The system-wide economics of a carbon dioxide capture, utilization, and storage network: Texas Gulf Coast with pure CO2-EOR flood

This letter compares several bounding cases for understanding the economic viability of capturing large quantities of anthropogenic CO2 from coal-fired power generators within the Electric Reliability Council of Texas electric grid and using it for pure CO2 enhanced oil recovery (EOR) in the onshore coastal region of Texas along the Gulf of Mexico. All captured CO2 in excess of that needed for EOR is sequestered in saline formations at the same geographic locations as the oil reservoirs but at a different depth. We analyze the extraction of oil from the same set of ten reservoirs within 20- and five-year time frames to describe how the scale of the carbon dioxide capture, utilization, and storage (CCUS) network changes to meet the rate of CO2 demand for oil recovery. Our analysis shows that there is a negative system-wide net present value (NPV) for all modeled scenarios. The system comes close to breakeven economics when capturing CO2 from three coal-fired power plants to produce oil via CO2-EOR over 20 years and assuming no CO2 emissions penalty. The NPV drops when we consider a larger network to produce oil more quickly (21 coal-fired generators with CO2 capture to produce 80% of the oil within five years). Upon applying a CO2 emissions penalty of 60$2009/tCO2 to fossil fuel emissions to ensure that coal-fired power plants with CO2 capture remain in baseload operation, the system economics drop significantly. We show near profitability for the cash flow of the EOR operations only; however, this situation requires relatively cheap electricity prices during operation.


Introduction
The capture and storage of carbon dioxide (CO 2 ) emissions from point source locations, such as coal-fired power plants, can play a major role in mitigating greenhouse gas (GHG) emissions (IPCC 2005). While the cost of the CO 2 capture is the majority of total system cost for carbon capture and storage (CCS), possibilities for utilization of that captured CO 2 (carbon capture utilization and storage, or CCUS) can create a broader economic picture. One existing example of utilizing CO 2 for economic purposes is CO 2 -based enhanced oil recovery (EOR). CO 2 -EOR has been occurring since 1972 in the Permian Basin of West Texas (NETL). To date, there has been very little use of anthropogenic CO 2 for EOR (e.g., Weyburn-Midale field in Saskatchewan, Canada) due to the lack of development of at-scale CO 2 capture facilities on fossil power plants.
The state of Texas has more coal-fired power plants and consumes more coal than any other state in the US. There are many mature oil fields relatively near these coal-fired power plants. Previous work at the Gulf Coast Carbon Center of the Bureau of Economic Geology categorized many of the mature oil fields in the Texas Gulf Coast as candidates for CO 2 -EOR (Holtz et al 1999, Nuñez-Lopez et al 2008. The present authors have published several previous analyses of the linkages and cash flows for integrating coal-fired power plants with CO 2 capture into the Electric Reliability Council of Texas (ERCOT) grid and using some of the captured CO 2 for EOR , 2009. These previous CCUS studies were limited to either generalizing costs and oil recovery rates over a large number of oil fields or linking one power plant to one EOR field.
For the present analysis, we build upon our previous works to model several coal-fired power plants and ten CO 2 -EOR candidate oil fields (see tables S1 and S2 in the supplemental information, available at stacks.iop.org/ERL/8/ 034030/mmedia) while considering individual heat rates for power plants and specific geologic characteristics of the EOR fields. This letter takes a 'system-wide' look at an entire CCUS system in Texas using: (1) a dispatch model for the actual power plant fleet of the ERCOT grid in which we model selected coal-fired power plants as being retrofitted to capture CO 2 emissions; (2) geologic characteristics of ten mature oil fields in the Texas Gulf Coast that are applicable for miscible CO 2 -EOR; (3) geologic characteristics of deep saline reservoirs above or below the EOR fields (a 'stacked storage' concept) in which all captured CO 2 that is not destined for EOR is injected for permanent storage; and (4) a pipeline network that connects the selected capture facilities with the 'stacked storage' locations.
In this analysis, the 'CCUS system' is composed of the following investments: (1) the capital and operating costs of the retrofitting of amine-based post-combustion CO 2 capture at selected coal-fired power plants; (2) the capital and operating costs of a new CO 2 pipeline network; and (3) the capital and operating costs for wells at stacked storage locations where drilling, production, and injection operating companies produce oil via CO 2 -EOR and inject all excess CO 2 into saline formations.
A couple of studies provide relevant background to this present letter. Ghomian et al (2008) evaluates the economics of CO 2 -EOR value chain in a similar fashion to our analysis. They consider costs of carbon capture, transportation of captured CO 2 to oil field and EOR operations in a discounted cash flow analysis. The most detailed part of their analysis is the reservoir simulations they conduct to differentiate across various reservoirs types, well spacing, and the use of water alternating gas (WAG) or continuous CO 2 injection methods. However, they do not conduct a system-wide analysis to test boundary conditions nor do they model the electricity sector independently. They find that CO 2 sequestration in oil fields would require additional incentives in the oil price environment of the mid-2000s. When they update their analysis with higher prices of 2007-2008, the need for incentives disappears (Ghomian et al 2008).
In another similar techno-economic analysis, Ravagnani et al (2009) evaluate the economics of capturing CO 2 from a fertilizer plant in Brazil, transporting and using it in EOR in a hypothetical oil field (Ravagnani et al 2009). In addition to economics, the authors also investigate energy and emission balances of the CO 2 -EOR value chain. Similar to Ghomian et al (2008) they conclude that with higher oil prices, CO 2 sequestration is economically viable even without CO 2 credits although they also point out that credits, if high enough, would have a significant impact on the project. They also conclude that EOR-sequestration is reducing overall emissions.
Our analysis adds value to the existing literature in that we consider the costs of sequestering CO 2 in saline reservoirs in addition to costs and revenues from EOR operations. Our case study reservoirs within the Gulf Coast of Texas provide relevant context given the possible use of 'stacked storage' where oil can be produced from one geologic layer while storing CO 2 in saline formations above or below that oil reservoir (Ambrose et al 2009, Zahid et al 2012. We describe a range of system-wide economic outcomes for linking multiple existing CO 2 sources and multiple sinks by using bounding scenarios on the deployment rate and extent of a CCUS network. In this way, businesses and government entities understand the range of possibilities regarding the system costs and benefits as well as different business relationships that are possible. Some previous work discusses CO 2 prices and business relationships among entities involved in CCUS for EOR (Agarwal andParsons 2011, Esposito et al 2011). Agarwal and Parsons (2011) discuss how combinations of assumed contract oil and CO 2 sales prices between a power plant and oil field operator shifts the profits from one to the other. Esposito et al (2011) layout how different business models ('self-build', 'joint venture', and 'pay at the gate') provide tradeoffs in risk and reliability between electric power and oil production. Our present work adds value by providing economic information for business entities to understand their role and risk in these or other types of cooperative CCUS business models. We realize that our target system exists effectively as a link between the ERCOT electric market and the (world) oil market. Our scenarios have impacts on electricity and petroleum product prices; they also involve large scale capital investment, which will have multiplier effects on the rest of the regional economy. We do not analyze these larger economic impacts in this research. Table 1 shows the four scenarios that are combinations of two operational scenarios and two economic scenarios. This approach provides a means to explore some bounding cases to provide perspective on different CO 2 price and emissions penalty scenarios such that information exists to contemplate possible business relationships and overall system economics.

Scenario descriptions
The 'slow' and 'fast' operational scenarios describe the rate of oil production from the modeled candidate EOR fields. The 'slow' scenarios are designed such that oil is developed over a 20-year time span and the annual needed delivery of CO 2 is approximately constant. In this sense, the 'slow' scenarios approximate an optimal situation for a coal-fired power plant owner that must invest in CO 2 capture capital equipment and wishes that investment to be fully utilized until fully amortized. The 'fast' scenarios are defined such that all EOR fields drill all wells, install all capital, and begin operations at all wells at the beginning of the 20-year period (within the first three years). In this sense, the 'fast' scenarios approximate an extreme case for capital investment of EOR developers in that they would produce the oil as fast as possible (though not necessarily a realistic business case). These 'fast' scenarios estimate an upper bound on the total quantity of capital investment for the whole system in that much more capital is needed to deliver larger quantities of CO 2 during the first few years relative to all years in the 'slow' scenarios. The 'fast' scenarios necessitate larger capital investment for all three parts of the CCUS network: CO 2 capture, pipelines, and oil/CO 2 storage field operations. For a given injection and production well pattern, our CO 2 -EOR model (using the Prophet program) models some delay in oil production relative to the delivery of CO 2 to allow for buildup of pressure.
The 'CO 2 sales price' and 'CO 2 emissions penalty' economic scenarios capture two commonly discussed options for reducing CO 2 emissions with different dynamics. They represent internal (CO 2 sales) or external (CO 2 emissions penalties) economic drivers to encourage investment in the CCUS network. In the 'CO 2 sales price' scenarios, there is no penalty (e.g. tax) for emitting CO 2 to the atmosphere, and the 'CO 2 sales price' is that price that an EOR operator pays to the coal-fired power plant for CO 2 . Thus, the customer that buys electricity and/or oil does not (in large) notice the effect of this CO 2 sales transaction (via impacts on electricity or oil prices) that is internal to our modeled CCUS network. On the other hand, the 'CO 2 emissions penalty' cases represent the impact upon the modeled CCUS system of a penalty for the CO 2 emissions from burning fossil fuels consumed (for electricity) or produced (for oil) by our modeled CCUS system. Thus, an electricity or oil products (e.g. gasoline) customer would more directly notice price increases in the 'CO 2 emissions penalty' cases, but we do not model any price-related feedbacks on energy demand.

CO 2 -enhanced oil recovery modeling and cost assumptions
To estimate the CO 2 injection and oil production profiles for an inverted five-spot pattern 6 , we used CO 2 -Prophet, a CO 2 flood prediction software that generates streamlines between injectors and producers and estimates displacement and recovery through a finite difference routine. One of the motivations of this work is to consider economic scenarios handling large volume flow rates of CO 2 , so our EOR modeling in Prophet is based upon continuous CO 2 injection, not the more common EOR practice of WAG where both water and CO 2 , the 'gas', are injected through the same wells in an alternating manner over the life of the well. The choice of EOR candidate fields comes from a coauthor's previous assessments and represents fields with a range of sizes and qualities (Nuñez-Lopez et al 2008). The fields are not specifically chosen to be the most or least profitable prospects. In addition, four of the EOR fields are in south Texas near the Corpus Christi area, and the other six EOR fields lie closer to Houston. Thus, we have included some geographic diversity to consider the connectivity of coal-fired plants in different regions of the state (see figure 1). Table  S1 of the supplementary information (available at stacks.iop. org/ERL/8/034030/mmedia) lists the EOR fields, along with the number of wells (oil production, CO 2 injection, and water injection) assumed for each field.
The assumed profiles for CO 2 and oil flows are based upon recent CO 2 -EOR activity in a Texas Gulf Coast oil field (Davis et al 2011). This particular oil field development is somewhat unique in that it employs continuous CO 2 injection, just like we do in our modeling. We do model some water injection wells, but these water injection wells are meant to create a 'curtain' of water and pressure at the edge of the displacement area that minimizes CO 2 loss from the targeted part of the reservoir (Davis et al 2011).
Based on limited CO 2 -EOR experience in the US Gulf Coast, specifically related to projects where only CO 2 is continuously injected (not WAG), our main assumptions for EOR are that (i) injection continues until either 20 years is reached or a total quantity of CO 2 injected is an amount equal to 500% of the hydrocarbon pore volume (HCPV) of each reservoir (Davis et al 2011), (ii) approximately 12-16% of the original oil in place (OOIP) is recovered from each field, and (iii) that CO 2 injection occurs at a pressure that is 90% of the fracture pressure of the reservoir. For each well pattern, the injection pressure assumption models the oil production in a close to 'as fast as possible' manner. By staggering the dates to start each pattern of wells, we can design various plans for the development for the entire oil field over time. The supplemental information (available at stacks.iop.org/ERL/8/034030/mmedia) details the capital and operating costs models for well drilling and operation of CO 2 injection and oil production wells. Table S2 of the supplementary information (available at stacks.iop.org/ERL/ 8/034030/mmedia) lists the oil recovery, CO 2 delivery, and time period of oil production for each EOR field.
Major assumptions for assessing the capital and operating costs for CO 2 -EOR are as follows (see the supplemental information for more details).
• Drilling costs. Drilling costs of both oil production and CO 2 -EOR injection wells are assumed at 50% of the cost of drilling a new well. This capital cost assumption translates to our assumed use of existing oil production and water injection wells at the EOR candidate fields and that reworking and/or side-track drilling the wells is estimated at half the cost of a new well. Analysis of data in the Joint Association Survey on Drilling Costs supports the idea that wells drilled by side-tracking cost 40%-80% of the cost of a newly drilled well.
• Lease costs. Lease costs are a combination of costs for primary oil production and additional lease costs for CO 2 -EOR. We largely follow the method described in ARI (2006) as outlined in the supplemental information.
• CO 2 recycling capital. We assume the full capacity capital costs ($700 000 per peak million cubic feet per day of recycled CO 2 ) of a CO 2 recycling plant at each EOR field occur during the year that CO 2 recycling is first required. We do not assume the CO 2 plant is upgraded or expanded at any time but it is installed a full capacity even if full capacity is not needed until several years later.
• Electricity and lifting. The quantity of electricity for compressing and pumping CO 2 follows the McCullom and Ogden method for CO 2 compression to supercritical and higher pressures as needed for EOR operations (McCullom and Ogden 2006). Other lifting electricity needs come from (ARI 2009).
• General and administrative costs. Additional costs administrative costs are added equal to 20% of lifting plus both water and CO 2 injection costs (per ARI 2006).

CO 2 storage in saline reservoirs modeling and cost assumptions
Major assumptions in assessing the capital and operating costs for CO 2 injection into saline reservoirs are as follows (see the supplemental information for more details).
• Drilling costs. The capital drilling costs for CO 2 saline injection wells are assumed the same as the full cost of drilling a new oil well.
• Lease costs (capital and O&M). Both capital and O&M lease costs are assumed equal to the additional lease costs for CO 2 -EOR.
• CO 2 pumping electricity. The CO 2 arrives to the saline injection site in supercritical condition in the pipeline, and additional CO 2 pumping electricity is assumed at 5 kWh/BBL of supercritical CO 2 (ARI 2009).

Pipeline network configuration modeling assumptions
Because the foci of this analysis are the coal-fired power plants with CO 2 capture and EOR/CO 2 storage operations, each scenario uses a relatively simple pipeline network assumption that is not a complete analysis of the requirements for operating the pipeline network. We do not calculate needs for booster pumps or specifically estimate pipeline operating electrical demands. For the construction costs of the pipelines, and determination of the diameter of the pipeline for a given CO 2 flow rate, we follow previous methods (Herzog and Javedan 2010). The sizing and capital cost model for the pipeline is detailed in the supplemental information. Figure 1 shows the conceptual layout of the pipeline network for both the 'slow' and 'fast' scenarios.

ERCOT electric grid dispatch and CO 2 capture at coal-fired power plants
This work uses a first-order electricity dispatch model to estimate the wholesale electricity price and dispatch of power plants on the ERCOT grid. In this model, the fixed and variable operating costs for each power plant in ERCOT are used as a basis for economic dispatch. At a given load, these operating costs determine the marginal operating power plant such that all power plants with higher operating costs are not dispatched. We do not model transmission expansion or transmission constraints. Previous work has used the same model to explore the implications of flexible CO 2 capture operations in the ERCOT and Great Britain grids (Cohen 2009, Ziaii et al 2008. Please refer to these publications for more detail. We do not impose retirement or fuel conversion of any ERCOT power plant over the 20-year time horizon of our scenarios. Coal-fired power plants that are likely to sign long-term contracts to provide CO 2 to EOR operators are not likely to retire. ERCOT includes many coal-fired power plants that we did not model with CO 2 capture, and these units not modeled in our scenarios are those that are (i) older, (ii) less efficient, and (iii) without as many criteria emissions controls. Thus, we chose coal-fired power plants that would be most likely to capture CO 2 within the scope of our scenarios. Most of the coal-fired generating units we model with CO 2 capture are less than 40 years old. Clearly, there is a risk of new regulations (e.g., on mercury emissions) that could force these plants to shut down. It is most likely that some coal fired and other power plants will retire during the time frame of our analysis, but considering these changes is outside of the scope of this present work.
We assume the demand for electricity in ERCOT is the same in each of our system scenarios. We ignore impacts on demand of changes in electricity price due to CO 2 sales revenues or CO 2 taxes. The ERCOT annual load for the first year (2012) is 328 TWh, and we forecast it to increase at approximately 2.3% yr −1 to 523 TWh in 2031 (ERCOT 2011).
For each scenario, we specifically chose the CO 2 sales price and CO 2 emissions penalty to equal the approximate level that makes the chosen subset of coal-fired power plants operate at baseload capacity factors. In other words, if the CO 2 sales price were lower, all of the chosen coal-fired power plants would operate at lower than baseload capacity over the 20-year simulation horizon, and if the CO 2 sales price were higher, they would not operate at any significantly higher capacity factor. The same rationale was used to set the level of the CO 2 emissions penalty: the penalty was chosen such that it was just at the level to enable our chosen subset of coal-fired power plants to operate at baseload conditions for the 20-year time horizon. Table S3 (available at stacks.iop.org/ERL/8/034030/ mmedia) of the supplemental information shows the chosen subset of existing ERCOT coal-fired power plants that were modeled with retrofitted amine-based post-combustion CO 2 capture. The amine capture unit is assumed to require Figure 2. The fossil fuel prices assumed for the models and cash flow analysis affect electricity generation dispatch, costs, and revenues, as well as revenues and costs for EOR oil production. Low sulfur light crude oil price is used. 0.27 MWh/tCO 2 to remove 90% of the CO 2 from flue gas, which translates to a 22-32% net power output penalty depending on the base plant efficiency (Ziaii et al 2008). Capacity lost to CO 2 capture energy requirements could be replaced by new generating units, demand response, or increased efficiency and conservation. Examining these options is outside the scope of this work, but other literature suggests replacement capacity might not be necessary if capture systems could be bypassed during infrequent peak demand periods. With the assumed electricity demand and power plant fleet, sufficient capacity exists to substitute for reduced coal-based generation (Chalmers et al 2009, Cohen et al 2010. For the 'slow' scenarios, fewer coal-fired generators need CO 2 capture facilities because less CO 2 is needed per year than in the first few years of the 'fast' scenarios. Thus, in the 'slow' scenarios, only the following three electric generation units are assumed to capture CO 2 : Fayette Power Project unit 1, J K Spruce unit 2, and WA Parish unit 7 (see table S3).

ERCOT grid dispatch
For the purposes of this analysis, the marginal generation cost from our dispatch model approximates the wholesale cost of electricity in ERCOT. The assumptions for fuel prices (see figure 2) and other operating costs drive the upward electricity price trends in figure 3. The projected fuel costs were taken from the EIA Annual Energy Outlook (AEO) 2011 (AEO  table 1  We present the modeled CO 2 emissions from the ERCOT electric grid (see figure 4) during the 20-year time span of our scenarios to create context for the large flows of CO 2 by our subset of coal-fired facilities. These CO 2 emissions originate primarily from coal and natural gas combustion, with minor quantities from other fossil fuels. The total modeled 20-year CO 2 emissions from the grid for scenarios 1-4 are 4300, 2800, 3200, and 2400 MtCO 2 , respectively. The baseline (business as usual) modeled emissions for the grid with no CO 2 capture and no CO 2 emissions penalty is 4500 MtCO 2 over the 20-yr time span.
Figures S1(a)-(d) (available at stacks.iop.org/ERL/ 8/034030/mmedia) show the electricity generation mix calculated for each scenario. Note that in all scenarios we assume that the demand for electricity in ERCOT is the same each year (e.g. the quantity of electricity generation that consumers purchase is not affected by prices). This assumption keeps the scenarios simple for straightforward comparison. In reality, the scenarios with $60/tCO 2 emissions penalty have significantly higher wholesale prices of electricity that would reduce demand to some degree.
Due to the variations in marginal price of electricity for each scenario and the assumption of inelastic electricity demand, the total operating profits earned by all generators in ERCOT varies considerably (see figure 5). Here we define operating profits as 'operating revenues minus operating costs', and revenues to generators in ERCOT come from selling electricity. Revenues also come from CO 2 sales for those coal-fired power plants with CO 2 capture in the 'CO 2 sales price' scenarios 1 and 3. Over the modeled 20-year time span, the operating profits for scenarios 1-4 are 96, 181, 107, and 179 billion $2009, respectively. We can reasonably deduce that total ERCOT operating profits are the same between each 'CO 2 sales price' scenario and between each 'CO 2 emissions penalty' scenario.
3.2. CO 2 enhanced oil recovery Figure 6 shows the oil production and net CO 2 delivered for EOR. One can clearly visualize the front-loaded nature To conceptualize the quantity of CO 2 captured in each scenario, we show two baseline results for comparison in which no CO 2 is captured. 'Baseline: no emissions penalty, no sales price, no CO 2 capture' compares to scenarios 1 and 3 in which there is no emissions penalty. 'Baseline: $60/tCO 2 emissions penalty with no CO 2 capture' estimates the emissions from ERCOT generators when an emissions penalty exists but no generators have CO 2 capture. Figure 5. The operating profits (all revenues from CO 2 and electricity sales minus all operating costs) for each scenario for all electricity generators in ERCOT show a considerable jump for the 'CO 2 emissions penalty' scenarios because of the assumption that consumers will not lower consumption at higher electricity prices. of investment for scenarios 3 and 4 where drilling and oil operations occur as early as possible. Our analysis ends after 20 years, but according to our field-specific modeling, there would still be significant oil production after 20 years for all scenarios, especially in scenarios 1 and 2. The total oil production within 20 years is 350 MMBBL for scenarios 1 and 2 and 480 MMBBL for scenarios 3 and 4. There are slight differences in the oil and CO 2 delivery profiles between scenarios 3 and 4, and these minor differences are driven by the modeled rate of CO 2 capture from the electricity dispatch model. In scenario 3, the CO 2 sales price affects dispatch of the coal-fired plants with CO 2 capture slightly differently than in scenario 4 with the CO 2 emissions penalty. The differences are minor and insignificant when interpreting cash flow results. Table 2 lists non-discounted capital, and operation and maintenance (O&M) costs per barrel (BBL) of produced oil, over 20 years, for each of the ten EOR fields and for the aggregated system. Different values exist for scenarios 1 and 2 versus scenarios 3 and 4 because our EOR cost model calculates capital and O&M costs as a function of the assumed exogenous oil price, the drilling sequence is significantly different in the 'slow' versus 'fast' scenarios, and more oil is produced in the 'fast' scenarios 3 and 4 compared to 'slow' scenarios 1 and 2. Table 2 summarizes capital and operating expenses for our EOR cash flow model. All values are reported in units of total (non-discounted) dollars per barrel of oil produced within our 20-year scenario. The capital costs are generally higher compared to (Kheshgi et al 2010) that report a range of 3-9$/BBL capital expenditures. This discrepancy is consistent with the fact that our $/BBL calculation is based on only 20 years of production, does not include the full expected lifetime oil production from each well pattern, and is not based on the quantity proved reserves associated with the capital investment. The O&M costs in table 2 do not include the cost of an oil producer purchasing CO 2 , and the costs are all significantly higher than the ranges quoted in Kheshgi et al (2010) of 6-15$/BBL.
The major parameter that affects operating costs is the cost of electricity to an operator. We assume a constant electricity price of 0.05$ kWh −1 as a lower bound on electricity price. Two higher electricity price assumptions, equal to an industrial and residential price, provide medium and upper bounds on the electricity price. The industrial and residential prices are equivalent to the ERCOT average wholesale cost plus 3 and 7 cents kWh −1 , respectively. Even in the lowest-price case, the total O&M costs per field are 18-38$/BBL. The main reason for our higher O&M cost is that we are modeling a CO 2 -only EOR operation, instead of WAG, so a much higher quantity of electricity is required to recycle and recompress a larger flow rate of CO 2 per quantity of produced oil. For perspective on the quantity of electricity needed for our modeled scenarios we consider only the separately modeled electricity for three purposes: CO 2 Figure 6. (a) The oil production for each scenario. (b) The quantity of CO 2 delivered specifically for EOR for each scenario. recompression during recycling, CO 2 pumping for injection for EOR and saline sequestration, and lifting of CO 2 and oil during EOR. These three needs peak at 12, 16, and 17 TWh yr −1 for scenarios 1 and 2 (year 19), scenario 3 (year 7), and scenario 4 (year 8), respectively. These peak annual electricity needs are significant on the scale of ERCOT as they equate to approximately 2.3%, 4.1%, and 4.3% of modeled ERCOT electricity demand in each of the respective peak years of need.
3.3. Results scenario 1: 'slow' EOR development, CO 2 sales price and scenario 2: 'slow' EOR development, CO 2 emissions penalty In order to approximate the delivery of CO 2 for EOR at a constant rate, we aggregated each well pattern into phases of operation equal to 1/6 the total number of wells needed to develop each field. For example, we assume that there will be 211 production and CO 2 injection wells sidetracked in the Conroe field. Thus, each phase of drilling involves approximately 35 wells. In this way we stagger the operation of each phase as necessary to approximate a need for constant CO 2 delivery from the coal-fired power plants.
Staggering drilling phases is done to optimize for constant CO 2 delivery, not profitability (i.e. the more profitable fields are not prioritized first with the less profitable fields later) (see figure 7). This approach is consistent with our system-wide scenario framework where we seek to understand bounding cases for a CCUS network. The cash flows for scenarios 1 and 2 are not significantly different for any ordering of EOR development phases, particularly within the accuracy of our capital and operating cost estimates. The oil production increases to 14 MMBBL yr −1 by year 4 and stays between 17 and 21 MMBBL yr −1 from years 5 to 20 (see figure 8). For the 'sales price' scenario 1, the price starts at 15$/tCO 2 in 2012 and increases linearly to 22$/tCO 2 in 2031, year 20 (see figure 8(a)). This CO 2 sales price is relatively small compared to the oil price, with price ratio of ($/tCO 2 ):($/BBL) of approximately 0.18:1 (or 3.4$ per thousand cubic feet (Mcf) of CO 2 per $/BBL of oil). This sales price largely offsets the increased operating costs of the coal-fired power plants with CO 2 capture and does not materially affect the marginal generation cost of the overall ERCOT grid.
Scenario 2 differs from scenario 1 in that there is an emissions penalty imposed upon CO 2 emissions from EGUs and from assumed combustion of produced oil (at 0.42 tCO 2 /BBL). There is also no CO 2 sales price that EOR operators pay to coal-fired power plants. In effect, the CO 2 is delivered to the EOR fields and saline sequestration sites 'for free' to the field operators. This assumption of 'free CO 2 ' is a simplification based upon the choice of the emissions penalty price. We do charge EOR and saline sequestration operators a pipeline transport fee (0.8$/Mcf = 15.2$/tCO 2 ).
For the 'emissions penalty' scenario 2, we determined that a constant penalty of 60$/tCO 2 each year incentivized the three coal-fired generators with CO 2 capture to operate (considering only electricity sales for revenue) at near baseload and capture a similar amount of CO 2 as in scenario 1. Because this emissions penalty induces the coal-fired power plants to operate while capturing CO 2 , those coal-fired power plants in ERCOT that do not capture CO 2 are at an economic disadvantage having to pay for CO 2 emissions. Economic dispatch modeling does not yet include capital costs for CO 2 capture infrastructure that we include in the system-wide cash flow.
The quantity of CO 2 emissions from the modeled system is shown in figure 8(b). Over 20 years, 240 Mt CO 2 are captured from coal combustion and sequestered in EOR and saline reservoirs, 146 MtCO 2 are emitted from the oil Figure 7. For 'slow' scenarios 1 and 2, the net amount of CO 2 injected (equal to CO 2 delivered for EOR) at each EOR field is distributed in time by assuming that each field is developed in six phases. The six phases at each field are started at different times in order to approximate a constant need for CO 2 delivery that closely matches the approximately constant rate of CO 2 /yr captured at three base load coal-fired generation units. produced from the CO 2 -EOR operations, and 27 MtCO 2 are emitted from the power plants with capture. Thus, our modeled system stores a net amount of 66 MtCO 2 .
3.4. Results scenario 3: 'fast' EOR development, CO 2 sales price and scenario 4: 'fast' EOR development, CO 2 emissions penalty For the same CO 2 sales price trajectory as scenario 1, more base load coal-fired power plants are assumed to engage in CO 2 capture and deliver CO 2 for EOR to the same ten EOR fields but at a much faster rate. The EOR production is assumed to occur at a maximum rate where all phases of oil field operations engage in EOR operations within the first three years. This scenario is meant to be an extreme case indicating the shortest timescale for EOR production. CO 2 capture operations continue after EOR production, necessitating that much higher flow rates of 'excess' CO 2 (compared to scenarios 1 and 2) are captured and injected into saline reservoirs. Figure 9 shows the net total CO 2 /yr, or delivered CO 2 /yr, for EOR operations in the 'fast' scenarios Figure 9. In the 'fast' scenarios 3 and 4, the net amount of CO 2 /yr injected (equal to CO 2 /yr delivered for EOR) at each EOR field over time is front-loaded as if all wells begin operations within the first 3 years of the analysis period.
3 and 4. The net CO 2 delivered for EOR has a slightly different profile between scenarios 3 and 4 that accounts for the different amount of CO 2 captured in early years. This minor difference exists because coal-fired power plants are dispatched differently between the 'CO 2 sales' (scenario 3) and 'CO 2 emissions penalty' (scenario 4) scenarios.
Scenario 4 has the same CO 2 emissions penalty as scenario 2. The number of oil production and CO 2 injection wells completed each year is approximately the same as scenario 3. In scenarios 3 and 4, the field location of the saline injection wells occurs approximately in proportion to the amount of modeled oil production at each field (with some adjustment as needed). There are 566 more saline CO 2 injection wells needed in scenario 4 than in scenario 3 because the smaller oil fields (e.g. not Hastings and Conroe) start production 1 or 2 years later than in scenario 3 to better match the CO 2 delivery profile. Figure 10 shows the annual oil production for scenarios 3 and 4, as well as the assumed oil and CO 2 prices. To continue utilizing the capital investment in CCUS infrastructure, we assume the coal-fired power plants with capture continue to capture CO 2 each year even if it is no longer needed for EOR. Figure 11 shows that the quantity of CO 2 flows for scenarios 3 and 4 are nearly the same, with 80 and 75 MtCO 2 yr −1 , Figure 8. (a) The EOR oil production follows the delivered quantity of CO 2 . The oil price taken from the EIA AEO 2011 increases over time from 85$/BBL in year 1-124$/BBL in year 20. (b) Over the 20-year span of the modeled scenarios 1 and 2 the total CO 2 emissions captured from the three coal-fired EGUs is approximately equal to that needed for EOR in the ten oil fields. Figure 10. For the 'fast' scenarios 3 and 4, there is a much larger quantity of CO 2 captured to serve the demand for EOR in the first two years. The assumed commodity prices (oil, CO 2 , natural gas, coal) are the same in scenarios 3 and 4 as in scenarios 1 and 2.
respectively. For scenario 3 and 4, respectively, 1540 and 1450 MtCO 2 of coal emissions are sequestered in EOR and saline reservoirs, 200 MtCO 2 are emitted from the oil produced from the CO 2 -EOR operations, and 170 and 160 MtCO 2 are emitted from the power plants modeled to have CO 2 capture. Thus, our modeled system stores a net amount of 1090-1170 MtCO 2 over 20 years.

Integrated cash flow analysis
We developed an integrated discounted cash flow model for the CCUS system described above. We include capital investment in (i) CO 2 capture at several Texas coal plants, (ii) development of a pipeline network to transport the captured CO 2 to stacked storage locations for EOR and sequestration, and (iii) development of the stacked storage EOR and sequestration facilities. There are no revenues realized by the saline operations; saline costs are assumed to be covered by EOR operators. Our goal is not to evaluate commercial viability of each segment but rather to investigate viability of the overall system. Nevertheless, we impose a 12% rate of return on the pipelines as these would likely be subjected to some rate control if they are accessible to third parties. If pipelines are dedicated to either power plants or EOR operators or both, their costs would be mostly internalized. In any case, the pipeline segment represents a relatively small portion of operating costs (5% in slow scenarios and 14% in fast scenarios) 7 and a small portion of capital costs in the system (8-10%).
In the 'CO 2 sales' scenarios 1 and 3, the EOR operators buy the CO 2 from the coal-fired power plants that capture it and then pay the pipeline company for transporting the CO 2 to the oil fields. In the 'CO 2 emissions penalty' scenarios 2 and 4, the EOR operators are only burdened by the cost of transporting CO 2 .
In the 'CO 2 emissions penalty' scenarios, we can consider two system boundaries. One boundary does not penalize CO 2 emissions from oil combustion because these emissions can be considered to occur outside our CCUS system where oil is refined and consumed as refined products (e.g. gasoline). We also present cash flow results using a second system boundary that includes the cost of emissions from using the EOR oil. These oil emissions assume 0.42 MtCO 2 emissions per barrel of oil consumed at 60$/tCO 2 , the same emission penalty used in the power dispatch model. Table 3 shows key input values for the integrated cash flow analysis of the CO 2 -EOR/sequestration value chain. The EOR cash flow model does not include operating costs for capture facilities because these costs are already included in the marginal cost calculations for power plants in our economic dispatch model of ERCOT. In that model, the power plants are dispatched only if revenues from electricity sales recover standard operating costs and the operating cost of CO 2 capture.
CO 2 sales and emission penalties produce different operating costs, especially in the 'fast' scenarios, because 7 The share is much higher in 'fast' scenarios because EOR operations have produced more than 70% of the oil after 10 years but pipelines continue to transport CO 2 to fields for sequestration purposes. Figure 11. Both scenarios 3 and 4 have very similar total CO 2 emissions captured from the 21 EGUs at 13 coal-fired power plants with CO 2 capture. Figure 12. (a) No scenarios present a system-wide CCUS network that has a positive net present value. (b) When considering only the EOR operations, the oil production is profitable in three cases, none of which assume a CO 2 emission penalty applies to emissions from oil combustion.
emission penalties have a more indirect impact on power plant economics (via raising operating costs) than selling CO 2 , which provides direct revenues to the coal-fired power plant. Accordingly, the dispatched amount of electricity, mix of generators on the ERCOT grid, and CO 2 emissions change significantly between the 'CO 2 sales price' and 'CO 2 emissions penalty' scenarios. The difference in ERCOT dispatch is less drastic between the 'slow' and 'fast' scenarios for a given 'CO 2 sales price' or 'CO 2 emissions penalty' scenario.
In addition, the financial assumptions are the following: 10% discount rate, 10-year loans at a rate of 6% (2.5% of the loan amount is paid as the up-front fee), and 0.6% interest during construction for capture plant and pipeline capital. Capital expenses in capture facilities and pipelines are realized over three years (20% first year, 60% second year, and 20% third year) and partially financed via borrowing (50% for capture facilities and 60% for pipelines).
EOR and saline sequestration expenditures are treated differently. In slow scenarios, 50% of annual capital expenditures in EOR facilities are financed for the first six years when more than 80% of total capital is spent; the remainder is not financed because relatively small amounts are needed per year for the following 14 years. In slow scenarios, there is no financing for saline infrastructure because total investment is realized via sporadic and relatively small installments over 20 years. In fast scenarios, 50% of all capital expenditures are financed since all of it is realized within the first three years; in contrast, 50% of annual saline capital expenditures are financed for the first 10 years when more than 80% of total capital is spent; the remainder is not financed because relatively small amounts are needed per year for the following 10 years.
Finally, we assume that 70% of EOR capital costs and 100% of saline capital costs can benefit from intangible drilling cost tax deductions. This is a fairly standard deduction for risky upstream oil and gas projects. There is not much history with application of such deductions to saline operations but we assume all of these costs could be eligible for intangible drilling cost allowances because the sequestration operation does not generate any revenues for the EOR operator. For most cases, these deductions provide marginal benefits in our model and are inconsequential in terms of changing an NPV from negative to positive (or an internal rate of return, IRR, from less than 10% to higher than 10%). The sole exception is when considering joint 'EOR + saline storage' operations under scenario 4 ('fast' scenario with CO 2 penalties) without emissions penalties on oil and industrial electricity rates.
3.6. Net present value summary 3.6.1. NPV summary (whole system). Figure 12(a) plots the net present value (NPV) of the CCUS network. In no modeled scenario does the total system have positive NPV (table 4). The 'slow' scenarios have higher NPV than the 'fast' scenarios. These differences in NPV are because the 'fast' scenarios have a large increase in CCUS capital expenditures to add capture to additional coal-fired power plants, drill significantly more CO 2 injection wells, and concentrate EOR capital expenditures in the first three years rather than spread them out over 20 years. Because of the electricity required to pump CO 2 at the EOR and saline sequestration sites, electricity price is also very influential in affecting the total NPV.
There is a difference between sales and emission penalty cases mainly because the CO 2 emissions penalties have a significant impact on profitability of coal-fired power plants.
To have an internally consistent analysis, the CO 2 emissions penalty is applied to all generating plants using fossil fuels (coal, natural gas, and oil). Thus, our emissions penalty scenarios favor natural gas dispatch over coal generation within ERCOT (renewable generation is already favored in all scenarios due to low operating costs). Recall that we have specifically chosen the penalty of 60$/tCO 2 as level that maintains base load generation for the coal-fired plants with CO 2 capture. As we have chosen only a subset of coal-fired power plants to capture CO 2 (those delivering CO 2 to EOR fields), all other coal-fired power plants shift higher in the dispatch order and earn less operating revenue (see figure S1 of the supplementary information). One way to interpret the system costs is by calculating the present value cost of CO 2 storage as the NPV divided by the total CO 2 captured and stored (see table 4). This cost of storage is an estimate of the additional money needed to make the system break even. The NPV storage costs range from 5 to 25$/tCO 2 . The cheapest values are for 'CO 2 sales' scenarios and the most expensive costs are for 'CO 2 emissions penalties' scenarios when oil emissions are also penalized. The highest cost of storage per tCO 2 occurs for scenario 2, the slow scenario with oil CO 2 emissions penalized. For scenario 4, the costs with oil CO 2 emissions penalized are only 3$/tCO 2 higher than without an emissions penalty for oil.
3.6.2. NPV of each segment of CCUS system. Given our assumptions, the analysis does not attempt to target a certain return on investment for the individual business segments (except for pipelines) within the CCUS network. It is, however, still insightful to understand the differences in economic value and returns generated across the segments of the system (tables 5-7). The pipeline will generate positive NPV in all cases since we impose a 12% return restriction on that segment (table 6); the pipeline achieves this return at a transportation tariff of about 80 cents per Mscf ($15/tCO 2 ) in the slow scenarios and just under 40 cents per Mscf (8$/tCO 2 ) in the fast scenarios.
Coal-fired power plant operating profits from CO 2 and electricity sales from the power dispatch model are not high enough to cover the large capital investment in these facilities.
Thus, the CO 2 capture investment does not break even in any scenario (table 5). Although we obtain positive IRRs for the 'CO 2 sales' scenarios (based on cumulative revenues of $4.8 billion in scenario 1 and $30.8 billion in scenario 3), they are below investment levels as indicated by negative NPVs and IRRs below our assumed 10% discount rate. Revenues are not realized soon enough in the plant life to recover large investments made in capture facilities in the first 3 years of the project life (including interest, $3.2 billion in scenario 1 and $20.7 billion in scenario 3). The emission penalty cases have much lower NPV, as there are no CO 2 -based revenues to capture facilities, but they incur the same capital investment costs.
Focusing only on the NPV of the EOR operations, there are four cases where NPV is positive (see figure 12(b)). These four cases are (from highest to lowest NPV): (1) 'fast' CO 2 emissions penalty (scenario 4) with no emissions penalty on oil and a 0.05$ kWh −1 electricity price, (2) 'fast' CO 2 emissions penalty (scenario 4) with no emissions penalty on oil and an industrial electricity price, (3) 'slow' CO 2 emissions penalty (scenario 2) with no emissions penalty on oil and a 0.05$ kWh −1 electricity price, and (4) 'fast' CO 2 sales (scenario 3) and a 0.05$ kWh −1 electricity price, (there are no emission penalties in 'CO 2 sales' scenarios).
The EOR operations can be profitable at low electricity costs. Considering EOR operations only, the present day situation in the United States is best represented by the 'CO 2 sales' scenarios that assume EOR purchases CO 2 and there are no emissions penalties. Yet only the 'fast CO 2 sales' scenario 3 has positive NPV (and 20% IRR) when assuming cheap electricity. However, the 'slow CO 2 sales' scenario 1 has a 10% IRR (and practically a zero NPV) with cheap electricity. Lower CO 2 sales prices will improve economics of these scenarios and may lead to positive NPV and 10% or higher IRR even at higher electricity prices. For example, a 40% reduction in assumed CO 2 price yields an NPV of zero and an IRR of 10% for scenario 1 with industrial electricity prices; in contrast scenario 1 with residential electricity prices does not yield positive NPV even when CO 2 is provided free. A 10% reduction in CO 2 prices is sufficient for a zero NPV in scenario 3 with industrial electricity price, but even free CO 2 does not yield a positive NPV for help scenario 3 with residential electricity prices. The 'CO 2 emissions penalty' scenarios 2 and 4 that do not internalize costs of CO 2 emissions from the EOR oil have the highest NPV of all cases. The EOR cash flows for scenarios 2 and 4 are profitable assuming the low electricity price, even when including saline storage costs.
Integrating the cost of saline sequestration into the EOR operator cash flow prevents the 'EOR + CO 2 sequestration' operations from being profitable at industrial or higher electricity costs, especially for the 'fast' scenarios (3 and 4). This result is expected as CO 2 sequestration, a pure cost activity for the operator, would be driven by anticipated benefits outside of our modeled system, just as with any general CO 2 mitigation activity. Without emission penalties on oil, saline operations have an insignificant impact at low electricity prices for scenario 2 (22% versus 20% IRR) but are more striking for scenario 4, reducing NPV from $4 billion to $0.4 billion and decreasing IRR decreasing from 47% to 22%.
3.6.3. NPV of EOR with CO 2 emissions from oil internal to system boundary.
If we consider consumption of oil products refined from the EOR production inside our CCUS system (e.g. passed on to the consumer), the CO 2 emission penalty on consumed EOR oil would have significant impact. We considered system NPVs with and without such penalties to guide our future work. Our modeled CCUS system generates negative NPVs and very low IRRs in the best cases even without emissions penalties on oil products (see table 4). In the 'slow' scenarios, the NPV drops significantly lower and there is no positive IRR at any electricity cost. In the 'fast' scenarios NPVs also worsen.
Adding emission penalties on oil produced from EOR operations deteriorates EOR economics in scenarios 2 and 4 (see table 7). Including emission penalties on EOR-produced oil reduces EOR IRR from 22% to negative in the slow case with low electricity price (scenario 2) and from 47% to negative in the fast case with low electricity price (scenario 4). Combining saline storage costs into EOR operations, 'EOR + saline storage', lowers NPV and IRR further relative to considering 'EOR only', especially for scenario 4. NPVs become negative for all scenarios applying emissions penalties on oil products with or without saline operation costs.

Future work
Our current analysis provides a valid conceptualization of the 'CO 2 sales price' scenarios because the sale of CO 2 from the coal-fired power plant to the EOR operator is an internal transaction. However, because we do not consider any indirect economic impacts outside of the modeled system in the 'CO 2 emissions penalty' scenarios, a fuller interpretation is needed via an analysis that can consider these indirect impacts. A primary indirect impact to consider is the effect of higher energy (electricity and oil) prices on lowering consumer demand. Thus, future work will more effectively compare the 'CO 2 sales price' (that impacts only the parties involved in our model) and the 'CO 2 emissions penalty' (which impacts entire economy) scenarios. This approach will necessitate enlarging our scope from our well-defined CCUS system to the rest of the economy, especially the power sector and consumers of oil products. Further calculations can also estimate the necessary spacing of injection wells storing CO 2 in saline reservoirs such that we can incorporate the associated capital costs of a distribution network. Recent papers indicate the costs can be several $/tCO 2 , and there are different well-spacing configuration and designs to consider (Eccles et al 2012, Pooladi-Darvish et al 2011.

Conclusions
The scenarios presented in this analysis provide some bounding cases for the cash flow of a CCUS system in the Texas Gulf Coast and ERCOT grid. The benefits of the analyses are that they use information and data related to Texas geography, geology, and electricity market in an integrated manner. The scenario results are not necessarily meant to present one scenario as more probable or preferable than another, yet most realistic scenarios for development of a CCUS network should fall within the boundaries of the four scenarios. Our system-wide perspective is meant to demonstrate the economics as viewed from outside the system versus inside the system. In this way, any business and government players that could be part of a similar CCUS network in Texas can use this study as a basis for understanding realistic possibilities for cooperation (e.g. sharing of costs and revenues under uncertain future conditions).
The major conclusions from the NPV analyses are: • The scenarios show a system-wide NPV range from −$23 billion (scenario 4: fast EOR development with CO 2 emissions penalty) to −$1.0 billion (scenario 1, slow EOR development with CO 2 sales price). These two scenarios sequester large quantities of CO 2 of 1450 MtCO 2 and 240 MtCO 2 , respectively, over 20 years.
• Because our system-wide net present values are all negative, our results can be broadly interpreted as the additional costs of sequestering large quantities of CO 2 while using oil revenues to pay for a large portion of the costs. These additional 'NPV costs' range from 5 to 25$/tCO 2 .
• The more CO 2 is captured, the lower the NPV of the system. This result stems from our assumption that a similar amount of EOR oil is produced no matter how much CO 2 is available. It is quite feasible that more CO 2 -EOR oil would be produced in a scenario with more available CO 2 .
• The 'CO 2 emissions penalty' scenarios generate less NPV than the 'CO 2 sales price' cases, especially in the fast development scenarios.
• If the cost of purchasing, recycling, and reinjecting CO 2 is low enough, it seems feasible for pure CO 2 -EOR operations to have positive NPV in the present economic environment with no CO 2 emissions penalty.