Transport modeling at multiple scales for the ... - Wiley Online Library

Modeling and Analysis

Transport modeling at multiple scales for the Illinois Basin – Decatur Project William R. Roy, Edward Mehnert, Peter M. Berger, James R. Damico, and Roland T. Okwen, Illinois State Geological Survey, Champaign, IL, USA Abstract: The application of reactive-transport models is essential to understand and predict the impacts of carbon dioxide (CO2) storage in deep saline reservoirs. This study was conducted to generate preliminary information in support of the Illinois Basin – Decatur Project (IBDP) using two modeling approaches: (i) flow and transport modeling of CO2 at the basin scale, and (ii) geochemical modeling of CO2-saturated brine interactions with the primary seal at the IBDP. Using the TOUGH2-MP simulator, a flow and transport approach was developed to evaluate possible impacts of carbon sequestration at the basin scale. These modeling results should provide useful geologic and hydrogeologic data for future developers of carbon sequestration projects in the Illinois Basin and serve as a template for evaluating geologic carbon sequestration in other deep saline reservoirs. The modeling results demonstrated the significance of the geologic model for understanding the distribution of CO2 and the predicted pressure changes with time. Geochemical modeling was applied to further understand potential interactions of CO2-saturated brine with the Eau Claire Shale. Geochemical simulations were conducted using TOUGHREACT, a numerical simulator that includes reactive chemistry, and Geochemist’s Workbench®, which contains kinetic and reactive-transport modules. Simulations conducted for a 1000-year time frame yielded a decrease in porosity throughout the profile because of mineral precipitation. It was concluded that the rate by which ions diffuse into the caprock had little impact on changes in porosity when compared to the rates of mineral reaction. © 2014 Society of Chemical Industry and John Wiley & Sons, Ltd Keywords: diffusion; geologic sequestration; Illinois Basin; modeling; monitoring; trapping

Introduction n this paper, two approaches to modeling geologic carbon sequestration (GCS) will be discussed – flow and transport modeling and geochemical modeling. For GCS, flow and transport modeling involves simulating the injection of CO2 through one or more wells and predicting how the CO2 will move through the pore space of the injection formation. Injecting CO2 into a brine-filled formation is a two-phase process where the brine is the wetting fluid and CO2 is

I

the non-wetting fluid. As CO2 is injected into the deep saline reservoir, it will increase the formation pressure, and begin flowing away from the well. The density difference between the CO2 and brine means that vertical transport will be a significant process to model. Other important processes to include in the flow and transport model are dissolution of the CO2 into the brine, capillary trapping, and mineral trapping.1 Geochemical modeling involves the same processes as flow and transport modeling with additional calculations for kinetic and equilibrium chemical reactions. After the

Correspondence to: Edward Mehnert, Illinois State Geological Survey, 615 E. Peabody Drive, Champaign, IL 61820, USA. E-mail: [email protected] Received January 24, 2014; revised February 21, 2014; accepted February 27, 2014 Published online at Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/ghg.1424

© 2014 Society of Chemical Industry and John Wiley & Sons, Ltd | Greenhouse Gas Sci Technol. 4:645–661 (2014); DOI: 10.1002/ghg

645

W R Roy et al.

Modeling and Analysis: Transport modeling at multiple scales for the Illinois Basin – Decatur Project

CO2 dissolves into the brine, the perturbation to the system can cause both mineral dissolution and precipitation. The processing time needed for these additional calculations limit the application of geochemical models to smaller spatial scales than possible with the flow and transport modeling.

evaluate GCS at the basin scale. Next, the conceptual geologic model and geologic model for GCS in the Illinois Basin will be discussed. The conceptual geologic model is used to develop the geologic model, which provides most of the input data needed for the flow and transport simulator.

Rationale

Simulating GCS

Engineers, geologists, and geochemists rely on models for two major reasons: to explain why a flow system behaved in the observed manner and/or predict how a flow system will behave in the future.2 The purpose of this flow and transport modeling is to predict future behavior. The goal is to address questions that commonly arise when discussing GCS with other scientists, regulators, and the general public. These questions fall into four general categories: (i) the migration and the chemical transformations of the injected CO2; (ii) the physical and chemical impacts of the injected CO2 on the native brine; (iii) the integrity of the injection formation and its caprock; and (iv) the effects on other industries and stakeholders. Geochemistry plays an important role when assessing the impact of CO2 storage.3 The predictions made from geochemical modeling complement those made for flow and transport modeling to address questions such as: Will the injection of CO2 cause a significant amount of the reservoir or caprock to dissolve and/or cause CO2 to precipitate out of solution as a carbonate mineral, trapping it as a solid phase? Do mineral reactions interfere with or enhance injection through porosity and permeability changes?

GCS involves injecting CO2 into a reservoir. For most GCS projects, CO2 will have a lower density than the native brine and will behave as a supercritical fluid at reservoir pressure and temperature. CO2 will flow horizontally and vertically in the saline reservoir until it is trapped by one of four trapping mechanisms. Structural or stratigraphic trapping involves stopping the flow of CO2 by a structural feature or stratigraphic change in the subsurface geology resulting from an increase in capillary pressure or a reduction in permeability. A shale caprock overlying a sandstone reservoir is an example of a stratigraphic trap. Residual trapping describes the physics of two-phase flow where some of the non-wetting fluid (CO2) remains in small pores as it is replaced by the wetting fluid (brine). Solubility trapping describes the process of CO2 dissolving into the brine. The final process – mineral trapping – describes the geochemical process where the free-phase4 or aqueous CO2 reacts to form a mineral (solid) such as calcium carbonate. Simulators used to predict the effects of GCS can be simple and include only one trapping mechanism or can be more elaborate and include all four of these mechanisms. To simulate GCS in saline reservoirs, Ennis-King and Paterson5 developed a simple, onedimensional, analytical simulator to evaluate the significance of solubility trapping. Likewise, Hesse et al.6 developed a one-dimensional, analytical tool to describe the effect of residual trapping in saline reservoirs. Analytical simulators commonly require limited input data to estimate the desired solution, which allows estimates to be developed in a shorter time frame. This simplicity may limit the utility of the modeling results. For these two examples, the modeling results only demonstrate the effect of a single process in one dimension. More often, the competing effects of two or more mechanisms need to be evaluated and results need to be defined for two or three dimensions. Numerical simulators allow more trapping processes to be evaluated simultaneously and have been used to evaluate GCS in two and three

Methodology Flow and transport modeling Flow and transport modeling requires a simulator and input data to use with the selected simulator. The utility of the modeling results depends on the simulator chosen and the quality, quantity, and spatial representation of the input data. Modeling often requires the use of simplifying assumptions to allow the simulator to perform properly or efficiently. In addition, subsurface geologic data can be expensive to obtain. Thus, flow and transport models rely on extrapolating and interpolating the limited amounts of geologic data to maximize the utility of the data available. The following discussion will give a brief overview of simulators that have been developed to

646



dimensions for generic settings7 or for specific geologic settings like the Illinois Basin8,9 and the Michigan Basin.10 Preliminary geochemical modeling indicated that geochemical trapping would not be a significant trapping process for GCS in the Mt Simon Sandstone of the Illinois Basin.11 Thus, a simulator capable of modeling the three other trapping mechanisms was needed. TOUGH2-MP12 with module ECO2N13 was selected and used to simulate the three-dimensional, flow and transport of CO2 and brine through the Mt Simon Sandstone, the basal sandstone in the Illinois Basin. The purpose of this GCS modeling is to evaluate the effects of future GCS activities; it will explore injecting a significant portion of the Illinois Basin’s atmospheric emissions of CO2. Stationary sources in the Illinois Basin emit 291 million tonnes of CO2 into the atmosphere.14 Specifically, the effects of injecting 50 or 100 million tonnes of CO2 per year using 20 injection wells located across the center of the basin were investigated (Fig. 1). The modeled injection period was

W R Roy et al.

50 years, followed by a post-injection period of variable duration (140 to 1450 years).

Developing the geologic model The injection zone is the Cambrian-age Mt Simon Sandstone. It is unconformably underlain by Precambrian bedrock and conformably overlain by the Eau Claire Shale.15 The Mt Simon is a fine- to coarsegrained, partly pebbly, poorly sorted quartzose to arkosic sandstone and is part of a vast sheet of basal Cambrian sandstone that covers wide areas of the mid-continental United States.15 The Mt Simon Sandstone is a source of groundwater in northern Illinois and southern Wisconsin (outside the basin) and is used to store industrial waste (outside the basin) and natural gas (outside and inside the basin). The effects of GCS on these uses need to be evaluated. Because the Mt Simon Sandstone contains no economic minerals or oil and gas deposits, the geologic data for Mt Simon within the Illinois Basin are limited. Thus, the geologic model developed for this project has been evolutionary and is continually

Figure 1. A plan view of the telescopically refined mesh used for the TOUGH2-MP simulator. The numerical grid near the model boundary is composed of elements 10 km by 10 km. The elements become progressively smaller toward each injection well, which is located at the center of each of the 20 circular features. The three-dimensional mesh contains 1 254 397 elements. Wells 14 and 15 are identified in the green inset map (modified from Zhou et al.8 ).


647

W R Roy et al.


revised as more geologic, geophysical, hydrogeologic, and geochemical data become available from the Illinois Basin – Decatur Project (IBDP) research. The Eau Claire Formation consists of dolomitic siltstones and sandstones with interbedded dolomite and shale in northern Illinois, but grades to dominantly dolomite and limestone in southern Illinois.15 The Precambrian bedrock is also known as the crystalline crust and primarily consists of granite plutons, granodiorite, and rhyolite.16 The conceptual model includes the Mt Simon Sandstone as the injection zone and the Eau Claire Shale as the upper confining unit and the Precambrian bedrock as the lower confining unit. The conceptual model subdivides the relatively thick Mt Simon Sandstone into three major stratigraphic units.17,18 The deepest unit was deposited in a braided river setting with some eolian deposition, and has the greatest porosity and permeability of the three stratigraphic units. The middle unit appears to be fluvial, but at the time of writing, there were few data for additional interpretations. The upper unit was deposited in a shallow-water, tidally influenced system and tends to have relatively consistent properties. The upper unit tends to have greater porosity and permeability than the middle unit. The Precambrian rock is subdivided into two layers. A thin layer represents weathered Precambrian bedrock which rests on top of the unweathered Precambrian bedrock, which is the bottom confining layer. Because of the lateral continuity of these formations, it is not possible to define lateral boundaries based on geologic structure or stratigraphic changes. Instead, lateral boundaries were selected to be at a sufficient distance to minimize their effects. The side and top boundaries had fi xed pressures while the bottom boundary was a no-flow boundary. The initial condition required a definition of the salinity and density of the brine and formation pressure and temperature throughout the model domain. Zhou et al.8 provided details regarding the development of the initial condition for this threedimensional model. All simulations were conducted assuming isothermal conditions. The geologic model required defining the threedimensional geometry of the rock units and defining properties needed by the simulator such as rock properties (e.g. porosity, permeability), fluid properties (e.g. density, salinity) and reservoir properties (e.g. relative permeability, residual saturation). The geometric data for the top and bottom elevations of the

648

formations included in the geologic model were retrieved from the GIS database of the Illinois State Geological Survey (ISGS). While the thickness of the various formations vary by location, their thicknesses at well 14 (Fig. 1) are 95 m for the Eau Claire, 609 m for Mt Simon, and 22 m for the Precambrian weathered rock. Zhou et al.8 described the process used to develop a telescopically refined grid for TOUGH2-MP based on the geologic data (Fig. 1). While the geometric component of the geologic model remained constant, the other data in the geologic model were varied with each scenario modeled. Four scenarios were modeled (Table 1 and Fig. 2). The first scenario (ILB01a) was developed by a partnership of ISGS and Lawrence Berkeley National Laboratory scientists, and has petrophysical properties determined from the Weaber-Horn Unit #1 well, which is the nearest well (82 km south of the IBDP) that fully penetrates the entire Mt Simon Sandstone. Modeling results for this scenario have been published.8 The geological model for this scenario was developed prior to data acquired from drilling the IBDP injection well (CCS 1) and monitoring well (VW 1), which are completed in the Mt Simon Sandstone. The second scenario (ILB01b) is a minor variation of the first scenario, where the vertical permeability was decreased to values equal to the horizontal permeability. The third scenario (ILB02a) includes static data from the IBDP injection and monitoring wells. Static data includes data obtained via geophysical logs, sidewall and whole cores, and other tests. In this scenario, the geologic data determined using the IBDP wells were applied throughout the entire model. The fourth scenario (ILB02b) is a minor variation of the third scenario where porosity and permeability values for the northern and western sectors of the far field (area covered by the coarsest grid in Fig. 1) were assigned values from the Illinois State Water Survey’s calibrated groundwater model of the Mt Simon Sandstone and bedrock aquifers beneath northeastern Illinois.19 Porosity and permeability data are shown for ILB01a (Fig. 2) and all four scenarios (Fig. 3). Figure 2 also shows the stratigraphy and the two injection zones at the base of Mt Simon. The major difference between scenarios ILB01a and ILB01b is the lower vertical permeability assigned to nine Mt Simon layers in ILB01b (Fig. 3). These vertical permeability values were reduced to match the horizontal permeability values for the nine layers. Porosity and permeability values for scenarios ILB02a and ILB02b were assigned



W R Roy et al.

Table 1. Input data used for the TOUGH2-MP simulator. The data include values describing the rock, fluid and reservoir properties for the near-well model domain. Parameter

Unit

Minimum value

Maximum value

Horizontal permeability

m2

2.6 ×

10−20

1.0 × 10−12

Vertical permeability

m2

1.0 × 10−20

1.0 × 10−12

Porosity

%

4.7

20.3

Pore compressibility Temperature

Pa−1 °C

1.83 ×

10−10

7.4 × 10−10

24.4

44.2

0.075

0.228

0

0

Exponent (λ)

0.412

0.90

Liquid saturation

1.00

1.00

Residual liquid saturation

0.15

0.30

Residual gas saturation

0.20

0.25

Exponent (λ)

0.412

0.412

Liquid saturation

0.999

0.999

0.00

0.03

Salt mass fraction Dissolved CO2 Relative permeability function (van Genuchten-Mualem)

Capillary pressure function (van Genuchten)

Residual liquid saturation Strength coefficient Maximum capillary pressure

−1

Pa

Pa

based on data collected from the IBDP wells, and are much lower than those assigned in the first two scenarios. These lower permeability values also required an adjustment to the injection zones and rate. For modeled scenarios ILB01a and ILB01b, a total of 100 million tonnes of CO2 per year were injected into Mt Simon through two injection zones. For modeled scenarios ILB02a and ILB02b, only 50 million tonnes of CO2 per year were injected into Mt Simon through a single injection zone. Note that these modeled injection volumes were much greater than the volume to be injected (1 million tonnes) at the IBDP. The total CO2 injection rate was evenly distributed across all wells. For each well, constant injection rates were assigned in TOUGH2-MP. These rates were set accounting for the permeability and thickness of the rock layers comprising the injection zone.

Geochemical modeling Geochemical modeling has the same basic requirements as flow and transport modeling but uses a finer-scaled version of the geological model used for larger-scale, flow and transport modeling. There are

1.26 ×

10−5

2.24 × 10−4

5.0 ×

105

5.0 × 105

also additional requirements of mineralogical and brine composition, and kinetic rate laws that define how quickly the minerals react. A brief description of the various types of geochemical models applicable to GCS are first given, followed by an example of an application to the Mt Simon Sandstone and Eau Claire Shale.

GCS geochemical models Initially, the chemical composition of the Mt Simon brine was input to equilibrium codes. These geochemical calculations are the simplest form of geochemical modeling, but can yield important information for characterizing reactive fluids, and serve as the basis of more complex model development. Programs such as SpecE820 or PHREEQC21 use water composition as input and then calculate the speciation of the ions in the liquid phase and mineral saturation indices (SI). The SI are indications as to what mineral phases were at equilibrium with the liquid phase before injection, and what water-mineral reactions control changes in chemistry during and after CO2 injection. Knowing what minerals are originally at equilibrium in a


649

W R Roy et al.


Figure 2. Porosity and permeability values assigned to the 24 vertical layers near the injection wells for the TOUGH2-MP simulator. Parameters are shown for the first modeling scenario (ILB01a). The stratigraphy for each of the layers and the two injection zones are also shown.

system is useful when setting initial constraints for more complicated computations. The batch models22 are slightly more complicated than equilibrium models in that they simulate changes over time in response to a perturbation, such as the introduction of CO2. In these models, there is no explicit transport although mass can be added or removed during the simulation. Every component is part of the same system, and there is no separation of distance between components. These models are useful in assessing geochemical interactions based on input data from laboratory studies in which core sample and brines are exposed to CO2 at various temperatures and pressures. Reactive transport models are flow and transport models coupled with geochemical models at each node. With programs such as TOUGHREACT,23 changes in brine properties and porosity can affect properties such

650

as viscosity and porosity. These models can simulate the full injection process, though they are often reduced to one or two dimensions to reduce processing time. The model presented here is a one-dimensional model built to evaluate the effects of CO2-rich brine coming in contact with and diffusing into the Eau Claire Formation from the Mt Simon Sandstone.

GCS geochemical set-up A conceptual model was developed in TOUGHREACT that was a vertical column divided into two sections, the Mt Simon Sandstone and Eau Claire Shale. The model was approximately 200 m tall with each end at a point where pressure data were available. These endpoints were assigned to the observed pressure values, and then a flow model was run that lacked geochemical calculations so that the equilibrium



W R Roy et al.

Table 2. Composition of the brines (as molality). pH Aluminum (Al) Bromine (Br) Calcium (Ca) Carbon as bicarbonate (HCO3) Chlorine (Cl)

6.36

2.98 −6

5.28 × 10 0.0069

5.28 × 10−6 0.0069

0.37

0.37

0.0015

0.972

2.75

2.75

0.0096

0.0096

Magnesium (Mg)

0.076

0.076

Oxygen, dissolved (O2) Potassium (K) Silicon (Si) Sodium (Na)

pressure distribution could be determined. After this effort, two different brines (Table 2) were used as input. The chemical composition Brine A was based on analytical data from field samples collected from the Mt Simon Sandstone. Brine B was based on the same chemical composition as Brine A, but had been equilibrated with CO2 at the average pressure and temperature (20.7 MPa and 323 K (50° C)) of the Mt Simon Sandstone using the geochemical program React®.20 The interactions between Brine B and the

Brine B

Iron (Fe)

Manganese (Mn)

Figure 3. Porosity and permeability values assigned to the 24 vertical layers near the injection wells for the TOUGH2MP simulator. Parameters shown for four modeling scenarios – ILB01a, ILB01b and ILB02a/ILB02b.

Brine A

0.0045

0.0045

1.0 × 10−16

1.0 × 10−16

0.046

0.046

2.03 × 10−4

2.03 × 10−4

1.80

1.80

Sulfur as sulfate (SO4)

0.0065

0.0065

Strontium (Sr)

0.0055

0.0055

sandstone were not considered. Brine A was placed into the Eau Claire nodes, and Brine B, the CO2-rich brine, was placed into the Mt Simon Sandstone nodes. Based on preliminary inspections of core sample collected at the IBDP, the upper zone of the Mt Simon Sandstone appeared to be predominantly quartz. Relative to many other minerals, quartz is resistant to dissolution, and can be considered as geochemically unreactive.24 The nodes that represented the upper Mt Simon Sandstone were assigned a porosity of 11% and a permeability of 6.8 × 10 −15 m2 and were treated as an unreactive mineral volume. The mineralogical composition of Eau Claire Shale core samples collected from the IBDP was characterized by XRD (Table 3). Porosity was set to 8%, which is a reasonable estimate for shales, and the permeability was assigned a value of 6.8 × 10 −17 m2. Equilibrium constraints controlled the saturation state of several minerals in the model (Table 4). For the other minerals, preliminary data were derived from kinetic experiments conducted with Eau Claire Shale samples. The samples plus a synthetic brine were placed in modified Parr reaction vessels then were heated at 316 K (43° C) and pressurized to 16.8 MPa.25 The mineralogical composition of the shale samples, and the chemical composition of the liquid phase were characterized before and after exposure to supercritical CO2. The results were then modeled using an optimization routine to define kinetic rates.


651

W R Roy et al.


Table 3. Mineralogical composition of Eau Claire Shale samples collected at the IBDP (volume %). Mineral

Volume %

Calcite

2.0

Table 5. Average, maximum, and standard deviation of the ΔP (Pascals) for all elements in the TOUGH2-MP model after 50 years of CO2 injection. ΔP (Pa)

Scenario

1.6

Maximum

Average

Std Dev

Dolomite

6.0

ILB01a

3.77 × 10

6

2.47 × 10

1.09 × 106

Illite

12.3

ILB01b

3.78 × 106

2.46 × 106

1.09 × 106

106

3.44 × 106 3.42 × 106

Chlorite

6

Illite-smectite mixture

2.9

ILB02a

1.78 × 10

4.75 ×

Kaolinite

0.8

ILB02b

1.79 × 107

4.71 × 106

Plagioclase feldspar

2.0

Potassium feldspar

22.8

Pyrite

3.0

Quartz

39.4

Siderite

6.0

It was assumed that advective flow from the Mt Simon Sandstone into the Eau Claire Shale would be insignificant. Therefore, the primary transport mechanism for ions would be diff usion. The tortuosity of the Eau Claire Shale was varied from 0.1 to 1 × 10 −5 to gain an idea as to how this variable will impact the rate of ion migration into the shale. All of the ions were assigned the same diffusion coefficient of 1 × 10 −5 m2/s with retardation resulting from only chemical reactions with the mineral phases.

Preliminary results Flow and transport modeling To evaluate the modeling results, the following output was reviewed for the four scenarios: (a) change in Table 4. Kinetic constraints on mineral dissolution and precipitation normalized to 1 cm2/g. Mineral

Mole/m2/s

Smectite

9.7565 × 10−7

Illite

1.2763 × 10−8

Kaolinite

1.5566 × 10−7

Chlorite

7.5125 × 10−7

Quartz

3.3112 × 10−14

K-feldspar

8.3114 × 10−7

P-feldspar

3.3469 × 10−8

Dolomite

8.3828 × 10−7

Siderite

2.1578 × 10−7

652

7

pressure (ΔP) and gas saturation as a function of time at the top of Mt Simon (injection zone), (b) change in pressure (ΔP) and gas saturation as a function of time along a cross-section between two wells, (c) tabulated data showing the maximum and average ΔP in different portions of the geologic model, and (d) mass balance data showing the distribution of free-phase and aqueous CO2 throughout the geologic model. The effect of injecting CO2 into the Mt Simon Sandstone can be evaluated by reviewing the pressure-change data, where ΔP is defined as the pressure at a given time minus the initial pressure (Table 5). The average, maximum, and standard deviation of the pressure change for scenarios ILB01a and ILB01b were similar, but were much less than these same statistics for scenarios ILB02a and ILB02b. Despite the lower injection rate used for scenarios ILB02a and ILB02b, the maximum and average change in pressure were greater for scenarios ILB02a and ILB02b than those observed in scenarios ILB01a and ILB01b. The change in pressure was also evaluated at all elements in the model, the bottom layer of the Eau Claire, the middle of Mt Simon, the top injection zone in Mt Simon, the bottom injection zone of Mt Simon, and the Precambrian layer (Fig. 4). The smallest ΔP values were observed in the bottom of the Eau Claire, while the largest ΔP values were observed in the top injection zone. For the first two scenarios (ILB01a and ILB01b), the top injection zone had the same injection rate and permeability as the bottom injection zone, but was approximately 4 m thinner which led to larger ΔP. The spatial distribution of ΔP revealed that ΔP at the top of Mt Simon had a similar spatial pattern for ILB01a and ILB01b and for ILB02a and ILB02b, but the patterns for ILB02a and ILB02b were more spatially restricted than ILB01a and ILB01b (Figs 5 and 6). The ΔP values for ILB02a and ILB02b showed



W R Roy et al.

Figure 4. Change in pressure in different segments of the geologic model after 50 years of CO2 injection.

higher maximum values, but the area covered by the smallest contour (0.1 MPa or 15 psi) was significantly smaller for scenarios ILB02a and ILB02b. Cross-sectional plots of ΔP and gas saturation between wells 14 and 15 were created to show how these parameters behave inside the larger well field. For scenario ILB01a (Fig. 7), the extent of free-phase CO2 (i.e. gas saturation = 1.0) after 50 years of injection is shown in the upper portion of the figure and the ΔP is shown in the lower portion. The darker shading shows the higher gas saturation opposite the well perforations and upward migration of free-phase CO2 is evident by the lighter shading. Modeling results demonstrate that free-phase CO2 did not reach the top of Mt Simon at this time or at any time during the simulation period. The effect of formation dip and buoyancy can be seen in the asymmetric shape of gas bubbles, as the up dip (left) side of the bubble is slightly larger than the down-dip (right) side of two different wells. The same figure was generated for scenario ILB02a (Fig. 8). This figure shows a much smaller footprint for the free-phase

CO2, while showing dramatically larger pressure changes. To evaluate the transport of CO2 through the various scenarios, the mass balance data were reviewed. The distribution of CO2 at the end of the injection period (Table 6) showed that CO2 can be present as free-phase CO2 (part of a CO2 bubble) or in the aqueous phase (dissolved in brine). In addition, most of the CO2 mass remained in the injection zone (Mt Simon) with minor amounts of dissolved CO2 in the Precambrian layer and trace amounts of dissolved CO2 in the Eau Claire (only scenario ILB01a). At the end of the injection period for scenario ILB01a, approximately 6.8% of the injected CO2 dissolved into the brine. The percentage of the dissolved CO2 varied with time and modeling scenario (Fig. 9). At very early times, a high percentage of the CO2 mass dissolved, but this declined to a value of 6 to 7% early in the injection period and generally increased with time (Table 7). Scenario ILB01a had the largest amount of dissolved CO2 with 10.0% of the CO2 dissolved at the end of the simulation period of


653

W R Roy et al.


Figure 5. Contour maps of ΔP at the top of Mt Simon (injection zone) for the four scenarios after 50 years of CO2 injection. ΔP is report in MPa in this figure (1 MPa = 145 psi). The small black markers in this figure show the location of the numerical grid used by TOUGH2-MP.

190 years. Two modeling results from scenario ILB01a appeared to be due to greater vertical transport resulting in a greater amount of dissolved CO2 throughout the model domain and dissolved CO2 being transported into the Eau Claire. This greater vertical transport was a direct result of the vertical permeability values exceeding horizontal permeability for nine Mt Simon layers in scenario ILB01a. One of the major differences between scenarios ILB01a and ILB01b was the vertical permeability data assigned for these nine Mt Simon layers (Fig. 3).

Geochemical modeling The effects of CO2-rich brine coming in contact with and diff using into the Eau Claire Formation from the Mt Simon Sandstone were simulated. An increase in

654

porosity resulting from mineral dissolution would increase the tortuosity and permeability of the Eau Claire, and reduce the entry capillary pressure of CO2. These physico-chemical reactions could potentially reduce the efficacy of the Eau Claire to act as a seal. However, modeling suggested that porosity would decrease slightly throughout the profi le (Fig. 10) because of mineral precipitation. The greatest extent of the decrease in porosity was not at the Eau Claire-Mt Simon interface, but at approximately 12 m above it. The decrease in porosity resulted from the diff usion of ions up into the Eau Claire, and the subsequent precipitation of solid phases. The changes in porosity were relatively minor because a large fraction of the Eau Claire is relatively unreactive quartz. The major reactions were various



W R Roy et al.

Figure 6. Zoomed-in contour maps of ΔP at the top of Mt Simon (injection zone) for the four scenarios after 50 years of CO2 injection. ΔP is reported in MPa in this figure (1 MPa = 145 psi). The small black markers in this figure show the location of the numerical grid used by TOUGH2-MP.

silicate minerals that formed illite in response to the decrease in pH: 5.75 K-feldspar + H2O + 3H+ + 0.625 Mg 2+ 4.25 K + + 2.5 illite + 8.5 SiO2

(1)

7 kaolinite +10 H+ + 2.4 K + + Mg 2+ 15 H2O + 4.8 Al 3+ + 4 illite

(2)

and the precipitation of calcite in response to the increase in CO2 Ca 2+ + CO2 + H2O calcite + 2 H+

(3)

The modeling results indicated that the variation of the Eau Claire tortuosity had little effect on the porosity and mineralogical reactions. In order to compare the impact of reaction rates with the rate of diff usion, the reaction rate of each mineral was arbitrarily increased by a factor of ten. The ten-fold increase in the

reaction rate yielded a greater impact on the profile than changes in tortuosity (Fig. 11). Therefore, it was concluded that the rate by which ions diffuse into the caprock had little influence on the changes in porosity when compared with the mineral reaction rates. Based on the modeling results to date, the Eau Claire caprock is expected to maintain its integrity as a seal when exposed to CO2-saturated brine. The modeled porosity decreased as calcite precipitated and feldspar minerals dissolved, then precipitated as illite. These results reinforce the conclusion that the Mt Simon Sandstone is a suitable reservoir for GCS in the Illinois Basin because of the presence of a resilient caprock.

Expected outcomes Flow and transport modeling The modeling results from these four scenarios demonstrate the significant effect that the


655

W R Roy et al.


Figure 7. Gas saturation (top) and ΔP (bottom) along a cross-section between two wells (wells 14 and 15) for scenario ILB01a after 50 years of injection (100 MT of CO2 injected). ΔP is reported in Pa for this figure (1000 Pa = 0.145 psi). The small black markers in this figure show the location of the numerical grid used by TOUGH2-MP.

geologic model can have on the predicted pressure changes in the injection zone and adjacent confining layers as well as the distribution of aqueous and free-phase CO2 . In addition, the modeling results demonstrated that pressure fronts from the injection wells will experience well interference, but the CO2 plumes will remain close to the injection wells. In the future, additional geologic models will be developed and refined as new data become available. At the IBDP, CO2 injection commenced in November 2011. Pressure data are being collected from nine vertical locations in Mt Simon using a monitoring well located 330 m from the injection well. In addition, other geologic data from the Illinois Basin have been collected and compiled that allow for the

656

refinement of the far-field portion of the geologic model. The basin-scale geologic model will continue to be refined as data become available. A similar approach has been advocated by Mattax and Dalton26 for simulating oil reservoirs and Haitjema27 for understanding shallow groundwater flow.

Geochemical modeling The future of geochemical modeling for the IBDP site depends on the future needs of the project and the data collected. Additional batch kinetic experiments are ongoing in the laboratory with both Mt Simon Sandstone and Eau Claire Shale samples. The results will be used in future kinetic modeling. Post-injection formation-fluid samples have been



W R Roy et al.

Figure 8. Gas saturation (top) and ΔP (bottom) along a cross-section between two wells (wells 14 and 15) for scenario ILB02a after 50 years of injection (50 MT of CO2 injected). ΔP is reported in Pa for this figure (1000 Pa = 0.145 psi). The small black markers in this figure show the location of the numerical grid used by TOUGH2-MP.

collected, and the chemical composition data will be used to update simulations used to interpret the observed geochemical changes. Core collection will continue, yielding data that are more detailed that in turn will provide potential opportunities for model revisions. Geochemical modeling in a large-scale project is an iterative process carried out at multiple scales.

Summary and lessons learned Flow and transport modeling Using the TOUGH2-MP simulator, the ISGS has developed a flow and transport model to evaluate possible GCS developments at the basin scale. These modeling results should provide useful geological and hydrogeological data to future developers of GCS

Table 6. Distribution of CO2 mass (kg) for each of the modeling scenarios. Model location

ILB01a

ILB01b

ILB02a

ILB02b

CO2 free-phase (kg)

CO2 aqueous (kg)

CO2 free-phase (kg)

CO2 aqueous (kg)

CO2 free-phase (kg)

CO2 aqueous (kg)

CO2 free-phase (kg)

CO2 aqueous (kg)

Eau Claire

0

1.7 × 101

0

< 1 × 10−3

0

< 1 × 10−3

0

< 1 × 10−3

Mt Simon

4.7 × 1012

3.4 × 1011

4.7 × 1012

3.3 × 1011

2.4 × 1012

1.3 × 1011

2.4 × 1012

1.3 × 1011

0

1.7 × 10

0

1.9 × 10

8.6 × 10

1.9 × 10

8.6 × 10

4

1.9 × 106

4.7 × 1012

3.4 × 1011

4.7 × 1012

3.3 × 1011

2.4 × 1012

2.4 × 1012

1.3 × 1011

Precambrian Total

7

7

4

6

1.3 × 1011


657

W R Roy et al.


Figure 9. CO2 distribution by phase (free-phase or aqueous) as a function of time.

projects in and around the Illinois Basin. In addition, the ISGS has developed the expertise needed to run and process output data from the TOUGH2-MP simulator, and has compiled the necessary input data for such simulations. This experience and expertise should be valuable for reviewing the feasibility of future GCS projects in the Illinois Basin and in similar geological formations. The modeling results also demonstrated the farreaching effects that pressure will have within the Mt Simon Sandstone if GCS develops to a scale envisioned in this paper. It would be extremely beneficial to add ‘distant’ monitoring wells to collect pressure data in Mt Simon. As described by Mehnert et al.,28 pressure data should be collected at monitoring wells located remotely, between 1 and 30 km, from the CO2

injection wells. Monitoring wells located between two or more injection wells would have additional value.

Geochemical modeling The interaction of CO2-saturated brine with the Eau Claire caprock was investigated using TOUGHREACT

Table 7. Temporal distribution of dissolved CO2 mass (kg) for the four modeling scenarios. Time

Percentage of injected CO2 dissolved in brine (%) ILB01a

ILB01b

ILB02a

ILB02b

Early

5.8

5.7

7.4

7.2

End of simulation

10.0

6.8

5.8

8.1

658

Figure 10. Porosity evolution over time resulting from CO2-saturated brine diffusing into the Eau Claire. The boundary between the Eau Claire Formation and the Mt Simon Sandstone is at approximately -100 m.



W R Roy et al.

Figure 11. Final porosity of the Eau Claire after diffusion of CO2-rich brines for several different tortuosity values and an increase in kinetic reaction rates. The boundary between the Eau Claire Formation and the Mt Simon Sandstone is at approximately -100 m.

to create a 1D model of diff usion into the formation. This model incorporated data gained from coring the Eau Claire and brine sampling of the Mt Simon Sandstone at the IBDP site. The kinetic constraints for the mineral reactions demonstrated the potential use of data from laboratory-scale reactor experiments which are currently being conducted.25 Geochemical simulations of the impacts GCS on the Eau Claire Formation can be refined as more observational and experimental data become available.

development of the geologic models described in this paper. GHGST reviewers provided comments that improved this manuscript. The Midwest Geological Sequestration Consortium is funded by the U.S. Department of Energy through the National Energy Technology Laboratory (NETL) via the Regional Carbon Sequestration Partnership Program (contract number DE-FC26-05NT42588) and by a cost share agreement with the Illinois Department of Commerce and Economic Opportunity, Office of Coal Development through the Illinois Clean Coal Institute.

Acknowledgments This research is partially supported with funds from the US Department of Energy under award number DE-FC26-05NT42588 (US DOE project manager: Darin Damiani, NETL) and from a US Environmental Protection Agency STAR grant #488220 (USEPA project manager: Barbara Klieforth). The modeling results presented in this work used the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by National Science Foundation grant number OCI-1053575. LBNL Scientists Quanlin Zhou, Jens Birkholzer and Keni Zhang contributed to the development of the modeling results for scenario ILB01a. Scott Frailey and Hannes Leetaru of the ISGS contributed to the

References 1. IPCC, IPCC Special Report on Carbon Dioxide Capture and Storage, ed by Metz B , Davidson O , Coninck H , Loos M and Meyer L. Intergovernmental Panel on Climate Change, Cambridge, United Kingdom and New York, USA, pp. 442 ( 2005). 2. Fetter CW, Applied Hydrogeology, 3rd edn. Macmillan College Publishing Company, New York (1994 ). 3. Gaus I, Audigane P, Andre L , Lions J, Jacquement N, Durst P et al., Geochemical and solute transport modeling for CO2 storage, what to expect from it? Int J Greenhouse Gas Control 2 ( 2 ): 605 – 625 ( 2008 ). 4. McGrail BP, Schaef HT, Glezakou VA , Dang LX and Owen AT, Water reactivity in the liquid and supercritical CO2 phase: Has half the story been neglected? Energ Procedia 1: 3415 – 3419 ( 2009 ). 5. Ennis-King J and Paterson L , Role of convective mixing in the long-term storage of carbon dioxide in deep saline formations. SPE J 10 ( 3 ): 349 – 356 ( 2005 ).


659

W R Roy et al.


6. Hesse MA , Orr FM and Tchelepi HA , Gravity currents with residual trapping. J Fluid Mech 611: 35 – 60 ( 2008 ). 7. Ghanbari S, Al-Zaabi Y, Pickup GE, Mackay E, Gozalpour F and Todd AC, Simulation of CO2 storage in saline aquifers. Chem Eng Res Des 84 ( A9 ):764 –775 ( 2006 ). 8. Zhou QL, Birkholzer JT, Mehnert E, Lin YF and Zhang K, Modeling basin – and plume-scale processes of CO2 storage for full-scale deployment. Ground Water 48 (4 ):494 – 514 ( 2010 ). 9. Person M, Banerjee A , Rupp J, Medina C, Lichtner P, Gable C et al., Assessment of basin-scale hydrologic impacts of CO2 sequestration, Illinois Basin. Int J Greenhouse Gas Control 4 ( 5 ): 840 – 854 ( 2010 ). 10. Barnes DA , Bacon DH and Kelley SR, Geological sequestration of carbon dioxide in the Cambrian Mount Simon Sandstone: Regional storage capacity, site characterization, and large-scale injection feasibility, Michigan Basin. Environ Geosci 16 ( 3 ):163 –183 ( 2009 ). 11. Berger PM, Mehnert E and Roy WR, Geochemical modeling of carbon sequestration in the Mt Simon Sandstone. Geological Society of America Abstracts with Programs, April 2–3, Rockford, IL ( 2009 ). 12. Zhang K, Wu YS and Pruess K, User’s Guide for TOUGH2MP– A Massively Parallel Version of the TOUGH2 Code, Report LBNL-315E. Earth Sciences Division, Lawrence Berkeley National Laboratory, Berkley, CA , pp. 108 ( 2008 ). 13. Pruess K, ECO2N: A TOUGH2 fluid property module for mixtures of water, NaCl, and CO2, Report LBNL-57952. Earth Sciences Division, Lawrence Berkeley National Laboratory, Berkley, CA , pp. 66 ( 2005 ). 14. National Energy Technology Laboratory, US Department of Energy, The United States 2012 Carbon Utilization and Storage Atlas, 4th edn [Online] ( 2012 ). Available at: http:// www.netl.doe.gov/technologies/carbon_seq/refshelf/atlasIV/ [22 January 2014] . 15. Kolata DR, Cambrian and ordovician systems (Sauk sequence and tippecanoe I subsequence), in Geology of Illinois, ed by Kolata DR and Nimz CK. Illinois State Geological Survey, Champaign, IL , pp. 136 –157 ( 2010 ). 16. McBride JH, Kolata DR, Sargent ML and Hildenbrand TG, The precambrian crust, in Geology of Illinois, ed by DR Kolata and CK Nimz. Illinois State Geological Survey, Champaign, IL , pp. 123 –135 ( 2010 ). 17. Leetaru HE and McBride JH, Reservoir uncertainty, precambrian topography, and carbon sequestration in the Mt Simon Sandstone, Illinois Basin. Environ Geosci 16 (4): 235 –243 ( 2009 ). 18. Leetaru HE and Freiburg JT, Reservoir characterization of the Mt Simon Sandstone at the Illinois Basin – Decatur project. Greenhouse Gas Sci Technol 4 ( 5 ): 580 – 595 ( 2014 ). 19. Meyer SC, Roadcap GS, Lin YF and Walker DD, Kane County Water Resources Investigations: Simulation of Groundwater Flow in Kane County and Northeastern Illinois. Illinois State Water Survey Contract Report 2009-07. Illinois State Water Survey, Prairie Research Institute, Champaign, IL, pp. 425 (2009). 20. Bethke CM, Geochemical Reaction Modeling. Oxford University Press, New York, pp. 397 (1996 ). 21. Parkhurst DL and Appelo CAJ, User’s guide to PHREEQC (Version 2)—A computer program for speciation, batch-reaction, one-dimensional transport, and inverse geochemical calculation, Report 99-4259. US Geological Survey WaterResources Investigations, Denver, CO, pp. 310 (1999 ).

660

22. Albarède F, Introduction to Geochemical Modeling. Cambridge University Press, Cambridge, pp. 543 (1995 ). 23. Xu T, Sonnenthal E, Spycher N and Pruess K, TOUGHREACT User’s Guide: A Simulation Program for Non-isothermal Multiphase Reactive Geochemical Transport in Variably Saturated Geologic Media, Report LBNL-55460. Earth Sciences Division, Lawrence Berkeley National Laboratory, Berkley, CA , pp. 192 ( 2004 ). 24. Knauss KG and Wolery TJ, The dissolution kinetics of quartz as a function of pH and time at 70° C. Geochim Cosmochim Acta 52 :43 – 53 (1988 ). 25. Yoksoulian LE, Freiburg JT, Butler SK, Berger PM and Roy WR, Mineralogical alterations during laboratory-scale carbon sequestration experiments for the Illinois Basin. Energ Procedia 37: 5601– 5611 ( 2013 ). 26. Mattax CC and Dalton RL , Reservoir Simulation. Society of Petroleum Engineers, Richardson, TX, pp. 173 (1990 ). 27. Haitjema HM, Analytic Element Modeling of Groundwater Flow. Academic Press Inc, San Diego (1995 ). 28. Mehnert E, Okwen RT, Frailey SM and Damico J, Far Field Pressure Monitoring Well Spacing in Open Basins with Multiple Sequestration Sites. Abstract H41K-05, Fall Meeting, December 5–9, American Geophysical Union, San Francisco, CA ( 2011).

William R. Roy William R. Roy is currently a professor in Nuclear, Plasma, and Radiological Engineering at the University of Illinois. He retired as a senior geochemist with the Illinois State Geological Survey (ISGS) where he had conducted research on the environmental fate of contaminants in soil and groundwater. He holds a BSc and an MA in Geology and a PhD in Soil Physical Chemistry.

Edward Mehnert Edward Mehnert PhD, is a senior geohydrologist at the ISGS, Prairie Research Institute, University of Illinois at Urbana-Champaign. He conducts applied research on a variety of topics including groundwater-surface water interaction and basin-scale modeling of geologic carbon sequestration.

Peter M. Berger Peter M. Berger is an assistant geochemist at the Illinois State Geological Survey. His research focuses on laboratory work and reactive transport modeling of carbon sequestration systems for both deep injection zones and shallow monitoring areas.



James R. Damico James Damico has been an assistant geologist at the ISGS since 2006. He focuses on using geostatistical methods to characterize heterogeneity in reservoirs and building 3D computer simulations of reservoir architecture using stochastic methods. He holds a BSc in Earth Sciences from Purdue University and an MSc in Geology from Wright State University.

W R Roy et al.

Roland T. Okwen Dr Roland T. Okwen is a reservoir engineer at ISGS. He is principal investigator of a US DOE funded research project and provides reservoir engineering expertise to the MGSC’s Phase II pilot and Phase III CO2 EOR performance curves studies in the Illinois Basin. He holds a BSc in Chemistry, an MSc in Petroleum Engineering, and a PhD in Civil Engineering.


661