SPE 124047

Laboratory and Field Scale Analysis of Steam Over Solvent Injection in

Fractured Reservoirs (SOS-FR) for Heavy-Oil Recovery
Al Muatasim Al Bahlani and Tayfun Babadagli, University of Alberta

Copyright 2009, Society of Petroleum Engineers

This paper was prepared for presentation at the 2009 SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 4–7 October 2009.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Tapping heavy-oil from fractured carbonates is a real challenge due to unfavorable rock and reservoir characteristics. We
introduced a new technique called Steam-Over-Solvent in Fractured Reservoirs (SOS-FR) for efficient heavy-oil recovery
from fractured reservoirs, more specifically carbonates. The process consists of cyclical injection of steam and solvent in the
following manner: Phase-1: Steam injection to heat up the matrix and recover oil mainly by thermal expansion, Phase-2:
Solvent injection to produce matrix oil through diffusion-imbibition-drainage processes, and Phase-3: Steam injection to
retrieve injected solvent and recover more heavy-oil. Our preliminary experiments under static (SPE 117626) and dynamic
(SPE 123568) conditions showed that, under very unfavorable conditions (oil-wet carbonate, ~4,500 cp crude), oil recovery at
the end of Phase-3 could be as high as 85-90% OOIP with 80-85% solvent retrieval.
This paper presents numerical modeling of the dynamic experiments and an upscaling study for reservoir size matrix.
Heptane was selected as the solvent to inject through a single-matrix/single-fracture oil-wet Berea sandstones saturated with
~4,500 cp oil. The experimental results were matched to a single matrix/single fracture numerical model and parameters
needed for larger scale simulation (matrix-fracture interaction parameters such as thermal diffusion, solvent diffusion and
dispersion coefficients) were obtained. The main focus was the matrix size and first an up-scaling study to field conditions
was performed.
Specific observations and conclusions as to the applicability of this technique in the field effectively were reported. It is
hoped this new technique will be an alternative for tapping heavy matrix oil from oil-wet, fractured, deep, carbonate fields.

Introduction
Finding an efficient way to produce heavy oil / bitumen production from tight naturally fractured carbonate reservoirs is one of
the most challenging tasks the petroleum industry. Due to geological constraints, i.e., discontinuous structure of reservoir
caused by fractures, steam injection can only be applied to heat matrix rather than conventional steam displacement (Briggs et
al., 1988, 1992; Reis, 1990; Babadagli, 1996a-b, 2002). Therefore, steam injection applications in fractured reservoirs are not
abundant and have been limited to a few pilot applications (Sahuquet and Ferrier, 1982; Nakamura et al., 1995; Snell and
Close, 1999; Macaulay et al., 1995; Al-Shizawi et al., 1997; Babadagli, et al., 2008). The common observation in all these
pilot attempts was the inefficiency of the process that was caused by the rapid movement of steam resulting in early
breakthrough without effectively heating the matrix.
Out of these pilot tests, only the Qarn Alam project has been switched to a full field scale application (Rawnsley et al.,
2005; Penney et al., 2005; Shahin et al., 2006). Babadagli and Bemani (2007) reported that the matrix drainage at relatively
favorable laboratory conditions do not exceed 42% for challenging Qarn Alam cores containing ~3,000 cp dead oil. It is
expected that this number will be reduced to lower values at field conditions. Shahin et al (2006) reported that the recovery
only increases to 26% by steam injection through the process called thermally assisted gas oil gravity drainage (TAGOGD),
which was expected to be only 4% without thermal assistance. Note that the drainage process is rather slow as pointed out by
Macaulay (1995).
With huge amounts of oil locked in fracture reservoirs, the need for a novel approach for efficient recovery is eminent.
Very recently, Al Bahlani and Babadagli (2008; 2009) proposed a new technique called steam-over-solvent injection in
fractured reservoirs (SOS-FR). They showed that starting with hot-water injection followed by solvent injection and
completing the process with how-water injection, up to 85-90% OOIP recovery for static conditions and 70% recovery under
dynamic conditions with 80-85% solvent retrieval can be achieved. In this paper, we present numerical modeling of the
dynamic experiments and an upscaling exercise for field scale matrix conditions
2 SPE 124047

Experimental Procedure
Three symmetrically fractured sandstone cores (cylindrical core with an induced fracture in the middle) were used to conduct
the experiment. Their wettability was altered from water to oil wet using a siliconizing wettability alteration agent. This agent
is a short chain, clear polymeric silicone fluid consisting primarily of dichlorooctamethyltetrasiloxane (Surfasil™). They were
then saturated with ~4,500 cp dead crude oil under vacuum at high temperature and left for a few weeks in oil for aging. The
core was placed into a core holder with an overburden pressure of 180 psi. Hot water was injected at ambient pressure and
fluids were collected and measured at the production end. The process consists of three phases as described before (Al
Bahlani and Babadagli, 2008; Al Bahlani and Babadagli, 2009). It starts with hot water injection (Phase 1) continued with
solvent injection (Phase 2). Finally, how water was injected (Phase 3) to retrieve the solvent injected and recover some more
extra oil. After Phase 1 was completed; the core was left to cool for a certain period until it reaches room temperature. Solvent
(Heptane) was then injected at ambient pressure. Three experiments were run for three different solvent injection rates (0.1,
0.3, and 0.5 cc/min). The hot water injection experiments were conducted at a constant rate of 2.0 cc/min and at 90 oC. The
experimental set-up is demonstrated in Figure 1 (Al Bahlani and Babadagli, 2009). Details of the experimental work can be
found in the same reference.

Numerical modelling of the experiments

Model description
The simulator used to run the experiments is a semi-compositional -commercial- simulator (STARS of CMG). The PVT
values of the solute (heavy oil) characteristics were calculated using PVTi package of another commercial simulator
(ECLIPSE). First contact miscible tests were also performed for the solvent (heptane) and the solute (heavy crude oil).
The CMG-STARS model is a single porosity cartesian model (Figure 2). The model length was fixed to the exact core
length and sensitivity runs were applied on the width and thickness to approximate the original amount of initial oil in place in
the cylindrical experimental sample. Two additional grid blocks at the inlet and outlet were included to act as a buffer zone
and mimic heat losses at the experimental connection points. A fracture was assigned in the middle of the model and the
perforations were created to allow the injected fluid flow through the fracture. This representation corresponds to the fracture
blocks creating a source / sink well model. The units used in the model were lab units. The IJK blocks for the 3D model were
16x21x12 with the dimensions of 9.62 x 2.248 x 2.238 cm representing the exact volume of the experimental system. The
matrix properties used in the simulation study are given in Table 1.
The main objective of this numerical simulation is to history match the experiments and obtain immeasurable parameters
for the upscaling exercises.

History matching
A 2D model was first initiated to study the factors affecting the controlling the process at all three phases. For Phase 1 (hot
water injection), the factors affecting the process are mainly (a) the first coefficient of thermal expansion (CT1), (b) injection
temperature, (c) relative permeability curves, and (d) heat loss parameters. Impacts of other parameters such as rock heat
capacity, and fluids thermal conductivities (oil, gas and water) were less pronounced. Therefore, the values from a producing
heavy oil field with the similar properties were used. For Phase 2 (solvent injection), the main parameters affecting the
process are (a) matrix molecular diffusion, and (b) matrix-fracture mechanical dispersion. For Phase 3, the only tuning
parameter left was the injection temperature as others were already set in the previous phases.
A 3D model was then built and a grid sensitivity analysis was made first. Fine grids seemed to have no major effect on the
process and the ultimate grid size was obtained to optimize the run time. The thermal oil expansion coefficient was estimated
from the previous static experiments (Al Bahlani and Babadagli, 2008). During the initial phase of the static experiment oil
was produced mainly from thermal expansion. The coefficient was obtained from the following equation:
ΔV 1
c= (1)
ΔT Vi
where ΔV is the difference in volume, Vi initial volume, and ΔT is the difference in temperature. After employing the
thermal expansion coefficient obtained from the experimental work and applying different sensitivity runs, the optimal number
was specified and used for all other models. A typical oil wet relative permeability table was created and a sensitivity analysis
on the relative permeability parameters was performed. The most critical uncertainty was in the injection temperature as the
temperature control due to heat loss during experiments was not stable. The temperature hardly stabilized at later stages in the
experiment and to achieve this, a heating tape was used. Thus, injection temperature was the last tuning parameter employed
and a single injection temperature for each experimental phase was specified.
In order to initiate Phase 2 (solvent injection) the model temperature was desired to be at ambient conditions. To model
that, a new model with the same dimensions and perforations were built. The saturation profiles (water saturation and oil
saturation) from the last time step were then imported from the previous model. Subsequently, solvent injection started at the
desired rates (0.1, 0.3, 0.5 cc/min) with three different models each with matched saturation profile from Phase 1. The
dominating process in the fracture was assumed to be dispersion based on previous experience (Trivedi and Babadagli, 2008,
SPE 124047 3

2009; Er and Babadagli, 2009). Hence, there was no molecular diffusion coefficient parameter employed for the fracture. The
dispersion coefficient was assigned for both matrix and fracture, bearing in mind that mechanical dispersion is modelled
explicitly in this model, unlike total dispersion which models implicitly both molecular diffusion and mechanical dispersion
coefficient lumped in total dispersion coefficient. The matching parameter for Phase 2 was the recovery factor. For Phase 3,
hot water was injected at the same rate in the experiment (2 cc/min) and the match parameters that yielded the best match were
used in the upscaling study.

Upscaling process
Scale dependency of the molecular diffusion and dispersion coefficients during miscible displacement and the upscaling of this
process have been critically discussed previously, especially in terms of heterogeneity (Arya et al. 1988). To define the
upscaling parameters, the production from Phase 3 of the models were matched with the lab experiments. One model gave the
closest match and its saturation profile was used for upscaling purposes. The upscaling methodology included enlarging the
grid size by three factors; x2 x5 and x10 of the lab scale size with a constant block density. The matching parameter was the
recovery factor over a certain time. The first upscaled parameter was the solvent injection rate. Since the fracture spacing (or
matrix size) was also doubled (x2), it was assumed that fluid velocity in both experimental and larger scales should be the
same. Thus, the rate was estimated by the following equation

Q 0 .1 cm3 / min
v= = = 4.47cm / min
Wb 2.238 × 0.01 cm ⋅ cm

Q = 4.47Wb cc/min

Note that this was only a rough initial estimate and the actual upscaling method will be introduced later.
The solvent and solute diffusion coefficients were considered to be the same in all cases but the dispersion coefficients
were the tuning parameters used to obtain the recovery factor match at a certain time.
After all three models gave a satisfactory recovery factor matches at a specified time, the relationship between the three
models was identified and tuning parameters were introduced.

Results and discussion

The SOS-FR process is a complicated process. The first phase (steam/hot-water injection) is commonly tested in labs.
However, the miscible displacement in fractured porous media filled with heavy-crude and some amount of immiscible fluid
has not been researched in detail. Its importance captured the attention of many researchers in the fields of CO2 sequestration
in waterflooded oil reservoirs, and in Expanding Solvent SAGD (ES-SAGD). Yet, the physics of the process was mainly
studied implicitly through total dispersion coefficient where diffusion coefficients were to be modelled explicitly. Reasons for
that are mainly due to tortuosity introduced in the matrix and its effect on the diffusion process. The effect of diffusion on
dispersion is likely to appear at low rates (laminar flow).
To test the effect of diffusion and dispersion coefficients on model, three runs were conducted: (a) both diffusion and
dispersion coefficients are present, (b) without dispersion coefficient, and (c) without diffusion coefficient. The results are
shown in Figure 3. By conducting these runs, we can estimate what controls the recovery process using the output curve
shape. When the diffusion coefficient is eliminated immediate increase in recovery is seen, and when dispersion coefficient is
eliminated a long time with constant low recovery is obtained as seen in Figure 3. This helped in tuning the model in history
match. It was also useful in explaining experimental observation.
Figures 4 through 6 show the history matches for Phase 1. After obtaining satisfactory history match, Phase 2 was then
initiated with different temperature (ambient). In order to reduce the run time to achieve ambient temperatures, new models
were created with saturation profiles imported from the last run of Phase 1. Figure 7 shows the saturation distribution after
Phase 1 for the core used for 0.1 cc/min solvent injection rate. By tuning both dispersion and diffusion coefficient parameters
a match for the recovery factor was achieved for all three rates. The results are shown in Figures 8 through 10.
As can be seen from Figure 11 that the solvent injection process undergoes three phases: (a) the first phase where a small
but similar amounts for all rates were observed, (b) the second phase where a constant rate recovery was achieved, and (c) the
third phase where very fast recovery was noticed. To explain this behaviour we refer back to Figure 3 where we can see the
effect of dispersion and diffusion on the process. For the first phase, oil seems to be produced mainly from the residual oil in
the fracture which is a very small amount. In the second phase, the dominant recovery mechanism is through diffusion into the
matrix which is a very slow process as indicated by a “plateau like” recovery curve. As solvent diffusion into the matrix is
completed and the stable volume (or density of mixture) is reached, the dispersion in the fracture starts dominating the
recovery process represented by thirds phase in Figure 11. The diffusion and dispersion coefficients obtained from history
matching Phase 2 are presented in Table 2.
Finally, to simulate Phase 3, we selected the semi-compositional model option for the sake of simplicity. It has certain
drawbacks as this type of model (not fully compositional) does not represent bi-products of the process, mainly asphaltene
4 SPE 124047

precipitation. During the lab experiments, gas (heptane) was produced at very high rates initially during Phase 3 (Al Bahlani
and Babadagli, 2008). What we noticed in the very beginning of Phase 3 is a sharp oil peak (first few minutes of Phase 3) as
shown in Figure 12. This sharp peak is highly likely the amount of gas (heptane) observed during lab experiments. Note that
the temperature (90 oC) is very close to the boiling point of heptane and when Phase 3 is started, heptane in the matrix starts
boiling and produced as gas bubbles. The produced amount during this stage is in good agreement with the experimental
observation. The rest of the recovery profile shows also a good match with the experimental results.
For the upscaling, three upsized grid blocks of the initial matrix size were tested (x2, x5, and x10). The initial guesstimate
of the oil rate was close to some extent, however, with tuning, it was found that there is an exponential relation between the
rate upscaling methodology and rates obtained from tuning. Thus, a power relation was obtained as it can be seen from Figure
13. The results were employed to obtain an upscaling equation for the oil rate with respect to fracture dimensions.
Figure 14 gives a good estimation of weather the upscaling parameters in hand are applicable or not. If the oil recovery
reaches the same recovery factor of what is reached in the lab then parameters in hand are applicable. Thus, upscaled model
can be tuned to design efficient solvent rate injection, well placement or any desired injection strategy. It is also apparent from
Figure 14 can be seen that the three phases seen in both lab experiments and simulation are apparent as explained earlier (oil
fracture depletion, diffusion dominated phase and dispersion dominated phase). The diffusion and dispersion coefficients for
tuned upscaled model are presented in Table 3 and the equations used to initially estimate them are presented in Table 4.
After Phase 2 in the upscaled matrices, hot water was injected again. As it can be noticed again, the recovery is inversely
proportional to the grid block size. This is because the recovery in this case is mainly due to thermal means and higher energy
injection rates are required as it can be seen from Figure 15. A peak is also noticed at the solvent production rate at this phase
as it can be seen from Figure 16. This, to some extent, strengthened the hypotheses that as solvent is heated and escapes the
matrix (gas phase in case of lab experiment) it pushes a good amount of oil with it causing a high initial recovery factor. The
solvent recovery factor is calculated and results are presented in Table 5. It can be clearly seen that solvent recovery factor is
quite high at very low time ranging between 51 and 70% OOIP (the smaller the matrix block size, the higher the solvent
retrieval). The decrease with enlarging grid size maybe explained in terms of rock properties, since heating a larger matrix
block takes larger energy which delays the solvent production from inner matrix. It should be emphasized that simulator
package in hand is a semi-compositional and results may vary when fully compositional simulator is used.

Conclusions
• The process of SOS-FR introduced in two previous papers (Al Bahlani and Babadagli, 2008 and 2009) with
experimental evidence was highly promising for heavy-oil/bitumen recovery from fractured reservoirs. The
numerical model study presented in this paper also showed that it is applicable at larger matrix sizes especially
when the time scale to reach the ultimate recovery was considered.
• A new methodology was proposed for upscaling lab data to field data by employing a dimensionless equation,
rate and dispersion coefficient provided in Table 4. These equations were obtained by employing fracture
dimensions of both lab experiments and upscaled model, and assuming the fact that the velocity at all scales are
the same.
• A relationship between mechanical dispersion and molecular diffusion at low rates was introduced and studied to
understand and maintain reasonable tuning for the lab experiment history match.
• During solvent injection, three phases were observed; (1) fracture oil depletion phase, (2) diffusion dominant
phase, and (3) dispersion dominant phase. This indicates that diffusion and dispersion process in highly water
saturated porous media is more distinctive than diffusion-dispersion process in oil saturated porous media.
• Large block sizes need larger solvent injection rate because of fluid velocity dispersion dependency.
• High and fast initial solvent rate is observed during the third phase, this coincides with the high gas bubbling seen
during phase three in lab experiments (Figure 16). This strongly suggests that gas retrieved is mainly heptane.
• Even for large matrices, solvent retrieval seems to be very high ranging from (50 – 70 %).
• The next step towards understanding this process is to up-scale to field scale and test the efficiency of the process
(Al Bahlani and Babadagli 2009 – IPTC).

Acknowledgements
This research was partly funded by an NSERC Grant (No: G121210595). The funds for the equipment used in the
experiments were obtained from the Canadian Foundation for Innovation (CFI) (Project # 7566) and the University of Alberta.
We gratefully acknowledge these supports. We would like to thank Dr. Suliman Al Hinai (Petroleum Development of Oman)
for his support and constructive comments on the numerical modelling of the process. The first author (MB) is also thankful
to Petroleum Development of Oman Co. for the financial support for this research as part of his PhD scholarship. Finally, we
would like to extend our gratitude to Computer Modeling Group (CMG) and Schlumberger (ECLIPSE) for providing the
simulation software package for research purposes.
SPE 124047 5

References
Al Bahlani A. and Babadagli T.: “Heavy-Oil Recovery in Naturally Fractured Reservoirs with Varying Wettability by Steam Solvent Co-
Injection”, SPE 117626, 2008 SPE/PS/CHOA International Thermal Operations and Heavy Oil Symposium, Calgary Canada, Oct.
2008.
Al Bahlani A. and Babadagli T.: “Steam-Over-Solvent Injection in Fractured Reservoirs (SOS-FR) for Heavy-Oil Recovery: Experimental
Analysis of the Mechanism”, SPE 123568, 2009 SPE Asia Pacific Oil and Gas Conference and Exhibition held in Jakarta, Indonesia,
4–6 August 2009.
Al-Shizawi, A., Denby, P. G. and Marsden, G.: “Heat-Front Monitoring in the Qarn Alam Thermal GOGD Pilot,” paper SPE 37781
presented at the 1997 SPE Middle East Oil Show, Bahrain, March 15-18.
Arya, A., Hewett, TA, Larson, RG and Lake, LW, 1988. Dispersion and reservoir heterogeneity, SPE Reservoir Eng., 3, 139- 148.
Babadagli, T.: “Temperature Effect on Heavy-Oil Recovery by Imbibition in Fractured Reservoirs,” J. Pet. Sci. and Eng. (1996a) vol. 14, no.
3-4, 197.
Babadagli, T.: “Heavy-Oil Recovery From Matrix During Thermal Applications in Naturally Fractured Reservoirs,” In Situ (1996b) vol. 20,
No. 3, 221.
Babadagli, T.: “Scaling of Capillary Imbibition under Static Thermal and Dynamic Fracture Flow Conditions” J. of Petroleum Science and
Eng., vol. 33, no. 4, 2002, 223.
Babadagli, T. and Al-Bemani, A.: “Investigations on Matrix Recovery During Steam Injection into Heavy-Oil Containing Carbonate
Rocks,” accepted for publication in J. Petr. Sci. and Eng, vol. 53, 2007, 259.
Babadagli, T., Sahin, S., Kalfa, U., Celebioglu, D., Karabakal, U. and Topguder, N.N.: “Development of Heavy-Oil Fractured Carbonate
Bati Raman Field: Evaluation of Steam Injection Potential and Improving On-Going CO2 Injection,” SPE 115400, 2008 SPE Annual
Tech. Conf. and Exh., Denver, CO, 21-24 Sept.
Briggs, P. J. et al.: “Development of Heavy-Oil Reservoirs,” JPT (Feb. 1988) 206.
Briggs, B.J. et al.: “Heavy Oil from Fractured Carbonate Reservoirs,” SPE Res. Eng., 7 (Feb. 1992), 179.
Er, V. and Babadagli, T.: “Miscible Interaction between Matrix and Fracture – A Visualization and Simulation Study,” accepted for
publication in SPE Res. Eval. and Eng., 2009.
Macaulay, R. C. et al.: “Design of a Steam Pilot in a Fractured Carbonate Reservoir – Qarn Alam Field, Oman,” paper SPE 30300 presented
at the 1995 SPE Int. Heavy Oil Symp., Calgary, Canada, June 19-21.
Nakamura, S., Sarma, H.K., Umucu, T., Issever, K. and Kanemitsu, M.: “A Critical Evaluation of a Steamflood Pilot in a Deep Heavy
Carbonate Reservoir in Ikiztepe Field, Turkey,” paper SPE 30727, presented at the SPE Annual Tech. Conf. and Exh., Dallas, TX, 267-
282.
Penney, R. et al.: “Steam Injection in Fractured Carbonate Reservoirs: Starting a New Trend in EOR,” IPTC 10727, Int. Petr. Tech. Conf.,
Doha, Qatar, 21-23 Nov., 2005.
Rawnsley, K. et al.: “Accelerated Understanding and Modeling of a Complex Fractured Heavy-Oil Reservoir, Oman, Using a New 3D
Fracture-Modeling Tool,” IPTC 10095, Int. Petr. Tech. Conf., Doha, Qatar, 21-23 Nov., 2005.
Reis, J. C.: “An Analysis of Oil Expulsion Mechanism from Matrix Blocks During Steam Injection in Naturally Fractured Reservoirs,” In
Situ, 16, No. 1 (1992), 43.
Sahuquet, B. C. and Ferrier, J. J.: “Steam-Drive Pilot in a Fractured Carbonate Reservoir: Lacq Superieur Field,” JPT (April 1982) 34, 873.
Shahin, G.T. et al.: “The Physics of Steam Injection in Fractured Carbonate Reservoirs: Engineering Development Options That Minimize
Risk,” SPE 102186, 2006 SPE Annual Tech. Conf. and Exh., San Antonio, TX, 24-27 Sept.
Snell, J.S. and Close, A.D.: “Yates Field Steam Pilot Applies Latest Seismic and Logging Monitoring Techniques,” SPE56791, 1999 SPE
Annual Tech. Conf. and Exh., Houston, TX, 3-6 Oct.
Trivedi, J. and Babadagli, T.: “Efficiency of Diffusion Controlled Miscible Displacement in Fractured Porous Media,” Transport in Porous
Media, vol. 71, no. 3, Feb. 2008, 379-394.
Trivedi, J. and Babadagli, T.: “Experimental Analysis of CO2 Sequestration Efficiency during Oil Recovery in Naturally Fractured
Reservoirs,” accepted for publication in Energy and Fuels, 2009.
6 SPE 124047

Table 1: Matrix properties

Matrix blocks in IJK direction 16x21x12
Fracture spacing 0.01 cm
Matrix porosity 0.147
Fracture porosity 1
Matrix permeability 550 mD
Fracture permeability 55 D
Heavy oil gravity 14 API
First thermal expansion coefficient CT1 0.0015

Table 2: Parameters obtained from the history matching of Phase 2

Solvent diffusion Solute diffusion Matrix dispersion Fracture dispersion
Rate coefficient coefficient coefficient coefficient
(cm2/min) (cm2/min) (cm) (cm)
0.1 cc/min 0.0002 0.000535 0.039 0.000888
0.3 cc/min 0.002 0.00535 1.7e-6 0.098
0.5 cc/min 0.0002 0.000535 0.0 0.00088

Table 3: Parameters obtained from upscaling at certain recovery factor (10 %) and at a certain time (480 min).
Upscaled Solvent diffusion Solute diffusion Matrix dispersion Fracture dispersion
model coefficient coefficient coefficient coefficient
(cm2/min) (cm2/min) (cm) (cm)
X2 0.0002 0.000535 0.078 0.003552
X5 0.0002 0.000535 0.159 0.0222
X 10 0.0002 0.000535 0.39 0.0888

Table 4: Upscaling equations for key parameters

Upscaling number l
N=
3.07W
Matrix dispersion coefficient upscale equation Di × N
Fracture dispersion coefficient upscale equation Di × N 2
Rate upscale equation 1.4196(4.4683Wb )
1.3886

Di : Dispersion coefficient obtained from the original model.

N : Upscaling number
W : Fracture width of the upscaled model.
b : Fracture spacing of the upscaled model.
l : Fracture length of the upscaled model
SPE 124047 7

Table 5: Solvent injection and recovery in the upscaled models.

x2 x5 x10
Total solvent injected 191.988 2399.98 16799.8
(cm3)
Total solvent produced @ 186.993 2305.89 16112.6
480 min (cm3)
Total solvent produced @ 190.499 2363.92 16463.8
540 min (cm3)
Amount of solvent 4.995 94.11 687.2
diffused in matrix (cm3)
Amount of solvent 3.506 58.05 351.2
produced in Phase 3 (cm3)
Solvent recovery factor 70 62 51
from Phase 3 (%)

Thermocouples Heating tape

Data acquisition
Core holder

Syringe pumps Heating unit Sensitive scale

Figure 1: Experimental set-up for three different solvent injection rates (Al Bahlani and Babadagli, 2009).

Figure 2: Single porosity cartesian model used to history match lab experiment dynamic results.
8 SPE 124047

2.00

1.50
Cumulative Oil(Heavy Oil) SC (cm3)

1.00

0.50

0.00
0 100 200 300 400 500
Time (min)

Diffusion and Dispersion

Only Diffusion coefficient present
Only Dispersion coeficient present

Figure 3: Three runs testing the effect of dispersion and diffusion coefficient on the simulation model.

100

90

80

70
Oil Recovery Factor %

60

50

40

30

20

10

0
0 50 100 150 200 250
Time (min)
Simulation
Experiment

Figure 4: History match of Phase 1 on the core sample later used for 0.1 cc/min solvent injection rate.
SPE 124047 9

100

90

80

70
Oil Recovery Factor

60

50

40

30

20

10

0
0 50 100 150 200 250
Time (min)
Simulation
Experiment

Figure 5: History match of Phase 1 on the core sample later used for 0.3 cc/min solvent injection rate.

100

80
Oil Recovery Factor %

60

40

20

0
0 50 100 150 200
Time (min)
Simulation
Experiment

Figure 6: History match of Phase 1 on the core sample later used for 0.5 cc/min solvent injection rate.
10 SPE 124047

Figure 7: Oil saturation at the last time step of Phase 1 for the core used in 0.1 cc/min solvent injection rate. This saturation
distribution was imported to the model for Phase 2.

100
Oil Recovery Factor (Heavy Oil) %

90
80
70
60
50
40
30
20
10
0

0 100 200 300 400 500

Time (min)

Experiment Simulation

Figure 8: History matching of Phase 2 for 0.1 cc/min injection rate.

SPE 124047 11

Oil Recovery Factor (heavy Oil) % 100

90
80
70
60
50
40
30
20
10
0
0 100 200 300 400 500 600
Time (min)

Experiment Simulation

Figure 9: History matching of Phase 2 for 0.3 cc/min injection rate.

100
Oil Recovery Factor (Heavy Oil) %

90
80
70
60
50
40
30
20
10
0
0 20 40 60 80 100 120 140
Time (min)

Experiment Simulation

Figure 10: History matching of Phase 2 for 0.5 cc/min injection rate.
12 SPE 124047

The first phase: Dispersion

dominated process (fracture oil only)
0.70

The second phase: Diffusion

0.60 dominated process
Cumulative Oil(Hvy Oil) SC (cm3)
The third phase: Dispersion
dominated process
0.50

0.40

0.30

0.20

0.10

0.00
0 100 200 300 400 500
Time (min)
0.1 cc/min
0.3 cc/min
0.5 cc/min

Figure 11: Demonstration of three distinctive phases during solvent injection process (Phase 2).

1.40

1.20

1.00
Oil Rate SC (cm3/min)

0.80

0.60

0.40

0.20

0.00
480 490 500 510
Time (min)
Simulation
Experimental

Figure 12: History match of Phase 3 for 0.1 cc/min solvent injection core.
SPE 124047 13

40
y = 1.4196x1.3886
35 R2 = 1
30
Tuned Rate

25

20

15

10

0
0 2 4 6 8 10 12
Calculated Rate

Figure 13: Relationship between the calculated upscaled solvent injection rate and tuned rate to obtain desired recovery factor.

50
45
Recovery Factor (Heavy Oil) %

40
35

30
25
20

15
10

5
0
0 50 100 150 200 250 300 350 400 450
Time (min)

x2 x5 x10
Figure 14: Recovery factor for three upscaled grid blocks sizes all reaching the same recovery factor at the same time.
14 SPE 124047

Recovery Factor (Heavy Oil) % 100

90
80
70
60
50
40
30
20
10
0
0 10 20 30 40 50 60 70
Time (min)
x2 x5 x10

Figure 15: Incremental recovery factor at Phase 3 from upsized models. Smaller models produced higher recovery.

100

80
Oil Rate(Heptane) SC (cm3/min)

60

40

20

0
480 485 490 495 500 505 510 515 520 525 530 535 540 545 550
Time (min)
x2 Phase 3
x5 Phase 3
x10 Phase 3

Figure 16: Oil rate during Phase 3 for the upscaled grid blocks.