A novel methodological approach for achieving £/MWh cost reduction of CO2 capture and storage (CCS) processes

Carbon capture and storage is widely recognised as essential for the cost effective decarbonisation of the power and industrial sectors. However its capital and operating costs remains a barrier to deployment, with significant reduction in the cost per unit of decarbonised product considered vital. In the context of power generation, this is best expressed in terms of cost per MWh of electricity generated. To achieve a meaningful reduction in the cost of low carbon electricity, capital costs must also be reduced. Thus, this work presents a novel approach for identifying system improvements via a combination of process integration and intensification based on minimisation of thermodynamic losses. Application of this methodology to an oxy-combustion CCS process led to a 3% increase of net efficiency and a 13% reduction of £/MWh of electricity.


Introduction
Anthropogenic carbon dioxide (CO 2 ) emissions from burning fossil fuels are currently recognised as the leading contributor to climate change, with 36.2 Gt being emitted in 2015 [1,2]. However, despite substantial investment in renewable energy, fossil fuels continue to play an integral role in the world's energy landscape [3]. Indeed, coal still plays a major role as a primary energy source [4] and although its global use is declining, some countries are highly reliant on this fuel, so it is expected that coal will keep being relevant in the future.
Carbon capture and storage (CCS) technologies have the potential to reduce these anthropogenic CO 2 emissions as part of a transition to a low carbon energy system [5][6][7]. These technologies are typically divided in three categories: precombustion, post-combustion, and oxy-combustion [6,8], and all are based on the idea of the capture and subsequent storage of CO 2 from the combustion of fossil fuels in either the power or industrial sectors. In all cases, high purity CO 2 has to be compressed to approximately 110 bar prior to transportation via pipeline to a storage site [9][10][11] Oxy-combustion is a promising technology where fuel is burnt in a highoxygen (O 2 ) environment, using O 2 obtained from an air separation unit (ASU), instead of with air, improving combustion efficiency [12]. Safe operation conditions are maintained by recycling a fraction of the flue gas back to the furnace, thus keeping the temperatures inside the boiler close to air-firing mode [9,[13][14][15]. Burning coal under these conditions generates an flue gas rich in CO 2 (60-70 mol%) with appreciable quantities of H 2 O (20-25 mol%), O 2 (3-4 mol%) and N 2 (0-10 mol%), which varies according to coal rank and process design [13]. This flue gas is then upgraded to transport specifications via a gas processing unit (GPU) [7,14,16]. 1 Actual compression pressure is a function of the design of the CO 2 transport infrastructure and the chosen CO 2 storage option.
Oxy-combustion can also be applied to natural gas combined cycle (NGCC), however the gas turbines need to be redesigned because the increased CO 2 concentrations in the flue gas alter its physical properties [9,12]. Unlike for pulverized coal oxy-combustion, O 2 must be compressed to the high operating pressures of the NGCC before delivered to the furnace [9].
Currently, the dominant technology for producing the quantities of oxygen required for oxy-combustion of pulverized coal (above 600 kg/MWh) 2 is cryogenic distillation [17,18]. This technology was originally commercialised by Carl von Linde in 1902 [19] and is based on separation of the constituents of air using distillation at cryogenic temperatures [20][21][22][23][24][25]. Despite its technical maturity, cryogenic distillation processes are still energy intensive consuming 200 kWh/t O 2 [26] which led to proposals for reducing this penalty, such as using selfheat recuperation [27]. This high energy requirement also promoted the development of alternative technologies for air separation, such as adsorption [28][29][30], ion transport membranes (ITM) [31][32][33][34][35], and chemical looping [36][37][38]. However, none of these technologies are suitable for the production of high purity oxygen at utility scale either because of high costs, as for adsorption processes, or the technology is still under development, as for ITM [17,39].
The requirement to add both an ASU and GPU increases the capital cost of the plant and imposes an 8 -12% efficiency penalty to the process [7,40].
One way of minimising the effects of this efficiency penalty is through heat integration, which can be optimised by minimising inefficiencies within the process via an exergy destruction analysis. This analysis is based on the second law of thermodynamics, aimed at identifying inefficiencies within a system due to irreversibility [41]. Exergy refers to the amount of work that can be generated by a system on a reversible process, leaving it in equilibrium with the environment [42].
Several studies have focused on reducing this parasitic power consumption by performing thermodynamic and techno-economic analyses on double and triple column ASUs, and different GPU units [43][44][45]. Skorek-Osikowska et al.
determined that low grade heat of compression could be used to pre-heat the feedwater reducing the number of feedwater heaters required [45]. Aneke et al.
simulated an oxy-combustion process with liquid air storage and determined there was an advantage to using this strategy as well as recovering waste heat of compression [46]. Stanger et al. and Li et al. both determined that SO x can have higher concentrations in oxy-combustion flue gas than in air-combustion due to recycling and a lack of dilution by N 2 [47,48]. This increase in SO x has the effect of increasing acid dew point from 116 • C in air-firing to 141.6 • C in oxycombustion [47], as well as changes in ash composition [48,49]. Oxy-combustion CCS has been demonstrated a number of times, including the Callide oxyfuel project [50][51][52], Lacq pilot plant [53], Compostilla OXYCFB300 circulating fluidised bed [54], and Vattenfall's pilot plant [55,56]. These projects proved the feasibility of oxy-combustion and provided further insights on operational performance of the technology.
Whilst improvements in process performance are important, it is vital that they do not result in increased capital cost, leading to an increased cost per MWh of low carbon electricity generated. This creates the need to develop a methodological approach that allows the evaluation of the efficacy of a process modification in this context.
In this work, we present a novel methodological approach for the identification and rational analysis of potential process performance improvements via system integration and process intensification. This approach is grounded in the application of the 1 st and 2 nd laws of thermodynamics coupled with a capital expenditure (CAPEX) analysis. The methodology proposed in this study is well-suited for application to other CCS technologies, or more generally to other complex industrial processes, such as liquefied natural gas processes.  A medium sulphur bituminous coal with a grindability of 0.664 g/rev [57] was used in this work with the composition detailed in Table 1. The power required, W (kWh/t), to grind coal with mean particle size of 2.7 mm (F 80 ) to a target size of 90 µm (P 1 ) and mean particle size of 77 µm (P 80 ) is given by Equations 1 and 2 [57],

Power plant
The ultra-supercritical power plant was first simulated operating in air-fired mode based on the Callide oxyfuel project [50] and Spliethoff [60], as seen in was specified to be 30% of the fuel mass, representing a high ash coal scenario, with Spero reporting 24% ash in Callide coal [50]. Fly ash goes into the boiler section along with flue gas while the remaining furnace ash falls down to an ash hopper [50]. Feedwater is heated against hot flue gas in the boiler, comprised of the Evaporator (Evap), Radiant Superheater (RS), Superheater 2 (SH2), Reheater 2 (RH2), Superheater 1 (SH1), Reheater 1 (RH1), and Economiser (Econ). These components were modelled as heat exchangers following the paths shown in the boiler section of Figure 3. Flue gas is then cooled against incoming air to 130 • C in a regenerative heat exchanger (RHX) [44,61]. This temperature is chosen to be above acid dew point [9,50] and allows a safe operation for baghouse filters [50]. The operating parameters of the power plant are presented in Table 3. Table 3: Boiler and flue gas pre-treatment operating conditions.

Parameter Unit Value
Feedwater temperature • C 299 Steam temperature at superheater exit • C 600 Reheated steam temperature at reheater exit • C 620 Steam pressure at superheater exit bar 300 Steam pressure at reheater exit bar 70 Excess air ratio 1.04 Amount of ash in coal wt% 30 Amount of fly ash in total ash wt% 90 Isentropic efficiency of fans % 85 Isentropic efficiency of compressors % 75

Feedwater enters the boiler through the economiser and leaves through SH2
where it is sent to the high pressure turbine (HP turb) at 600 • C and 300 bar.
Part of the steam is used to pre-heat feedwater and the remainder is reheated in the boiler. The pressure drop of the flue gas inside the boiler was assumed to be 2.1 mbar, while feedwater had a pressure drop of 24.1 bar for the economiser, evaporator, and superheaters sections, and 5 bar for the reheater section [9].
Steam exits the boiler from RH2 at 620 • C and is expanded inside an intermediate pressure turbine (IP turb) followed by a low pressure turbine (LP turb) to 0.04 bar. The amount of steam bled from each turbine, as well as their conditions, is presented in Table 4. Steam leaving the LP Turb is condensed and pumped to the feedwater heating train, where it is heated to 299 • C against the turbine bleeds, and sent back to the boiler.

Gas Processing Unit (GPU)
A GPU process based on a combination of compression and distillation was simulated following the work of Posch et al. [62]. where more water is removed from the gas phase. This gas phase is compressed to 28 bar in compressor LP Comp3, cooled again to 25 • C in HX3 and sent to another flash separator (flash-3) for more water removal.
Compressed flue gas is sent to a dehydrator for water removal [62] and subsequently cooled against cold products from the distillation column to −31 is further compressed to 120 bar and 5 • C by a pump (P-1) and sent to the pipeline, as represented in Figure 4. The operating parameters of the GPU are presented in Table 5.

Oxy-combustion
The simulation of an oxy-combustion process was modelled following the work of Spero et al. [50] and Stanger et al. [66]. The power plant model de- 37% of this flue gas is recycled as the primary RFG and used to pulverise ground coal to the furnace [50,66]. The non-recycled flue gas is subsequently purified and compressed to transport specifications [68] on the gas processing unit (GPU) [16,50]. The resulting PFD of the boiler operating in oxy-mode is presented in Figure 3.

Heat integration
Heat integration was performed by pre-heating feedwater leaving the condenser using low grade heat of compression of the ASU and GPU as represented in Figure 3 [44,45]. The minimum allowable temperature at the regenerative heat exchanger outlet was defined as 150 • C to be above acid dewpoint [47].
Medium temperature feedwater incoming from the ASU can be used to further cool down the flue gas before particulate removal. Using this heat integration strategy allows elimination of feedwater heaters 1 to 4, requiring less turbine bleeds and hence more steam to expand in the turbines to generate more electricity. The bleed fractions for the process with heat integration are presented in Table 6.

Thermodynamics analysis
A thermodynamic analysis was performed in order to determine the potential for an improvement to the gross and net efficiency of the process using Equations 3 and 4, The power generated by the turbines (P turbines ) is the sum of the power generated by the HP, IP and LP turbines. The power consumed (P consumed ) is the sum of the power demand of the ASU, the GPU, and from the fans and pumps circulating flue gas and feedwater respectively. The heat input of the fuel is taken by multiplying the lower heating value (LHV) and coal mass flow (ṁ f uel ). In this study, gross and net efficiencies were determined on a LHV basis.
The net efficiency was compared against the maximum thermodynamic efficiency that could be achieved using an ideal Rankine cycle and the minimum thermodynamic separation work for both ASU and GPU. The ideal Rankine cycle is determined using the Carnot efficiency with Equation 5, The Carnot efficiency reflects the amount of thermal energy that could be transformed to electricity assuming a hot source (Superheater) and a cold sink (condenser) with no losses in the system. It states that the limit of efficiency is driven by the source and sink absolute temperatures, T Superheater and T condenser respectively. With Equations 3 and 5 it is possible to calculate the maximum amount of power generated by the turbines (P M ax Rankine )from a Rankine cycle. The minimum thermodynamic work required to separate O 2 from air is given by Equation 6 and to separate CO 2 from flue gas by Equation 7, Using the maximum power produced by a Rankine cycle and the minimum separation work determined by Equations 6 and 7, the maximum thermodynamic efficiency of oxy-combustion can be determined with Equation 9,

Exergy analysis
The previous analysis quantifies by how much it is possible to improve the process efficiency, but it does not give a clear indication of the units responsible for most irreversible losses. This insight can be obtained via an exergy destruction (ED) analysis for each unit operation using Equation 10, The term n(h − T 0 s) refers to the exergy of the stream, the term Q(1-T 0 /T s ) refers to the thermal energy that could be transformed into exergy using a Carnot cycle, n is the molar flow of the stream, h is the specific molar enthalpy, s is the specific molar entropy, T 0 is the reference temperature, T s is the system temperature, and W s is the work done to or by the system. Performing this analysis to each unit operation of oxy-combustion allows for the identification of the most suitable candidates for process improvement.

Simulation and model validation
Combustion is assumed to take place at an adiabatic flame temperature (T ad ) of 2167 • C, similar to the value reported by Lackner for anthracite of 2180 • C [69]. The simulated oxy-combustion process has a gross efficiency of 47.2% and a net efficiency of 34.6%, resulting in a 12% efficiency loss when compared with the unabated power plant. A good agreement between the simulated process and the results from IECM [58] and Callide oxyfuel project [51] was obtained, as presented in Table 7. The disagreement observed for O 2 demand and fuel burned is due to the assumption that complete combustion took place with no carbon monoxide (CO) formation. The parasitic energy demand incurred by the ASU, GPU, and power plant are presented in Table 8 showing a good agreement with the values reported by Tranier et al. [26]  On this basis, the ASU and GPU were determined to be operating with a thermodynamic efficiency of 24% and 5%, respectively.
Assuming an increase in thermodynamic efficiency of the ASU, GPU, and boiler it is possible to observe from Figure 5 that the parasitic power losses from these systems tend to decrease. A 5% increase in efficiency of the ASU results in a 17% decrease of power consumption while for the GPU results in a 46% reduction. This shows a higher increase in process efficiency with an initial improvement of the GPU, however it will be preferential to start improving the ASU once the GPU achieves a separation efficiency of 22%. observed that improving the GPU will lead to the greatest improvement in the overall process.

Process improvement and heat integration
The net efficiency of the simulated power plant was 34.6%. Then, assuming ideal separation of O 2 from air and CO 2 from the flue gas, the net efficiency improves to 43.1%. Tranier et al. [26] stated that Air Liquide can improve separation efficiency of the ASU by 10%. However, an improvement of up to 21% relative to the base simulation could be achieved by increasing the operating conditions of the boiler. The maximum thermodynamic efficiency of the power plant was 67% with a theoretical power output of 720 MW, and the minimum thermodynamic separation work for O 2 and CO 2 was 24 and 16 kJ/mol, respectively. This results in a maximum theoretical efficiency of 63% for oxy-combustion, showing that the simulated process is 55% thermodynamically efficient, represented in Figure 6. Advanced ultra-supercritical technology with steam parameters operating up to 700 • C and 350 bar are currently being developed [70,71]. Such technology is able to achieve net efficiencies up to 52% on a LHV basis however, operating under such conditions is limited to current material selection [70,71].
The exergy destruction analysis identified the boiler as the largest source of inefficiency. Due to material limitations, there are limited opportunities to improve this process element, however the feedwater heating train shows po-tential for reducing the inefficiencies of the plant. Here, the low temperature feedwater heaters 1 to 4 were identified as an important source of inefficiency.
Therefore, this is an excellent point to reuse low grade waste heat of compression from both ASU and GPU to increase the temperature of condensed feedwater.
After using this low grade heat, it was found that the feedwater temperature increased sufficiently to completely bypass feedwater heaters 1 to 4, allowing for their removal as illustrated by the heat integration path in Figure 3.
The exergy destruction associated with the feedwater heating train has been significantly reduced by 53% as shown in Figure 7 due to the removal of the low pressure heaters, FWH1 to 4. This reduces the heat requirement from coal, reducing its consumption to 0.36 kg coal /kWh, and increasing gross and net efficiencies to 49 and 38%, respectively. The main exergy losses of the ASU were determined to be in the MHX, also identified by Taniguchi et al. as responsible for more than half of the exergy destroyed in the ASU [72]. Although possessing small losses (4% of total losses), the MAC was identified as a good target for heat integration as compressed air is discharged at high temperatures. An alternative is to use adiabatic compressors with one inter-and one after-cooler instead of isothermal compressors. Feedwater can be used to cool this air stream showing potential to reduce capital costs and cooling water [43].
Most exergy is destroyed in the MHX due to the high heat duty. The difference in pressure of air, at 4.2 bar, compared with the pressure of products, which are at 1.2 bar, also contribute to the exergy destruction of this unit. Decreasing the temperature at which the air feed enters the exchanger reduced the exergy destruction of the ASU as shown in Figure 8, however that also reduces the O 2 temperature leaving the heat exchanger. Although this can negatively affect combustion efficiency, this is not a concern for oxy-combustion, and adopting this strategy will increase operational safety [14,43].
Changing the MAC from an isothermal compressor to a two-staged adiabatic compressor has led to an increase in both exergy destruction and power consumption. This change allows to use the low grade heat of compression to pre-heat a portion of the feedwater while cooling down the compressed air stream. By using this strategy it is possible to reduce the number of feedwater heaters, as previously described.
The power consumption of both the ASU and GPU is primarily associated with the compression work. As the compressors are operating at 90% adiabatic efficiency, it will not be likely that significant increase in thermodynamic efficiency will come from improving them. One possible way of reducing the power consumption from the ASU is sending a greater amount of compressed air to the expansion turbine. Applying this strategy, the second stage of the booster was removed because the required compression could be provided using the amount of work obtained in the expander. The number of stages from the HP column was reduced to 10, and the LP column was reduced to 23 stages in order to lower the oxygen purity to 97 wt%. This strategy reduces the power consumption of the unit to 197 kWh/t O 2 , similar to previous work [26,[43][44][45]73]. The MHX was found to be where most exergy was destroyed on the GPU, as for the ASU. Low grade heat of compressed gases can be used to pre-heat feedwater [43][44][45]50] however, Spero et al. [50] stated that some care has to be taken when employing this strategy. This is because the unit is only started when CO 2 concentrations on the flue gas are high enough to allow a successful separation. This operation could be put at risk if high concentration of impurities are present in the flue gas. One possible way of avoiding this issue is to use heat from the low pressure compression train and while CO 2 concentrations are not sufficiently high to run the process send this flue gas to the stack. This approach has been shown by Skorek-Osikowska et al. [45] to provide a greater increase of net efficiency instead of using all heat of compression available in the GPU.
Changing the first compressor from a one-stage to a two-staged compressor with inter-cooling has reduced the exergy destruction of the unit. This strategy also reduces the power required by the compressor and avoids increasing the gas temperature to levels that could reduce the equipment lifetime. Low grade heat of compression is used to pre-heat another portion of the feedwater and cool down the compressed flue gas. The second compressor uses the rest of the feedwater to cool down the compressed flue gas, while using its heat to increase the feedwater temperature. This allowed for a 6.4% decrease of exergy destruction in the GPU obtaining a final CO 2 stream with 99 wt% purity at a specific power consumption of 137 kWh/t CO 2 , considering the final CO 2 compression to pipeline pressure.
In summary, a 4% decrease of exergy destruction is observed after process optimisation and heat integration between the power plant and both ASU and GPU.

Techno-economic analysis
The results from the thermodynamic and exergetic analysis show that the changes proposed to the process have improved the efficiency of the plant. To analyse if these changes would be economically appealing, the efficiency was plotted against CAPEX as represented in Figure 9. This enables a comparison with the baselines, chosen to be the unabated power plant CAPEX and the maximum thermodynamic efficiency determined earlier. The ideal path for technological innovation is represented by the arrow in Figure 9 going from the oxy-combustion before improvement towards where the baselines cross. A reduction of CAPEX from £436M to £369M and an increase of net efficiency from 34% to 37% is observed, resulting in a decrease of 13% in £/MWh. The process improvement using this analysis reduced CAPEX and increased efficiency, but showed a sharper reduction of CAPEX than increase in efficiency.

Conclusions
The methodology proposed in this study is illustrated in Figure 10 and was successful in identifying opportunities for the concurrent improvement in process efficiency and reduction in capital cost of an oxy-combustion power plant. In this study, heat integration between the feedwater heating train and the compressors in the ASU and GPU allowed the removal of the low pressure feed-water heaters, reducing the amount of steam bleed from the LP steam turbine, thus increasing the amount of electricity generated. This has the dual benefits of reduced capital cost and improved power generation efficiency. Moreover, this heat integration strategy would reduce the water intensity of the CCS plant, resulting in a smaller cooling tower, land use, and capital cost.
The minimum thermodynamic separation work for the ASU to obtain the desired oxygen purity of 97 wt% was found to be 49.9 kWh/t O 2 and for the GPU to obtain a CO 2 purity of 99.9 wt% was 5.4 kWh/t CO 2 . After heat integration the power consumption of the ASU was reduced by 3.4% to 197 kWh/t O 2 and the GPU by 2.1% to 137 kWh/t CO 2 , representing a second law efficiency of 24% and 5%, respectively.
On the basis of this analysis, the maximum Rankine efficiency of the boiler simulated in this study was 66.8% and the maximum theoretical efficiency of the oxy-combustion process was 62.3% LHV. Owing to the low efficiency of the GPU relative to the ASU, focusing on improving the Second Law efficiency was observed to be a promising option for improving the efficiency of the oxycombustion process.
Finally, using this approach, the net efficiency of the oxy-combustion was increased by 3% to 37% LHV, with a CAPEX reduction of 15%, and reduced the £/MWh by 13%.
This work is therefore of general use to anyone proposing a new power generation or storage technology and provides a rational basis for its evaluation and comparison with incumbent options.