Annual02 PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 288

Heavy and Thermal Oil Recovery

Production Mechanisms
By

Anthony R. Kovscek, Principal Investigator


Louis M. Castanier, Technical Manager

Technical Report

Annual report for the period


September 1, 2001 – August 31, 2002

Work Performed Under Contract DE-FC26-00BC15311

Prepared for
U.S. Department of Energy
Assistant Secretary for Fossil Energy

Jerry Casteel, Project Manager


National Petroleum Technology Office
P.O. Box 3628
Tulsa, OK 74101

Prepared by
Stanford University
Department of Petroleum Engineering
Green Earth Sciences Bldg., Room 080B
367 Panama Street
Stanford, CA 94305-2220
Disclaimer

This report was prepared as an account of work sponsored by an agency of the


United States Government. Neither the United States Government nor any agency
thereof, nor any of their employees, makes any warranty, express or implied, or assumes
any legal liability or responsibility for the accuracy, completeness, or usefulness of any
information, apparatus, product, or process disclosed, or represents that its use would not
infringe privately owned rights. Reference herein to any specific commercial product,
process, or service by trade name, trademark, manufacturer, or otherwise does not
necessarily constitute or imply its endorsement, recommendation, or favoring by the
United States Government or any agency thereof. The views and opinions of authors
expressed herein do not necessarily state or reflect those of the United States Government
or any agency thereof.

ii
Acknowledgements

This work was prepared with the support of the U.S. Department of Energy, under Award
No. DE-FC26-00BC15311 These contributions are gratefully acknowledged. However, any
opinions, findings, conclusions, or recommendations expressed herein are those of the authors
and do not necessarily reflect the views of the DOE.

iii
Summary

The Stanford University Petroleum Research Institute (SUPRI-A) conducts a


broad spectrum of research intended to help improve the recovery efficiency from
difficult to produce reservoirs including heavy oil and fractured low permeability
systems. Our scope of work is relevant across near-, mid-, and long-term time frames.
The primary functions of the group are to conduct direction-setting research, transfer
research results to industry, and educate and train students for careers in industry.

Presently, research in SUPRI-A is divided into 5 main project areas. These


projects and their goals include:

• Multiphase flow and rock properties—to develop better understanding of the physics
of displacement in porous media through experiment and theory. This category
includes work on imbibition, flow in fractured media, and the effect of temperature
on relative permeability and capillary pressure.

• Hot fluid injection—to improve the application of nonconventional wells for


enhanced oil recovery and elucidate the mechanisms of steamdrive in low
permeability, fractured porous media.

• Mechanisms of primary heavy oil recovery—to develop a mechanistic understanding


of so-called “foamy oil” and its associated physical chemistry.

• In-situ combustion—to evaluate the effect of different reservoir parameters on the


insitu combustion process.

• Reservoir definition—to develop and improve techniques for evaluating formation


properties from production information.

What follows is a report on activities for the past year. Significant progress was made
in all areas.

iv
TABLE OF CONTENTS

Acknowledgements iii

Area 1. Multiphase Flow and Rock Properties 1

A Technique for Measuring Two-Phase Relative Permeability in Porous 3


Media via X-ray CT Measurements
(Josephina Schembre and A.R. Kovscek)

A Micro-Visual Analysis of Matrix to Fracture Transfer 25


(Edgar Rangel-German and A.R. Kovscek)

An Experimental Study of Imbibition in Diatomite and Chalk 47


(S. S. LeGuen and L.M. Castanier)

Area 2. Hot Fluid Injection 63

Steam Injection into Light-Oil Diatomite Reservoirs 65


(B. Todd Hoffman)

Experimental Study of Heavy Oil Production from Diatomite by Water


Imbibition at Elevated Temperatures 97
(Tom Tang and A.R. Kovscek)

Area 3. Mechanisms of Primary Heavy Oil Recovery 127

Mechanistic of Solution Gas Drive in Viscous Oils 129


(Yi Tak Leung)

Gas Bubble Coalescence in Heavy Oil Reservoirs Undergoing Solution


Gas-Drive 145
(Omar Hayat)

Area 4. Insitu Combustion 169

Insitu Combustion, Handbook Chapter 171


(Louis Castanier and W.E. Brigham)

Upgrading Via Insitu Combustion 209


(L.M. Castanier and W.E. Brigham)

v
Area 5. Reservoir Definition 225

Integrating Production History into Reservoir Models Using


Streamline-Based Time of Flight Ranking 227
(Yuandong Wang and A.R. Kovscek)

Pore Microstructure and Fluid Distribution in a Diatomaceous Reservoir 261


(Cynthia Ross and A.R. Kovscek)

Improved Reservoir Definition Through Integration of Multiscale


Petrophysical Data 279
(Liping Jia)

vi
MULTIPHASE FLOW AND ROCK PROPERTIES

1
2
A Technique for Measuring Two-Phase Relative Permeability
in Porous Media Via X-ray CT Measurements
J. M. Schembre and A. R. Kovscek

ABSTRACT

A novel method for computing two-phase relative permeability curves from the
results of spontaneous imbibition experiments is presented. Using a specially constructed
imbibition cell and an X-ray Computed Tomography (CT) scanner, we obtain accurate
measurement of saturation profiles along the length of cores as a function of time. The
saturation profile history allows direct computation of the relative permeability for both
phases from a single experiment when used in combination with a previously measured
capillary pressure curve. Results are unique within experimental error. The proposed
procedure works equally well for spontaneous and forced cocurrent imbibition. It was
tested thoroughly using synthetic and experimental data, for water-air and water-oil
systems. Test results are described within the paper. Advantages include the
incorporation of capillary forces and no requirement for steady-state conditions. This
method is useful to measure imbibition relative permeability curves, especially in low
permeability rocks at relatively low wetting phase saturations. In such systems, it is
laborious to reach multiple steady states and capillary forces are significant so that
classical unsteady- state techniques do not apply.

1. INTRODUCTION

Simulation of multiphase flow in porous media requires knowledge of relative


permeability functions. These functions are necessary to make estimates of productivity,
injectivity, and ultimate recovery from oil reservoirs for evaluation and planning of
production operations (Honarpour and Mahmood, 1988). Therefore, measurements of
relative permeability in the laboratory and/or empirical and theoretical models are an
important subject in reservoir modeling.

The laboratory methods used to calculate relative permeability functions are


grouped into centrifuge, steady- and unsteady-state techniques. The centrifuge method
has been improved (Hirasaki et al., 1992; 1995), yet, it has many potential pitfalls in the
experimental and interpretation parts. It also poses limitations including loss of
information on the low saturation region that can not be gained from the production data
at low mobility ratio (Hirasaki et al., 1995). Additionally, concerns regarding the
replacement of viscous forces with a range of centrifugal forces still remain for unsteady
state displacement processes that are rate dependent (Ali, 1999). Steady- state methods
offer disadvantages, especially in the case of low permeability rocks, in which it is
laborious to reach multiple steady states and, capillary forces and capillary end effects are
significant (Firoozabadi and Aziz, 1988; Kamath et al., 1993). Steady-state techniques

3
have been improved in order to make corrections for capillary effects but they still
require successive measurements for different total flow rates (Virnosvky et al., 1995).
Capillary pressure has a significant effect on saturation distribution and recovery, and
capillary forces dominate multiphase flow in low-permeability rocks and fractured
reservoirs. In developing unsteady-state methods for low-permeability systems, it is
necessary to account for capillary pressure when obtaining the relative permeability
curve. Akin et al. (1999) proved that the use of the JBN technique, an unsteady
technique, leads to inaccurate measurements of relative permeability. Thus, most
unconventional unsteady techniques do not apply.

There are also methods that infer relative permeability by history matching observable
parameters such as fluid production, pressure drop, and saturation profile history. These
methods assume that the relative permeability and capillary pressure curves behave
according to a pre-determined function. The limitation imposed on the description of the
actual shape of the relative permeability curve causes a bias error and a variance error
(Watson et al., 1986; Hirasaki et al., 1995). When the number of parameters in the
functional representation increases, the bias error tends to decrease. However, because
more parameters are estimated on the basis of the same amount of information, the
variance error tends to increase.

In the method we propose, two-phase relative permeability curves are computed


from experimental in-situ, saturation profiles obtained from spontaneous imbibition
experiments. Each relative permeability value at a particular saturation is treated
independently when the experiment begins with low or zero initial water saturation. For
experiments where the core is partially saturated with water initially, the water saturation
increases almost uniformly in time throughout the core. In this event, the imbibition
method becomes quite similar to that proposed by Sahni et al. (1998) for gravity drainage
in three-phase systems. Fitting to a pre-determined functional shape is not necessary in
either case. This paper proceeds by first describing the set up used for experiments. Next,
the various equations needed for interpreting the data are developed. Synthetic CT data
are then used to first verify the technique before application to experimental data. In all
cases, the relative permeability curve obtained is input for numerical simulation and the
agreement between input and computed saturation profiles checked.

2. EXPERIMENTAL APPARATUS

Brigham and coworkers were among the earliest researchers to apply X-ray CT to
monitor the progress of multiphase displacement inside porous media (Liu et al. 1990,
Demiral et al. 1991, Liu et al. 1992). They showed the critical importance of measuring
the evolution of fluid phase saturation in situ for interpretation of experiments. The
experimental program used in this method applies these foundational studies. Figure 1
shows a schematic of the experimental apparatus. Imbibition in this study is cocurrent.
The experimental cell is constructed from acrylic tubes, and it consists of two separate
chambers. The main chamber is the core holder and it is surrounded by water, contained
in the second chamber. The objective of the second chamber is to reduce possible beam

4
hardening artifacts and provide temperature control. Core holder and water jacket fluids
never mix.

The scanning plane is along the entire length of the core in the direction of the
flow, as illustrated in Fig. 1. Hence, the porosity and saturation fields are measured on a
single vertical volume section in the center of the core as a function of time. The total
time of measurement is short enough to capture accurately the position of each saturation
value. As a consequence of the scan geometry, many saturation data points are obtained
and, importantly, a single scan provides the position of every saturation at a specific time.
Images obtained are flat and do not exhibit detectable positioning errors. Scanning is
performed every 50 seconds for the early stage of imbibition, afterwards the interval
between images is longer because the change in saturation becomes slower with time.
The duration of the early stage varies according to the rock and viscosity, and other
properties, of the fluids present in the core.

The raw CT-number data is converted to a map of porosity. The water saturation
profiles are obtained from subtracting images of raw CT data. One-dimensional
saturation profiles are obtained from two-dimensional images by averaging the saturation
values at a particular location. The procedure and expressions used for the calculation of
saturation profiles depend on the saturation history of the core and fluid saturations at the
beginning of the experiment. Further information about the experimental equipment,
procedure, pure fluid CT numbers, and data processing is provided by Akin and Kovscek
(1999) and Akin et al. (2000).
In the experiments to follow, diatomite cores come from a block of stone obtained
from the Grefco Quarry (Lompoc, CA). Berea sandstone samples were obtained from
Cleveland Quarries (Amherst, OH). The sandstone samples are fired at 450 °C for 12
hours in an oven to stabilize any clay present and prevent any fines migration. The quarry
diatomite samples display no signs of clay swelling, and consequently were not fired.

3. RELATIVE PERMEABILITY FROM CT DATA

The equations necessary for data processing of experiments are developed next.
First cases with little or no initial water saturation are considered. Then, equations
appropriate to cases with medium to high initial water are developed. The method differs
slightly among these two cases.

The mathematical model begins from the standard one-dimensional equation for
flow of two immiscible phases in porous media. We neglect the effect of gravity, but this
is not a necessary assumption. For the wetting-phase,

kkrw ∂pw
uw = − (1)
µw ∂x

5
where uw is the Darcy velocity, k is the absolute permeability, krw is the relative
∂p
permeability, µw viscosity, and w is the pressure gradient of the wetting phase.
∂x
The Darcy velocity is, in fact, a superficial velocity, that is identically equal
dx
to φ S w . Substituting this expression for uw in Eq. 1 results in
dt S w
dx kkrw ∂pw
φSw =− (2)
dt Sw µw ∂x
dx
where is the velocity of a particular saturation Sw as a function of time and is the
dt Sw

porosity. Integrating with respect to time, the relative permeability of a particular Sw is


 
 
 x Sw Sw  µ wφ (3)
k rw =   k
t ∂p
 − ∫  w 
 dt 
 0  ∂x  S 
 w 
In this function, the porosity, viscosity and, permeability are independently measurable
and constant. The pressure gradient history and the position of a particular saturation as a
function of time, x s , are measured during the experiment. Thus, Eq. 3 allows direct
w

computation of krw solely from measured data. A similar development for the non-
wetting phase yields
 
 
µ φS  x S nw 
k rnw = nw nw   (4)
k  t ∂pnw 
 ∫0 ∂x dt 
 S nw 
∂pnw
where is the non-wetting phase pressure gradient. Equations 3 and 4 are related by
∂x
the summation of phase saturations to 1.

It is also noted that the capillary pressure function can be obtained uniquely, given
the water relative permeability function, as

x S µ φ t 
 S w w w − ∫ ∂pnw dt 
 kkrw 0 ∂x

∂pc
= − 
sw Sw 
t ∂S  (5)
∂Sw
 ∫
w
dt 
 ∂x 
 0 Sw

6
The extension of the method to forced imbibition is given in the appendix. In cases with
initial water saturation and where the water saturation shows diffuse-like behavior, it
might be difficult to implement the method above. Multiple locations for a given
saturation value at a specific time lead to uncertainty in the assessment of the velocity of
the saturation value. For cases with initial water, a variant of the method of Sahni et al.
(1998) for gravity drainage is developed.

The Darcy velocity and conservation equation for each phase are again used.
Rather than integrating Darcy’s Law, the material balance for each phase is integrated.
The in-situ saturation measurement and an independently measured capillary pressure
relationship are still needed. The result for the wetting phase is
∂Sw
∫ dx
µ wφ ∂t x,t
krw =− (6)
k ∂pw
∂x x ,t
and similarly, for the non-wetting phase,

∂Snw
∫ dx
µnw φ ∂t x ,t (7)
k rnw =−
k ∂pnw
∂x x,t

4. VALIDATION

Next, we test and validate the method embodied in Eqs. 3, 4, 6, and 7, using
synthetic experimental data. Known relative permeability functions and petrophysical
properties are input to the simulator (ECLIPSE 100). Saturation histories are created that
can be analyzed. Relative permeability curves obtained from the simulated data are
compared to the corresponding input curves. The rock properties used in the synthetic
cases are descriptive of diatomite (Akin et al., 2000). Characteristics of the porous
medium and fluid properties are shown in Table 1. Water-air and water-oil spontaneous
cocurrent imbibition cases are considered. Further description of the models and results
obtained for each case follows.

4.1. Water-Air
The first case is water-air spontaneous imbibition with zero initial water
saturation. A one-dimensional grid with 1002 cells is used to generate the synthetic
saturation data and later to check to the relative permeability functions obtained. The fine
grid and small time steps limit, but do not eliminate, numerical artifacts.
We assume that the contribution of the non-wetting phase pressure gradient is not
significant to the determination of krw. Air is inviscid and infinitely mobile compared to

7
the wetting phase. Consequently, the wetting phase pressure gradient is approximately
equal to − ∂p c . The relative permeability functions for water and gas phases follow the
∂x
functions of Fatt and Klikoff (1959) with power-law coefficients of 3 for each phase.
The capillary pressure curve is given by a polynomial fit to the Leverett function (1941):

[ ]
1/ 2
φ 
Pc =   σ cos θ 4.013 − 13.68S w* + 21.313( S w
* 2 * 3
) − 11.548( S w ) (8)
k

where σ is the fluid interfacial tension, θ is the contact angle, and the reduced saturation
is given as

* Sw − Swc
Sw = (9)
1 − Sw

The subscript, wc, refers to connate water. A plot given by this function and the data in
Table 1 is shown in Fig. 2. There is no initial water saturation, no connate water
saturation, and no residual gas saturation. As these are synthetic cases, we are not too
concerned with the shape of the capillary pressure curve. That is, the capillary pressure
curve in Fig. 2 applies strictly to drainage, but it is convenient for use in this verification
exercise.

Although the non-wetting-phase pressure gradient is neglected when finding the


wetting-phase relative permeability, the method yields the non-wetting phase relative
permeability via Eq. 4. A comparison of the input relative permeability curves with those
obtained from applying Eqs. 3 and 4 to the synthetic input data is shown in Fig. 3. Solid
lines are input and symbols represent output data. The curves shown for the water relative
permeability include the case where the gas pressure gradient is neglected and the case
where it is included in the calculations. The semi-logarithmic scale is used to accentuate
any differences, especially in the low water saturation range. The computed curves follow
the input curves with an error of approximately 12 per cent. The largest error is attributed
to the computation of the integrals numerically. A second cause of error is due to
uncertainty in locating the position of a particular saturation from the saturation profiles
as a function of time. Neglecting the pressure gradient of the air leads to insignificant
errors.

To test the ability of the calculated curves to reproduce the input saturation
profiles, simulations were conducted. Figure 4 compares the input saturation history and
that obtained using the computed relative permeability curves. The larger values of krw
computed at low water saturation affect slightly the leading edge of the saturation fronts
at late times. Otherwise, the saturation history is matched exactly.

4.2. Water-Oil
In this case, the properties of n-decane describe the non-wetting fluid and the
initial water phase saturation is zero. The 1002 cell one-dimensional grid is again used to

8
generate synthetic CT data. The non-wetting phase pressure gradient is included in Eq.
(3). Despite the variation in interfacial tension among water-air (~70 mN/m) and water-
oil (~50 mN/m) systems, the capillary pressure curve from the previous case is used
again (Fig. 2). Figure 5 shows the input relative permeability curves and the output data
as symbols. Note the ability of the method to reproduce the non-functional shape of the
curve. Relative permeability is also computed and shown for the case where we chose to
neglect the oil-phase pressure gradient. The relative permeability obtained when we
neglect the oil-phase pressure gradient agrees with the input curve with the exception of
water saturation values greater than 0.7. At high water saturations, the gradients in Pc
and po are roughly equal in magnitude. The water relative permeability obtained when we
include the oil-phase pressure gradient in the calculations matches the input data at large
Sw somewhat better. The nonwetting phase is not especially viscous allowing the close
match with or without incorporating the oil-phase pressure gradient.

The oil relative permeability computed from Eq. (4) is compared to the input
curve in Fig. 5 as well. The inaccuracy is larger in this case than in the water-air case,
especially at low oil (i.e., large water) saturations. The relative permeability values for
the highest Sw values no longer follow the same trend as the input curve. A close
examination of the computational output revealed spurious data reported by the simulator
for the oil-phase pressure gradient at early times. Initial time-step size is set to the
minimum value allowed by the simulator, yet the oil-phase pressures are not converged at
early times. This numerical error affects the integral term in Eqs. 4 and 7. In short, the
input-output mismatch for oil relative permeability is not inherent in the method proposed
here, but lies in numerical artifacts generated by the simulator.

The input water saturation history and that obtained from simulation using the
computed relative permeability curves are compared in Fig. 6. Similar to the water-air
case, the larger values of krw at low water saturation affect slightly the leading edge of the
saturation fronts at late times. The mismatch in the oil relative permeability curves for the
largest water saturations affects the overall speed and shape of the imbibition front.
Hence, the disagreement in the position of Sw as a function of time for the largest values
of Sw.

4.3. Water-Oil, Initial Water


Next, the method as applied to cases with initial water was validated using
synthetic CT data. The capillary pressure curve is shown in Fig. 7 and the input relative
permeability curves are shown in Fig. 8. The initial water saturation is uniform and equal
to 0.6. This gives a core that fills slowly by spontaneous imbibition with little of the
frontal-like displacement of the previous cases. The properties of the rocks are those of
diatomite, Table 1, and the non-wetting fluid has the characteristics of decane. The water
relative permeability obtained, using Eq. 6 and the oil relative permeability computed
from Eq. (7) are compared to the input curve in Fig. 8. In both cases, the curves obtained
follow the corresponding input curves closely, with some scatter. The scatter is mainly
related to the fact that the relative permeability curves are obtained from the scatterplot of

9
Kp(x) vs Sp(x). Figure 9 compares the synthetic saturation history with that obtained
using the measured relative permeability curves; the results match well.

5. EXPERIMENTAL CASES

The verification exercises indicate that the method yields accurate measurement
of relative permeability for water-oil and water-air experiments provided that the non-
wetting phase pressure gradient versus time is measured. A previously determined
capillary pressure is also needed. Next, the method is applied to experimental data. We
use water saturation profiles measured by X-ray CT during spontaneous imbibition water
imbibition into air-filled Berea sandstone and diatomite. The properties of each system
are summarized in Table 2. The column "pixel resolution" refers to the number of CT-
measured saturation data.

5.1. Berea Sandstone


The imbibition capillary pressure curve given by Sinokrot et al. (1969), was
modified for different fluid system and petrophysical properties (i.e., k, φ) by using the
Leverett J-function (Leverett, 1941; Ma et al., 1991) That is, the required curve was first
non-dimensionalized and then rescaled according to k,φ, and σ (θ = 0). The capillary
pressure relationship used in this case is shown in Fig. 10. The relative permeability
curve obtained is shown in Fig. 11. The values of krw obtained are reasonable. Moreover,
the endpoint krw obtained agrees with endpoints found in the literature for Berea
sandstone (Lerdahl et al., 2000; Oak et al., 1990).

A comparison of the CT-determined experimental water saturation profiles versus


time and the profiles obtained by simulating the core with the measured relative
permeability is shown in Fig. 12. The results match especially well given the scatter in
the experimental data. Notice that the shape and position of the imbibition front are well
reproduced. Since the pressure gradient within the core was not measured in the
experiments, it was not possible to find the non-wetting phase relative permeability. A
linear krg relationship is used for simulation. Because the gas is inviscid, the shape of krg
versus Sw has little effect on the computed saturation profiles in Fig. 12.

5.2. Diatomite
There is very little known about petrophysical properties in diatomite. The use of
the classical unsteady state method in the determination of relative permeability for
diatomite surely leads to inaccurate results because of low permeability and
corresponding capillary effects. The same experimental method as in the case of the
Berea sandstone was used to obtain the water saturation profiles. The core is initially air-
filled.

The input imbibition capillary pressure curve was measured with the porous-plate
method as reported by Akin and Kovscek (1999). It is similar to another curve shown by

10
Kumar and Beatty (1995). The capillary pressure used is that shown in Fig. 7. The
relative permeability curve obtained from the measured Sw profile is shown in Fig. 13.
At high water saturation, the relative permeability is found to decrease as Sw increases.
The reason for this artifact is shown in Fig. 14. Here, the experimental water saturation
profile and the profile obtained by simulating the imbibition process are compared. High
water saturations within the experiments advance slowly at earlier time. Note the first
profile for a time of 0.0069 hr. This is indicative of flow that is not one dimensional at
small times. Therefore, retardation exists for saturation values greater than 0.7. The solid
line gives the relative permeability curve used for simulation:

k rw (Sw ) = 5E −5 exp(8.3355 Sw ) (10)

Despite the difficulties in measuring krw at large Sw, the results obtained from
simulation (solid line) with Eq. 10 match the saturation-history data well, as shown in
Fig. 14. Additionally, the data points in Fig. 13 and the extrapolation at high water
saturations indicate a low endpoint value of roughly 0.2 for krw. This is indicative of and
consistent with the strong water-wetness of these core samples (Akin et al., 2000; Zhou et
al., 2002).

6. CONCLUSIONS

A method has been developed to determine dynamic relative permeability from


the water saturation profile history obtained during X-ray CT monitored spontaneous
imbibition experiments. The method was tested using synthetic data for the water-air and
water-oil cases. In addition, the method was adapted for cases with large initial water
saturation. Computed relative permeability values matched well the input curves,
suggesting reliability for these systems. In the water-oil case, discrepancies among the
measured and input oil relative permeability values were attributed to numerical
inaccuracies in the simulator. The air-phase pressure gradient can be neglected in the
calculations of water relative permeability. A sensitivity study showed that neglecting the
oil-phase pressure gradient might lead to error in the high water saturation values for
viscous oils.

The procedure was used to determine the water relative permeability for two
experimental cases with Berea sandstone and quarried diatomite. In both cases, the
relative permeability curves obtained were used to simulate the water saturation profile
versus time. The computed saturation history reproduced the experimental behavior well.
Importantly, this method allows measurement of relative permeability from a single
experiment. Potentially this represents a great savings of time.

7. NOMENCLATURE

h height of the water column, L


k absolute permeability, L2

11
kp relative permeability of phase p, dimensionless
p phase pressure, ML-1T-2
pc capillary pressure, ML-1T-2
Sp saturation for phase p, dimensionless
Sw* rescaled water saturation, dimensionless
t time, T
uw flow velocity of phase p, LT-1
x length, L

Greek Symbols
σ interfactial tension, MT-2
ρ density, ML-3
φ porosity, dimensionless.
µ fluid viscosity, mL-1T-1
θ contact angle

Subscripts
o oil
nw non wetting
w water, wetting

12
8. REFERENCES

1. Akin, S. and Kovscek, A.R.: “Imbibition Studies of Low-Permeability Porous


Media.” SPE 54590, paper presented at the SPE Western Regional Meeting,
Anchorage (May 1999) pp. 26-29.

2. Akin, S., Schembre, J.M., Bhat, S.K. and Kovscek, A.R.: “Spontaneous Imbibition
Characteristics of Diatomite,” J. Pet. Sci. Eng. 25 (2000) pp. 149-165.

3. Ali, J.K.: “ Development in Measurement and Interpretation Techniques in Coreflood


Tests to Determine Relative Permeabilities.” SPE 39016, paper presented at the Fifth
Latin American and Caribbean Petroleum Engineering Conference and Exhibition,
Rio de Janeiro, Brazil (Aug. 30 -Sept. 3, 1997).

4. Dake, L.P.: Fundamentals of Reservoir Engineering. Elsevier, Amsterdam, pp. 114-


116 (1978).

5. Demiral, B.M.R., Castanier, L.M. and Brigham, W.E.: “CT Imaging of Steam and
Steam/Foam Laboratory Experiments.” SPE 22644, paper presented at the 66th Ann.
Tech. Conf. and Exhibition of the SPE, Dallas (October 6-9, 1991).

6. Fatt, I. and Klikoff, W.A.: “Effects of Fractional Wettability on Multiphase Flow


Through Porous Media.” Pet. Trans. AIME 216 (1959) p. 246.

7. Firoozabadi, A. and Aziz, K.: “Relative Permeabilities from Centrifuge Data.” SPE
15059, paper presented at the 56th California Regional Meeting of the SPE, Oakland
(April 2-4, 1988).

8. Garg, A., Kovscek, A.R., Nikravesh, M., Castanier, L.M. and Patzek, T.W.: “CT
Scan and Neural Network Technology for Construction of Detailed Distribution of
Residual Oil Saturation During Waterflooding.” SPE 35737, Proceedings of the
Western Regional Meeting, Anchorage (May 22-24, 1996).

9. Hirasaki, G.J., Rohan, J.H. and Dudley, J.W: “Modification of Centrifuge and
Software for Determination of Relative Permeability Curves.” SPE 25290, paper
presented at the Annual Technical Conference, Washington D.C. (October 4-7, 1992).

10. Hirasaki, G.J., Rohan, J.H. and Dudley J.W.: “Interpretation of Oil-Water Relative
Permeabilities from Centrifuge Experiments.” SPE 24879. SPE Adv. Tech. Series 3,
1 (1995) pp. 66-75.

11. Honarpour, M. and Mahmood, S.M.: “Relative Permeability Measurements: An


Overview.” Journal of Petroleum Technology 40 (August 1988) pp. 963-966.

12. Kamath, J., deZabala, E.F. and Boyer, R.E.: “Water/Oil Relative Permeability
Endpoints of Intermediate-wet, Low Permeability Rocks.” SPE 26092, proceedings
of the Western Regional Meeting, Anchorage (May 26-28, 1993).

13
13. Kumar, M. and Beatty, F.D.: “Cyclic Steaming in Heavy Oil Diatomites.” SPE
29623, proceedings of the Western Regional Meeting, Bakersfield (March 8-10,
1995).

14. Lerdahl, T.R., Øren, P. and Bakke, S.: “A Predictive Network Model for Three-
Phase Flow in Porous Media.” SPE 59311, proceedings of the SPE/DOE Improved
Oil Recovery Symposium, Tulsa (April 3-5, 2000).

15. Leverett, M.C.: “Capillary Behavior in Porous Solids.” Pet. Trans. AIME 142 (1941)
p. 152.

16. Liu, D.B., Castanier, L.M. and Brigham, W. E.: “Analysis of Transient Foam Flow in
1-D Porous Media with CT.” SPE 20071, proceedings of the SPE California Regional
Meeting and Exhibition, Ventura (April 2-6, 1990).

17. Liu, D.B., Castanier, L.M. and Brigham, W.E.: “Displacement by Foam in Porous
Media.” SPE 24664, paper presented at the 67th Ann. Tech. Conf. and Exhibition of
the SPE, Washington D.C. (October 4-7, 1992).

18. Ma S., Jiang, M. X. and Morrow N.R.: “Correlation of Capillary Pressure


Relationships and Calculations of Permeability.” SPE 22685, proceedings of the 66th
Annual Technical Conference and Exhibition, Dallas (October 6-9, 1991).

19. Oak, M.J., Baker, L.E. and Thomas, D.C.: “Three-Phase Relative Permeability of
Berea Sandstone.” Journal of Petroleum Technology 42 (August 1990) pp. 1054-61.

20. Sahni, A., Burger, J. and Blunt, M.: “Measurements of Three-Phase Relative
Permeability During Gravity Drainage Using CT Scanning.” SPE 39655, proceedings
of the SPE/DOE Improved Oil Recovery Symposium, Tulsa (April 19-22, 1998).

21. Sinnokrot, A.A., Ramey, H.J. and Marsden, S.S.: “Effect of Temperature Level Upon
Capillary Pressure Curves.” SPE 2517, proceedings of the SPE 44th Annual Fall
Meeting, Denver (Sept 28-Oct. 1, 1969).

22. Virnosvky, G.A., Skjaevaland, S.M., Surdal, J. and Ingsøy, P.: “Steady-State
Relative Permeability Measurements Corrected for Capillary Effects.” SPE 30541,
paper presented at the 1995 SPE Annual Technical Conference and Exhibition, Dallas
(October 22-25, 1995).

23. Watson, A.T., Kerig, P.D., Richmond, P.C. and Tao, T. M.: “An Improved Method
for Estimating Relative Permeability from Displacement Experiments.” SPE 15064,
proceedings of the 56th California Regional Meeting, Oakland (April 2-4, 1986).

24. Zhou, D., Jia,. L. Kamath, J. and Kovscek, A.: “Scaling of Counter-Current
Imbibition Processes in Low Permeability Porous Media.” Journal of Petroleum
Science and Engineering (to be published, 2002).

14
APPENDIX: EXTENSION OF THE METHOD FOR FORCED IMBIBITION

Thus far, we have assumed spontaneous imbibition with no applied head in the
analysis of the water saturation profile. One of the main concerns when analyzing the
water saturation profiles is the maintenance of water supply such that the experimental
results represent purely spontaneous imbibition behavior. In this section, we present the
numerical solution for the case of slight forced imbibition.

For the case of forced imbibition, we recall that the psi-potential for any fluid is
defined as (Dake 1978)
Ψ = p + ρgh (A-1)
Assuming that the water level is constant and geometry is as shown in Fig. A-1, the water
potential gradient can be approximated by
dΨ w dp ρ gh
=− c+ w (A-2)
dx dx x S , t
w

where, x|Sw,t is the position of a particular water saturation at a given time, and h is the
level of the water column measured from the water-core contact.

Substituting Eq. (A-2) in Eq. 3 and making the same assumptions, we express
relative permeability at any given water saturation and time as
x S w Sw µ wφ
k rw = (A-3)
t  ∂p ∂S ρ gh  k
∫  −  dt
c w
0  ∂Sw ∂x x  S
w
Equation A-3 shows that neglecting the height of the water column in the method
leads to larger values in the relative permeability curve. The size of this error depends on
the magnitude of the water column compared to the derivative of capillary pressure. In
general, the derivative of the capillary pressure is more sensitive to these effects for high
water saturation values. This result extends the application of this method to data
collected during forced imbibition experiments.

15
Table 1: Properties of Rock and Fluids Used in Validation of the Method

Permeability [md] 6.6


Rock Properties Porosity 0.65
Length [cm] 8.8
Compressibility 5E5
Water Viscosity [cp] 1.0
Density [g/cm3] 1.0
Fluids Air Viscosity [cp] 0.02
Properties Density [g/cm3] 0.002
Oil Viscosity [cp] 0.86
Density[g/cm3] 0.8

Table 2: Characteristic of the Berea Sandstone and Diatomite Cores

Berea Sandstone Diatomite


Porosity[fraction] 0.22 0.68
Permeability[md] 780.0 6.6
Diameter[cm] 2.43 2.43
Length 8.8 9.02
Pixel resolution 210 225

16
Fig. 1. Schematic of the imbibition cell.

600000

500000

400000
Pc [pa]

300000

200000

100000

0
0 0.2 0.4 0.6 0.8 1
Sw

Fig. 2. Capillary pressure relationship used for the synthetic water-air case.

17
1

R e la tive P e rm ea b ility 0.1

0.01

0.001
Krw including dpg/dx

Krw neglecting dpg/dx

krg obtained
0.0001
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Sw

Fig. 3: Relative permeability curves for the synthetic water-air case: solid lines are input
and symbols are the output data obtained.

1 time [hours]

0.013686
0.9

0.108706
0.8

0.7 0.195416

0.6
0.363416
Sw

0.5

0.4

0.3

0.2

0.1

0
0 1 2 3 4 5 6 7 8
X[cm]

Fig. 4. Water saturation profiles obtained for reference and input relative permeability
curves: solid lines are input and the symbols are obtained from computed relative
permeability curve.

18
1

0.1

R e la t iv e P e rm e a b ilit y

0.01

0.001

Output kro
0.0001
Output krw: dpodx
neglected
Output krw with
dpo/dx
0.00001
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Sw

Fig. 5. Input and output relative permeability curves obtained for the synthetic water-oil
case: solid lines are input and symbols are the output data obtained.
1

0.089416
0.9

0.8 0.389416

0.7
0.714416

0.6

1.16942
Sw

0.5

0.4

0.3

0.2

0.1

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
X[cm]

Fig. 6. Water saturation profiles obtained for the water-oilcase using input (solid) and
output (symbols) relative permeabilities.

19
1.8E+06

1.6E+06
Capillary Pressure used in method

1.4E+06 Capillary pressure reported in


SPE54590
1.2E+06

Pc [pa] 1.0E+06

8.0E+05

6.0E+05

4.0E+05

2.0E+05

0.0E+00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Sw

Fig. 7. Capillary pressure curve used for both the synthetic case with initial water
saturation and the diatomite imbibition experiments. Data points are from Kumar and
Beatty (1995) and Akin and Kovscek (1999).

Fig. 8. Relative permeability curves for the synthetic water-air case with initial water
saturation: solid lines are input and symbols are the output data obtained.

20
1

0.95

0.9

0.85
Sw

0.8

0.75

time [mins]
- 0.000
0.7 - 8.00478
- 20.0048
- 32.0042
- 62.0042
0.65 - 92.0048
-182.005
-242.005

0.6
0 1 2 3 4 5 6 7 8
X[cm]

Fig. 9. Water saturation profiles obtained for reference and input relative permeability
curves for the synthetic water-air case with initial water saturation: solid lines are input
and the symbols are obtained from computed relative permeability curve.

60000

50000

40000
Pc(pa)

30000

20000

10000

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Sw

Fig. 10. Capillary pressure for Berea sandstone.

21
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

0.1

0.01
Krw

0.001

0.0001

0.00001

0.000001
Sw

Fig. 11. Experimentally determined water relative permeability curve for Berea
sandstone.

0.8

0.7

0.6

0.5
Sw

0.4

0.3

0.2 0.033333333 hr

0.108333333 hr
0.1
0.166666667 hr

0.233333333 hr
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
X[cm]

Fig. 12. Saturation profile obtained by simulation with measured krw compared to the
experimental profile for Berea Sandstone.

22
1

0.1
Krw

0.01

0.001

0.0001
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Sw

Fig. 13. Experimentally determined water relative permeability curve for outcrop
diatomite.

1
0.006944444
0.9 0.4125
0.6625
0.8
0.995833333
1.679166667
0.7

0.6
Sw

0.5

0.4

0.3

0.2

0.1

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1

X[cm]

Fig. 14. Saturation profile obtained by simulation with measured krw compared to the
experimental profile for outcrop diatomite.

23
Figure A-1: Outline of the experimental setup for the method with forced imbibition.

24
Micro-Visual Analysis of Matrix-Fracture Interaction
E. R. Rangel-German and A. R. Kovscek

ABSTRACT

To understand fully how fluids communicate between high conductivity


fractures and lower conductivity matrices and how matrix-fracture mass transfer
occurs, one needs to analyze the problem at different scales. In this paper, a number of
micromodel experiments at constant injection rates were performed to study the flow
behavior of fractured media at the pore-scale and to help understand similar
macroscopic experiments conducted at the core-scale reported previously in the
literature (Rangel-German and Kovscek, 2001). The ultimate goal is to provide realistic
information on how to model or simulate matrix to fracture transfer. Exhaustive pore-
scale imaging of fluid flow driven by capillary imbibition was achieved. Pictures of the
entire micromodel were also obtained to visualize the macroscopic behavior of the
experiments. Filling-fracture and filled-fracture regimes were analyzed. Results showed
that the mode of imbibition is clearly counter-current within the instantly-filled regime,
whereas imbibition is co-current while the fractures fill (non-wetting fluid is expelled
from a matrix-block to a fracture ahead of the water front in the fracture). It was also
found that counter-current imbibition took place in “channels” at discrete locations
along the fracture as opposed to counter-current flow with a pore or set of pores.

1. INTRODUCTION

Understanding and modeling flow behavior in naturally fractured rocks,


including oil and gas reservoirs, geothermal, and environmental sites requires a detailed
analysis of these systems at various length scales. To understand how fluids
communicate between the high conductivity fracture and the lower conductivity matrix
requires both pore- and core- scale analysis of the problem. Several experiments have
focused on matrix-fracture interaction at pore-scale. Rangel-German and Kovscek
(2001) reported a number of experiments exploring matrix-fracture interaction covering
a wide range of fracture apertures and flow rates using different fluids. They reported
two imbibition regimes. The first one is the named filling-fracture regime which shows
a variable length plane source due to relatively slow water flow through fractures.
Water advance in the fracture is controlled by the interaction between the matrix and the
fracture as shown in Fig. 3. It was found that for this early-time of imbibition (filling-
fracture) the water front advance in a fracture that interacts with a porous-rock block
scales with the square-root of time. The second regime is named instantly-filled
fracture, in which little water imbibes before the fracture fills with water. The second

25
image in Fig. 3 is nearly one-dimensional water advance. They found that the mass of
water imbibed scaled linearly with the square root of time for the instantly-filled
fracture regime. These experiments and the proposed analytical model indicated that
during the filling-fracture regime the overall rate of wetting fluid imbibed into the rock
is proportional to time and that a considerable amount of water is imbibed. For a low
injection rate, virtually all of the non-wetting fluid can be expelled from a square matrix
block with all sides no-flow except the fracture face by the time the water front reaches
the end of the fracture. This is because co-current imbibition apparently represents a
more efficient displacement force than counter-current imbibition (Rangel-German and
Kovscek, 2001; Pooladi-Darvish and Firoozabadi, 2000). Whereas the mode of
imbibition is clearly counter-current within the instantly-filled regime, imbibition may
be co-current while the fractures fill.

Water propagation and multiphase flow in fractured reservoirs depends upon,


the combined, nonlinear effects of hydraulic connectivity and wettability of fractures,
rock matrix permeability and porosity, matrix-block size, capillary pressure, and the
interfacial tension between the resident and the imbibing phase. The rate of mass
transfer between the rock matrix and fractures is significant, and calculation of this rate
with dual-porosity reservoir simulations depends on matrix-fracture transfer functions
incorporating shape factors. It is apparent that the transfer shape factors for totally
immersed blocks (classic approach) will be different to those for blocks partially
covered by water.

The ability of a fracture to conduct fluids at partial saturation is still one of the
biggest unknowns in numerical reservoir simulation. In real reservoirs, wetting and non-
wetting fluids will flow simultaneously in both fractures and the matrix, and the
combination of these two flow processes results in a ‘combined’ relative permeability
behavior that has not been determined. The difficulty of modeling this behavior comes
from the fact that matrix and fracture interact in a specific manner depending on a
number of parameters such as: fluid properties, rock properties, fracture properties, and
the ‘combined’ properties. These combined properties include capillary pressure,
relative permeability, and fracture continuity, and should be analyzed with the matrix
and the fracture coupled.

To study the physics of this matrix-fracture transfer and to better understand


some of the core-scale experiments regarding matrix-fracture interaction in the
literature, we designed and built a micromodel and micromodel holder that mimic,
macroscopically speaking, the set-up used in the core scale experiments of Rangel-
German and Kovscek, 2001. This set-up allows observation of the water advance in the
fracture and the pattern, extent, and mode of imbibition in the matrix at the pore scale.
Micromodels are transparent networks of pores and constrictions that simulate some of
the complexities of natural porous media.

We believe that the filling-fracture regime is mostly controlled by co-current


imbibition, whereas the filled-fracture regime is controlled by counter-current
imbibition. The objective of this paper is to prove the hypotheses and allow us to

26
observe whether the expulsion of non-wetting fluid from a matrix-block to a fracture
takes place at pore-by-pore scale all along the fracture or if it occurs as “channels” that
organize themselves at specific locations along the fracture to carry wetting fluid in one
set of channels and non-wetting fluid in the other. This paper should also help in the
efforts to derive general matrix-fracture transfer functions and to provide better
recommendations on the use of non-zero fracture capillary pressure or the correct type
of fracture relative permeability assumptions.

2. EXPERIMENTAL SET-UP AND PROCEDURE

Experiments at constant water injection rates into the fracture were performed
using the micromodels and the general set-up described below.

2.1 Micromodels

Micromodels are transparent networks of pores and constrictions that simulate


some of the complexities of natural porous media, as shown in Fig. 1a. Note the grains
that appear as islands and the etched pores and throats. The etch depth is 25 µm .
Generally speaking, micromodels have been developed to fill the need of researchers to
see flow processes at the pore level (Buckley, 1999). They are produced with the
objective of directly observing fluid flow through porous media. They contain an etched
flow pattern that can be viewed with a microscope. The flow pattern can be a lattice of
straight or constricted channels, or the pore network can be modeled after the pores
found in naturally occurring rock. One limitation that is inherent to all micromodels is
that they are only two-dimensional (George, 1999). Extrapolation of results to the three-
dimensional flow problem occurring in real porous media must be done carefully.
However, micromodels provide the best means to visualize fluid movement at the pore
scale while honoring both geometric and topologic properties of real rocks. A detailed
literature review on micromodels is provided by Buckley (1999).

2.1.1. Fabrication and Bonding of Stanford Micromodels

Etched-silicon-wafer micromodels of the type used by Castanier and Sagar


(1998), initially developed by Hornbrook et al. (1991), were used in this work.
Typically micromodels are made up of a repeated pattern of an SEM (Scanning
Electron Microscopy) image of a reservoir rock thin section. Usually, digital
modification of the image is necessary to ensure continuity in the porous medium. The
micromodels here are 5-cm squared etched pore patterns. These patterns have grains
ranging from the size of 30 to 200 um (Fig. 1a). The porosity of these micromodels is
roughly 0.2. The permeability of the micromodel is approximately 0.1 mD. These types
of micromodels have been used to study flow behavior in multi-phase flow, oil-foam
interaction, solution gas drive, contaminant hydrogeology, etc.

27
For the current study, an inlet/outlet port is located in every corner of the pore
network. Holes are drilled into the silicon wafer at these ports, providing locations that
allow fluid to enter and exit the micromodel. These micromodels contain some
modifications as compared to those used by Castanier and Sagar (1998) and George
(1999): narrow etched channels adjacent to the edges of the pore network carry fluid to
the inlet/outlet ports. These channels allow fluid to enter and exit the porous medium at
any point along the side of the pore network. The two parallel channels etched on the
micromodels, thus, act as fractures adjacent to the porous medium as shown in Figure
1b.

The micromodels used for this work were fabricated at the Stanford
Nanofabrication Facility (SNF), where we adapted fabrication techniques used for
computer industry, such as microchips. The micromodels created this way offer 1:1
scaling of typical sandstone pores and, for this work, were made to represent Berea
sandstone. There are two main steps in the construction of a micromodel:

2.1.2. Fabrication

The fabrication of a micromodel starts with the design and construction of a


mask, which is a glass-chrome plate where the pattern (a digitalization of a rock section
into a computer from a high quality photograph) is engraved. The mask is placed on a
silicon wafer previously coated with a photosensitive chemical. The pattern of the mask
is impressed on the wafer by means of passing ultraviolet light through the mask. The
silicon wafer is developed and later etched with hydrofluoric acid (HF) to a depth of
around 25-30 µm. In this way pore cavities are created. Flow channel depth is up to a
maximum possible etching depth of 30-35 µm.

2.1.3. Cleaning And Anodic Bonding

The silicon wafer with the etched pattern is cleaned in a sulfuric acid (HSO4)
cleaning solution (Nochromix crystals and sulfuric acid) air dried, and then bonded to a
glass plate to create a flowing medium. The procedure of bonding is simple, yet
extremely important to guarantee a flow medium useful for the experiments. Firstly, the
cleaned micromodel is placed with the pattern facing up on a previously heated hotplate
at 600° F. This hotplate is connected to one of the electrodes (anode) of a power supply.
The micromodel is left on the plate for around 45 minutes. The idea of this step is that
the silicon becomes oxidized (because of the air at high temperature) resulting in the
formation of a film of silicon dioxide, which is water wetting. The temperature is then
dropped to 500° F. After the hotplate has stabilized to this temperature, a clean glass
plate (cleaned with soap solution, then sulfuric acid solution, and then air dried) is
placed on the micromodel. The other electrode (cathode) is placed on the top of the
plate and the voltage on the power supply is brought to 1000Volts. After 50 minutes
bonding should be achieved. Electricity is turned off and the new micromodel is
allowed to cool to room temperature.

28
2.2. Micromodel Holder

A holder to mount the micromodels for unobstructed viewing under the


microscope was specially designed. This aluminum (6061 T6) micromodel holder has
conduits connected to the injection and production systems. It has four ports (one for
each hole of the micromodel) that are sealed with viton O-rings. See Fig. 2.

2.3. Microscope

A Nikon Optiphot-M reflected light microscope with a photo tube allowing for
video imaging was used in the apparatus. The properties of the optical lenses used are
outlined in Table 1. Working distance is the distance between the tip of the lens and the
focal plane of the objective. Numerical aperture relates to the light gathering ability and
is a dimensionless number between 0 and 1. The higher the value, the better the light
gathering. The lens 50X magnifications were used to track the motion of the fluids in
the micromodel, while the other lenses were used to focus on a specific portion of a
micromodel to study fluid motion there.

TABLE 1. PROPERTIES OF THE OBJECTIVE LENSES

Model Total Working Distance Numerical Viewable


Magnification (mm) Aperture µm)
Diameter (µ
5X 50 X 20.0 0.1 3000
0.4LWD 200 X 6.0 0.4 800
0.5ELWD 400 X 10.1 0.5 375

2.4. Microscope Imaging

A television camera was used to record the images through the microscope into
a video cassette recorder (VCR) that recorded at a speed of 30 frames per second. Then
the images from the VCR were transferred to a Macintosh/PC computer system with the
aid of imaging video/recording software.

2.5. Macroscopic Imaging

A commercial home-video digital camera was used as a second imaging device


to obtain images of the entire micromodel (macro-scale) at different stages of the
experiments. In these pictures the pattern and extent of imbibition in the entire
micromodel was observed.

29
2.6. Pump, Filter and Fluids

An ISCO 100DM syringe pump (Instrumentation Specialties Co.) was used. This
pump is very accurate for very low flow rates; i.e. 0.001 cc/min to 0.0001 cc/min. A
Swagelock filter with a cartridge of 2-micron pores was used to filter out any small
particles that might clog the micromodels. The properties of the fluids used in the
experiments are shown in Table 2.

TABLE 2. FLUID PROPERTIES

Property Water Decane Air

Specific Gravity 1.0 0.73421 0.0012


Viscosity 1.0 cp 0.95 ~0.02 cp

2.7. Experimental Procedure

Experiments at different constant water injection rates into the fracture were
performed. Injection rates were set at two values: one low and one high relative to the
micromodel (0.001 cc/min and 0.01 cc/min). The fracture aperture was constant in
depth and width as per the fabrication etching. This system mimics length, width, and
approximate fracture aperture of the core scale set-up used with the CT Scanner to
image in situ saturation reported in a previous work (Rangel-German and Kovscek,
2001). We performed experiments similar to those at the core-scale trying to achieve
filling-fracture and instantly-filled fracture regimes using the low and high flow rates,
respectively. Figure 3 displays typical core CT experiment results and illustrates the two
different fracture-flow regimes. Both images were taken after 0.32 PV of water had
imbibed into the core. White shading indicates water and black shading indicates air.

Setting the injection flow rate into the fracture, we observed the progress of
multidimensional imbibition at two levels. At the first level, the pore scale was
observed, and at the second level, images of the entire micromodel (macro-scale) were
obtained. For the experiments reported here, water was injected on the left at constant
rate and oil and water produced from the right-hand side. The objective was to have
uniform advance of water along the fracture as was done with the core CT experiments.
We performed experiments with water and air, water and CO2, oil and CO2; and water
and oil. In the case of experiments with oil, CO2 was first injected into the system to
displace any gas and this was followed with oil. The CO2 is relatively soluble in oil and
the combination allowed the micromodels to be fully saturated with oil.

30
3. EXPERIMENTAL RESULTS
Following the aforementioned procedure, experiments were conducted to
achieve filling-fracture and instantly-filled fracture regimes using low and high
injection rates, respectively. The goal of the micromodel experiments was to observe
how the non-wetting phase displacement and matrix-fracture interaction occurred at the
pore level for two cases: when fractures were filling and when they were rapidly filled.

3.1. “Filling-Fracture” Regime

This regime shows a variable length line source due to relatively slow water
flow through fractures. Water advance is controlled by the interaction between the
matrix and the fracture. Figure 4 shows the images obtained of the entire micromodel
for different times for the air-water system. Water is darkly shaded prior to water
reaching the end of the fracture at 49 min., the imbibition pattern is two-dimensional.

The behavior of the filling-fracture regime is very similar to that observed


during core CT experiments reported previously. Micromodel experiments allowed us
to visualize how the non-wetting fluid is expelled within the pore structure.

Figure 5 shows a series of pictures obtained at early stages of the experiment


focusing on the matrix-fracture interface close to the inlet. One can see that a fairly
large amount of water imbibes into the matrix. Air is not trapped or by-passed by the
water front to a sizeable extent, and no air moves counter-currently. That is, no air
bubbles are seen in the watch filled portion of the fracture. Figure 6 shows a sequence
of images of the filling of a section of the fracture close to the outlet. No air is expelled
in the region covered by water but a small amount of air seems to be trapped close to
the outlet. Figure 7 shows images of the water advance in the matrix in the direction
perpendicular to the fracture. The water front has advanced further on the left than it has
done on the right. Figs. 4 and 7 illustrate that there is an elliptical imbibition pattern in
the matrix. This lends support to the assumption of a linear superposition of one-
dimensional solutions of the diffusion equation proposed in the analytical model of
Rangel-German and Kovscek (2001). The filling-fracture regime is dominated by co-
current imbibition as witnessed by little or no escape of air from the region upstream of
the water front in the fracture.

A similar experiment was performed at this low flow rate (0.001 cc/min), CO2
as the gas phase, but having all four ports open. The behavior of this experiment, during
the filling-fracture regime, was found to be almost identical to the previous one. After
the water front reached the end of the fracture, the water imbibed in a one-dimensional
direction (perpendicular to the fracture) displacing the CO2 and displaying a well
defined front parallel to the fracture. In this case, the gas is not trapped (since all the
ports are open) so the imbibition continues in a co-current mode. The advance of the
front was found to be faster than that of the previous experiment (gas 'encapsulated')

31
and very little gas was trapped. These results support the assertion that co-current
imbibition is a more efficient mechanism.

Because the advance of water in both directions (parallel and perpendicular to


the fracture) each scale linearly with the square root of time, we anticipate that the mass
of water imbibed scales linearly with time. That is, the mass imbibed in this two-
dimensional geometry is proportional to the product of these two length scales.

3.2. “Instantly Filled” Regime

In this regime, little water imbibes before the fracture fills with water. Figure 8
includes the images obtained with the digital camera of the entire micromodel for
different times for a water-air experiment. It shows that the water advance in the
direction perpendicular to the fracture is nearly one-dimensional. Another important
feature of this type of displacements is the production of non-wetting fluid through
“channels”. Refer to image labeled 25 min. on Fig. 8, for instance.

Figure 9 shows microscope images during the “Instantly-filled fracture" regime


focusing on the matrix-fracture interface. Almost no water is imbibed while the fracture
fills. The water fills the fracture very fast and the air becomes “encapsulated” inside the
micromodel; therefore, the non-wetting fluid can only be expelled by counter-current
imbibition. Figure 10 shows images of the water advance in the matrix in the direction
perpendicular to the fracture during “Instantly-filled fracture” regime. As it was
observed macroscopically in Fig. 8, there is a well-defined spreading front parallel to
the surface exposed to the imbibing fluid. Figure 11 shows the same water advance at
100X magnification close to the “channels”. One can see that there is a large amount of
gas trapped/bypassed by water in these areas.

Similar experiments to the instantly-filled fracture regime were performed with


water displacing CO2 with very similar results. In this case some of the gas was also
dissolved into the water so we could achieve 100% water saturation. Fig. 12 shows the
trapped gas being dissolved and displaced at the end of this water-CO2 experiment. This
was done with the idea of eventually having a matrix saturated with oil to the
irreducible water saturation. Once the matrix was completely saturated with water, oil
was injected to displace the water.

The last experiment under instantly-filled fracture regime, and probably the
most interesting, was conducted by injecting water into an oil saturated matrix with
irreducible water saturation. Water was injected at 0.01 cc/min and no water imbibed
while the fracture was filling. Once the water front reached the end of the fracture, it
penetrated into the matrix displacing oil by counter-current imbibition. Fig. 13 shows
how the oil is displaced through "channels" at discrete locations along the fracture.
One can see how blobs of oil grew as more oil is pushed into them. Once the blobs have
reached some critical volume they are displaced by the viscous forces in the fracture.
This process was found to be intermittent; i.e., a blob grows, it is displaced, and after

32
some time a new blob starts to grow in the same location. However no constant
periodicity was found.

The behavior of this second regime is very similar to that observed during both
counter current and co-current imbibition experiments reported previously in the
literature (Akin et al., 2000; Handy, 1960; Cil and Reis, 1996; Reis and Cil, 1999; Reis
and Haq, 1999; Rangel-German and Kovscek, 2000(a,b); Li and Horne, 2000; Zhou et
al., 2001; Rangel-German and Kovscek, 2001). The instantly-filled fracture regime is
dominated by counter-current imbibition.

4. FUTURE WORK
As it was mentioned above, the objective of this paper was to prove a
hypotheses on the mode of imbibition for filling- and instantly-filled fractures regimes.
The ultimate goal of this ongoing research is to help in the efforts to derive general
matrix-fracture transfer functions and to provide better recommendations on the use of
assumptions while numerically simulating fractured systems.

One of our biggest concerns is the fact that current numerical


simulators/formulations use transfer functions that include constant shape factors. It is
clear from the results presented in this paper and the work by Rangel-German and
Kovscek (2001) that the use of a constant shape factors for matrix block under filling-
fractures regime will lead to erroneous results; most likely to pessimistic results. We
have observed that the co-current imbibition nature of the filling-fracture regime is a
much more efficient displacement process than the instantly-filled fracture (classic
approach: totally immersed blocks). Mathematical analysis of the filling-fracture regime
and the correct expression for the matrix-to-fracture transfer functions is our current
focus of study.

A second concern is the case when a matrix block is adjacent to two fractures,
but these fractures are not necessarily connected to each other. This is a very common
case of clusters of parallel and disconnected fractures. This situation implies that neither
the pressure nor the saturation within the fractures will be the same as it is supposed in
the classic approach of totally immersed fractures. We believe that constant shape
factors can be used for this kind of problems when dealing with instantly-filled
fractures. However, we also believe that those shape factors will be quite different from
those presented in the literature. This is because having different boundary conditions
would eliminate the symmetry property under which previous shape factors have been
developed. It is readily clear that the shape factors will not be constant when dealing
with filling-fracture regimes.

For the case of parallel (and disconnected) fractures under instantly-filled


fracture regime, we are undertaking the process of improving a method to calculate the
water saturation profile as a function of distance from the fracture based on the
micromodel images. Sets of images where obtained every two patterns (a pattern is the

33
unit used repeatedly to create the micromodels) from the injecting fracture to the
opposite one. Three sets were obtained: one at a quarter of the distance from the inlet to
the outlet; another set in the middle; and one more set at three quarters of the distance
from the inlet to the outlet.

The calculation of the water saturation based on the micromodel images requires
a complex processing of the images. Fig. 14 shows an example of the original image
obtained through the microscope and the post-processing image obtained with the
program written for this task.

The major difficulty to calculate the water saturation is the fact that the images
obtained from the microscope are nearly monochromatic, which means that different
tones of the same color are obtained. This makes it virtually impossible for any code to
distinguish the rock grains from the liquids. However we can divide the images into two
categories: gas and water and rock. Knowing that the micromodels have constant
porosity we can work out the equations to calculate both the gas and water saturations.
The value shown as Sw in the right image of Fig. 14 indicates the ratio of volume of
rock plus water over the bulk volume.

5. CONCLUSIONS
A micromodel and micromodel holder were built that allow detailed
visualization of fluid movement, matrix-fracture interaction, water advance in the
fracture and the pattern, extent, and mode of imbibition in the matrix at the pore scale.
The existence of two different modes of matrix and fracture fill-up found previously
were verified and analyzed at the pore-scale. The ”filling-fracture” regime represents a
slow flow through fractures where fluid transfer is relatively rapid, fracture aperture is
wide, and/or water injection is slow. Recovery scales linearly with time under this
particular co-current imbibition mode. The "instantly-filled fracture" regime represents
a rapid flow through fractures where relatively low rates of fracture to matrix transfer,
narrow apertures, and/or high water injection rates. Recovery scales with the square-
root of time on this counter-current imbibition mode. Experimentally, co-current
imbibition was found to be more efficient than counter-current.

For a low injection rate, virtually all of the non-wetting fluid can be expelled
from the matrix by the time the water reaches the end of the fracture. Although this
filling-fracture regime may take longer than a fast injection/production rate, it
represents a more efficient displacement in terms of recovery. Reinjection in a fractured
reservoir that had an early breakthrough caused by high injection rate can become very
expensive. When the fractures surrounding blocks are filled by water, non-wetting fluid
from these blocks can only be produced by counter-current imbibition.

It was found that the reason for the more efficient displacement of the filling
fracture regime comes from the fact that it takes place in a co-current imbibition mode.

34
Thus, little non-wetting fluid is trapped as the water front advances. Most of the non-
wetting fluid is produced in the non-wetting phase.

Since matrix-fracture interaction is controlled by the injection rate, it is apparent


that the relative permeability relationships are rate dependent. Wetting and non-wetting
fluids will flow simultaneously in both fractures and the matrix, and the combination of
these two flow processes results in a combined relative permeability behavior that has
not been determined. These experiments help in the understanding of matrix-fracture
interaction and maybe useful while modeling real fracture relative permeability and
matrix-fracture transfer functions.

6. REFERENCES
1. Akin, S., Schembre, J.M., Bhat, S.K., and Kovscek, A.R.: "Spontaneous
imbibition characteristics of diatomite," Journal of Petroleum Science and
Engineering, 25 (2000) pp. 149-165.

2. Buckley, J. S.: "Multiphase displacements in micromodels," in Interfacial


Phenomena in Petroleum Recovery, Morrow, N. R. (ed.), Chapter 5, Marcel
Dekker Inc., New York, NY, ISBN 0-8247-8385-9 (1991) pp. 157-189.

3. Cil, M. and Reis, J.C.: "A Multi-dimensional, analytical model for counter-current
water imbibition into gas saturated matrix blocks," J.Pet. Sci & Eng., 16 (1996)
pp. 61-69.

4. George, D.S.: “Visualization of solution gas drive in viscous oil,” M. S. Thesis,


Stanford University (September 1999).

5. Handy, L.: "Determination of effective capillary pressure for porous media from
imbibition data." Pet. Trans. AIME 219 (1960) pp. 75-80.

6. Hornbrook, J.W., Castanier, L.M. and Pettit, P.A.: “Observation of foam/oil


interactions in a new, high-resolution micromodel,” SPE 22631, paper presented
at the 66th Annual Technical Conference and Exhibition of the SPE, Dallas
(October 6-9, 1991).

7. Kazemi, H. and Gilman, J.R.: "Multiphase flow in fractured Reservoirs," in Flow


and contaminant transport in fractured rock, Bear, J.R., Tsang, C. F., and de
Marsily, G (eds.), Academic Press, San Diego (1993) pp. 267-323.

8. Li, K. and Horne, R.: “Characterization of spontaneous imbibition into gas-


saturated rocks.” SPE62552, paper presented at the SPE/AAPG Western Regional
Meeting, Long Beach (June 19-23, 2000).

35
9. Pooladi-Darvish, M. and Firoozabadi, A.: "Cocurrent and countercurrent
imbibition in a water-wet matrix block." Soc. Pet. Eng. J. 5 (2000) pp. 3-11.

10. Rangel-German, E. and Kovscek, A.: “Matrix-fracture interaction in single matrix


blocks.” paper presented at Twenty-fifth Workshop on Geothermal Reservoir
Engineering, Stanford University, Stanford, CA (January 24-26, 2000).

11. Rangel-German, E. and Kovscek, A.: "Experimental and analytical study of water
infiltration in fractured systems," paper presented at the Annual Geothermal
Resources Congress, San Francisco (September 2000).

12. Rangel-German, E. and Kovscek, A.: "Water infiltration in fractured systems:


Experiments and analytical model", SPE 71618, paper presented at the SPE
Annual Technical Conference and Exhibition, New Orleans (October 2001).

13. Reis, J. and Cil, M.: “Analytical models for capillary imbibition: One-dimensional
matrix blocks.” In Situ, 23(3) (1999) pp. 243-270.

14. Reis, J.C. and Haq, S.A.: "Water advance in a single fracture in the presence of
capillary imbibition into adjacent matrix blocks." In Situ, 23, 3 (1999) pp. 271-
295.

15. Sagar, N.S. and Castanier, L.M.: “Oil-foam interactions in a micromodel,” SUPRI
TR 110, Stanford University (November 1997).

16. Zhou, D., Jia, L., Kamath, J. and Kovscek, A. R.: "Scaling of counter-current
imbibition processes in low-permeability porous media," Journal of Petroleum
Science and Engineering, accepted for publication (2002).

36
5 cm
fracture

25 µm

matrix 5 cm

Inlet Outlet

(a) (b)

Figure 1. Schematic of the micromodel: The network acts as the porous medium

and the adjacent channels act as the high conductivity fractures.

Observation Inlet/outlet
Window ports

Inlet/outlet
to MM ports

Perforations for
screws O ring gap

Figure 2. Schematic of the micromodel holder. Top and Bottom pieces.

37
Sw

Figure 3. CT saturation images showing the imbibition regimes: “Filling-fracture" and

“Instantly-filled fracture”. (After Rangel-German and Kovscek, 2001).

17 min 20 min 30 min 49 min

1 hr 13 min 1 hr 53 min 2 hr 45 min 3 hr 5 min

Figure 4. Images of the micromodel showing “filling-fracture" regime.

38
Matrix

Fracture

Figure 5. Microscope images of the micromodel during “Filling-fracture" regime.

Close to inlet.

Figure 6. Filling of a section of the fracture close to the outlet.

39
Figure 7. Water advance in the matrix in the direction perpendicular to the fracture.

10 min 12 min 18 min 23 min

25 min 28 min 45 min

Figure 8. Images of the micromodel showing “Instantly-filled fracture" regime.

40
Figure 9. Microscope images of the micromodel during “Instantly-filled fracture" regime.

Figure 10. Water advance in the matrix in the direction perpendicular to the fracture

during “Instantly-filled fracture” regime.

41
Figure 11. Water advance in the matrix in the direction perpendicular to the fractur

during “Instantly-filled fracture” regime. 10X Magnification.

42
Figure 12. Gas dissolution in water in the matrix.

43
Figure 13. Production of oil by counter-current imbibition of water.

“Instantly-filled fracture” regime.

44
Figure 14. Calculaton of water saturation from micromodel images.

45
46
An Experimental Study of Imbibition in Diatomite and Chalk

S.S. LeGuen, A.R. Kovscek and L.M. Castanier

ABSTRACT

Studying fluid displacement in low permeability rocks allows petroleum


engineers to better handle oil production. So researchers try to understand the
imbibition process which plays a leading role in oil recovery. Thanks to an X-ray
computerized tomography scanner saturation profiles can be drawn for oil-air and
water-oil experiments. A special imbibition cell has been designed in order to hold
diatomite, sandstone and chalk cores. Furthermore, scanning electron microphotographs
and porosity scans were used to interpret our experimental results. Lastly, we confirmed
the established theory that states that recovery increases proportionally with the square
root of time.

1. LITERATURE REVIEW

In order to improve oil recovery, petroleum engineers try to better understand


the process of displacement of oil towards the producing wells. Primary recovery
corresponds to the volume of hydrocarbons produced with the energy available in the
reservoir. Secondary recovery consists of the injection of fluids, which leads to an
increase in the natural energy of the reservoir fluid system. The success of an injection
project depends on well placement, fluid characteristics and formation quality.

This study deals with the comparison of imbibition between diatomite and chalk.
Imbibition defines the process by which a wetting fluid displaces a non-wetting fluid.
The capillary suction forces the wetting fluid to remain around the pore structure of the
rock while the non-wetting fluid is extruded.

In the case of fractured systems, the stronger the imbibition forces are, the more
efficient the recovery becomes. If imbibition cannot play a major part in the waterflood,
water will be led through the fracture network between injector and producer and the
waterflood will be a failure.

After studying the properties of diatomite and chalk, a brief presentation of the
experimental set-up and its features will be made. Then it will be possible to compare
those results with the theory regarding imbibition.

1.1. Diatomite and Chalk

The process of fluid injection such as waterflooding is linked to the fundamental


properties of the porous medium. Imbibition depends on capillary pressure and

47
properties of fluid and rock. Porosity represents the proportion of the void space
available in the rock. Permeability measures the medium’s capacity to transmit fluids.
Saturation is the fraction of gross void space occupied by a fluid. Although diatomite
and chalk look alike because of their color and texture, their compositions and
properties could account for their different behaviour with regard to imbibition.

Diatomite is a sedimentary rock composed of the fossilized skeletal remains of


diatoms, one-celled algae-like plants ranging in size from 10 to 200 microns. The
honeycomb silica structure gives diatomaceous earth a high absorptive capacity,
chemical stability and low bulk density. So its permeability fluctuates between 0.1 mD
and 10 mD and its porosity varies from 35 % to 65 %. The United States is the world’s
largest producer and consumer of diatomite. The major end applications are in filtration,
absorbent and fillers.

Chalk is a porous marine limestone composed of fine-grained remains of micro-


organisms with calcite shells, coccolithophores. Whereas diatomite is a hydrous form of
silica SiO2, chalk consists mainly in calcium carbonate CaCO3. The Austin Chalk of the
United States Gulf coast is a prolific, fractured oil reservoir that spurred widespread
horizontal drilling activity.

1.2. Imbibition

As crucial as the rock characterization may be, the properties of fluids must be
known in order to interpret the efficiency of the spontaneous imbibition. Thus the mass
of water imbibed by the core depends on all the fundamental properties of the rock and
the fluid (Handy, 1960).
1
 2 Pc kΦS w  2 12
m = ρ w A  t (1)
 µw 
In Eq. 1, m is the mass of imbibed water, ρ w is the density of water, A is the cross-
sectional area of the core, Pc the capillary pressure, Φ is the porosity, k is the
permeability, Sw is the aqueous phase saturation, µ w is the wetting-phase viscosity and
t is the time. So, the mass of imbibed water varies according to the square root of time.
Our experiments will be compared to this theory.

2. EXPERIMENTAL APPARATUS AND METHODOLOGY

In this section we briefly describe the equipment and procedures used for our
experiments.

2.1. Experimental Set-up


To obtain saturation profiles and porosity images of cores we use a Picker™ 1200
SX CT-scanner. This give an inside view of the core without stopping the experiment

48
or destroying the sample. The computerized X-ray tomography (CT) scanner is
composed of 1200 detectors located in a circular fashion within the gantry. The X-ray
source rotates around the object in a circular path with an angle of 398°. After
determining the best settings for our experiments, the tube current is 95 mA and the
voltage is 130 kV. The scanner analyses a slice of the core of 5mm thickness. The gray
scale images displayed on the monitor represent CT numbers in Hounsfield units. Those
pictures are processed quantitatively with the software Fpviewer™, which allows
determination of saturation profiles and porosity during the imbibition experiments.

As illustrated in Fig. 1, the whole length of the core is scanned and a series of
scanning pictures show the progress of the fluid inside the rock. To minimize artifacts, a
special imbibition cell was designed as shown in Fig. 2. The cell is composed of two
chambers : the water jacket contains the first chamber which is the core holder. First of
all, the rock is drilled so that a cylindrical core with a circular cross section was built.
The core measures 8 cm in length and 2.6 cm in diameter. In the length, the rock surface
is covered with a layer of epoxy (Epoxy 907, Miller-Stephenson Chemical) and potted
in an acrylic tube. Then the two endcaps of the core holder are fixed on the two clear
areas of the sample, the top and the bottom of the core, which constitutes a hermetic
chamber. The core holder is introduced into the cylindrical water jacket in a vertical
position. Lastly, the second chamber is filled with distilled water and tightly closed. It
ensures a constant temperature in the rock during experiments. The whole system is
mounted on a positioning system on the couch of the scanner. Two positioning lasers
are used to ensure proper centering and horizontality of the experiment.

2.2. Imbibition Experiments

2.2.1. Oil-Air Experiments

First, the core is saturated with oil by injecting n-decane using a liquid
chromatography pump which creates a flow rate of 5 cc/min. The endcap on the bottom
of the core is closed and oil is introduced by the top. Oil goes through the fracture dug
in the endcap, decane is imbibed by the core and the non imbibed oil and air come out
by the other pipe on the top. While the imbibed oil goes down because of gravity, air
has to rise through the pores of the rock to go out the core. This process is called
counter-current imbibition. On the contrary, for the second diatomite sample, oil enters
the core through the bottom and the top pipe is opened so that oil goes up and pushes air
outside. This is called co-current imbibition.

The oil whose properties are mentioned in the Table 1, is recycled to ensure a
constant pressure in the capillary pipes. At the beginning of the experiment, scanning
images are taken every minute for the three first minutes, then it is possible to increase
the interval up to one picture every hour. Actually, the process becomes slower because
in imbibition, displacement of fluid depends on the square root of time.

49
By subtracting CT images, porosity scan and saturation profiles can be created.
It appears that the image of the dry core is essential for these calculations. That is the
reason why before starting any experiment, all cores were placed in a vacuum oven for
24 hours or more in order to obtain completely dry samples.

Equation 2 gives the calculation of the porosity from the CT images :


CTwet − CTdry
Φ= (2)
CToil − CTair

where CT means the CT value for a voxel. The subscripts wet and dry refer to the oil
saturated core and the dry core. The subscripts a and o refer to the CT values of air and
oil phase.

The calculations for oil saturation profiles need to use the formula (Eq. 3):
CTobject− CTdry
So = (3)
CTwet − CTdry

where the subscript object refers to the image being processed. Before beginning water-
oil experiments, it is useful to wait 8 hours to be sure that the core is fully saturated with
oil.

2.2.2. Water-Oil Experiments

Water is then introduced in the same way as oil and the recycling system is
stopped. 2 wt % of potassium bromide or sodium bromide are added to water so as to
better distinguish the oil phase from water on profiles. Given the difficulty of fully
saturating the core by draining off the imbibed oil, it is better to use the Eq. 4 to draw
water saturation profiles:
CTobject− CTwet
Sw = (4)
Φ (CTwater − CToil )

where Φ refers to the porosity scan. The subscripts o and w refer to CT values of the
oil phase and water. The subscripts object and wet refer to the image being processed
and the image of the fully oil saturated core.

Table 1. Fundamental Properties of Fluids.

Water N-Decane Air


CT-number -290 -570 -1275
(H)
µ (mPa.s) 1.0 0.84 0.018
ρ (kg.m-3) 1000 730 1.82

50
3. RESULTS

We first discuss rock characterization via various methods, then the imbibition
results.

3.1. Rock Characterization


3.1.1. Scanning Electron Microphotograph
The scanning electronic microphotograph (SEM) reveals the pore structures of
rocks and the size of diatomite and chalk pores. Put on a piece of carbon, the samples
are covered with a palladium-gold (Pd-Au) coating, which acts as a conductor. The
slightest area on the surface of the samples could be observed and enlarged up to a 100
000 ✕ magnification. On Figs. 3 and 4, the SEM images of diatomite and chalk at
several magnifications 500 ✕ , 2000 ✕ and 5000 ✕ show pore structures that appear
thorny. We can clearly see significant structural differences between the rocks in terms
of pore shape and size.

3.1.2. Porometer

We used a Ruska Universal Porometer (model 1051-801) which is an accurate


apparatus for determining porosity of cores by using the mercury injection method. A
clean dry sample of the cylindrical core with a diameter of 2.5 cm and a length of 3.2
cm is placed in the chamber of the stainless steel pycnometer. After having filled the
pump cylinder with approximately 4 pounds of mercury, the plunger is advanced so that
the capillary pressure reaches 910 psi, which provides the saturation of the sample with
mercury. Mercury is injected at different pressures and the volume of mercury injected
in the core recorded. The scale of volume has been translated in a scale of saturation:
the largest measured volume has been considered as representing a mercury saturation
of the core equal to 0 and a wetting phase saturation equal to 1. Figure 5 shows the
chart giving the capillary pressure versus the wetting phase saturation. The curve shows
a plateau as the Leverett correlation for air and water drainage behavior of clean sands.

3.2. Imbibition Results

Three experiments have been run, two of them with diatomite cores and the last
one with a chalk core.

3.2.1 Saturation Profiles for Diatomite in Counter-Current Imbibition

Small fractures between the rock and the layer of epoxy were visible to the
naked eye. Thus it is not really astonishing to observe that oil enters the rock by the side
where there should be a fracture.

51
Figure 6 shows the saturation profiles for an oil-air experiment. On these pictures
white means the rock is saturated by oil and black means the core is dry. At the
beginning of this experiment, oil enters the core on the whole clear area at the top and
the marked oil front goes down in a piston manner. It can be observed very quickly that
oil has already begun to enter by the side of the core at the same time. Decane is
progressively imbibed in an uneven way until it reaches the middle of the core. Then the
oil front becomes less and less sharp. Some colors appear on the images, which proves
the diffusion to increase.

In Fig. 8, times beneath the images are given in minutes. Water imbibition appears
as a slower process than oil imbibition. For 22 hours, water has spread through the core.

3.2.2. Saturation Profiles for Chalk in Counter-Current Imbibition

The same experiment has been realized with chalk. On Fig. 8, white means oil
saturated core and black means dry core. The front is very sharp at the beginning and
diffusion can be observed after 426 minutes of experiment. As Fig. 9 illustrates, the
water-oil experiment is slower than the oil-air experiment and the full saturation is not
obtained. However, contrary to the imbibition with diatomite, the diffusion is not very
obvious. The two last pictures look the same.

3.2.3. Saturation Profiles for Diatomite in Co-current Process for Oil Imbibition

First, the diatomite core is saturated by oil with a co-current imbibition process.
Decane imbibes the core from the bottom. By putting the oil reservoir at a higher level
than the cell, oil faster flows into diatomite. It reaches the top of the core after 9 hours
as shown on Fig. 10. The oil front is initially sharp and gradually diffuses. After waiting
15 hours to ensure complete saturation of the core, water is introduced by the top. In
Fig. 11 we can see that the water diffusion appears more considerable than the oil
diffusion. The last image has been taken 50 hours after water was introduced so the core
is supposed to be saturated by water. Water imbibition appears slower than oil
imbibition in air saturated cores.

3.2.4. Porosity Scan

Porosity represents the percentage of the void space in the core, Fig. 12. The
average porosities for diatomite cores are 38 % and 56 %, which shows that diatomite
porosity really depends on the chosen sample. For chalk, the average porosity is 16 %.
So, diatomite contains more void than chalk. Furthermore, black points on chalk
porosity scan prove the presence of clay in the rock. So it becomes harder to foresee the
displacement of oil or water in the rock. On the second diatomite core, there is a darker
layer at the end of the core which looks like a fracture. During the experiment, oil and
water go faster into the core through that irregularity, which could be easily seen on the
CT images.

52
3.2.5. Recovery

As Eq. 1 shows, the mass of imbibed fluid is linked to the saturation of this
fluid. Thus, the recovery can be compared to the saturation. By using the saturation
average of each picture, it is possible to draw the recovery versus the square root of
time in order to check theory with experiments. As Fig. 13 illustrates for diatomite, the
relation between recovery and the square root of time is linear in the case of oil-air and
water-oil experiments. Regarding experiments with chalk on Fig. 14, the linear relation
is not so obvious. In the oil-air experiment, the linear property of the curve has been
checked only for two hours. Moreover, the first 30 minutes do not give any results of
recovery. It could be due to the imbibed oil. Actually, decane creates a kind of film on
the top of the core and the flow rate of the pump is not powerful enough to destroy it in
order to let water imbibe the rock. So waiting some hours is necessary to observe water
imbibition.

4. CONCLUSION

The experimental apparatus composed of the CT-scanner and the special imbibition
cell provides the possibility to observe quantitative saturation profiles. These CT images
represent an efficient means to observe fluid displacements inside the rock.

The imbibition process in diatomite appears to be faster than in chalk. The imbibed
fluid enters the rock in piston like manner so the saturation front is sharp. Next diffusion
can be observed while imbibed fluid is going on spreading in the core. Thanks to
calculations based on CT images, theory is confirmed since mass of fluid imbibed
depends on the square root of time.

5. NOMENCLATURE

A cross-sectional area
CT CT number
k absolute permeability
m mass of water imbibed
µ viscosity
Pc capillary pressure
Φ porosity
ρw water density
So oil saturation
Sw water saturation
t time

53
6. REFERENCES

1. Akin, S., Schembre, J.M., Bhat and S.K, Kovscek, A.R.: “Spontaneous Imbibition
Characteristics of Diatomite.” J. Pet. Sci. Eng. 25 (2000) pp. 149-165.

2. Akin, S. and Kovscek, A.R,: “Imbibition Studies of Low Permeability Porous


Media.” SPE 54590, paper presented at the SPE Western Regional Meeting,
Anchorage (May 1999) pp. 26-29.

3. Handy, L.L.: “Determination of Effective Capillary Pressure for Porous Media


from Imbibition Data.” (1960).

4. Zhou, D., Jia, L., Kamath, J. and Kovscek, A.R.: “An Investigation of Counter-
Current Imbibition Processes in Diatomite.” SPE 68837, Soc. Pet. Eng. (2001).

54
X-rays CT-scanner
pump

core

water

computer
reservoir detector

Figure 1. Schematic of the experimental set-up.

Figure 2. Schematic of the imbibition cell.

55
Figure 3. Scanning electron microphotographs of diatomite pore structure
at 500 ✕ magnification (left), 2000 ✕ magnification (center) and
5000 ✕ magnification (right).

Figure 4. Scanning electron microphotographs of chalk pore structure at 500 ✕ magnification (left), 2000 ✕ magnification
(center) and 5000 ✕ magnification (right).

56
1000
900
Diatomite

Capillary pressure, psi


800
Chalk
700
600
500
400
300
200
100
0
0 0.2 0.4 0.6 0.8 1

Wetting phase saturation, fraction

Figure 5. Capillary pressure versus wetting phase saturation


for diatomite and chalk samples.

Oil
saturation

1 3 7 16 41 66 96 130 191 261 349 446 0

Figure 6. CT-derived oil saturation images of oil imbibition in dry core


of diatomite. Times beneath the images are given in minutes.

57
1

Water
saturation

0
2 8 1145 1265 1645 2610 2835 3105 4037

Figure 7. CT-derived water saturation images of water imbibition in oil saturated


core of diatomite. Times beneath the images are given in minutes.

Oil
saturation

1 2 6 14 30 62 126 246 426 685 1346 0

Figure 8. CT-derived oil saturation images of oil imbibition in dry core of


Chalk times beneath the images are given in minutes.

Water
saturation

1 2 15 47 95 187 307 949 1125 0

Figure 9. CT-derived water saturation images of water imbibition in oil saturated


core of chalk. Times beneath the images are given in minutes.

58
1

Oil
saturation

0
1 5 12 30 62 120 210 270 330 450 510

Figure 10. CT-derived oil saturation images of oil imbibition in dry core of
diatomite. Times beneath the images are given in minutes.

Water
saturation

4 14 48 130 240 480 878 1341 1814 2969 0

Figure 11. CT-derived water saturation images of water imbibition in a oil saturated
core of diatomite. Times beneath the images are given in minutes.

59
Diatomite Chalk Diatomite
φ=38% φ=16% φ=56%
1 +0.25
1

0 0 0

Figure 12. Porosity scans for diatomite (left and right) and
chalk (middle) with porosity scale.

1 0 ,1 6

0 ,9
0 ,1 4
Oil recovery, fraction

0 ,8
Water recovery, fraction

0 ,1 2
0 ,7
0 ,6 0 ,1

0 ,5 0 ,0 8
0 ,4
0 ,0 6
0 ,3
0 ,0 4
0 ,2
0 ,1 0 ,0 2

0 0
0 5 10 15 20 0 10 20 30 40 50
S q u a r e r o o t o f t im e , m in ^ 0 .5 S q u a r e r o o t o f t im e , m in ^ 0 .5

Figure 13. Recovery versus square root of time for oil-air experiment (left)
and water-oil experiment (right) with diatomite core.

60
0 ,3
1

0 ,9 0 ,2 5

Water recovery, fraction


Oil recovery, fraction
0 ,8

0 ,7 0 ,2

0 ,6
0 ,1 5
0 ,5

0 ,4
0 ,1
0 ,3

0 ,2
0 ,0 5
0 ,1

0 0
0 10 20 30 0 5 10 15 20 25 30

S q u a re r o o t o f t im e , m in ^ 0 .5 S q u a re ro o t o f tim e , m in ^ 0 .5

Figure 14. Recovery versus square root of time for oil-air experiment (left) and
water-oil experiment (right) with chalk core.

61
62
HOT FLUID INJECTION

63
64
Injecting Steam into Light Oil Diatomite Reservoirs
B. Todd Hoffman

ABSTRACT

Steam injection into heavy oils has been characterized very well over the last 40
years, and while steam has been injected into light oils almost as long, this process is
much less understood. When this lack of understanding is coupled with the
complexities of low permeability fractured reservoirs, even less is known. This study
sheds some light onto the subject at hand through the use of thermal simulation. By
studying the recovery mechanisms, this work shows that thermal expansion of the
hydrocarbon fluids accounts for about two-thirds of the incremental recovery early in
the steam drive; afterwards, the incremental recovery is equally split among thermal
expansion, vaporization, and viscosity reduction. Three separate fluid banks form in
this type of steam injection process: a cold water bank, a combined hot water and
distillate bank, and the steam front. The combined bank is a complex mixture of
hydrocarbon components and condensed steam. Since this work uses diatomite, which
is known for its high porosity, as the fractured medium, the issue of steam front stability
is also considered. Although both analytical and modeling information show that as
porosity increases the stability of the steam front decreases, the actual results show a
small variation between sandstone (25% porosity) and high porosity diatomite (70%)
that is typified by a 10% increase in break through time. The significance of this work
can easily be extended to include all low permeability fractured reservoirs that have low
primary and waterflood recoveries. For appropriate reservoirs, significant additional
recovery could be obtained by implementing a steam drive into a low permeability
fractured reservoir that contains light or medium weight oils.

1. INTRODUCTION

A simulation study of steamflooding diatomite reservoirs that contain light to


medium weight oils has been carried out. The work was accomplished using the
commercial thermal simulator STARS. This paper begins by giving some background
on the problem and why it needs to be studied. Next, the models that are used are
described in moderate detail. Then, the results are displayed and discussed, and finally,
the applicability of this work to other areas is illustrated.

1.1. Background

By way of review, steam injection, as an enhanced oil recovery technique, is


discussed first, and then the challenges of unconventional reservoir are considered.

65
Both steam injection and unconventional reservoirs have a long and interesting history
in the oil industry.

1.2. Steam Injection

In 1952, steamflooding was initiated as an enhanced recovery technique. The


famous pilot at Yorba Linda Field in California began the steam drive process into oil
reservoirs. Through the 1950’s and 1960’s steamflood technology was in the initial
experimental phase with a number of pilot tests. This early work determined that steam
injection is generally most efficient in highly permeable rock (k > 1 Darcy) and thick
sands, usually greater than 30 feet. Once industry had gained enough confidence from
these pilot tests, it was ready to make significant investments in steam flooding. By the
1970’s a number of fields had benefited from steamdrive technology, and production
peaked (Fig. 1) in the mid 1980’s and has since been fairly constant. The important
aspect of Fig. 1 is this production has been almost exclusively from the heavy oil realm
(Blevins, 1990).

Light or medium oil steam flood (LMOSF) also had its roots in California. In
the 1960’s, one of the first LMOSF field trials was initiated at the Brea Field near Los
Angeles. Brea consisted of steeply dipping sands that had an average permeability
around 70 md and a typical porosity value about 22% (Volek and Pryor, 1972). The
development of LMOSF was not as rapid as with heavy oil. Heavy oils benefit
tremendously from steam injection, compared to waterflood, because of the oil viscosity
reduction caused by the increase in reservoir temperature. In many instances, heavy oil
can only be produced with enhanced thermal recovery such as steamflooding. Unlike
heavy oils, light oils typically produce very well from water drives, and because water
drives are seen as less risky with less initial investment than steam injection, LMOSF
were developed more sparsely (Hanzlik, 1981). Nevertheless, there are a number of
applications where LMOSF seem to be a good alternative to waterflooding, for example
highly dipping reservoirs where steam assisted gravity drainage improves the recovery
process, and confined unconsolidated sandstones where residual oil saturations can be
very low, around 5%. Additionally, application in low-permeability highly-fractured
reservoirs, such as diatomite, is also promising (Johnston and Shahin, 1995).

1.3. Non-Conventional Reservoirs

Low-permeability fractured reservoirs contain a large volume of oil around the


world, yet production from these reservoirs has been modest. Primary production is
notoriously low, usually 5% - 10% of the original oil-in-place. Furthermore,
waterfloods in these reservoirs usually suffer from low injectivity, poor sweep, and
injector-to-producer linkage which all tend to degrade the waterflood’s effectiveness.
All of these factors have facilitated an attempt to use enhanced recovery techniques,
such as steam injection, to increase the recoverable reserves. The Teapot Dome Field in
central Wyoming was an early attempt to steamflood a light oil (32º API) in a low-

66
permeability fractured reservoir (Olsen et al., 1993). While this reservoir did not
perform up to initial expectations, it showed that steam injection into these types of
reservoirs is feasible. Also, the importance of steam quality and φ So was more clearly
understood.

One specific type of the low-permeability fractured reservoir is diatomite. Pure


diatomite is a hydrous form of silica composed almost entirely of the skeletal remains of
unicellular aquatic plants called diatoms. These skeletal remains collect at the sea
bottom along with silts, sands, and mud. A general observation is that the fewer the
impurities with the diatoms, the higher the porosity and permeability. Porosity can
range from 25% for the silty diatomite to 65% for the clean diatomite. An interesting
property of diatomite is that it can have very high porosity, but unfortunately, its
permeability is extremely low, ranging from 0.01 – 10 md. This is due to the extremely
small size of the diatoms and diatom fragments. Another important feature of diatomite
is a high pore-volume rock compressibility, yet some formations may contain a
network of permeable natural fractures. This apparent inconsistency is because of the
brittle nature of diatomite when it encounters non-compressive loads. Huge reserves lie
in diatomites in the San Joaquin Valley of California. These diatomite reservoirs
consist of a series of stacked layers separated by virtually impermeable shale layers.
This makes fluid and heat communication among layers in the vertical direction
exceptionally poor (Kovscek et al., 1996 I, II) Diatomite is a complex rock-type that
makes understanding the oil recovery mechanism very challenging.

Steamflood pilots in diatomites have shown positive results for at least three
different operations. First, in the Cymric Field in San Jouquin Valley, CA, cyclic steam
injection has been successfully used to produce a heavy oil (14ºAPI) (Kumar and
Beatty, 1995). In such a reservoir, heavy-oil and very low permeability, there is little to
no primary or waterflood production therefore steamflooding makes an unproductive
field productive. In the South Belridge Field also in the San Joaquin Valley, two
separate continuous steam injection pilots were initiated. One is in the heavy oil part of
the reservoir, and one in a more intermediate weight oil. The heavy oil pilot showed
positive response to steamflooding but was economically unsuccessful due to high
steam to oil ratios (Murer et al., 2000). The pilot in the intermediate part of the reservoir
was very promising. Early results on a single pattern showed that steamflooding could
dramatically improve the amount of oil recovered from waterfloods (Johnston and
Shahin, 1995). A multi-pattern project is currently underway to verify the previous
results and determine the economic viability of a full-scale implementation.

Throughout the U.S. and the world there is a significant amount of light and
medium oil in low-permeability fractured reservoirs. The success of fields like South
Belridge show that a considerable amount of oil can be recovered from this type of
reservoir. However, much of the recovery process is not well known in these reservoir
applications. This study attempts to better describe this recovery process. Even though
a diatomite reservoir is used as a model, this work can be extended to other low-
permeability fractured reservoirs. The only difference with diatomite may be its high
porosity, which will be treated in the next section.

67
1.4. Issues

Three separate subjects are to be covered in this work. First the stability of the steam
front will be analyzed. The second issue will deal with the importance of individual
recovery mechanisms, and the third will show the temperature and fluid profiles for
steam drive in this type of reservoir.

1.5. Steam Front Stability

Steamflooding is considered to be a stable displacement process on a


microscopic and core scale which leads to increased stability on the reservoir scale.
The tendency of steam to “finger” beyond the front is suppressed by condensation of
steam within the finger as it comes in contact with the lower temperature reservoir
(Farouq, 1982). However, when the porosity is high, there is less rock to dissipate the
heat; therefore less suppression of the instabilities. Depending on the porosity and other
reservoir properties, the typical assumption of stability in steamfloods may not be valid.

Burger et al. (1985) published a method that relates steam front stability and
rock properties. It starts with the mobility divided by Darcy velocity for the
displaced fluid, and that is divided by the mobility divided by the Darcy velocity for the
displacing fluid, so that you have the following equation:

k rw
µ w • Vw
> 1 (1)
k rs
µ s • Vs

where the subscripts w and s stand for water and steam, respectively. If the result is
greater than one, then the process is considered stable. Upon assumption that the
relative permeability for the displacing fluid is approximately equal to that of the
displaced fluid, useful information can be drawn from

µ S • VS
> 1 (2)
µW • VW

Burger et al. presented an expression to calculate the ratio of Darcy velocities, VS/VW,
and it is displayed below.

1 − φ ( hc ) s (Th −Tr)
1+
Vv φ ρ v,TrH vTr,Th
= (3)
Vw 1 − φ (hc ) s (Th − Tr )
1+
φ ρ w,TrH vTr,Th

68
where:

φ is porosity,
(hc)s is average heat capacity per unit volume of solid matrix,
Th is steam temperature,
Tr is reservoir temperature,
HTrv,Th is steam vapor enthalpy from reservoir temperature to steam temperature,
ρ v,Th is vapor density at steam temperature,
ρ w,Tr is water density at reservoir temperature,
µ v is vapor viscosity, and
µ w is water viscosity.

This method was used to determine if steamflood in diatomite is a stable


displacement process. The enthalpy and temperature of the saturated steam, along with
the density and viscosity of both water and vapor, are calculated from correlations from
Ejogu and Fiori (1987). The correlation for most of these properties is only a function
of pressure. The viscosity correlations are a function of both temperature and pressure.
Consequently, the results of the equation are only a function of porosity, heat capacity
of the solid matrix, reservoir temperature, and pressure. Reservoir temperature has little
effect on displacement stability. Heat capacity, within its known range, also changes
the calculations only a small amount. Therefore, the stability parameter versus pressure
is plotted for the range of porosities 25% to 65%. The outcomes are presented
graphically in Figure 2, and lead to some interesting results. For a pressure of 500 psi,
the steam front is stable for all porosity values in our range. At 2500 psi, the front will
be unstable for all porosity values. Steam injection pressures ranging from 500 to 1200
psi have been reported for diatomite projects (Kovscek et al., 1996 I, II). Figure 2
indicates that the more porous diatomite layers could be close to the stability limit and
unstable at the largest injection pressures.

Due to this apparent instability in the highest porosities, the issue of steamflood
stability will be examined in detail with the model. Multiple cases with increasing
porosities will be completed and viewed to see if and where instabilities occur. Where
the instabilities occur we will see how they affect our reservoir performance and in what
magnitude.

1.6. Mechanisms

Whether steamflooding is applied to heavy or light oils and high or low


permeability reservoirs, the general mechanisms remain the same. The difference
occurs in the amount a particular mechanism effects a particular hydrocarbon reservoir
(Blevins et al., 1984). Figure 3 qualitatively displays one view of the roles certain
mechanisms play for light to heavy oils. In the literature there is a significant detail
paid to steamflooding mechanisms. Wu (1977) presented a critical review of

69
steamflood mechanisms. He and others have identified the following, which should be
considered a thorough but not exhaustive list:

viscosity reduction;
distillation (vaporization);
distillate (in situ solvent) drive;
steam (gas) drive;
thermal expansion;
relative permeability and capillary pressure variation; and
gravity segregation.

However, it can be seen from Fig. 3 that the three most important mechanisms
are viscosity reduction, thermal expansion and vaporization. In high viscosity
reservoirs, the objective of steamflooding is to increase oil production by reducing oil
viscosity; thus allowing oil drainage at significantly higher rates. Conversely, for low
viscosity oils the primary objective is to reduce residual oil saturation below that
obtainable by ordinary waterflooding, which is primarily accomplished through
vaporization. In fractured systems or highly heterogeneous reservoirs, thermal
conduction allows heat to sweep areas of the reservoir not contacted by steam. In this
case, thermal expansion is an important recovery mechanism.

While Fig. 3 presents a nice qualitative description of the recovery mechanisms


for a general reservoir description, the relative importance of each mechanism is only
speculated. This work attempts to quantify the relative importance of each mechanism
for LMOSF in diatomite. It is important to understand the role of each mechanism
because that information gives insight on how to best develop the steamflood.

1.7. Saturation Profiles

When steam is constantly injected into a continuous oil-bearing reservoir layer,


several separate zones develop in the reservoir. Figure 4 displays one interpretation of
how this profile may form. It includes four zones (steam, distillate, hot water, and oil)
in a manner that only slightly takes into account gravitational forces. The way these
zones are formed is of particular interest to LMOSF. When the initial steam contacts
the cold reservoir, the steam is condensed to hot water. As the reservoir is heated up,
some steam will remain in the vapor phase and a steam zone begins to develop. In this
steam zone, the high temperature causes some of the hydrocarbons to be vaporized. As
more steam is injected, the vaporized oil is carried forward along with the steam. A
portion of the steam and distillate condense at the steam front causing a distillate bank
to form.

Remembering that Fig. 4 is only one interpretation for a general reservoir, this
research will attempt to find how the zones develop for our particular model. In a
diatomite reservoir, each discrete layer is relatively thin so that gravitational forces do
not greatly affect the profile; therefore, gravity can be ignored in this case. Because the

70
well spacings are close in most diatomite reservoirs (85 - 340 ft.), it has been
hypothesized that the distillate zone does not have enough time to form. The critical
time and length the specific zones begin to occur and when each particular zone reaches
the producer will be determined. Also, we will find the size of different zones and if all
the speculated zones actually develop.

1.8. Model Framework

The simulator STARS from the Computer Modeling Group will be used to perform the
runs. The following sections layout the framework for the model that will be used. At
first describing the general rock and fluid properties along with the initial conditions.
Then the 4 different reservoir descriptions will be covered, and last to be covered will
be what is changed for the different issues being studied.

1.9. General Set-up

This model attempts to represent a injection-production well pair development


typically seen in diatomite reservoirs. Wells are 116.5 feet apart from each other.
Much of the rock and fluid data used in this model comes from a previous reservoir
model AERA Energy LLC had used. The fluid is characterized by three hydrocarbon
components, a light, medium and heavy component and a water component. The fluid
properties are developed from upscaling crude oil compositional analysis and fluid
property evaluation. Table 1 shows the properties for each component.

Table 1. Component Fluid Properties


W ater Heavy Medium Light
Comp. Comp. (1) Comp. (2) Comp. (3)
Molecular W eight 18 530.7 152.7 27.89
Critical Temperature 705 10000 458.2 550
Critical Pressure 3217 10000 710 728
Density 62.4 58.51 52.28 34

Thermal expansion, liquid compressibility, and viscosity are all treated with
equations using inputs for each component. Mole-fraction weighted oil viscosity as a
function of pressure for the original composition is shown in Fig. 5.

Relative permeability for the oil-water and liquid-gas relationships is shown in


Fig. 6 as the black lines. This relative permeability curve comes from one of twenty-
eight different diatomite packages represented in the AERA model. In the that model,
relative permeability was adjusted to match historical production data. The basecase
curve roughly represents a median of the 28 separate curves. A second set of relative
permeability curves was used for a sensitivity analysis, and they are displayed in Fig. 6
as the lighter curves. These curves represent the oil-wet end of the 28 available curves.

71
Initial pressure of the system is equal to 400 psi and initial oil saturation is equal
to 0.50. The product of oil saturation and porosity is equal to 0.25, indicating that this
is a good candidate for an enhanced oil recovery process. This is typical of a system
that has produced on before steam injection commences. Also, there is no initial free
gas saturation present; however, for a sensitivity case one run was completed with some
initial free gas. The difference between the injection pressure and producing bottom-
hole pressure (BHP) was held constant at 725 psi. In the 2 dimensional cases the
injection pressure is 800 psi while the BHP is 75 psi. For the 1D case the injection
pressure is 1125 and the BHP is equal to 400 psi. The 1D case’s BHP had to be set to
the initial pressure to prevent pressure depletion by the producer, which would have
confused interpretation of mechanisms. The quality of the injected steam is 0.70.
Initial reservoir temperature is 100ºF and increases to above 500ºF near the injector.
The duration of the simulation runs are different for different models, but all are at least
1.0 hydrocarbon pore volume injected (HCPVI) and typically around 1.5 HCPVI.

1.10. Model Dimensions

Four separate model descriptions were used. The first was a 1D model with
homogeneous permeability (2 md) and porosity (50 %). There is one injector and one
producer that are 116.5 feet apart. The model is discritized with 50 blocks. The 1D
model is used as part of the mechanisms study and to determine the saturation and
component profiles.

The second model was also homogeneous with the same permeability and
porosity as the 1D case, but it is 2D in the x-y direction. Again, the two wells are 116.5
feet apart, but in this case, one half of a hydraulic fracture (136.5 feet) in the orthogonal
direction of the wells is also included in the model (Patzek, 1992). The total length of
the model in that direction is 166.5 feet (Fig. 7).

There are 2500 total gridblocks; 50 in the x-direction and 50 in the y-direction.
The height of the model is constant at 2.33 feet, so that we are modeling only one thin
layer in a series of possibly 30 layers in the actual reservoir. This 2D model was used
for the mechanism study, and the stability runs were completed on this model as well,
however due to the homogeneous nature of the model, little information is gained from
the stability element.

The third model is very similar to the second model (2D and same size), but the
permeability and porosity are variable. The porosity is created using unconditional
sequential Gaussian simulation (SGS) with a mean of 0.50, and a standard deviation of
0.20, and a minimum cut-off value of 0.25 (Fig. 8). In the SGS, no preferential
direction was selected for the porosity. The permeability is calculated from the porosity
using the following equation:

k = 100(φ − 0.25) 3.5 (4)

72
and its mean was 1.8 md. Current thought says that this type of model is most
representative of what is actually in a layer of a diatomite reservoir. This 2D
geostatistical model gives good insight to both the stability issue and the importance of
the various mechanisms.

The last model is a 2D geostatistical model with preference in the y-direction, a


“thief zone” model. The mean and standard deviation of both the porosity and
permeability were the same as the isotropic case, it is just distributed in a different
manner. The porosity model is displayed in Fig. 9. This model represents a common
feature of diatomite reservoirs, injector to producer link up. In reservoirs this is
modeled here as high permeability streaks. This model is a description of reservoir
conformance caused by injection induced fractures, but it is; here it signifies poor areal
conformity, but it could also represent unconformities in the vertical direction, which
are known to exist. This model offers some information about the mechanisms and it is
very important when looking at the stability of the front.

1.11. Stability

Stability runs were made for all three 2D cases, however the two geostatistical
cases provide the most information. For each model description, three separate runs
were completed. They were exactly the same except for their porosity. One was a
sandstone case with porosity of 25%; one was the diatomite basecase with 50%
porosity, and one was a high porosity diatomite at 70 %. For the homogeneous case
these values were constant and for the geostatistic cases these were the mean porosity
values. Even though the porosity changed for each run, permeability remained constant
so that the flow paths would not be affected. To insure a conservative solution, all the
heat injected into the layer, stayed in the layer (heat loss to the over/underburden was
turned off). The three cases are visually compared to see if the higher porosity cases are
less stable.

1.12. Mechanisms

Three primary recovery mechanisms (vaporization, thermal expansion and


viscosity reduction) are compared in all four model descriptions to see the relative
importance of each mechanism in LMOSF in diatomite reservoirs. Five separate runs
are needed for each description to quantify the impacts of each mechanism. One run
where all the mechanism are turned on is the basecase. Then each mechanism is turned
off individually and the production from the subsequent runs is less than the basecase.
The difference between the run and the basecase shows the importance of that
mechanism for that case. Finally, a waterflood case, representing the low end case, will
serve as a base line for this study. The difference between the basecase and waterflood
case should be very close to the sum of the three primary mechanisms. The small
differences may be explained by secondary mechanisms, such as relative
permeability/capillary pressure affects. These mechanisms along with others like

73
thermal expansion of initial gas saturation are examined but in less detail than the three
primary mechanisms. The different mechanisms are all compared on a HCPVI basis.

2. RESULTS

Data for the three primary study areas is set out in the following sections. The
stability of the steam front will be discussed first, followed by the relative importance of
the three mechanisms, and end with a discussion on the profile for the three phases and
the three component.

2.1. Stability

Stability cases were run on all three 2D models which resulted in some
interesting findings. As hypothesized, instabilities did increase as the porosity
increased; however, for these well conditions and reservoir properties, the instabilities
are not very significant. Due to the dominating hydraulic fractures and close well
spacing, finger growth is similar for both the average porosity sandstone case and the
high porosity diatomite case. However, as the geologic descriptions became more
complex and the relative permeability became more oil-wet, instabilities were more
prevalent and there existed slightly larger differences between the sandstone case and
high porosity diatomite case.

2.2. 2D Homogeneous

A quick look at the 2D homogeneous case (Fig. 10) shows at best a very small change
between the three porosity runs. As porosity increases the front appears to be less
smooth; however, no significant streaks of the displacing fluid are noticeable. In
simulation of pure homogeneous rocks, instabilities tend not to occur because there are
no permeability differences where a finger can begin to grow, therefore, we did not
expect to see significant differences among the three porosities. The representations in
Fig. 10 are all at a HCPVI of 0.15.

2.3. 2D Isotropic

In a pure homogeneous reservoir, porosity will not affect the stability of displacement in
the reservoir; however, very few reservoirs can be treated as homogeneous. When even
small heterogeneities exist in the reservoir, instabilities may begin to occur. The next
case to be examined is the isotropic permeability description (Fig. 11). In this case, the
graphics are shown at a HCPVI of 0.24. When looking at these three runs, we start to
see some of the effects of the porosity on stability. In the high permeability regions,
displacement tends to finger more for the high porosity cases; additionally, in the low
permeability regions (the two unswept regions in the bottom-right) more oil has been

74
recovered in the low porosity case than in the higher cases. This shows that the effects
of heat on oil recovery are more efficient for low porosity systems.

2.4. 2D Thief

Now let us move to the thief case where, as one might expect, there is the greatest
amount of instability. In addition, this case has the largest difference between the
sandstone porosity run and the high porosity diatomite run (Fig. 12). The following
plots are again all shown at 0.15 HCPVI. For the lowest porosity, water is just breaking
through at the producer at 0.15 HCPVI, while for both of the higher porosity runs, water
has broken through earlier. For the 50% mean run, breakthrough occurs at 0.13 HCPVI
and for the 70% mean run, it is 0.11 HCPVI. When comparing these runs it is easy to
visualize the difference between all three different porosity cases. By examining the
high permeability streak on the left, you can see that the front is progressively farther as
the porosity increases.

2.5. Relative Permeability Sensitivity

When the more oil-wet relative permeability curves were used, instabilities
increased for all model descriptions, and of course this is most apparent in the case with
the highest heterogeneity, the thief case. Figure 13 shows displacement for the three
porosity values at a HCPVI of 0.12. When this is compared with the thief case for the
original relative permeability data (Fig. 12), all runs are a little less efficient. However,
for the new relative permeability cases, porosity still does not play an important role.

2.6. Summary

Table 2 compares breakthrough time for all three model descriptions and for the
three different porosity cases. To be able to compare the three model descriptions with
each other the last column compares the higher porosity diatomite runs to the sandstone
porosity runs as a percentage. As you can see, the biggest difference is for the most
heterogeneous case.

Table 2. Breakthrough Time Comparisons in HCPVI for the Three Increasing Porosity Cases

% before Sandstone
25 % 50 % 70 % 50 % 70 %
Original 2D Homogeneous 0.28 0.26 0.26 95% 94%
Relative 2D Isotropic 0.27 0.26 0.26 95% 93%
Permeabilty 2D Thief 0.15 0.13 0.11 87% 73%
2D Homogeneous 0.23 0.22 0.21 93% 88%
New Relative
2D Isotropic 0.16 0.14 0.13 93% 86%
Permeabilty
2D Thief 0.10 0.08 0.07 85% 72%

75
Overall, there were noticeable differences between the stability of the displacing
front as the porosity increased. However, the difference varied considerably depending
on geologic description and relative permeability conditions. The largest difference
observed was in the thief cases where the high porosity runs broke through in about
three-quarters of the time the sandstone runs took.

2.7. Mechanisms

All four reservoir descriptions evaluate the mechanisms that were most
important for oil recovery when injecting steam into diatomite reservoirs. Each
description gives slightly different results as to which of the three primary recovery
mechanisms is most important. These differences along with the similarities will be
discussed in detail in the following sections. To begin, a couple generalities can be
asserted. Early in the injection (less than 0.30 HCPVI) thermal expansion is the
dominant mechanism, but as total injection increases the recovery is split more equally
among vaporization, thermal expansion and viscosity reduction.

Two important aspects need to covered before examining the results. First, the
waterflood recoveries are higher than would be expected from typical diatomite
reservoirs. In the models, the waterflood recovery is as high as 50 percent; whereas,
typical diatomites usually produce less than 20 percent. One major difference between
this model and the real reservoirs is that we have only one layer while the reservoir is
many stacked layers. Those stacked layers have large permeability variations. One
layer may show good waterflood recovery while the remaining layers demonstrate
poorer recovery. It is hypothesized for steam injectionthat heat from the high
permeability layers conducts into the low permeability regions and helps recover the
fluid that is not swept within the lower permeability layer. This should increase the
steamflood recovery relative to waterflood. To confirm this hypothesis, future work
will include many stacked layers in a model.

The other key point is with the treatment of the results. As previously stated in
the Model Set-up section, the summation of the effects of all three mechanism does not
exactly equal the difference between the thermal basecase and the waterflood. This
difference can be attributed to other secondary mechanisms and/or the non-
independence of the three primary mechanisms. This difference is represented by
“Other” in the mechanism results plots and is basically insignificant for all but the 1D
case.

2.8. 1 Dimension Homogeneous

The 1D description will be discussed first. Figure 14 shows the cumulative oil
recovery for all the cases. The top line is the basecase thermal run and the lowest is the
cold waterflood recovery. The three curves between the waterflood and basecase
represents the recovery of the thermal case without that mechanism included.

76
Although this plot shows the ultimate recovery for each run without a particular
mechanism, it is hard to see the relative importance of each one; therefore, plots like
Figure 15 were created to better visualize the recovery. In this plot the incremental
recovery of steamflood above waterflood is displayed, and the relative importance of
each mechanism is also illustrated.

For this case, there was a rather large difference between the summation of the
support from the three mechanisms and the total incremental recovery over waterflood.
This discrepancy is noted by the large “Other” category. This phenomenon was only
observed in the 1D case and can most likely be attributed to the sharp change in the
relative permeability from the water-oil curves to the liquid-gas curves. To confirm this
idea, an additional run was completed where the relative permeability curves were
changed to straight line curves (keeping the end points the same). In this case the
“Other” mechanism has increased which shows that sharper changes in the curves lead
to increased relative permeability effects.

Ignoring the other affects, thermal expansion is the dominant recovery


mechanism for the first 0.50 HCPVI. The large peak around 0.40 HCPVI corresponds
to water breakthrough at the producer. After that, vaporization plays an increasingly
more important role. This is because the distillate bank breaks through after the water
bank. The heavier components begin to mobilize and viscosity reduction plays a
significant role after water breakthrough, as well. The oil modeled has a significant
proportion of heavy ends (14%), so that the relative importance of viscosity reduction is
expected.

2.9. 2D Homogeneous

The next reservoir description to be examined is the 2 dimensional


homogeneous model. The incremental plot is shown in Fig. 16. As noted before, the
“Other “ mechanism is much smaller for the 2D case. Once again, thermal expansion is
the primary recovery mechanism for the first 0.50 HCPVI, and vaporization and
viscosity reduction increase after water breakthrough.

2.10. 2D Isotropic

It was expected that as the models became more heterogeneous and complex, there
would be an increase in separation between the thermal case and the waterflood case;
however, this expectation did not hold true. As Fig. 17 demonstrates, the waterflood
recovery is still above 50 percent, and the thermal basecase recovery has come down
about 15 percent from the 1D case. There are still very efficient sweeps for the
waterflood case. The wells are all fracture stimulated and this is modeled as shown in
Fig. 7 in the Model Set-up section. These fractures extend the majority of the model
width, giving a very good initial displacement front for the waterflood. This type of
displacement front is too optimistic and provides another reason for the waterflood to be

77
higher than expected. In future work, the fracture length will be shortened as a
comparison to the current work to understand the sensitivity of fracture length.

The incremental plot is quite similar to the previous two models as far as relative
importance of the mechanisms (Fig. 18). Thermal expansion is most important early in
the steam flood life. After 0.70 HCPVI, the importance of viscosity reduction and
thermal expansion are constant and the importance of vaporization increases at a
constant rate as the distillate bank moves through to the producer.

2.11. 2D Thief

The major difference between the thief case and the previous cases is the water
breakthrough time is much quicker (around 0.15 HCPVI) and the incremental peak
associated with the water breakthrough is much smaller (Fig. 19). Nevertheless, the
principal recovery mechanism early on is still thermal expansion of the oil phase. In this
case the importance of all three mechanisms are fairly equally distributed for the middle
time data than in the previous models. For the data after 1.0 HCPVI, viscosity
reduction plays an increasing role as the heat migrates to the unswept regions. On the
contrary, vaporization is not as significant as it was in the previous models. This is
because the majority of the injected fluid channels through the high permeability streaks
and there is less area in contact with the high temperature steam that causes the
vaporization.

2.12. Summary

Figure 20 is a summary of the previous incremental plots at two different


hydrocarbon pore volumes injected on a cold water basis. The top bar graph (0.30
HCPVI) shows the general observation that thermal expansion is the most important
recovery mechanism in these reservoirs early in the injection. For all the 2D cases,
thermal expansion accounts for more than half of the incremental oil recovery. Later in
the injection time (1.0 HCPVI), each mechanism contributes approximately equal
amounts to the recovery. As more heat is injected, more of the heavy ends are
mobilized; consequently, viscosity reduction goes through the largest increase between
the two injection times.

This is a dynamic process that changes dramatically as time and reservoir


description change, and therefore it is difficult to pinpoint the exact relationship
between mechanisms and recovery. However, it should be noted that for most
diatomite systems the target total steam injection for the field is around 0.30 HCPVI.
This is because the permeability and therefore the injectivity is very low in diatomites.
Assuming that this criterion is proper, thermal expansion will be the primary recovery
mechanism for diatomite reservoir containing light/medium weight oils.

78
2.13. Profiles

The 1D case was used to characterize the saturation, composition and


temperature fronts as they moved through the homogeneous reservoir. Figure 21 shows
the saturation profile for the water saturation (solid line) and minus the steam
saturation (dotted line) at four injection times. The difference between the two curves
represents the total hydrocarbon saturation. For 0.33 HCPVI, about one-quarter of the
model is still at original reservoir conditions, which are 50 % water and 50 % oil.

These saturation profiles clearly display three separate saturation fronts.


Starting farthest from the injector there is a cold water bank that breaks through to the
producer some time between 0.33 HCPVI and 0.67 HCPVI. Closer to the injector is a
mixed distillate and hot water bank. It can be seen most clearly for at 1.00 HCPVI
where it forms from 15 to 42 distance units. This bank breaks through between 1.00
and 1.19 HCPVI. Finally, closest to the injector, a steam bank forms and is shown here
as the dotted line. In this simulation run, the steam bank has not yet made it to the
producer. The small dip in the water saturation curve seen in front of the hot water
bank corresponds to a distillate bank that is formed by component one, and the large dip
seen behind the hot water bank corresponds to another distillate bank formed by the
medium weight component. In the simplified 3 component system, there is some
distillate produced before the hot water bank and some after. In the complex reservoir
where there is more than 3 components, there is probably no distinguishable boundary
between the hot water and distillates. Most likely there is just one mixed zone that has
lighter ends closest to the producer and the heaviest ends closest to the steam front.

Figure 2 provides the information to be able to determine how the hydrocarbons


are distributed in the reservoir and consequently where the banks are forming. It is a
plot of the fractions of each composition, again at the four different HCPVI. The peaks
of both components 2 and 3 relate to the forming of their respective distillate banks.
This can be seen in Fig. 21 where there are increases in the hydrocarbon fraction. Each
individual bank is not constant in character either. The end of the bank closest to the
steam zone is primarily made up of vapor phase while the end closest to the producer is
primarily in the liquid phase.

Another concern was the timing of the bank formation; this plot shows
unmistakably that the banks occur very early in the injection time and fairly close to the
injector. The final figure shows the temperature profile in the reservoir at the four
different HCPVI. There is a smooth and rather rapid change from original reservoir
temperature to the temperature at the injector. It can be seen that where the temperature
change occurs is also where the distillate and hot water bank is forming in the previous
plots.

79
2.14. Conclusions

A pattern reservoir simulation study of steam injection into a fractured diatomite


formation has been completed. The information gained includes a better understanding
of steam front stability in high porosity mediums, primary recovery mechanisms in
steam drives of light and medium oils, and front profiles in steam drives. The model
consisted of two hydraulically fractured wells (one injector and one producer) that are
set 116.5 feet apart from each other. There are multiple reservoir descriptions that vary
from homogeneous to quite heterogeneous. They all contain the same three-
hydrocarbon-component fluid. Viscosity, relative permeability and other rock and fluid
properties come from a reservoir model developed by AREA Energy, LLC.

Analytic work showed that for high porosity, steam fronts would become
unstable, and while simulation work confirmed this, the overall effects were minimal.
In the most dramatic case, the 2D thief case, the breakthrough time decreased from 0.15
HCPVI for the sandstone case to 0.11 HCPVI for the 70 % porosity diatomite case.

For these reservoir descriptions and fluid properties, thermal expansion of the
hydrocarbon fluids accounts for about two-thirds of the incremental recovery until
breakthrough of the cold water bank to the producer. At which point vaporization,
viscosity reduction, and thermal expansion all contribute approximately equal amounts
to the recovery.

A cold water bank breaks through to the producer first. Then a mixture of
distilled and condensed hydrocarbons and heated water forms another bank, which is
closely followed by the steam front. This second bank is a complex combination that is
characterized by the lightest hydrocarbon component at the front and the heaviest
vaporized components closest to the steam front.

Diatomites are one type of low permeability fractured reservoir where steam
injection can improve light to medium weight oil recovery; however, other low
permeability fractured reservoirs that suffer from the same problems as diatomites such
as below average primary recovery and poor waterflood performance, can have
improved performance with steam injection. Injecting steam into low permeability
fractured reservoirs that contain light and medium weight oil has been tried on a limited
basis, and is being considered in additional areas

80
3. REFERENCES

1. Blevins, Ted R.: “Steamflooding in the U.S.: A Status Report,” J. Pet. Tech. (May
1990) pp. 548-554.

2. Blevins, T. R., Duerksen, J. H., and Ault J. W: “Light-Oil Steamflooding—An


Emerging Technology,” J. Pet. Tech. (July 1984) pp.1115-1122.

3. Burger, J., Sourieau P., and Combarnous, M.: Thermal Methods of Oil Recovery,
Editions Technip, Paris, France, 1985, pp. 89-125.

4. Ejogu, G.C., and Fiori, M.: “High-Pressure Saturated-Steam Correlations,” J. Pet.


Tech. (Dec. 1987) pp. 1585-1590.

5. Farouq Ali, S. M.: “Steam Injection Theories—A Unified Approach,” SPE 10746,
paper presented at California Regional Meeting in San Francisco (March 24-26,
1982).

6. Hanzlik, Edward J.: “Steamflooding As an Alternative EOR Process for Light Oil
Reservoirs” SPE 10319, paper presented at the 56th Annual Fall Technical
Conference and Exhibition, San Antonio (Oct. 5-7, 1981).

7. Johnston, R.M., Shahin, G.T.: “Interpretation of Steam Drive Pilots in the Belridge
Diatomite,” SPE 29621, paper presented at Western Regional Meeting, Bakersfield
(March 8-10, 1995).

8. Kovscek, A. R., Johnson, R. M., and Patzek, T.W.: “Interpretation of


Hydrofracture Geometry During Steam Injection Using Temperature Transients I.
Model Formulation and Verification,” Marcel Dekker, Inc. (1996) pp. 251-285.

9. Kovscek, A. R., Johnson, R. M., and Patzek, T.W.: “Interpretation of


Hydrofracture Geometry During Steam Injection Using Tempreature Transients II.
Asymmetric Hydrofractures,” Marcel Dekker, Inc. ,1996, pp. 289-309.

10. Kumar, M., Beatty, F.D.: “Cyclic Steaming in Heavy Oil Diatomite,” SPE 29623,
paper presented at Western Regional Meeting, Bakersfield (March 8-10, 1995).

11. Murer, A.S., et al.:“Steam Injection Project in Heavy-Oil Diatomite,” SPE 60853,
paper orginally presented as paper SPE 38302 at Western Regional Meeting, Long
Beach ( June 25-27, 1997).

12. Moritis, Guntis: “Biennial EOR Production Report – EOR weathers low oil
prices”, Oil and Gas Jour., Tulsa (March 20, 2000) p. 15.

81
13. Olsen, D. K., et al: “Case History of Steam Injection Operations at Navel
Petroleum Reserve No. 3, Teapot Dome Field, Wyoming: A Shallow
Heterogeneous Light-Oil Reservoir,” SPE 25786, paper presented at Internation
Thermal Operations Symposium in Bakersfield (Feb. 8-10, 1993).

14. Patzek, T.W.: “Surveillance of South Belridge Diatomite,” SPE 24040, paper
presented at Western Regional Meeting, Bakersfield (March 30-April 1, 1992).

15. Volek, C. W., and Pryor, J. A.: “Steam Distillation Drive—Brea Field,
California,” J. Pet. Tech. (Aug. 1972) pp. 899-906.

16. Wu, Ching H.: “A Critial Review of Steamflood Mechanisms,” SPE 6550, paper
presented 47th Annual California Regional Meeting, Bakersfield (April 13-15,
1977).

82
Steamflooding History

500000
450000
400000

Production B/D
350000
300000
250000
200000
150000
100000
50000
0
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000
Year

Figure 1. Steamflood Production from initiation at Yorba Linda through 2000


(Moritis, 2000)

Stability of Steam Drive in Diatomite


(Porosity)

6
φ = 0.25

5 φ = 0.35
µ w*Vw
µ v*Vv

4
φ = 0.45

3
φ = 0.55

2
φ = 0.65

0
0 500 1000 1500 2000 2500 3000
Pressure (psi)

Figure 2. Steam drive stability in diatomite.

83
Vaporization
Viscosity
reduction

Thermal Wettability
Expansion

Water/Oil interfacial tension

Figure 3. Contributions of different mechanisms to the improvement


of oil displacement by a heated fluid instead of cold water (Burger et
al., 1985)

Heat Losses

Steam Zone Oil

Oil Distillate Hot Water


Bank Zone

Figure 4. Schematic of light-oil steamflooding (Blevins et al., 1984)

84
10.0

Viscosity (cp)

1.0

0.1
0 100 200 300 400 500 600 700
Temperature (ºF)

Figure 5. Viscosity as a function of temperature.

basecase sensitivity

1.0 1.0

0.8 0.8

0.6 0.6 kr
kr
o w
0.4 0.4

0.2 0.2

0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Sw

basecase sensitivity
1.0 1.0

0.8 0.8

0.6 0.6 kr
krl
g
0.4 0.4

0.2 0.2

0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Sl

Figure 6. Relative Permeability.

85
Producer

166.5’

116.5 Fractures

136.5
Injector

Figure 7. General model schematic.

Figure 8. Porosity for geostatistical model. The injector is at the lower left corner, while the
producer is in the upper left corner.

86
Figure 9. Porosity for geostatistical thief model.

87
Figure 10. Front comparisons for three different
porosities in 2D homogeneous case. All figures
correspond to 0.15 HCPVI on a cold water basis.

88
Figure 11. Front comparisons for three different
porosities in 2D isotropic permeability case. All
figures correspond to 0.24 HCPVI on a cold water
basis.

89
Figure 12. Front comparisons for three
different porosities in 2D thief permeability
case. All figures correspond to 0.15 HCPVI
on a cold water basis.

90
Figure 13. Front comparisons for three different
porosities in 2D thief permeability case with
more oil-wet relative permeability curves. All
figures correspond to 0.12 HCPVI on a cold
water basis.

91
Cumulative Production 1D
Basecase No Therm Exp No Vap No Vis Red W aterflood

1.0
0.9
0.8

Fraction of Recovery
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
HCPVI

Figure 14. 1 dimension oil recovery for waterflood,


steamflood and steamflood without each mechanism.

Incremental Thermal Recovery - 1D

0.25
Vaporization
Thermal Expansion
0.20 Viscosity Reduction
Other
Fraction of Recovery

0.15

0.10

0.05

0.00
00

08

16

24

32

40

48

56

64

72

80

88

96

04
0.

0.

0.

0.

0.

0.

0.

0.

0.

0.

0.

0.

0.

1.

HCPVI

Figure 15. Incremental oil recovery for steamflood over waterflood in 1D.

92
Incremental Thermal Recovery - 2DHomog

0.14

0.12 Vaporization
Thermal Expansion
0.10 Viscosity Reduction

Fraction of Recovery
other
0.08

0.06

0.04

0.02

0.00
0

10

20

30

40

50

60

70

80

90

00

10

20

30

40

50
0.

0.

0.

0.

0.

0.

0.

0.

0.

1.

1.

1.

1.

1.

1.
HCPVI

Figure 16. Incremental oil recovery for steamflood over waterflood in


2D. Homogeneous case.

Cumulative Production 2D Isotropic

Basecase No Vap No Therm Exp No Vis Red W aterflood

0.8

0.7

0.6
Fraction of Recovery

0.5

0.4

0.3

0.2

0.1

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
HCPVI

Figure 17. 2 dimension oil recovery for waterflood, steamflood and


steamflood without each mechanism with isotropic permeability model.

93
Incremental Thermal Recovery - 2D Isotropic

0.16
Vaporization
0.14 Thermal Expansion
Viscosity Reduction
0.12 Other

Fraction of Recovery
0.10

0.08

0.06

0.04

0.02

0.00
0

5
0.

0.

0.

0.

0.

0.

0.

0.

0.

0.

1.

1.

1.

1.

1.

1.
HCPVI
Figure 18. Incremental oil recovery for steamflood over waterflood in 2D.
Isotropic case.

Incremental Thermal Recovery - 2D Thief

0.16

0.14 Vaporization
Thermal Expansion
0.12 Viscosity Reduction
Fraction of Recovery

other
0.10

0.08

0.06

0.04

0.02

0.00
0

5
0.

0.

0.

0.

0.

0.

0.

0.

0.

0.

1.

1.

1.

1.

1.

1.

HCPVI

Figure 19. Incremental oil recovery for steamflood over waterflood in 2D.
Isotropic case.

94
HCPVI = 0.30

0.16

other
0.14
viscosity reduction
vaporization
0.12 thermal expansion

Fraction of Recovery
0.10

0.08

0.06

0.04

0.02

0.00
Theif Isotropic Homogeneous 1D

HCPVI = 1.0

0.20

0.18 other
viscosity reduction
0.16 vaporization
0.14 thermal expansion
Fraction of Recovery

0.12

0.10

0.08

0.06

0.04

0.02

0.00
Theif Isotropic Homogeneous 1D

Figure 20. Incremental oil recovery for steamflood over


waterflood in 2D. Isotropic case.

Saturations

0.33 HCPVI 0.67 HCPVI 1.00 HCPVI 1.19 HCPVI

1.0 1.0

0.9 0.9

0.8 0.8
1-Ssteam

Sw

0.7 0.7

0.6 0.6

0.5 0.5
0 10 20 30 40 50
Distance

Figure 21. Saturation Profiles for water (solid) and 1-steam (dashed).
95
Hydrocarbon Compositions
Comp1-Heavy Comp2-Medium Comp3-Light
1.0 1.0

0.8 0.8

C om position Fraction
0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0 10 20 30 40 50
Distance

Figure 22. Fraction of hydrocarbon components.

Temperature
0.33 HCPVI 0.67 HCPVI 1.00 HCPVI 1.19 HCPVI
600

500

400
Temperature (ºF)

300

200

100

0
0 10 20 30 40 50
Distance

Figure 23. Temperature profile for steam injection at four different HCPVI on a
cold water basis.

96
Experimental Study of Heavy Oil Production From Diatomite
by Water Imbibition at Elevated Temperatures

Tom Tang and A.R. Kovscek

ABSTRACT

Oil production from many fractured reservoirs is frustrated by not only low matrix
permeability but also large oil viscosity and a matrix wettability state that is not
sufficiently water wet to favor water imbibition. Some diatomaceous reservoirs,
especially those containing heavier crude oil, appear to exhibit these attributes. Thermal
recovery using hydraulically fractured wells is one process to improve oil recovery and
unlock these heavier resources. Steam injection is typically carried out under saturated
conditions and initial heating of a reservoir is accompanied by significant condensation
and flow of the resulting hot water away from the injector. Thus, hot-water imbibition is
an important component of thermal recovery in fractured, low permeability porous
media.

This work presents the results from a series of spontaneous, counter-current water
imbibition tests using outcrop diatomite cores (2 to 4 md) and reservoir diatomite cores
(0.8 to 2 md). The experiments are isothermal and temperatures from 20 to 180 °C are
explored. Decane (nC10), two white mineral oils with differing viscosities, and heavy
diatomite crude oil were all used as oil phases. At room temperature, oil-phase viscosities
ranged from 0.9 to 6400 mPa-s. An X-ray computed tomography (CT) scanner, in
combination with a high-temperature and high-pressure core holder, was used to visualize
fluid movement and production of these oils. It is observed that temperature has a
significant effect on water imbibition rate and residual oil saturation to spontaneous
imbibition. In reservoir core filled with crude oil, the spontaneous oil recovery increased
from 3% at 40 °C (reservoir temperature) to more than 40% of oil in place at 180 °C. The
effect of temperature is to reduce oil-phase viscosity and increase water wetness of the
rock. Finally, it is found that the experimental data for mineral oils can be nearly scaled
according to dimensionless groups available in the literature. Inclusion of the mobility
ratio as a function of temperature during scaling is necessary to collapse data.

1. INTRODUCTION

The ultimate goal of this work is to improve our understanding of multiphase


fluid and heat flow characteristics of diatomite. Diatomite is a low permeability (0.1–10
md) and high porosity (35-70%) siliceous rock. We conduct X-ray computerized
tomography (CT) monitored free and forced displacement experiments at various
elevated temperatures to characterize oil recovery response to hot fluid injection and any
possible enhancements due to wettability alteration with temperature. Diatomite field
cores (Kern Co., CA) and relatively clean outcrop diatomite cores (Grefco Quarry, Kern

97
Co., CA) are the porous media employed. The major tool for analysis and interpretation
of experimental data across the temperature range from 40 to 180 °C is an appropriate
scaling group that accounts for water-oil mobility ratio as a function of temperature.

By the way of background, the diatomite resources in the San Joaquin Valley, CA
are estimated at 12 to 16 billion barrels of initial oil in place (Ilderton et al., 1996). In
recent years, water flood has been implemented in diatomite reservoirs containing
relatively light oil (°API>22) for improved recovery and to mitigate subsidence (Patzek,
1992). Steam injection in either drive (Kovscek et al., 1996a, 1996b), or cyclic modes
(Murer et al., 2000, Kumar and Beatty, 1995) is also being tested for light and heavy oil
diatomite reservoirs.

For both steam injection and water flood, production and injection wells are
hydraulically fractured to overcome low reservoir permeability and improve productivity
and injectivity. Thus, virtually all recovery processes for diatomite, and other tight
fractured rocks, rely to an extent on free and/or forced imbibition. In imbibition,
nonwetting fluid (i.e., oil) is spontaneously expelled as wetting fluid is drawn into rock
by capillary forces. In strongly and moderately water wet samples of diatomite, capillary
forces are significant and partially offset low formation permeability. Thus, the
imbibition potential for water is large and, in some instances, imbibition rates rival those
of much more permeable sandstone samples (Akin et al., 2000). Capillary phenomena are
equally important during steam injection into diatomite. Injected steam, especially at
short times, is accompanied by condensation and flow of the resulting hot water away
from the injector. In steam drive, a condensed hot-water bank precedes the steam zone
within the reservoir; whereas, in cyclic steam operations the steam zone generally
collapses during soak and production leaving a region filled with heated oil and
condensed steam. Similarly, steam injection is carried out under saturated conditions with
some fraction of the injected steam in the liquid phase. Thus, a necessary precursor to
elucidating thermal oil recovery processes in diatomite is to understand the imbibition
characteristics of hot water.

Diatomite is relatively unstudied in the laboratory because of a host of difficulties


associated with low permeability. Most previous studies were performed on sandstone
(Bourbiaux and Kalaydjianor, 1988, Jadhunandan and Morrow, 1991, Reis and Cil, 1993,
and Tang and Morrow, 1997) or (Cuiec et al., 1994, and Milter and Φxnevard, 1996).
Diatomite is composed of the siliceous, depositional, consolidated skeletal remains of
colonies of unicellular aquatic plankton (Schwartz, 1988). As such, diatomite is fine
grained, exhibits a variety of pore length scales, and flow pathways are complex (Akin et
al., 2000). Thus, injectivity is low (unless samples are severely fractured), the rock matrix
is relatively fragile, and generalization of results across rock type from chalk or sandstone
to diatomite is not currently possible. To our knowledge, the first X-ray CT images of
diatomite undergoing water injection are given by Wendel et al. (1988). Subsequently,
spontaneous co-current (Akin et al., 2000) and counter-current (Zhou et al., 2001)
imbibition characteristics were imaged in outcrop with X-ray CT for a variety of mobility
ratios, at room temperature. The pattern of water advancement and oil recovery during
counter-current versus co-current imbibition appears to be more sensitive to core-scale

98
heterogeneity. Regardless of imbibition mode, oil was recovered at the laboratory scale at
appreciable rates with sensibly zero pressure drop. On the other hand, Kamath et al.
(1995) conducted long-term forced displacements revealing that water-oil relative
permeability end points could be increased by as much as 31% with increased flow rate
and/or pressure drop.

This paper proceeds by discussing the experimental set up constructed to conduct


imbibition experiments at elevated temperature with an applied confining pressure. After
the experimental results are given, they are explained and contrasted via a dimensionless
scaling group. This simplifies the discussion that follows.

2. EXPERIMENTAL APPARATUS AND METHOD

Experimentally, we wish to obtain the change of average water saturation,


measure the rate of imbibition, and determine fluid displacement patterns within the rock.
X-ray CT meets these goals, is noninvasive, and measurements can be made without
stopping an experiment. Some care, however, must be exercised in the choice of
materials and in apparatus design. The major components of the apparatus are described
next.

2.1. Coreholder

In order to perform spontaneous imbibition tests at elevated temperatures and


high pressure, we fabricated a coreholder that resembles a Hassler-sleeve from aluminum
(6061-T6) which is relatively transparent to X-rays and thereby CT-scanable. Cores with
lengths from 2.5 to 7.5 cm (1 to 3 in) and 2.5 cm (1 in) diameter can be held. The
apparatus is assembled by coating the core with a thin film of silicone gel (Ultra Copper
High Temp RTV Silicone Gasket Maker) and placing the core inside FTP heat-shrink
tubing (Plastic Professional). The silicone-gel is nonwetting and does not penetrate the
core. It serves to prevent any fluid bypass during injection once a confining pressure is
applied; thus, the silicone gel-tubing combination serves as a high-temperature sleeve.
Mineral oil was used as a confining fluid in the annular space between the sleeve and the
aluminum coreholder. Heating tape is wrapped around the outside of the coreholder and
then covered with high temperature insulation. A temperature controller maintains
constant coreholder temperature. This system was designed and tested for service up to
260oC (500 °F) and a maximum confining pressure of 7 MPa (1,000 psia). In addition to
being compatible with a CT scanner, the main advantage of this design is that the effect
of heating on rock compaction is minimized. The system is brought to the temperature of
interest at a constant difference between confining and pore pressure.
The coreholder is attached to an L-shaped mounting bracket. This assembly is
bolted to a stepper-motor driven precision positioning system (Compumotor RP240,
Parker-Hannifin) that is used to position the core inside the CT gantry. Any position
along the length of the core can be revisited with an accuracy of ±0.01 mm.

99
2.2 CT Scanner

The scanner is a fourth generation Picker 1200SX with 1200 fixed detectors and a
scan angle of 398°. The voxel dimension is 0.35 mm by 0.35 mm by 3 mm. The tube
current is 95 mA and the energy level of the radiation is 130 keV. The acquisition time of
one image is roughly 6 s, while the processing time is about 40 s. Images can be acquired
in series and processing delayed until all the desired images are acquired. Thus, the total
acquisition time to scan the length of the core is short enough to characterize the
saturation profile along the core at a particular time.

2.3 Rock

Both field and outcrop cores are studied. Outcrop diatomite cores (Grefco Quarry,
Lompoc, CA) have a porosity, φ, of 62-67% and an air permeability, k, of 2-4 md, as
shown in Table 1. They are almost pure white and the porosity is relatively uniform, as
shown by the representative CT image in Fig.1a. This quarried diatomite represents an
end member of so-called "clean" reservoir rock resulting from Late Miocene
diatomaceous sediments (Ross and Kovscek, 2002). It provides a convenient and
appropriate contrast to the reservoir samples used here. A previous study (Akin et al.,
2000 and Zhou et al., 2001) showed that this outcrop diatomite is strongly water wet.
However, the outcrop diatomite core samples used in this work (from the same region,
but not the same block of quarried stone) were not strongly water-wet. The relevant data
are shown later. We used a similar process to cut and shape the outcrop diatomite core
samples as described in (Akin et al., 2000 and Zhou et al., 2001). The cores were dried
and preserved in an oven under house vacuum at a temperature of 50-60 oC.

Field diatomite (Kern Co., CA) cores are also used. Their porosity ranges from 55
to 59 % and permeability from 0.8 to 1.3 md, as indicated by Table 1. The field cores
were sampled from depths between 650 to 980 ft. F-1 depth=681.3 ft, F-15 depth=920.8
ft, and F-25 depth=928.1 ft. Cores were cleaned by injection of a mixture of methanol
and toluene. After cleaning, they were preserved in synthetic reservoir brine and sealed in
an airtight aluminum bag until used. The composition of the synthetic brine is listed in
Table 2. Prior to an experiment, the brine was removed by drying carefully at 35 oC under
house vacuum to avoid inducing fractures. Measurement of permeability after drying
verified that cores had not fractured, see Table 1. Figure 1b shows a representative
porosity image obtained for a field core. The bedding structure is obvious. The length of
the field cores was about 4.0 to 6.5 cm (1.6 to 2.6 in) and the diameter was 2.5 cm (1 in).

2.4 Fluids

Three white oils, n-decane, blandol (Witco Chemical), and PAO40 (Chevron
Lubricants), and filtered/dewatered crude oil from Midway-Sunset were used as the oil
phases. The viscosity at room temperature (22 °C) for these oils is 0.93 mPa-s, 35 mPa-s,

100
407 mPa-s, and 6,400 mPa-s, respectively. The decane was chosen because it has a
viscosity similar to water for a range of temperature. With nC10, we attempted to study
water imbibition under favorable viscosity ratio (µw/µnw≅1) conditions. In the next stages,
a modestly viscous mineral oil, blandol, and more viscous PAO40 were used. Hence, the
viscosity ratio is progressively less favorable (µw/µnw <1) at elevated temperatures. The
chief reason for using the mineral oils was to identify the effect of temperature on
recovery due to oil viscosity reduction without contribution to recovery from wettability
alteration via crude oil-brine-rock interaction. Lastly, we used crude oil to study further
water imbibition behavior that could be attributed to both reduction of oil viscosity and
wettability alteration accompanying an increase in temperature.

Figure 2 shows the measured viscosity versus temperature for the oils used. It
shows that the crude oil is the most sensitive to temperature. The viscosity of nC10 and
water versus temperature was obtained from “81st edition Handbook of Chemistry and
Physics” (Lide 2000). Oil viscosity was measured with a Brookfield (LVT) viscometer.
The same synthetic reservoir brine was used as the water phase for all tests. It was
prepared by dissolving salts in house distilled water as listed in Table 2. It was assumed
that the brine had the same viscosity versus temperature relation as pure water.

2.5 Procedures

Figure 3 shows how the individual pieces were assembled for imbibition tests.
Brine flows past one face of the core at 0.5 cm3/min via pump (reciprocating or syringe
pump at high temperature). We study counter-current behavior so the outlet of the
coreholder was closed for all tests during the imbibition phase. A backpressure regulator
(model R3A-C, N-norgren Inc) elevated the system pressure to 2.7 to 4.1 MPa (400 to
600 psia) such that water remained in the liquid phase throughout the duration of an
experiment. The confining pressure was set from 4.1 MPa to 5.5 MPa (600 to 800 psia)
so as to maintain roughly 1.4 MPa (200 psia) of net effective stress. All experiments were
conducted under isothermal conditions. Produced oil was collected in a graduated
cylinder to provide a check on the mass balance and average water saturation as
measured by X-ray CT scanning. After the produced water was separated from the oil it
was returned to a 300 cm3 water cylinder maintained at the experimental temperature.
The core was placed horizontally within the CT scanner gantry, as indicated by
Fig. 3. Cylindrical volume sections were monitored at 5 mm intervals along the central
axis of the core. About 8 to 10 images were collected sequentially to span along the core
from inlet to outlet at any particular time. For the tests with Swi=0, images of the dry air-
saturated cores were taken first. Then oil was injected into the core until 100% saturated
with oil and a second set of images acquired. The voxel-by-voxel values of porosity were
calculated (Akin et al., 2000, and Zhou et al., 2001) as
CT − CTar
φ = or (1)
CTo − CTa
where CT represents CT number (Hounsfields). The subscripts or and ar signify oil-
saturated rock and air saturated rock, respectively; whereas o and a represent the oil and

101
air bulk fluid phases. Average core porosity is obtained by averaging porosity across each
cross section and then averaging cross-sectional porosities.

Once the initial condition was established, the core was subjected to counter-current
water imbibition. Water was supplied at the core face until no further change in water
saturation was measured. The voxel-by-voxel water saturation, Sw, of each cross-sectional
area was computed from
CTowr − CTor
Sw = (2)
φ (CT
w − CT
o)
where φ is the independently measured porosity from Eq. 1, the subscript owr refers to
rock containing oil and water, the subscript w denotes the bulk water phase.

For the tests with Swi≠0, images of the dry air-saturated cores were taken first.
Then the core was 100% saturated with water and a second set of images was acquired.
Thereafter, oil was injected into the core to establish initial water saturation; when the
initial water saturation was established, a third set of images were taken.

3. RESULTS

Results are described in order of increasing complexity, that is, from outcrop
cores and simple fluids to field cores and crude oil. In total, 18 experiments were
conducted.

3.1 Water/Air/Outcrop Core

We first performed a water imbibition test with a dry outcrop diatomite core to
characterize spontaneous water imbibition behavior under strongly water-wet conditions.
This also served to verify that the experimental apparatus was operating properly. The
test temperature was 25 oC and initial water saturation was zero. The viscosity ratio of
displacing phase to displaced phase is greater than unity and a piston-like displacement
was expected (Zhou et al., 2001). The experimentally measured recovery is presented in
Fig. 4. Because water wets diatomite strongly as compared to air, the water imbibition
rate was very fast. It took about 270 minutes to reach a final gas recovery of 0.79 PV
which is less than the value of 0.94 reported by Zhou et al. (2001). This case will provide
a base line for studying water imbibition under different viscosity ratios of displacing
phase (water) to displaced phase (oil).

3.2 Water/nC10/Outcrop Core

Under this circumstance nC10 was used as the initial phase (non-wetting). The
tests were run at T=25, 48, and 88 oC, respectively. Initial water saturation was again
zeroing to provide a constant well-defined initial condition. No backpressure was applied
and the confining pressure was about 1.4 MPa (200 psia). The same core was used for the

102
three tests. We assumed that the core was initially water–wet and that any wettability
shift introduced by contacting the rock surface directly with decane was negligible.
Regardless, any wettability change associated with the procedure is constant from
experiment to experiment, as the core, brine, and oil are identical. The experiments were
conducted from low to high temperature progressively. At the termination of each test,
the oil and water in the core were removed by heating the core at about 50 oC under
vacuum. After the core was dry, it was re-saturated with oil at the next test temperature;
the oil volume was measured at room temperature using an ISCO pump and material
balance. Note that the oil production was also measured at room temperature. The water
saturation within the core was calculated after correction for thermal expansion of oil and
water phases. Rock thermal expansion was not incorporated because all tests are
isothermal and the effect was thereby constant.

The oil recovery by spontaneous counter-current water imbibition for the tests
with normal decane is also presented in Fig 4. At room temperature, both water
imbibition rate and the extent of oil recovery are somewhat low; the final oil recovery is
about 0.27 PV, suggesting weak water-wetness (WWW) at 25 °C. However, both
imbibition rate and the extent of recovery increase with increase in temperature
systematically. The final oil recovery is about 0.57 PV for 48 oC and 0.67 PV for 88oC. A
water imbibition curve obtained from a strongly water-wet (SWW) outcrop diatomite
core at 22 °C (Zhou et al., 2001) is also presented in this figure for comparison. It shows
that the recovery curve at 88oC is more close to that obtained for a SWW core. This
indicates that wettability alteration to increased water-wetness may occur as temperature
increases (Al-Hadhrami and Blunt, 2000). This phenomenon is in agreement with the
results obtained from Berea sandstone by Tang and Morrow (1997) and from carbonate
rocks by Dangerfiedl and Brown (1985). In addition, the decrease in oil viscosity
contributes to the increased rate of oil recovery.

3.3 Water/Blandol/Outcrop Core

A modestly viscous mineral oil (blandol) was used as an oil phase to study the
effect of temperature on water imbibition under a circumstance of viscosity ratio of water
to oil phase less than one. The test temperatures were 57, 120, and 180 oC, respectively.
The viscosity ratio of displacing phase (water) to displaced phase (oil) was about 0.062 at
57oC, 0.15 at 120 oC, and 0.217 at 180 oC. In order to prevent the formation of live steam,
a backpressure of 4.1 MPa (600 psia) was employed at the outlet of the water-circulation
loop; a confining pressure of 5.5 MPa (800 psia) was applied. After the backpressure
regulator, a water-cooled condenser reduces the temperature of the produced liquids
before they reached the oil/water separator. Blandol oil was difficult to remove from the
core following an experiment. A fresh core was used at each test temperature.

Similar to the results with n-decane/water, spontaneous imbibition recovery of


blandol increases with temperature, as summarized in Fig. 5. The final oil recovery is
about 0.49 PV at 57 oC, 0.53 PV at 120 oC, and 0.62 PV at 180 oC. The time to achieve
residual oil saturation is about 9,000 minutes at 57 oC, and about 5,000 and 4,000 minutes

103
at 120 and 180 oC, respectively, indicating a faster water imbibition rate at high
temperatures.

3.4 Water/PAO40/Outcrop Core

A mineral oil with a viscosity of about 400 cp at room temperature was used next
as the oil phase. The test temperatures were 120 and 180 oC, respectively. Initial water
saturation was again zero. A backpressure that varied between 2.8 to 4.1 MPa (400-600
psia) and a confining pressure of 5.5 MPa (800 psia) were employed. Similar to tests with
blandol, a different core was used for each temperature.

Figure 6 shows the imbibition curves for PAO40 tests. Again, both water
imbibition rate and the extent of oil recovery increase significantly with temperature as
observed previously. The final oil recovery increases from 0.39 to 0.49 PV as
temperature increases from 120 to 180 °C. At 120 °C, an apparent induction time of 300
min is observed prior to the onset of recovery. Induction time is defined as a time before
water begins to imbibe into a core; it is usually observed when the viscosity ratio of
displacing to displaced phase is far less than unit, or the displacing phase weakly wets the
core (Zhou et al., 1995 and Tang and Firoozabadi, 2001). Here, induction is likely related
the period of time required to wash a small amount of oil from the face of the core so that
water can access the rock matrix directly and imbibe. There is no induction time at 180
°C where the mineral oil viscosity is roughly 5 mPa-s, as shown in Fig. 2. Figure 6 also
contrasts recovery versus time as determined by produced oil volume and X-ray CT. Note
the pleasing agreement among the two methods of conducting material balance.

Figure 7a presents the one-dimensional water saturation profile versus time for
PAO40/water imbibition at 180 oC as measured by X-ray CT. Water begins to imbibe
instantly and reaches the far end of the core in about 400 min. A diffuse water-advancing
front is obvious during the early imbibition period consistent with previous observations
(Zhou et al., 2001). Water saturation builds quickly at the inlet reaching 0.58 PV at 200
minutes. Thereafter, the water saturation close to the inlet increases slowly to roughly 0.7
in about 1000 min. Likewise, the water saturation far from the inlet gradually increased
with time during this period. The water saturation profile changes little after 2275 min.;
however, the water saturation distribution along the core is not uniform at the end of
spontaneous imbibition. The region of high water saturation near the core inlet indicates
low connectivity of oil; hence, oil relative permeability is probably low. Combined with
the somewhat elevated oil viscosity, flow resistance is high. The nonwetting phase has
difficulty exiting the interior of the core. Figure 7b presents water saturation distribution
(CT-Images) versus time at the cross section XD=0.64 for the water/PAO40 system at 180
o
C. Note that black shading corresponds to Sw equal to 0 and that white corresponds to Sw
equal to 0.7. The diffuse nature of the imbibition front is apparent through the relatively
slow increase in water saturation. The distribution of water saturation at the cross-section
is relatively uniform consistent with the homogeneous porosity distribution in outcrop
samples as shown in Fig.1a.

104
3.5 Water/Crude Oil/Outcrop Core

To increase system complexity in a logical fashion, a heavy crude oil was used
next as the oil phase in outcrop cores. The test temperatures were 42, 120, and 180 oC,
respectively for cores with Swi=0 and was 180 oC for a core with Swi=21.7%. A separate
core was used for each test. Because the crude oil is very viscous at room temperature,
the oil and coreholder were heated to 80oC to accelerate the oil injection process. After
the core was saturated with the crude oil, the core was heated or cooled to test
temperature and allowed reaching thermal equilibrium. Each core was aged with oil at the
test temperature for 1to 2 days before it was subjected to water imbibition. Because the
oil is very viscous at room temperature, the oil flow from the tubing to the oil/water
separator was slow. This caused inaccuracy in the collection of oil in the separator and a
time-shift error in volumetric. Thus, we accounted for the imbibition rate and recovery
strictly from CT-images. The final oil recovery from the volumetric method, however,
was used to check the CT data.

Figure 8a shows the oil recovery curves for the crude oil at the various
temperatures. When Swi=0, oil recovery at 42 oC is about 0.03 PV. However, as the
temperature was increased from 42 to 120 and 180 oC, the water imbibition rate and oil
recovery increased significantly. The final oil recovery is about 0.42 PV at 180 oC.
Compared to the oil recovery for PAO40 at180oC in Fig. 6, both final oil recovery and
water imbibition rate are lower for the water-crude oil/outcrop core system. As Fig. 2
teaches, crude oil and PAO40 viscosity are nearly equal at this temperature. This
indicates a possible adsorption of polar oil components from the crude oil onto rock
surfaces that reduces water-wetness slightly in this case (Kovscek et al.,1993). At T=180
o
C, the water imbibition rate for the core at Swi=21.7% is initially (t < 1500 min) similar
to the core at Swi=0. Thereafter, water imbibition rate remains low for the core with initial
water saturation, indicating a less capillary driving force. Fig.8b shows the water
saturation vs. time for cores with and without initial water saturation at T=180 oC. It
shows that water saturation for the core with initial water saturation remains higher all the
time. Thus, the capillary force is weak for the core with initial water saturation, especially
at the later period of the test. Figure 9 shows the water saturation profiles along the core
for the test at 180 oC. Again, water enters the core instantaneously and the imbibition
front is relatively diffuse. However, water advance is slower compared to previous cases;
it took about 4,000 min to reach a stable water saturation profile. Similar to PAO40,
water saturation along the core is not uniform at the end of the experiment. Near the inlet
the final water saturation is 0.55 PV while near the outlet it is roughly 0.35 PV.

3.6. Field Core

Four field cores from the same diatomite reservoir as the crude oil sample were
used for the remaining tests. Two of them (F-a, and F-15), from a depth of 920 ft, were
used at T=43, 120, and 180 °C and zero initial water saturation. Another two cores (F-25
and F-1) from different depth were used for experiment at 180 °C with Swi=27-30%. Core
F-25 is from a depth of 920 ft and core F-1 is from a depth of 650 ft.

105
3.6.1. Swi=0

The first test (with core F-a) at T=43oC did not show any water imbibition. This
core was then heated to 120oC and allowed reaching a thermodynamic equilibrium. Then
it was subjected to water imbibition. The second core (F-15) was used at 180oC. All three
of these tests were conducted at Swi=0. At T=180 oC, two tests with initial water
saturation (F-25 with Swi=30% and F-1 with Swi=27.1%) were conducted. Figure 10
shows the recovery data for the crude oil/field core ensemble. For the tests without initial
water saturation (cores F-a and F-15), the water imbibition is nearly zero at T=43 oC (a
reservoir temperature), even after 12,000 minutes. However, when the temperature was
increased to 120 oC, water did imbibe. About 0.22 PV oil was produced by spontaneous
water imbibition at 120oC. Compared to outcrop, the final oil recovery is lower. This
indicates, possibly, that field cores are less water wet than outcrop cores. When the
temperature increased further to 180 oC and a new experiment was conducted, the water
imbibition rate at times less than 500 min increased significantly. The rate decreased
somewhat in the later period, but it is still greater than rates for the 120 °C case. The final
oil recovery is about 0.43 PV. Figures. 11a and 11b show the water saturation profile
along the core versus time at T=120 and 180 oC, respectively. At 120 oC, the waterfront
stopped advancing at XD equal to 0.6; whereas at T=180 oC, the waterfront reached the
far end of the core. Final water saturation at the core end is about 0.35 PV. Reduction in
oil viscosity from 12 cp at 120 oC to 4.5 cp at 180 oC and an increase in water-wetness
from 120 to 180 oC may contribute to the improved recovery behavior among these two
temperatures. Similar to the previous cases, the final water saturation at 180 °C is not
uniform. Figure 12 shows water saturation images versus time as obtained by CT for the
cross section XD=0.35. Again, dark shading indicates low Sw while light indicates high
Sw. Due to core heterogeneity, the water saturation is not uniform and water has not yet
swept all of the cross-sectional area. It is apparent that water advance through some of the
bedding layers is responsible for a major portion of the imbibition response. Note the
white shading in this area compared to the darker shading in the remainder of the cross
section increases with time.

3.6.2. Swi≠0

To establish the initial water saturation, the coreholder was heated to 80oC, to
reduce oil viscosity, and then oil was pumped into the brine-saturated core. After oil
breakthrough, the flow direction was reversed and oil was injected until no further water
was produced. The produced water was collected and used to estimate the initial water
saturation. The core F-25 was aged for 10 days at room temperature; core F-1 was aged
only 2 days at room temperature. At the end of spontaneous imbibition, the cores were
subjected to a forced imbibition. The inlet pressure was about 4.03 MPa (585 psia) and
outlet pressure was about 3.89 MPa (565 psia). Hence, the pressure drop across the core
was about 138 kPa (20 psi). Forced water imbibition continued for about two days until
no further oil was produced. The relevant results are shown in Fig.10. The total oil

106
recovery by spontaneous water imbibition is about 27% of OOIP for core F-25 and about
40% of OOIP for core F-1. After the cores were subjected to forced water imbibition, the
oil recovery increased to about 52% of OOIP for core F-25 and 54% of OOIP for core F-
1 quickly (<400 min). Thus, forced displacement is effective for the diatomite system;
however, the pressure gradient is roughly 1.8 MPa/m (80 psi/ft). For a reservoir system,
the imposed pressure gradient will obviously be less and oil production will occur by
both spontaneous and forced water imbibition.

From the imbibition curves for T=180 oC (cores F-1, F-15, and F-25), we could
not conclude a systematic effect of initial water saturation on water imbibition. This may
be caused by heterogeneous feature of the field cores. However, with respect to residual
oil saturation, we found the residual oil saturation is less for the core with initial water
saturation (see Fig. 13). The Sor is about 0.46 PV for core F-1, 0.51 PV for core F-25, and
0.58 PV for core F-15, respectively. With forced water imbibition, the Sor could be as low
as 0.35 PV. One may imply that the presence of initial water saturation is favorable to
reduce residual oil saturation in cases such as these. Our companion work on the pore
network properties of diatomite should shed light on the possible reasons for this
behavior (Ross and Kovscek, 2002).

The water saturation profile along the core for core F-25 (Swi=30%) experiment is
presented in Fig. 14. The dashed lines are for forced water imbibition. This figure shows
that (1) the water-front is even more diffuse than what was observed for the experiments
without initial water saturation, and (2) the forced water imbibition (following free
imbibition) does not result in a clearly definable water front moving through the core.

4. SUMMARY OF EXPERIMRNTS

All experiments followed a linear trend (not shown) when plotted as the average
aqueous phase saturation versus the square root of time, as expected (Handy, 1960). The
slope is proportional to the imbibition potential of each fluid/rock system (Akin et al.,
2000) and is useful in sorting the macroscopic results. Table 1 summarizes all
experiments by listing the best-fit slope of the line obtained by plotting average water
saturation versus t0.5. All lines are forced through the initial saturation. Fortunately, the
outcrop sample permeabilities are all nearly equal allowing generalization without
considering the role of permeability. Also tabulated is the nonwetting phase residual
saturation to spontaneous imbibition, Sorsi. The air-water/outcrop system exhibits the
largest imbibition potential (IP) consistent with low gas viscosity and the extreme
preference of the solid surface for water as opposed to air. For each fluid system
(water/nC10, water/blandol, water/PAO40, water/crude oil), an increase in temperature
results in more rapid recovery consistent with reduced oil viscosity as temperature
increases. Similarly, at a fixed temperature, the recovery rate decreases as the ratio of
wetting to nonwetting phase viscosity decreases. Presence of initial water saturation also
shows a less imbibition potential value.

107
In general, results for experiments with crude oil lag behind those for the white
mineral oils even when crude and mineral oil viscosity are similar at 180°C and the
difference in permeability among field and outcrop is considered. This indicates possible
interaction of crude oil constituents with rock that alters wettability toward mixed
wetness in a fashion; perhaps, similar to that suggested by Kovscek et al. (1993).
Residual oil saturation to spontaneous imbibition is largest in the crude oil systems.
Nevertheless, the residual oil saturation to spontaneous imbibition for a specific fluid
system decreases as temperature increases. Forced displacement following spontaneous
imbibition significantly reduced oil saturation in the water-crude oil/outcrop system. This
indicates that a substantial portion of the oil phase remains continuous at the end of
spontaneous recovery that could be recovered during forced imbibition.

5. DISCUSSION

Experimental results as a function of temperature appear to exhibit some


influence from changing wettability state. However, the viscosity and mobility ratios and
interfacial tension are also a function of temperature and must be considered before any
inferences can be made regarding wettability. Research on rock wettability through
spontaneous imbibition tests has been extensively performed (Morrow and Mason, 2001).
Spontaneous imbibition tests, as conducted here, provide not only final recovery data, but
also the imbibition rate and pattern of water invasion. A more precise characterization of
wettability is possible as compared to the Amott Index (Amott, 1959) that is based only
on spontaneous and forced imbibition end points. Wettability is assessed through the
construction of dimensionless groups that include water and oil mobilities, capillary and
viscous forces, core dimensions, permeability, and porosity.

Several scaling groups for counter-current imbibition have been proposed (Zhou
et al., 2001, Zhang et al., 1996, and Wang, 1999). In this work, we used Zhou et al.’s
scaling group (Zhou et al. 2001) to examine our experimental results. It explicitly
incorporates the mobility ratio. Given the extreme variations in viscosity and mobility
with temperature, it is most appropriate. Beginning from the equations of motion in

porous media and restricting them to counter-current conditions, it was derived:

k σ t
tD = λ*rw λ*rnw (3)
φ L2
c M* +
1
M*
where λ*r = /µ is a characteristic mobility for the wetting (w) and non-wetting (nw)
*
kr
phases and M* = λ*rw /λ*rnw is a characteristic mobility ratio, and σ is interfacial tension.
End-point relative permeabilities are used for calculating λ*r and M * . Numerical values
of kro* and krw* are set equal to 0.16 and 0.60, respectively, consistent with refs (Zhou et

108
al., 2001 and Schembre and Kovscek, 2001). The characteristic length scale, Lc, is
expressed as

Vb
Lc = Ai (4)

li

where Vb is the bulk volume of the rock sample, Ai is the area of a face open for
imbibition in the i direction and li is the distance from the open face to the no-flow
boundary in the i direction.

The interfacial tension varies with temperature. Unfortunately, a high-temperature


tensiometer was not available. Measured, at room temperature, IFT were extrapolated to
the temperature of interest via the corresponding states relationship (Reid, 1987).
1.22
 1− Tr 1
σ (T
2)= σ (T
2)  (5)
 1− Tr 2
where Tr (=T/Tc) is the reduced temperature. For blandol and PAO40, the critical
temperature is obtained from correlation (Pedersen et al., 1989).

Tc = 24.279Tb0.588 SG 0.36 (6)

where Tb is the normal boiling temperature and SG is the specific gravity. Table 3
summarizes IFT measurements and extrapolations.

Figure 15 presents the results from renormalizing the spontaneous imbibition


behavior of the outcrop/white mineral oil systems. Scaled results appear to separate into
two groups. The water/air/outcrop and water/PAO40/outcrop systems fall into one group.
In a dimensionless time of roughly 2 these systems achieve residual non-wetting phase
saturation. This is consistent with prior results of counter-current imbibition in very
strongly water wetting diatomite outcrop (Zhou et al., 2001). The second requires a
dimensionless time of order 10 to come to residual oil saturation. Note that with the
exclusion of the water/nC10 at 48° C and water/blandol at 57° C data sets, the second
group of results collapses well. In the future, we intend to repeat these particular fluid
systems at the specified temperature to determine whether some experimental error or
inaccuracy exists. Nevertheless, the existence of two distinct groups within the
dimensionless data indicates that core samples are not uniform with respect to wettability.
Evidently, the air/water/outcrop and PAO40/water/outcrop systems are more strongly
water wet than the remaining systems. This is consistent with the fact that cores were
drawn from two different blocks of the same outcrop that were quarried at different times
and locations.

We also attempted to scale the water imbibition data for crude oil tests and the
results are presented in Fig. 16. We assumed the interfacial tension between crude oil
and water is about 30 mN/m at 120 oC and 20 mN/m at 180 oC, respectively. Obviously,
the experimental data for crude oil/water/outcrop core system are similar to those shown

109
in Fig. 15. However, the scaled data for crude oil/water/field core do not show a
consistent trend and not collapsed. Further study of water imbibition for crude
oil/water/field rock system is underway to explain these trends.

6. CONCLUSIONS

Temperature affects water imbibition and oil recovery from oil-filled diatomite
cores significantly. Imbibition rate increases and the residual oil saturation to
spontaneous imbibition decreases systematically with increasing temperature. The
experimental results indicate that improvement in recovery with temperature occurs
through both a more favorable mobility ratio and wettability alteration. The significant
reduction in oil viscosity as a function of temperature decreases the mobility ratio and
this effect always favors an increase in the rate of spontaneous imbibition. Wettability
alteration toward a more water-wet state with increase in temperature, as evidenced by a
decrease in Sor, also favors imbibition. However, there is competition between the
increase in water-wetness and decrease in interfacial tension with temperature. Over the
range of temperatures examined here, the decrease in interfacial tension is a lesser effect
as compared to wettability change.

An interesting observation from the CT images is that water saturation within the
cores was not uniform, and became more nonuniform as the viscosity ratio of water to oil
became much less than one. Water saturation at the completion of spontaneous recovery
was high near the inlet and decreased with distance from the inlet. In this regard, some
portion of the oil phase remained continuous at the end of spontaneous imbibition. Forced
displacement, following spontaneous imbibition recovered significant oil. In field core,
recovery increased from roughly 30 % of OOIP following spontaneous recovery to 50%
with forced displacement.

110
7. NOMENCLATURES

A area, cm2
CT CT number, Hounsfields
IP imbibition potential, min-0.5
k, kr permeability and relative permeability, md
L length, cm
M mobility ratio
PV pore volume
S saturation, fraction
T temperature, oC
Tb boiling temperature, K
Tc critical temperature, K
t time, min
XD dimensionless length, fraction
V volume, cm3
φ porosity, fraction
µ mPa.s
σ interfacial tension, mN/m
λ mobility

Subscripts

a air
o oil
r rock
w water

111
8. REFERENCES

1. Ilderton, D.C., Patzek, T.W., Rector, J.W.,and Vinegar, H.J.: “Passive Imaging of
Hydrofractures in the South Belridge Diatomite,” SPEFEE, 11. (March 1996) pp. 46-
54.

2. Patzek, T.W.: “Surveillance of South Belridge Diatomite,” SPE 24040, paper


presented at the Western Regional Meeting, Bakersfield (March 30-April 1, 1992).

3. Kovscek, A.R., Johnston, R.M., and Patzek, T.W.: “Interpretation of Hydrofracture


Geometry During Steam Injection Using Temperature Transients I. Model
Formulation and Verification,” In Situ 20, 3, (1996) pp. 251-288.

4. Kovscek, A.R., Johnston, R.M., and Patzek, T.W., “Interpretation of Hydrofracture


Geometry During Steam Injection Using Temperature Transients I. Model
Formulation and Verification,” In Situ 20, 3 (1966) pp. 289-309.

5. Murer, A.S., McClennen, K.L., Ellison, T.K., Larson, D.C., Timmer, R.S., Thomsen,
M.A., and Wolcott, K.D., “Steam Injection Project in Heavy-Oil Diatomite,” SPERE
3, 1 (Feb. 2000).

6. Kumar, M., and Beatty, F.D.: “Cyclic Steaming in Heavy Oil Diatomite,” SPE 29623
paper presented at the Western Regional Meeting, Bakersfield (March 8-10, 1995).

7. Akin, S., Schembre, S.K., Bhat, S.K., and Kovscek, A.R.: “Spontaneous Imbibition
Characteristics of Diatomite,” JPSE 25 (2000) pp. 149-165.

8. Bourbiaux, B.J. and Kalaydjian, F.J.: “Experimental Study of Co-current and


Counter-current Flows in Natural Porous Media,” SPE 18283, paper presented at the
1988 SPE Annual Tech. Conf. And Exhibition, Houston (October 2-5, 1988).

9. Jadhunandan P.P and Morrow, N.R.: “Spontaneous Imbibition of Water by Crude


Oil/Brine/Rock Systems,” In Situ 15, 4 (1991) pp. 319-345.

10. Reis, J.C. and Cil, M.: “A Model for Oil Expulsion by Counter-Current Water
Imbibition in Rocks: One Dimensional Geometry,” J. Pet. Sci. & Eng. 10 (1993) pp
97-107.

11. Tang, G., and Morrow, N.R.: “Effect of Temperature, Salinity and Oil Composition
on Wetting Behavior and Oil Recovery by Waterflooding;” SPERE (1997).

12. Cuiec, L., Bourbiaux, J.B. and Kalaydjian, F.J.:. “Oil Recovery by Imbibition in Low
Permeability Chalk,” Soc. Pet., Eng., Form, Eval. 9 (Sept. 1994) pp. 200-208.

112
13. Milter, J. and Φxnevard, R.I.: “Spontaneous Imbibition in Two Different Chalks
Facies,” Petroleum Geoscience 2 (1996) pp. 231-240.

14. Schwartz, D.E.: “Characterizing the Lithology, Petrophysical Properties, and


Depositional Setting of the Belridge Diatomite, South Belridge Field, Kern, County,
California,” Studies of the Geology of the San Joaquin Basin, S.A. Graham and H.C.
Olson (eds.), The Pacifica Section of Soc.Eco. Paleontologists and Mineralogists, Los
Angeles (1988) pp. 281-301.

15. Wendel, D.J., Kunkel, L.A., and Swanson, G.S., “Waterflood Potential of Diatomite:
New Laboratory Methods,” SPE 17439, paper presented at the SPE California
Regional Meeting, Long Beach (March 23-25, 1988).

16. Zhou, D, Jia, L. Kamath, J., and Kovscek, A.R.: “An Investigation of Counter-
Current Imbibition Processes in Diatomite,” SPE 68837, paper presented at the 2001
SPE Western Regional Meeting, Bakersfield (March 26-30, 2001).

17. Kamath, J., deZabala, E.F., and Boyer, R.E.: “Water/Oil Relative Permeability
Endpoints of Intermediate-Wet, Low-Permeability Rocks,” Soc. Pet. Eng. Form.
Eval. 10 (March 1995) pp. 4-10.

18. Ross, C.M. and Kovscek, A.R.: “Pore Microstructure and Field Distribution in a
Diatomaceous Reservoir,” SPE 75190, paper to be presented at the SPE/DOE
Thirteenth Symposium on Improved Oil Recovery, Tulsa (April 13-17, 2002).

19. “Handbook of Chemistry and Physics,” David R, Lide (ed.) Rubber Publishing
Company, 81st edition, 2000-2001.

20. Dangerfield, J.A. and Brown, D.A.: “The Ekofisk Field,” North Sea Oil and Gas
Reservoirs,” J. Kleppe et al. Graharm and Trotman (eds) London (1985) pp. 3-22.

21. Al-Hadhrami, H.S., and Blunt, M.J.: “Thermally Induced Wettability Alteration to
Improve Oil Recovery in Fractured Reservoirs,” SPE 59289, paper presented at 2000
SPE/DOE Symposium on Improved Oil Recovery, Tulsa (April 3-5, 3000).

22. Zhou, X., Torsaeter, O, Xie, X., and Morrow, N.R.: “The Effect of Crude–Oil Aging
Time and Temperature on the Rate of Water Imbibition and Long-Term Recovery by
Imbibition,” SPE Form. Eval. (1995) pp.259-265.

23. Tang, G. and Firoozabadi, A.: “Effect of Pressure Gradient and Initial Water
Saturation on Water Injection in Water-Wet and Mixed-Wet Fractured Porous
Media,” SPE Res. Eval. & Eng. (Dec. 2001) pp.516-523.

24. Kovscek, A.R., Wong, H., and Radke, C.J.: “A pore-Level Scenario for the
Development of Mixed Wettability in Oil Reservoirs,” Environmental and Energy
Engineering 39, 6 (June 1993).

113
25. Handy, L.L., “Determination of Effective Capillary Pressure for Porous Media form
Imbibition Data,” Trans, AIME 219 (1960) pp. 75-80.

26. Morrow, N.R. and Mason, G.: “Recovery of Oil by Spontaneous Imbibition”, Current
Opinion in Colloid and Interface Science 6 (2001) pp. 321-337.

27. Amott, E.: “Observations Relating to the Wettability of Porous Media,” Trans. AIME
216 (1959) pp. 156.-162

28. Zhang, X., Morrow, N.R., and Ma, S.: “Experimental Verification of a Modified
Scaling Group for Spontaneous Imbibition,” SPERE (Nov. 1996) pp. 280-285.

29. Wang, R.: “Gas Recovery from Porous Media by Spontaneous Imbibition of Liquid,”
MS thesis, University of Wyoming, Laramie, 1999.

30. Schembre, J. M. and Kovscek, A.R.: "Direct Measurement of Dynamic Relative


Permeability from CT-Monitored Spontaneous Imbibition Experiments," SPE 71484,
paper presented at the SPE Annual Technical Conference and Exhibition, New
Orleans (Sept. 30 - Oct. 3, 2001).

31. Reid, R.C., Prausnitz, J.M., and Poling, B.E.: “The Properties of Gases and Liquids,”
McGraw-Hill, New York, 1987.

32. Pedersen, K.S., Fredenslund, Aa., and Thomassen, P.: “Properties of Oils and Natural
Gases,” Contributions in Petroleum Geology & Engineering 5, Gulf Publishing
Company, Houston, (1989).

114
Table 1 Core Properties and Test Conditions

Cor k φ L Swi T Non- µn Sor,si IP


ea wetting
o
md % cm % C phase mPas PV min-0.5
O-1 2.5 67 7.0 0 22 air 0.018 0.21 0.053
O-2 2.7 65 7.0 0 25 n-C10 0.93 0.73 0.0059
O-2 2.7 65 7.0 0 48 n-C10 0.62 0.43 0.0091
O-2 2.7 65 7.0 0 88 n-C10 0.24 0.33 0.016
O-3 2.9 67 3.5 0 57 blandol 9.0 0.51 0.0064
O-4 2.5 67 6.5 0 120 blandol 1.7 0.47 0.0073
O-5 2.3 65 7.0 0 180 blandol 0.7 0.38 0.012
O-6 2.5 67 6.5 0 120 PAO40 12.6 0.61 0.010
O-7 2.5 67 6.5 0 180 PAO40 4.7 0.51 0.018
O-8 3.9 63 6.5 0 42 Crude 640 0.97 0.00040
O-9 2.6 65 6.5 0 120 Crude 18.2 0.66 0.0042
O-10 2.7 67 6.4 0 180 Crude 4.5 0.48 0.0062
O-11 2.1 70 6.0 22 180 Crude 4.5 0.49 0.0053
F-a 0.8 55 6.5 0 43 Crude 640 1 0
F-a 0.8 55 6.4 0 120 Crude 18.2 0.88 0.00020
F-15 1.1 58 4.5 0 180 Crude 4.5 0.57 0.0045
F-1 1.5 69 5.1 27 180 Crude 4.5 0.46 0.0028
F-25 1.2 60 4.0 30 180 Crude 4.5 0.51 0.0025
Note: a-O and F denote outcrop and field core, respectively.

Table 2 Brine Composition for Diatomite Reservoir Brine, mg/L

Na+ Ca++ Mg++ Cl- HCO3- SO4-- TDS pH


2884 56 3 3575 1494 108 8100 7.5

Table 3 Estimation of Interfacial Tension (mN/m) at Elevated Temperatures

Temperature, °C 22 48 57 88 120 180


(measured)
n-C10/water (mN/M) 50 46 44 38 32 22
blandol/water (mN/M) 42 38 37 33 29 21
PAO40/water (mN/M) 40 37 36 32 28 22

115
0.7

XD=0.23 XD=0.88
(a) Outcrop Diatomite Core

XD=0.23 XD=0.88
0
(b) Field Diatomite Core (#15)
φ (PV)

Figure 1: Porosity images for outcrop and field diatomite cores.

10000

1000 Crude Oil


blandol
PAO40
Oil viscosity (cp)

100

10

0.1
10 100 1000
o
Temperature ( C)

Figure 2: Viscosity of crude oil and mineral oils vs. temperature.

116
Oil collector

Pressure transducer
readout X-ray

Confinin
pressure cylinder

Back pressure
regulator core

Circulating pump

N2

detector
Temperature
controller
CT-scanner
Water cylinder

Figure 3: Schematic of experimental

0.9

0.8 T=22oC (air)


T=88oC

0.7 T=48oC
Recovery (fraction of PV)

T=22oC
data of zhou et al
0.6

0.5
Swi=0
0.4

0.3

0.2

0.1

0
0 1000 2000 3000 4000 5000 6000 7000 8000
Imbibition Time (min)

Figure 4: Spontaneous water imbibition in air or oil (n-C10) filled outcrop diatomite
core.

117
1

0.9
T=180oC
0.8 T=120oC
Oil Recovery (fraction of PV)

T=57oC
0.7

0.6 Swi=0

0.5

0.4

0.3

0.2

0.1

0
0 2000 4000 6000 8000 10000 12000 14000
Imbibition Time (min)

Figure 5: Spontaneous water imbibition in oil filled outcrop diatomite core at elevated
temperatures: blandol.

0.9 T=180oC
T=180oC (CT)
0.8 T=120oC
Oil Recovery (fraction of PV)

Swi=0
0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 500 1000 1500 2000 2500 3000 3500
Imbibition Time (min)
Figure 6: Spontaneous water imbibiiton in oil filled outcrop diatomite core at elevated
temperatures: PAO 40

118
0.8
t=45 min
0.7 t=130 min
t=200 min
t=330 min
0.6 t=400 min
Sw (fraction of PV)

t=630 min
0.5 t=740 min
t=1340 min
t=1550 min
0.4 t=1780 min
t=2275 min
0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1
Dimensionless Length, XD
Figure 7a: Water saturation profile for water imbibition in oil filled outcrop core at
T=180oC: PAO40.

0.7

t=200 min t=300 min t=400 min t=630 min

t=740 min t=1340 min t=1550 min t=2880 min


0

Sw
Figure 7b: Water saturation vs. time at XD=0.64 for PTO40/outcrop diatomite core at
o
T=180 C.

119
0.45

0.4

0.35
Oil Recovery (of OOIP)

0.3

0.25

0.2
T=42oC,Swi=0
0.15 T=120oC, Swi=0
T=180oC, Swi=0
0.1 T=180oC and Swi=21.7%

0.05

0
0 2000 4000 6000 8000 10000 12000
Imbibition Time (min)
Figure 8a: Water imbibition in outcrop diatomite core at elevated temperatures: crude oil.

0.7

0.6

0.5
Swi (fraction of PV)

0.4
T=180oC, Swi=0

0.3 T=180oC, Swi=21.7%

0.2

0.1

0
0 2000 4000 6000 8000 10000 12000
Imbibiiton Time (min)
Figure 8b: Water imbibition in outcrop diatomite core with zero and non zero initial
water saturation at T=180oC: crude oil.

120
1

0.9
t=80 min t=200 min t=320 min
t=420 min t=1600 min t=1780 min
0.8
t=2190 min t=2950 min t=5710 min
Water Saturation (fraction of PV)

0.7 t=6130 min t=7080 min t=7480 min


t=8510 min t=9100 min t=9920 min

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless Length, XD
Figure 9: Water saturation profile for water imbibition in oil filled outcrop core at
T=180oC: crude oil.

0.6

F-15, T=180oC, Swi=0


F-a, T=120oC, Swi=0
0.5 F-a, T=45oC, Swi=0
F-25, T=180oC, Swi=30%
F-1, T=180oC, Swi=27.1%
Oil Recovery (of OOIP)

0.4

begin a forced
imbibition
0.3

0.2

0.1

0
0 2000 4000 6000 8000 10000 12000 14000 16000
Imbibition Time (min)

Figure 10: Water imbibition in oil filled fIeld core at elevated temepratures: curde oil

121
1

0.9 t=0 t=730 min


t=1300 min t=1480 min
0.8
Water Saturtation (fraction of PV)

t=1860 min t=2840 min


t=3200 min t=4413 min
0.7 t=7370 min t=10250 min
t=14250 min
0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless Length, XD
Figure 11a:
3 Water saturation profiles for water imbibition in oil filled field core at
o
T=120 C: crude oil

1
t=45 min t=130 min
0.9 t=270 min t=340 min
t=750 min t=1465 min
Water Saturation (fraction of PV)

t=1970 min t=2880 min


0.8
t=3185 min t=3385 min
t=4350 min t=4800 min
0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless Length, XD
Figure 11b: Water saturation profile for water Imbibition in
oil filled field core at T=180oC: crude oil.

122
0.7

t=230 min t=2940 min t=4300 min

t=5700 min t=7020 min t=8400


0
Sw

Figure 12: Water saturation distribution at XD=0.35 vs. imbibition time for crude oil/field
o
diatomite core at T=180 C.

0.8

0.7
Change in Swi (fraction of PV)

Forced Imbibiiton
0.6

0.5

0.4

0.3
Swi=30%, core#25
Sw=27.1%, core#1
0.2 Swi=0, core#15

0.1

0
0 2000 4000 6000 8000 10000 12000
Imbibiiton Time (min)
Fig.13: Change in average water saturation vs. time for water imbibition in oil filled field core at
T=180oC: crude Oil.

123
1
t=10080 min t=8940 min
0.9 t=8760 min t=8520 min
t=7080 min t=5700 min
0.8 t=3840 min t=2820 min
t=2421 min t=1041 min
0.7 t=351 min t=11300 min
Swi (fraction of PV)

0.6

0.5

0.4

0.3

0.2
Swi=30%
0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless Length, XD
Figure 14: Water saturation profile for water imbibiiton in oil filled field diatomite core at
T=180oC and Swi=30%: crude oil.

1.2

1
Normalized Oil Recovery

Swi=0
0.8
T=88oC, nC10
T=48oC, nC10
T=22oC, nC10
0.6 T=180oC, blandol
T=120oC, blandol
T=180oC, PAO40
0.4 T=120oC, PAO40
T=22oC, air
T=100oC, blandol
T=57oC, blandol
0.2

0
0 5 10 15 20 25 30 35
tD
Figure 15: Scaled spontaneous water imbibition in oil filled outcrop diatomite core at eleavted
temperatures using tD2:mineral oils.

124
1.2

1
Normalized oil recovery (fraction)

0.8

0.6
T=120oC, outcrop at Swi=0
T=180oC, outcrop at Swi=0
0.4 T=120oC, field at Swi=0
T=180oC, field at Swi=0
T=180oC, field at Swi=30%
0.2 T=180oC, outcrop at Swi=21.7%
T=180oC, field at Swi=27.1%

0
0 5 10 15 20 25 30 35
tD
Figure 16: Scaled spontaneous water imbibition in oil filled diatomite cores at eleavted
temperatures using tD:crude oil.

125
126
MECHANISMS OF PRIMARY
HEAVY OIL RECOVERY

127
128
Pressure Depletion Experiments in Highly Viscous Systems

Yi Tak Leung

1. INTRODUCTION

Primary production in some heavy-oil reservoirs in Canada and Venezuela has


been found to be higher than that estimated by conventional calculations. There have
been studies done to determine the reason for this unexpected phenomenon and they are
discussed in the literature review. Conventionally, the main driving force behind primary
recovery is pressure depletion through solution gas drive. Solution gas drive is the
mechanism, whereby the lowering of reservoir pressure through production in an
undersaturated reservoir will reach a point where gas starts to evolve from solution
(bubble point pressure). Because gas is quite compressible, it acts as the primary driving
force by keeping reservoir pressure high. The evolved gas does not begin to flow until
the critical gas saturation has been reached. Once the critical gas saturation point is
reached, there is an increase in rate of pressure drop due to the production of the gas-
phase. It has been noted that the oil at the wellhead of these heavy-oil reservoirs
resembles the form of foam, hence the term “foamy oil” (1995). This foamy appearance
is due to its high content of bubbles dispersed in the oil, but is not directly the cause for
high oil recovery.

There already have been studies performed demonstrating the positive effects of
increasing flow rates on recovery. This corresponds to decreases in gas-phase mobility
and thus gas relative permeability. A reduction in gas-phase relative permeability is the
most likely cause for increased recovery. The motivation, through the process of
experimentation, is to better understand the effects of high viscosity and differing
depletion rates on relative permeability of the fluids in the system, particularly gas
relative permeability.

Estimated recovery of oil and production forecasts are dependent upon knowledge
of reservoir parameters. By having more knowledge of this heavy-oil mechanism, better
predictions and more informed decisions can be made. Knowledge of optimal flow rates
can then be used to obtain the best recovery and total cumulative production before other
secondary or tertiary methods are implemented. Relative permeability plays an important
role in determining optimal flow rates and should be researched further.

The objective of this work is to characterize fluid relative permeability with


respect to flow rate and viscosity. This would further our understanding regarding fluid
behavior during the solution gas drive mechanism of a heavy oil system.

129
2. LITERATURE REVIEW

Smith (1988) was the first to address this issue of anomalous behavior in heavy-
oil reservoirs during primary production. He hypothesized that the oil and gas flowed
together, where the gas is in the form of small bubbles interspersed throughout the oil.
This simultaneous flow of oil and gas has a lower apparent viscosity as compared to oil
alone, thereby increasing productivity. Smith also looked at the theory of sand
production creating “wormholes”, thereby increasing permeability in the neighborhood of
the wellbore. It has been documented that initially, some new primary production wells
produce oil at high sand cuts, which decreases as production continues until finally the
sand cut stabilizes. This sand production is believed to increase the permeability near the
wellbore, contributing to a negative skin. Smith determined that this still would not
account for the drastic increase in total recovery. He also theorized that the asphaltenes
in the oil could act as nucleation sites for bubbles to form.

Claridge and Prats (1995) go one step further, basing an entire model on the
process of asphaltene adsorption. They performed simulations but did not actually do
laboratory experiments to confirm their theory where the asphaltenes coat the gas bubble
preventing bubble growth and coalescence. Thus, the bubbles stay small enough to travel
with the oil. This continual adsorption of asphaltenes onto the bubble’s surface causes
the oil viscosity to decrease.

Experiments were later done by Bora et al. (1997) using micromodels to test
several theories. The main objective of their experiments were to determine the effects of
asphaltenes, pressure depletion rates, and sand wettability on the process of solution gas
drive. Their use of micromodels, though not as representationally accurate as the use of
cores or sandpacks, permits them to examine pore-scale events. One concern about their
apparatus design is the structure of their micromodels. They used a very symmetric and
structured design probably not representative of the rough grains and tortuosity of the
pores and throats encountered at the pore-scale. We should be cautious if we are to
extrapolate their findings to processes that happen in the actual reservoir. They found
that the asphaltenes tended to increase the supersaturation pressure and decrease the
amount of bubble coalescence, but the asphaltenes did not increase the number of
nucleation sites. They then found that higher depletion rates tended to keep bubbles
dispersed as they flowed to the outlet and the effect of wettability on the depletion
process was negligible.

Possibly from the same results as Bora et al. (1997), Maini (1999) presents a
mechanistic model of foamy solution gas drive with possible explanations for the
improved performance and recovery. In his mechanistic model, he compares the foamy
flow and conventional solution gas drive. The main difference happens after nucleation
and growth of bubbles up to pore size. The bubbles will continue to grow, staying
immobile in the conventional solution gas drive process until it forms a continuous gas-
phase by coalescing with other nearby bubbles. Once the continuous gas-phase is
formed, it will then flow towards the producing end. But in the foamy flow situation, the
bubbles will flow with the oil once it reaches a certain size. This bubble size is dictated
by the equilibrium of capillary forces and viscous forces. As the bubbles flow,

130
coalescence is offset by the breakup of bubbles keeping them dispersed. This results in a
low gas oil ratio and high recovery. He then did experiments with micromodels and
sandpacks to support his theories. Maini (1999) also notes the independence of
wettability on the solution gas drive process. Since the solution gas drive process seems
to be independent of wettability, we can extend and simplify our experiments by using an
oil-gas system rather than go through the complicated process of conducting three-phase
(water-oil-gas) experiments. As to be expected, in Maini’s (1999) sandpack experiments,
the increase in flow rate corresponds to a higher pressure gradient, which leads to
reduced coalescence, increased bubble dispersion and higher recoveries. Also, an
increase in oil viscosity meant that these high recoveries could still be obtained at lower
depletion rates, as compared to the lower viscosity oils.

A more representational experiment was done by Urgelli et al. (1999) using a


sandstone core to investigate gravity and depletion rate effects. They found that recovery
was lower when the core was in a horizontal position as compared to when it was in a
vertical position. This was further compounded by the fact that there was higher gas
saturation found at the top of the core. It was not determined whether this gas-phase was
continuous. In any event, results indicate that there can be a gravity effect. These
findings compete with the idea that small bubbles remain entrained in the oil. It should
be noted that oil with viscosity 1500 cp was used. In their depletion rate tests, they found
that higher depletion rates corresponded to higher oil recovery.

In contrast to the results of Urgelli et al. (1999), Pooladi-Darvish and Firoozabadi (1999)
did similar experiments using a sandpack with light and heavy-oil and found that gas
mobility in heavy-oil was much less than in light oil. Intuitively, this makes sense. The
heavy-oil with higher viscosity will of course inhibit gas mobility. Other findings
include, a lack of microbubble gas-phase flow with the oil and lack of increased liquid
mobility upon nucleation of bubbles. So contrary to Claridge and Prats (1995), the
evolution of gas-phase, did not increase liquid mobility. But Claridge and Prats (1995)
specifically indicated asphaltenes as the reason for increased liquid mobility. The
presence of asphaltenes in the heavy-oil used by Pooladi-Darvish and Firoozabadi (1999)
is not known. Pooladi-Darvish and Firoozabadi (1999) then did simulations, varying
relative permeability functions to fit experimental results. They found that there was a
very large decrease in relative permeability to gas for the heavy-oil case as compared to
the light-oil case.

Tang and Firoozabadi (1999) set out to disprove the idea of foamy oil as the reason for
increased recovery. Their experiments compared recovery of mineral oil and crude oil.
The mineral oil is taken to not have the ability to be in a foamy state. They found that the
increased recovery was not due to the results of the oil’s ability to be foamy, but rather
the effect of low gas mobility. They then formulated a mathematical model to determine
the relative permeability of gas and oil, but they cannot compare these relative
permeability values to other independent measurements (to ensure that their mathematical
model was accurate).

131
Kumar et al. (2000) takes the next step by changing depletion rates and observing
its effect on gas mobility in an unconsolidated sandpack. They then match their
experimental results using Eclipse-100 Black Oil Simulator to obtain gas relative
permeability. They found that gas relative permeability decreases as depletion rate
increases.

3. APPARATUS SETUP

The equipment is setup as in Fig. 1. The sandpack holder is 16 inches long, 2


inches in diameter, and has 11 pressure ports attached to it9. A circular water jacket
surrounds the sandpack holder thus allowing for experiments to be conducted at a
specified temperature. The water jacket also helps to reduce X-ray diffraction. The
pressure ports are connected using high pressure plastic tubing (EFTE) to a 12 port
multiplexer (Scannivalve 12L8-175), which leads to a central pressure transducer
(Scannivalve). This system is hooked up to a computer via the digital interface unit
(Scannivalve) where automatic remote port measurements are recorded using a visual
basic program. The position of the pressure ports is such that they do not interfere with
axial CT scans of the sandpack. The CT scanner is a Picker 1200 SX X-ray scanner and
will be used to scan the sandpack lengthwise. A video camera is positioned at the outlet
to monitor instantaneous GOR as the effluent is produced out of the clear plastic tubing.
A high pressure piston accumulator with holding capacity of 882 mL is used to combine
the CO2 with dead oil. An ISCO Model 500D syringe pump is used in conjunction with
an accumulator. This accumulator is filled with water and is used at the outlet as an
intermediary oil depletion device. This setup allows us to conduct constant volumetric
depletion at a set rate.

4. EXPERIMENTAL PROCEDURE

The sandpack holder is initially packed with clean Ottawa sand with grain sizes
ranging between 75-150 µm. Before the inlet cap is placed back on, some glass wool is
added to the top of the sand to prevent sand movement and slumping. The apparatus is
then pressure tested to determine whether or not the O-ring on the inlet cap was sheared
during the closing process. If the O-ring is sheared, then the sandpack holder will leak.
If the sandpack does not leak, then it is flushed with CO2 for approximately 20 minutes to
displace any air in the system. The dry sandpack is then locked into place on the
positioning table in the CT scanner. The sandpack will remain immobile after the first
scan of the dry image is taken. The tubing for the pressure ports, inlet and outlet are then
connected and a dry image of the sandpack is taken with the CT scanner. The settings
used in the CT scanner are shown in Table 1.

Saturation of the mineral oil with CO2 is done in the piston accumulator. The
amount of gas needed in solution is first determined from the GOR we desire and the
volume of live oil we wish to create. This amount of gas is then converted into a
corresponding pressure that the gas will exist at when the gas is at a volume equivalent to
that of the piston accumulator’s, via the real gas law. CO2 is then injected into the

132
accumulator until pressure stabilizes to our calculated value. A liquid pump (LDC
Analytical constaMetric 3200 Solvent Delivery System) is used to inject the dead oil into
the accumulator. The oil is situated above the piston while water is used below the piston
as the displacement fluid. There are only 2 orifices, one at the top for the oil and one at
the bottom for the water. Once the oil has been injected, the valve is closed and an ISCO
pump is connected to the water side and is used to bring the pressure in the accumulator
above saturation pressure. This system is left for several days to equilibrate while the
pump is left on constant pressure mode to keep the system consistently pressurized. Prior
to injecting the live oil, several samples are taken from the piston accumulator and tested
to determine whether the live oil is truly at the target GOR.

The liquid pump is used to inject the dead oil, which saturates the sandpack.
Once oil begins to be produced at the outlet, then the back pressure regulator (BPR),
located at the outlet of our system, is set to the same pressure as the pressure we wish to
start the experiment at. We also try to saturate the live oil at this pressure. A total of 1.5
pore volumes of dead oil is injected to achieve a saturated state. The pressurization of the
sandpack using the BPR removes any gas pockets that the injection fluid may have
missed displacing. Any CO2 left still in the system will dissolve into the oil and be
produced out. Absolute permeability of the sandpack to oil can be determined at this
point. Once the 1.5 pore volumes of dead oil has been injected, flow to the inlet of the
sandpack is then switched from dead oil to live oil. The ISCO pump is set to slow
constant flowrate injection and should displace the dead oil with piston-like
displacement. 1.5 pore volumes of live oil is injected into the sandpack and an image of
the sandpack fully saturated with live oil is taken. Once injection of the live oil has
finished, the inlet is valved off and the ISCO pump is then connected to the bottom
(outlet orifice) of the outlet accumulator while the other end of the accumulator goes to
the effluent orifice of the sandpack. This outlet accumulator system is then brought up to
the same pressure as that of the sandpack and the outlet valve is switched from the BPR
to the outlet accumulator. The ISCO pump is then set to constant refill mode, which
begins the depletion experiment. Throughout the experiment, the effluent is monitored
for gas bubbles. Pressure measurements at all 11 pressure ports are recorded and CT
images are taken at various time intervals. The experiment is concluded when effluent is
all gas and it appears that no more oil is produced.

5. RESULTS

Three experiments have been conducted so far. All the experiments have used
Kaydol and CO2, with the live oil containing a GOR of roughly 43 vol/vol. The
following are details of data processing and observations for each experiment.

The following formulas are used to determine porosity and gas saturation in the
sandpack from raw CT data:
CT − CTdry
φ = oilwet
CToil − CTgas

133
CTobj− CTdry
Sg = 1 −
CToilwet − CTdry

All CT values have the units of Hounsfield. CTdry denotes the CT value of the dry
sandpack. CToilwet is for the fully oil saturated sandpack while CTobj is for the image
being processed. CTgas is the CT value for pure gas, in our case CO2 with a value of –
1000 Hounsfield. CToil is the CT value for pure oil, in our case Kaydol with a value of –
182 Hounsfield.

Kaydol Experiment 1: This first experiment was not conducted exactly according
to the procedure as previously outlined. For this first experiment, a volume of Kaydol
was allowed to saturate with a large amount of CO2 at a pressure of 600 psi for several
days. It was initially hope that the GOR would be high, around 110 vol/vol (from Fig. 2),
but the GOR measurements obtained at the end of the experiment indicated a GOR of
44.3 vol/vol. The experiment began at 615 psi using a constant volumetric depletion rate
of 0.160 mL/min (0.035 PV/hr). From Fig. 3, we see that compressibility of the liquid
accounts for the steep pressure drop of 5.03 psi/min above the bubble point pressure.
Once the bubble point pressure of 300 psi is reached, we enter a zone of supersaturation,
where pressure drop is still dependent on liquid compressibility, until pressure reaches
266 psi. The pressure then rebounds back to 276 psi and then slowly decreases at a
gentler slope, which is dependent on gas compressibility. The GOR value correlates well
with the saturation pressure in Fig. 2. The first bubble seen in the CT scan and from the
outlet tubing was observed at the same time of 1:22:00 into the experiment. Strangely,
the pressure difference between the inlet and outlet does not drop until 1:39:00. From CT
images, we can conclude that mobilization of the gas happened soon after the gas
nucleated. Looking at Fig. 6, we can deduce that this gas production is from a continuous
gas phase because of the sharp increase in fractional flow of gas soon after production of
the first gas bubble. This suggests that gas entrained in oil flow, where the gas does not
act as a free phase, is unlikely as the latter scenario would be indicated by a gradual
increase or a constant fractional flow of gas for a period of time soon after the first
bubble. The experiment was stopped after 17 hours.

Looking at the CT images of the sandpack, we immediately see a problem with


packing in the upper right corner. Glass wool was not used in the sandpack for this
experiment because it was not foreseen that slumping would be a problem. For
subsequent experiments, a disk of glass wool with a diameter 2 mm smaller than that of
the inlet cap and a thickness of 3 mm is used as a cushioning device to reduce movement
of sand and slumping.

It was decided that the sandpack should be repacked for the 2nd experiment. This
time, an over-ambitious attempt was made to prevent sand movement. The handle of a
large screwdriver was used to compact the sand as the sandpack was being filled. A disk
of glass wool was placed over the sand before the inlet cap was inserted. The intermittent
manual reciprocating compaction led to an undesirable heterogeneous pack with areas of
alternating high and low porosities.

134
Kaydol Experiment 2: The experimental procedure for this experiment is congruent
with the process described in the experimental procedure section. The target GOR was
set to 44 vol/vol and samples of the live oil were taken after saturation to determine
GOR. The GOR was found to be 43.8 vol/vol. Initial pressure at the beginning of the
experiment was 650 psi and the depletion rate is ten times slower than Kaydol
Experiment 1. Depletion rate for this experiment was 0.016 mL/min (0.0035 PV/hr).
The pressure drop in the liquid expansion region is 0.502 psi/min. Since Kaydol is again
used in this experiment and the GOR is similar to the last experiment, we expect to see a
bubble point close to 300 psi. Because of this estimated bubble point, we decreased the
frequency for pressure measurements below 500 psi. Unfortunately, the pressure
rebound was found to be in the vicinity of 420 psi. Pressure drop in the gas expansion
region was 0.0463 psi/min. The first gas seen in the CT images is at 5:31:00. Unlike the
last experiment, the first bubble seen at the outlet end is at a much later time (36:00:00)
as compared to time when gas is seen in the CT image. The critical gas saturation for the
slow rate case is much higher. The regions of high gas saturation resemble the shape of
the heterogeneities. From the CT images, gas nucleation and growth happens in the
locations of high porosity. The capillary pressure in these areas are lower, thus
facilitating growth of gas bubbles, whereas the low porosity regions may have
proportionately a larger amount of gas nucleation due to the numerous crevices but the
growth of these nucleated bubbles is hindered by higher capillary pressure. In Fig. 11,
the gas first nucleates near the center of the sandpack and growth of the gas is fairly equal
in directions to both the outlet and inlet. In Experiment 1, it is clearly seen that the gas
preferentially nucleates and grows near the outlet. Fractional flow of gas at the outlet
was unable to be obtained due to a failure of the video camera.

The discrepancy of bubble point pressure between the first experiment and the
second experiment is strange and the only plausible reason to explain it is that Kaydol is
an industrially obtained oil. The properties may differ somewhat from batch to batch
because there are no stringent guidelines. Since the time period between Experiments 1
and 2 was two months, it is possible that the Kaydol used in both experiments came from
different batches. In association to this, it is quite possible that the saturation properties
for the oil changed from batch to batch. The bubble point for the second experiment was
confirmed to be 425 psi, by doing a volumetric expansion on the remaining saturated oil
in the piston accumulator. Since the bubble point pressures are so different between
experiments, we decided that to conduct Experiment 1 again, this time using the same oil
as for Experiment 2. The oil was resaturated with CO2 and such a method would
guarantee the same bubble point pressure between Experiments 2 and 3.

Kaydol Experiment 3: The same sandpack was used in Experiment 3 as in Experiment


2. Again 1.5 pore volumes of dead oil was injected into the depleted sandpack (the status
of the sandpack at the end of Experiment 2 ). The system was brought up to 600 psi and
live oil was injected into the sandpack. The depletion rate is the same as in Experiment.
As expected, the bubble point pressure is almost exactly the same for Experiment 3 as for
Experiment 2 . Pressure depletion rate in the single phase liquid zone is 4.6 psi/min and
0.27 psi/min in the gas expansion region. The results between Experiment 1 and
Experiment 3 are very similar besides the difference in bubble point pressure. Gas

135
nucleation is seen in the CT images at 00:44:17 and gas production from the outlet is at
00:47:49. Again, in this fast depletion rate case, mobilization of the continuous gas phase
happens soon after gas formation. This leads to a low critical gas saturation. Looking at
Figs. 3, when gas nucleates and is produced, we are situated near the peak of the
differential pressure curve. As more free gas is produced, the differential pressure drops.

Figure 5, illustrates the difference in supersaturation between the slow and fast depletion
rate experiments. The average pressure in the lower rate experiment dips lower and
rebounds higher than that of Experiment 3. There are two possible reasons for this
behavior. The first, models show a constant rate with reduced number of bubbles
displaying a larger supersaturation, while a higher depletion rate with constant number of
bubbles will also show a larger supersaturation (Arora and Kovscek, 2001). These two
effects compete and if the number of bubbles dominate over the depletion rate, then we
would see a larger supersaturation in Experiment 2 as opposed to Experiment 3. Another
possible reason is due to the heterogeneity in the sandpack. This heterogeneity affects
capillary pressure, especially when the depletion rate is slow, so that viscous forces are
low. This capillary pressure could trap the gas, leading to a higher supersaturation.
When the gas finally is able to mobilize and coalesce, it may do so abruptly causing a
large rebound in pressure.

There have been difficulties calculating gas saturations because of anomalous CT


numbers in the region of gas formation. The CT number inside the gas bubbles are larger
than the saturated CT number and for pixels immediately adjacent to these gas bubbles,
the CT numbers are lower than the dry CT numbers. This leads to gas saturations in
certain areas which are less than zero and greater than one. Methods to work around this
are currently being looked at.

6. CONCLUSION AND FUTURE WORK

With the experiments so far, it is shown that depletion rate does have an effect on
the apparent supersaturation and critical gas saturation. The ability to apply these results
to relative permeability remains elusive, but using normalized gas saturations for the
relative permeability calculations may yield some results.

The objectives and motivation remain the same thus far. No one has tried to
obtain a concrete analytical definition of how relative permeability is affected by
viscosity and depletion rates of oil. The objective of the laboratory experiments will be
to use the experimental results and try to formalize an analytical model relating flow rate
of heavy-oil at their respective viscosity’s with relative permeability curves, in particular
the gas relative permeability curve. Quantifying effects of viscosity and depletion rate on
critical gas saturation and apparent supersaturation would also be useful.

This understanding is important because gas and oil relative permeability have
significant effects on the solution gas drive mechanism in heavy-oil reservoirs. We see
this from Darcy’s Law. By decreasing gas relative permeability, flow rate of gas
decreases thereby enabling more gas to stay in the system prolonging the energy in the

136
reservoir, and creating a solution gas drive mechanism that is more effective than
expected.

At least two more experiments are planned for the future. This time, the effect of
oil viscosity will be examined. Two mineral oils will be used, PAO-40 and PAO-100,
with viscosity of 900 cp and 2500 cp, respectively. The experiments will be conducted at
the slow depletion rate of 0.016 mL/min (0.0035 PV/hr) and the sandpack will be
repacked so that it is more homogeneous. Revisiting Experiment 2 with a homogeneous
sandpack would also be beneficial in the comparison studies.

137
7. REFERENCES

1. Smith, G.E.: “Fluid Flow and Sand Production in Heavy-Oil Reservoirs Under
Solution Gas Drive,” SPERE, (May 1988) pp. 169-180.

2. Claridge, E.L. and Prats, M.: “A Proposed Model and Mechanism For Anomalous
Foamy Heavy Oil Behavior,” SPE 29243, paper presented at the International Heavy
Oil Symposium held in Calgary (June 1995) 19-21.

3. Bora, R., Maini, B.B., and Chakma, A.: “Flow Visualization Studies of Solution Gas
Drive Process in Heavy Oil Reservoirs Using a Glass Micromodel,” SPE 37519,
paper presented at the SPE International Thermal Operations and Heavy Oil
Symposium, Bakersfield (February 10-12, 1997).

4. Maini, B.B.: “Foamy Oil Flow in Primary Production of Heavy Oil Under Solution
Gas Drive,” SPE 56541, paper presented at the SPE Annual Technical Conference
and Exhibition, Houston (October 3-6, 1999).

5. Urgelli, D., Durandeau, M., Foucault, H., and Besnier, J-F.: “Investigation of Foamy
Oil Effect from Laboratory Experiments,” SPE 54083, paper presented at the SPE
International Thermal Operations and Heavy Oil Symposium, Bakersfield (March 17-
19, 1999).

6. Pooladi-Darvish, M. and Firoozabadi, A.: “Solution-Gas Drive in Heavy Oil


Reservoirs,” JCPT (April 1999) pp. 54-61.

7. Tang, G-Q. and Firoozabadi, A.: “Gas and Liquid-Phase Relative Permeabilities for
Cold Production from Heavy Oil Reservoirs,” SPE 56540, paper presented at the SPE
Annual Technical Conference and Exhibition, Houston (October 3-6, 1999).

8. Kumar, R., Pooladi-Darvish, M., and Okazawa, T.: “An Investigation into Enhanced
Recovery under Solution Gas Drive in Heavy Oil Reservoirs,” SPE 59336, paper
presented at the SPE Improved Oil Recovery Symposium, Tulsa (April 3-5, 2000).

9. Akin, S. and Kovscek, A.: “Heavy Oil Solution Gas Drive Reservoirs: An
Experimental Study,” Journal of Petroleum Science and Engineering (to be
published, 2002).

10. Arora, P. and Kovscek, A.: “A Mechanistic Modeling and Experimental Study of
Solution Gas Drive,” SPE 69717, paper presented at the 2001 SPE International
Thermal Operations and Heavy Oil Symposium held in Prolaman Margarita Island,
Venezuela (March 12-14, 2001).

11. Sheng, J.J., Maini, B.B., Hayes, R.E., and Tortike, W.S.: “Critical Review of Foamy
Oil Flow,” Transport in Porous Media (1999) pp. 157-187.

138
Table 1 – CT Scanner Settings

Pilot scan Off MA 124


Field of View 48 cm Focal Spot LARG
Image Matrix 512 x 512 Anode Speed Low
Sampling 1024 X-Ray Filter 3
Scan Rotation 398 degrees Fast Process Off
Scan Speed 6 Dyn Ref On
Slice Thickness 10 mm Dyn Cal on
Algorithm 7
Resolution High
Set KV 140

Table 2 - Experiment Properties

Experiment 1 Experiment 2 Experiment 3


Porosity 0.33 0.2 0.2
Permeability (D) 20 5 5
GOR (vol/vol) 44.3 42.0 43.3
Depletion rate (mL/min) 0.16 0.016 0.16

139
Figure 1 – Apparatus for primary depletion experiments.

GOR vs. Pressure

900 160

800
140

700
y = 0.1843x - 15.974 120
2
R = 0.9804
600
100
GOR (scf/stb)

GOR (m3/m3)

500

80

400

60
300

40
200

20
100

0 0
0 100 200 300 400 500 600 700 800 900
Pressure (psi)

Figure 2 – Saturation pressure for Kaydol. The gas is CO2.

140
Average P vs. t

700

600

500
Average P (psia)

400
Exp1
Exp2
Exp3
300

200

100

0
0:00:00 2:24:00 4:48:00 7:12:00 9:36:00 12:00:00 14:24:00 16:48:00 19:12:00 21:36:00
t (hrs:mins:secs)

Figure 3 – Average pressure vs. time.

Delta P vs. t

20

18

16

14

12
Delta P (psi)

Exp 1
10 Exp 2
Exp 3

0
0:00:00 2:24:00 4:48:00 7:12:00 9:36:00 12:00:00 14:24:00 16:48:00 19:12:00 21:36:00
t (hrs:mins:secs)

Figure 4 – Delta P vs. time.

141
Average P vs. Volume Expanded

700

600

500
Average Pressure (psi)

400
Exp1
Exp2
Exp3
300

200

100

0
0 5 10 15 20 25 30
Volume Expanded (cm3)

Figure 5 – Average pressure vs. volume expanded.

Fg vs. t (Experiment 1)

0.9

0.8

0.7

0.6
Fg

0.5

0.4

0.3

0.2

0.1

0
0 10000 20000 30000 40000 50000 60000 70000
t (s)

Figure 6 – Fractional flow of gas at outlet for Experiment 1.

142
Fg vs t (Experiment 3)

0.9

0.8

0.7

0.6
Fg

0.5

0.4

0.3

0.2

0.1

0
0 20000 40000 60000 80000 100000 120000 140000
t (secs)

Figure 7 – Fractional flow of gas at outlet for Experiment 2.

0.3

Figure 8 – Experiment 1 Porosity.

Figure 9 – Experiment 2 & 3 Porosity.

143
For the next 3 figures, the outlet is on the left side.

00:00:44 1:21:34

1:46:55 2:20:16

4:38:35 16:46:00
Figure 10 – Gas saturation in sandpack (Experiment 1).

00:16:00 9:52:00

11:30:00 13:20:00

18:47:00 59:00:00
Figure 11 – Gas saturation in sandpack (Experiment 2).

00:02:30 00:44:17

01:06:24 01:40:55

05:03:40 34:41:00
Figure 12 – Gas saturation in sandpack (Experiment 3).

0 Sg 1

Figure 13 – Gas saturation bar.

144
A Study of the Effect of Viscosity on Gas Bubble
Coalescence in Viscous Oils

Omar Hayat

1. INTRODUCTION

Primary production from heavy-oil reservoirs operating under a solution-gas-drive


mechanism exhibits unexpectedly high primary recovery with slower pressure decline
rate, lower than expected producing gas oil ratios, and higher oil production rates (Bora et
al, 1997).

Heavy-oil reservoirs in Canada and Venezuela exhibit this phenomenon. The


process of solution gas drive involves nucleation of gas bubbles as the pressure in the
reservoir falls below the bubble point. It is postulated these small bubbles are trapped and
tend to remain dispersed in the oil due to the large viscosity of the continuous oil phase.
Initial development of many of these reservoirs under solution-gas drive has a great
economic advantage because implementation of traditional thermal methods can be
avoided or delayed, resulting in cost saving. At Frog Lake, Alberta Canada a study
(Huang et al., 1997) concluded that cold production, the simultaneous production of oil
and sand with the aid of progressive cavity pumps, can produce heavy oil (12o API,
20,000 to 50,000 cp) without the introduction of any external thermal energy. Another
study in the Hamaca area of the Orinoco Belt in Eastern Venezuela (Mirabal, 1997)
attributed the high productivity found in the reservoir to foamy-oil behavior of the crude.
It was reported that primary production mechanism of the Hamaca reservoirs is
controlled by the in-situ formation of a non-aqueous oil foam under solution gas drive
conditions, improving oil mobility, therefore leading to high well productivity. The
contribution of foamy-oil behavior as a primary production mechanism is up to 10% of
the OOIP in the Hamaca Area (Mirabal et al., 1996).

Several of these reservoirs are said to have shown a “foamy oil” behavior. This
“foam” is oil continuous with the gas fraction close to solution gas-oil ratio and the
consistency of a “chocolate mouse” at the wellhead. This kind of reservoir shows
abnormally high production rate as well as effective recovery factor (Urgelli et al., 1999).

To explain the above phenomenon a number of mechanisms have been suggested


(Pooladi-Darvish and Firoozabadi, 1999) which can be divided into two main categories.
Geomechanical effects such as sand dilation and development of wormholes comprise the
first category. The second category suggests that the special properties of the flowing
fluids, the gas and heavy oil, are the main reasons for high production. Both of these
categories are not mutually exclusive and are described in some detail below.

145
The first theory postulates that this favorable behavior of heavy-oil reservoirs
driven by solution-gas-drive is because of sand production. Economical production of
heavy oil is possible by allowing formation sand to be produced along with the fluid.
Dusseault (1993) shows that heavy oil (10-15o API) in cohesionless sandstones (>30%
porosity) can be exploited by permitting sand to enter the wellbore along with fluids. He
showed that with 1-3% sand production the flow rate goes up from 0.3 – 2 m3/d to 5-15
m3/d. This higher rate with sand production is attributed to four major factors enhanced
drainage radius, grain movement, gas bubble expansion, and continuous pore deblocking.

The second theory suggests that sand production is not the predominant
mechanism responsible for high oil rates and recoveries in heavy oils. It focuses more on
the way the two phases i.e. oil and gas interact. According to this theory retention of the
evolved gas phase in a dispersed form in the oil would lead to maintaining the reservoir
energy. Although Dake (1978) does not explicitly discuss heavy oils he has shown that to
obtain a high primary recovery, as much gas as possible should be kept in the reservoir.
Maintaining the cumulative gas oil ratio the ratio of cumulative gas production to
cumulative oil production) as low as possible, helps keep the gas in the reservoir. By
keeping the gas in the reservoir the total reservoir system compressibility is large and
results in greater cumulative production for a given pressure drop.

The second category, is the focus of this research. Smith (1986) argues that it is
certainly the case that extensive sand production modifies the reservoir in the vicinity of
the casing, but it has been shown in the laboratory that no amount of sand remolding or
fines removal would increase the permeability to the extent observed in pressure analysis.
It is more likely that the laboratory determined core permeabilities are overestimates of
the undistributed in-situ values and it is most commonly observed that the fines migration
in the reservoir leads to reservoir plugging and not reservoir cleaning.

Its now an accepted fact that in the solution gas drive process, the main source of
energy driving the oil towards the wellbore is the evolution and expansion of the gas
initially dissolved in the oil (Bora et al., 1997). The role of the gas bubbles in the oil
displacement process is a focus of many research efforts, for the simple reason that
understanding the pore level physics of how these bubble nucleate, grow, and coalesce is
the key to developing better oil recovery techniques from heavy oil reservoirs.

In studying the bubble nucleation and growth process, research efforts have been
directed towards trying to examine directly the pore level events by using micromodels.
These micromodels are mostly made of glass and in some cases etched out in silicon.
Such visualization experiments have also been accompanied by numerical simulation of
the experiments. Li and Yortsos (1995) carried out experiments of bubble growth by
pressure depletion in 2-D etched-glass micromodels and in Hele-Shaw cells (bubble
growth in bulk liquid) . The work also included development of a pore network numerical
model to simulate the growth of multiple gas clusters under various conditions. Du and
Yortsos (1999) also studied the dependence of critical gas saturation (gas saturation at
which the onset of bulk gas flow occurs) on the geometric parameters of the porous
media in detail. Bora and Maini (1997) carried out a series of flow visualization

146
experiments to examine the pore scale behavior of the solution gas drive process in heavy
oils.

Experimental and modeling effort has also focused on understanding the


nucleation mechanism for the gas bubbles. There are two models explaining the
nucleation of bubbles. One is instantaneous nucleation (IN), which states that all the
bubbles nucleate at essentially the same time and do not originate from preexisting
bubbles trapped in crevices or the roughness of pore walls. Firoozabadi and Kaschiev
(1995) did significant work to explore IN. The progressive nucleation process assumes
that the bubbles are released from sites such as crevices or roughness on pore walls with
poor liquid wetting properties that become activated when the local supersaturation
exceeds the capillary pressure of the site (Li et al., 1995).

The most important issue, however, is the understanding of bubble coalescence in


porous media. The presence of dispersed gas bubbles in the oil during pressure depletion
results in increased production rates in heavy oil reservoirs. Bora et al. (1997) made
several observations about bubble coalescence. They described bubble coalescence as a
three-step process. According to them, first, the bubbles have to come close to each other.
This occurs naturally during migration of the bubbles in fast depletion tests and growth of
adjoining bubbles by diffusion in slow depletion tests. When the bubbles are close
enough and collide, the second step of the coalescence process begins. This involves
draining and thinning of the liquid film separating the two bubbles. This thinning can be a
slow process, which requires a finite length of time. According to them whether or not
bubbles will coalesce depends whether the hydrodynamic forces pull the two bubbles
apart before the liquid film has drained to a critical film thickness. At critical thickness
the final step leads to coalescence of two adjoining bubbles.

George and Kovscek (1999) developed an apparatus to allow the use of silicon-
wafer micromodels over a wide range of pressures, as required for solution gas drive
experiments. Expected pore-level phenomena, such as snap-off and “breathing” of liquid
films were successfully observed. Visual observations of gas bubble formation and
evolution at a nucleation site were recorded. Experimental results indicated that oil-phase
viscosity affects the rate at which pressure changes during pressure decline. Viscosity
was also found to affect possibly the ability of bubbles to coalesce.

Building on that, Arora and Kovscek (2000) developed a mechanistic population-


balance model to describe solution gas drive behavior in viscous oils and address the
question of gas-bubble microstructure. The objective of this work was to incorporate
pore-level bubble nucleation, growth, and transport mechanisms into a continuum
displacement model consistent with standard simulation of multi-phase flow in porous
media and to verify the theoretical predictions by comparison to experimental results.
The results for the IN and PN models were compared to experimental data reported
elsewhere for light oil and to new data for viscous oils. It was found that PN model
matches the experiments somewhat better, but is more demanding computationally.

147
The work presented in this paper is directed towards developing a better
understanding of solution gas drive behavior in heavy oils. While there is experience and
data from micromodels and numerical simulation, the pore-level physics needs to be
explored further. An interesting phenomenon observed in previous micromodel
experiments was the effect of viscosity on bubble coalescence (George, 1999). A study
by Kumar and Pooladi-Darvish (1999) on the effects of viscosity and diffusion
coefficient on the kinetics of bubble growth in heavy oil, reported that the effect of oil
viscosity on bubble growth can be ignored. It concluded that viscous forces and in
general hydrodynamic forces have little or no effect on bubble growth in heavy oils at
late times. This observation was based on studying the numerical model for growth of a
single bubble in heavy oil and light oil for gradual decline in pressure. These seemingly
different observations make it all the more necessary to explore the effect of viscosity on
bubble coalescence in porous media.

The objective of this work is to understand the effect of viscosity on the


coalescence of gas bubbles trapped in narrow throats filled with heavy oil. Bubble
coalescence is divided into two broad time regimes, an earlier time regime, after
nucleation, when coalescence is diffusion dominated and a later time regime when
coalescence is governed by pressure gradient across the bubbles, pushing them together.

The earlier, diffusion dominated time regime, is explored by borrowing from the
work of Firoozabadi et al. (1995). The effect of viscosity on diffusion dominated bubble
coalescence is studied in detail.

A physical model is developed to study the effect of viscosity on bubble


coalescence in a narrow, non-circular throat filled with oil, at a later time. The model
computes the time required for the oil lens between the bubbles to drain, which are being
pushed together by pressure gradient across the bubbles. The computed time for
coalescence is compared with visual data from micromodel and capillary tube
experiments.

2. DIFFUSION DOMINATED BUBBLE COALESCENCE

Once the bubbles nucleate, they grow in size by diffusion of dissolved gas in oil.
Bubbles nucleating in close vicinity therefore coalesce by growing in size through
diffusion.

Firoozabadi et al. (1995) suggested that bubble nucleation rate J, and growth, G,
can be represented by the following general expressions:

J(t ) = Aexp[ − B / s 2 (t )] (1)


G(t ) = (ω + ν ) Ksω (t )tν −1 (2)

148
In the above Eqs. A and B are the kinetic and thermodynamic constants of bubble
nucleation, K is the kinetic constant of bubble growth, ω and ν are positive numbers of
the order unity, s is supersaturation and is defined as

s(t ) = pe (t ) − p(t ), s ≥ 0 (3)

Pe is the liquid pressure at which liquid and gas phases are at equilibrium, and P is the
pressure in the liquid, which maybe supersaturated. However for calculations in this
paper, we considered a constant value of s.

Firoozabadi used ω = 1/2 and ν= 1/2 to derive an expression for Kinetic constant
K. Here as we are considering diffusive growth of the bubble trapped in narrow pore, and
so ω = 2/5 and ν= 2/5 is used to approximate the restraint in growth from the pore walls.
Based on the derivation presented by Firoozabadi, an expression for K is derived for new
values of ω and ν.
5
K = K′[ st ]0.1 (4)
4
1
RTDks 2
where K = [′ ] (5)
2 Pb
In Eq. 5, R is the gas constant, T is temperature, D is diffusion coefficient, Ks is solubility
constant and Pb is bubble point pressure.

The diffusion coefficient can be calculated using the empirical correlation of


Hayduk and Minhas (1982) according to this correlation, the mutual diffusion coefficient
is given by
T1.47 µ ε
DAB = 13.3 × 10 −8 (6)
VA0.71
The above expression is used to calculate diffusion of an infinitely dilute solution
of A in B. Which implies that each A molecule is in an environment of essentially pure
B. This may need correction of calculated diffusion coefficient. However according to
Reid et al. (1987), DAB can be assumed to be a representative diffusion coefficient even
for concentrations of A up to 5 and perhaps 10 mole percent. Therefore in this work, DAB
is used to represent D in Eq. 5. Here CO2 is the solute dissolved as a solution gas in oil
that acts as a solvent where µ is viscosity , VA is molar volume and ε is given by
10.2
ε= − 0.791
VA
From Tyn and Calus (1975), molar volume can be calculated from the following
relationship
VA = 0.285VC1.048
Equation 2 can be solved for radius r, as
dr
= (ω + ν ) Ksω (t )tν −1 (7a)
dt

149
After integration the expression for the radius of the bubble is
r = 2.5K′[ st ]0.5 (7b)
The values for different parameters used to make the calculations are given in Table 1.
Values of different parameters are chosen for illustrative purposes only.

Table 1: Parameters for Calculation of Diffusive Bubble Growth

Parameters Values
R 8.3124 J/mol K
T 298K
Ks 5x10-6 cm2/sec
Pb 1500 psi
S 90 psi
ω 2/5
ν 2/5

Figures 1 and 2 show the variation of diffusion coefficient with viscosity and change in
bubble radius with time respectively, as more and more gas diffuses into the bubble.
Result shows strong dependence of diffusion coefficient on viscosity of the solute. It
drops significantly for higher viscosity oils. Bubble growth follows the same trend and
show strong dependence on oil viscosity.

3. PRESSURE INDUCED BUBBLE COALESCENCE

At a later time after bubble nucleation and the onset of gas flow, coalescence is
mainly controlled by pressure gradient across the bubble that pushes them together, until
they come close enough to coalesce. A physical model is developed to describe the
coalescence of gas bubbles trapped in the originally oil filled pores in reservoir media.
Consider Fig. 3a to understand the geometry of the trapped bubbles in the pore, this
geometry is the basis for developing the model.

The following assumptions are made to develop this model: Oil is the wetting
phase and gas is the non-wetting phase and there is no aqueous phase present. The pore
has a square cross-sectional area. This is a valid assumption because real pores in porous
media have corners. Pore geometry is shown in Fig 4. Bubbles are acted on by the oil-
phase pressure gradient and diffusional bubble growth at this stage is slow enough to be
neglected. Pore cross-sectional areas in our models are also square. The flow out of the
oil lens is through the four pore corners, and oil films that line the straight section of the

150
bubble-pore wall interface are thin enough to be ignored in flow calculations. Any effect
of gravity is ignored considering the very small pore diameters in reservoir media. Finally
we assume that no thin-film forces are present to stabilize gas-oil interfaces.

The problem is solved for one of the trapped bubbles and the solution is valid for
the second bubble by virtue of symmetry. Fig 3b shows the geometry of flow lines as the
two trapped bubbles move towards each other under the pressure gradient across the
bubble. At the plane of symmetry the flow lines interact to generate a stationary flow
plane with respect to the z-axis. At this point the problem becomes analogous to a
“squeeze film”. Denn (1980) explores the problem of a squeeze film in detail.

In the problem under discussion, gas bubbles move towards the symmetry plane
and the oil flows out through the four corners. Figure 3b shows the geometry of the flow
problem. The boundary Conditions for the problem illustrated in Fig 3 are that,
vz = vr = 0 (8)
at the symmetry plane.

The oil-gas bubble interface moves towards the symmetry plane, with velocity
V(t). Therefore at the moving surface z = H(t):
vz = V (t) (9a)
vr = 0 (9b)

Solving the Navier stokes creeping flow equation for the above boundary conditions, the
following expressions for vz, vr and Vt are reached (Denn, 1980).

vz = -3V(t) [ z/H(t)] 2 [1-2z/3H(t)] (10)

vr = [3rzV(t) / H2(t)] [1 – z/H(t)] (11)

V(t) = - dH/dt = [2F/3πµR4 ] H3 (12)

In the above equations, H(t) is the distance between the symmetry plane and the
bubble front at time t, vz is the velocity of fluid in z-direction, vr is the velocity of fluid in
the radial direction, V(t) is the velocity of the moving bubble, µ is the viscosity of the
fluid, F is the imposed force, responsible for bubble movement, and R is the radius of the
bubble.

At this point, boundary conditions and the flow equations are established for this
problem.

The task now is to calculate the expression for time t, needed for the two bubbles
each of radius R and separated by distance Ho, to drain the oil lens so that coalesce may
occur. The answer comes from Stefan equation (Denn et al., 1980), derived for the
squeeze-film problem using the boundary conditions and flow equations described above:

H(t) = [(1/Ho2) + (4Ft/ 3πµR4)] –1/2 (13)

151
Solving for t gives

t = (3πµR4/4F) [ 1/H2 – 1/Ho2] (14)

In the above equation R, H and Ho can be measured and quantified from visual
micromodel experiments. The viscosity of the oil, µ, can be readily measured. The only
unknown that needs some mathematical treatment is the imposed force F.

In short, F is simply the pressure difference times the cross-sectional area of the
pore and is given by
F = (∆P - ∆Pc) πR2 (15)
Where ∆P is the total pressure gradient across the bubble and ∆Pc is the pressure gradient
of oil in four pore corners. The resistance to flow in the pore corners opposes thinning of
the lens.

The expression for the term ∆Pc can be calculated from the expression for flow
resistance coefficient β, given by Ransohoff and Radke (1988),

∆Pc = (qµβL) / (a2Aw) (16)

where q is the volumetric oil flow rate through the corners, β is the dimensionless flow
resistance coefficient, L is the length of the bubble, and Aw is the area of the corners
occupied by the oil.

The volumetric flow rate q out of the oil lens is given by


H (t )
qw = ∫ 2πRvr dz (17)
0
Substituting Eq. 11 for vr and integrating, we obtain

qw = [6πR2V (t) / H2(t) ] [z2/2 – z3/3H(t)] (18)

Putting in the limits from 0 to H(t), Eq. 18 reduces to

qw = πR2V (t) (19)

Substituting Eq. 12 for V(t) gives;

qw = 2FH3(t) / 3µR2 (20)

Putting the value of qw back into the expression for ∆Pc , Eq. 16, we obtain

∆Pc = (2F H3(t)βL) / (3 a2 R2 Aw) (21)

152
Where a is the interfacial radius of curvature as shown in Fig 2b. Substituting F with (∆P
- ∆Pc) πR2 in the above expression gives

∆Pc = [(2πH3(t)βL) / (3 a2Aw)] [∆P - ∆Pc ] (22)

Let λ = [(2πH3βL) / (3 a2Aw)] and rearranging the above equation

∆Pc = [λ(t)/(1+λ(t))] ∆P (23)

At this point we have established a relationship between ∆P and ∆Pc and can find
the time required to drain any lens. The next step is to establish the values for each of the
remaining parameters. Once this is done, calculations can be made for comparison with
experimental findings.

Values of parameters are established from experimental observations. The


pressure gradient across the bubble, ∆P can be obtained by measuring the gradient across
the entire micromodel and equating the ∆P/Lmodel with ∆P/Lpore. For example the
micromodel used here is a 1.6-inch square with pressure drop of about 70 psi across its
length. The ratio ∆P/Lmodel for this model is about 44.45 psi/in, for a typical pore of
length 100 µm the value of ∆P is about 0.2 psi.

Ransohoff and Radke (1998) have computed the value of β as a function of


surface viscosity, pore half angle, contact angle, and degree of roundedness in both
graphic and tabular form. For this problem, a no slip condition is assumed and the value
for sharp corners is used (degree of roundedness, R0 = 0). From Ransohoff and Radke, for
surface viscosity = 0, β = 93.93. See Fig 4b for the geometry of the problem.

The value for interfacial radius, a, can be calculated from the work of Ransohoff
et al. (1987). This work tabulates the values for the equilibrium dimensionless interfacial
entry curvature, Cdm = CmR, in a variety of noncircular cross sections, where R is the
local radius for the largest inscribed circle of the capillary. For the case under discussion,
it is the radius of the throat. As the throat is assumed to have a square cross-section, Cdm
= 1.89 from Ransohoff et al.. Interfacial radius a is related to Cm by simple relationship a
≅ 1/Cm. For typical pore radius of 10 µm, a = 5.3 µm.

The area wetted by the oil at the four corners, Aw is related to interfacial radius a
by simple expression Aw = 4[a2 –π a2/4]. For typical pore dimensions mentioned above,
Aw equals 24.02 µm2.

Its clear that value of β will remain independent of pore geometry for this
problem because of the assumption that the throat has sharp corners. Other parameters are
dependent on pore geometry and will change from one case to another. The procedure to
evaluate them, however, remains unchanged.

153
4. MICROMODEL EXPERIMENT

The experimental apparatus consists of a micromodel, pressure vessel, optical


equipment and a system of hardware that controls fluid flow and pressure [George,
1999). The micromodel is a 5 cm by 5 cm etched pore pattern in silicon-wafer, which is
anodically bonded to a borosilicate glass plate. Figure 5a is an optical micrograph of the
model. Figure 5b is a SEM micrograph of the model and shows the shape and pattern of
the pore network. Note the pore wall roughness and sharp unrounded pore corners.

In order to perform solution gas drive experiments at elevated pressure, a pressure


vessel is designed to house the micromodel. The pressure vessel has an operating
pressure of 1000 psig. It has a thick sapphire window located directly above the
micromodel. See Fig. 6 for pressure vessel configuration. Confining fluid is present on
both surfaces of the micromodel.

A microscope is positioned on top of the vessel and pore events in the


micromodel can be observed through the sapphire window of the pressure vessel.
Kaydol, a 200 cP mineral oil, was saturated with CO2 at the desired pressure for a
minimum of 20 hours. If the micromodel contained air, CO2 was injected and allowed to
sweep the micromodel and displace the air. The upstream pressure was increased to the
desired injection pressure, and the injection of live liquid was initiated. Sufficient time
was allotted for the micromodel to fill with the liquid, and for some fluid to exit the
micromodel. The three-way valve on the downstream side of the pressure vessel was
turned such that the outlet of the micromodel was connected to the syringe pump. The
pump was then run until the outlet pressure reached the inlet pressure. The valve on the
upstream side of the pressure vessel was shut. At this point the volume between the valve
and the syringe pump, including the porous medium, pressure gauges, and tubing was
closed. Once the inlet and outlet pressure equilibrated (∆p = 0), the syringe pump was run
in refill mode at a low flow rate, to provide a controlled volume expansion.

The upstream and downstream pressures were recorded against time, and the
VCR recorded pore-level visual observations. These visual observations are used to
verify the physical model developed in this work.

5. RESULTS OF CALCULATIONS FOR MICROMODEL EXPERIMENT

Equations derived previously are used to calculate the time for coalescence of gas
bubbles trapped in a narrow pore. The geometrical dimensions used are taken from a
particular micromodel experiment see Fig 7. The results are plotted in Fig 7 for the
parameters tabulated in Table 2.

154
Table 2. Parameters for Calculation of Coalescence in a Micromodel.

Parameter Value
Viscosity, µ 220 cP
Bubble Length, L 100 µm
Bubble Radius, R 10 µm
Interfacial Radius, a 5.29 µm
Wetting Area, Aw 24.02 µm2
Pressure Gradient, ∆P(across bubble) 0.2 psi

The time required for the bubble to travel from its initial position to the plane of
symmetry is calculated. There are no chemical agents present in the viscous mineral oil to
stabilize the interface and prevent coalescence. Thus coalescence is assumed to occur
when the bubble interfaces (gas-oil) touch. The time interval for coalescence is found by
computing time versus interface position according to Eq. 14 from the initial position to
1µm of separation. We then extrapolate the solution to an interface separation zero. At
very small thickness of the oil lens, the squeeze film approximation of this problem
breaks down and infinite coalescence times are predicted. With this procedure the
calculated time for coalescence is roughly 34s, whereas the experimentally measured
time is 32s. The two agree well.

The dependence of time for coalescence on viscosity of oil is in agreement with


observed high recoveries for heavy oil reservoirs under solution gas drive. These heavy
oil reservoirs are able to maintain their high-energy state because gas bubble coalescence
is delayed by viscous oil. The more time these bubbles take to coalesce the more time is
available for the reservoir to stay energetic and produce at high rates, resulting in higher
productivity and higher recovery. Any surface agents present in the crude oil enhance the
effect of viscosity by further preventing coalescence.

6. CAPILLARY TUBE EXPERIMENT

Micromodel experiments as employed here, have one significant flaw for the
problem that is being analyzed. The problem is with the control of events within a
micromodel. It is hard to capture frequently a situation where one can observe bubble
coalescence under pressure gradient across the bubble. Capillary tube experiments
provide a suitable alternative without compromising the basic assumptions made, to solve
the problem.

155
Triangular capillaries with sharp corners were used to conduct the experiments.
The experimental setup is shown in Fig 8. The Capillary tube is held vertical with its
lower end immersed in the beaker filled with the oil. An oil lens is introduced from the
top of the capillary using a syringe. The trapped air between oil bank at the bottom of the
capillary and the oil lens simulates the gas bubble trapped in narrow pore. The oil in the
lens drains itself by flowing down the capillary corners, across the bubble, into the
beaker. As shown in the Fig 8, the problem set up is practically the same as for the
micromodel experiments. The time it takes for the oil lens of given length to drain
completely is calculated. This is the time that a bubble required to reach the plane of
symmetry, as described in the previous section. In other words this is the time for
coalescence of two bubbles trapped in a narrow pore. Experiments were repeated for a set
of oils with different viscosity.

The effect of gravity has to be considered as the capillary tubes are held in a
vertical position during the drainage of the oil lens. The force pushing the bubble up,
therefore is given by

F = (∆P - ∆Pc + ρogh) πR2

ρo is the density of the oil, g is gravity and h is the length of the bubble. In the above
equation the net imposed pressure ∆P can be calculated by the following expression

∆P = (ρo-ρg) gh

Where ρg is the density of the gas. As gas density is negligible compared to oil density, F,
can be written as

F = (2∆P - ∆Pc) πR2

The rest of the equations are the same as were used to calculate the time for
coalescence in case of micromodel experiment. The next section describes some of the
results from the experiments and their comparison to the calculations made using the
physical model.

7. RESULTS OF CALCULATIONS FOR CAPILLARY TUBE EXPERIMENT

Effort is made to maintain the height of the oil lens (i.e., the distance between the
bottom of the oil lens and the top of the oil bank at the bottom of the capillary) and the
length of the oil lens itself is roughly the same for every experiment. The upper interface
of the oil lens remains stationary during the drainage and its acts as a plane of symmetry.
A set of five different visocities was tested and the results compared with the physical
model in Fig. 9.

The following equation was used to calculate the time for coalescence in this case.
t = Cf × (3πµR4/4F) [ 1/H2 – 1/Ho2]

156
where Cf is the correction factor based on the correlation between experimental results
and the physical model. It has a value of 1.63.

Effect of viscosity on time for coalescence is apparent and it takes longer for
bubbles trapped in high viscosity oils to come close and coalesce.

8. CONCLUSIONS

The effect of viscosity on bubble coalescence in heavy oils is studied .It is shown
that viscosity of oil has a profound effect on coalescence dynamics at both earlier and
later time regimes.

In the earlier time stage when the coalescence occurs primarily owing to bubble
growth by diffusion, high viscosity of the oil slows down diffusion of dissolved solution
gas in the oil thus reducing the rate of bubble growth. This result in longer time for
coalescence for bubbles trapped in more viscous oils. This is shown by making
calculations for the rate of growth of bubbles trapped in oils with different viscosity.

At a later stage coalescence is governed by the pressure gradient across the


bubbles, pushing them together, draining the oil lens between the bubbles and finally
resulting in coalescence. Intuitively it should be hard to drain the highly viscous oil
between the bubbles and this is verified by using a physical model based on a rather
simple set of equations.

Although at this stage physical model is capable of predicting the effect of


viscosity on coalescence but it needs improvement to predict accurately the time for
coalescence. Ideally it should be able to predict the time for coalescence without using
the correction factor Cf, coming from the experimental observations. This may require an
even closer look at the physics of the model.

157
9. REFERENCES

1. Bora, R. and Maini, B.B.: “Flow Visualization Studies of Solution Gas Drive
Process In Heavy Oil Reservoirs Using a Glass Micromodel,” SPE 37519 paper
presented at the SPE International Thermal Operations and Heavy Oil Symposium,
Bakersfield (Feb. 10-12, 1997).

2. Huang W.S., Marcum, B. E., Chase, M.R., Yu, C.L.: “Cold Production of Heavy
Oil from Horizontal Wells in the Frog Lake Field.,” SPE 37545, paper presented at
the SPE International Thermal Operations and Heavy Oil Symposium, Bakersfield
(February 10-12, 1997).

3. Mirabal, Rodriguez, H. and Gordillo, R.: “Production Improvement Strategy for


Foamy Hamaca Crude Oil: A Field Case,” SPE 37544, paper presented at the SPE
International Thermal Operations and Heavy Oil Symposium, Bakersfield
(February 10-12, 1997).

4. Mirabal, Rodriguez, H.: “Impact of Foamy Oil Mechanism on the Hamaca Oil
Reserves,” Orinoco Belt-Venezuela, SPE 36140, Fourth Latin American and
Caribbean Petroleum engineering Conference, Port-of-Spain, Trinidad and Tobago
(April 23-26, 1996).

5. Urgelli, D., Durandeau, M., Foucault, H., and Besnier, J.F “Investigation of Foamy
Oil Effect from Laboratory Experiments,” SPE 54083, paper presented at the SPE
International Thermal Operations and Heavy Oil Symposium, Bakersfield (March
17-19, 1999).

6. Pooladi-Darvish, M., and Firoozabadi, A.: “Solution-gas Drive in Heavy Oil


Reservoirs,” JCPT 38, 4 (April 1999).

7. Dusseault, M.: “Cold Production and Enhanced Oil Recovery,” JCPT(Nov. 1993)
p. 16.

8. Dake, L.P.: Fundamentals of Reservoir Engineering, Elsevier ,1978, p. 84.

9. Smith G.E.: “Fluid Flow and Sand Production in Heavy Oil Reservoirs Under
Solution Gas Drive.” SPE 15094, paper presented at the Sixth California Regional
meeting of SPE, Oakland (April 2-4 1986).

10. Li, X. and Yortsos, Y.C.: “Visualization and Simulation of Bubble Growth in Pore
Networks.” AICHE Journel (Feb 1995), pp. 1247-1271

11. Du, C. and Yortsos, Y.C: “A Numerical Study of the Critical Gas Saturation in a
Porous Media.” Transport in Porous Media 35 (1999) pp. 205-225.

158
12. Firrozabadi A.and Kaschiev D.: “Pressure and Volume Evolution During Gas
Phase Formation in Solution Gas Drive Process.” SPE 26286 (paper submitted to
SPE Journal) (Sept 28, 1995).

13. George, D.S.: “Visualization of Solution Gas Drive In Viscous Oils,” MS Thesis,
Stanford University, September 1999.

14. Arora P.: “A Mechanistic Modeling and Experimental Study of Solution Gas
Drive.” MS Thesis, Stanford University, September 2000.

15. Kumar, R.and Pooladi-Darvish, M.: “Effect of Viscosity and Diffusion Coefficient
on the Kinetics of Bubble Growth in Solution-gas Drive in Heavy Oil.” Eighth
Saskatchewan Petroleum Conference of CIM, Regina, Canada (Oct 18-20, 1999).

16. Hayduk, W. and Minhas B.S.: “Correlation for Prediction of Molecular Diffusion
in Liquids, Can. J. Chem. Eng. 60 (1982) p. 295.

17. Reid, R.C., Prausnitz, J.M., and Poling, B.E.: The Properties of Gases & Liquids,
4th Ed., McGraw-Hill, Inc., p. 598.

18. Denn, Morton M.: Process Fluid Mechanics, Prentice Hall, 1980, pp. 255-261.

19. Ransohoff, T.C. and Radke, C.J.: “Laminar Flow of a Wetting Liquid along the
Corners of a Predominantly Gas-Occupied Noncircular Pore.” Journal of Colloid
and Interface Science 121, 2 (February 1988) pp. 392-400.

20. Ransohoff, Radke and Gauglitz: “Snap-Off of Gas Bubbles in Smoothly


Constricted Noncircular Capillaries.” AIChE Journal 33, 5 (May 1987) pp. 753-
765.

159
Viscosity vs Diffusion Coefficient

8.E-06

7.E-06

6.E-06

5.E-06
D (cm/sec2)

4.E-06

3.E-06

2.E-06

1.E-06

0.E+00
0 100 200 300 400 500 600 700 800 900 1000
viscosity (cP)

Fig 1 Variation in diffusion coefficient with viscosity.

Bubble Radius vs T ime (after nucleation)

16.00

14.00

12.00
radius (microns)

10.00

8.00

6.00

4.00

2.00

0.00
0 20 40 60 80 100 120 140
tim e (min)

51 cP 95 cP 178cP 490 cP 950 cP

Fig 2 Variation in radius of the bubble with time

160
Symmetry Plane

r
Bubble Movement
Gas

z
Oil

Symmetry Plane

r=R

F
r=0
Gas Oil

V(t)

b z=0 z = H(t)

Figure 3 a) Geometrical configuration of two gas bubbles moving towards each other in a
narrow pore.
b) Schematic representation of fluid flow and how the plane of symmetry acts as a
static plate.

161
thin film of
Gas
Oil

oil in the
corners

Center of
Interfacial circle

Gas

a
Oil

α = half Ψ = contact
angle angle

Figure 4 a) Cross-sectional view: wetting-liquid distribution in a square throat withthe


trapped gas bubble in it.
b) Geometry of the corner flow problem. Showing the half angle,α, the contact
angle,Ψ and the primary radius of curvature, a. Flow is in the positive z-
direction into the plane of the figure.

162
Figure 5 a) Optical micrograph of the micromodel.
b) SEM micrograph of the micromodel.

163
Pressure Vessel Lid

O-Ring Gland

O-Ring Gland

Sapphire Window

Micromodel

Pressure Vessel

Confining Liquid Inlet/Outlet

Micromodel Fluid Inlet/Outlet

Figure 6 Exploded view of the pressure vessel components

164
0 10

Time for coalescence 32 sec


Time vs Distance

40.00

35.00

30.00

25.00
T im e ( s e c )

20.00

15.00

10.00
5.00

0.00
0 10 20 30 40 50 60
42
Distance (um)

Figure 7. Video shots of bubble coalescence in a micro model with 220 cP White oil.

165
gas bubble

z Stationary
Symmetry
oil lens Plane

Oil Lens

gas bubble
Gravity

Bubble
Movement

Gas
(trapped air)

Figure 8 Experimental setup for capillary tube experiments.

166
P h y sic a l M o d e l o f b u b b le c o a l e sc e n c e

30 0

25 0
51c P

20 0 95c P

178c P
15 0
time (minutes)

490c P
10 0

950c P

50

0
0 5 10 15
d ista n c e (m m )

Figure 9. Experimental observations and results from physical model.


Dotted lines show the experimental observations, solid lines
are results from the physical model.

167
168
INSITU COMBUSTION

169
Chapter 17 (Reservoir Engineering Section)

INSITU COMBUSTION
William E. Brigham and Louis M. Castanier

This chapter is going to be the combustion part of the SPE handbook.


This is an advanced draft and will be published in 2003.

170
Chapter 17 (Reservoir Engineering Section)

Insitu Combustion

William E. Brigham and Louis Castanier

1. INTRODUCTION

Insitu Combustion (ISC) is the oldest thermal recovery technique. It has been
used for over nine decades with many economically successful projects. Nevertheless, it
is regarded as a high-risk process by many, primarily because of many failures of early
field tests. Most of those failures came from application of a good process (ISC) to the
wrong reservoirs or to the poorest prospects. An objective of this chapter is to clarify the
potential of ISC as an economically viable oil recovery technique for a variety of
reservoirs. This chapter is a summary containing a description of ISC, a discussion of
laboratory screening techniques, an illustration of how to apply laboratory results to field
design, a review of performance prediction methods, a discussion of operational
practices and problems, and an analysis of field results. For a more complete review, the
work of Sarathi1, Prats2 and Burger et al.3 should be consulted.

2. PROCESS DESCRIPTION

In situ combustion is basically injection of an oxidizing gas (air or oxygen-enriched


air) to generate heat by burning a portion of resident oil. Most of the oil is driven toward
the producers by a combination of gas drive (from the combustion gases), steam and
water drive. This process is also called fire flooding to describe the movement of a
burning front inside the reservoir. Based on the respective directions of front
propagation and air flow, the process can be forward, when the combustion front
advances in the same direction as the air flow, or reverse, when the front moves against
the air flow.

2.1. Reverse Combustion

This process has been studied extensively in laboratories and has been field
tested. The idea is that it could be a useful way to produce very heavy oils with high
viscosity. In brief, it has not been successful economically for two major reasons.

First, combustion started at the producer results in hot produced fluids that often
contain unreacted oxygen. These conditions require special, high-cost tubulars to protect
against high temperatures and corrosion. More oxygen is required to propagate the front
compared to forward combustion, thus increasing the major cost of operating an insitu
combustion project.

171
Second, unreacted, coke-like heavy ends will remain in the burned portion of the
reservoir. At some time in the process the coke will start to burn and the process will
revert to forward combustion with considerable heat generation but little oil production.
This has occurred even in carefully controlled laboratory experiments.

In summary reverse combustion has been found difficult to apply and


economically unattractive.

2.2. Forward Combustion

As only forward combustion is practiced in the field we will only consider this
case. Forward combustion can be further characterized as “dry” when only air or enriched
air are injected or “wet” when air and water are co-injected.

2.2.1. Dry Combustion

The first step in dry forward insitu combustion is to ignite the oil. In some cases
auto-ignition occurs when air injection begins if the reservoir temperature is fairly high
and the oil reasonably reactive. This often occurs in California reservoirs. Ignition has
been induced using down hole gas burners, electrical heaters, and/or injection of
pyrophoric agents or steam injection. Ignition will be discussed in more detail later.

After ignition the combustion front is propagated by a continuous flow of air.


Rather than an underground fire, the front is propagated as a glow similar to the hot zone
of a burning cigarette, or to hot coals in a barbecue. As the front progresses into the
reservoir, several zones exist between injector and producer as a result of heat and mass
transport and the chemical reactions. Figure 17.14 is an idealized representation of the
various zones and the resulting temperature and fluid saturation distributions. In the field
there are transitions between zones, however the concepts illustrated provide insight on
the combustion process.

2.2.2. Zone Definitions

Starting from the injector, seven zones have been defined:

A. The burned zone is the volume already burned. This zone is filled with air and
may contain small amounts of residual unburned organic solids. As it has been
subjected to high temperatures, mineral alterations are possible. Because of the
continuous airflow from the injector, the burned zone temperature increases from
injected air temperature at the injector to combustion front temperature at the
combustion front.

172
Figure 17.1 Schematic Diagram Of Temperature And Saturation
Profiles And Zones In Insitu Combustion4

173
B. The combustion front is the highest temperature zone. It is very thin, often no
more than several inches thick. It is in this region that oxygen combines with the
fuel and high temperature oxidation occurs. The products of the burning
reactions are water and carbon oxides. The fuel is often misnamed coke. In fact it
is not pure carbon but a hydrocarbon with H/C atomic ratios ranging from about
0.6 to 2.0. This fuel is formed in the thermal cracking zone just ahead of the front
and is the product of cracking and pyrolisis which is deposited on the rock
matrix. The amount of fuel burned is an important parameter because it
determines how much air must be injected to burn a certain volume of reservoir.

C/D. The cracking/vaporization zone is downstream of the front. The crude is


modified in this zone by the high temperature of the combustion process. The
light ends vaporize and are transported downstream where they condense and
mix with the original crude. The heavy ends pyrolize, resulting in CO2 , CO,
hydrocarbon gases and solid organic fuel deposited on the rock.

E. The steam plateau. This is the zone where some of the hydrocarbon vapors
condense. Most of those condense further downstream as the steam condenses.
The steam plateau temperature depends on the partial pressure of the water in the
gas phase. Depending on the temperature the original oil may undergo a mild
thermal cracking, often named visbreaking that usually reduces oil viscosity.

F. A water bank exists at the leading edge of the steam plateau where the
temperature is less than steam saturation temperature. This water bank decreases
in temperature and saturation downstream, with a resulting increase in oil
saturation.

G. The oil bank. This zone contains most of the displaced oil including most of the
light ends that result from thermal cracking.

H. Beyond these affected areas is the undisturbed original reservoir. Gas saturation
will increase only slightly in this area because of the high mobility of
combustion gases.

2.2.3. Wet Combustion

A large amount of heat is stored in the burned zone during dry forward insitu
combustion (Fig. 1), because the low heat capacity of air cannot transfer that heat
efficiently. Water injected with the air can capture and advance more heat stored in the
burned zone.

During wet combustion injected water absorbs the heat from the burned zone,
vaporizes, moves through the burning front and condenses, expanding the steam plateau.
This results in faster heat movement and oil displacement.

174
Depending on the water/air ratio, wet combustion is classified as 1) incomplete
when the water is converted into superheated steam and recovers only part of the heat
from the burned zone, 2) normal when all the heat from the burned zone is recovered, and
3) quenched or super wet when the front temperature declines as a result of the injected
water.

When operated properly, water assisted combustion reduces the amount of fuel
needed, resulting in increased oil recovery and decreased air requirements to heat a given
volume of reservoir. Up to a 25% improvement in process efficiency can be achieved5.
Determination of the optimum water/air ratio is difficult because of reservoir
heterogeneities and gravity override that can affect fluid movement and saturation
distributions. Injecting too much water can result in an inefficient fire front, thus losing
the benefits of the process.

Some authors recommend, as a best practice, injecting water at high rates to


achieve “partially quenched combustion”. This method has limited application. A high
temperature burn is preferred but is difficult to achieve with oils that are not highly
reactive. Injecting large amounts of water can lower combustion temperatures resulting in
a greater fraction of oil burned and higher costs for oxygen. At the same time these types
of burns only partially oxidize the oil. This partial oxidation results in a much more
viscous liquid, which in turn lowers the flow rate. So, in brief, if water injection is used,
great care should be taken to assure that liquid water never reaches the high temperature
combustion front. A discussion of heat and material balance calculations that include
chemical reactions and the effect of injected air and water, is presented later in some
detail.

3. LABORATORY STUDIES

In situ combustion mechanisms are largely a function of oil composition and rock
mineralogy. The extent and nature of the chemical reactions between crude oil and
injected air, as well as the heat generated, depend on the oil-matrix system. Laboratory
studies, using crude and matrix from a prospective ISC project, should be performed
prior to the design of any field operation.

3.1. The Reactions

The chemical reactions associated with ISC are complex and numerous. They
occur over a broad temperature range. Most researchers group them into three classes in
ascending temperature ranges:
• Low temperature oxidation (LTO) - heterogeneous gas/liquid reactions producing
partially oxygenated compounds and few carbon oxides.
• Medium temperature reactions - cracking and pyrolisis of hydrocarbons to form
fuel.

175
• High temperature oxidation (HTO) - heterogeneous H/Cbond breaking reactions
in which the fuel reacts with oxygen to form water and carbon oxides.

A more recent and more accurate kinetics model has been developed6. Only two reactions
are used, but in addition the geometry of the reacting residual fuel in the pore spaces is
taken into account as indicated in Figure 17.2. This figure represents the fuel remaining
on two sand grains at different times in the combustion process, as discussed by Mamara
et al. 6 The crude oil oxidation consists of two stages, low temperature oxidation forming
an oxygenated hydrocarbon fuel, and high temperature combustion of this fuel. A
detailed description of the different reaction regimes is outside the scope of this
handbook; some practical comments on the role of LTO, however, are appropriate at this
stage.

Figure 17.2. Schematic Diagram Of Varying-Fuel-Geometry6.

Low temperature oxidation (LTO) can be described as oxygen addition to the


crude oil. LTO yields water and oxygenated hydrocarbons such as ketones, alcohols, and
peroxides. A good description of LTO can be found in Burger7. LTO generally increases
original oil viscosity, boiling range and density. LTO increases the amount of fuel. LTO
is promoted by low air flux in the oxidation zone. Poor crude oxidation characteristics
can also play a role. In heavy oil reservoirs (API gravity < 20o), LTO tends to be more
pronounced when oxygen rather than air is injected in the reservoir8.

Research has shown that, for heavy oils, LTO reactions must be minimized.
Figure 17.3 shows the oxygen uptake as the temperature of a typical heavy oil is raised
linearly with time. Notice the negative temperature gradient region that is the region
where oxygen rate uptake decreases with temperature increase. If the temperature of the
ISC process stays at or below the negative temperature gradient region, the oil
displacement efficiency will be very low. This is because LTO increases the oil viscosity

176
Figure 17.3. Schematic of dry combustion temperature profile
showing the general effect of temperature on oxygen
uptake rate for heavy oils and the negative temperature
gradient region8.

and fuel content. The injected air flux in a heavy oil project should be maintained at a
value well above the value needed to maintain the reactions in the high temperature
oxidation regime. LTO generally has almost no effect on light oils in terms of mobility
or recovery despite the fact that light oils are more susceptible to LTO than heavy oils.

Fuel deposition determines the feasibility and economic success of a combustion


project. It occurs at intermediate temperatures after the LTO reactions. Numerous studies
have been conducted aimed at understanding fuel formation and deposition at
intermediate temperatures. The oil type and chemical structure determine the rate and
extent of the different reactions. Catalytic effects from the matrix and/or injected
solutions of metals may affect the type and amount of fuel formed. Again all laboratory
experiments must include not only the crude to be tested but also representative core
material from the reservoir of interest.

3.2. Kinetics

Kinetics of combustion reactions can be defined by how fast the chemical


reactions occur and how much of the oil is affected. It is important to study kinetics for
several reasons:
• Characterization of oil reactivity
• Determination of ignition conditions
• Insight on the nature of the fuel and its combustion characteristics

177
• Use of kinetic parameters as input for possible numerical simulation of the
process.

As crude oils contain hundreds of different compounds, it is impossible to


accurately represent all the reactions occurring during ISC. Even if it were possible to
detail all the reactions, the use of such information in numerical models would be
impossible because of cost and computer limitations. Consequently we will concentrate
on useful simple models describing ISC reaction kinetics that have been published in the
literature. Most studies use the Arrhenius reaction expressions defined as follows. The
model assumes a functional dependency on fuel concentration and oxygen partial
pressure. It is given by,
dC f
Rc = = K p a Cb (17.1)
dt O2 f
where
Rc = reaction rate of the crude, kg/m 3 s
Cf = the concentration of fuel, kg/m 3

p = oxygen partial pressure, Pa


O2

The exponent constants, a and b, are the orders of the reactions with respect to oxygen
partial pressure (a), and fuel concentration (b). Data shows that “a” ranges between 0.5
and 1.0, while “b” is close to 1.0. The reaction rate, K , is the Arrhenius constant,
expressed as a function of temperature, as follows,

K = Aexp( − E / RT) (17.2)


where
A= Arrhenius rate constant
E = activation energy, kJ/mole
T = absolute temperature, ºK
R = universal gas constant, kJ/mole ºK

When using literature values one has to be careful because the parameters in Eqs. 17.1
and 17.2 vary depending of the system of units used.

A variety of experimental techniques can be used to determine the kinetics of ISC


reactions. Among those are differential thermal analysis, thermogravimetric analysis,
accelerating rate calorimetry and effluent analysis. The reference list contains several
descriptions of various methods and results.

The effluent analysis method, also called the ramped temperature method is
quantitative and consists in heating a sample of oil and rock while flowing oxygen (for
oxidation) or nitrogen (for pyrolisis) through the sample. The kinetic parameters can be
calculated from effluent gas evolution with temperature, and chemical analysis of post
test cores. Details of the analysis techniques can be found in the references.3,6,9

178
3.3. Combustion Tube Studies

Although the kinetic studies can provide useful insight on ISC reactions,
combustion tube experiments are mandatory to determine the parameters needed to
design and implement field projects. These data are used to make predictions of field test
performance. As Sarathi1 points out, “Combustion tube studies are the necessary first step
in the design of an ISC project.”

Combustion tubes aim at representing a small volume of the reservoir. They are
usually packed with native reservoir cores or representative samples of matrix material
and oil, placed in vertical position to minimize gravity effects and heated to reservoir
temperature. Ignition is usually started at the top by electrical heaters and the combustion
front is propagated downward. This allows propagation of a combustion front and the
associated chemical reactions at conditions close to those in a reservoir.

Temperature profiles, pressures, gas and liquids injection and production rates,
and composition histories at the inlet and outlet are recorded. ISC tube runs are unscaled
and direct correlation of combustion tube results to the field is not possible. However, as
long as the runs are performed with reservoir rock and fluids at reservoir conditions, the
reactions of fuel deposition and combustion will be similar in both tube and reservoir.
Tube runs will not provide information on ISC sweep efficiency. They adequately model
the chemistry of the process but not the flow behavior in the reservoir, and only partially
model the heat transfer processes. Flow behavior in the reservoir is affected by gravity
override, well spacing and geometry and reservoir heterogeneities, and tube runs cannot
reproduce these phenomena. Heat transfer from the tube to the surroundings can be much
higher than reservoir heat losses.

Two different schools of thinking exist on this heat transfer problem. Many
experimenters use strip heaters around the tube to lower the temperature gradient between
the tube and the surroundings. This reduces heat losses and allows front propagation at
fluxes similar to those in the field. It can, however, lead to overestimation of water/oil
ratios in wet combustion if the strip heaters provide too much energy to the system, as
they often do. Information on front cooling by injected water may also be masked by the
heaters. As a result, the extent of the steam plateau may not be correct. Most of these
types of experiments are bulky and time consuming and require extensive
instrumentation.

The other solution is to increase the air flux and minimize heat losses by insulation
alone. This may slightly overestimate air requirements and fuel content but is much
simpler and easier to operate. As a result, it is widely used. Description of various setups
for combustion tube studies have been provided5,6,10,11,12.

The information that can be acquired from tube runs includes:


• Fuel burned
• Air required to burn a unit volume of reservoir
• Atomic H/C ratio of burned fuel

179
• Excess air and oxygen utilization
• Air/Fuel ratio
• Oil recovery from the swept zone
• Optimization of water/air ratio in wet combustion
• Composition of produced fluids
• Front temperature and stability

This last information is quite important in heavy oils to determine if the process is
operating properly in the desired high temperature regime. If high temperature cannot be
achieved in ideal laboratory conditions it is likely that field results would be worse.

3.3.1. Data Analysis

The following is a simple analysis of data from tube runs. It assumes that the
combustion occurs at high temperature where the fuel exclusively combines with oxygen
to produce water and carbon oxides. The stoichiometric equation13 is then:
 2 m + 1  n   m   1  n
CHn +   +  O2 =  CO2 +   CO +   H 2O (17.3)
 2 m + 2  4   m + 1  m + 1 2

where n = hydrogen/carbon atomic ratio of fuel


m = CO2 /COconcentration ratio produced
The other symbols indicate the various components in the chemical balance equation.

This equation is only an approximation of the process. It neglects LTO reactions,


oxygen /minerals reactions and water/organic fuel reactions. Alternate analysis when
some of these reactions are important are detailed in Sarathi1 based on work from Moore
and Mehta10. Assuming Eq. 17.3 to be valid the apparent H/C ration, n , can be estimated
from the concentration of exhaust gases and the injected oxygen concentration13.

4[O2 − CO2 − CO / 2 − (O2 ) prod ]


n= (17.4)
CO2 + CO
where (O 2 ) prod = Oxygen concentration produced

It is prudent to normalize the concentrations by making a balance on the nitrogen, which


in these conditions can be considered inert. The basic chemical equation is then:

CHn + aO2 + RaN2 = bCO2 + dCO + fO


2 + RaN2 (17.5)

where R is the molar ratio of nitrogen to oxygen in the feed gas and a , b, d , and fare
stochiometric coefficients similar to those in Eq. 17.3.

180
The range of the ratio, n , for high temperature reactions should be from 0.5 to 2.
Calculation of an unusually high value of n indicates that low temperature oxidation is
important. In the very early stages of field projects high “n” values are often observed
because of the solubility of the combustion gases, particularly CO 2 , in the oil.

Once n and m are known, the amount of air required to burn one unit weight of
fuel is found from Eq. 17.3. The heat generated by burning a unit weight of fuel can be
calculated by simple addition of the heat generated by each reaction as described in the
stoichometric equation (Eq. 17.3). The calculation of heat produced must take into
account the production of carbon monoxide. The following formula13 can be used to
estimate heating values of fuels as a function of n and m.
m (174,000) 52,500 61,500
Hc = + +n (17.6)
( m + 1) ( n + 12) ( m + 1) ( n + 12) ( n + 12)

where
Hc = heating value, Btu/lb fuel

To convert to Joules/kg multiply by 2,326.

The air required to burn a given volume of reservoir is of course a very important
design parameter and one of the keys to the economics of the combustion process. This is
directly calculated from the experimental data by dividing the amount of oxygen
consumed by the volume swept during the tube run. The mass of fuel burned in a unit
volume of reservoir can be calculated from the oxygen consumed by a unit volume and
applying Eq. 17.3. All the other relevant parameters can be estimated13,14. It is prudent to
perform multiple laboratory tube runs prior to field implementation.

4. COMBINING MATERIAL AND HEAT BALANCE CALCULATIONS

Many useful and reasonably accurate calculations can be made on ISC to predict
the behavior of a proposed project. These ideas will be explained in the following
diagrams and example calculations. They start with a very simple heat balance, and are
then extended to more closely represent what happens in the reservoir.

4.1. First Assumptions

Start by assuming no combustion data is available to get an initial idea of the


feasibility of a project. This preliminary work gives the engineer a sound basis to decide
whether further work has economic promise. Assume a sandstone formation with a
porosity of 22%, a temperature of 100°F, a 24° API oil at a saturation of 65%, and an
injection pressure of 300 psia. Also assume the CO2 /COatomic ratio, m, will be,
CO2
m= = 20 (17.7)
CO

181
This is a reasonable ratio to assume, based on both laboratory and field experience. Since
there is no tube run data, generalized correlation curves13, Figs. 17.4 and 17.5, will be
used to calculate expected results. From Figure 17.4, the fuel availability, W, for 24°API
crude is,

W = 0.95 lbC / 100 lb Rock (17.8)

Figure 17.4. Fuel Availability vs Air/Gravity13

182
The apparent H/Catomic ratio (n) of the fuel is also needed. This is a function of
the combustion front temperature13 as shown on Figure 17.5. Selected data from the
graph are listed in Table 17.1.
20

18 21.8o API Crude


Ottawa Sand

16

14
Apparent H/C Ratio of Fuel - Atomic

12

10

0
0 200 400 600 800 1000 1200
o
Combustion Temperature - F

Figure 17.5. H/C ratio vs. combustion temperature13.

Table 17.1 Effect of Temperature on H/C Ratio, 21.8° API Crude


H / C Ratio, n Comb. Temp (°F)
0.40 1,000
0.80 800
1.40 700
2.40 600

These data for 21.8° API crude are close enough to 24°API crude for initial estimates.

183
4.2. Calculate Initial Heat Balances and Temperatures

Start by assuming all heat generated is used to heat the rock formation through
which the combustion front has moved. This assumption is incorrect, but simplifies
understanding of the mathematics and concepts involved in heat balance calculations. A
sketch of the temperature profile generated is shown in Fig. 17.6. Corrections to this heat
balance calculation will be discussed later.

Assuming 1.0 ft 3 of rock formation burned and the front temperature is 1000°F,
from Eq. 17.8, and Table 17.1, we get,

lb Fuel lb Carbon lb Hydrogen


= +
100 lb Rock 100 lb Rock 100 lb Rock
(17.9)
0.95(0.40)
= 0.95 + = 0.9817
12

Quartz weighs about 164 lb/ft 3 . The amount of fuel for a cubic foot of formation equals
(0.9817) (164) (1 − 0.22)
= 1.256 lb Fuel/ft 3 Rock (17.10)
100

Figure 17.6 Idealized Temperature Profile Assuming All


Heat Stays In The Burned Zone.

Using the heat of combustion, Eq. 17.6, with the appropriate parameters, it becomes,

( 20) (174,000) 52,500 (0.4) (61,500)


Hc = + +
(21) (12.4) ( 21) (12.4) 12.4 (17.11)
= 15,550 BTU/lb Fuel

184
Thus the total heat generated is,

Heat generated BTU


= (15,550) (1.256) = 19,530 (17.12)
ft 3 Rock ft 3 Rock
Next, calculate the temperature rise of the formation behind the front to see if it
matches the temperature assumed by performing a heat balance behind the front. Since,
for practical purposes, the only fluid in the formation behind the front is air, which has a
very small volumetric heat capacity, we only need to calculate a heat balance on the
sandstone itself. A good equation for average sandstone heat capacity is13,

T + 2000 T2 − T1
cs = 1 + (17.13a)
10,000 20,000
1000 − 100
= 0.21 + = 0.255 BTU/lb − !F (17.13b)
20,000

where T1 = Initial reservoir temperature, ºF,


T2 = Final reservoir temperature, ºF

From a heat balance calculation, the reservoir sand temperature is as follows,

19,530
T2 − 100 = = 598.7
(0.255) (164) (1 − 0.22)
or T2 = 699!F (17.14)

The result from Eq. 17.14 does not agree with the assumed temperature of
1000oF. Calculations using other assumed temperatures result in calculated temperature
values shown in Table 17.2.

Table 17.2 Assumed and Calculated Temperatures as a Function of H/C Ratios

Assumed H /C Calculated
Temperatures Ratio Temperature
(°F) (°F)
1000 0.40 699
800 0.80 802
700 1.40 939

185
The tabular data is graphed as circles in Figure 17.7. The two temperatures match at
801°F. This is the calculated combustion front temperature if all the heat generated is
used to heat the formation behind (upstream of) the combustion front.

1000

900
Calculated Temperature - F
o

First Pass Heat


Balance
800 801 oF
Heat Balance
With Water
Effects
788 oF

700

600
600 700 800 900 1000 1100
o
Assumed Temperature - F

Figure 17.7. Assumed and calculated combustion zone temperatures.

4.3. Correction for Water of Combustion

These results don’t include all the processes occurring in the reservoir. First, the
water formed by combustion will condense beyond the combustion front, absorb some
heat of combustion and reduce the heat of the formation behind the combustion front.
This effect can be calculated as follows.

From Eq. 17.9, 0.0317 lb of H2 are formed per 100 lb of rock at 1000o F.
Assuming that a pound of steam will release 1,000 BTU when cooling from combustion
temperature and condensing (This number is not exactly correct, but is adequate for

186
estimation purposes.) the amount of heat carried forward by the steam is calculated using
concepts similar to Eqs. 17.10 and 17.12,

 18  (1000) (164) (1 − 0.22)


Heat carried ahead by steam = 0.0317  
2 100 (17.15)
= 365 BTU/ ft 3 Rock burned

In this equation, 18/2 is the ratio of molecular weight of water and hydrogen. The heat
given up by the steam is 1000 BTU/lb, and the other numbers are similar to those in Eq.
17.10. Thus, the calculated temperature is lower than it was in Eq. 17.14 as shown below,
19,530 − 365
T2 − 100 = = 587.5
(0.255) (164) (1 − 0.22)
or, T2 = 688! F (17.16)

Other temperatures were calculated similarly and the results, graphed as triangles in
Figure 17.7, show a corrected combustion temperature of 788ºF. At this temperature the
H/Cratio is 0.85 as indicated in Figure 17.8 below.

2.5

2
H/C Ratio (Atomic)

1.5
H/C Ratio 0.85 Atoms H/Atom C
At 788 oF
1

0.5

0
600 700 800 900 1000 1100
o
Temperature - F

Figure 17.8. H/C atomic ratio versus combustion temperature13

187
4.4. Calculating the Volume and Temperature of the Steam Plateau

No calculations on the steam plateau were necessary in the above calculations.


The steam plateau temperature and volume directly affect the volume of oil moved as a
result of the combustion process. To calculate these terms, use the H/Cratio of 0.85 and
calculate the partial pressure of the water as follows.

The fuel composition is CH0.85 . From Eq. 17.3 the moles of oxygen used per
mole of fuel are,
2m + 1 n 2( 20) + 1 0.85
O2 = + = +
2m + 2 4 2( 20) + 2 4 (17.17)
= 1.189 moles O2

Combustion products are calculated in a similar way,

m 20
CO2 = = = 0.952 Moles CΟ 2 (17.18)
m + 1 21
1 1
CO = = = 0.048 Moles CΟ (17.19)
m + 1 21
n 0.85
H 2O = = = 0.425 Moles H 2O (17.20)
2 2
 79 
N2 =   (O2 ) = (1.189 ) = 4.473
79
Moles N2 (17.21)
 21  21

The operating pressure is 300 psia. The partial pressure of H 2O in the combustion gas
is,
0.425 (300)
p water =
0.952 + 0.048 + 0.425 + 4.473
= 21.6 psia

From steam tables, the saturation temperature for 21.6 psia is 232ºF. This is the
temperature of the steam plateau.

The volume of the steam plateau is a function of the amount of H 2O formed.


Knowing that there are 0.95 lb C/100 lb rock burned, and knowing from Figure 17.8, that
the H / C ratio is 0.85, an equation similar to Eq. 17.10 yields the amount of water
formed per cubic foot of rock burned.

 0.85   18  (164) (1 − 0.22)


Water formed = 0.95   
 12   2  100
(17.22)
lbH 2O Formed
= 0.775
ft 3Rock burned

188
Thus the total heat carried forward by the water formed, is,

Heat carried by steam = 0.775 (1,000)


(17.23)
= 775 BTU/ft 3 rock burned

Using Eq. 17.13a, the heat capacity of the formation is,

232 − 100
c s = 0.21 + = 0.2166 BTU/ lb− o F (17.13c)
20,000

The amount of heat needed to raise a cubic foot of sand from 100ºF to 305ºF, from a heat
balance, is,
0.2166 BTU 164 (1 − 0.22) lb
 (232 − 100) F
∆H = !
o  3
lb- F  ft  (17.24)
= 3,701 BTU/ft 3

Thus the volume of rock heated by condensing steam is Eq. 17.23 divided by Eq. 17.24,

775 0.209 ft 3 Steam zone


Volume of steam heated rock = = (17.25)
3,701 ft 3 Rock burned

A sketch of the resulting temperature profile is shown in Figure 17.9.

Figure 17.9. Idealize Temperature Profile With Steam


Plateau Ahead of Combustion Front.

189
4.5. Calculating Effects of Injected Air and Water

Further corrections are needed to the temperature profile in Figure 17.9. Injected
air will partially cool the burned zone, rise in temperature as it approaches the
combustion front, and carry heat forward. This will have little effect on the combustion
kinetics or the amount of heat generated by combustion; so, in essence, this amount of
energy is merely carried forward to extend the size of the steam plateau.

A sketch of this idea is shown in Figure 17.10 below. In this sketch, the area
marked 1 is the temperature profile behind the burning front; area 2 is the steam plateau,
which is now larger than calculated before because of the heat carried forward by the
combustion gases.

Figure 17.10. Schematic of Combustion Temperature


Profile Including Cooling by Injected Air.

This temperature profile can be approximated as indicated in Figure 17.11 where


the profiles of the burned zone and steam plateau are treated as square waves that have
been adjusted so that the total heat in Areas 1 and 2 are the same as in the Figure 17.10.

Figure 17.11. Idealized Temperature Profile Including Injected Air Cooling.

190
There are several reasons for using this square wave concept. One is that it makes
it easier to calculate heat losses to be expected from either a laboratory or field
combustion operation using superposition calculations similar to those discussed by
Ramey 14 as seen in Prats’ Thermal Recovery Monograph2. These references also
indicate that the heat losses calculated using Figure 17.11 are quite adequate.

When wet combustion is used, the temperature behind the front tends to be a
sharp front, as shown in Figure 17.11. As a result, heat and material balances of the sort
discussed next, can be used to calculate the movement of the resulting cooling front, burn
front, and steam plateau.

The amount of air injected per cubic foot of rock burned and the heat capacity of
air are needed to calculate this heat transfer process for dry combustion. The volume of
combustion gas should also theoretically be calculated, but normally this isn’t necessary,
for its volume is nearly identical to the air volume. Further its heat capacity is nearly the
same – remember that most of the combustion gas is nitrogen.

The moles of air injected are calculated by adding the O2 from Eq. 17.17 to N2
from Eq. 17.21,
5.662 Mole air
Air injected = 1.189 + 4.473 = (17.26)
Mole fuel

Combining a heat capacity for air of 7.00 BTU/lb Mole - ºF with previously determined
factors of 0.95 lb of carbon burned for 100 lb rock, 164(1 - 0.22) pounds of rock per
cubic foot of rock, 12 lb of carbon per mole, a combustion zone temperature of 788ºF
(Figure 17.7), and the results of Eq. 17.26 yields the amount of heat carried forward by
the injected air as follows.

 0.95   (164) (1 − 0.22) 


 (5.662 )(7.00 )(788 − 100 )
Heat carried by air
= 
Cubic ft rock burned  100   12
(17.27)
2761 BTU
=
3
ft Rock burned

This heat extracted behind the burned zone is deposited into the steam plateau. The
resulting size of the steam plateau can be calculated in a way similar to Eq. 17.25 by
adding the heat carried by the combustion gas to that by the water, as follows,

775 + 2761
Vol of steam heated rock = = 0.901 ft 3 (17.28)
3701

This calculation shows the condensing steam front is far enough ahead of the combustion
front to displace oil efficiently and it is unnecessary to have the combustion front cover
the entire reservoir to get good recovery. Recovery can be estimated by knowing the

191
amount of fuel burned, by estimating the residual oil saturation in the steam plateau, and
by estimating the sweep efficiency of the process.

4.6. Heat Losses

An estimate of heat losses using the superposition concepts seen in Prats’ Thermal
Recovery Monograph2 based on Ramey’s 14 work, will make the calculations just
presented more accurate. These estimates are particularly important if a laboratory heat
balance indicates significant heat losses. The temperature profiles just calculated
assuming no heat losses can be used to make a first estimate of heat losses and a
recalculated steam plateau size. This is a reasonable way to handle the heat balance.
Since all heat transferred was assumed to be in the steam plateau, any reduction in that
transferred heat because of losses will reduce the amount of heat in the plateau.

As the size of the steam zone and the size of the calculated heat losses are
interdependent, iterative calculations are necessary until the assumed and calculated heat
balances match. This will usually require only two to three interations.

Data used in the previous calculations were based on generalized predictions of


combustion behavior, i.e., the amount of fuel per cubic feet of formation, and the H/C
ratio of the fuel. If combustion tube runs are made in the laboratory, those parameters are
known and can be used in the calculations. In addition, accurate temperature and
saturation profiles versus time will allow reasonably accurate heat balance calculations to
determine the heat losses from the experiment. As an alternative, reasonable assumptions
about the heat losses can be used to check the heat balance calculations and indicate if
there is significant experimental error.

Computer assisted tomography scanner measurements produce the most accurate


saturation histories. Alternatively, accurate measurements of temperature profiles and
accurate oil, water and gas production data also make it possible to estimate reasonable
saturation histories. These are the major sources of error in the overall heat balance
calculations, but are fairly small compared to the amount of heat stored in the hot matrix.

5. DESIGN CONSIDERATIONS

Conditions favoring the use of ISC rather than steam include the following: 1)
high reservoir pressure where steam is not efficient, 2) potential for severe well bore heat
losses (i.e., depth, offshore, permafrost), 3) reservoir clay swelling in contact with fresh
water, 4) limited water supply and 5) environmental regulations prohibiting steam
generation.
Like any other injection process, the design of ISC projects must consider
injection pressure limitations and reservoir flow resistance. These are especially
important in heavy oil reservoirs where combustion must occur in the high temperature
regime to be successful. The minimum air flux needed to maintain high temperatures at

192
the front is estimated to be 0.125 ft/day (0.04 m/day)15. As the burn zone growth is
directly proportional to the injected air, the maximum air injection rate determines the
minimum lifetime of the project. Ways to increase the air injection rate are often needed,
especially in heavy oil reservoirs. They may include reduced well spacing, cyclic
steaming of injectors and producers and an increase in injection pressure. These factors
will determine the compressor pressure and volume output.

There has often been some controversy over whether ISC projects should be
developed using patterns or line drives. Many early projects were started as pilots with a
single injector. Usually this resulted in a five-spot pattern. These pilots behaved contrary
to plan with the combustion front moving in only one direction because of permeability
variations, gravity effects, well spacing differences or a combination of these factors.

Attempts to correct the unbalanced flow included stimulating unresponsive wells


and limiting withdrawal rates of wells that producing excessive volumes of combustion
gas. Generally these efforts did not have the desired effect.

In retrospect, this reservoir behavior makes sense. Once a combustion front is


even slightly asymmetric, the higher temperature and thus higher mobility will cause
greater flow in that direction. Thus the flow will become more asymmetric, finally
resulting in flow principally in only one direction.

Since it is often difficult to decide, a priori, which direction the major flow will
take, operating plans should remain flexible until field performance indicates what
injection scheme best utilizes the flow directions.

For the above reasons many of the more successful ISC projects have been line
drive operations that start near the top of the reservoir and move downdip. In such an
operation, the direction of the fire front is known. The operating engineers can then plan
their completion and operating history in a rational way that will mirror the front
movement and breakthrough history. This operating practice can be seen in most of the
successful ISC field projects that will be discussed later.

6. PERFORMANCE PREDICTION

Predicting the production response to ISC has been the topic of various studies.
Complete numerical simulation of ISC is difficult because of the complex reactions and
the thin burning front that requires small grid blocks for representation. Simulators range
from tank models to complex three-dimensional simulators. In addition to simulation,
empirical models, hybrid models and correlation methods have been developed. A
discussion of some of these methods follows.

The easiest method is essentially a tank balance15 adapted by Prats2. The oil and
water produced are given by,
N p = φVb( Soi − S f) + 0.4 Soi (Vp − Vb) (17.29)

193
and
W p = Vbφ ( S wi + S wf) (17.30)

Where Soi = initial oil saturation, fraction


Sf = oil saturation burned, fraction

Vb = volume burned, m 3
Np = oil produced, m 3
Wp = water produced, m 3
φ = porosity, fraction
Vp = volume of the pattern, m 3
S wf = water saturation resulting from the combustion process, fraction

If the volumes are in acre-feet and the production terms are in barrels, a multiplication
factor of 7,758 must be used. The estimate of 40 percent of the oil produced coming from
outside the burned volume is an empirical value based on experience. This is the 0.4 term
in Eq. 17.29.

Figure 17.12, presented by Gates and Ramey16, combines laboratory results and
field observations from the Belridge ISC projects. It shows the effect of initial gas
saturation on the oil recovery history. Oil production rates and instantaneous air/oil ratios
can be estimated from the slopes of the curves. At late times the above two techniques
give similar results.

Brigham et al.17 used data from dry combustion field tests to obtain two empirical
correlations. Those are presented in Figs. 17.13 and 17.14. The terms in the ordinates are:
CIOP, cumulative incremental oil produced; OOIP, original oil in place; FB, fuel burned;
and OIP, oil in place at start of the project.

In addition to original oil saturation, So , thickness, h , oil viscosity, µo , and


porosity φ ; the abscissa includes, CAI, cumulative air injected, OIP, oil in place at start
of the project, and fraction oxygen utilization. The second correlation, Figure 17.14, is
the most accurate except for oils of less than 10 cp viscosity where the first correlation
must be used. These correlations were generated from pilot floods; thus, they would not
be expected to be accurate for pattern flooding. However, the narrative in the previous
section points out that pattern flooding is generally not the best way to operate an ISC
project, and these correlations are expected to be reasonably accurate for line drive
projects.

194
Figure 17.12. Gates and Ramey Method16

X axis legend numerator should


read
CAI x (Fraction O2 Utilization)

Figure 17.13. First Satman correlation17

195
X axis numerator
should read
fraction rather
than percent

17.14. Second Satman correlation17

7. OPERATING PRACTICES

In addition to the standard field equipment for oil production, ISC requires
particular attention to air compression, ignition, well design, completion, and production
practices.

7.1. Compressors

Air compression systems are critical to the success of any ISC field project.
Failures of the past can often be traced to poor compressor design, faulty maintenance or
operating mistakes. A detailed discussion of compressors and sizing considerations
appears in Chapter 7 of the Facilities Section of this handbook. Other discussions are
available in Sarathi1.

The factors to be considered when selecting compressors include peak air


requirements, injection pressure, capital cost, power requirements, operation and
maintenance costs and other relevant technical and economic parameters specific to the
field considered. Compressor terminology varies among manufacturers. It is best to
obtain a complete description including compressor, driver, interstage cooling system,
and all ancillary equipment including control and safety systems from each vendor being
consulted.

Air compression causes high temperatures because of the high c p / cv ratio.


Compressor design must consider these high temperatures to ensure continuous,
sustained operations free from the corrosive effects of air and the explosion hazards of
some lubricating fluids. Mineral oils are not recommended. Synthetic lubricants

196
withstand the higher temperatures and offer lower volatility and flammability than
conventional lubricants.

7.2. Ignition

Ignition and maintenance of high combustion temperatures, especially in heavy


oil projects, are the most critical factors of a ICU project. Shallcross18 presented a
complete review of ignition methods. The following is a summary of this study.

Ignition can occur spontaneously if the oil is reactive, the reservoir temperature
high enough, and the reservoir is reasonably thick. Various models have been proposed to
determine the time for spontaneous ignition19,20.

When spontaneous ignition does not occur or is not desired (i.e., in heavy oil
reservoirs where it is important to maintain high combustion temperatures), the most
appropriate ignition method to use depends on the reservoir and the equipment available
on site.

Down hole gas-fired burners allow good control of the temperature of injected
gases and may be operated at a greater depth than other methods. The disadvantages
include the need to run multiple tubing strings in the injection wells. Some particles, such
as soot, may be carried into the formation if the gas does not burn cleanly.

Catalytic heaters run at lower temperatures but are sometimes prohibitively


expensive. Electrical heaters can be lowered with a single cable, and can provide
excellent temperature control. They can be reused repeatedly. There is, however, a depth
limitation because of electrical power losses in the cable.

Chemically enhanced ignition does not have a depth limitation but may require
handling and storage of dangerous materials. Fuel packs are not recommended because of
poor temperature control and nonuniform ignition across the entire reservoir thickness.
Well damage from elevated temperatures and plugging by particulate matter may occur.

Steam may be used to locally increase reservoir temperature and facilitate auto
ignition. It suffers from depth limitation because of wellbore heat losses, but when the
conditions are right it can be a very simple and effective method for ignition.

The Papers and Reports Section includes details of design and implementation of
the above methods.

7.3. Well Design and Completions

ISC wells must be designed to account for several factors amplified by the
combustion process, namely high temperature, corrosive environment and sand and clay
control. Safe operations should be the primary concern.

197
Typical well designs for injection and production are shown in Figs. 17.15 and
1
17.16 . Completion type and design depends on the reservoir being considered.
Laboratory testing for sand control and completions can help to determine the best
completion technique for a given field. Care has to be taken to properly cement the wells.
There are cement formulations that are stable at high temperatures21. Open hole
completions may be used in conjunction with slotted liners, screens, gravel packs or
various other sand and clay control methods. To maximize productivity, producing wells
should be completed toward the bottom of the zone of interest to take advantage of
gravity drainage and avoid hot gases as long as possible. Rat holes have been used
successfully in certain heavy oil combustion projects to increase the effect of gravity
drainage22.

Figure 17.15. Typical injection well design1

7.4. Injection and Production Practices

Safe air injection requires that the surface injection equipment and the injection
well are free of hydrocarbons. All lubricants used in compression and down hole
operations should be synthetic or non hydrocarbon types. All equipment, tools, lines,
tubing, work strings and injection strings must be clean and hydrocarbon free. Personnel
at all levels should be aware of the importance of preventing hydrocarbons in the
injection wells. As a safety measure to protect injection wells if the compressor is shut
down, a system to prevent backflow of oil from the formation must be present at every
injection well.

Downhole temperatures in producing wells increase as displaced oil, hot water


and steam fronts reach the well. Producers are preserved by downhole cooling and proper
material selection. Figure 17.17 provides an estimate of the water requirements to
maintain bottom hole temperature no higher than 250ºF as a function of oil and water in
the lower section of the producing zone to maximize gravity segregation in the

198
Figure 17.16. Typical production well design1

reservoir. In many cases, after the combustion front has moved through the well it is
possible to convert the former producer to a new air injector, thus realizing significant
cost reductions over the life of the project.

Monitoring is crucial for proper combustion operations. In addition to testing


individual producers for oil and water rates, injected fluids must be measured. Also,
produced gases must be measured and analyzed to determine the efficiency of the
combustion operation. Down hole temperature measurements are essential to calculate
the size and location of the burned zone. Flow line temperatures can indicate thermal
stimulation or down hole problems.

Figure 17.17 Water needed to cool hot wells1

199
Combustion projects generate waste water, flue gases and pollutants from
compression and oil handling equipment. Local pollution disposal regulations must be
consulted prior to the design of any insitu combustion operation.

In general, environmental problems are similar to those posed by steam injection.


The produced water may contain H 2 S and/or CO2 which may require special handling
and anti-corrosion equipment. Flue gases may contain hydrocarbons, H 2 S , CO2 , CO
and other trace amounts of sulfur gases. Table 17.31 summarizes the various pollution
control systems suitable for combustion projects and their recommended applications.
Sarathi1 also provides detailed descriptions of the various types of systems and their uses.
Some other problems that can be encountered are sand production, corrosion, emulsions,
well failures or compressor failures.

Table 17.3. Application of Pollution Control Systems to a Fireflood Project1

Gas Pollutant Method Suggested


Equipment Treated Removed Application

Flare stack Flue gas None Vent to atmosphere Flue gas meets air
quality regulations
Flare stack with Flue gas H/C (1) ,CO (1) , Burn Flue gas with
flame burner enough H/C to
S-gases (1,2 ) support
combustion
(>200Btu/Scf)
Thermal Flue gas H/C, CO, Burn Flue gas not
incinerator S-gases (1,2 ) suitable for a flare
with burner or a
catalytic
incinerator (heat
value > 85 Btu/Scf
but < 200 Btu/Scf)
Catalytic Flue gas H/C, Co, Burn Flue gas with heat
incinerator (excess S-gases ( 2 ) value < 85 Btu/Scf
air)
H 2 S scrubber Flue gas H 2 S (4) Chemical reaction Flue gas
containing H 2 S
but acceptable
amounts of other
S gases
SO 2 scrubber Incinerator SO 2
(4) Chemical reaction When H 2 S
exhaust removed from flue
gas is inadequate
or impractical
1. Removal efficiency may be poor
2. S-gases are converted to SO 2
3. May increase amount of CO
For flue gas with heat values > Btu/Scf
4. Typical removal efficiency is 90-95%.

200
8. FIELD EXPERIENCE

Insitu combustion has been used in the field since 1920. In the U.S. over 230
projects have been implemented. Many of those were technically and economically
successful. Unfavorable reservoir and fluid characteristics, poor design and engineering
or operational problems caused failures. Most of the failed projects were small pilot
projects implemented in unfavorable reservoirs. Worldwide, combustion accounts for
about 10% of the oil produced by thermal methods. Steam injection accounts for the rest
and disussed in Chapter 16. Twenty nine projects are active as of 19981. Most of the
projects outside of the U.S. are large heavy oil projects while the current trend in the U.S.
is to use ISC in deep, lighter oil reservoirs where water flooding or steam flooding are
not effective. Brief comments on these projects follow.

8.1. Heavy Oils

For oil with 20º API gravity or less, 19 projects using ISC were active in 1998.
Some general comments apply:
• Most of these projects last a long time; projects initiated in the 1960’s are still
active. Economics of successful projects are favorable even as compared to steam
and water flooding23.
• All of the successful projects operate in the high temperature mode.
• Gravity override and channeling do occur. Gravity drainage of the hot oil is an
important mechanism and should be maximized. Frequently, improved production
of oil continues after the air injection has been terminated.
• Line drive projects, starting at the top of the reservoir and moving downward,
exhibit superior performance compared to repeated pattern projects.
• Most of the projects failed when air injection was attempted in different layers of
the reservoir at the same time. Air injectivity is a critical parameter and injectivity
contrasts between layers is usually too difficult to overcome.

As a detailed description of all or even a few of the projects is outside the scope
of this narrative the reader can find additional information on several current projects in
the following references:
• Projects in Romania: Supplacu de Barcau is the world’s largest combustion
project. It started in 1964 and is operated in a line drive mode from the top
downward. Videle and Balaria are other ISC projects.24,25
• In India, Balol, started as a pilot in 1990, was expanded to the whole field.
Designed as a wet combustion project, the water injection rate had to be cut in
half because of too much cooling. This project was also changed from patterns to
updip line drive because of premature breakthrough in the producers.26,27
• Projects in Russia, Kasakhstan, and Azerbaijan, are not very well described in
western literature.28
• The Albanian project of Kasnice is described in Marko et al.29.
• Batrum in Canada is a successful Mobil project using horizontal wells as
producers. Eyehill field is another Mobil project with horizontal wells.27,30
Wabaska also uses the same concept of horizontal wells as producers.31

201
Horizontal wells in insitu combustion operations have been successfully used in
Canada since 1993.

Cyclic applications such as pressure up and blow down, have been described31,32.
This operational technique allows production from very low mobility oil fields or tar
sands where fracturing or cyclic steaming are needed prior to air injection. The Wabaska
project is a cyclical combustion project with horizontal wells. This type of pressure up,
blow down, technique has also been successfully implemented at Wolf Lake32. Air was
injected until the front arrived at the producers. When the front reached a given producer,
this well was shut down and cooling water circulated. When all the producers were shut
down, injection was stopped and the producers reopened to blow down the reservoir. This
process was repeated for several cycles. Operating combustion in this fashion allows
production from fields where injectivity is low because of a high crude viscosity at
reservoir conditions.

U.S. projects at Bellevue and Midway-Sunset have been described27. More details
on the Midway-Sunset project can be found in Hoffmann.30 Ramey, et al.22describe the
Belridge project as an economic success.

Insitu combustion in heavy oil reservoirs has been successful in both the dry and
wet mode. Dry combustion early in the life of the project is the preferred method to form
the desired high temperature regime. When the process is well established, moderate
amounts of water can be added to improve efficiency. Quenched or super wet combustion
seems to have limited success except when used at the end of a field operation to
scavenge the heat remaining in the rock.

Another operating variation includes the use of enriched air or pure oxygen.
Oxygen enriched combustion presents technical and economic advantages for reservoirs
with high pressure or very low injectivity. It has been successfully demonstrated in the
field34. The additional literature covering the special handling methods and additional
precautions needed for enriched air injection is listed in the Paper and Reports Section of
the bibliography. Commercial application of the oxygen technology has been limited
because of oil price variations.

8.2. Light Oils

In situ combustion is used in light oil reservoirs for two purposes:


• To reduce the viscosity of unconventional light oils such as Demjen in
Hungary36or Niemangu in China1. In these cases, thermal effects are
important. In the case of Demjen, a catalyst had to be injected to promote
combustion. Iron was used to increase the amount of fuel burned because the
light oil by itself was not depositing enough fuel to sustain combustion. The
oil is parafinic and almost solid at reservoir temperature despite an API
gravity of 32º.

202
• To produce from light oil reservoirs where water flooding or other enhanced
oil recovery methods are not attractive. Combustion is used to generate flue
gases for reservoir pressure maintenance and production by gravity drainage.
Thermal effects are only minor for this process. An interesting case is the
West Hackberry double displacement process37 in which the gas cap is
expanded for gravity drainage to recover residual oil after waterflood.

Combustion is successful in tight carbonate reservoirs, located in the Dakotas,
such as Medicine Pole Hill, Buffalo, west and south Buffalo38 and Horse Creek.39 In
those cases, combustion allows exploitation of thin reservoirs with large well spacing.

8.3. Screening Guidelines

Insitu combustion is a complex process. It combines effects of steam drive,


hydrocarbon miscible and immiscible flood, immiscible gas drive and hot and cold water
flood. Because of its complexity there is a misconception that combustion has a low
probability of success. The truth is that combustion is an economically attractive, proven
recovery process, capable of economically recovering a large fraction of the oil in place.

Insitu combustion can be applied to many different reservoirs. Some suggested


screening guidelines are:
• Nature of the Formation : The rock type is not important providing that the
matrix/oil system is reactive enough to sustain combustion. As in any drive
process, high permeability streaks are detrimental. Swelling clays may be a
problem in the steam plateau area.
• Depth: Depth be large enough to ensure containment of the injected air in the
reservoir. There is no depth limit, except that this may affect the injection
pressure.
• Pressure: Pressure will affect the economics of the process, but does not
affect the technical aspects of combustion.
• Temperature: Temperature will affect auto ignition but is otherwise not
critical.
• Reservoir Thickness: Thickness should be greater than 4m (15 ft) to avoid
excessive heat losses to surrounding formations. Very thick formations may
present sweep efficiency problems because of gravity override.
• Permeability: This has to be sufficient to allow injection of air at the designed
air flux. The air injectivity is especially important for heavy oil reservoirs.
Conditions are favorable when kh / µ is greater than 5md m/cp3.
• Porosity and Oil Saturation: These have to be large enough to allow
economic oil recovery. The product, φ So , needs to be over about 0.08 for
combustion to be economically successful.
• Oil Gravity: This parameter is not critical. Insitu viscosity has to be low
enough to allow air injection and resulting oil production at the design rate.

203
• Oil Nature: In heavy oil projects the oil should be readily oxidizable at
reservoir and rock matrix conditions. This relationship must be determined by
laboratory experiments. The same laboratory experiments can also determine
the amount of air needed to burn a given reservoir volume. This is key to the
profitability of the process.

9. CONCLUSIONS

In situ combustion can be applied to a wide array of reservoirs. In fact it is the


only thermal method that can presently be applied to deep reservoirs, though deep
downhole steam generation is being tested. It can be used at any stage of reservoir
depletion. It can be used in special situations such as offshore or in arctic regions.
Because of the lack of heat losses at the surface and in the injection wells, it is the most
thermally efficient thermal recovery method. The injectant (air) is readily available.
Combustion allows wider well spacing than steam. Economic results are comparable to
those of steam injection.

Several aspects of operating ISC projects are important. First is the large
compression ratio and associated costs required to inject air into the formation. Second is
planning and design requirements for a combustion project that are more difficult than
steam. Third is extensive laboratory work to assess fuel availability, air requirements and
burning characteristics of the crude that are required before designing insitu combustion
projects. Fourth is the high degree of technical sophistication, and careful monitoring
needed to ensure proper operation of a project. Fifth is the limitation of numerical
simulation and other techniques that makes predictions of recovery more difficult than
most other enhanced oil recovery methods.

Considerable improvements in the application of insitu combustion have been


made since the early projects. New developments, such as application to light oil
reservoirs, and the use of horizontal wells are reviving interest in ISC. This process
deserves consideration for many reservoirs including those in hostile environments or not
amenable to other recovery methods.

204
10. REFERENCES

General

1. Sarathi, P.: “In-Situ Combustion Handbook Principles And Practices”, Report


DOE/PC/91008-0374, OSTI ID 3175, January 1999.

2. Prats, M.: “Thermal Recovery”, Society of Petroleum Engineers Monograph 7


(1982).

3. Burger, J., Sourieau P. and Combarnous M.: “Thermal Methods Of Oil Recovery”
Editions Technip, Paris, France, 1985, re-edited in 1989.

Other References in Text

4. Wu, C.H. and Fulton, F.P.: “Experimental Simulation of the Zones Preceding the
Combustion Front of an In-situ Combustion Process.” Soc. of Pet. Eng. Jour. (March
1971) pp. 38-46.

5. Boussaid, I.S.: “Oil Recovery by Multiple Quenched In-situ Combustion.” SPE


16739, paper presented at the 62nd Annual SPE Technical Conference, Dallas
(September 27-30, 1987).

6. Mamora D., Ramey, H.J., Jr, Brigham, W.E. and Castanier, L.M.: “Kinetics of In-
situ Combustion,” SUPRI TR 91, US DOE Report DOE/BC/14600-51
(DE93000152) July 1993; also Ph.D thesis, Stanford University.

7. Burger, J. and Sahuquet B.: “Chemical Aspects of In-situ Combustion –Heat of


Combustion and Kinetics.” SPEJ 12, 5 (October 1972).

8. Moore, R.G.: “New Strategies for In-situ Combustion.” J Can. Pet. Tech. 32, 10
(December 1993).

9. Fassihi, M.R., Brigham, W.E. and Ramey, H.J., Jr.: “The Reaction Kinetics of In-
situ Combustion” SPE 9454, paper presented at the 55th Annual SPE Meeting, Dallas
(September 1980).

10. Moore, R. G., Laureshen, C.J., Belgrave, J., Ursenbach, M., and Jha, K.N.: “A
Comparison of the Laboratory In-situ Combustion Behaviour of Canadian Oils.”
Proceedings 6Th UNITAR Conference, Houston (February 1995).

11. Bardon, C. and Gadelle, C.: “Essais de Laboratoire pour l’etude de la Combustion In-
situ,” French Soviet Symposium on Enhanced Oil Recovery, Moscow (May 1977).

205
12. Leaute, R.P. and Collyer, C.J.: “Laboratory Studies of Insitu Combustion with Cold
Lake Crude,” Paper No 5, Annual conference on Upgrading Technology and
Petroleum Recovery, Calgary, Canada (1984).

13. Dew, J. N. and Martin, W.I.: “How to Calculate Air Requirements for In-situ
Combustion,” Petroleum Engineer (December 1964 and January 1965).

14. Ramey, H.J., Jr.: “Discussion of Reservoir Heating by Hot Fluid Injection” Trans.
AIME 216 (1959) pp. 364-365.

15. Nelson, T.W. and McNeil, J.S.: “How to Engineer an In-situ Combustion Project,”
Producer Monthly (May 1961); also Oil and Gas Jour. (June 5, 1961).

16. Gates, C.J and Ramey, H.J., Jr.: “A Method for Engineering In-situ Combustion Oil
Recovery Projects,” J. Pet Tech. (February 1980).

17. Brigham, W. and Satman, A.: “Recovery Correlations for In-situ Combustion Field
Projects and Application to Field Pilots.” J. Petroleum Technology 32 (1980) pp.
2132-2138.

18. Shallcross, D. C.: “Devices and Methods for In-situ Combustion Ignition ,” SUPRI
TR 69, US DOE report DOE/BC/14126-12 (DE 89000766) October, 1989.

19. Burger, J. G.: “Spontaneous Ignition in Oil Reservoirs,” Soc. Pet. Eng. Jour. 16
(January 1976).

20. Tadema, H.J and Weidjeima, J.: “Spontaneous Ignition of Oils,” Oil and Gas Jour.
68, 50.

21. Smith, D.K.: “Cementing,” SPE Monograph Volume 4 (1976).

22. Ramey , H. J., Jr., Stamp, V.V., Pebdani, F.N.: “Case History of South Belridge,
California, Insitu Combustion Oil Recovery,” SPE 24200, paper presented at the 9th
SPE/DOE EOR Symposium, Tulsa (April 21-24 1992).

23. Ramage, W.E., L.M. Castanier and H.J. Ramey, Jr.: "Economic Evaluation of
Thermal Recovery Projects," SPE 16852, paper presented at the 62nd Annual
Technical Conference and Exhibition, Dallas (September 27-30, 1987).

24. Machedon, V.: “Suplacu de Barcau Field: 28 Years of Insitu Combustion.”


Proceedings 7th European Symposium on Enhanced Oil Recovery, Moscow (1993).

25. Machedon, V.: “Romania 30 Years of Experience in Insitu Combustion.”


Proceedings NIPER/DOE Symposium on In-situ Combustion, Tulsa (April 21-22,
1994); also US/DOE report NIPER/BDM-0086 (January 1995).

206
26. Roychaudurhi, S.: “Experience with In-situ Combustion Pilot in Presence of Edge
Water.” Proceedings, 6th UNITAR Conference on Heavy Crudes and Tar Sands,
Houston (1995).

27. Moritis, G.: “EOR Oil Production Up Slightly.” Oil and Gas Jour. (April 20,1998)
pp. 49-77.

28. Mamedov,Y.G. and Bocserman, A.A. “Development of Heavy Oils and Bitumen in
the Former Soviet Union and Eastern and Central Europe.” Proceedings 6Th
UNITAR Conference on Heavy Crudes and Tar Sands, Houston (February 12-17,
1995).

29. Marko, D.: “Oil Production History in Albanian Fields and their Perspective.”
proceedings. 6Th UNITAR conference on Heavy Crudes and Tar Sands, Houston
(1995).

30. Ames, B.G., Gramms, R.E. and Pebdani, F.N.: “Improved Sweep Efficiency Through
Application of Horizontal Well Technology in a Mature Combustion EOR.”
Proceedings NIPER/DOE Symposium on In-situ Combustion, Tulsa (April 21-22,
1994); also US/DOE report NIPER/BDM-0086 (January 1995).

31. Fassihi, M.R. and Marjerisson, D.M. “Performance of Morgan Pressure Cycling
Insitu Combustion Project,” SPE 27793, paper presented at the 9th SPE/DOE
Symposium on Enhanced Oil recovery Tulsa (April 17-20, 1994).

32. Hallam, R. J. and Donelly, J.K.: “Pressure-Up Blow Down Combustion.” SPE
18071, paper presented at the 63rd Annual SPE Fall Meeting, Houston (1988).

33. Hoffmann, S. J.: “Successful Application of Insitu Combustion to a Dipping Heavy


Oil Reservoir Midway Sunset field.” SPE/DOE 39639, paper presented at the
SPE/DOE Symposium, Tulsa (1998).

34. Hvizdos, L.J., Howard, J.V. and Roberts, G.W. “Enhanced Oil Recovery Through
Oxygen Enriched In-situ Combustion.” J.P. Tech. (June 1983) pp. 1061-1070.

35. Moore,R.G., Bannion,D.W., Belgrave, J.D. and Ursenbach, M.G.: “New Insight into
Oxygen Enriched In-situ Combustion.” SPE 16740, paper presented at the 62nd
Annual SPE Conference, Dallas (September 27-30, 1987).

36. Racz, D.: “Development of a Thermocatalytic Insitu Combustion Process in


Hungary.” Proceedings 1985 European Meeting on Improved Oil Recovery, Rome,
Italy (April 1985).

37. Gilham, T.H., Cerveny, B.W., Turek, E.D. and Yanimaras, D.V.:; “Keys to
Increasing Oil Production from a Light Oil Reservoir via Air Injection.” SPE 38848,
paper presented at the SPE annual conference San Antonio (October 5-9, 1997).

207
38. Miller, R.J.: “Koch’s Experience with Deep In-situ Combustion in Williston Basin.”
Proceedings NIPER/DOE Symposium on In-situ Combustion, Tulsa (April 21-22,
1994); also US/DOE report NIPER/BDM-0086 (January 1995).

39. Germain, P, and Geyelin, J.L.: “Air Injection Into a Light Oil Reservoir; The Horse
Creek Project SPE 37782, paper presented at the SPE Middle East Oil Show,
Manama, Bahrein (March 15-18, 1997).

208
Upgrading Via In-situ Combustion

L. M. Castanier and W.E. Brigham

ABSTRACT

Crude oil upgrading is of major economic importance. Heavy crude oils exist in
large quantities in the western hemisphere but are difficult to produce and transport
because of their high viscosity. Some crude oils contain compounds such as sulfur and/or
heavy metals causing extra refining problems and costs. In-situ upgrading could be a
very beneficial process.

In-situ combustion has long been used as an enhanced oil recovery method. For
heavy oils, numerous field observations shown upgrading of 2 to 6 degrees API for heavy
oils undergoing combustion (Ramey, 1992) . During in-situ combustion of heavy oils,
temperatures of up to 700 degrees C can be observed at the combustion front. The
purpose of this note is try to examine the changes caused by the various reactions
occurring during combustion with emphasis on the upgrading potential for various oil
types. We will first describe the process with an emphasis on the temperature changes
and chemical reactions which can either upgrade or degrade the oil in terms of its API
gravity and viscosity. In the second part, we present results from three combustion tube
runs aimed at sulfur removal. We will end by discussing possible in-situ upgrading
strategies using combinations of in-situ combustion and other improved recovery
methods.

1. BACKGROUND AND PROCESS DESCRIPTION

In-situ combustion is injection of an oxidizing gas (air or oxygen enriched air) to


generate heat by burning a portion of the oil. Most of the oil is driven toward the
producers by a combination of gas drive (from the combustion gases) and steam and
water drive. This process is also called fire flooding to describe the movement of the
burning front inside the reservoir. Based on the respective directions of front propagation
and air flow, the process can be forward, when the combustion front advances in the
same direction as the air flow, or reverse, when the front moves against the air flow.

As only forward combustion is practiced in the field we will only consider this
case. Forward combustion can be further characterized as “dry” when only air or
enriched air is injected or “wet” when air and water are co-injected.

Air is injected in the target formation for a short time, usually a few days to a few
weeks and the oil in the formation is ignited. Ignition can be induced using down hole
gas burners, electrical heaters, and/or injection of pyrophoric agents ( not recommended)

209
or steam injection. In some cases auto-ignition occurs when the reservoir temperature is
fairly high and the oil reasonably reactive. This often occurs in California oils.

After ignition the combustion front is propagated by a continuous flow of air. As


the front progresses into the reservoir, several zones can be found between the injector
and the producer as a result of heat and mass transport and the chemical reactions
occurring in the process. Figure 1 (Wu and Fulton, 1971) shows a schematic
representation of the locations of the various zones and temperature and fluid saturation
distributions. In the field there are transitions between all those zones so Fig. 1 is only an
idealized representation.

Starting from the injector seven zones can be defined:

(A) The burned zone is the volume already burned. This zone is filled with air and
may contain small amounts of residual unburned organic solids. As it has been
subjected to high temperatures, mineral alterations are possible. Because of the
continuous air flow from the injector to the burned zone, temperature increases
from injected air temperature at the injector to near combustion front temperature
near the combustion front. There is no oil left in this zone.

(B) The combustion front is the highest temperature zone. It is very thin, often no
more than several inches thick. It is in that region that oxygen combines with the
fuel and high temperature oxidation occurs. The products of the burning reactions
are water and carbon oxides. The fuel is often misnamed coke. In fact it is not
pure carbon but a hydrocarbon with H/C molar ratios ranging from about 0.6 to
2.0. This fuel is formed in the thermal cracking zone just ahead of the front and is
the product of cracking and pyrolysis which is deposited on the rock matrix. The
amount of fuel burned is an important parameter because it determines how much
air must be injected to burn a certain volume of reservoir. One should note that
the fuel burnt is the heaviest part of the oil. As we will see later it is formed by
the steam distillation, cracking and pyrolysis.

(C) The cracking/vaporization zone is downstream of the front. The oil remaining in
that zone is the residual oil left behind the steam plateau. The crude is modified
in this zone by the high temperature of the combustion process. The light ends
vaporize and are transported downstream where they condense and mix with the
original crude. The heavy ends pyrolyse, resulting in CO2 , hydrocarbon gases
and solid organic fuel deposited on the rock.

(D) The steam plateau is next downstream. Most of the oil is displaced forward. The
immobile oil undergoes steam distillation. Further downstream some of the
hydrocarbon vapors condense, mostly as the steam condenses. The steam plateau
temperature depends on the partial pressure of the water in the gas phase.
Depending on the temperature the original oil may undergo a mild thermal
cracking, often named visbreaking, in this zone. That usually reduces the oil
viscosity.

210
(E) At the leading edge of the steam plateau, where the temperature is less than steam
saturation temperature, a water bank is formed which decreases in temperature
and saturation as we look downstream, with a resulting increase in oil saturation.

(F) An oil bank is next. This zone contains most of the oil displaced from upstream,
including most of the light ends that resulted from the thermal cracking upstream.

(G) Beyond lays the undisturbed original reservoir. Gas saturation will only increase
slightly in this region because of the high mobility of combustion gases. The
production mechanism in this area is gas drive from the combustion products.

2. CHEMICAL REACTIONS

Chemical reactions are of two main categories. Oxidations occur in the presence
of oxygen while pyrolysis is caused mainly by elevated temperatures.

Oxidation can be of two different natures (Mamora, 1993). In general at low


temperatures oxygen combines with the oil to form oxidized hydrocarbons such as
peroxides, alcohols or ketones This generally increases the oil viscosity but could
increase oil reactivity at higher temperature. When oxygen contacts the oil at higher
temperature, combustion occurs resulting in production of water and carbon oxides.
Pyrolysis is the chemical alteration of hydrocarbons under the effect of heat. The
reactions depend on the chemical structure of the oil and can be dehydrogenation where
the number of carbon atoms in the molecule remains the same, cracking which reduces
the number of carbon atoms or condensation which increases the number of carbon
atoms. Dehydrogenation is usually preponderant for short chain hydrocarbons while
cracking is the main reaction for longer chain molecules. At temperatures over 500
degrees centigrade, the products of pyrolysis are light hydrocarbons and a coke like
residue.

In the steam plateau, the composition of the oil will change (Venkatesan and Shu
1984; Aksinat, 1983). Sulfur removal by formation of H2S often occurs at this stage.
Figure 2 (Monin and Audibert, 1984) shows the effect of steam and of the rock matrix on
the changes in composition of two oils heated to 300ºC. One can notice a decrease in
asphaltene content and the formation of a solid coke like substance called pyrobitumen.

Distillation allows transport and production of the light fractions of the oil (Lim et
al., 1992) leaving behind the heavy ends. These heavy ends often contain the majority
of undesirable compounds which may contain sulfur or metals.

Of all the reactions that can occur during in-situ combustion, only low
temperature oxidation can increase the viscosity of the oil. (Moore et al., 1994) If the
fireflood is conducted properly low temperature oxidations are minimized because most
of the oxygen injected is consumed at the burning front.

211
Abu Khamsim et al. (1985) presented a good literature review of fuel formation
and pyrolysis during in-situ combustion . They also present results of experiments aimed
at understanding fuel formation. They emphasized the importance of distillation and of
the nature of the crude. They found that fuel formation started with distillation followed
by visbraking and pyrolisis. They confirmed that the fuel left is mainly composed of the
heavy fractions of the oil.

From a multi-year study (Castanier and Brigham, 1996) we know that water
soluble metallic additives cause changes in the nature and the amount of fuel formed
during in-situ combustion. These changes appear to depend on the type of oil used.
Various crude oils are affected differently by different additives. The efficiency of
various metals as additives does not follow the usually cited order for cracking reactions
of coke Cu<V<Fe<Cr=Zn<Ni. For the cases studied, iron and tin seem to be the
additives which best increase fuel deposition and promote combustion. Unfortunately no
additive was found as yet to decrease fuel lay down. This method was successfully used
in the field in Hungary (Racz et al., 1985) using iron to increase fuel content.

3. EXPERIMENTS FOR SULFUR REMOVAL

Sulfur can be a major problem for refiners. To investigate the possible use of in-
situ combustion to improve sulfur content of produced oil three combustion tube runs
were performed using a matrix made of 95% Ottawa sand and 5% fire clay. All three
runs were made using a light Middle East oil containing 6% by weight sulfur. In the
runs, oil saturation was about 20% and water saturation about 20%. The oil, water and
matrix were premixed and packed in the combustion tube. A schematic of the equipment
is shown in Fig. 3 (after Mamora et al., 1993). The first run was a simple dry forward
combustion run. As the combustion was not good, we decided to use metallic additives.
The second run was made using a solution of 5% by weight of stannous chloride in the
water to increase the amount of fuel deposited. The third run contained 5% iron nitrate in
the water.

3.1. Run Without Additive

Temperature profiles are shown on Fig. 4. One can observe some fluctuations
and an average front temperature of about 450ºC. Figure 5 shows the effluent gas
composition. Again fluctuations are seen and the oxygen utilization is poor with an
average of 7 mole % fraction of oxygen produced. This shows that the combustion is not
efficient and not enough fuel is burned. Figure 6 shows the front locations and fuel
concentrations as a function of time. Again, wide fluctuations in the amount of fuel
burned are observed. The original oil contained 6% by weight of sulfur, the average
sulfur content of the produced oil was about 3%. Most of the removed sulfur was found
as sulfuric acid in the produced water.

212
3.2. Runs with Water Soluble Metallic Additives

Tin chloride and Iron nitrate were dissolved in the water at 5% by weight
concentration. The experimental procedure and parameters were similar to those of the
previous run. As the behavior of both runs was similar we will only present in detail the
results of the run where tin chloride was used.

The temperature profiles as a function of time are shown on figure 7. The average
front temperature was 500ºC . This represents an increase of about 50º over the run
without additive. Figure 8 shows the produced gas composition. The combustion was
stable with only minor fluctuations. Almost all of the oxygen injected was consumed .
Fuel concentration (Fig. 9) was also stable at a value around 0.04 grams per cubic
centimeter of the pack. This run was stopped after 4 hours because a leak developed in
the thermal well.

The run with iron nitrate was made after repairing the leak and the results were
similar to the run made with tin chloride. In both cases, the amount of sulfur in the
produced oil was reduced to less than 1% by weight from the original 6%.
Although it was necessary to use metallic additives to promote fuel deposition and obtain
a good burn of this light oil, sulfur removal by in-situ combustion appears promising.
We can also postulate that heavy metals, sometimes present in the oil, can be removed in
a similar manner.

3.3. Possible In-Situ Upgrading Strategies

Forward in-situ combustion is already an effective in –situ upgrading method.


Field improvements in gravity by as much as 6º have been mentioned in the literature. In
the following we will present some hybrid methods that could be used to further upgrade
the oil in-situ.

In a recent article (Moore et el., 2001), a method of upgrading Athabasca bitumen was
proposed. It consists in a combination of low temperature oxidation followed by higher
temperature oxidation. They showed that dividing the oxidation process into two steps
can result in an overall reduction of viscosity. Low temperature oxidation will occur
during the first step, and by raising the temperature the pre oxidized oil will react more
easily at higher temperature, and cracking of the heavy oxygenated compounds will
become easier.

This process can be achieved, for example, by injecting air and then by following
the air slug with steam to raise the temperature. Xia and Graves (2001) describe
experiments combining air injection and horizontal wells. Using SARA (Saturates,
Aromatics, Resins, Asphaltenes) analysis they show significant upgrading of bitumen
during their three dimensional experiments. From a bitumen of 1 million Mpas viscosity
they produce oil of viscosity between 500 and a 1000 Mpas. The I gravity improved by

213
up to 8 degrees API. The SARA analysis showed a dramatic increase in saturates for the
produced oil. Although direct application of these laboratory results to the field is
difficult, such technology deserves further research.

Another possible in-situ upgrading technique involves a combination of solvent


injection and combustion. Cyclic oil recovery has numerous advantages both technically
and economically. It can also be easily optimized in a given oil reservoir. We propose
cyclic injection of solvents, either gas or liquid, followed by in-situ combustion of a small
part of the reservoir to increase the temperature near the well bore but also to clean the
well bore region of all the residues left by the solvents. Alternate slugs of solvent and air
will be injected and production will occur after each solvent slug injection and after each
combustion period. The process can be repeated until an economic limit is reached. One
important fact to note is that both solvent injection and in-situ combustion have been
proven to be effective in a variety of reservoirs, however, the combination of the two
methods has never been tried to our knowledge.

We expect that in most cases, the amount of solvent to be injected will be a small
fraction of the pore volume of the reservoir considered. By mixing with the oil the
solvent will decrease the viscosity of the mixture below the viscosity of the original oil.
The improvement in oil production can be estimated using an approach similar to the one
used by Boberg (1998) for steam cycling and taking into account only the change of
viscosity caused by the solvent . Production after the injection of the first slug of solvent
can thus be easily estimated.

At the end of the first solvent cycle, one can expect some damage to the near
wellbore area. Most significant will be the precipitation and/or deposition of heavy
hydrocarbons such as asphaltenes or paraffins. The produced oil is expected to be
slightly upgraded by the solvent cycle. Unlike the classic well to well in-situ combustion,
we would only try to improve near well bore conditions by burning the solid residues left
after the solvent cycle. The amount of air (or oxygen) to be injected can easily be
determined by laboratory experiments and numerical simulation or simple calculations.
The benefits of using combustion at this stage are expected to include:

• Productivity improvement through removal of the heavy ends left from the
solvent cycle.
• Possible deactivation of the clays near the well bore due to the high temperature
of the combustion.
• Reduced viscosity of the oil due to temperature increase.
• Effect of CO2 production causing swelling and viscosity reduction of the oil
When CO2 partial pressure is high enough
• pressure maintenance

One must note that combustion burns only the heaviest part of the crude and
hence the upgrading already observed during the solvent cycle should be maintained and
even improved. A soak period will probably be necessary at the end of the air/oxygen
injection to ensure that all of the oxygen injected will have reacted in the reservoir and

214
will not be produced during the subsequent production cycle. The recovery predictions at
this stage of the process will be difficult and probably require a fine grid
thermal/compositional simulation. It is however possible to obtain some estimates based
on simple correlations such as Brigham, et al. (1981), or to perform simple analytical
approximations similar to the well known steam cyclic models (Boberg, 1998).

The process can be repeated by alternating solvent injection and near well in-situ
combustion as long as the economics warrants it. Optimization of slug sizes and duration
needs to be addressed to ensure maximum economic efficiency. Reservoir conditions
will, of course, dictate these parameters.

4. CONCLUSIONS

Because of the nature of the process, in-situ combustion is a viable in-situ


upgrading method. For heavy oils significant upgrading occurs by the fact that the heavy
ends are burned while most of the light ends are produced. The use of water soluble
metallic additives allows in- situ combustion of light oils. This can be quite beneficial for
removal of sulfur or other metals. More research is needed in that area. Several possible
new in-situ upgrading technologies have been discussed. They include use of steam and
air injection, combinations of horizontal wells and air injection and combinations of in-
situ combustion with solvent injection. All of those techniques should be further
examined for technical and economic feasability.

5. BIBLIOGRAPHY

Books

1. Boberg, T.C.: Thermal Methods of Oil Recovery, New York. John Wiley and Sons
(1998).

2. Burger, J., Sourieau P. and Combarnous M.: Thermal Methods of Oil Recovery
Editions, Technip, Paris, France (1985) re-edited in 1989.

3. Prats M.: “Thermal Recovery,” Society of Petroleum Engineers Monograph 7


(1982).

4. Sarathi P. “In-situ Combustion Handbook Principles and Practices” Report


DOE/PC/91008-0374 OSTI ID3175 (January 1999).

Papers

5. Aksinat, M.H. “Gas Evolution and Change of Oil Composition During Steam
Flooding of Oil Reservoirs” Jour. Petroleum Geology, 5 (1983) pp. 363-388.

215
6. Alexander, J.D, Martin W.L and Dew J.N. “ Factors Affecting Fuel Availability and
Composition During In-situ Combustion “ J. Pet. Tech. Trans AIME 225 (October
1962).

7. Brigham, W., and Satman, A.. “”Recovery Correlations for In-situ Combustion
Field Projects and Application to Field Pilots.” J. Petroleum Technology 32 (1980)
pp. 2132-2138.

8. Burger J. and Sahuquet B.: “Chemical Aspects of In-situ Combustion –Heat of


Combustion and Kinetics,” SPEJ, 12, no. 5 (October 1972).

9. Burger J. G.” Spontaneous Ignition in Oil Reservoirs,” Soc Pet Eng. Jour., 16
January (1976).

10. Castanier L.M. and Brigham W.E “Modifying In-situ Combustion with Metallic
Additives” In situ (January 1996).

11. Dew J. N. and Martin W.I “How to Calculate Air Requirements for In-situ
Combustion” Petroleum Engineer (December 1964 and January 1965).

12. Fassihi, M.R., Brigham, W.E. and Ramey, H.J., Jr.: “The Reaction kinetics of In-
situ Combustion” SPE 9454, Proceedings 55th Annual SPE Meeting, Dallas
(September 1980).

13. Hoffmann, S. J.: “Successful Application of Insitu Combustion to a Dipping Heavy


Oil Reservoir Midway Sunset field.” SPE 39639, Proceedings Symposium (Tulsa,
(1998).

14. Leaute R.P. and Collyer C.J.: “Laboratory Studies of Isitu Combustion with Cold
Lake Crude.” Paper No 5, Annual conference on Upgrading Technology and
Petroleum Recovery, Calgary, Canada (1984).

15. Lim K.T., Ramey H.J.m Jr. and Brigham W.E.: “Steam Distillation Effects and Oil
Quality Changes During Steam Injection.” US DOE report DOE/BC/14600-18
(January 1992).

16. Mamora D., Ramey H.J Jr, Brigham W.E. and Castanier L.M.: “Kinetics of In-situ
Combustion.” SUPRI TR 91, US DOE Report DOE/BC/14600-51 (DE93000152)
(July 1993); Also Ph.D thesis, Stanford University.

17. Monin, J.C. and Audibert, A.: “Alteration of Heavy crudes under High Pressure
Steam Conditions. ” Am chem. Soc. Div. Petroleum Chem. Preprints 29, No 4
(1994).

18. Moore R.G.: “New Strategies for in-situ combustion” J Can. Pet.Tec. 32, No 10
(December 1993).

216
19. Moore R. G., Laureshen C.J., Belgrave, J., Ursenbach, M., and Jha, K.N.: “A
Comparison of the Laboratory In-situ Combustion Behaviour of Canadian Oils.”
Proceedings 6Th UNITAR Conference, Houston (February 1995).

20. Moore R.G., Xu, H.H., Okazawa N.E., Mehta S.A., Laureshen, C.J, Ursenbach
M.G., and Mallory D.H.: “In-situ Upgrading of Heavy Oils.” Journal of Canadian
Petroleum Technology (August 2001).

21. Moore R. G., Laureshen, C.J., Mehta, S.A., Belgrave, J., Ursenbach, M, Weissman,
J. and Kessler, R.: “A Downhole Catalytic Upgrading Process for Heavy Oil Using
In-situ Combustion.” JCPT 38, No. 13 (1999).

22. Ramey , H. J., Jr, Stamp, V. V., Pebdani, F.N.: “Case History of South Belridge,
California, Insitu Combustion Oil Recovery.” SPE 24200, Proceedings 9th SPE/Doe
EOR Symposium,Tulsa (April 21-24 1992).

23. Racz, D.: “Development of a thermocatalytic in situ combustion process in


Hungary,” Proceedings 1985 European Meeting on Improved Oil Recovery, Rome
(April 1985)

24. Venkatesan, V.N. and Shu, W.R.: “Alteration of Heavy Oil Characteristics During
Thermal Recovery” CIM 84-35-108, Proceedings Annual meeting CIM (June
1984).

25. Wu, C.H. and Fulton, F.P.: “Experimental Simulation of the Zones Preceding the
Combustion Front of an In-situ Combustion Process” Soc. of Pet Eng Jour. (March
1971) pp. 38-46.

26. Xia, T.X. and Greaves, M.: “Downhole Upgrading Athabasca Tar Sand Bitumen
Using THAI –SARA Analysis.” SPE 69693, paper presented at SPE 2001 Heavy
Oil Operation Symposium, Porlamar, Venezuela (March 2001).

217
Fig. 1. In-Situ Combustion: Schematic Diagram of Temperature and
Saturation Profiles and Zones (After Wu and Fulton, 1971).

218
Fig. 2. Chemical alteration of heavy crudes after 200 h heating at 350°C;
atmosphere of helium at about 4 Mpa, or atmosphere at 18-20 Mpa
containing argon or water vapor. Oil density at 25°C is 999 kg/m 3 for Oil A
and 946 kg/m 3 for oil B. Oil B contains essentially no sulfur.
(a) Material balance (in weight percent of the initial crude oil).
(b) Composition of oil fractions C 12 + .

219
Fig. 3. Schematic diagram of kinetic and combustion tube apparatus.

220
Fig. 4. Average gas saturation in the core versus time for an oil withdrawal rate
of 1.44 cm3/day (5.1x10-5 ft3/day), IN population balance model. The
number of bubbles nucleated is 10,000 bubbles/ft3 of rock (0.353
bubbles/cm3).

221
Fig. 5. Gas bubble size with time for an oil withdrawal rate of 1.44 cm3/day
(5.1x10-5 ft3/day), IN population balance model. The number of
bubbles nucleated is 10,000 bubbles/ft3 of rock (0.353 bubbles/cm3).

222
Fig. 6. Plot of average pressure versus time illustrating the effect of varying the expansion rate,
IN population balance model. The number of bubbles nucleated is 10,000 bubbles/ft3 of
rock (0.353 bubbles/cm3).

223
Fig. 7. Match of the IN bubble population balance model to the experimental pressure
24
data of Firoozabadi et al. . Withdrawal rate is 1.44 cm3/day (5.1x10-5 ft3/day).

224
Fig. 8. IN population balance model prediction of average gas saturation along length
of core. Withdrawal rate is 1.44 cm3/day (5.1x10-5 ft3/day).

225
Fig. 9. Match of the IN bubble population balance model to the experimental pressure
24
data of Firoozabadi et al. . Withdrawal rate is 7.2 cm3/day (2.5x10-4 ft3/day).

226
RESERVOIR DEFINITION

227
228
Integrating Production History into Reservoir Models Using
Streamline-Based Time-of-Flight Ranking

Y. Wang and A. R. Kovscek

ABSTRACT

The reservoir models generated by geostatistical techniques provide equally


probable reservoir descriptions that honor observed geology. However, flow simulation
results on these models may vary widely indicating uncertainty. Constraining
geostatistical models to known production history reduces uncertainty. To this end, a
streamline-based algorithm for ranking geostatistical realizations with regard to
production history is proposed. First, a rapid, streamline-based inversion method is
applied to obtain a history-matched reservoir model. Then, unit mobility ratio streamline
geometries are computed, without full flow simulation, for the history-matched model
and the geostatistical models considered. Each model is then compared to the history-
matched model with regard to streamline properties, such as time-of-flight. In this way, a
reservoir model that matches production history and honors known geological
information is obtained. An example using 500 distinct reservoir models demonstrates
computational efficiency and also shows that the method easily selects the most
appropriate permeability fields. Flow simulations confirm that selections are satisfactory.

1. INTRODUCTION

Geostatistical techniques are used primarily to generate realistic equally-probable


numerical reservoir models that honor known geological information. However,
significant uncertainty exists in the detailed distribution of reservoir properties such as
permeability (Deutsch and Srinivasan, 1996) because the known data is relatively sparse.
If flow simulation is performed on various realizations to predict reservoir performance,
the geological reservoir model uncertainty, in turn, is transferred into uncertainty of
simulation results. In many instances, the variability in flow simulation results among
geostatistical realizations is large.

When no production history is available, uncertainty in reservoir performance can


be evaluated by conducting flow simulations on various geostatistical models.
Approaches to speed up the process of comprehensive flow simulation on a great number
of models have been discussed in the literature. Deutsch and Srinivasan (1996) reviewed
three categories of methods for ranking geostatistical reservoir models. They are: (1)
statistical methods, (2) methods based on approximations to flow simulation, and (3)
flow-simulation-based methods for a simpler process, e.g., on coarsened models. In the

229
second category, Ballin et al. (1992) selected models for comprehensive simulation by
using a fast simulator to rank realizations. The fast simulation can be, for example, a
tracer simulation or a simplified flow model. To address further the issue of uncertainty,
Cruz et al. (1999) introduced the concept of a quality map that is obtained by running
flow simulation many times with a single well at varying locations. A number of quality
maps can be generated for different geostatistical models, and uncertainty can be
evaluated. The quality map can be then used for decision-making. This method can be
computationally intensive.

If production history is present, integration of the data with geostatistical


realizations can reduce uncertainty. In turn, future reservoir performance should be better
predicted. However, no general and computationally efficient method for production data
integration exists currently. Barker et al. (2000) discussed the application of three
approaches to constraining geological realizations to production data: rejection sampling,
pilot-point method, and Markov Chain Monte Carlo (MCMC). The latter two approaches
use inverse modeling techniques in conjunction with finite difference based flow
simulator. In general, on the order of thousands of flow simulations were needed to
achieve tens of matched models for all three methods. Similar to the pilot point method,
the sequential-self-calibration technique has been adapted to the inverse problem of
integrating multiwell transient pressure and production data with reservoir models (Wen
et al. 1998a, b). On the other hand, Yoon et al. (2000) presented a multiscale approach to
production-data integration using streamline models. In their method, multiscales of
coarseness within reservoir descriptions are used to account for the varying resolving
power of different production data. A scale by scale inversion of production response is
achieved with an increase in computational efficiency. A streamline simulator is used for
flow simulation and a streamline approach is used to compute the sensitivities of
production data to reservoir model parameters.

In other work, Wang and Kovscek 2000 proposed a streamline approach for
history-matching production data (Wang and Kovscek, 2000) Streamline properties are
used in the inverse method to obtain the permeability changes needed to match
production data. The method works in two steps: (1) the error between measured and
computed producer water cut is related to an error in effective permeability of
streamlines; and (2) the perturbation in streamline effective permeability is computed
analytically and mapped into a change of permeability for individual grid cells. This
method applies well in water-flood cases with few or no infill wells to be matched. A
streamline flow simulator, 3DSL (Batycky et al., 1997), is used for forward flow
simulation in the history matching process, but a standard finite difference based
simulator could also be used. Agarwal and Blunt (2001) applied this same method to
cases with gravity and compressibility. Data from a North Sea Field was used to test the
technique with reasonable history match and prediction of future performance. In both
cases, the history-matched reservoir models were not constrained to geological
information. Subsequently, Caers et al. (2001) combined Wang and Kovscek’s method
with a Gauss Markov random function (GMrf) technique to constrain reservoir models
jointly to production data and geological information. Other streamline-based history-

230
matching methods are discussed in literature (Gupta et al., 1994; He et al., 2001; Vasco et
al., 1998; Wen et al., 1998).

It is commonly known that both the history matching process and geostatistics are
inherently non-deterministic. It is not appropriate to use only one geologically
constrained and history-matched reservoir model to characterize reservoir performance.
It is desirable to have a suitable number of such models so that the uncertainty of the
prediction can be evaluated. Even though the method of Caers et al. (2001) is efficient in
generating a reservoir model that honors both the production history and geology, the
generation of many such models may still be computationally involved. Thus, we propose
a rejection sampling type approach to rank or sort geostatistical realizations. This
provides an additional process to constrain our streamline-based history matching method
(Wang and Kovscek, 2000) to observed geology. Further, it provides a means to select
rationally a subset of reservoir models that honor observed production data from a larger
suite of realizations.

It is proposed here that a reservoir model be obtained by unconstrained history


matching to the known production as shown by Wang and Kovscek (2000). Then a
desired number of reservoir models are generated using geostatistical methods (e.g.
Deutsch and Journel, 1992). These geostatistical models are thus consistent with any
known geology, whereas the unconstrained history match provides a model consistent
with production. Next, streamlines and their time-of-flight (SL-TOF) distributions are
computed (Blunt et al., 1996) for the unconstrained history-matched model and the
reservoir models to be ranked. Full flow simulation is not required for this step. The SL-
TOF distribution of each geostatistical model is then compared to that of the history-
matched model. The models are ranked based on the SL-TOF cdf errors. This ranking is
proven below to be the same as the ranking in terms of water-cut error at corresponding
producers, Errors in the flow-rate, pressure and cumulative oil production are ranked in a
similar way. We proceed by first defining a number of terms and then turn to the task of
correlating the distribution of time of flight among streamlines to observed water-cut
behavior. A synthetic example is used to illustrate the method.

2. TERMINOLOGY

Before continuing, some terms are defined so that the discussion can be concise.

PVI: Pore volume injected. In this context, it refers to displaceable pore


volume replaced by water.

SL-TOF: Streamline time of flight. More precisely, it refers to the


dimensionless time of flight. SL-TOF is equivalent to the pore-
volume-injected (PVI) along the streamline.

231
cdf: Cumulative distribution function. The cdf characterizes the probability
that the value of a variable is less than a particular value. It is the integral
of the probability density function.

CFS: Comprehensive flow simulation on a reservoir model.

M: Mobility ratio. The endpoint mobility of the displacing phase divided by


the displaced phase endpoint mobility.

SL: Streamline.

Water-cut/fractional-flow: The injected-phase fractional-flow or injected


component concentration at a producer. The two terms are used
interchangeably.

CIPP: Cumulative initial-phase production, that is, cumulative oil production.

The ideal set-up for our method requires information in the following two
categories, First, known geology, including direct measurements of reservoir properties at
discrete locations. Therefore, many equal-probable reservoir models can be generated
using any appropriate geostatistical approach. Secondly, known well production history,
including water-cut and injection data. The ideal cases are incompressible water flooding
and tracer flow study.

3. EXHAUSTIVE SIMULATION AND ERROR CORRELATION

Streamline trajectories and other properties are correlated with permeability


distribution, and so streamline properties bear on reservoir performance. The various TOF
of streamlines that connect an injector and producer can be summarized as a cdf.
Intuitively, the streamline time of flight (SL-TOF) cdf is correlated with water-cut history.
Similarly, flow rate and pressure distribution at one time correlates with that at other
times. As cumulative initial-phase production can be computed from water-cut and flow-
rate, a term that is equivalent to CIPP, equivalent CIPP, can be computed by combining
SL-TOF cdf and flow-rate. Compared to the comprehensive flow simulation, computation
of streamlines, SL-TOF cdf, and equivalent CIPP takes significantly less execution time.
Substantial computational work can be saved if reservoir models are ranked (i.e.,
rejected) according to streamline properties rather than simulation results. However, the
direct correlation between ranking realizations with streamline properties and with CFS
must be established. This is the purpose of the remainder of this section. First, exhaustive
simulations on 51 geostatistical realizations are conducted. One realization is used as
"truth". Unit and unfavorable mobility ratio flow simulations are performed to explore
the variation in results among the models and truth. Phenomenonlogically, a direct
correlation between SL-TOF cdf and producer water cut is shown. Second, a brief
theoretical explanation is given that confirms the correlation.

232
3.1. Correlation of TOF Results

Throughout, we refer to the difference between a realization and truth as "error".


Within streamline computations, errors include SL-TOF cdf, flow rate, pressure drop, and
equivalent CIPP. The error of a physical term of a model with respect to that of the
reference model is computed as follows. Take the SL-TOF cdf error as an example.
Compute the SL-TOF distributions for the reference model and the model to be checked.
Discretize the two distributions at identical TOF values. Then the error is the average of
the absolute values of the differences at all the discrete points. Therefore, each type of
the error of a model is quantified by a single value.

To verify the correlation of errors numerically, fifty-one models, represented by


permeability fields, are generated using sequential-Gaussian simulation (SGSIM)
(Deutsch and Journel, 1998) with fixed permeability values at well locations as shown in
Table 1. Each permeability field is 50 by 50 grid-blocks. Table 2 lists the other details of
the flow simulations. Each model is ranked compared to the truth case for three pairs of
error type: SL-TOF cdf and water-cut from CFS, flow-rate from SL simulation and flow-
rate from CFS, and equivalent-CIPP and CIPP. Streamline properties for ranking models
are always computed for unit mobility ratio. A confined five-spot pattern is used for all
displacements

Results for forward simulations with M=1and M=2.5cases are presented below.
For unit mobility ratio, total flow-rate in CFS is constant and the same as that in the
ranking method. Streamlines do not vary in time. The only concern is the correlation of
water-cut from CFS and SL-TOF cdf. Figure 1a shows the correlation between SL-TOF
cdf error and water-cut error for M=1. Figure 1b shows the correlation between SL-TOF
cdf error rank and water-cut error rank. The correlation coefficients are both above 0.97.
The error among models correlates well between CFS and unit-mobility ratio streamlines.
Importantly, the relative rank also correlates. Models that are far from the truth case
produce large error whether computed from CFS or strictly from SL-TOF.

For mobility ratio M=2.5 , Figs. 2a, 2b, and 2c show the correlation between SL-
TOF cdf error and fractional-flow error from CFS, between the error of the equivalent
flow-rate in ranking and the error in flow-rate from CFS, and between the error of
equivalent-CIPP and error of CIPP from CFS, respectively. Figure 2d shows the
correlation of error rank between each pair. The correlation coefficients are all above
0.97 as shown in Fig. 2. Therefore, the correlation is almost perfect in terms of both error
and error rank. Again, the relative rank correlates among these pairs of errors.

In Fig. 1, SL-TOF cdf error and fractional flow-error correlate with unit slope for
M=1. That is, the difference in performance between two reservoir models is gauged
from either the difference in water-cut curves or the difference in the distribution of SL-
TOF. For non-unit mobility ratio, the slope is not unity. However, as shown in Fig 2, the
errors among CFS and SL calculations correlate linearly with a bias close to zero.
Properties, such as TOF and streamline geometry, computed for unit mobility ratio are

233
indicative of the results on the same permeability field when the displacement is not unit
mobility.

3.2. Correlation--Theoretical Development

The correlation of different error pairs is made obvious by the substantial flow
simulation results and ranking results as shown in Figs. 1 and 2. The equations derived
below give a quantitative explanation of this correlation.

For unit mobility ratio cases, streamline trajectories and the pressure distribution
are constant. Flow is equivalent to miscible displacement or tracer injection. Due to the
fact that each streamline carries the same amount of flow, the amount of injectant that is
produced increases by 1/nwhere( n is the number of streamlines) when each streamline
breaks through. The SL-TOF cdf also increases by 1/nwhen each streamline breaks
through. Therefore, the SL-TOF cdf is the same as the fraction of injected phase
produced. As shown in Fig. 3, the SL-TOF cdf agrees with fractional-flow very well at
all producers for a confined five-spot pattern where permeability heterogeneity is present.
The same argument holds for non-unit mobility ratio cases because the
breakthrough of each streamline contributes an incremental amount to the water cut if
each streamline carries the same flow-rate. However, in the ranking process, the error
SL-TOF cdf is computed under unit mobility ratio while the error of water-cut in
simulation result is computed under true (non-unit) mobility ratio. Therefore, we need to
show that the SL-TOF cdf error for unit mobility case is linearly correlated with the SL-
TOF cdf error for non-unit mobility case that is equivalent to the water-cut at the same
mobility ratio.

Assume n streamlines are launched between an injector-producer pair. A


streamline can be treated as the center of a streamtube. Throughout this discussion,
streamline and streamtube are used interchangeably. Equations are derived, first for unit
and then for non-unit mobility ratio, by treating streamtubes as non-communicating layers
and following the work of Dykstra and Parsons (1950) (Dykstra and Parsons, 1950; Lake,
1989).

For unit mobility ratio and piston-like displacement, the total PVI when the first
streamtube breaks through is the same as the dimensionless TOF of the first streamline,
τ D1 ,
n
v1 + ∑ vi x
i= 2
fDi
v1 + v'
τ D1 = = (1)
V V

where vi is the pore volume of the streamtube i, xfDiif the dimensionless position of the
displacement front in streamtube i, V is the total pore volume, and v' is total swept pore

234
volume of all but the first streamtubes, which is defined as
n
' ' '
v = ∑ vi x fDi= V x fD (2)
i =2

where x 'fDis the pore- volume averaged displacement front of those streamtubes, and V'is
the total pore volume of all but the first streamtubes as defined below.
n n
'
V = ∑ vi = v1 + ∑ vi = v1 + V (3)
i=1 i= 2

Each streamline carries the same volumetric rate q. Hence, the flow-rate of the
first streamline is qand that of the rest is (n-1)q
. For homogeneously distributed porosity,
the following can be obtained from material balance
'
v (n − 1)q (4)
= = n −1
v1 q

Combining the Eqs. 2-4, we obtain

1 n −1 (5)
v1 = V, v' = V
1 + (n − 1) / x 'fD 1 + (n − 1) / x 'fD

Plug Eq. 5 into Eq. 1 and rearrange,

n−1
x ' fD = τ D1 ≈ τ D1 , ∀ large n (6)
n − τ D1
Then
τ D1 τ D1
v1 = V, v' = (1 − )V (7)
n n

For non-mobility ratio, the dimensionless position of the displacing front can be
expressed in terms of mobility ratio and the dimensionless distance of displacing front for
unit mobility case
M   1   (8)
x
*
= 1 − 1 + x '  − 1  fD
fD
M − 1  M 2  
where superscript * denotes the non-unit mobility ratio case.

The PVI at the first streamtube breakthrough is


v1 + v' x *fD
τ *
D1 = (9)
V
Plug Eqs. (7) and (8) into (9) and simplifying for large n, we obtain

M   1  
τ *D1 = 1 − 1 + τ D1  2 − 1  (10)
M − 1  M  

235
For moderately favorable or moderately unfavorable mobility ratios, this
approximation is accurate. For example, τ D1 and τ D* 1 are 0.425 and 0.335 respectively in
the simulation result on a permeability field for M=2.5 . The τ D* 1 computed using the
above formula is 0.330.

Given the difference in τ D1 from that of the reference model, τ D1 R


, for unit mobility
ratio, the formula to compute the error of τ D* 1 from τ D*R1 can be derived from Eq. 10,

∆τ *D1 = τ *D1 − τ *R
D1 =
1
M −1 [ M (1 − τ
2
D1 ) + τ D1
R R
− M (1 − τ D1 ) + τ D1
2
] (11)

where ∆τ D* 1 is the error of dimensionless TOF of streamtube 1. Superscript R denotes the


reference.

Note that ∆τ D1 = τ D1 − τ DR1 where ∆τ D1 is the error of dimensionless TOF of


streamtube 1at unit mobility-ratio, and let c = M 2 (1 − τ DR1 ) + τ DR1 , we obtain

c  (M − 1)(M + 1) 
∆τ *D1 = 1 − 1 − ∆τ D1  = a(1− 1− b∆τ D1 ) (12)
M − 1  c 

where c (M − 1)(M + 1) with c defined above. The terms a, b, and c are all
a= , b=
M −1 c
constant. Based on 0 < τ D* 1 < 1, we know that b∆τ D1 < 1. Using a power series
expansion, we obtain

1 1×1 1×1 × 3 
∆τ D* 1 = a  (b∆τ D1 ) + (b∆τ D1 ) 2 + (b∆τ D1 ) 3 + • • • (13)
 2 2 × 4 2 × 4 × 6 

Usually, b∆τ D1 is small and the higher order terms can be omitted, thus
ab M +1 (14)
∆τ *D1 ≈ ∆τ D1 = ∆τ D1
2 2 M (1 − τ D1
2 R
) + τ D1
R

Therefore the relationship between ∆τ D* 1 and ∆τ D1 is approximately linear. Their


absolute values are mutually rank-preserving, i.e., the error rank in ∆τ D* 1 is the same as
that in ∆τ D1 . This correlation is confirmed by Fig. 2.

The above equations are derived for the first-breakthrough streamline. This
relation also holds for other streamlines. That is, the difference of the ith streamline
breakthrough-time of two given reservoir models at mobility ratio M is a linear function
of that when computed using M=1. The errors computed for different mobility ratios are
rank preserving. As the fractional-flow at a producer is the same as the SL-TOF cdf at
the real mobility ratio, the error of the fractional-flow at true mobility ratio and the error

236
of SL-TOF cdf computed using M=1is also rank preserving. To compute the fractional-
flow error, a CFS is required. However, the only computational effort required to
compute the SL-TOF cdf errors is equivalent to that of one time-step of a full simulation.
Obviously, the SL-TOF cdf error ranking is a tremendous savings of computation. This
savings is due to conversion of the complexity in the temporal coordinate as in water cut
history to complexity in the spatial coordinate as in SL-TOF spatial distribution. The
conversion is possible because they are highly correlated.

4. STREAMLINE-BASED RANKING – ALGORITHM

As elaborated above, the error rank in SL-TOF cdf, flow-rate, pressure, and
equivalent-CIPP are correlated with error in the fractional-flow, flow-rate, pressure, and
CIPP as computed by CFS, respectively. The ranking procedure requires computation of
the pressure field and streamlines only once for each model. Above, the difference in
results was computed between a realization and some reference. Below differences (i.e.
errors) in SL-TOF are computed between a realization and a history-matched model.

The ranking procedure is as follows:


• Start from an initial reservoir model and perform a history-matching process (Wang
and Kovscek, 2000) to obtain an unconstrained history-matched reservoir model.

• For the history-matched model and each of the models to be ranked, compute pressure
distribution for unit mobility ratio, compute streamlines geometries, and then compute
SL-TOF cdf for each producer. We used Pollock’s method to compute streamline
geometries (Pollock, 1998).

• For each of the reservoir models, compute the difference between its SL-TOF cdf, or
the other quantities of interest, and that of the history-matched model. This
computation is performed for each injector-producer pair.

• Rank the reservoir models based on the error type of interest. For instance, if we
desire to match the fractional-flow at all the producers, then SL-TOF cdf ranking
should be used. Also, combinations of the three terms can be used for ranking. For
example, flow-rate and fractional-flow can be combined in ranking to indicate fits in
cumulative oil production. Some combinations make more physical sense than the
others. The following four pairs are perfectly correlated as proven above: (1) SL-TOF
cdf and fractional-flow in CFS; (2) flow-rate in ranking and the flow-rate in CFS; (3)
pressure in ranking and the pressure in CFS; and (4) equivalent CIPP and CIPP.

• Select a suitable number of models within a specified maximum error and run CFS on
them to evaluate the match in simulation results.

The selected models can be then used to predict future reservoir performance and
to evaluate the uncertainty of this prediction.

237
5. EXAMPLE APPLICATIONS

In the examples below, we concentrate on heterogeneity in the permeability field.


Five hundred reservoir models are generated by the geostatistical routine, SGSIM
(Deutsch and Journel, 1998). They are conditioned to the permeability values at wells as
listed in Table 1. The reference field is excluded from the 500 samples. Permeability in
all the models is isotropic. The simulator, 3DSL (Batycky et al., 1997), is used for
forward simulation. Note that any reservoir simulator might be used here. The choice of
3DSL is based on its rapid execution times and availability. Two cases, one with unit
mobility ratio and the other with unfavorable mobility ratio (M=2.5 ), are tested and
presented below.

5.1. Five-Spot, M=1

First, the reference production history is obtained by running a CFS on a phantom


reference reservoir model (Fig. 4a) with the simulation parameters shown in Table 2.
This model is not available in practice and here is used only to obtain production data.
Then starting from another initial model generated by SGSIM, the fast streamline-based
history matching method is used to generate a history-matched model. No geological
constraints are applied while history matching. Figure 4 shows the phantom reference
reservoir model and the history-matched model as expressed by permeability
distributions. Figure 5 shows the match of the fractional-flow at all four producers
obtained in six iterations. Other terms, i.e. pressure and flow-rate are also well matched.
Obviously, the match to the production history is excellent, but the history-matched
model does not honor the geology as expressed by the permeability distribution.

The error of each geostatistical model with respect to the history-matched model
is computed. Here, 500 such models are evaluated. For unit mobility ratio, the match in
fractional-flow is most difficult and important, and therefore, only the ranking results in
SL-TOF cdf are discussed in this case. After ranking, CFS was conducted on the five
best-matched models. Figure 6 shows these fractional-flow curves and the reference
fractional-flow curve. The match is satisfactory. Figure 7 shows the fractional-flow of
five worst matched models. The difference between the best and worst is obvious.

5.2. Five-Spot, M=2.5

The production history is generated by the same method and the same phantom
reference model as in the previous case, but with end-point mobility ratio equal to 2.5.
Simulation parameters are listed in Table 2. A history-matched permeability field is
obtained in the same way as for the previous case. This permeability field is very similar
to that in Fig 4 (b). The same ranking procedure is performed. The ranking result in
fractional-flow error is very similar to that of the first case. Thus, we concentrate here
only on the match in equivalent CIPP to avoid repetition. Figure 8 shows the CIPP (in

238
PVI) versus time (in PVI) of the five best-matches as well as the five worst-matches to
the reference data.

Figure 9 shows the permeability fields of the best and worst matched models in
SL-TOF cdf and equivalent-CIPP. The model shown in Fig 9a and other models that are
best-matched in TOF cdf have a common feature—there is a high permeability region
connecting the injector in the center to Producer 1 at the lower-left corner of the model.
This results in earlier breakthrough of the injected-phase at Producer 1as shown in Fig. 6.
This trend is also observed in the reference model in Fig 4a. On the contrary, the worst-
matched models have a low-permeability barrier between this injector-producer pair as
shown in Fig. 9b.

While the SL-TOF cdf indicates the correlated heterogeneity of the permeability
distribution, the equivalent-CIPP represents swept-area, and the average permeability
between each injector-producer pair. Figure 9c shows the model that is best-matched in
equivalent-CIPP. The permeability is relatively high in the lower-left quarter of the
model. It agrees with the reference and the history-matched model in Fig. 4. Other well-
matched models have the same feature. However, in the worst-matched models (Fig. 9d,
the permeability in this region is relatively low.

6. DISCUSSION

The method just described has the potential to reduce greatly the uncertainty in the
prediction of future reservoir performance. Consider again a confined five-spot pattern
after 0.585 PV of water has been injected. The simulation parameters are the same as the
M=2.5case, except that a new reference permeability field is used to generate synthetic
production data. At 0.585 PVI, the average water-cut of the four producers is about 0.3.
Again, 500 reservoir models are ranked based on the match to SL-TOF cdf following the
procedure above and using water-cut data only up to 0.585 PVI. Figure 10 shows the best
matches to the partial production history. Evidently, the match to the history is very
accurate and the future prediction is satisfactory. Figure 11 shows the reference
permeability field and the five models selected as best matching the production history up
to 0.585 PVI. The common feature among the models is the lack of a high permeability
channel or streak connecting the injector in the center to any of the producers at the four
corners. This results in a fairly uniform breakthrough time for each producers of about 0.5
PVI, as shown in Fig. 10.

We observed, not surprisingly, that the accuracy of predicting reservoir future


performance is a monotonically increasing function of the amount of production history
used to select the models. To ensure satisfactory prediction, production history up to the
point of an average water-cut of 0.3 was needed to select models with accuracy. This
translates roughly to 0.6 PV, or more, of injection.

239
The remainder of this discussion is devoted to issues related to error distribution
and computational efficiency. It is observed from the ranking results that the error of
each term follows approximately a normal distribution. Therefore, the criteria for
selecting good models or rejecting bad models can be expressed by a constant k. That is,
models that have errors k times standard deviation lower than the average error are
accepted, while the rest are rejected. Given that the distribution is normal, the percentage
of models that prove acceptable can be determined solely by k. If the number of desired
models within a tolerance is known, then the number of models to rank is easily
determined. For instance, we can choose those models whose error is two standard
deviations lower than the average error. The probability for this selection is 2.23 percent
given the normal distribution and the skewness correction, if needed. Take the fractional-
flow and the cases above as an example. The mean of the error in SL-TOF cdf
is µ = 0.0736 and the standard deviation is σ = 0.0149 . Set the criteria for accepting a
model as µ − 2σ = 0.0438 . Eleven out of the 500 (2.2%) models sampled meet the criteria.
Table 3 gives the number of models that meet the selection criteria when a different
number (50, 100, 150, 200, 250, 300, and 500) of models are ranked. Therefore, once the
number of models to be selected and the criteria are determined, the number of models
needed for sorting is also determined. Obviously, once a suitable number of models are
generated and ranked, increasing the number does not contribute significantly to error
reduction. For this study, 300~500 is a suitable number.

The computational work in terms of CPU time for the two cases is listed in Table
4. The jobs were run on a 600MHz DEC machine. The difference between the two cases
lays in the history matching part. The history matching procedure requires, in each
iteration, a flow simulation to check the error and the streamline inverse process to
modify the permeability (Wang and Kovscek, 2000). Six to eight iterations are usually
needed to obtain a satisfactory match to the production history. Each iteration takes about
8 seconds for unit mobility ratio case, and 39 seconds for non-unit mobility ratio for a 50
by 50 grid. Sorting one model involves solving the pressure field, computing a sufficient
number of streamlines and their TOF distributions. However, this procedure is
exceptionally fast. It takes about one second for each model for a 50 by 50 case.
Generating a model using SGSIM takes negligible CPU time. As shown in Table 4,
sorting 500 models (50 by 50 grid-blocks) takes about ten minutes all together. On
average, each model takes about 1.1 seconds for unit mobility ratio case and 1.4 seconds
for non-unit mobility ratio case. However, if CFS is run on each model with the same
number of grid-blocks (50 by 50) to compute the error, it takes about 6 seconds for unit
mobility ratio case and up to 35 seconds for the non-unit mobility ratio case, to ensure
accurate results. Therefore, the proposed ranking method is, on average, a magnitude
faster than if the CFS is used for ranking. The ranking method would appear even more
efficient compared to CFS if a finite-difference based simulator was used for forward
simulation. While the CPU time consumed by sorting increases with the number of
models, the cost for history-matching is fixed and independent of the number of models
to be sorted.

240
It is worth-noting that the computational work is an approximately linear function
of the number of grid-blocks, n. More specifically, the computation time of the ranking
procedure and streamline inverse part of the history-matching process is strictly linear
with respect to n , while the flow simulation part of history-matching is approximately
proportional to n lg n . For instance, the CPU time for ranking five hundred models with
10,000 grid-blocks takes 40 minutes for unit mobility ratio case and 55 minutes for
M=2.5case. If a full flow simulation is used to compute the errors of a model and to rank
the model, the computational effort is at least proportional to n lg n . Disregarding the
computation for history-matching, the ranking effort is equivalent to one or two time-
steps of the full flow simulation.

Currently, the algorithm does include a number of assumptions or limitations: no


gravity effects, two-phase flow, incompressible or slightly compressible fluid or rock,
uniform initial condition, and some accurate production history showing injected phase
breakthrough available. Further study is underway to relax these assumptions. For
instance, it appears to be well suited for three-dimensional cases with slight gravity.

7. CONCLUSIONS

This study revealed the linear correlation between each of the following pairs of
streamline computed and reservoir flow properties: (1) SL-TOF cdf and producer water
cut, (2) unit mobility flow-rate or pressure at one time and that in comprehensive flow
simulation, and (3) equivalent-CIPP and CIPP. These correlations are proven
approximately via analytical expressions and confirmed numerically. Based on this
observation, a streamline based ranking algorithm is proposed. Functionally, it is a
conversion of matching in the temporal coordinate to matching in the spatial coordinate.
For instance, matching of the producer water cut versus time is converted to matching the
desired SL-TOF distribution. This conversion substantially speeds up the matching
process. For unit mobility ratio, such as in a tracer flow study, it is shown that the
fractional-flow curve is the same as the SL-TOF cdf. In this case, the history-matching
process is not necessary. That is, the reference fractional-flow curve can be directly used
as reference SL-TOF cdf. In general, the computational effort is equivalent to only two
time-steps of a comprehensive flow simulation of the same model. More importantly, the
computational work for calculating streamlines scales linearly with the number of grid-
blocks.

It is observed that the distribution of the difference in results between a realization


and the reference reservoir model is approximately a normal distribution. Therefore, the
number of models that must be sampled to obtain a given number of acceptable models
(i.e., within a specified range of the reference) can be estimated accurately. For instance,
it is known that about two percent of the models to be ranked have errors two standard
deviations lower than the average error.

241
8. NOMENCLATURE

k permeability, L2
krw, kro relative permeability of water and oil
M end point mobility ratio
n number of streamlines
q flow-rate, L3/t
so, sw oil and water saturation
v, V pore volume of streamtube(s), L3
xfD dimensionless distance the displacing front has traveled along a streamline
τD dimensionless time of flight

Subscripts:
D dimensionless
f displacing front
1,2 streamline index

Superscripts:
R reference
• non-unit mobility ratio case

9. REFERENCES

1. Agarwal, B. and Blunt, M. J.: “Full-physics, Streamline-Based Method Method for


History Matching Performance Data of a North Sea Field.” SPE 66388, paper
presented at the SPE Reservoir Simulation Symposium, Houston (February 11-14,
2001.

2. Baker, J. W., Cuypers, M., and Holden, L. : “Quantifying Uncertainty in Production


Forecasts: Another Look at the PUNQ-S3.” SPE 62925, paper presented at the
Annual SPE Technical Conference and Exhibition, Dallas (October 1-4, 2000).

3. Ballin, P. R., Journel, A. G., and Aziz, K.: “Prediction of Uncertainty in Reservoir
Performance Forecast”, Journal of Canadian Petroleum Technology , 31 No. 4
(April 1992) pp. 52-62.

4. Batycky, R. P., Blunt, M. J., and Thiele, M. R.: “A 3D Field-Scale Streamline-


Based Reservoir Simulator,” SPE Reservoir Engineering,(November 1997) pp.
246-254.

5. Blunt, M. J., Liu, K., and Thiele, M. R.: “A Generalized Streamline Method to
Predict reservoir flow.” Petroleum Geoscience, 2 (1996) pp. 259-269.

242
7. Caers, J., Krishnan, S., Wang, Y., and Kovscek, A. R.: “A Geostatistical Approach
to Streamline-based History Matching” submitted to SPE Journal(2001).

8. Cruz, P. S., Horne, R. N., and Deutsch, C. V.: “The Quality Map: A Tool for
Reservoir Uncertainty Quantification and Decision Making”, SPE 56578, paper
presented at the Annual SPE Technical Conference and Exhibition, Houston
(October 3-6, 1999).

9. Deutsch, C. V. and Journel, A. G.: GSLIB, Geostatistical Software Library and


User’s Guide, Second Edition, Oxford University Press, New York (1998).

10. Deutsch, C. V. and Srinivasan, S.: “Improved Reservoir management Through


Ranking Stochastic Reservoir Models,” SPE 35411, paper presented at SPE/DOE
Tenth Symposium on Improved Oil Recovery, Tulsa (April 21-24, 1996).

11. Dykstra, H. and Parsons, R. L.: “The Prediction of Oil Recovery by Waterflood,”
Secondary Recovery of Oil in the United States, Principles and Practice, 2d ed.,
American Petroleum Institute (1950) pp. 160-174.

12. Gupta, A. D., Vasco, D. W. and Long, J. C. S.: “Detailed Characterization of


Fractured Limestone Formation Using Stochastic Inverse Approaches,” SPE Ninth
Symposium (1994).

13. He, Z., Datta-Gupta, A. and Yoon, S.: “Streamline-based Production Data
Integration Under Changing Field Conditions”, SPE 71333, paper presented at the
Annual SPE Technical Conference and Exhibition in New Orleans, Louisiana
(September 30 - October 3, 2001).

14. Lake. L. W. : Enhanced Oil Recovery, Prentice Hall, Englewood Cliffs, NJ (1989).

15. Pollock, D. W. 1998. “Semianalytical Computation of Path Lines for Finite-


Difference Models”, Ground Water, 26, 6, (1998) pp. 743-750.

16. Vasco, D. W., Yoon, S. and Datta-Gupta, A.: “Integrating Dynamic Data Into High-
Resolution Reservoir Models Using Streamline-Based Analytic Sensitivity
Coefficients,” SPE 49002, paper presented at the Annual SPE Technical Conference
and Exhibition in New Orleans, Louisiana (September 27-30, 1998).

17. Wang, Y. and Kovscek, A. R. : "Streamline Approach for History Matching


Production Data," SPE Journal(October 2000) pp. 353-362.

18. Wen, X. H., Clayton, D. V. and Cullick A. S: "Integrating Pressure and Fractional
Flow Data in Reservoir Modeling with Fast Streamline-Based Inverse Method,"
SPE 48971, paper presented at the Annual SPE Technical Conference and
Exhibition in New Orleans, Louisiana, September 27-30, 1998a.

243
19. Wen, X. H., Clayton, D. V. and Cullick A. S.: “High Resolution Reservoir Models
Integrating Multiple-Well Production Data,” SPE Journal, (December 1998) pp.
344-355.

20. Yoon, S., Malallah, A. H., Datta-Gupta, A., and Vasco, D. W.: “A Multiscale
Approach to Production-Data Integration Using Streamline Models”, SPE Journal
(June, 182-192, 2000).

244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
Pore Microstructure and Fluid Distribution
in a Diatomaceous Reservoir

C.M. Ross and A.R. Kovscek

ABSTRACT

This study was undertaken to improve our understanding of fluid distribution and
flow behavior by quantifying pore type, pore body size, shape attributes, and throat sizes
of pore networks in diatomaceous reservoir rock. This work is a first step toward
providing a means of integrating different measurement types and establishing an
interpretive tool for understanding laboratory results and field behavior. This project lays
the groundwork for developing a predictive tool that can be utilized to generate input for
reservoir simulator studies.

The pore microstructure of 11 diatomite samples from Lost Hills Field (Kern Co.,
CA) and one quarry sample was quantified using image analysis of photomicrographs
collected with a Scanning Electron Microscope (SEM) in backscattered electron (BSE)
mode. Six distinct pore types were identified ranging from small, irregularly shaped pores
to large moldic and intraskeletal pores as well as large interconnected, intergranular
pores. The throat size average and size range for each pore type was established by
combining image analysis information with high-pressure mercury injection
measurements. Once this relationship is determined, the pore body size and throat size
serve as the basis for a predictive model allowing capillary pressure curves and fluid
saturations to be calculated from pore type data.

In this study, the relationship between pore types, fluid flow, and fluid
distribution indicate a dual porosity system. Permeability is primarily controlled by three
pore types, which represent 1 to 28% of the total core porosity and 0.67 to 9.75% of the
total volume. Fluid saturation models were developed to provide an accurate means of
calculating oil and water saturations as well as their distribution within the pore network
for both the reservoir and core samples. From the model, we found that the most
permeable pore types have oil saturations that are 5 to 35% lower than reservoir
saturations. This indicates that the oil is moveable and can be swept from permeable pore
types. Alternatively, there was no net change in saturations for the remaining pore types
between core- and log-derived saturations. For these pore types, spontaneous imbibition
is the most likely production mechanism. The smallest pore type is predominately water
filled with oil saturation values of about 32% in both the reservoir and core samples.

The proportion of each pore type can be predicted from wireline log data and used
as input for the predictive capillary pressure model and fluid saturation models. This
allows the models to be applied to the rest of the field providing data for field
development decisions and simulator input.

261
1. INTRODUCTION

reservoirs rely on either


Virtually all recovery processes for diatomite and other low permeability
spontaneous or forced imbibition. Wells are usually hydraulically fractured to overcome
low matrix permeabilities, thereby improving injectivity as well as productivity. These
fractures increase the contact area between boreholes and the reservoir. Injection pressure
for diatomite must be limited, because the fields are relatively shallow, the permeabilities
are very low, and diatomites often have low cohesive strength. Thus for any process
involving water injection, the role of matrix to fracture transfer via imbibition is
important in determining oil recovery.

The rate and extent of water flow during spontaneous and forced imbibition depends
primarily on the capillary pressure versus water saturation curve, the initial water
saturation, relative permeability curves, and the viscosity of the nonwetting and wetting
fluid phases. In turn, the above rock properties depend on factors such as pore structure,
pore connectivity, pore throat to body size aspect ratio, wettability, and interfacial
tension. Pore type, pore roughness, and connectivity are especially important to
imbibition, because water simultaneously moves in the corners or crevices of pores,
advances in a piston-like fashion, and fills pore bodies cooperatively (Lenormand and
Zarcone, 1984; Lenormand, et al., 1988).

Prior work on diatomites has focused on establishing the macroscopic effects of these
more microscopic factors. The main tools have been experiments to monitor the extent
of imbibition in tight rocks samples with X-ray CT (Akin et al., 2000; Zhou et al., 2002.
In strongly water-wet samples of diatomite, capillary forces are significant and partially
offset low formation permeability. Similarly, experimental imbibition results in a
combination with a scaling analysis teach that the Leverett J-function for diatomite is
greater than that for sandstones (Schembre et al., 1998). Limited qualitative, microscopic
characterization of diatomite using mercury porosimetry and SEM images indicates that
pore structure and pore roughness are varied (Akin et al., 2000). Many pathways for
water are available and these might either aid or hinder the imbibition process.

Thus, a first order solution to improving oil recovery from diatomite is to develop a
quantitative understanding of the relationship between pore structure, permeability,
porosity, wettability, and macroscopic oil transport. An equally important understanding
remains to be developed of how pore structure and other properties evolve with stress and
silica diagenesis. Any of these tasks are monumental and beyond the scope of a single
study. Nevertheless, we have begun characterizing quantitatively the pore structure of
diatomites and their variability from two-dimensional (2D) images collected from epoxy-
impregnated polished samples. The resulting pore types are used to model high-pressure
mercury injection data and fluid saturations. These models further our understanding of
fluid flow behavior at the pore level in diatomaceous reservoirs and are scalable to the
field level in that these models can be applied via wireline log analyses.

262
2. BACKGROUND, SAMPLING, AND AVAILABLE DATA

Twelve opal-A diatomite samples were analyzed in this study; eleven samples are
reservoir diatomites and one sample is from a coastal quarry. The quarry sample is from
Grefco quarry (Santa Barbara Co., CA) and consists of relatively pure biogenic material
deposited during the late Miocene (Mohnian) in the Santa Maria Basin (Graham and
Williams, 1985). The quarry sample serves as a relatively pure diatomite end member
and helps relate current and previous laboratory studies, conducted using readily
available quarry samples, to reservoir diatomites (Akin et al., 2000; Schembre et al.,
1998).

The remaining 11 samples are from well OB-7 in Lost Hills Field. These samples
are from the uppermost unit of the Monterey Formation, the Belridge Diatomite Member,
consisting of biogenic silica (opal-A), clay, and detrital silt as well as sand. The late
Miocene (Mohnian) Belridge Diatomite is a thick section of diatomites and diatomaceous
shales that occur in the central portion of the San Joaquin Basin and may be included as a
facies within the Reef Ridge Member of the Monterey Formation (Graham and Williams,
1985). Although Lost Hills Field was originally discovered in 1910 and production
began in 1913, aggressive development of the field did not occur until the mid-1980’s
following advancements in hydraulic fracturing (Wallace and Pugh, 1993). Waterflood
development began in 1991 to further improve recovery and slow subsidence rates
(Wallace and Pugh, 1993). The original oil in place (OOIP) is estimated at 2.6 billion
barrels (Perri et al., 2000). To date, only 5% of the OOIP or approximately 135 million
barrels of oil have been produced (Perri et al., 2000).

Due to the variability in diatomite composition within Lost Hills Field, diatomites
are classified as sandy diatomites, clayey diatomites, and clean diatomites based upon
their composition and properties Fossum and Fredrich, 2000). Four sandy diatomites,
four clayey diatomites, and three clean diatomites were provided for study (Table 1). The
quarry sample is classified as “quarry diatomite” and is considered an extreme member of
the clean diatomite class from Lost Hills Field, as shown by the continuum between
reservoir and quarry data in the porosity and permeability plot (Fig. 1).

Conventional core data, high-pressure (60,000 psi) mercury injection


measurements, wireline log data, core plugs, additional core material, and SEM mounts
were provided for ten reservoir samples. Much of this data and sample material was also
available for the quarry sample, except for wireline log data and fluid saturation data. For
sample 1450.00, only core data and an SEM mount were available.

3. PORE MICROSTRUCTURE

Characterization of pore microstructure and its variability improves our


understanding of oil, water, and gas transport within diatomite reservoirs. Image analysis
allows us to identify pore types and quantify their size range. By modeling capillary
pressure behavior and fluid saturation, we are able to further our understanding of

263
reservoir dynamics on the pore level with the potential to apply this knowledge to the
entire field via wireline log data.

3.1. Image Analysis

For each sample, a small amount of core material was cleaned and impregnated
with epoxy, first under vacuum and then under pressure (up to 1500 psi). We estimate
that pores as small as 0.04 m or less should be filled at this pressure. After the epoxy
hardened, the samples were finely polished to provide an optically flat surface for
subsequent BSE-SEM imaging. Photomicrographs were collected using a JEOL JSM-
500LV SEM with a BSE detector and an EDAX X-ray spectroscopy system.
Uncalibrated BSE images of polished surfaces are relative electron density maps where
epoxy-filled pores have the lowest average electron density and are dark in the grayscale
images. Silica, carbonate, and pyrite have increasingly higher electron densities and are
correspondingly brighter within an individual image (Fig. 2).

70 to 100 images from each sample were collected at a magnification of 500X. At


500X, large pores in the sandy diatomite samples could be imaged while theoretically
detecting pores as small as 0.1µm. Each image covers a 237.6 by 177.6µm area and is
represented by a matrix of 2048 by 1600 pixels. In most samples, the rare occurrence of
high porosity views (often associated with large pores; e.g., Fig. 2) increased the number
of images needed to generate stable SEM porosity averages. (For comparison, typically
15 to 30 images are collected at 40 to 200X for sandstone and carbonate reservoirs Ehrich
et al., 1991) In the limited area visible at 500X, it is difficult to distinguish between
naturally occurring fractures and those induced during coring and sample preparation;
therefore, fractures were not imaged.

Each grayscale image was segmented into a binary image (Fig. 2) consisting of
black pore pixels and white non-pore pixels, based upon visual inspection of the image as
well the grayscale histogram. Incomplete epoxy impregnation of larger pores and surface
roughness can interfere with the thresholding process causing pixels to be misclassified,
which is corrected by “painting” the appropriate pixels either white or black. For the
study samples, the amount of porosity resolved at 500X ranges from 5.77 to 18.12%
(Table 1). In this study, SEM porosity represents only 9.3 to 28.9% of the measured core
porosity. The remaining porosity is either too small to be resolved at the study
magnification or matrix pores were not filled with epoxy. For each portion of this study,
we found that the unresolved porosity does contribute to permeability, capillary pressure
modeling, and fluid saturation modeling. Therefore, permeability, capillary pressure, and
fluid saturation analyses were performed with the detected pore types alone.

Binary pore images were processed using erosion-dilation (E-D) procedures that measure
the size, shape, and connectivity of pores as viewed in two dimensions (2D) (Ehrlich et
al., 1991). In particular, E-D subdivides the porosity into smooth and roughness
components and records the size distribution and relative abundance of each component.
The smooth component is equivalent to either a circular or ellipsoid pore body, whereas
the roughness components are portions of the pore that deviate outside of the smooth

264
component (Ehrlich et al., 1991). In images with interconnected 2D pores, the largest
pore body will be attributed to smooth porosity and the remaining porosity (as well as
smaller pore bodies) will be attributed to roughness components based upon their size10.
A pattern recognition algorithm then determines smooth and roughness components that
occur together in constant proportion (Ehrlich et al., 1991). Attributes that occur in
constant proportion are considered different 2D slices of a three-dimensional (3D) pore
type (Ehrlich et al., 1991; Graton and Fraser, 1935). The algorithm also determines the
number of 3D pore types and their relative abundance within each study sample.

3.2. Pore Types

The pore classification procedure identified six distinct pore types in the study
samples (Fig. 3), determined the smooth and roughness size and shape characteristics of
each pore type, and the proportion of SEM porosity contributed by each pore type per
sample (Table 2). The pore body width range is the size range of 2D cross sections
through the pore body (smooth porosity; Table 3) (Ehrlich et al., 1991a; Ehrlich et al.,
1991b). The average minimum pore body diameter, as determined from E-D smooth and
roughness measures, serves as a 3D pore size measure and is used to rank the pore types
by size (Table 3). Pore Type 1 (PT1) is the smallest detected pore type whereas the
largest pore type is Pore Type 6 (PT6).

Pore type 6 consists of relatively large, well-connected intergranular pores that


occur in sandy diatomite samples, 1450.00, 1516.80, and 1526.85 (Tables 1-3). PT6
occurs in interstitial spaces between very fine-grained sand to coarse-grained silt grains in
regions that lack pore-occluding matrix material (Fig. 3). Due to the irregularity of the
pore body shapes and the high degree of connectivity as observed in 2D, PT6 primarily
consists of moderate- to large-scale roughness features (82%; ≥ 0.457µm). The larger
scale roughness features are associated with somewhat smaller, interconnected pore
bodies rather shape irregularities within a single pore.

In comparison, the other large pore type, PT5, consists mainly of smooth porosity
(71%; 0.456-120µm) with 3% moderate roughness components ranging from 0.229 to
0.342µm and 26% large scale roughness components (≥ 3.649µm), that are deviations
from either a spherical or ellipsoid pore shape. PT5 is primarily composed of moldic
pores (Fig. 3) as well as relatively large intraskeletal pores (diatom and foraminifer; Fig.
2) and occasionally equant intragranular pores. In Fig. 3, the example image for PT5 is
the largest pore observed in this study and has a minimum pore width of about 120µm.
Typically, PT5 consists of smaller pores ranging in diameter from 10 to 25µm. PT5 is
most abundant in the remaining sandy diatomite sample, 1484.95 (Tables 1-2). Although
classified as a sandy diatomite via wireline log analysis, sample 1484.95 does not contain
much sandy and silty detrital material and occurs as an outlier in the porosity-
permeability plot (Fig. 1; Table 1). The bulk density of this sample was most likely
increased by gypsum (observed bridging fractures and replacing matrix material), thereby
causing this sample to be misclassified via log analysis. Alternatively, the
misclassification could be attributed to the thin interbedded nature of diatomites.

265
The remaining pore types are comparatively small with an average minimum pore
body diameter ranging from 0.82 to 1.74µm (Table 3). Pore type 4 consists of small
intergranular pores that occur between irregularly shaped silt grains (Fig. 3) and are
primarily found in sandy diatomite samples, except for one clayey diatomite sample,
1707.00 (Table 1-2). These irregularly shaped pores have a small proportion of smooth
porosity (31%; 1.482-4.332µm) with a broad size range of roughness features (0.115-
1.254µm) that comprise 69% of PT4.

The remaining three pore types are, for the most part, abundant in clean and
clayey diatomite samples. Pore type 3 is composed of small- to moderate-sized
intraskeletal pores, primarily occurring in test fragments as well as some whole tests (Fig.
3). The shape characteristics for PT3 are bimodal with smooth porosity (24%) sizes of
0.114 to 0.456µm and 1.142 to > 4.334µm (Table 3). The smaller sized pores occur in
broken test fragments, tests that have their surface exposed, and cross sections of the test
walls (Fig. 3). The larger pores occur within the tests. These two pore sizes occur
together in constant proportion and are, therefore, classified as a single pore type.
Samples 1553.65, 1880.00, and Quarry1 contain the highest percentage of PT3 (42-44%
SEM porosity; Table 2). The roughness component of PT3 ranges in size from 0.114-
3.648µm.

Pore type 2 consists of small equant pores (1.35µm; Table 3) that occur in regions
that are relatively devoid of matrix porosity (Fig 3). Of the pore types in this study, PT2
is unique due to their relative isolation from neighboring pores. These low porosity areas
are associated with somewhat linear fabric disruptions in the matrix and may be
compaction features. PT2 is most abundant in the clean diatomites with an average of
46% of the SEM porosity (Tables 1-2). In the other diatomite types, PT2 comprises an
average 32% of the SEM porosity for clayey diatomites, 24% of the quarry sample, and
an average of 14% for sandy diatomites (Tables 1-2).

The last pore type, PT1, is the smallest pore type that was detected at 500X magnification
(0.82µm; Table 3) and consists of 44% smooth porosity with small roughness elements
(0.114-0.912µm). In sandstone and carbonate studies, the smallest pore type typically
consists of pores that are 1 to 2 pixels wide, do not survive past the first E-D cycle, and
are unable to distinguish shape characteristics (Ehrlich et al.; 1991; Ross et al., 1995). In
this study, PT1 has a multiple E-D cycle size range (0.114-2.508µm; Table 3) due to the
higher magnification and image resolution (i.e., pixel density). PT1 consists of very small
intraskeletal and intramatrix pores as well as very small intraparticle pores. PT1 occurs in
each diatomite type (Tables 1-2).

3.3. Pore Types, Porosity, and Permeability

The need to characterize pore networks in diatomaceous reservoirs is due in part


to the complex relationship between porosity and permeability (r = -0.14; Fig 1) as well
as capillary pressure and aqueous-phase saturation (Akin et al., 2000). For any given
porosity value, permeabilities have a range of two to three decades. Comparison of SEM
porosity with permeability does not provide an improvement on this relationship

266
(r = -0.16; Table 1). This indicates that the type of porosity (i.e., pore type distribution)
is more important in determining permeability than either the total amount of porosity or
the amount of SEM porosity.

The six pore types were grouped according to their relationship to permeability.
Pore types that do not correlate positively to permeability and, therefore, do not
contribute significantly to permeability are PT1, PT2, and PT3. Each of these pore types
have either low or negative correlations with permeability (PT1, r = 0.17; PT2, r = -0.78;
PT3, r = -0.52). Although PT2 has a particularly strong negative relationship with
permeability (Fig. 4), it does not indicate that pores classified as PT2 decrease
permeability. Rather, it is the association of PT2 with low porosity regions (Fig. 3),
which serve as permeability barriers, that causes this negative relationship with
permeability. In fact, the presence of these relatively non-porous permeability barriers as
indicated by PT2 provides good predictor of permeability [log(K) = 1.626 – 3.890(PT2);
R2 = 0.74]. PT1 has a complex relationship with permeability depending upon the
diatomite type. In clean and clayey diatomites, PT1 does not contribute to permeability;
however, PT1 has a strong positive relationship with permeability in sandy and quarry
diatomite samples (Fig. 4).

The remaining pore types have positive relationships with permeability (PT4, r =
0.35; PT5, r = 0.51; PT6, r = 0.47). Individually, these pore types do not have a strong
predictive relationship with permeability. By summing the relative abundance of these
pore types, the correlation between permeability and pore type information increases (r =
0.66; Fig. 5). If PT1’s contribution to permeability is added for sandy and quarry
diatomite samples, this relationship improves significantly (r = 0.84; Fig. 5) and provides
a good predictive relationship [log(K) = 0.244 + 1.893(permeable PT); R2 = 0.77]. In
these samples, porosity volumes of 0.67 to 9.75%, that represent only 1.2 to 28% of the
total core porosity, control permeability.

4. CAPILLARY PRESSURE MODEL

The size range of pore throats (i.e., constrictions connecting pore bodies) associated with each
pore type is determined by modeling mercury injection measurements using the pore type
distributions for each sample (McCreesh et al., 1991). Capillary pressure data reveals the
volume of porosity accessed through throats of a given size. By using the pore type
distribution to predict saturation changes within the capillary pressure curves, the pore
body and throat size relationship can be established. In addition, the resulting series of
equations can be used to construct synthetic or predicted capillary pressure curves.

Six inflection points were selected based upon the location of significant slope
changes in the capillary pressure curves provided for 11 study samples. For each sample,
the change in saturation between inflection points was regressed against the proportion of
pore types (R2 = 0.73-0.95). The resulting series of equations serve as the capillary
pressure model (Table 4). In the model, the coefficients summed for each pore type
(column) should approach unity (1) if all of the porosity attributed to the pore type is

267
being filled. On average, 96% of the total porosity is filled according to the model (Table
4).

The range of throat sizes associated with each pore type is determined using the
model (Tables 4-5). For example, the throat size range for PT1 is 0.0018 to 0.236µm. The
average throat size is a weighted average based upon each pore type’s relative saturation
and throat size range for each pressure interval. In sandstone and carbonate studies,
average pore body widths are typically 2 to 15 times larger than their average throat
diameter and may be as much as 600 times larger (Ross et al., 1995; McCreesh et al.,
1991; James, 1995; Bowers et al., 1994). In diatomites, the size ratio between average
pore body and throat diameter range is often much smaller (e.g., 1.6 for PT3, 1.4 for PT4,
and 0.8 for PT5; Table 5).
By definition, pore throats cannot be larger than pore bodies as shown for PT5. In
light of this, the applicability of throat and/or pore size averages developed on
conventional sandstone and carbonate reservoirs should be reassessed. In perfectly
aligned sphere packs, multiple throat sizes and multiple pore body sizes per unit pore
(pore type) occur depending upon the packing pattern (Ciftcioglu et al., 1988). Further
work is needed to establish the optimal averaging method to characterize pore body and
throat sizes in diatomites. If the data warrants, multiple pore body and throat sizes should
be attributed to each pore type for both interpretation and network modeling applications.

Synthetic capillary pressure curves were constructed by converting the saturations


predicted by the equations into cumulative form (McCreesh et al., 1991). The resulting
values are plotted with the measured capillary pressure curve to determine the accuracy
of the model (Fig. 6). All of the synthetic capillary pressure curves show good agreement
with the measured curves (r = 0.98-1.0). Sample 1642.25 shows the greatest discrepancy
between the measured and predicted capillary pressure curves (Fig. 6). This sample
contained many small fractures that were not imaged as discussed previously. If the
fractures were induced by either coring or sample preparation, the predicted capillary
pressure curve better indicates reservoir behavior. If the fractures are naturally occurring,
the predicted curve indicates the capillary pressure behavior of the matrix.

For sample 1450.00, capillary pressure measurements were not available. The
capillary pressure model allows a synthetic curve to be constructed using pore type
information alone (Fig. 6). Once calibrated, the model allows the limited number of
capillary pressure measurements to be extended to any sample with enough material for
image analysis sample preparation. Capillary pressure modeling provides an alternative
when sample materials fail during measurement and when measurements are particularly
time consuming and costly (Ross et al., 1995). In addition, this technique is particularly
useful when pore type proportions are predicted using wireline log response, which
allows the model to be applied to the entire reservoir interval throughout the field without
performing additional measurements and without the expense of collecting cores (Ross et
al., 1995; Callender et al., 1998). In this study, additional samples and measurements are
needed to develop a field-wide model for Lost Hills. It may also be possible to develop a
diatomite model for California reservoirs; this would require substantially more samples
and data from dozens of fields.

268
Oil saturations determined from wireline log data should be comparatively
unaffected by drilling operations, but log-derived saturations have a much lower
resolution compared to plug data. Despite the decreased resolution and reduced impact of
drilling, the same relationship between pore types and oil saturation was found using log-
derived saturations (Fig. 8). This relationship indicates that the oil saturation decreases as
the proportion of permeable pore types increases. The study well, OB-7, is only 20 ft
from a water injection well, 10W-9, that has been in operation since 1994 (Wilt et al.,
1995); the permeable pore types were likely swept by injection water.

Crosswell electromagnetic survey results indicate that the proximity to the


injector well did not affect the oil saturations above the injection interval (~1510 ft) (Wilt
et al., 1995) where two of the more permeable samples originate (1450.00 and 1484.95).
By comparing data from 10W-9 (drilled in 1994) and OB-7 (drilled in 1999), a 5%
reduction in oil saturation was observed between 1540 and 1750 ft, whereas a 4 to 6%
increase in oil saturation was observed in diatomite intervals below 1790 ft (Wilt et al.,
1995). Wilt and others (Wilt et al., 1995) were especially interested in intervals
containing high percentages of opal-A (i.e, not sandy diatomites) because of their higher
oil saturations as indicated by their higher resistivities. Their observations correspond to
the distribution of fluids as determined in this study.

In order to better understand the distribution of fluids within individual pore


types, saturation models were developed for both core- and log-derived fluid saturations.
Each model was constructed using 11 equations with known saturations for each sample
and six unknowns as the saturation coefficients for each pore type. The resulting
coefficients are presented in Table 6 for fluid saturations in core samples (Swc and Soc)
and in the reservoir (Swl and Sol). Using pore type proportion data from either image data
or predicted from wireline log data, fluid saturations can be calculated. For the study
samples, core fluid saturations are more accurately predicted (r = 0.92-0.94) than log-
derived fluid saturations (r = 0.85-0.87). Figures 9-10 provide an alternative way of
displaying the resulting saturation coefficients.

The saturation coefficients may be compared between models. For example, the oil
saturations of PT4, PT5, and PT6 are higher in the reservoir (34-73%) than observed in
core samples (13-59%; Figs. 9-10). The lower oil saturations of core- versus log-derived
values may be due in part to flushing of the core by drilling, expulsion (drainage) of the
oil during handling, shrinkage of the oil due to out gassing, and expansion of the core
once confining stress was removed (Bilodeau, 1995). There was not a net change in
saturation for the remaining pore types, PT1, PT2, and PT3.

For demonstration purposes, the core saturation model will be treated as if it were
a waterflood model based on laboratory measurements. In this case, we would conclude
that 5% of the oil was flushed from PT6, 34% from PT5, and 14% from PT4 during the
tests. The actual volume of oil produced during the waterflood could be calculated based
upon the SEM porosity and pore type distribution. In this manner, researchers could
better compare laboratory results with respect to different production mechanisms (e.g.,
waterflood, steamflood, CO2 flood, etc.). This knowledge may be applied to the rest of
the reservoir via simulations by predicting pore types from wireline logs and applying the

269
physical models that were developed based upon core data. Intervals could be targeted
with the optimal production mechanism based upon their pore network configuration.

5. CONCLUSIONS

1. Image analysis techniques originally developed for sandstone and carbonate


reservoir characterization are applicable to diatomaceous reservoir description.
Although their mean pore body size is small, pore types in diatomites carry
significant information about the pore network and their influence on physical
properties as well as fluid flow and distribution.

2. Permeability is more accurately predicted using pore type criteria than core
porosity values. Permeability prediction using core porosity has a range of two
to three decades whereas the pore type-based relationships with permeability
has only a one-decade range.

3. The amount and distribution of fluids within the pore network can be
determined using pore type characterization. In this case, we found that the
more permeable pore types contain oil saturations of 34 to 73% in the reservoir
and 13 to 59% in core plugs. These comparisons can be used in conjunction
with laboratory measurements to future our understanding of multiphase fluid
behavior.

4. Pore type 2 exhibits a strong relationship with many reservoir quality measures
such as permeability (r = -0.78) and oil saturation (r = 0.83).

5. Pore typing studies provide input parameters for equations modeling physical
properties such as permeability, capillary pressure, imbibition potential,
formation factor, and relative permeability. Some of these parameters include
throat size and distribution, pore body size and distribution, number of pores
per unit area, and surface area. Predicting pore types from wireline log data
extends the applicability of these parameters beyond the original sample set.

6. Some aspects of diatomite topology and morphology differ from those of


conventional sandstone and carbonate reservoirs. Due to the small pore
dimensions and extreme heterogeneity within diatomite samples, three to five
times more images are needed to produce stable SEM porosity values,
compared to conventional reservoir types.

7. Additional samples are needed to better characterize the pore network


variability within the Lost Hills diatomite reservoir and to serve as a training set
in order to predict pore type distribution using wireline log data.

270
6. APPLICATIONS AND FUTURE WORK

Pore typing may be the key to understanding diatomite reservoir behavior and
integrating disparate data collected across a variety of different length scales. For
example, samples can be characterized with respect to their pore type, size, and
connectivity information as well as macroscopic flow properties such as permeability and
relative permeability (Zhou et al., 2002; Etris, 1991). Likewise, samples can be correlated
with their log response. Hence, pore type can be inferred from log data. Then, flow
behavior from reservoir regions that have only been logged can be identified from pore
characteristics.

In addition to further refinement of pore types presented here, our work continues
in the implementation of pore network for opal-A diatomites. Diatomite network
simulators can be used to probe flow physics and oil production as well as estimate
relative permeability. Studies are underway on opal-A samples from Midway-Sunset
Field. We have begun to compare the pore structure of opal-CT samples with opal-A
diatomites. Another potential research direction is to study the effects of stress and
compaction upon pore networks of diatomites.

7. NOMENCLATURE

2D = two dimensions, area


3D = three dimensions, volume
∅ = porosity, % volume
BSE = back-scattered electron
BVO = bulk volume oil, % volume
BVW = bulk volume water, % volume
K = permeability, md
OOIP = original oil in place, barrels
PT = pore type, proportion
r = correlation coefficient
R2 = regression coefficient
SEM = scanning electron microscope
Soc = oil saturation, core (% pore volume)
Sol = oil saturation, log (% pore volume)
Swc = water saturation, core (% pore volume)
Swl = water saturation, log (% pore volume)
X = times, magnification

271
8. REFERENCES

1. Akin, S., J.M. Schembre, S.K. Bhat, and A.R. Kovscek: “Spontaneous Imbibition
Characteristics of Diatomite,” J. of Petroleum Science and Engineering 25 (2000)
pp. 149-165.

2. Bilodeau, B.J.: “Determining Water Saturation in Diatomite Using Wireline Logs,


Lost Hills Field, California,” SPE 29653, paper presented at the SPE Western
Regional Meeting, Bakersfield (March 8-10, 1995).

3. Bowers, M.C., R. Ehrlich, and R.A. Clark: “Determination of Petrographic Factors


Controlling Permeability Using Petrography Image Analysis and Core Data, Satun
Field, Pattani Basin, Gulf of Thailand,” Marine and Petroleum Geology 11/2 (1994)
pp. 148-156.

4. Callender, C.A., J.J. Funk, C.M. Ross, and J.B. Turbeville: “Generating Pore Types
and Synthetic Capillary Pressure Curves from Wireline Logs Using Neural
Networks,” U.S. Patent Number 5,828,981 (October 27, 1998).

5. Ciftcioglu, M., D.M. Smith, and S.B. Ross: “Mercury Porosimetry of Ordered
Sphere Compacts: Investigation of Intrusion and Extrusion Pore Size
Distributions,” Powder Technology 55 (1988) pp. 193-205.

6. Ehrlich, R., S.J. Crabtree, K.O. Horkowitz, and J.P. Horkowitz: “Petrography and
Reservoir Physics I: Objective Classification of Reservoir Porosity,” AAPG
Bulletin (Oct. 1991a) pp. 1547-1562.

7. Ehrlich, R., E.L. Etris, D. Brumfield, L.P. Yuan, and S.J. Crabtree: “Petrography
and Reservoir Physics III: Physical Models for Permeability and Formation Factor,”
AAPG Bulletin (Oct. 1991b) pp. 1579-1592.

8. Etris, E.L.: “Investigations Into the Control of Pore/Throat Structure on Capillary


Pressure Imbibition, and Its Relevance to Resistivity Index,” Ph.D. Dissertation,
University of South Carolina (1991) 338 p.

9. Fossum, A.F., and J.T. Fredrich: “Constitutive Models for the Etchegoin Sands,
Belridge Diatomite, and Overburden Formations at the Lost Hills Oil Field,
California,” Sandia Report SAND2000-0827, Sandia National Laboratories,
Albuquerque (2000).

10. Graham, S.A., and L.A. Williams: “Tectonic, Depositional, and Diagenetic History
of Monterey Formation (Miocene), Central San Joaquin Basin, California,” AAPG
Bulletin (March 1985) pp. 385-411.

11. Graton, L.C., and H.J. Fraser: “Systematic Packing of Spheres – With Particular
Relation to Porosity and Permeability,” J. of Geology
(Nov. 1935) pp. 785-909.

272
12. James, R.A.: “Application of Petrographic Image Analysis to the Characterization
of Fluid-Flow Pathways in a Highly Cemented Reservoir: Kane Field,
Pennsylvania, USA,” J. of Petroleum Science and Engineering 13 (1995) pp. 141-
154.

13. Lenormand, R., and C. Zarcone: “Role of Roughness and Edges During Imbibition
in Square Capillaries,” SPE 13264 presented at the SPE Annual Technical
Conference and Exhibition, Houston, TX, 16-19 September 1984.

14. Lenormand, R., C. Zarcone, and E. Touboul: “Numerical Models and Experiments
on Immiscible Displacements in Porous Media,” J. of Fluid Mechanics 189 (1988)
pp. 165-187.

15. McCreesh, C.A., R. Ehrlich, and S.J. Crabtree: “Petrography and Reservoir Physics
II: Relating Thin Section Porosity to Capillary Pressure, the Association Between
Pore Types and Throat Size,” AAPG Bulletin (Oct. 1991) pp. 1563-1578.

16. Perri, P.R., M.A. Emanuele, W.S. Fong, and M.F. Morea: “Lost Hills CO2 Pilot:
Evaluation, Design, Injectivity Results, and Implementation,” SPE 62526, paper
presented at the SPE/AAPG Western Regional Meeting, Long Beach (June 19-23,
2000).

17. Ross, C.M., C.A. Callender, J.B. Turbeville, and J.J. Funk: “Modeling Capillary
Pressure Behavior Using Standard Open Hole Wireline Log Data: Demonstrated on
Carbonates from the Middle East,” SPE 30543, paper presented at the SPE Annual
Technical Conference and Exhibition, Dallas (Oct. 23-25, 1995).

17. Schembre, J.M., S. Akin, L.M. Castanier, and A.R. Kovscek: “Spontaneous Water
Imbibition into Diatomite,” SPE 46211, paper presented at the SPE Western
Regional Meeting, Bakersfield (May 10-13, 1998).

18. Wallace, N.J., and E.D. Pugh: “An Improved Recovery and Subsidence Mitigation
Plan for the Lost Hills Field, California,” SPE 26626 presented at the SPE Annual
Technical Conference and Exhibition, Houston, TX, 3-6 October 1993.

19. Wilt, M., P. Zhang, M. Morea, D. Julander, and P. Mock: “Using Crosswell
Electromagnetics to Map Water Saturation and Formation Structure at Lost Hills,”
SPE 68802, paper presented at SPE Western Regional Meeting, Bakersfield (March
20-30, 2001).

19. Zhou, D., L. Jia, J. Kamath, and A.R. Kovscek: “Scaling of Counter-Current
Imbibition Processes in Low-Permeability Porous Media,” J. of Petroleum Science
and Engineering, in press.

20. Zhou, D., J. Kamath, F. Friedmann, and M. Morea: “Identifying Key Recovery
Mechanisms in a Diatomite Waterflood,” SPE 7514, paper presented at the SPE
Improved Oil Recovery Symposium, Tulsa (April 13-17, 2002).

273
Table 1. Study Samples and Measured Core Properties

DEPTH ROCK SEM ∅ ∅ K SO SW BVO BVW


(ft) TYPE (% area) (% vol) (md) (% ∅) (% ∅) (% vol) (% vol)
1341.12 Clayey 14.72 56.1 1.1 46.3 51.9 25.97 29.12
1341.23 Clayey 12.27 55.4 0.9 46.3 51.9 25.65 28.75
1450.00 Sandy 10.34 46.3 43.42 36.7 63.3 16.99 29.31
1484.95 Sandy 5.77 62.2 30.4 35.3 57.5 21.96 35.77
1516.80 Sandy 14.25 49.3 21 38.1 60.2 18.78 29.68
1526.85 Sandy 9.74 46.2 2.8 54.3 45.7 25.09 21.11
1553.65 Clayey 9.91 52.5 1 49.1 45.8 25.78 24.05
1642.25 Clean 11.36 61.7 1.1 50.1 47.1 30.91 29.06
1706.00 Clayey 12.73 48.2 2.1 58.4 40.7 28.15 19.62
1801.00 Clean 10.65 59.4 1.2 61.3 38.4 36.41 22.81
1880.00 Clean 8.43 54.9 0.4 56.2 40.3 30.85 22.12
Quarry1 Quarry 18.12 76.6 14.8

Table 2. Proportion of SEM Porosity Attributed to Each Pore Type

Sample PT1 PT2 PT3 PT4 PT5 PT6


1341.12 0.4407 0.1794 0.1775 0.0942 0.0884 0.0197
1341.23 0.2186 0.3753 0.2294 0.0832 0.0444 0.0491
1450.00 0.2161 0.0813 -0.0247 0.2632 0.0779 0.3862
1484.95 0.2838 0.1312 0.1452 0.0594 0.3609 0.0194
1516.80 0.0749 0.0426 0.0002 0.3061 0.0386 0.5376
1526.85 -0.0418 0.2811 0.1178 0.2307 -0.0141 0.4262
1553.65 0.0320 0.3874 0.4157 0.0436 0.0747 0.0466
1642.25 0.3404 0.3994 0.0599 0.1153 0.0320 0.0530
1706.00 0.0900 0.3437 0.2357 0.3198 0.0078 0.0029
1801.00 0.0461 0.4589 0.4255 -0.0030 0.0707 0.0017
1880.00 0.0539 0.5304 0.3360 0.0455 0.0415 -0.0072
Quarry1 0.1487 0.2370 0.4420 0.0803 0.0772 0.0148

Table 3. Pore Body Attributes for Each Pore Type

Pore Average Minimum Pore Pore Body Width


Type Body Width (µm) Range (µm)
PT1 0.82 0.114 - 2.508
PT2 1.35 0.230 - > 4.334
PT3 1.64 0.114 - 0.456;
1.142 - > 4.334
PT4 1.74 1.482 - 4.332
PT5 4.02 0.456 - 120
PT6 6.50 2.510 - > 4.334

274
Table 4. Capillary Pressure Model

µ)
Throat Radius (µ PT1 PT2 PT3 PT4 PT5 PT6
≥ 9.9441 0.0267 0.0512 0.0162
9.9440 - 1.5364 0.0173 0.0030 0.0151 0.0070
1.5363 - 0.2361 0.4974 0.7701
0.2360 - 0.0718 0.3510 0.3954 0.4847 0.8893 0.6613
0.0717 - 0.0220 0.3142 0.3933 0.2936
0.0219 - 0.0018 0.1615 0.2711 0.1625
sum 0.8267 1.0865 0.9994 0.7731 0.9556 1.1406

Table 5. Throat Size Attributes for each Pore Type

Pore Average Minimum Pore Body Width Average Throat Throat Radius
Type Pore Body Width (µm) Range (µm) Radius (µm) Range (µm)
PT1 0.82 0.114 - 2.508 0.0631 0.0018 - 0.236
PT2 1.35 0.230 - > 4.334 0.0734 0.0018 - 0.236
PT3 1.64 0.114 - 0.456; 0.5244 0.0718 - 9.944
1.142 - > 4.334
PT4 1.74 1.482 - 4.332 0.6042 0.2361 - 9.944
PT5 4.02 0.456 - 170 2.2409 0.0718 - >9.944
PT6 6.50 2.510 - > 4.334 0.7843 0.0219 - >9.944

Table 6. Fluid Saturation Models

Saturation PT1 PT2 PT3 PT4 PT5 PT6


Swc 68 29 47 42 68 71
Soc 33 66 54 59 13 29
Swl 69 24 57 27 53 66
Sol 31 76 43 73 47 34

275
1000
Well Data
Sandy
Clayey
Clean
100 Quarry
Quarry Plugs
Permeability (md)

Pore Type 1 Pore Type 4


10

0.1
40 50 60 70 80
Pore Type 2 Pore Type 5
Porosity (% volume)

Fig. 1 Porosity and permeability values measured on 12 study


samples (black) and over 1000 core plugs from Lost
Hills well OB-7 (gray) and Grefco Quarry (open circles)
near Lompoc, CA.

Pore Type 3 Pore Type 6


Fig. 3 Representative binary images of the six pore types.
µm.
Each image is 237.6 by 177.6µ

100 100
Sandy
Clayey
Clean
Quarry
Permeability (md)

Permeability (md)
10 10

1 1

r = 0.17 r = - 0.78
0.1 0.1
Fig. 2 Grayscale image (left) and its corresponding binary -0.1 0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6
Pore Type 1 Pore Type 2
image (right) after segmentation. The CaCO3 foraminifer
µm. SEM porosity for this view is 17.16%.
test is 60 by 110µ Fig. 4 Measured permeability versus proportion of PT1 (left)
and PT2 (right).

100 100
r = 0.84
Permeability (md)

Permeability (md)

10 10

1 1
Sandy
Clayey
Clean
r = 0.66 Quarry

0.1 0.1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
PT4, PT5, and PT6 Permeable Pore Types

Fig. 5 Measured permeability versus proportion of PT4, PT5,


and PT6 summed (left) and permeable pore types with
contributions from PT1 added for sandy and quarry
diatomite samples (right).

276
0.001 0.001 0.001
70 70

Log Oil Saturation (% porosity)

Log Oil Saturation (% porosity)


1341.12 1341.23 1450.00
Sandy
0.01 0.01 0.01 Clayey
r = - 0.53
65 Clean 65
Pore Throat Radius (microns)

Pore Throat Radius (microns)

Pore Throat Radius (microns)


0.1 0.1 0.1
60 60

1 1 1
55 55
10 10 10
50 50
100 100 100
100 80 60 40 20 0 100 80 60 40 20 0 100 80 60 40 20 0 45 45
Mercury Saturation (%) Mercury Saturation (%) Mercury Saturation (%) r = 0.73
0.001 0.001 0.001
40 40
1484.95 1516.80 1526.85 0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.2 0.4 0.6 0.8 1
0.01 0.01 0.01 Pore Type 2 Permeable Pore Types
Pore Throat Radius (microns)

Pore Throat Radius (microns)

Pore Throat Radius (microns)


0.1 0.1 0.1 Fig. 8 Oil saturation from wireline log data versus pore type 2
1 1 1
(left) and permeable pore types (PT1 in sandy and
quarry samples, PT4, PT5, and PT6; right).
10 10 10

100
100 80 60 40 20 0
100
100 80 60 40 20 0
100
100 80 60 40 20 0 PT6 66% 34%
Mercury Saturation (%) Mercury Saturation (%) Mercury Saturation (%)
0.001 0.001 0.001
1553.65 1642.25 1706.00
PT5 53% 47%
0.01 0.01 0.01
PT4 27% 73%
Pore Throat Radius (microns)

Pore Throat Radius (microns)

Pore Throat Radius (microns)

0.1 0.1 0.1

PT3 57% 43%


1 1 1

10 10 10
PT2 24% 76%

100 100 100 PT1 69% 31%


100 80 60 40 20 0 100 80 60 40 20 0 100 80 60 40 20 0
Mercury Saturation (%) Mercury Saturation (%) Mercury Saturation (%)
0.001 0.001 0.001 0 20 40 60 80 100
1801.00 1880.00 Quarry1 Fluid Saturations for Each Pore Type
0.01 0.01 0.01
Pore Throat Radius (microns)

Pore Throat Radius (microns)

Pore Throat Radius (microns)

Fig. 9 Oil (black) and water (gray) saturations per pore type
0.1 0.1 0.1
within the reservoir.
1 1 1

10 10 10 PT6 71% 29%


100 100 100
100 80 60 40 20 0 100 80 60 40 20 0 100 80 60 40 20 0 PT5 19% 68% 13%
Mercury Saturation (%) Mercury Saturation (%) Mercury Saturation (%)

Fig. 6 Measured (line) and predicted (dots) mercury injection PT4 42% 59%
curves for each study sample.
PT3 47% 54%
65 65 PT2 29% 66%
r = - 0.74
Oil Saturation (% porosity)

Oil Saturation (% porosity)

Sandy
60 Clayey 60
Clean
PT1 68% 33%
55 55

50 50 0 20 40 60 80 100
45 45
Fluid Saturations for Each Pore Type
40 40 Fig. 10 Oil (black), water (dark gray), and unfilled (light gray)
35
saturations per pore type within core samples.
r = 0.83 35
30 30
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.2 0.4 0.6 0.8 1
Pore Type 2 Permeable Pore Types

Fig. 7 Measured oil saturation from core plugs versus pore


type 2 (left) and permeable pore types (PT1 in sandy and
quarry samples, PT4, PT5, and PT6; right).

277
278
Improving Reservoir Definition Through Integration
of Multiscale Petrophysical Data

Liping Jia and Anthony Kovscek

1. INTRODUCTION

In this research, integration of multi-scale data is studied in order to improve


reservoir definition. There are three different scales of data considered: pore-scale, core-
scale, and well-scale data. Pore network modeling is used to integrate pore-scale and
core-scale data. A neural network analysis is proposed to correlate wireline logs with
pore-scale data. Integration of different scale petrophysical data can supply more
accurate input for reservoir simulation, which can provide more accurate ultimate oil
recovery and other predictions for the reservoir.

A diatomite reservoir is chosen to study in this report because of its low


permeability and large volumes of oil in place. The method is equally applicable to
sandstones, chalks, and other carbonates. The available data for reservoir study include
well log, core scale measurement, pore type proportions and production data. Because
these measurements are made at different length scales and the correlation among them is
generally not known, it is likely that interpretation of the reservoir and its performance is
inconsistent. Therefore, it is suggested integrating all the petrophysical data available be
accomplished when formulating a reservoir model. In this report, it is considered how to
better link pore type proportion, core scale permeability and relative permeability
measurements with well log data. The outline of approach is shown in Fig. 1.

Pore network modeling is a technique to model multiphase flow at the small scale
and to predict average properties, such as permeability, capillary pressure, relative
permeability and ultimate oil recovery. A pore network model is developed to calculate
permeability for Lost Hills core samples. Image analysis of petrographic thin sections is
applied to obtain the pore size, pore type, and proportion, throat size, type, and
proportion. Six pore types are acquired from analysis of diatomite core samples from
Lost Hills reservoir and six throat types correspond to each pore type. The pore-scale
information will be the input of the pore network modeling. A three-dimensional pore
network model is developed and used to predict the permeability for different core
samples. In order to calculate absolute permeability, the pore network is filled with a
single fluid first. Under steady state flow of incompressible fluid, a mass conservation
equation for each pore body is developed. The conductance of a throat is calculated using
Poisseuille’s law for flow in a pipe. After the pressure distribution for each pore body is
obtained, the total flow rate flowing through the pore network model can be calculated.
Then Darcy’s law is applied to calculate absolute permeability. Relative permeability and
capillary pressure calculations are under development. The preliminary results are shown
in Table 1. The samples we choose include sandy, clayey and clean diatomite lithologies.

279
Well logs Data
K field
Porosity
0.9 9
0.8 8
0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.7
Kri
7
Empirical 500
0.6
0.5
Krw
6
5 Pc
0.4
0.3
Kro
4
3
Correlation 700
0.2 2
0.1 1 900
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
1100
Netw Sorw prediction

Depth
mod ork 1300
eling
Scaling 1500

1700

1900

Pore size, shape, frequency,


Throat size, shape, frequency
Core scale imbibition study

Figure 1 Outline of approach

The samples of depth 1341.12 ft and 1706.00 ft are clayey diatomite. The samples of
depth 1450.00 ft and 1516.80 ft are sandy diatomite. The sample of depth 1880.00 ft is
clean diatomite. The porosity and permeability data in the second and third column are
from the lab measurement. The permeability (meas.) is air permeability. Permeability
(calculate) is from pore network modeling study for water/brine flow.

Table 1. Preliminary results for diatomite

Sample (ft) Porosity (meas.) K (meas.) K (calculate)


1341.12 0.561 1.1 1.656
1450.00 0.463 43.42 1.7546
1516.8 0.493 21.0 1.0047
1706.00 0.482 2.1 3.537
1880.00 0.549 0.4 2.046

In order to anchor the results from pore network modeling, spontaneous


imbibition experiments are performed, on both outcrop core samples and field core
samples from the Lost Hills reservoir. The important idea of laboratory measurements is
to relate these measurements to the well log and reservoir scales. Therefore, scaling-up
groups for counter-current and co-current imbibition are also investigated.
The objectives of experimental study are to obtain the saturation profile and rate
of imbibition in order to determine fluid displacement mechanisms during counter-

280
current and co-current water imbibition into air- and oil-filled cores. X-ray CT scanning
is used for visualization of displacement patterns. The diatomite outcrop sample is a well-
characterized porous medium and has no initial oil or water saturation. The rock is
consolidated and almost pure white indicating little silt or mud. Some evidence of
bedding planes is visible to the naked eye. However the field samples contain connate
water and residual oil, which creates difficulties for the experiment. The entire length of
the core is imaged because we scan along the direction of flow using an apparatus
designed specially for imbibition study, shown in Fig. 2. A single slice provides us with a
picture of the progress and saturation pattern of imbibition. Different viscosity fluids are
used for outcrop experiments in order to observe viscosity effects. For field sample
experiment, crude oil is applied to simulate the reservoir condition as much as possible.
The whole setup is kept close to reservoir temperature for experiments with field cores
and room temperature for outcrop experiments.

water and produced


nonwetting
fluid

water
in

water jacket
detector

core

X-rays
endcap

Figure 2 Schematic of imibibition cell set up for countercurrent imbibition

From the results for outcrop experiments, it is shown that the water front is
relatively sharp initially and becomes more diffuse as the front moves through the core.
One set of images for an outcrop experiment is shown in Fig. 3 (Note the satuation scale
on the right). To summarize the outcrop sample experimental results, imbibition
performance as gauged by speed of recovery and amount recovered decreases as the non-
wetting phase viscosity increases. Water movement through the cores becomes
increasingly diffuse with viscosity also. From the field sample results, it is noticed that
the displacing front is no longer obvious and becomes more diffuse. It takes much longer
to imbibe water for field core experiment even with higher temperature. The recovery for
a field sample experiment is lower than the results for outcrop experiment.

281
Sw
1

0 77 130 243 850 3327

Figure 3 CT-derived water saturation images of counter current water/blandol imbibition in


diatomite. Time is given in minutes beneath the image. Viscosity of blandol is 25 cp.

In order to scale-up the experimental results to predict the reservoir behavior, a


theoretical analysis of the flow characteristics of imbibition processes was also
conducted. A new set of scaling groups for counter-current and co-current imbibition is
acquired. Both the simulation and the experimental data are used to verify the new
scaling group, which shows that the new dimensionless time that can improve
significantly the scaling of spontaneous imbibition in low permeability porous media.
The new scaling is used to correlate experimental and simulation results. It improves
correlation significantly by taking end-point fluid-phase mobilities and the mobility ratio
into account. The scaling also explains the empirical observation that water/oil imbibition
time scales inversely with the geometric mean of water and oil viscosity, whereas
water/air imbibition scales inversely with water viscosity. The field sample results are
still under investigation and will be reported soon.

2. FUTURE WORK

The pore network model will be used to calculate relative permeability and
capillary pressure curves for diatomite. The field sample experiments need to be
completed and analyzed. The scaling group for co-current imbibition needs to be verified
using simulation data and experimental data if possible. The correlation between pore
network model and wireline logs will be studied also. A novel method will be developed
to integrate core-scale data with wireline log and pore network modeling results. An
accurate and consistent data set will be generated to be used in simulation study, which
will give us more accurate prediction of ultimate oil recovery for diatomite reservoirs.
The study will also be extended to other kinds of reservoirs as well.

282

You might also like