Core-scale sensitivity study of CO2 foam injection strategies for mobility control, enhanced oil recovery, and CO2 storage

This paper presents experimental and numerical sensitivity studies to assist injection strategy design for an ongoing CO2 foam field pilot. The aim is to increase the success of in-situ CO2 foam generation and propagation into the reservoir for CO2 mobility control, enhanced oil recovery (EOR) and CO2 storage. Un-steady state in-situ CO2 foam behavior, representative of the near wellbore region, and steady-state foam behavior was evaluated. Multi-cycle surfactant-alternating gas (SAG) provided the highest apparent viscosity foam of 120.2 cP, compared to co-injection (56.0 cP) and single-cycle SAG (18.2 cP) in 100% brine saturated porous media. CO2 foam EOR corefloods at first-contact miscible (FCM) conditions showed that multi-cycle SAG generated the highest apparent foam viscosity in the presence of refined oil (n-Decane). Multi-cycle SAG demonstrated high viscous displacement forces critical in field implementation where gravity effects and reservoir heterogeneities dominate. At multiple-contact miscible (MCM) conditions, no foam was generated with either injection strategy as a result of wettability alteration and foam destabilization in presence of crude oil. In both FCM and MCM corefloods, incremental oil recoveries were on average 30.6% OOIP regardless of injection strategy for CO2 foam and base cases (i.e. no surfactant). CO2 diffusion and miscibility dominated oil recovery at the core-scale resulting in high microscopic CO2 displacement. CO2 storage potential was 9.0% greater for multi-cycle SAGs compared to co-injections at MCM. A validated core-scale simulation model was used for a sensitivity analysis of grid resolution and foam quality. The model was robust in representing the observed foam behavior and will be extended to use in field scale simulations.


Introduction
A major challenge in carbon dioxide (CO2) enhanced oil recovery (EOR) is poor macroscopic CO2 sweep efficiency caused by the low viscosity and density of injected CO2 [1,2]. These adverse properties can result in viscous fingering and gravity override, greatly hindering oil recovery and sweep efficiency [3,4]. Reservoir heterogeneity can also cause injected CO2 to channel through high permeability zones (layers or fractures) resulting in early CO2 breakthrough, high CO2 recycling, and low incremental oil recoveries. CO2 foam injection is a laboratory verified and field-validated technique proven to mitigate poor CO2 sweep efficiencies through effectively increasing CO2 viscosity and reducing its relative permeability [5].
Foam is a dispersion of gas (CO2, in this case) in liquid where the liquid phase is continuous and at least some of the gas phase is discontinuous and separated by lamella [6]. The discontinuous CO2 phase becomes trapped between lamella [7], effectively reducing its relative permeability. Foam apparent viscosity is used as a measure of foam strength in laboratory experiments, where a stronger foam corresponds to a higher apparent viscosity. Entrapment of CO2 in foam and CO2-surfactant emulsification also increases CO2 apparent viscosity and reduces CO2 mobility [8,9]. These combined effects are capable of diverting flow from high permeability, wellswept regions, into less permeable areas with higher oil saturations, thereby increasing macroscopic displacement and oil recovery.
CO2 foam generation and stability is influenced by oil composition and wettability. The presence of oil can destabilize some foams [10] and it has also been reported that foam can only be generated at strongly water-wet conditions [11]. At strongly water-wet conditions, waterwet films covering the pore surfaces maintain the continuous foam structure [12]. A shift to oil-wet can cause the lamellas to detach from the pore walls and foam may be destabilized. Schramm and Mannhardt (1996) confirmed reduced foam effectiveness at intermediate to oil-wet conditions [13] and Fredriksen et al. (2019) induced surfactant wettability alteration in oil-wet fractures to generate CO2 foam in matrix below a critical oil saturation [14]. The impacts of wettability and the presence of oil on foam stability are areas under active investigation.
Foam is generated in porous media by injecting a water (or CO2) soluble foaming agent (surfactant) with gas, such as CO2, either simultaneously (co-injection) or in alternating slugs (SAG). The injection strategy must balance in-situ foam generation, propagation, and injectivity constraints. In practice, there are two main injection strategies for in-situ CO2 foam generation [15,16]. The first is simultaneous injection of CO2 and surfactant solution known as co-injection. In this case, the quality of the foam is determined by the fraction of gas (fg) and is a function of flow rate [17]. The second injection strategy is surfactant-alternating gas (SAG), where the surfactant solution and CO2 are injected in alternating slugs and the quality of the foam depends on the slug sizes of CO2-to-surfactant solution. At laboratory scale, alternating slugs are not typically used due to small fractional flows of gas and the inability to achieve steadystate [18]. At the field scale, operational constraints influence aspects of the injection strategy. For example, co-injection can be difficult to implement due to extremely low injectivity and associated pressure increases, which minimize throughput [19]. In addition, downhole corrosion can occur from carbonic acid during co-injection of CO2 and surfactant solution. This has led to a majority of field tests using SAG injection for better injectivity control, especially when operating close to the fracture pressure [20,21,22].
Few attempts have been made to characterize unsteadystate in-situ CO2 foam behavior. Therefore, this study attempts to begin establishing a knowledge base for investigating core-scale CO2 foam injection strategies. The aim is to reduce CO2 mobility, through the generation of foam, in experiments that are representative of the near wellbore region. This work is part of an ongoing field pilot research program utilizing CO2 foam for mobility control, EOR, and CO2 storage in the Permian Basin of West Texas. Miscible CO2 foam injection will be implemented to reduce high producing gas-oil-ratios, CO2 recycling, and to provide mobility control for increased CO2 sweep efficiency. The reservoir pressure in the pilot area is 220 bars, which is close to the formation fracture pressure of 269 bars [23]. An injection strategy must be selected that will not fracture the formation, which would cause problems beyond any remediation by the foam. Thus, the primary objective of this work is to evaluate coinjection and a variety of SAG injection strategies for CO2 foam mobility control, EOR, and CO2 storage to assist in the design of the field pilot. A secondary objective is to investigate the impacts of miscibility conditions on in-situ foam generation and stability. For more detailed information on the field pilot program see Alcorn et al. (2019).
CO2 foam was evaluated based upon apparent foam viscosity and its impact on oil recovery. In addition, CO2 storage potential was measured for each injection strategy. Experimentally, foam behavior by co-injection, single-cycle SAG, and multi-cycle SAG were performed in brine saturated systems to generate foam in-situ without the presence of oil. CO2 foam EOR corefloods were then conducted at first-contact miscible (FCM) and multi-contact miscible (MCM) conditions using n-Decane and crude oil, respectively, after waterflooding. A localequilibrium foam model was fitted from foam stability scans, and experimental results from a CO2 foam EOR coreflood were used to validate a core-scale numerical model to investigate effects of grid resolution, foam quality, and surfactant. The main objective of the numerical sensitivity study was to validate the foam model in representing the observed foam behavior, which is being extended for use in ongoing field-scale simulations.

Materials and Fluids
Outcrop limestone core plugs were used as reservoir analogues due to limited reservoir cores available from the field. Core plugs were drilled with a 2-inch diameter from larger slabs, cut, cleaned and dried before being 100% saturated with synthetic Permian Basin brine under vacuum. Values of porosity and pore volumes were calculated based on weight differential before and after fluid saturation. Absolute permeability was measured by injecting brine until a stable differential pressure was obtained for three different flood rates. Permeability measured for the single core plugs varied between 8mD to 73mD and porosities from 19 to 35%.
Synthetic Permian Basin brine was made based upon water analysis from the field, whereas a light North Sea crude oil was used. The North Sea crude oil has an API of 33.6° (calculated from specific gravity), a little above the API gravity of the Permian Basin crude of 31° at standard conditions. Compositions of brine and crude oil are reported elsewhere [14,24]. The North Sea crude oil is considered MCM with CO2 at 60°C and 180 bar, with a minimum miscibility pressure (MMP) of 125 bar [25]. Reservoir conditions for the field test are well above MMP for CO2 and crude oil. To investigate the effects of miscibility conditions on foam generation and stability a refined oil, n-Decane (C10H22), was also selected for the first set of EOR corefloods to obtain FCM conditions with CO2.
A non-ionic surfactant (i.e. Huntsman L24-22) was previously selected for the CO2 foam field pilot from a surfactant screening study that quantified adsorption on reservoir rock with and without CO2 present [26]. The non-ionic surfactant was also screened for its ability to alter wettability of oil-wet carbonate rocks to weakly oilwet conditions in favor of foam generation [14]. The nonionic surfactant was used at a 1.0 wt% concentration in Permian Basin Brine.

Coreflood Set-up
The core-scale system was composed of two stacked core plugs providing a total nominal length of 25 cm to generate foam in-situ. Cores with similar porosity and absolute permeability were paired and stacked. An overview of experiments and their stacked system properties are presented in Table 1. The cores were mounted in a horizontally oriented Hassler type core holder and experimental conditions were set to 60°C, to avoid crude oil wax precipitation. Pore pressure was set to 180 bars for CO2 to be MCM with the North Sea crude oil. At these conditions CO2 is supercritical and will create an emulsified phase (liquid-in-liquid) with the surfactant solution. A differential pressure transducer and two absolute pressure transducers (i.e. one downstream and one upstream) measured pressure response. The standard setup for high-pressure/high-temperature CO2 foam injection can be found in [27].
In-situ foam generation, strength, and stability is evaluated by its apparent viscosity, which is quantified from flow rate and pressure drop during foam injection [28]. Foam apparent viscosity is calculated by; where, is the absolute permeability of the porous media, ∇ is the pressure gradient measured, and and are the superficial velocities of liquid and gas, respectively [17].

Unsteady-State Foam Stability Scan by SAG
Foam generation and stability was investigated for two modes of SAG injection in a foam stability scan. Singlecycle SAG was run at 100% brine saturated conditions prior to multi-cycle SAG on the same core (E2). A waterflood to re-establish initial conditions with close to zero CO2 saturation was run in between the foam stability scans. For both SAG injections, pre-determined slug sizes were injected for 4.0 pore volumes (PVs) targeting a gas fraction (fg) of 0.70. For single-cycle SAG, a single slug of surfactant solution was injected (1 PV) before CO2 was introduced for 3 PVs. For multi-cycle SAG, 12 rapid cycles were run: each cycle consisting of a surfactant slug of 0.11 PVs and a CO2 slug of 0.22 PVs. Un-steady state apparent foam viscosities were calculated as a function of time (i.e. PVs injected) using Equation 1.

Steady-State Foam Quality and Rate Scans by Coinjection
Foam generation and stability during co-injection was assessed by foam quality and rate scans. Tests were run in 100% brine saturated cores, where one system was composed of a single core (D1) and the other a stacked system (D2-D3). Scans were also performed at residual oil saturation after CO2 foam EOR to investigate the effect of residual oil on foam stability (E13-E14), and equivalently for a base case without surfactant (E15-E16). See Table 1 for experimental overview.
Foam quality scans determine the optimal gas fraction (fg) that will generate the highest apparent viscosity during coinjection. CO2 fractions were changed from 0.0 to 0.90 for drainage-like co-injection (i.e. increasing CO2 fraction) at a total superficial injection rate of 1.0 ft/day. Each fraction was injected until steady state pressure drop was achieved before increasing to the next fraction. The apparent foam viscosity was calculated from Equation 1. Rate scans were performed following foam quality scans on the same core(s) to estimate rate-dependency on shear-thinning behavior. Rate scans for co-injection were run by increasing the total injection rate at the optimal CO2 fraction from quality scans. Starting at a superficial velocity of 1.0 ft/day, the injection rate was increased to 2.0, 3.0, and 4.0 ft/day after reaching steady-state conditions.

CO2 Foam EOR
Core plugs were initially 100% brine saturated, stacked, and drained with either n-Decane or dead crude oil to irreducible water saturation (Swi) at a constant pressure drop of 2 bar/cm. A waterflood was performed for 1 PV prior to CO2 foam injection. Injection strategies during CO2 foam were either co-injection, single-cycle, or multicycle SAG. CO2 foam was injected for no more than 2 PVs, considering volumetric and economic limitations at the field-scale. Produced fluids were measured at ambient conditions as stock tank volumes. Density and compressibility were considered negligible for the specific dead crude oil and refined oil used in these experiments. Saturations were calculated based upon the difference in initial oil in place and oil produced divided the total pore volume.
CO2 foam EOR was performed at FCM conditions using n-Decane and at MCM conditions using North Sea crude oil. At FCM conditions, CO2 foam co-injection used two injection rates (1.0 ft/day for E3-E4 and 2.0 ft/day for E5-E6) to investigate the shear-thinning behavior of foam during EOR. Co-injection results were then compared with single cycle (E7-E8) and multi cycle SAG (E9-E10). The most promising (highest apparent foam viscosity and oil recovery) injection strategies were evaluated at MCM conditions, with crude oil, and compared with base case experiments without surfactant solution.

Core-Scale Model Set-up
Laboratory data from a co-injection experiment (E13-E14) was utilized for a core-scale simulation model. The model was initialized to represent the stacked system during waterflood and co-injection. For the waterflood, simulations were conducted with ECLIPSE 100 Blackoil simulator while the compositional simulator E300 (Schlumberger, 2015.2) was used for co-injection. Experimental data validated the model through matching bottom hole pressure (BHP) and cumulative oil/water production. The validated model was used to conduct a sensitivity analysis on the effect of grid cell size, foam quality, and surfactant on oil recovery and CO2 mobility reduction. The main objective was to ensure model robustness at representing the observed foam behavior, which is extended to use in ongoing field scale simulations.
The base case model consisted of a rectangular grid with dimensions 1 x 1 x 100 (x, y, z). The length of the model was identical to the experimental core system length of 24.7 cm. The model and individual grid cells were also consistent with the core diameter of 4.8 cm thus, each individual grid cell measured 4.8 x 4.8 cm in the x-and ydirections. The injector was located in the first grid block (inlet), while the producer was located in the last grid block (outlet). The production well was placed on BHP control and the injector was controlled by rate both of which were measured in the laboratory. Relative permeability data for the waterflood was derived from JBN analysis of oil and water displacement during the laboratory experiment [29]. Oil and water densities and viscosities were available from PVT-analysis of the crude oil. The model was initiated with Swi of 0.24 at a system pressure of 182 bars.
The co-injection was initialized at pressure and saturations from the history-matched waterflood. The grid size, orientation, well completions and controls were kept identical, except two injection wells were used to represent the single co-injection well from the experiment (one for CO2 and one for surfactant solution). A compositional simulation case was generated which contained 14 oil components and 2 water components (water and surfactant). Relative permeability curves were derived from CO2/brine displacement experiments on similar core material described elsewhere [23]. Despite their influence on foam behavior, capillary pressure effects were not included in this study. This is an area, which merits further investigation.
The injection schedule was identical to the experimental procedure and a foam quality of 70% was targeted using a surfactant solution concentration of 1.0 wt%. The effect of foam was modeled using an empirical local-equilibrium approach where the gas relative permeability in the presence of foam is modified by multiplying the gas relative permeability without foam by a mobility reduction factor (MRF) [30]. The mobility reduction factor is dependent upon water saturation, oil saturation, surfactant concentration and shear rate. A maximum gas mobility reduction (fmmob) is also used to set the maximum mobility reduction that can be achieved by foam. These parameters were derived from foam quality and rate scans and fit to the empirical model by curve fitting regression [23,30].

Results and Discussion
Apparent viscosities and incremental oil recoveries were used to evaluate foam generation, stability, and EOR performance during CO2 foam injection. Several mechanisms contribute to the foam apparent viscosity (and pressure response) including foam generation, trapped gas, and CO2-surfactant emulsification.

Unsteady-State Foam Stability by SAG
CO2 foam stability scans by single-cycle and multi-cycle SAG injections were compared in initially brine saturated systems without oil present (Figure 1). Foam apparent viscosity during single-cycle SAG was 18.2 ± 2.7 cP (dashed curve, Figure 1), and 120.2 ± 0.3 cP for multicycle SAG (solid curve, Figure 1). Based upon the increase in apparent viscosity, foam was generated when the first CO2 slug was injected for both injection modes ( Figure 1).
During single-cycle SAG, foam generation occurred within the first PV of CO2 injected and foam remained stable for the next 2 PV with only a slight dry-out effect towards the end of injection (black dashed curve, Figure  1). Injecting multiple alternating slugs of surfactant solution and CO2 improved conditions for foam generation and stability. During multi-cycle SAG, surfactant solution was introduced to the system in an imbibition process, which caused a decrease in capillary pressure, likely triggering foam generation. This change in capillary pressure is advantageous for foam generation since the creation of lamella requires exceeding a minimum pressure gradient. Hence, a decrease in capillary pressure during the surfactant slugs improved conditions for foam generation. Increased wetting-phase saturation during surfactant solution slugs also may have mitigated foam dry-out. The growth and propagation of a stable high apparent viscosity foam was clear during multi-cycle SAG (solid curve, Figure 1). Despite efforts to return the core to 100% water saturation, the presence of trapped CO2 from the previous single-cycle SAG experiment likely reduced CO2 relative permeability and increased the pressure drop during multi-cycle SAG. This would result in higher foam apparent viscosities. A decrease in effective water permeability was observed for E2 from 42.3 mD to 19.8 mD by capillary trapped CO2 prior to multi-cycle SAG foam stability. However, during multi-cycle SAG, the step-wise increase in differential pressure and significantly increased apparent viscosity compared to single-cycle SAG suggest that the alternating injection scheme did improve conditions for foam generation and stability. Figure 2 shows co-injection foam quality scans, D1 and D2-D3 (orange and green curves), which have previously been published [14,31] and are shown here for comparison. Peak apparent foam viscosities at steady state for fg = 0.70 was 34.9 ± 1.0 cP (D1), and 56.0 ± 1.6 cP (D2-D3), higher than single-cycle SAG (18.2 ± 2.7 cP), but lower than multi-cycle SAG (120.2 ± 0.3 cP) in Figure 1. The foam strengths measured during the two co-injections are relatively similar, considering the heterogeneous core, but discrepancy remains when comparing the three coinjection experiments to the SAGs. It is likely that the injection scheme is not the only explanation for the variation in apparent viscosities.

Steady-State Foam Quality and Rate Scans by Coinjection
Foam quality scans with residual oil also investigated the influence of crude oil on foam stability. Experiment E13-E14 in Figure 2 (purple curve) was performed after CO2 foam EOR at a residual oil saturation (Sor) = 0.05 (purple curve, Figure 2). A lower apparent foam viscosity was observed for every fg compared to foam quality scans conducted with 100% initial water saturation (i.e. strongly water-wet) in Figure 2 (orange and green curves). However, apparent viscosities at fg = 0.70 were 13.20 cP for experiment E13-E14 with surfactant (purple curve, Figure 2) compared to 4.19 cP for the base case (E15-E16, blue curve, Figure 2) at Sor = 0.06, without surfactant. Hence, these experiments show that this foam system is capable of generating foam with low apparent viscosities when crude oil is present.

First-Contact Miscible Conditions
Co-injection and SAG injection strategies were evaluated based upon their apparent viscosity and oil recovery at FCM conditions ( Figure 3). The initial waterfloods recovered between 31.2 % and 47.7% OOIP and a clean water cut was observed for all experiments (left of the vertical black dashed line). The range in waterflood recovery can be attributed to core heterogeneity [32]. Recovery factors and apparent viscosity values are listed in Table 2.
CO2 foam co-injections were performed at two different injection rates (1.0 ft/day and 2.0 ft/day) to investigate shear-thinning behavior during EOR corefloods ( Figure  3). Foam was generated earlier at higher rate (at 1.7 PVs injected, E5-E6, orange dashed curve) compared to the lower co-injection rate (at 2.3 PVs injected, E3-E4, orange solid curve) from dynamic observations of apparent viscosity. The average apparent viscosity, however, was higher at 1.0 ft/day (28.1 cP, E3-E4) than for 2.0 ft/day (18 cP, E5-E6) for the last 0.5 PVs injected ( Table 2). This is an indication of the shear-thinning behavior of foam at increasing flow rates. Shear-thinning behavior was also observed in foam rate scans shown in Figure 2 (bottom).  Apparent viscosity trends during SAG EOR (green curves, Figure 3) were the same as for foam stability scans without oil present ( Figure 1). The increase in apparent viscosity was more rapid for each SAG injection (green curves) than either of the co-injection experiments (orange curves), likely due a faster increase in CO2 saturation. However, it is possible that a higher apparent viscosity could have been obtained by continuing the coinjection at 1.0 ft/day (Figure 3, solid orange curve). CO2 foam apparent viscosity for multi-cycle SAG (Figure 3, dashed green curve) increased continuously for each cycle and reached an average value of 100.7 cP at the end of injection. In terms of mobility control, multi-cycle SAG was superior in creating conditions for high viscous displacement during EOR corefloods at FCM conditions ( Figure 3).
The presence of the alkane n-Decane oil did not negatively influence foam generation or propagation for either injection strategy (Figure 3). N-Decane is a nonpolar oil molecule with has no ability to alter wetting-state like that of heavy polar molecules (i.e. asphaltenes and resins). Aging carbonate rock in n-Decane does not alter wettability [33] and so stable foam was able to be created in-situ, even in the presence of oil.
Accelerated oil recovery rate was observed from the start of both co-injections ( Figure 3, open and closed orange circles) and most of the oil was produced after 1.0 PV of CO2 foam injected. For single-cycle and multi-cycle SAG (solid green circles), no oil was recovered during the initial surfactant slug before CO2 was injected. Thus, diffusion dominated oil recovery above that of viscous displacement by foam at FCM conditions. CO2 diffusion is a dominant recovery mechanism at core-scale with the potential to recover nearly 100% of the oil [34,35]. At constant fg = 0.70, CO2 diffusion recovered the same amount of oil regardless of injection mode, on average 29.7 ± 2.2% OOIP. As observed in Figure 3, however, total recoveries did not reach the ultimate recovery potential of 100% OOIP. This is because stable foam lamellas can create barriers that hinder direct contact between the discontinuous CO2 phase and unrecovered oil, negatively impacting oil recovery. Values of incremental oil recovery during CO2 foam are listed in Table 2.

Multiple-Contact Miscible Conditions
Viscous forces are desirable for mobility control and fluid diversion during CO2 foam EOR. Co-injection at 1.0 ft/day and multi-cycle SAG gave best results in terms of both apparent viscosity and EOR at FCM conditions. These injection strategies were therefore used to determine sensitivity in presence of multicomponent crude oil at MCM conditions. Base case experiments were also performed without surfactant solution as a reference.
Results are presented in Figure 4.
The initial waterflood recovered on average 62.0 ± 2.8% OOIP (blue curves, Figure 4) with two-phase production, a characteristic of less water-wet conditions [36]. The measured apparent viscosities during CO2 foam experiments demonstrated little to no in-situ foam generation. No increase in foam apparent viscosity was observed above that of the base cases without surfactant. Two possible explanations are offered for the absence foam generation:  Oil composition is known to influence lamella creation and foam stability, and its presence can make some foams more unstable than others. Foam stability can be reduced with decreasing carbon content in crude oils. It has also been observed elsewhere that a lower number alkanes are more destabilizing than higher number alkanes, because their shorter hydrocarbon chains are more easily imbibed into foam plateau borders to solubilize with surfactants leading to oil spreading and foam instability [10,[37][38][39].
 Wettability alteration by crude oil towards less waterwet conditions can hinder foam generation and strength. It has previously been reported that foam cannot be generated at wettability conditions other than strongly water-wet due to the lack of water-wet snap-off sites [40,16,13]. The multi-component crude oil likely induced a shift towards oil-wet conditions, which caused the lamellas to detach from the pore walls, restricting foam generation. CO2 foam oil recoveries were on average 30.6 ± 3.0% OOIP for all injection strategies with and without surfactant, at both FCM and MCM conditions. Therefore, incremental oil recoveries were also driven by CO2 diffusion at MCM conditions ( Figure 4). Bernard and Holm observed the same effect in their core-flood experiments with CO2 foam [44]. This restricts evaluation of injection strategies in terms of production efficiency. Longer induction periods were observed, however, for continuous oil banks to develop under MCM conditions in Figure 4. Between 0.3 to 0.4 PVs of CO2 foam were injected before oil production was observed (red curves). Total recoveries were higher at MCM conditions (93.2 ± 2.7% OOIP) compared to FCM conditions (71.2 ± 3.2% OOIP) as the waterfloods at MCM conditions left behind lower residual oil saturations ( Table 2). Further sensitivity analysis on the effect of CO2 fractions on foam behavior was considered using the history-matched core-scale simulation model.

Associated CO2 Storage
A secondary objective for implementing CO2 foam for mobility control in EOR is the potential for storing CO2 as a part of carbon capture, utilization, and storage (CCUS). To calculate CO2 stored, the volume of CO2 produced was measured from volumetric conservation of injected and produced fluids, and subtracted from the volume of CO2 injected. Calculations were made for the FCM and MCM experiments. Based upon mass balance, the value for CO2 storage (given as a fraction of total PV) is equal to the saturation change of oil and water in the core after CO2 foam. Figure 5 shows CO2 storage in fraction of PV for both coinjections (top) and multi-cycle SAGs (bottom). A direct correlation between the amounts of CO2 stored and the water/oil saturation change was observed regardless of injection strategy. For every amount of fluid produced, the equivalent amount of CO2 was stored. CO2 storage potential was 9.0% greater for multi-cycle SAGs compared to co-injections at MCM. However, CO2 storage potential was 17.1% greater at FCM conditions (for either injection strategy), compared to MCM, due to improved CO2 foam displacement and increased CO2 trapping by capillary forces in more water-wet core plugs.

Core-Scale Model Validation and Sensitivity Study
The core-scale simulation model utilized the foam stability measurements in Figure 2 to fit the empirical local equilibrium foam model by curve fitting regression [23,30]. The value for the maximum gas mobility reduction, fmmob, however, was reduced to 41.5 in agreement with previous findings on field core material to reflect more realistic conditions for the field system. An acceptable history-match for experiment E13-E14 was obtained for both waterflood and co-injection. The waterflood match was achieved by tuning the oil relative permeability curve to match oil/water production rate and BHP. The co-injection was matched by tuning the oil and water relative permeabilities to match cumulative oil/water production and BHP. Figure 6 shows the history-match. Fig. 6. Observed (open circles) and modeled (curves) cumulative oil/water production during waterflood and coinjection of the history-matched experiment. Production well BHP is shown on the secondary y-axis.
The validated model was first used to investigate the effect of grid resolution during co-injection. The generated case was identical to the base case historymatch, but used the fine scale grid (5 times finer in the x, y, and z directions) to evaluate change in CO2 mobility reduction and oil recovery. CO2 mobility reduction is generally inferred from delayed CO2 breakthrough and an increased response in injection pressure. However, simulation results showed limited effect of grid resolution on co-injection in terms of injection well pressures, CO2 breakthrough, and cumulative recoveries. As limited variations were observed, with the fine grid model, further sensitivity studies utilized the coarser base grid.

Effect of Foam Quality
The effect of foam quality (CO2 fraction) was studied using the base grid to determine its impacts on oil recovery and CO2 mobility reduction. Cases injecting higher CO2 fractions were set to assess model sensitivity on amounts of CO2 injected and its influence on oil recovery. The base case history-matched model used the optimal CO2 fraction of 0.70 measured in laboratory. Further sensitivity cases were set to CO2 fractions of 0.80, 0.90, and 0.95. Figure 7 shows cumulative oil production (solid curves) for the base case (fg=0.70) and CO2 fraction sensitives. With increasing CO2 fractions, the oil recovery rate accelerated but all cases recovered the same volume of cumulative oil. Hence, CO2 miscibility dominated oil displacement as also observed in the CO2 foam EOR corefloods experiments in Figure 3 and 4.
Injection pressures for the various CO2 fractions are shown in Figure 7 (dashed curves). All injection pressures followed the same trend. Higher pressures, however, were observed as the fraction of CO2 decreased and the fraction of surfactant solution increased, creating a higher apparent viscosity foam (blue and red dashed curves, Figure 7). This is consistent with trends observed in foam quality scans where the optimal CO2 fraction and highest apparent viscosity was observed at 0.70 (cf. Figure 2, left, green and orange circles). At fractions above this, apparent viscosities declined.

Effect of Surfactant
CO2 diffusion and miscibility are evidently the dominant oil recovery mechanisms in both simulation and laboratory experiments. Therefore, a similar case to the history-matched experiment was set-up without surfactant to investigate pure CO2 versus CO2 foam displacement. The injection schedule was kept identical to the base case co-injection, except only water was injected as the aqueous phase. Figure 8 shows injection pressure (dashed curves) and cumulative CO2 production (solid curves) during co-injection with surfactant (blue curves) and without surfactant (red curves). The injection well pressure was significantly lower for the case without surfactant compared to the base case with surfactant ( Figure 8, red dashed curve). This indicated higher CO2 mobility and increased CO2 production in absence of foam ( Figure 8, red solid curve). Analysis of liquid production showed the same cumulative volume of oil recovered in both cases (green curves, Figure 9), but additional water was produced during CO2 foam injection (blue dashed curve). Hence, a larger storage potential for CO2 was obtained with CO2 foam displacement. Similar behavior was also observed in Figure 5, where CO2 storage was higher for multi-cycle SAG because of increased water displacement during foam injection compared to CO2 alone.
The similar volumes of oil produced with and without surfactant, demonstrated the dominance of CO2 miscibility over viscous displacement by foam (green curves, Figure 9). The core-scale model is consistent with laboratory observations indicating that miscibility and diffusion are the governing displacement forces in small core-scale systems. This creates a challenge when upscaling core-scale foam behavior to the field-scale as reservoir heterogeneity and gravity effects will likely be more dominant in the field. Fig. 9. Cumulative liquid production during CO2/water coinjection (solid curves) and CO2/surfactant solution (dashed curves). Green curves correspond to cumulative oil produced and blue to cumulative water.

Conclusions
This work investigated various injections strategies (coinjection and SAG) for CO2 foam mobility control, EOR and CO2 storage to assist in the design of a CO2 foam field pilot. The sensitivity study involved both experimental laboratory work and numerical modeling. Supercritical CO2 foam behavior in brine-saturated systems was investigated to generate foam in-situ, without the presence of oil. Additionally, CO2 foam EOR corefloods were conducted in the presence of refined oil at first-contact miscible (FCM) conditions and in the presence of crude oil at multiple-contact miscible (MCM) conditions after waterflooding to investigate the impact of oil and miscibility on foam generation and stability. Key findings from this work are:  Multi-cycle SAG resulted in the highest apparent viscosity foam of 120.2 cP during in-situ CO2 foam stability scans compared to co-injection (56.0 cP) and single-cycle SAG (18.2 cP) without oil present. Multicycle SAG also achieved the highest apparent viscosity foam of 100.7 cP for the CO2 foam EOR corefloods, with refined oil present, at FCM conditions.  Incremental oil recoveries during tertiary CO2 foam injections were on average 30.6% OOIP for all injection strategies, with and without surfactant, at both FCM and MCM conditions. At MCM conditions, CO2 foam was not generated as a result of wettability alteration by crude oil and foam destabilization in presence of crude oil.  CO2 diffusion and miscibility with oil were the dominant recovery mechanisms as observed in laboratory corefloods and numerical core-scale sensitivity studies on foam quality.  A validated numerical core-scale model captured the observed foam behavior from laboratory corefloods. The foam model was not sensitive to grid resolution and corroborated laboratory observations of corescale foam behavior. Further work is ongoing to upscale the laboratory observations to field-scale simulations.  A direct correlation between the amounts of CO2 stored and water/oil saturation change was observed in the laboratory corefloods. For every amount of fluid produced, the equivalent amount of CO2 was stored regardless of injection strategy. CO2 storage potential was 17.1% greater at FCM conditions, compared to MCM, due to increased displacement by CO2 foam and CO2 trapping by capillary forces in more waterwet core plugs. Core-scale simulations indicated higher CO2 storage potential with CO2 foam because of increased water displacement, compared to cases without foam.
Reservoir heterogeneity and gravity dominate displacement at the field-scale, therefore the high apparent viscosities and viscous displacement forces provided by multi-cycle SAG in the foam stability scans are favorable. The shear-thinning behavior of the foam system can also mitigate injectivity loss near the injection well, where flow rates are high. Additionally, a SAG injection scheme provides better injectivity control, when operating close to formation fracture pressure, due to the ability to switch to CO2 injection for foam dry-out. Reported oil recoveries, from CO2 foam EOR corefloods, cannot be upscaled to predict field performance as CO2 diffusion will have less effect on displacement at the length scales existing in the field.