Proceedings of the 1st Annual Gas Processing Symposium H. Alfadala, G.V. Rex Reklaitis and M.M. El-Halwagi (Editors) © 2009 Elsevier B.V. All rights reserved.
1
Simulation of an Acid Gas Removal Process Using Methyldiethanolamine; an Equilibrium Approach Hassan E. Alfadala, and Essa Al-Musleh Department of Chemical Engineering,Qatar University, P.O. Box. 2713 Doha-Qatar
Abstract Adopting a rigorous equilibrium stage model for simulating an amine-based Acid Gas Removal (AGR) process is not straightforward technique. It may also be a frustrating exercise in simulation. The complexity of the system is mainly attributed to the necessity of considering both chemical and phase equilibrium issues to characterize such an electrolytic system. This paper discusses an effective approach for simulating a Methyldiethanolamine (MDEA) system using Aspen Plus RadFrac equilibrium stage model. For thermodynamic modeling, the approach uses the electrolyte-NRTL model, Redlich-Kwong-Soave equation of state, and Henry’s law. Component Vaporization efficiencies were incorporated in the simulation to account for the departure from equilibrium. To test its validity, the latter approach was tested against real design data obtained from a plant located in the State of Qatar. Furthermore, another process simulator namely ProMax and developed by Bryan Research and Engineering, Inc. was used for the purpose of comparison. Challenges faced during this practice (e.g., unit operation convergence, recycle convergence, etc.) and troubleshooting are also considered. Keywords: MDEA, Aspen Plus, Simulation, equilibrium, reactive distillation.
1. Introduction The removal of acid gases, H2S and CO2, from gas streams (sweetening) is essential for environmental, operational and health reasons. In general, acid gas pipeline specifications is 4.0 ppm H2S and 1.0% CO2. Processes utilizing aqueous alkanolamines as an absorption agent have become well established in this area. Compared to the most common alkanolamines, the tertiary MDEA (methyldiethanolamine) amine is known for its lower regeneration cost, its thermal and chemical degradation resistance, and lower corrosion rate. In addition, it has capability for selective H2S removal in the presence of CO2 and removal of both H2S and CO2. The objective of this work is to use Aspen Plus equilibrium approach model RadFrac for simulating an actual MDEA process located in the state of Qatar. Generally, such processes can be simulated using two methods; equilibrium stage method and rate based method. The letter method is known for its capability of handling actual trays and height of packing without the use of stage efficiencies. On the other hand, stage efficiencies are essential for the equilibrium stage method. Using an equilibrium stage model could be non practical to model the system into consideration. This is due to the
2
Alfadala and Al-Musleh
difficulties that might be associated with finding suitable efficiencies. Therefore, this work is pointing on an effective way to predict such efficiencies adequately. Fig. 1 shows a typical process flow sheet for such process. The system mainly consists of a reactive absorption and desorption (regenerator) columns. Acid gases are removed in the absorber via the direct contact of the gas stream with the amine in which H2S and CO2 react exothermally. The rich amine in acid gases (leaving the bottom of the absorber) is then fed to a low pressure flash drum to flash the hydrocarbons coabsorbed in the amine solvent. Such hydrocarbons may be used as fuel. The rich amine is then regenerated in the desorber by mean of stripping steam supplied by the column reboiler. In addition, the desorber is equipped with a reflux system to improve the purification of the amine. The regenerator column performs better at high temperature and low pressure. Therefore, before entering the column, rich amine is heat exchanged with the amine leaving the column reboiler. Prior to recycling the lean amine back to the absorber, the amine is cooled to a temperature that is 5 °C higher than the sour gas. Lower temperatures could increase the rate of hydrocarbon coabsorption. This temperature reduction is achieved by using the amine rich/lean heat exchanger and air coolers.
2. Model theory 2.1 Stage efficiency In most cases, the use of the phase equilibrium relation (eq. 1) in modelling such electrolytic system is not practical. Vapour phase composition should not be assumed to be at phase equilibrium. Therefore, incorporating another term, namely efficiency, to the relation is essential. This work uses the vaporization efficiency concept. yij = K ij xij (1) where
y , x , K , i, and j are vapor and liquid phase's composition, thermodynamic K-value, component index and stage index. Many efficiency relations are function of plate behavior (ex. Geometry, hydraulic, etc.) and physical properties (ex. Viscosity, volatility, etc.). On the other hand, vaporization efficiency represents positive multiple numbers introduced to the equilibrium relations (eq. 2). These numbers convert vapor phase compositions from equilibrium compositions to non equilibrium compositions. Identifying adequate efficiency values are discussed in the subsequent section. yij = ηij K ij xij (2) where ηij is vaporization efficiency of component i in stage j.
Simulation of an Acid Gas Removal Process Using Methyldiethanolamine; an Equilibrium Approach 3
2.2 Solution reactions K-values and enthalpies calculations need to take into accounts all the ionic and nonionic species existing in the liquid phase. Consequently, identifying the reactions
Figure 1: Typical AGR process using alkanolamine
involved in the system is extremely important. Reactions occurring in MDEA-water liquid phase, may be classified into two groups, kinetic controlled reactions and equilibrium controlled reactions (Bolhar-Nordenkampf et al., 2003). Those reactions are:
Table 1:Typical reactions in AGR process
Reaction
Type
+
-
2H2O ↔ H3O + OH
ionization of water (equilibrium)
+
-
H2O + H2S ↔ H3O + HS -
+
-
+
dissociation of hydrogen sulfide (equilibrium)
2-
H2O + HS ↔ H3O + S
H2O + HCO3 ↔ H3O + CO3
dissociation of bisulfide (equilibrium) 2-
+
dissociation of bicarbonate (equilibrium) +
H2O + MDEA ↔ MDEA + H3O +
CO2 + H2O ↔ H3O + HCO-3 CO2 + OH- ↔ HCO-3
dissociation of protonated alkanolamine (equilibrium) dissociation of carbon dioxide (kinetic) bicarbonate formation (kinetic)
4
Alfadala and Al-Musleh
The dissociation of carbon dioxide is very slow (Austgen and Rochelle, 1989); therefore, it may be neglected for such system. Aspen Plus built in power law expression was used for simulating the bicarbonate reaction. The kinetic parameters in the law were all retrieved from Aspen data base. 2.3 Phase equilibrium Phase equilibrium is concerned with the distribution of each species between the liquid and vapour phases. Due to the highly non-ideal behaviour of the system, such distribution should be characterized by an activity coefficient approach (ACM). The electrolyte NRTL and Redlich-Kwong-Soave equation of state are used in this work. Since the system into consideration contains dilute gases (ex. CH4, and C2H6) and supercritical species (ex. H2S, CO2), Henry's law is also considered with ACM model. It should be noted that in this work, binary parameters for the above models were all retrieved from Aspen data base.
3. Simulating an actual MDEA process The simulation of Qatar's MDEA plant was based on the design data of the plant. The operating conditions and characteristic of the process are summarized in Table 2. 3.1 Simulation approach Fig. 2, 3, and 4 demonstrate the approach used for simulating the entire AGR process. The approach shown in Fig. 2 was followed in simulating the absorber section of the process. Aspen data base does not have Henry's binary coefficients for n-pentane and nhexane in water. Therefore, those coefficients were obtained from Mokraoui al et., 2007 and defined to the simulator. The number of stages, column pressure profile, sour gas and lean MDEA conditions, and modelling approach (equilibrium or rate) need all to be specified for the simulator. In addition, since the unit into consideration is an absorber, the user needs to remove the reboiler and condenser specifications. Finally, reactive stages and their correspondence liquid holdup were specified to the absorber model. Liquid holdups were identified using the relation developed by Bennett et al., 1983. In this work all the stages were assumed to be reactive. In this work the absorber did not converged. However, supplying the absorber with a temperature profile as an estimate for a new run was found to be adequate. In this work, the profile from the unconverged case was used. Moving to the MDEA regenerator, Fig. 3 , although, the distillate rate for the regenerator column is known (Table 2), the flow rate of the components constituting the distillate stream ( H2S, HS-, S-2, CO2, HCO3-2, CO3-2 and some water), were imported from the stripper (regenerator) feed stream and exported to be the stripper distillate via the use of a calculator block. The main reason behind this calculator block is as follow: so far the work is under equilibrium criteria; therefore, the amount of acid gases absorbed are higher than those in reality. Consequently, the distillate rate in Table 2 should not be reasonable. Thus the calculator block is identifying a suitable distillate specification. In case, the original rate was defined to Aspen, Aspen will predict dried
Simulation of an Acid Gas Removal Process Using Methyldiethanolamine; an Equilibrium Approach 5
stages (liquid or vapour flow approaches zero). In other words, the column will not converge. To complete the column specifications, number of stages, feed stage, column pressure profile, reflux ratio, calculation mode, and type of condenser were all defined to Aspen. In this work, the regenerator has converged directly without any problems. Table 2: Qatar's MDEA plant operating conditions and process characterstcs
Parameter Sour gas flow rate (kmol/h) Sour gas temperature (°C) Sour gas molar composition H2 S mol.% CO2 mol.% COS mol.%
Absorber 20,000.00 35 0.92 2.39 -4
N2+H.C’s mol.% Solution flow rate (kmol/h) Solution temperature (°C) MDEA composition (wt%) Acid gas flow (kmol/h) Reflux ratio Pressure (barg) Number of stages *Flash drum pressure is 10.00 barg.
5.15×10 96.68 10,000.00 40 45 72 25
Stripper 103.6 514 1.6 2.35 24
Define thermodynamic model, components, reactions, and missing Henry's low parameters Number of stages Reactions & liquid holdup
S.G & L.A conditions RadFrac (absorber)
Model: equilibrium
Column pressure Profile Insert auxiliary equipments (ex. pumps, coolers, etc)
Temperature profile Run No Converged Yes Regenerator
Figure 2: Simulation approach for the MDEA abosrber
6
Alfadala and Al-Musleh
After inserting the heat exchangers, MDEA storage, and pumps downstream the regenerator, lean MDEA (LMDEA) needs to be recycled back to the absorber. This recycling requirement was found to be accompanied with convergence problems. The convergence bottleneck was the water lost from the MDEA stream as it is processed from the absorber to the regenerator. Fig 4 shows the block diagram for the simulated process. The calculation starts with the absorber in which the tear stream LMDEA* conditions were defined from Table 2. Each complete cycle (from the absorber to the convergence criteria block) is one iteration.
Absorber Insert calculator for stripper distillate rate Number of stages Reboler &. partial cond.
Reflux ratio
RadFrac (Regenerator) Model: equilibrium
Feed stage Column pressure profile
Run
Recycling the lean MDEA
Figure 3: MDEA regenerator simulation approach
Aspen will be iterating and updating the tear stream LMDEA* from the previous LMDEA (from step 7) till LMDEA meets tear stream convergence tolerance (i.e | LMDEAT, P, flow rates – LMDEA*T, P, flow rates | ≤ Tolerance ). As number of iteration increases, the more the water looses from the absorber, MDEA flash drum, and regenerator. Consequently, the LMDEA tear stream will face convergence problems in terms of water content (i.e. water in LMDEA < water in LMDEA*). To overcome this, water lose from the solvent in each unit operations was exported to a calculator block (Fig. 4) to identify a makeup water amount (i.e WMkUp= WSweet Gas+WFlush Gas+WAcid GasWSour Gas). This amount is then exported to the storage tank. Applying this strategy had resulted in converging the flow sheet in three iterations in which the tolerance was 0.0001. 3.2 Columns vaporization efficiencies Aspen plus built in sensitivity analysis tool was utilized to identify values for the absorber and regenerator efficiencies. Both H2S and CO2 efficiencies were varied against the flow rates of H2S and CO2 in the sweet and acid gas streams. Fig. 5 shows how this material balance (for the absorber) behaves against different efficiency values.
Simulation of an Acid Gas Removal Process Using Methyldiethanolamine; an Equilibrium Approach 7
StopCalculation Step Criteria met [8] Convergence criteria LMDEA*-LMDEA< З Start calculation Design LMDEA
Step [7]
LMDEA
Water make up Storage Tank
Step [6] Calculator for H2OMukeup Total water - Water Sour Gas Water
Criteria was not met LMDEA*
Sour Gas
Step [1]
RadFrac Absrober
Water Sweet Gas
RMDEA Water Sour Gas
Step [5]
Step [2]
Acid Gas Water
Rich MDEA
Step [3] Amine lean/reach exchanger
Flash 2 MDEA Flash Drum
RadFrac Stripper
Distillate rate
RMDEA
Acid Gas Calculator for stipper Distillate rate Step [4]
Water Water Water
Figure 4: Block diagram for the simulated process
One may see that as the H2S efficiency increases (at different CO2 efficiencies), the more the H2S content in the sweet gas. On the other hand, sweet gas CO2 content is not sensible to the H2S efficiencies. It is only influenced by those for the CO2.
4. Results and discussion 4.1 Material and energy balance Main simulation outputs (Table 3 and 4) for the MDEA absorber and regenerator do match very closely the plant data. However, predicted rich and lean MDEA temperatures were deviating from the plant data by approximately 8.6 and 12.0 °C, respectively. 0.09 170 165 160
0.08 CO2 in Sweet Gas (kmol/h)
H2S in Sweet Gas (kmol/h)
0.085
0.075 0.07 0.065 1.5
1.6
1.7
1.8
0.06 1.9
0.055
2
2.1
155 1.5 1.6 1.7 1.8 1.9
150 145
2 2.1 2.2
140 135 130
2.2
125 0.05 2
2.1
2.2
2.3
2.4
2.5
H2S efficiency
2.6
2.7
2.8
120 2
2.1
2.2
2.3
2.4
2.5
2.6
H2S Efficiency
Figure 5: Acid gas content in the sweet gas vs H2S efficiency at different CO2 efficiencies
2.7
2.8
8
Alfadala and Al-Musleh
It should be noticed here that a temperature of 144.0 °C in the regenerator reboiler can result in amine degradation (Pandey, 2005). Moreover, rich MDEA and acid gas streams were found to contain hydrocarbons less than those reported in the design data. Table 3: Simulation results for the MDEA absorber
MDEA Absorber Parameter Plant 3.7 H2S in sweet gas (ppm) 0.9 CO2 in sweet gas (mol.%) Sweet gas Temperature (°C) 47 Rich MDEA Temperature (°C) 50.8 H.C's co-absorbed (mol. %) 0.15
Aspen 3.8 0.89 48 42.2 0.064
Table 4: Simultion results for the MDEA regeneragtor
MDEA Regenerator Parameter Plant H2S in Acid gas (mol. %) 34.8 CO2 in sweet gas (mol.%) 57 Acid gas Temperature (°C) 53 Lean MDEA Temperature (°C) 132.1 H.C's in Acid gas (mol. %) 0.89
Aspen 35 57.2 59 144 0.24
4.2 Columns profiles Temperature and composition profiles are presented in Fig. 6 and 7 for both columns. Due to the lack of actual data the literature and a process simulator namely ProMax (which is the most popular process simulator for amine sweetening, Kohl and Nielsen, 1997) were used for the purpose of comparison. For the absorber, both ProMax and Aspen temperature profiles have similar behaviour. A temperature bulge was identified by both simulators. Nevertheless, each simulated stage in ProMax is 10.0 °C higher than that in Aspen. The behaviour of the temperature in the absorber may be explained as follow: as the MDEA solution flow down the absorber, it reacts exothermically with H2S and CO2 resulting in raising the temperature of both the vapour and liquid phases. Near the middle of the tower (where the temperature is maximum), the temperature starts to drop down as a result of the heat absorbed by the cooled gas entering the contactor. Such behaviour was shown by Shiveler et al., 2006 in which they have performed a thermal scanning on an industrial MDEA absorber. Regarding the composition profiles in the absorber, the rate of reaction of H2S with the MDEA solution is fast. Therefore, H2S bulk removal in the first five stages of the
Simulation of an Acid Gas Removal Process Using Methyldiethanolamine; an Equilibrium Approach
9
absorber was expected. Fine cleaning of the gas stream from H2S is mainly being achieved in the rest of the stages.
40.0
Temperature, C 60.0 70.0
50.0
80.0
0
90.0
1
0
5
5
9
0.025
0.03
10
ProMax Aspen plus
13
Mole Fraction‐vapor 0.01 0.015 0.02
0.005
Stage
Stage No.
30.0
ProMax (H2S)
15
17
Aspen Plus (H2S) ProMax (CO2)
20
21
Aspen (ProMax)
25
25
30
Figure 6: Absorber temperature and composition profiles
Such behaviour was obtained in Aspen and ProMax. Furthermore, Bolhar-Nordenkampf work does confirm such attitude. For CO2, all RadFrac stages were found to be necessarily to achieve CO2 pipe line specifications (about 1.0% mol.). This trend does mach those in ProMax and Bolhar-Nordenkampf. 50
60
70
80
90
Temp. C 100 110
120
130
140
0
150 0 0
5
5
0.1
0.2
Mole Fraction (Vapor) 0.3 0.4
Temp. C 130 5
10
135
140
0.5
0.6
0.7
Mole Fraction 145
150
0
10
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.01
10
20
Aspen ProMax
14
15
16
Aspen ProMax
18
20
20
Stage
15
Stage
10
15 Stages
Stages
12
20
Aspen (H2S) Aspen (CO2) ProMax (CO2) ProMax (H2S)
25
22
25
30
25
24
26
28
30
30
Figure 7: Regenerator temperatue and composition profiles
Moving to the regenerator, both ProMax and Aspen plus predicted profiles having similar overall trends for both the composition and temperature. However, Aspen's
Aspen (H2S) Aspen (CO2) ProMax (CO2) ProMax (H2S)
10
Alfadala and Al-Musleh
temperature profile is shifted by around 10.0 °C to the right of ProMax's profile. In addition, both profile shows that the change of temperatures from stage 6 down to the bottom are small. Therefore composition change in these stages was also found to be small. Fig. 7 shows that bulk removal of H2S and CO2 are achieved in the first ten stages and purification of the traces is carried out in the rest.
5. Conclusion The use of Aspen plus RadFrac model for simulating Qatar's MDEA process was found to be able to predict good results. The electrolyte-NRTL model, Redlich-Kwong-Soave equation of state, and Henry’s law were used for the thermodynamic framework. To account for the non-ideality behaviour of the system, vaporization efficiencies for H2S and CO2 were defined for the MDEA absorber and regenerator. The material and energy balances predicted by Aspen were found to be close to the design data. Regarding the temperature and compositions profiles, they were identified to have similar behaviour as those predicted by ProMax and those reported in the literature.
References A. Kohl and R. Nielsen, 1997, Gas Purification, Gulf Publishing Company, Houston, Texas. Bennett, D. L., Agrawal, R., & Cook, P. J. New pressure drop correlation for sieve tray distillation columns. American Institute of Chemical Engineers Journal, 29, 1983, 434-442. D. Austgen and G. Rochelle, Model of vapour-liquid equilibria for aqueous acid gasalkanolamine systems using electrolyte-NRTL equation, Ind. Eng. Chem., 28, 1989, 1060-1073. G. Shiveler, G. Solis, L. Gonzalez, and M. Bueno, 2006, Retrofit of a H2S selective amine absorber using MellapakPlus structured packing, Asia-Pac. J. Chem. Eng., 2, 2007, 232-244. M. Bolhar-Nordenkampf, A. Friedl, U. Koss, and T. Tork, 2003, Modelling selective H2S absorption and desorption in an aqueous MDEA-solution using a rate-based non-equilibrium approach, Chem. Eng. Proc.,43, 2004, 701-715. M. Pandey, Process Obtimization in Gas Sweetening Unit-A Case Study, International Petroleum Technology Conference, 2005.