You are on page 1of 18

SPE 90938

Dynamic Simulations With an Improved Model for Foam Generation


S.I. Kam*, Q. P. Nguyen,† Q. Li†, and W. R. Rossen†
* - The University of Adelaide

- The University of Texas at Austin

Copyright 2004, Society of Petroleum Engineers Inc.


Introduction
This paper was prepared for presentation at the SPE Annual Technical Conference and Foam can improve sweep efficiency in gas-injection improved
Exhibition held in Houston, Texas, U.S.A., 26–29 September 2004.
oil recovery (IOR) processes,1-3 redirect acid flow in matrix
This paper was selected for presentation by an SPE Program Committee following review of
information contained in a proposal submitted by the author(s). Contents of the paper, as
acid stimulation,4-6 and increase the efficiency of
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to environmental remediation of aquifers.7,8 A continuing goal of
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at foam research is the development of a fully mechanistic,
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
predictive model. This paper describes efforts toward such a
for commercial purposes without the written consent of the Society of Petroleum Engineers is model and insights gained from application of the model to
prohibited. Permission to reproduce in print is restricted to a proposal of not more than 300
words; illustrations may not be copied. The proposal must contain conspicuous dynamic displacements. Before providing a detailed
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
description of the model, it is worthwhile to review the
mechanisms of foam in porous media and the experimental
observations that the model attempts to reproduce.
Abstract
We present the first 1D simulations of dynamic foam Water and Gas Mobility
displacements with a new population-balance model When strong foam forms in porous media the gas phase is
incorporating bubble creation controlled by pressure gradient. broken into discrete bubbles separated by liquid films called
For the first time, a population-balance model is fit to steady- lamellae.1,2,9 Bubbles are thought to be as large as pores, or
state experimental data for both the three foam states (coarse larger. Liquid and gas continue to flow as separate phases.
foam, intermediate state, and strong foam) and the two strong- Numerous studies find that foam in porous media does not
foam regimes (low-quality and high-quality) observed in directly alter the relative permeability krw or viscosity µw of
laboratory studies. Simulations confirm the stability of the the aqueous phase in the foam.10-13 Specifically, the flow of
coarse-foam and strong-foam states to small perturbations, and aqueous phase in the midst of foam is given by
the instability of the intermediate state.
In dynamic displacements, the model shows foam u w = kk rw ( S w ) / µ w ∇p .................................................. (1)
generation as injection rates increase, or as liquid fraction of
injected fluids increases, in agreement with laboratory where uw is water superficial velocity, k absolute permeability,
observations. ∇p the magnitude of pressure gradient (neglecting gravity for
When coarse foam is created instead of strong foam, there simplicity) and krw the same function of water saturation Sw as
is a narrow region of finer foam predicted near the gas in the absence of foam
displacement front. This region appears to play a role in foam Foam does greatly reduce the mobility of gas, and to a
generation. However, in the cases examined, foam generation greater extent the smaller the bubble size of the foam.1,2,9,14,15
occurs at roughly the same injection rate as predicted by local- In mechanistic simulators,9,13,16-20 the bubble size is quantified
steady-state theory. Because of this narrow region of finer- as nf, the number of lamellae per unit volume of gas phase.
textured foam, fronts can be sharper than estimated from The smaller the bubbles, the larger nf, and the lower the gas
fractional-flow theory assuming a constant gas viscosity at its mobility. The lamellae impart an effective yield stress to the
steady-state value behind the displacement front. gas,15 trapping a significant portion of the gas in place even at
If a strong foam forms in the low-quality-regime, the high pressure gradient,13,21 and resist the flow of the gas that
kinetics of foam generation and destruction affect the length of does flow. Because of both the yield stress and a shear-
the entrance region in which foam forms. Therefore, the length thinning drag on the moving bubbles, the effective viscosity of
of the entrance region can be used to calibrate the kinetic gas in foam is shear-thinning.
parameters in the model. The displacement front and the bank
behind it, however, are essentially what one would have Foam Generation and Destruction
predicted from local-steady-state modeling. The complexities While gas mobility depends on bubble size, bubble size
of population-balance modeling are not necessary, if it is itself depends on several processes that create and destroy the
known that strong foam will be created. lamellae that separate bubbles. The destruction of lamellae
increases as capillary pressure increases; in particular, foam is
observed to go through an abrupt collapse as capillary pressure
2 SPE 90938

rises to a limiting value, the "limiting capillary pressure" superficial velocities or even between complete plugging and
Pc*.22-24 This implies that the rate of lamella destruction has a flow.
singularity, approaching infinity at Pc*. As a practical result, If one fixes injection rates rather than pressure drop, one
foam passes through a wide range of properties, from strong observes a jump from coarse foam to strong foam, i.e. the
foam to much weaker foam, in the immediate vicinity of Pc*, abrupt change in state called "foam generation." This jump
which tends to hold Pc near Pc* over a range of superficial occurs at lower injection rate if the liquid fraction of the
velocities of gas and liquid. Because capillary pressure and injected fluids is increased. Rossen and Gauglitz29 contend
water saturation are related in porous media, the value of Pc* that this results from a larger initial population of liquid lenses
corresponds to a limiting water saturation, Sw*. Because Pc is and lamellae available for mobilization at higher injected
nearly constant near Pc* over a range of superficial velocities, liquid fraction.
Sw is likewise nearly constant near Sw* under these same In addition, in the laboratory it is often seen that a
conditions.23 As a result (Eq. 1), pressure gradient ∇p is nearly sufficient, through temporary, perturbation in injection rates or
independent of gas superficial velocity when Pc is near Pc*. other conditions can cause a jump from coarse foam to strong
An example is shown in the upper-left portion of Fig. 1.25 For foam that would not have occurred with steady injection rates.
modeling purposes it may be easier to represent the rate of
lamella destruction as a function of water saturation and Sw* Steady-State Strong-Foam Regimes
than in terms of Pc and Pc*. If strong foam is created, it exists in two steady-state flow
Much less is known about lamella destruction at other regimes.8,12,25,32,33,34 One is described above: If Pc is close to
conditions than in the vicinity of Pc* and Sw*. Pc* (Sw close to Sw*), then pressure gradient ∇p is nearly
Lamellae can created by several processes,26,27 including independent of gas superficial velocity. This is called the
"snap-off," i.e. creation of a new lamella by capillary "high-quality regime," or the "coalescence regime."35 At lower
processes in a pore throat; by "leave behind" (lenses of liquid foam qualities, there is a transition to another regime
stranded in pore throats during drainage of adjacent pores); illustrated in the lower-right portion of Fig. 1. Here pressure
and "lamella division," which occurs when foam is flowing gradient is nearly independent of liquid superficial velocity.
and lamellae pass unoccupied pore throats. Details of these It is thought that in this regime bubble size is fixed at some
processes can be found elsewhere.26-28 Recent studies29-32 lower limit,8,25 possibly at roughly pore size. Therefore nf is
indicate that the crucial step in foam generation is the fixed, constant, at some maximum value nfmax which it cannot
mobilization of lamellae as a function of pressure gradient ∇p, exceed. (In contrast, in the high-quality regime nf < nfmax, and
setting off the process of lamella division. The creation of new nf is very sensitive to injection rates and especially to Pc and
lamellae by division leads to a reduction in gas mobility; if Sw. In the low-quality regime Sw is not so near to Sw* as in the
injection rates are fixed, this in turn leads to higher pressure high-quality regime and Sw is not constant.) Photographs of
gradient, more mobilization and division, and an abrupt effluent foam25,34 provide qualitative evidence of a small
change of state from weak or "coarse" foam to strong foam, bubble size, independent of injection rates, in the low-quality
with orders of magnitude reduction in gas mobility. Gauglitz regime. It is plausible that such a minimum bubble size might
et al.31 show that if one fixes pressure drop across the porous exist, for two reasons. First, diffusion would tend to eliminate
medium instead of injection rates, one observes behavior bubbles smaller than pores, if formed.2 Second, new bubbles
illustrated in Fig. 2. In Fig. 2, liquid injection rate is fixed and cannot easily be created by division or snap-off if bubbles are
gas flows at a pressure drop set by a pressure regulator already smaller than pores. Bubbles smaller than pores do not
upstream of the beadpack. Pressure drops are moderate in Fig. divide easily at splits in the pore pathway, and snap-off would
2 because the beadpack permeability is high (30.4 darcy); have to occur before the trailing lamella reached the pore
nonetheless, the pressure drop with strong foam is of order throat. Strictly, then, this maximum value nfmax is a result of a
100 times higher than without foam. If gas injection rate were sharp increase in lamellae destruction rate and a sharp
fixed instead of pressure drop across the beadpack, upon decrease in lamella creation rate as nf nears nfmax. In practice, it
increasing gas injection rate, one would observe a point of is easier to model the low-quality regime simply by
sudden creation of strong foam, with an increase in pressure postulating a maximum value of nf, nfmax, rather than the
drop by a factor of about 100. Gauglitz et al.31 show similar complex changes in lamella-creation and -destruction
behavior holding injected quality (injected gas volume processes for nf near nfmax.32
fraction) fixed rather than liquid injection rate. Kim et al.34 report a modified low-quality regime for some
Fig. 2 illustrates three possible steady states at some foams made with liquid or supercritical CO2. They show that
injection rates: (1) a coarse-foam state, where bubbles (if they this behavior can be explained in the context of the low-
exist) are large, foam texture nf small, gas mobility relatively quality regime if gas viscosity depends on capillary pressure.
high (though perhaps reduced somewhat from that without any It is not clear how widely this sort of behavior extends.
foam), and ∇p small. With high gas mobility, there is little The distinction between the two strong-foam regimes
pressure gradient to mobilize existing lamellae, and the high- shown in Fig. 1 is crucial. First, the rheology of the two
mobility state is stable to small perturbations. (2) A strong- regimes is completely different, as illustrated in Fig. 1.
foam state, with small bubbles, large nf, low gas mobility, and Moreover, because the high-quality regime is sensitive to
large ∇p. (3) Between them, an intermediate state. Gauglitz et capillary pressure but the low-quality regime is not, diversion
al.31 found that this intermediate state was difficult to control between layers differing in permeability is expected for foam
and appeared to be unstable, reverting to higher and lower gas in the high-quality regime but not foam in the low-quality
regime. The foam quality at the transition between high-
SPE 90938 3

quality foam and low-quality foam depends on the factors that formation.5,20,37-44 Then one can represent foam texture
control the two regimes, and in particular the limiting capillary explicitly, if desired,20 or implicitly, as reflected in the change
pressure of the particular foam formulation in the given in gas mobility for Sw near Sw*, for instance. Cheng et al.38
formation and on the permeability of the porous medium. show that a local-steady-state model can represent the two
Further details on the low-quality and high-quality strong- strong-foam regimes without explicitly representing nf. One
foam regimes are in Alvarez et al.25 can allow for jumps between steady states when one state
reaches the limit of its range (foam generation or foam
Modeling Foam Behavior collapse) in the context of a local-steady-state model.42 With
There have been two approaches to modeling dynamic additional assumptions, local-steady-state modeling lends
foam displacements in porous media. In the population- itself to fractional-flow methods36,38,43,44 that provide useful
balance approach,9,13,16-20 foam texture nf is modeled insights and simplicity of use. The local steady-state approach
explicitly, using a balance equation on lamellae similar to the is valid as long as the foam quickly attains the given local
mass balances for surfactant or water. The mechanisms steady state on the time and length scales of the displacement.
considered crucial to lamella creation and destruction are Rossen et al.37 show that in some cases, even on the core
represented explicitly in the balance on nf. An example is scale, population-balance simulations can be modeled nearly
given below. Meanwhile, gas mobility is expressed as a equivalently with fractional-flow methods, i.e. assuming local
function of nf, Sw and other factors. The population balance steady state.
can represent dynamic changes in foam texture without Local-steady-state modeling is much less complex, and
assuming that foam texture has reached local steady state. more numerically stable, than population-balance modeling. It
Until now, most population-balance models9,17-20 have been is important therefore to determine the conditions under which
based on lamella creation by repeated "Roof" snap-off, this added complexity is required. A model that assumes that
governed by an aspect ratio between pore throats and adjacent foam immediately attains local steady state as strong foam is
pore bodies. This explanation for foam generation is clearly inadequate in cases where generation of strong foam is
problematic,36 and several studies have shown that in in doubt. In this paper we present dynamic simulations of
homogeneous porous media foam is created by mobilization steady co-injection of gas and liquid using a population-
of lamellae and subsequent division rather than primarily by balance model modified from that of Kam and Rossen.32 We
Roof snap-off.29-31 examine when and how foam generation occurs, and compare
Fig. 3 shows water saturation and foam texture nf from a foam generation in dynamic simulations to the steady-state
population-balance simulation of co-injection of gas and version of the same model and to trends in published
surfactant solution into a core filled with surfactant solution, laboratory studies. When strong foam is created, we compare
from Kovscek et al.18 The fit of this simulation to scans of the effect of kinetic parameters excluded from a local-steady-
water saturation and pressure-drop data (not shown here) was state model on the simulation results at different length scales.
good. Foam texture in the model is nearly uniform behind the The model presented here is not predictive and does not
foam front, but there is a maximum in nf near the gas include all known foam mechanisms, though it does fit a
displacement front. (Other simulations from Kovscek et al.17 variety of steady-state data and qualitative trends observed in
do not show this maximum in nf). The maximum is physically foam-generation experiments. The model allows one to
justified, though it is too localized to be confirmed directly by consider the separate roles of lamella generation and
laboratory data. At the front, bubbles are convected convection on foam propagation, compare steady-state
downstream to the front, and furthermore foam generation is behavior to dynamic displacements, and examine the role of
ongoing. Lamella destruction rate is low until Sw approaches local perturbations on foam generation and the stability of
Sw* and Pc approaches Pc*. Thus in the leading edge of the steady states.
displacement front, foam texture overshoots its steady-state Before presenting simulation results, we tune the model to
value, until local water saturation falls enough to trigger both the foam-generation data of Fig. 2 and new steady-state
greater lamella destruction near Pc*, whence foam texture falls strong-foam data like those of Fig. 1 for the same foam
to a roughly steady-state value in the foam bank. formulation and porous medium. This represents the first
Fig. 3 also shows a region near the inlet of the core in tuning of a population-balance model to such a range of data.
which nf rises and foam is created from the injected liquid and The experimental data fitted here are for foam in a high-
gas, in agreement with laboratory data.35 permeability (30.4 darcy) beadpack. Therefore the pressure
Kam and Rossen32 show that one can represent the high- gradients reported are much lower than would be observed in
and low-quality strong foam regimes and the three foam states lower-permeability, consolidated rock. Beadpacks and
(coarse, strong and intermediate) using a population-balance sandpacks are appropriate for mechanistic studies of foam,
model implementing foam destruction at Sw* and a maximum however, because the key features of foam behavior in
value of nf, nfmax. The model was not as successful in consolidated rock (three foam states,31 two strong-foam
representing the nearly complete plugging of foam in the regimes,8,12,25,33,34 abrupt foam generation upon increasing
intermediate regime. This same model predicts the existence injection rates,26,29,30,31 foam collapse at a limiting capillary
of two steady-state regimes within the strong-foam state, pressure,22,24 foam generation at a transition in
illustrated in Fig. 4 (cf. Fig. 1), though Kam and Rossen did permeabilities26,30) are seen in beadpacks and sandpacks as
not fit the model directly to data for the two regimes. well as consolidated rock.
An alternative to population-balance modeling is to
assume local steady state at all locations in the
4 SPE 90938

Experimental Method and Results where rc is the rate of foam coalescence per unit volume of gas
Linear-flow experiments were conducted to characterize the phase and nf the number of lamellae per unit volume of gas. Cc
flow behavior of strong foam. Gas and surfactant solution and n are model parameters.
were co-injected into a vertically-mounted 1-ft long beadpack For gas viscosity in the presence of foam, we use the
(k=30.4 darcy and φ=0.31) with three pressure taps. The gas shear-thinning effective viscosity of Hirasaki and Lawson14
phase was nitrogen and the surfactant solution was prepared derived from the movement of long bubbles in tubes:
with 1.6 wt% of Aerosol MA 80I (Sodium dihexyl
sulfosuccinate, CYTEC, NJ) added into 3 wt% NaCl and 0.01 Cf nf
µg = µg +
f o
........................................................ (5)
wt% CaCl2 brine solution. These experimental conditions are ug
1/ 3
identical to those of the "Type 3" experiments in Gauglitz et
al.31 (Fig. 2) except that the gas and surfactant solutions were where µgo is gas viscosity in the absence of foam and Cf is a
injected at fixed rates in this study. Steady-state pressure model parameter. Parameter Cf varies widely among
gradient was calculated from pressure drop across in the third theoretical and experimental studies in tubes, beadpacks and
pack section. rock.45 We make no explicit account in this model for gas
Fig. 5 is a plot of pressure gradient as a function of trapping or gas yield stress. Separating these two factors is
superficial velocities of gas (ug) and water (uw) in the strong- complicated, because the yield stress is also the origin of gas
foam state. In other words, Fig. 5 represents the ∇p(uw,ug) trapping.46 Different population-balance models treat these
surface formed by combining the upper branch of the curve in effects differently.9,13,16-20
Fig. 2 with upper branches of curves at other values of uw. In the absence of a simple way to represent the effects of
These new data reflect the same types of steady-state strong- gas trapping and yield stress, here we assume
foam regimes shown in Fig. 1.
o
kk rg
Population-Balance Model ug = ∇p ............................................................... (6)
µg
f
It is not clear exactly how the rate of lamella creation in
porous media depends on pressure gradient and other factors.
Kam and Rossen32 suggest the following simple relationship: where krgo is gas relative permeability in the absence of foam
and µgf is given by Eq. 5. Thus Cf (Eq. 5) accounts both for a
m
rg = C g (∇p ) ............................................................(2) constant reduction in gas relative permeability with foam and
for gas effective viscosity with foam.
where rg is the rate of lamella creation per unit volume of gas Transport of liquid is governed by Darcy's law as well (Eq.
phase, ∇p is the magnitude of pressure gradient in Pa/m, and 1). The liquid-phase relative-permeability function and liquid-
Cg and m are model parameters. The theory of Rossen and phase viscosity are both assumed to be unaffected by foam.
Gauglitz29 for foam generation implies that lamella creation Relative-permeability functions for gas and liquid phases
depends on pressure gradient, but also on water saturation or in the absence of foam were estimated by curve-fitting data for
capillary pressure, which governs the presence of lenses or unconsolidated sandpacks:47
lamellae available to be mobilized. Specifically, foam 2.2868
generation is easier at higher water saturation, because there ⎛ 1 − S w − S gr ⎞
= ⎜⎜ ⎟⎟
o
are more liquid lenses on the pore network, and these lenses k rg ....................................... (7)
can be mobilized at lower ∇p because of their arrangement on ⎝ 1 − S wc − S gr ⎠
the network.29 Therefore, here we have modified the rate of
lamella creation as follows: 1.9575
⎛ S w − S wc ⎞
k rw = 0.7888 ⎜ ⎟⎟ for Sw < 1 .............. (8)
rg = C g S w (∇p)
m
. ..................................................(3) ⎜1− S − S
⎝ wc gr ⎠
This relation is still not a complete description of all with Swc = 0.04 and Sgr = 0.
mechanisms of lamella creation, but does include the crucial This model has five adjustable parameters (Cg, m, Cc, n,
role of pressure gradient on lamella creation by mobilization. and Cf): two for lamella creation, two for lamella destruction,
Beyond that, our approach shares many features with and one for effective gas mobility as a function of bubble size.
previous population-balance models.9,13,16-20 Foam coalescence For modeling steady-state behavior, there are four independent
is governed by limiting capillary pressure, Pc*, and the parameters: (Cg,/Cc, m, n, and Cf). Only the ratio of Cg to Cc
corresponding water saturation, Sw*.22-24 No one knows how matters to steady-state properties.
the rate of coalescence (lamella destruction) should depend on
Sw or Pc, as long as it diverges toward infinity as Sw Model Fit to Steady-State Data
approaches Sw* or Pc approaches Pc*. We adopt the following The fitted values of model parameters are given in Table
expression: 1. This set does not represent a unique fit to the data or even
n necessarily an optimized fit. Steady-state behavior of the
⎛ 1 ⎞ model is determined by setting rc = rg. The model fit to the
rc = C c n f ⎜ ⎟ ...............................................(4)
⎜ S − S* ⎟ three foam states is shown in Fig. 2. The new model does a
⎝ w w ⎠ better job of fitting the near-plugging (low gas superficial
SPE 90938 5

velocity) in the intermediate state than the previous model fit of gravity effects. Gravity would have been insignificant in
of Kam and Rossen.32 cases where strong foam formed, but would have been an
The model fit to the steady-state low-quality and high- important factor with coarse foam (leading to override in a
quality strong-foam regimes is shown in Fig. 6 for comparison horizontal beadpack or to differences between ∇p in the liquid
with Fig. 5. Pressure gradient is nearly independent of gas and gas phases with vertical flow). In all cases gas and liquid
superficial velocity (nearly vertical ∇p contours) in the high- are co-injected as separate phases, not as foam; that is, the
quality regime, and nearly independent of liquid velocity injected foam texture nf is zero.
(nearly horizontal ∇p contours) in the low-quality regime. In
the context of this model, obtaining absolutely vertical Simulation Results
contours in the high-quality regime would require a steady- The parameter values in Table 1 apply to all simulations, and
state water saturation that does not change at all with all simulations assume a 0.3-m long porous medium
superficial velocity of gas: for instance, a steady-state (beadpack).
saturation constant and nearly equal to Sw*. In fact, over most
of this figure, Sw is only about 0.000002 higher than Sw* Stability of the Three Foam States
(0.0585 (Table 1)). (Sw* itself is so low because connate water By analogy to catastrophe theory,48 one would expect from
saturation is low in this high-permeability beadpack.) A the shape of the data and model fit in Fig. 2 that the strong-
steady-state Sw so close to the singularity in rc at Sw* (Eq. 4) and coarse-foam states would be stable to small perturbations,
gives nearly unchanging Sw and the desired vertical contours, while the intermediate state would be unstable. Simulation
but leads to numerical problems in simulations involving the results confirm this expectation, based on the response of
high-quality strong-foam regime, as described below. simulations initially at steady state to perturbations in foam
Within the low-quality strong-foam regime, nf = nfmax. texture nf. These simulations were conducted without
Making ∇p contours absolutely horizontal in this regime considering capillary-pressure gradients, to better reflect
would require completely de-coupling gas mobility from water stability on the field scale, where perturbation wavelengths
saturation, as suggested by Rossen and Wang.33 In this model, could be long and capillary-pressure gradients insignificant.
gas mobility depends weakly on water saturation through In each case injection rates and all properties throughout
krgo(Sw) (Eqs. 6 and 7). A value of 8 x 1013 is selected for nfmax the medium were set at the values corresponding to a given
for modeling and simulation, based on the characteristics of steady state, based on the fit illustrated in Figs. 2 and 6.
beadpack used in the experiments, i.e. bubble size roughly Values of these properties are given in Table 2. Then a
equal to pore size. Combined with Cf (Eq. 5), nf determines ∇p perturbation of ±50% in steady-state nf was imposed in the
in the low-quality strong-foam regime. middle of the pack, injection continued, and the response to
The model parameters in Table 1 and fits in Figs. 2, 5 and the perturbation was simulated. In additional cases not shown
6 represent the first population-balance model fit to both the here, we find that varying Cc and Cg (holding their ratio fixed)
three foam states and the high-quality and low-quality regimes does not affect the results of stability tests but does determine
in the strong-foam state. how rapidly perturbations diminish or grow. A variety of
For fitting steady-state data, only the ratio of the kinetic values of Cc are selected for stability tests shown below.
parameters Cg/Cc matters; as long as this ratio is fixed, the Fig. 7 shows how Sw and nf respond to a perturbation in nf
magnitudes of Cg and Cc do not matter. Therefore, to probe the for a foam in the coarse-foam regime. This condition
importance of the rates of kinetic processes of lamellae corresponds to point A in Fig. 2, with steady-state ∇p = 0.005
destruction and creation, we hold (Cg/Cc) fixed and vary Cc. psi/ft. Note that there is no initial perturbation in Sw, but Sw
The magnitude of Cg changes proportionately to Cc. does change in response to the perturbation in nf. The
perturbation shrinks with time as it moves downstream,
Numerical Simulation Approach however, indicating the stability of the coarse-foam state to
The numerical approach used for simulating dynamic this perturbation. Every case we examined in the coarse-foam
displacements is summarized in the Appendix. Briefly, we state showed similar stability. A sufficiently large
used a simple upstream-weighted forward-difference perturbation, of course, or starting from a state sufficiently
technique, similar to that of Rossen and Bruining,42 extended close to the limit of the coarse-foam state (where ug starts to
to a second dependent variable, nf. In these simulations, decrease upon increasing ∇p in Fig. 2) could in principle
surfactant concentration is uniform in initial resident and trigger formation of strong foam and a shift to that steady
injected fluids, so there is no need for a material balance on state.
surfactant. Our simulations assume incompressible phases. Fig. 8 shows how Sw and nf respond to a perturbation in nf
We were unable to simulate displacements in the high- for a foam in the low-quality strong-foam regime. The initial
quality regime with the model parameters in Table 1. As state is represented by point B in Fig. 6 and listed in Table 2
noted, steady-state water saturation in the high-quality regime (∇p = 4.2 psi/ft). nf is already at nfmax in this case, so only a
is extremely close to the singularity in rc at Sw*. This greatly negative perturbation in nf is possible. In this case as well, the
destabilized the numerical simulations, which tended to perturbation shrinks with time. Every case we examined in the
alternate in the foam bank between strong foam and complete low-quality, strong-foam regime showed similar stability.
foam collapse. We discuss this further below. Fig. 9 shows an example in the intermediate state, which is
In all cases shown, the simulations reflect 1D foam marked as point C in Fig. 2 (∇p = 0.5 psi/ft); see also Table 2.
displacement through a 0.3 m (~1 ft) beadpack, in the absence Here the perturbation in Sw grows with time as it moves
6 SPE 90938

downstream, and nf quickly attains its maximum value nfmax in corresponding to coarse foam at these injection rates of gas
the low-quality strong-foam state. In all cases in the and liquid. Within the coarse-foam bank gas viscosity is
intermediate state, we observed unstable behavior if the hardly altered by foam (µgf = 2.21 x 10-5 Pa s (0.00221 cp)),
perturbation was given sufficient time to grow. The reason, and ∇p is only 0.003 psi/ft.
evidently, is that, near the intermediate foam steady state, net The sharp, narrow peak in nf in Fig. 10 raises concerns
foam generation rate (rg - rc) increases upon a perturbation about grid effects. Fig. 11 shows the effect of grid refinement,
increasing the value of nf. This would eventually drive the still neglecting capillary-pressure gradients. The peak in nf
perturbed system either to the coarse-foam or strong-foam reaches virtually the same value as with the coarser grid, but
steady state. Moreover, we never observed the intermediate the peak is narrower. Gas is most viscous within the
foam state in dynamic simulations of fixed-rate gas and liquid displacement front, which makes the fall in Sw at the front
injection, except when we chose that state as the initial sharper than one would expect from foam viscosity in the
condition of the medium. Displacements did not come to this coarse-foam bank, especially with grid refinement. The
state spontaneously. displacement front is sharper than one would predict from
fractional-flow theory assuming a constant gas viscosity at its
Dynamic Displacements value in the coarse-foam bank, because gas viscosity is higher
The steady-state behavior of the population-balance model within the displacement front. In some cases with coarse foam,
indicates that at sufficiently low injection rates one ought to illustrated below, foam at the displacement front is viscous
reach a coarse-foam state throughout the medium, in enough to drive Sw below its steady-state value, and Sw then
agreement with experimental studies of foam rebounds to its higher, steady-state value behind the front.
generation.13,26,29-32 Starting from the coarse-foam state, as Figs. 12 and 13 show the corresponding cases to Figs. 10
injection rates increase, either in separate experiments or in and 11, accounting now for capillary-pressure gradients. The
one continuing experiment, there should be an abrupt change peak in nf is broader, though nf reaches virtually the same
to a strong-foam steady state with much lower total mobility - value at the peak, and grid refinement has relatively little
i.e., "foam generation." In steady-state analysis, this jump effect on behavior.
occurs as superficial velocity first starts to decrease with
increasing pressure gradient (Fig. 2). This point represents the Strong-Foam State. Fig. 14 shows a case at the same foam
limit of intrinsic stability of the coarse-foam regime; an quality as Fig. 10, with injection rate 8.15 times higher. Note
infinitesimal perturbation there would convert coarse foam to that the changes in injection rate at fixed foam quality in Figs.
strong foam. How sudden or abrupt this change might be in 10 ff. can be represented by the dashed line in Fig. 6. In the
real time in a coreflood experiment would depend on kinetic laboratory, increasing injection rate promotes generation of
effects excluded from the steady-state model (the magnitudes strong foam.29 At these conditions the steady-state analysis
of Cg and Cc, as distinct from their ratio). Steady-state analysis predicts that only the strong-foam state can exist. There is an
of nearly the same model as used here32 indicates that the extremely sharp front at the leading edge of the displacement.
jump from coarse foam to strong foam occurs at lower Bubble are convected forward at the maximum texture nfmax;
injection rate the higher is the liquid fraction of injected fluids. there can be no further generation at the front, even though rg
All these trends in the population-balance model at steady > rc according to Eqs. 3 and 4, because nf is already at its
state are consistent with laboratory observations. Simulations maximum value. Of course there can also be no overshoot in
confirm all these trends. the value of nf at the displacement front, as in Figs. 10 to 13,
for the same reason. Behind the displacement front, nf = nfmax,
Coarse-Foam State. Fig. 10 shows a simulation of co- which sets the value of Sw some distance above Sw* (Sw =
injection of gas and liquid into a medium saturated with 0.075 v. Sw* = 0.0585). The distance between Sw in the foam
surfactant solution, conditions where the steady-state analysis bank and Sw* in this case helps the numerical stability of the
predicts a coarse-foam steady state (cf. Fig. 2). In the simulation.
simulation of Fig. 10 there are 25 grid blocks, and capillary- Including capillary-pressure gradients in the calculations
pressure gradients are neglected. The injection conditions are made no difference to these results for strong foam with high
consistent with point D in Fig. 2 (uw = 1.0157 x 10-6 m/s ∇p.
(0.288 ft/d); ug = 2.571 x 10-6 m/s (0.729 ft/d); foam quality Fig. 14 shows no entrance region within which strong
=72% (cf. dashed line in Fig. 6); ∇p = 0.003 psi/ft) as steady- foam is created; Evidently strong foam is created almost
state coarse foam. At the leading edge of the displacement instantaneously from injected liquid and gas in the first grid
front, there is a steep rise in foam texture nf. Bubbles are block. Various studies suggest that this region should be of
convected forward to the front, and additional lamellae are order a few inches.13,35 The lack of an entrance region
created there by pressure gradient as well (Eq. 2). The highest therefore suggests that the values of kinetic parameter Cc and
∇p in this case occurs at the leading edge of the foam bank Cg are too large in this simulation. The effects of Cc and Cg are
where nf is higher and ug small (Eq. 6). Both this and the Sw- discussed further below.
dependence of generation (Eq. 2) promote lamella creation at
the displacement front. Meanwhile, lamella destruction rate is Transition from Coarse Foam to Strong Foam. Figs. 15
low at the front because Sw is high, far from Sw* (Eq. 4). Once to 19 show a sequence of simulations with increasing total
nf rises and Sw falls, however, lamella destruction overtakes injection rate, each time into a pack saturated with liquid. At
creation and the net rate of lamella creation (rg - rc) falls below this foam quality (72%), steady-state analysis predicts that the
zero. Foam texture falls back to its steady-state texture
SPE 90938 7

coarse-foam state becomes intrinsically unstable at an potentially adverse effect on foam generation at higher
injection rate a little more than 6.03 times that in Fig. 10. injection rates. Much-larger values of the kinetic parameters
In Fig. 15, injection rate is 1.69 times that in Fig. 10; the (Fig. 21), however, leave the peak in nf much as in Fig. 10,
peak in nf is higher than in Fig. 10, but coarse foam still and suggest that larger values of kinetic parameters would not
results, though Sw is a little lower in the foam bank than in alter the onset of strong-foam generation.
Fig. 10 (as expected for steady-state strong foam at this Fig. 22 shows the effect kinetic parameters on
injection rate and quality). In Fig. 16, with injection rate 2.86 displacements with strong foam: specifically, the simulation of
times that in Fig. 10, the peak in foam texture at the front Fig. 14 with Cc and Cg both 10 times smaller than in Fig. 14.
increases further. In Fig. 17, with injection rate 4.83 times that Now there is an entrance region roughly corresponding to that
in Fig. 10, gas viscosity in the front is so high that it drives Sw one might observe in the laboratory. With both parameters
below the steady-state coarse-foam value; but as foam another factor of 10 smaller (not shown), the entire 0.3-m pack
degrades behind the front Sw rebounds to the steady-state lies in the entrance region. Thus the length of the entrance
coarse-foam value corresponding to this injection rate. region offers one means of calibrating the magnitude of
In Fig. 18, with injection rate another 5.79 times that in kinetic parameters. It is noteworthy that the entrance region
Fig. 10, the peak value of nf grows as the front moves has foam texture much coarser than strong foam, and the
downstream; in a separate simulation (not shown) for a 0.6-m transition to strong foam in the simulation is quite abrupt,
pack, strong foam results before the gas bank reaches the evidently triggered by a sufficient local pressure gradient.
outlet. Thus strong foam is created in a dynamic displacement Although the entrance region where foam is created is
at an injection rate a little lower than that predicted from extended with smaller values of Cc and Cg, the properties and
steady-state analysis. In a laboratory experiment, an injection mobility of the foam bank that eventually forms are not altered
rate near that in Fig. 18 would correspond to the critical from that predicted by steady-state theory. The leading edge of
velocity for foam generation at this foam quality. In laboratory the displacement front is still sharp, because this front is not a
experiments, near the threshold for generation of strong foam, front of foam generation - it is a front governed by convection
duplicate experiments give different results.29 We find that the of bubbles. Therefore the width of the front is not sensitive to
simulation results are also sensitive to small changes in the kinetic parameters in the model.
assumptions, discretization scheme or calculation method near Fig. 22 suggests a principle: If strong foam is created in
the threshold injection rate for foam generation. the medium from co-injected gas and liquid, the magnitude of
In Fig. 19, with injection rate 6.26 times that in Fig. 10, the kinetic parameters in the population balance have
strong foam results soon after foam injection. This is expected, relatively little effect on the displacement except in a small
since at this injection rate the coarse-foam state is intrinsically region near the injection face. Most of the displacement could
unstable. be predicted by local-steady-state modeling, if one knows
When capillary-pressure gradients are accounted for (not strong foam will be created.
shown), at the higher injection rates of Figs. 17 and 18, the The effect of smaller kinetic parameters in Fig. 20, with
peak in nf at the displacement front is higher and broader than coarse foam, however, suggests that near the threshold for
with Pc gradients neglected. The formation of strong foam still generation of strong foam, kinetic parameters can affect the
occurs between the injection rates of Figs. 18 and 19, magnitude of the peak in nf and thereby possibly whether
however. Including capillary-pressure gradients did not strong foam is created.
significantly alter the critical velocity for foam generation. Kinetic parameters may play a more important role when
We also simulated one long period of co-injection at foam is created in situ by alternate injection of gas and liquid
stepwise-increasing injection rates. This case more closely ("SAG" injection).39 In that case, a much weaker foam may be
matches the experimental protocol of Rossen and Gauglitz.29 present behind the displacement front, with convection of
In this case the porous medium is already partially drained by bubbles to the front playing a smaller role than in Fig. 22,
gas before injection rate increases in each step. As in Figs. 15 where there is a strong-foam bank behind the front. In that
to 19, strong foam was created when injection rate was raised case the kinetics of foam creation within the foam front may
from 4.91 ft/d (cf. Fig. 17) to 6.40 ft/d (cf. Fig. 19). In other be crucial. Research on this topic is ongoing.
words, in the simulations, the injection rates required to create
strong foam in a pack initially saturated with liquid was fairly High-Quality Strong-Foam Regime. As noted, we were
close to that for one already partially drained with gas. unable to simulate displacements in the high-quality strong-
In a separate series of simulations, not shown here, we foam regime. For the given parameters, steady-state analysis
held gas injection rate fixed and increased liquid injection rate indicates that Sw in the high-quality regime is extraordinarily
in a series of displacements. Foam generation occurred at a close to Sw* (e.g., Sw = 0.058502 at steady state, compared to
lower total injection rate than in Figs. 15-19 at the lower foam Sw* = 0.058500). It is extremely difficult to avoid reaching the
quality, in agreement with Rossen and Gauglitz.29 singularity in rc at Sw* (Eq. 4), and complete foam collapse, in
this case. Simulations would require extraordinarily short time
Sensitivity to Kinetic Parameters. Figs. 20 and 21 steps to avoid driving Sw below Sw*. Moreover, the singularity
illustrate the effect of kinetic parameters on displacements in in rc causes wild fluctuations in nf (from nf = 0 to nf = nfmax in
the coarse-foam state: specifically, the simulations of Fig. 10 adjacent grid blocks) even if Sw stays above Sw*.
with both Cc and Cg 100 times smaller and larger than in Fig. We believe a singularity in rc, or something approximating
10, respectively. With much smaller kinetic parameters (Fig. a singularity, is real; this singularity, and the abrupt change in
20), the peak in nf is lower than in Fig. 10, suggesting a foam properties it produces, lies at the heart of most foam
8 SPE 90938

models over the last decade5,9,3,16-20,36-40 and has ample indirect 6. The model described here is not yet predictive, and the
experimental support.22,35 It may not be necessary in general parameter set fit to the data is not necessarily unique.
that Sw be as close to Sw* as in our model fit to fit the high- Moreover, we had difficulty simulating displacements in
quality regime, however. Cheng49 found with a local-steady- the high-quality strong-foam regime, because the steady-
state foam model that if foam collapses over a range in Sw of state water saturation with foam is extraordinarily near the
no more than a about 0.002, one observes ∇p virtually singularity in lamella destruction rate at Sw*. Research on
independent of gas superficial velocity ug in the high-quality these issues is continuing.
strong-foam regime. Previous local-steady-state simulations in
our group used a wider range of Sw over which foam Nomenclature
collapses, 0.02,37,49 to speed up simulation times. Larger SI units are assumed in all equations in the text, but customary
differences between steady-state Sw and Sw* should make it units are cited in the text and figures.
easier to simulate foam in the high-quality strong-foam Cc = model parameter (Eq. 4)
regime. In addition, more-sophisticated numerical techniques Cf = model parameter (Eq. 5)
may allow more efficient convergence on the steady state in Cg = model parameters (Eq. 3)
the foam bank for high-quality-regime foams. fg = gas fractional flow (gas injected volume
fraction), usually expressed as %
Conclusions fw = water fractional flow (injected liquid volume
1. For the first time, a population-balance model is fit to fraction), usually expressed as a fraction < 1
experimental data for both the three foam states (coarse G = mobility term defined by Eq. A3
foam, intermediate state, and strong foam) and the two k = permeability, units m2 in calculations but
strong-foam regimes (low-quality and high-quality). reported in text as darcy
2. In dynamic displacements, the population-balance model krgo = gas relative permeability in absence of foam
shows a transition from coarse foam to strong foam as krw = water relative permeability
injection rates increase at fixed foam quality, or as liquid m = model parameter (Eq. 3)
fraction of injected fluids increases at fixed gas injection n = model parameter (Eq. 4)
rate, in agreement with previous laboratory studies. In nf = foam texture or density (inversely related to
dynamic displacements, foam generation occurred at an bubble size); lamellae/unit volume of gas (m-3)
injection rate a little lower than that predicted by steady- nfmax = foam texture in the low-quality regime; maximum
state analysis. This appears to reflect complex dynamics at attainable value of nf
the gas front. N = total number of grid blocks in finite-difference
3. In steady co-injection of gas and surfactant solution, if a simulation
bank of low-quality-regime strong foam forms, the kinetics Pc = capillary pressure (Pa)
of foam generation and destruction affect the length of the Pc* = limiting capillary pressure (Pa)
entrance region in which foam forms. Thus the length of ∇p = magnitude of pressure gradient (Pa/m)
the entrance region can be used to calibrate the kinetic R = net rate of lamella creation per unit volume of
parameters in the model. The displacement front is porous medium
practically unaffected by the magnitude of the kinetic rc = rate of lamella destruction per unit volume of gas
parameters, though. The displacement front, and the bank phase (Eq. 4)
behind it, are essentially what one would have predicted rg = rate of lamella creation per unit volume of gas
from local-steady-state principles, if one knows which phase (Eq. 3)
steady state obtains in the foam bank. Local-steady-state Sgr = residual gas saturation
modeling would fit the displacement on the large scale. Sw = water saturation
4. When coarse foams are created, there is a narrow region of Swc = connate water saturation
finer foam predicted near the gas injection front. This Sw* = water saturation at limiting capillary pressure
phenomenon has two important implications. First, it is Pc* (Eq. 4)
possible that this transient peak in texture near the ∆t = time increment in finite-difference simulation
displacement front can perturb the system from one state to ug = gas volumetric flux or superficial velocity (m/s)
another, i.e. trigger "foam generation." These dynamics did ut =total volumetric flux or superficial velocity (m/s),
not greatly alter the threshold injection rate for foam i.e. (ug+uw)
generation in this study, however. Second, gas viscosity uw = water volumetric flux or superficial velocity (m/s)
varies within the displacement front, and this affects the Vg = gas interstitial velocity,(ug/φ) (Fig. 4)
change in water saturation across the front. Fronts can be Vw = liquid interstitial velocity,(uw/φ) (Fig. 4)
sharper than estimated from fractional-flow theory ∆x = grid-block size in finite-difference simulation
assuming a constant gas viscosity at the steady-state value
behind the displacement front. Greek Symbols
5. Simulation results confirm the stability of the coarse-foam φ = porosity
and strong-foam states to small perturbations, and the µgf = effective gas viscosity in presence of foam (Eq. 5)
instability of the intermediate state. In the intermediate (Pa s)
state, a perturbation in foam texture grows with time.
µgo = gas viscosity in absence of foam (Pa s)
SPE 90938 9

µw = water viscosity 13. Friedmann, F., Chen, W. H., and Gauglitz, P.A.: "Experimental
σ = gas-liquid surface tension and Simulation Study of High-Temperature Foam Displacement
in Porous Media," SPERE (Feb. 1991) 37-45.
14. Hirasaki, G. J., and Lawson, J. B.: "Mechanisms of Foam Flow
Subscripts and Superscripts Through Porous Media - Apparent Viscosity in Smooth
ι = grid block Capillaries," SPEJ (April 1985), 176-190.
t+∆t
= updated value at next time step 15. Falls, A. H., Musters, J. J., and Ratulowski, J.: "The Apparent
J
= property of injected fluids Viscosity of Foams in Homogeneous Bead Packs," SPERE.
(May. 1989) 155.
Acknowledgments 16. Falls, A. H., Hirasaki, G. J., Patzek, T. W., Gauglitz, P. A.,
This work was conducted with the support of the National Miller, D. D., and Ratulowski, J: "Development of a
Mechanistic Foam Simulator: The Population Balance and
Petroleum Technology Office of the U.S. Department of
Generation by Snap-Off," SPERE (Aug. 1988) 884-892.
Energy, through contract DE-FC26-01BC15318 and also with 17. Kovscek, A. R., Patzek, T. W., and Radke, C. J.: "A
a grant from Schlumberger, Inc., and support from the Small Mechanistic Population Balance Model for Transient and
Research Grant Scheme of the University of Adelaide. Steady-State Foam Flow in Boise Sandstone," Chem. Eng. Sci.
50, 3783-3799 (1995).
References 18. Kovscek, A. R., Patzek, T. W., and Radke, C. J.: "Mechanistic
1. Schramm, L. L. (ed.): Foams: Fundamentals and Applications Foam Flow Simulation in Heterogeneous and Multidimensional
in the Petroleum Industry, ACS Advances in Chemistry Series Porous Media," SPE J. (Dec. 1997), 511-526.
No. 242, Am. Chem. Soc., Washington, DC (1994). 19. Myers, T. J. and Radke, C. J.: "Transient Foam Displacement in
2. Rossen, W. R.: "Foams in Enhanced Oil Recovery," in Foams: the Presence of Residual Oil: Experiment and Simulation Using
Theory, Measurements and Applications, R. K. Prud'homme and a Population-Balance Model," Ind. Eng. Chem. Res. 39, 2725-
S. Khan (eds.). Marcel Dekker, New York (1996). 2741 (2000).
3. Terdre, N.: "Foam-Assisted Injection Trials Could Spread to 20. Bertin, H. J., Quintard, M. Y., and Castanier, L. M.:
Other North Sea Fields," Offshore 63 (Aug. 2003), 70. "Development of a Bubble-Population Correlation for Foam-
4. Gdanski, R. D.: "Experience and Research Show Best Designs Flow Modeling in Porous Media," SPE J (Dec. 1998), 356-362.
for Foam-Diverted Acidizing," Oil and Gas Journal (Sept. 6, 21. Gillis, J. V., Radke, C. J.: “A Dual Gas Tracer Technique for
1993), 85. Determining Trapped Gas Saturation During Steady Foam Flow
5. Cheng, L., Kam, S. I., Delshad, M., and Rossen, W. R.: in Porous Media,” paper SPE 20519 presented at the 1990 SPE
"Simulation of Dynamic Foam-Acid Diversion Processes," SPE Annual Technical Conference and Exhibition, New Orleans,
J. (Sept. 2002), 316-324. Sept. 23-26.
6. Nguyen, Q. P., Currie, P. K., and Zitha, P. L. J.: "Determination 22. Khatib, Z. I., Hirasaki, G. J., and Falls, A. H.: "Effects of
of Foam Induced Fluid Partitioning in Porous Media Using X- Capillary Pressure on Coalescence and Phase Mobilities in
ray Computed Tomography," presented at the 2003 SPE Foams Flowing Through Porous Media," SPERE (Aug. 1988)
International Symposium on Oilfield Chemistry held in 919-926.
Houston, Texas, U.S.A., 5-8 February. 23. Rossen, W. R., and Zhou, Z. H.: "Modeling Foam Mobility at
7. Hirasaki, G. J., Jackson, R. E., Jin, M., Lawson, J. B., the Limiting Capillary Pressure," SPE Adv. Technol., 3, 146
Londergan, J., Meinardus, H., Miller, C. A., Pope, G. A., (1995).
Szafranski, R., and Tanzil, D.: "Field Demonstration of the 24. Aronson, A. S., Bergeron, V., Fagan, M. E., and Radke, C. J.:
Surfactant/Foam Process for Remediation of a Heterogeneous "The Influence of Disjoining Pressure on Foam Stability and
Aquifer Contaminated with DNAPL," in NAPL Removal: Flow in Porous Media," Colloids Surfaces A: Physicochem.
Surfactants, Foams, and Microemulsions, S. Fiorenza, C. A. Eng. Aspects 83 109, (1994).
Miller, C. L. Oubre, and C. H. Ward, (eds.), Lewis Publishers, 25. Alvarez, J. M., Rivas, H., and Rossen, W. R.: "A Unified Model
Boca Raton (2000). for Steady-State Foam Behavior at High and Low Foam
8. Mamun, C. K., Rong. J. G., Kam, S. I., Liljestrand, H. M., and Qualities," SPE J (Sept. 2001), 325-333.
Rossen, W. R.: "Extending Foam Technology from Improved 26. Ransohoff, T. C., and Radke, C. J.: "Mechanisms of Foam
Oil Recovery to Environmental Remediation, paper SPE 77557 Generation in Glass-Bead Packs," SPERE (May 1988), 573-585.
presented at the 2002 SPE Annual Technical Conference, San 27. Chambers, K. T., and Radke, C. J.: "Capillary Phenomena in
Antonio, TX, 29 Sept.-2 Oct. Foam Flow Through Porous Media," in Interfacial Phenomena
9. Kovscek, A. R., and Radke, C. J.: “Fundamentals of Foam in Oil Recovery, N. R. Morrow, (ed.), Marcel Dekker, New
Transport in Porous Media,” in Foams: Fundamentals and York, 1990.
Applications in the Petroleum Industry, L. L. Schramm (ed.), 28. Chen, M., Yortsos, Y. C., and Rossen, W. R.: "Pore-Network
ACS Advances in Chemistry Series No. 242, Am. Chem. Soc., Study of Mechanisms of Foam Generation," SPE 90939,
Washington, DC (1994). presented at the 2004 SPE Annual Technical Conference and
10. Bernard, G. G., Holm, L. W., and Jacobs, W. L.: "Effect of Exhibition, Houston, TX, 26-29 Sept. 2004.
Foam on Trapped Gas Saturation and on Permeability of Porous 29. Rossen, W. R., and Gauglitz, P. A.: "Percolation Theory of
Media to Water," SPE J. (Dec. 1965), 195-300. Creation and Mobilization of Foam in Porous Media," AIChE J.
11. Sanchez, J. M., and Schechter, R. S.: "Surfactant Effects on the 36 (1990), 1176-1188.
Two-Phase Flow of Steam-Water and Nitrogen-Water Through 30. Tanzil, D., Hirasaki, G. J., and Miller, C. A.: “Conditions for
Permeable Media," J. Petr. Sci. Eng. 3, 185-199, (1989). Foam Generation in Homogeneous Porous Media,” paper SPE
12. de Vries, A. S. and Wit, K.: "Rheology of Gas/Water Foam in 75176 presented at the 2002 SPE/DOE Symposium on
the Quality Range Relevant to Steam Foam," SPERE (May Improved Oil Recovery, Tulsa, OK, 13-17 April.
1990) 185-192. 31. Gauglitz, P. A., Friedmann, F., Kam, S. I., and Rossen, W. R.:
“Foam Generation in Homogeneous Porous Media,” Chem.
Eng. Sci. 57 (2002), 4037-4052.
10 SPE 90938

32. Kam, S. I., and Rossen, W. R.: "A Model for Foam Generation surfactant. Here we use a simple forward-difference,
in Homogeneous Porous Media," SPE Journal 8 (Dec. 2003), upstream-weighted finite-difference solution for water
417-425. saturation Sw and foam texture nf. The continuous position
33. Rossen, W.R. and Wang, M.-W.: "Modeling Foams for Acid variable x is discretized into N grid blocks i (25, 50, 100 or
Diversion," SPEJ (June 1999), 92-100.
34. Kim. J. S., Dong, Y., and Rossen, W. R.: "Steady-State Flow 200 in cases shown below) of length ∆x = L/N, where L is the
Behavior of CO2 Foam," presented at the 2004 SPE/DOE length of the porous medium (0.3 m in all cases here).
Symposium on Improved Oil Recovery, Tulsa, OK, 17–21 April
2004. Mass Balance on Water
35. Ettinger, R. A., and Radke, C. J.: "The Influence of Texture on Values of water saturation are updated first. The mass
Steady Foam Flow in Berea Sandstone," SPERE (Feb. 1992) 83- balance on water
90.
36. Rossen, W. R., "A Critical Review of Roof Snap-Off as a ∂S w ∂u w
φ + = 0 .........................................................(A1)
Mechanism of Steady-State Foam Generation in Homogeneous ∂t ∂x
Porous Media," Colloids Surfaces A: Physicochem Eng.
Aspects, 225 1-24 (2003). is discretized as follows. Given N values Sw,i and nf,i at time t
37. Rossen, W.R., Zeilinger, S.C., Shi, J.-X., and Lim, M. T.: (along with all other local properties at t), each value of Sw,i is
"Simplified Mechanistic Simulation of Foam Processes in updated at time (t + ∆t) as follows:
Porous Media," SPE J, 4, 279-287 (Sept. 1999).
38. Cheng, L., Reme, A. B., Shan, D., Coombe, D. A., and Rossen,
W. R., "Simulating Foam Processes at High and Low Foam
S
t +∆t
w,i
=S
u ∆t
+ t
w,i φ∆x ( ) ( )
⎡f S
⎢⎣ w w,i−1
− fw S

w,i ⎥⎦

⎪⎪ ( ) ⎡⎢⎣ ( ) ( )⎤⎥⎦ ⎪⎫⎪ ..(A2)


Qualities," paper SPE 59287 presented at the 2000 SPE/DOE
Symposium on Improved Oil Recovery, Tulsa, OK, 3-5 April. ⎧
G S Pc S − Pc S
39. Shan, D., and Rossen, W. R.: “Optimal Injection Strategies for ∆t w,i w,i w,i −1
⎨ ⎬
( ) ⎢⎣ ( ) ( )⎥⎦⎭⎪
Foam IOR,” paper SPE 75180 presented at the 2002 SPE/DOE +
φ (∆x)2 ⎪ ⎡ ⎤⎪
Symposium on Improved Oil Recovery, Tulsa, OK, 13-17 April; −G S Pc S − Pc S
SPE J., in press. ⎩⎪ w,i+1 w,i +1 w,i
40. Shrivastava, V. K., Belgrave, J. D. M., Singhal, A. K., and
Coombe, D. A., "Numerical Simulation of Foam Flooding for with
Sweep Improvement," J. Can. Pet. Tech., 38 special issue #13 λw
(1999), p45. fw (Sw ) ≡
41. Vassenden, F., Holt, T., Ghaderi, A., and Solheim, A.: "Foam λw + λ g
Propagation on Semi-Reservoir Scale," SPEREE (Oct. 1999),
λwλ g . ..........................(A3)
436-441. G(Sw ) ≡
42. Rossen, W. R., and Bruining, J.: "Foam Displacements With λw + λ g
Multiple Steady States," presented at the 2004 SPE/DOE
−0.2
0.5 ⎛ ⎞
Symposium on Improved Oil Recovery, Tulsa, OK, 17–21 April
S w − S wc
2004. Pc = σ (φ k ) ⎜⎜ ⎟⎟
43. Martinsen, H. A., and Vassenden, F.: “Foam-Assisted Water ⎝ 1 − S wc − S gr ⎠
Alternating Gas (FAWAG) Process on Snorre”, presented at the
10th European IOR Symposium, Brighton, UK, Aug. 18-20, All terms in Eq. A2 are evaluated at time t, except for Sw,i t+∆t
1999. on the left-hand side. When capillary-pressure gradients are
44. Rossen, W. R., and Z. H. Zhou: "Applying Fractional-Flow
neglected, σ is set to zero in Eq. A3.
Theory to Foam Processes at the 'Limiting Capillary Pressure',"
SPE Adv. Technol., 3, 154 (1995). For grid block 1 at the inlet, the injected fractional flow fwJ
45. Xu, Q., and Rossen, W. R.: "Effective Viscosity of Foam in is substituted for the fractional flow in the upstream grid
Periodically Constricted Tubes," submitted to Colloids and block, and, in effect, the capillary-pressure gradient upstream
Surfaces A: Physicochem. Eng. Aspects. is assumed to be zero:

( )⎤⎥⎦
46. Rossen, W.R.: "Rheology of Foam in Porous Media at the
t +∆t u ∆t ⎡ J
'Limiting Capillary Pressure'," Revue de l'Institut Français du S =S + t f − fw S
Pétrole 47, 68-80 (1992). w,1 w,1 φ∆x ⎢⎣ w w,1
. ...(A4)

{ ( ) ⎣⎢ ( ) ( )⎦⎥}
47. Collins, R. E.: Flow of Fluids Through Porous Materials,
Research & Engineering Consultants, Inc., Englewood, CO +
∆t
−G S
⎡ ⎤Pc S − Pc S
(1961). φ (∆x)2 w,2 w,2 w,1
48. Poston, T. and Stewart, I.: Catastrophe Theory and Its
Applications, Dover Publications, Inc., Mineola, (1978). For the Nth grid block at the outlet, two adjustments are
49. Cheng, L.: “Simulation Studies of Foam-Acid Diversion,” Ph.D. made to satisfy the capillary end effect: a) a capillary pressure
dissertation, The University of Texas at Austin, 2002. of zero is assumed downstream at the outlet, a distance (∆x/2)
from the center of the grid block; and b) the liquid flux out of
Appendix: Discretized Equations and Numerical the outlet uw,N is computed first (assuming zero Pc at the
Techniques outlet), and if this flux would be less than zero, indicating
A population-balance simulation for incompressible foam in liquid flow into the porous medium from the outlet, uw,N is set
porous media requires a material-balance equation for water to zero instead:
(or gas) and a population balance for lamellae.9,13,16-20 No mass
balance is needed for surfactant here, because here we model
cases in which all liquids have a uniform concentration of
SPE 90938 11

( ) m
( ) ( )
⎡ 0− Pc S w, N ⎤ ⎛ 1 ⎡P ⎤ ⎞
⎜ 2 ⎢⎣ w, N −1 + Pw, N ⎥⎦ − Poutlet
⎣ ⎦
u w , N = f w S w, N u t + G S w, N ⎟
∆x / 2 rg , N = C g S w, N ⎜ ⎟ . (A12)
⎜ ∆x ⎟
n n ⎝ ⎠
if u w, N < 0 , then u w, N = 0 ................... (A5)
For the purpose of pressure-gradient calculations, water
pressure is calculated as follows:
t +∆t u ∆t
Sw,N = Sw,N + t fw Sw, N −1
φ∆x ( ) uw,i µ w
Pw,i = P + ∆x .......................................(A13)
+
∆t
φ (∆x)2
( ⎡

) ( ) ( ⎤
G Sw,N ⎢Pc Sw, N −Pc Sw, N −1 ⎥ . .. (A6)
⎦ )
w,i +1 kk
rw,i
with the inlet pressure given by
∆t
− u
φ∆x w,N f J
J ∆x µ g u w
Pw = Pw,1 + . .........................................(A14)
Once all N values of Sw,i are computed, two checks are 2 kk
conducted: if the change in any Sw,i exceeds a tolerance value, rw,1
or any value of Sw,i has fallen below Sw*, the time step ∆t is Gas pressure is given by
reduced, and all Sw are recomputed for the smaller time step.
The process is repeated until the two conditions are satisfied. Pg = Pc + Pw . ..........................................................(A15)

After the N updated values of nf,i are computed, they are


Population Balance on Lamellae
The population-balance equation9,13,16-20 checked to verify that none have fallen below zero in that time
step. If they have, the time-step size is reduced, and the
φ

∂t(Sg n +
f

) ( ) (
n u g = φ S g rg − rc ................. (A7)
∂x f
) algorithm returns to calculating the set of Sw,i, followed by the
set of nf,i.
Once the set of Sw,i and nf,i are set for the new time step,
is discretized as follows: one must iterate for gas viscosity and local superficial
t +∆t velocities. Gas viscosity depends on gas superficial velocity,
⎡S n ⎤
⎢⎣ g f ⎥⎦i ⎢⎣ f ⎥⎦i (
= ⎡ S g n ⎤ + ∆t S g ⎡⎣ rg − rc ⎤⎦
i ) while gas superficial velocity depends on gas viscosity. An
iteration loop is used to resolve this interdependence for each
.......... (A8)
∆t ⎛ ⎡ ⎤ ⎡ ⎤ ⎞ grid block. At the end of this iteration loop, all properties are
+ ⎜ n u − n u ⎟
φ∆x ⎝ ⎢⎣ f g ⎥⎦i −1 ⎢⎣ f g ⎥⎦i ⎠ determined for this new time step. The algorithm then
increments time again and proceeds as above.
with the lamella-creation and -destruction functions given by
Eqs. 3 and 4. Note that Sg,i=(1-Sw,i) on the left-hand side of Eq. Table 1. Model Parameters Fit to Data in
A8 uses the updated value of Sw,i, while all terms on the right-
Figs. 2 and 6
hand side refer to time t. For Eq. 3, the pressure gradient is
Foam parameters Other parameters
discretized as follows:
Cg/ Cc 15,488.894 S w* 0.0585
m k (m2) 30.4x10-12
⎛P −P ⎞ m 2.4
w,i −1 w,i +1
rg ,i = C g S w,i ⎜ ⎟ . ........................ (A9) n 0.28 φ 0.31
⎜ 2∆x ⎟ Cf 1.007x10-16 µw (Pa s) 0.001
⎝ ⎠
µgo (Pa s) 0.00002
Foam texture injected into the first grid block is zero: Swc 0.04
t +∆t
⎡S n ⎤
⎢⎣ g f ⎥⎦1 ⎢⎣ f ⎥⎦1 (
= ⎡ S g n ⎤ + ∆t S g ⎡⎣ rg − rc ⎤⎦
1 ) Sgr 0.0
........ (A10)
∆t ⎛ ⎡n u ⎤ ⎞ Table 2. Steady-State Properties Predicted
+ ⎜ 0 − ⎟
φ∆x ⎝ ⎣⎢ f g ⎦⎥1 ⎠ by the Model.
Coarse foam Intermediate Strong foam
with (Fig.7) foam (Fig. 9) (Fig. 8)
m uw (m/s) = 1.0157x10-6 1.0157x10-6 1.0157x10-5
⎛ P J − 1 ⎡P + P ⎤ ⎞ ug (m/s) = 1.742x10-5 2.836 x10-5 5.324x10-6
2 ⎣ w,1 w,2 ⎦ ⎟
rg ,1 = C g S w,1 ⎜
w
. .......... (A11) nf (1/m3) = 7.013x108 3.158x1012 8x1013(nfmax)
⎜ ∆x ⎟
⎝ ⎠ ∇p (psi/ft) = 0.005 0.5 4.2
Sw = 0.625 0.09564 0.10083
For the Nth grid block, Eq. A8 applies, with µgf (Pa s) = 0.0000227 0.0104 0.461
12 SPE 90938

Fig. 3. Transient foam-texture profiles from a population-balance


simulation of a laboratory coreflood, from Ref. (18).
Fig. 1. Pressure gradient (psi/ft) as a function of gas and water
superficial velocities ug and uw, from study of strong foam in
Berea sandstone, from Ref. (25). The high-quality regime 140
corresponds to the upper-left portion of the plot, where ∇p is
nearly independent of ug; the low-quality regime corresponds to
the lower-right portion of the plot, where ∇p is nearly independent 120 o
of uw. Dark points represent data. oo

100 20o
Model fit
a) 16o
5 strong foam Experiment
80 12
o
4.5
strong foam oo
Pressure Gradient, psi/ft

4 o
o
3.5 60 oo
o o
3 o
intermediate state o o
2.5 40 o
o o o o o o

2 oo
o 20
oo o o o o
o
1.5 20 o
16
1 coarse foam o
o 12
o o o o o o
0.5
0
0
0 0.0001 0.0002 0.0003
0 5 10 15 20 25 30 35 40
Interstitial Liquid Velocity (Vw), ft/day
Superficial Gas Velocity (Ug), m/s
Fig. 4. Pressure gradient (psi/ft) as a function of gas and water
interstitial velocities Vg and Vw predicted by model of Kam and
32
Model fit Rossen without fitting model directly to data for the two
b) Experiment regimes. Open circles represent calculated values, not data.
10 strong foam
Pressure Gradient, psi/ft

generation”

70
1 C
“foam

gas superficial velocity (ug) x106, m/s

0.000001 0.00001 0.0001 0.001


intermediate 60
0.1 78
state
50 6
9 10
0.01 A
D coarse foam 40
45
0.001
30
Superficial Gas Velocity (Ug), m/s 9
20 8
7
Fig. 2. Model fit (smooth curve) to data (dark diamonds) in which
31
10 6
5
liquid injection rate and pressure drop were fixed across a 1-ft
beadpack; a) linear axes. b) log-log axes. In an experiment with 0
fixed injection rates rather than fixed pressure drop, "foam 0 1 2 3 4 5 6 7 8 9 10
generation" would be observed at the point where coarse foam
would convert to strong foam, with jump in ∇p of about a factor of water superficial velocity (uw) x10 6, m/s
100. Points A, C and D, with pressure gradients of 0.005, 0.5 and Fig. 5. Pressure gradient (psi/ft) as a function of gas and water
0.003 psi/ft, respectively, correspond to the conditions in Figs. 7, superficial velocities ug and uw, for the same surfactant
9 and 10 below. formulation and porous medium as Fig. 2. Open circles represent
data.
SPE 90938 13

70 0
0.125
gas superficial velocity (ug) x106, m/s

0.006772213
60 0.12
78 0.013544426
50 6 0.115 0.02031664

Water saturation, Sw
9 10
0.11 0.033861067
40 B
45 0.105
30 fg ≈72%
9 0.1
20 8
0.095
7
10 6
5 0.09

0 0.085
0 1 2 3 4 5 6 7 8 9 10
0.08
water superficial velocity (uw) x10 6, m/s
0.4 0.6 0.8 1
Fig. 6. Model fit to data in Fig. 5, using same parameters (Table 1)
as used in Fig. 2. Open circles represent calculated values, not dimensionless position, x/L
data. Point B corresponds to the conditions of Fig. 8. Line
represents a fixed foam quality of 72%, the conditions in Figs. 10
ff.
0
a) 0.6262 9.00E+13
0.004032305
0.00806461 8.00E+13
0.012096915
7.00E+13
Water saturation, Sw

0.6257
0.020161524
Foam Texture, nf

6.00E+13
0
0.6252
5.00E+13
0.006772213
4.00E+13 0.013544426
0.6247
3.00E+13 0.02031664
2.00E+13 0.033861067
0.6242
0.4 0.45 0.5 0.55 0.6 0.65 1.00E+13
dimensionless position, x/L
0.00E+00
0
b) 0.4 0.6 0.8 1
1.20E+09 0.004032305
0.00806461 dimensionless position, x/L
1.00E+09 0.012096915
Foam Texture, nf

0.020161524
8.00E+08 Fig. 8. Response of a low-quality-regime strong foam to
perturbation in foam texture nf. Times listed are time since
perturbation, in PV injected. There is no initial perturbation in
6.00E+08 water saturation. Perturbations in Sw and nf both decrease with
time. Steady-state properties, listed in Table 2, correspond to
4.00E+08 point B in Fig. 6. Cc = 0.001. 200 grid blocks.

2.00E+08

0.00E+00
0.4 0.5 0.6
dimensionless position, x/L

Fig. 7. Response of a coarse foam to perturbation in foam texture


nf. Times listed are time since perturbation, in PV injected. There
is no initial perturbation in water saturation. Perturbations in Sw
and nf both decrease with time. Steady-state properties, listed in
Table 2, correspond to point A in Fig. 2. Cc = 0.1. 200 grid blocks.
14 SPE 90938

0.1

0.098

0.096
Water saturation, Sw

0.094
0
0.092
6.42568E-05
0.09 0.000128514
0.00019277
0.088
0.000321284
0.086

0.084

0.082
0.4 0.5 0.6 0.7 0.8
dimensionless position, x/L

0
6.42568E-05
9.00E+13
0.000128514
8.00E+13
0.00019277
7.00E+13
0.000321284
Foam Texture, nf

6.00E+13
5.00E+13
4.00E+13
3.00E+13
2.00E+13
1.00E+13
0.00E+00
-1.00E+13 0.4 0.5 0.6 0.7 0.8
dimensionless position, x/L

Fig. 9. Response of foam in the intermediate state to perturbation


in foam texture nf. Times listed are time since perturbation, in PV
injected. There is no initial perturbation in water saturation. The
perturbations in Sw grows in magnitude, while nf reaches nfmax,
indicating a low-quality strong foam. Steady-state properties,
listed in Table 2, correspond to point C in Fig. 2. Cc = 1. 200 grid
blocks. Fig. 10. Simulation of co-injection of gas and liquid into a medium
saturated with surfactant solution, leading to a coarse-foam state.
-6 -6
ug = 2.57 x 10 m/s, uw = 1.0157 x 10 m/s (ut = 1.02 ft/d, fg = 0.72).
Steady-state properties correspond to point D in Fig. 2. 25 grid
blocks, and capillary-pressure gradients neglected. Cc = 0.1.
SPE 90938 15

Fig. 11. Simulation of the case in Fig. 10, with 100 grid blocks.

Fig. 13. Simulation of the case in Fig. 12, with 100 grid blocks.

Fig. 12. Simulation of the case in Fig. 10 (25 grid blocks), with
capillary-pressure gradients included.
16 SPE 90938

Fig. 15. Simulation of co-injection of gas and liquid into a pack


saturated with surfactant solution at the same foam quality as Fig.
10, at 1.69 times higher injection rate than Fig. 10 (1.72 ft/d). 25
grid blocks, and capillary-pressure gradients neglected.

Fig. 14. Simulation of co-injection of gas and liquid into a pack


saturated with surfactant solution, leading to a strong-foam state.
Foam quality is the same as in Fig. 10, with injection rate 8.15
times greater (8.31 ft/d). 50 grid blocks, and capillary-pressure
gradients neglected. Cc = 0.1.

Fig. 16. Simulation of co-injection of gas and liquid into a pack


saturated with surfactant solution at the same foam quality as Fig.
10, at 2.86 times higher injection rate than Fig. 10 (2.91 ft/d). 25
grid blocks, and capillary-pressure gradients neglected.
SPE 90938 17

Fig. 17. Simulation of co-injection of gas and liquid into a pack


saturated with surfactant solution at the same foam quality as Fig. Fig. 19. Simulation of co-injection of gas and liquid into a pack
10, at 4.83 times higher injection rate than Fig. 10 (4.91 ft/d). 25 saturated with surfactant solution at the same foam quality as Fig.
grid blocks, and capillary-pressure gradients neglected. 10, at 6.27 times higher injection rate than Fig. 10 (6.37 ft/d). 25
grid blocks, and capillary-pressure gradients neglected.

Fig. 18. Simulation of co-injection of gas and liquid into a pack


saturated with surfactant solution at the same foam quality as Fig.
10, at 5.79 times higher injection rate than Fig. 10 (5.89 ft/d). 25 Fig. 20. Simulation of the case in Fig. 10, with both Cc and Cg 100
grid blocks, and capillary-pressure gradients neglected. times smaller than in Figs. 10 and 11. 100 grid blocks, capillary-
pressure gradients neglected.
18 SPE 90938

Fig. 21. Simulation of the case in Fig. 10, with both Cc and Cg 100 Fig. 22. Simulation of the case in Fig. 14, with both Cc and Cg both
times larger than in Fig. 10. 100 grid blocks, capillary-pressure 10 times smaller than in Fig. 14. 100 grid blocks, capillary-
gradients neglected. pressure gradients neglected.

You might also like