Improving Prediction Accuracy of a Rate-Based Model of an MEA-Based Carbon Capture Process for Large-Scale Commercial Deployment

Abstract

Carbon capture and storage (CCS) technology will play a critical role in reducing anthropogenic carbon dioxide (CO₂) emission from fossil-fired power plants and other energy-intensive processes. However, the increment of energy cost caused by equipping a carbon capture process is the main barrier to its commercial deployment. To reduce the capital and operating costs of carbon capture, great efforts have been made to achieve optimal design and operation through process modeling, simulation, and optimization. Accurate models form an essential foundation for this purpose. This paper presents a study on developing a more accurate rate-based model in Aspen Plus^® for the monoethanolamine (MEA)-based carbon capture process by multistage model validations. The modeling framework for this process was established first. The steady-state process model was then developed and validated at three stages, which included a thermodynamic model, physical properties calculations, and a process model at the pilot plant scale, covering a wide range of pressures, temperatures, and CO₂ loadings. The calculation correlations of liquid density and interfacial area were updated by coding Fortran subroutines in Aspen Plus^®. The validation results show that the correlation combination for the thermodynamic model used in this study has higher accuracy than those of three other key publications and the model prediction of the process model has a good agreement with the pilot plant experimental data. A case study was carried out for carbon capture from a 250 MWe combined cycle gas turbine (CCGT) power plant. Shorter packing height and lower specific duty were achieved using this accurate model.

Content

《1.Introduction》

1.Introduction

《1.1. Background》

1.1. Background

Increasing concentrations of greenhouse gases (GHGs) caused by anthropogenic activities are responsible for most of global warming [1]. Carbon dioxide (CO₂) is a main GHG, accounting for 76% of total GHG emissions in 2010 [2]. The International Energy Agency (IEA) set up a BLUE Map scenario with 14 Gt of CO₂ emissions in 2050 compared with 57 Gt of CO₂ emissions in the baseline scenario [3]. In order to achieve this target, carbon capture and storage (CCS) technology will play a vital role in delivering 19% of cumulative CO₂ emission reductions between 2015 and 2050 in the power sector [3].

Among the three main approaches envisaged for CO₂ capture from power plants—pre-combustion capture, post-combustion capture, and oxyfuel capture [4]—the solvent-based post-combustion carbon capture (PCC) process is regarded as the most promising technology for commercial deployment [5,6]. In solvent-based carbon capture technology, CO₂ is separated from flue gas after combustion by chemical absorption; monoethanolamine (MEA) is regarded as a benchmark solvent for this process.

《1.2.Previous studies》

1.2.Previous studies

A complex electrolyte aqueous solvent is involved in the MEA-based PCC process [7], which requires accurate thermodynamic modeling and physical properties calculations for its modeling. Thermodynamic data, especially regarding CO₂ solubility, have been reported for 30 wt% MEA aqueous solutions [8,9] and for a wider MEA solution concentration range [10,11]. For the parameterization and validation of physical properties calculation methods of an MEA-H₂O-CO₂ mixture, experimental data on MEA aqueous solutions are valuable, especially with different CO₂ loading. Correlations for the calculation of the density and viscosity of MEA-H₂O-CO₂ mixtures at different temperatures and MEA concentrations can be found in the literature [12–14]. In terms of mass transfer and thermal performance of the integrated MEA-based PCC process, several experimental campaigns [15,16] have been conducted.

For a highly nonlinear electrolyte MEA-H₂O-CO₂ solution, the electrolyte non-random two-liquid (eNRTL) model [17,18] is the most widely adopted model [10,19]. Recently, some studies [20,21] have also used the perturbed-chain statistical association fluid theory (PC-SAFT) [22,23] equation of state (EOS) for the vapor phase of an MEA-H₂O-CO₂ mixture, with a system temperature of up to 500 K and a system pressure of up to 15 MPa.

For this capture process, significant energy is consumed for solvent regeneration [6]. Thus, the cost of carbon capture is high when PCC is added to the emitters. Great research efforts have been taken to reduce the carbon capture cost through process modeling and simulation approaches. Most early studies were carried out for the parametric sensitivity analysis of solvent-based PCC processes in the context of coal-fired power plants [24–27]. Some studies were carried out on integrations between power plants and carbon capture plants [28–30]. Several studies focused on optimizing the whole plant through process optimization [31–34].

However, obvious inconsistencies in the literature were found for key equipment design features and key operational variables. For example, the packing height varies from 13.6 m [35] to 30.6 m [32] for the absorber and from 7.6 m [35] to 28.15 m [21] for the stripper for similar capture tasks. The optimal lean loading range is equally wide from 0.132 mol_CO2· [31] to 0.234 mol_CO2· [36], with corresponding specific duty in a range from 3.77 to 4.35 .Those inconsistencies cause confusion for future research in this field. They may also cause some trouble for the engineering design of a large-scale commercial deployment.

The main reasons for the abovementioned knowledge gaps may be related to conflicts between the complexity of the integrated system and the accuracy requirement of the modeling and simulation studies. Firstly, the models used in some publications were relatively simple. For example, equilibrium models were used for the mass transfer and reaction in both the absorber and the stripper [37]. For a rate-based model, the correlations for calculations of mass transfer coefficients, interfacial area, liquid holdup, and pressure drop inside packing beds also have a large impact on the prediction accuracy [38,39]. For the kinetics-controlled reactions, it is found that the values of the kinetics of reverse reactions for bicarbonate formation are different for the absorber and the stripper [40]. Inappropriate correlations used in the models would significantly affect the accuracy of model predictions.

《1.3. Aim and novel contribution》

1.3. Aim and novel contribution

In order to address the abovementioned knowledge gaps, this study aims to improve the accuracy of the rate-based model in Aspen Plus^® for the MEA-based carbon capture process. The novel contributions of this paper can be justified by the following: ① A new combination of correlations was selected after comparing model predictions with the experimental vapor-liquid phase equilibrium (VLE) data; ② the correlations for predicting the liquid density of the mixture and the effective vapor-liquid interfacial area were improved by coding Fort an subroutines in Aspen Plus^®; ③ different kinetics parameters were used for reverse reactions for bicarbonate formation in the absorber and the stripper, respectively, thus reflecting the nature of the different operating conditions in the absorber and the stripper; and ④ the rate-based process model was validated with the experimental data and pilot plant data at three different stages, including thermodynamic modeling, physical properties calculations, and process model development at the pilot scale.

《2.Framework of modeling of the solvent-based carbon capture process》

2.Framework of modeling of the solvent-based carbon capture process

Using an amine solvent to absorb CO₂ from exhaust gases is a reactive absorption process involving an electrolyte aqueous solvent [6]. The modeling of this non-ideal multi-component system is a systematic work at different levels. Fig. 1 outlines the modeling framework for such a PCC process. Although the software package Aspen Plus^® was used for the modeling and simulation of the process, it is important to check the calculation methods with their corrections in order to ensure the accuracy of the process simulation and optimization.

《Fig. 1》

Fig. 1. Framework of the modeling of a solvent-based PCC process. L: liquid phase; V: vapor phase.

Accurate prediction of the physical properties of pure components and mixtures is one of the basic prerequisites in process modeling and simulation. As the first step, the thermodynamic model should be developed to present VLE and to calculate the state parameters of the MEA-H₂O-CO₂ mixture, such as the temperature, pressure, and composition of the liquid and vapor phases. The solubility of CO₂ in the MEA-H₂O-CO₂ mixture is a key parameter, and is normally used for validation purposes for the calibration of the correlations or for selection for VLE calculation.

The physical properties are part of the correlations for heat transfer, mass transfer, interfacial area, liquid holdup, and pressure drop. It is important to choose the right physical property models to ensure the success of process modeling and simulation.

At the process level, both absorption and desorption in the packed columns are key processes. A rate-based model offers better accuracy than an equilibrium model for the absorption performance of the columns [41]. This accuracy is a function of the appropriate correlations used for liquid and vapor phase mass transfer coefficients, the effective vapor-liquid interfacial area, and the pressure drop in the rate-based model.

This framework shows that the rate-based model for this solvent based carbon capture process is a highly nonlinear model, which has numerous parameters, correlations, and equations. Therefore, it is not realistic to completely repeat the published models with the same input conditions. This is also the main consideration behind the choice to use three-stage validations in this study, and to update some correlations by coding a Fortran subroutine in Aspen Plus^® to ensure model accuracy, rather than directly comparing the process performance with those of other published models. Using this three stage model validation method, the model was dissected in detail based on the logical structure of numerical modeling, allowing more insights to be obtained.

《3.Thermodynamic modeling of the MEA-H2O-CO2 system》

3.Thermodynamic modeling of the MEA-H₂O-CO₂ system

《3.1.EOSs and relevant model parameters》

3.1.EOSs and relevant model parameters

In this study, the PC-SAFT EOS [22,23] is used to calculate the properties of the vapor phase, and the eNRTL [18] method is used to model the electrolyte system of an MEA-H₂O-CO₂ mixture.

3.1.1. The PC-SAFT EOS for the vapor phase

Compared with some typical cubic EOSs such as the Peng-Robinson (PR) EOS and the Soave-Redlich-Kwong (SRK) EOS, the PC-SAFT EOS is able to accurately estimate vapor phase fugacity coefficients at high pressures [20,42], which is important for accurate performance predictions of CO₂ compression with an outlet pressure as high as 136 bar^† [43]. Table 1 [20,23,44] summarizes the PC-SAFT parameters of pure components, and Table 2 [45,46] lists the binary interaction parameters (k_ij) of MEA-H₂O and CO₂-H₂O.

^†1 bar = 10⁵ Pa.

《Table 1》

Table 1 PC-SAFT parameters of pure components.

《Table 2》

Table 2 Binary parameters for the PC-SAFT EOS.

3.1.2. The eNRTL method for the liquid phase

The liquid phase of an MEA-H₂O-CO₂ mixture is a typical electrolyte solution [19]. The eNRTL method has been validated and used to model electrolyte solutions in many publications [19,20,27,40,47].

Table 3 [20,44,46] summarizes the model parameters used for this study, and their sources. Most of the parameters were obtained from the SRK-ASPEN databank [44], and some were updated by recent studies by regression using new experimental data [46].

《Table 3》

Table 3 Model parameters for eNRTL.

《3.2. Physical solubility and Henry’s law constant》

3.2. Physical solubility and Henry’s law constant

Physical solubility is the equilibrium between CO₂ molecules in the vapor phase and those in liquid solutions; it is calculated using Henry’s law. The Henry’s law constants for CO₂ with water and with MEA are required, and can be calculated using Eq. (1):

where H_i−j is the binary Henry’s law constant between pure components i and j; T is the system temperature; and C₁, C₂, C₃, and C₄ are correlations for Henry’s law constants. Table 4 [46,47] summarizes the available binary Henry’s law constants for an MEA-H₂O-CO₂ mixture. For the system of an MEA-H₂O-CO₂ mixture, most publications only take gas components such as CO₂ and N₂ as Henry components In addition, most studies only consider Henry’s law constants for CO₂ with H₂O [19]. The Henry’s law constants for CO₂ with H₂O have been well studied by Yan and Chen [46] by examining extensive quantities of experimental VLE data for the CO₂-H₂O binary system. Liu et al. [47] considered Henry’s law constants for CO₂ with MEA.

《Table 4》

Table 4 Correlations for the calculation of Henry’s law constants (on the molality scale).

《3.3. Chemical reaction equilibrium》

3.3. Chemical reaction equilibrium

Liquid phase chemical reactions involved in the MEA-H₂O-CO₂ system can be expressed as follows:

R1: water dissociation

R2: dissociation of CO₂

R3: dissociation of carbonate

R4: dissociation of the protonated amine

R5: carbonate formation

The chemical equilibrium constants of these reactions were calculated using Eq. (2), and the related correlations are shown in Table 5 [19,48,49].

《Table 5》

Table 5 Correlations for chemical equilibrium constants (on the molality scale).

where K_j refers to the chemical equilibrium constants for each reaction j; T is the system temperature; and C₁, C₂, C₃, and C₄ are correlations for the chemical equilibrium constants.

Once the chemical equilibrium constants are determined, the chemical equilibrium of each reaction is determined using Eq. (3) [19].

where i denotes the reactant component; n denotes the product component; x denotes the model fraction of each component in the liquid phase based on true species, molecular and ionic; γ denotes the activity coefficient; and ν denotes the stoichiometric coefficient of each component in reaction j.

《3.4.Validation of CO₂ solubility prediction》

3.4.Validation of CO₂ solubility prediction

3.4.1.Case setup

In order to compare and select appropriate correlations for this study, several combinations of correlations [19,20,47] were chosen for carrying out the validation against the experimental data. Table 6 [19,20,46,47,50,51] provides the model details.

《Table 6》

Table 6 Different combinations of correlations for validation.

《Table 7》

Table 7 MAPE of validation with CO₂ partial pressure of the MEA-H₂O-CO₂ system.

3.4.2.Validation results

For model validation purposes, the model predictions were compared with the experimental data in terms of the CO₂ partial pressure and/or total pressure in the vapor phase for different CO₂ loading in an MEA aqueous solution. In this study, the experimental data from Ref. [11] were chosen because these data cover a wider range of MEA concentrations than other publications, as well as wider ranges of system temperatures and pressures.

Fig. 2 depicts comparisons between model predictions and experimental data for the partial pressure of CO₂ in the vapor phase of MEA-H₂O-CO₂ mixtures for different concentrations of MEA. Table 7 [19,20,47] presents the mean absolute percentage error (MAPE) of validation results at different MEA concentrations. Generally, the deviations between experimental data and model predictions become bigger at the lower (15 wt%) and higher (45 wt%–60 wt%) MEA concentrations, compared with the 30 wt% MEA concentration. It is noticeable that the model predictions of this study at 15 wt% concentration are worse than those of Ref. [47]. The reason is that some of the correlations used in this study were inherited from Ref. [20]. Furthermore, none of these four combinations produced good predictions that covered low to high MEA concentrations, which reflects an inherent limitation of the correlation method: A correlation should not go beyond the conditions of the data for its regression. However, most of existing correlations that were used for the thermodynamic modeling of the MEA-H₂O-CO₂ system were regressed based on experimental data at 30 wt% MEA concentration solvent.

《Fig. 2》

Fig. 2. CO₂ partial pressure as a function of CO₂ loading with: (a) 15 wt% MEA solvent, (b) 30 wt% MEA solvent, (c) 45 wt% MEA solvent, and (d) 60 wt% MEA solvent. Exp: experimental data; TS: this study.

《4. Physical properties of the MEA-H₂O-CO₂ system》

4. Physical properties of the MEA-H₂O-CO₂ system

《4.1.Physical property model》

4.1.Physical property model

The physical properties include: ① thermodynamic properties, such as density, enthalpy, and heat capacity; and ② transport properties, such as viscosity, surface tension, thermal conductivity, and diffusivity. Table 8 lists the chosen models for the property calculation for the mixture in this study. It should be noted that the correlations for the density of the liquid mixture are obtained from Ref. [14] by coding a Fortran subroutine in Aspen Plus^®.

《Table 8》

Table 8 Correlations used for the property calculation of the mixture.

《4.2.Available experimental data for validation》

4.2.Available experimental data for validation

Table 9 [14,52–54] provides the available experimental data from the literature for the physical properties validation of MEA-H₂O-CO₂. The vapor phase of the MEA-H₂O-CO₂ mixture under the operating temperature (20–150 °C) and pressure (1–2 bar) of the absorber and stripper is not an issue, and no experimental data are available for those properties of the vapor phase. Available experimental data for the thermal conductivity of the liquid phase were not currently found. Furthermore, direct measurement of CO₂ diffusivity in an MEA aqueous solution is impossible because CO₂ reacts with MEA. The NO₂ analogy method was used to produce the data for CO₂ diffusivity [55].

《Table 9》

Table 9 Available experimental data for the physical properties of the liquid phase.

《4.3.Validation results》

4.3.Validation results

Fig. 3 and Fig. 4 present comparisons between the model predictions and the experimental data for different properties of the MEAH₂O-CO₂ mixture at different concentrations of MEA.

Table 10 presents the deviations of the validation results for the physical properties plotted in Fig. 3 and Fig. 4. Both MAPE and the maximum absolute percentage error (APE) are given. For the liquid density, the model predictions are in good agreement with the experimental data over the full range of system conditions. For the specific heat capacity (Fig. 4(a)), the deviations gradually increase as CO₂ loading rises. For the surface tension, the experimental data themselves have large deviations (Fig. 4(c)).

《Fig. 3》

Fig. 3. Liquid density of MEA-H₂O-CO₂ with different CO₂ loading ( ) at: (a) 30 wt% MEA solvent, (b) 40 wt% MEA solvent, (c) 50 wt% MEA solvent, and (d) 60 wt% MEA solvent. Model: model predictions.

《Fig. 4》

Fig. 4. Physical properties of MEA-H₂O-CO₂ at different MEA concentrations and 298.15 K for: (a) specific heat capacity, (b) liquid viscosity, and (c) surface tension.

《Table 10》

Table 10 Deviations of the model predictions from the experimental data.

《5. Process model development and validation at the pilot scale》

5. Process model development and validation at the pilot scale

《5.1. Introduction of the pilot plant》

5.1. Introduction of the pilot plant

In this study, the pilot plant located at the University of Kaiserslautern [56] was chosen for model validation, for the following reasons: ① Both the absorber and the stripper use Mellapak 250Y packing, which is regarded as appropriate structured packing for industrial deployment [57]; and ② the experimental data are comprehensive and well presented in the publications from this plant [16], which enables more comprehensive validation that can be compared with other studies. Table 11 summarizes the equipment features and the ranges of the key operation variables. (For more details about this pilot plant, refer to Ref. [16].)

《Table 11》

Table 11 Main specifications of the pilot plant.

《5.2. Process model development》

5.2. Process model development

5.2.1Model flowsheet and process description

Fig. 5 shows the flowsheet of this steady-state process model in Aspen Plus^®. The flue gas leaving the power plant goes to a gas blower and is then cooled to 40–50 °C before entering the absorber, in order to improve the absorption efficiency [25]. The scrubbed flue gas is emitted to the atmosphere and the CO₂-rich solvent is discharged from the bottom of the absorber and enters the stripper. The CO₂-rich solvent is regenerated inside the stripper with heat input to the reboiler. The regenerated solvent is cooled and re-circulated to the absorber for reuse.

《Fig. 5》

Fig. 5. Process flowsheet in Aspen Plus^®. MU: make up; REB: reboiler; CONDSPL: condensate splitter; DESUPSTM: desuperheater steam; INTHEX: internal heat exchanger; DCC: direct contact cooler; WAT: water desuperheater.

5.2.2.Kinetics-controlled reactions

In Section 3.3, the equilibrium reactions of the MEA-H₂O-CO₂ mixture were described during the thermodynamic modeling. In the rate-based model, the reaction of dissociation of CO₂ and the reaction of carbonate formation should be considered as kinetics-controlled reactions [27]:

R2*: dissociation of CO₂

R5*: carbonate formation

Power law expressions were used for the kinetics-controlled reactions. The reaction rates of reactions R2* and R5* can be calculated by Eq. (4) [27].

where r_j is the reaction rate for reaction j in mol·(min·m³ )^-1; k^o_j is the pre-exponential factor in kmol·(m³·s)^–1; T is the system temperature in K; n is the temperature factor; E_j is the activation energy in kJ·mol^-1 ; R is the gas constant; C_i is the mole fraction of species i; and α_ij is the reaction order of component i in reaction j. k^o_j and E_j for the reactions were calculated using the experimental data shown in Table 12 [40].

《Table 12》

Table 12 Parameters k_j^o and E_j in Eq. (4) [40].

5.2.3. Rate-based mass transfer

The modeling of the absorber and the stripper was based on two film theory [58], which is used to describe the mass transfer of components between the gas phase and the liquid phase. According to two-film theory, a vapor film and liquid film with a phase equilibrium interface are assumed between the bulk gas and bulk liquid phases. Chemical reactions are assumed to occur in the liquid film only.

For the Rate Sep model in Aspen Plus®, Zhang et al. [27] provided very detailed discussions about correlations and settings. In this study, the VP lug flow model was chosen in order to model the bulk properties with reasonable accuracy, since the “Countercurrent” model sometimes causes oscillations in the temperature profile even though it is the closest approximation of the real situation [59]. It was also noted that the discretization points of the liquid film need to be over 10, in order to achieve accuracy; otherwise, the simulation results may exhibit over-predictions of the rate of mass transfer.

For the correlations related to mass transfer, Razi et al. [59] validated 12 correlation combinations with experimental data from the CO₂ enhanced separation and recovery (CESAR) pilot data; the results show that Billet and Schultes [60] give an accurate correlation. Table 13 [59–62] lists the parameters and correlations related to mass transfer that were used in this study. Here, a Fortran subroutine was used to implement the correlation of Ref. [61] for the calculation of interfacial area.

《Table 13》

Table 13 Parameters and correlations selection for mass transfer in the RateSep model.

《5.3.Model validation》

5.3.Model validation

For the MEA-based carbon capture process, the key operational parameters affecting performance are the CO₂ concentration in the flue gas, the MEA concentration in the solvents, the lean loading, and the liquid-to-gas (L/G) ratio. Thus, four sets of experiments from Ref. [16] were chosen for model validation purposes. These include: ① Experiments A1–A6, with different CO₂ concentrations in the flue gases; ② Experiments A24–A27, with different MEA concentrations at two different CO₂ concentrations in the flue gases; ③ Experiments A28–A33, with different solvent flow rates at high CO₂ concentrations in the flue gases; and ④ Experiments A34–A39, with different solvent flow rates at low CO₂ concentrations in the flue gases. Model validations were carried out based on the same feed conditions, and the CO₂ loading in the lean solvent (lean loading) was targeted by varying the reboiler duty of the stripper. Experimental data and model predictions for CO₂ loading in a rich solvent (rich loading), the CO₂ capture level, and the stripper reboiler duty could then be compared.

Fig. 6 illustrates the bias between experimental data and model predictions for CO₂ capture level, rich loading, and specific duty under the same input conditions. The validation results for CO₂ capture level and rich loading show a good agreement. Regarding the specific duty, reboiler duty in the experiments was affected by heat loss from the equipment and pipelines, which could not be measured directly. Although the values for specific duty provided in Ref. [16] were corrected, the deviations themselves could not be evaluated, which may be the reason for high APEs for the validation results of the specific duty. MAPEs of the model predictions for the CO₂ capture level, the stripper reboiler duty, and the rich CO₂ loading, as compared with the experimental data from Ref. [16], are 1.78%, 1.54%, and 7.49%, respectively.

《Fig. 6》

Fig. 6. The bias between experimental data and model prediction for: (a) CO₂ capture level, (b) rich loading, and (c) specific duty.

Validations were also conducted to compare the temperature profiles and the CO₂ composition profiles inside the absorber and the stripper based on Experiments A1, A2, and A3 [16]. CO₂ concentrations in the flue gases are 8.5 mol% for A1, 16.5 mol% for A2, and 5.5 mol% for A3. Fig. 7 shows that the model predictions are in very good agreement with the experimental data. One statement is that the total packing height is 2.25 m inside the stripper; the 3 m position of the packing in Fig. 7(b) and Fig. 7(d) represents the reboiler of the stripper. The comparison results show that the model predictions are in very good agreement with the experimental data.

《Fig. 7》

Fig. 7. Validation results between model predictions and experimental data. (a) Temperature profile of the absorber; (b) temperature profile of the stripper; (c) CO₂ composition profile inside the absorber; (d) CO₂ composition profile inside the stripper.

《6.Case study》

6.Case study

《6.1.Methodology for model scale-up from pilot scale to commercial scale》

6.1.Methodology for model scale-up from pilot scale to commercial scale

To match the capacity requirement of handling the flue gas from power plants at an industry scale, the model of the carbon capture process at a pilot scale was scaled up, based on chemical engineering principles regarding the estimation of column diameter and pressure drop [63].

As initial inputs to the process model at an industrial scale in Aspen Plus^®, first-guess diameters are required for both the absorber and the stripper. The column diameters can be calculated from the maximum flooding vapor. In this study, a generalized pressure drop correlation (GPDC) figure (Fig. 8) [64] was used to estimate the maximum flooding vapor. The abscissa and ordinate are presented in Eq. (5) and Eq. (6) [64], respectively.

《Fig. 8》

Fig. 8. Generalized pressure drop correlation [64].

In Eq. (5), F_LV is a flow parameter. For the absorber, the liquid feed is the lean solvent. Its flow rate can be estimated using Eq. (7) [21].

where F_Lean is the mass flow rate of the lean solution; F_Flue is the mass flow rate of the flue gas; x_CO2 is the mass fraction of CO₂ in the flue gas; ψ_CO2 is the required carbon capture level; M_MEA is the molar weight of MEA; α _Rich and α_Lean are the CO₂ loading () in rich solvent and lean solvent, respectively; and ωMEA is the MEA mass fraction in solvent.

From Eq. (6) , V^*_W (vapor mass flow rate per unit cross-sectional area) is calculated; next, the total cross-sectional area can be obtained given the flue gas flow rate. In this equation, K₄ is a load parameter obtained from Fig. 8, according to the value of F_LV and the specified pressure drop; F_P is a packing factor.

In order to achieve good liquid and gas distribution and to avoid flooding inside packing beds, a pressure drop of 15–50 mmH₂O^† per meter of packing for absorber and stripper was recommended [64]. In this study, a maximum pressure drop per unit height of 20.83 mmH2O [65] was used, considering the formation of the MEA solvent [21]. It should be noted that the design of the column internals, such as gas/liquid distributors and re-distributors, is crucial in order to ensure good gas and liquid distribution inside the absorber and regenerator for such large diameters.

^† 1 mmH₂O = 9.8066136 Pa.

《6.2. Carbon capture from a 250 MWe CCGT power plant》

6.2. Carbon capture from a 250 MWe CCGT power plant

In this case study, the carbon capture plant was scaled up to handle the flue gas from the 250 MW_ecombined cycle gas turbine (CCGT) power plant described in Ref. [66]. For comparison purposes, the input conditions of the flue gas and the operating conditions of the columns were chosen to be same as for the case without exhaust gas recirculation (EGR) in Ref. [66]; these conditions are presented in Table 14.

《Table 14》

Table 14 Boundary conditions of the solvent-based PCC process.

It is noticeable that the upper limit of the pressure drop per height of packing bed is 42 mmH₂O in Ref. [66]; at this value, the columns may have serious flooding [21]. In this case study, the flooding factors of the columns were set up to 65%. Another consideration is that structured packing is preferred for a large-diameter absorption column, due to the possibility of serious mal-distribution of both the liquid and vapor phases inside the random packing bed [67]. Therefore, Mellapak 250Y packing, which is regarded as an appropriately structured packing type for industrial deployment [57], was chosen for the packing beds inside the absorber and the stripper in this case study.

The first-guess diameters of the absorber and the stripper can be calculated using the method presented in Section 6.1. Starting from that point, these parameters were simulated using the improved rate-based model developed in Aspen Plus^®. Table 15 summarizes a comparison of the results of this study with those presented by Canepa et al. [66], in terms of equipment size and process parameters. It shows that the design of this study achieved smaller equipment size and lower thermal duty.

《Table 15》

Table 15 Comparison results between this study and the literature.

^aTwo columns with same diameter of 9.5 m.

One significant difference is that the packing heights in both the absorber and the stripper in this study are significantly smaller than those in Ref. [66]. One contributor is that higher efficiency structured packing was used in this study, which allows a lower height of transfer unit (HTU); the total packing height was then reduced, while keeping the same number of transfer units (NTU). At the same time, using the more accurate model provides more confidence in the simulation results, such that a conservative margin may not be needed. After all, capture levels reached as high as 90%–95% during experiments with short packing beds for both the absorber and the stripper of the pilot plants. In the experimental study by Dugas [15], the packing heights of both the absorber (with IMTP No. 40 random packing) and the stripper (with Flexipac 1Y) are 6.1 m. Notz et al. [16] reported the packing heights of the absorber and the stripper (with Mallepak 250Y) to be 4.2 m and 2.25 m, respectively, for their pilot plant.

It is also noticeable that the specific duty in this study is lower than that in Ref. [66]. The results in Table 15 show that the rich loading of this study is slightly higher, which may reflect the impact of thermodynamic modeling. With the same CO₂ concentration in the flue gas, the solubility of CO₂ that is calculated by this improved thermodynamic model may be slightly higher when close to its saturation. This situation also results in a smaller flow rate of the solvent entering the stripper. Thus, the heat requirement for solvent evaporation in the stripper decreases. At the same time, using more accurate kinetics for the reactions inside both the absorber and the stripper could also have a large impact on the predictions of heat requirement for solvent regeneration, although it is hard to dissect those highly nonlinear relations between those factors.

《7.Conclusions》

7.Conclusions

This paper presented a study on the development of an accurate rate-based steady-state model in Aspen Plus^®, with some elements implemented in Fortran subroutines, for the MEA-based carbon capture process, along with a case study using this model. It was found that the correlations of the thermodynamic model have a significant impact on the prediction accuracy of the VLE of the MEAH₂O-CO₂ mixture. A new combination of correlations was selected in this study, and shows better prediction performance. Following this step, the model from Ref. [14] was used, by coding a Fortran subroutine in order to improve the prediction accuracy of the liquid mixture density. The rate-based process model was improved by setting different kinetics for the reverse carbonate formation reactions in the absorber and the stripper, respectively, and by coding a Fortran subroutine for the effective gas-liquid interfacial area using the model from Ref. [61]. The model validation results show that the model predictions appear to be in good agreement with the experimental data from the pilot plant.

Using this accurate model, a case study was carried out for carbon capture fitted to a 250 MW_e CCGT power plant. The results show that this study achieved smaller equipment size and lower energy consumption than the previous study; these results may translate into significant savings in both capital investment and utility cost for the carbon capture plant.

《Acknowledgement》

Acknowledgement

This work was supported by the EU FP7 Marie Curie International Research Staff Exchange Scheme (PIRSES-GA-2013-612230).

《Compliance with ethics guidelines》

Compliance with ethics guidelines

Xiaobo Luo and Meihong Wang declare that they have no conflict of interest or financial conflicts to disclose.

《Nomenclature》

Nomenclature

C Correlations for property calculations

E_j Activation energy

F_p Packing factor

F_LV Flow parameter

H Henry’s law constant

k^o_j The pre-exponential factor

K_j Chemical equilibrium constants of reaction j

L Total liquid flow rate in mass

P Pressure

R Ideal gas constant

r_j Reaction rate of reaction j

T Temperature

V Total vapor flow rate in mass

V^*_W Vapor mass flow rate per unit cross-sectional area

x Liquid-phase model fraction based on true species, molecular and ionic

y Vapor-phase mole fraction

《Greek Letters》

Greek Letters

α CO₂ loading in lean solvent or rich solvent

α_ij Reaction order of component i in reaction j

ρ Density

ψ_CO2 CO₂capture level

ω_MEA MEA mass fraction in solvent

g Activity coefficient

v Stoichiometric coefficient of each component in reactions

《Superscripts》

Superscripts

o Standard state

《Subscripts》

Subscripts

CO₂ CO₂ component

Flue Flue gas

Lean Lean solvent

L Liquid phase

MEA Monoethanolamine

V Vapor phase

i Reactant component

j Chemical reaction

n Product component

Show More