Experimental Study of Crossover from Capillary to Viscous Fingering

Jun 5, 2012 - State Key Laboratory for Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy of. Sciences ...
0 downloads 0 Views 3MB Size
Article pubs.acs.org/est

Experimental Study of Crossover from Capillary to Viscous Fingering for Supercritical CO2−Water Displacement in a Homogeneous Pore Network Ying Wang,† Changyong Zhang,*,‡ Ning Wei,† Mart Oostrom,‡ Thomas W. Wietsma,‡ Xiaochun Li,† and Alain Bonneville‡ †

State Key Laboratory for Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy of Sciences, Wuhan 430071, China ‡ Pacific Northwest National Laboratory, 902 Battelle Boulevard, P.O. Box 999, MSIN K8-96, Richland, Washington 99352, United States S Supporting Information *

ABSTRACT: Carbon sequestration in saline aquifers involves displacing brine from the pore space by supercritical CO2 (scCO2). The displacement process is considered unstable due to the unfavorable viscosity ratio between the invading scCO2 and the resident brine. The mechanisms that affect scCO2−water displacement under reservoir conditions (41 °C, 9 MPa) were investigated in a homogeneous micromodel. A large range of injection rates, expressed as the dimensionless capillary number (Ca), was studied in two sets of experiments: discontinuous-rate injection, where the micromodel was saturated with water before each injection rate was imposed, and continuous-rate injection, where the rate was increased after quasi-steady conditions were reached for a certain rate. For the discontinuous-rate experiments, capillary fingering and viscous fingering are the dominant mechanisms for low (logCa ≤ −6.61) and high injection rates (logCa ≥ −5.21), respectively. Crossover from capillary to viscous fingering was observed for logCa = −5.91 to −5.21, resulting in a large decrease in scCO2 saturation. The discontinuous-rate experimental results confirmed the decrease in nonwetting fluid saturation during crossover from capillary to viscous fingering predicted by numerical simulations by Lenormand et al. (J. Fluid Mech. 1988, 189, 165−187). Capillary fingering was the dominant mechanism for all injection rates in the continuous-rate experiment, resulting in monotonic increase in scCO2 saturation.



INTRODUCTION Geological sequestration of CO2 in deep saline aquifers and depleted oil and gas reservoirs represents an important approach to mitigate greenhouse gas emission into the atmosphere.1,2 Such an approach involves injecting compressed CO2 into deep permeable geological formations and causes large volumes of formation fluids (e.g., brine) to be displaced from the pore space during the initial stage of the deployment. Among the various factors that may limit the storage efficiency, hydraulic properties of the porous media and interfacial properties of CO2 and formation brine are expected to play important roles in affecting fluid flow, plume development, and CO2 saturation levels. In particular, the viscosity of supercritical CO2 (scCO2) under reservoir conditions (i.e., high temperature and pressure) is much lower than the formation brine being displaced from the pore space, resulting in a potentially unstable displacement process and low scCO2 saturation. Extensive research on geological storage of CO2 has been performed over multiple spatial scales by means of both numerical models and laboratory experiments.32 Numerical models have been developed to simulate large-scale deployment of CO2 sequestration.3−9 These modeling studies have © 2012 American Chemical Society

shown the importance of both porous media properties and fluid properties in affecting CO2 flow pathways, plume development, and storage capacity. However, uncertainties still exist for reliable predictive modeling due to the limited data available for model validation. Recent laboratory experiments on CO2−water or CO2−oil displacement provided valuable insights into processes that control CO2 migration in porous media. Several studies primarily focused on core flooding experiments using various visualization techniques and revealed that, among other factors, structural heterogeneity of rock core samples influences CO2 migration and saturation.10−14 Due to the extreme conditions required for studying scCO2−water displacement processes at the microscopic scale (i.e., pore scale), only a limited number of experiments have been performed to investigate the fundamental interfacial processes Special Issue: Carbon Sequestration Received: Revised: Accepted: Published: 212

April 11, 2012 June 4, 2012 June 5, 2012 June 5, 2012 dx.doi.org/10.1021/es3014503 | Environ. Sci. Technol. 2013, 47, 212−218

Environmental Science & Technology

Article

at this scale.15−18 These pore-scale studies have demonstrated the importance of small-scale processes that are often neglected in continuum-scale models, such as unstable fingering17 and wettability alteration,18 on CO2 migration and saturation. Porescale experiments also permit evaluation of constitutive relationships that are used in large scale simulators.28 Over the last three decades, fundamental understanding of the mechanisms that control immiscible two-phase flow under ambient pressure and temperature conditions has been gained through pore-scale experiments using microfabricated physical models of porous media (i.e., micromodels) and a variety of pore network modeling approaches.1,19−22 Displacement of a wetting phase by a nonwetting fluid in a two-dimensional (2D) pore network, commonly referred to as main drainage, is often described by two dimensionless numbers: (i) the capillary number Ca = (μn × un)/(σnw × cosθ), where μn and un are the viscosity and velocity of the invading nonwetting fluid, respectively, σnw is the interfacial tension, θ is the contact angle; and (ii) the viscosity ratio M = μn/μw, where μw is the viscosity of the initially residing wetting fluid. Depending on Ca and M, different mechanisms, including capillary fingering, viscous fingering, and stable displacement, have been shown to control immiscible displacement at the pore scale.1,23 Capillary fingering occurs in the form of wide forward and lateral nonwetting phase flowpaths when the capillary force of the wetting fluid is the dominant force. Viscous fingering occurs when the viscosity of the nonwetting fluid is lower and is characterized by narrow forward progressing flowpaths. For immiscible fluid displacement with low viscosity ratios (i.e., logM < 0) consistent with scCO2−water systems, a crossover zone with coexisting capillary and viscous fingering has been predicted numerically using invasion percolation with trapping and diffusion-limited aggregation models.1,19,20 More specifically, using a pore network simulator, Lenormand et al. predicted, for a fluid pair with logM = −4.7, that the saturation of the invading nonwetting fluid decreases significantly during crossover from capillary to viscous fingering.1 These authors have also qualitatively demonstrated the crossover phenomenon for oil displacement by mercury (logM = −1.8) in an oil-wet pore network micromodel.1 Because of the undesirable effect on the injected nonwetting phase saturation, it is important to investigate the potential impact of crossover from capillary to viscous fingering for scCO2−water displacement (logM < −1). The objective of this study was to investigate the unstable displacement mechanisms relevant to scCO2 sequestration at the pore scale under reservoir conditions. Specifically, we focused on scCO2−water displacement in a homogeneous pore network with well-defined surface properties and pore geometry over a broad range of injection rates (logCa = −7.61 to −4.73) under 9 MPa pressure at 41 °C. The displacement process was imaged using fluorescent microscopy, and the scCO2 saturation was determined from fluorescent images. Results obtained from experiments were compared with the pore network model simulations results of crossover from capillary to viscous fingering obtained by Lenormand et al.1 In a recent study, we demonstrated that unstable capillary and viscous fingering dominated liquid CO2 displacement of water in a dual permeability pore network micromodel.17 The present study expands our previous work17 as experiments were conducted at a temperature above the supercritical point over a broader range of injection rates. The use of higher temperatures allows us to investigate impacts of different injection methods and injection rates on displacement

mechanisms and pore-level scCO2 saturations that are directly relevant to geological CO2 sequestration. The large range in flow rates corresponds to flow conditions that may be experienced at various distances from an injection well.



EXPERIMENTAL METHODS Micromodel. The micromodel used in this study was fabricated in a silicon wafer using microfabrication methods involving photolithography, plasma etching, and anodic bonding as described in the literature.24,25 The micromodel consists of a 1.2 × 1.2 cm2 homogeneous network of cylindrical grains (200 μm diameter) and pore spaces (120 μm pore bodies, 26.7 μm pore throats), an inlet, and an outlet chamber (Figure S1a in the Supporting Information). The porosity of the micromodel is approximately 0.40 and the average pore depth is 35 μm with a variation of ±1 μm as a result of nonuniform plasma etching.23 The micromodel was thermally oxidized26 to obtain hydrophilic surfaces, which was confirmed by contact angle measurements (Table 1 and Figure S2 in the Supporting Information). Table 1. Summary of Experimental Conditions, Fluid Properties, Volumetric Flow Rates, and Corresponding Darcy Velocities and Capillary Numbers Q (μL/h)

fluid properties pressure [MPa/psi] temperature [°C] viscosity scCO2/water [mPa·s] logM [-] interfacial tension [mN·m−1] density scCO2/water [g/cm3] scCO2/water contact angle [degree] scCO2 entry pressure [Pa] a

30 b

9/1305 41 ± 1 0.036a/ 0.64 −1.25 26.2b 0.49a/0.99 14.7 ± 0.3c 3346d

v (m/d)

logCae (-)

10 50 100

0.57 2.83 5.67

−7.61 −6.91 −6.61

500 1000 2500 5000

28.33 56.67 141.67 283.35

−5.91 −5.61 −5.21 −4.91

7500

425.03

−4.73

31 c

Nordbotten et al. Bachu and Bennion. Measured in this study. An image of the scCO2−water contact angle with micromodel surface in the outlet channel is included in the Supporting Information (Figure S2). dComputed using equation Pc = (1/r1 + 1/r2)σnw cos θ. e Ca number determined using IFT value obtained from ref 31 and measured contact angle value.

ScCO2−Water Displacement Experiment Setup and Procedures. A high-pressure micromodel experimental system17 was modified to allow the entire system to be temperature controlled (Figure S1b). The system includes the following three main components: (i) a high-pressure fluid delivery system containing three ISCO pumps (Teledyne ISCO Inc., Lincoln, NE) and a stainless steel overburden pressure cell with a sapphire viewing window (viewable diameter = 3.6 cm); (ii) two heating elements including a temperature chamber (In Vivo Scientific, St. Louis, MO) surrounding the microscope stage thermostatted by circulating hot air, and a water heater (Fisher Scientific), allowing all pump cylinders and fluid delivery lines to be thermostatted by circulating hot water; (iii) an epifluorescent microscope equipped with an automated stage, inverted objectives, and a digital CCD camera. The micromodel sits on the sapphire window in the overburden pressure cell, which is placed horizontally on the microscope stage (Prior Scientific Inc., Rockland, MA) inside the temperature chamber. The micromodel inlet is connected to pump A (ISCO 100 DM) containing scCO2, and the outlet is 213

dx.doi.org/10.1021/es3014503 | Environ. Sci. Technol. 2013, 47, 212−218

Environmental Science & Technology

Article

Figure 1. ScCO2 (blue) distribution in the micromodel for the discontinuous-rate experiments. The scCO2 flow is from left to right. The numbers in parantheses indicate the logCa and injected scCO2 pore volumes, respectively.

connected to a back pressure pump B (ISCO 100 DM) filled with water. The overburden pressure cell is filled with glycerol (Sigma Aldrich, St. Louis, MI) and connected to pump C (ISCO 500 D). Before each experiment, the micromodel was first cleaned using a basic solution (DI water:NH4OH:H2O2 at 5:1:1) and then saturated with degassed DI water. Pump A was filled with supercritical grade CO2 to an initial pressure of 5.9 MPa and pressurized to 9.0 MPa while valve 1 was closed. The pressure inside the micromodel and the overburden pressure cell were sequentially increased to 9.0 and 9.5 MPa, respectively, by simultaneously pressurizing Pumps B and C at 172 KPa increments over approximately a 1.5 h interval. The entire system was thermostatted to 41 °C and the temperature was allowed to equilibrate for >1.5 h before an experiment. The scCO2 was injected from pump A to the micromodel inlet at a constant volumetric flow rate (Q). The back pressure pump B connecting to the outlet was simultaneously switched to refill mode with the same constant volumetric flow rate as pump A. The displacement process was imaged through the microscope and camera. A total of eight different volumetric injection rates ranging from 10 to 7500 uL/h were investigated (Table 1). The imposed rates correspond to a range in Darcy velocity from 0.57 to 425 m/day, and a range in logCa from −7.61 to −4.73. Two independent sets of experiments were conducted according the following injection methods: (i) discontinuousrate injection: scCO2 was injected into the micromodel at a certain constant flow rate until quasi-steady state was reached (i.e., the saturation of the scCO2 did not change over time); the experiment was then stopped and the micromodel was thoroughly cleaned and saturated with water before the next experiment at a higher injection rate; and (ii) continuous-rate injection: scCO2 was first injected at the lowest flow rate which was sequentially increased to the highest flow rate once quasisteady state was reached at each rate. The discontinuous-rate experiments were studied with an aim to evaluate pore-scale displacement mechanisms and scCO2 saturation levels under different flow conditions that may occur at various distances from an injection well. For example, lower injection rates in the discontinuous-rate experiments correspond to flow rates at locations 100 m away from an injection well (Table S1 in the Supporting Information). The continuous-rate experiments represent an alternative injection approach that can be explored

in order to increase pore-level scCO2 saturation. All experiments were performed in the same micromodel, with two replicates for selected injection rates. A summary of the experimental conditions is included in Table 1. It should be noted although the viscosity of formation brine and scCO2− brine interfacial tensions are higher than listed in the table, the viscosity ratio (M) is not expected to be substantially different. Although buoyancy is an important force affecting CO2 migration and plume development, this force is not considered in this study because of its small magnitude relative viscous force,1,12,20 the 2D horizontal configuration of the micromodel, and its limited vertical dimension. Image Acquisition and Analysis. Direct visualization of the scCO2 distribution and quantification of saturation in the micromodel pore network is made possible by adding a fluorescent dye, Coumarin 153 (99.99%; Alfa Aesar, Ward Hill, MA), to the scCO2 phase with a concentration of ∼0.01 mM. The low dye concentration is not expected to change the interfacial properties24 or the surface wetting property, as reflected in the measured contact angle values in Table 1. Fluorescent images of the scCO2 in micromodel were acquired using a Nikon Eclipse-2000TE (Nikon, Melville, NY) epifluorescent microscope through a 4× inverted objective (1.6 μm pixel size) and a Blue GFP filter set with excitation and emission wavelengths of 379−401 nm and 435−485 nm, respectively. At each flow rate, an image of the micromodel pore network was obtained by montaging a total of 48 (6 × 8) separate images acquired using a CoolSnap HQ2 monochrome CCD camera (Photometrics Inc., Tucson, AZ). The percentage of the pore space occupied by scCO2 (i.e., scCO2 saturation, SscCO2) was quantified using a method combining image segmentation and pixel counting described in previous studies.17,23 All images were acquired during scCO2 injection after quasi-steady state saturations were reached. Due to the wide range of flow rates and the time required to collect an image covering the entire micromodel, it was impractical to capture images at a fixed injection volume for all injection rates. To demonstrate that the quasi-steady state saturations did not increase with injection volume for a given rate, a sequence of images acquired for the relatively slow 10 μL/h experiment showed minimal change in scCO2 saturation after approximately one pore volume (PV) of scCO2 was injected into the pore network (Figure S3). Evidence that the saturations also do not change at a higher injection rate (5000 μL/h) is presented 214

dx.doi.org/10.1021/es3014503 | Environ. Sci. Technol. 2013, 47, 212−218

Environmental Science & Technology

Article

Figure 2. ScCO2 distribution in the micromodel for the continuous-rate experiments. The scCO2 flow is from left to right. The numbers in parantheses indicate the logCa and cumulative injected scCO2 pore volumes, respectively.

flowpath transitioned into one gradually narrowing finger leading to the outlet. The preferential flowpath is attributed to small variations in pore depth. Correspondingly, SscCO2 increased slightly from ∼0.43 at logCa = −7.61 to 0.53 at logCa = −6.61 (Figure 3). At higher injection rates (logCa > −5.21), displacement is controlled by the viscous force of the scCO2, and the scCO2 entered the pore network at several locations in the form of narrow flowpaths (i.e., 1−3 pore bodies) distributed over the entire width of the micromodel and mainly progressed forward from inlet to outlet, indicating that viscous fingering is the dominant displacement mechanism. The SscCO2 remained nearly constant (∼0.50) over the three highest imposed injection rates (Figure 3). At the intermediate injection rates (logCa = −5.91, −5.21), crossover from capillary to viscous fingering is observed: the scCO2 front entered the pore network in the form of wide forward and lateral flowpaths, as in the lower displacement rates, that gradually transitioned to narrower forward progressing flowpaths, similar to what was observed at the higher rates (Figure 1). A large decrease in SscCO2 occurred in the crossover zone, from 0.53 at logCa = −6.61 to less than 0.30 at logCa = −5.91 to −5.61 (Figure 3). The behavior of scCO2 saturation vs injection rate observed in this study is consistent with pore network simulations obtained by Lenormand et al., who showed when crossover occurred, the nonwetting fluid saturation decreased between two plateaus corresponding to capillary and viscous fingering, respectively.1 These predictions were included as an inset in Figure 3. It is worth noting that logM for scCO 2 −water under the experimental conditions of this study is −1.25, which is more than 3 orders of magnitude higher than the numerical model study (logM = −4.7), suggesting crossover phenomena may occur over a wide range of logM values. The difference in the nonwetting fluid saturation and range of logCa values between model and experiments results from differences in fluid properties and pore network inhomogeneity. Several authors have also qualitatively demonstrated the crossover behavior in air−water displacement21 and mercury−oil displacement experiments1 at ambient conditions. These results indicate potentially profound impacts of pore-scale processes on unstable displacement of water by scCO2. Large-scale simulation of CO2 sequestration using constitutive relative permeability and capillary pressure relations may benefit from incorporation of pore-scale processes through consideration of

in Figure S4, where images acquired after different volumes of scCO2 were injected for two replicate experiments showed good agreement, i.e., saturation values within 3% of each other.



RESULTS AND DISCUSSION ScCO2−Water Displacement Mechanisms vs scCO2 Saturations. Fluorescent images of scCO2 distribution in the micromodel pore network after quasi-steady state was reached at each injection rate (expressed as logCa) are shown in Figures 1 and 2 for the two injection scenarios. The corresponding volumes of scCO2 injected (expressed as pore volume, PV) are also shown in the figures. Results of scCO2 saturation (SscCO2) versus injection rate are shown in Figure 3 for both scenarios.

Figure 3. ScCO2 saturation vs injection rate (logCa) for both the continuous- and discontinuous-rate experiments. The starting data point in both injection methods was obtained from the same experiment. The inset shows numerical simulation results obtained by Lenormand et al.1

Characteristics of three different displacement patterns are observed in the discontinuous-rate injection experiments (Figure 1). At low injection rates (logCa < −6.61), the capillary forces associated with water dominate the displacement process: scCO2 entered the pore network as a relatively uniform front, immediately followed by randomly distributed forward and lateral flowpaths with clusters of entrapped water. Such a displacement pattern indicates capillary fingering. Approximately halfway through the micromodel, the scCO2 215

dx.doi.org/10.1021/es3014503 | Environ. Sci. Technol. 2013, 47, 212−218

Environmental Science & Technology

Article

relative permeability hysteresis and independent evaluation of pore geometry factors.27,28 In the continuous-rate experiments, capillary fingering dominated all injection rates (Figure 2). At the lowest rate (i.e., logCa = −7.61), the scCO2 entered the micromodel pore network evenly and transitioned into one finger halfway through the micromodel. As the injection rate was increased, the front of the scCO2 moved further through the pore network until it broke through all the way to the end at the highest injection rate (i.e., logCa = −4.73). The SscCO2 increased monotonically with injection rate (Figure 3): the increase is small between logCa = −7.61 (SscCO2 = 0.4) and −5.61 (SscCO2 = 0.5), while a larger increase occurred between logCa = −5.21 (SscCO2 = 0.55) and −4.91 (SscCO2 = 0.75) when the front of the main capillary finger progressed toward the end of the pore network (Figure 2). In core flooding experiments under similar experimental conditions (10 MPa, 40 °C), Shi et al. showed increasing injection rate resulted in an increase in CO2 saturation.12 Continuous vs Discontinuous Injection. Results shown in Figures 1−3 showed the same injection rate used in two different injection methods can result in different displacement mechanisms and saturations, and that an increase in injection rate does not always result in an increase in scCO2 saturation. To further compare these two injection methods, local scCO2 saturations along the length of the micromodel pore network were determined by averaging saturation over each transverse cross-section of the pore network and subsequently over 1.5x grain diameters along the length of the micromodel to smooth the profile.29 In the discontinuous-rate experiments at low injection rate (logCa = −6.91), the local scCO2 saturation was high (∼0.8) in the first half of the pore network and gradually decreased to