Professional Documents
Culture Documents
Optimal Injection Strategies For Foam Ior - Shan Spe
Optimal Injection Strategies For Foam Ior - Shan Spe
Mobility During Liquid Injection After Gas. Many studies find Gravity Override in SAG Processes
that foam does not change the water relative permeability func-
Gravity override results from a competition between gravity (and
tion,19–22 but none of these studies examined specifically the case
density difference) and horizontal pressure gradient.36,38 For a
of post-foam injection of liquid.46–48
foam process with continuous foam injection, water saturation in
Here, the same mobility functions are used during foam injec-
the foam bank is nearly uniform and constant.36 For a process that
tion and gas and liquid injection in SAG processes. Refinement of
perfectly obeys the fixed-Pc* model, water saturation in the foam
this assumption must await definitive experimental studies.
bank is absolutely uniform and constant.40 Foam mobility is very
low throughout the foam bank. Especially in a cylindrical reser-
voir, high velocity and low foam mobility near the injection well
cause high injection pressure, which risks fracturing the forma-
tion.9 Moreover, velocity and pressure gradient at the displacement
front decrease as the displacement front moves away from the
well, promoting gravity override.36
In contrast, for SAG processes, during gas injection, water
saturation declines and mobility increases toward the injection
well (Figs. 2 through 6), helping to avoid fracturing the formation.
Fig. 7 shows total relative mobility as a function of position after
0.2 PV gas injection into a large bank of surfactant in a cylindrical
reservoir using the Fisher model at 0.02% surfactant concentration.
The shock front is near the middle of the reservoir at r⳱0.5 Re (or
xd⳱0.25 on Fig. 5). High mobility near the injection well results
from drying out and partial foam collapse there; it is not the result
of any shear-thinning rheology, which has been excluded from the
model.43 There is a low-mobility zone behind the shock front, a
high-gas-mobility zone near the well, and a single gridblock with
extremely low mobility at the shock front. According to Fig. 7, a
SAG process can employ higher injection rate to prevent gravity
override without exceeding the allowed injection pressure, because
foam mobility is high near the well during gas injection. With
relatively high mobility ahead of and behind the displacement
front, a SAG process with fixed injection pressure can focus much
of the reservoir pressure drop at the displacement front to fight
gravity override. In a SAG process with fixed injection rate, how-
ever, decreasing frontal velocity in radial flow reduces pressure
gradient at the front and promotes gravity override. Shi and Ros-
sen38 compare continuous foam injection and fixed-injection-rate
and fixed-injection-pressure SAG processes in rectangular and cy-
Fig. 8—Water saturation in a 2D cylindrical reservoir using the Fisher model with 0.02% surfactant concentration. (a) Continuous
foam injection (fw=0.2, qt=108 ft3/D, maximum pressure drop=350 psi); (b) gas is injected at fixed rate of 2,205 ft3/D, maximum
pressure drop=350 psi; (c) gas is injected at fixed pressure of 2,350 psi; (d) gas is injected at fixed rate of 315 ft3/D, maximum
pressure drop=50 psi; (e) gas is injected at fixed pressure of 2,050 psi.
Fig. 9—Water saturation in a 2D cylindrical reservoir using the Fisher model with 0.5% surfactant concentration. (a) Continuous
foam injection (fw=0.2, qt=1 ft3/D, maximum pressure drop=350 psi); (b) gas is injected at fixed rate of 120 ft3/D, maximum pressure
drop=350 psi; (c) gas is injected at fixed pressure of 2,350 psi; (d) gas is injected at fixed rate of 17 ft3/D, maximum pressure
drop=50 psi; (e) gas is injected at fixed pressure of 2,050 psi.
tion of 0.05 PV surfactant solution is followed by gas injected at regation in this region. For processes with sufficient pressure gra-
a fixed injection pressure of 2,100 psi. After 0.2 PV gas injection, dient, surfactant propagation limits sweep efficiency. For pro-
gas breaks through the surfactant front at about 0.38 Re, quickly cesses with sufficient surfactant, gravity and pressure gradient
rises, and overrides the rest of the reservoir. As in the SAG foam limit sweep efficiency.38
project in the Snorre field,18,41 the effectiveness of foam degrades
when gas breaks through the surfactant front. Thus, the ability of Multiple SAG Cycles. Comparison Between One SAG Cycle and
foam to overcome gravity override cannot extend beyond the front Two SAG Cycles. Once SAG injection is selected for field appli-
of the surfactant bank. cation, the slug sizes are important design considerations. Fig. 13
Thus, there are two necessary conditions for foam to overcome shows surfactant concentration and water saturation after 0.6 PV
gravity override: Sufficient surfactant solution to sweep the region total injection for one SAG cycle and two SAG cycles, for a SAG
of interest, and sufficient pressure gradient to avoid gravity seg- process at fixed injection rate. For the case with one SAG cycle,
Fig. 11—Water saturation in a 2D cylindrical reservoir using the 10% progressive-collapse model. (a) Continuous foam injection
(fw=0.2, qt=16 ft3/D, maximum pressure drop=225 psi); (b) 0.4 PV gas is injected at fixed rate of 922 ft3/D, maximum pressure
drop=225 psi; (c) 0.4 PV gas is injected at fixed pressure of 2,225 psi; (d) 0.2 PV gas injected at fixed pressure of 2,050 psi (different
gray scale; saturations are similar to other cases).
0.2 PV surfactant solution is injected first at an injection rate of Thus SAG with fewer, larger slugs gives a somewhat better
31.4 ft3/D, followed by 0.4 PV gas injection with an injection rate sweep efficiency than the case with many smaller slugs. The rea-
of 1,260 ft3/D. During injection of the first liquid slug there is no son, we believe, is a coupling between gravity segregation and
gas in the reservoir, so no liquid slumping occurs initially. For the surfactant propagation. In the two-SAG-cycle process, surfactant
case with two SAG cycles, 0.1 PV surfactant solution with an fills the lower portion of the reservoir not penetrated by gas, with
injection rate of 31.4 ft3/D is followed by 0.2 PV gas with an water saturation Sw⳱100%. Where gas sweeps, in this process,
injection rate of 1,260 ft3/D in the first cycle, followed by a second Sw≈20%, and a given volume of foam requires only one-fifth
cycle the same as the first. There is gravity segregation during the as much surfactant solution as a region with Sw⳱100%. In
injection of the second liquid slug in this process. Fig. 13 shows other words, a relatively small underride zone with Sw⳱100%
that the process with one SAG cycle gives nearly piston-like dis- represents what could have been a much larger region filled with
placement, while gas override has already occurred at 0.6 PV foam. In the two-cycle SAG process, a large portion of the injected
injection for the process with two SAG cycles. The reason is clear surfactant slug lies useless in the underride zone. Meanwhile,
from the surfactant front in Fig. 13: The surfactant front propagates in Figs. 13a and 14a, the piston-like sweep with the one-slug
further along the top of the reservoir, where override occurs, with process helps to push surfactant ahead and maximizes the propa-
only one SAG cycle. gation of surfactant.
Fig. 14 shows water saturations for processes with one SAG Liquid Slumping and Optimal Gas-Injection Strategy. In prin-
cycle and two SAG cycles with fixed injection pressure of gas in ciple, liquid slumping only a short distance downward could de-
both. Liquid is injected at a fixed injection rate of 31.4 ft3/D and prive the top of reservoir of the surfactant solution needed for
gas is injected at fixed pressure of 2,200 psi (the maximum injec- foam. Surfactant is treated as an aqueous-phase tracer in the simu-
tion pressure of the case in Fig. 13). After 0.6 PV gas and liquid lator,40 with no change in aqueous-phase density; the only factor to
injection, the process with one SAG cycle gives slightly better cause gravity slumping here is the large density difference between
sweep efficiency than the process with two SAG cycles. liquid and gas.
Fig. 13—Surfactant concentration (top) and water saturation (bottom) for SAG processes at fixed injection rate in a 2D cylindrical
reservoir using the Fisher model with 0.02% surfactant concentration. (a) One SAG cycle: 0.2 PV liquid is injected at 31.4 ft3/D, then
0.4 PV gas is injected at 1,260 ft3/D. (b) Two SAG cycles: 0.1 PV liquid is injected at 31.4 ft3/D, followed by 0.2 PV gas injected at
1,260 ft3/D, followed by a second cycle the same as the first.
Fig. 15 shows the performance of four cases after 0.6 PV lution has the same density as water. The first gas slug is injected
injection with two SAG cycles and different gas-injection pres- at 2,010 psi. Gas largely overrides the liquid slug because of in-
sures: 10, 30, 50, and 100 psi above reservoir pressure. The res- sufficient pressure gradient. Then, 0.1 PV surfactant solution is
ervoir is initially saturated with surfactant-free water with injected into the reservoir. By this point, gas override is significant,
Sw⳱100%. The liquid-injection strategy is same for all cases, with and liquid slumping is starting to deplete the top of the reservoir.
an injection rate of 31.4 ft3/D. Surfactant-concentration contours Then, 0.2 PV more gas is injected at a larger fixed pressure of
and water saturations in Fig. 15 illustrate that higher gas-injection 2,050 psi. During this period, the high pressure gradient pushes
pressure gives both better gas sweep efficiency and less surfactant liquid back up to the top of reservoir. Gas injection with sufficient
slumping. Comparison of these cases indicates that gas override pressure gradient can partially reverse gravity slumping during
and surfactant slumping go together. Better sweep of gas, with injection of liquid.
higher pressure gradient, helps the liquid propagate further along Optimal Liquid Injection Strategy. For the SAG project in the
the top of the reservoir. Snorre field,18,41 the high well pressure experienced during injec-
Thus, injecting gas at the maximum allowable injection pres- tion of the liquid slug caused fracturing of the formation, with loss
sure for SAG processes not only prevents gas override within the of surfactant, and reduced effectiveness of foam. Fortunately, Figs.
surfactant bank, but also decreases surfactant slumping, both of 16 and 17 suggest that liquid-injection strategy is not crucial to
which help gas sweep further. success against gravity override (as long as fracturing is avoided),
Redistribution of Liquid During Gas Injection. A worst-case because even if slumping of surfactant occurs, the following period
scenario for slumping of the surfactant slug occurs during injection of gas injection with sufficient pressure gradient can at least par-
of surfactant into a field with a gas present, but no foam. Highly tially heal this problem. This is especially important because the
mobile gas can redistribute, facilitating slumping of the surfactant mobility of a liquid slug injected after foam is still not fully re-
front. Fig. 16 shows a two-SAG-cycle process with gas, but no solved.46–48 Nonetheless, one can still make the following gener-
surfactant or foam, initially present in the reservoir (Swi⳱0.8). The alization: It is best to inject liquid at fixed pressure, just below
first cycle comprises 0.1 PV liquid injected at an injection rate of fracturing pressure, to accelerate the process and minimize the
31.4 ft3/D, followed by 0.2 PV gas injected at a modest 50 psi time for gravity segregation during liquid injection. Shan35 shows
above reservoir pressure. The second cycle is same as the first one. an example in which liquid injection at fixed injection rate doubles
During the first 0.1 PV liquid injection, severe gas override and the injection pressure, while injection at fixed, lower pressure
surfactant slumping occur, because gas mobility is initially high. gives nearly the same sweep efficiency.
However, the following period of gas injection with 50-psi pres-
sure drop across the reservoir partially redistributes surfactant so- Sensitivity to Missing Data at Low Water Fractional Flow. For
lution, with surfactant solution propagating further along the top of continuous foam injection in a given reservoir, gravity override
the reservoir. During injection of the second gas slug, liquid again depends only on injection-well pressure.36 Shi and Rossen suggest
propagates along the top of the reservoir, thus helping to prevent that sweep efficiency in fixed-pressure SAG processes may like-
gas override. It appears that foam can push surfactant back toward wise be independent of foam properties.38 Fractional-flow theory
the top of reservoir to mitigate gravity slumping that occurs during predicts that the effectiveness of SAG processes depends on foam
injection of liquid. Thus fixed-pressure SAG injection with suffi- strength at extremely high foam quality (low fw),39,40 but it is
cient pressure drop can at least partially heal a process of gravity difficult to conduct steady-state corefloods at this condition.51,52
slumping during liquid injection. Thus, if process performance were independent of this behavior, it
Fig. 17 illustrates a case in which insufficient injection pressure would simplify field process design.
in a previous SAG cycle led to gravity override. At the end of Figs. 8 through 11 compare the performance of four foams with
injection of the first liquid slug, the contour lines of the surfactant very different mobilities at low fw (Figs. 3 to 6) under similar
concentration are vertical, because (by assumption) surfactant so- conditions. These differences in foam models represent the differ-
ences of foam properties between different foam formulations, and than for 0.5% concentration, because with 0.02% surfactant con-
the uncertainty in foam behavior at the low fw that controls SAG centration, the foam bank has higher mobility (Figs. 5 and 6).
floods. The sweep efficiency with continuous foam injection (Figs. Therefore, the case with 0.5% surfactant concentration focuses
8a to 11a) depends only on injection-well pressure, as expected.36 more of the well-to-well pressure drop on the gas displacement
The sweep efficiency of fixed-pressure SAG processes with the front than that with 0.02% concentration.
same injection pressure is also remarkably similar. For instance, Sweep With Complete Foam Collapse in Shock Front. The
the Fisher foam with 0.5% surfactant concentration is approxi- data of Persoff et al.44 are also consistent with a model in which
mately 100 times stronger than with 0.02% concentration (see foam collapses completely in a shock front at the leading edge of
Figs. 5 and 6), but performance of the foams in fixed-pressure the foam bank (Fig. 2).38,40 In finite-difference simulation, foam
SAG floods at the same injection pressure (Figs. 8c and 9c, or 8e cannot form and collapse in a shock front with zero width; the
and 9e) is similar. minimum width of the front is one gridblock (see the lowest-
Injection rate is sensitive to foam strength, however, as noted mobility gridblock in Fig. 7). Therefore, artificial numerical dis-
above. Sweep efficiency with a fixed-injection-rate SAG process persion is introduced by finite-difference simulation. As the gas
also depends more on foam model than with fixed-pressure SAG. front moves across one gridblock, or, more precisely, drains the
It appears that Shi and Rossen’s conclusion38 should be modi- water from one gridblock and moves on to the next, foam is
fied: The performance of a fixed-pressure SAG process changes created and then collapses as the gridblock drains. For the fixed-
remarkably little with differences between foam models; pressure Pc* model (Fig. 2), there should be no low-mobility bank at all,39
gradient is a more important factor than details of foam behavior but finite-difference simulation artificially introduces at least one
in controlling gravity override in fixed-injection-pressure SAG gridblock with low mobility.40
processes. We propose an explanation below. For now, we note Comparing the two Fisher models (Figs. 8 and 9), the one with
that for the 0.02%-concentration Fisher model, a larger fraction of the narrower low-mobility zone (0.02% concentration) has similar
the well-to-well pressure drop is dissipated across the water bank sweep efficiency but much higher injectivity than the case with the
broader bank of much lower mobility (0.5% concentration). For a An Idealized Model for SAG Displacements
process in which foam collapses in the shock front, the width of the From the previous examples, it appears that the ideal SAG process
low-mobility bank is one gridblock; as gridding is refined, the introduces a spreading wave with low mobility at the gas displace-
width of this bank decreases. Sweep efficiency does not at first ment front and very high mobility near the injection well (Figs. 3
suffer. Fig. 18 shows the sweep efficiency of a foam that obeys the to 6). Shan35 shows that the pressure drop across the spreading
fixed-Pc* model, with 50, 200, and 500 gridblocks along the radial wave in a fixed-rate SAG process in a 1D cylindrical flow is
direction. It is remarkable that gravity override does not worsen constant in time. This observation motivates the following ideal-
more rapidly as the gridblock size decreases. An explanation is ized model for SAG foam displacements, in which all resistance to
suggested here. flow is concentrated in a narrow (but slowly spreading) zone at the
Nevertheless, the similarity in sweep efficiency masks an arti- gas displacement front, with infinite mobility ahead of and behind it.
fact in these simulations that is apparent in a plot of injection rate
vs. time. Fig. 19 shows cumulative gas injection for the same three Model Formulation. Fig. 20 shows a schematic of the idealized
cases. The simulation is not approaching a constant behavior as model of foam SAG processes. We assume that that the mobility
gridblock size is reduced, but injection rate is increasing with ahead of and behind the displacement front is much (infinitely)
increasing grid refinement, apparently without limit in the range of greater than mobility in a low-mobility zone at the displacement
grid refinements shown. In these simulations, injection rate is front. Water saturation ahead of the front is uniform at 100%.
regulated largely by the single low-mobility gridblock at the lead- Reservoir pressure is fixed at the top of reservoir at the production
ing edge of the gas bank. As the gridblock size shrinks, injectivity well, and, because mobility is infinite in the water bank, the pres-
increases. This symptom is a warning sign that the simulation as a sure in the water bank is fixed by hydrostatics. For the gas bank
whole is controlled by numerical dispersion at the gas shock front behind the displacement front, water saturation is uniform at Swf,
and is unreliable. gas density is close to zero, and, because mobility is infinite there,
the gas-phase pressure is uniform. Therefore, the whole well-to- reservoir. For simplicity, horizontal position variable x represents
well pressure drop ⌬P is focused on a low-mobility zone with either x (rectangular reservoir) or r (cylindrical reservoir).
thickness , which is assumed for simplicity’s sake to be narrow Applying Darcy’s law to the low-mobility zone, the Darcy
compared to the length of the reservoir. As in a spreading wave in velocity of gas inside the low-mobility zone is
1D (as in Fig. 4), the thickness of the front is assumed to increase
proportionately to the distance that the front has traveled, s. The ug = krfⱍⵜPⱍ, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
following derivation applies to either a rectangular or cylindrical
where ug⳱gas Darcy velocity within the zone, rf⳱gas relative vidual points in the front at positions [x(t),z(t)]. For each point
mobility in low-mobility zone, assumed to be constant throughout [x(t),z(t)], there is the distance s(t) that this point has moved since
the displacement, and |ⵜP|⳱the magnitude of the pressure gradi- the beginning of the displacement. The movement of the front is
ent in the low-mobility zone. Pressure gradient across the front is everywhere perpendicular to the front, with angle ␣ defining the
determined by the width of the front local normal to the front as illustrated in Fig. 20:
ⱍⵜPⱍ = ⌬P Ⲑ , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
and , the width of the low-mobility front, is proportional to s, the
␣ = arctan 冉冊 ⭸x
⭸z t
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5)
distance the front has traveled: where (⭸x/⭸z)t is the slope of the front at a given location at a given time;
= *s Ⲑ L*, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3) dx krf
vx = = ⱍⵜPⱍ ⭈ cos共␣兲 . . . . . . . . . . . . . . . . . . . . . . . . (6)
where the characteristic front width *⳱a parameter of the model dt 共1 − Swf兲
and L*⳱a length scale to be determined below. (*⳱the width of dz krf
the front when it has traveled a distance L*.) For 1D rectilinear or vz = = ⱍⵜPⱍ ⭈ sin共␣兲 . . . . . . . . . . . . . . . . . . . . . . . . (7)
cylindrical flow, s⳱x or r, respectively, but for 2D flow we give dt 共1 − Swf兲
a more complex definition. ⌬P − ⌬gz 共⌬P − ⌬gz兲L*
ⱍⵜPⱍ = = . . . . . . . . . . . . . . . . . . . . . (8)
Therefore, the interstitial velocity v of displacement front is *s
krf ⱍⵜPⱍ ds
v= , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4) = 共vx2 + vz2兲1 Ⲑ 2, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
共1 − Swf兲 dt
where ⳱porosity of reservoir. The Darcy velocity ug, interstitial where vx and vz⳱the velocity of the front (perpendicular to the
velocity v, and pressure gradient across the low-mobility zone are front) in the x and z directions, respectively, at a given point;
perpendicular to the interface between the gas and water zones. To ⌬P⳱well-to-well pressure difference at the top of the reservoir,
track the movement of the front, we track the movement of indi- ⌬⳱the density difference between phases, g⳱gravitational ac-
Fig. 19—Cumulative-injection profiles for the cases in Fig. 18. Fig. 20—Schematic of the idealized model for SAG displacements.
IC: at t = 0, x = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10) where x, the position variable in the model, stands in for r.
IC: at t = 0, s = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11) To allow for gas trapping, in all cases we forbid the front to
retreat [i.e., forbid (dXD/dtD)<0 and (dZD/dtD)<0 for any point on
BC: at z = 0, ␣ = 0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (12) the front] even if ⌬P decreases during the injection process.
The dimensionless differential equations for the positions of For a fixed-pressure SAG process, a single family of curves
individual points on the displacement front are [XD(ZD)] as a function of tD (Fig. 21) describes all displacements.
In addition, the characteristic width * or mobility rf of the low-
dXD 共1 − ZD兲 ⭈ cos共␣兲 mobility zone affects the scaling of tD (and the injection rate), but
= . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13)
dtD sD not the shape of the front. Gravity override for this idealized fixed-
pressure SAG process depends only on density difference between
dZD 共1 − ZD兲 ⭈ sin共␣兲
= . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (14) phases (⌬g) and well-to-well pressure difference (⌬P). In the
dtD sD idealized model, for given ⌬P, sweep efficiency is independent of
dsD
dtD
= 冑冉 冊 冉 冊dxD
dtD
2
+
dzD
dtD
2
, . . . . . . . . . . . . . . . . . . . . . . . . . . (15)
the details of foam behavior [i.e., (rf/*)] as long as model as-
sumptions are satisfied—much lower mobility near the leading
edge of the gas bank than ahead of the bank or near the injection
where well. This explains much of the response of the fixed-injection-
冉 冊
pressure SAG process to the different foam models in Figs. 8
⭸XD ⌬gx through 11 above.
␣ ≡ arctan , XD ≡ . . . . . . . . . . . . . . . . . . . . (16)
⭸ZD tD ⌬P At the start of the displacement, the front is vertical; it begins
to tip over because the pressure difference between gas and water
⌬gz krf⌬g is greater at the top of the reservoir than at the bottom, because of
ZD ≡ , tD ≡ t . . . . . . . . . . . . . . . . . . . . . . . . (17)
⌬P 共 − Swf兲*
1 hydrostatic effects present in the liquid but not the gas. A fixed-
⌬gs ⌬P pressure SAG process is so successful in the idealized model be-
sD ≡ L* ≡ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18) cause all that is needed is a well-to-well injection pressure that is
⌬P ⌬g well above the hydrostatic pressure difference in the liquid phase
These definitions of dimensionless position and time differ from across the reservoir height. Even if ⌬P is not much larger than this,
those in fractional-flow theory.1 The boundary conditions are and the front starts to tip over, the displacement is aided by the fact
that the front moves normal to its direction. Thus even as the front
IC: at tD = 0, XD = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (19) tilts over, the front moves normal to this direction (i.e., downward
IC: at tD = 0, sD = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (20) toward the unswept portions of the reservoir).
where W⳱the width of the reservoir in the y direction, and H⳱its Comparison With Fixed-Rate SAG Process. To model a fixed-
height in the z direction. In a cylindrical reservoir, rate SAG process in the context of the idealized model, one must
Fig. 21—Dimensionless progress of gas injection in the idealized model for fixed-injection-pressure SAG displacement.
adjust ⌬P over time to maintain injection rate fixed. In our calcu- 12 psi above reservoir pressure, the injection pressure at the end of
lations, we approximated fixed-qt injection as follows. At each the fixed-qt process. In the fixed-⌬P case, injection rate decreases
timestep, we compute the advance of the front by forward differ- with time, but is always higher than 1 ft3/D. The fixed-⌬P process
ence and iterate on ⌬P until qt is within 1% of its fixed value. We gives somewhat better sweep efficiency than the fixed-qt process
used 105 timesteps in a typical model run. The calculation diverges because of higher ⌬P at the start of injection in the latter case. In
at the start, when sD⳱zero (Eqs. 13, 14, and 18); therefore, we this example, neither process sweeps to the bottom of the reservoir,
start with sD⳱0.001 at tD⳱0 for all ZD. This small error at the because 12 psi injection pressure is insufficient to overcome hy-
start rapidly becomes insignificant as the front advances over drostatic pressure difference over a height of 30 ft. (For simplicity,
larger distances. gas density is taken as zero in calculations with the idealized model.)
Figs. 24 and 25 compare fixed-rate SAG injection to fixed- In a cylindrical reservoir (Fig. 25), the maximum pressure for
injection-pressure SAG in a rectangular and a cylindrical reservoir. the fixed-qt process occurs near the start of injection, while the
In these figures, Swf⳱0.1515. The dimensionless front shapes are injection rate of the fixed-⌬P process increases with time. Still, the
identical for fixed-pressure SAG in rectangular and cylindrical injection rate and injection pressure of the fixed-⌬P process are
geometries for the same values of XD, but the PV injected at a always greater than for the fixed-qt process. In the fixed-qt process,
given value of XD differs in the two geometries. In a fixed-qt injection pressure decreases over time, and gas stops entering the
process in a rectangular reservoir (Fig. 24a), and ⌬P increase lower injection interval after a certain time. Injection rate in a
with time until the low-mobility front reaches the production well. fixed-⌬P process increases to keep pressure drop constant. Gas
The increase in ⌬P with time in this case helps gas sweep toward sweeps the entire formation in the fixed-⌬P process. The improve-
the bottom of reservoir later in the flood. In the fixed-⌬P process ment in sweep efficiency would be even greater in a process with
in the same rectangular reservoir, gas is injected at a fixed pressure lower injection rate and correspondingly lower injection pressure.
Fig. 23—Displacement front position after 0.2, 0.4, and 0.6 PV of A Model for Effects of Numerical Dispersion at the Shock
gas injection into a rectangular reservoir of Fig. 22 using the Front. In a 1D cylindrical or rectilinear displacement, the width of
idealized model. Gas is injected at 50 psi above reservoir pres- a spreading wave increases proportionately to distance the wave
sure. Other parameters are listed in Table 1. has traveled, as assumed for front width above. Numerical dis-
persion, on the other hand, spreads a shock front into a front of IC: at tD = 0, XD = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (28)
finite, relatively fixed width (one or two gridblocks) as the dis-
placement proceeds. As noted above, for a foam process obeying BC: at ZD = 0, ␣ = 0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (29)
the fixed-Pc* model, the only low-mobility zone is that provided As in the original model, rf and appear only in the ratio
by numerical dispersion at the shock front. For analyzing the effect (rf/). The model is unchanged as → 0 as long as this ratio holds
of numerical dispersion on simulation results, it is useful to con- constant.
sider a different model, in which is fixed and independent of s. Fig. 26 shows the behavior of this model. Sweep efficiency is
For this model, comparable to that in Fig. 21, though the sweep at the bottom of
the reservoir is a little lower in Fig. 26 than in Fig. 21. In the
dXD original model, the slow advance of the front at the bottom of the
= 共1 − ZD兲 ⭈ cos共␣兲 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (24)
dtD reservoir gives small s, small , and larger |ⵜP|, partially mitigating
dZD the effects of poorer sweep there. This effect is missing when front
= 共1 − ZD兲 ⭈ sin共␣兲 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (25) width is fixed. Processes where ⌬P is much greater than ⌬gH,
dtD
冉 冊
however, would correspond to small values of ZD (see Figs. 21 and
⭸XD ⌬gx 23) and show excellent sweep.
␣ ≡ arctan , XD ≡ . . . . . . . . . . . . . . . . . . . . (26)
⭸ZD tD ⌬P This result helps explain the spurious high sweep efficiency
observed with the fixed-Pc* model with abrupt, complete foam
⌬gz krf⌬g collapse (Fig. 18). There, a low-mobility zone one gridblock wide
ZD ≡ , tD ≡ t . . . . . . . . . . . . . . . . . . . . . . (27)
⌬P 共1 − Swf兲 is provided by numerical dispersion at the displacement front (Fig.
7). Nevertheless, the narrow low-mobility zone, a numerical arti- It is easily modeled in the context of fractional-flow theory.1,39
fact, gives reasonably good sweep efficiency and little deteriora- With adsorption, one needs larger slugs of surfactant to sweep a
tion upon refinement of the grid; as noted, the model predicts that given volume of reservoir with foam.
the width of the low-mobility zone does not affect the shape of This study has also excluded the effects of oil on foam.3,4,53
the front (see Fig. 18) but does affect injectivity (see Fig. 19). In Most crude oils significantly weaken foam, especially in oil-wet
reality, sweep efficiency must deteriorate eventually with increas- reservoirs. At this time, there is no consensus on the best way to
ing grid refinement, as the pressure drop across the single low- account for the effects of oil, which can depend on oil composition
mobility gridblock becomes insignificant compared to the pressure and can be complicated by movement of oil during the foamflood.
drop in the liquid bank ahead of the foam. In principle, fractional-flow theory can handle multiphase, multi-
component foam displacements with oil present, though the analy-
sis becomes more complex.54 If the oil remains immobile during
Discussion the foam displacement,53 it would be relatively simple to accom-
Both the idealized model and the simulations presented throughout modate the weakening effect of oil on foam in the context of
this report assume that injection and production rates can be ad- two-phase (gas/water) fractional-flow theory.
justed as needed to maintain fixed injection and production pres- As a first step toward understanding gravity override in real,
sures, with corresponding (potentially large) increases in rates of heterogeneous reservoirs, this study has considered homogeneous
produced fluids. There may be operational barriers to implement- reservoirs. In addition to its benefits in overcoming gravity over-
ing this strategy in practice. ride, foam can be stronger in higher-permeability media and
For this study, we have excluded the effects of surfactant ad- thereby mitigate the effects of permeability variations.3,4,26 More-
sorption. Surfactant adsorption slows the propagation of surfactant over, foam created as gas attempts to migrate vertically across
and is an important economic consideration in foam applications. layer boundaries24,43,55,56 would help suppress gravity override in
heterogeneous reservoirs. Fuller investigation of these mechanisms has much lower mobility than both the water ahead and the dried-
on the field scale must await further study. out region near the injection well. The following conclusions de-
rive from this model:
Conclusions 9. In the ideal SAG process, the well-to-well pressure drop is
focused entirely in the displacement front. Then, the well-to-
The following conclusions derive from our simulation study:
well pressure drop need only compete with the hydrostatic
1. A SAG process at fixed injection pressure better controls grav-
pressure difference over the height of the reservoir; thus, mod-
ity override in homogeneous reservoirs than either continuous
erate injection pressures give very good sweep efficiency. In
foam injection or a fixed-rate SAG process.
addition, even as the displacement front tilts over from over-
2. Sufficient surfactant slug size and sufficient pressure gradient
ride, it then tends to sweep downward from the override zone.
are both necessary conditions for foam to overcome gas over-
10. Fixed-rate SAG processes suffer from declining injection pres-
ride. Surfactant slug size determines the potential region of
sure in radial flow, reducing pressure drop at the front over
sweep and sufficient pressure gradient prevents gravity segre-
time compared to the hydrostatics with which it competes.
gation in this region.
Gravity override worsens as a result. The fixed-⌬P SAG pro-
3. The case with one large SAG cycle gives somewhat better sweep
cess gives marginally better sweep efficiency than a fixed-rate
efficiency than that with more numerous, smaller SAG cycles.
SAG process in a rectangular reservoir because of higher ⌬P at
4. For cases where gravity override has taken place during the first
the start of injection.
SAG cycle, better injection strategies in the following SAG cycles
11. Because the idealized model takes no account of details of
can reduce surfactant slumping and improve sweep efficiency.
foam behavior, it explains why sweep efficiency in a fixed-
5. Gas should be injected at the maximum allowable injection
pressure SAG process is relatively insensitive to the details of
pressure; this not only increases gas sweep efficiency within
foam behavior. It also helps explain why even a spurious low-
the surfactant bank, but also decreases surfactant slumping.
mobility zone, created by numerical dispersion, can by itself
6. Liquid should likewise be injected at maximum allowable in-
give reasonable sweep efficiency in simulation. If, however,
jection pressure; this puts liquid most rapidly into the reservoir
the foam process creates a weak foam, so that the water bank
and minimizes segregation during liquid injection.
ahead of the foam accounts for a significant portion of the
7. Differences between foam models (i.e., differences between
well-to-well pressure drop, then process efficiency suffers
surfactant formulations) have only a moderate effect on gravity
somewhat. These inferences are supported by the simulations
override in SAG processes with the same fixed injection pres-
with five foam models shown here.
sure. The differences in sweep observed among foam models
may reflect differences in the fraction of the well-to-well pres-
Nomenclature
sure drop dissipated in the water bank ahead of foam. Although
different foam models give similar sweep efficiency in fixed-pres- Consistent (for instance, SI) units are assumed in Eqs. 1 through
sure SAG injection, they do differ significantly in injection rate. 29, but customary units are cited in the text and figures. Both SI
8. If fractional-flow theory indicates that foam both forms and and customary units are given below.
collapses within the shock front in a SAG displacement, con- fw ⳱ water fractional flow, fraction
clusion 7 does not apply. Ironically, in such a case, a finite- g ⳱ gravitational acceleration, L/t2, N/kg [lbf/lbm]
difference simulation still gives a low-mobility bank at least H ⳱ height of reservoir (Fig. 20), L, m [ft]
one gridblock wide. This numerical artifact then provides k ⳱ absolute permeability, L2, m2 [md]
sweep improvement in the simulation, and the sweep effi-
L ⳱ length of rectangular reservoir (Fig. 20), L, m [ft]
ciency may not change much upon grid refinement. The symp-
tom of the artifact is that injectivity increases without bound L* ⳱ length set equal to ⌬P/⌬g in rendering position
upon increasing grid refinement. variable dimensionless
An idealized model, in which all resistance to flow is confined to Pc ⳱ gas/water capillary pressure, m/Lt2, Pa [psi]
a narrow bank between high-mobility fluids ahead of and behind Pc* ⳱ limiting capillary pressure for foam stability, m/Lt2, Pa
the displacement front, matches simulation results if the foam bank [psi]