SPE-169575-MS

Simulation Study of CO2 Huff-n-Puff Process in Bakken Tight Oil Reservoirs Wei Yu, HamidReza Lashgari, Kamy Sepehrnoori, The University of Texas at Austin

Copyright 2014, Society of Petroleum Engineers This paper was prepared for presentation at the SPE Western North American and Rocky Mountain Joint Regional Meeting held in Denver, Colorado, USA, 16–18 April 2014. This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract Advancements in horizontal drilling with hydraulic fracturing have enabled commercial oil production from Bakken tight oil reservoirs. However, the primary recovery factor remains very low (less than 15%), resulting in the high volume of oil remaining in place. Hence, it is extremely important to investigate the application of enhanced oil recovery methods. Carbon dioxide (CO2) injection as a huff-n-puff process is a preferred approach to improve oil recovery in tight reservoirs. In this work, we present the effect of CO2 molecular diffusion and performed a series of sensitivity studies to quantify the impacts of reservoir properties such as permeability and fracture properties such as fracture half-length, fracture conductivity, number of fractures, and operation parameters such as CO2 injection rate, injection time, soaking time, number of cycle of CO2 huff-n-puff and CO2 diffusivity on the CO2 huff-n-puff process for enhanced oil recovery in the Bakken Formation. A numerical model was built using the typical reservoir and fracture properties from Middle Bakken to simulate CO2 huff-n-puff process. In this process, we consider CO2 molecular diffusion term to swell oil in matrix since Darcy velocity is negligible due to low permeability. The numerical model was validated with field production data from a horizontal well in Middle Bakken. Based on the history matching results, the relative permeability curves such as water-oil relative permeability and liquid-gas relative permeability, are obtained. Furthermore, the wettability for the Middle Bakken is found to be weak water wet. Simulation results show that the most important parameter is CO2 injection rate, followed by CO2 injection time, number of cycle, CO2 diffusivity. The other parameters such as fracture conductivity, CO2 soaking time, permeability and fracture half-length are less sensitive based on the range investigated in this study. The range for the incremental oil recovery factor at 30 years of production is obtained as 2.5% - 9.4%. This work can provide fundamental understanding of the key parameters controlling the CO2 huff-n-puff process for enhanced oil recovery in the Bakken Formation. Introduction U.S. Energy Information Administration (EIA) reported that the United States has experienced a rapid increase in dry shale gas production, shale and tight oil production in the past few years, as shown in Fig. 1. Bakken Formation contributes significantly on the tight oil production. The Bakken Formation is a lithologically complex and mixed carbonate-clastic petroleum system with low porosity and permeability, consisting of Upper and Lower Bakken Shales, Middle Bakken, Sanish and Three Forks. Numerous horizontal wells with multi-stage hydraulic fracturing are required to make the economic development of the Bakken Formation. The Williston Basin oil province was discovered in 1951 by the Amerada Petroleum Company, and the Bakken Formation was first formally named in the H.O. Bakken No.1 well in 1953 (Pilcher et al., 2012). The United States Geological Survey (USGS, 2008) estimates the average technologically recoverable resource for the Bakken Formation to be 3.65 Bbbl of oil, 1.85 Tcf of associated/dissolved natural gas, and 148 MMbbl of natural gas liquids. The Williston Basin is an elliptical intracratonic basin spanning the western side of North Dakota, northwester South Dakota, the eastern part of Montana, and the southern part of the Canadian provinces of Saskatcewan and Manitoba, as shown in Fig. 2 (Pilcher et al., 2012); the limit of the Bakken Formation is indicated by the dashed line, and significant fields are indicated by P (Parshall Field) and EC (Elm Coulee Field), structural features are indicated by NA (Nesson Anticline), A (Antelope Field), LK (Little Knife Anticline), BN (Billings Nose Anticline), PD (Polar Dome) and CC (Cedar Creek Anticline), and the average area is approximately 345,000 km2 (Gerhard et al., 1990). The Upper and Lower Bakken are the black, organic-rich source shales with similar lithologies, high total organic carbon (10-20% by weight), and relatively low clay mineral content;

2 SPE-169575-MS

the Middle Bakken, the upper Three Forks Formation and the Sanish unit are three reservoirs with averaging approximately 6% of porosity and 0.005 md of permeability (Pilcher et al., 2012). The average depth of the Middle Bakken is about 10,000 ft. The porosity of the Middle Bakken is about 6% and the air permeability is on the order of microdarcies (Nojabaei et al., 2013). The water saturation ranges between 25% and 50% in the Middle Bakken (Cherian et al., 2012). The average oil gravity of the Bakken Formation is about 42 oAPI, which indicates the crude oil is light with low viscosity. Nojabaei et al. (2013) reported that the range for gas oil ratio (GOR) is from 507 to 1712 SCF/bbl and bubble point pressure varies from 1617 to 3403 psi based on the different location in the Bakken Formation. The Bakken Formation is variably overpressured with pressure gradients up to 0.73 psi/ft in the central part of the basin (Meissner, 1978). The Middle Bakken and Upper Three Forks are the main targets for current field development since they have the best reservoir qualities such as porosity and oil saturation (Iwere et al., 2012). In recent years, with the advancements in horizontal drilling and hydraulic fracturing technologies, the development of Bakken resources has been very rapid. Although with the advanced technologies, the primary oil recovery factor remains very low and has been estimated to be less than 15% of original oil in place (OOIP) (Iwere et al., 2012; Cherian et al., 2012; Hoffman, 2012). The key recovery mechanism for the primary recovery in tight oil reservoirs is depressurization and solution gas drive, resulting in a very low recovery factor. Hence, it provides a highly potential for enhanced oil recovery (EOR). It is reported that a minor improvement in oil recovery factor such as 1% could yield additional 1.6 to 9 billion barrels of recovered oil (Hawthorne et al., 2013). If we assume the average oil price of $90 per barrel, then the additional 1% increase of oil recovery factor can bring revenue of $144 to 810 billion. Accordingly, it is essential to evaluate enhanced oil recovery methods to maximize the recovery factor. For conventional reservoirs, the main concern for the secondary oil recovery is pressure maintenance when the reservoir is depleted after the primary production, and the rock matrix will control injectivity and sweep efficiency. However, for unconventional reserovirs with multiple hydraulic fractures along with a horizontal wellbore, the induced fractures play an important role in determining the injectivity and sweep efficiency. Although water flooding is the most widely used recovery technique for enhanced oil recovery after primary production for conventional oil reservoirs, it is not the most applicable option for the Bakken Formation because of low injectivity (low permeability and capillary forces below minimum miscibility pressure) or poor sweep efficiency (induced fracture networks) or clay swelling problems. In contrast, gas injection is recognized to be more suitable than water flooding for enhanced oil recovery in the Bakken Formation due to lower gas viscosity and higher injectivity than water. In addition, the depth of the Bakken Formation (average of 10000 ft) and high oil saturation are beneficial to gas injection (Iwere et al., 2012). The most commonly used gases for injection includes Carbon dioxide (CO2), nitrogen (N2), natural gas or the mixture of them. For CO2 injection, it is very necessary to maintain the reservoir pressure above the minimum miscibility pressure (MMP) of CO2, so that CO2 injection is operated as a miscible flooding process, which will yield CO2 swelling and viscosity reduction, resulting in a higher oil recovery. Gas injection is typically a multiple contact process since it is hard for the injected gas to be miscible with the in-situ oil at the beginning, especially for the light and medium oil reservoirs (Wang et al., 2010). Hawthorne et al. (2013) proposed five conceptual steps for CO2 injection for enhanced oil recovery in the Bakken Formation: (1) CO2 flows into and through the fractures, (2) unfractured rock matrix is exposed to CO2 at fracture surfaces, (3) CO2 permeates the rock driven by pressure, carrying some hydrocarbon inward; however, the oil is also swelling and extruding some oil out of the pores, (4) oil migrates to the bulk CO2 in the fractures via swelling and reduced viscosity, and (5) as the CO2 pressure gradient gets smaller, oil production is slowly driven by concentration gradient diffusion from pores into the bulk CO2 in the fractures. Also, they conducted CO2-exposure experiments to investigate these concepts with rock samples from Middle Bakken, Upper and Lower Bakken, and a conventional reservoir at 230 oF and 5000 psi, and demonstrated that CO2 is capable of recovering hydrocarbons from Bakken source and reservoir rock effectively. In addition, many attempts have been made to evaluate CO2 injection in the Bakken Formation (Shoaib and Hoffman, 2009; Wang et al., 2010; Chen et al., 2013; Wan et al., 2013), and it is recognized that CO2 injection is a technically promising method for enhanced oil recovery in the Bakken Formation. However, the impacts of key reservoir and fracture properties and CO2 operation parameters on the process of CO2 injection for enhanced oil recovery are not studied systematically and still poorly understood. Hence, it is very critical to quantify the key parameters controlling the mechanisms of CO2 injection in the Bakken Formation. In this work, we built a numerical model using typical reservoir and fluid properties from the Middle Bakken to simulate cyclic CO2 injection as a process of huff-n-puff. The reservoir model is validated with field production data from one horizontal well in the Middel Bakken. After that, we performed a series of sensitivity study to quantify the impacts of reservoir properties such as permeability and fracture properties such as fracture half-length, fracture conductivity, number of fractures, and operation parameters such as CO2 injection rate, injection time, soaking time, number of cycle of CO2 huff-n-puff and CO2 diffusivity on the practical feasibility of CO2 injection for enhanced oil recovery. The goal of this paper is to provide insights into a better understanding of key parameters controlling the CO2 huff-n-puff process for enhanced oil recovery in the Bakken Formation.

SPE-169575-MS 3

Physical Mechanism CO2 molecular diffusion term plays an important role in capturing the real physics in numerical simulation study in the Bakken Formation. In principle, many laboratory experiments (McKay, 1971; Thomas et al., 1991; Le Romancer et al., 1994; Lagalaye et al., 2002; Darvish et al., 2006a, 2006b; Karimaie and Torsæter, 2010) illustrated that CO2 injection can improve recovery of conventional and unconventional oil reservoirs. Solubility of CO2 in oil causes the oil swelling and viscosity reduction, which are the main contributing mechanisms of CO2 injection into oil reservoirs. Since permeability of the matrix in the Bakken Formation is very low, convection flux cannot transport CO2 from the fracture to the matrix. The only driving force is concentration gradient of CO2 (Fick’s law) in oil and gas phases. But there is a challenging part in the diffusion coefficients for the non-ideal multicomponent mixtures in the oil and gas phases. This mechanism may occur in fractured media when fractures saturated with CO2 get in contact with under-saturated oil in the surrounding matrix. CO2 diffuses from gas phases in fracture to oil in matrix based on Fick’s law. In this study, we use CMG-GEM (CMG, 2012) compositional simulator to model CO2 molecular diffusion and its role in increasing oil recovery from the Bakken Formation. This simulator uses Sigmund model (Sigmund, 1976a, 1976b) to define the effective diffusivity of each component at each phases rather than diffusivity tensor. Fig. 3 shows a schematic of a matrix saturated with oil (very low permeability) and a fracture saturated with CO2 gas. It is assumed that there are no capillary, gravity and viscous forces between two phases in the matrix and fracture. There are only differences of gradient mass or concentration of CO2 and components in the oil and gas phases, which result in Fick’s flux occurrence. CO2 moves from the gas phase to the oil phase, and components move from the matrix to the fracture. CO2 concentration in the oil phase increases with time, leading to oil swelling and viscosity reduction. Mathmatical Formulation The total mass balance for each component in the oil and gas phases is described by the continuity equation as

( )1 1 1

1 .p p pN N N

s is i j i i iw S w w u S K w S rt

∂ φ ρ φ ρ ρ φρ φ∂ = = =

⎛ ⎞ ⎛ ⎞− + + ∇ − ∇ =⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟

⎝ ⎠ ⎝ ⎠∑ ∑ ∑l l l l l l l l l l l ll l l

r (1)

where φ is matrix porosity, sρ is matrix density, isw is mass fraction of component i that precipitates on the matrix rock,

iw l is mass fraction of component i in the phase l. iK l is the dispersivity of component i in the phase l, which is defined as

ii

uDKα

τ φ= +l l l

l (2)

αl is the dispersivity of fluid l in the longitudinal direction and two transverse directions. Since matrix has a very low

permbeality, so that Darcy’s velocity (ul ) for each phase in the matrix is negligible. Therfore, iK l is simplified as

ii

DKτ

= ll (3)

If no reaction term is assumed, then the equation is simplified as

( )1 1

1 . 0p pN N

s is i i iw S w S D wt

∂ φφ ρ φ ρ ρ∂ τ= =

⎛ ⎞ ⎛ ⎞− + − ∇ ∇ =⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟

⎝ ⎠ ⎝ ⎠∑ ∑l l l l l l ll l

r (4)

where τ is tortuosity of the matrix and iD l is the effective diffusivity of component i in phase l. In this equation the only driving force is mass (or concentration) gradient of components and the other driving forces such as viscous force, gradient force and capillary are neglected due to very low permeability of the matrix. This equation represents a partial differential equation only with mass transport due to the gradient of mass or concentration. In this work, the impact of CO2 diffusivity in

gas and oil phases ( iD l ) on the CO2 injection for enhanced oil recovery from the shale matrix is investigated.

Numerical Simulation Although there is very little experience of pilot tests in the Bakken Formation, reservoir simulation approach provides an effective way to evaluate CO2 injection for enhanced oil recovery and contribute to make decisions in the early stage of field developments. In this simulation study, the GEM simulator is used to model multiple hydraulic fractures and fluid flow in tight oil reservoirs. GEM is Computer Modeling Group’s (CMG, 2012) advanced general equation-of-state compositional simulator. In our simulation model, local grid refinement (LGR) with logarithmic cell spacing is used to accurately model

4 SPE-169575-MS

fluid flow from shale matrix to hydraulic fractures. This method has been extensively used to model transient gas flow in hydraulically fractured shale reservoirs (Rubin, 2010; Cipolla et al., 2010; Yu and Sepehrnoori, 2013a, 2013b, 2014; Yu et al., 2014a, 2014b). In the GEM simulator, the Sigmund method (Sigmund, 1976a, 1976b) is used to model CO2 molecular diffusion in oil and gas phases. The coefficient of CO2 diffusivity in oil and gas phases can be determined through laboratory measurements (McKay,1971; Thomas et al., 1991; Le Romancer et al., 1994; Lagalaye et al.,2002; Darvish et al., 2006a, 2006b, Karimaie and Torsæter, 2010). History Matching Once a reservoir model including multiple hydraulic fractures is built, it is very significant to validate the model with field production data to make sure a high reliability of simulation results. When a good history matching result is achieved, this reservoir model can be used to investigate CO2 injection for EOR in the Bakken Formation. In this work, the field production data of a hydraulically fractured horizontal well from the Middle Bakken in North Dakota is used to perform history matching for validation of the reservoir model (Kurtoglu and Kazemi, 2012). In this field case study, the horizontal well was stimulated with fifteen hydraulic fracturing stages. In the model, it is assumed that there are four effective fractures per single stage, so there are totally sixty hydraulic fractures along the horizontal well. The fracture width is set at 0.001 ft. An area of about 326 acres was simulated by setting up a basic 3D reservoir model with dimensions of 10500 ft × 2640 ft × 50 ft, which corresponds to length, width, and thickness, respectively. The other detailed reservoir and fracture properties about this well are listed in Table 1. During history matching, bottom hole pressure (BHP) measured from the field is used for a constraint input, as shown in red line in Fig. 4. Oil and gas flow rate are the targets for history matching. The history matching results for oil and gas flow rate are shown in Figs. 5 and 6, respectively. As shown, a reasonable match between the numerical simulation results and the actual field data is obtained for oil and gas flow rate, respectively. During history matching, the main tuning parameters are fracture half-length, fracture conductivity, matrix permeability, CO2 diffusivity in oil and gas phases, water saturation and relative permeability. Hence, based on a good history matching result, we obtained matrix permeability of 5 µD, fracture half-length of 215 ft, fracture conductivity of 50 md-ft, and initial water saturation of 41%. These values are consistent with the existed studies in the literature. In addition, relative permeability for tight oil reservoirs is extremely hard to measure, which is very sensitive for history matching. In this field case study, the relative permeability curves, such as water-oil relative permeability and liquid-gas relative permeability, are obtained by tuning these curves to fit a good history match, as shown in Fig. 7. Furthermore, it can be seen that the wettability for the Middle Bakken is weak water wet. These relative permeability curves are used in the subsequent CO2 injection simulation study. Fig. 8 presents the pressure distribution in 2D view at 1.2 years of field production. It can be seen that the effective drainage area is clearly illustrated in the horizontal fractured Bakken well. Simulation Study for CO2 Huff-n-Puff The CO2 huff-n-puff process in a horizontal tight oil well with multi-stage fractures consists of three main stages: CO2 injection, CO2 soaking, and production (Fig. 9). In this work, initially the horizontal well produces for five years, and then it is converted to a CO2 injector with injection rates of 100 MSCF/day. After six months of CO2 injection, the well is shut-in and soaking for three months. Finally, the well is put back into production for one year. This is one cycle of CO2 huff-n-puff. After that, another cycle of CO2 huff-n-puff continues. The total production period is 30 years. In the simulation study, four cases with different number of fractures from one to four within single stage are investigated because of the expensive computational time for the entire well with multiple stages, as shown in Fig. 10. Two cycles of CO2 huff-n-puff are used for each case. The dimensions for this segment are 340 ft × 1300 ft × 40 ft, which corresponds to length, width, and thickness, respectively. For each case, fracture half-length is 350 ft, fracture height is 40 ft and fracture conductivity is 10 md-ft. Fracture spacing is set at 120 ft for case 2, and 80 ft for cases 3 and 4. Case 1 is only with single hydraulic fracture. CO2 diffusivity coefficient of 0.0008 cm2/s is used for each case. The typical fluid and rock properties from Middle Bakken are used for the simulation study, as listed in Table 2. The reservoir is assumed to be homogeneous and the fractures are with stress-independent porosity and permeability. The reservoir oil is carefully divided into seven different pseudocomponents, i.e., CO2, N2-C1, C1-C4, C5-C7, C8-C12, C13-C19, C20-C30, and their corresponding molar fractions are 0.01%, 22.03%, 20.63%, 11.7%, 28.15%, 9.4%, and 8.08%, respectively. The other detailed input data required for the Peng-Robinson equation-of-state (CMG-WinProp) are listed in Table 3. The binary coefficient used for flash calculation is listed in Table 4. Comparison of oil recovery factor with and without CO2 injection for each case is shown in Fig. 11. For the case without CO2 injection means that well is under shut in corresponding to the well is under CO2 injection and soaking for the case with injection. The total shut in time is same as the time of CO2 injection and soaking. As shown, the oil recovery factor for the case with CO2 injection is higher than the case without CO2 injection for the four cases. This is because the injected CO2, operated as a miscible process, yields CO2 swelling and oil viscosity reduction, resulting in a higher oil recovery. Comparison of incremental oil recovery factor at 30 years of production for the four cases is shown in Fig. 12. It can be seen that case 2

SPE-169575-MS 5

with two fractures has the highest incremental oil recovery factor compared to the other cases, and cases 3 and 4 have the similar incremental oil recovery factor, which is larger than case 1. The reason for this is case 2 has the largest CO2 invasion area compared to the other cases, as shown in Fig. 13. In addition, the distance for CO2 invasion along the fractures decreases with the increasing number of fractures, which might be resulted from the fracture interference. Sensitivity Study for CO2 Huff-n-Puff There is high uncertainty in Bakken tight oil formation because of many uncertain parameters such as reservoir permeability, fracture half-length, number of fractures, and fracture conductivity. In addition, parameters related to CO2 huff-n-puff is also uncertain, such as CO2 diffusivity, CO2 injection rate, and injection time and soaking time, and number of cycle of CO2 huff-n-puff. Accordingly, in the subsequent simulation study, we perform a series of simulations to investigate the impacts of these uncertain parameters on CO2 huff-n-puff process. We investigated eight uncertain parameters, and each parameter is given a reasonable range, as listed in Table 5. For the base case, we selected case 2 with two effective hydraulic fractures. The other parameters are the same as that listed in Table 2. Comparison of oil recovery factor with and without CO2 injection for each uncertain parameter is shown in Figs. 14-21, respectively. As shown, the oil recovery factor increases with increasing CO2 injection rate, CO2 injection time, number of cycle, CO2 diffusivity, CO2 soaking time, permeability, and fracture half-length, while decreases with increasing fracture conductivity. In addition, the impacts of all uncertain parameters on the incremental oil recovery factor at 30 years of production are shown in Tornado plot (Fig. 22). It can be seen that the most important parameter is CO2 injection rate, followed by CO2 injection time, number of cycle, CO2 diffusivity. The other four parameters such as fracture conductivity, CO2 soaking time, permeability and fracture half-length are less sensitive compared to the other four parameters based on the range investigated in this study. The range for the incremental oil recovery factor at 30 years of production is obtained as 2.5% – 9.4%. In addition, it can be suggested that in the Bakken Formation experimental measurements for CO2 diffusivity coefficient is required for a better understanding of the mechanism of CO2 injection for enhanced oil recovery. Summary and Conclusions We performed a series of simulations for the CO2 huff-n-puff process for enhanced oil recovery in the Bakken Formation. The following conclusions can be drawn from this work: (1) The relative permeability curves, such as water-oil relative permeability and liquid-gas relative permeability, are

obtained based on the history matching with a fractured well from the Middle Bakken. Furthermore, the wettability for the Middle Bakken is weak water wet.

(2) The case with two effective hydraulic fractures within one perforation stage has the highest incremental oil recovery factor compared to the other cases with one, three and four fractures within one perforation stage.

(3) Molecular diffusivity of CO2 is an important parameter in the numerical simulation study to capture the real physics during CO2 injection into tight oil reservoirs.

(4) The most important parameter for CO2 huff-n-puff process is CO2 injection rate, followed by CO2 injection time, number of cycle, CO2 diffusivity.

(5) The other parameters such as fracture conductivity, CO2 soaking time, permeability and fracture half-length are less sensitive based on the range investigated in this study.

(6) The range for the incremental oil recovery factor at 30 years of production is obtained as 2.5% - 9.4%. Acknowledgement The authors would like to acknowledge Computer Modeling Group Ltd. for providing the CMG-GEM software for this study. We would also like to express our gratitude for financial support from the Chief Oil & Gas Company.

Nomenclature BHP = Bottom hole pressure CMG = Computer Modeling Group LGR = Local grid refinement EOR = Enhanced oil recovery GOR = Gas oil ratio OOIP = Original oil in place MMP = Minimum miscibility pressure, psi MMSCF = 106 standard cubic feet, ft3 φ = Porosity

sρ = Matrix density

isw = Mass fraction of component i that precipitates on the matrix rock

6 SPE-169575-MS

iw l = Mass fraction of component i in the phase l τ = Tortuosity of the matrix

iD l = The effective diffusivity of component i in phase l

iK l = The effective dispersivity of component i in phase l

SI Metric Conversion Factors ft × 3.048 e-01 = m ft3 × 2.832 e-02 = m3

(oF-32)/1.8 = oC cp × 1.0 e-03 = Pa·s psi × 6.895 e+00 = kPa acre × 4.047 e+03 = m2 bbl × 1.589873 e-01 = m3 References Chen, C., Balhoff, M., and Mohanty, K.K., 2013. Effect of Reservoir Heterogeneity on Improved Shale Oil Recovery by CO2 Huff-n-Puff.

Paper SPE 164553, Unconventional Resources Conference-USA, The Woodlands, TX, April 10-12. Cherian, B.V., Stacey, E.S., Lewis, R., Iwere, F.O., Heim, R.N., and Higgins, S.M., 2012. Evaluating Horizontal Well Completion

Effectiveness in a Field Development Program. Paper SPE 152177, SPE Hydraulic Fracturing Technology Conference, The Woodlands, TX, February 6-8.

Cipolla, C.L., Lolon, E.P., Erdle, J.C., and Rubin, B., 2010. Reservoir Modeling in Shale-Gas Reservoirs. SPE Res. Eval. & Eng. 13 (4): 638-653.

CMG, 2012. IMEX User’s Guide. Computer Modeling Group Ltd. Darvish, G.R., Lindeberg, E.G.B., Holt, T., Kleppe, J., and Utne, S.A., 2006a. Reservoir Conditions Laboratory Experiments of CO2

Injection into Fractured Cores. Paper SPE 99649, SPE/DOE Symposium on Improved Oil Recovery, Tulsa, OK, April 22-26. Darvish, G.R., Lindeberg, E.G.B., Holt, T., Kleppe, J., and Utne, S.A., 2006b. Reservoir Conditions Laboratory Experiments of CO2

Injection into Fractured Cores. Paper SPE 99650, SPE Europe/EAGE Annual Conference and Exhibition, Vienna, Austria, June 12-15.

Gerhard, L.C., Anderson, S.B., and Fischer, D.W., 1990. Petroleum Geology of the Williston Basin, in Leighton, M.W., Kolata, D.R., Oltz, D.T. and Eidel, J.J., eds., Interior cratonic basins: AAPG Memoir 51, 507-559.

Hawthorne, S.B., Gorecki, C.D., Sorensen, J.A., Steadman, E.N., Harju, J.A., and Melzer, S., 2013. Hydrocarbon Mobilization Mechanisms from Upper, Middle, and Lower Bakken Reservoir Rocks Exposed to CO2. Paper SPE 167200, SPE Unconventional Resources Conference, Calgary, Canada, November 5-7.

Hoffman, B.T., 2012. Comparison of Various Gases for Enhanced Recovery from Shale Oil Reservoirs. Paper SPE 154329, SPE Improved Oil Recovery Symposium, Tulsa, OK, April 14-18.

Iwere, F.O., Heim, R.N., and Cherian, B.V., 2012. Numerical Simulation of Enhanced Oil Recovery in the Middle Bakken and Upper Three Forks Tight Oil Reservoirs of the Williston Basin. SPE Americas Unconventional Resources Conference, Pittsburgh, PA, June 5-7.

Karimaie, H., and Torsæter,O., 2010. CO2 and C1 Gas Injection for Enhanced Oil Recovery in Fractured Reservoirs. Paper SPE 139703, SPE International Conference on CO2 Capture, Storage, and Utilization, New Orleans, LA, November 10-12.

Kurtoglu, B., and Kazemi, H., 2012. Evaluation of Bakken Performance Using Coreflooding, Well Testing, and Reservoir Simulation. Paper SPE 155655, SPE Annual Technical Conference and Exhibition, San Antonio, TX, October 8-10.

Lagalaye,Y., Nectoux,A., and James, N., 2002. Characterization of Acid Gas Diffusion in A Carbonate Fractured Reservoir through Experimental Studies, Numerical Simulation and Field Pilots. Paper SPE 77339, SPE Annual Technical Conference and Exhibition, San Antonio, TX, September 29-October 2.

Le Romancer, J-F.X., Defives, D., and Fernandes, G., 1994. Mechanism of Oil Recovery by Gas Diffusion in Fractured Reservoir in Presence of Water. Paper SPE 27746, SPE/DOE Improved Oil Recovery Symposium, Tulsa, OK, April 17-20.

McKay, W.N., 1971. Experiments Concerning Diffusion of Multicomponent Systems at Reservoir Conditions. J. Can. Petrol. Technol. 10(2): 25-32.

Meissner, F.F., 1978. Petroleum Geology of the Bakken Formation, Williston basin, North Dakota and Montana, in Estelle, D. and Miller, R,. eds, The economic geology of the Williston Basin, 1978 Williston Basin Symposium, Billings, Montana: Montana Geological Society, 207-230.

Nojabaei, B., Johns, R.T., and Chu, L. Effect of Capillary Pressure on Phase Behavior in Tight Rocks and Shales. SPE Reserv. Eval. Eng. 16 (3): 281-289.

Pilcher, R.S., Ciosek, J.M., McArthur, K., Hohman, J., and Schmitz, P.J., 2012. Ranking Production Potential Based on Key Geological Drivers-Bakken Case Study. Paper IPTC 14733, International Petroleum Technology Conference, Bangkok, Thailand, February 7-9.

Rubin, B., 2010. Accurate Simulation of Non-Darcy Flow in Stimulated Fractured Shale Reservoirs. Paper SPE 132093, SPE Western Regional Meeting, Anaheim, CA, May 27-29.

Shoaib, S., and Hoffman, B.T., 2009. CO2 Flooding the Elm Coulee Field. Paper SPE 123176, SPE Rocky Mountain Petroleum Technology Conference, Denver, CO, April 14-16.

Sigmund, P.M., 1976a. Prediction of Molecular Diffusion At Reservoir Conditions. Part 1- Measurement and Prediction of Binary Dense Gas Diffusion Coefficients. J. Can. Petrol. Technol. 15(2): 48-57.

SPE-169575-MS 7

Sigmund, P.M., 1976b. Prediction of Molecular Diffusion at Reservoir Conditions. Part II - Estimating the Effects of Molecular Diffusion and Convective Mixing in Multicomponent Systems. J. Can. Petrol. Technol. 15(3): 53-62.

Thomas, L.K., Dixon,T., and Pierson, R., 1991. Ekofisk Nitrogen Injection. SPE J. 6(2): 151-160. U.S. EIA, 2013. EIA to Release New Drilling Productivity Report. http://www.eia.gov/todayinenergy/detail.cfm?id=13451#. U.S. Geological Survey, 2008. Assessment of Undiscovered Oil Resources in the Devonian-Mississippian Bakken Formation, Williston

Basin Province, Montana and North Dakota: USGS Fact Sheet 2008-3021, http://pubs.usgs.gov/fs/2008/3021/pdf/FS08-3021_508.pdf. Wan, T., Sheng, J.J., and Soliman, M.Y., 2013. Evaluate EOR Potential in Fractured Shale Oil Reservoirs by Cyclic Gas Injection. Paper

SPE 168880/URTeC 1611383, Unconventional Resources Technology Conference, Denver, CO, August 12-14. Wang, X., Luo, P., Er, V., and Huang, S., 2010. Assessment of CO2 Flooding Potential for Bakken Formation, Saskatchewan. Paper SPE

137728, Canadian Unconventional Resources and International Petroleum Conference, Calgary, Canada, October 19-21. Yu, W., and Sepehrnoori, K., 2013a. Simulation of Proppant Distribution Effect on Well Performance in Shale Gas Reservoirs. Paper SPE

167225, SPE Unconventional Resources Conference, Calgary, Alberta, Canada, November 5-7. Yu, W., and Sepehrnoori, K., 2013b. An Efficient Reservoir Simulation Approach to Design and Optimize Unconventional Gas

Production. Paper SPE 165343, SPE Western Regional & AAPG Pacific Section Meeting, Monterey, CA, April 19-25. Yu, W., and Sepehrnoori, K., 2014. Simulation of Gas Desorption and Geomechanics Effects for Unconventional Gas Reservoirs. Fuel,

116: 455-464. Yu, W., Luo, Z., Javadpour, F., Varavei, A., and Sepehrnoori, K., 2014a. Sensitivity Analysis of Hydraulic Fracture Geometry in Shale Gas

Reservoirs. J. Petro. Sci. Eng. 113: 1-7. Yu, W., Al-Shalabi, E.W., and Sepehrnoori, K., 2014b. A Sensitivity Study of Potential CO2 Injection for Enhanced Gas Recovery in

Barnett Shale Reservoirs. Paper SPE 169012, SPE Unconventional Resources Conference, The Woodlands, TX, April 1-3.

8 SPE-169575-MS

Table 1: Parameters used for history matching (Middle Bakken)

Parameter Value Unit The model dimensions 10500 × 2640 × 50 ft Initial reservoir pressure 7800 psi Production time 1.2 year Reservoir temperature 245 oF Initial water saturation 0.41 value Total compressibility 1×10-6 psi-1 Matrix permeability 5 µD Matrix porosity 0.056 value Horizontal well length 8828 ft Number of stages 15 value Cluster spacing 118 ft Total number of fractures 60 value Fracture conductivity 50 mD-ft Fracture half-length 215 ft Fracture height 50 ft

Table 2: Parameters used for the CO2 huff-n-puff process

Parameter Value Unit The model dimensions 340 × 1300 × 40 ft Initial reservoir pressure 8000 psi Production time 30 year Reservoir temperature 240 oF Initial water saturation 0.25 value Total compressibility 1×10-6 psi-1 Matrix permeability 5 µD Matrix porosity 0.08 value Stage spacing 340 ft Fracture conductivity 10 mD-ft Fracture half-length 350 ft Fracture height 40 ft

Table 3: Compositional data for the Peng-Robinson EOS in Bakken

Component Molar fraction

Critical pressure

(atm)

Critical temperature

(K)

Critical volume (L/mol)

Molar weight

(g/gmol)

Acentric factor

Parachor coefficient

CO2 0.0001 72.80 304.20 0.0940 44.01 0.2250 78.0 N2-C1 0.2203 45.24 189.67 0.0989 16.21 0.0084 76.5 C1-C4 0.2063 43.49 412.47 0.2039 44.79 0.1481 150.5 C5-C7 0.1170 37.69 556.92 0.3324 83.46 0.2486 248.5 C8-C12 0.2815 31.04 667.52 0.4559 120.52 0.3279 344.9 C13-C19 0.0940 19.29 673.76 0.7649 220.34 0.5672 570.1 C20-C30 0.0808 15.38 792.40 1.2521 321.52 0.9422 905.7

Table 4: Binary interaction parameters for Bakken oil

Component CO2 N2-C1 C1-C4 C5-C7 C8-C12 C13-C19 C20-C30 CO2 0 0.1013 0.1317 0.1421 0.1501 0.1502 0.1503

N2-C1 0.1013 0 0.0130 0.0358 0.0561 0.0976 0.1449 C1-C4 0.1317 0.0130 0 0.0059 0.0160 0.0424 0.0779 C5-C7 0.1421 0.0358 0.0059 0 0.0025 0.0172 0.0427 C8-C12 0.1501 0.0561 0.0160 0.0025 0 0.0067 0.0251 C13-C19 0.1502 0.0976 0.0424 0.0172 0.0067 0 0.0061 C20-C30 0.1503 0.1449 0.0779 0.0427 0.0251 0.0061 0

SPE-169575-MS 9

Table 5: Eight uncertain parameters used for sensitivity study

Parameter Minimum Base case Maximum CO2 injection rate, MSCF/day 50 100 500 CO2 injection time, month 3 6 9 Number of cycle 1 2 3 CO2 diffusivity, cm2/s 0.0001 0.0008 0.01 Fracture conductivity, md-ft 5 10 50 CO2 soaking time, month 1 3 5 Permeability, µD 2 5 8 Fracture half-length, ft 250 350 450

Fig. 1. Shale and tight oil and dry shale gas production in North America (Source: U.S. EIA).

Fig. 2. Location map of the Williston Basin with structure contours (Pilcher et al., 2012).

10 SPE-169575-MS

Fig. 3. Mass transfer between oil and gas components at the same pressure and temperature.

Fig. 4. Bottom hole pressure used as a constraint input for history matching.

Fig. 5. History matching for oil flow rate.

SPE-169575-MS 11

Fig. 6. History matching for gas flow rate.

(a) Water-oil relative permeability curve (b) Liquid-gas relative permeability curve Fig. 7. Relative permeability curves for a good history match.

Fig. 8. Pressure distribution at end of the field production.

12 SPE-169575-MS

Stage 1: CO2 injection

Stage 2: CO2 soaking

Stage 3: Production

Fig. 9. CO2 huff-n-puff process in a horizontal tight oil well with multi-stage fractures.

Case 1 Case 2 Case 3 Case 4

Fig. 10. Four cases with different numbers of fractures within single perforation stage.

SPE-169575-MS 13

(a) Cases 1 and 2 (b) Cases 3 and 4 Fig. 11. Comparison of oil recovery factor with and without gas injection.

Fig. 12. Comparison of incremental oil recovery factor for the four cases.

Case 1 Case 2 Case 3 Case 4

Fig. 13. CO2 molecule fraction distribution for the four cases.

14 SPE-169575-MS

Fig. 14. Effect of CO2 injection rate on oil recovery factor.

Fig. 15. Effect of CO2 injection time on oil recovery factor.

Fig. 16. Effect of number of cycle of CO2 huff-n-puff on oil recovery factor.

SPE-169575-MS 15

Fig. 17. Effect of CO2 diffusivity coefficient on oil recovery factor.

Fig. 18. Effect of fracture conductivity on oil recovery factor.

Fig. 19. Effect of CO2 soaking time on oil recovery factor.

16 SPE-169575-MS

Fig. 20. Effect of reservoir peremability on oil recovery factor.

Fig. 21. Effect of fracture half-length on oil recovery factor.

Fig. 22. Rank of the impacts of eight uncertain paramters on incremental oil recovery factor.