Pressure Maintenance and Improving Oil Recovery by Means of Immiscible Water-Alternating-CO Processes in Thin Heavy-Oil Reservoirs
Pressure Maintenance and Improving Oil Recovery by Means of Immiscible Water-Alternating-CO Processes in Thin Heavy-Oil Reservoirs
Pressure Maintenance and Improving Oil Recovery by Means of Immiscible Water-Alternating-CO Processes in Thin Heavy-Oil Reservoirs
Summary meaning that only a small fraction of oil can be displaced and
Techniques have been developed to experimentally and numeri- recovered, especially when low-temperature oxidation is the dom-
cally evaluate performance of water-alternating-CO2 processes in inant reaction model (Islam et al. 1989; Moore et al. 1999; Shen
thin heavy-oil reservoirs for pressure maintenance and improving 2002). Among the thermal methods proposed for heavy-oil recov-
oil recovery. Experimentally, a 3D physical model consisting of ery, by far the most successful ones are based on steam injection;
three horizontal wells and five vertical wells is used to evaluate however, large water consumption is another major constraint of
the performance of water-alternating-CO2 processes. Two well implementing the steam-based processes in certain areas. Chemi-
configurations have been designed to examine their effects on cal flooding has been considered one of the promising enhanced-
heavy-oil recovery. The corresponding initial oil saturation, oil- oil-recovery (EOR) techniques in heavy-oil reservoirs (Thomas
production rate, water cut, oil recovery, and residual-oil-satura- et al. 2001). For example, polymer flooding has been successfully
tion (ROS) distribution are examined under various operating applied in the Pelican-Lake heavy-oil field (Strauss et al. 2010).
conditions. Subsequently, numerical simulation is performed to In addition to consumption of a large amount of chemical agents,
match the experimental measurements and optimize the operating such as alkali and surfactants, which must be transported and
parameters (e.g., slug size and water/CO2 ratio). The incremental handled safely in the field, several environmental problems may
oil recoveries of 12.4 and 8.9% through three water-alternating- result from chemicals that are injected into the formation (Stoll
CO2 cycles are experimentally achieved for the aforementioned et al. 2010). Conventionally chemical flooding techniques are sen-
two well configurations, respectively. The excellent agreement sitive to brine salinity, leading to high cost or infeasibility in
between the measured and simulated cumulative oil production many cases (Flaaten et al. 2009). High levels of adsorption within
indicates that the displacement mechanisms governing water- the porous medium lead to some chemical EOR techniques asso-
alternating-CO2 processes have been numerically simulated and ciated with surfactants being uneconomical (Fletcher et al. 2012).
matched. It has been shown that water-alternating-CO2 processes Regarding heavy-oil reservoirs discovered in western Canada,
implemented with horizontal wells can be optimized to signifi- although the shallow depth makes it extremely difficult to achieve
cantly improve performance of pressure maintenance and oil re- miscible CO2 displacement, CO2 has been found to be an effec-
covery in thin heavy-oil reservoirs. Although well configuration tive agent for recovering heavy-oil resources through an immisci-
imposes a dominant impact on oil recovery, the water-alternating- ble displacement process, and satisfactory results have been
gas (WAG) ratios of 0.75 and 1.00 are found to be the optimum achieved in heavy-oil reservoirs, such as the Bati Raman field and
values for Scenarios 1 and 2, respectively. Alaska North Slope fields (Issever et al. 1993; Sahin et al. 2008;
Ning et al. 2011). It is of practical and fundamental importance
to evaluate suitability of pressure maintenance and improving
Introduction heavy-oil recovery with immiscible CO2 injection in thin pay
With escalating oil consumption and depleting conventional oil zones in which other EOR techniques are not applicable.
resources, the vast heavy-oil resources discovered in western Can- Compared with natural gas and air, CO2 still has high solubil-
ada (i.e., Alberta and Saskatchewan) have attracted much more ity in heavy oil and bitumen, though it is difficult for CO2 to
attention recently. Approximately 62% of Canada’s total heavy- achieve miscibility with heavy oil. The mechanisms associated
oil resources are located in Saskatchewan, including 3.4 billion with oil recovery include viscosity reduction, swelling effect,
m3 of proven reserves (Bowers and Drummond 1997). As for interfacial-tension reduction, and blowdown recovery (Beecher
these resources, more than 90% of the proven reserves are con- and Parkhurst 1926; Jha 1986; Rojas and Farouq Ali 1988). The
tained in reservoirs with pay zones of less than 10 m, approxi- injected CO2 not only acts as a pressure-maintenance agent but
mately 55% of which have pay zones of less than 5 m (Huang also enhances oil recovery during the pressure-maintenance pro-
et al. 1987). It is a great challenge to recover such heavy-oil cess in heavy-oil reservoirs. In recent years, the WAG process has
resources economically and effectively because of the high vis- been found to be much more effective than gas injection alone,
cosity and fast depletion of reservoir pressure. Thermal recovery because it combines the improved microscopic displacement effi-
techniques, such as cyclic steam stimulation, steamflooding, ciency of the gasflooding with improved macroscopic sweep effi-
steam-assisted gravity drainage, and in-situ combustion, appear to ciency by injecting water (Christensen et al. 2001). Although
be effective for recovering heavy oil because the oil viscosity is numerous immiscible or miscible WAG processes with use of
very sensitive to temperature. However, characteristics of thin res- CO2 (CO2-WAG) have been implemented in light- or medium-oil
ervoirs are such that thermal-based processes are ineffective or reservoirs (Asghari et al. 2007; Righi and Pascual 2007; Shi et al.
uneconomical, mainly because of excessive heat losses to adja- 2008; Chen et al. 2010a, b), few attempts have been made to eval-
cent formations or aquifers. In-situ combustion processes may uate performance of CO2-WAG processes under the well configu-
yield more viscous or immobile components in the formation, rations associated with horizontal wells for pressure maintenance
in a heavy-oil reservoir.
In general, the CO2-WAG performance is significantly
Copyright V
C 2013 Society of Petroleum Engineers
affected by reservoir heterogeneity, rock wettability, fluid proper-
This paper (SPE 157719) was accepted for presentation at the SPE Heavy Oil Conference ties, miscibility conditions, trapped gas, injection techniques, and
Canada, Calgary, 12–14 June 2012, and revised for publication. Original manuscript
received for review 7 April 2012. Revised manuscript received for review 25 September
well operational parameters. Performance of a CO2-WAG process
2012. Paper peer approved 14 November 2012. is largely affected not only by the injection parameters, including
CO2
3D model
BPR Collector
initial oil saturations for Scenarios 1 and 2 are measured to be The porosity of the sandpacked model is measured by means
93.1 and 92.5%, respectively. of the imbibition method. After completion of the displacement
process, the blowdown process is carried out. Once a blowdown
Experimental Procedures recovery is completed, ROS is measured. From top to bottom, the
sandpack model is divided into four layers and marked as Layers
Before conducting the experiments, several preparations must be
1 through 4 in sequence, each of which is approximately 3.0 cm
completed, including sandpacking, porosity measurement, perme-
thick. In each layer, mixtures of sand and reservoir fluids are col-
ability determination, and initial-oil-saturation determination. As
lected from 13 locations to measure the ROS. First, the weight of
for the sand-packing process, the 3D displacement model is placed
each collected sample that contains sand, residual oil, and water is
horizontally before filling the void space with a given amount of
measured. Second, the collected sample is heated in an oven at
the Ottawa sand. Subsequently, the 3D model is shaken for a tight
100.0 C for 4 hours to ensure that water is completely evaporated.
packing with a pneumatic vibrator (NP 35, Northern Vibrator) for
Then the remaining sand/oil mixture is weighed. It should be
at least 10 hours. The lid is then removed, and more sand is added
noted again that the heavy-oil sample used in this study does not
to the formed void space in the model; the model is then covered,
contain any hydrocarbon components lighter than C9, of which
tightened, and shaken again. The same process is normally re-
the normal boiling point is 143.0 C. As such, hydrocarbon com-
peated four to five times until no void space is observed at the top
ponents will not be distilled when the heating temperature is set to
of the physical model. During the packing process, the bottom of
be 100.0 C for evaporating water. Third, toluene and kerosene are
the sandpack is packed more tightly than the top layer.
used as solvents to completely remove the residual heavy oil in
the sample. Finally, the sand is heated in the oven at 110.0 C and
then weighed when it is completely dried. Consequently, ROS is
calculated on the basis of the weight differences, porosity of the
sandpack model, oil density, and sand density. Detailed proce-
dures for the aforementioned preparations can be found elsewhere
(Zheng et al. 2011).
The sandpacked 3D physical model is first flooded with oil
until irreducible water saturation is reached. The physical model
is then left undisturbed for 24 hours to equilibrate the distribution
of fluids. For each scenario, water injection is started from the
(a) injector(s) with a constant injection rate of 2.0 cm3/min, which is
in accordance with the typical average pore velocity of 1.0 ft/D in
an oil field. Water injection is terminated once no more oil is
produced.
After the waterflooding process with a specific well configura-
tion, several cycles of water and CO2 are injected into the physi-
cal model alternatively to evaluate their performance of pressure
maintenance and improving oil recovery. During the CO2-WAG
processes, CO2 is introduced into the physical model from a CO2
cylinder, in which a gas regulator is used to control the injection
(b) pressure of 3900 kPa. A backpressure regulator (BPR) (EB1HP1,
Equilibar) is connected with the production port to maintain the
Fig. 2—Schematic of well configurations. (a) Scenario 1: one physical model at the prespecified operating pressure of 3860 kPa
horizontal injector and one horizontal producer; (b) Scenario 2: during WAG processes. Therefore, the pressure inside the 3D dis-
four vertical injectors and one horizontal producer. placement model has been maintained at approximately 3860 kPa
1.15
1
Viscosity, 103 cP
Swelling factor
1.10
0.1
1.05
0.01 1.00
0.0 0.2 0.4 0.6
Solubility of CO2 in heavy oil, mole fraction
Fig. 3—Calculated viscosity (Emera and Sarma 2008) and swelling factor (SF) (Li et al. 2011) of CO2-diluted heavy oil as a function
of CO2 solubility.
with the use of the gas regulator and BPR during the displacement Because it is not reliable to generate the EOS model for composi-
process. All the experiments are performed at the ambient temper- tional simulator without sufficient laboratory pressure/volume/
ature of 21.0 6 1.0 C. temperature data, as recommended by the CMG, genetic-algo-
Conical-bottom glass centrifuge tubes (Kimble) are used to rithm (GA) -based correlations (Emera and Sarma 2008) have
collect the produced fluids. The capacity of the tube is 100 cm3. been used to calculate CO2 solubility, swelling factor, and the vis-
There are several subdivisions, among which the largest toler- cosity of the oil/CO2 mixture in this study. The GA-based viscos-
ance is 1.00 cm3. The volume of produced liquids can be meas- ity correlation can reproduce 130 viscosity data for oil/CO2
ured accurately through the use of the tubes. A gas flowmeter mixtures, with an accuracy of 6.0%, whereas the swelling-factor
(XFM17S, Aalborg) is connected with the collector to measure correlation is able to reproduce 85 swelling-factor data for oil/
the produced gas. The gas-flow rates and gas cumulative volume CO2 mixtures, with an accuracy of 0.61% (Emera and Sarma
are logged and stored automatically in a computer at a preset time 2008). Fig. 3 shows the predicted viscosity and swelling factor of
interval. an oil/CO2 mixture, as a function of CO2 solubility through the
use of the GA-based correlations. A significant viscosity reduction
and swelling effect can be achieved with the dissolution of CO2 in
Numerical Simulation heavy oil (Beecher and Parkhurst 1926; Jha 1986). Fig. 3 also
To better understand the mechanisms of CO2-WAG processes shows the calculated swelling factor with the use of a newly
with the purpose of pressure maintenance in heavy-oil reservoirs, developed binary-interaction-parameter correlation in the vol-
numerical simulation has been conducted to history match the ume-translated PR EOS, which has been experimentally validated
waterflooding and CO2-WAG processes. As mentioned previ- with an oil sample collected from the same area (Li et al. 2011).
ously, no hydrocarbon components lighter than C9 are found in As seen in Fig. 3, there exists an excellent agreement between the
the dead heavy-oil sample in this study, whereas heavy compo- predicted swelling factor by use of the GA-based correlations and
nents account for a large proportion of the composition. The oper- the predicted swelling factor with the PR EOS model. This also
ating pressure (less than 4000 kPa) is far below the minimum confirms the reliability of the GA-based correlations used in this
miscibility pressure of 15.5 MPa at temperature of 21.0 C, which study. Because a black-oil model has also been considered to be
is calculated by the CMG WinProp module (Version 2009.11). an effective tool for simulating performance of immiscible flood-
Neither light-components extraction nor miscible flooding has ing processes (Cobanoglu 2001), the IMEX black-oil simulator
been achieved at the experimental conditions in this study. (Version 2009.11, CMG Limited, Canada) has been used in this
study. A grid system of 13 13 6 gridblocks is used in the nu-
merical model to represent the physical model, resulting in a fine-
Injector #1
gridblock size of 2.34 2.34 2.10 cm (see Fig. 4).
Injector #3 Uniform distribution of porosity and oil saturation is used in
Injector #2 10.6 the model; the absolute permeability of each layer varies with the
9.5 depth. Because of the relatively high absolute permeability, the
Injector #4 capillary pressure effect is neglected. Because the relative perme-
8.5
abilities cannot be directly measured from experiments, they are
7.4 to be determined by use of the history-matching techniques
6.4 instead (Watson et al. 1980; Eydinov et al. 2009; Li et al. 2009;
5.3 Li and Yang 2011). In principle, trapping of the nonwetting phase
4.2 (CO2) significantly affects the hysteresis in relative permeability
Producer (Spiteri and Juanes 2004). It is experimentally observed, however,
3.2
that not much CO2 is trapped in the sandpacked model during the
2.1 immiscible CO2-WAG process. It is reasonably assumed that hys-
1.1 teresis effect is not significant in this study, and thus the hysteresis
0.0 effect on relative permeability has not been taken into account. In
this study, two sets of similar relative permeability curves are
Fig. 4—3D view of the simulation model of Scenario 2. used to accurately match the experimental measurements, though
H2O
CO2
Water cut, %
CO2 H2O 60
CO2 H2O
20.0
Cycle 1 Cycle 2 Cycle 3
40
Scenario #1
Scenario #2
10.0
Scenario #1 20
Scenario #2
0.0 0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0
Injected fluids, PV Injected fluids, PV
Fig. 5—Oil recovery during displacement experiments for Sce- Fig. 6—Water cut during displacement experiments for Scenar-
narios 1 and 2. ios 1 and 2.
one set of the average relative permeability curves would provide water is able to mobilize the contacted oil that is to be displaced
a reasonable match. Regarding field applications, however, one to the producer along these paths (Mai and Kantzas 2009).
set of average relative permeability curves is used because these As shown in Fig. 5, oil recovery of the two scenarios stays
two sets of relative permeability curves are similar. close at the early stage. With a further increase of injected water
To simulate the experimental displacement process, the well PV, the well configuration associated with two horizontal wells
operational conditions defined in simulation models are set the (i.e., one horizontal injector and one horizontal producer, Scenario
same as those used in the experiments. For example, in the experi- 1) tends to yield a higher oil recovery compared with Scenario 2.
ments, the production port is connected with the BPR that is used For instance, oil recovery of Scenario 1 is approximately 1.3%
to maintain the pressure of model at the prespecified operating higher than that of Scenario 2 when 1.00 PV of water is injected.
pressure of 3860 kPa. In the simulation models, the bottomhole This can be attributed to the fact that a horizontal well is efficient
pressure of the producer is then set to be 3860 kPa. During a water- to achieve a better sweep efficiency.
injection process, the water-injection rate in the simulation model
is set at the constant value of 2.0 cm3/min, which is the same as the
CO2-WAG Processes. After waterflooding, three cycles of water
experimental water-injection rate. For the CO2-injection process, a
and CO2 injection are alternately implemented to examine per-
gas regulator connected with the CO2 cylinder is used to control
formance of pressure maintenance and improving oil recovery.
the injection pressure of 3900 kPa. Accordingly, the CO2-injection
During CO2-WAG processes, a BPR is used to maintain the pres-
pressure is set to be 3900 kPa in the simulation model.
sure of the physical model at 3860 kPa, which is higher than the
original reservoir pressure of 3500 kPa, whereas the CO2-injec-
Results and Discussion tion pressure of 3900 kPa is controlled through a gas regulator
Waterflooding. Waterflooding is started in Scenarios 1 and 2. connected with the CO2 cylinder. For Scenario 1, the same slug
The oil-recovery profiles for the two scenarios are shown in sizes (0.40 PV) of water and CO2 are used; that is, the WAG ratio
Fig. 5. After 1.00 pore volume (PV) of water injection, oil recov- is 1:1. The slug sizes of water and CO2 of Scenario 2 are 0.25 and
eries of 24.3 and 23.1% are achieved for Scenarios 1 and 2, 1.00 PV, respectively, resulting in a WAG ratio of 1:4.
respectively. A similar recovery profile for heavy-oil waterfloods Fig. 5 illustrates that the majority of the produced oil in the
can be found in the literature (Mai and Kantzas 2009). Fig. 6 CO2-WAG processes is recovered during the first cycle. After the
shows water cut in the produced liquid during displacement proc- second cycle, there is no significant increment in oil recovery.
esses of the two scenarios. The fact that the heavy-oil sample is Similar results are also documented in the literature (Sohrabi et al.
dramatically more viscous than the injected water leads to a high 2000). It is obvious that the oil recovery of Scenario 1 levels out
mobility ratio of water to heavy oil. In this way, viscous fingering during the third cycle, whereas an additional 2.0% in oil recovery
is severe during waterflooding processes, resulting in a very early is achieved during the third cycle for Scenario 2. As depicted in
water breakthrough and low oil recovery. Fig. 5, for these CO2-WAG cycles, CO2-injection processes
For light- and medium-oil waterfloods, as water is injected, oil recover a small amount of heavy oil, whereas more oil is pro-
is continuously produced until breakthrough, and a relatively high duced during the subsequent water-injection processes. Although
recovery factor can be achieved before breakthrough (Hadia et al. there is not much heavy oil produced during the water-injection
2007). As shown in Fig. 6, however, it is obvious that break- phase, the injected CO2 still contributes to oil recovery. The
through occurs very early in a heavy-oil waterflooding process, injected CO2 dissolves into heavy oil, leading to a swelling effect
and water cut increases sharply to high values at the early stage. and oil-viscosity reduction. In addition, because gasflooding leads
For these two scenarios, water cut approximates 60% at 0.15 PV to higher microscopic displacement efficiency than that of water-
of injected water, whereas water cut for Scenarios 1 and 2 is flooding, the injected CO2 changes the distribution of liquids in
measured as 95.6 and 97.5%, respectively, at 1.00 PV of injected the pores and mobilizes much more oil that is not previously con-
water. Although more heavy oil is produced after 0.20 PV of tacted by water. Subsequently, the injected water displaces the
water injection, water cut is high. This means that the injected mobilized oil to the producer. In this study, Scenario 1, with a
water may still keep contacting the in-place oil and displacing the larger water-slug size, achieves a higher oil recovery than Sce-
oil to the producer. The difference of water cut is minor, whereas nario 2, with a smaller water-slug size, when 3.0 PV of fluids is
the high water cuts indicate that much water is produced, and injected. Oil recovery of Scenario 1 reaches the plateau in a short
waterflooding is found to be inefficient at the late stage of water period, demonstrating that it reaches the production limit quickly.
injection. Figs. 5 and 6 demonstrate that considerable heavy oil is In Scenario 2, oil is still produced in the third cycle, indicating
still produced after water breakthrough. This is ascribed to the that production can be sustained for a much longer time. The bet-
fact that water flows through the continuous low-resistance path- ter performance of Scenario 1 is partially attributed to the fact
ways that are generated by water after breakthrough. The flowing that a significantly larger drainage volume can be controlled by a
60
40
20
0
0.0 1.0 2.0 3.0 4.0 5.0
Blowdown recovery, %OOIP
Blowdown Recovery. After three cycles of water and CO2 are Fig. 8—Distribution of ROS in the top layer for (a) Scenario 1
injected into the 3D physical model under each well configura- and (b) Scenario 2.
tion, pressure inside the model is decreased to atmospheric pres-
sure for blowdown recovery. As shown in Fig. 7, the blowdown ROS in the top layer can be observed directly. Digital images of
process contributes an increment of 2.8 and 4.1% in oil recovery the top layer for the two scenarios are shown in Fig. 8. In Scenario
for Scenarios 1 and 2, respectively. The blowdown recovery is 1 (see Fig. 8a), it is obvious that the color of sand along the hori-
mainly attributed to the solution-gas-drive mechanism (Jha 1986). zontal producer (Well 2) is light, demonstrating that heavy oil
CO2 gas has a good ability to dissolve into oil. During the blow- along the horizontal producer has been displaced effectively. It
down process, with pressure decline, energy stored by CO2 disso- also can be seen that the color of the bottom-left portion for Sce-
lution is then released and continues to drive the oil to the nario 1 is much darker than that of the top-right portion, indicat-
wellbore (Mangalsingh and Jagai 1996). Besides the expanded ing that a stable displacement front along the horizontal injector
gas exerting a displacement force to drive oil to producers, such a was difficult to establish. A large portion of injected fluids invades
high efficiency of foamy-oil flow under solution-gas drive in vis- the sandpack through the heel of the horizontal injector and then
cous oil is mainly the result of (1) a significant reduction in oil displaces the heavy oil to the producer, resulting in a well-
viscosity, because small gas bubbles are formed in the oil, and (2) displaced triangle area (i.e., top-right portion). This observation
simultaneous flow of continuous oil phase and discontinuous gas implies that it is possible to improve the sweep efficiency by
phase in the form of tiny bubbles (Smith 1988). changing the position of heels and toes of these two parallel hori-
It is experimentally found that little fluid is produced when zontal wells. In Scenario 2 (see Fig. 8b), the color of sand for the
pressure is decreased from 3800 kPa to 2000 kPa, whereas most top layer is very dark, indicating that much heavy oil is still at the
of the fluids are produced at pressure less than 1500 kPa. The top layer and the heavy oil at the top layer cannot be swept effi-
blowdown process is found to last for a relatively long period ciently, although approximately 5.00 PV of fluids has been
(i.e., more than 2 hours) in this study. Approximately 40 L of CO2 injected. Such observed dark-sand distribution in the 3D model
(ambient temperature and atmospheric pressure) is produced dur- corresponds to its low oil recovery with the displacement proc-
ing the blowdown process, whereas gas is still producing even esses previously applied.
though the pressure of the physical model approximates to the To examine the ROS in the physical model, ROS is measured
atmospheric pressure. This implies that CO2 storage also occurs in this study. Distribution of ROS in the sandpack for Scenarios 1
as the injected CO2 dissolves into reservoir fluids and occupies and 2 is illustrated in Figs. 9 and 10, respectively. As mentioned
the pore spaces that were previously filled by reservoir fluids. previously, from top to bottom, the sandpack is divided into four
layers, marked as Layers 1 through 4 in sequence. For each layer,
Distribution of ROS. After the blowdown process, the steel lid 13 samples were collected and measured from different locations
on the top of the 3D model is removed and the distribution of that are marked by the circle symbols in Figs. 9 and 10. The ROS
Y, cm
0.0 0.0
–5.0 –5.0
–10.0 –10.0
–10.0 –5.0 0.0 5.0 10.0 –10.0 –5.0 0.0 5.0 10.0
X, cm X, cm
(a) (b)
0.20 0.20
10.0 0.25 10.0 0.25
0.30 0.30
0.35 0.35
0.40 0.40
0.45 0.45
5.0 0.50 5.0 0.50
0.55 0.55
0.60 0.60
0.65 0.65
Y, cm
Y, cm
0.0 0.0
–5.0 –5.0
–10.0 –10.0
–10.0 –5.0 0.0 5.0 10.0 –10.0 –5.0 0.0 5.0 10.0
X, cm X, cm
(c) (d)
Fig. 9—Distribution of ROS for Scenario 1 in (a) Layer 1, (b) Layer 2, (c) Layer 3, and (d) Layer 4.
contour of each layer is then generated on the basis of these meas- tem and a gas/liquid system are determined by history matching
ured values. As shown in Fig. 9, for Layer 1, the bottom-left por- the experimental measurements (Watson et al. 1980; Spiteri and
tion shows a higher ROS, which is consistent with the dark color Juanes 2004; Eydinov et al. 2009; Li et al. 2009, 2011). More spe-
of the bottom-left portion in the digital image for Scenario 1 (see cifically, the initial inputs of the three-phase relative permeability
Fig. 8a). For Layer 2 of Scenario 1, the ROS along the horizontal curves are generated with the commonly used correlation (i.e.,
producer is low, indicating heavy oil along the producer has been Stone II correlation). Subsequently, the relative permeability
efficiently displaced. That is attributed to the fact that the horizon- curves are tuned and determined once the experimented results
tal producer controls a large reservoir area and has high sweep effi- have been history matched. Such determined relative permeability
ciency. The average ROSs of Layers 3 and 4 for Scenario 1 are curves for two scenarios are shown in Figs. 11 through 14. As
relatively low (See Fig. 9). This is ascribed to the fact that two par- can be seen from Figs. 11 and 13, water relative permeability
allel horizontal wells are at the bottom of the physical model, lead- increases slightly when the water saturation is less than 0.3. The
ing to high sweep efficiency in the bottom layers of the model. relative permeability values are low at a high water-saturation
As shown in Fig. 10 for Scenario 2, Layer 1 shows that the range. The performance of waterflooding is very sensitive to the
ROS in the top layer is high, which is consistent with the dark relative permeability values at a low water saturation range (less
color as seen in Fig. 8b, whereas the average ROS decreases from than 0.2). For gas/liquid relative permeability curves (see Figs. 12
top to bottom layers. Layer 4 illustrates low ROS; a clear dis- and 14), oil relative permeability values decrease sharply. This
placement front from four vertical injectors to the horizontal pro- implies that gas fingering is severe because of the high mobility
ducer is formed because the horizontal producer is at the bottom ratio of gas to heavy oil, resulting in an early breakthrough. The
of the model in Scenario 2. Therefore, the horizontal producer at relative permeability of oil reduces significantly as gas saturation
the bottom layer is beneficial for recovering the bottom heavy oil increases.
in reservoirs with gas-cap drive. The oil viscosity also affects the heavy-oil/water relative per-
meability curves. With an increase in oil viscosity, ROS is
increased and relative permeabilities are decreased at a higher
Numerical Simulation and Performance Optimization. water-saturation range (Wang et al. 2006; Nejad et al. 2011). Nu-
Because no direct interpretation methods have been made avail- merical simulation of the displacement processes is conducted
able to determine relative permeability for such displacement with injection rate, injection pressure, and production pressure as
experiments, the relative permeability curves for a water/oil sys- input constraints. The cumulative oil production and water cut are
Y, cm
0.0 0.0
–5.0 –5.0
–10.0 –10.0
–10.0 –5.0 0.0 5.0 10.0 –10.0 –5.0 0.0 5.0 10.0
X, cm X, cm
(a) (b)
0.20 0.20
0.25 0.25
10.0 0.30 10.0 0.30
0.35 0.35
0.40 0.40
0.45 0.45
0.50 0.50
5.0 0.55 5.0 0.55
0.60 0.60
0.65 0.65
Y, cm
Y, cm
0.0 0.0
–5.0 –5.0
–10.0 –10.0
Fig. 10—Distribution of ROS for Scenario 2 in (a) Layer 1, (b) Layer 2, (c) Layer 3, and (d) Layer 4.
compared with the experimental data in Figs. 15 and 16. The indicating that the waterflood becomes less efficient afterward.
good matching results show that the numerical simulation cap- After CO2 injection, however, a large amount of oil is still pro-
tured the overall behavior of both the waterflooding and CO2- duced during the subsequent water-injection process. This is
WAG processes. As can be seen in Figs. 15 and 16, oil recovery attributed to the fact that the injected CO2 with high microscopic
approaches the plateau after 1.00-PV injection (2,000 minutes), sweep efficiency can swell the heavy oil, reduce oil viscosity, and
1.0
1.0 Krg
Krog
Krw
Krow 0.8
0.8
0.6
0.6
Kr
Kr
0.4
0.4
0.2
0.2
0.0
0.0 0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0 Sg
Sw
Fig. 12—Liquid and gas relative permeability curves for Sce-
Fig. 11—Water and oil relative permeability curves for Scenario 1. nario 1.
0.6 0.6
Kr
Kr
0.4 0.4
0.2 0.2
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Sw Sg
Fig. 13—Water and oil relative permeability curves for Scenario 2. Fig. 14—Liquid and gas relative permeability curves for Sce-
nario 2.
1600 100
1400
80
Cumulative oil production, cm3
1200
1000
Water cut, %
60
800
40
600
Measured water cut
400 Simulated water cut
Measured oil production 20
Simulated oil production
200
0 0
0 1000 2000 3000 4000 5000
Time, min
Fig. 15—Measured and simulated cumulative oil production and water cut for Scenario 1.
1600 100
1400
80
Cumulative oil production, cm3
1200
1000
Water cut, %
60
800
40
600
Measured water cut
Simulated water cut
400
Measured oil production 20
Simulated oil production
200
0 0
0 1000 2000 3000 4000 5000
Time, min
Fig. 16—Measured and simulated cumulative oil production and water cut for Scenario 2.
1400
Cumulative WOR
8
1000
800 6
600 WAG ratio: 2.00; Injected PV: 3.6
WAG ratio: 1.00; Injected PV: 3.6 4 WAG ratio: 2.00; Injected PV: 3.6
400 WAG ratio: 0.75; Injected PV: 3.5 WAG ratio: 1.00; Injected PV: 3.6
WAG ratio: 0.50; Injected PV: 3.6 WAG ratio: 0.75; Injected PV: 3.5
WAG ratio: 0.25; Injected PV: 4.0 2 WAG ratio: 0.50; Injected PV: 3.6
200 Continuous Waterflooding; Injected PV: 2.0 WAG ratio: 0.25; Injected PV: 4.0
Experiment
0 0
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
Time, min Time, min
Fig. 17—Production profiles with different water-slug sizes for Fig. 18—Cumulative WOR with different water-slug sizes for
Scenario 1. Scenario 1.
mobilize the heavy oil in the pores, resulting in a higher oil recov- ratio of 0.5. As such, WAG ratio of 0.75 leads to a relatively higher
ery with the subsequent water injection that displaces the mobi- oil recovery and a lower cumulative WOR. For Scenario 2, the
lized heavy oil. Therefore, the CO2-WAG process is a promising slug size of CO2 is set to be 1.00 PV for all cases (see Figs. 19 and
approach to maintain pressure and improve oil recovery in heavy- 20). The production profiles for different WAG ratios follow a
oil reservoirs because of the positive and composite effects of oil similar trend, whereas a higher water-slug size yields a higher oil
swelling and viscosity reduction by CO2 dissolution, high micro- recovery. The water-slug sizes of 1.00 PV and 2.00 PV (i.e., WAG
scopic sweep efficiency of CO2, and good macroscopic sweep ef- ratios of 1.00 and 2.00, respectively) are sufficient to control the
ficiency of water. mobility of CO2 to achieve high oil recovery. However, the WAG
As mentioned previously, oil is mainly produced during the ratio of 1.0 can keep a lower cumulative WOR than that of WAG
water-injection process for each CO2-WAG cycle. In this study, ratio of 2.00 when the same volume of fluids is injected (6.00 PV).
effects of water-slug size are examined by means of the history- Therefore, the WAG ratios of 0.75 and 1.00 are determined to be
matched reservoir models for Scenarios 1 and 2, respectively. In the optimum values for Scenarios 1 and 2, respectively.
Scenario 1, the slug size of CO2 is kept to 0.40 PV for each cycle, It is found from both the experimental measurements and nu-
whereas the slug size of water is changed from 0.10 to 0.80 PV merical simulation that well configuration imposes a dominant
(i.e., WAG ratio varies from 0.25 to 2.00). Although the CO2 slug impact on oil recovery. The well configuration associated with
size of Scenario 2 (1.00 PV) is larger than that of Scenario 1 (0.40 two horizontal wells tends to yield a higher oil recovery under the
PV), oil recovery of Scenario 2 is still low. For convenient com- same WAG ratio. Regarding each well configuration, WAG ratio
parison, continuous waterflooding is used as a reference case. As also affects oil recovery and cumulative WOR, although there
shown in Fig. 17, compared with other schemes, the scheme with exists an optimum value for each scenario.
a WAG ratio of 0.25 produces less oil, though 4.00 PV of fluids
has been injected. This is attributed to the fact that compared with
the CO2-slug size, the water-slug size is too small to effectively Conclusions
control the CO2 mobility in heavy-oil reservoirs. Regarding the
same CO2 slug size, a larger water-slug size yields a higher oil re- Techniques have been developed to experimentally and numeri-
covery. As shown in Fig. 18, if the water-slug size is very large, cally evaluate the performance of CO2-WAG processes under
the extra amount of water injected leads to trapping of oil and well configurations associated with horizontal wells in thin heavy-
high cumulative water/oil ratios (WORs) (Rojas and Farouq Ali oil reservoirs for pressure maintenance and improving oil recov-
1986). It should be noted that oil recovery increases when the ery. Experimentally, a 3D physical model is used to examine the
WAG ratio is increased from 0.50 to 2.00, whereas the cumulative suitability of pressure maintenance and improving oil recovery by
WOR at WAG ratio of 0.75 remains close to the value at WAG means of an immiscible CO2-WAG process in heavy-oil
12
1600
1400 10
Cumulative oil production, cm3
1200
Cumulative WOR
8
1000
800 6
0 0
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
Time, min Time, min
Fig. 19—Production profiles with different water-slug sizes for Fig. 20—Cumulative WOR with different water-slug sizes for
Scenario 2. Scenario 2.