You are on page 1of 9

3yA7~

The Development, Testing, and Application Of a Numerical Simulator for Predicting Miscible Flood Performance
. ---W. J.

M. R. Todd, SPE-AIME, Shell Development Co. Longstafi, SPE-AIME, Sheij canada Ltd.

Introduction
When designing or evaluating a miscible displacement project, one would like to be capable of accurately forecasting reservoir performance for a variety of operating conditions. In the past, physical model * studies have provided the buiic of the data used in * these evaluations. Unfortunately, proper scaling of miscible displacement is difficult to obtain and model construction is time-consuming and expensive. Consequently, considerable effort has been devoted in recent years to the development of numerical reservoir simulators capable of predicting miscible flood performance. In this paper, a method is ciescfiDeci for modifying an existing three-phase simulator so that it may be used to simulate miscible flooding. The model employed is unique in that the essential characteristics of miscible displacement are described without reproducing the fine structure of unstable frontal advance. This feature is of particular importance as it allows the reservoir to be represented by a fairly coar$e numerical grid. We shall describe here three- and four-component versions of the miscible flood simulator. In the threecomponent version, wetting (water) and nonwetting (hydrocarbon) phases are considered with twocomponent miscible flow of oil and solvent in the hydrocarbon phase. With the four-component version, slug as well as continuous solvent injection may be examined. The validity of the miscible simulator is demonstrated by comparing predicted performance with experimental results from both linear and areal displacement in laboratory models. Predictions are also compared with reported results of miscible displacements carried out in a shallow, water-bearing reservoir. Finally, possible application of the simu,. . . 3..-.-. ., A-J L.. :.. ---- :- ALIa[or ]s aemonsmdteu Dy IL> usc m LIIC cviiIua LIuII UI alternative flooding configurations and operational policies for a field example.
-. ..-1..-.:--47

Simulator Requirements
A common characteristic of miscible recovery processes is unstable frontal advance, in the form of either viscous fingering or gravity tonguing. Tlnese itistabiiities are the natural result of the highly adverse viscosity ratio and large density difference that generally exist between an oil and the displacing solvent. Fig. 1, for example, depicts the swept zone at solvent breakthrough for a five-spot pattern flood as actually observed in a Hele-Shaw (parallel plate) model. The model is simulating the miscible displacement of oil by solvent at an adverse mobility ratio of about 15. The effect that an unstable frontal advance exerts upon oil recovery might be considerably worse than actually observed if it were not for the fact that solvent disperses in the oil, which renders the effective viscosity and density differences less extreme than those of the pure components. Thus, a successful miscible-flood simulator should allow for the possibility of unstable frontal advance and must describe the dispersion phenomenon, at least as related to the determination of effective fluid properties.

Here is a method for modifying an existing to forecast miscible flood performance. The essential features of miscible displacement miscible flow unresolved, making it possible coarse numerical grid.

three-phase simulator so that it may be used simulator is capable of modeling the while leaving the fine structure oj unstable to represent the reservoir by a jairly

Past simulation effort has been directed toward the numerical solution of those transport equations that specifically include the mutual dispersion of miscible fluids during flow in porous media. -9 Unfortunately, attempts at solutions based on these formulations have met with only limited success. This is partly because standard finite-difference techniques, such as those used in immiscible reservoir simulators, introduce a numerical dispersion into the solution, which tends to mask or distort the effects of the true physical dispersion of the solvent in the oil. 3 A more formidable obstacle, however, is that rigorous formulation of the problem requires that the fine structure of the unstable frontal advance be reproduced in detail by the simulation; otherwise, mixing cell effects will considerably dampen the simulated flow instabilities, resulting in optimistic forecasts of oil recovery performance. This requirement necessitates the use of impractically large numerical grids for simulating all but the simplest of reservoir systems. For example, the grid structure overlying the displacement shown in Fig. 1 contains 121 grid blocks. Since this is only one five-spot, this grid would be considered fairly large with respect to the requirements of a full-scale reservoir simulation; however, the grid is far too coarse to reproduce in detail the viscous fingering shown in the figure. Since the required size of a grid block is on the order of that of a finger width or smaller, about 10 to 50 times this number of blocks might be required for accurate numerical simulation fivr==nmt Aicl. nr.rn-+ of th~~ ~~~ob b.- ...Oyw. u&oylcL&ti,l,tiUL. We have found that the effects of dispersion on effective fluid properties can be adequately represented by means of an empirical model. Consequently, we find that accurate simulation of miscible displacement performance can be obtained without reproducing the fine structure of the flow. This, then, f,=ncihlta +hmothc.mot;n.ml .: . . ...1-+.--c t..ll ~z~.~c u .wbAaAuAe .Lti .L&a L1ltillla L1bal mL1lula L1uIl U1 lull-

scale miscible flooding projects.

Model Formulation
In this section we develop a model for two-component miscible flow for use in a three-component reservoir simulator. A four-component model, developed so that solvent slugs as well as continuous solvent injection might be investigated, is described in the Appendix. As the miscible formulations are merely variants of immiscible models, the actual solution of the conservation equations will not be discussed. Eqs. 1a through lC are the familiar expressions for the superficial velocities of water, oil, and gas used in th ree-pha~e re~en~ir ~i,mu!ater~ for the calculation of component fluxes. kk,,c u,~ = (VP,C flogVD),
/4.

.
.

.
.

(la) (lb)
(lC)

kk,o MO=
PO Ug =

(VP.

pogVD) ,

1%

(VP,

f%gv~)

Here, the relative permeability, k,,,, is a single-valued function of the water saturation, S,,, as is the difference in phase pressures, P,.t~O = P. P,,,. Similarly, the gas relative permeability and the capilla~ pressure, P~,,~ = pg PO, are single-valued functions of the total liquid saturation, (S,c + SJ. The oil relative permeability is normally modeled as a function of both the water saturation and the total liquid saturation. These functions of saturation are assumed to be known and are generally entered into a three-phase immiscible simulator in tabular form. For our purposes, gas is identified with a selected miscible drive agent (or solvent). Hence, oil and gas are considered miscible components of the nonwetting phase. Consequently, the capillary pressure E...+ha. iliGi~, Ske the ilKN~in~ilt Of Ori~ -vCOO is zerc. J w L1lGi component through a porous medium is not impeded by the presence of other components in the same phase except for a reduction in the area available for flQw,each of the rnkcih!e cmnponents asswnes a frrx= tion of the nonwetting phase relative permeability equal to its volume fraction in that phase. Thus, the relative permeabilities of the miscible oil and gas may be determined as follows:

k,o+. k,n, . . . . . . .
n

(2a)
@bj

s, 1. firg=-= s.

km,.

. . . . . . .

Fig. lSwept
pattern

zone at solvent breakthrough flood as observed in a Hele-Shaw model. M ~ 15.

for a five-spot

(parallel plate)

where in this case S. = S,, + S~ is the nonwetting phase saturation and k, = k,. (SJ is the imbibition ni=rmt=ahilitv nf t~e nnnmmttinm nh. s=. r~!ati~e ~-.. ----... -J . . . ...* ..w. bu,5 ~,,= b. For the development of the viscosity model, we consider the miscible displacement of oil by gas (solvent) in a two-diinefi~icmal system. It is assumed that a rectangular grid such as that overlying the displacement shown in Fig. 1 has been prescribed for numerical simulation. Fig. 2 depicts possible fluid distributions in an arbitrary grid block during frontal passage. The clear areas represent zones were the oil

JULY:

19?2

and solvent have been mixed by dispersion. Whether the displacement is stable or unstable, we must choose for the viscosity and density of each component a single value that reflects the effective fluid property averaged over the entire grid block. These values are required for the numerical calculation of component fluxes with Eq. 1. If the rate of dispersion of solvent and oil were such that the size of the dispersed zone is large with respect to the size of the grid block, the miscible components could be considered completely mixed in the block. Both components would have the same effective density and viscosity, as determined by appropriate mixing rules. However, if the dispersion rate were so low that the mixed zone is negligible with respect to the size of the block, the effective fluid properties of solvent and oil would be those of the pure components. It is reasonable to expect that the actual effective fluid properties will fall somewhere between these mixing limits. Eqs. 3a and 3b describe the effect of partial mixing on effective viscosities in terms of an empirical model suggested by K. S. Lee and E. L. Claridge of Shell Development Co.
POe = /LO*w/bLu , . . . . . . .

where M = mobility ratio p.~/;Lg,and Sg/sn S0/S,,. Rearranging, we have

= 1

As Eq. 7 is valid for any mixture viscosity, it is valid for mixing at the effective fractional saturations; thus

(-)
so zoe ()

so s,,

M% (po/poJ$~
M?A_l
Y . . . .

g. = _

(8a) (8b)

M% (p./poJ~ M~_~Y..

where the subscripts ge and oe are used to denote effective fractional saturations in the gas and oil, respectively. The effective densities may now be expressed by

oe= O(i_)Oe
[s. Pw=P.~~) n )
ge

Pg[1 - (*).,
1

ga)

Pv

[l-(*)

II

,e

].

@J

(3a) (3b)

=pn,w . . I.% = /%

A value of the mixing parameter, O, of zero corresponds to the case of a negligible dispersion rate. Note that the effective viscosities equal those of the pure components. A value of u = 1 corresponds to complete mixing within the grid block. Here, the effective viscosity of each component equals the mixture viscosity, which in our model is determined by the %-power fluidity-mixing rule, Eq. 4 or Eq. 4a.

Note that Eq. 8 is indeterminate for a unit gas/oil mobility ratio. In other words, no information can be obtained from the mixing parameter model for unit viscosity displacements. For these cases, we suggest the following alternative mixing model for the effective densities:

FLOW

DIRECTION -

(i) WY (WY 9 ~ 4) U )4. /(


l-k =
l-w%

a. STABLE DISPLACEMENT

s, s _Opo%+JLO S,j s.

(4a)

An independent model has not as yet been developed for evaluating the effective densities of the miscible components. However, a density model can be derived that is at least consistent with the viscosity model. The only assumption is that mixing is ideal. If the gas and oil were completely mixed in a twocomponent miscible system, they would have the --A-.; +., a> h., Gn < o. rlo+-.-~--~ aau16 uwmILy, UkLWLInLLLL. U Uy J+. d.
F%= PO*$+P9 n

~DISPERSED
FlnW ------nlRFf .. . . TinN . .. . .

ZONE

G4

;... u

(5)

b.

UNSTABLE

However, as with the component viscosities, we will assume that the fluids are only partially mixed or, alternatively, that the fluids are completely mixed, but at the effective saturation fractions that would result in the effective viscosities derived by Eq. 5. To determine the necessam effective saturations in the miscible tluids, the mixing rule (Eq. 4) is expressed alternatively as
Pm =
P.. /[

DISPLACEMENT

~
Fig. 2Schematic

DISPERSED

ZONES

M%

(MM1); n 1

((j)

representation of solvent displacing in a grid block.

oil

876

JOURNAL

OF PETROLEUM

TECHNOLOGY

poe=(l pve=(l

*J)po+~p,,i o))pg+copm

, .

. .

. .

. .

. .

. .

(lOa) (lOb)

using fluids of equal density. Simulation of Physical Model Performance Fig. 3 depicts the recovery performance for a linear displacement of oil by solvent at an oil:solvent viscosity ratio of 86. The experimental data are the reported results of Blackwell et al.fi for a horizontal, two-dimensional linear displacement in a sand pack. The computed results were determined using onedimensional simulations. Grid requirements will be discussed in a later section. Note that the computed results are quite sensitive to the value of the mixing parameter, and that good correspondence between the experimental and computed recovery performance is obtained with ., = 2/3. The mixing parameter model would be of limited utility if it had to be calibrated for each displacement; however, as shown in Fig. 4, we have found that one value of w can be used to predict oil recovery performance for this system over a wide range of unfavorable mobility ratios. To evaluate the areal performance of the simulator, computed resuits were compared with experimental data for displacements carried out in a quarter fivespot symmetry element. As shown in Fig. 5, the predicted oil recovery performance is in good agreement with the reported experimental data of Lacey [1 ail; for mobility ratios of 10. 40, and 85. The computed results were obtained using a two-dimensional grid and a mixing parameter of 2/3. Fig. 6 also depicts the recovery performance for a five-spot displacement, but all data shown here are for a mobility ratio near 10. Note that the recovery curves determined from the three experimental investigations are considerably different. Lowest production performance reported is for a Hele-Shaw

In an immiscible simulator, values of water, oil, and gas relative permeabilities, viscosities, and densities, together with water/oil and oil/gas capillary pressures are normally taken from tables of input data as mentioned previously. For miscible flood simulation, these values are modified by Eq. 2 for oil and gas relative permeabilities, Eq. 3 for viscosities, and either Eq. 9 or Eq. 10 for densities. The only additional input variable required is the value of the mixing parameter, o). In addition, the input values of the capillary pressure between the oil and gas, PCOO, must be set equal to zero. This completes the derivation of the three-component model. An analogous derivation for a fourcomponent model, intended for the study of miscible slug flooding, is presented in the Appendix.

Program Testing Simulation of Experimental Data l-h. -.,4,.1 di%~tibd 1llG llluuCL iii tk ~~~@di~fj

~i?di~rl

for

evaluating apparent miscible fluid viscosities, densities, and relative permeabilities has been incorporated into Sheii Development Co.s compressible, three-phase, three-dimensional, black-oil reservoir simulator. To test the efficacy of the resultant miscible simulator and to determine the appropriate magnitude of the mixing parameter, ~),simulator predictions were compared with results from physical model experiments reported in the literature. -I ~In addition, predictions were compared with reported results of miscible floods carried out in a shallow, watersaturated reservoir.] All results are for two-component, miscible displacements in horizontal systems

//
G

.7=2
QJ ,

T
r

1/2

1 -Y
G

M=86

/M=150

EXPERIMENTAL

G EXPERIMENTAL

COMPUTED

COMPUTED

LINEAR M=B6

LINEAR (lJ= 2/3

{, 0

u
VOLUMES

1 1 1.0
PORE VOLUMES OF SOLVENT 2.0 INJECTED for

0 PORE Fig.

1.0 OF SOLVENT

2.0
INJECTED
performance

3Linear

miscible displacement for M = 86.

Fig. 4-Linear miscible displacement performance various unfavorable mobility ratios.

JULY,

1972

877

(parallel plate) model. Performance is better for the unconsolidated bead-pack model,l; and is best for an artificially consolidated sand pack.1~ The variation in performance is thought to result from differing rates of transverse dispersion of the solvent in the oil for the three experimental models. Dispersion is held ~l~Q~tfi w.., .- th~ ...- rn.ojecu]ar !eve! in the He!e-Shaw model. With increasing pore complexity, convective dispersion increases. This increase results in wider mixed ~-. ~ild h~ii~~ idid~d t?4LGW U ~~rles bringing about improved recovery performance. In the numerical simulator, the influence of transverse dispersion on effective fluid properties is specified by means of the mixing parameter. Thus, it is in accord with the reported experimental results that the computed oil recovery performance improves with increasing values of w Note that all of the experimental data shown in Fig. 6 are bracketed by the computed recovery curves for the mixing limits of ~ = O and ,,, = 1. Further note that if we do not take into account the mutual dispersion of solvent and oil as, for example, when attempting to use an immiscible reservoir simulator directly to forecast miscible flood performance the predicted oil recovery would probably be pessimistic, as indicated by the computed curve for ~,,= O. On the other hand, if we use a simulator that explicitly includes diffusive mass transfer and do not reproduce unstable frontal advance in the simulation, the predicted oil recovery will probably be optimistic, as indicated by the computed curve for o, = 1. We he!iev~ the Hele-Shaw model to be most closely scaled to reservoir conditions for miscible displace#Tf.+1..l V & ..; CP* V I~&&.y it., TQt;rl , C&U, G For excellent descriptions of the requirements and tiiiiicu ities associated with the construction of scale models for studying miscible displacements. !JIease refer to tJa Ders by Pozzi and Black. well! and by Heller. ]g

ment;* consequently, we believe that the appropriate value of OJ to be used for miscible flood simulation is on the order of 1/3. Simulation of Reservoir Performance The Chandler Experiments As a further test of the miscible simulator, the Chandler experiments were simulated. These experiments were miscible floods carried out in a confined five-spot
~Q@lJ~~~~Q~ ~~ ~ Sha]]ow water-bearing reservoir.

Adjustment of the brine viscosity with Dextran allowed the investigation of two-component miscible floods having mobility ratios of 0.1, 1.0, and 10.0. The experiments were conducted by Esso Production Research Co. and are fully described in Ref. 15. The Iithology of the Chandler sand is described in Refs. 16 and 17. Note that although thickened brines were actually used for the Chandler experiments, for clarity the resident fluid will be referred to here as oil and the invading fluid as solvent. An initial investigation indicated that introducing into the simulator the measured areal variation of transmissibility at Chandler would have little effect on the computed flood performance. Consequently, a uniform quarter five-spot symmetry element was used for the remainder of the study. Fig. 7 shows the oil recovery performance for the unit mobility ratio displacement in the Chandler reservoir. Note that because of the vertical reservoir heterogeneity, the observed recovery performance is considerably poorer than that predicted for a homogeneous five-spot. Our computed recovery curve for a homogeneous reservoir. by the way, is in good agreement with the analytical solution of Morel13e~tow,19

To include reservoir heterogeneity in the simulations, the method of Hearn was used to generate U=l. o
5-SPOT

fAJ %3
~:1/ 3

QJ, ()

G EXPERIMENTAL COMPUTED

7
E VOLUMES

5-SPOT ~ . 2/3 COMPUTED


HELE-SHAW BEAD PACK 5AiND PACK MODEL

I
I
2.0
INJECTED Fig. 6--Miscible

G
=

CO NS.0Lii2ATEi2

1 1.0 OF SOLVENT

PORE VOLUMES

I 1.0 OF SOLVENT

2.0
INJECTED

Fig. &Miscible displacement performance in a confined, five-spot pattern for various unfavorable mobili~ ratios. 878

displacement performance in a confined, fivespot pattern for M ~ 10.

JOURNAL

OF PETROLEUM

TECHNOLOGY

pseudo oil/solvent relative permeability curves (Fig. 8) for the permeability distribution of the Chandler sand.; Use of those pseudo relative permeability curves describing Lithotype A, thought to be the dominant lithotype at Chandler, resulted in the best agreement between the computed and the observed reservoir performance as shown in Fig. 7. The mixing parameter model, of course, has no influence on these results because the viscosities of the oil and solvent are equal. Fig. 9 shows the experimental and computed oil recovery performance for the unfavorable mobility ratio displacement at Chandier. H-ere, the pseudo relative permeability curves that describe Lithotype A were used for all three simulations. Use of ,, = 1/2 yields the best agreement between computed and -,--- --J omerveu reservoir perfOMiatlW. lk3G3:i tii~i gd correspondence between experimental and computed results is obtained with (,J = 2/3 when forecasting miscible flood performance in a laboratory sand pack or bead pack model, and that we believe 6) = 1/3 is the approximate value to be used for simulating full-scale displacements (say 10 to 40 acres/well). Because of the smaii pattern size (0.03 acre/weiij and relatively low injection rate, scaling considerations dictate that the performance of the Chandler flood lie between that of a sand pack model and that of a full-scale reservoir flood. Thus. the value of ,,) = 1/2 found here to give the best correspondence between computed and observed Chandler performance seems reasonable. It should be noted, however, that Hearns model for pseudo relative perrneabilities is strictly valid for unit mobility ratio displacements only. It is possible, therefore, that use of the Hearn method in simulating the unfavorable Chandler flood is inappropriate.

Z?:al,.1 A lGIU

Appllwluull

A--l:nn+:n

-F

WI

tha

Lll&

LrmLci&lul&

M:. n;hla

c;m*#l!3+n*
uxllluLu

Lwa

To demonstrate possible applications of the miscible simulator, we examine briefly the prospects for high-pressure gas miscible flooding of a watered-out oil r,, n. nela. rrlmary depietiori, together with a iiigiii~ siiCcessful peripheral line drive waterflooding operation, is assumed to have reduced the oil saturation to waterflood residual. Laboratory testing has confirmed that the high-pressure gas will become sufficiently enriched at reservoir temperature and pressure to miscibly dispiace practically aii of the oii contacted. The gas:oii mobility ratio is 16 and the immiscible gas:water mobility ratio is 70. The viscous: gravity-force ratio for the displacement is large enough so that gravity -... -.:.J - ul -c &d> -- uum d--- uut --- . . . ..-. -s..+ . . ..a L1...111 auul.-. AA; uvcllluc plcxu Ld pIUUIGIIi. T.. tion, the reservoir is practically uniform and isotropic. Consequently, simulations were carried out using a two-dimensional areal grid. A mixing parameter of 1/3 was employed. A peripheral line drive was considered the prime candidate for the flooding configuration because of the success of the writer_i300dittgOperadOii. lHOW~V~i, laboratory study has indicated that pattern flooding could be more efficient for highly unfavorable displacements. Consequently, an inverted, confined nine-spot pattern was also selected for study. For the line drive, gas is injected into wells adjacent to the reservoir boundary. and two or three downstream wells are put on production. The injection of high-pressure gas effects an immiscible displacement of the water injected during the prior watetiooding operation while reconnecting and banking up the waterflood residual oil. A large portion of this oil is recovered as the oil bank is swept downstream past

HOMO.

/
A
1.0

A+B 0.8 -

LITHOTYPES A68

EXPERIMENTAL COMPUTED

~ m $ 0.6 s a w o. u : 0.4 a u ~

krs
HOMOGENEOUS RESERVOIR

~HANDl&R

M=l 0.2 /

()
o 1.0

I o 0.2

2.0
INJECTED ratio

0.4
OIL

0.6
so

0.8

to
from

PORE VOLUMES

OF SOLVENT

SATURATION,

Fig. 7Recovery performance for the unit mobility displacement in the Chandler reservoir.

Fig. 8-Pseudo relative permeability curves derived the vertical permeability distribution of the Chandler reservoir.

1111 V

1q77

x79

the production wells. When production from a given producer reaches an economic GOR limit, the well is shut in and the next well downstream is added to the line of producers. For an inverted, confined ninespot flooding pattern there are effectively three producers surrounding each injector. As these are nearly equidistant from the injector, all may be expected to simultaneously produce displaced tertiary oil even though their behavior will not be identical. Some results of the numerical evaluation of these flooding schemes are shown in Figs. 10 and 11. A comparison of the predicted recovery efficiencies for the line drive and the confined nine-spot is given in Fig. 10. These results show that the nine-spot pattern flood is clearly superior, recovering almost 70 percent of the tertiary oil before reaching the economic limit. Predicted recovery for the line drive, on the other hand, is less than 45 percent. Once a preferred flooding pattern has been determined, alternative operational schemes may be evaluated using the numerical simulator. For example, continued gas injection in the confined nine-spot until the economic limit is reached results in a recovery of 1.3 million STB of oil in 2,000 days under the field conditions as shown by Curve a. Fig. 11. (The pattern area originally contained almost 1.9 million STB of tertiary oil.) Blowdown of the reservoir to recover a portion of the gas remaining after the miscible flood might recover an additional 160.000 STB of oil. if blowdown is effected through the comer production well, Curve c. Alternatively, water injected after 1,200 days of gas injection results in only 1 million STB of oil recovered, Curve d. The results of various simulator predictions can now be evaluated economically to aid in selecting the optimum flood design. In addition to assisting in the selection of a preferred flooding scheme and operational policy, the simulator can be further employed for the design and subsequent surveillance of both pilot operations

and fieldwide expansions.

Program Limitations
In evaluating the applicability of the miscible displacement model presented here for a specific reservoir problem, it is important to recognize that this model does not represent an attempt to describe rigorously the thermodynamic and transport phenomena that determine the details of the local fluid composition and flow characteristics. Instead. it is a computational expedient that allows for the engineering evaluation of miscible flood performance. Even so, the explicit and implicit assumptions inherent in the model do restrict to a certain degree the practical application of the miscible simulator. We shall describe those limitations with which we are familiar. The model does not include diffusive flow between grid blocks due to spatial gradients of concentration. Consequently, it is not applicable to systems characterized by dispersed zones perpendicular to the local flow vector, which are large with respect to the size of the grid blocks used for the simulation. This drawback might be particularly important in modeling highly stratified reservoirs (with communicating layers) using multilayer grid systems. A combination of the miscible and immiscible simulator analogy described by Lantzt and the model presented in this paper might result in a miscible simulator without this limitation.

-1

GAS

lN,

fC140N

lHVD@0CAR80N

POtf

VOLUME)

Fig. 10-Comparison of predicted tertiary oil recovery performances for peripheral line drive and inverted nine-spot flooding patterns.

-1600

1
[b] [.)

(. I CONTINUOUS ~LOWDOWN BLOWDOWN

GAS

INJCC11ON ORIGINAL CORNE@ INJECTOR PRODUCER

THROUGH lMROUGH

[d) WAIEt INJECTIONFROM 1200DAYS PROM 700 DAYS l.) WATtt INJEC1OON

CHANDLER M=IO 1
1.0

1
2.0
INJECTED

PORE VOLUMES

OF SOLVENT

.o~

400

Km

IIW low,

Ibm DAM

mm

2400

00

Fig. 9-Recovery performance for the unfavorable mobility ratio displacement in the Chandler reservoir. 8R0

Fig. 1 lTertiary oil reeovery from an inverted ninespot pattern.

JOURNAL

OF PETROLEUM

TECHNOLOGY

The model does not include mass transfer effects; oil and solvent are assumed to be completely miscible on first contact. This treatment might seem to p,reclude the use of the miscible simulator for forecasting displacements where multiple contacts are required before miscibility is obtained; for example, in condensing or vaporizing gas drives. However, so long as the distance required to obtain miscibility is short with respect to the well spacing, the simulator should be adequate for determining oil recovery performance. The model inherently assumes that the invading solvent has equal chance of contacting all of the oil in a grid block. Consequently, the model may not be applicable to systems exhibiting strong gravity segregation where an attempt is made to represent the system with a single layer of grid blocks by partial analytical integration of the flow equations. 2 If, for this system. the oil saturation is at waterflood residual when the miscible displacement is started, Eqs. 2 through 4 will exaggerate the influence of the oil properties in attenuating solvent mobility. This will result in optimistic forecasts of oil recovery performance. Although this paper does not treat the actual numerical techniques used in a simulator for solving the conservation equations. some observations pertaining to the selection of the numerical grid are warranted. It is well known that the normal nlanifestation of truncation error in finite-difference-based numerical simulators is a nonphysical dispersion at the flood front that results in an apparent smearing or degradation of the saturation profiles. For a specified time-step size the magnitude of the truncation error decreases with decreasing grid spacing and increases with increasing frontal velocity and increasing saturation gradient at the front. Thus, it requires fewer grid blocks to simulate accurately a
hiphlv . .. C...=
low

mobility ratio (viscosity ratio for miscible fluids) ~= pressure P= C!ro gas-oil capillary pressure P CKO = water-oil capillary pressure s= saturation u= superficial fluid velocity /.t = viscosity = density P 1,) = mixing parameter gradient operator v= Subscripts e= g=

M.

effective gas in four-component simulator, solvent in three-component simulator yn = mixture n= nonwetting (hydrocarbon) o = oil s = solvent w= wetting (water)

References
1. Pozzi. A. L. and Blackwell, R. J.: Design of Laboratory Models for Study of Miscible Displacement, SOc.
.2.

P@f ..-.

F,,

. (J.

j.

(March,

]963

2!3-40.

Lantz. R. B.: -Rigorous Calculation of Miscible Displacement Using immiscible Reservoir Simulators. Sot. Per, ,51AI. J. (June, 1970) 192-202. 3. Lantz, R. B.: Quantitative Evaluation of Numerical Diffusion (Truncation Error) , SOC. Pet. Eng. J. (Sept., 1971 ) 315-320. 4. Peaceman, D. W. and Rachford. H. H., Jr.: Numerical Calculation of Multidimensional Miscible Displacement, .Soc. Per. Eng. 1. (Dec.. 1962) 327-339. 5. Blackwell, R. J., Rayne, J. R. and Terry, W. M.: Factors Influencing the Efficiency of Miscible Displacement. Trans., AIME ( 1959) 216, 1-8. 6. Garder. A. O.. Peaceman. D. W. and Pozzi, A. L.: Numerical Calculation of Multidimensional Miscible Displacement by Melhod of Characteristics, Sot. Per.
F,, r> #.-rr A. . 1 .

~nfav~ra~!e

n~~~ilitx, ... ..

ratin

.-...

Aicnlactmm-nt (.hnl. CI, . ..oy.utiti fifi. b., t {DA,

[M.

, 1.1-IL!.

--h

10 ,,

CA\ -?,

?L.lz= A

.JU.

saturation gradient) than to simulate one whose mobility ratio is near unity (steep saturation gradient). The computed results shown for a linear system in Figs. 3 and 4 were obtained using a one-dimensional grid. The results for mobility ratios of 86 and 150 exhibited no sensitivity to the grid spacing for grids comprising 10 elements or more. However, it took 40 grid elements to obtain a convergent solution for the M = 5 displacement. This grid requirement still compares favorably with the two-dimensional, 800point mesh used by Peaceman and Rachford to obtain detailed simulations of linear sand-pack displacements. For our quarter five-spot simulations. a two-dimensional, 5 X 5 point grid was adequate for the AI = 40 and M = 85 displacements; however, a 7 < 7 point grid was required for the M = 10 displacements. A 10 X 10 point grid was required for simulating the unit mobility ratio displacements shown in Fig. 7.

Nomenclature
D = depth, measured positive downward from

sea level
k = rock permeability

k, = relative permeability-component fied with additional subscripts


JULY. 1972

speci-

7. Perrine, R. L. and Gay, G. M.: Unstable Miscible Flow in Heterogeneous Systems, SOc. Pet. Eng. J. (Sept., 1966) 228-238. 8. Price, H. S. and Donahue, D. A. T.: Isothermal Displacement Processes with interphase Mass Transfer, Sot. Pet. Eng. J. (June, 1967) 205-220. 9. Chaudhari, N. M.: An Improved Numerical Technique for Solving Multidimensional Miscible Displacement Equations, Sot. Pet. ErIR. J. (Sept., 1971) 277-284. 10. Koval. E. J.: A Method ~or Predicting the Performance of Unstable Miscible Displacement ;n Heterogeneous Media, SOC. Pef. Eng. J. (June, 1963) 145-154. 1. Brigham, W. E., Reed, P. W. and Dew, J. N.: Experiments on Mixing During Miscible Displacement in Porous Media, SOc. Pd. Eng. J. (March, 1961) 1-8. 2. Mahaffey, J. L., Rutherford, W. M. and Matthews, C. S.: Sweep Efficiency by Miscible Displacement in a FiveSpot, .SOC.Pet. Ew. J. (March, 1966) 73-80. 3. Lacey. J. W., Faris, J. E. and Brinkman, F. H.: Effect of Bank Size on Oil Recovery in the High Pressure GasDriven LPG-Bank Process. J. Pet. Tech. (Aug., 1961) 806-816. 4. Habermann, B.: The Efficiency of Miscible Displacement as a Function of Mobility Ratio, Trans., AIME ( 1960) 219, 264-272. 5. Greenkorn, R. A.. Johnson, C. R. and Haring, R. E.: Miscible Displacement in a Controlled Natural System, j. Per. Tech. (Nov.. 1965) 1329-1335. L6. Greenkorn, R. A., Johnson. C. R. and Shallenberger, L. K.: Directional Permeability of Heterogeneous Anlsotropic Porous Media, Sot. Per. Eng. J. (June. 1964) 115-123. 17. Johnson, C. R. and Greenkorn, R. A.: Description of 88 I

Gross Reservoir Heterogeneity by Correlation of Lithologic and Fluid Properties from Core %mpie, Buii., Intl. Assoc. Scientific Hydrology ( 1963). 18. Heller, J. P.: The Interpretation of Model Experiments for the Displacement of Fluids Through Porous Media, AIChE Jour. (July, 1963). 19. Morel-Seytoux: H. J.: Unit Mobility Ratio Displacement Calculations for Pattern Floods in Homogeneous Medium, Sot. Pet. Eng. J. (Sept., 1966) 217-227. 20. Hearn, C. L.: Simulation of Stratified Watertlooding by Pseudo Relative Permeability Curves, J. Pet. Tceh.
(Jdy, 1971) 805-813. 21. Coats, K. H., Nielsen,

R. L., Terhune, Mary H. and Weber, A. G.: Simulation of Three-Dimensional, TwoPhase Flow in Oil and Gas Reservoirs, Sot. Pet Eng. J. (Dec., 1967) 377-388.

APPENDIX

the dispersed zones. As the dispersed zones will generally be small relative to the grid biock size, an approximate IOW value (say, S,/S,, = 0.01) should suffice for simulation purposes. Thus, when the solvent saturation at a grid point falls below this critical value, the miscible model discussed below is bypassed. The remaining solvent is assumed immobile, and the oil and gas are treated as immiscible phases. Thus, the capillary pressure and the component viscosities, relative permeabilities, and densities are taken from tables of input data in the normal fashion. While miscibility is maintained, the relative permeabilities for miscible oil, solvent, and gas are given as follows: k,. =$*k,%,
n krg=~% n ,

Four-Component

Model
.
.

In order that miscible slug displacements as well as continuous solvent injection may be examined, the equations developed for two-component miscible flow may be extended to allow for three-component miscible flow in the nonwetting phase. Here it is assumed that a slug of solvent (denoted by subscript s) is miscible with the oil and with the gas driving the s!yg. The drive gas, however, is not miscible directly with the oil. Water is again considered the wetting phase. The four-component model requires the addition of a fourth conservation equation to an existing three-phase simulator. However, if no mobile water is present in the system being simulated, a three-phase simulator could be modified so that the conservation equation for water is used instead for the solvent. So long as miscibility is maintained, oil, solvent, and gas are all considered miscible components of the nonwetting phase. The capillary pressure Pc~o is thus zero as for the three-component simulator. However, when the oil-solvent and solvent-gas dispersed zones contact (note Fig. 12) miscibility is lost. This occurs when the local solvent saturation drops below a value sufficient to maintain both dispersed zones. This critical value will vary with both time and space in a manner consistent with the growth of

.
.

.
.

.
.

.
.

.
.

.
.

(A-la) (A-lb) (A-lc)

k,, = 5*
n

k,,,,

where, now, S. = So + S. + So and the nonwettingphase reiative permeabiiii~, k,,,, k again assumed % be a known, single-valued function of the water saturation. For the application of the mixing parameter model, we assume that oil viscosity and density are modified only by the presence of the solvent, as are the viscosity and density of the gas. The viscosity and density of the solvent, however, are modified by both the oil and the gas since the solvent shares dispersed zones with both of these components. Thus, in analogy with Eqs. 3 and 4, we have:
Poe =

pol-u ,L1.,,,08~ , .
P8-J/1,,,~
l.(c-wp,,lv

, ,
.

. .
.

. .
.

. .
.

. .
.

(A-2a) (A-2b) (A-2c)

P.e

,
,

.
.

pg. =

F1OW DIRECTION

.
pmsg = hpg

.
7j--89

.
p.%

(A-3a)

/(
.

s. G
. .

pv%+-.

s,
s 89

4 ,

. Pm =
-

(A-3b)

P0P8P9

....

/(

so
~

s, ~~%pgti + s.

..e~. lmmlaL,

..c eLK

OISSIACEMENT OF OIL BY GAS

Eq. A-3c is the M-power fluidity-mixing rule for three miscible hydrocarbon components. The effective densities for the three miscible components may be calculated in a manner analogous to that described in the body of the paper. -T
f-GAS/WLVENT
DISPERSED ZONE

so~E~T/o,~ DISPERSED ZONE

Original manuscript received in Society of Petroleum Engineers office July 23, 1971. Revised manuscript received Feb. 29, 1972. Paper (SPE 3484) waa presented at SPE 46th Annual Fall Meeting, held in New Orleans, Oct. 3-6, 1971. 0 Copyright 1972 American Institute of Mining, Metallurgical, end Petroleum Engineers, Inc. This paper cover 1972. will be printed in Tranaactiona volume 253, which will

Fig. 12Schematic representation of a three-component, miscible displacement in a grid block.

882

JOURNAL

OF PETROLEUM

TECHNOLOGY

You might also like