Challenges and uncertainties of ex ante techno-economic analysis of low TRL CO2 capture technology Lessons from a case study of an NGCC with exhaust gas recycle and electric swing adsorption

(cid:129) The challenges of techno-economic analysis of very low TRL technologies are explored. (cid:129) Hybrid approaches to project future performance of low TRL technologies are proposed. (cid:129) Electric swing adsorption is infeasible to economically capture CO 2 from an NGCC. this work addresses the methodological challenges of undertaking techno-economic assessments of very early stage (technology readiness level 3 – 4) CO 2 capture technologies. It draws lessons from a case study on CO 2 capture from a natural gas combined cycle with exhaust gas recycle and electric swing adsorption technology. The paper shows that also for very early stage technologies it is possible to conduct techno-economic studies that give a sound ﬁ rst indication of feasibility, providing certain conditions are met. These conditions include the availability of initial estimates for the energy use of the capture technology, either from bench scale measurements, or from rigorous process models, and the possibility to draw up a generic (high level) equipment list. The paper shows that for meaningful comparison with incumbent technologies, the performance of very early stage technologies needs to be projected to a future, commercial state. To this end, the state of the art methods have to be adapted to control for the development and improvements that these technologies will undergo during the R&D cycle. We call this a hybrid approach. The paper also shows that CO 2 capture technologies always need to be assessed in sympathy with the CO 2 source (e.g. power plant) and compression plant, because otherwise unreliable conclusions could be drawn on their feasibility. For the case study, it is concluded that electric swing adsorption is unlikely to become economically competitive with current technologies, even in a highly optimised future state, where 50% of the regeneration duty is provided by LP steam and 50% by electricity: the net e ﬃ ciency of an NGCC with EGR and optimised ESA (49.3% LHV ; min – max 45.8 – 50.4% LHV ) is lower than that of an NGCC with EGR and standard MEA (50.4% LHV ). Also, investment and operational costs are higher than MEA, which together with ESA ’ s lower e ﬃ ciency leads to an unfavourable levelised cost of electricity: 103 € /MWh (min – max 93.89 – 117.31 € /MWh) for NGCC with ESA, versus 91 € /MWh for NGCC with MEA.

A B S T R A C T this work addresses the methodological challenges of undertaking techno-economic assessments of very early stage (technology readiness level 3-4) CO 2 capture technologies. It draws lessons from a case study on CO 2 capture from a natural gas combined cycle with exhaust gas recycle and electric swing adsorption technology. The paper shows that also for very early stage technologies it is possible to conduct techno-economic studies that give a sound first indication of feasibility, providing certain conditions are met. These conditions include the availability of initial estimates for the energy use of the capture technology, either from bench scale measurements, or from rigorous process models, and the possibility to draw up a generic (high level) equipment list. The paper shows that for meaningful comparison with incumbent technologies, the performance of very early stage technologies needs to be projected to a future, commercial state. To this end, the state of the art methods have to be adapted to control for the development and improvements that these technologies will undergo during the R & D cycle. We call this a hybrid approach. The paper also shows that CO 2 capture technologies always need to be assessed in sympathy with the CO 2 source (e.g. power plant) and compression plant, because otherwise unreliable conclusions could be drawn on their feasibility. For the case study, it is concluded that electric swing adsorption is unlikely to become economically competitive with current technologies, even in a highly optimised future state, where 50% of the regeneration duty is provided by LP steam and 50% by electricity: the net efficiency of an NGCC with EGR and optimised ESA (49.3% LHV ; min-max 45.8-50.4% LHV ) is lower than that of an NGCC with EGR and standard MEA (50.4% LHV ). Also, investment and operational costs are higher than MEA, which together with ESA's lower efficiency leads to an unfavourable levelised cost of electricity: 103 €/MWh (min-max 93.89-117.31 €/MWh) for NGCC with ESA, versus 91 €/MWh for NGCC with MEA.

Introduction
Carbon capture and storage (CCS) is a necessary technology towards deep decarbonisation of the energy and industrial sectors, thereby mitigating severe global warming [1]. To progress technical implementation of CCS, the past decade and a half have seen the discovery and development of a wide portfolio of carbon capture technologies [2][3][4][5], many of which are still in early stage of development, i.e. technology readiness level (TRL [6]) = < 4.
For CCS to succeed as a CO 2 mitigation strategy, it is necessary to especially advance the technologies that are most promising in terms of technical, economic and environmental performance. To reach a commercially ready portfolio of the most promising CCS technologies on time, and in an efficient way, targeted technology selection and investment are required [7]. This selection requires performance assessment of the different technology alternatives, by analysing the performance of the carbon capture technology in an integrated system (CO 2 source, carbon capture installation, transport, and storage) already during the early stages of development. These analyses will also point out key improvement options, thereby supporting RD & D.
The current best practice for techno-economic analysis of (near) commercial energy technologies involves well established methods. Typically, the technical performance is evaluated using process models [8][9][10], and the economic performance is estimated based on detailed equipment lists and mass and energy balances that are derived from the process model [11][12][13][14]. As we will show in this paper however, there are challenges to use these best practices for technologies down the TRL ladder, and shortcuts or simplified approaches may be needed to produce the desired performance metrics. Inherently, uncertainties in the analyses will be substantial and need specific attention, especially because they can point out hotspots for technology improvement.
In this context, the aim of this paper is threefold. First, to identify the key challenges when using state of the art methods for developing techno-economic assessments of very early stage CCS technologies (TRL [3][4]. Second, to identify and develop shortcuts and/or other methods that can be used to deal with these challenges, and third, to extract lessons for performing meaningful techno-economic assessments for very early stage technologies. In this way the paper advances the stateof-the-art of very early stage ex-ante technology assessment. Its novelty is especially in presenting hybrid approaches to techno-economic performance assessment, approaches that have not been published before. To this end, this work will consider the performance assessment of an emerging carbon capture technology that is at low TRL (3−4). In this paper, we will use electric swing adsorption (ESA) as case study. ESA is proposed as a postcombustion CO 2 capture technology analogous to temperature swing adsorption [15][16][17][18]. The differentiating aspect of ESA is that the heat required for the temperature swing is provided by electricity using the Joule effect, instead of by low pressure steam, which is used in TSA. The ESA process in this work is analysed in the context of CO 2 capture from a natural gas fired power plant, as suggested by the developers of the technology [16].
Although this work focuses on power plants with CO 2 capture technologies, its lessons are equally applicable to other emerging energy technologies, especially the ones down the TRL ladder. Examples include solar fuels, biofuels and bioproducts, and concentrating solar power.

General approach
This work systematically investigated the steps that were undertaken in the very early stage techno-economic assessment of NGCC with ESA. First, a generic literature scan was performed to select an emerging technology that could function as case study (ESA, see introduction). Second, an in depth literature study was conducted to investigate the technology characteristics and establish the case study scope (Sections 3.1-3.2). The literature study specifically investigated the availability of data for modelling purposes, and the development potential of the technology. The case study scoping also included selection of a reference technology for comparison of the techno-economic performance: also for low TRL technologies a benchmark is relevant to put their techno-economic performance into perspective. Third, available methods for techno-economic assessment of CO 2 capture technologies were reviewed, and suitable methods were introduced for this case study, notably hybrid approaches (Section 4). Fourth, the performance of the low TRL technology and the benchmark technology were modelled and the results, including sensitivities, were analysed (Section 5). Last, the identified challenges and lessons for this kind of (low TRL) technology analysis were summarized in the conclusions (Section 6).

Technology description
The aim of the literature review was to identify the technology characteristics and to set the scope for the case study. This amongst others meant investigating the availability and strength of data for modelling purposes, which was required for model selection (Section 4). In this respect, and given the low TRL, there were two options: the available knowledge base was sufficient to undertake a classic technoeconomic analysis that included integration of the technology with the power plant (e.g. like [11,13,19,20]). This would require as a minimum A) the possibility to make an estimate of mass and energy balances and B) a rough understanding of equipment and size. Or, a classic technoeconomic assessment was not possible due to data limitations, and a preliminary assessment method was required, for instance as described in [21,22]. Questions that needed to be answered included "are estimates available of CO 2 separation efficiency and purity or can these be produced?", "Are estimates of separation energy requirement available or can these be produced?", and "is a process design available or can this be produced?"

Electric swing adsorption
The investigated CO 2 capture technology (ESA) is a specific type of temperature swing adsorption (TSA) technology. It consists of trains of parallel columns with solid sorbents that are alternatively cooled and heated to respectively adsorb and desorb the CO 2 ( Fig. 1). In a standard TSA process, the heat is provided by steam that is extracted from the power plant steam cycle [23]. However, this poses engineering issues for large scale application due to slow heat transfer of steam to adsorbent, resulting in long cycle times and large sorbent requirements [15,16,24,25]. ESA is meant to address this challenge, using electricity to heat the sorbents through the Joule effect [15][16][17][18].
The cycle of a basic ESA or TSA process consists of four steps: (1) feed, (2) heating, (3) desorption, and (4) purge [15,26]. But more advanced cycles have been proposed that may benefit from step and heat integration [18,25]. With respect to separation efficiency and energy requirement, Grande et al. [16] estimated a regeneration duty of 2.04 GJ e /t CO 2 for an ESA process using zeolite 13X sorbent and a five step cycle. They reported approximately 80% CO 2 recovery and 80% CO 2 purity, a significant achievement, but not yet fully representative of the performance of a commercial CO 2 capture process, where 90% capture at > 95% purity is required. For this 5-step cycle, also column sizes were reported which may form the basis of capital cost estimations.
The above concise technology description shows that the ESA technology is developed enough to simulate mass and energy balances, and that preliminary PFD's and equipment sizes are available. This means that classic techno-economic analysis could be undertaken. However, because the technology is at TRL 3 (or proof of concept stage), performance results will change as it moves towards commercialisation. To provide a fair representation of potential, future, performance, it is necessary to assess the process taking into account potential performance advances over its development cycle. This requires knowledge on which advances are likely to take place and how these will influence the process.
The development potential of T/ESA will likely come from more efficient sorbents that rely on chemical bonding with CO 2 [2,25,27]. Chemical sorbents may have higher working capacities than typically used physical sorbents, such as zeolite 13X and zeolite 5 A [27][28][29]. As an additional benefit, chemical sorbents often perform better in humid environments such as flue gas streams, whereas physical sorbents quickly deactivate in the presence of water [23,25,30], and thus require an additional drying unit upstream the CO 2 capture step. An example of a chemical adsorbent that has been well characterised, and that has favourable properties in the presence of water is amine grafted pore expanded silica (TRI-PE-MCM-41) [28,31,32], but many others have been developed (e.g. [27]). TRI-PE-MCM-41 has shown a high CO 2 capacity of up to 2.94 mol/kg at a CO 2 partial pressure of 5 kPa, and in the presence of water [28], which represent typical conditions of NGCC flue gas streams. Other process improvements may come from optimising the steps of the adsorption-desorption cycle.

Natural gas combined cycle
Natural gas combined cycles (NGCC) are commercial technologies. The selected configuration of the NGCC was based on the EBTF framework [14] and includes two generic F-class gas turbines (GT's), two heat recovery steam generation (HRSG) units, and one train of HP, IP, and LP steam turbines (Fig. 1). The power plant size is approximately 830 MW. To follow the current advancements in NGCC with CCS development, the NGCC was equipped with exhaust gas recycling (EGR), which reduces the flue gas volume and increases the flue gas CO 2 content, thereby decreasing capital costs of the CO 2 capture plant and simplifying CO 2 separation [33]. When standard gas turbines are used, their efficiency will decrease slightly because the EGR changes the composition of the air inlet. In this study, we however assumed the same efficiency values as suggested by EBTF, similar to [34], assuming GT designs can be modified to render the same efficiency with EGR.

CO 2 compression
CO 2 compression was considered a commercial technology because it is used in numerous US enhanced oil recovery projects. This study uses a 5-stage CO 2 compressor and a liquid CO 2 pump (Fig. 1), increasing the pressure to 110 bar. During compression the CO 2 stream is dried to an H 2 O content of 200 ppm v using knock out vessels and a drying unit [8].

Reference technology
The general purpose of a feasibility study is to investigate whether or not a novel technology performs better, or has advantageous characteristics, when compared to an incumbent technology. In the case of CO 2 capture, postcombustion capture with standard MEA technology is often seen as the incumbent CO 2 capture technology, and is typically used as reference [8,10,11,13,34]. Recently constructed CO 2 capture plants like Boundary Dam and Petra Nova use more advanced postcombustion solvent technologies than MEA, (Cansolv and MHI KS1,), that show improved energetic performance over MEA. Therefore, for a very novel technology like ESA to be techno-economically "feasible", it should at least outperform MEA technology in terms of parasitic impact on the NGCC and cost impact. Because Cansolv and KS1 are proprietary blends, and their exact formulation is unknown, this paper uses MEA as a reference technology, acknowledging that any emerging technology should show drastically improved performance with respect to MEA to be able to compete with commercially used solvent systems.

System scope
Because the ESA technology was developed far enough to undertake classic techno-economic assessment, the scoping was similar to that of advanced technologies. It included setting the location (NW-Europe), currency (€), and temporal (2014), boundary conditions such as ambient characteristics and cooling water specifications (Appendix A), and the units/plants that are included in the system. With respect to the latter, note that to assess the technical performance of any CO 2 capture technology, it is essential to analyse it in the context of the other units it is connected to, i.e. the power plant (or industrial CO 2 source) and the compression plant (further discussed in Section 5.1). For economic  performance analysis, also CO 2 transport and storage should preferably be included [12].

Technical performance assessment
The main purpose of the technical performance analysis of any CO 2 capture technology is to find the energy required for separation of the CO 2 from the power plant flue gas stream, given a specified CO 2 yield and purity. For solid adsorbents, this comes down to determining the energy requirement for adsorbent regeneration. The metric for this is the specific regeneration duty (SRD: energy used per tonne of CO 2 captured), analogous to the metric used for postcombustion solvent technology.
Selecting a suitable modelling method to estimate the technical performance is a problem that many process modellers face. Typically, a choice can be made between rigorous, shortcut, or simplified process models [9,37]. Rigorous methods have the potential to provide more detailed and/or accurate results. However, time, knowledge base, and sometimes skills may be limited, therefore limiting model selection to more simplified methods. To aid the selection process, an attribute complexity matrix and/or pedigree analysis of the knowledge base can be used [9,37]. These methods screen the (scientific) knowledge base of a technology on criteria such as empirical basis (which knowledge is available, from which kind of, and from how many sources was it derived); theoretical understanding (how strong is the theory on this technology, and the level of consensus between scientists); methodological rigour (how was the knowledge generated), and validation process (was the data validated and how sound was the validation exercise). The gained understanding of the technology's knowledge base can then be used to select modelling methods that fit the available knowledge and the purpose of the modelling study. For example, if the knowledge base strength is low to medium, one may need to select simplified modelling methods.

Electric swing adsorption unit
For solid sorbents, the existing best practice is to perform rigorous analysis of the adsorption-desorption cycle, using dynamic modelling methods. This was shown amongst others by Grande et al. [16,24,38] for ESA and is equivalent to cyclic modelling of PSA and TSA systems (e.g. [39,40]). This rigorous method considers the time aspect of adsorption cycles as well as mass transfer limitations into the sorbent pores. Models like these consist of a set of partial differential equations that are solved numerically. The construction of such a rigorous cyclic adsorption model is, however, a laborious processmodel development times of over a year are not uncommonand requires a great deal of skill from the modeller.
The second option is to use a short-cut method analogous to the one developed by Joss et al. for TSA [26]. The differential equations in this type of model can be solved analytically by treating the columns as isothermal rather than adiabatic, thereby neglecting temperature fronts. Also, it excludes mass transfer limitations, but rather calculates the CO 2 adsorption based on adsorption equilibrium. Although this kind of model is easier to construct, and simulation time can be greatly reduced compared to rigorous cyclic models, model development is still time and resource intensive, which can be an issue for rapid technology portfolio screening.
The third option is to use a simplified method. This method estimates the SRD of an adsorption process using simple (non-differential) equations to calculate the adsorption heat, sensible heat, gas (adsorbate) heating, and water vaporization, and sums these values to a total SRD estimate [41]. This method can be suited to estimate initial performance of continuous adsorption processes (such as found in sorbent systems using two fluidized beds), but tends to overestimate the SRD in a cyclic process, because it fails to include interactions between the steps in a cycle (e.g. sensible heat generated during the adsorption step that already partially heats the column for the desorption step).
In this specific case, the NGCC with ESA assessment was part of a larger European research project where ten CCUS technologies were screened in four years, something not uncommon in energy system analysis. This, however, limited the available modelling/assessment time. Therefore, in this paper we used a hybrid option to project the performance of the ESA technology: we based the SRD estimate of the ESA process on existing, preliminary, rigorous modelling results (the aforementioned zeolite 13X modelling results by Grande et al. [16]). Then, we constructed a performance estimate of a future advanced/ commercial ESA process, extrapolating Grande's results to a state where 90% CO 2 at 96% purity is captured using an advanced amine grafted sorbent (TRI-PE-MCM-41, Section 3.1.1). This means that preliminary results of a rigorous modelling method were used, and were adapted using the analogy of the simplified method.
The chosen technical assessment strategy required adjusting Grande's results in two ways. First, an adjustment was needed to increase the CO 2 recovery and purity to respectively 90% and 96%. To this end we assumed that these levels could be reached by two system improvements: 1) increasing the flue gas CO 2 concentration to 6.4% vol by means of the EGR (in the original study by Grande et al. this was 3.5% vol leading to a higher required enrichment factor), and 2) using the aforementioned TRI-PE-MCM-41 sorbent, which is much more selective towards CO 2 than the original zeolite sorbent. This sorbent has the additional advantage that it is able to operate effectively in humid flue gas, contrary to zeolite 13X [28].
The second adjustment involved the value of the SRD. We projected the SRD of the future ESA process on the value found by Grande et al. [16], accounting for differences in sorbent properties (Table 1) and process design (Table 2), and including water vaporization. This was done in three steps (Fig. 2), using the analogy of the SRD of a continuous (non-cyclic) adsorbent process [41]: Where Q des is the total heat requirement (kJ/mol), ΔH a is the enthalpy of adsorption (kJ/mol), m is the sorbent mass (kg), C p is the sorbent heat capacity (kJ/kgK), T ads and T des are the respective adsorption and desorption temperature, and Q H2O is the heat of water vaporization.
Step 1 encompassed approximating the division between regeneration energy used for CO 2 desorption and for sensible heat. This division was required for step 3. The results from Grande et al., that  [16]. 2 same amount of binder assumed as in the zeolite monolith. 3 [28]. The highest measured loading capacity was used to represent an advanced ESA process. 4 [25]: working capacities of 0.18 to 0.26 (% max capacity) were reported for an amine-impregnated polymeric resin. A high value was selected to represent an advanced ESA process. This working capacity was validated with TRI-PE-MCM-41 adsorption-desorption cycles presented in Serna-Guerrero et al. [42]. 5 An average value of 70 kJ/mol was used as reported by [32]. This value falls with within the range of 43-92 kJ/mol that was measured/calculated for similar amine impregnated sorbents [2,41,43]. 6 [43]. This value is relatively conservative compared to heat capacity of activated carbon and SiO 2 (both∼700 kJ/kg K), which constitute about 65% wt of the solid sorbent.
were used as a basis specified a total SRD of 2.04 GJ e /t CO 2 captured, of which 0.13 GJ e /t for heating of the adsorbate (gas), and the remaining 1.91 GJ e /t for heating of the adsorbent. This means that the 1.91 GJ e /t includes adsorption enthalpy and sensible heat. A division between the twoexpressed as their contributions (%) to total heat -was estimated using the analogy of the continuous process (Eq. (1)), assuming the division of sorbent heating in the cyclic process was similar to the continuous process (Fig. 2b).
Step 2. Second, the heat of water vaporization was added. Zhang et al. [41], estimated this was 0.38 GJ/t for a continuous adsorption process. Again, this value was expressed as the contribution of the total heat of adsorption and analogously added to the cyclic SRD (Fig. 2c).
Step 3. Third, the SRD value of the future ESA process with the TRI-PE-MCM-41 sorbent was projected based on the SRD terms of the current, zeolite sorbent process (Fig. 2d), incorporating the differences in sorbent properties and process design following Eqs. (2)-(5): Where Q des,TRI and Q des,ZEO are the enthalpies of ad/desorption of the advanced sorbent and the zeolite sorbent, Q sens,TRI and Q sens,ZEO are the sensible heat requirements of the advanced sorbent and the zeolite sorbent, Q gas,TRI and Q gas,ZEO are the adsorbate heating requirements of the advanced sorbent and the zeolite sorbent, Q H2O,TRI and Q H2O,ZEO are the water vaporization heat requirements of the advanced sorbent and the zeolite sorbent, and F 1 , F 2 , F 3 , F 4 , and F 5 , are factors representing the differences in sorbent properties and process design.
Finally, the total SRD (GJ electricity per tonne of CO 2 captured) was calculated by summing the individual heat components of the ESA process:

NGCC with exhaust gas recycling
The NGCC model was specified in Aspen Plus V8.4., following the EBTF design specifications [14]. Because NGCC technology is commercial, and many such models exist, this was a straightforward activity, contrary to ESA modelling. An exhaust gas recycle (EGR) ratio of 35% was assumed. Flame stability in the GT combustion chamber may allow higher recycle ratios, but tests suggest that above 35% recycle excessive NO x and CO start to form which is undesired from an environmental perspective [44]. The EGR includes a direct contact cooler (DCC) to cool the flue gas to 25°C before re-entering the combustion chamber.

CO 2 compression unit
Like the NGCC, the compression unit was modelled with Aspen plus V8.4., using the same design as in [8,9].

Benchmark MEA absorption unit
The benchmark MEA absorption unit was modelled using a ratebased model that was specified in Aspen Plus V8.4. The model used kinetic reactions for the bicarbonate (Eq. (7)) and MEA carbamate (Eq. (8)) formation following Kvamsdal and Rochelle [45], applying their reported kinetic parameters for the Arrhenius equation. The reaction equilibria of other reactions were calculated from the Gibbs free energy. The flowsheet included a standard absorber-stripper configuration with water washes on top of absorber and stripper [9]. Advanced process configurations like lean vapour compression and split flows were excluded from this study. The MEA process was earlier reported in [9].

Economic performance assessment
Economic performance assessment of CCS technologies typically aims to find estimates of capital and operational costs, and uses these to calculate the Levelised Cost of Electricity (LCOE) and the Cost of CO 2 Avoided (CCA) ( [12,13], see Appendix B for equations).
The main challenge for economic analysis of early stage technologies is to find the investment costs of a potential N th of a kind (NOAK) plant, and to estimate the technology specific operational costs. Estimating the generic operational costs and selecting financial parameters is more straightforward because often standard values are used. Therefore, the description below focusses mainly on methodological aspects of capital cost estimation and technology specific operational costs.  [12,[46][47][48].
Exponent methods can be used when a (reliable) cost estimate of the same (or similar) technology already exists. The exponent method uses a cost estimate of a reference study and scales this to the size of the plant in a new study: Where C and C ref are the capital costs of the equipment/plant in respectively the new study and the reference study; Q is the equipment capacity in the new study, Q ref is the equipment capacity in the reference study, and n is the scaling exponent.
A typical characteristic of very low TRL technologies is that reference cost estimates are unavailable. In that case, (only) a bottom up method can be used, which calculates the capital costs based on an equipment list of the studied process.

Direct methods versus indirect methods for
Nth of a kind cost estimations. Bottom up capital costing methods rely on a detailed equipment list and purchased equipment cost estimates, which are then multiplied with factors for installation, engineering, procurement and contracting, and contingencies [12,49]. This direct method of estimating the cost of an NOAK plant is suitable for technologies that are close(r) to commercialisation, and of which the design is welldefined. Early stage technologies (TRL 3, 4, sometimes 5) often lack this level of design detail, and are usually described by simplified process flow diagrams and basic equipment lists. The estimated costs of these technologies typically escalate in the period between the lab stage and the first large demonstration plants (first of a kind (FOAK); TRL 7, 8) (Fig. 3) [50,51]: during upscaling from lab to demonstration the design is further detailed and unforeseen technological issues are uncovered which need additional design solutions, typically increasing the costs of the technology. After this point, the costs start to fall as more commercial plants are built (TRL 9) and technological learning commences (Fig. 3). The Nth of a kind plant is reached when cost decline starts to level out, i.e. when a significant part of the learning has taken place.
The direct bottom up method directly calculates the costs of the NOAK plant but disregards the cost curve that very early stage technologies follow and the effect this may have on NOAK costs. Instead, an indirect method or hybrid method may be more suitable. Rubin [50,51] proposed such an indirect method including the following three steps [51]: (1) estimate the preliminary engineering, procurement and contracting (EPC) costs bottom up, based on a (simplified) equipment list of the novel technology, (2) find the FOAK total plant cost (TPC) costs by adding appropriate process and project contingencies, and (3) find the NOAK total plant cost using learning curves. This novel hybrid method provides an elegant solution to low TRL technology cost estimation, because it includes the large cost escalations to the first built plants, but also because it allows more detailed evaluation of uncertainties during the different stages of the development curve.

4.2.1.3.
Capture unit capital costs. The capital costs of the ESA (and to a lesser extent of the EGR) equipment could not be calculated using an exponent method because a reference cost for thisor a similartechnology was lacking in available literature. Therefore, we chose to estimate the NOAK capital costs bottom up with the indirect method, using a preliminary equipment list of the ESA process. As a reference, the NOAK TPC costs of the CO 2 capture equipment were also estimated using the direct bottom up method, to highlight the differences and similarities in outputs between the two approaches.
The three steps of the indirect method were taken as follows: Step 1. Bottom up estimation of the EPC costs based on a simplified process flow diagram and equipment list. At the heart of the ESA process are the adsorption columns, flue gas and CO 2 fans, and some gas storage tanks (Fig. 1). The ESA column sizes were estimated based on the process design presented by Grande et al. [16]. The height of the adsorption columns was kept equal to Grande's study, while the column diameters were varied to match the flue gas flow in this work, taking into account the working capacity and density of TRI-PE-MCM-41 (see Table 3). This ensured the gas speed in the columns to remain the same as in Grande's work. The size of the other ESA equipment (fans, storage vessels, and valves) was defined based on the ESA mass balance and engineering rules of thumb.
Based on the simplified equipment list, the purchased equipment costs were calculated using the Aspen capital cost estimator (V8.4). The cost of each equipment was then multiplied with an individual installation and EPC factor that were retrieved from an in-house database [13]. This led to an engineering, procurement, and contracting (EPC) cost estimate. The same approach was applied to the EGR equipment.
Step 2. The second step included estimation of the total plant cost of the FOAK plant, based on the EPC cost estimated in step 1 (Fig. 3). To this end, the EPC costs were escalated with process and project contingencies. Guidelines exist for the amount of process and project contingencies to apply, based on technology readiness and type of cost estimate, respectively ( [12], following EPRI and AACE) as displayed in Table 4Table 5. Note that both tables give ranges for the process and project contingencies, rather than single values, representing the uncertainty in cost estimate escalations. This work used these ranges as uncertainty margins for the FOAK cost estimates. For the ESA (+EGR) technology, process contingencies from 30 to 70% were applied with a nominal value of 50%, representing its status as a "concept with bench scale data" (TRL 3/4). For the MEA (+EGR) technology, process contingencies from 5 to 20 were applied with a nominal value of 12,5%, representing its "full-sized modules have been operated" status. We considered both cost estimates AACE class 4 estimates [46] and hence 30-50% project contingencies were applied with a nominal value of 40% (see also Section 4.3 sensitivity analyses).
Step 3. The NOAK costs were calculated using a single factor learning curve [53]: Where C represents the capital costs (total plant cost, TPC), N is the number of installed plants, and b is the learning rate coefficient, calculated from: Where LR is the learning rate. This single factor learning curve 1 combines learning by doing, learning by searching (continued RD & D in the commercial stage of technology deployment), and scale factors [7].
The three determining parameters in Eqs. (10) and (11), other than the costs of the FOAK plant, are the values of N FOAK , N NOAK , and LR. It is common practice in literature on learning curves to use value for the installed capacity rather than installed number of plants, N [53]. However, Greig et al. [54] provided a rather appealing and useful definition for FOAK, early mover, and NOAK CCS plants, based on the number of installed plants, rather than installed capacity. Their definition of N FOAK and N NOAK includes the number of installed plants worldwide and defines FOAK as < 10 demonstration and/or commercial scale plants wold-wide, early movers as > 10 -< 20 commercial scale plants world-wide, and NOAK as > 20 commercial scale plants world-wide. We adopted Greig's definition in this study: for N FOAK the range of 1-10 built plants was used with a nominal value of 5 (see also Section 4.3 sensitivity analyses). For N NOAK the value of 20 built plants was used.
Because learning rates are based on the costs of built and operating plants, they are still unknown for CCS technologies. Instead, learning rates from analogous technologies can be used as a proxy. Rubin [51] and Van den Broek et al. [53] proposed to use the learning rate from flue gas desulphurisation (LR = 0.11) as a proxy for wet CO 2 capture systems, because of process similarities. Table 6 presents a selection of learning rates found for the process industry, including a range of gas processing technologies. The table shows that most gas processing technologies have learning rates between 0.10 and 0.14, but that also outliers and inconsistencies exist: for example, Rubin et al. [55] report a learning rate of 0.27 for steam methane reformers, while Schoots et al. [56] report a value of 0.11 ± 0.06 for the same technology.
For MEA technology, we adopted a learning rate of 0.11 as proposed by Rubin [55] and Van den Broek [53]. For ESA technology, it was more challenging to decide on an appropriate learning rate. As said, an ESA process most resembles a pressure swing adsorption process as used in e.g. hydrogen production. Learning rates for hydrogen production have been reported (   Process is used commercially 0-10 1 Also so-called multi factor learning curves exist where the single factor is disaggregated to separately treat learning by doing, learning by searching, and cost reductions by economies of scale [7]. However, data scarcity often inhibits this division in separate cost reduction drivers, and even when data is available, there may be significant overlap between learning by doing, searching, and scale, leading to question the validity of such a division [7]. methane reforming, water gas shift, and PSA, rather than the PSA alone. A second complication was that the reported hydrogen production learning rates vary significantly (0.11 and 0.27). A third consideration in the learning rate decision was that all gas processing technologies in Table 6 (with exception of the SMR value reported by [55]) are in the order of 0.10 to 0.14, while only the liquids processing technologies show higher learning rates (ammonia: 0.29; bulk polymers: 0.18-0.37). Given these three considerations, for the ESA technology we also adopted a learning rate with a nominal value of 0.11, and applied a minimum of 0.10 and a maximum of 0.14 in the sensitivity analyses (see Section 4.3). For the NGCC with CCS cases (ESA and MEA), the different equipment of the NGCC plant -GT, HRSG, ST, heat rejection -were scaled to their respective required sizes using Eq. (9). Based on DOE/NETL guidelines [59], an exponent of 0.8 was used for the gas turbines and steam turbines, and exponent of 0.7 was used for the HRSG and cooling water sections. After scaling, the costs were escalated from 2008€ to 2014€ using the Eurozone Harmonized Index of Consumer Prices (HICP).

CO 2 compression capital costs.
Finally, the TPC of the CO 2 compression plants was calculated bottom-up using the direct estimation method, because this technology is considered mature and has already gone through the learning curve shown in Fig. 3 [53].

Operations and maintenance costs
Operational costs of the CO 2 capture, EGR, and CO 2 compression equipment were estimated including labour (1 extra shift and 1 extra process technologist in comparison to NGCC w/o CCS), maintenance costs, and variable costs (e.g. process water) ( Table 7). Calculation of sorbent costs and replacement frequency are provided in Appendix C.
For the power plant, labour costs were taken from the EBTF study [14] and escalated to 2014€ using the HICP. Maintenance and insurance costs were calculated as a percentage of TPC ( Table 7). The costs of water make-up and fuel were calculated based on their respective feed flows and multiplied with unit costs (Table 7). To complete the O & M cost estimate, transport and storage costs were added based on the ZEP reports [67,68] (Table 7), similar to [13].

Sensitivity analyses
The previous sections highlighted some of the methodological choices, trade-offs, and simplifications that need to be made when analysing low TRL CO 2 capture technologies. A low technology development stage inherently leads to many uncertainties in technology performance, and because the use of simplifications is unavoidable, uncertainties in performance results will increase. Study of the available literature, general engineering knowledge, and attention to the particular characteristics of the technology can provide insights into the range of many uncertainties. I.e., an informed, educated value for input uncertainties can be derived, and hence their effect on performance results can be analysed in a sensitivity analysis. This may lead to a preliminary, but substantiated understanding, of potential performance ranges.
The analysis in this case study focused especially on ESA specific parameters but included sensitivity analysis of some generic parameters too, to provide a reference for the ESA specific uncertainties. The inputs to the technical sensitivity analyses are provided in Table 8.
To investigate the uncertainties in ESA capital cost results, the values of EPC cost, N FOAK , LR, process contingencies, and project contingencies where varied. Also, a scenario with maximum EPC costs and contingencies and minimum learning, as well as a scenario with minimum EPC costs and contingencies and maximum learning were calculated (see Section 4.2.1, EPC, FOAK, LR, and contingencies varied simultaneously). This led to a capital cost range as an addition to the point estimate. The LCOE sensitivity analysis also included variation of net system efficiencya result of the technical sensitivity analysis -and sorbent replacement frequency, as well as more generic economic parameters (Table 9).

Results and discussion
This section presents the results of this work. Section 5.1 presents the technical performance of the ESA case study, after which Section 5.2 deals with the methodological insights on technical performance  Transport (180 km offshore) €/tonne 6 j Storage (offshore depleted oil/gas field) €/tonne 10 k a 2014 average industrial consumer price of natural gas in The Netherlands [60]. b [14]. c [61]. d [62]. e Wage information retrieved from the Norwegian Confederation of Trade Unions [63].
1 additional operator assumed in 6 shift rotation. f Wage information retrieved from the Confederation of Norwegian Enterprises [64].
projection of very early stage technologies. Section 5.3 then shows the economic results of the ESA case and draws conclusions on its economic feasibility, and on very early stage cost estimation in general.

Technical performance results
Following the approach outlined in Section 4.1.1 and using the factors presented in Table 10, the different contributions to the ESA regeneration duty were calculated. The SRD estimate of the advanced ESA process amounts to 1.9 gigajoule electricity per tonne of CO 2 captured (Table 11). In first instance, this seems low compared to the regeneration duty of the benchmark MEA process (3.6 GJ steam/t CO 2 , Table 12). However, because ESA uses electricity instead of steam as energy input for regeneration, the ESA process has a considerable impact on the net output of the NGCC (see Table 12 and Fig. 4). In fact, the net output and net efficiency of the NGCC + ESA system are substantially lower than those of the NGCC + MEA system. This is  [16]. b The value of 1.9 GJe/t was calculated using a more advanced adsorption cycle [24]. c TRI max adsorpbent loadings were based on the experimental range reported in [28]. d Heat capacity of pure activated carbon assumed as minimum value. e The minimum and maximum values of TRI working capacity were based on [25].
This source reports working capacities for amine modified polymeric resins, which were used as a proxy for amine grafted pore expanded silica. f Minimum and maximum values taken from the range reported in [32]. Values depend amongst others on sorbent loading. g Temperature assumed well below sorbent degradation temperature. h Assuming complete heat recovery of the water vaporization energy. i Value derived from [30]. j Plus and minus 50% of the nominal value assumed. k Adiabatic working capacity reported in [69]. l Fifty percent higher than nominal value assumed. m Heat capacity of amine impregnated solid sorbent used in [41]. n Max regeneration temperature due to sorbent degradation restrictions. o No heat recovery of water vaporization energy. p Minimum and maximum reported values for GT efficiency in [14]. q Same value as nominal value, other sources often report higher values. r Educated guess, generator efficiency seems well established figure. s Average ambient temperatures as they are estimated to be found within Europe. t Educated guess. u Steam turbine efficiencies and steam pressures as reported in [49].  [47,61], a maximum value of 6 percent was used to illustrate the high uncertainty of a novel process. g A maximum value of 4 replacements per annum was used assuming a case where max deactivation is 15% instead of 30% (Section 4.2.2), a minimum of 1 replacement per 2 years was used to represent a case where sorbent is further developed to deactivate slower, or where sorbent can be regenerated. This value is more in line with industrial practice where sorbent beds are replaced less frequently. Based on ZEP [67,68]. h In techno-economic studies the economic life time is often assumed the same as the technical life time, 25 years in case of an NGCC plant. In reality, operators depreciate their assets faster, and an economic life time of less than 25 years is more realistic. Hence the nominal and maximum value are equal (25 years), and the minimum value is taken as half of that. i Based on Dutch bureau of statistics [70]. j Based on [14,49]. k See Table 4. l See Table 5. m See Section 4.2.1

Table 10
Factors used in SRD projection of advanced ESA process (using Equation (2) -Equation (5)). explained by the use of high quality electricity instead of low quality LP steam for regeneration (LP steam is typically transformed to electricity in steam turbines with an efficiency of 20%-30%). The picture slightly improves if it is assumed that part of the regeneration duty of the ESA system could be supplied by steam instead of electricity. In how far this would be possible is subject to further investigation. The NGCC + ESA net efficiency increases from 45.9% to 47.5% under an assumption of 25% regeneration with steam (NGCC ESA 25/75), and to 49.3% under an assumption of 50% regeneration with steam (NGCC ESA 50/50). This is however still 1 %-point lower than the net efficiency of and NGCC with MEA configuration. From a technology perspective, these results thus show that NGCC with ESA capture does not have an energetic advantage over NGCC with MEA. Fig. 5 presents the sensitivity analysis of SRD and net system efficiency of the NGCC ESA 50/50 system. The SRD sensitivity analysis shows that the base SRD value is on the low end and that it is likely to be higher than lower. The mainly higher SRD values lead to mainly lower values of net system efficiency: this indicator could be reduced with over 3 %-point compared to the base case. Also, the ESA input parameters generally have a higher impact than the NGCC parameters, stressing the importance of the SRD estimate.

Methodological insights on future technical performance projection
The results show that it is possible to make a rough but sound technical performance projection of a future advanced ESA process, despite its low TRL. This is mostly due to the availability of results from rigorous modelling (Section 4.1.1) of a preliminary ESA process, and the availability of lab data for advanced solid adsorbents. Two general methodological insights can be extracted from this. First, at least some level of rigorous modelling work -or laboratory SRD measurementsare required to produce meaningful technical performance estimates of CCS technologies. And second, basic lab data of advanced sorbents need to be available to project the future performance. These requirements are unlikely to be met at TRL < 3. It is therefore unlikely that performance estimates of lower TRL technologies will lead to reliable results.
Furthermore, the results showed the importance of system analysis for understanding the performance of capture technologies. In the case of ESA, the capture technology has been reported as very promising due to its low regeneration duty [16]. This conclusion was based on the performance of the capture unit alone. However, results change when the full system is analysed, indicating that this system cannot compete with MEA technology. CCS technology screening thus requires the connection between power plant and capture plant, and is otherwise inconclusive.
Last, the SRD sensitivity analysis showed that the use of simplified models may lead to physically impossible outputs. In this sensitivity analysis the SRD sometimes went below the ideal CO 2 separation energy. This stresses the care that must be taken when projecting technical performance with simple methods.

Economic performance
At the interface of technical and economic evaluation is the equipment sizing and costing. Table 13 shows that for the studied ESA design a total of 16 columns were required, divided over 4 trains. The total amount of adsorbent required to fill these columns equalled 5936 tonnes, 371 tonnes per column. Table 13 also gives the costs of the columns and of the other equipment.
Based on the estimated equipment costs, Fig. 6 shows the progression of capital costs from currently estimated EPC, to FOAK TPC, and finally NOAK TPC. For the ESA plant, the addition of process and project contingencies escalates the EPC from 305 M€ to first-of-a-kind TPC of 641 M€. Technological learning then reduces the FOAK TPC to an NOAK TPC of 508 M€. This is 19% higher than when the TPC was estimated with the direct costing method. For the MEA plant, the indirect TPC estimate is actually 11% lower. Fig. 6 also shows the uncertainty ranges of the direct and indirect capital cost estimates. The ranges applied to the direct TPC estimates are simply the -30%/+50% accuracy which is typical for an AACE class 4 estimate. Because the AACE does not specify a typical accuracy for EPC costs (before the addition of contingencies), the same range was used to display the margin on EPC. The ranges of the indirect NOAK TPC estimates doubled in size compared to the EPC margins, because the uncertainty in contingencies and learning was added (outside uncertainty ranges in Fig. 6). If no uncertainty on the EPC was assumed, the NOAK TPC margins were around ± 40% (inside uncertainty ranges in Fig. 6).
Finally, Fig. 6 shows that the ESA capital costs are likely to be higher than the MEA capital costs. This means that also from an investment cost perspective ESA is likely the lesser option when compared to postcombustion solvent technology.
From a methodological point of view, the capital cost analysis shows Table 11 Breakdown of specific regeneration duty calculated using Eqs. (2)-(5) and the factors in Table 10.  Fig. 7 highlights that especially the values of the EPC estimate are relevant to the final value of TPC. This means that the indirect method, like the direct method, relies heavily on accurate bottom up estimation of equipment costs and installation, and the other costs included in the EPC estimate. The values used for N FOAK and contingencies have less impact on the value of NOAK TPC. These parameters have an impact of 7-20%. The learning rate only has a minor impact on the TPC result.    Last, Fig. 8 shows that the uncertainty ranges of total plant cost (extracted from Fig. 6) and net efficiency have a significant impact on the levelised cost of electricity. They rank as second and fifth most influential parameters in the LCOE sensitivity analysis, but are outranked by fuel price (and net efficiency also by two general financial parameters). Note however, that the total plant cost ranges are a combination of maximum/minimum EPC, maximum/minimum contingencies and minimum/maximum learning (i.e. EPC, FOAK value, process and project contingencies, and learning rate were varied together). The individual influence on LCOE of these input parameters is much smaller (Fig. 8); they all fall in the lower half of the tornado diagram. So where these separate indicators have a high influence on the capital cost estimate, they have less impact on the LCOE estimate. Operational cost uncertainties like maintenance and labour costs are among the less influential LCOE input parameters.

Lessons and conclusions
This work aimed to explore the key challenges when using state-ofthe-art methods to analyse the techno-economic feasibility of a very early stage (TRL 3, 4) CO 2 capture technology; to develop other methods that could help overcome these challenges, notably novel hybrid approaches; and to draw lessons on very early stage technoeconomic performance assessment, lessons that can also be used for the assessment of other emerging energy technologies. To this end, a case study of CO 2 capture with electric swing adsorption was used and its current (lab stage/model) performance was projected to a future, advanced, commercial ESA process.
As expected, the key challenges lie in constructing performance estimates that represent a commercial process, while the process itself is in very early stage of development. At low TRL, it is challenging to construct a conclusive rigorous process model due to data, time, or resource limitations. This makes it hard to estimate conclusive mass and energy balances, however, it often is possible to produce preliminary mass and energy balances. A challenge following from this is to project these preliminary results to a future state of commercial technology. Also, the use of state-of-the-art direct bottom up capital cost estimation may prove challenging, and may lead to too optimistic results, mainly because process design is still preliminary, prohibiting to identify (the size of) every equipment. Using an indirect bottom up costing method was found to provide a solution, but this came with its own difficulties, for instance identifying technology specific learning rates. A last key challenge identified was to estimate technology specific operational cost estimates, such as sorbent replacement costs, although this work and our previous work [13] found that sorbent/solvent replacement costs are among the less important cost items for feasibility studies.
To address these challenges, it proved useful to apply novel hybrid methods to estimate the future performance of ESA technology, based on the currently available knowledge. For instance, we projected the future technical ESA performance based on earlier reported preliminary (rigorous) modelling results, accounting for improvements in sorbent and process design. In a similar fashion, we projected the capital costs of a commercial ESA plant, based on a preliminary equipment list, and using cost escalations and technological learning. This may also be a solution for other low TRL CCS technologies.
Key lessons on very early stage performance assessments and the use of hybrid assessment methods include the following: first, every CCS technology assessment, also at very early stage, is inconclusive unless it is performed in connection with the CO 2 source (power or industrial plant) and the CO 2 compression plant. Failing to include this may lead to invalid conclusions on feasibility. Second, it is unlikely that classic techno-economic studies can be undertaken for CCS technologies at TRL < 3, because sound SRD estimates based on rigorous modelling or lab measurements are lacking, as well as the required basic data to make development projections. Third, the results of very early stage technology screening comeas expectedwith higher uncertainty margins than more advanced technologies. In the case of ESA we found ranges of minus 15% to plus 50% for SRD; and minus 60% to plus 100% for total plant cost. These TPC ranges are double that of commercial technology. Fourth, the direct and indirect capital costing methods employed in this paper provided similar results (max 19% difference). This could be an indication that both methods are valid and reliable, or that they have the same bias. The indirect method provided a more conservative estimate than the direct method for the low TRL technology (ESA), but a more optimistic estimate for the high TRL technology (MEA). From this we tentatively draw that the indirect method may be more suitable for very early stage assessments, but this remains subject to further study. Finally, for CO 2 capture technologies, real plant based learning rates are still lacking, and definitions for "first of a kind" and "N th of a kind plants" are preliminary. However, in the last two years, four CO 2 capture plants have come online [35,36,72,73], with others currently under construction, providing the opportunity to start tracking cost development, and thus filling knowledge gaps on learning rate and definitions of N FOAK and N NOAK .
Based on the analyses presented in this paper, it can be concluded that electric swing adsorption is infeasible of economically capturing CO 2 from power plants. An advanced, future ESA process was projected to have a specific regeneration duty of 1.9 GJ electricity per tonne of CO 2 (min-max: 1.6-2.9 GJ e /t CO 2 ). This may seem low compared to the standard MEA postcombustion SRD of 3.6 GJ/t CO 2 , but because the ESA process uses electricity instead of steam for regeneration, the impact on net power plant efficiency is much larger than that of MEA: 45.9% LHV versus 50.4% LHV for NGCC with ESA and NGCC with MEA respectively. If the ESA regeneration duty can be supplied fifty percent by steam, the NGCC with ESA system efficiency can be cranked up to 49.3% LHV , which is still lower than the NGCC with standard MEA system. As a result, it is unlikely that ESA will be able to compete with commercial, second generation solvent systems. The total plant costs of the N th of a kind ESA plant have a range from 190 to 1050 million € 2014 , with an expected value around 500 million € 2014 . This is higher than the expected total plant costs of the MEA plant: 323 M€ 2014 (min-max 129-629 M€ 2014 ). If a configuration is assumed where the ESA regeneration duty is supplied 50/50 by steam and electricity, and the plant is running at base load, the system has an LCOE of 103 €/MWh, versus 91 €/MWh for the NGCC MEA system. At part load the difference is bigger: 134 €/MWh versus 113 €/MWh for NGCC ESA and NGCC MEA respectively.
Concluding, this paper has shown that also for very early stage technologies (TRL 3-4) it is possible to conduct techno-economic stu-dies that give a sound, first indication of feasibility, for instance by using hybrid analysis methods that combine preliminary rigorous estimates with projections of further development. This will help in the process of selecting and progressing the most feasible (CO 2 capture) technologies given constraints on R & D resources, and will aid to the development of a portfolio of the most promising CO 2 capture technologies, as well as other emerging energy technologies.   (12) Where I i is the investment cost in year i, O & M i are the operations and maintenance costs in year i, r is the real discount rate (%), and E i is the electricity production (MWh) in year i.
Cost of CO 2 avoided: Where LCOE cc is the levelised cost of electricity in the plant with CCS, LCOE ref is the levelised cost of electricity in the plant without CCS, C ref is the CO 2 intensity (t CO 2 /MWh) in the plant without CCS, and C cc is the CO 2 intensity (t CO 2 /MWh) in the plant with CCS.
The economic indicators were calculated for a 25 year lifetime, including 3 years of construction. Because gas-fired combined cycles are used as base loaders as well as cycling plants [74][75][76], the economics were calculated for a high power plant utilisation scenario (85% capacity factor) and for a low power plant utilisation scenario (45% capacity factor) [13]. All costs are reported as constant (real) costs, using a discount rate of 7,5%, similar to [13].
Appendix C. Calculation of sorbent cost and replacement frequency Sorbent costs: the ESA sorbent is a mixture of activated carbon (30%wt), SiO 2 (35%wt) and triamine silane (35%wt) formed into a monolith. The costs of activated carbon are well-known, but triamine silane costs were not available. Therefore, a proxy price of the monolith was estimated using Lichtenberg's method [77,78] (14) Where, P low is the price of the cheapest component, P high is the price of the most expensive component, and P average is the average price of the components included in the analysis. The unit costs of four materials were used as basis for the estimate: activated carbon, zeolite, silica gel, and polymeric resin. These were retrieved from the Aspen capital cost estimator (V8.4) and used to estimate the monolith material costs for this study. Monolith manufacturing costs from the raw materials were excluded. Sensitivity analysis (Sections 4.3,5.3) showed that this simplification was justified within the scope of this study.
Replacement frequency: because TRI-PE-MCM−41 deactivation studies are scarce and/or inconclusive, the deactivation of solid sorbent impregnated with another amine -polyethylene imine (PEI) -was used as a proxy [31]. From reference [31], monolith replacement was estimated to take place twice a year (PEI deactivation rates measured in a lab environment were found to be 1.79% per 300 cycles. Given the cycle duration of 40 min and assuming a max allowable deactivation of 30%, the bed lifetime equals 3360 h, or half a year).