Article pubs.acs.org/est
Effects of Well Spacing on Geological Storage Site Distribution Costs and Surface Footprint Jordan Eccles,* Lincoln F. Pratson, and Munish Kumar Chandel Nicholas School of the Environment and Earth Sciences, Division of Earth and Ocean Sciences, Duke University, Durham, North Carolina 27708, United States S Supporting Information *
ABSTRACT: Geological storage studies thus far have not evaluated the scale and cost of the network of distribution pipelines that will be needed to move CO2 from a central receiving point at a storage site to injection wells distributed about the site. Using possible injection rates for deepsaline sandstone aquifers, we estimate that the footprint of a sequestration site could range from 100 000 km2, and that distribution costs could be $10/tonne. Our findings are based on two models for determining well spacing: one which minimizes spacing in order to maximize use of the volumetric capacity of the reservoir, and a second that determines spacing to minimize subsurface pressure interference between injection wells. The interference model, which we believe more accurately reflects reservoir dynamics, produces wider well spacings and a counterintuitive relationship whereby total injection site footprint and thus distribution cost declines with decreasing permeability for a given reservoir thickness. This implies that volumetric capacity estimates should be reexamined to include well spacing constraints, since wells will need to be spaced further apart than void space calculations might suggest. We conclude that site-selection criteria should include thick, low-permeability reservoirs to minimize distribution costs and site footprint.
1. INTRODUCTION Global warming stemming from anthropogenic greenhouse gas (GHG) emissions threatens to harm ecosystems and alter human civilization.1 GHG emissions might be mitigated through carbon capture and storage (CCS), the storage component of which may involve the injection of CO2 into geological formations below the Earth’s surface. Candidate reservoirs are relatively abundant and their estimated storage capacity is large,2 but due to geological heterogeneity, the cost for this form of storage remains uncertain and a focus of considerable research.3,4 For example, cost estimates for storing CO2 in deep saline aquifers, the largest potential onshore reservoirs, include simple averages that are generalized to all such aquifers3,5 to highly detailed techno-economic modeling of individual reservoirs.6−8 Neither approach, however, includes the cost of the distribution system for a geologic storage site, i.e. the network of pipelines needed to distribute CO2 arriving at a central receiving point to multiple injection wells spaced about the storage site. Previous studies suggest multiple wells are likely to be the norm for sequestering CO2 emissions from large point sources, such as coal-fired power plants.4,7−9 Whereas current models (e.g., SimCCS) include methods for evaluating CO2 distribution costs,10,11 these do not incorporate the effects of geological variability.12 The cost of a storage-site distribution system depends upon the mass of CO2 to be stored, the injection rate at the wells, and their spacing. If well spacing is large, the footprint of the © 2012 American Chemical Society
distribution system may come into conflict with previous land uses. Some land uses may be amenable to pipeline and well development while others, such as urban areas, may not. Property owners must also be comfortable with and/or properly compensated for risks associated with CO2 storage, such as leakage. Our objective is to explore the impact of well spacing on the surface footprint and cost of a storage-site distribution system. We do this by developing two models for spacing injection wells. In the first, which we refer to as the volume-limited model, the thickness and porosity of a reservoir are used to calculate the overlying surface area and thus well spacing needed to fully sequester the volume of CO2 that would be pumped down each injection well over the lifetime of the sequestration project (Figure 1A). The second approach, which we call the interference model, is to space wells so as to minimize subsurface pressure interference between the wells (Figure 1B). Comparison of these models demonstrates the contrast between large-scale capacity assessments of CO2 storage potential using capacity factors and more realistic site screening and evaluation methods that model reservoir dynamics. The results from our interference model suggest that the Received: Revised: Accepted: Published: 4649
October 6, 2011 March 20, 2012 March 21, 2012 March 21, 2012 dx.doi.org/10.1021/es203553e | Environ. Sci. Technol. 2012, 46, 4649−4656
Environmental Science & Technology
Article
reservoir where a plume of injected CO2 would initially have a circular footprint but evolve to become cuboid as the edges interacted with plumes from other wells. We thus calculate the average well spacing, Wcap (m) as Wcap =
A well
(3)
2.1.2. Interference Model. In the volume-limited model, injection rate is assumed to be constant and unaffected by injection of CO2 about the storage site (eq 2). In reality, subsurface pressures beneath a well will be raised by injection of CO2 elsewhere in the reservoir, slowing the rate of injection if pressure is not reduced by some form of well management taken (Figure 1B).16 McCoy and others come to a similar conclusion exploring well interference,16−18 and our approach is theoretically similar to theirs except that instead of determining the impact of well spacing on injection rate, we want to establish the inverse relationship and use reservoir characteristics to determine the appropriate well spacing. The underground pressure situation can be expressed mathematically as Figure 1. Schematic representation of volume-limited and interference models. The methods for evaluating good well spacing using different premises are shown. The failure mode for each method is shown schematically. In the volume-limited model, wells that are too close together will not have enough bulk volume to inject the required CO2. In the interference model, wells that are spaced too closely will experience the increase in background pressure caused by nearby wells. The injection environment displayed is highly simplified.
Pwell − P0 = ΔP = Q massf (t ) + P(t , ri−1)
In this equation, the mass flux of CO2 being injected into the reservoir, Qmass, is a function of the difference between the pressure at which CO2 is being pumped down the well, Pwell (MPa), and the ambient pressure of reservoir prior to injection, P0 (MPa). Note that this difference, ΔP, is a fixed constant unless the well operator changes the borehole pressure, Pwell. P(t,ri) (MPa) is the added reservoir pressure produced by injection at the other wells. This added pressure increases with time t (s) and decreases with increasing distance to these neighboring wells, ri, (m) where i is well index. Simplifying and rewriting eq 4 as
distribution system has the potential to be a significant component of the total cost of geological storage, and that large-volume storage sites could have either extensive surface footprints or networks of small, geologically isolated reservoirs linked by mesoscale transport systems.
ΔP − P(t , ri−1) ∝ Q mass
2. MODELS 2.1.1. Well Spacing. Volume-Limited Model. The volume-limited model calculates well spacing by first evaluating the mass-per-unit-volume of CO2 that can be stored in the reservoir, Cunit, in kg CO2/m3 bulk reservoir volume (see Supporting Information S1 for a full list of model parameters, definitions, and units). Following the approach of others2,13−15 Cunit = ρCO (P , T )ϕe 2
(1)
Q massl Cunit b
(5)
it can be seen that Qmass at an injection well will decrease as neighboring wells are located closer by (or are operated over a longer time). Options for preventing this decline include increasing the injection pressure at the well (Pwell) or pumping out pore water from the reservoir to make more room for the CO2.19 The amount well pressure can be increased will be limited by the capacity of the well pump, the fracturing threshold of the reservoir, and cost. Cost will also affect the feasibility of brine extraction, which at present is not well constrained and may end up being heavily dependent on local regulations and waste management practices. Given these uncertainties, we do not include well pressure management in our model, and instead limit it to producing baseline cost estimates against which the costs of options for managing well pressures can be compared. To determine a well spacing for which the impact on the performance of an injection well by surrounding wells does not exceed a tolerance threshold, It (the ratio of total pressure interference from other wells to the borehole pressure of a given well as a decimal or percent), we introduce the following dimensionless pressure ratio
where ρ (kg/m3) is the density of CO2 as a function of reservoir temperature (T, K) and pore fluid pressure (P, MPa), ϕ is the porosity or void per bulk volume of the reservoir (ranging from 0 to 1), and e is the volume of CO2 per void known as the capacity factor (which also ranges from 0 to 1). Next we determine the surface footprint associated with Cunit, Awell (m2), which is the minimum area under which CO2 from one injection well could theoretically be stored
A well =
(4)
(2)
In this equation, Qmass (kg/yr) is the annual injection rate of the well, l (yr) is the project lifetime, and b (m) is the injection layer thickness. Due to spatial variations in reservoir properties, Awell can have an irregular shape. For simplicity, we approximate the area as being square. This is not unrealistic for a homogeneous
It =
P(t , ri) ΔP
(6)
where the subscript t denotes the time by which It is reached. 4650
dx.doi.org/10.1021/es203553e | Environ. Sci. Technol. 2012, 46, 4649−4656
Environmental Science & Technology
Article
end member reservoir properties of the U.S. saline aquifers analyzed in Eccles et al.4 Figure 2 portrays the model results for storage site distribution costs. These costs range over several orders of
The right-hand side of eq 6 can be recast in terms of well spacing and key geologic properties of the reservoir using the Theis solution, a radially integrated application of Darcy’s law that describes the change in reservoir pressure over time around a pumping well.20 This revised equation, which is derived in Supporting Information S2, is ⎛ f (Wint)Ss ⎞ i ⎟⎟ 4kt ⎠ ⎝ i=2 n
It =
∑ ei⎜⎜
(7)
where f i is a function given in Supporting Information S2 that transforms Wint (m) into the pair-specific distances between any injection well and each of its neighbors (i.e., ri such as in eq 4), which in our case is based in the same rectangular well grid used in the capacity model. Ss is the specific storativity of the reservoir, and k (mD) is permeability. In our model, we assume that the pore fluids have the compressibility of water (which is true beyond the plume boundary), that the reservoir is homogeneous, and that there is no limit on the pressure that can be introduced to the reservoir, though our per-well injection rate model (see Section 2.2) limits borehole injection pressure to 90% of the fracturing pressure of the formation. Equation 7 is difficult to solve for Wint. In the equation, however, only k varies substantially enough (it can span several orders of magnitude) to significantly alter the value of I. This allows us to develop a simpler approximation for Wint by fitting a regression model to results from eq 7 using the average specific storativity and range of permeabilities for the U.S. saline sandstone aquifers analyzed by Eccles et al. as equation inputs.4 Furthermore, we solve the regression model for I at time t = 20 y, an arbitrary period that corresponds to the planning horizon for energy capital projects.21 In other words, I20 is assumed to be the maximum increase in background pressure counteracting the injection pressure that the well operator is willing to tolerate by year 20 in the project. The regression model, which is formulated in Supporting Information S3, is
Figure 2. Cost of distribution pipeline network produced by the transport model as a function of well spacing and injection rate. Note that cost varies over several orders of magnitude; small jumps in the cost result from changes in discrete pipeline sizes available.
magnitude, from $100/t CO2. The costs increase with increasing well spacing (x-axis, Figure 2) because longer pipelines are needed to distribute CO2 to the wells. Cost rise even further where well injection rates are lower (y-axis, Figure 2) because more wells are needed to handle the incoming flux of CO2. The costliest scenarios occur when the average injection rate per well is low and the well spacing is high. Next, we examine the effect of well spacing and injection rate on surface area requirements for the same type of site using the same ranges for variables as above. The surface area or footprint we discuss is the spatial footprint of the well pattern, not the subsurface plume. The range of areal footprints for the storage site is plotted in Figure 3. If both the spacing and number of wells are small, the footprint of the storage site ends up being relatively small as well. But if the spacing and/or number of wells are large, the footprint becomes huge, reaching >100 000 km2 for a significant subset of the well spacings and injection rates (or number of wells required). Of course, any sequestration site that requires too much land area will probably not be developed, but even a much smaller storage site than that assumed here has the potential to be quite large. For example, a single 500-MW coal-fired power plant produces ∼3.5 MtCO2/y. If this is injected underground by a grid of 50 wells spaced 10 km apart at a rate of 200 t CO2/day/well, the storage site would have a surface footprint larger than the state of Rhode Island. It is important to note that these results are based on our assumption that the underground reservoir for storing CO2 is geologically homogeneous over tens of kilometers or more. This appears to be the case for some potential reservoirs such as the Utsira sandstone in the North Sea, the Mt. Simon Formation in the central U.S., and various Cretaceous strata along the U.S. eastern seaboard.22−24 Most reservoirs, however, are likely to be heterogeneous leading us to caution against universal application of the interference model results. Although our model may produce large well spacings for a given reservoir, geologic heterogeneity can lead to compartmentalization of the reservoir, such as was caused by zones of low permeability in the sandstone injected with CO2 at the
Wint = −445.1*In(I20 + 0.1888ln(k) − 14.221) + 2656.5)* k
(8)
2.2. Distribution Pipeline Model. Once the well spacing is established (either Wcap or Wint), we calculate the cost of distributing CO2 about the storage site using a transport model published by Chandel et al. that we have modified for a storagesite distribution system. This model, most of which has already been presented elsewhere, is described in detail in the Supporting Information.
3. RESULTS 3.1. Ranges of Distribution Pipeline Lengths and Cost. We begin by using the cost model to explore possible distribution pipeline costs for a storage site receiving 10 MtCO2/y. This is a moderate influx of emissions equivalent to that produced annually by ∼1.1 GW of coal-fired power generation. The geology of the storage site is assumed to be uniform. The injection rates extend from zero to the maximum rate designed for the Sleipner West project’s pump systems, 3500 t/day,4 an engineering technology limit Eccles et al. use in exploring other storage system costs.12 Note that injection rate, total mass flux, and number of wells are related in ref 9. The well spacings range between the minimum and maximum values produced by the capacity and interference models for the 4651
dx.doi.org/10.1021/es203553e | Environ. Sci. Technol. 2012, 46, 4649−4656
Environmental Science & Technology
Article
Frequency plots of the distribution costs and well spacings produced by the volume-limited and interference models are shown in Figure 4. The volume-limited model generally
Figure 3. Storage site footprint as a function of number of injection wells and well spacing. Footprints range from 100 000 km2. Sites may have dozens or hundreds of wells because of either low injectivity per well or large site capacity. Well spacings range from $25/tonne for the Nagaoka Project. The latter is so expensive because the reservoir layers at Nagaoka are thin and have low permeability. Consequently the volumelimited model predicts that the large number of injection wells need to be spaced far apart in order to handle the incoming CO2. The opposite holds for the Sleipner Project and the Mt. Simon where the reservoirs are thick, creating a large bulk 4653
dx.doi.org/10.1021/es203553e | Environ. Sci. Technol. 2012, 46, 4649−4656
Environmental Science & Technology
Article
ideal reservoirs (Utsira and the Mt. Simon) this total cost is still quite low. However, Table 1 shows that the Mt. Simon has the smallest increase as a proportion of total cost, especially in comparison to Sleipner, a factor that will be important to site operators.
Combined, the volume-limited and interference models constrain the sizes and costs of distribution systems that will be needed at geological storage sites if CCS is ever deployed on a commercial scale. Of the two models, however, we believe the interference model is more realistic, since the volume-limited model relies on an assumption about the capacity factor. The Theis solution4 makes clear that if a site operator were to space injection wells according to the volume-limited model, the wells would interfere with one another, driving up the reservoir pressure and lowering injection rates, as indicated by previous research.16 One way site operators could overcome this is by increasing the borehole pressure in the injection wells, but only if these higher pressures do not fracture the reservoir and compromise its integrity. For example, injection rates for demonstration projects with reservoirs of relatively low permeability (e.g., the Frio and Nagaoka, Table 1) approach this fracturing threshold, so site operators would have little leeway to increase borehole pressure. The alternative is to drill more injection wells (expanding the site), but at added cost. To overcome pressure interference, operators would require additional compressor capacity at the site. However, roughly half of the BEG data we analyzed had reservoir characteristics in which injection rate would be limited by fracture pressure, meaning that increasing pressure in the bore holes is not an option. In these cases, well operators would need to resort to other options, such as pore water removal. They could also drill additional injection wells or even horizontal wells, which could reduce site footprint and distribution costs, especially if there is vertical compartmentalization in the reservoir. A third alternative is to allow for a higher level of pressure interference between the initial wells and thus declining rates of injection over the project lifetime. A higher tolerance, however, will not translate into a proportionate decrease in well spacing and injection cost. For example, tripling the tolerance used for the Sleipner-type reservoir in Table 1 from 25% to 75% only decreases well spacing and distribution costs by ∼40%. Finally, we believe that our analysis has important implications for the estimation of reservoir storage capacity. Van der Meer39,40 was among the first to contend that in increasing reservoir pressure, CO2 injection would reduce the storage capacity of a reservoir. While he and others have continued to raise this issue,41−43 greater emphasis has been placed on estimating the volumetric capacity inherent in candidate CO2 reservoirs prior to injection.14 Our analysis supports the argument that reservoir pressure management must be taken into account in capacity estimates, particularly where multiple injection wells are involved.41−43 Since wells communicate pressure far beyond the actual migration front of the injected CO2, storage sites will have effective capacity factor that, depending on the permeability of the reservoir, can be significantly lower than the 4% modeled by Doughty et al.,25 and thus lower than the 0.5−5.5% used in U.S. National Carbon Sequestration Database.14 For example, our use of 4% in the volume-limited model resulted in a well spacing of 4.6 km for the highly permeable Sleipner reservoir (Table 1). The interference model, on the other hand, produced a well spacing for the reservoir of 40.6 km. Using this latter spacing and working backward through the volume-limited model equations suggests that a more appropriate capacity factor for Sleipner may be as low as 0.05%, a figure that is notably close to that reported in early work on pressure management by van der Meer.40 For reservoirs with much lower permeability than Sleipner, the revision to the
4. DISCUSSION The results produced by both the volume-limited and interference models make clear that storage site distribution systems can have extremely large physical footprints. For example, our use of a 100 Mt CO2/y injection scenario (Table 1) may seem excessive, but this equates to only 4% of current U.S. annual CO2 emissions from the electric industry and
