155. The result shows the importance of selecting
fore it can be said that the validation in the outlet of the system proper mass transfer coefficient model in the rate-base modeling
296 T.N.G. Borhani et al. / Chemical Engineering Science 122 (2015) 291–298

Table 3
Summary of non-equilibrium and equilibrium model predictions and the experimental conditions (Karaj Petrochemical Complex, 2010).

Natural gas stream temperature (K) 363.15


Lean liquid stream temperature (K) 380.15
Exp. treated gas stream temperature (K) 377.15
Calc. treated gas stream temperature (K) 382.78
Absorber operation pressure (kPa) 6000
Natural gas stream flow (kmol/h) 16716.79
Mole fraction of CO2 0.07994
Mole fraction of H2O 0.00070
Lean liquid stream flow (kmol/h) 45797.26
Mole fraction of K2CO3 0.04038
Mole fraction of KHCO3 0.02104
Mole fraction of DEA 0.00728
Mole fraction of KVO3 0.00053
Mole fraction of H2O 0.93077
Exp. treated gas flow (kmol/h) 15377.28
Calc. treated gas flow (kmol/h) 15800.04
Exp. CO2 in treated gas stream (mole fraction) 0.00196
Calc. CO2 in treated gas stream (equilibrium with ME ¼0.1) 0.00399
Calc. CO2 in treated gas stream (equilibrium with ME ¼0.2) 0.00313
Calc. CO2 in treated gas stream (equilibrium with ME ¼0.3) 0.00250
Calc. CO2 in treated gas stream (non-equilibrium Grester et al.) 0.00197
Calc. CO2 in treated gas stream (non-equilibrium Scheffe and Weiland) 0.00213

of the absorber column. The AAD% for the equilibrium model is 0.09
E.M. with M.E.=0.1
more than 27.6 showing a significant difference between the 0.08 E.M. with M.E.=0.2
CO2 mole fraction in gas stream

equilibrium and non-equilibrium approaches. It should be noted E.M. with M.E.=0.3


0.07
that the equilibrium model over-predicted the CO2 amount in the R.B.M with Scheffe & Weiland
0.06
treated gas stream considerably since in the equilibrium model it R.B.M. with Grester et al.

is assumed that the liquid and gas phases are at equilibrium. 0.05 Experimental

Figs. 4 and 5 show a comparison between the predicted results 0.04


of non-equilibrium and equilibrium models. Unfortunately, the
0.03
experimental data from Karaj Petrochemical Complex only include
the data of inlet and outlet streams (the data along the column 0.02

were not available). Profiles in these figures reveal the changes of 0.01
CO2 gas concentration and temperature of the liquid phase along
0
the tray column. It should be noted that in these figures E.M. is 1 2 3 4 5 6 7 8 9 10 11 12 13
equilibrium model, ME is Murphree efficiency, and R.B.M. is rate- Stage Number
based model. Fig. 4. CO2 concentration profile in the natural gas stream along the column.
In Figs. 4 and 5, the bigger bulges can be seen in profiles for
non-equilibrium rate-based model in comparison with the equili-
brium model in the middle of column. These bulges reveal good 415
predictions of model (Zhang et al., 2009). 410
The profiles achieved by the Grester et al.'s (1958) mass
405
transfer coefficient model show the biggest bulge even bigger
Temperature K

than the predictions obtained by Scheffe and Weiland (1987) mass 400
transfer coefficient model. In equilibrium model the best results
395
were obtained by using the Murphree efficiency value of 0.3 for E.M. with M.E.=0.1

five stages in the bottom of the column and fixed value of 0.4 for 390
E.M. with M.E.=0.2

E.M. with M.E.=0.3


other stages of the column. There are various exothermic chemical
385 R.B.M. with Scheffe & Weiland
reactions during the absorption of CO2 in the liquid phase. These
R.B.M. with Grester et al.
reactions cause the high absorption of CO2 at the bottom of the 380
Exprimental
column. Therefore, various small amounts of Murphree efficiencies 375
are used in the bottom of the column, and selection of Murphree 1 2 3 4 5 6 7 8 9 10 11 12 13
efficiency has a great significance in the predictions of the Stage Number
equilibrium model. As can be seen in Fig. 4, most of the CO2 Fig. 5. Temperature profile of liquid phase along the column.
captured in the middle stages of the tray column and final stages
of the column removed small amount of CO2. In general, in aspect
of final CO2 amount in the treated gas which leaves the top of the Fig. 8 illustrates the profile of mole fraction of water in the gas
column, the equilibrium model has acceptable predictions but phase during the column. As can be seen, at the bottom of the
non-equilibrium model has better result. column some amount of water evaporates from the liquid solution
In Figs. 6–8, profiles of liquid flow rate, gas flow rate, and mole and moves to the gas phase. A portion of it condenses along the
fraction of water have been predicted along the column using the column, and the remained amount leaves the column with treated
non-equilibrium and equilibrium models, respectively. Figs. 6 and gas from the top. The evaporation of the water is due to the high
7 depict the profiles of liquid and gas flow rate. CO2 is absorbed by temperature of the inlet gas and the heat of absorption between
DEA-promoted potassium carbonate solution as the gas goes to the the CO2 and liquid solution. Moreover, in the non-equilibrium rate
top of the column encountering the flow rate reduction. based models, higher amount of water remains in the gas phase
T.N.G. Borhani et al. / Chemical Engineering Science 122 (2015) 291–298 297

14.25 DEA–K2CO3–KHCO3–H2O–CO2 has been calculated using the ther-


14.2 modynamic model. The best results for the process model have
Liquid Flow Rate (kmol/s)

been obtained by the non-equilibrium rate-based approach and by


14.15
using the Grester et al.'s (1958) mass transfer coefficient model
14.1
with AAD% equal to 0.92683. The results of equilibrium models
14.05 show considerable deviation from the experimental data. In
14
E.M. with M.E.=0.1
general, it can be concluded that by having the accurate and
E.M. with M.E.=0.2
E.M. with M.E.=0.3
suitable model for potassium carbonate systems the detailed
13.95
R.B.M. with Scheffe & Weiland analysis and improvement of the process is achievable.
13.9 R.B.M. with Grester et al.

13.85
1 2 3 4 5 6 7 8 9 10 11 12 13
Stage Number Nomenclature

Fig. 6. Liquid flow rate profile along the tray column.


AARD average absolute relative deviation
DEA di ethanol amine
5
DEAH þ protonated DEA
E.M. with M.E.=0.1 DEACOO  DEA carbamate ion
4.9 E.M. with M.E.=0.2 Electrolyte NRTL electrolyte nonrandom two liquid
Gas Flow Rate (kmol/s)

E.M. with M.E.=0.3


4.8 R.B.M. with Scheffe & Weiland
E activation energy
R.B.M. with Grester et al. E.M. equilibrium model
4.7
k pre-exponential factor
4.6 K equilibrium constant
M.E. Murphree efficiency
4.5
n temperature exponent
4.4 r rate of reaction
4.3
R universal gas constant
1 2 3 4 5 6 7 8 9 10 11 12 13 R.B.M. rate-base model
Stage Number T absolute temperature
Fig. 7. Gas flow rate profile along the tray column.

0.07
Acknowledgments
0.06
Water Mole Fraction

0.05 Financial support of Universiti Teknologi Malaysia under Grant


no. Q.J13000.2444.00G52 is gratefully acknowledged.
0.04

0.03
E.M. with M.E.=0.1 References
0.02 E.M. with M.E.=0.2
E.M. with M.E.=0.3
R.B.M. with Scheffe & Weiland Afkhamipour, M., Mofarahi, M., 2013. Comparison of rate-based and equilibrium-
0.01
R.B.M. with Grester et al. stage models of a packed column for post-combustion CO2 capture using
2-amino-2-methyl-1-propanol (AMP) solution. Int. J. Greenh. Gas Control 15,
0
186–199.
1 2 3 4 5 6 7 8 9 10 11 12 13
Al-Ramdhan, H.A., 2001. A Rate-Based Model for the Design and Simulation of a
Stage Number Carbon Dioxide Absorber Using the Hot Potassium Carbonate Process. Colorado
School of Mines, Department of Chemical Engineering Petroleum, Refining,
Fig. 8. Water mole fraction profile along the tray column.
Colorado School of Mines, Golden, CO.
Al-Rashed, O.A., Ali, S.H., 2012. Modeling the solubility of CO2 and H2S in DEA–
due to the mass and energy transfer relations considered in these MDEA alkanolamine solutions using the electrolyte–UNIQUAC model. Sep.
Purif. Technol. 94, 71–83.
models. Aseyev, G.G., 1999. Electrolytes, Equilibria in Solutions and Phase Equilibria.
Calculation of Multicomponent Systems and Experimental Data on the Activ-
ities of Water, Vapor Pressures, and Osmotic Coefficients. Begell House, New
York.
4. Conclusion and remarks Aspen Technology, 2001. Aspen Plus Reference Manual. Aspen Physical Property
System. Physical Property Methods and Models 11.1. Cambridge, MA, USA.
In this study, a thermodynamic model has been proposed to Aspen Technology, 2001. Aspen Plus Reference Manual. Aspen Plus 11.1 Unit
Operation Models. Cambridge, MA, USA.
represent the speciation and vapor–liquid equilibrium for the Aspen Technology, 2001. Aspen Plus Reference Manual. Aspen Plus 11.1 User Guide.
DEA-promoted potassium solution. In addition, non-equilibrium Cambridge, MA, USA.
and equilibrium models have been developed to simulate the CO2 Asprion, N., 2006. Nonequilibrium rate-based simulation of reactive systems:
simulation model, heat transfer, and influence of film discretization. Ind. Eng.
absorption by DEA-promoted potassium carbonate solution in a Chem. Res. 45, 2054–2069.
tray column. In the non-equilibrium model, two different correla- Astarita, G., Savage, D.W., Bisio, A., 1983. Gas Treating with Chemical Solvents. John
tions have been used to calculate the mass transfer coefficients in Wiley & Sons, NY.
Astarita, G., Savage, D.W., Longo, J.M., 1981. Promotion of CO2 mass transfer in
the gas and liquid phases and effective interfacial area in the tray
carbonate solutions. Chem. Eng. Sci. 36, 8.
column. For the equilibrium model various Murphree efficiencies Austgen, D.M., 1989. A Model of Vapor–Liquid Equilibria for Acid Gas–Alkanola-
have been considered. mine–Water Systems. University of Texas, Austin.
For the thermodynamic model, the prediction accuracy was Austgen, D.M., Rochelle, G.T., Chen, C.C., 1991. Model of vapor–liquid equilibria for
aqueous acid gas-alkanolamine systems. 2. Representation of hydrogen sulfide
acceptable for K2CO3–KHCO3–H2O–CO2 solutions. The values and carbon dioxide solubility in aqueous MDEA and carbon dioxide solubility in
of AARD% were 10.7 and 7.4. In addition, the speciation of the aqueous mixtures of MDEA with MEA or DEA. Ind. Eng. Chem. Res. 30, 543–555.
298 T.N.G. Borhani et al. / Chemical Engineering Science 122 (2015) 291–298

Barreau, A., Blanchon, l.B.E., Habchi, T.K.N., Mougin, P., Lecomte, F., 2006. Absorption combustion capture of CO2: results from the solvent absorption capture plant at
of H2S and CO2 in alkanolamine aqueous solution: experimental data and hazelwood power station using potassium carbonate solvent. Energy Fuels 26,
modelling with the electrolyte-NRTL model. Oil Gas Sci. Technol. 61, 17. 138–146.
Bennett, D.L., Agrawal, R., Cook, P.J., 1983. New pressure drop correlation for sieve Oexmann, J., Hensel, C., Kather, A., 2008. Post-combustion CO2-capture from coal-
tray distillation columns. AIChE J. 29, 434–442. fired power plants: preliminary evaluation of an integrated chemical absorp-
Borhani, T.N.G., Akbari, V., Hamid, M.K.A., Manan, Z.A., Rate-based simulation and tion process with piperazine-promoted potassium carbonate. Int. J. Greenh. Gas
comparison of various promoters for CO2 capture in industrial DEA-promoted Control 2, 539–552.
potassium carbonate absorption unit. J. Ind. Eng. Chem., http://dx.doi.org/10. Park, S.B., Shim, C.S., Lee, H., Lee, K.H., 1997. Solubilities of carbon dioxide in the
1016/j.jiec.2014.07.024, in press. aqueous potassium carbonate and potassium carbonate—poly(ethylene glycol)
Cullinane, J.T., Rochelle, G.T., 2004. Carbon dioxide absorption with aqueous solutions. Fluid Phase Equilib. 134, 141–149.
potassium carbonate promoted by piperazine. Chem. Eng. Sci. 59, 3619–3630. Pérez-Salado Kamps, Á., Meyer, E., Rumpf, B., Maurer, G., 2007. Solubility of CO2 in
Cullinane, J.T., Rochelle, G.T., 2005. Thermodynamics of aqueous potassium carbo- aqueous solutions of KCl and in aqueous solutions of K2CO3. J. Chem. Eng. Data
nate, piperazine, and carbon dioxide. Fluid Phase Equilib. 227, 197–213. 52, 817–832.
Endo, K., Nguyen, Q.S., Kentish, S.E., Stevens, G.W., 2011. The effect of boric acid on Pinsent, B.R.W., Pearson, L., Roughton, F.J.W., 1956. The kinetics of combination of
the vapour liquid equilibrium of aqueous potassium carbonate. Fluid Phase carbon dioxide with hydroxide ions. Trans. Faraday Soc. 52, 1512–1520.
Equilib. 309, 109–113. Puchkov, L.V., Kurochkina, V.V., 1970. Saturated vapor pressure over aqueous
Faramarzi, L., Kontogeorgis, G.M., Thomsen, K., Stenby, E.H., 2009. Extended solutions of potassium carbonate. Zhur. Priklad. Khim. 43, 181–183.
UNIQUAC model for thermodynamic modeling of CO2 absorption in aqueous Qi, G., Wang, S., Yu, H., Wardhaugh, L., Feron, P., Chen, C., 2013. Development of a
alkanolamine solutions. Fluid Phase Equilib. 282, 121–132. rate-based model for CO2 absorption using aqueous NH3 in a packed column.
Fosbøl, P.L., Maribo-Mogensen, B., Thomsen, K., 2013. Solids modelling and capture Int. J. Greenh. Gas Control 17, 450–461.
simulation of piperazine in potassium solvents. Energy Procedia 37, 844–859. Rahimpour, M.R., Kashkooli, A.Z., 2004. Enhanced carbon dioxide removal by
Fürst, W., Renon, H., 1993. Representation of excess properties of electrolyte promoted hot potassium carbonate in a split-flow absorber. Chem. Eng.
solutions using a new equation of state. AIChE J. 39, 335–343. Process.: Process Intensif. 43, 857.
Grester, J.A., Hill, A.B., Hochgraf, N.N., Robinson, D.G., 1958. Tray Efficiencies in Rinker, E.B., Ashour, S.S., Sandall, O.C., 1996. Kinetics and modeling of carbon
Distillation Columns. In: Report, A. (Ed.), AIChE Report. dioxide absorption into aqueous solutions of diethanolamine. Ind. Eng. Chem.
Hilliard, M., 2005. Thermodynamics of Aqueous Piperazine/Potassium Carbonate/ Res. 35, 1107–1114.
Carbon Dioxide Characterized by the Electrolyte NRTL Model within Aspen Sanyal, D., Vasishtha, N., Saraf, D.N., 1988. Modeling of carbon dioxide absorber
Plus, Technical Report. University of Texas at Austin, Texas p. 251. using hot carbonate process. Ind. Eng. Chem. Res. 27, 2149–2156.
Hilliard, M., 2008. A Predictive Thermodynamic Model for an Aqueous Blend of Savage, D.W., Sartori, G., Astarita, G., 1984. Amines as rate promoters for carbon
Potassium Carbonate, Piperazine, and Monoethanolamine for Carbon Dioxide dioxide hydrolysis. Faraday Discuss. Chem. Soc. 77, 17–31.
Capture from Flue Gas. The University of Texas, Austin p. 1061. Scheffe, R.D., Weiland, R.H., 1987. Mass-transfer characteristics of valve trays. Ind.
Kale, C., Górak, A., Schoenmakers, H., 2013. Modelling of the reactive absorption of Eng. Chem. Res. 26, 228–236.
CO2 using mono-ethanolamine. Int. J. Greenh. Gas Control 17, 294–308. Seader, J.D., Henley, E.J. et al., 2010. Separation Process Principles. New York, John
Karaj Petrochemical Complex, 2010. Natural Gas Plant. Operating Data of Potassium Wiley & Sons.
Carbonate Process. Smith, K., Anderson, C.J., Tao, W., Endo, K., Mumford, K.A., Kentish, S.E., Qader, A.,
Kenig, E.Y., Schneider, R., Górak, A., 2001. Reactive absorption: optimal process Hooper, B., Stevens, G.W., 2012. Pre-combustion capture of CO2—results from
design via optimal modelling. Chem. Eng. Sci. 56, 343–350. solvent absorption pilot plant trials using 30 wt% potassium carbonate and
Kim, Y.E., Choi, J.H., Nam, S.C., Yoon, Y.I., 2012. CO2 absorption capacity using boric acid promoted potassium carbonate solvent. Int. J. Greenh. Gas Control 10,
aqueous potassium carbonate with 2-methylpiperazine and piperazine. J. Ind. 64–73.
Eng. Chem. 18, 105–110. Taylor, R., Krishna, R., 1993. Multicomponent mass transfer. New York, John Wiley
Knuutila, H., Juliussen, O., Svendsen, H.F., 2010. Kinetics of the reaction of carbon and Sons.
dioxide with aqueous sodium and potassium carbonate solutions. Chem. Eng. Thee, Suryaputradinata, Y.A., Mumford, K.A., Smith, K.H., Silva, G.d., Kentish, S.E.,
Sci. 65, 6077–6088. Stevens, G.W., 2012. A kinetic and process modeling study of CO2 capture with
Kohl, A., Nielson, R., 1997. Gas Purification. Gulf Publishing Company, Houston, MEA-promoted potassium carbonate solutions. Chem. Eng. J. 210, 271–279.
Texas. Thiele, R., Faber, R., Repke, J.U., Thielert, H., Wozny, G., 2007. Design of industrial
Krishnamurthy, R., Taylor, R., 1986. Absorber simulation and design using a reactive absorption processes in sour gas treatment using rigorous modelling
nonequilibrium stage model. Can. J. Chem. Eng. 64, 96–105. and accurate experimentation. Chem. Eng. Res. Des. 85, 74–87.
Maddox, R.N., Bhairi, A.H., Diers, J.R., Thomas, P.A., 1987. Equilibrium Solubility of Tosh, J.S., Field, J.H., Benson, H.E., Haynes, W.P., 1959. Equilibrium study of the
Carbon Dioxide or Hydrogen Sulfide in Aqueous Solutions of Monoethanola- system potassium carbonate, potassium bicarbonate, carbon dioxide, and
mine, Diglycolamine, Diethanolamine and Methyldiethanolamine. GPA water. p. 27.
Research Report. Wang, M., Lawal, A., Stephenson, P., Sidders, J., Ramshaw, C., 2011. Post-combustion
Maddox, R.N., Elizondo, E.M., 1989. Equilibrium Solubility of Carbon Dioxide or CO2 capture with chemical absorption: a state-of-the-art review. Chem. Eng.
Hydrogen Sulfide in Aqueous Solutions of Diethanolamine at Low Partial Res. Des. 89, 1609–1624.
Pressures. GPA Research Report, No. 124. Zhang, Y., Chen, H., Chen, C.C., Plaza, J.M., Dugas, R., Rochelle, G.T., 2009. Rate-based
Mumford, K.A., Smith, K.H., Anderson, C.J., Shen, S., Tao, W., Suryaputradinata, Y.A., process modeling study of CO2 capture with aqueous monoethanolamine
Qader, A., Hooper, B., Innocenzi, R.A., Kentish, S.E., Stevens, G.W., 2011. Post- solution. Ind. Eng. Chem. Res. 48, 9233–9246.

You might also like