Oil-recovery predictions for surfactant polymer fooding

a b s t r a c t
There is increasing interest in surfactantpolymer (SP) fooding because of
the need to increase oil production from depleted and water fooded
reservoirs. Prediction of oil recovery from SP fooding, however, is complex
and time consuming. Thus, a quick and easy method is needed to screen
reservoirs for potential SP foods. This paper presents a scaling model that
is capable of producing reasonable estimates of oil recovery for a SP food
using a simple spreadsheet calculation. The model is also useful for initial SP
design.
We present key dimensionless groups that control recovery for a SP food. The
proper physics for SP foods including the optimal salinity in the three-phase
region and the trapping number for residual oil saturation determination has
been incorporated. Based on these groups, a BoxBehnken experimental
design is performed to generate response surface ts for oil recovery
prediction at key dimensionless times. The response surfaces derived can be
used to estimate the oil recovery potential for any given reservoir and are
ideal for screening large databases of reservoirs to identify the most
attractive chemical fooding candidates. The response function can also be
used for proper design of key parameters for SP fooding. Our model will aid
engineers to understand how key parameters affect oil recovery without
performing time consuming chemical simulations. This is the rst time that
dimensionless groups for SP fooding have been derived comprehensively to
obtain a response function of oil recovery as a function of dimensionless
groups.
Published by Elsevier B.V.
1. Introduction
Surfactantpolymer (SP) fooding processes involve the injec- tion of a
surfactantpolymer slug followed by a polymer buffer and
chase
water
injection. If designed correctly, the surfactant increases the capillary number,
which is crucial for the mobilization and recovery of tertiary oil. Polymer increases the sweep efciency by
lowering the mobility ratio. If the reservoir crude oil has sufcient saponiable
components, soap (surfactant) is gen- erated in situ by the reaction of these
components with the injected alkali, thus adding more surfactant to the food
(Lake, 1989).
Recovery predictions for SP foods involve numerous para- meters and complex
simulations. One way to simplify the process and predict oil recovery is to use a
screening model based on a few key input variables or dimensionless groups.
Dimensionless groups strategically combine properties so that their units cancel
out. If done correctly, a reservoir with the same dimensionless groups should have
similar dimensionless oil recovery curves. Dimensionless groups are typically
attained in two ways: dimen- sional analysis and inspectional analysis (Shook et
al., 1992). The dimensional analysis approach is based on Buckingham's theorem. Dimensional analysis is the only option in problems where equations and
boundary conditions are not completely articulated. It computes sets of
dimensionless parameters from given variables, even if the form of the equation is
still unknown. However, the choice of dimensionless parameters is not unique:

Buckingham's theorem only provides a way of generating sets of dimensionless

parameters. Inspectional analysis takes advantage of the problem's full
mathematical specication based on physical laws, and reveals a higher degree of
similarity than dimensional analysis (Sonin, 2001). We use the University of
Texas Chemical Flooding Simulator
(UTCHEM, 2000) for our simulations, for a typical one dimensional
surfactantpolymer food UTCHEM requires around 170 parameters, carrying out
Buckingham's dimensional analysis on this would result in over 160 groups. Using
these 160 groups for creating a response surface for oil recovery in SP food
would be extremely time consuming and impractical. This was another reason why
we chose inspectional analysis over dimensionless analysis as our method for
obtaining dimensionless groups for SP fooding.
Previous screening models such as that of Paul et al. (1982) did not consider
gravity and salinity effects. Pope et al. (1979) and Shook (1988) carried out
sensitivity studies on SP foods and showed oil recovery as a function of RL, Ng,
and M, but did not attempt to correlate oil recovery to the parameters studied.
Gupta et al. (1988) showed oil recovery as a function of RL, Ng, M, TDs, and NTD.
Thomas et al. (2000) described scaling criteria for the micellar fooding process
from the basic mass balance equations using inspectional and dimensional
analysis. Micellar fooding experi- ments were carried out in sandstone cores of
two different sizes, and the scaled up recovery curves were compared. The
agreement between the predicted and actual recoveries was good in some
cases, but poor in others. Poor agreement is likely because they did not consider
important effects such as heterogeneity.

This paper presents a screening model that is capable of quickly producing

quantitative estimates of oil recovery for a given surfactantpolymer food
including the effects of
heterogeneity and salinity. It also develops the
important dimensionless groups necessary to scale SP foods using inspectional
analysis. The derived groups are used in a BoxBehnken experimental design to
produce response surfaces for dimensionless oil recovery during SP fooding. We
also give an approximate method to correct the scaling groups for the addition
of alkali in a SP process. Thus, when analysis from laboratory experiments of
chemical interaction of fuids and rocks are available, this model can help to
quickly screen reservoirs for their surfactantpolymer food potential.
2. Procedure
The rst step in creating a screening model is to determine the important
dimensionless groups that govern recovery for SP foods. Inspectional analysis was
carried out on the mass balance equations and boundary conditions for SP
fooding, and auxiliary relations included in the UTCHEM simulator (Technical
Documentation,2000). We also included important groups for heterogeneity. The
objectives for any SP design are given below:
Select the surfactant molecule and concentration that can increase the
capillary number several orders of magnitude.
Such an increase may mobilize and produce the residual oil saturation. This
objective is usually investigated with a parti- cular oil, brine, and rock using
salinity scanning pipette tests.
Reduce surfactant and polymer adsorption
water and oil. This objective is met by design

and

interfacial tension between

of surfactant formulation.
Determine surfactant and polymer slug size based on the mobility ratios,
heterogeneity, and reservoir geometry.
Predict the SP potential in incremental oil recovery.

The response functions based on the dimensionless groups developed in this

paper will provide a good estimate for the last two objectives. The main idea in
design is to minimize the amount of surfactant required, but still achieve high oil
recovery.
2.1. Dimensionless groups
Fig. 1 represents the schematic of the reservoir under con- sideration. As
presented in the gure, the reservoir modeled is a two-dimensional, Cartesian,
multi-layered reservoir. The initial oil saturation is the saturation after
waterfood, assumed to be uni- form spatially. Inspectional analysis was carried
out on the SP fow equations and boundary conditions. The governing equations
for the miscible displacement of remaining oil by the microemulsion phase are
conservation equations, Darcy's equations, capillary pressure equations, phase
equilibrium equations, and auxiliary equations described in UTCHEM (2000).
The conservation equations, Darcy's equations, capillary pres- sure, saturation
constraint, and initial and boundary conditions were expanded by replacing the
variables with their respective scaling factors. From inspectional analysis of the
governing equa- tions and boundary conditions, 11 dimensionless groups were
obtained. The detailed inspectional analysis procedure and the governing
equations can be found in Rai (2008).
Inspectional analysis does not tell us how to include the auxiliary relations, that
is, equations that are included in UTCHEM to model certain physical behavior but
that are not a part of the conservation equation, Darcy's or capillary pressure
equations and thus would not be a part of the conventional inspectional analysis
procedure. Thus a more empirical approach was taken in this research. Additional
groups were recognized from the auxiliary relations used in UTCHEM (2000). For
example, we added the Bond number to our groups since both gravity and
buoyancy can affect mobilization of oil. Bond number does not appear in the
inspectional analysis, but was included when considering the auxiliary equations
used in UTCHEM. Without the inclusion of the Bond number, we found that the
simulation dimensionless recoveries for the same values of the scaling groups
were not sufciently equal, especially for larger pore volumes injected. The
relative permeability functions are dened as linear inter- polations between the
given inputs of relative permeabilities at low and high trapping numbers. That
is,
l

Tl
The capillary desaturation curve is a function of the trapping number, NTl, and
the T-parameter, Tl. NTl includes both capillary number and Bond number (Jin,
1995). The capillary number is dened as
!
The Bond number is dened as
NBl
for l 1; ; Np
4
where k is the permeability and g is the gravitational force constant (UTCHEM,
2000).
The
above
equations
were
analyzed to generate additional
dimensionless parameters that were included in the earlier set of groups. These

are model parameters used to t the capillary desaturation curve (CDC) with
the experimental data. The T parameter was also included as a dimensionless
group. The T parameter values are a function of the wettability and the rock
type. This way, the wettability and rock type are included in the set of
dimensionless groups, for a given set of constant T parameters (T1, T2, T3). We
assumed that the T parameter value for the microemulsion phase is equal to
the water phase T parameter value (T1 T3) as this assumption is often made
in SP food simulations. Some groups such as initial and residual oil and water
saturations also occur as dimensionless groups through the initial and boundary
conditions.
The DykstraParsons coefcient was included as one of the groups to scale
heterogeneity. Correlation lengths were not needed since a perfectly layered
reservoir model is used. The permeability layers, however, can be arranged for
the same value of DykstraParsons coefcient in ascending downward, descending downward or a random sequence. Reservoirs, however, often exhibit
upwards or downwards coarsening sequences. If perme- ability is greater at the
top (upward coarsening) or at the bottom (downward coarsening) the recoveries
can be impacted owing to the combined effect of gravity (or buoyancy) and
channeling. To approximately account for the arrangement of the layers, we
introduce one additional parameter, the slope of the dimension- less horizontal
permeability with dimensionless depth. A linear t
The 18 dimensionless groups used to scale SP fooding served as the basis for
creating the screening model. Experimental design and response surfaces were
used to achieve efcient and compre- hensive results. BoxBehnken experimental
designs were used to specify the input group values and reservoir properties for
each of the 624 simulations used in the model. The results of these simulations
were then used to generate response surface ts for each parameter of the
model. The t-statistics were used to under- stand the signicance of each group.
Normalized values of the groups were used in the response surface ts to assist
in reducing the number of coefcients in each t. The number of coefcients
was reduced from 189 to between 9 and 11 for each surface response function
while
still maintaining an acceptable level of error in the ts. Additional
conclusions are
Surfactantpolymer (SP) has been
scaled for the rst time including
all important physical processes
so that reservoirs can now be
quickly screened for oil recovery.
The response functions can also
be used for proper engineering
design of key parameters.
Equations
and
physical
phenomena modeled in UTCHEM
were used as the basis for
carrying out inspectional analysis
and determining the nal set of
the
most
signicant
dimensionless
groups.
Effects

that are not modeled in UTCHEM

were excluded from this model.
The effects of heterogeneity and
rock type were included in the
set
of
dimensionless
groups
through
the
DykstraParsons
coefcient, T parameters, and
the dimensionless vertical permeability
trend.
The
groups
served as the basis for scaling
heterogeneity.
The
most
important
dimensionless
groups
were
determined using t-statistics and
the values for the coefcients in
the response surface function.

The
most
important
dimensionless groups for ultimate
recovery
were:
initial
oil
saturation, kO;L =kO;H and the
heterogeneitygroups
assuming
adequate mass of chemicals are
injected at favorable salinity
gradient.

(6) While 18 dimensionless

groups
are
needed
to
completely
scale
surfactant
polymer foods, only 911 are
required for good
accuracy
depending on the dimensionless
time (pore volumes injected).
The screening model is ideal for
use by operators in quickly
screening large databases of
reservoirs
for
the
most
attractive
surfactantpolymer
fooding candidates.

The model could also be used in designing a SP fooding process to optimize the
recovery factor by changing the controllable or operational parameters.
Acknowledgments
The authors are grateful for the nancial support provided by the participating
companies of the Chemical EOR research pro- gram at The University of Texas
at Austin. Dr. Russell T. Johns is the Beginning Professor of Petroleum and Natural
Gas Engineering in the John and Willie Leone Family Department of Energy and
Mineral Engineering at the Pennsylvania State University. He is also a member

of the Earth and Mineral Sciences Energy Institute at Penn State. Larry Lake
holds the W.A. (Monty) Moncrief
Centennial Chair at The University of Texas.