Pressure maintenance and improving oil recovery with immiscible CO2 injection in thin heavy oil reservoirs

Pressure maintenance and improving oil recovery with immiscible CO2 injection in thin heavy oil reservoirs

Journal of Petroleum Science and Engineering 112 (2013) 139–152 Contents lists available at ScienceDirect Journal of Petroleum Science and Engineeri...

11MB Sizes 5 Downloads 86 Views

Journal of Petroleum Science and Engineering 112 (2013) 139–152

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering journal homepage: www.elsevier.com/locate/petrol

Pressure maintenance and improving oil recovery with immiscible CO2 injection in thin heavy oil reservoirs Sixu Zheng, Huazhou Li, Daoyong Yang n Petroleum Systems Engineering, Faculty of Engineering and Applied Science, University of Regina, Regina, Saskatchewan, Canada S4S 0A2

art ic l e i nf o

a b s t r a c t

Article history: Received 2 October 2012 Accepted 30 October 2013 Available online 7 November 2013

Techniques have been developed to experimentally and numerically evaluate performance of CO2 injection in heavy oil reservoirs for pressure maintenance purpose. More specifically, a threedimensional (3D) physical model consisting of five vertical wells and three horizontal wells is used to examine the effect of well configurations on pressure maintenance and oil recovery with CO2 injection in heavy oil reservoirs. The initial oil saturation, oil production rate, water cut, gas–oil ratio, ultimate oil recovery, and distribution of residual oil saturation are examined under three well configurations, which can be optimized to maximize heavy oil recovery when CO2 injection is employed for pressure maintenance purpose. Subsequently, numerical simulation has been performed to match the experimental measurements. It has been found that pressure maintenance with CO2 injection is beneficial for oil recovery in heavy oil reservoirs. Compared with the conventional five-spot well configuration, the well configurations associated with horizontal well(s) are found to achieve a better performance, while the well configuration of two horizontal wells (Scenario ♯3) yields the highest oil recovery. The oil recovery of Scenario ♯3 is experimentally determined to be 38.6% of original oil in place (OOIP) after CO2 injection. & 2013 Elsevier B.V. All rights reserved.

Keywords: pressure maintenance immiscible CO2 injection heavy oil well configuration

1. Introduction Heavy oil has been considered as a key resource to prop up the world's energy demand because the global energy consumption is escalating and conventional petroleum reserves are depleting. There are tremendous heavy oil reserves all over the world, especially in Western Canada (i.e., Alberta and Saskatchewan). About 62% of Western Canada's total heavy oil reserves are located in Saskatchewan, including 0.35 billion m3 of proved reserves (Bowers and Drummond, 1997). As for these reserves, more than 90% is contained in reservoirs with the payzones of less than 10 m, about 55% of which have the payzones less than 5 m (Huang et al., 1987). In recent years, thermal-based techniques, such as cyclic steam stimulation (CSS) and steam-assisted gravity drainage (SAGD), have conventionally been utilized to enhance heavy oil recovery because heavy oil viscosity is very sensitive to temperature. However, characteristics of such thin reservoirs result in excessive heat losses to adjacent formations. In addition, such great heat and large water consumption make thermal techniques ineffective and uneconomical in such heavy oil reservoirs. In situ combustion has been proposed to be a potential enhanced oil recovery (EOR) technique for heavy oil reservoirs. However, more

n

Corresponding author. Tel.: þ 1 306 337 2660; fax: þ 1 306 585 4855. E-mail address: [email protected] (D. Yang).

0920-4105/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.petrol.2013.10.020

viscous or immobile components may occur in formation during in situ combustion process because the high temperature oxidation reaction model is difficult to be achieved or maintained. As a result, only a small proportion of oil can be produced when low temperature oxidation is the dominant reaction model (Moore et al., 1999; Shen, 2002). Chemical flooding has been considered as one of promising EOR techniques, whereas conventional chemical flooding is susceptible to formation brine. The requirement of soft water may not always be satisfied, thus, the brine with high salinity and divalent ions leads to chemical flooding ineffectiveness in many cases (Flaaten et al., 2009; Hsu et al., 2012; Novosad et al., 1982). In addition, due to high level of adsorption within the porous media, some chemical techniques associated with surfactants are uneconomical (Fletcher et al., 2012). Ethane and propane can be used as solvents to recover heavy oil reservoirs; however, their respective vapor pressures are relatively low. Furthermore, ethane and propane are expensive commodities. Alternatively, CO2 has been found to be an efficient agent for recovering heavy oil resources through an immiscible displacement process (Sahin et al., 2008). Although pressure maintenance via gas injection has been applied in light and medium oil reservoirs, few attempts have been made to evaluate performance of CO2 injection for such purpose in a heavy oil reservoir. It is of practical and fundamental importance to evaluate suitability of pressure maintenance and improving heavy oil recovery with CO2 injection in thin payzones where other EOR techniques are not applicable.

140

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

After being injected into an oil reservoir, CO2 leads to swelling the oil, reducing oil viscosity, exerting an acidic effect on rock, and vaporizing and extracting light-components of crude oil (Holm and Josendal, 1974). Due to the shallow depth of heavy oil reservoirs discovered in Western Canada, it is difficult to achieve miscibility between the injected CO2 and heavy oil in formation. Although it is less effective, immiscible CO2 injection can still contribute to increasing oil recovery by initiating viscosity reduction, oil swelling, interfacial tension reduction, and blowdown recovery (Jha, 1986; Rojas and Farouq Ali, 1988). CO2 immiscible displacement includes continuous CO2 injection, the CO2 slug process, injection of CO2 alternating water, and simultaneous injection of CO2 and water (Rojas and Farouq Ali, 1986). The simultaneous injection of CO2 and water has been seldom used in field applications. This is mainly due to severe corrosion in the injection wells at high pressures and injectivity loss associated with simultaneous injection. In addition, the other three processes are operationally easier than simultaneous injection of CO2 and water. Viscous force was experimentally found to dominate the immiscible CO2 injection for heavy oil recovery in thin payzones, while viscous instability occurred because of the high CO2–oil mobility ratio (Rojas and Farouq Ali, 1986). Moreover, an increase in viscosity of the crude oil leads to a decrease in CO2–oil mass transfer, resulting in a remarkable reduction in heavy oil recovery by immiscible CO2 injection. Bagci (2007) conducted immiscible CO2 flooding by using three groups of well combinations, i.e., (1) a vertical injector and a vertical producer (VI–VP), (2) a vertical injector and a horizontal producer (VI–HP), and (3) a horizontal

injector and a horizontal producer (HI–HP). Oil recovery of the VI– HP well configuration was higher than that of the other well configurations during a continuous CO2 injection process. CO2 breakthrough occurred very early in all runs mainly due to the dominant viscous forces and relatively small effect of mass transfer between CO2 and oil. Although effects of well configuration have been investigated by Bagci (2007), the contribution of CO2 dissolution on oil recovery has not been well examined due to the low injection pressures (less than 200 kPa). Parasiliti Parracello et al. (2012) experimentally evaluated the performance of CO2 injection in a heavy oil reservoir by using coreflooding tests. Their results indicate that CO2 injection is an effective method to increase oil recovery; however, the effects of well configuration and CO2 displacement efficiency have not been revealed due to the small scale. The effectiveness of CO2 injection in Alaska North Slope reservoirs has been numerically evaluated (Al-Quraini et al., 2007; Ning et al., 2011). It is found that continuous liquid CO2 injection produces almost the same amount of oil compared to waterflooding, while the combination of water and CO2 injection shows better results than waterflooding without considering well configurations associated with horizontal wells. Although the immiscible CO2 has been widely applied and satisfactory results have been achieved in the Bati Raman field, the potential benefits of horizontal wells during the CO2 injection process have not been investigated because of the technical and economical difficulties (Sahin et al., 2008). In this paper, a three-dimensional (3D) displacement model with five vertical wells and three horizontal wells is developed

Fig. 1. Schematic diagram of well configurations: (a) five-spot well pattern (Five-spot); (b) four vertical injectors and one horizontal producer (4VI–HP); and (c) one horizontal injector and one horizontal producer (HI–HP).

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

and used to evaluate performance of pressure maintenance and improving oil recovery with immiscible CO2 injection in heavy oil reservoirs. Three well configurations have been designed to examine their effects on heavy oil recovery. They are (a) fivespot well pattern (Scenario ♯1), (b) four vertical injectors with one horizontal producer (Scenario ♯2), and (c) one horizontal injector and one horizontal producer (Scenario ♯3). The schematic diagram of these well configurations is shown in Fig. 1. For each well configuration, water is first injected into the 3D displacement model to simulate the waterflooding process, and then continuous CO2 injection is conducted to maintain reservoir pressure. Subsequently, numerical simulation is performed to match experimental measurements. The well configurations associated with horizontal well show a better performance than the conventional five-spot well pattern, while the highest oil recovery of 38.6% is achieved from Scenario ♯3. The good agreement between numerical and experimental results indicates that numerical simulation has captured the overall behavior of both waterflooding and CO2 injection in heavy oil reservoirs.

Table 1 Compositional analysis result of oil sample.

2. Experimental 2.1. Materials The heavy oil sample and reservoir brine are collected from Lashburn area, Canada. The initial reservoir temperature is 21.0 1C and reservoir pressure is about 3500 kPa. The density of oil sample is measured to be 976.0 kg/m3 at 15.6 1C and atmospheric pressure. Viscosity of the oil sample is measured to be 5861 mPa·s at 20.0 1C and atmospheric pressure with a Brookfield DV-IIþ Programmable Viscometer (Brookfield Engineering Laboratories, USA). The asphaltene content of the original heavy oil is measured to be 14.4 wt% (n-pentane insoluble). As can be seen from Table 1, the compositional analysis indicates that there are no hydrocarbon components lighter than C9 in this heavy oil, while the heavy components of C30 þ account for 61.33 wt%. Density of reservoir brine is measured to be 1051.7 kg/m3 at 20.0 1C. The primary ions of the reservoir brine are chloride, sodium, calcium, and magnesium. The brine viscosity is determined to be 1.03 mPa·s at 20.0 1C and atmospheric pressure. According to the compositional analysis of the reservoir brine, synthetic brine is prepared and used in this study, while its composition is shown in Table 2. In this study, CO2 with a purity of 99.998% is purchased from the Praxair, Canada. Densities of CO2 at different pressures and temperatures are calculated by using the CMG WinProp module (Version 2009.11, Computer Modelling Group Ltd., Canada) with the Peng–Robinson equation of state (Peng and Robinson, 1976). Ottawa sand ♯710 (Bell & Mackenzie, Co., Ltd) is used to pack the 3D displacement model. The Ottawa sand is found to be water-wet and has not been treated in any way. The corresponding screen analysis is listed in Table 3. 2.2. Experimental setup The schematic diagram of experimental setup is shown in Fig. 2. This setup is composed of the fluid supply system, the 3D displacement system, and the fluid production system. The fluid supply system is used to supply both CO2 and brine. As for supplying CO2, the system includes a CO2 cylinder, a gas regulator, and a digital pressure gauge, while, as for supplying brine, a high pressure syringe pump (500HP, Teledyne ISCO Inc., USA) and two transfer cylinders are used. The CO2 cylinder is used to provide continuous CO2 to the 3D displacement model, while the synthetic brine is introduced from the transfer cylinders to the 3D model by using the syringe pump at a constant flow rate.

141

Carbon number

wt%

C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16 C17 C18 C19 C20 C21 C22 C23 C24 C25 C26 C27 C28 C29 C30 þ

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.96 0.99 1.26 1.54 1.95 2.18 2.55 2.29 2.29 2.60 2.23 2.02 2.39 1.44 1.98 1.67 1.67 1.70 1.72 1.75 1.50 61.33

Table 2 Composition of synthetic brine. Component

Concentration (mg/L)

Chloride Sodium Calcium Magnesium Potassium

43,677 24,700 1500 850 413

Table 3 Typical screen analysis of ♯710 Ottawa sand. USA STD sieve size

Retained

Mesh

mm

(wt%)

40 50 70 100 140 200 270

0.425 0.300 0.212 0.150 0.106 0.075 0.053

Trace 0.3 7.0 52.0 32.0 8.0 0.7

The 3D displacement model with dimension of 304 mm  304 mm  127 mm consists of five vertical wells and three horizontal wells. The wall of the 3D model has a thickness of 25.4 mm. A total of 28 bolts are used to fasten the steel lid and a piece of rubber gasket is applied for sealing the model. Three parallel horizontal wells are located at the bottom of the void space, while five vertical wells are distributed as a conventional five-spot well pattern in the model (see Fig. 1). This 3D displacement model can

142

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

Digital pressure

3D physical model

Brine

Brine

CO2

Injection valve

BPR

Production valve

Gas flow meter To atmosphere

N2

Syringe pump

Oil sample collector Personal computer Fig. 2. Schematic diagram of the experimental setup.

be used to perform experiments under five-spot, line drive, and other well patterns associated with horizontal wells. All the wells have a diameter of 6.4 mm, and small holes are perforated around circumference to allow fluids to flow. The wells are wrapped with a 250-mesh wire-screen (Ferrier Wire Goods, Canada) to prevent sand production during the tests. The maximum operating pressure of the 3D displacement system is 6.0 MPa at room temperature. The production system is comprised of a digital pressure gauge, a back pressure regulator (BPR) (EB1HP1, Equilibar, USA), an oil sample collector, and a gas flow meter (XFM17S, Aalborg, USA). The BPR is used to maintain a pre-specified pressure inside the physical model during each test, while a nitrogen cylinder is applied to exert reference pressure on the BPR. The separated gas is measured by the gas flow meter before being exhausted to the atmosphere, and produced liquids are collected by using conical-bottom glass centrifuge tubes (Kimble, USA). The produced CO2 flow rate can be logged and stored in a computer, while the total produced gas volume is read directly from the flow meter after separation during the test.

2.3. Experimental preparation 2.3.1. Sand-packing The physical model is placed horizontally for filling the void space with the Ottawa sand prior to any experiments. Once being fully packed, the physical model is covered and tightened by the steel lid with the rubber gasket in place. In order to achieve a tight packing, the physical model is shaken with a pneumatic vibrator (NP 35, Northern Vibrator, USA) for at least 10 h. While removing the cover and adding more sand to the formed void space in the model, the model is subsequently covered, sealed, tightened, and shaken again. The same process is normally repeated 4–5 times until no more void space is observed at the top of the physical model. Finally, leakage test is conducted up to the pressure of 5.0 MPa at room temperature for 30 min once the sand-packing has been finished.

2.3.2. Porosity measurement The imbibition method is used to measure porosity of the sandpacked physical model (Dong and Dullien, 2006). Firstly, the physical model is connected to a vacuum pump (M12C, Fisher Scientific, Canada) for evacuation through the injection valve. During this process, the injection valve is left open, while the production valve is kept tightly closed. Secondly, the production valve is submerged into a container filled with the synthetic brine whose weight and density are known. Thirdly, the injection valve is closed, while the production valve is opened to allow brine to imbibe into the model and saturate the sandpack. Fourthly, weight of the remaining brine sample in the container is measured. Finally, total volume of brine is calculated based on the weight difference of water measured before and after imbibition, and then porosity of the sandpack is calculated. In this study, the porosities of Scenarios ♯1, ♯2, and ♯3 are determined to be 34.2%, 33.2%, and 33.3%, respectively. 2.3.3. Permeability determination The model is placed in the horizontal position to measure the absolute permeability. Two diagonally opposite vertical wells are used to inject and produce the synthetic brine. During the test, different flow rates and cumulative produced volume are measured at corresponding injection pressures. Then, the CMG IMEX module (Version 2009.11, Computer Modelling Group Ltd, Canada) is used to determine absolute permeability by history matching the production profile. In this study, absolute permeabilities of Scenarios ♯1, ♯2 and ♯3 are determined to be 1700 mD, 1600 mD, and 1800 mD, respectively. 2.3.4. Initial oil saturation measurement After absolute permeability determination, the heavy oil sample is introduced to the 3D displacement model through its three bottom horizontal wells at a rate of 0.4–0.6 cm3/min from a transfer cylinder by using a syringe pump. At the same time, all the other ports located at the lid are left open to release fluids

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

from the physical model. Due to the high absolute permeability of the sandpack and a low injection rate, the injection pressure drop is less than 250 kPa. The injection is terminated when no further water production is observed from the open ports. This means that the model has been saturated with heavy oil sample to ensure reasonable initial oil saturation under irreducible water saturation. The volume of the heavy oil saturated into the model is then recorded. The initial oil saturation is calculated to be the ratio of the volume of saturated oil to the pore volume of sandpack in the physical model. After initial oil saturation being established, the experimental model is left undisturbed for 24 h to equilibrate the distribution of fluid. Effect of temperature on the relative permeability endpoints has been previously studied. It is found that irreducible water saturation increases with temperature, while residual oil saturation decreases with temperature (Etminan et al., 2008; Maini and Okazawa, 1987; Poston et al., 1970; Weinbrandt et al., 1975). As such, the temperature under which the heavy oil is saturated into the sandpack should not be very high in order to achieve a high initial oil saturation. In addition, saturating oil is such a process that heavy oil is the displacing phase, while water is the displaced phase. The increased temperature will decrease heavy oil viscosity, which may induce early oil breakthrough, resulting in a low and heterogeneous initial oil saturation. In this study, the increased temperature is not used to initially saturate sandpack in the experiments. In order to obtain similar initial oil saturation for the three scenarios, displacement of water with heavy oil is conducted at the same temperature of 21.0 1C, resulting in similar initial oil saturations of 95.5%, 96.5%, and 93.9% for Scenarios ♯1, ♯2, and ♯3, respectively. 2.3.5. Asphaltene content measurement Precipitation of asphaltene from the heavy oil sample has been determined by using the standard ASTM D2007-03 method (ASTM D2007-3, 2007). More specifically, one volume of the heavy oil sample is poured into a beaker and mixed with 40 volumes of liquid n-pentane that is used as a precipitant. The mixture is stirred by utilizing a magnetic stirrer (120SQ, Fisher Scientific, USA) for 12 h and then filtered through filter paper with 2 μm pore size. Prior to filtering the oil–precipitant mixture, the filter paper is weighed with an electronic balance (SP2001, Ohaus Corporation, USA). The filtration is terminated until the mixed fluids after passing through the filter paper remain colorless. The filter cake is formed at the end the filtration, which is primarily composed of precipitant asphaltene. The filter paper with the precipitated asphaltene is gently dried at temperature of 100.0 1C in an oven until the total weight does not change from the reading of the electronic balance. According to the weight change of the filter paper measured before and after filtration, the asphaltene content of the oil sample is determined to be 14.4 wt%. 2.3.6. Residual oil saturation measurement Once a blowdown process is completed, residual oil saturation has been measured. From top to bottom, the sandpack is divided into four layers and marked as Layers ♯1–♯4 in sequence, each of which is about 3.0 cm thick. Mixtures of sand and reservoir fluids are collected from 13 locations in each layer to measure the residual oil saturations. Firstly, the weight of each collected sample that contains sand, residual oil, and water is measured. Secondly, the collected sample is heated in an oven at 100.0 1C for 4 h to ensure water is completely evaporated. Then the remaining sand– oil mixture is weighed. It should be noted the heavy oil sample used in this study does not have any hydrocarbon components lighter than C9 whose normal boiling point temperature is 143.0 1C and heavy components (C30 þ ) account for a large proportion (more than 60 wt%). Therefore, it can be reasonably assumed that

143

trace of hydrocarbon components has been evaporated and the measurements will not be significantly affected by such evaporation when the heating temperature is set to be 100.0 1C for evaporating water. Thirdly, toluene and kerosene are used as solvents to completely remove the heavy oil from the sample mixture of heavy oil and sand. Finally, the sand is heated in the oven at 110.0 1C and then weighed until it is completely dried. Consequently, residual oil saturation is calculated based on the weight differences, porosity of the sandpack model, oil density and sand density. 2.4. Experimental scenarios Three well configurations, i.e., five-spot well pattern (Scenario ♯1), four vertical injectors and one horizontal producer (Scenario ♯2), and one horizontal injector and one horizontal producer (Scenario ♯3) are designed in this study, as shown in Fig. 1. As for Scenarios ♯1 and ♯2, water injection is first initiated from the four vertical injectors located in the four corners of the 3D model to displace heavy oil, while the oil is produced from the central vertical producer (Scenario ♯1) or horizontal producer (Scenario ♯2). Water is injected through the horizontal injector and the other horizontal well is used as producer for Scenario ♯3. The synthetic brine is injected into the 3D displacement model at a constant rate of 2.0 cm3/min that corresponds to an average pore velocity of 1.0 ft/day for oilfield implementation (Hadia et al., 2008). Water injection is terminated when the water cut is higher than 99%. After waterflooding, CO2 is injected into the 3D model to examine the effect of well configurations on pressure maintenance and oil recovery. Following the preceding well configurations, CO2 is injected into the displacement model through the injector(s). A gas regulator connected with a CO2 cylinder is set at a prespecified injection pressure of 3900 kPa. The production port is connected with the BPR, while the reference pressure of 3860 kPa on the BPR is applied by using a nitrogen cylinder connected with a gas regulator. In consequence, the pressure inside the physical model is maintained at the pre-specified pressure of 3860 kPa during the CO2 injection process. Once a sufficient pressure drop is achieved, fluids start producing and filling the oil sample collector. The produced gas is flashed off from the oil sample collector and then measured by using a gas flow meter (XFM17S, Aalborg, USA). All the flow rates and accumulative gas production are recorded and stored automatically in a computer at a preset frequency.

3. Numerical simulation In order to better understand the overall behavior of waterflooding and pressure maintenance with CO2 injection in heavy oil reservoirs, numerical simulation has been performed to match the experimental measurements of waterflooding and CO2 injection processes. The composition of heavy oil and experimental conditions are crucial factors to determine a numerical simulator. As shown in Table 1, C9 is the lightest hydrocarbon component of the heavy oil sample, while heavy components are in a large proportion. The operating pressure (about 4.0 MPa) at temperature of 21.0 1C is far below the minimum miscibility pressure (MMP) of 15.5 MPa that is calculated by the CMG WinProp module (Version 2009.11, Computer Modelling Group Ltd. (CMG), Canada). Neither light components extraction nor the miscible flooding has been achieved at the experimental conditions in this study. Since it has been proven that the black oil simulator is an effective tool for simulating performance of immiscible flooding processes (Cobanoglu, 2001), the black oil simulator of IMEX (Version 2009.11, Computer Modelling Group Ltd., Canada) has been used

144

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

in this study. Because it is not reliable to generate the equation of state (EOS) model for simulation without sufficient laboratory PVT data, as recommended by the CMG, the genetic algorithm (GA)based analytical correlations (Emera and Sarma, 2008) have been used to calculate CO2 solubility and the viscosity of the oil–CO2 mixture in this study. The GA correlations can be applied over a wide range of conditions to predict oil swelling factor and viscosity with a good accuracy (Emera and Sarma, 2008). The simulation model is created by using the CMG Builder module (CMG, Version 2009.13) with 13 grid blocks in the I-direction, 13 grid blocks in the J-direction, and 6 grid blocks in the K-direction, resulting in a fine grid block size of 2.34 cm  2.34 cm  2.10 cm. Uniform distribution of porosity and oil saturation are used in the model. The capillary pressure effect is neglected in this paper, because it has been pointed out that capillary pressure has negligible effect on oil recovery behavior for highly viscous oil displacement from unconsolidated sands (Pujol and Boberg, 1972; Rojas and Farouq Ali, 1988). In order to simulate the experimental displacement process, the well operational conditions defined in simulation models are set same as those used in the experiments. For example, the production port is connected with the BPR that is used to maintain the inside pressure of the physical model at 3860 kPa. Then, in the simulation models, the bottomhole pressure (BHP) of producer is set to be 3860 kPa. During a water injection process, the water injection rate in simulation model is same as the experimental water injection rate of 2.0 cm3/min. For CO2 injection process, the CO2 injection pressure in the simulation model is set to be 3900 kPa, because the CO2 injection pressure in the experimental system is regulated to be 3900 kPa by using a gas regulator that is connected with the CO2 cylinder.

4. Results and discussion

is measured to be 14.9%, 16.1%, and 17.1% after 1.0 pore volume (PV) water injection for Scenarios ♯1, ♯2, and ♯3, respectively. The oil recovery after 1.0 PV water injection is not high. This is mainly due to the fact that viscosity of heavy oil sample is noticeably higher than that of water. The mobility ratio of the injected brine to heavy oil is high, leading to severe viscous fingering. As a result, water breakthrough occurs very early and the efficiency of the waterflooding is correspondingly low. As shown in Fig. 3, oil recovery for Scenarios ♯1 and ♯3 remains close at the early stage, while Scenario ♯2 yields a higher oil recovery. With a further increase of water injection, the well configurations associated with horizontal well (i.e., Scenarios ♯2 and ♯3) achieve much higher oil recovery than the conventional five-spot configuration (Scenario ♯1). The highest oil recovery is found to be 18.5% at 1.2 PV of water injection for Scenario ♯3. This can be attributed to the fact that a horizontal injector and a horizontal producer are used in Scenario ♯3, which facilitates achieving a better sweep efficiency. For light and medium oil waterfloods, as water is injected, oil is continuously produced until breakthrough, and a relatively high recovery factor can be achieved prior to breakthrough (Hadia et al., 2007). The waterflooding behavior in a heavy oil reservoir is very different from that of light or medium oil reservoirs. As can be seen in Fig. 4, it is obvious that, due to early breakthrough, water cut is increased rapidly to around 80.0% when 0.2 PV water is injected. It should be noted that considerable heavy oil is still produced at such a high water cut, though water breakthrough has already occurred. This is ascribed to the fact that continuous channels from injector(s) to producer have been generated by water after its breakthrough. Subsequently, the injected water prefers to continue to follow in such channels with low resistance. Along these channels, water can contact oil to displace it to the producer. This finding is also reported in the literature (Mai and Kantzas, 2009). Water cuts for Scenarios ♯1, ♯2, and ♯3 are measured to be higher than 90.0% when 1.0 PV of water is injected.

4.1. Waterflooding

4.2. CO2 injection

Oil recovery profiles for the three scenarios during waterflooding process are shown in Fig. 3. As can be seen, oil recovery

After waterflooding with a specific well configuration, CO2 is injected into the 3D displacement model to examine the effect of

Fig. 3. Oil recovery during displacement experiments for three well configurations.

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

145

Fig. 4. Water cut of produced liquid during displacement experiments for three well configurations.

well configurations on pressure maintenance and oil recovery. Due to the fact that the performance of CO2 injection with a larger amount of PV injection has not been well examined in the literature, a higher PV of CO2 injection is performed in the experiments in this study. As mentioned previously, during the CO2 injection process, the BPR is used to maintain the pressure of the physical model at 3860 kPa that is slightly higher than the original reservoir pressure of 3500 kPa, and injection pressure is set to be 3900 kPa by means of gas regulator. The injected CO2 is still in gas phase, while the viscosity of injected CO2 at 3900 kPa and 21.0 1C is determined to be 0.0167 mPa·s by CMG WinProp module (Version 2009.11). It is experimentally found that gas breakthrough occurs early after CO2 injection is initiated for all three scenarios. This is attributed to the low resistance channel generated during waterflooding and the high mobility ratio between CO2 and reservoir fluids (water and heavy oil), both of which lead to severe CO2 fingering. Such early CO2 breakthrough indicates that viscous forces dominate the immiscible CO2 injection and there exists relatively small effect of mass transfer between CO2 and heavy oil (Bagci, 2007; Rojas and Farouq Ali, 1986; Tüzünoǧlu, Baǧci, 2000). By contrast, CO2 injection in heavy oil reservoirs with thick payzone may confront earlier CO2 breakthrough due to the heterogeneity (Sahin et al., 2008). As can also be seen in Fig. 4, water cut is declined sharply for all three well configurations during CO2 injection processes. When 20.0 PV of CO2 is injected into the model, water cut is decreased to 5.8%, 0.8% and 0.3% for Scenarios ♯1, ♯2, and ♯3, respectively. This means that CO2 moves quickly in the channels that are generated during waterflooding, while the injected CO2 contacts heavy oil along these channels and displaces oil out. In addition, CO2 shows its preference to displace heavy oil, though there still exists a significant amount of water in the porous media. This results in a very low water cut at a later time of CO2 injection. Obviously, there is a singular point (about 40.0 PV of CO2 injection) for the measured water cut in Scenario ♯1 (see Fig. 4), where the water cut increases abruptly rather than continues to decrease. This may be ascribed to the fact that the water that is trapped and/or bypassed by CO2 at early stage has now been mobilized and produced with the heavy oil. Subsequently, water cut keeps decreasing and remains almost constant after 62.0 PV CO2 is injected. Also, there exist similar points at 1.1 PV and 11.3 PV of CO2 injections for Scenarios ♯2 and ♯3, respectively, though they are much less obvious. Oil recovery during the CO2 injection process has also been plotted in Fig. 3. It can be seen from this figure that CO2 injection

Fig. 5. Cumulative gas–oil ratios of three well configurations during the CO2 injection process.

has a favorable effect on pressure maintenance and oil recovery, though waterflooding has been implemented. In addition, oil recovery is found to be strongly dependent on well configuration during the CO2 injection process. After CO2 injection, oil recovery is measured to be 19.5% for Scenario ♯1, 32.2% for Scenario ♯2, and 38.6% for Scenario ♯3, respectively. Furthermore, it can be observed that Scenarios ♯2 and ♯3 show a later breakthrough than Scenario ♯1 through visual observation during experiments, though breakthrough occurs at early stage of all three scenarios. Such a large difference is due to a better sweep efficiency resulting from the well configuration with horizontal well(s). In general, a horizontal producer controls a large area than that of a vertical well. The presence of the horizontal producer located at the bottom of the physical model alleviates CO2 override and delays early CO2 production, leading to gas accumulating at the top of the model and exerting an extra force to drive heavy oil from top to the producer. It is experimentally found that tiny gas (CO2) bubbles evolve from the produced heavy oil, though the operating pressure is not very high. Such dissolution of CO2 leads to oil swelling, viscosity reduction, and interfacial tension reduction, contributing to enhancing oil recovery (Jha, 1986; Rojas and Farouq Ali, 1988). The injected CO2 not only acts as a pressure maintenance agent,

146

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

but also enhances oil recovery during the pressure maintenance process in heavy oil reservoirs. The poor sweep efficiency of Scenario ♯1 implies that CO2 override is severe, resulting in early gas breakthrough and the minor effect of mass transfer between CO2 and heavy oil (Tüzünoǧlu, Baǧci, 2000). It can also be found that, although oil recovery reaches its plateau after 1.1 PV of water injection, CO2 injection can still enhance oil recovery. The injected CO2 is in gas phase under the experimental conditions (i.e., about 4.0 MPa and

21.0 1C). The microscopic displacement of the oil by gas is normally better than by water (Christensen et al., 2001). Therefore, the increased oil recovery by CO2 injection is not only resulted from the dissolved CO2 in heavy oil, but also due to a better microscopic displacement efficiency of CO2 injection than that of waterflooding. Fig. 5 shows the cumulative gas–oil ratios during the CO2 injection process. The cumulative gas–oil ratio for Scenario ♯1 is extremely high, which is another indication for poor sweep efficiency of the five-spot well configuration. As for Scenarios ♯2 and ♯3, the relatively low cumulative gas-ratios imply that the well configuration associated with horizontal wells can produce much more oil at the same PV injection of CO2. 4.3. Blowdown recovery

Fig. 6. Water cut as a function of blowdown recovery.

After termination of CO2 injection, the blowdown process is performed, in which no fluid is injected into the 3D model, while the bottomhole pressure (BHP) of producer is set to be atmospheric pressure. The reservoir fluids are produced until the pressure of the physical model is depleted to atmospheric pressure for blowdown recovery. As shown in Fig. 3, the blowdown process contributes an incremental oil recovery of 6.7%, 5.1%, and 2.3% for Scenarios ♯1, ♯2, and ♯3, respectively. Solution gas drive is the main mechanism that contributes to blowdown recovery (Jha, 1986). CO2 readily dissolves into heavy oil, while energy is stored by CO2 dissolution. During a blowdown process, the stored energy releases and continues to drive oil to producers. At the same time, when pressure is decreased, fluids (heavy oil, CO2, and water) tend to expand in the formation. Such an

Fig. 7. Distribution of residual oil saturation in the top layer: (a) Scenario ♯1, (b) Scenario ♯2, and (c) Scenario ♯3.

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

expansion of fluids exerts displacement force to drive oil to producers, while evolution of the dissolved CO2 from heavy oil due to sharp pressure decrease also agitates the distribution of heavy oil. The synergistic effect allows more oil to be displaced and produced. Fig. 6 illustrates water cut versus oil recovery during the blowdown process for Scenarios ♯2 and ♯3. It can be seen that water cut is low at early stage, then increases rapidly to reach its peak, and finally decreases. Residual water is not mobilized at the early stage of blowdown recovery, while much oil is agitated and displaced by the evolved CO2. It is experimentally found that little fluids are produced when pressure is decreased from 3800 kPa to 2000 kPa, whereas most of fluids are produced at pressure below 2000 kPa and even below 1000 kPa. The blowdown process lasts more than 2 h, while about 100 L of CO2 is produced at ambient temperature and atmospheric pressure. Gas keeps producing from the displacement model even though the inside pressure of the model approximates to the atmospheric pressure.

147

It is worthwhile mentioning that the tiny bubbles are observed to evolve from the produced liquids. This is another indication that CO2 has been dissolved into heavy oil during CO2 injection processes, though the operating pressure is not very high. The produced CO2 with a large volume during a blowdown process and the evolved tiny bubbles from heavy oil indicate that CO2 storage occurs as the injected CO2 dissolves into reservoir fluids, displaces the fluids from formation, and occupies some of pore spaces that are previously filled by reservoir fluids. Therefore, a large amount of CO2 can be sequestrated by controlling the reservoir pressure at a relatively high value during the process of CO2 injection in heavy oil reservoirs. 4.4. Distribution of residual oil saturation After the blowdown process, the steel lid is removed and the distribution of residual oil saturation in the top layer can be observed. Digital images of the top layer for three scenarios are

Fig. 8. Distribution of residual oil saturation for Scenario ♯1 in (a) Layer ♯1, (b) Layer ♯2, (c) Layer ♯3, and (d) Layer ♯4.

148

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

shown in Fig. 7. It is clearly found that the color of the top layer is dark, indicating that there still exists much oil in the top layer for Scenario ♯1. This implies the heavy oil in top layer of Scenario ♯1 has not been swept effectively, though about 1.2 PV of water and 76.0 PV of CO2 have been injected into the 3D displacement model in sequence. By contrast, the color of sands along the horizontal producer (Well ♯5) is very light for Scenario ♯2. It indicates that the area along the horizontal producer has been well swept, leading to a low residual oil saturation. This is ascribed to the fact that, as CO2 is injected into the model, it accumulates at the top of the model, and then displaces the heavy oil to the horizontal producer. The bottom-left portion of Scenario ♯3 is much darker than its topright portion, indicating that it is difficult to establish a stable displacement front along the horizontal injector. A large portion of the injected fluids invades the sandpack through the heel of the horizontal injector, and then displaces the heavy oil to the producer, resulting in a well-displaced triangle area (i.e., top-right portion). This observation implies that it is possible to improve the sweep

efficiency by changing the position of heels and toes of these two parallel horizontal wells. The adverse effect of gas override is alleviated by using the horizontal producer that is located at the bottom of the physical model. It adequately explains the advantage of the well configuration containing the horizontal producer that yields high oil recovery and low cumulative gas–oil ratio for Scenarios ♯2 and ♯3. In order to examine the residual oil saturation for different layers of the 3D displacement model, distribution of residual oil saturation in the model is measured and illustrated in Figs. 8–10. As aforementioned, 13 samples are collected from different locations for each layer. The residual oil saturation contours are generated based on their measurements at 13 locations, marked as the circle symbols in the figures. As can be seen, for Scenario ♯1, the residual oil saturation is high in Layer ♯1, which is consistent with the dark color observed from Fig. 7a. Residual oil saturation in Layer ♯4 is much higher than that of Layers ♯1 and ♯2. This implies that bottom of the model has not been well swept by CO2

Fig. 9. Distribution of residual oil saturation for Scenario ♯2 in (a) Layer ♯1, (b) Layer ♯2, (c) Layer ♯3, and (d) Layer ♯4.

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

149

Fig. 10. Distribution of residual oil saturation for Scenario ♯3 in (a) Layer ♯1, (b) Layer ♯2, (c) Layer ♯3, and (d) Layer ♯4.

so that a high residual oil saturation is resulted from gas override. This finding is consistent with the phenomena of early breakthrough and high gas-ratio discovered for Scenario ♯1. As for Scenario ♯2, the residual oil saturation for each of four layers is shown in Fig. 9. It is clearly found that the residual oil saturation along the horizontal producer is low, while the areas around four vertical injectors show high residual oil saturation. As shown in Fig. 9, the low residual oil saturation along the horizontal producer is another indication that the horizontal producer located at the bottom of reservoir has the advantages of contributing to a better sweep efficiency and alleviating gas override. Meanwhile, it is similar to Layer ♯4 of Scenario ♯1, much residual oil is found at the bottom of the model for Scenario ♯2. The comparison between Figs. 7 and 9 demonstrates that the generated residual oil saturation profiles have successfully described the real distribution of residual oil saturation in the 3D displacement model. Fig. 10 shows the residual oil saturation of Scenario ♯3 under the well configuration of HI–HP. It is obviously found that

the stripe area along the horizontal producer has low residual oil saturation, while the residual oil saturation along the horizontal injector is much higher. The reservoir area near the horizontal producer has been efficiently swept by injected fluids. The average oil saturation of the bottom layer is higher than that of top layers. It is experimentally observed that gas override occurs during the CO2 injection process and the horizontal wells located at the bottom of the model can alleviate the influence of gas override in the experiments. In comparison with the performance of CO2 injection in thin payzones, gas override is more severe due to gravity segregation in thick reservoirs, resulting in a strong tendency for the injected CO2 to migrate upward until an impermeable barrier is encountered (Ning et al., 2011; Spivak et al., 1990). As a result, the bottom payzone will not be well displaced under the conventional well configurations with vertical wells. Therefore, the CO2 injection under the well configurations associated with horizontal wells has a great potential to improve oil recovery in thin payzones.

150

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

Fig. 11. Relative permeability curves for Scenario ♯1: (a) the water–oil system and (b) the liquid–gas system.

Fig. 12. Relative permeability curves for Scenario ♯2: (a) the water–oil system and (b) the liquid–gas system.

4.5. History matching

relative permeability curves are tuned and determined once the experimental results have been matched. Such determined relative permeability curves for Scenarios ♯1, ♯2 and ♯3 are shown in Figs. 11, 12 and 13, respectively. It can be seen that the residual oil saturation is high, and water and oil relative permeabilities are low in the high water saturation range. These are in accordance with the previous study (Nejad et al., 2011; Wang et al., 2006). As can be seen from water–oil relative permeability curves, water relative permeability increases slightly at low water saturation range (less than 0.3). With an increase in water saturation, oil relative permeability decreases rapidly. The relative permeability of water is not very high at high water saturation range. The performance of waterflooding is very sensitive to the relative permeability values when water saturation is less than 0.2. As for liquid–gas relative permeability curves, liquid and gas relative permeabilities change dramatically in low gas saturation range (less than 0.1). The relative permeability of oil reduces sharply as gas saturation increases, while gas relative permeability increases quickly. This implies that gas has high mobility in heavy oil, leading to adverse gas–oil mobility ratio and severe gas fingering. The cumulative oil production of simulation is compared with the experimental data in Figs. 14–16. The good matching results

As for Scenarios ♯1–♯3, waterflooding and CO2 injection processes have been history matched in this study. Since the relative permeabilities cannot be directly measured from experiments, they are to be determined by using the history matching techniques instead (Eydinov et al., 2009; Li et al., 2012; Li and Yang, 2012; Watson et al., 1980). In order to achieve the history matching, the well operational conditions defined in simulation models are set the same as those used in the experiments, while the relative permeability curves are tuned manually. More specifically, the production port is maintained at the prespecified operating pressure of 3860 kPa by using the BPR in experiments. In simulation models, the bottomhole pressure of the producer is then set to be 3860 kPa. During the water injection process, the water injection rate of 2.0 cm3/min is used in the experiments. Accordingly, the water injection rate in the simulation models is set at the same value. For the CO2 injection process, the CO2 injection pressure is set to be 3900 kPa in the simulation models, which is the same as the CO2 injection pressure that is controlled by a gas regulator in experiments. The initial inputs of the threephase relative permeability curves are generated with the commonly used correlation (i.e., Stone's II model). Subsequently, the

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

151

Fig. 15. Measured and simulated cumulative oil production for Scenario ♯2.

Fig. 13. Relative permeability curves for Scenario ♯3: (a) the water–oil system and (b) the liquid–gas system.

Fig. 16. Measured and simulated cumulative oil production for Scenario ♯3.

efficient afterwards. These are consistent with the high water cut after 1.0 PV of water injection shown in Fig. 4. As for the CO2 injection following the waterflooding process, oil recovery is enhanced although oil production has reached the plateau at the late stage of the waterflooding process. Therefore, pressure maintenance with CO2 injection is favorable for oil recovery in heavy oil reservoirs.

5. Conclusions

Fig. 14. Measured and simulated cumulative oil production for Scenario ♯1.

show that the numerical simulation has captured the overall behavior of the waterflooding and CO2 injection processes. As shown in Figs. 14–16, oil recovery flattens off after 1.0 PV water injection (2000 min), indicating that waterflooding becomes less

Experimentally, a 3D physical model is developed and successfully used to investigate the suitability of pressure maintenance and improving oil recovery with immiscible CO2 injection in heavy oil reservoirs. Three scenarios are designed to evaluate the effect of well configurations on pressure maintenance and oil recovery via CO2 injection. The well configuration shows a considerable effect on pressure maintenance and oil recovery with CO2 injection in heavy oil reservoirs. The well configuration with a horizontal producer (Scenarios ♯2 and ♯3) not only achieves a higher oil recovery during the waterflooding process, but also shows a better performance in the CO2 injection process than the conventional five-spot well pattern (Scenario ♯1). Scenario ♯3 (HI–HP) yields the highest oil recovery, indicating that two horizontal wells located at the bottom of the reservoir are favorable for oil recovery. It is

152

S. Zheng et al. / Journal of Petroleum Science and Engineering 112 (2013) 139–152

experimentally found that gas override is severe under well configuration of five-spot (Scenario ♯1), while the well configuration in Scenarios ♯2 and ♯3 with horizontal wells that are located at the bottom of the 3D displacement model alleviates gas override, leading to a better sweep efficiency and a higher oil recovery. The excellent agreement between the experimental and simulated results indicates that the overall behavior of waterflooding and CO2 injection processes have been numerically simulated and matched. CO2 breakthrough occurs early due to the adverse mobility ratio of CO2 to heavy oil. Viscous forces are found to dominate the displacement process, though viscous fingering is severe. CO2 injection enhances oil recovery because of not only the high microscopic displacement efficiency of CO2, but also the enhancing oil recovery mechanisms, such as the swelling effect, viscosity reduction, and interfacial tension reduction. Acknowledgments The authors acknowledge a Discovery Grant and a CRD Grant from the Natural Sciences and Engineering Research Council (NSERC) of Canada and an innovation fund from the Petroleum Technology Research Centre (PTRC) to D. Yang and EHR Enhanced Hydrocarbon Recovery Inc. for financial support. References Al-Quraini, A., Sohrabi, M., Jamiolahmady, M., 2007. Heavy oil recovery by liquid CO2/water injection. Paper SPE 107163, presented at the SPE Europec/EAGE Conference and Exhibition, London, UK, June 11–14. ASTM D2007-03, 2007. Standard Test Method for Characteristic Groups in Rubber Extender and Processing Oils and other Petroleum-Derived Oils by the Clay–Gel Absorption Chromatographic Method. ASTM International, West Conshohocken, PA. Bagci, A.S., 2007. Immiscible CO2 flooding through horizontal wells. Energy Sources Part A 29 (1), 85–95. Bowers, B., Drummond, K.J., 1997. Conventional crude resources of the Western Canada sedimentary basin. J. Can. Pet. Technol. 36 (2), 56–63. Christensen, J.R., Stenby, E.H., Skauge, A., 2001. Review of WAG field experience. SPE Reserv. Eval. Eng. 4 (2), 97–106. Cobanoglu, M., 2001. A numerical study to evaluate the use of WAG as an EOR method for oil production improvement at B. Kozluca Field, Turkey. Paper SPE 72127, presented at the SPE Asia Pacific Improved Oil Recovery Conference, Kuala Lumpur, Malaysia, October 6–9. Dong, M., Dullien, F.A.L., 2006. Porous Media Flows in Multiphase Flow Handbook. In: Crowe, C.T. (Ed.), 2006. Taylor & Francis Group, New York. Emera, M.K., Sarma, H.K., 2008. A genetic algorithm-based model to predict CO2–oil physical properties for dead and live oil. J. Can. Pet. Technol. 47 (2), 52–61. Etminan, S.R., Maini, B.B., Kharrat, R., 2008. The role of connate water saturation in VAPEX process. J. Can. Pet. Technol. 47 (2), 8–12. Eydinov, D., Gao, G., Li, G., Reynolds, A.C., 2009. Simultaneous estimation of relative permeability and porosity/permeability fields by history matching production data. J. Can. Pet. Technol. 48 (12), 13–25. Flaaten, A.K., Nguyen, Q.P., Pope, G.A., Zhang, J., 2009. A systematic laboratory approach to low-cost, high-performance chemical flooding. SPE Reserv. Eval. Eng. 12 (5), 713–723. Fletcher, P., Cobos, S., Jaska, C., Forsyth, J., Crabtree, M., Gaillard, N., Favero, C., 2012. Improving heavy oil recovery using an enhanced polymer system. Paper SPE 154045, presented at the SPE Improved Oil Recovery Symposium, Tulsa, OK, April 14–18. Hadia, N., Chaudhari, L., Mitra, S.K., Vinjamur, M., Singh, R., 2007. Experimental investigation of use of horizontal wells in waterflooding. J. Pet. Sci. Eng. 56 (4), 303–310.

Hadia, N., Chaudhari, L., Mitra, S.K., Vinjamur, M., Singh, R., 2008. Waterflood profiles and oil recovery with vertical and horizontal wells. Energy Sources Part A. 30 (17), 1604–1618. Holm, L.W., Josendal, V.A., 1974. Mechanisms of oil displacement by carbon dioxide. J. Pet. Technol. 26 (12), 1427–1438. Hsu, T., Lohateeraparp, P., Roberts, B.L., Wan, W., Lin, Z., Wang, X., Budhathoki, M., Ben Shiau, B.J., Harwell, J.H., 2012. Improved oil recovery by chemical flood from high salinity reservoirs-single-well surfactant injection test. Paper SPE 154838, presented at the SPE EOR Conference at Oil and Gas West Asia, Muscat, Oman, April 16–18. Huang, S.S., Pappas, E.S., Jha, K.N., 1987. The carbon dioxide immiscible recovery process and its potential for Saskatchewan reservoirs. Paper SS-87-1, presented at the Technical Meeting/Petroleum Conference of the South Saskatchewan Section, Regina, SK, October 6–8. Jha, K.N., 1986. A laboratory study of heavy oil recovery with carbon dioxide. J. Pet. Technol. 25 (2), 54–63. Li, H., Chen, S., Yang, D., Tontiwachwuthikul, P., 2012. Estimation of relative permeability by assisted history matching using the ensemble kalman filter method. J. Can. Pet. Technol. 51 (3), 205–214. Li, H., Yang, D., 2012. Estimation of multiple petrophysical parameters for hydrocarbon reservoirs with the ensemble-based technique. Adv. Pet. Explor. Dev. 4 (2), 1–17. Mai, A., Kantzas, A., 2009. Heavy oil waterflooding: effects of flow rate and oil viscosity. J. Can. Pet. Technol. 48 (3), 42–51. Maini, B.B., Okazawa, T., 1987. Effects of temperature on heavy-oil–water relative permeability of sand. J. Pet. Technol. 26 (3), 33–41. Moore, R.G., Laureshen, C.J., Ursenbach, M.G., Mehta, S.A., Belgrave, J.D.M., 1999. A Canadian perspective on in situ combustion. J. Can. Pet. Technol. 38 (13), 1–8. Nejad, K.S., Berg, E.A., Ringen, J.K., 2011. Effect of oil viscosity on water/oil relative permeability. Paper SCA 2011–12, presented at the International Symposium of the Society of Core Analysts, Austin, TX, September 18–21. Ning, S., Jhaveri, B., Jia, N., Chambers, B., Gao, J., 2011. Viscosity reduction EOR with CO2 & enriched CO2 to improve recovery of Alaska North Slope viscous oils. Paper SPE 144358, presented at the SPE Western North American Regional Meeting, Alaska, USA, May 7–11. Novosad, J., Maini, B., Batycky, J., 1982. A study of surfactant flooding at high salinity and hardness. J. Am. Oil Chem. Soc. 59 (10), 833A–839A. Parasiliti Parracello, V., Bartosek, M., De Simoni, M., Mallardo, C., 2012. Experimental evaluation of CO2 injection in a heavy oil reservoir. Paper IPTC 14869, presented at the International Petroleum Technology Conference, Bangkok, Thailand, February 7–9. Peng, D.Y., Robinson, D.B., 1976. A new two-constant equation of state. Ind. Eng. Chem. Fundam. 15 (1), 58–64. Poston, S.W., Ysrael, S., Hossain, A.K.M.S., Montgomery III, E.F., Ramey Jr., H.J., 1970. The effect of temperature on irreducible water saturation and relative permeability of unconsolidated sands. SPE J. 10 (2), 171–180. Pujol, L., Boberg, T.C., 1972. Scaling accuracy of laboratory steam flooding models. Paper SPE 4191, presented at the SPE California Regional Meeting, Bakersfield, CA, November 8–10. Rojas, G.A., Farouq Ali, S.M., 1986. Scaled model studies of carbon dioxide/brine injection strategies for heavy oil recovery from thin formations. J. Can. Pet. Technol. 25 (1), 85–94. Rojas, G.A., Farouq Ali, S.M., 1988. Dynamics of subcritical CO2/brine floods for heavy-oil recovery. SPE Reserv. Eng. 3 (1), 35–44. Sahin, S., Kalfa, U., Celebioglu, D., 2008. Bati Raman field immiscible CO2 application-status quo and future plans. SPE Reserv. Eval. Eng. 11 (4), 778–791. Shen, C., 2002. Limitations and potentials of in-situ combustion processes for heavy oil reservoirs. Paper SPE 2002-217, presented at the Canadian International Petroleum Conference, Calgary, AB, June 11–13. Spivak, A., Garrison, W.H., Nguyen, J.P., 1990. Review of an immiscible CO2 project, Tar zone, Fault Block V, Wilmington Field, California. SPE Reserv. Eng. 5 (2), 155–162. Tüzünoǧ lu, E., Baǧ ci, S., 2000. Scaled 3-D model studies of immiscible CO2 flooding using horizontal wells. J. Pet. Sci. Eng. 26 (1–4), 67–81. Wang, J., Dong, M., Asghari, K., 2006. Effect of oil viscosity on heavy oil–water relative permeability curves. Paper SPE 99763, presented at the SPE/DOE Symposium on Improved Oil Recovery, Tulsa, OK, April 22–26. Watson, A.T., Seinfeld, J.H., Gavalas, G.R., Woo, P.T., 1980. History matching in twophase petroleum reservoirs. SPE J. 20 (6), 521–532. Weinbrandt, R.M., Ramey Jr., H.J., Casse, F.J., 1975. The effect of temperature on relative and absolute permeability of sandstones. SPE J. 15 (5), 376–384.