Evaluating the effects of CO2 Injection in Faulted Saline Aquifers

Available online at www.sciencedirect.com

ScienceDirect Energy Procedia 63 (2014) 3012 – 3021

GHGT-12

Evaluating the effects of CO2 Injection in Faulted Saline Aquifers David Alexandera,* Donnie Boodlalb * a,b

The University of Trinidad and Tobago, Esperanza Road, Brechin Castle, Point Lisas, Trinidad, West Indies

Abstract If carbon capture and storage (CCS) is to be an effective option for decreasing greenhouse-gas emissions, commercial-scale storage operations will require large storage capacities for carbon dioxide (CO2). Saline aquifers have been reported to have very large storage capacities. It is important for research to be conducted in this area, since the possibilities for storage in depleted oil reservoirs are reduced by the fact that they still contain large quantities of hydrocarbon, reducing the amount of pore spaces that could have potentially be available. Disposal of CO2 into deep saline aquifers involves CO2 being injected as a supercritical fluid that is less dense and less viscous than the formation water. Due to density differences, the CO2 becomes buoyant in the water and has a high tendency to migrate from the storage location once the possible pathways exist. This may lead to the critical problem of seepage from the storage area. In Trinidad and Tobago (T&T) many potential reservoirs are highly faulted. Some faults form an integral part of the structural traps whilst others are leaky and provide migration pathways for the injected CO2 to return to surface. A simulation study was conducted using the commercial compositional simulator CMG-GEM. The model described within this paper seeks to optimize the injection of CO2 into saline aquifers in the presence of faults Sensitivities for fault transmissibility, fault throw, fault permeability and bottomhole pressure were modelled to see their effects on the amount of CO2 stored with time. The model included the effects capillary entry pressure of the fault zone. We studied two fault types, i.e. normal faulting and reverse faulting. It was observed that fault throw and transmissibility of the fault had the most impact on the amount of CO2 that can be injected into a reservoir and ultimately stored. Solubility trapping played an essential role in securing CO2 in deep saline aquifers. For transmissibility values less than 0.01 all the CO2 injected into the reservoir remains stored. However the quantities that were stored are relatively small. Most of the stored CO 2 remains in the supercritical phase whilst approximately 10% remains dissolved in the aqueous phase. CO2 stored due to hysteresis was relatively small i.e. <1%. © 2014 byby Elsevier Ltd.Ltd. This is an open access article under the CC BY-NC-ND license © 2013 The TheAuthors. Authors.Published Published Elsevier (http://creativecommons.org/licenses/by-nc-nd/3.0/). Selection and peer-review under responsibility of GHGT. Peer-review under responsibility of the Organizing Committee of GHGT-12

Keywords: CO2 injection; fault types; saline aquifer; CMG-GEM; CO2 leakage

* Corresponding author. Tel.:868-370-3800; fax: 868-636-3339 E-mail address: [email protected]

1876-6102 © 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/). Peer-review under responsibility of the Organizing Committee of GHGT-12 doi:10.1016/j.egypro.2014.11.324

David Alexander and Donnie Boodlal / Energy Procedia 63 (2014) 3012 – 3021

1.0 Introduction 1.1 Main problem In recent times, there has been a growing concern that anthropogenic CO2 emissions released into the atmosphere globally may have a significant role to play in climate change. This has led to many researchers have placing immense emphasis on the development of safe and economical Geological Carbon Sequestration (GCS) technology. Suggested sites to store CO2 must be validated using demonstration projects to ensure that public acceptance can be gained. Although GCS is one of the most promising technologies to address the problem of anthropogenic globalwarming due to CO2 emissions, the detailed mechanisms are not well-understood [1]. As a result, there remain many uncertainties in determining the sequestration capacity of the formation and the safety of sequestered CO2 due to leakage. The goal of GCS is to maximize the sequestration capacity and minimize the plume migration by optimizing the GCS operation before proceeding with its large scale deployment [1]. Successful geological storage and sequestration of CO2 also require efficient monitoring of the migration of CO 2 plume during and after largescale injection [2]. CO2 injected into geological formations such as saline aquifers can be effectively immobilized by structural trapping, residual trapping, solution trapping and mineralization. Deep saline aquifers are reported to have the largest estimated capacity for CO2 sequestration [3, 4]. The sequestration process can be broadly broken into two phases. The first phase is the gas injection phase which typically lasts from 10-100 years depending on the size of project. During this phase, the CO2 displaces the brine in the pore space. A portion of the CO 2 dissolves into the brine, though most of the injected gas remains in the gaseous phase. The brine with the dissolved CO2 is denser than the original in situ brine and sinks towards the bottom. In the second phase, there is no more CO 2 injection. The density difference between the CO2 and the brine causes the CO2 to migrate upwards to the top of the geologic structure. The sealing faults and cap rock can stop the further upward movement of the gas, trapping it within the formation. This type of trapping is not the preferable trapping mechanism for long-term CO2 storage because the CO2 is still mobile and any loss in the seal integrity of the cap rock could cause it to leak from the formation. There is another trapping mechanism which is important in this phase called residual trapping. As the CO2 migrates upwards, it displaces the water while water replaces the CO2. We therefore have both imbibition and drainage occurring simultaneously. Due to hysteresis in the relative permeability curves and the residual gas saturation, generally a significant amount of CO2 gets trapped in the pores as an immobile phase [5, 6, 7, 8].

Nomenclature Fk T Tf BHP Cum Inj Accum SGR

fault permeability, mD fault throw, ft fault transmissibility Bottomhole pressure, psi Cumulative CO2 injected, MMSCF Amount of CO2 stored in the reservoir, MMSCF Shale Gouge Ratio

1.2 Background T&T has been a producer of hydrocarbons for over 100 years [9]. Simultaneously, T&T has emitted over 56 million metric tons of CO2 per annum, due to its growing petrochemical industry and power generation [10]. Whilst this figure on the global scale is less than 1%, the impact of climate change on small island states such as T&T can be detrimental. In T&T, faults and sands in the area have been reasonably well characterized for oil and gas exploration. Coincidentally, the islands have several surface seeps of hydrocarbons, some of which are associated with leakage from deeper reservoirs [11]. Trinidad has one of the most complex geology in the world. It is located

3013

3014

David Alexander and Donnie Boodlal / Energy Procedia 63 (2014) 3012 – 3021

on the edge of three plates. The geology is complex and made up of folded (anticlines) and faulted structures [12]. Some faults and anticline structures are structural traps, which prevent upward and lateral movement of fluid in the reservoir. Over time, sediments such as clay and silt, under high pressures and temperatures move and become compressed within the fault, changing its permeability and forming a barrier or seals. However, according to Harris et al., (1999) [13] sealing faults are temporary. They showed that hydrocarbons can migrate across an inactive fault zone, when the fluid pressure difference is above the critical entry pressure [13]. Faulted reservoirs increase the complexity in determining the outcome of CO2 flooding along with these parameters mentioned above. This paper presents a preliminary case study on a field “X” using data analogous to a field located in the southeast coast of Trinidad and comprises a series of stacked sands in separate major fault blocks. This paper concentrates on the outcome of injected CO2 into a deep saline aquifer where the reservoir is faulted by a major fault. For this reservoir, we take into consideration the effects of fault transmissibility, variable permeability in the fault zone, a range of fault throw and varying bottomhole pressure. 1.3 Fault description It is important to describe the role of faults in this paper because of the significant impact it may have on CO2 storage. Caine et al., [14] suggest that faults that form in brittle host rock (e.g., sandstone) comprise of two distinct components: the fault core and the damaged zone. The fault core, where most fault displacement is accommodated, contains fault rocks that have undergone the greatest degree of deformation. Faults can restrict fluid flow as is often observed in petroleum reservoirs [15]. Buoyancy driven CO 2 migration is stopped by a fault, and an accumulation forms behind it. This accumulation grows until eventually the capillary pressure matches the capillary threshold pressure of the fault, and then CO2 will continue to migrate into other adjacent formations. Fault zone properties are incorporated in production flow simulators using transmissibility multipliers. These are a function of properties of the fault zone and of the grid-blocks to which they are assigned [16]. Faults influence flow in a reservoir simulation model in that hey alter the connectivity of sedimentological flow units. Displacements across faults can cause partial or total juxtaposition of different flow units, possibly connecting stratigraphically disconnected high permeability units, as well as juxtaposing high against low permeability units. For the fault zone permeability, we consider the equation below developed by Manzocchi et al., [16]: 1 5 (1) log k 4SGR  log  D 1  SGR f

Where

4

kf

fault permeability, mD

D

fault displacement, m

Gibon and Bentham [17] evaluated the controls on fault seal using shale gouge ratio (SGR) for the Columbus basin. It was found that a transition between sealing and non-sealing faults occurring in the SGR = 0.15– 0.25 range. We also followed after Manzocchi et al., [16] that the transmissibility multiplier is calculated as a function of the dimensions and permeability of the grid-blocks and the thickness and permeability of the fault: ª «1  t f ¬«

Where

§ 2 k f  1 ki 1 k j T ¨¨ © Li ki  L j k j T = transmissibility multiplier t = fault thickness L = grid block length k = fault permeability

·º ¸¸ » ¹ ¼»

1

(2)

The maximum residual gas saturation Sgrm was calculated using the correlation developed by Holtz [18]: S grm

0.9696I  0.5473

(3)

David Alexander and Donnie Boodlal / Energy Procedia 63 (2014) 3012 – 3021

This correlation is used to calculate the HYSKRG value in the simulation model. We modelled the solubility of CO2 in water using Henry’s correlation; see [18]. 2.0 Approach for the simulation model In this model, part of a field analogous to those offshore Trinidad was built using two fault blocks labelled FB1 and FB2. Each fault block has 3 different sands having different permeabilities labelled S1,S2 and S3.

CO2 inj 1 producer

Figure 1: Sand shale sequence with different horizontal permeabilities (normal fault)

CO2 inj 1

producer

Figure 2: Sand shale sequence with different horizontal permeabilities (reverse fault)

3015

3016

David Alexander and Donnie Boodlal / Energy Procedia 63 (2014) 3012 – 3021

The three sand bodies had average permeabilities of 700, 500 and 300 mD respectively, whilst that of the shale was 0.01 mD; see Figure 1 (illustrates the model for normal faulting). Figure 2. above illustrates how the model will look when reverse faulting occurs. The overall porosity used was 30% for the sand, shale and fault. This field is assumed to be located on the southeast coast of Trinidad in the Columbus Basin within the Eastern Venezuelan Basin. The faults in this area can be viewed as the main conduits for flow of the injected CO2 and the displaced reservoir fluids. For convenience in this preliminary study we consider a 2D reservoir using Cartesian grids to represent this field. The width, thickness and depth to the top were 4200, 2000 and 8600 ft respectively. The temperature and pressure gradients were 25oC/Km and 0.465 psi/ft respectively. The model was built with a sand-shale sequence. The blue areas in Figure 1 and Figure 2 represented the shale layers. The reservoir is gridded using 10 grid blocks in the x-direction and 1 in the y-direction and 20 in the z-direction. A total of 200 grid blocks are used in the simulator. Fault grid blocks with sides as small as 2 feet in the x-direction are included in the model. This fault zone in this model accounted for the fault core and the damaged zone. Such fine grid blocks were used to evaluate and capture the small physical size of fault zone details. We use two cases for the fault permeability: 1) a high-permeability fault (100mD and 2) a low-permeability fault (10mD). The model assumes that the reservoir is not tilted and the only fluid present initially is a large aquifer. The boundaries of the domain were closed. In this study we assume that the injection well was located in fault block 1 (FB1) whilst the producer was in fault block 2 (FB2). CO2 was injected continuously, in all three sands in FB1 for a period of 25 years whilst the producer, perforated in S1, S2 and S3, was allowed to produce simultaneously for the same time period. After this, the simulation was allowed to run a total of 1000 years to see the migration of the CO2 due to buoyancy and the impact of different trapping mechanisms over time. This enabled us to observe the effects of injected CO2 in the presence of a fault on the production of CO 2 and also the long term migration of the said gas due to the petrophysical properties of the fault rock (permeability along the fault, and permeability across the fault). These different sensitivities were done for a normal fault and a reverse fault to observe the effects on the amount of CO2 stored in the reservoir with time. We used the relative permeability curves described in Figures 3-5. 1

1

1.2 1

0.8

Krw Krg

0.8

Krw

0.6 0.4 0.2

0.6

0.6

0.4

0.4

0.2

0.2

0

0.2

0.4

0.6

Krg

0

0

0 0

Krw

0.8

Krg

0.8

0.1

0.2

0.3

0.4

0.5

Gas saturation

0

0.2

0.4

0.6

0.8

1

Gas saturation

Gas saturation

Figure 3—Relative permeability curves for the sand region

Figure 4: Relative permeability curves for the shale region

Figure 5. Relative permeability curves for the fault region.

The relative permeability data used for the sand and shale regions were taken from data published in Bennion and Bachu [19] shown in Figures 3 (data for the Basal Cambrian Sandstone) and 4 (data for the Wabamun Carbonate; low relative k) respectively. A fracture gradient of 0.72 psi/ft was used to estimate the fracture pressure. We also varied bottom-hole pressures (BHP) to represent the different injection rates of 4500 psia, 5000 psia and 5500 psia. These values were selected to represent injection rates below the fracture pressure. We varied the fault throw using the range 50 - 400 ft. We injected at a constant surface gas rate of 80 MSCF/day. Two types of fault-related seals are considered here i.e. whether the fault plane itself acts as the seal or whether a sealing unit is juxtaposed against a trapped CO2 column across a fault (juxtaposition fault). We modelled two types of faulting i.e. normal faults and reverse faulting using all the variables mentioned above.

David Alexander and Donnie Boodlal / Energy Procedia 63 (2014) 3012 – 3021

In this model, we used capillary curves for the fault zone as represented in Figure 6. 3.5 3 2.5 log Pc

2 1.5

Sand Shale

1

Fault

0.5 0 0

0.2

0.4

0.6

0.8

1

Sw Figure 6: log Pc vs. Sw

3.0 Results and Discussion. A base case with no faults existing in the reservoir was run initially so that a comparison could be made between a non-faulted reservoir, one that has normal faulting and another that has a reverse fault to see relative effects on the amount of CO2 stored in a 2D reservoir; see Figure 7.

Figure 7: Bottomhole pressure varied – (no faulting)

The only parameter varied for the base case was the BHP. For this base case, it was observed that changing the BHP between 4500-5500 psi didn’t make any significant impact on the amount of CO2 trapped in the reservoir. 33% of the injected CO2 remained in the reservoir. Over 90% of the accumulated CO2 trapped in the reservoir remained in the supercritical phase, whilst 9% was dissolved in the aquifer. In this base case, no CO2 was trapped due to hysteresis. We then placed a fault into the model which was located approximately 2500ft away from the injector. Sensitivities were conducted on fault permeability, fault throw, fault transmissibility and bottomhole pressure for

3017

3018

David Alexander and Donnie Boodlal / Energy Procedia 63 (2014) 3012 – 3021

both a normal fault and a reverse fault. Both models had the same physical and petrophysical properties except for the type of faulting; see Figure 1 and Figure 2. When fault permeability was varied for the normal fault, there was no significant difference in the amount of CO2 stored in the reservoir. For this case, as fault permeability increased, there was a slight decrease (from 45-42%) in the amount of CO2 stored; see Figure 8.

Figure 8: Fault permeability varied – (normal faulting)

Figure 9: Fault permeability varied – (reverse faulting)

This may be as a result of the fault being more susceptible to fluid flow and hence a lower amount remaining in the reservoir. However, for the same parameters for the reverse fault, there was a slight reduction in the amount of CO2 injected and hence stored; see Figure 9. The overall % of CO2 stored for both scenarios was between 40-45%. We then varied fault throw. The general trend for normal faulting was, as fault throw increased, the amount of CO2 that could be injected also decreased gradually. The same trend followed for the amount stored. When the same was applied to the reverse fault, there was an exponential decrease in the amount of CO2 that could have been injected and hence the amount stored; see Figure 10 and Figure 11 respectively below.

Figure 10: Fault throw varied – (normal faulting)

Figure 11: Fault throw varied – (reverse faulting)

From further investigation into the graphical results for the model, it was seen that as fault throw varied, it also varied the juxtaposition of sand/sand and hence according to how large the fault throw was, juxtaposed sand on sand now because juxtaposed sand on shale. This would have significantly reduced the ability of CO2 to flow out of the storage area. It can therefore be concluded that sand on shale juxtapositions are better traps however, they significantly reduce the amount of CO2 that can be stored in a reservoir.

David Alexander and Donnie Boodlal / Energy Procedia 63 (2014) 3012 – 3021

We then varied the fault transmissibility. A value of 0 transmissibility implies a sealing fault and hence flow across the fault zone is considered restricted. A value of 1 for the transmissibility implies that the fault zone is free flowing, similar to the rock matrix. For small transmissibility values less than 0.01, the amount of CO 2 injected was equivalent to the amount stored. This is because the injected CO 2 did not reach the producer well (used in our model to create void pore space for CO2 storage) hence all remained in the reservoir. However, as transmissibility became greater than 0.01, the amount of CO 2 injected and hence stored increased linearly with a sharp gradient; see Figure 12 and Figure 13.

Figure 12: Fault transmissibility varied – (normal faulting)

Figure 13: Fault transmissibility varied – (reverse faulting)

It’s noteworthy that for transmissibility values between 0.01 and 0.1 seems to be the best range for CO 2 storage. For transmissibility values greater than 0.1, although the amount of CO2 injected increases with a high gradient, the amount of CO2 stored remained almost the same as 0.1. Faults with high transmissibility values i.e. greater than 0.1 therefore do not auger well for storage. When BHP was varied, the general trend was as BHP increased, the cumulative CO 2 injected and hence the amount stored increased; see Figure 14 and Figure 15 below.

Figure 14: Bottomhole pressure varied – (normal faulting)

Figure 15: Bottomhole pressure varied – (reverse faulting)

3019

3020

David Alexander and Donnie Boodlal / Energy Procedia 63 (2014) 3012 – 3021

It was important to note however, for the normal faulting, more CO 2 was stored in the lowest BHP compared to that of the similar reverse faulting model. For the lower BHP, it was observed that the percent stored was much greater than that of the highest BHP at 5500 psia. For this model, 5000 psia would have been optimum since the greatest absolute amount of CO2 was stored even though at 4500 psia we had the highest ratio of cumulated CO 2 injected/CO2 stored. It was observed that with most models, the cumulative CO 2 injected and hence stored was generally less for the reverse fault when compared to the normal fault. 4.0 Conclusion Fault throw and transmissibility of the fault have the most impact on the amount of CO2 that can be injected into a reservoir and ultimately stored. Solubility trapping is essential for securing CO2 in deep saline aquifers. For transmissibility values less than 0.01, the all the CO2 injected into the reservoir remains stored. However, the quantities that are stored are relatively small. From our models, faults with high transmissibility values i.e. greater than 0.1 do not augur well for storage. Most of the stored CO2 remains in the supercritical phase whilst approximately 10% remains dissolved in the aqueous phase for most cases. For our models, CO 2 stored due to hysteresis was relatively small i.e. <1% of the total amount stored. Acknowledgements This research was supported by the University of Trinidad and Tobago. We thank CMG Ltd. for making available the GEM-GHG simulator used in this research. References [1] Zhang, Z. (2013). "Numerical Simulation and Optimization of CO2 Sequestration in Saline Aquifers". Electronic Theses and Dissertations. Paper 1097. [2] Lu, C., Zhang, C, Hunag, H., and Johnson, T. C. (2014). Monitoring CO2 sequestration into deep saline aquifer and associated salt intrusion using coupled multiphase flow modeling and time-lapse electrical resistivity tomography. Society of Chemical Industry and John Wiley & Sons. Ltd | Greenhouse Gas Sci Technol. 4:1–16 (2014); DOI: 10.1002/ghg. [3] Soong, Y., Howard, B. H., Hedges., S. W., Haljasmaa, I, Warzinski, R. P., Thomas G.I., McLendon, R. (2014). CO2 Sequestration in Saline Formation. Aerosol and Air Quality Research, 14: 522–532, 2014. [4] Gale, J. (2002). “Overview of Sources, Potential, Transportation and Geological Distribution of Storage Possibilities,” Presentation at IPCC Workshop on Carbon Capture and Storage, Regina, Canada, November 18-21. [5] Juanes, R., Spiteri, E. J., Orr, F. M., Jr., and Blunt, M. J. (2006). “Impact of Relative Permeability Hysteresis on Geological CO2 Storage,” Water Resour. Res., 42,W12418, DOI: 10.1029/2005WR004806. [6] Kumar, A., Noh, M., Pope, G. A., Sepehrnoori, K., Bryant, S., and Lake, L.W. (2005).“Reservoir Simulation of CO2 Storage in Deep Saline Aquifers,” SPE J., 10(3),336 – 348. [7] Mo, S., and Akervoll, I. (2005). “Modeling Long-Term CO2 Storage in Aquifer with a Black-Oil Reservoir Simulator,” paper SPE 93951 presented at SPE/EPA/DOE Exploration and Production Environmental Conference, Galveston, March 7–9. [8] Deepanshu K. (2007). Optimization of Well Settings to Maximize Residually Trapped CO 2 in Geologic Carbon Sequestration.MS thesis.Stanford University. [9] Geological Society of Trinidad and Tobago “Petroleum Geology” retrieved from www.gstt.org on 1 April 2014. [10] Boodlal D.V. and Al Taweel. A. (2012). “Options for Sustainable Reduction of GHG: A case study for Trinidad and Tobago” Presented at the 62nd Canadian Chemical Engineering Conference, Vancouver, Monday 15th October, 2012. [11] Alexander, D. and Bryant, S.L. (2009). “Evaluating storage and leakage scenarios for carbon dioxide sequestration in Trinidad and Tobago”. Science Direct Energy Procedia, 2761-2768. [12] Ministry of Energy and Energy affairs Trinidad and Tobago “Petroleum Geology” retrieved from www.energy.gov.tt on 3March 2014. [13] Harris. S.D., Elliott and L. Knipe, R.J. (1999): “The pulse migration of hydrocarbons across inactive faults”. University of Leeds. Hydrology and Earth System Sciences, 3(2), 151-175. [14] Caine, J.S., Evans, J.P., and Forster, C.B.. (1999). “Fault Zone Architecture and Permeability Structure”, Geology, v. 24,1025-1028. [15] Watts, N. L. (1987). Theoretical aspects of cap-rock and fault seals for single- and two-phase hydrocarbon columns. Marine and Petroleum Geology, 4, 274-307. [16] T. Manzocchi, Walsh, J.J., Nell, P., and Yielding, G. (1999). Fault Transmissibility for Flow Simulation Models. Petroleum Geoscience, Vol., 5, pp. 53-63. [17] Gibson., R. G. and Peter A. and Bentham. P. A. (2003). “Use of fault-seal analysis in understanding petroleum migration in a complexly

David Alexander and Donnie Boodlal / Energy Procedia 63 (2014) 3012 – 3021 faulted anticlinal trap, Columbus Basin, offshore Trinidad.” AAPG Bulletin, v. 87, no. 3 (March 2003), pp. 465–478. [18] Holtz. M.H. (2002). Residual Gas Saturation to Aquifer Influx: A Calculation Method for 3-D Computer Reservoir Model Construction. SPE 75502. [18] GEM Manual, CMG software. (2013). [19] Bennion, D.B. and Bachu, S. (2008). Drainage and Imbibition Relative Permeability Relationships for Supercritical CO2/Brine and H2S/Brine Systems in Intergranular Sandstone, Carbonate, Shale and Anhydrite Rocks. SPE 99326.

3021