Environ. Sci. Technol. 2008, 42, 563–569
Cycle Development and Design for CO2 Capture from Flue Gas by Vacuum Swing Adsorption JUN ZHANG AND PAUL A. WEBLEY* Cooperative Research Centre for Greenhouse Gas Technologies, Department of Chemical Engineering, Monash University, Victoria 3800 Australia
Received March 19, 2007. Revised manuscript received July 31, 2007. Accepted September 27, 2007.
CO2 capture and storage is an important component in the development of clean power generation processes. One CO2 capture technology is gas-phase adsorption, specifically pressure (or vacuum) swing adsorption. The complexity of these processes makes evaluation and assessment of new adsorbents difficult and time-consuming. In this study, we have developed a simple model specifically targeted at CO2 capture by pressure swing adsorption and validated our model by comparison with data from a fully instrumented pilot-scale pressure swing adsorption process. The model captures nonisothermal effects as well as nonlinear adsorption and nitrogen coadsorption. Using the model and our apparatus, we have designed and studied a large number of cycles for CO2 capture. We demonstrate that by careful management of adsorption fronts and assembly of cycles based on understanding of the roles of individual steps, we are able to quickly assess the effect of adsorbents and process parameters on capture performance and identify optimal operating regimes and cycles. We recommend this approach in contrast to exhaustive parametric studies which tend to depend on specifics of the chosen cycle and adsorbent. We show that appropriate combinations of process steps can yield excellent process performance and demonstrate how the pressure drop, and heat loss, etc. affect process performance through their effect on adsorption fronts and profiles. Finally, cyclic temperature profiles along the adsorption column can be readily used to infer concentration profiles-this has proved to be a very useful tool in cyclic function definition. Our research reveals excellent promise for the application of pressure/vacuum swing adsorption technology in the arena of CO2 capture from flue gases.
Introduction With the increasing focus on global warming caused by greenhouse gas emissions, Carbon Capture and Storage (CCS) is seen as a promising mitigation initiative and has attracted considerable research development and demonstration efforts over the last two decades (1–3). In particular, since the capture cost is the major component of the overall cost of CCS, many capture technology solutions have been proposed and investigated (3). Among the capture technology candidates, pressure/vacuum swing adsorption technology (referred to as PSA or VSA) has been frequently investigated * Corresponding author phone: (+61 3) 9905 3628; fax: (+61 3) 9905 9602; e-mail:
[email protected]. 10.1021/es0706854 CCC: $40.75
Published on Web 12/07/2007
because of its low energy requirements and relative simplicity. Many CO2-selective adsorbents have been developed and tested and various pressure swing cycles have been designed and investigated (4, 5). Most of the reported studies rely on numerical modeling to reach conclusions; these models are reliant on a variety of simplifying assumptions and boundary conditions (6–11). However, justifications for the use of specific adsorption cycles are generally not given and the detailed analysis of the role of specific cyclic component steps in the context of CO2 capture, such as feed, pressure equalization, evacuation, product purge, and repressurization (co- and counter-current), have not been conducted. Although many detailed pressure/vacuum swing adsorption models have been developed these models are often difficult to use, run slowly, and rely on an expert user to determine appropriate operating parameters. As a result, there is a very wide range of reported performance data from pressure swing adsorption simulations with the same adsorbent, ranging from recoveries of less than 20% and purities of 30% up to recoveries of 80% and purities in excess of 95%. It is difficult for the inexpert PSA engineer to make sense of these reports especially if a new adsorbent has been developed and the engineer wishes to evaluate this adsorbent. While simple equilibrium analysis can yield useful data (e.g., the analytical solution to PSA by Chang and Knaebel and Hill with the assumptions of isothermal operation and linear isotherms (12, 13), Tezel’s approximate isothermal adsorbent screening (14), and the more elaborate equilibrium analysis of Pigorini and LeVan (15)), there is a need for a simple but reliable method for designing a PSA cycle for CO2 capture and assessing a new adsorbent quickly using this cycle. In particular, it is of great interest to understand the power requirements for a CO2 capture process and hence a rapid indicator would be extremely useful. It is the goal of the current study to develop and use a simple, fast, yet reliable model for selection and evaluation of a cycle and operating conditions for CO2 capture from process streams using pressure/vacuum swing adsorption. A similar effort for more general PSA cycles was made by Serbezov and Sotirchos (16) who developed semianalytical solutions for 4-step PSA cycles assuming linear isotherms and isothermal operation. Our model is also a local-equilibrium-based one but includes nonisothermal operation-this is an essential component since the heat of adsorption during CO2 capture by PSA can be significant. In addition, we do not restrict ourselves to linear isotherms since CO2 adsorption on virtually all adsorbents is strongly nonlinear and indeed it is this nonlinearity which determines the optimal operating conditions. Finally, the role of coadsorption of nitrogen, a major constituent of flue gas, on CO2 recovery must be included. After development of this model, we illustrate the role of each of the various process steps in PSA cycles for CO2 capture and introduce the concept of composition front management. We also show a comparison of the model predictions with experimental data from our pilot-scale pressure swing adsorption plant. The results from our models can be readily linked to economic models allowing rapid estimation of capture costs and the relative benefits of new adsorbents and process conditions on capture cost.
Experimental Section To validate our model and to generate data for the capture of CO2 from flue gas by VSA, a three-column pilot-scale apparatus was constructed (17). From our experience, it is difficult to scale up benchtop PSA data reliably because the
U.K. or Canada Published 2008 by the American Chemical Society VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 563
TABLE 1.
Adsorbent and Apparatus Properties parameter
value
kg/m3
640.74 0.201 1.89 2.20 1200 1000 76 5 3.61
bulk density, micro pore volume, m3/g particle density, g/cm3 skeletal density, g/cm3 bed length, mm effective working length, mm internal diameter, mm wall thickness, mm adsorbent loaded in each bed, kg
large difference in surface to volume ratio between a benchtop apparatus and a full-scale plant leads to large differences in thermal effects and correspondingly large differences in process performance. Although benchtop PSA equipment tends to operate closer to isothermal than plant-scale, this does not always lead to improved performance. While it is well-known that an isothermal working capacity is generally larger than an adiabatic one, this phenomenon does not translate obviously to improved process performance. In particular, multilayered PSA systems with strongly adsorbing components (such as CO2 and water) are subject to regenerative thermal effects which can lead to dramatic temperature “cold spots” (18) and large thermal gradients. If carefully managed, these thermal profiles can be exploited to advantage by placing suitable adsorbents at locations of “optimal” temperature operation. In general then, PSA systems are most reliably designed with adiabatic operation in mind. For this reason, our adsorption beds are well insulated with a packed length of 1 m and an internal diameter of 7.7 cm ensuring close to adiabatic performance. Although this leads to high gas processing rates, the reliability of the data is well worth the additional cost of high gas throughput. The presence of 3 beds in our apparatus as well as a flexible upper and lower gas valving manifold allows us to investigate a large variety of process cycles and steps. More details about experimental apparatus and operating procedure can be found in the Supporting Information (Page S3 and Figure S1). We used UOP 13X as the adsorbent but the model is valid regardless of adsorbent type provided the correct isotherms are used for CO2 and N2. The simple adsorption model built to permit quick cycle assessment for CO2 capture was validated by experiments through changing of the operating parameters. Comprehensive examinations of cyclic functions were conducted by use of experiments and simulations. A summary of adsorbent and process conditions is presented in Table 1. Development of a Simple PSA model for CO2 Capture. As discussed earlier, a simple PSA model based on equilibrium adsorption theory is needed to permit rapid screening of potential adsorbents and to help assess the economic benefit of adsorbent and process improvement. Such a model is not intended to replace more sophisticated adsorption simulators (such as ADSIM and MINSA-a robust numerical simulator developed at Monash University) and is not intended to give exact quantitative agreements with experiments. Nevertheless, it is important to capture “first order” effects in a simple model such as the effects of adsorbent CO2 capacity and CO2/N2 selectivity as well as nonisothermal behavior, changes in amount of purge, etc. In addition, the simple model must be capable of providing trends in power requirements and bed size factors. “Second-order” effects such as mass transfer resistance and bed pressure drop can be neglected in such a simple model. The need to use realistic isotherm data precluded the use of simple Binary Linear Isotherm (BLI) (13, 19) based models or batch models. A set of assumptions were made in this model and are explained in more detail in Supporting Information, Page S4. 564
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 2, 2008
FIGURE 1. Basic step components for CO2 PSA process showing the profiles of CO2 in the bed at the start and end of each step. Our model is composed of several basic step modules according to the functions in a cycle, which includes a feed step, heavy component (CO2 product) cocurrent purge step, counter-current blowdown or evacuation step, countercurrent waste (containing N2) repressurization step, and a pressure equalization step which is split into the “provide” cocurrent pressure equalization part and a “receive” countercurrent pressure equalization part (Figure 1). Different cycles with different combinations of these step modules were constructed. The working mechanisms of these step modules are explained in Supporting Information (Supporting Information, Page S4). PSA Model Validation. To validate our model, a series of experiments was conducted with our 3-bed apparatus to test the basic functional steps discussed above. As mentioned before, the purpose of this model is not aimed at achieving accurate prediction of the process, but at predicting the correct trend changes with the change of cycles and process conditions. The following parameters (which are considered to be the most important for validation purposes) were examined: evacuation pressure, feed concentration and temperature, extent of purge, and the magnitude of temperature swing in the bed. Effect of Evacuation Pressure. From our experimental and simulation work, we have determined that evacuation pressure is the most important parameter in a vacuum swing adsorption process, and significantly affects the purity, recovery, and power. A detailed comparison of experimental and simulation results is provided in the Supporting Information, Figure S2. CO2 recovery decreased with the increase of evacuation pressure and this trend was confirmed in both experiments and simulations. Since the isotherm for CO2 is quite “steep” at low pressures,
small reductions in evacuation pressure lead to relatively large increases in moles of CO2 removed from the bed. This leads to an increase in CO2 purity and recovery. The recovery predicted by the model is higher than that of the experiments, because, in practice, there was pressure drop along the column which could be as large as 15 kPa in the adsorption step and 5 kPa in desorption step. Since the desorption pressure is only 3 kPa (measured at the bottom of the bed), a pressure drop of 5 kPa is significant. This would lead to a reduction in working capacity of the adsorbent. In addition, the model assumes complete utilization of the bed during the feed step. In practice, a mass transfer zone (MTZ) prohibits complete bed usage and some small amount of breakthrough is permitted leading to reduced recovery. The recovery difference between simulation and experiment was also reflected in the power requirements which also increased with evacuation pressure (Supporting Information, Figure S3). Therefore, evacuation pressure is a very sensitive parameter and needs to be controlled in relation to power and recovery, and, more importantly, the regeneration of the adsorbent. Lastly, the effect of nonisothermal simulation should be noted. Had we assumed isothermal operation in our model (and not accounted for the drop in temperature of up to 15 °C on desorption), we would have overestimated the amount of CO2 removed in the evacuation step by up to 20% leading to a large overprediction of process recovery and throughput. Effect of Feed Concentration and Temperature. The feed concentration of the flue gas condition varies from plant to plant and the model needs to be able to account for this variability if it is to have value as an assessment tool. In our experiments, the feed CO2 concentration was adjusted through the control in the CO2 inlet line. Product recovery was found to increase with feed concentration in both simulation and experiments, although quantitative differences of up to 10% existed between simulation and experiments (Supporting Information, Figure S4). The relatively low experimental recovery is due to the incomplete utilization of the column in the adsorption step and the pressure drop in the column. As a result of lower recovery, specific experimental power consumption is higher than that of the simulation, however, the trends of power change with feed concentration in simulation and experiments agreed quite well (Supporting Information, Figure S5). Those trends indicate that with higher feed concentration, less work would be required to separate the CO2 from the flue gases which is consistent with the minimum work of separation. It also suggests that process gas from the IGCC process which has considerably more CO2 than conventional pulverized coal fired plant flue gas (up to 40%) would be considerably easier to separate and capture. Another important parameter is feed gas temperature which, in practice, would determine the requirement for heat exchange before separation and inevitably affect the performance and process economics. While the flue gas leaving the combustion process can range in temperature from 50 to 200 °C (after heat recovery), in our model (and process), we assume that a wash tower (direct contact) is used to control the flue gas temperature to the capture plant and to knock out SOx and NOx contaminants. The model predicted a slight increase of recovery with feed gas temperature, which was in good agreement with experiments in terms of the changing trend (Supporting Information, Figure S6). In both simulation and experiments, small fluctuations in the power with temperatures were observed as a result of recovery change and other factors (Supporting Information, Figure S7). The effect of flue gas heat exchange on the power consumption has not been included in this graph.
FIGURE 2. Effect of purge ratio on specific power.
FIGURE 3. Temperature changes during each step in the 6-step cycle. Effect of Purge. In this study, the purge effect was studied through the change of purge/product ratio (defined as % of product used to purge, see x-axis Figure 2) instead of purge/ feed ratio which is an index generally used in stripping cycles, since the cycles used for CO2 capture are generally rectifying cycles. It was found that the specific power increased slightly with higher purge/product ratio (Figure 2). More purge gas fed into the column before conducting the blowdown step requires more power to be expended in the subsequent evacuation step. In practice, the purge amount is determined by product purity constraints-the target purity for CO2 transport and sequestration is a minimum of 95%. Provided the purity requirement is met, excessive purge is not needed considering overall capture economics. It is important to note that the purge stream leads to a large increase in temperature in the bed (indeed this thermal front during the purge step acts as a useful diagnostic to control the extent of purge). Once again, had we assumed isothermal operation in our model, this fact would have been ignored and we would not have been able to predict the effect of purge on recovery and purity. Observed Temperature Variation in a Cycle. In our experiments, the adsorption front movement was monitored through temperature swings at different locations along the column. Therefore, the model must be able to predict the temperature changes in a cycle if it is to mimic the movement of the concentration fronts. The temperature variations agree very well between simulation and experiment which demonstrates the ability of the model to capture non-isothermal effect (Figure 3 and Supporting Information, Figure S8). The simulated temperature swings are somewhat greater than that of experiments as a result of pressure drop (and reduced working capacity), heat loss, and somewhat lower working capacity in the experiments produced through mass transfer effects. Through the parametric comparison above, it is concluded that our simple model is capable of capturing first-order VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
565
FIGURE 4. Pyramid of cycle hierarchy. FD ) feed, feed pressurization; EV ) counter-current evacuation/blowdown; WP ) counter-current waste pressurization (also called LPP); PG ) heavy component (product gas) cocurrent purge (also called heavy reflux or HR or HR-IP); PE/RPE ) cocurrent provide pressure equalization and counter-current receive pressure equalization; and WPG ) counter-current light component (waste gas) purge (also called light reflux or LR). effects for fast simulation of CO2VSA performance and is a useful tool for rapid evaluation of CO2 capture by adsorption processes. Comparison of Cyclic Steps. As mentioned in the Introduction, many cycles, including stripping, rectifying, and dual-reflux, have been designed and tested by simulation and/or experiments (20). However, among all those rectifying cycles (designed for heavy component recovery), some cyclic steps/functions are very basic and commonly adopted. Based on those functions, a pyramid of cycle hierarchy (Figure 4) was developed and studied in this research. Different combinations of the basic cycle steps would create a great number of cycles, some of which have been reported in the literature as indicated earlier. However, justifications for certain cycle steps and corresponding configuration of the cycle are often neglected. For example, much effort has been put into use of the Skarstrom cycle and its modified forms which, in essence, are stripping cycles. These are not appropriate cycles for heavy component purification, i.e., CO2 capture. Our goal here was to understand the exact function/mechanism of the basic steps in a cyclic process to permit quick construction of a cycle for a given separation. In the following, the functions of the cycle steps were investigated by both experiments and simulation. Central to our discussion and cycle design is the concept of front management of CO2. It should also be noted that development of cycles is bound by scheduling constraints which in turn may lead to more complex cycles having longer cycle times (or more beds) than simpler cycles. This is turn reduces the “bed size factor” or bed productivity unless the recovery of the process is dramatically improved by the more complex cycle. On the other hand, more complex cycles with more beds improve the opportunities for multiple equalization steps and energy recovery and hence the final cycle design is closely coupled to the economic model for the capture process. Function of the FEED Step. In the FEED step, which is generally the starting point of a cycle, the heavy components and light components (resulting from the fresh feed as well as any recycled effluents from the process) are fed to the column which initially contains CO2 at a lower concentration than the feed gas. Conditions of pressure, temperature, and 566
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 2, 2008
flow rate all affect the amount of CO2 adsorbed from the feed gas. The primary role of the feed step is to provide adsorption of CO2 from the feed stream. It is essential that the maximum use is made of the column under the specified feed gas conditions, however, the adsorption front should be controlled properly to avoid the occurrence of breakthrough during this and subsequent steps. After all, the role of this step is to adsorb as much desired component/components as possible while avoiding breakthrough which may lead to low recovery. For a given feed gas mixture, the adsorption front control can be achieved through adjusting the feed pressure, flow rate, and/or the step time. The feed pressure can in turn be controlled by adjusting feed or waste line valve settings. Initial column breakthrough experiments are a very good indicator for the influence of feed flow rate and determination of mass transfer zone lengths. Different flow rates lead to different movement speed of the adsorption front and hence, different mass transfer zone length (MTZL), which is an important parameter for column design (e.g., MTZL < 0.2 length of column). To assess the role of feed time (i.e., length of mass transfer zone) on performance, VSA experiments were conducted at feed times of 30, 50, 70, and 100 s. The location of the adsorption front can be inferred through study of the thermal front location (Supporting Information, Figure S10). During these experiments, the cycle used was a simple 4-step cycle with pressure equalization but without purge, and the evacuation step time was also adjusted to ensure scheduling constraints were met. In each instance the final evacuation pressure was held constant. For a 30-s feed step, the adsorption front reached about 500 mm into the bed where the temperature swing had declined from about 12 °C to about 0 °C. With the increase of the feed time to 100 s, the adsorption front gradually moved to the exit of the column where the temperature swing at the top of the column was 4 °C, indicative of breakthrough. A feed step time of 30 s therefore corresponds to underutilization of the bed during feed while 100 s corresponds to excessive use and breakthrough. With fixed evacuation conditions, the product purity increased with longer feed step (Figure 5). The reason is that the average CO2 concentration in the column is increased at higher bed utilization which leads to higher product purity. However, as breakthrough occurred at the longer feed step, the loss in CO2 corresponds to higher CO2 in the waste stream and thus lower recovery which leads to an increase in the specific power. These trends can also be explained by the feed step illustration in the simple model. Different feed times determine the front position, and different temperature and feed concentrations determine the gas concentration in the column or the “height” of the adsorption front. Therefore, in a specific cycle and equipment design, the specific feed step parameters, including flow rate, temperature, heat of adsorption, feed step time, feed pressure, and temperature, should be thoroughly understood to realize the best use of column and maximize recovery subject to purity constraints. Function of the Evacuation Step (EV). The role of the evacuation step is to desorb adsorbates from the adsorbent into a product tank(s). We employed counter-current evacuation although it is possible to employ cocurrent or dualend evacuation. The latter may be beneficial especially if low (1 kPa) vacuum levels are required and pressure drop must be minimized. It was found that the purity increases with deeper evacuation while such an increase remains quite flat beyond a certain evacuation point (Supporting Information, Figure S11). In addition, the absolute cyclic power increases with
FIGURE 5. Effect of feed time on purity, recovery, and specific power. deeper evacuation but specific power may be lower as a result of higher recovery and higher product purity. There is a very important tradeoff among evacuation pressure, purity, recovery, and specific power. The duration of the evacuation step plays a similar role in determining product purity and recovery. A longer evacuation time was found to provide more CO2 product. Since the same vacuum pressure is achieved at the end of the step in all cases (long or short evacuation times), the cause for higher recovery and purity for longer evacuation time cannot simply be equilibrium capacity. Rather, with the same vacuum level reached, longer evacuation time implies smaller vacuum flow rate, which leads to lower pressure drop in the column than short evacuation times. Thus longer evacuation results in lower average pressure in the column (the pressure gauge is placed at the bottom of the bed) than shorter evacuation times and hence the effect of short evacuation time is equivalent to higher average vacuum level with a corresponding effect on performance. In experiments, it was observed that longer evacuation step time leads to more desorption (lower average pressure) and consequently a larger temperature change (Supporting Information, Figure S12). It is also found that the purity increased slightly with longer evacuation time while recovery has a greater increase as expected (Supporting Information, Table S1). Specific power for the 200-s evacuation cycle is smaller than that of the 100-s cycle though the absolute work needed per cycle for the 200-s evacuation cycle is greater than that of 100-s evacuation cycle. The specific power can be calculated as Specific Power )
Wtotal × τ
nCO2,cycle ×
1 ) 24 × 3600 × 44.01 ⁄ 106 τ Wtotal 1 × (1) nCO2,cycle 3.8016
Equation 1 indicates that specific power is proportional to the total work per cycle and inversely proportional to CO2 moles produced per cycle. Therefore, any factors affecting the total work or the CO2 moles may influence the specific power. Longer evacuation steps times means more cyclic work but also brings in more CO2 which in fact may offset
each other to give an unpredictable specific power number that needs to be optimized according to the specific process and equipment. Function of the Purge Step (PG). A heavy component purge step is conducted to enrich the column in CO2 prior to evacuation. Front management becomes very important during this step since excessive heavy component purge is not beneficial to purity or recovery. The effects of different purge/product ratio on specific power have already been mentioned in the model validation (Figure 2). With more product purge gas fed (cocurrently) into the column, the purge front moves further and further into the bed displacing the gas from the previous step and eventually, if not controlled, would break through the column as indicated by the temperature profiles (Supporting Information, Figure S13). The large increase in temperature observed in the purge step is the result of high-concentration CO2 product displacing the high nitrogen content adsorbent. Thermocouple T7 is located near the end of the bed and shows that the CO2 purge front has clearly broken through the bed. Combining with the performance data shown in Figure 2, it is quite clear that the purge front needs to be controlled accurately to avoid purge breakthrough which would result in a lower recovery as the top of the column is contaminated with highconcentration CO2. To further clarify the function of the purge, a 5-step cycle without purge (FD-PE-EV-RPE-WRP) and a 6-step cycle with purge (FD-PE-PG-EV-RPE-WRP) were examined by both simulation and experiments. Both cycles were run under 12% feed, PH ) 120 kPa, PL ) 5 kPa.abs, 40 °C feed. Temperature profiles of both experiments and simulations showed fairly good qualitative agreement (Supporting Information, Figure S14 and Figure S15). However, due to the great amount of heat loss in the experiment (ambient temperature 15 °C and the column is metal), the simulation temperature curves are above those of experiments and due to the pressure drop in the column (neglected in the simulation), the feed and evacuation temperature swing of the experiments are not as great as the simulation. It is clear that the simulation has qualitatively represented the purge step quite well. The purge step can be considered as a “second-stage” separation (of the product) which dramatically increases the product purity at the expense of recovery. VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
567
TABLE 2.
Comparison of Cycles run
purity, %
recovery, %
power, kW/TPDc
without PG
experiment simulation
85.7 88.0
77.6 78.6
9.8 6.4
with PG
experiment simulation
95.2 95.0
66.9 77.6
12.1 8.9
without PE
experiment simulation
88.5 95.0
68.8 78.1
11.4 6.6
with PE
experiment simulation
95.2 95.0
66.9 77.6
12.1 8.9
The performance results are shown in Table 2. As expected, the purge step would displace the light component and increase the initial CO2 concentration in the column before commencement of the evacuation step which consequently increases the average CO2 purity in the product. However, on the other hand, because of purge, the CO2 concentration in the column after purge, evacuation, and repressurization is higher than that without purge and this may cause the feed gas front during the subsequent feed step to move further toward the top of the column, even breaking through the column, leading to a reduction in CO2 recovery. As a result of lower recovery, the power cost for cycle with purge is higher than a cycle without purge. Some adjustment of the feed time may offset this reduction in recovery. This complex coupling between process steps and operating conditions make P/VSA operation a challenging system to optimize and design. Function of Pressure Equalization (PE). Pressure equalization (PE) is commonly regarded as a step for improving purity, recovery, and reducing power cost in a PSA operation. Performances from simulations and experiments are shown in Table 2. Both show that the PE step increases purity and recovery while reducing the power consumption. For cocurrent pressure equalization, the pressure is reduced and the gas adsorption front is stretched toward the top of the column. More low-concentration gas will flow from the column leading to higher CO2 concentration. This process may be
FIGURE 6. Effect of light reflux on performance parameters. 568
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 2, 2008
considered a cocurrent evacuation process without power consumption. However, excessive pressure equalization leads to difficulty in maintaining a “clean” top of bed affecting purity of the CO2 product. This is in contrast to stripping cycles in which light reflux can be used to clean the bed and thus reduce the effect of the pressure equalization step on concentration profiles. Function of Waste Repressurization (WRP). The waste repressurization (counter-current) step functions to “push” the heavy component (CO2) front downward (toward the feed end) so as to clean the top part of the column. Its working principle is similar to light component (waste gas) purge in stripping cycles. It was observed that, in the WRP step the temperatures along the column dropped to varying degrees (Supporting Information, Figure S16). The purpose of this step is 2-fold: one is to increase the column pressure from vacuum to atmospheric and the second is to clean the top of column. Repressurization with feed (the alternative) will accomplish the first but not the second purpose. Both simulation and experiments confirm that waste repressurization may cause the readsorption of carbon dioxide as a result of relatively high CO2 partial pressure, which is also reflected by the shape of isotherm. It is found that the presence of waste repressurization on the performance does improve the recovery though the increase is small (Supporting Information, Table S2). In addition, the effect of WRP on purity is negligible in the range of conditions investigated here. Therefore, the function of WRP is to recycle some CO2 from the waste gas and to clean the top of the column, though its effect is not as obvious as light component purge/light reflux discussed next. Function of Waste Purge/Light Reflux (WPG/LR). To further test the light component rinse effect, a waste gas purge step was included in the cycle. Waste purge causes more CO2 to desorb which was indicated by a drop in the temperature in the WPG step. The cleaning of the column is effective along the whole column. However, as light component gas N2 reached the bottom of the column, (N2 breakthrough into the product line), the product purity decreased. It is found that with less waste purge, the product purity increases while more waste purge brought more recovery, and consequently resulted in lower power consumption (Figure 6). A tradeoff between purity, recovery, and power should be considered
when conducting waste purge. With proper control of the waste purge front, higher purity and higher recovery can be realized simultaneously. The very short amount of waste purge required is an indication of the vastly different isotherm shapes for CO2 and N2. The slope of the CO2 isotherm is much greater than that of N2. Consequently, the proportionate pattern induced during light reflux leads to rapid movement of N2 through the bed and breakthrough into the CO2 product stream. Front management during this step is essential to avoid contamination of CO2 product. Different sequential arrangements of the basic steps makes a great difference to the movement of gas composition fronts which should be managed carefully according to a specific cycle and a particular set of conditions. For such a highly coupled system, one single parameter change should be accompanied by proper changes of other parameters by assessing adsorption front movement as determined through temperature profiles. Especially for CO2 capture from flue gas of major emission sites, specific power, as an indicator of capture cost, should be taken into account when designing and configuring cycles.
Acknowledgments We acknowledge financial support for this work from the Cooperative Research Centre for Greenhouse Technologies (CO2CRC), which is established and supported under the Australian Government’s Cooperative Research Centres Program, and Monash University.
Appendix A Cs ∆H k m n P Q R T TPDc V W •
W y Z
adsorbent heat capacity (Joule kg-1K-1) heat of adsorption (J gmole-1) ratio of heat capacities (Cp/Cv) mass of adsorbent (kg) gas moles adsorbed (gmol/kg adsorbent) pressure (kPa) volumetric flow rate (m3/s) universal gas constant temperature (K) tonnes of carbon dioxide per day volume (m3) work (kJ) instantaneous power (kJ/s) mole fraction in gas phase column length (m)
Greek Letters void fraction η vacuum pump and compressor efficiency τ cycle time (seconds)
Supporting Information Available Part of experimental and modeling details, Figures S1-S16 and Tables S1 and S2. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Stern, N. Stern Review on the Economics of Climate Change; Cambridge University Press: Cambridge, 2006. (2) Gielen, D. The future role of CO2 Capture and Storage Results of the IEA-ETP Model; International Energy Agency: Paris, 2003. (3) IEA. The Prospects for CO2 Capture and Storage; International Energy Agency: Paris, 2004; p 250. (4) Reynolds, S. P.; Ebner, A. D.; Ritter, J. A. Stripping PSA cycle for CO2 recovery from flue gas at high temperature using a hydrotalcite-like adsorbent. Ind. Eng. Chem. Res. 2006, 45, 4278– 4294. (5) Park, J.-H.; Beum, H.-T.; Kim, J.-N.; Cho, S.-H. Numerical analysis on the power consumption of the PSA process for recovering CO2 from flue gas. Ind. Eng. Chem. Res. 2002, 41, 4122–4131. (6) Baek, J. I.; Eom, H. M.; Lee, H. U.; Lee, J. G.; Na, B. G.; Sim, J. G.; Song, H. G.; Yang, G. S.; Yoon, J. H. Pressure swing adsorption system for recovering carbon dioxide using activated carbon and zeolite; Patent KR 2002003963; 2002. (7) Chang, D.; Min, J.; Moon, K.; Park, Y.-K.; Jeon, J.-K.; Ihm, S.-K. Robust numerical simulation of pressure swing adsorption process with strong adsorbate CO2. Chem. Eng. Sci. 2004, 59, 2715–2725. (8) Ebner, A. D.; Ritter, J. A. Equilibrium theory analysis of dual reflux PSA for separation of a binary mixture. AIChE J. 2004, 50, 2418–2429. (9) Hirose, M.; Omori, I.; Oba, M.; Kawai, T. Carbon dioxide separating and recovering system; Patent JP2005262001; 2005. (10) Matsumura, Y.; Sadagata, K.; Kakigami, H.; Inoue, G.; Minemoto, M.; Yasutake, A.; Oka, N. Simulation of CO2 recovery system from flue gas by honeycomb type adsorbent II (optimization of CO2 recovery system and proposal of actual plant). Kagaku Kogaku Ronbunshu 2006, 32, 146–152. (11) Chou, C.-T.; Chen, C.-Y. Carbon dioxide recovery by vacuum swing adsorption. Sep. Purif. Technol. 2004, 39, 51–65. (12) Chiang, A. S. T. An analytical solution to equilibrium cycles. Chem. Eng. Sci. 1996, 51, 207–216. (13) Knaebel, K. S.; Hill, F. B. Pressure Swing Adsorption: Development of an Equilibrium Theory for Gas Separations. Chem. Eng. Sci. 1985, 40 (12), 2351–2360. (14) Harlick, P. J. E.; Tezel, F. H. Equilibrium Analysis of Cyclic Adsorption Processes: CO2 Working Capacities with NaY. Sep. Sci. Technol. 2005, 40, 2569–2591. (15) Pigorini, G.; LeVan, M. D. Equilibrium Theory for Pressure Swing Adsorption. 5. Separation and Purification in Multicomponent Adsorption. Ind. Eng. Chem. Res. 1999, 38, 2439–2449. (16) Serbezov, A.; Sotirchos, S. V. Semianalytical solution for multicomponent PSA: application to PSA process design. Sep. Purif. Technol. 2003, 31, 203–223. (17) Chaffee, A. L.; Knowles, G. P.; Liang, Z.; Zhang, J.; Xiao, P.; Webley, P. A. In CO2 Capture by Adsorption: Materials and Process Development,8th International Conferences on Greenhouse Gas Control Technologies; Trondheim, Norway, 2006. (18) Wilson, S.; Webley, P. A. Cyclic Steady-State Axial Temperature Profiles in Multilayer, Bulk Gas PSA-The Case of Oxygen VSA. Ind. Eng. Chem. Res. 2002, 41, 2753–2765. (19) Ruthven, D. M.; Farooq, S.; Knaebel, K. S. Pressure Swing Adsorption; VCH Publishers Inc.: New York, 1993; p 352. (20) Kearns, D. T.; Webley, P. A. Modelling and evaluation of dualreflux pressure swing adsorption cycles: Part I. Mathematical models. Chem. Eng. Sci. 2006, 61, 7223–7233.
ES0706854
VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
569
