Professional Documents
Culture Documents
jxz y
Mathematical Simulation of Polymer Flooding in
Complex Reservoirs
P. L, BONDOR
G. J. HIRASAKI
SHELL DEVELOPMENT CO.
JUNIOR MEMBERS AIME
HOUSTON, TEX.
M. J. THAM
MEMBER AIME
%
ABSTRACT include the presence of a gas cap or a desaturated
zone due to gravity drainage in a dipping formation,
Simulation of polymer flooding in many complex the presence of an aquifer, irregular well spacing
reservoirs has requirements that preclude the use and reservoir boundaries, multiple zones, reservoir
of either three-phase stream tube or two-phase heterogeneities, and a well performance that is
finite-difference simulators. The development of a limited by state proration, injectivity, and
polymer flooding model used in a three-phase, productivity. These reservoir features are being
four-component, compressible, finite-difference represented by most compressible, three-phase,
reservoir simulator that allows the simulation of a three-dimensional simulators. However, to model
variety oi complex situations is discussed. polymer flood projects, it is necessary to include a
Tbe polymer model represents tbe polymer conservation equation for the polymer, and to
solution as a fourth component that is included in
represent the adsorption of polymer, the reduction
the aqueous phase and is fully miscible with it.
of the rock permeability to the aqueous phase after
Adsorption of polymer is represented, as is both
contact with the polymer, rhe dispersion of the
(1) the resulting permeability reduction of the
polymer slug, and the non-Newtonian flow behavior
aqueous phase and (2) the resulting lag of the
of the polymer solution.
polymer injection front and generation of a stripped
water bank. Tbe effects of fingering between the PREVIOUS SIMULATOR DEVELOPMENT
water and polymer are taken into account using an
Previous simulator development of polymer
emp in”cal “mixing parameter” model.
flooding has been reported in two different general
~,~g ?.esz!~ing simgjg~g? ~~ Cgpgbje o~ rne&@
categories: three-phase stream tube rnod.e!s and
reservoirs with nonuniform dip, multiple zones, one- or two-phase, incompressible, finite-difference
desaturated zones, gravity segregation, and simulators.
irregular well spacing and reservoir shape.
Jewett and Sc hurz 1 developed a two-phase,
Two examples are presented. The first illustrates
multilayer Buckley -Leverett displacement simulator
the polymer flooding of a multizone dipping reservoir
capable of modeling either linear or five-spot
with a desaturated zone due to gravity drainage.
patterns. A mobile gas saturation also could be
The second illustrates the flooding of a reservoir
specified, but this was treated as void space and
with a gas cap and an oil rim with polymer injection
did not affect the flow characteristics of the
near the oil-water contact. In this example, the
system. Gravitational and capillarity effects were
effects of nonuniform dip, irregular well spacing
neglected. The residual resistance of the brine
and field shape, and gravity segregation of the flow
following a water slug was modeled as an increase
are all taken into account. The two examples
in its viscosity; the viscous fingering of the brine
presented illustrate the versatility of tbe simulator
through the polymer slug was treated by altering
and its applicability to a wide range of problems. empirical relative permeability relationships to
specify a more adverse mobility ratio.
INTRODUCTION Slater and Farouq-Ali2 modeled five-spot patterns
The design of a polymer flood for a complex with a two-phase, two-dimensional, finite-difference
—l. .— —-—!-—.:----- ..:------ J--- :11 --:... -I-L -..
reservoir requires a model that represents the simulator, neglccung gIav ILy am Lap IL Idt LLy. IIICy
reservoir features that have a significant ef feet on obtained an empirical expression for the resistance
the performance of the flood. These features may factor of the porous medium as a function of a
time-dependent mobility ratio.
Patton, Coats and Colegrove3 developed a
Original manuscript received in Society of Petroleum Engineers
office July 28, 1971. Revised manuscript received April 5, 1972. finite-difference model utilizing a stream tube
Paper (SPE 3524) was presented at the SPE 46th Annual Fall
Meeting, held in New Orleans, Oct. 3-6, 1971. ~ Copyright 1972 approach that could be used to simulate linear or
American Institute of Mining, Metallurgical, and Petroleum five-spot polymer floods in either a single sand
Engineers, Inc. reservoir or a reservoir with several noncommunicat-
1Ref~~~~ceS given at end of PaPer.
~ ( @bwSw) - –
‘ads
—
o
. . . . (1)
and water saturations
of polymer from polymer solution
are adjusted. The stripping
generates water in
at at front of the leading
TzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA edge of the polymer, and thus
forms a stripped water bank berween the connate
water and the polymer solution banks. The details
V . [ Aobo(vpo - pog~)] + do of the calculations are described in Appendix A.
Sp kkraq(Saq)
k.. =— . . . . . . (12)
p
s R
aq k
Sw kkraq(Saq)
.- ... . . /,7\
kw=— . {12)
‘/+., + PSEUDOPIASTIC~pm,n~ DllATANT
s
aq ‘k
10G U
OCTOBER, 1972
371
pseudoplastic and the viscosity is decreasing with power-law coefficient, K, and exponent, n, were
increasing velocity. 7-1s At a higher velocity there determined from the viscometric data (the actual
~~ ~ ,,,..’
m;m;.,,’-..!
.n,, m .t;cfiac;*y
. .Us”u.. $bla$ is -“,,=1
.yy... rn
.“ ~~ g~ea~er VC+IIIF. C urt=r~
. . . . ...-” . . . . Ar=rerminc=d
-------------- IOQ-IOP
f~Q.rn. a ._. --0 =-.DIOt . of shear
than the solvent viscosity. lo! 11,14 At very high stress vs rate of strain). The coefficient, H,
velocities the viscoelastic effects become important determined from Eq. 20 and the power-law exponent,
and the viscosity increases with increasing n, were used to compute the apparent viscosity from
velocities.1 1,14-17 In this region, the theological Eq. 19. The calculated apparent viscosity using
behavior is “dilatant”. the modified Blake-Kozeny model is compared with
The pseudoplastic behavior of the flow of the the experimental values on Fig. 3. The maximum
,.-l., . ..-. ~fil...:f,., ●h.,.,. mh I-.,-..,.,le .“. EJ:. ~afi ~~ 4:C1-.---- LC k=+..,a-.. ●L- nnl,-iiln*,aA .amA ewmarin-.t=nral
IJ’J’Y11’=’ ‘U’u”u” ““VW5” PJ’--- 11~~-~- ULLICLC1l V= LWG=ll !..~~ ~=.ti-.-.~= -.’.’ -.-y ---. --&=---
modeled over a wide range of flow rates with a values was less than 10 percent.
power-law model. The model that we have used is The Blake-Kozeny model accurately reproduced
the modified Blake-Kozeny model for power-law the apparent viscosity behavior of Kelzan-M
fluids.12~ 18 This model has been used to correlate solutions for the systems considered here. Kelzan-M
data in packed beds over a wide range of Reynolds is a type of polymer that provides a mobility
numbers.12~ 13 The apparent viscosity can be reduction by an increase in viscosity alone.
expressed with this model as However, with polymers such as partially
hydrolyzed polyacrylamide, 21 a significant factor
n-1 in the mobility reduction is the reduction of the
P =Hu>. .o. o....- (19)
app permeability to the aqueous phase. If the parameters
H and n in Eq. 19 are determined from flow
where u is the superficial fluid velocity. The values
experiments through a core sample of the reservoir
of the parameters, H and n, can either be calculated
rock through Eqs. 16 and 19, then the effect of the
directly from data on flow through core material or
be estimated from viscometric measurements. If the permeability reduction is included in the parameters.
power-law coefficient and exponent, K and n, are If the parameters are estimated from viscometric
calculated from viscometric data, then H can be measurements through Eq. 20, then the apparent
estimated as viscosity calculated from Eq. 19 will not represent
the effect of the permeability reduction. However,
the development of the non-Newtonian well model
‘=:(9n:3)n(’’0k4=”
in the next section requires only the ratio of
’20) viscosities
will
and, thus,
be canceled
any permeability
in the final
reduction
expression. The
‘
in a region that contained original oil, and is Sw
= 1.0 in any region that was originally 100 percent
water. 2
The modified Blake-Kozeny model was applied to
data on the flow of Kelzan-M solutions through cores
I I I I 1 I I I I I I
1
with a residual oil saturation present. The core and 1 10 lo~
fluid data are summarized in Tables 1 and 2.19 The SHEAR RATE, yhOC-’]
viscometric data for the three polymer solutions are FIG. 2 — VISCOMETRIC BEHAVIOR OF KELZAN-M.
shown in Fig. 2. All three solutions showed
pseudoplastic behavior to a different extent. The
Core No.
kw (red) sor
1 17.0 0,188 G
2 7.7 0.20 0.20
3 22.8 0.20 0.20
Cone. ~brine
I I I I I I I I II I 1 I J
Core ,.-4 ,.-3
No. (ppm)
—— (CP) (Cp S:&-l) _n_
u (cmkec)
1 Kelzan-M 200 0.84 7.6 0.67
2 Kelzan-M 300 0.45 2.25 0.85 FIG. 3 — THEOLOGICAL BEHAVIOR OF KELZAN-M
3 Kelzan-M 300 0.45 5.5 0.69 IN CORES.
viscoelastic behavior can be included in the Representing the non-Newtonian behavior in ordy
modified Blake-Kozeny model. The resultantthe intergrid-block fluid transmissibility coefficients
apparent viscosity with this model can be expressed will not properly include the much more significant
as effects near the wells. Unless the flow rate is so
n-1 large that the dilatant effect becomes significant,
Hu
= , (21) the low polymer viscosity near injection wells will
P result in a higher injectivity than if rhe injectivity
app
0<1. u were based on an average polymer viscosity in the
1- reservoir. Thus, all the non-Newtonian effects are
represented in the simulator by a well injectivity
d @w
(1 -zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
c&) IsokwlzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
model. The required increase in injectivity over
that based on an average Newtonian viscosity is
where 61 is the fluid relaxation time. (Similar obtained by introducing a negative, rate-dependent
models for the viscoelastic behavior have recently skin factor into a radial-flow equation for each
been developed. z“~zl) The viscoelastic effect injection well.
becomes significant as the second term in the The well model relates the Weii rates and
denominator (which is equivalent to the Deborah pressure to the grid block pressure and saturations
number introduced by Marshall and Metzner 16) by assuming radial flow in an annular region as
becomes significant compared to unity. It will be shown in Fig. 5. The external radius, re, is the
assumed in the following that the polymer flood radius of a circle that has the same area as the
wi!! be designed such that the detrimental effect of grid b!Qck.
the viscoelastic behavior does not occur in the
reservoir.21
Ax Ay
The power-law model and the resulting Blake- r . (23)
?- --- 1 . . . . . . . . .
Kozeny model represents only the pseudoplastic ‘e
range of flow rates. The maximum and minimum
IIT
viscosities at low flow rates and high flow rates,
The rate of injection of polymer from the well into
respectively, could either be represented empiri-
a grid block is modeled as
cally or with a four-parameter model such as the
Meter model. 22 However, for our purpose the
apparent viscosity was modeled to be Newtonian at 10
I 1 I 1
low and high flow rates and to follow the Blake-
----- -----
Kozeny model in the pseudoplastic region.
z
~s -
( P
max
, low velocities *
n-1 I
v= Hu, pseudoplastic region 1 1 1 1
app 600 800 1000
200 400
high velocities .(22)
r(ft)
w
min ‘
[ FIG, 4 — VISCOSITY PROFILES.
if
NV<V
+4 I
I
I
i
,
I
I
V’’C”’’TR”F”LELE and producer
viscosity
apparent viscosity
the injector
is 1,000 ft. Since enough data were
not given in Fig. 3 to estimate
at low velocity,
the value
and producer,
the maximum apparent
choose as the maximum
at the midpoint between
i.e., re = s00 ft. From
I Fig. 4, the viscosity at this point is 7.3 cp. Suppose
that the well radius is rw = 0.25 fr. The apparent
skin factor from Eq. 27 is then sp = -4.81. This
represents an effective well radius, r;, of 31.0 ft.
The ratio of injectivity with the non-Newtonian
effects and the injectivity assuming a Newtonian
viscosity of pmax is
1 I 1
r, r=’ r, lNN in re/rw
—. = 2.73
10G r-
in re/rw + s
FIG, 6 — RADIAL PROFILES. lN P
SOCIETY OF PETROLEUM ENGINEERS JO URiVAL
TABLE 3 — DATA FOR LINEAR SIMULATION displacement and no adsorption, polymer solution
System dimensions 2ooftxloftx loft would break through at 0.78 PV injected. With
a&sorption, breaicti-lroiigh --- . ..-
uLLkLLcd at Q.g~ pv
Reservoir data:
!=i~ja! ~jl saturation 0.75
injected.
Connate water 0.25
0,22
These figures also illustrate the effect of
Residual oil to WF
Pore volume 1,069 bbl numerical dispersion on the polymer saturation
Original oil in place 778.7 bbl profile. Even though a marked effect is evident, the
Waterflood movable oil 544 bbl
indicated recovery given by the simularor is affected
Po,osity, 0.3
4
-- . ..-.. h.l)*w k 1 darcy
to a much lesser extent; the oil recovery curves for
m
‘b--”=Wldl-‘- F.?rqo!=””.. . . . .-
& at connate water 0.9
Fluid doto:
20 Cp
.2 PV POLYMER
Oil viscosity
Y%ter Wisccsi?y 0.64 Cp ‘!.3 .4 PV WATER
Polymer viscosity 2.9 Cp
Ad S 10 pg/g
R,, i.d
GJ 0.666
----- -. ....- -
.2 PV PULYMt K
.2 Pv w&TE!?
FIG. 7C —EFFECT OF ADSORPTION AND DIsPERSION
ON POLYMER SATURATION, LINEAR MODEL, 0.8 PV
INJECTED.
a. 3 .
(n
.2 PV POLYMER
z-
0 .8 PV WATER
DISTANCE,
!___ .6
X/L
.8 1.0
tii
.2 .4
DISTANCE,
.6
X/L
.8 1.0
FIG. 7A — EFFECT OF ADSORPTION AND DISPERSION FIG. 7D — EFFECT OF ADSORPTION AND DISPERSION
ON POLYMER SATURATION’, LINEAR MODEL, & 4 Pv ON POLYMER SATURATION, LINEAR MODEL, 1.0 PV
INJECTED. INJECTED.
constraints on both injection and production wells & at residual oil 0.3
,.,. -/- , R=5~~u=! =~! .!”*,,.”*
-“.”,”, ;-”
,“. , Q, ~~
(lu D/u Daianceri injection and production). Time 0,25
Connate water saturation
steps used were: 0.2 days for the first 10 days, 0.5 Initial saturation distribution:
days for the next 10, and l-day steps for the In gas zone:
less useful. An example will be presented in which An experimental program was undertaken to
ir was necessary to model the recovery process evaluate the effectiveness of polymer flooding in
with our mathematical reservoir simulator. such a reservoir. A cross-sectional bead pack model
was constructed that was designed such that
POLYMER FLOODING IN A VISCOUS OIL nondimensional parameters of the system were the
RESERVOIR WITH A DESATURATED ZONE same in both model and prototype. A series of
experiments were performed for both waterflooding
Waterflooding a viscous oil reservoir with a
continuous gas zone (due to gravity drainage), Fig.
9, is characterized by the tendency of the gas zone PROOUCING
to act as a thief zone until most of the gas has WELL A WELL B WELL
?5
‘U ‘
.2
FIG, 9 —
‘ RESERVOIR WITH DESATURATED
CAUSED BY GRAVITY DRAINAGE.
v- ZONE
I/
oo~
PORE VOLUMES INJECTED
confirmed the qualitative behavior of the desaturated gas zone as a dispersed phase, and
experimental model. assuming negligible gravity segregation of the
Fig. 10 illustrates an experimental waterflood in aqueous and oil phases, yields results that are
the cross-sectional model. Figs. 11 and 12 are. comparable to the five-Iayer model in which the gas
water and oil saturations obtained by the simulared zone is distinct. For this problem a single-layer
waterflood at the same time. The overrunning of the model using the same relative permeability curves
oil by the water in the desaturated zone is evident.. as the five-layer model is sufficient.
No oil bank is formed by the flood water, and very
SIMULATION OF A FLOOD IN A
lirtle oil has been moved.
RESERVOIR WITH A DESATURATED ZONE
Fig. 13 illustrates an experimental 0.25 PV
polymer flood where the slug has been followed by A simulation study was carried out for a reservoir
water. Figs. 14 and 15 are polymer and oil with six noncommunicating layers, each with a
saturations obtained by the simulation of this desaturated zone that varied areaiiy. Tine study
poiymer fiOOd. Note the formation of an oii banit ~--m. . . ..:.-,-1
&Aca L&u rha
. ..s ~ff,=rr
------ ~f che &~~rgK~t~~ z~e on oil
ahead Qf ~he PQi~rne~, NQte aisQ Lhat 50rne Oii iS recovery and on the oprimal flood pattern, and
moved upward into the desaturated zone; if the included the variable reservoir continuity and
desaturated zone has a residual oil saturation less heterogeneity. We include here the results of a
than waterflood residual, there will be resaturation five-spot pattern flood that was modeled.
losses even in the swept portions of the reservoir. Figs. 18A and 18B illustrate the five-spot pattern.
Observed and calculated production curves are Each producing sand was treated as a single layer
presented in Fig. 16 for floods terminated at 98
percent water cut. The extent of the vertical and
horizontal lines on the experimental data points
indicates the standard deviation of the experimental
results. The agreement between experiment and
simulation is reasonable. Note that, while both
studies indicate that the waterflood will recover
more of rhe movable oil in place (68 percent vs 61
------- *L- --1.? --- ficc~ ““.nh,n;a.L.
”cs ~~~ Qii ~~ ~.~
l-’--u), ‘“= PJJY lll=L
~~~o Iq _ POLYM.ER SP.TIURA.TION DISTR.IEKJTION,
CROSS-SECTIONAL SIMULATION.
● MATH
I
I
MODEL
O PHYSICAL MODEL } ‘ATERFLOOD
I I 1 f 1
~ FLOW
1
a
> 20
‘e
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
CONNAIE WAIE* f,ooo WAIE* 4
~ a
GAS
PO, ”MER we oo-
SLUG 1.0 2.0 3.0 4.0
=i PV THROUGHPUT
FIG. 13—EXPERIMENTAL 0.25 PV POLYMER FLOOD, FIG, 16 — OIL RECOVERY EFF1C2ENCY, CROSS-
CROSS-SECTIONAL MODEL. SECTIONAL MODEL.
I
40 +
u.
z
20:’
10 $
303
5
10 +
20
block, psi
FIG. 19A — OIL RECOVERY RATE, WATERFLOOD.
Pruf = bottom-hole injection pressure at
POLYMER FLOOD - RECOVERY FACTOR .45 W. F. M.O. datum, psi
WAIIR !NJfC!19?$. ........ .. ... ........ .. . . . .....=.s..~ ~.’”
q= injection rate, B/D
; zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA ~=
-%..lN,.,y”;E&@@’Q~-------------
,. ..,,, (STB or Mcf)/D/(unit source term,
a.’ .
NzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
volume)
J
o 500 Wo 1500 20D0 2500 3000 3500 4000 Qad, = mass of adsorbed polymer, pg/bbl of
TIME, DAYS pore volume
FIG. 19B- OIL RECOVERY RATE, POLYMER FLOOD. R, = resistance factor
Rk . permeability reduction factor
R,, = residual resistance factor
R. = solution gas ratio, Mcf/STB
‘e =
external radius, ft
Tw = well radius, ft
/
‘w = effective well radius, ft
s= saturation
s. skin factor due to well impairment or
stimulation
Sp =
pseudo skin to represent non-Newtonian
.2 - effects
u= superficial fluid velocity, cm/sec or
ft/day
.1 - y. shear rate, see-l
Ax, Ay, Az = dimensions of grid block, ft
(j, = fluid relaxation time, sec
0
0 .2 .4 .6 .8 1.0
PORE VOLUMES INJECTED TASi E 5 — RELAT:VE PERMEABILITY DATA
y
● PRODUCER
‘liquid k ro ~
0.25 z 0.55
FIG. 21 — FAULT BLOCK RESERVOIR WITH GAS CAP 0.99 0.85 0,00
AND AQUIFER. 1.OO 0,90 0.OO
f;; the use of his reservoir study. Solutions, ” ]. Pet. Tech. (March, 1971) 391-401.
22. Meter, D. M. and Bird, R. B.: “Tube Flow of Non-
REFERENCES Newtonian Polymer Solutions: Part I, Laminar Flow
and Theological Models, ” J. Am. Inst. Cbem. Eng.
1. Jewett, R. L. and Schurz, G. F.: “Polymer Flooding (1964) Vol. 10, No. 6, 878.
— A Current Appraisal, “ J. Pet. Tech. (June, 1970)
23. Lee, K. S. and Claridee, E. L.: “Areal Sweep
675-684.
Efficiency of Pseudopla;tic Fluid. in a Five-Spo_t
2. Slater, G. E. and Farouq-Ali, S. M.: “Two-Dimensional Hele-Shaw Model, ” Sot. Pet. Eng, J, (March, 1968)
Polymer Flood Simulation, ” paper SPE 3003 52-62.
presented at the 45th SPE Annual Fall Meeting,
Houston, Oct. 4-7, 1970.
3. Patton, J. T., Coats, K, H, and CO!e=O}W, G. T’. : .APDENllTY
. . . e.. ”... . A.
. . . . . . . . . . . . . . . . . .(A-2)
The radial velocity profile can be expressed as
At the end of a time step, the amount of polymer
avaiia”bie for adsorption, in micrograms per barrel ~~
of pore volume, is u =— —.. . . . . . . . . .(B-l)
2* r
6(polymer) = Es . . . . . . . . . .(A-3)
P The viscosity profile with the velocity profile of
Eq. B-1 is
while the adsorptive capacity of the rock in rhe grid
block is given by the maximum adsorptive capacity n-1
()
~
multiplied by the fraction of mobile aqueous phase rl-n
present in the gridblock: ~(r) = H — J
, 211h
I
Sw+s .Swc
c = AZ .(A-4)
/zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
1
ads IJ w(r) < IA .(B-2)
1 -s-s
Min ~
\ max “ “
/zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
orw Wc
1 ’7 :.L -- ~~~
~i jii St etloii~!l G$ t~~ poly,mer to The radii, r, and r~. where the viscosity reaches
J2.1 L1 lCL
H
_ . . . . .(B-3)
c - Qads > 6(polymer) , ‘1 =
ads w 211h
mi n
then
1
ti(sp)= s
P / H \n-l q — . . . .
/-,.
.( B-4)
and ‘2 =
P 2 rrh
6 (Q ads) H ma x
= 6(polymer)
The pressure drop from 12 to ~1 is
and if
qn
() -
H
c
ads
-Q ~d5 < 6(polymer)
P2 -pi. .—
211h k(l - n)
then
c -Q
ads (l-n)
6(SP) = ads r(l-n) . .(B-j)
and
E
●
✻
‘2 1
1
The pressure drop from rl to rw is
6(Q ~d5) = cad5 -zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Qads “
()
q IJmin rl
In both cases, pl-pw =-— —in—.
6(SW) = -6(SP) . 2rrh k r
w
{n <N
. . . . . . . . . . . . . . . . (u-u)
1
/\ qn H
P2-Pw=-
. [ in (u”N~l-v)) - u + l]) k(l - n)
\–1211h
(1 - n)
. . . . . . . . . . . . . . . . . . (B-8) ●
[
‘2
-#n)..
(l-n)zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQP
(B-12)
1 w
where
Pp . Ln N -Nw+l .
c1
max () PJ J
(1 - n) [
The condition that rl L Tw is sarisfied when . . . . . . . . . . . . . . . . . (B-13)
***