Experimental study of two-phase flow properties of CO₂ containing N₂ in porous media

Abstract

In preliminary analyses, the co-injection of CO₂ with H₂S, SO₂ and N₂ impurities has been shown to reduce total carbon capture and storage (CCS) cost. The multiphase flow properties of impurities in the CO₂–brine system in porous media are the key to understanding the mechanisms and nature of geological CO₂ sequestration projects. In this study experiments were performed on the multiphase flow process of CO₂/N₂/brine system at conditions similar to aquifer pressure and temperature using the X-ray CT technique. Experiments at various rates of CO₂ injection that affect saturation and spatial distribution of injected gas were conducted in this experiment. The results indicate an strong relationship between gas saturation and porosity distribution in porous media, and the increasing capillary number leads to lower saturation in downward injection. Small capillary numbers and higher fractional flows in the gas phase both result in uniform saturation maps in the core. The CO₂ clusters seem larger at high capillary numbers and the high CO₂ collection regions extend based on the saturation distribution in the lower CO₂ fraction as the flow pattern stays similar as the same capillary number. CO₂ containing N₂ tends to retain more correlative relationships at different gas injection rates compared with the pure CO₂ stream. Even though both distribution and saturation are storage concerns, the N₂ component has little effect on gas distribution, whereas it brings about an overall increase in the saturation for most experiments. Thus the N₂ enhanced the storage performance of CO₂.

---

1. Introduction

Carbon capture and storage (CCS) is considered a potentially effective, economic and fossil energy-compatible option for reducing CO₂ emissions to mitigate the greenhouse effect. A number of studies have indicated that deep saline aquifers have a large and permanent capacity for CO₂ geological storage.^1–3 During and after CO₂ capture processes in energy-intensive factories, gaseous impurities (e.g., N₂, O₂, NO_X, SO_X, H₂S, and H₂O) usually exist in the CO₂ stream. To reduce impurities in the CO₂ stream, several major approaches are recommended for adoption in industrial CO₂ capture, such as post-combustion (capture of CO₂ from flue gas), pre-combustion (capture of CO₂ before combustion) or IGCC (integrated gasification combined cycle), and oxyfuel combustion. Although commercial operation of the CO₂ capture technology has already been achieved, after these purification processes the CO₂ stream usually still contains the abovementioned impurities with lower CO₂ concentration—for example, the CO₂ concentration in oxyfuel flue gases varies from 80 to 90 vol% (dry).⁴ Impurities also exist during the transit of the CO₂ stream. Most of the CO₂ pipelines in the United States transport CO₂ with impurities such as nitrogen, hydrocarbons, water and hydrogen sulphide,⁵ and the increasing impurities increase risks lead to higher risks in pipeline transportation. However, the cost of capturing CO₂ out of flue gas is significantly higher than that of transporting and injecting CO₂ in geological reservoirs.⁶ Thus preliminary analyses indicate that the co-injection of CO₂ with impurities show that total CCS costs may be reduced by lowering the capture requirement of CO₂, especially from high-purity emission sources.^7,8

According to several previous studies,^9–11 impurities like H₂S and SO₂ will affect the geo-chemistry, trapping mechanism, and migration of the CO₂ plume. However, few studies have focused on the effect of non-condensable gas such as N₂, which influences properties of the gas mixture such as density, viscosity, and phase behaviour, among others. Recent studies show similarities in experimental results between N₂ and CO₂. Kneafsey et al. (2013) reported that silica sand nitrogen displaced more brine than CO₂, but CO₂ formed a thicker and more saturated layer.¹² In calcite sand the invasion patterns were similar for the two invading phases with a higher CO₂ saturation. Experiments regarding the effects of N₂ on CO₂ are necessary to obtain a better understanding of the N₂ effect in the gas mixture. If impurities including but not limited to N₂ have beneficent effects on CO₂ storage, implementation of CCS projects would be more easily implemented.

After CO₂ is injected into reservoirs, the CO₂ plume spreads along the impermeable structures and only a small fraction of the reservoir would be used for CO₂ physical storage. If the CO₂ distributes uniformly with relatively low saturation in the reservoir, snap-off might occur at the early stage of buoyancy-driven migration after gas injection. The snap-off causes CO₂ to behave in a discontinuous phase, allowing it to be trapped by capillarity in porous rock. Therefore, CO₂ would be trapped at the injected location without migration to a reservoir scale. The co-injection scheme also allows use of open, dipping aquifer structures for CO₂ storage^13–15 and provides huge storage potential in closed dome structures. In order to improve our understanding of multi-phase flow and trapping mechanisms in CO₂–brine systems, more experimental data, numerical simulations and theoretical studies are needed. Although the literature describing multi-phase flow of oil and water, and CO₂ and oil is extensive,^16–21 very few laboratory experiments have been performed on CO₂–brine systems with N₂.

In this paper, we present gas–liquid, two-phase flow experiments using X-ray computed tomography technology at reservoir pressure and temperature. The gas distribution and saturation calculated from X-ray images during co-injection processes were determined at a fine resolution corresponding to the volume of pores. The study investigates and analyses the effect of N₂ on the CO₂ trapping mechanism and to understand the effect of total injection capillary numbers on residual gas saturation distributions.

2. Experimental methods

2.1. Experimental setup and materials

The experimental setup is shown in Fig. 1. An Inspexio SMX-225 CT system (Shimadzu, Japan) was used to visualize internal pore structure and gas saturation. The syringe pumps (A1 and B1, 260D Syringe Pump, Teledyne ISCO, USA) injected brine and gas in the porous media, respectively, and another syringe pump (C1) controlled the back pressure of the system. The glass beads (particle diameter of 0.15–0.25 mm, As-One Co., Ltd., Japan) were packed into the holder as the porous media. The holder placed in the turntable center is made of aluminium alloy (inner diameter 15 mm, length 150 mm, thickness of cylinder wall 3 mm). To minimize end effect, the field of view (FOV) is chosen at the upper center of the holder. Temperatures of the holder were controlled by electric heating tape. In addition, the temperatures of syringe pumps were controlled by the heating circulator (JULABO, F-250) with an accuracy of 0.1 °C. The impure gas used in the experiment is made up by 79.98% CO₂ and 20.02% N₂ (Dalian Special Gases Co., Ltd, China). Table 1 displays the properties of fluids at the experimental condition. The impure gas was injected under 40 °C and 8 MPa during the experiment as supercritical status. Potassium iodide (KI, mass fraction of sodium chloride 3%) was used within the brine phase to enhance contrast between brine and gas in the CT images.

Fig. 1 Experimental apparatus.

Table 1 Properties of CO₂, impure gas and brine in experiments

Condition	Properties	sc CO2	Gas mixture	Brine
40 °C, 8 MPa	Density	277.9 kg m^−3	239.4 kg m^−3	1172.8 kg m^−3
	Viscosity	0.0224 cp	0.0219 cp	0.722 cp

---

2.2. Experimental procedure

After the pipeline was flushed with N₂, glass beads were filled in the holder, and then the holder was placed on the X-ray CT turntable. After vacuuming the porous media for 30 min, the brine filled into the porous media until 100% brine saturation was achieved at 8 MPa. Fluid was injected downward with different capillary numbers (see Table 2) at different gas fractions of 0.2, 0.4, 0.7 and 1. In the following discussion, the fg = 1, which represents the gas drainage brine process, is defined as single-phase injection, while the other fractional gas injection experiments where the injection flows consist of gas and brine streams, are defined as co-injection. Two sets of experiments were conducted to make comparisons between CO₂ and impure gas. For each injection experiment the gases were injected continuously for about 6 PV (1 PV is about 9 ml) to achieve steady state. CT scanning was then conducted and then the core was depressurized and vacuumed again for the next circulate experiment. Next, the pipeline was flushed with N₂ and refilled with brine again to perform the subsequent set of experiments. All experiments were restricted to a single sample, which means once we filled glass beads into the core holder, the experiments were replicated on it without sand filling again. Since each sand filling will generate a completely different pore distribution, the porosities and pore throat forms may induce results that cannot be compared. To reduce errors resulting from heterogeneity, we used the same sand core in all experiments. The parameters of the X-ray CT are shown in Table 3. We obtained 452 slice images with a spatial resolution of 512 × 512 pixels in each scan, and the thickness of every slice is 0.033 mm. In each experiment, scanning was performed at 150 kV and 40 μA. The spatial resolution was 55 μm per voxel.

Table 2 Capillary numbers of CO₂ and gas mixture in single-phase injection^a

Carbon dioxide	Impure gas
Ca	Ca
a Note: In co-injection the total injection Ca of gas–brine stream remains the same.
5.81429 × 10^−7	5.80088 × 10^−7
2.90714 × 10^−7	2.90044 × 10^−7
1.9381 × 10^−7	1.93363 × 10^−7
1.16286 × 10^−7	1.16018 × 10^−7
5.81429 × 10^−8	5.80088 × 10^−8

---

Table 3 Parameters for X-ray CT scanning in drainage experiment

Parameter	Value
Tube current	40 μA
Tube voltage	150 kV
Spatial resolution	55 μm per voxel

---

2.3. Experimental parameters

To obtain a porosity map of the packed bed, two sets of images were needed: the dry images, where the packed bed is dry and the pore space is filled with air; and the brine saturated images, where the packed bed is fully saturated with brine at reservoir conditions. The porosity Φ of each voxel element is then calculated as follows:²²where CT is a numerical value obtained from the image acquired by the CT scanner. To obtain the saturation map at different CO₂ fractional flows, three sets of images are needed: brine saturated images, CO₂ saturated images and experimental images. The CO₂ and brine saturations (S_CO2 and S_brine) were then calculated for each voxel as follows:²²

3. Results and discussion

Calculations of porosity along the length of the core and the three-dimensional image are shown in Fig. 2. The average FOV porosity was 0.35 according to eqn (1), and the porosity distributions are uniform along the length of the core holder with a deviation of less than 1%. The porous media was relatively uniform because of the uniform porosity distribution. However, the bead filling method raised the porosity at the upper end face. Some large pore throats exist in the middle and upper end of the area in the three-dimensional map. The length of the Z axis along the FOV was 22.3 mm, occupying 15% of the height of the core holder. This imaging region prevents the capillary end effect and impact of buoyancy on the saturation at the core bottom, which could significantly disturb the saturation profile in a downward injection. To obtain 3-D saturation spatial distribution in representative experiments, the CT images stack with 452 slices was imported into commercial software (VG Studio Max 2.1) in each set according to the downward flow sequence.

Fig. 2 Slice-averaged porosity and three-dimensional map of core.

3.1. Gas distribution

CO₂ injection. The steady-state, three-dimensional views of CO₂ distribution in the core at different CO₂ fractional flows are shown in Fig. 3. After the stable state was achieved in the CO₂ drainage brine process, gas distribution in the porous media appeared to be continuous slugs. After the co-injection of CO₂ and brine was completed, most of the gas clusters were also continuous; however, gas blobs also existed. Layer stacks of glass beads tended to be heterogeneities and pore throats that inclined in the flow direction seemed to be more crooked. As a result, CO₂ was easily trapped in larger pore clusters along the flow direction. This feature significantly affected CO₂ distribution during drainage and demonstrated the importance of small-scale capillary heterogeneity in non-wetting phase distribution during drainage. High-porosity regions have significantly higher CO₂ distribution than regions with lower porosity.

Fig. 3 Comparisons of different total capillary numbers and CO₂ fraction.

As seen in Fig. 3, CO₂ distribution tends to be more uniform at low Ca. Overall residual saturation increased and especially in the low saturation regions, suggesting that CO₂ entered into pores gradually and uniformly along the flow direction and trapping capacity increased. Injecting gas at low Ca reduced instability of displacement during the downward drainage process, which agrees well with previous studies.^16,23 With the Ca increasing, CO₂ occupied large pores in the middle of the core vertically and preferentially. Thus the CO₂ clusters were larger than those in the low Ca experiments. Also, in the core center CO₂ distribution was generally greater at higher Ca.

With higher fraction injection, CO₂ gradually permeated into more pore spaces at the edge of the core, leading to a higher saturation distribution. This feature was enhanced at higher Ca, indicating that the displacement processes were more stable. Throughout all fraction experiments, high CO₂ collection regions extended from low to high gas fraction since the flow channel patterns remained similar.

Impure gas injection. As seen in Fig. 4, gas distribution in the FOV seemed to be almost the same as in the CO₂ injection experiment, and the irreducible brine tended to be more concentrated. In the distribution map, the red and yellow zones represent high saturation regions, whereas the green and blue zones represent low saturation ones. Generally speaking, the number of red and green zones increased compared with those in CO₂ experiments, indicating that the surrounding regional saturation may increase with highly compact, non-wetting phase ganglia. Compared to CO₂ injection, more gas was trapped at the core edge. Considering the fractal morphology of gas saturation distribution, multiple loosely connected or disconnected fingers mainly progressed forward toward the downstream boundary with limited or no lateral flows (referred to as “viscous fingers”).^24,25 Also the gas was trapped more separately in the center of the porous media, resulting in larger gas plumes in the core. Because the CO₂ containing N₂ is less viscous and instability at the displacement front was lower, more heterogeneous propagation led to greater reinforcement of buoyancy between cross-sectional slices. Thus the N₂ component has little effect on CO₂ distribution.

Fig. 4 Comparisons of different total capillary numbers and impurities gas fraction.

Regarding fractional flow, the impure gas distribution became more uniform as fraction increased, and this phenomenon was more distinguishable at high Ca. In the case of low fraction, the CT images imply that large gas slugs were trapped in the center of the porous media, while around the porous media, there was little CO₂ because of the low-porosity artifact pore structure. Such phenomena were similar to CO₂ injection.

In the CO₂ set at low fraction, the number of discrete and small bubbles increased along the flow injection direction, and the low saturation area shrank as capillary number increased. However, for the impure gas this tendency was not as evident. Viscous fingering is characteristic of the flow channel.^24–26 As the viscosity in downward injection could be restrained more obviously by buoyancy at lower Ca, branches of the flow channel combine and furcate to generate various sizes of discontinues gas clusters resulting in cellular saturation distribution. The pattern also implies that horizontal gas-spreading channels may exist. Compared with CO₂, the lower viscosity of the impure gas will improve front displacement to distribute over the entire core cross-section, disturbing the trapping saturation trends slightly with the co-injection of large-volume brine. At low fraction, the impure gas saturation changed a lot with Ca, while its saturation became similar as the fraction gradually increased.

3.2. Gas saturation in the case of CO₂ injection and impure gas injection

Ca and fraction effect on CO₂ saturation. Regarding the Ca effect, in porous media the two-phase flows are influenced by interfacial tension, viscous sheer stress and gravity. The capillary number (Ca) represents the ratio of viscous sheer stress to the capillary force and is defined as:where σ is interfacial tension, μ is the viscosity, v is the displacing fluid velocity and subscript w denotes the wetting phase. Interfacial tension and viscosity are constant in the CO₂–brine and impure gas–brine system. Thus the capillary number (Ca) is proportional to the injection fluid velocity.

Based on eqn (2) and Fig. 3, CO₂ residual saturations are obtained. The sensitivity of Ca and gas fraction on the residual saturation were investigated. Fig. 5 shows the saturation profile with Ca and fraction of CO₂. In general, the CO₂ saturations increased as Ca decreased in downward injection. The smooth CO₂ saturation profile along the flow direction decreased with the Ca. At high Ca, the curves follow the porosity curves and the capillary barriers could be observed by the saltation of curves. However, as Ca decreased the curves become smoother and such regulation fell as flow developed. But when fraction was reduced to 0.2, the effects of capillary number became complex. The residual gas saturation increased with Ca and peaked at 1.9 × 10⁻⁷. In these cases, CO₂ is dramatically affected by the co-injection brine at the center of the core (Fig. 3).

Fig. 5 Comparisons of slice-averaged CO₂ saturation between different capillary numbers at fraction flow of CO₂: (A) f_CO2 = 0.2, (B) f_CO2 = 0.4, (C) f_CO2 = 0.7, (D) f_CO2 = 1.

Under such low fraction and Ca, the buoyancy force caused by both injection brine and in situ brine dominates the flow; therefore a rapid descent occurs in the upper part of the CO₂ saturation profile along the flow direction. At the end of the porous media, CO₂ saturation decreases sharply because of the buoyancy force. However, with Ca increasing, the effect of buoyancy declines. After the peak at 1.9 × 10⁻⁷, the separations of low CO₂ reservations tend to be more and more uniform, indicating elimination of the impact of buoyancy on the saturation decline along the core. Thus in the case of Ca 5.814 × 10⁻⁷ and f_CO2 = 0.2, CO₂ saturation gets stable and even increases. With low fraction and Ca, dissolution occurs at the beginning of the displacement process. This occurs because of the variable amount of dissolved CO₂ in the phase flow injected into the core.^27–29 Typically lower Ca leads to more CO₂ to be solved in two-phase flow flooding, especially at the displacement front, and the gravity effect appears more significant. Since the CO₂ fraction is 0.2, experiments with Ca below 5.814 × 10⁻⁸ have a substantially low saturation and steep curve. Also the higher Ca results in smoother curves because dissolved quantities decrease.

Considering the fractional effect, along the flow direction the residual saturation increases with the CO₂ fraction independently of pore porosity profile. For most of the Ca, the CO₂ saturation decreases along with gas fraction reduction, since the quantity of gas filled into the porous media was controlled by the fraction and injected flow rates. At a given Ca, the saturation profiles are similar and are mainly induced by the pore structure.

The gravity effect is not obvious for high fractional flow. Large fractions of CO₂ stabilize the displacement and smooth the curve, improving the trapped amount in the low porosity region. For the situation of f_CO2 = 0.4 and f_CO2 = 0.7, the saturation maps have linear decline with increasing capillary numbers. The saturation curves are more regular and accordant with the two experiments that show similar curve paths, and this will be enhanced by the increasing CO₂ fraction.

At high total Ca, the CO₂ saturation profiles for co-injection get unsmooth comparing with single phase injection (Fig. 5), but between different fractions the differences tend to be relative low and have a large decline when single phase injection. At high Ca the co-injection of brine dramatically removes the trapping CO₂ comparing with the single injection of it, especially in low porosity region (Fig. 3). With Ca decreasing, the saturation increases and the D-values between fractions becomes smaller and smaller. As the drainage happens rapidly the viscous fingering appears faster as the pressure drop increase in more viscous wetting fluid, and the later injection will prefer to flow through these channel. As a result, more gas occupies the larger pores and less gas enters in the pores next to the fingering along with the decrease in CO₂ fraction. The impact of co-injected brine is to displace some unattached bubbles and drive them into lager pores as the gas is readily to converge and be trapped by pore throats nearby with relative high capillary entry pressure. So the effect of brine at high Ca could not be neglected because of the loss of drainage efficiency. At low capillary numbers through the co-injection, some gas bubbles are discontinuous and likely to be trapped by capillary force. Under such trapping mechanism gas saturations in co-injection and single-phase injection are nearly the same at low Ca.

Even at the total Ca of 5.814 × 10⁻⁸, the lowest Ca experiment, the result in co-injection experiment won't along with the linear increase of the fraction whose curves cross together and almost exceed the single-injection result. However, for CO₂ fraction of 0.2 the trends aren't regular with the linear change of total capillary number. The low CO₂ component makes the pressure drop insufficient to overcome the capillary pressure and is displaced by the following injection brine, leading to a random CO₂ saturation distribution.

Ca and fraction effect on impure gas saturation. In Fig. 4, the 3D impure gas distributions are visualized clearly. Fig. 6 shows the saturation profile with Ca and fractionals of impure gas.

Fig. 6 Comparisons of slice-averaged impurities gas saturation between different capillary numbers at fraction flow of: (A) fg = 0.2, (B) fg = 0.4, (C) fg = 0.7, (D) fg = 1.

Apart from the 0.2 fraction, the saturation trends are similar under similar total capillary numbers for both CO₂ and CO₂ impure gas. Saturations still increase as the total capillary numbers decrease in most cases. At the same fraction in co-injection, the resulting contrasts are mostly obvious and lead to various effects. Because of N₂, the viscous of impure CO₂ decrease a little (Table 1). The saturation distribution and profile appear to be different from that of CO₂. Especially at low fraction and Ca, the gravity effects almost disappear since the curves are no longer decrease at the later part of the FOV. It could be concluded that the stability of downward drainage and the saturation result are largely improved so that the small fraction of co-injection are predictable.

In the case of fg = 0.2, though the saturation is still low at Ca of 5.801 × 10⁻⁷, it increases a lot at lower Ca. The peak is achieved at 5.801 × 10⁻⁸, whereas in the CO₂ experiment the value is extremely low. Also the saturation values in the 1.934 × 10⁻⁷ and 2.900 × 10⁻⁷ sets increase so that the tendencies between capillary number and saturation are more regular. Also the curves among these Ca are much smoother than those of CO₂–brine injection. Another phenomenon is that the saturation profiles are consistent with porosity profiles at high Ca, which is different from the CO₂ set.

Next, at a given Ca the effect of nitrogen is different according to low or high total capillary number. For a fraction of 0.4 and 0.7, compared with the CO₂ set, the saturation at low and high Ca divided into higher and lower, respectively.

At high Ca, the D values between saturations in co-injection are far smaller than those compare with single injection. The fractions of 0.2 and 0.4 have almost identical saturation cause the amount of brine is relative large in co-injection. The D-values of saturation between the two sets are smaller than those in CO₂ experiment. Even in the 0.7 set, a relative high fraction, the D-values between saturation profiles are far less than CO₂ set. This implies the brine has a great inhibition on the trapping amount of impure gas at high flow rates. As the Ca decrease to 1.934 × 10⁻⁷, the D-values between different fractions are more hierarchical whereas the saturation is still lower than those in former CO₂ experiments. Also the displacements at high Ca with less viscous fluid easily produce viscous fingers. Hence, with the brine inject into the pore throat, the trapped gas which appears to be more slippery is more readily displaced by the following brine in the viscosity fingering channel. Then the less viscous gas is brought downward before they penetrate into the downstream crooked pore throats which are against the flow direction.

Besides, at low Ca almost all of the saturation values are higher than CO₂ set and the D-value between fractions decrease again. At low Ca the co-injection will increase the buoyancy effect, and more stable displacement front occurs with higher drainage efficiency. This indicates the effects of fractions are slight when the high trapped gas amounts have already achieved the utmost of the pores capacities. The appearance that different saturation curves in fractional flow will cross at some parts of the core also proves this. In CO₂ experiment, for large fraction of CO₂ in the two phase flow, such phenomenon will only occur at the total capillary number of 1.938 × 10⁻⁷. However, in impure gas set when the capillary numbers are below 1.160 × 10⁻⁷, the saturations between various fractions will be analogous. This indicates that for impure gas a relatively slow injection rate will be satisfactory with residual storage of the gas no matter the displace front and plume are gas or gas–brine in reservoir.

The saturation peak is achieved at capillary number 1.160 × 10⁻⁷ when fg = 1, and then the saturation decreases slightly. Since the impure gas isn't as viscous as pure CO₂, the bubbles are easy to be diverted and can spread further into the low capillary entry pressure region. If the separation of injected gas appears at the inlet, the edge part of the core will enter more gas. So the distributions among the cross section are more uniform, resulting in higher gas saturation. Above all, the relationship between total Ca and saturation is single: the former decrease, the later increase.

Considering the fraction, the effect of brine between the porosity and saturation profiles seems disappeared. Also the D-values between saturations are almost vanished between low total capillary numbers. However, as in single phase injection the saturation are higher than those in CO₂ experiments, in co-injection the brine impel the saturation to be lower, so the inhibitions of brine on impure gas are intense. At the edge of the core as the low porosity region the saturations have significant reductions, implying the less viscous gas is more easily displaced through pore throats by the following brine. The highest saturation among the whole sets of experiments is achieved at the capillary number of 5.801 × 10⁻⁸. This indicates lower injection rate and moderate N₂ amount in CO₂ stream will improve the storage efficiency in CCS project.

3.3. Stability analysis

For gravity drainage processes which correspond to vertically downward injections of gas into the packed glass beads in our experiments, the critical gravity drainage velocity is defined by Blackwell and Terry (1959) and Dumore (1964) as follows:where ρ is the density, g is gravity acceleration, k is the absolute permeability, and subscripts w and n denote the wetting phase and non-wetting phase, respectively. According to the Carman–Kozney expression, the permeability of packed glass beads with a diameter of 200 μm is 24.7 Darcy.¹⁶ The ratio of the critical gravity drainage velocity to the displacement fluid velocity vc/v is referred as the stability parameter. Typically if vc/v is less than 1, the downward drainage become unstable and the saturation result may have different values that deviate from regulation of the Ca-saturation tendency. It must be noted that in previous research, the vc/v can adequately predict displacement stability in downward single-phase injection, whereas its application to the co-injection displacement phenomenon has not yet been confirmed.

In CO₂ experiments, at Ca 5.814 × 10⁻⁷ the decreasing tendency of saturation distribution between diverse fractions of CO₂ disappear apart from the other two experiments with relative low Ca. Considering that the vc/v is 0.835 and 0.477 with CO₂ fraction of 0.4 and 0.7, respectively, it reflects the buoyancy effect with a relatively high gas fraction and implies unstable displacement with buoyancy against the viscous force. This phenomenon also exists in impure gas experiments as interpreted above.

In the case of fg = 1 (Fig. 5d), the highest saturation is achieved at Ca 1.163 × 10⁻⁷, and as Ca increases the saturations decrease along with fluctuating profiles. The stability parameter decreased from 1.67 to 0.334, giving an appropriate criterion for stability of a drainage interface. Thus displacement interface stability increases as the stability parameter increases. A turning point at Ca 1.163 × 10⁻⁷ occurs, since saturation at capillary number 5.814 × 10⁻⁸ is still lower than for the 1.163 × 10⁻⁷ set. Considering Ca 5.814 × 10⁻⁸ and 1.163 × 10⁻⁷ at fg = 0.4, saturation of the former is much higher than the latter. But with the fraction increasing, the saturation difference gradually declines and is finally opposite. As seen in Fig. 3, Ca 1.163 × 10⁻⁷ and 5.814 × 10⁻⁸ experiments correspond to vc/v 1.67 and 3.34, respectively; at the edge of the core, the former contains more yellow and less blue, which correspond to high and low gas saturation regions, respectively. Similar results were achieved in previous research.²³ One of the possible reasons could be that the vc/v is not precise enough to predict a residual saturation trend in downward drainage between small Ca intervals, and stability will be disturbed by small-scale heterogeneity in pores media. In Fig. 5d at the beginning of a few slices, the saturation values at Ca 5.814 × 10⁻⁸ gradually decrease whereas saturation values at Ca 1.163 × 10⁻⁷ remain in the trend, leading to the value difference. The relatively low porosity at these regions results in higher capillary entry pressure, especially for the pore throats at the edge of the core. So for Ca 1.163 × 10⁻⁷ viscous fingerings exist at these regions, and continuously permeate downward since the pressure drop is high enough against capillary entry pressure. However, at Ca 5.814 × 10⁻⁸, the numbers in the blue zone, which represents low gas saturation, are relative greater than those in the Ca 1.163 × 10⁻⁷ experiment, preventing viscous fingering from developing further as the effective pressure drop could not be established.

In the case of impure gas experiments, the opposite tendency occurs between Ca 5.8 × 10⁻⁸ and Ca 1.16 × 10⁻⁷, comparing different fractions. A similar pattern could be predicted as seen in the blue zone at the edge of the core. For co-injection, at Ca 5.8 × 10⁻⁸ the vc/v is 7.26 and 4.14 corresponding to fg = 0.4 and fg = 0.7, respectively. The curves shown in Fig. 6 are relatively flat despite the high volume of brine injected together with gas. Also, the curves for Ca 1.16 × 10⁻⁷ with vc/v from 3.63 to 2.07 are both unsmooth, which indicate the Ca effect on downward drainage stability is larger than that of vc/v. The drainage process is controlled by viscous force that can barely overcome the capillary barrier in the low-porosity region when permeating vertically along the flow direction. As the viscous effect decreases, the unstable drainage occurs. It could be predicted that the brine interferes with principal stability parameter for CO₂ containing N₂, as the injection of brine disturbs detainment of less viscous gas in viscous fingering channel and induces the loss of slip gas at a certain level. But for single-phase injection, comparing the saturation of Ca 5.8 × 10⁻⁸ and Ca 1.16 × 10⁻⁷, the former is higher. And the vc/v is 2.9 and 1.45, respectively. This means that the N₂ component enhances indication of the saturation trend in downward drainage between small Ca intervals since the less viscous gas is more likely to form a stable displacement front via buoyancy.

4. Conclusion

Two sets of experiments are implemented to explore the difference in storage capacity between the CO₂ and the CO₂ containing N₂ by single injection and co-injection in the porous media. The three-dimensional image stacks are obtained by high resolution X-ray CT. Also the co-injection fractions of gas phase as well as the total injection capillary number are altered to investigate their influence on gas saturation in the steady displacement procedure.

The three-dimensional images of the dry core showed moderate heterogeneity, as the slice-averaged porosity was in the 35% range. In the two experiments the saturation results and distributions in the core were similar except for the 0.2 fractional set. As impure gas is less viscous, the displacement fronts were more readily reformed by buoyancy and achieved equilibrium status. Thus N₂ seems to assist CO₂ to express further and permeate into the crooked pores that slant to the flow direction. The distributions then are more uniform and have higher displacement stability, which induces higher gas saturation. Since viscous fingerings dominated under our Ca condition, establishing the preferential route for the gas resulting in less viscosity to enhance drainage performance proved more difficult. The total injection rates decrease, resulting in the slice-averaged saturations increasing for both sets of experiments. And for CO₂ containing N₂ the saturations will be higher at low flow rates and high fractions of gas in co-injection.

It should be noted that our results show the brine disturbing detainment of less viscous impure gas in pore throats so that between small intervals of Ca the saturation values in downward co-injection experiment of CO₂ containing N₂ are slightly irregular vis-a-vis the common rule of Ca-saturation. Meanwhile at low total injection flow rates, the saturations achieved substantially high values for pore gas-capture capacity, so the fraction study results are comparable with each other for saturation estimation. However the N₂ sets appear to have higher saturation results and the highest slice-averaged saturation among these experiments had the capillary number of 5.801 × 10⁻⁸ in single-injection impure gas. This indicates that the N₂ in gas mixture improved CO₂ storage and a low injection rate should be adopted when injecting the gas into reservoirs where the underground sealed condition will be beneficial regardless of whether the displacement front is single gas plume or gas–brine mixture.

Acknowledgements

This study has been supported by the National Natural Science Foundation of China (grants 51506024, 51436003), and the National Basic Research Program of China (973, grant 2011CB707300). It has been also supported by Fundamental Research Funds for the Central Universities (DUT13LAB01).

Notes and references

1. A. Manancourt and J. Gale, A review of capacity estimates for the geological storage of carbon dioxide Conference:Proceedings of 7th International Conference on Greenhouse Gas Control Technologies, ed. E. S Rubin, D. W Keith and C. F Gilboy, 2004.
2. S. Bachu, D. Bonijoly, J. Bradshaw, R. Burruss, S. Holloway, N. P. Christensen and O. M. Mathiassen, Int. J. Greenhouse Gas Control, 2007, 1, 430–443.
3. E. S. Rubin and H. de Coninck, IPCC special report on carbon dioxide capture and storage, RITE International workshop on CO₂ geological storage, 2006.
4. B. Paschke and A. Kather, Energy Procedia, 2012, 23, 207–215.
5. P. N. Seevam, J. M. Race, M. J. Downie and P. Hopkins, Transporting the next generation of CO₂ for carbon, capture and storage: the impact of impurities on supercritical CO₂ pipelines, 7th International Pipeline Conference, 2008.
6. M. C. N. D. Mago, Master of Science, Texas A&M University, 2005.
7. I. Gaus, P. Audigane, L. Andre, J. Lions, N. Jacquemet, P. Durst, I. Czernichowski-Lauriol and M. Azaroual, Int. J. Greenhouse Gas Control, 2008, 2, 605–625.
8. X. Li, N. Wei, Z. Fang, Q. Li, R. Dahowski and C. Davidson, Energy Procedia, 2011, 4, 6029–6036.
9. S. Bachu and D. B. Bennion, Int. J. Greenhouse Gas Control, 2009, 3, 458–467.
10. T. Xu, J. A. Apps, K. Pruess and H. Yamamoto, Chem. Geol., 2007, 242, 319–346.
11. D. B. Bennion and S. Bachu, SPE Reservoir Eval. Eng., 2008, 11, 487.
12. T. J. Kneafsey, D. Silin and J. B. Ajo-Franklin, Int. J. Greenhouse Gas Control, 2013, 14, 141–150.
13. I. Akervoll, P. Zweigel and E. Lindeberg, CO₂ storage in open, dipping aquifers, Proceedings of GHGT-8 8th international conference on greenhouse gas control technologies, Trondheim, Norway, 2006.
14. S. Georgescu, E. Lindeberg and T. Holt, Co-injection of CO₂ and water into non-sealed saline aquifers, Proceedings of the 8th International Conference on Greenhouse Gas Control Technologies, Elsevier, CD-ROM, 2006.
15. K. Michael, P. Neal, G. Allinson, J. Ennis-King, W. Hou, L. Paterson, S. Sharma and T. Aiken, Energy Procedia, 2011, 4, 4267–4274.
16. T. Suekane and K. Okada, Energy Procedia, 2013, 37, 5545–5552.
17. T. S. a. H. Ushita, in Computed Tomography – Special Applications, ed. E. b. D. L. Saba, InTech, 2011, ch. 10.
18. Z. Na, T. Suekane, T. Hosokawa, S. Inaoka and Q. Wang, Transp. Porous Media, 2011, 90, 575–587.
19. T. Suekane, N. Zhou, T. Hosokawa and T. Matsumoto, Transp. Porous Media, 2009, 82, 111–122.
20. J.-C. Perrin and S. Benson, Transp. Porous Media, 2009, 82, 93–109.
21. L. Zuo and S. M. Benson, Geophys. Res. Lett., 2014, 41, 2014GL059653.
22. S. Akin and A. Kovscek, Spec. Publ. – Geol. Soc. London, 2003, 215, 23–38.
23. Y. Mikami, Y. Deguchi and T. Suekane, J. Flow Control, Meas. Visualization, 2014, 2014.
24. Y. Wang, C. Zhang, N. Wei, M. Oostrom, T. W. Wietsma, X. Li and A. Bonneville, Environ. Sci. Technol., 2012, 47, 212–218.
25. H. Liu, A. J. Valocchi, C. Werth, Q. Kang and M. Oostrom, Adv. Water Resour., 2014, 73, 144–158.
26. H. Liu, A. Valocchi, Q. Kang and C. Werth, Transp. Porous Media, 2013, 99, 555–580.
27. C. Chang, Q. Zhou, L. Xia, X. Li and Q. Yu, Int. J. Greenhouse Gas Control, 2013, 14, 1–14.
28. S. Pistone, R. Stacey and R. Horne, The significance of CO₂ solubility in geothermal reservoirs, Proceedings Thirty-Sixth Workshop on Geothermal Reservoir Engineering, 2011.
29. S. Berg, S. Oedai and H. Ott, Int. J. Greenhouse Gas Control, 2013, 12, 478–492.

---