Closing carbon cycles Evaluating the performance of multi-product CO2 utilisation and storage configurations in a refinery

Carbon capture and utilisation (CCU) has the potential to provide business cases as CO 2 waste streams are turned into feedstock for the synthesis of marketable products. Although CCU could reduce fossil resource demand, its capability as a climate change mitigation option is under debate. In contrast to single-product CCU, this pro- spective study explores the techno-economic and environmental feasibility of novel systems that include more than one CO 2 utilisation product. The combination of multi-product CCU with CO 2 storage is also investigated. Two con ﬁ gurations have been designed, in which CO 2 is captured in a re ﬁ nery and converted into dimethyl ether (DME) and polyols, simultaneously ( parallel con ﬁ guration) or in two consecutive cycles ( cascade con ﬁ g- uration). Compared to a reference system without capture, results show that the largest direct CO 2 emission reductions are achieved with CCS without utilisation ( − 70%) but at the expenses of higher total costs (+7%). Multiproduct CCU systems show lower fossil depletion and costs than the reference without capture ( − 10% and − 9%, respectively) because of feedstock replacement by the CO 2 utilised. Combination of multi-product CCU with storage turns to be the best alternative for reduced climate change potential ( − 18% relative to the re- ference) while still been economically feasible. In addition to lower upstream emissions due to fossil feedstock replacement by utilising CO 2 , process direct emissions diminish owing to storage. No signi ﬁ cant di ﬀ erences were found between the cascade and the parallel con ﬁ gurations. The extra e ﬀ ort to recycle CO 2 in the cascade con ﬁ gurations is neither penalised nor rewarded.


Introduction
Carbon capture and utilisation (CCU) concepts are increasingly been researched, since in addition to reduce CO 2 emissions they could result in lower fossil resource demand [1,2]. Moreover, the large capital investment associated with carbon capture could (partially) be compensated because CO 2 is converted into valuable products that provide revenues [3,4]. However, the potential of CCU as a pathway to obtain large emission reductions has been debated, due to the short CO 2 storage time of many applications, and the difficulties to assess potential displacement effects.
Previous literature studies have investigated CCU options mainly focusing in CO 2 conversion into fuels including techno-economic aspects and simple carbon metrics. These studies show that the production of liquid hydrocarbon fuels with commercially proven CCU technology is not yet economically viable [5]. Methanol synthesis from CO 2 and renewable hydrogen from electrolysis is only economically feasible for large plant capacity, when by-products are sold and methanol has a high selling price [6], or when the feedstock costs are lower and the CO 2 value is high [7]. However, methanol production from captured CO 2 has the potential of net reduction of CO 2  process. Other CCU options consider formic acid as final product, which can be used as hydrogen carrier or as fuel for fuel cells [8]. When using renewable electricity and steam, this CO 2 utilisation alternative has lower CO 2 emissions than the corresponding conventional process. In spite of its environmental attractiveness and technically feasibility, CO 2 -based formic acid is not yet financially attractive.
CCU has been typically regarded as a technology towards a single end product. The earlier studies indicated that the major drawback is a no yet economically practicability. However, integrated CO 2 conversion into multiple fuels and chemicals in the same facility could benefit from synergies, such as process and equipment integration and more efficient energy and resource use, analogous to the refinery concept. Systemwide and detailed assessments of CCU configurations for co-production of fuels and chemicals are not available in literature.
This paper aims to assess whether configurations that combine multi-product carbon capture and utilisation (CCU) and multi-product carbon capture, utilisation and storage (CCUS), are feasible concepts to close carbon cycles in the petrochemical industry. In this study, we investigate the potential benefits and trade-offs of such multi-output configurations.
To answer the research goal, a case study in a refinery was used, where CO 2 is captured at a steam methane reforming (SMR) facility and is utilised into dimethyl ether (DME) and polyols, different target products than those that received more attention in preceding research (i.e., methanol). This case was selected as: • SMR processes contribute to 20% of the CO 2 emissions at refineries, and refineries account for 10% of the global industrial CO 2 emissions [9].
• CO 2 conversion into fuels is considered as an attractive option to achieve large CO 2 emissions reduction due to its high fuel global market demand (100-times higher than that of chemicals; [10]. Whereas liquid hydrocarbon fuels, methanol or formic acid are still not economically attractive [5,7,8,6], DME has been reported as a cost-effective option to replace conventional transportation fuels [11,12]. Moreover, DME is a sulfur-free fuel with higher cetane number than diesel and leads to very low emissions of particulate matter, NOx, and CO during its combustion [13]. CO 2 -based DME appears as a more efficient alternative compared to conventional DME synthesis [14]. CO 2 is used in a methane dry reforming process to produce syngas, which is then directly transformed into DME [15]. This option has large market potential but the CO 2 is stored for a short period of time. • The second product in the configuration considers CO 2 conversion into chemicals. Urea and salicilic acid synthesis using CO 2 are wellestablished industrial processes [3]. CO 2 -based polyols are a CCU alternative with high potential for market growth, so they can contribute to meet emissions reduction targets. These polyols are already at commercial stage [16,17] and used as precursors of polyurethane flexible foams. The CO 2 incorporated into the polyol is limited to 20 wt% in order to meet the right flexibility of the final product [18]. This option has a lower market potential than fuels but the CO 2 is stored for a longer period of time (decades vs. days).
Process modelling of the commercial-scale CO 2 source (SMR unit in the refinery), the CO 2 capture unit and CO 2 conversion processes (DME and polyols) serves as basis for an integrated techno-economic and environmental assessment. The environmental evaluation follows a life cycle-assessment approach incorporating climate change and fossil depletion indicators since fuel savings could be a relevant benefit of CCU options (Pérez-Fortes et al., [6]. A comparison among the different CCU and CCUS configurations is carried out to identify the economic and environmental hotspots of each system.

Scoping
For the purpose of assessing multi-product CCU and multi-product CCUS systems, two configurations were defined. The CO 2 source in all cases is a hydrogen manufacturing unit of a refinery equipped with carbon capture. In the first configuration (parallel) the synthesis of DME and polyol happens in parallel, using the CO 2 stream captured at the hydrogen unit. In the second configuration (cascade) the CO 2 captured in the hydrogen unit is first used in the synthesis of DME. The CO 2 released during DME processing is then re-captured and used as feedstock for polyols synthesis. In addition, two reference configurations were considered. One reference case consisting of H 2 production at the refinery without carbon capture, and a storage case that only incorporates CO 2 capture, transport and storage. In total, six different systems were defined (see below). Note that it is assumed that the CCU products would displace fossil based counterparts, and therefore in the systems where there is no utilisation (REF and CCS cases), DME and polyol are still produced but in the conventional manner.
-Case 2. Storage-only (CCS): H 2 unit of a refinery with CO 2 capture and storage + conventional DME + conventional polyol production (no CO 2 utilisation). -Case 3. Multi-product CCU, Parallel: H 2 unit of a refinery with CO 2 capture + CO 2 -based DME + CO 2 -based polyol production. After capture, the CO 2 stream is split into two parts. One part of the CO 2 is used for CO 2 -based polyol synthesis and the rest is used for CO 2 -based DME production. There is no CO 2 storage. -Case 4. Multi-product CCU, Cascade: H 2 unit of a refinery with CO 2 capture + CO 2 -based DME + CO 2 -based polyol production. The CO 2 captured from the H 2 unit is first used for DME production. During DME production 90% of the used CO 2 is re-emitted [19]. Part of the CO 2 released in the DME process is then re-captured, and utilised in CO 2 -based polyol synthesis. The rest of the CO 2 is released to the atmosphere. There is no CO 2 storage. -Case 5. Multi-product CCUS, Parallel: H 2 unit of a refinery with CO 2 capture + CO 2 -based DME + CO 2 -based polyol production. After capture, the CO 2 stream is split into two parts. One part of the CO 2 is used for CO 2 -based polyol synthesis and the rest is used for CO 2 -based DME production. The CO 2 released during DME synthesis is re-captured and sent to storage. -Case 6. Multi-product CCUS, Cascade: H 2 unit of a refinery with CO 2 capture + CO 2 -based DME + CO 2 -based polyol production. The CO 2 captured from the H 2 unit is first used for DME production. The CO 2 released in the DME process is then re-captured, a part of it is utilised for CO 2 -based polyol synthesis and the rest is stored.
The total systems were divided into system areas (SA) as shown in Fig. 1 and Table 1. Each SA corresponds to a part of the value chain or process type (e.g., natural gas production and transport, hydrogen manufacturing unit, polyol synthesis, etc.). The division on SAs allows transparently communicate differences in the type and level of modelling complexity among SAs and clearly identify the sub-processes with the largest contributions to the costs and environmental impacts.
In all cases, three final products (H 2 , DME and polyol) are produced. To size the configuration we chose a SMR unit with a typical commercial-scale production capacity (59 kt/a of hydrogen; [20,21]. From this unit, 95% of the direct CO 2 emissions (337 CO 2 kt/a) are captured via chemical absorption [21]. CO 2 -DME synthesis requires 1.76 kg CO 2 /kg of DME [19], thus 192 kt/a CO 2 -DME could be produced from the CO 2 captured at the SMR unit. Because it is not realistic to have a plant that uses all the CO 2 for polyol, we assumed the capacity of the   polyol process to be 250 kt/a, based on a commercial scale plant [22]. For a fair comparison of the performances among the six case studies, the same process capacities were used regardless of whether the final products are produced from CO 2 or fossil fuels ( Table 1). Note that a particular situation occurs for DME production in the parallel configurations (cases 3 and 5). These are utilisation cases, and therefore all DME should be CO 2 based. However, in these configurations, the CO 2 captured from the H 2 unit is also used for polyols synthesis resulting in a lower availability of CO 2 for DME synthesis (compared with the cascade cases, in which all CO 2 captured at the H 2 unit is converted into DME). To maintain the same final product output as in the other systems, in the parallel cases 159 kt/a DME are produced from the CO 2 available and the deficit is compensated by conventional DME (33 kt/a DME), (Table 1).

Process assessment
The mass and energy balances and equipment list of the production processes of hydrogen, DME and polyol, the CO 2 capture units and CO 2 compression before transport and storage were based on previous inhouse research [21,19,23,22]. These models were developed based on the scope of the assessment, the state of the art literature review and an evaluation of the knowledge base available, which determined the level of detail and complexity of each model. The details on the process models are briefly described in this section.
No specific models were developed for SA 1. NG production and transport, SA 6. Chemicals and CO 2 transport and storage of SA 3. The amounts of NG, chemicals and CO 2 included in these system areas were determined from the mass of NG and chemicals needed and CO 2 produced in the other system areas.

H 2 production
Hydrogen production via SMR is a well-known established process. In this process, methane and steam are pre-heated and fed to the reformer (560°C, 30 bar), (Fig. 2a). The syngas is further converted to H 2 and CO 2 in a water gas shift (WGS) reactor (400°C, 25 bar). The stream from the WGS reactor (220°C, 25 bar) contains some water, which is removed, so only CO 2 and H 2 enter the pressure swing adsorption unit (PSA). In the PSA unit (220°C, 25 bar), H 2 is obtained as main product, while the offgas is burned in the furnace section of the reformer to aid the endothermic reaction. The extra heat available after the reforming reaction is used for steam generation. The specific SMR models for the H 2 unit included in this study were developed in Aspen Plus V8.4. based on [22].

CO 2 capture and compression
The CO 2 capture unit in the SMR facility was adapted from Meerman et al. [21], which identified the optimal techno-economic configuration of CO 2 capture at SMR facilities using currently available technologies. The authors indicated that the most efficient location for the capture unit is after the WGS reactor (Fig. 2b). For solvent selection, energy requirement, loading capacity, corrosiveness, vapour pressure and chemical stability were analysed. In consultation with industry experts, ADIP-X solvent (a mixture of methyl diethanolamine (MDEA) and piperazine in water) was selected to capture 95% of the overall CO 2 produced in the WGS reaction.
In the cascade and CCUS-Parallel case studies, the CO 2 released in the synthesis of CO 2 -DME is re-captured in a second unit (SA 7). The CO 2 concentration in this stream is 33 mol% (as opposed to 20 mol% in the first capture unit). Instead of using the very specific model of the first capture unit (developed to capture CO 2 after WGS in SMR units), the second capture unit was modelled based on a generic model using more conventional MEA solvent to capture 90% of the CO 2 [23]. The steam demand in the reboilers of both capture units is covered by the steam generated in the H 2 unit (Section 2.2.1).
The CO 2 captured is then compressed to 110 bar for transport via pipeline to an offshore aquifer, where it is stored. The compression train is made up of four compression stages with intermediate cooling and a final pump (Fig. 2b). Specific compression trains for the different case studies were modelled in Aspen Plus V8.4.

DME production
The Aspen Plus models of conventional DME and CO 2 -DME production processes developed in Schakel et al. [19] have been used in the current study. In the conventional route for the synthesis of DME, liquid methanol is pre-heated to 160°C and fed to the dehydration reactor, where is directly converted to DME at 15 bar of pressure (Fig. 3a). After the reaction, the produced DME, the water and the unconverted methanol are depressurized to 10 bar and cooled down to 50°C. This mixed stream is fed to a distillation column. DME at > 99.5% is Complete details can be found in Schakel et al. [19].
In the CO 2 -based DME production process CO 2 and methane are pre-  heated and fed to the dry reforming reactor (800°C, 2 bar), (Fig. 3b). The syngas leaving the reactor is cooled down and compressed to 79 bar for direct DME synthesis (250°C). Methanol, syngas, water, CO 2 and methane are the impurities leaving the direct DME reactor. After depressurization to 10 bar and cooling down to 32°C, this stream is fed to a distillation column. CO 2 , CH 4 , CO and H 2 are separated at the top of the column (-48°C, 10 bar). DME, methanol and water leaving the column at the bottom (45°C, 10 bar) are directed to a second distillation column to obtain highly pure DME (> 99.5%, 40°C). The syngas, methane and CO 2 separated at the top of the first distillation column, together with the methanol and water stream separated at the bottom the second column (70°C, 10 bar), are fed to a furnace and burnt for heat recovery.
In the parallel configurations, the flue gas stream containing CO 2 is either directly released to the atmosphere (case 3. CCU-Parallel) or captured and sent to storage (case 5. CCUS-Parallel). In case 4. CCU-Cascade, only the amount of CO 2 needed for polyol synthesis is captured and the rest is released to the atmosphere. However, in case 6. CCUS-Cascade all CO 2 from the DME process is captured. Partly is used in polyol synthesis and the rest is stored.

Polyol production
The models for polyol synthesis developed in Fernández-Dacosta et al. [22] were used in the present study. These spreadsheet models of the conventional polyol processes were built upon input data on reaction parameters, polyol properties and process line-ups described by industrial experts. Due to the novelty of the process and lack of real process data available, the CO 2 -based polyol model was based on literature [24][25][26] and patents [27][28][29].
Conventional polyethercarbonate polyol (PPC) is produced from the reaction of propylene oxide (PO), glycerol (G) and monopropylene glycol (MPG), which takes place at 135°C and 3 bar (Fig. 4a). Double metal cyanide (DMC) is used as catalyst and recovered in a filter after the reaction. In a vacuum-stripping step (140°C, 25 mbar) odours and other impurities are separated, and polyol is obtained as final product (40°C, 1 bar).
PO, MPG and glycerol are also starting materials for the synthesis of CO 2 -based polyol. However, part of the PO needed in the conventional process is replaced by CO 2 (Fig. 4b). The maximum content of CO 2 in the final product polyol is 20 wt% to ensure the right flexibility for flexible foam application [16]. After the reaction (135°C, 20 bar), the CO 2 that has not been converted to polyol is flashed out at 3 bar and recycled. The by-product cyclic propylene carbonate (cPC) is removed with the odours in the vacuum stripping step (140°C, 25 mbar), and thus polyol at 99.9 wt% is obtained as final product (40°C, 1 bar). For full details of the models see Fernández-Dacosta et al. [22].

Economic evaluation
Based on the equipment sizes and mass and energy balances from the process models, capital and operational costs were calculated for each of the systems investigated. Capital costs and operational costs were used to determine the total systems costs and net present value (NPV). Finally, H 2 , DME and polyol costs were calculated per case study to evaluate how the introduction of CO 2 capture units would affect the final products costs.
The geographical location of this analysis is North West Europe, the scope is 20 years, and the reference year is 2016. Cost data was corrected for inflation using the Consumer Price Index (CPI) for the raw materials and the Chemical Engineering Plant Cost Index (CEPCI) for the equipment. To convert US$ to €, an exchange rate of 1.11 US$/€ was used [30].

Capital and operational costs
Capital expenditure (Capex) is divided into fixed capital costs (FCC) and working capital costs (WCC). FCC are the initial investment needed to build the plant and the WCC are the additional investment needed to start up the plant. FCC were estimated using the factorial method (Eq. (1) and Eq. (2) from [31]. In this method, the capital costs are estimated based on the purchased cost of equipment (PCE), and the physical plant costs (PPC). The cost items that contribute to the FCC are estimated as factors of the PCE (f1 to f12 in Eq. (2)). The factors per cost item used in this study are shown in the Supplementary material (S.2.1). The WCC was assumed 15% of the FCC [31].
The PCE was estimated applying the exponent method [32]. In this method a cost correlation (Eq. (3)) is used that relates specific cost constants (a, b) and exponents for each type of equipment (c) with its characteristic size (S).
The characteristic sizes of the equipment were derived from the process assessment. For some sub-process, the capacity varies depending on the case study. Since the equipment of these sub-processes are the same regardless the case study, equipment sizes were adjusted using the exponent method Towler and Sinot, [32]. A typical scale exponent of 0.65 for petrochemical processes was used in this study. Specific details of the PCE estimation are reported in the Supplementary material (S.2.1).
The total operational expenditure (Opex) consists of indirect (IOC) and direct operating costs, which in turn are divided into fixed (FOC) and variable operating costs (VOC), Eq. (4): Fixed operating costs include maintenance, capital charges, rates, insurance, license fees and royalty payments. These costs are estimated as a percentage of the FCC. Labour costs are also a fixed operating expenditure and estimated as a percentage of the total production costs. Laboratory, supervision and plant overheads costs are based on labour costs Sinnot, 2005. The specific percentages assumed for the FOC estimation are described in the Supplementary material (S.2.1). The indirect operating costs were assumed 25% of the direct operating costs [31].
Variable operating costs cover raw materials, utilities and miscellaneous operating material costs. Miscellaneous operating materials are plant supplies required to operate the plant such as safety clothing, instrument charts and accessories, pipe gaskets or cleaning materials. Miscellaneous materials were assumed 10% of maintenance costs [31]. Raw materials and utilities costs were estimated based on the mass balances (from the process assessment) and the material prices. Utilities prices were estimated based on the correlations proposed by Ulrich and Vasudevan [33] and the industrial gas price of 2016 [34]. Raw materials prices used in this study are shown in Appendix A. Specific details in the estimation of raw materials prices are reported in the Supplementary material (S.2.2).
Natural gas (NG) production and transport only has operational costs made up of the costs for NG exploration, operation, transportation, storage and refueling. These costs were adopted from IEA [35] and [36]. CO 2 transport and storage costs were taken from ZEP [37][38][39]. It is assumed that the CO 2 is transported a total distance of 100 km, 5 km onshore to the port and 95 km offshore to the saline aquifer were it is stored. Specific costs used in this study and details in the calculation of CO 2 transport and storage costs and NG production and transport costs are shown in the Supplementary material (S.2.3).

Total system costs and NPV
Total costs were estimated as the sum of the annualised Capex plus Opex. 20 years of plant lifetime (n) and a discount rate (d) of 10% were assumed to annualise the capital costs (ACC) following Eq. (5) Towler and Sinot, [32]: The NPV was calculated based on the initial investment (I), which equals the total Capex and the net cash inflow expected to be received in each period (R), Eq. (6).
The net cash inflows were calculated based on the purchase price of the products (PP, Appendix A) and their annual produced amount (PA, process capacity in Table 1) (Eq. (7)):

Final product costs
Besides estimating total costs at system level, the final product costs (H 2 , DME and polyol) were calculated. Final product costs were estimated by adding the costs of the system areas needed for the production of each product (Fig. 1) and dividing it by its total production capacity ( Table 1). Capture costs were allocated to the CO 2 source or to the CO 2 product, taking into account whether the main purpose of the CO 2 capture unit was storage or utilisation. The assumptions to estimate final product costs based on mitigation/utilisation goals were: • Case 2. CCS: CO 2 from the H 2 unit at the refinery is captured, transported and stored for mitigation goals. Therefore, capture, transport and storage costs are allocated to H 2 .
• Case 3. CCU-Parallel: CO 2 from the H 2 unit at the refinery is solely captured for utilisation goals. Thus, capture costs are allocated to the utilisation products i.e., CO 2 -DME and CO 2 -polyol, based on the mass of CO 2 incorporated to each product.
• Case 4. CCU-Cascade: CO 2 from the H 2 unit at the refinery is captured and utilised in CO 2 -DME synthesis. The costs of this capture unit are completely allocated to CO 2 -DME. A second capture unit only captures the CO 2 emitted during CO 2 -DME production that is used in the synthesis of CO 2 -polyol. Consequently, cost of this unit is allocated to the CO 2 -polyol.
• Case 5. CCUS-Parallel: as in case 3.CCU-Parallel, the costs of the first capture unit are allocated to CO 2 -DME and CO 2 -polyol based on the mass of CO 2 incorporated in each product. In this CCUS case, the CO 2 emitted during CO 2 -DME production is captured in a second capture unit and transported to storage for mitigation purposes. Thus, CO 2 -DME pays for the second CO 2 capture unit, transport and storage.
• Case 6. CCUS-Cascade: as in case 4.CCU-Cascade, CO 2 -DME pays for the first capture unit. A second capture unit captures all CO 2 emitted during CO 2 -DME production. Part of the CO 2 captured in this second unit is used in the synthesis of CO 2 -polyol and the rest is sent to storage. Therefore, the costs of the second capture unit are allocated to CO 2 -polyol, based on the mass of CO 2 incorporated to the polyol and to CO 2 -DME, based on the amount of CO 2 sent to storage. CO 2 -DME pays for the transport and storage.
The equations used to estimate the final product costs in each case study and the allocation factors used are in the Supplementary material (S.2.4).

Environmental assessment
Given the complexity of the systems and the explorative approach of this research, two impact categories i.e., climate change (CC), fossil depletion (FD) were estimated following a cradle to gate life cycle assessment (LCA) perspective.
One of the most challenging aspects for LCA of multi-output configurations is the selection of the functional unit. In this research, a system expansion approach was used by defining a harmonized basket of products (H 2 , DME and polyol), which are produced in the same amounts in all systems. To create this harmonized basket of products, the departure point is 1 MJ of H 2 produced at the refinery. The amount of CO 2 captured per MJ of H 2 produced is calculated, and based on this CO 2 captured, the corresponding amounts of CO 2 -DME and CO 2 -polyol produced were estimated. The functional unit used in the LCA is therefore 1 MJ H 2 + 0.78 MJ DME + 0.04 kg polyol.
Material and energy balances from the process assessment were used as input to create a life cycle inventory (LCI). The LCI was combined with system process data from Ecoinvent 3. The ReCiPe impact characterization method was used [40] with a hierarchist perspective (ReCiPe Midpoint (H) v1.12) and European normalization (Europe Recipe H).
Key assumptions taken for the environmental assessment are shown below (a complete list of assumptions is in Appendix B): • Heat integration in the H 2 unit generates steam, which is used to cover the steam demand in the capture units and in the DME and polyol processes. The surplus of steam is credited Schwarz et al., 2002.
• Methane fugitive emissions from upstream natural gas production processes are included in the LCA, as they are reported to contribute to large impacts on the overall greenhouse gas emissions (GHG) of a fossil fuel based production process [41], Appendix B).
• Impacts of chemical plant infrastructure are excluded from the analysis due to limited data available and because the share of these impacts is typically small in this sector [26].
• The impacts of NG and CO 2 transport and storage infrastructure are included in the LCA (Appendix B). Table 2 shows the most relevant results, which are discussed in the next sections. Full mass and energy balances of the sub-processes included in each case study are in the Supplementary material (S1).

• CO 2 flows
The largest CO 2 emission reductions are achieved when all CO 2 that is captured is stored (Case 2.CCS-only). Direct CO 2 emissions of the CCS system are reduced by 70% compared to the reference case with no carbon capture.
Multi-product CCU systems (cases 3 and 4) show low CO 2 direct emissions reductions compared to the reference case without capture, mainly due to the CO 2 emissions in CO 2 -DME synthesis, which accounts for 90% of the CO 2 utilised in this process (337 kt/a). The CCU-Cascade configuration (Case 4) has slightly lower CO 2 emissions than the CCU-Parallel (Case 3) because a small fraction of the CO 2 released during CO 2 -DME synthesis is re-captured in a second unit and used in polyol synthesis.
In order to achieve large mitigation goals, multi-product CCU needs to be combined with storage. Lower amounts of CO 2 are emitted in the CCUS cases compared to the CCU cases because all the CO 2 released in the CO 2 -DME process is captured and either stored (Case 5.CCUS-Parallel) or partly used in polyol synthesis and partly stored (Case 6.CCUS-Cascade). Less CO 2 is captured in the second unit in the CCUS-Parallel system (case 5, 230 kt/a) respect to the CCUS-Cascade (case 6, 277 kt/a) due to lower CO 2 -DME production (Section 2.1).

• Fossil feedstock replacement (NG, MeOH, PO)
One of the advantages of CO 2 utilisation is its potential to displace and therefore reduce fossil resource demand. Propylene oxide (PO) is used as feedstock in the conventional synthesis of polyol (section 2.2.4). In the production of CO 2 -based polyol, 17% of PO is replaced by CO 2 ( Table 2).
Methanol is used as feedstock in the conventional synthesis of DME (section 2.2.3), and instead of methanol, natural gas is used in the production CO 2 -based DME. Therefore, in the CCU and CCUS systems (cases 3 to 6), higher amounts of natural gas are needed (310 to 331 kt NG/a in the CCU and CCUS cases compared to 210 kt NG/a in the REF and CCS cases). No methanol is used in the cascade configurations because all DME production is based on CO 2 . However, a small amount of methanol (46 kt/a) is needed in the parallel systems (cases 3 and 5) to produce conventional DME (33 kt/a) in order to keep an equal total DME amount produced among all the case studies (192 kt/a, Section 2.1).

• Energy
In all case studies steam is generated in the hydrogen unit from heat integration (indicated by the minus sign in Table 2). The net steam generated in the REF case is the highest among all cases. The reason is that the steam produced in the hydrogen unit is used to fulfill the demand in the capture units and the production processes. The net steam generated in the CCUS systems (cases 5 and 6) is the lowest (10-2% of the steam generated in the REF case, respectively) because they include a second capture unit after CO 2 -DME synthesis. A higher amount of CO 2 -DME is produced in the CCUS-Cascade system than in the CCUS-Parallel (Section 2.1). Subsequently more CO 2 is emitted and higher amount of steam is consumed in the second capture unit. Overall steam generated in the CCU-Cascade system is 18% lower than in the CCU-Parallel system because it includes a small second capture unit that only captures the CO 2 needed for polyol synthesis.
The introduction of a compression train for CO 2 transport to storage increases the electricity demand in the systems including CO 2 storage (cases 2, 5 and 6). The energy required in the CO 2 -DME process to compress the syngas leaving the dry reformer reactor to 79 bar for direct DME synthesis (Section 2.2.3) largely increases the electricity consumed in all CCU and CCUS systems. Electricity consumption in the REF and CCS cases is 5% and 27% of the electricity of the CCU and CCUS systems, respectively ( Table 2).

Total system costs
The costs at system level are shown in Table 3. All cases have higher Capex than the REF case, being the Capex of the multi-product CCU systems (cases 3 and 4) higher than the CCS-only system (case 2), and the Capex of the multi-product CCUS systems (cases 5 and 6) the highest across all alternatives.
However, only the CCS-only system (case 2) has higher Opex than the REF case because of the energy and materials requirements for CO 2 capture, compression, transport and storage. This extra operational costs due to CO 2 capture are partly compensated in the CCU and CCUS cases because CO 2 is utilised in the synthesis of DME and polyol and therefore this cases benefit from lower feedstock demand than in the REF and CCS, in which DME and polyol are manufactured via the conventional routes.  Total system costs follow the same trends as the Opex. Note that no CO 2 price (tax) was assumed in the estimation of the total costs at system level (H 2 unit + DME production + polyol production). From a refinery perspective, taking into account a CO 2 allowance price from the emissions trading system will further motivate the introduction CO 2 capture units.
The NPV analysis shows that in economic terms, the most preferable systems are the multi-product CCU configurations (cases 3 and 4), followed by the multi-product CCUS (cases 5 and 6). CCUS systems have higher total costs and lower NPV than the CCU cases because they include a second capture unit, CO 2 compression, transport and storage. Only the CCS system (case 2) shows higher total costs and lower NPV than the reference system (case 1) with no capture. This is an interesting finding since previous studies have shown than other CCU plants producing methanol or formic acid considering renewable H 2 from electrolysis, do not easily achieve a positive NPV [7,8,6].

• Capex contribution per system area
The differences in the Capex between the different cases are better understood by analysing the SA contribution, as depicted in Fig. 5a.
In all case studies the capital costs increase with respect to the REF case due to the incorporation of the first capture unit. However, the most capital-intensive system areas making a difference between the case studies are SA 3. CO 2 compression, transport and storage and SA 4b. CO 2 -DME synthesis, both of them due to compression costs. The Capital costs of SA 3. CO 2 CTS and SA 4b. CO 2 -DME synthesis only differ by 9%. Therefore, the total Capex of the CCS-only system (case 2) and the CCU systems (cases 3 and 4) is rather similar because they include only one of these two system areas. However, the CCUS systems (cases 5 and 6) include both SA 3 and SA 4b. Consequently, the CCUS systems show the highest capital costs.
• Opex contribution per system area Fig. 5b. shows that the Opex of SA 1. NG production and transport increases for the CCU and CCUS systems (cases 3 to 6) respect to the REF and CCS-only systems (cases 1 and 2). NG is used for the production of H 2 , which is included in all systems investigated, but also in the synthesis of CO 2 -DME, which is only included in the utilisation systems (cases 3 to 6). SA 6. Chemicals has the highest share in the Opex of all cases. There are two major contributors to the Opex of SA 6. Chemicals i.e., methanol used as feedstock in the synthesis of conventional DME and PO used as feedstock in the manufacture of polyols. No methanol is needed in the cascade cases since all DME production is based on the CO 2 route. In the parallel cases, only a small amount of MeOH is needed for the production of conventional DME (46 kt/a MeOH to produce 33 kt/a conventional DME, Table 2, Section 2.1). 17% of the PO needed for conventional polyol production is replaced by CO 2 in the CCU and CCUS Chemicals for the parallel and cascade systems are reduced by 34% and 38%, respectively.

Final product costs
The production costs of H 2 , DME and polyol were estimated per case study ( Table 4). Note that different cost allocation factors were used in the case studies as explained in (Section 2. H 2 costs in the CCS case increase by 30% compared to the reference system because the capture costs are allocated to H 2 due to mitigation purposes. The CCU and CCUS cases are driven by utilisation goals, and therefore H 2 costs remain the same as in the reference case without capture. The results indicate that even when including capture expenses, DME costs in the CCU cases are slightly lower than in the REF and CCS cases, indicating that conventional DME production is more expensive than CO 2 -DME production. Although CO 2 -DME has higher Capex due to compression costs, it has lower feedstock costs than conventional DME, which have a larger share in the total costs. In the CCUS systems all CO 2 emitted during CO 2 -DME synthesis is captured and stored (case 5) or partly used and partly stored (case 6). Allocation of capture costs increase DME costs in the CCUS systems by 23-25% relative to the REF case.
The largest variation is found in the final product costs of polyols. CO 2 -based polyol costs in the CCU and CCUS cases, are 14-15% lower than conventional polyol costs in the REF and CCS cases. This shows a clear business case for CO 2 -polyol even when the cost of capture are allocated to the final product as in this study, which is mostly due to replacing part of the PO feedstock with CO 2 . Table 5  • Climate change Contrary to expectations, there is no significant difference among the storage and utilisation cases at system level. Upstream emissions are those related to all the industrial activities from the point of resource extraction to the process were it is used. Upstream emissions in this case correspond to SA 1. NG production and transport and SA 6. Chemicals. They have a significant share of the climate change potential in all systems, especially in the systems without utilisation. Fig. 6 shows the contribution of upstream and process emissions to the climate change indicator. Although the values at the system level are similar, significant differences are found between the different stages (upstream vs. factory gate). The results show that the displacement impacts (e.g., which fossil-based product is replaced) matters. PO is a very energy intensive chemical and thus, PO replacement in polyol synthesis results in large decrease in upstream emissions in all utilisation cases. Factory gate CO 2 -DME climate change potential is higher than conventional DME due to compression requirements and direct emissions. Therefore, process-related emissions of the utilisation systems are larger than the CCS case.

Environmental assessment
At system level, upstream climate change reduction by PO replacement in CO 2 -polyol synthesis is not enough to compensate the increase in climate change potential driven by CO 2 -DME production, relative to the conventional DME route. Consequently, the utilisationonly systems (cases 3 and 4) show slightly higher climate change than the storage-only system (case 2). The systems that combine utilisation and storage (cases 5 and 6) show the largest potential for climate change mitigation, since in addition to PO replacement by CO 2 in polyol synthesis, CO 2 released in DME synthesis is re-captured and stored.
Finally, the results show no significant differences between the cascade and parallel configurations. The extra effort needed to recycle CO 2 in the cascade systems does not show a penalty nor a benefit.

• Fossil depletion
Upstream contributions (SA 1. NG production and transport and SA 6. Chemicals) determine the total fossil depletion in all systems investigated (Fig. 7). Natural gas is used as feedstock in CO 2 -DME synthesis. Therefore, the fossil depletion contribution of SA 1 in all CCU and CCUS systems is higher than in the REF and CCS cases that include conventional DME production.
Lower methanol and PO feedstock requirements in the CCU and CCUS systems are the main cause for a smaller contribution of SA 6. Chemicals. Consequently, all CCU and CCUS cases have lower fossil depletion impact than the REF case. CCUS systems (cases 5 and 6) have a somewhat (3%) higher fossil depletion than the CCU systems (cases 3 and 4) because of the second capture unit.
The CCS system does not benefit from feedstock replacement and therefore it shows the highest fossil depletion among all systems.

Comparative assessment of multiproduct CCU and CCUS systems. Potential and limitations
The aim of this study was to examine the technical, economic and environmental feasibility of multi-product CCU and CCUS implementation and compare their performance with more conventional CCS alone. Each system presents trade-offs between technical, economic and environmental aspects, which highlights the importance of performing integrated assessments. Table 6 summarises the outcomes of the comparative assessment, taking as a reference a H 2 unit at a refinery without carbon capture (case 1.REF). The arrows show whether the indicator increases or decreases with respect to the REF case. Multiple arrows emphasise the magnitude of the increase or reduction. Zero is given when the indicator value is equal to the reference. • Carbon capture and storage. CCS-only system Implementation of carbon capture and storage alone (case 2) is the best option to reduce direct CO 2 emissions in a refinery (−70% direct emissions relative to the REF case). However, compared to the reference case without carbon capture, CCS alone is not economically profitable (+7% total costs) neither environmentally advantageous in terms of fossil depletion.
• Multi-product CCU Multi-product CCU systems (cases 3 and 4) show lower direct CO 2 emissions and climate change than the reference system (case 1) but higher than the storage-only system (case 2). The reason lies in the large emissions during CO 2 -DME synthesis. Therefore, from a climate change mitigation perspective, it is preferable CO 2 storage than utilisation.
Multi-product CCU systems benefit from replacing fossil feedstock by CO 2 . Multi-product CCU systems show the best economic and fossil depletion performance of all systems evaluated. The total costs of the CCU systems are reduced by 9% and 15% with respect to the REF and CCS systems, respectively, and they have the highest NPV of all cases evaluated. In contrast, a positive NPV is not readily attained in other CCU options such as methanol or formic acid using renewable H 2 from electrolysis as feedstock [7,8,6]. The fossil depletion of the multi-product CCU cases is about 12% lower than the REF and CCS cases.
When assessing the parallel vs. cascade configuration, no differences are found in the total costs, climate change and fossil feedstock potential and very minor differences appear in the final product costs of polyol and DME.
• Multi-product CCUS The lowest climate change potential is achieved in the CCUS systems because the CO 2 released in DME production is re-captured and stored (case 5) or partially used and partially stored (case 6). CCUS systems reduce climate change by 18% with respect to the REF case.
Although the total costs and fossil depletion of the CCUS systems is lower than the reference system with no carbon capture, larger energy and materials demand due to the second capture unit, CO 2 compression, Fig. 6. Climate change per system area. Upstream and process contribution. * total system climate change. transport and storage increases the total system costs and fossil depletion of CCUS respect to the utilisation-only (CCU) systems. Furthermore, CCUS systems show the highest DME production costs of all alternatives and the lowest polyol costs (+26% and −15% relative to REF). DME and polyol are produced based on CO 2 in the CCU and CCUS cases, but the allocation of the capture costs changes depending on whether it is stored or utilised.

Conclusions
The goal of this article was to explore the practicability of multiproduct carbon capture and utilisation (CCU) and multi-product carbon capture, utilisation and storage (CCUS) concepts to aid closing carbon cycles in refineries.
To answer this research question, a case study including a hydrogen manufacturing unit with CO 2 capture and utilisation into dimethyl ether (DME) and polyols was defined. Two multi-product CCU and multi-product CCUS systems were evaluated in this research: simultaneous CO 2 utilisation in parallel configuration and 2-cycle CO 2 utilisation in cascade configuration. The techno-economic and environmental performance of these multi-product configurations were compared with a reference case without carbon capture and a case with carbon capture and storage-only (CCS, no CO 2 utilisation).
The integrated techno-economic and environmental assessments carried out in this study show that multi-product CCU and multi-product CCUS systems are cost-effective options that can contribute towards closing the carbon cycles in refineries. The CCU and CCUS parallel and cascade configurations show lower total system costs and environmental impacts than a reference case without carbon capture.
This explorative research points out the importance of performing holistic analysis of CCU and CCUS complex configurations including the CO 2 source, economic and environmental attributes: although the largest direct CO 2 emission reductions are achieved with carbon capture and storage-only (CCS) without CO 2 utilisation, the economic evaluation shows that CCS-only is not an economically viable option. Feedstock replacement by CO 2 is the key to make CCU and CCUS configurations feasible options to reduce CO 2 emissions from the petrochemical sector while having an interesting business case.
The results show trade-offs between the economic and environmental performance in the CCU and CCUS systems. The CCUS systems present higher direct CO 2 emissions reductions than the CCU systems, and the lowest climate change potential among all systems studied. Nevertheless, compared to CCU systems, CCUS present a more limited economic potential and higher fossil depletion due to the incorporation of a second capture unit, CO 2 compression, transport and storage.
This approach also aimed to explore comparative advantages of different multi-product CO 2 utilisation (and CO 2 storage) configurations. The results however show no conclusive differences between the economic and environmental performances of the parallel and cascade configurations. The studied systems are rather complex, and differences at unit levels compensate each other at the system level, and therefore both configurations showed similar costs and environmental impacts.
This prospective study serves to identify cost-effective mitigation alternatives that should be further explored and considered in the research efforts for closing the carbon cycles in the petrochemical industry.

Acknowledgments
This project was carried out as part of the EDDiCCUT project (www. eddiccut.com). EDDiCCUT is a 4-year research project (2012-2016) supported by the Norwegian Research Council under grant number 218952 and by an industrial consortium (Bharat Petroleum, Uniper, and Norske Shell).

Appendix A. Economic evaluation
The raw materials prices were taken from difference sources and updated to € 2016 . The values used in the economic evaluation are in Table A1. The product purchase prices assumed for the estimation of the net present value (NPV) are shown in Table A2.

Appendix B. Life Cycle Assessment Assumptions
The assumptions taken during the life cycle inventory assessment (LCIA) were: -Natural gas production and imports are those of the Netherlands, adopted from the Eurogas report [53].
-Share of onshore/offshore gas production in the Netherlands adopted from [54].
-Given the absence of data and considering that the gas fields for the Netherlands are also located in the North Sea as the Norwegian ones, the LCI for offshore transport in Norway was used for transport of Dutch gas as well. -The lengths of NG pipelines adopted, assuming that the plant is located in the Rotterdam area, are shown in Table B1. -CO 2 transport and storage infrastructure was modelled by adopting the NG pipeline and well infrastructure in the Ecoinvent database. It was assumed that CO 2 is stored offshore in one well of 1000 m depth. The length of the transport pipeline was assumed to be 100 km in total, 5 km onshore to the port and 95 km offshore to the saline aquifer. -The transport of chemicals to plant was modelled according to the model split of inland transport modes of freight transport in the EU-28 given by [56], see Table B2. -For chemicals the 'transport to disposal' and 'disposal phases' are included in the SA that utilizes the chemicals. The impacts of 'production of chemicals' and 'transport to plant' are included in SA 6. Chemicals. The mode and transport distances are adopted from The Ecoinvent Report No.1 [57] where the transport mode is road transport (lorry), the distance for residual material landfill is 15 km, and distance for residual material landfill is 50 km. Due to absence of separate datasets for waste disposal of the chemicals it was assumed that all waste, except zeolite, goes to residual material landfills, where inorganic, industrial waste is deposited [58]. Zeolite disposal is modelled to inert material landfills.
-The amount of CO 2 fugitive emissions during CO 2 transport to storage adopted is 0.01% of transported CO 2 [41].
-The amount of CH 4 fugitive emissions is in the range of 0.2% to 6% of the natural gas produced [41]. In this study, an average of the range (3.1%) was assumed for CH 4 fugitive emissions. Fugitive emissions associated with unconventional shale gas are on average higher than the ones associated with conventional natural gas. Conventional natural gas is utilised in this work. Therefore, the value assumed sets a conservative approach since values lower than 3.1% are found in the literature for CH 4 fugitive emissions from shale gas production and transport [59]. The impacts of CH 4 fugitive emissions (FEI) were calculated using ReCiPe characterization factors [60] following Eq. (B1). The impact categories affected by CH 4 fugitive emissions are CC, POF and FD.
-The electricity mix is that of the Netherlands.
-Due to absence of data for treatment of wastewater as an effluent of the processes in question, the impacts of wastewater are assumed to be those of wastewater as an effluent of soft fireboard production. This process was chosen as it has the median impacts out of a group of class 3, industrial wastewater treatment processes in the Ecoinvent database. -Two components were taken into account for calculating the impact of chilled water: industrial process water and the heat energy (MJ) required for lowering the temperature of water from room temperature (∼20°C) to 10°C. -Due to absence of data, the impacts of MDEA production are modelled as those of MEA production since LCA data for MEA is regarded as proxy for amine-based solvents [26]. In addition, even though the solvent of the H 2 capture unit is a mixture of MDEA and PZ, only the impacts of MDEA are modelled because the ratio of MDEA/PZ is 9/1 [21]. -For γ/Al 2 O 3 the impact of Al 2 O 3 was adopted because it is a different phase of Al 2 O 3 and they are both used as catalysts [61].
-Due to absence of data, the impacts of DMC are modelled as a combination of impacts of the catalyst components: sodium cyanide, Zn and Co.
Sodium cyanide was chosen as a conservative assumption as it has the highest impacts of the cyanides in the Ecoinvent database.
Cora Fernández-Dacosta is a researcher at the Copernicus Institute of Sustainable Development at Utrecht University, the Netherlands. Her research focuses in the design, technoeconomic and environmental assessment of processes to convert waste streams into products. She is a Chemical Engineer with a Professional Doctorate in Engineering, Process and Equipment design from Delft University of Technology.
Viktorija Stojcheva is a Sustainable Industries and Services Consultant at Ecofys. Her work focuses on impact assessments, value chain carbon foot-printing of multinational companies, estimating projected emissions and emission intensity reductions in line with climate science and science based emission reduction target setting for companies. She has a joint bachelor degree in Physics and Environmental Science and a master's degree in Energy Science.
Andrea Ramírez is full professor of Low Carbon Systems and Technologies at Delft University of Technology and associate professor at the Copernicus Institute, Utrecht University. Her research aims at developing methodologies and tools that build bridges between fundamental research at laboratory scale and the insights needed by stakeholders on potential benefits and impacts of these technologies when implemented. Her research portfolio includes projects that assess the sustainability of novel bio-based materials and chemicals, and novel CO2 capture and utilisation technologies.