In 2009, Bechtel carried out a front-end engineering design (FEED) study for the Kårstø CO₂ capture and compression plant project in Norway (Bechtel for Gassnova, 2009). The design considered a PCC plant, using MEA as the solvent at a 35% w/w concentration, attached to an existing 420 MW CCGT power plant. The PCC plant was designed to capture at least 85% of the CO₂ contained in the flue gas emitted by the CCGT plant, giving an annual capture rate of just over 1 MtCO₂/yr.

In 2021, Bechtel delivered a similar FEED study (Elliott et al., 2021a), funded by the US DOE, for a PCC plant that captures approximately 85% of the CO₂ in a slipstream of 68% of the flue gas (an amount corresponding roughly to the flow at minimum stable generation) from a larger CCGT plant located in Sherman, Texas, United States, commissioned in 2014 and with a rated capacity of 758 MWe.

Similar to Kårstø, the Sherman design is based on using a low-cost generic solvent, i.e., MEA at 35% w/w. The flue gas from the HRSG is cooled from 92 to 51°C using a fogging system in the connecting duct; no direct contact cooler (DCC) is employed. Two identical absorbers of 11.8 m in diameter each with a packing height of 15 m (6 m upper bed; 9 m lower bed) operating in parallel are used to capture the CO₂. Solvent is regenerated in a stripper of 6.67 m in diameter with a packing height of 20 m. In addition, a (20%) slip-stream of rich solvent is heated using heat from the CO₂ compressor intercoolers, then flashed and the semi-lean solvent is sent back to the absorber between the upper and lower packing beds. The design of the equipment and the quantification of material and energy flows in the Bechtel FEED study have been established by means of simulation using ProMax^®. Some of the main plant design parameters are presented in Table 1 and the predicted performance is shown in Table 2; more details can be found in the comprehensive FEED study for the Kårstø plant (Bechtel for Gassnova, 2009) and in (Elliott et al., 2021a); a full FEED study for the Sherman plant is also expected to be published in due course.

We undertook process modelling in Aspen Plus V10 utilising the add-on MEA Steady State Model developed by the DOE Carbon Capture Simulation Initiative (CCSI) (Morgan et al., 2018; CCSI, 2021). The various developed user models contained in the library include physical property models for viscosity, density, surface tension, and diffusivity; the hydraulics model; the interfacial area model; and the reaction kinetics model. As part of the CCSI MEA Steady State Model development, these sub-models have been integrated into a complete process model, and tuned and validated against data from the PCC unit at the National Carbon Capture Center (NCCC) in Wilsonville, Alabama (Morgan et al., 2018). The hydrodynamic models developed in the CCSI/Aspen work include models for pressure drop and hold-up. The Billet and Schultes correlation was regressed with data from Tsai (2010) for MellapakPlus™ 250Y packing.

A key feature of our study is the consideration of lower lean loadings (<0.2) and L/G ratios (<1) than have usually been employed in previous PCC studies. This combination would not be appropriate for coal flue gases due to their higher CO₂ concentration, because it would lead to excessive temperature bulges in the absorber, but it appears to be useful for CCGT flue gases, especially at higher capture levels. Fortunately, the experimental programme at NCCC used to support the development of the CCSI MEA models has been deliberately designed to cover a wide range of operating conditions, and a number of low-load data points are available for comparison with model predictions. As shown in Table 2, CCSI/Aspen predictions for stripper energy requirements with low lean loadings are generally higher than experimental values inferred from NCCC tests, suggesting that CCSI thermodynamic correlations, if anything, underestimate the ease of achieving low lean loadings.

A comparison between CCSI/Aspen model predictions and the results derived by Bechtel using ProMax^® simulation software for the Sherman FEED study (using the same rich flash and semi-lean return arrangements) is shown in Table 3. The CCSI/Aspen kinetic model is observed to predict a lower rich loading than the ProMax^® equilibrium model for a lean loading of 0.254 molCO₂/molMEA and hence a higher SRD. Increased rich loadings are, however, predicted by the CCS/Aspen model at lean loadings of 0.2 molCO₂/molMEA and below, with the absolute minimum SRD, which is only 2% higher than the ProMax^® values, predicted for a lean loading of 0.12 molCO₂/molMEA at 15 m packing height and 85% capture level.

The current study adopts the Sherman flue gas properties (flow and composition) and the twin absorber design with an internal diameter of 11.8 m, but absorber packing heights of 18 m (2 × 9 m) and 24 m (3 × 8 m) are used, as well as 15 m (9 m lower, 6 m upper), to allow for more efficient high CO₂ capture. The stripper is identical to the Sherman design. Both the absorber and the stripper use Sulzer MellapakPlus™ 250Y structured packing (Sulzer, 2021). In order to identify optimal operation windows for capture levels above 90% up to 99%, the lean loading and the liquid to gas ratio have been varied within the ranges of 0.1–0.25 molCO₂/molMEA and 0.6–1.1 kg/kg, respectively. Similar to Sherman, for all runs the MEA concentration is 35% w/w.

The process flow diagram, as implemented in Aspen, is depicted in Figure 1. The rich split is heated, as in the Sherman case, using heat recovered as LP steam at 126°C from the CO₂ compressor intercoolers, but is then sent to the stripper with the other rich solvent instead of going to flash. This approach avoids the need for (perhaps infeasible, due to different packing bed dimensions) reinjection of the semi-lean solvent from the flash at a suitable point in the absorber while still allowing for the use of available waste heat and a lower rich flow to the cross-flow heat exchanger, and hence a hot rich stream at a higher temperature.

Stripper performance at low lean loadings (down to around 0.1 molCO₂/molMEA for MEA) is the key factor in determining PCC performance at higher capture levels. Operation at low lean loadings has not been a priority for “traditional” PCC applications because capture levels have not usually been above 90% and also, when coal power plants were the main target application, because of the relatively high CO₂ concentrations in the flue gas. With high-CO₂ coal flue gases, relatively high L/G ratios (>1) are desirable to limit solvent temperature rise; this in turn implies a relatively low lean-to-rich loading interval and, hence, since the optimal rich loading is as high as reasonably possible to minimise energy requirements, a relatively high (typically >0.2 molCO₂/molMEA) optimal lean loading.

For a given amount of CO₂ captured and constant rich loading, lower lean loadings also increase specific energy consumption, as shown in Figure 2, which, for a capture level of 95% using the modified Sherman case (i.e., a fixed CO₂ capture rate of 39.985 kg/s), reports calculated specific energy consumption for three rich loadings, 0.422, 0.446, and 0.463 molCO₂/molMEA. These rich loading values have been selected from cases that will be presented in the following sections. For all rich loading levels in Figure 2, the increase in SRD is, however, relatively small for lean loadings down to an inflexion point at approximately 0.12 molCO₂/molMEA, while below this a sharp increase is observed (and, as will be shown below, in practice the coupled absorber characteristics can lead to slight increases in rich loading in a full PCC system as lean loading is reduced).

The overall reason for this initial insensitivity to reduced lean loading can be inferred from Figure 3, showing stripper performance data for a constant capture level of 95% (corresponding to a capture rate of 39.985 kgCO₂/s) and a constant rich load of 0.446 molCO₂/molMEA. Under these conditions, the lean solvent flow (Figure 3A) reduces as the lean loading reduces. Less sensible heat is then required and, as far as the inflexion point, this offsets the increased energy demand due to the slight increase in the temperature of the CO₂/vapour mixture leaving the top of the stripper (Figure 3B) and the slight increase in water vapour levels at the top of the stripper (TWV) (Figure 3C). At lean loadings below the inflexion point, however, the TWV rises more rapidly, and the additional latent heat of vaporisation required is not offset by a corresponding decrease in solvent flow; as a consequence, the overall SRD (Figure 3E) exhibits the rapid rise already noted.

The reason for the rapid rise in TWV at the inflexion point is tentatively interpreted as being due to the inability of the stripper column to effectively use all of the available latent heat in the water vapour entering with the CO₂ at the bottom (bottom water vapour, BWV) for heating the down coming solvent. This is a thermodynamic and not a kinetic limitation in these modelling results; the CCSI package uses an equilibrium model for the stripper. The Sherman PCC design deliberately provided what was assumed to be ample packing height (20 m), so a close approach to equilibrium is likely to be reasonable in this case. More detailed modelling and pilot tests for verification would be needed to explore the lower limits on stripper packing height.

Figure 3D shows the variation in BWV with lean loading. The BWV flow increases with reducing lean loading (and temperatures at the bottom of the stripper also increase), as required by the H₂O/MEA/CO₂ vapour/liquid equilibrium (VLE) in the stripper sump/reboiler, until it reaches a critical value (∼3.7 mol/mol CO₂ stripped in this example). At lower BWV flows, the sensible heat that can be recovered by condensing rising water vapour in the stripper column is less than that which could be applied to heating the down coming solvent and releasing some of the CO₂, with the balance of the heat required being supplied directly in the reboiler. Beyond the inflexion point, however, the heat available from the BWV is greater than required and the excess goes into unproductively raising the temperature and water vapour content (TWV) of the exiting CO₂.

In line with the Sherman design, full PCC plant performance results in this paper are reported for a total reboiler pressure of 2.4 bar. Stripper performance results at lower reboiler pressures are, however, also shown in Figure 3. As can be seen, trends at reboiler pressures down to 1.5 bar follow those at 2.4 bar, although the VLE requirements mean there is a trend for lower temperatures and higher vapour/CO₂ ratios as the total pressure falls. SRD values at all pressures exhibit an inflexion point that occurs at a similar BWV flow (Figure 3D) but, because of the VLE characteristics, at a lean loading that increases with decreasing pressure.

Operation at higher reboiler/stripper pressures is therefore advantageous for achieving very low lean loadings to allow higher capture levels and also slightly reduced SRD values generally. The drawbacks are expected to be higher levels of thermal degradation of the MEA and increased reclaiming requirements. As discussed in the Sherman FEED study, long-term testing would, however, be required to determine actual solvent management costs with the combination of elevated stripper pressures and semi-continuous reclaiming proposed, noting also that MEA replacement costs were reported as being relatively low (Elliott et al., 2021b) (for a MEA price of 1.15 $/kg, which corresponds to 2.3 $/tCO₂ captured at the estimated solvent consumption rate of 2 kg MEA/tCO₂).

The rich loading will also have a significant effect on the energetic performance of the PCC plant. The partial pressures of CO₂ and water in equilibrium with the liquid in the stripper affect the total heat input required for water evaporation. Additionally, lower rich loadings are associated with increased L/G ratios and hence a higher sensible heat requirement. To examine the impact of rich loading alone, values were varied for three constant lean loadings of 0.25, 0.2, and 0.12 molCO₂/molMEA. To do this, an effectively infinite absorber height was assumed and L/G ratios were increased from the value corresponding approximately to the thermodynamic limit (for the 4.19% v/v dry flue gas CO₂ concentration in this study and CCSI thermodynamics) of 0.47 molCO₂/molMEA to give lower rich loadings. Figure 4 depicts the effect of rich loading on the SRD; it can be seen that for all three cases, there is an almost linear relationship between the rich loading and the SRD. This increase arises because, as Figure 5 confirms, both L/G ratios, and hence sensible heat requirements, and also TWV levels increase approximately linearly with decreased rich loading. Similar trends are observed in the case of a stripper operated at 1.5 bar in which higher energy demands are observed for the same or similar (i.e., low) lean loadings (see Figure 4) and, as depicted in Figure 6 higher amounts of TWV are required compared to Figure 5, even though the lean loading is higher.

Based on the sensitivity of SRD to low and rich loadings shown above, it can be concluded that, at the design stripper pressure of 2.4 bar, changes in lean loading per se will not have a major effect, down to around 0.12 molCO₂/molMEA. If, however, a change in lean loading (increase or decrease) results in a lower rich loading through its effect on processes in the absorber, then a significant increase in SRD can be expected.

Figure 7 shows the effect on the SRD of varying the lean loading at discrete 0.01 molCO₂/molMEA intervals for capture levels of 90%, 95%, 96%, 97%, 98%, and 99% and for absorber packing heights of 15 m (assumed to be two beds, 6 and 9 m), 18 m (two beds, 9 m each), and 24 m (three beds, 8 m each). Detailed results can be found in the Supplementary Material.

At lower capture levels, the SRD values are relatively insensitive to lean loading, but as the capture level increases, the SRD varies more significantly with lean loading, with the minimum value occurring at lean loadings in the range of 0.10–0.13 in all cases. As would be expected, though, at the lower packing heights, lower rich loadings are achieved, resulting in significantly increased energy requirements.

A summary of the optimal energy performance at varying capture levels is presented in Figure 8. The minimum (within the limits of the discrete L/G values used) SRD values for the 15 m packing height absorber are achieved at a lean loading value of 0.13 molCO₂/molMEA for 90% up to 97% capture, and consequently, the L/G ratio has to be increased from 0.67 to 0.84 over this range to meet the desired capture level. Even lower lean loading is necessary for 98% and 99% capture, i.e., 0.12 and 0.1 molCO₂/molMEA, respectively. In addition, the available packing area limits rich loading values for 98% and 99% capture, and higher L/G ratios (0.88 and 0.96, respectively) are necessary to capture the additional CO₂. The SRD therefore exhibits a sharp increase beyond the 97% capture level with 15 m of absorber packing.

For 18 m of packing, low leans of 0.13 molCO₂/molMEA are necessary for the absolute minimum SRD values at all capture levels except for the 99% case, for which the minimum occurs at 0.11 molCO₂/molMEA. Compared to 15 m of packing, the SRD vs. L/G trend is more flat, since more mass transfer area is available to allow the rich loading to approach its asymptotic value. While a sharp increase in SRD is also observed for 99% capture, the absolute increase is smaller than that for 15 m of packing, at roughly 4.7 GJ/tCO₂ instead of around 6 GJ/tCO₂.

For 24 m of packing and 90% capture, a relatively higher lean of 0.17 molCO₂/molMEA gives the minimum SRD, but for higher capture levels, lean loadings of 0.13 molCO₂/molMEA are required, apart from the 99% capture case, where a lean loading of 0.11 molCO₂/molMEA is optimal. In addition, the L/G ratios and the energy consumption trend are almost flat for capture levels up to 97%, which might be taken as an indication that 24 m is more than enough packing for this range of capture levels. Small increases in the L/G ratio and the SRD are observed for 98% and 99% capture, however, suggesting that this packing height might be appropriate for ultra-high capture levels without excessive increases in SRD. SRD trends as a function of lean loading for the three packing heights are also plotted for comparison in Figure 9.

Possible benefits at higher capture levels from greater pre-cooling of the inlet streams to the absorber and from intercooling the (limited to low lean loadings) have also been investigated.

Current United Kingdom guidance on post-combustion capture for power applications (Environment Agency, 2021) recommends designing PCC systems for a capture level of 95% but, given that market forces can be expected to incentivise higher capture levels in order to achieve net-zero emissions, the scope to operate plants designed for higher capture levels is worth considering. Limits on increasing the capture level beyond 95% are somewhat different, depending on whether they are to be achieved on a plant during:

a) Continuous full-load operation;

For continuous operation at full load, limits on capture level might include:

i. Heat input—additional reboiler steam will be required; in the cases above, the minimum increase, with 24 m of packing, from 95% to 99% capture level was an additional 12%.

Table 7 shows how these factors vary between 95% and 99% capture for the three packing heights used in this study. As might be expected, the 24 m packing case shows the greatest potential for extension from 95% to 99% capture with minimal or no modifications, although it must be emphasised that project-specific factors will determine actual characteristics.

a) Variable-load operation.

In practice, a CCGT + PCC plant may not be required to operate at full load all of the time due to electricity market constraints (e.g., variable demand, intermittent renewable supplies), with frequent periods with the plant either shut down or operating at minimum stable generation (MSG). This raises the prospect of additional operating modes in which higher capture levels could be desirable. In most cases, variable operation will be predictable for at least a day ahead, and if unpredicted grid requirements demand more urgent action to give an unexpected boost in electricity supplies, then any short-term reduction in capture level due to unpredicted operating requirements should not be an issue since:

a) In a situation like this, generation to ensure supply security will always be more valuable than the extra CO2 emitted.

Operation with higher capture levels might generally be expected to be facilitated at MSG. A PCC plant designed for 100% CCGT output will obviously be operating at below its design levels in all or most respects at MSG. The main problem is likely to be obtaining adequate amounts of reboiler steam at a high enough pressure (and hence saturation pressure)—this may require modifications to the steam cycle, including tapping steam from the hot or cold reheat instead of the IP/LP crossover.

For starts and stops, full capture can be achieved using rich and lean solvent storage, as described in (AECOM, 2020). The AECOM study was based on modelling at 95% capture but, provided sufficient lean solvent at a low enough loading is available, then any capture level would be feasible. If additional PCC plant capacity is, however, also required to regenerate stored rich solvent during the period of operation following storage-supported start-up, as envisaged in this report, then the necessary capacity to shift to running at higher capture levels than the nominal design value may be constrained.

Further analysis is beyond the scope of this paper; detailed project-specific design work would be required to investigate all of the above features and make a judgement on the basis of future expected operating patterns with respect to any additional provisions involving capital expenditure.

Although a more detailed assessment of capital expenditure issues around increased capture levels is beyond the scope of this paper (and will depend on site-specific factors and also the terms of any support mechanisms for CCS deployment), an illustrative estimate can be made of the conditions under which the costs associated with the reduced electricity output due to higher CO₂ capture levels would be offset by the reduction in CO₂ emission costs. This corresponds to the operating decisions that would be made in a suitable plant, where any investment costs to enable higher capture levels have already been incurred and so are no longer a factor. Even if actual capture cost support mechanisms may differ, exposure to market electricity and carbon prices is a (GOV.UK, 2021) basic principle. Table 8 depicts the basis for the calculations, and Figure 14 depicts the break-even CO₂ emission cost to move from 95% to ultra-high capture levels for different packing heights for an assumed electricity selling price of £100/MWh. Above 98% capture, for all packing heights, an exponential increase (but with different slopes) is observed for the break-even CO₂ emission cost. The taller absorber (24 m packing) requires only 80.22 £/tCO₂ to break even with the 95% base case. For a packing height of 18 m a value close to 210 £/tCO₂ would be needed to move to 99% capture. As a purely illustrative comparative benchmark, the Climate Change Committee’s net-zero scenario studies for the United Kingdom’s Sixth Carbon Budget (CCC, 2020) assume carbon dioxide removal (CDR) costs of 90–300 £/tCO₂ for BECCS and 180–300 £/tCO₂ for DACCS in 2050, suggesting that operation at capture levels higher than 95% and possibly up to 99% would potentially be cost effective if CDR is the alternative.