Interest of Membrane Contactors for CO2 Stripping

Get Complete Project Material File(s) Now! »

Membrane Contactors for Post-combustion Carbon Capture

Membrane contactors are considered as one of the most promising strategies for inten-sified gas–liquid absorption processes. The key concept of a membrane contactors lies in making use of a permeable, hydrophobic membrane material that separates the gas and the liquid streams in a gas absorption process. Due to the membrane, a supplementary mass-transfer resistance needs to be considered, when compared to a classical gas–liquid absorption situation in which there is direct contact between the two phases. However, possible development of much higher interfacial areas per unit volume than current refer-ence technologies, i.e., gravity-driven tray- or packed columns, is expected to significantly counterbalance the increase of the mass-transfer resistance, making significant process in-tensification possible. For instance, in acid gas scrubbing, significant reduction factors of equipment size, ranging from 10 to 30, have been estimated [9], [10].
Among the various potential utilizations of membrane contactors for gas–liquid absorp-tion processes that have already been explored, post-combustion carbon dioxide capture by absorption in a chemical liquid absorbent are currently the most intensively investi-gated application [11]–[13]. For this process, the major development targets are decreasing equipment capital cost and size, and minimizing the energy requirement in order to meet technical and economic constraints. Most investigations to date have focused on the absorption section of the process [11], [14]. Indeed, this step is characterized by mild operating conditions, i.e. near ambient temperatures and almost atmospheric pressure, suitable for membrane use. For the regen-eration step by means of vapour stripping, higher temperatures are required. Hence, polymeric membranes contactors are difficult to implement as available choices of suitable membrane materials able to withstand such conditions are limited [15]–[17]. In order to reduce the operating temperature, regeneration through membrane vacuum processes has been studied [18]–[20]. More recently, using membrane contactors has been suggested as an additional step in the absorption/desorption loop, dedicated to process enthalpy recov-ery by means of water evaporation or condensation [21], [22].
Various liquid formulations have been used in membrane contactors, mainly aimed at minimizing the process energy requirement. However, in the selection process, membrane compatibility issues pose an additional constraint [23]–[27].

Module and Fibre Characteristics

The membrane may be implemented in the contactor as either a flat or a spiral wound sheet, or as a bundle of hollow fibres. The latter arrangement is preferred for gas-liquid contacting, as it allows for developing the highest interfacial areas per unit volume.
Hollow fibre membrane contactors (HFMC, Figure 1.5) are made up of a bundle of hollow fibres arranged in a geometrical array inserted in a shell. The bundle is supported by a sealing ring. One of the fluids flows through the lumen of the hollow fibres, whereas the other flows through the shell surrounding the fibres. Analogous to shell and tube heat exchangers, the flow in the shell of HFMC can be orientated in different directions with respect to the direction of the flow through the fibre lumen. In Co- and counter-current parallel flow, both crossflow and multipass arrangements are encountered. In the latter, the fluid passing through the lumen is fed to an adjacent fibre bundle (e.g. two passes, as shown in Figure 1.5.b).
In a laboratory environment, the fibre bundle can stand by itself. However, an industrial membrane contactor for post-combustion carbon capture would contain thousands of fibres grouped in the same bundle. Therefore, a grid must be included in order to support the entire bundle during sealing, which also maintains the bundle structure [28].
Typical values of HFMC key geometrical characteristics used in industry are presented in Table 1-1.

Advantages and Drawbacks of Membrane Contactors

The principal advantages and drawbacks of HFMC technology in post-combustion car-bon capture are summarized in Table 1-3. The intensification potential of this technology resides essentially in the large interfacial area per unit volume, roughly 20 times higher than that attained in packed columns, which would eventually lead to equipment size reduction. This confers important advantages, particularly in offshore applications where weight and volume considerations play a key role. Another interesting feature of hollow fibre membrane contactors lies in the physical separation of the gas and liquid phases, thereby avoiding the flooding and liquid entrainment that limits the operational range of packed columns. The main issue that has to be addressed in order to develop the technol-ogy is the stability of the membrane mass-transfer resistance under severe operating con-ditions and over long-term operation times. Indeed, the membrane mass-transfer resistance can be tremendously enhanced through membrane aging, wetting or solid plugging. The scale-up of the HFMC is an a priori straightforward numbering-up procedure. How-ever, the significantly high throughputs involved in post-combustion carbon dioxide cap-ture lead to important distribution issues. But this is also the case in conventional tech-nology, i.e. packed columns. As a result, the scale-up problematic is not a discriminating feature between technologies.

Investigations on the Pilot-Plant Scale

In the context of industrial application, hollow fibre membrane contactors have been in use for the past decade in various pilot plant tests performed in the Netherlands, France, China and Australia. Available operating conditions together with the most important results of pilot plant tests are summarized in Table 1-8. Mainly commercial contactors were used for these tests (Liqui-cell ® and Celgard ® for PP and Polymem contactor for PTFE). The operation mode was consistently counter-current, while the local flow was either transversal or parallel, depending on the technology used. The total membrane surface varied between 0.27 and 24 m2, as much as 200 times higher than that used on the laboratory scale.
Most of the investigations were performed using aqueous solutions of pure MEA or reactant mixtures containing MEA. However, Scholes et al. [33] used a commercial absorp-tion liquid formulation and Feron et al.[23] used a patented formulation. Lab-gases as well as flue-gases were used. The investigations were performed by setting lower liquid-to-gas flow rate ratios than those found in the laboratory scale investigations, with some of them using partially loaded liquid absorbents. These conditions combined with high membrane areas correspond to more severe conditions than those applied on the laboratory scale, but were not as severe as would be found in industrial application.
Earlier work conducted in a pilot plant focused on economic membrane material, i.e. PP and PVDF, and diluted solutions in order to reduce investment costs while preventing wetting [23], [27]. Despite this fact membrane wetting was frequently observed. As dis-cussed in Section 1.2.5, laboratory experiments showed that the more expensive PTFE fibres can prevent wetting when operating with low liquid reactant concentrations [38]. However, low concentrations are not relevant from an economical point of view. More recent pilot investigations were performed using PTFE membrane contactors together with high reactant concentrations. Again, membrane wetting was reported. Further still, Scholes et al. [33] and Chabanon et al.[63] noticed liquid droplets in the outflowing gas.
Another issue addressed by the different investigators is pertinent to shell-side distribu-tion, whether or not the liquid or the gas-phase flows through this compartment. Axial dispersion, channelling, as well as dead-zones appear to reduce mass-transfer performances.
In most of the investigations, the average overall mass-transfer coefficient is not reported. However, its value seems to be relatively low which is likely linked to membrane wetting. The average CO2 absorbed flux per unit volume is comprised between 0.6 and 4 molm-3s-1 which is still higher than that reached in industrial packed columns operation [64]. This information should be considered with some reserve, as lean liquid loading as well as liquid and gas reagent conversions attained in HFMC pilot plants are low in general when com-pared to industrial scale packed-column operations.

Thermodynamic Models of CO2 Solubility

The thermodynamic equilibria of the CO2-N2-MEA-H2O system is illustrated in Figure 2.1. The molecular species in a gaseous state are absorbed or condensate at the vapour-liquid interface, at which point they react in the liquid phase until reaching chemical equilibrium. Therefore, thermodynamic models consist in the description of both vapour-liquid and chemical equilibrium. Thermodynamics and reaction kinetics of carbon dioxide absorption in aqueous amine solutions Optimal modelling strategies require simplicity and accuracy at every stage, including the thermodynamic description of the system. Models with too many parameters are com-plex, time consuming and cumbersome. That said, models which are too simple in form can provide less accurate results. Thus, there is a trade-off between complexity and accu-racy.
The simplest model links gaseous CO2 concentrations proportionally to liquid concen-trations solely by means of a solubility constant (Henry s law), which is estimated using empirical correlations [68]. This approach cannot be applied to kinetically controlled sys-tems, since chemical reactions make concentrations near the interface different to those of the bulk liquid. In addition, these models cannot estimate further properties such as amine volatility, heat of absorption or the distribution of chemical composition in the liquid phase. The latter is known as true liquid speciation and is necessary to compute reaction rates.
The true speciation can only be estimated by thermodynamic models of these systems. It can be computed by simultaneously solving the equations corresponding to the molar balances of each species, the electric charge balance and the chemical equilibrium associ-ated with each chemical reaction [69]. Thus the problem resides in the estimation of the chemical equilibrium, which can be described using chemical equilibrium constants [70]. The equilibrium constants available in literature, usually expressed in polynomial form as a function of temperature, correspond to those of diluted species, relating the concentra-tions (e.g. molar concentration or molar fraction) of the reactants and products [71]. True speciation calculation using these equilibrium constants leads to significant deviations from the experimental data, within the interval of interest for concentrations [72].
Thermodynamic models based on activity coefficients take into account mixture inter-action effects, thus correcting deviations that occurred using expressions for diluted species. Indeed, these models are based on the excess Gibbs free energy formulations [71], [73]–[75]. The mixture interactions may be classified by local (combinatorial and/or residual)6 and long-range contributions (Debye-Hückel ion-ion interactions). The simplest of the excess Gibbs free energy thermodynamic approaches which has been used successfully for systems such as amine7-CO2-H2O, is based on the specific ion interaction theory (SIT), which only takes into account long-range contributions. However, this approach has not yet been used within rate-based transfer models for CO2 absorption, where a physically meaningful rep-resentation of the mixture interactions is believed to be crucial. Moreover, prediction of additional properties, such as heat of absorption, is not clear and has not been proved to be suitable. Adding to their complexity, electrolyte-NRTL and UNIQUAC models, which Thermodynamics and reaction kinetics of carbon dioxide absorption in aqueous amine solutions take into account the group of interactions mentioned above, are able to predict the true speciation as well as all thermodynamic properties with physical meaning, and thus are reliable means for extrapolating. Moreover, application of both approaches within rate-based transfer models has been proved to be suitable [64], [76], [77]. Nevertheless, fitting for a large number of parameters required for these models has been found to be compli-cated. Further, the results are strongly dependent on the experimental methods used to obtain the required data.
Despite the different formulations, the excess Gibbs free energy models have the follow-ing points in common: 1) iterations are required since the activity coefficients depend on true speciation, which itself depends on these coefficients, 2) equations of molar balances, electric charge balance and chemical equilibrium associated with each chemical reaction need to be solved and their convergence is strongly sensitive to the initial point, and 3) in addition to the interaction parameters, it has been found that the equilibrium constants have also been fitted . In order to prevent iterations and thus reducing calculation time of the CO2 absorption model with amine aqueous solutions by means of membrane con-tactors, Hoff et al. [50], [78] have stated that, with an adequate formulation, speciation can be solved in a straightforward manner. In addition, they introduced activity coeffi-cients in the form of polynomials in order to tune the thermodynamic model, fitting ex-perimental data. By contrast, Aboudheir et al. [69] introduced apparent equilibrium con-stants to take into account the nonidealities of the system and rigorously solved the system of equations corresponding to speciation. Similarly, Puxty et al. [73] adjusted the diluted species equilibrium constant but included computation of long-term interactions. The mod-els of Hoff et al., Aboudheir et al. as well as Puxty et al. have shown excellent agreement between the experimental data and the model predictions, with significantly fewer param-eters to fit than more complex models while maintaining a physical representation.
The approach implemented in this present work is an adaptation of the three investiga-tions previously mentioned. Indeed, the formulation to solve the chemical speciation with-out iterations is employed, and the equilibrium constants are fitted to lump the nonideal-ities of the system. The introduction of long-term interactions had no positive effect on the accuracy of the model. The approach is detailed in the following sections.

Heat of Absorption of CO2

An important thermodynamic parameter in gas treatment processes is the heat of CO2 dissolution and the reaction between CO2 and amine. These are often combined and re-ferred to as the heat of absorption. The magnitude of the heat of absorption is a significant factor in determining the thermal effects observed in both absorption and stripping sec-tions. Since CO2 solubility, reaction kinetics and diffusion coefficients are strongly tem-perature dependent, estimating the heat of absorption is essential for predicting the per-formance of both absorption and desorption processes.
Several model complexities of heat of absorption are proposed in literature, varying from constant values [82], empirical correlations [68] and more elaborated excess enthalpy mod-els [74], [75], [83]. The constant values and empirical correlations are not adequate for modelling the process in conditions outside the range of validity, especially for high tem-peratures. Models with physical meaning, such as those for excess enthalpy, are preferred when extrapolations are required. The approach used in this work to estimate the heat of the absorption of carbon dioxide in aqueous amine solutions, is based on the formulation proposed by Kim et al. [84]. Indeed, while it is not an excess enthalpy model, its thermodynamic formulation is sufficiently solid to allow for extrapolations and was appropriate for implementation into the present work, given that the solubility model approach was used. The concept of the approach is that each chemical reaction involved in the speciation, contributes to the overall enthalpy of the CO2 absorption.

Equilibrium Constants for CO2 Solubility

The parameter values of the equilibrium constants obtained by the parameter adjust-ment are highlighted in bold in Table 2-19. The parameters corresponding to K1, K2 and K3 are taken from [71], since they are not adjusted. Regression analysis results are consid-erably different from those for equilibrium constants of diluted solutions published in oth-ers studies [71], [73]. The difference may be attributed to two factors: first, the nonidealities of the system were lumped into the expressions of K4 and K5 and second, the minimization is strongly dependent on the initial guess. Fluctuations in the results due to the second factor were smoothed by setting the literature values, those of the diluted solution, as for the initial estimation.

Equilibrium Constants for CO2 Heat of Absorption

The influence of excess enthalpy on the enthalpy of absorption in the MEA-H2O-CO2 system is known to be significant [75]. The most reliable method for predicting this pa-rameter is by implementing a Gibbs free energy thermodynamic model. Using the equilib-rium model featured in this work which considers apparent equilibrium constants (with no activity coefficient), expressions fitted for CO2 solubility would not be expected to account for this excess property. Therefore, temperature dependence of the equilibrium constants of the amine protonation and the carbamate formation were fitted to experi-mental data. A similar procedure was used in [84].
To perform the parameter adjustment, the objective function (Equation 2.49) was mod-ified by replacing the CO2 partial pressures with the heat of absorptions. The experimental data here were taken from [88]. The results are shown in Erreur ! Source du renvoi in-trouvable.. Figure 2.3 illustrates model predictions for different temperatures. Despite the simplicity of the model, it produced acceptable predictions for the variation of heat of absorption with temperature and CO2 solvent loading. Extrapolations at this stage are considered suitable due to the thermodynamic fundamentals of this procedure.

Table of contents :

Chapter 1. General Context
CO2 Emissions, Capture and Storage
1.1.1. CO2 Emissions
1.1.2. CO2 Capture and Storage
1.1.3. CO2 Post-combustion Capture by Chemical Absorption
Membrane Contactors for Post-combustion Carbon Capture
1.2.1. Module and Fibre Characteristics
1.2.2. Membrane Characteristics
1.2.3. Advantages and Drawbacks of Membrane Contactors
1.2.4. Membrane Wetting
1.2.5. Investigations on the Laboratory Scale
1.2.6. Investigations on the Pilot-Plant Scale
Summary and Conclusions
Résumé et conclusions
Chapter 2. Thermodynamics and reaction kinetics of carbon dioxide absorption in aqueous amine solutions
Thermodynamic Models of CO2 Solubility
Vapour-Liquid Equilibria
Chemical Equilibrium
2.2.1. Chemical Reactions Involved
2.2.2. Mathematical Formulation based on the Extent of Reaction
Heat of Absorption of CO2
Parameter Estimation of Equilibrium Constants
2.4.1. Equilibrium Constants for CO2 Solubility
2.4.2. Equilibrium Constants for CO2 Heat of Absorption
Auxiliary Operations of the CO2 Capture Process
Reaction Kinetics of the CO2-MEA-H2O System
Summary and Conclusions
Résumé et conclusions
Chapter 3. Isothermal Modelling of Membrane Contactors
Mass-transfer Modelling of Microporous Membranes
3.1.1. Convection inside Membrane Pores
3.1.2. Membrane Mass-transfer in Wetted Membranes
Isothermal One- and Two-Dimensional Modelling of HFMC
Summary and Conclusions
Résumé et conclusions
Chapter 4. Adiabatic Modelling of Membrane Contactors
Adiabatic One-Dimensional Modelling of HFMC
Verification of the Rate-Based Model of the Contactor
4.2.1. Packed Column Model Equations
4.2.2. Simulation Results and Model Validation
Influence of Total Trans-membrane Flux on the Model Predictions
Adiabatic Two-Dimensional Modelling of HFMC
4.4.1. Model Assumptions
4.4.2. Model Equations and Boundary Conditions
4.4.3. Simulation Results in an Industrial Context
4.4.4. Comparison of Experimental and Simulation Results
Influence of Axial Distribution of the Membrane Mass-transfer Coefficient.
Flow Configuration Analysis
4.6.1. Co-current Flow Arrangement
4.6.2. Liquid in-shell Flow Configuration
Summary and Conclusions
Résumé et conclusions
Chapter 5. Interest of Membrane Contactors for CO2 Stripping
Stripping using Packed Columns
5.1.1. Model Validation
5.1.2. Simulation Results of the Energetic Optimum
High Temperature Stripping using Membrane Contactors
5.2.1. Simulation Results of HTS
Low Temperature Stripping using Membrane Contactors
5.3.1. Simulation Results of LTS
Summary and Conclusions
Résumé et conclusions
General Conclusions and Outlook
General Conclusions
Outlook
Conclusions générales et perspectives
Conclusion générale
Perspectives
Bibliography

GET THE COMPLETE PROJECT