Energy-efficient solvent regeneration in enzymatic reactive absorption for carbon dioxide capture

Applied Energy xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Applied Energy journal homepage: www.elsevier.com/locate/apenergy

Energy-efficient solvent regeneration in enzymatic reactive absorption for carbon dioxide capture ⁎

Mathias Leimbrinka, , Stephanie Sandkämpera, Leigh Wardhaughb, Dan Maherb, Phil Greenb, Graeme Puxtyb, Will Conwayb, Robert Bennettb, Henk Botmab, Paul Feronb, Andrzej Góraka,c, Mirko Skiborowskia a b c

TU Dortmund University, Laboratory of Fluid Separations, Emil–Figge–Straße 70, 44227 Dortmund, Germany CSIRO Energy, PO Box 330, Newcastle, NSW, Australia Lodz University of Technology, Faculty of Process and Environmental Engineering, Department of Environmental Engineering, Wólczanska 213, Lódz 90-924, Poland

H I G H L I G H T S experimental investigation of solvent regeneration with loaded MDEA solutions. • Extensive IR analytic method successfully applied to loaded MDEA solutions. • Innovative model showed good agreement with experimental data. • Developed • 40% improvement in energy requirement compared to MEA as baseline solvent.

A R T I C L E I N F O

A B S T R A C T

Keywords: Solvent regeneration Pilot scale testing Rate-based modeling Energy efficiency Carbon capture Enzyme carbonic anhydrase

Although recent studies on the application of enzyme-catalyzed reactive absorption of carbon dioxide (CO2) with thermodynamically favorable solvents such as tertiary amine N-methyldiethanolamine (MDEA) have demonstrated competitiveness with kinetically favorable solvents such as primary amine monoethanolamine (MEA), experimental data on the desorption of CO2 in MDEA are scarce. However, these data are necessary to validate the energetic benefit expected from an enzyme-catalyzed reactive absorption process with an aqueous MDEA solvent. To bridge this gap, the current work presents the experimental results of aqueous MDEA solvent regeneration at the pilot scale with consideration of different solvent flow rates, CO2 loadings and applied reboiler duties. Furthermore, a process model that accurately describes the experimental data was developed to evaluate the energy requirements in a closed-loop absorption-desorption process. For this purpose, the desorption process model was extended using a previously validated enzymatic reactive absorption model to determine the energy efficiency of the overall enzymatic reactive absorption-desorption process. Although the MEA benchmark pro−1 , it was found that this value could be reduced cess requires a specific reboiler duty of approximately 3.8 MJ ·kgCO 2 −1 by more than 40% to 2.13 MJ ·kgCO with use of the enzymatic reactive absorption process based on aqueous 2 MDEA solvent.

1. Introduction A strong decrease in the greenhouse gas emissions from various industrial processes, especially fossil-fueled power plants, is a major climate goal that is expected to occupy academia, industry and policymaking in the coming decades [1]. The need to reduce greenhouse gas (GHG) emissions such as carbon dioxide (CO2) by 80–95% before 2050

relative to the 1990 emissions has been proposed by the European Union as a roadmap to a low-carbon economy [2]. Innovative process concepts and novel routes must be developed to meet this ambitious goal while simultaneously enabling reliable and environmental-benign generation of target products and services. In addition, existing plants must be retrofitted with efficient technologies capable of mitigating climate relevant emissions [3]. In this scenario, post-combustion carbon

Abbreviations: CO2, carbon dioxide; CSIRO, Commonwealth Scientific and Industrial Research Organisation; DI, deionized; DOR, degree of regeneration; GHG, greenhouse gas; IR, infrared; MDEA, N-methyldiethanolamine; MEA, monoethanolamine; N2, nitrogen; NEQ, non-equilibrium; PCC, post combustion carbon dioxide capture; PDF, process development facility; PLSR, partial least-squares regression; SRD, specific reboiler duty; VLE, vapor-liquid-equilibrium ⁎ Corresponding author. E-mail address: [email protected] (M. Leimbrink). http://dx.doi.org/10.1016/j.apenergy.2017.10.042 Received 20 June 2017; Received in revised form 7 September 2017; Accepted 6 October 2017 0306-2619/ © 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: Leimbrink, M., Applied Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.10.042

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

Decarbonized Ňue gas

Condenser CO2 Cooler

Absorber

Lean solvent

Fig. 1. Simplified process scheme of typical amine-based reactive absorption processes.

Rich solvent

Desorber

Heat exchanger

Flue gas

Rich solvent

Regenerated solvent

Reboiler

rate because a certain temperature difference between inlet and outlet streams of a desorber column is inevitable due to the underlying separation principle of reactive absorption, which is exploitation of the temperature-dependent solubility of CO2 [10]. The solvent flow rate can be reduced if the CO2 loading capacity of the solvent is increased. The other two contributors can also be considered as important factors of influence when aiming for more energy-efficient solvent regeneration and thus lowering the energy requirement and capture costs associated with CO2 capture. Referring to the detailed energetic assessment reported by Feron [11], the potential improvement in the energy performance of CO2 capture processes is quite large. Therefore, significant reductions in the energy requirement can be expected from future generations of innovative absorption-based capture technologies, and according to Feron, bicarbonate forming solvents such as tertiary amines and alkalicarbonates offer a promising alternative [11]. In particular, tertiary amine N-methyldiethanolamine (MDEA) has several advantageous thermodynamic properties compared with primary amines, e.g., MEA and is known to operate well based on industrial experience from other acid gas removal processes [12]. It is worth mentioning that development of novel energy-efficient solvents such as 1-diethylamino-2-propanol [13,14] or the application of acid solid catalysts that facilitate amine-based desorption processes [15–17] can further contribute to reductions in the energy requirement for CO2 capture processes. MDEA has a significantly lower absorption enthalpy −1 (ΔHabs,CO2 ≈ 55 kJ ·molCO ) compared with the value for MEA (approx. 2 −1 ΔHabs,CO2 ≈ 85 kJ ·molCO2 ) [18]. As clearly explained by Oexmann and Kather, a solvent with a low absorption enthalpy alone does not always lead to improved energy efficiency [10]. However, MDEA exhibits favorable temperature-dependent CO2 solubility, which enables significant reductions in the amount of steam needed as a stripping agent in the desorber (lower Q vap,H2 O ) as well as high cyclic loading capacities that consequently reduce the required solvent flow rate [8,19]. Despite such promising thermodynamic properties, absent a promoter such as piperazine, aqueous MDEA is usually not considered as a viable option for CO2 separation from flue gases due to slow reaction kinetics, which currently limit its application to processes with high CO2 partial pressures (e.g., natural gas sweetening) [12]. To make MDEA ratewise competitive in a CO2 capture from power plant flue gas scenario, several research groups have successfully demonstrated that high reaction rates can be achieved using carbonic anhydrase enzyme as a biocatalyst [9,20–26]. Although the majority of these studies have investigated the absorption performance of enzyme-accelerated absorption solvents, virtually no attention has focused on actual experimental evidence and evaluation of the assumed energy-efficient solvent regeneration. Therefore, the current work addresses this lack of

dioxide capture (PCC) represents the most mature mitigation technology for simple and effective retrofit of existing power plants. Currently, most PCC processes use amine-based reactive absorption in capture unit operation [4]. Industrial PCC plants are few in number, and most are located in North America where the captured CO2 can be used in enhanced oil recovery. A simplified process scheme the aminebased reactive absorption process is shown in Fig. 1. However, as recently noted by the International Energy Agency, a large discrepancy still exists between the planned reductions in GHG emissions and the currently applied PCC processes that consequently endanger the achievement of global climate goals [5]. The major reason for this lack of application is the large difference between the capture costs of CO2 and the value of CO2 certificates. Although the latter concept was originally introduced to promote CO2 capture, it is currently its biggest handicap. Therefore, to increase the number of PCC applications, more energy-efficient CO2 capture processes must be developed to make capture costs more competitive compared with CO2 certificate prices. Among the few currently applied amine-based reactive absorption processes, which use primary or piperazine promoted amine solvent systems, the high energy requirement for solvent regeneration is an important factor because it could add up to 80% to the costs of electricity [6,7]. Despite its fairly detrimental thermodynamic properties, especially low cyclic loadings and high heat of reaction, monoethanolamine (MEA) is accepted as a baseline solvent because it offers high absorption rates that lead to reduced equipment size. Nevertheless, other solvents such as tertiary amines offer much higher CO2 loading capacities and lower heat of reactions, thus enabling significant potential for energy requirement reductions [8,9]. According to Oexmann and Kather [10], the energy requirement usually supplied by the reboiler (Qreb) can be described as the sum of three main contributors:

Qreb = Qsens + Qdes,CO2 + Q vap,H2 O

(1)

•Q • •

sens : Sensible heat used to increase the solvent temperature from the desorber inlet temperature (rich solvent) to the desorber outlet temperature (regenerated solvent temperature). Qdes,CO2 : Heat of CO2 desorption (non-ideal mixing + heat of dissolution + heat of reaction). Q vap,H2 O : Heat of evaporation used to produce the portion of water steam that serves as a stripping agent for CO2 and does not recondense along the length of the column but is ultimately condensed in the top condenser.

In addition to fine tuning of the column pressure and liquid-to-gas ratios, Qsens can be optimized primarily by reduction of the solvent flow 2

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

Fig. 2. Simplified flow scheme of PDF at CSIRO.

CO2

purified gas

reflux drum cooler into wash tank

absorber

desorber heat exchanger

reboiler

raw gas RICH SOLVENT

LEAN SOLVENT

CO2-rich solvent

CO2-lean solvent

method functions by finding a linear regression model capable of linking the predicted and observed variables. In this case, the predicted variables are the individual amine concentration and the concentration of CO2, and the observed variables are the measured spectra. PLSR is a suitable technique for analysis of multivariate spectra that contain many linearly dependent contributions resulting from chemical speciation as a function of amine concentration and CO2 content. If the predicted variable (in this case, the concentration of each amine and the CO2 concentration) are arranged into a column matrix Y, and the response at specific wavelengths of associated spectra into a row matrix X , PLSR solves the following matrix equation for B such that Y can be predicted from X with a minimum error E . A data set with samples of known composition was used to build the model and determine the elements in B , known as the calibration data set. Once B is determined, it can be used to estimate Yunk from the spectra of samples of unknown composition Xunk .

information and evaluates the potential energy reduction by means of experimentally validated models for enzymatically catalyzed absorption and the absorption-desorption process in a closed loop. First, an experimental investigation into solvent regeneration of loaded MDEA solutions was performed in a pilot-scale testing facility in accordance with loadings that can be obtained by enzymatic reactive absorption of CO2 from flue gas. Second, the experimental data were used in model validation of a developed process model in Aspen Plus®. Finally, this model was applied to a distinct process analysis with the aim of determining the minimal energy requirement for loaded MDEA solvents. 2. Materials and methods This section describes the materials used, analytical methods implemented, and equipment applied for performance evaluation.

Yunk = Xunk ·B

2.1. Materials

(2)

Further details on this method can be found in a recent article published by Puxty et al. [28]. In addition, gas phase analytics were conducted using two VA3000/VS-3000 single-component analyzers (Horiba) with measurement ranges of 0–100 vol% CO2 and 0–25 vol% CO2, respectively. The composition of the feed gas stream entering the absorber and the outlet gas stream leaving the absorber at the top was continuously analyzed.

The solvent was an amine solution based on tertiary amine MDEA purchased from Huntsman (batch purity: 99.3 vol%) and diluted with water to amine concentrations of 30 wt% MDEA. An MDEA concentration of 30 wt% was determined to be most effective for enzymatic CO2 capture in a previous study [27]. CO2 (purity ≥99.9 vol%) and nitrogen (N2) (purity ≥99.95 vol%) were sourced from storage tanks. Deionized (DI) water was supplied by the local DI water grid.

2.3. Process development facility 2.2. Analytics The process development facility (PDF) used in the experimental investigations in this study was established at the Energy Centre of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) in Newcastle, Australia. This facility mainly serves to investigate novel and promising solvent systems in an industrially representative technical environment. The PDF consists of two packed columns, one for absorption and one for desorption, which were operated with synthetic flue gas supplied from liquefied bulk storage. The plant was designed for a total gas flow rate of 72 Nm3 h−1 with a CO2 capacity of 12 Nm3 h−1. A small water flow was added to the synthetic

The concentrations of MDEA and CO2 in liquid samples were determined using a combination of infrared (IR) spectroscopy (with an attenuated total reflectance probe) and partial least-squares regression (PLSR) [28]. This technique is based on a collection of spectra of a calibration dataset in which the composition of each sample is known. A PLSR model was built using this dataset, which allowed the amine and CO2 content of an unknown sample to be predicted from its IR spectrum. PLSR is a powerful technique for analysis of complex data sets. The 3

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

Condenser

discharged from the desorber could be calculated as follows

CO2 + H2O (v) Rich solution Tdes,top

CO2

Gas,out

)·ρCO2

(3)

with ρCO2 representing the density of CO2 and

H2O (l)

Rich

Gas,in

Gas,in Gas,out ̇ ̇ ṁ CO2,des = ṁ CO2,abs = ṁ CO −ṁ CO = (VCO −VCO 2 2 2 2

V,top

pCO2,top

Gas,in

̇ VCO 2

Gas,in

in ̇ = vCO ·Vtotal , 2

(4)

and Gas,out

̇ VCO 2

out out ̇ Gas,out = vCO ̇ Gas + VCO ̇ Gas,out ) = vCO ·VTotal ·(Vinert 2 2 2 Gas,in

Gas,out

out in ̇ ̇ = vCO ·(VTotal ·(1−vCO ) + VCO 2 2 2

Tdes,bot

Lean

Lean solution

V,bot

).

(5)

In this equation, V̇ [m3·s−1] represents the gas volume flow, and v [vol. %] is the gas volumetric fraction. For validation of the mass balance, liquid samples collected from the rich and lean solvent line were further analyzed via the IR method with respect to the present loading. This process allowed for additional evaluation of ṁ CO2,des based on the determined rich and lean loadings −1 −1 αRich [molCO2·molMDEA ] and αLean [molCO2·molMDEA ] as follows

pCO2,bot

Reboiler

̇ ṁ CO2,des = nMDEA ·(αRich−αLean)·MCO2,

(6)

with consideration of the molar flow of MDEA

Fig. 3. Parameter changes in the desorber and corresponding energy requirements during solvent regeneration. Based on Schäffer (2013).

̇ nMDEA =

raw gas inlet stream via an evaporator. Fig. 2 shows a simplified process flow diagram of the PDF and indicates the stages in which the streams enter or leave the columns. Because this study focuses on solvent regeneration, the absorber was used only as auxiliary equipment to produce a steady flow of solvent with a pre-defined CO2 loading. Along the height, both columns were segmented into two different section types, i.e., a packed section filled with 16 mm metal Pall rings and a non-packed sample section that allowed for the installation of temperature probes and gas sample ports. The section types were arranged in an alternating sequence starting and ending with a sample section. The absorber column consisted of 12 packing and 13 sample sections, and each had an identical height of 0.308 m and a diameter of 0.148 m, corresponding to a total height of 7.70 m. The desorber had a setup similar to that of the absorber with 9 packing and 10 sample sections installed, yielding a total height of 5.85 m. CO2-loaded solvent (rich) and regenerated solvent (lean) were sampled at the liquid outlet lines from the bottom of the absorber and desorber, respectively. The bottom of the desorber was connected to a reboiler via a pump, which circulated a defined flow rate of solvent through the reboiler and back into the desorber bottom. The reboiler was supplied with heat from an external steam source to heat and partially evaporate the solvent. The vapor product leaving the top of the desorber was passed through a condenser to condense the water and separate it from CO2 in a flash drum. Both the CO2 product and purified gas were fed to a packed bed wash column to remove residual solvent before release to the atmosphere. Prior to the wash column, the mass flow of the CO2 stream was measured. The evaporated water feed and the condensate reflux to the desorber were controlled to maintain the water balance in the system.

wMDEA·ṁ lean solvent MMDEA

(7)

[kg ·kg −1]

is the mass fraction of MDEA in water and is where wMDEA adjusted to 0.3, and MMDEA [g ·mol−1] is the molar mass of MDEA. All reported values for ṁ CO2,des refer to the calculation using Eq. (3), whereas all values were validated by the results obtained from Eq. (6). The suitability of both approaches is demonstrated in Section 3.3. 2.4.2. Determination of the specific reboiler duty (SRD) As described in Section 2.3 the reboiler was operated with heat from an external steam source, for which data on flow rates and temperature was available. However, due to significant heat losses on the steam side, the heat released from the steam was not equivalent to the heat uptake by the solvent in the reboiler. Consequently, the SRD had to be determined indirectly by an energy balance around the desorber. As shown in Fig. 3, multiple parameter changes occur simultaneously over the column height during desorption and thus represent the complexity of this process step. The energetic assessments made by Oexmann and Kather [10] were previously introduced and the SRD was calculated analogously to Eq. (1) as the sum of the three individual contributors

SRD = qreb = qsens + qdes,CO2 + qvap,H2 O.

(8)

In this equation, the SRD is determined based on the specific heat −1 q [MJ ·kgCO ] calculated by relating the heat Q [MW ] to the desorbed 2 mass of CO2 ṁ CO2,des [kg ·s−1]

q=

Q . mCO2,des

(9)

In accordance with the approach described by Oexmann and Kather [10], the sensible heat qsens was determined from the temperature difference between the rich solvent entering the top of the desorber and the lean solvent leaving the reboiler at the bottom of the desorber (see Fig. 3).

2.4. Calculation methods The following sections describe the calculation of the process parameters used to evaluate the experimental results.

qsens = 2.4.1. Determination of the desorbed mass flow of CO2 Because all experiments were evaluated during steady-state operation, the mass of CO2 ṁ CO2,des discharged from the desorber had to equal the absorbed mass of CO2 ṁ CO2,abs in the absorber. The latter was easier ̇ Gas,in to access, considering the measurement of the total gas inlet flow Vtotal and the gas phase analysis of the gas streams that entered and left the absorber, as presented in Eqs. (3)–(5). Consequently, the mass of CO2

ṁ lean solvent ,out ·Tlean solvent ,out ·cP,lean−ṁ rich solvent ,out ·Trich solvent ,out ·cP ,rich ṁ CO2,des (10)

In this equation, cP [MJ ·(kg ·K )−1] represents the specific heat capacity, and T [K ] denotes the temperature. Because the sensible heat of the CO2 gas stream that leaves the desorber via the top flash drum is much smaller than the sensible heat of the lean and rich solvent, it is neglected in the calculation. The specific heat qvap,H2 O , which is required 4

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

to condense the water in the top vapor stream leaving the desorber, is equivalent to the increase in the sensible heat of the used cooling water qcond , which was calculated by the following

qvap,H2 O = qcond =

ṁ cooling water ·cP,H2 O·(Tcooling water ,out −Tcooling water ,in) ṁ CO2,des

Table 1 Experimental overview of conducted experiments.

. EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP

(11) Finally, the specific heat of CO2 desorption qdes,CO2 was determined from the literature data with consideration of the calorimetric measurements. This process is conducted under the reasonable assumption that the heat of absorption is similar to the heat of desorption under equal operating conditions for tertiary amine systems [29–32]. Therefore, based on the study by Gupta et al. [33], an integrated average −1 ] for typical desorber value of the absorption enthalpy ΔHabs [kJ ·molCO 2 temperatures in the range between 373 and 393 K was considered and described as follows −1 −1 qdes,CO2 = −ΔHabs ·MCO = 1.3 MJ ·kgCO 2 2

where MCO2

[g ·mol−1]

(12)

represents the molar mass of CO2.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Lean solvent

Rich loading

flowrate kg·h−1

−1 molCO2·molMDEA

110 110 110 110 110 150 150 150 150 150 200 200 200 200 200 200 200 200

0.36 0.37 0.37 0.41 0.41 0.22 0.22 0.27 0.35 0.35 0.19 0.27 0.31 0.32 0.32 0.32 0.34 0.35

Steam pressure kPa g 210 190 182.5 175 182.5 190 210 175 190 182.5 175 190 210 175 210 210 182.5 190

2.5. Estimation of expected statistical errors this study, the conducted experiments are representative for an application of immobilized enzyme in the absorber [27] or recovery of the enzyme, e.g., by means of ultrafiltration preceding the desorber [35]. The operating conditions resulting in rich loadings that could be theoretically achieved with enzyme present in the absorber were not easy to predict a priori for steady-state operation. Table 1 shows the finally obtained rich loadings that were realized in the absorber, and these −1 . Liquid flow rates values varied between 0.19 and 0.49 molCO2·molMDEA were adjusted to the planned values of 200, 150 and 110 kg ·h−1 in the solvent flow rate. The reboiler steam pressure was adjusted in a predefined range between 175 and 210 kPa g., which was chosen as equivalent to previous studies published by Saimpert et al. using MEA as a solvent [36].

The statistical error was quantified for each experiment based on the error of the relevant measurement instruments. Instruments subject to uncertainties in the measurement were the CO2 gas analyzers, with a specified relative accuracy of ± 2%, and the IR method used to de−1 termine the liquid loading, with a mean error of 0.0383 molCO2·molMDEA −1 and a standard deviation of 0.0345 molCO2·molMDEA . As the SRD is calculated theoretically based on abovementioned assumptions there is only a minor influence by statistical errors. The reproducibility of the measurements is verified in Section 3. 3. Experimental investigation An overview of the conducted experiments is presented to supply an accurate impression of the experimental investigations performed in this study. The obtained results and the quality of the produced data are discussed in detail.

3.2. Results of experiments For all of the experiments, determination of the desorber performance was the main objective, whereas the degree of regeneration (DOR) was determined by measuring the lean loading via the IR method and the desorbed mass flow of CO2 from gas analysis, as described in Section 2.4.1. Calculation of the DOR is given by the equation below

3.1. Overview of experiments The major objective of the experimental campaign was to conduct a sufficient number of experiments over a wide range of operating conditions to generate a meaningful database for model validation. The process variables that have a major influence on solvent regeneration were first identified as the rich loading of the solvent stream, the desorber pressure, the solvent flow rate and the reboiler duty. The desorber pressure was held constant at 198 kPa during all experiments for operational reasons. The remaining process variables were varied over a predefined range to reflect the variety of the representative absorber outlet conditions with reference to an enzymatic reactive absorption scenario for CO2 capture from power plant flue gases [21,34] in accordance with a systematic experimental plan consisting of 18 experiments, as shown in Table 1. The applied steam pressure was chosen as a surrogate for the reboiler duty, which was subsequently calculated according to the description in Section 2.4.2. The amine concentration in the aqueous solvent was fixed to 30 wt % of MDEA for all experiments. This decision was made in accordance with a previous analysis of the absorption performance of an enzymatically catalyzed reactive absorption process [21]. The rich solvent loading was difficult to precisely adjust in the experiments. Initially, −1 rich loadings of 0.35, 0.275 and 0.2 molCO2·molMDEA were considered as representative with respect to an absorption temperature of 313 K and a CO2 partial pressure in the flue gas of 15 kPa for the enzyme-accelerated absorption process. However, because no enzyme was actually used in

DOR =

αRich−αLean , αRich

(13)

where α is the CO2 loading of the solvent in apparent moles CO2 per apparent moles MDEA. The SRD was calculated from the experimental data according to the calculation method described in Section 2.4.2. The raw data used to calculate SRD are reported in the Supplementary Information. To gain further insights into the regeneration process and for model validation, temperature profiles over the height of the desorber were recorded. Exemplary temperature profiles are shown in Fig. 4 for experiments EXP 6, EXP 7 and EXP 9, which all had a lean solvent flow rate of 150 kg ·h−1 but different rich loading and steam pressure settings. As shown in Fig. 4, the recorded temperature profiles have highly similar trajectories but differ in average column temperature. The difference between EXP 6 and 7 can be explained by the higher steam pressure applied in EXP 7. Thus, additional heat was supplied, resulting in a higher temperature. The difference between EXP 6 and EXP 9 is the different rich loading. The rich loading of EXP 9 is significantly higher than of EXP 6. Hence, higher heat was used in EXP 9 to desorb CO2 than to heat up the solvent, resulting in a lower column temperature compared with EXP 6. These example temperature profiles visualize the mutual influences of rich loading, steam pressure and solvent flow rate 5

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

6

Fig. 4. Temperature profiles over column height for experiments EXP 6, EXP 7 and EXP 9.

Height / m

5 4 3 2 1 0 90

95

100

105

110

115

120

125

130

Temperature / °C EXP 7

EXP 9

Consequently, a significantly higher cyclic loading capacity can be used that ultimately reduces the required solvent flow rate. The experimental results are also shown in Fig. 5. It can be observed that the desorbed mass flow of CO2 varies between 3 and 7 kg ·h−1, and −1 with a mean value the SRD is fairly stable at values of 5 ± 0.5 MJ ·kgCO 2

Table 2 Results of the experiments performed according to conditions described in Table 1.

EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP EXP

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Lean loading

Desorbed mass flow of

−1 molCO2·molMDEA

CO2 kg·h−1

0 0 0.04 0.01 0.05 0.01 0 0.04 0.05 0.02 0.04 0.02 0 0.08 0 0 0.08 0.03

4.47 4.49 3.98 4.40 4.38 3.46 3.89 3.83 4.65 5.28 3.06 5.08 6.48 4.91 6.88 6.87 5.49 6.30

−1 SRD MJ ·kgCO 2

−1 of 4.76 MJ ·kgCO . As listed in Table 1, EXP 15 and 16 were conducted 2 under equivalent operating conditions. Therefore, those results can be used to show the reproducibility of the conducted measurements. As shown by the results supplied in Table 2, both experiments agree quite well and verify the reproducibility of the chosen experimental setup. In addition to the results shown above, the DOR calculated from Eq. (13) is shown together with the SRD in Fig. 6. The previous results and those indicated in Fig. 6 show that the solvent flow rate and the rich loading have a significant influence on the DOR and that consideration of the reboiler duty alone is not sufficient for correct analysis of desorber performance. These results support the previous finding that MDEA is easy to regenerate from an energy perspective because all experiments achieve a DOR greater than 70%, the majority exceed 90%, and certain values even approach 100%. This complete regeneration of the solvent was observed for the experiments at the highest steam pressure in the reboiler, which resulted in a temperature of greater than 398 K, at which the equilibrium data verify the possibility of completely lean MDEA solvent for CO2 partial pressures below 20 kPa (corresponding illustration given in the Supplementary Information). It must be emphasized that the investigated experimental conditions were not designed to determine the optimal energy conditions based on the experimental results. However, the results offer a sufficiently large database for model validation such that a model-based analysis is planned to determine the most favorable operating conditions from an energy efficiency perspective.

4.59 4.71 4.67 4.93 4.66 5.01 4.74 4.66 4.87 4.69 5.39 4.71 4.61 4.53 4.57 4.67 4.8 4.77

desorbed mass flow of CO2 / kg h-1 SRD / MJ kgCO2-1

on the solvent regeneration. The results highlight the importance of a detailed experimental investigation of the solvent regeneration and also emphasize the need for a reliable mathematical model of the entire process for a meaningful evaluation.An overview of the experimental results obtained at the operating conditions in Table 1 is provided in Table 2. Obviously, notably low lean loadings representing a high DOR were obtained throughout all experiments. These findings indicate the advantageous solvent properties of MDEA compared with the regeneration results obtained with MEA solvent, which can usually be −1 regenerated to lean loadings of 0.2 molCO2·molMEA [36,37]. 8

0.18

7

0.16

6

0.14 0.12

5

0.10

4

0.08

3

0.06

2

0.04

1

0.02

0

0.00

desorbed mass flow of CO2

SRD

lean loading

6

lean loading / molCO2 molMDEA-1

EXP 6

Fig. 5. Overview of experimental results in terms of desorbed mass flow of CO2, SRD and lean loading for all experiments.

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

120%

5.6

5.2 80%

5.0

60%

4.8 4.6

40%

4.4 20%

SRD / MJ kgCO2-1

5.4

100%

DOR / %

Fig. 6. Overview of DOR and SRD for all conducted experiments.

4.2 4.0

0%

DOR

SRD

model of the desorption process with an aqueous solutions of 30 wt% MDEA was developed based on the experimentally gained insights and findings. The selected modeling approach is discussed in detail, including refinement of the model and flowsheets. Finally, the developed model is validated against experimental data to verify its applicability for process analysis purposes.

3.3. Comparison of liquid and gas analysis As described in Section 2.4.1, two different analytical methods were applied to determine the desorbed mass flow of CO2 and validate the obtained results. Either a CO2 mass balance of the desorber by measuring loadings of lean and rich solvent according to Eq. (7) or of the absorber by measuring CO2 volume fraction in inlet and outlet gas streams according to Eq. (3) can be used. To satisfy the mass balance of the overall process and validate the results of each method, both methods should lead to the same results. To illustrate how well the results of both methods support each other, a graphical comparison of the results obtained from both methods is presented in Fig. 7. Considering the parity plot on the left side of Fig. 7, it is obvious that both methods deviate by 10% at maximum, thus resulting in reasonably good fulfillment of the overall mass balance. Moreover, for the bar diagram, it is noticeable that for the majority of experiments, the liquid analysis shows slightly higher desorbed mass flows than the gas analysis.

4.1. Modeling approach, reactions and properties Extensive literature exists on the modeling of amine-based absorption-desorption-processes and generally recommends rate-based models that allow explicit calculation of reaction kinetics, film reactions and electrolyte speciation [38,39]. Therefore, such a model is also developed in this study. Because the model library in Aspen Plus® already contains process models that allow use of pre-implemented reaction models for the MDEA-H2O-CO2 system and thermophysical models to predict transport and component specific properties [40,41], this model was selected as an initial starting point in the current study. Most important for accurate modeling of the solvent regeneration of loaded MDEA solutions is an accurate reaction kinetic model. To this end, the set of reactions and the corresponding kinetic models that were originally developed by Pinsent et al. [42] and Rinker et al. [43], as summarized in Table 3, are implemented in the model. However, because the kinetic models reported by Rinker et al. [43] are only validated up to temperatures of 313 K , whereas desorption processes are

4. Modeling solvent regeneration of MDEA

8

8

7 +10 %

6 -10 %

5 4 3 2 1 0

6 5 4 3 2 1 0

0

1

2

3

4

5

6

7

8

desorbed mass flow of CO2 via gas analytics / kg h-1

EXP 1 EXP 2 EXP 3 EXP 4 EXP 5 EXP 6 EXP 7 EXP 8 EXP 9 EXP 10 EXP 11 EXP 12 EXP 13 EXP 14 EXP 15 EXP 16 EXP 17 EXP 18

7

desorbed mass flow of CO2 / kg h-1

desorbed mass flow of CO2 via liquid analytics / kg h-1

Reliable modeling of desorption is essential to precisely predict the required amounts of energy, which is required for a techno-economic process analysis. Hence, subsequent to the conducted experimental investigations and as a necessary step towards a final process analysis, a

Liquid analytic

Gas analytic

Fig. 7. Graphical comparison of liquid and gas phase analysis to determine the desorbed mass flow of CO2; left: parity plot; right: bar diagram.

7

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

Table 3 Overview of pre-implemented reactions. Reaction ID

Reaction

Reaction type

R1 R2 R3

MDEAH+ + H2 O↔ MDEA + H3O+ 2H2 O↔ H3O+ + OH−

Equilibrium Equilibrium Equilibrium

2− + HCO− 3 + H2 O↔ H3 O + CO3 CO2 + OH− → HCO− 3 − HCO− 3 → CO2 + OH

R4 R5 R6 R7

MDEA + H2 O+ CO2 → MDEAH+ + HCO− 3

Kinetic Kinetic Kinetic

MDEAH+ + HCO− 3 → MDEA + H2 O+ CO2

Kinetic

k [–]

EA [Cal mol−1]

EA [J mol−1]

Reference

R4 R5 R6 R7

4.32e+13 2.38e+17 2.22e+07 1.06e+16

13249 29451 9029 25424

55434 123223 37777 106374

[42] [43]

usually operated at temperatures that exceed 373 K , the desorption model must be validated. Although all reactions are considered to be reversible, reactions R1 to R3 involve only proton transfer and are assumed to reach reaction equilibrium instantaneously. Reactions R4 to R7 are kinetically limited, and thus calculation of the actual reaction rates is required for these four reactions. Equilibrium constants for reactions R1 to R3 were calculated based on minimization of the standard Gibbs free energy. The required reaction rates of R4 to R7 were calculated using a power law approach as follows

−E r = k ·T n·exp ⎛ A ⎞ ⎝ RT ⎠

N

∏

[42] [43]

components in the reaction, ci is the concentration of component i, and νi is the stoichiometric coefficient of component i. With respect to application of Eq. (14), the temperature exponent n is zero for all considered reactions, and the values for the pre-exponential factor k and the activation energy EA are listed in Table 4. It should be noted that the equilibrium constants, which were calculated from data collected by Austgen et al. [44], were used together with the forward reaction rates of R4 and R6 to determine the missing parameters of R5 and R7, respectively. The Henry constants for selected gaseous components (CO2, N2, CO, O2, H2S) were taken from the Aspen database. The Henry constants for these components were automatically specified with water and MDEA. The ELECNRTL property method was chosen for simulation of the liquid phase because it is specifically developed and is suitable for electrolyte systems such as aqueous MDEA solutions [41]. The nonideality of vapor was considered by the Redlich-Kwong equation of state [41]. The suitability of the chosen property method was verified by comparison of simulated vapor-liquid-equilibrium (VLE) with experimental data from the literature, which are supplied in the Supplementary Information.

Table 4 Reaction parameters k and EA . Reaction ID [–]

Reference

4.2. Model of the pilot scale test system

ciνi

(14)

i=1

An appropriate flowsheet model was developed to represent the experimental setup. Based on the description of the PDF in Section 2.3, the flowsheet model of the PDF that was developed with consideration of the process scheme shown in Fig. 2 is illustrated in Fig. 8.

where r is the rate of reaction, k is the pre-exponential factor, T is the absolute temperature, n is the temperature exponent, EA is the activation energy, R is the universal gas constant, N is the number of To absorber

CO2 Cold lean solvent

Pump From absorber

Cold rich solvent

Condenser CO2 + H2O(v)

Hot rich solvent

Desorber

Recycled H2 O

Heat Loss

Heat exchanger

CO2 + H2O(l)

Reboiler Hot lean solvent

Lean solvent Recycled lean solvent 85 kg h-1 Cooler ®

Fig. 8. Flowchart of the pilot scale test system model created in Aspen Plus .

8

Flash

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

resistance for the liquid film was calculated with a film discretization ratio of 5 (Discrxn), as recommended for modeling of reactive absorption processes [39]. The gas film resistance was calculated using the film method under consideration of the diffusional resistance. Heat transfer coefficients were calculated based on the Chilton-Colburn analogy [46]. Because random packings were used, the correlation of Billet and Schultes (1993) was selected for calculation of the mass transfer coefficients and interfacial area [47]. A liquid inventory volume of approximately 40 L was present in the sump of the desorber and was in continuous contact with the vapor phase introduced by the reboiler. Considering the highest flow rate of 200 kg ·h−1, the residence time of the liquid in the sump can be calculated as approximately 12 min, a value that was increased for lower solvent flow rates. The residence time in the sump is therefore significantly larger than that of all other stages such that it can be assumed to be sufficient to reach thermodynamic and reaction equilibrium. Thus, contrary to all stages above, the bottom stage of the desorber (38) was modeled as a single equilibrium stage. In accordance with the experimental setup, the reboiler and condenser are represented by two heat exchanger models (“Reboiler” and “Condenser”). The reboiler was constantly supplied with a constant solvent flow rate of 85 kg ·h−1, taken as a side stream from the lean solvent line exiting the desorber at the bottom. The duty of the reboiler was specified by the calculated reboiler duty, as explained in Section 2.4.2, and the pressure was directly taken from experimental data. The condenser at the top of the desorber was specified using the pressure and the outlet temperature of each experiment. The reflux drum in which the gaseous CO2 is separated from the condensed water is represented by a flash model for which the temperature and pressure were specified according to experimental data.

100°C / 20.5 wt.-% MDEA

10

2

pCO / kPa

100

1

0,1 0,0

0,1

0,2

loading / mol CO2 mol MDEA-1

simulation with flash

experiment

0,3

simulation with column

Fig. 9. Comparison of model predictions for the ternary VLE H2O-MDEA-CO2 between the equilibrium stage and the flash model.

As previously introduced, the absorber of the PDF was used only to supply the rich loaded solvent with constant CO2 loading such that the desorber could be operated at steady-state conditions. Because this step does not contribute to the analysis of the energy efficiency, it is simply neglected in the developed model by directly specifying the composition of the entering liquid stream in Aspen Plus® (represented by the stream “From absorber” in Fig. 8). This stream is compressed by the pump to an experimental specified discharge pressure and enters the process heat exchanger model (“Heat exchanger”), where it is heated by the hot lean solvent from the desorber. In accordance with the experimental facility, the heat exchanger model was specified with a heat exchanger area of 3.6 m2 and a heat transfer coefficient of 850 W m–2 K–1. The heated rich solvent (“hot rich solvent”) passes another heat exchanger (“Heat Loss”), which was be implemented in the model to account for the considerable heat losses between the actual process heat exchanger and the desorber in the piping system of the real plant. This “Heat Loss” block was represented by a simple heater model, which reduced the solvent temperature to match the experimentally measured temperature at the feed stage of the desorber column. The desorber itself is modeled in Aspen Plus® using the rate-based mode of the RadFrac column model, not including the reboiler and condenser. In total, the number of non-equilibrium (NEQ) stages was specified as 38, with one column section in the real column represented by two NEQ stages with a diameter of 0.148 and a height of 0.154 m. For each stage, the mass and heat transfer balances were solved by considering the actual transport and reaction rates. All model stages were simulated as packed stages with 16 mm Pall Rings. Known from the operational experience of the CSIRO team, the mass transfer performance of the empty sections corresponds to approximately 30% of the mass transfer performance of the packed sections. This information was transferred to the model, and accordingly, the model stages representing the empty column sections were weighted to only 30% mass transfer performance of a packed section. For this purpose, in the chosen rate-based model, the liquid-phase mass transfer coefficient predicted by the selected correlation is multiplied with a scaling factor, i.e., the so-called liquid mass transfer coefficient factor, and this scaling factor was set to 0.3. Analogously, the same setting was applied to the vapor mass transfer coefficient factor. For the general model equations of the NEQ model of the reactive absorption, the interested reader is referred to excellent textbooks [45,46], and a detailed description of the rate-based mode of the Aspen Plus® RadFrac model can be found in the articles by Zhang et al. [40,41]. To account for the rapid reaction in the liquid phase, the film

4.3. Model refinement Although it was initially assumed that the above-described model supplies an accurate representation of the experimental setup sufficient to simulate the conducted experiments, it quickly became obvious that the measured desorbed mass flow rates of CO2 could not be achieved with the SRD, as calculated from experimental data (see Supplementary Information for an illustration). After careful analysis of the model and considering various potential explanations, e.g., the calculation of the reboiler duty and the validity of the reaction kinetics at the considered temperature range, accurate prediction of the VLE of the ternary system H2O-MDEA-CO2 was identified as the source of the deviations between simulation and experimental results. Although the VLE had been validated prior to setting up the flowsheet model using an evaluation of single flash calculations (cf. Supplementary Information), it was surprisingly found that the equilibrium stage calculations for the bottom stage of the column were not in alignment with the literature data. It was observed that although the same chemistry and property method is used, the column equilibrium model produces different VLE results than the flash model. This finding was checked by performing the same simulation that was conducted with the flash model to validate the predictability of the ternary VLE with a single equilibrium stage column model. The corresponding results are presented in Fig. 9, where the predicted partial pressure of CO2 is significantly higher at low CO2 loading levels (because they occur in the desorber bottom) when applying the column equilibrium model versus the flash model. As a consequence, the last stage of the column was not simulated by an equilibrium stage but was replaced by a flash model, as illustrated in Fig. 10. To simulate the flash as the bottom portion of the column, no specifications were applied to the flash model (the pressure drop and duty were specified as zero). This process indicates adiabatic mixing of the liquid coming down the column and the solution heated by the reboiler with the exiting liquid and gas streams in thermodynamic equilibrium. Due to this modification, the previously observed deviation between simulation and experimental results was resolved. 9

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

Cooler

To absorber Cold lean absorbent Pump From absorber

Cold rich solvent

CO2 + H2O(v)

Hot rich solvent

CO2

CO2 + H2O(v)

CO2 + H2O(l) Condenser

Desorber Heat exchanger

Flash

Recycled H2O

Heat Loss

Flash EQ-stage Hot lean solvent

Lean solvent

Reboiler Recycled lean solvent 85 kg h-1

Fig. 10. Refined flowchart of the pilot scale test system model created in Aspen Plus®.

Therefore, the deviations in the prediction of the VLE in the column model appear to have a minor influence on the NEQ stage calculations.

4.4. Model validation After refinement, the presented model was successfully validated against the experimental data. For this purpose, the experimentally determined desorbed mass flows of CO2 were compared with simulation results using exactly the same operating conditions. Deviations of less than ± 5% were determined, thus verifying the excellent predictive capability of the developed model (cf. Supplementary Information for an illustration). In a second validation scenario, 13 randomly selected experiments were re-evaluated. Instead of specifying the SRD to the value calculated from experimental data, the experimentally obtained lean loading was specified by an additional constraint, and the necessary SDR, i.e., the reboiler duty, was determined from simulation of the model. The agreement between the simulated and experimentally determined reboiler duties is shown in Fig. 11 in a parity plot diagram. Fairly good agreement between the simulated data and experimental results was achieved. However, it is noticeable that most reboiler duties predicted by the model were lower than those determined from experimental data. However, because additional heat losses to the environment at various locations of the PDF piping and equipment (especially the column shell) are not specifically considered in the process model, the resulting underestimation with deviations mostly below 20% of the reboiler duty is quite reasonable. To further illustrate the accuracy of the model, three selected temperature profiles from the simulation are compared with temperature measurements from the experimental setup in Fig. 12. Note that all other temperature profiles were predicted with similar accuracy by the developed model. The temperatures in the sump and at the top of the column show larger deviations (6–7 K) than the remainder of the

Fig. 11. Parity plot for simulated and experimentally determined reboiler duty.

trajectories. Noticeably, the temperature of the top is overestimated by the model, and the temperature in the sump is underestimated by the model. It is known from previous testing that the heat loss of the desorber sump is much more significant than of the remainder of the column due to the high liquid inventory. Furthermore, these deviations can be explained by the trajectory of the ternary VLE in Fig. 9, which also shows an underestimation at low loadings (as in the sump) and an overestimation at high partial pressures (as at the top). Nevertheless, the developed model is capable of precisely predicting the temperature profiles in the desorber column and accurately describing the regeneration of loaded MDEA solutions. The developed model can be used in detailed analysis and assessment of the MDEA-based capture processes. This process is conducted for the case of enzymatic reactive absorption in the following Section 5. 10

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

6

Table 5 Overview of flue gas flow rate (10% from original scale), conditions and compositions taken from [23].

column height / m

5

4

3

2

Flue gas

Unit

Value

Mass flow rate Temperature Pressure Composition CO2 H2O N2 O2

kg·s−1 °C Bar

82.1 58 1

Mol.% Mol.% Mol.% Mol.%

13.5 15.37 68.75 2.38

1

8 m in diameter and desorber dimensions of 15 m packing height and 9 m diameter were considered for process analysis. Sulzer BX packing was specified for both columns modeled with the rate-based RadFrac model and with consideration of 40 NEQ stages. Compared with the validated model, this approach results in a greater height per NEQ stage, but for the considered scenario and the performed calculations, it was found that the influence of the number of NEQ stages greater than 40 is insignificant (cf. Supplementary Information). SRD and lean and rich loadings were analyzed as function of the solvent flow rate. The obtained results are presented in Fig. 13. It is obvious that a lower solvent inlet temperature in the absorber (which can be effectively exploited by the enzymatic reactive absorption) offers the potential to significantly reduce the energy requirement for solvent regeneration. This adjustment also leads to a reduced solvent demand compared with the higher temperature. The minimum energy requirement at the 40 °C absorber inlet temperature is close to −1 2.6 MJ ·kgCO at a flow rate of 650 kg ·s−1, which is in agreement with a 2 comparable study of Penders van Elk and Versteeg [23], who considered a different form of carbonic anhydrase enzyme and a non-va−1 lidated desorber model, resulting in a minimum SRD of 2.8 MJ ·kgCO at 2 a similar liquid to gas ratio. At a reduced absorber solvent inlet tem−1 perature of 20 °C, the lowest SRD of 2.13 MJ ·kgCO was obtained for a 2 solvent flow rate of only 450 kg ·s−1. This result represents a significant reduction of the energy requirement compared with that of the conventionally used solvents such as MEA, which requires approximately −1 3.7 MJ ·kgCO for solvent regeneration [51] and is a real breakthrough in 2 amine-based capture technologies. Considering that a conventional process concept was considered, even lower reboiler duties can be expected if applying process modifications that allow for significantly improved energy efficiency, as discussed by Wang et al. [52], Oh et al. [53] or Le Moullec et al. [54]. The corresponding lean and rich loadings obtained in the process analysis for both temperatures are depicted as a function of the solvent

0 90

100

110

temperature / °C experiment

120 simulation

Fig. 12. Comparison of simulated and experimentally determined temperature profiles (in black EXP 5, in dark gray EXP 12 and in light gray EXP 7).

5. Model-based process analysis for an enzymatic reactive absorption process As a final step, the developed and validated desorption model was combined with the enzymatic reactive absorption process model previously developed by Leimbrink et al. [21] to determine the operating conditions of an enzymatic reactive absorption process that results in the minimum energy duty required for solvent regeneration. Only the combination of both process steps (absorption and desorption) allows for evaluation of mutual influences on the overall process performance. Both models were built using a common basis in Aspen Plus®, as presented in Section 4.1. Within this process analysis, the enzyme-catalyzed reaction is considered only for the absorption column and not for solvent regeneration due to the limited thermal stability of the enzyme. Hence, either a well-performing immobilized enzyme for application in the absorber, as published by Reardon et al. [48], or an appropriate reclaiming strategy for a dissolved enzyme, as published e.g., by Gundersen et al. [35], was assumed to be applied. Overall energy savings are expected for the following reasons: 1. A reduction in Qsens due to exploitation of a higher cyclic loading capacity accessible by accelerated absorption rates at lower temperatures [21], 2. Exploitation of the lower Qdes,CO2 of MDEA compared with that of conventionally used primary amine solvents such as MEA [8] or blends such as MDEA + piperazine [49], 3. Reduced Q vap,H2 O due to favorable temperature-dependent solubility of CO2 in aqueous MDEA solutions [34] and thus lower amounts of water steam needed as a stripping agent.

3.6

SRD / MJ kg CO2-1

3.4

To facilitate a comparison with the available results in the literature, the same scenario for CO2 absorption as that evaluated by Penders van Elk and Versteeg [23] was selected for process analysis. This scenario considers 90% CO2 capture from the flue gas of an advanced supercritical pulverized-coal Rankine-cycle power plant and originates from the report by Gerder et al. [50]. Note that only 10% of the fullscale flue gas stream was considered for this simulation study, which does not affect the relative energy requirements per kg of CO2 separated. The conditions, flow rate and composition of the flue gas considered in this study are summarized in Table 5. A 30 wt% aqueous MDEA solution was applied as the solvent, and the liquid inlet temperature in the absorber was varied between 20 °C and 40 °C. In determining the minimum energy requirement for solvent regeneration, fixed absorber dimensions of 50 m in packing height and

3.2 3.0 2.8 2.6 2.4 2.2 2.0 300

400

500

600

700

800

900

1,000

solvent flow rate / kg s-1 30 wt.-% MDEA, 40 °C

30 wt.-% MDEA, 20 °C

Fig. 13. Energetic evaluation of enzymatic reactive absorption showing SRD as function of the solvent flow rate and absorber solvent inlet temperature.

11

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

solvent flow rate / kg s-1 300

0.8

400

2.7

500

600

100

0.6 0.5 0.4 0.3 0.2 0.1 0.0 300

400

500

600

700

800

900

Lean loading (40 °C)

Rich loading (20 °C)

Lean loading (20 °C)

60

2.1

40

1.9

20

353

453 condenser heat

567 heat of desorption

0 SRD at 20 °C

Fig. 16. Distribution of the three contributors to SRD at a solvent inlet temperature of 20 °C for three exemplary flow rates.

desorber, and consequently, more water vapor must be produced as a stripping agent to achieve the required lean loadings shown in Fig. 14. With increasing solvent flow rate, the required cyclic loading decreases and therefore the required qcond also decreases, while qsens simultaneously increases because additional solvent must be heated along the desorber. Although the heat of CO2 desorption (qdes,CO2 ) is insensitive to the solvent flow rate, the two opposing trends of qsens and qcond result in a minimum SRD at a certain solvent flow rate, which is determined as −1 approximately 2.13 MJ ·kgCO at a solvent flow rate of 450 kg ·s−1 for the 2 current example, considering the considered equipment dimensions.

0.5

cyclic loading capacity / mol CO2 mol MDEA-1

2.3

sensible heat

Fig. 14. Rich and lean loading at absorber inlet temperature of 20 °C and 40 °C of a 30 wt % MDEA solution as function of the solvent flow rate.

0.4 0.3 0.2

6. Conclusions and outlook

0.1 0.0 300

80

1.7

1,000

solvent flow rate / kg s-1 Rich loading (40 °C)

2.5

fraction of SRD

0.7

SRD / MJ kg CO2 -1

loading / mol CO2 mol MDEA-1

0.9

400

500

600

700

800

900

The current article presents an elaborate experimental investigation of the influences of solvent flow rate, rich loading and reboiler duty on the regeneration of aqueous MDEA solvent in CO2 absorption processes. Based on the experimental results, a rate-based model was developed and validated, and this model can be further used in process analysis of any MDEA-based CO2 capture process. This study applied a modelbased process analysis in combination with a similar previously developed model of an enzymatic reactive absorption column [21]. The availability and validity of such models is crucial for correct approximation of the energy requirements and process performance during conceptual design of a capture plant that makes use of the novel enzyme-solvent system and requires recovery of the aqueous MDEA at large scale. For the considered aqueous MDEA solvent system, an increased cyclic loading capacity becomes available due to the reaction rate acceleration by the enzyme carbonic anhydrase, allowing for operation at lower absorber temperatures. Furthermore, operation with significantly reduced solvent flow rates becomes feasible. Considering a common CO2 capture scenario that requires 90% capture of CO2 from the flue gas of an advanced supercritical pulverized-coal Rankine-cycle power plant, using the specified packed column equipment, a minimum spe−1 cific reboiler duty of 2.13 MJ ·kgCO was determined when supplying the 2 solvent at a temperature of 20 °C to the absorber. This situation corre−1 sponds to a 40% improvement compared with MEA (∼3.7 MJ ·kgCO ) as 2 the benchmark solvent. If technical realization of such low temperatures (20 °C) becomes challenging due to geographical location, application of an increased absorption temperature of 40 °C could still result in an improvement of 30% compared with the MEA benchmark. Nevertheless, a final decision on the optimum process conditions should be made with consideration of the total annualized costs of the process, including operating costs and capital investment subject to site-specific constraints. In addition to consideration of different solvent systems, further process modifications (e.g., interstage cooling, advanced flash stripper or lean vapor compression) as demonstrated for MEA or piperazine solvent systems [53,55] should be included to supply an

1,000

solvent flow rate / kg s-1 40 °C

20°C

Fig. 15. Cyclic loading capacity of a 30 wt% MDEA solution as function of solvent flow rate and temperature.

flow rate in Fig. 14, illustrating the expected higher rich loadings that are accessible at the lower temperature. To emphasize the increased cyclic loading capacity, it is illustrated as a function of the absorber temperature and solvent flow rate in Fig. 15. Although the curves for both temperatures are superimposed in the range of equal solvent flow rates (which is reasonable because the absorbed molar flow of CO2 is fixed due to the specified 90% capture rate and the amount of MDEA is equivalent at the same solvent flow rate), a significant increase in cyclic loading capacity becomes feasible for an absorber inlet temperature of 20 °C and is exploited at lower solvent flow rates. It is important to note that it is only because of the increased cyclic loading capacity that the operation at reduced solvent flow rates becomes feasible, which facilitates reduced operating costs and also reduced investment costs because the pumps and heat exchangers are reduced in size for lower solvent flow rates. However, for the SRD, the minimum energy requirement for the considered equipment is not reached at the highest loading capacity and lowest solvent flow rate. This situation is illustrated in Fig. 16, which shows the course of the SRD for the various solvent flow rates and indicates the three individual contributors of the SRD for three exemplary flow rates at solvent inlet temperatures of 20 °C. Apparently, the heat of CO2 desorption qdes,CO2 represents the largest contribution to the SRD at all flow rates. However, the relative shares of the sensible heat qsens and the condenser heat qcond change with increasing solvent flow rate. At lower solvent flow rates, high cyclic loadings are required, higher gas-to-liquid ratios are required for the 12

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

accurate estimate of the true potential for energy efficiency improvements for the CO2 capture process. This potential is supported by the assessment by Feron in 2010 [11], which envisioned the development of third (G3) and fourth (G4) generation capture technologies. The thermal energy requirement of G3 −1 technologies was estimated as 2.29 MJ ·kgCO , which is highly similar to 2

[16]

[17]

[18]

−1 the 2.13 MJ ·kgCO reported in this study. However, G4 technologies are 2 −1 MJ ·kgCO 2

foreseen to further reduce the thermal energy input to 0.95 by involving “chemical absorbents making use of bicarbonate formation” [11], which exactly matches the reaction mechanism of the enzyme carbonic anhydrase. A potential solvent candidate that could enable even lower energy requirements than MDEA is tertiary amine 1-dimethylamino-2-propanol, as recently published by Liu et al. [13]. Therefore, the presented methodology can be applied to such novel solvents to determine the energy requirements, which can be compared with those supplied in this study. As such, a reliable and fair assessment of novel solvents can be guaranteed. In summary, the presented study represents a breakthrough in improved energy efficiency, leading into a new generation of carbon capture technologies as estimated by Feron in 2010 [11]. Nevertheless, further significant improvements with use of enzymatic reactive absorption can be expected from future research and development.

[19]

[20]

[21]

[22]

[23]

[24]

[25]

Acknowledgements [26]

The research leading to these results was funded by the European Union Seventh Framework Programme FP7/2007-2013 under grant agreement n° 608535. The authors also acknowledge support from CSIRO’s Low Emissions Technology Research Program.

[27]

[28]

Appendix A. Supplementary material [29]

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.apenergy.2017.10.042.

[30]

References [31] [1] Gunter W, Wong S, Cheel D, Sjostrom G. Large CO2 Sinks: Their role in the mitigation of greenhouse gases from an international, national (Canadian) and provincial (Alberta) perspective. Appl Energy 1998;61(4):209–27. [2] da Graça Carvalho M. EU energy and climate change strategy. Energy 2012;40(1):19–22. [3] Friedlingstein P, Houghton RA, Marland G, Hackler J, Boden TA, Conway TJ, et al. Update on CO2 emissions. Nat Geosci 2010;3(12):811–2. [4] Feron PHM, editor. Absorption-based post-combustion capture of carbon dioxide. Duxford, UK: Woodhead Publishing is an Impring of Elsevier; 2016. [5] International Energy Agency. 20 years of carbon capture and storage - Accelerating future deployment. [March 02, 2017]; Available from: http://www.iea.org/ publications/freepublications/publication/20YearsofCarbonCaptureandStorage_ WEB.pdf. [6] Li B, Duan Y, Luebke D, Morreale B. Advances in CO2 capture technology: A patent review. Appl Energy 2013;102:1439–47. [7] Goto K, Yogo K, Higashii T. A review of efficiency penalty in a coal-fired power plant with post-combustion CO2 capture. Appl Energy 2013;111:710–20. [8] Yildirim Ö, Kiss AA, Hüser N, Leßmann K, Kenig EY. Reactive absorption in chemical process industry: A review on current activities. Chem Eng J 2012;213:371–91. [9] Kunze A-K, Dojchinov G, Haritos VS, Lutze P. Reactive absorption of CO2 into enzyme accelerated solvents: From laboratory to pilot scale. Appl Energy 2015;156:676–85. [10] Oexmann J, Kather A. Minimising the regeneration heat duty of post-combustion CO2 capture by wet chemical absorption: The misguided focus on low heat of absorption solvents. Int J Greenhouse Gas Control 2010;4(1):36–43. [11] Feron PH. Exploring the potential for improvement of the energy performance of coal fired power plants with post-combustion capture of carbon dioxide. Int J Greenhouse Gas Control 2010;4(2):152–60. [12] Kohl AL, Nielsen R. Gas purification. 5th ed. Houston, Tex.: Gulf Pub; 1997. [13] Liu H, Gao H, Idem R, Tontiwachwuthikul P, Liang Z. Analysis of CO2 solubility and absorption heat into 1-dimethylamino-2-propanol solution. Chem Eng Sci 2017. [14] Liu H, Xiao M, Liang Z, Tontiwachwuthikul P. The analysis of solubility, absorption kinetics of CO 2 absorption into aqueous 1-diethylamino-2-propanol solution. AIChE J 2017;63(7):2694–704. [15] Osei PA, Akachuku A, Decardi-Nelson B, Srisang W, Pouryousefi F, Tontiwachwuthikul P, et al. Mass transfer studies on catalyst-aided CO2 desorption

[32]

[33]

[34] [35]

[36] [37]

[38] [39] [40]

[41]

[42] [43]

[44]

13

from CO2-loaded amine solution in a post-combustion CO2 capture plant. Chem Eng Sci 2017. Srisang W, Pouryousefi F, Osei PA, Decardi-Nelson B, Akachuku A, Tontiwachwuthikul P, et al. Evaluation of the heat duty of catalyst-aided aminebased post combustion CO2 capture. Chem Eng Sci 2017. Zhang X, Zhang X, Liu H, Li W, Xiao M, Gao H, et al. Reduction of energy requirement of CO 2 desorption from a rich CO 2 -loaded MEA solution by using solid acid catalysts. Appl Energy 2017;202:673–84. Kim I, Hoff KA, Hessen ET, Haug-Warberg T, Svendsen HF. Enthalpy of absorption of CO2 with alkanolamine solutions predicted from reaction equilibrium constants. Chem Eng Sci 2009;64(9):2027–38. Saghafi H, Arabloo M. Modeling of CO2 solubility in MEA, DEA, TEA, and MDEA aqueous solutions using AdaBoost-decision tree and artificial neural network. Int J Greenhouse Gas Control 2017;58:256–65. Gladis A, Gundersen MT, Fosbøl PL, Woodley JM, von Solms N. Influence of temperature and solvent concentration on the kinetics of the enzyme carbonic anhydrase in carbon capture technology. Chem Eng J 2017;309:772–86. Leimbrink M, Tlatlik S, Salmon S, Kunze A-K, Limberg T, Spitzer R, et al. Pilot scale testing and modeling of enzymatic reactive absorption in packed columns for CO 2 capture. Int J Greenhouse Gas Control 2017;62:100–12. Migliardini F, de Luca V, Carginale V, Rossi M, Corbo P, Supuran CT, et al. Biomimetic CO2 capture using a highly thermostable bacterial α-carbonic anhydrase immobilized on a polyurethane foam. J Enzyme Inhibition Med Chem 2014;29(1):146–50. Penders-van Elk N, Versteeg GF. Enzyme-enhanced CO2 absorption. In: Feron PHM, editor. Absorption-based post-combustion capture of carbon dioxide. Duxford, UK: Woodhead Publishing is an Impring of Elsevier; 2016. p. 225–58. Russo ME, Olivieri G, Marzocchella A, Salatino P, Caramuscio P, Cavaleiro C. Postcombustion carbon capture mediated by carbonic anhydrase. Sep Purif Technol 2013;107:331–9. Salmon S, House A. Enzyme catalyzed solvents for CO2 separation. In: Morreale B, Shi F, editors. Novel materials for carbon dioxide mitigation technology. Amsterdam, Netherlands: Elsevier; 2015. p. 23–86. Vinoba M, Bhagiyalakshmi M, Grace AN, Kim DH, Yoon Y, Nam SC, et al. Carbonic anhydrase promotes the absorption rate of CO2 in post-combustion processes. J Phys Chem B 2013;117(18):5683–90. Leimbrink M, Limberg T, Kunze A-K, Skiborowski M. Different strategies for accelerated CO2 absorption in packed columns by application of the biocatalyst carbonic anhydrase. Energy Procedia 2017;114:781–94. Puxty G, Bennett R, Conway W, Maher D. A comparison of Raman and IR spectroscopies for the monitoring and evaluation of absorbent composition during CO2 absorption processes. Int J Greenhouse Gas Control 2016;49:281–9. Kim YE, Lim JA, Jeong SK, Yoon YI, Bae ST, Nam SC. Comparison of carbon dioxide absorption in aqueous MEA, DEA, TEA, and AMP solutions. Bull Korean Chem Soc 2013;34(3):783–7. McCann N, Maeder M, Hasse H. Prediction of the overall enthalpy of CO2 absorption in aqueous amine systems from experimentally determined reaction enthalpies. Energy Procedia 2011;4:1542–9. Quang DV, Rabindran AV, El Hadri N, Abu-Zahra MRM. Reduction in the regeneration energy of CO2 capture process by impregnating amine solvent onto precipitated silica. Eur Sci J 2013;9(30). Wang T, Yu W, Le Moullec Y, Liu F, Xiong Y, He H, et al. Solvent regeneration by novel direct non-aqueous gas stripping process for post-combustion CO 2 capture. Appl Energy 2017;205:23–32. Gupta M, da Silva EF, Hartono A, Svendsen HF. Theoretical study of differential enthalpy of absorption of CO2 with MEA and MDEA as a function of temperature. J Phys Chem B 2013;117(32):9457–68. Rho S-W, Yoo K-P, Lee JS, Nam SC, Son JE, Min B-M. Solubility of CO 2 in aqueous methyldiethanolamine solutions. J Chem Eng Data 1997;42(6):1161–4. Gundersen MT, Gladis A, Fosbøl PL, von Solms N, Woodley JM. Operating considerations of ultrafiltration in enzyme enhanced carbon capture. Energy Procedia 2017;114:735–43. Saimpert M, Puxty G, Qureshi S, Wardhaugh L, Cousins A. A new rate based absorber and desorber modelling tool. Chem Eng Sci 2013;96:10–25. Rao AB, Rubin ES. A technical, economic, and environmental assessment of aminebased CO2 capture technology for power plant greenhouse gas control. Environ Sci Technol 2002;36(20):4467–75. Kenig EY, Schneider R, Górak A. Reactive absorption: Optimal process design via optimal modelling. Chem Eng Sci 2001;56(2):343–50. Kucka L, Müller I, Kenig EY, Górak A. On the modelling and simulation of sour gas absorption by aqueous amine solutions. Chem Eng Sci 2003;58(16):3571–8. Zhang Y, Chen C-C. Modeling CO2 absorption and desorption by aqueous monoethanolamine solution with aspen rate-based model. Energy Procedia 2013;37:1584–96. Zhang Y, Chen C-C. Thermodynamic modeling for CO2 absorption in aqueous MDEA solution with electrolyte NRTL model. Ind Eng Chem Res 2011;50(1):163–75. Pinsent B, Pearson L, Roughton F. The kinetics of combination of carbon dioxide with hydroxide ions. Trans Faraday Soc 1956;52:1512–20. Rinker EB, Sami S, Sandall OC. Kinetics and modelling of carbon dioxide absorption into aqueous solutions of N-methyldiethanolamine. Chem Eng Sci 1995;50(5):755–68. Austgen DM, Rochelle GT, Chen CC. Model of vapor-liquid equilibria for aqueous acid gas-alkanolamine systems. 2. Representation of hydrogen sulfide and carbon dioxide solubility in aqueous MDEA and carbon dioxide solubility in aqueous mixtures of MDEA with MEA or DEA. Ind Eng Chem Res 1991;30(3):543–55.

Applied Energy xxx (xxxx) xxx–xxx

M. Leimbrink et al.

Research/Energy%20Analysis/OE/BitBase_FinRep_Rev2a-3_20130919_1.pdf. [51] European Benchmark Task Force. D 1.4.3 European best practice guidelines for assessment of CO2 capture technologies. [April 05, 2017]; Available from: https:// www.sintef.no/globalassets/project/decarbit/d-1-4-3_euro_bp_guid_for_ass_co2_ cap_tech_280211.pdf. [52] Wang M, Joel AS, Ramshaw C, Eimer D, Musa NM. Process intensification for postcombustion CO 2 capture with chemical absorption: A critical review. Appl Energy 2015;158:275–91. [53] Oh S-Y, Binns M, Cho H, Kim J-K. Energy minimization of MEA-based CO2 capture process. Appl Energy 2016;169:353–62. [54] Le Moullec Y, Neveux T, Al Azki A, Chikukwa A, Hoff KA. Process modifications for solvent-based post-combustion CO2 capture. Int J Greenhouse Gas Control 2014;31:96–112. [55] Rochelle GT. Conventional amine scrubbing for CO2 capture. In: Feron PHM, editor. Absorption-based post-combustion capture of carbon dioxide. Duxford, UK: Woodhead Publishing is an Impring of Elsevier; 2016. p. 35–67.

[45] Kenig EY, Górak A. Reactive Absorption. In: Sundmacher K, Kienle A, SeidelMorgenstern A, editors. Integrated chemical processes: Synthesis, operation, analysis, and control. Weinheim: Wiley-VCH; 2005. p. 265–311. [46] Taylor R, Krishna R. Multicomponent mass transfer. New York: Wiley; 1993. [47] Billet R, Schultes M. Predicting mass transfer in packed columns. Chem Eng Technol 1993;16(1):1–9. [48] Reardon J, Bucholz T, Hulvey M, Tuttle J, Shaffer A, Pulvirenti D, et al. Low energy CO2 capture enabled by biocatalyst delivery system. Energy Procedia 2014;63:301–21. [49] Gupta M, Svendsen HF. Theoretical study of temperature dependent enthalpy of absorption, heat capacity, and entropy changes for protonation of amines and amino acid solvents. Energy Procedia 2014;63:1099–105. [50] Gerder K, Haslbeck J, Kuehn N, Lewis E, Pinkerton LL, Woods M, et al. Cost and performance baseline for fossil energy plants - volume 1: bituminous coal and natural gas to electricity: revision 2. Technical report DOE/NETL-2010/1397. [April 25, 2017]; Available from: https://www.netl.doe.gov/File%20Library/

14