You are on page 1of 14

5?

9305

Fractured Reservoir Simulation


~. Kent Thomas, SPE, Phillips Petdeum Co. Thomas N. Dixon, SPE. Phillips Pctrdeum Co. Ray G. Pierson, SPE, Phillips Petmieum Co.

Abstract
This paper describes the development of a th~edimensional (3D), three-phase model for simulating the flow of water, oil, and gas in a naturally fractured reservoir. A dual porosity system is used to describe the fluids pment in the fractwes and matrix blocks. Primary flow in the resezvoir occurs within the fractures with local exchange of fluids between the fractute system and matrix blocks. The matrixlfmctwe transfer function is based on an extension of the equation deveioped by Warren and Root and accounts for capihry pressure, gmvity, and
viscous forces.

Both the fmctwe flow equations and matrix/fracture flow are solved implicitly for pressure, water salutation, gas saturation, and salutation pressure. We present example problems to demonstrate the utility of the model. These include a comparison of our tesults with previous tesuk.s: comparisons of individual block matrix/fractute transfers obtained using a detailed 3D grid with results using the fractwe models matrbdfractm transfer function; and 3D field-scale simulations of two- and thee-phase flow. The thteephase example illustrates the effect of free gas satumtion on oil recovery by watertkmdhtg.

Introduction
Simulation of naturally f!actumd reservoirs is a chaHenging task from both a nzservoir description and a numerical standpoint. F1OW fluids through the reseof rvoir primarily is through the high-permeability, loweffective-porosity fractures summnding individual matrix blocks. The matrix blocks contain the majority of
0197-7s20/s3m02-9wsso0.zs Copynghl19S3SOCNIY 01Pewdaum Erg-

d AIME

the ~servoir PV and act as source or sink terms to the fMCNIW. The rate of recovery of oil and gas from a fractured mervoir is a timction of several vtiables, including size and properties of matrix blocks and pmssum and saturation history of the fmcture system. Ultimate recovexy is influenced by block size, nettability, and pressure and saturation history. Specific mechanisms controlling matrix/fracture flow include water/oil im,.. -. . ..* :L: .----- ~ttw UII .4..,,; s a..- %L-I e~--./.-.:1 *A ..-.= Cnoltlon, oil lmuibhiuu, UUW86W, pansion. The study of naturally fmctured reservoim has been the subject of numemus papers over the last four decades. These inciuue laboratory iiivestigations d oil tecovery fmm individual matrix blocks and simulation of single- and multiphase flow in fractured resemoim. Wanen and Root i presented an analytical soiution for single-phase, unsteady-state flow in a namrally fmctwed resemoir and introduced the concept of dual porosity. Their work assumed a continuous uniform fracture system parallel to each of the principal axes of permeability. Superimposed on this system was a set of identical tectartgtdar parallelepipeds mpmsenting the matrix blocks. Manax and Kyte2 presented experimental results on water/oil imbibition in labomtory core samples and defined a dimensionless group that relates ~covery to time. This work showed that recovexy time is proportional to the squa~ root of matrix ptmneability divided by porosity and is inversely pmpottional to the square of the characteristic matrix i~iigtt. Yamamoto et af. 3 developed a compositional model of a single matrix block. Recovery mechanisms for various-size blocks surrounded by oil or gas wete studied. Gas/oil capillary pressure hysteresis was inciudSOCIETY OF PETROLEUM ENGINEERS JOURNAL

42

Conclusions
The

new mathematical model can provide useful information on the design and operation of sucker pumping installations, especially in the case where the tluid physical properties of viscosity and inertia ate important. These properties have been shown to exert a significant effect on plunger stroke, peak and minimum polishedrod loads. polished-rod horsepower, and peak-plunger !oad. The model also is useful in predicting the effects that different choices of tubing or rod sizes would have on the viscous damping of the pumping system.

s= sp = SRL = [=

Nomenclature
Ac = td coupling area, L2
Ap = plunger area, L2

polished-td stroke, L plunger stroke length, L length of one sucker md. L time, t Vf = fluid velocity. Lt - Vr = rod velocity. Lt - x= depth below polished rod, L h, = fluid friction factor AZ = rod coupling friction factor w = fluid viscosity, mL -t - Pf = fluid density, mL3 P, = md density, mL3 O = ctank angle, radians

References
lle Technology of Artificial Lsfi Merhads. Petroleum PublishingCo., Tulsa (1980) 2a. 1-94. 2. Gibbs, S.G.: A Genend Method for predicting Rod Pumping SystemPerfomranee. paperSPE 6850 presented the SPE 1977 at Annual Technical Conferenceand Exhibition. Denvq, Oct. 9-12. 3. Gibbs, S.G.: predicting the Behavior of Sucker-Rod Pumping Systems, J. Per. Tech. (July 1%3) 769-78. 4. Adantache. I.: Analytical Methods for Computer Diagnostic Analysis of Sucker-fhl pumping Well Performance, Bolferlin~
1. Brown. K. E.: deila Associazione hlinemria Subtdpina Anno X111(~mo-giugno

A, = rod atea. L* A, = tubing area, L2 Dc = coupling diameter, L D, = rod diameter. L D, = tubing diameter, L E= modulus of elasticity, mL -1 t2 fr = rod tension, mLt2 Fc = viscous fotce per rod coupling, mLt2 F@ = viscous fotce per rod coupling per unit length of rod, mt2 FJ = viscous force per unit length of tubing, mt2 Fti = viscous fotce per unit length of rod, mt2 Fm = frictional fotce between tubing and rods per unit length of rod. mt2 F. = gross plunger load, mLt2 F, = peak polished-rod load, mLt2 F2 = minimum polished-rod load, mLt2 = gmvitational constant, Lt2 k: = spring constant for md string, mt2 L = length of sucker rod string, L L* = pump plunger location teiative to the casing, L L, = ctank length, L L2 = fixed-bar length, L L3 = driving-bar length, L L4 = pitrnan length, L L5 = driven-bar length, L N = pumping speed, t- NRe = Reynolds number for the fluid NR., = Reynolds number fOr the rods No = natuml frequency of the md string, t- Pc = casing ptessum at plunger level, mL -t -2 Pf = fluid pressutv, mL -1 t2 Pth = tubing head pressure, mL - t2

5.

6.

7.

8.

10

1976) n. 1-2, 65-92. Adarnache,I., er al.: Reseatdws on DeepWell Pumping Made by Means of art AnaiogueComputer.paper presentedat the 8th World Pet. Cong.. Moscow. 1971. Vaieev. M.D. and Repin, N. N.: IZVESTIYA VYSSHIKH VCHEBNYKH ZAVEDENII, NEml GAZ.( !976) 8.394$. in Russian. Gray, H. E.: Kinerrsatics f Oil-Well Pumping Units, paper o presented at the 1%3 API Mideontinerrt District Meeting. Amarillo, Mamh 27-29. Gibbs, S.G.: Computing Gearbox Toque and Motor Loading for Beam Units Wklr Considerationof Irreitia Effects. J. Per. Tech. (Sept. 1975) 1153-59. Ames. William F.: Numencal Methods for Parrial Dr~erenrial Equations, Barnesand Noble hsc.. New York City ( 1%9) 165-86. Watters, Gary Z.: Modem AM@is and Conrroi of unsteadyFIOW
m

.-

f-fpefmes. msn

.,..

_ . A..
Amor

aacikm

e.:-

--- n.. Ll:. L-

ruuuxtcis

k.,

A... A&. -,lS1 ~....

MI .. . .

47-82. 11 API Recommendedpractice for Design Calculations for Sucker Rod PtsmninrrSvstertss(Conventional Units) API Rp 11L. Dallas. Feb.197;.

( 1979)

S1 Metric Conversion Factors


141.5/(131 .5+ API Cp x 1.0* ft X 3.048 in. X 2.54* Ibf X 4.448 222 Conwfsian Is exact faoror API) E-03 E-01 E+OO E+OO = = = = = g/cm3 Pans m cm N

SPEJ

Wlml mmwaerim mcawad in.sociOfYP.rrofeum of Enqmaara Orrieeufy18. lam J PSW aamg.rad forPubfkmm May 11,1582.Rev- manusenpr NCS@ 15. 1982. Papar (SPE10249) first praaenradt tfra1881 SPEAnnual eclmieal a T Con~wenca ndExrsibition in Sm Anfonii Oer.5-7. a IsOM

FEBRUARY

1983

41

~d to ~ccQuntfor the effect of ttapped gas on recovery. Braester4 gave an analytical solution for lD, horizontal flow of two immiscible fluids through a fractured porous medium. Capillary pressure was assumed equal to zero in the fracture and nonzero in the matrix. The effect of pressure drop across a matrix block was included in the attdysis. Previous multiphase, multidimensional simulations of natumlly fractured resemoirs include papers by Kazemi et af. 5 and Rossen. 6 Kazemi treated two-phase water/oil flow using block Gauss-Seidel to solve iteratively first for fracture pressunx and water saturations using last iterate matrix pressures and water satumtions. Next matrix pressures and water satumtions were calculated using last itenwe values of fmcture pressures and water saturations. This process was repeated until convergence was teached. No gravity terms were included in this work for matrkdfractme flow and they used the same fmctu~ capiilaty pmssute and upstteam relative pemneability for both matrix/fracmre and fmctum flow. Twodimensional (areal) water/oil examples were included. Rossens model used input recove~ ctuves that wete a function of p~ssure, matrix block pmpetties, and the fluid enviromnent of the fmcttue system. These recovery curves we~ differentiated with respect to pressure and salutation and then were inciuded in a sequetttiai p/Sw -Sg simulator as semi-implicit source terms. Complete segregation of fluids was assumed in the fracture system. Matrix blocks surrounded by gas were assumed not to tmnsfer water from the matrix to the fracture, and matrix blocks surrounded by water were assumed not to tmnsfer gas fmm the matrix to the fracttue. An example was given simulating gas injection into an oil reservoir with an initiaJ gas cap ~presented by a 2 x 25 grid with no communication except in the gas cap. This paper presents the development of a ~D threephase, dual-porosity, finitediffenmce model for simulating naturally fractuted reservoirs. The formulation is implicit in pmssute, water saturation, imd gas saturation or saturation pressure for both mattix/fractu= flow and fmcture flow. We include provisions for calculating gravity, capillary pressure, and viscous fomes for both matrix/fracture and fmcture flow, and hysteresis effects on both capillary pressure and relative permeability. We pment single-block studies to demonstmte matritifmcture flow and simulation of this term by the model. The model can be used to simulate primary depletion, gas injection, and water injection in a natumlly fractured teservoir. The degree of fracturing within the resemoir is described by input data and can vary fmm highly fmctured regions in portions of the resewoir to unfractumd regions in other areas. Examples are included of 3D field-scale water and gas injection.

Following are the equations describing three-phase, 3D flow in the fmcture system written in finite difference form. Fracture Flow Equations Water.
A[Tw@p,, - ~wAD)]+Aw( .pwrn -.pw)qw

S:WLIWSW),

. . . . . . . . . . . . . . . . . . . . . . . . . . (1)

od. AITo(Apo-ToAD)l +ho( Pom Po)-90

=:q+boso),

. . . . . . . . . . . . . . . . . . . . . . . . ...(2)

Gas. A[Tg(@g -~g~)l+~g(


+A(Toi?~@po

pgm -Pg)

-TOAD)] +AoRJ pm Po)9g

=;~(&@g+@oR,SO).

. . . . . . . . . . . . . . ...(3)

The A(pm -p) terms teptesent matrix/fracture fluid exchange and act as source or sink terms in the fmctute
CV.tnwl 0, =.,ss. With . . ,. the ~~~qg~~rt of ~~e rn.ittm:x.!fm!aue

transfer tetms, the fractwe flow equations am the same as those normally found in a conventional beta-type simulator. Finite diffemmce equations describing matrix/fracture flow am as follows. Matrix/Fracture Flow Equations Water.
. . . . . . . . (4) -Aw(pw-pw)=;3(@bwsw)m.

011. s:w$boso)m.
. . . . . . . . (5)

-~o(Pom-PO)

Development of Flow Equations


The teservoir is assumed to comprise a continuous fracture system filled with discontinuous matrix blocks. Primary flow in the rese~oir occurs within the fractures with local exchange of fluids between the fractwe system and marnx blocks. Each block is assumed to have known properties and geometric shape, and all blocks within a given grid block me assumed identical.
FEBRUARY 1983

Gas. ~g( p~m -P~)-~.Rs(


=:s(t$bgSg

P.m p.)

+i$boR,So).

.....

. . . . . . . . (6) q~

The tmnsmissibilities between the matrix and fracture blocks are calculated as follows for the water equation (7)

The CU coetllcients mpment partial derivatives of the accumulation temns evaluated at k-level conditions. For example, (14)

The coefficient u is a geometric factor that accounts for the surface area of the matrix blocks per unit volume and a characteristic iength associated with matrixifractwe flow. This coefficient was introduced first by Warren and Root who presented an equation for u for singlephase quasi-steady-state flow. The preceding matritifracture flow equations assume ID, horizontal flow between block centers of the matrix and fracture. To include the effect of gmvity in the flow terms, we use pseudorelative permeability and capillary nressuns curves 7.s for both the matrix and fracture. This integration effectively reduces the physical flow from 3D to 2D. Fracture/fracture flow is evaluated by using input fractwe capillary pressures, usually set equal to zero. Eqs. 1 through 6 represent six equations in 12 unknowns. Six additional equationsthree for the fracture, and three for the matrix expressing the sum of sammtions Sw+so+sg=l, . . . . . . . . . . . . . . . . . . . . . . . ...(8)

c22=-:&b5
CZ3=xtj)bo, and CZ4= b@kS$b;~, At
bk~

........................(i5j
. . . . . . . . . . . . . . . . . . . . . . ..(16)

. . . . . . . . . . . . . . . . . . . . . . . (17)

where formation volume factor for oil is always calculated as a function of pressure and salutation pmsuxe using the equation
bo=bo~+bJ~p~).

. . . . . . . . . . . . . . . . . . . ..(18)

and capillary pressure relationships


Pw=po-pw . . . . . . . . . . . . . . . . . . . . . . . . . ...(9)

and P cgo=Pg-PCJ> . . . . . . . . . . . . . . . . . . . . . . . . . ..(10) complete the set of 12 model equations. Accumulation Terms The right sides of both the fmctute and matrix equations am expanded as follows for the fmctum equations.
v~

For conditions above and below the bubble point, only three unknowns exist: P,, P2, and P3, where PI and P2 are ~p and 6SW, respectively, and P3 is equal to 6Sg for saturated oil or dp~ for undersaturated oil. Thus, for saturated cells, dp~ is set equal to ~p and the Cw coefficients are added to the Cil coeftlcients. When a cell is undersaturated, ii$g is set equal to zero and the C~3coefficients are set equal to the Cig vahes. Fracture Flow Terms Fracmre flow terms a~ evaluated implicitly as follows for the oil phase fracture flow term.

AITO(&O-TOM)l=AITO(& OT.W)lk
+A[T$ (Ac5p-A13Pcgo)]

~t(4bwSw)=C116p+ +(obwsw)~-(r$bwsw)n. :N46JO)=C2,

Ci26Sw+C136Sg ............... (11)

+(Ap-APcgo -~oAD)~aTo, where 6P Cgo aP,,o


=-6s8,

. . . . . . . . . . . . . . .(19)

. . .... .. ... .. ... . ... . ..

l?sg

ap+czzbsw +c~@sg and

C246p~+(@boSo)k -(4boSo)n.

..........

[12)

13T0 f5sg+ = asg or

aro

aTo as.

Ww,p<p.

. . . . . . . .. (21)

:8(4bgsg
+C33(5Sg

+@boRsSo)=C316p+C326Sw
+CMtip,

aTo 6T0 =6p, ap,

aTo + 15sw, as.

p>p,.

. . . . . . . ..(22)

+(@bgSg+q$boR$So)k ................ (ig) Relative permeabilities and PVT properties all are evaluated at upstream condhions using latest iterate vaiues of ptessure, satumtions, and satwation pressure. SOCIETY OFPETROLEUM ENGINEERS JOURNAL

-(@$gsg+@oJRJo)n.
1 I

44

M.++.

Lvmu.. .A,

lFm..-JAm
m--..
-

~llW ~~~m~ ..
.

TABLE l-DATA

FOR SINGLE BLOCK EXAMPLES 1 0.3

Matrix/fracture flow terms in Eqs. 4, 5, and 6 am calculated implicitly in a manner similar to the fracture flow terms with two exceptions: the aTo/ap$ derivatives are not included for flow above the bubble point, and special considerations a~ given to upstream relative permeabilities when flow is fmm the ftacture to the matrix since flow is governed essentially by matrix properties. Relative permeability to water is limited to matrix kw evaluated at zero Pcw. kw =kw(Pm=O)S@. . . . . . . . . . . . . . . . . . . ..(23)

Permeability, rnd Matrix porosity, % One-foot block Grid Olmensions Grid Spacing q : Ten-foot block Grid Dimensions Grid Spscing q :
Fracture

7x7x8 Ax= Ay=o.ol , 0.1, 0.2, 0.4,0.2, 0.1, 0.01 tt Az=o.ol, 0.1, 0.2, 0.2,0.2,0.2,0.1, 0.01 ft

7x7x
&f=

8
Ay=(t.Ool,

1,2,4,2, 1,0.001 ft AZ%().L3- * -J n ,6, ,I, U.? * nfi~, !. WI, I,<, C,A .-l values areincluded.

The term Sw accounts for the fractional covemge of a grid block by water. Oil relative permeability is calculated as km =km(Sw,Sg)mSo. . . . . . . . . . . . . . . . . . . . ..(24) using Stones9

where krO(SW ,Sg )~ is evaluated equation: km =( kw+kmw)(k,g

Rate Calculations Production rates are calculated semi-implicitly as a function of water and gas satumtions and saturation pmssum, and, optionally, as a function of pressure and wellbore pmssute. For example, oil production fmm a saturated well producing against a constant bottomhole pmsstm (BHP) is evaluated as aqo tip+ ap aqo Mw+as. aqo *S asg

+k,og)-(knv+krg).

. . . . (25) qon+l q:+ (29)

Gas relative permeability is calculated at a matrix gas satumtion of one minus residual oil to gas minus irreducible water saturation,
km=

krg(l-SOrg-SWc)S@.

. . . . . . . . . . . . . . .. (26)

For flow from the matrix to the fracture, matrix saturations are used to evaluate dative permeabilities and capillaty pressures from input values. Hysteresis in both the matrix gadoil and water/oil telative permeability and capillary pmssum is incorporated in the model. The hysteresis model, which includes dminage, imbibition, scanning curves, tmpped gas, trapped oil, residual oil, and residual gas, is similar to that proposed by Killough.0 In this work, however, the scanning functions are assumed reversible. Trapped-gas samtation is defined as the gas saturation at which the gss/oil capillary pressure becomes zero, and trapped-oil saturation is defined in a similar manner for a water/oil system. Residual-gas samrations are calculated using the equation developed by Land, 11

=Pfj, ap

aqo
aqo

. . . . . . . . . . . . . . . . . . . . . . . . . . . .. (30)

as. and aqo asg

b&df#&t,

. . . . . . . . . . . . . . . . . . . ..(31)

=b:q$fo/asg.

. . . . . . . . . . . . . . . . . . . . . (32)

For wells producing a specified constant oil rate, water and gas rates are evaluated as follows for the water mte aqw
%WI+I= %$+-NW!

as.

. . ..-. ..- $.. .. (33)

where

s =

1+ csg/j

Sgu . ....... . . . . . . . . . . . . . . . . . . . (27)

%=(%59-s (34)
An option also is included in the model to perturb rates from ~.rnen m~er ~ from k. Thefonmdation is identical to the one preceding except that all quantities evaluated at level k ate evaluated at time n, and 6 terms such as ~p are written as 6P.

The term SgH i: ~ maximum historical gas satumtion. Drainage, imbibmon, and scanning gasioii Capiiiav pressures are evaluated as a function of gadoil surface tension 12,13at block pressure

P -a- =zPcgo!. .-a u]

. . . . . . . . . . . . . . . . . . . . . . . . . . (28)

Solution of Flow Equations -1 ne u- mud cquuhulla &= m.+,.-.n~ &ennn.atinqc in 6 - .-l- ----- :-... -- J=Mu&W .Y..w.. ... .0
q q

The term P.gol corresponds to input capillary pressuns values calculated using surface tension, u I.
FEBRUARY 1983

unknowns by substitution of Eqs. 8, 9, and 10 for the matrix and fracture into Eqs. 1 through 6. A tiuther reduction in the number of unknowna can be made by eliminating matrix unknowns in terms of fractute values.

TABLE 2-PVT Pressure (psig) 1674.0 2031.0 2530.0 2991.0 3553.0 4110.0 4544.0 4935.0 5255.0 5545.0 7000.O BO (RBISTB) 1.3001 1.3359 1.3891 1.4425 1.5141 1.5938 1.6630 1.7315 1.7953 1.6540 2.1978

DATA FOR SINGLE-BLOCK

AND 3D EXAMPLES
P,j.

TABLE 3-RELATIVE PERMEABILITY DATA FOR SINGLE-BLOCK AND 3D EXAMPLES P Cwo i%) 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.60 0.70 Q.?s
km k &!L

(R%) 0.00198 0.00162 0.00130 0,00111 0.000959 0.000855 0.000795 0.000751 0.000720 0.000696 0.000600

(sc%TB) 367.0 447.0 564.0 679.0 832.0 1000.o 1143.0 1265.0 1413.0 1530.0 225s.tl

(Cp) ~ 0.467 0.436 0.397 0.351 0.310 0.278 0.246 0.229 0.210 0.109

& 0.0162 0.0171 0.0164 0.0197 0.0213 0.0230 0.0244 0.0255 0.0265 0.0274 C.0330

(dyn~/cm) 6.0 4.7 3.3 2.2 1.26 0.72 0.444 0.255 0.155 0.090 C.05Q 0.0 0.005 0.010 0.020 0.030 0.045 0.060 0.110 0.180 g.~~Q 1.& 0.860 0.723 0.600 0.492 0.392 0.304 0.154 0.042 0.0

50.0 9.0 2.0 0.5 0.0 -0.4 -1.2 -4.0 -10.0 -40.0 Pw (psi) 0.075 0.065 0.095 0.115 0.145 0.255 0.386

Original bubble point, pS19 SIOPS Of b. above Pb >vO1/vOkPsi Density of stock-tank oil, lbm/cu ft slope Of jL. a~ve Pb ! c@Psi Gas density at standard conditions, lbm/cu ft Water formation volume factor, psig Water compressibility, voUvol-psi Water viscosity, cp Water density at standard conditions, Ibm/cu ft Matrix compressibility, vol/vol-psi Fracture compressibility, vol/vol-psi

5,545 0.000012 51.14 0.0000172 0.058 1.07 3.5(10 -6) 0.35 65 3.5(10 -6) 3.5(10 -3)

0.0

(t)

km

kW

0.1 0.2 0.3 0.4 0.5 0.55

0.0 0.015 0,050 0.103 0.190 0.310 0.420

1.00 0.70 0.45 0.25 0.11 0.028 0.0

Elimination of Matrix Unknowns After expanding the matrix/fracture flow equations in totally implicit form, the resulting equations are written mathematically in matrix form as
~ms,#+& . . . . . . . . . . . . . . . . . . . . . . . . ,:.

@5)

where Pm is the column vector with elements 6Pm, 6s ~, and 6Sg~, and P is the column vector with elements 6P. MW, and CWg.Using Gaussian elimination, Pm is expressed in terms of P as
P

basic recovexy mechanisms and to demonstmte the validity of the postulated matrix/fracture tmnsfer equations. One- and ten-ft (0.3- and 3.O-m) cubic matrix blocks of intermediate nettability and absolute permeability of 1.0 md were modeled. Grid dimensions and matrix properties for these suns an? shown in Table 1. PVT data me presented in Tabie 2 and matrix reiative permeability and capilla~ pressure data an given in Table 3. Gas/oil capillary pmssute was calculated as a function of pressure according to the variation of surface tension with presswe. Water/Oil Water/oil imbibition above the bubble point was mod,, . J.__ .leiea oy surraunaing me m2iriXA1--1-.....L .------ -,& ~m uiuwi WIUIWUWIa\ W, psig (43 MPa). Fracture water/oil capillary ptesswe was set equal to zem and fmcture water xelatlve pemneability was set equal to the value corresponding to zem capillary pressure to simulate the boundary condition at the face of the block. Oil ~covery vs. time for the 1-and 10-ft (0.3and 3-m) blocks is shown in Figs. 1 and 2, n?spectively. Recove~ tiom the l-ft (0.3-m) block essentially was complete at the end of 10 days, while the imbibition process was still active in the 10-ft (3-m) black after 2 years. Note that the ultimate recovery from the 10-ft (%m) ,- .... block is ------ -- 34,2%, compared with 26% for the l-ft (0.3-m) block. This additional ~covery occurs because of a larger gravity head for the lo-ft (3-m) block. Ultimate recovery comespands to a water satumtion where watdoil capillary pressme equals minus one-half the block height times the difference in water and oil densities. Calculated recovery using the fracnue model with a single cell --- values of u equal to 25 and 0.25 for the 1and ---and lo-tt (U.+ and 3-mj biacks, tispectiveiy, is iii excellent agreement with the 3D, single-block results. Warren and Roots equation for u, 4N(N+2) u= . (37) L*
SOCIETY

-m =/! P+El.

. . . . . . . . . . . . . . . . . . . . . . . . . . .

(36)

The m3tIiX/fmCNre flow terms, Eqs. 4, 5, and 6, now can be written in terms of fracture unknowns. Substitution of either the right or left sides of Eqs. 4 through 6 into Eqs. 1 through 3 can be made. In this work, the right side accumulation terms were used to allow easy modification of the model for additional transfer mechanisms such as those discussed in the Appendix. Solution Technique A42.- .4.. -:-- LUG , ~a,,f.nn. +fi~h- _q_..w... an~ nwst ieuuvmig ,i.~ +1~. QYW.IUUS. ..-- ,am,l~~;nm= Mu+ -three fracmm unknowns, a simultaneous solution for all

three unknowns is obtained using the reduced bandwidth direct solution method ptvsented by I%ce and Coats. 14 Next, matrix unknowns m evaluated using Eq. 36. Following each itemtion, saturation checks are made and coefficients are m-evaluated. Convergence criteria am based either on maximum pmssum change over the last itemtion or on the absolute sum of residuals divided by total production/injection.

SiigIe Matrix Block Studies Detailedsimulations of single matrix blocks surrounded


by water or gas are presented in this section to illustrate
46

OF PETROLEUM ENGINEERS JOURNAL

gives a value of 60 for the l-ft (0.3-m) block and 0.6 for the 10-ft (3-m) block. The above difference in u values is not surprising since Warren and Roots vaiue is for single-phase quasi-steady-state fiow and the values shown here are for transient countemwrent flow. Calculating

50 t

~r

- 3G tiGcEL

FRACTURE MODEL J I

A u =.
LVb

............ (38) . . . . . . . . . .$..,.,<,

... ;-- .-. --- .w..~. vaiU~~ ~~~~ponding ualbs~utiu -id l~nurh

the centmid of one of the six equilateral pyramids in a cube of side L, yields
to

o~ o

20

40 TIME, DAYS

60

80

Fig. l-Water

imbibition recovery, l-ft block.

36.6 =. u
Thus.

L2

. . . . . . . . . . . . . . . . . . . . . . . . . . . . ,. . (39) 601

. ..--, v~ues for u of 36.6 and 0.366 are obtained from this equation for the 1- and 10-ft (0,3- and 3-m) blocks. Kazemi proposed an equation for u,

50 L
r

0=4

++ L: L;

111

L? )

, . . . . . . . . . . . . . . . . . (40)

which was derived on the basis of a length between matrix and fracture nodes and gives values for u of 12 for the i -ft @.3-mj bbcic and 0. i2 frJrtiie iO-ft @-t@ biOCk. The single-cell fracture model results were obtained using pseudo water/oil dative permeability and capillary plessule. GaalOil Oil recovery from matrix blocks with gas in the fracttms was modeled using a depletion mte of 0.75 psi/D (5.2 I&a). Initial ptessum was set equal to 5,540 psig (38 MPa). Gravity drainage is the primary producing mechanism and occurs when the difference between oil and gas densities times block height is greater than the gas/oil capillary tlueshold pressure. 15Ultimate recovery corresponds to a saturation distribution whete capilla~ and gmvity forces are equal. Recovery fmm the 1- and 10-fi (0.3- and 3-m) blocks vs. time is shown in Figs. 3 and 4. Ultimate tecovety for the 1-ft (0.3-m) block is 10%, which is obtained in fewer than 60 days, while ultimate tecovery for the Io-ft (3-m) block is equal to 46%, which is teached in 2.5 years. Excellent agreement between recoveries from the single+ell fmcture model and the 3D values was ob.. :--4 .. --- u ...1..*. 01# &.U ~t,= -.va frw fh~ 1- ~0~ i~fi - VmUGa ? n --r! n (1? .W. . .. . tamcu using (0.3- and 3-m) blocks, respectively. These geometric coefficients were derived assuming lD vertical flowl which chamcterizes the gravity drainage process. For this case, L in Eq. 38 is evaluated as one-half of the matrix height and A is equal to the area of the base. The single-cell fmctute model results wete obtained using gadoil capillary pseudopressums as a function of both gas saturation and pressure, and pseudo gaa/oil relative permeabilities at the initial pmssum. Table 4 gives the pseudodata at 5,545 psig (34 MI%).
FEBRUARY 1983

o~
1

[r
q q q

]
1

I
I

FRACTURE MODEL 30 MODEL

o
Fig. 2-Water

1
TIME, YEARS

imbibition reeovery, ltl-ft block.

40 M

*g 30 > Q g 20 s

FRACTURE MODEL 30 MODEL I

10 0 0 Fig. 20 40 TIME, DAYS So So

3-Gaa/oil

drainage raeovary, l-ft block.

47

TABLE 4-GASIOIL PSEUDODATA AT 5,545 paig

TABLE 5GA!MOIL lMBIBITtON SCANNING CURVES AT 4,500 paig sgH =0.1 SQH=0.2
P

(2)
7i?i0.05 0.10 0.20 0.30 0.40 0.50 0.55

(;5

Sq =0.3
P
km (:% km

S*M=0,4
P,& (psi)

& 0.0 0.026 0.062 0.14 0.21 0.29 0.37 0.420

km 1.0 0.69 0.80 0.62 0.44 0.25 0.070 0.0


;J) 0.076 ~ 0.0

-0.74 -0.47 -0.33 -0.06 0.21 0.48 0.75 1.27

(p%) -2.0 -0.91 -0.17 0.36 0.57 0.73

& 0.0 0.037 0.045


Q,Q~

(;$ -To -0.68 0.25


~,~~

0.0 0.022 0:028 0.103 0.19 -2.0 -0.58 -0.20 0.57 0.73

0.0 0.022 0.090 0.19

-2.0 -1.04 0.018 0.73

0.130 0.167 0.193 n 9tl --0.30 0.40

0.015 0.025 0.032


~,Q34

b.bq

0.103 0.190

0.103 0.19

0.57 0.73

Three-Phase Water imbibition below the bubble point into the 1047 (3-m) block was modeled at 4,500 psig (34 MPa) stafiing with input gas saturations of 0.4, 0.4, 0.3, 0.2, 0.1, and 0.1 in Layers 1 through 6, respectively, by surrounding the block with water. The gas/oil imbibition scanning curves for each of the above saturations are given in Table 5. For the single block calculations, residual gas was calculated using a value of C equal to 2.667 in Lands equation. Straight-line scanning curves between maximum gas saturation and residual gas saturation we~ used for gas/oil relative permeability and capillary pressule. Fig. 5 shows oil recovery vs. time. Ultimate oil recovery for this example was 47%, which is approximately13% higher than the recovery from the water/oil example above the bubble point. Fig. 5 also shows a comparison between calculated recovery fmm the singiecell fracture model and the detailed 3D model and tepmsents an excellent match. A geometric coefficient, u, of 0.02 was used during the period in which the block ...-. . ...- .. ..A-A &Ia, alms a u U1 w.Ad ..,,.. ,..-.-l alLG1 k.. . . . .-A . - .-.+- 9K wait UmGU .,&a. n Wat SU1lUU1lUGU Uy surrounding the block with water. In general, u is calculated in the model using the gas/oil value, Ug, if S@ is equal to zero, Sti is greater than zero, and pOmis greater than pO~.Otherwise, u is calculated using the water/oil value, u ~.

Example Problems
Kazemi et d Five-Spot Exsmple A comparison of nxults from this paper with previous mults is pnzsented in this example which was given first by Kazemi et al. Water is injected into onequaner of a five-spot at a rate of 200 STB/D (31.8 m3/d) and productionis set at a total liquid rate of210 STB/D (33.4 m3/d). The reservoir is assumed to be fractured unifoonly and was modeled using a 2D grid. Reservoir dimensions and properties are given in Tables 6 and 7. Effective fracture permeability was calculated as
Ke=~Kf.

(pf+r$m

. . . . . . . . . . . . . . . . . . . . . . . ..(41)

Upstream n?lative penneabilities were used by Kazemi et al. with ~w . -.:.. _..-.-.:_- as a direct function of ~d ..L.C--L. k ~~ calculated emner rmcturc or mawrx sdturauun. uiiputary prcasuim given in Table 7 were used for both matrix/fmctum flow and flow from one fracture cell to another. Timesteps wem calculated automatically using a maximum fractute
---:11 ---. ..-

cstilmtinn -.. -....

rhmmm -..-.~-

nf(l 11~ ad an anrmn=nt . .. ...- -. -y~-------

mnximnrn . -------------

cten ---r

size of 20 days. A total of 81 timesteps were ~quired. Water breakthrough for this example occumed after approximately 30 days, but water production remained relatively low for the first 2 years because of imbibition of water into the matrix rock and countercurrent flow of

::[ a

UJ

./
-

FRACTURE MODEL 30 NOOEL

:30$ a203--1 10

r
q

.~

FRACTURE MODEL 30 MOOEL

0 o
1
Fig. &-Gas/oil 48

2
TIME, YEARS

o~
o

2
TIME, YEARS

4
104t block.

drainage recovery, 104 block.

Fig. 5-Thrae-phaee,

water imbibition recovev,

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

TABLE 6-KAZEMI et al. FIVE-SPOT EXAMPLE Initial pressure, psia Thickness, ft Gnd dimensions Grid spacing, Ax= Ay (ft) Fracture porosity, % Matrix porosity, % Fracture permeability (effective), md Matrix permeabilii, md Matrixshape factor, aq ft Water compressibility, voilvol-psi Bubble point pressure, psia Water and oil formation volume factor at the bubble point, RB/STB Slops of b. above p~, vol/voi-psi Fracture compressibility, vollvol-psi Water viscosity, cp Oil viscosity, cp Water density, psi/ft Oildensity, pailft Water injection rate, STB/D Total production rate, STS/D 3,959.89 30 8x8 75 0.01 0.19 500 0.0; 3.03(10-6) o 1.0 O.OOOO1O3O83 3(10-6) 0.5 2
(%)

TABLE 7-RELATIVE FOR KAZEMI


PM kti

PERMEABILITY

DATA

et al. EXAMPLE
~ 0.000 0.020 0.055 0.100 0.145 0.200

0.4444
0.3611 210

TF 0.1 0.2 0.25 0.3 0.4 0.5 0.6 0.7 0.6 0.9 1.0

0.05 0.11 0.145 0.180 0.260 0.355 0.475 0.585 0.715 0.850 1.000

kti 1.000 0.770 0.587 0.519 0.450 0.330 0.240 0.173 0.102 0.057 0.021 0.0

(psi) 4.00 1.85 0.90 0.725 0.55 0.40 0.28 r3.2(3 0.16 0.11 0.05 0.0

0.920 0.705 0.420 0.240 0.110 0.0 4.00 2.95 1.65 0.65 0.30 0.00

oil into the fracture system. By the end of 1,200 days, water saturation in the fractwe at the injection cell was 0.97 and water saturation in the matrix was 0.58. Similar values were obtained by Kazemi et al. A comparison of water/oil ratio from this work with that calculated by Kazemi et af. is shown in Fig. 6 and is quite good considering that we used inteqrolated values of dative permeability and capillary psessu~. An additional run using a maximum step size of 3 months was made to test the stability of the model. The number of timesteps used in this run was 48 and the oil recovery from the two runs was essentially identical. Both simulations presented here tequimd two iterations per timestep to converge to a pmssum tolemnce of 0.1 psi (0.7 kPa) and the material balances wem 1.0000 for both water and oil at the end of the nuts.

Three-Dimensional Examples Several 3D, field-scale examples are presented to demonstrate the utility of the model and to illustrate the nature of fluid flow in fractured resewoirs. A linear section of reservoir between a production and injection well was modeled in each example. Areally, the msewoir was divided into thee cress sections with widths of 800,400, and 800 ft (244, 122, and 244 m), respectively. Vertically, five layers each having a thickness of 50 fi ( 15.2 m) wexe used. In the x ditection, the resetvoir was divided into ten 200-fi (61-m) grid blocks. Because of the symmet~ line through the wells, only one-half of the reservoir was actually modeled using a 10x2x 5 grid. Matrix pmneability and porosity wem set equal to 1 md and 0.29, and the effective fmcture permeability and porosity were 10 md and 0.01, respectively. Matrix

PRODUCTION WELL +

WATER-OIL EXAMPLE FRACTURE 50% SW CONTOUR

INJECTION WSLL J

0
-o
1 2 TIME, YEARS
five-spot example.

Fig. 6-Kazemis

I
Fig. 7-Water/oil FEBRUARY 1983 example fracture 50% S ~ contour.

I
49

100
/ /

q
,O

/ q

. /.

/ q

20

o
t

l_/
0

/ q

,
4 -6

1 10

2
Fig. S-Water

TIME, YEARS
injection example.

PRODUCTION

WELL

GAS - OIL EXAMPLE,

INJECTION WELL

2 VEARS ---l

flowing bottomhole pressure of 7,400 psig (51.02 MPa) was used to control water injection after pressure buildup near the well. Total liquid rate was set equal to 4,000 STWD (636 m3 /d). Water/oil rdative permeability and capillary pressure for the matrix is shown in Table 3. We used straight-line fracture dative pemteabilities with zero and one as endpoints and capillaty pmssums equal to zero. Rock and water/oil pseudocurves for this example are essentially :>--.: --1 iiUu Ullly -....1. W VGS Wa *,.-.-I . luenu~~ -_A .--1.. IVLfi_..-.a matrix shape factor equal to 0.25, which was derived from the single block studies for a 104I (3-m) block surrounded by water. p~~~~re ~Q~ wa~~r ~tumtion contoum in both the center and outside cross sections at 1, 3, and 7 years are shown in Fig. 7. Note the gravity segregation that occurs in the fracture system, resulting in water undernmning oil. Matrix saturation contouts are similar in shape and illustrate the water/oil imbibition process for blocks surrounded by water. Fig. 8 shows water cut vs. time for this example. Water breakthrough occurred at approximately 1.5 years, Watercut increased gradually, having a value of 56% at 6 years and 92% at 10 years. Total recovery at the end of 10 years was 35 % OOIP. This tecovery figure is essentially the same as that obtained in the 10-ft (3-m) water/oil single-block study and indicates that recovery is complete. This example took 114 timesteps with an average of 2.18 itemtionslstep. The example was repeated using a maximum satumticm change of 0.3 and a maximum timestep size of 0.50 years. This run took 41 steps (3.1 itemtionslstep) with an avemge step size of 0.24 years. Water cut at the end of 10 years was 90.5% and oil recove~ was within 0.7% of the smaller timestep nm. Water and oil material balances at the end of the mns we~ both i .0000.
aAG oh. -..., SIIUWil. VV S, U-U u

Gas Inj-ion In this example, 90% of the produced gas was teinjected, resulting in partial pressure maintenance. Injection rates wem based on gas production mtes at time n. Initial production mte was setequal to 2,000 STB/D Fig. 9-Gardoil eXam@e. fMture WwSg contour. (318 m3/d) and a maximum drawdown of 150 psig (1 MPa) was applied to the well to simulate declining production with increasing GOR. Rock gas/oil relative permeabilities and capillay pressutes are given in Table blocks were assumed to be 10-fi (3-m) cubes. A detailed 3. Pseudocuwes were input as a function of both saturalist of fluid and reservoir propefiies is given in Table 2. tion and ptessuR, and two-way inteqmlation was used to Initial pmsum in the reservoir was set equal to 6,200 calculate specific values. Data at 5,545 psig (38 MPa) psig (43 MPa) at the center of Layer 1. Both wells were are shown in Table 4. We used a value of u equal to 0.02 perforated in all five layers. Layer productivity indices derived from the single-block study for 104I (3-m) (excluding kr/Bp) wem set equal to 1.0 darcy-ft (0.3 m). blocks sumounded by gas. Timesteps we~ controlled automatically using a maxInjected gas in the fracture system near the well quickimum salutation change of 0.10, an initial and minimum ly sammted the oil. Afier the free gas phase was formed, step size of 0.01 year, and a maximum step size of 0.2 gravity segregation occurred with gas rapidly moving vt=ar T~aM~Sq ~~~ ~n~re~s ftDrn.one step to the next ~cm~~~$e ~ep iayer of Lbemwrv~ir, Ga_sbM_kth_mugh J.. .. was limited to 1.5 times the previous step size. A resulted after only 0.2 years, and, by the end of 1 year, pressure tolerance of 0.5 psi (3.45 kPa) was used in dl the gas saturation in the top producing cell was 0.12. mns. Saturation of the matrix blocks took considembly longer, with some of the bottom blocks becoming saturated only Water Injection after the memoir pnessure was reduced to the original bubble-point pressure of 5,545 psig (38 MPa). Contours In this example, water was injected into the nxervoir at of 50 % gas saturation at 2,4, and 8 years for the fracture an initial rate of 7,000 B/D (1113 m3 /d). A maximum 50 SOCIETY PETROLEUM OF
ENGINEERS JOURNAL

,,

are given in Fig. 9. GOR vs. time for this example is shown in Fig. 10. GOR increased sharply after 3 years operation and reached a value of 118,000 scf/STB (21 254 std m3/stock-tank m3) after 9 years. Average pressure in the reservoir at this time was approximately 3.500 psig (24 MPa). Oil recovery after 9 years was 15.4%. This mn took a total of71 timeste~s, averaging 3 iterations/step, corresponding to an avemge step ~ize of 0.13 years. Gas and Water Injection The effect of gas injection for a period of time followed by water injection is demonstrated in this example. The first 4 years simulation are identical to the gas injection example presented earlier. By the end of 4 years, the GOR in the producing well had risen to approximately 13,000 scf/STB (2342 std m 3/stock-tank m 3), and the average pressure had dropped to 4,700 psig (32 MPa). Starting at the end of the fowth year, the injection well was switched fmm as to water injection at a rate of 7,000 B/D (1113 m ! /d). A maximum flowing BHP of 6,000 psig (41 MPa) was employed. At the producing well, a minimum bottomhole flowing pmssun of 4,300 psig (30 MPa) and a maximum drawdown of 150 psig ( 1 MPa) were applied. Following the initiation of water injection, the GOR at the producing well declined sharply (Fig. 11), until it reached approximately 1,350 scf/STB (243.2 std m3/stock-tank m3), corresponding to solution GOR. Water breakthrough in the producing well occumd in approximately 1.5 yearn. Residual gas for this example was calculated using Lands equation with a C value of 2.667. Thus, a matrix block with a maximum gas saturation of 0.30 would have a midual gas saturation of approximately 0.17 after water imbibition at constant pressure. Gas saturation contours in the matrix blocks at 10 years am shown in Fig. 12. This figure illustmtes the higher trapped gas at the top of the resemoir where significant gravity dminage had occtmed. Oil recove~ at the end of 16 years was equal to 33.8 ?4 and the water cut was equal to 75%. The 12 years of water flooding in this example took i48 timesteps using a maximum fmcture saturation change of 0.1, and 99 timesteps using a maximum fracNR salutation change of 0.2. In the latter run, fmm 6 to 16 years the timestep size was limited by the maximum step size of 0.2 years. The number of iterations per timestep for the two mns was 2.6 and 3.1, respectively. Oil recoveries fmm the two runs were within 0.3%.

120000 m 1<100000

& Q 0 ~- 80000
t4 =

~ o
& a a

60000
40000 20000

1/
/
0

0246810

/ I
q

.-.4-1

TIME,

YEARS

Fig. 10Gas injection example.

80

024681012 TIME, YEARS

1416

Fig. 1l-Three-phaee

example.

PRODUCTION INJECTION THREE PHASE EXAMPLE, WELL WELL ? MATRIX 10% Sg CONTOUR AT 10 YEARS 1 .

Discussion
Itis interesting to compare the mcove~ from the lo-ft (3-m) single-block run surrounded by gas with the recovery from the 3D field-scale gas injection simulation. The single-block example resulted in an ultimate recovery of 47 %, while the 3D field example recovered only 15%. The primary reason for this disparity is that gas segregation to the top of the reservoir allowed only the upper part of the formation to dtain oil. Also, gassing zones in the lower pat of the memoir potentially could imbibe oil.
FEBRUARY 1983

I
4

~...
I

.I
51

Fig. 12-Three-phaee years.

example, matrix 10% Sg contour at 10

To study the effect of better gas coverage, which might be obtained in a muitiiayereci reservoir with limited crussflow, an additional run was madeusing a vettical/horizontal permeability ratio of 0.1. This run exhibited less gravity segregation than the previous one, but as a result of early gas breakthrough in all layetw recove~ was only 10.7% by the time a 100.000 GOR was reached after only after 6 years operation. Oil recovery from the water-injection example was ~~~nti~!!y Lie ~atrn.ess ~h~ @!e-block results, indicating that gravity segregation was not a significant problem for this case and that good vefiical and anxd coverage by water wem obtained. The 3D, thee-phase example illustmtes the effect of trapped gas on oil recovery. By the end of 12 years water injection following 4 years gas injection, a recovery almost equal to uitimate water flood recovery was ~ached. The water cut at this time was 75% and ultimate recovery was projected to be37%. Resuits from the time-phase single-block run indicated 47% recove~. Obviously, this value of recovery can be obtained in a field-scale project only if complete gas coverage is attained prior to waterflooding and if pressure is not increased to a level where the residual gas goes back into solution.

D=

depth measured positive downward,


ft (in)

Conclusions
This paper ptesents the development of a 3D, threephase model for simulating the flow of fluids in a naturally fractured reservoir. There a~ four specific advancements in this work. 1. We describe a highly stable fommlation that tteats both fracture flow and matrix fracture flow implicitly in pressure, water satumtion, gas saturation, and satmation pressulw 2. The implicit treatment of matrixhcture flow
nmcmntd

p------

h-m ..-.=

r~n

-=.

hP

.-

rwwfnrmed ~-.--....-_

W~Lh ~S.~n~~i~

aO ad-

ditional work per iteration compamd to the sequential approach used by Kazemi et af. and results in a more stable and ei%cient model. 3. The matrix/fmcture flow equation developed in this paper accurately matches detailed simulations of twophase, water/oil and gas/oil flow processes as well as three-phase flow. Provisions are included for modeling the gravity term, for properiy calculating dative petmeabilities as a function of both up- and downstream conditions and hysteresis, and for evaluating the appropriate geometric factor depending on the environment of the fracnue system. 4. We give two- and three-phase, 3D examples that demonstrate the utility of the model and provide insight into the nature of multiphase flow in natukally fractured resemoirs.

f= fractional flow k= formation permeability, md k, = dative permeability L = length, ft (m) N= number of normal sets of fractures p= pressute, psia (kPa) p= = capillary pressure, psi (kPa) pi = vector of unknowns Ps = saturation pressure, psia (kPa) Pwf = flowing bottomhole pressure, psia (kPa) q = production rate, STB/D or scf/D (std m3Id) qT = total production rote, RB/D (res m3 Id) qmF = matrix/fmcture flow, STB/D or scf/D (std m3/d) R, = solution gadoil ratio, scf/STB (std m3/m3) s = saturation, fraction s@ = maximum gas sanuation for hysteresis calculation t = time, days T = fmctwe transmissibility, 0.001127 (kA/L)bkJp, STBID-psi [2.6 (10 5) (kA/L)bkr/~, ml .-. --- m 3 Id. ~_Pa] v~ = bulk volume, res bbl (res m3) Vp = pore volume, res bbl (ms m3) w= fiactum width, cm (in.) X,y,z = Cartesian cocmhates Y= specific weight, psi/ft (kPa/m) & = iteration difference, &=xk+ 1-xk 3 = time step difference, a=ai=xn+[ +Xn At= time increment, rn+ 1- tn A(TAp) = AX(TXAXP)
+Ay(TyAyP)+AZ(TZAZP) A(TXAXP) = Ti+%(pi+l -pi) -Ti-lA@ipi-l) A = matrix/fmcture tmnsmissibility,

SITI/D-psi (m3/d. kPa) viscosity, cp (Pa. S) density, iiimicu tl (kghn3 j matrix shape factor, l/sq ft (1/m2 ) or surface tension, dyne/cm (mN/m) porosity @i = porosity at initial conditions, fraction

Nomenclature
A = ma, sq fi (m*) b = formation volume factor, STWRB or scf/RB (m3 /m3) B = fortnation volume factor, RB/STB or RB/scf (m3/m3) C = accumulation term pmiai derivatives ~ = ~Q~-P~.~~ihi!i~ Qi/vQi-~i , (vol/vol kPa)

superscripts
k = iteration level # = derivative at kth iteration level

subscripts e = effective
f = fmctufe

or fomkation PETROLEUM ENGINEERS


JOURNAL

g = gas
SOCIETY OF

i? J* k = grid block indices

m = n = o = r = s = w=

matrix time step level oil residual saturation pressure water

mulation offers definite advantages over a strictly sequential approach, especiall for conventional nonfractured resetvoir simulation. #

Matrix/Fracture Transfer
In addition to flow contributions from capillary ptessure, gravity, and viscous forces within matrix blocks, matrix/fracture flow may result fmm a pressure gtadient across the matrix block and from diffusion of gas fmm a satutated ftactttre cell to an undemammted matrix block during gas injection. Inclusion of these terms in the matrix/ftactute flow equations results in the following equations.
L= ?wnsf=hw( Pwm P.)

Acknowledgments
We thank Phillips Petroleum Co. for permission to publish this paper.

References
1. Wan-en, J.E. andRoot,P.J.:

The Behavior Naturally of Fracin<?.*.C <= mredReservoirs, Per. Eng. J.(*pi. 1YOJ L4J-JJ. ) Sot.

2. Matrax. C.C. and Kyte. J.R.: Imbibition Oil Recovesy frum Fractured, Water-Drive Reservoir, Sot. Per. &rg. J. (June 1%2) 177-84. 3. Yarnarnoto, R.H.et aL: -CompositionalReservoir Simulator for m.-.,.-,.! 0, .,alsC.,..m P., t70 j T1.- .J,,,&vw..l. ~;,.1-.llln,-l. .Mn,i.al . . , ,, kc. ,-,,W,SU 1 ,,t. ... . . . . -.. a. . .
(June 1971) 113-28. 4. Bmester, C.: Simultaneous Flow of Immiscible Liquid Through Porous Media, SoC. er. Errg. J. (Aug. 1972) 297-305. P 5. Kazemi, H. er aL: Numcricai Simulation of Water-Oil Flow in Natumlly Fractured Reservoirs. .%x. Per. Eng. J. (Dec. 1976)

7WJ! +}. A =

LB

. . . . . . . (A-1) . . .. .

LC
qomf=Aoi Pom -po)+~o &ofl LB . . . . . . . . . (~-~j

and
qgmf=hgt Pgm Pg)+hg ~%f

317-26. 6. Rossen. R. H.: Simulation of Namrally Fractured Reservoirs with Semi-Implicit SourceTerms, Sac. Per. Eng. J. (June 1977)

201-10.
7. Coats. K.H. er al.: Simulation of Three-Dimensional, TwoPhaseFlow in Oil and Gas Reservoirs,SoC. Per. Eng. J. @cc. 1%7) 377-g8. 8. Coats, K.H., Dempsey, J.R., and Henderson.J.H.: The Use of Vertical Equilibrium in Two-Dimensional Simulation of ThreeDimensional Reservoir Performance,Sot. Per. Eng. J. (March 1971) 63-71. 9. Stone, H. L.: Estimation of Three-Phase Relative Perrneabdity and Residual Oil Data, J. C&s. Per. Tech. (Oct. 1973) 53-61. 10. Killough, J.E.: Reservoir Simulation with History-Dependent SaturationFunction. Sot. Per. Eng. J. (Feb. !976) 374. 11. Land, C. S.: Calculation of Imbibition Relative Permeability for Two- and Three-phase Flow Fmm Rock Properties,Sot. Per. Eng. J. (June 1%8) 149-56. 12. KaQ, D. L., Monroe, R. R., and Ttainer, R. P.: Surface Tension of Crude Oils Containing Dissolved Gases, Per. Tech. (Sept. 1943). 13. Sugden, S.: A RelationshipBetween SurfaceTension, Density, and Chemical Composition, J. Chcm. ot. (1924) 12s, 1,177. S 14. Price, H.S. and Coats, K. H.: Dkect Methcds in Reservoir Simulation,Sot. Per. Eng. J. (June 1974) 295-308. 15. Thomas, L. K.. Katz. D. L., and Tek. M. R.: Threshold Pressure Phenomena in Porous Media, Sot. Per. Eng. J. (June 1%8) 174-84. 16. Coats, K. H.: A Fully Implicit Steamflood Model, &c. Per. Eng. J. (Oct. 1978) 3&-83. 17. Amyx, J.W.. Bass, D.M. Jr., and Whiting, R. L.: Perrofeum Reservoir Engineering, McGmw-Hill Book Co.. Inc., New York City (1960) 83-85. 18. Parsons,R. W.: Permeability of Ideaiized FracturedRock, Sot. Per. Eng. J. (June 1966) 126-36.

+WJ

P..

Po)+~o&&wof

+Ag~(boR,m -b oR ~, . . . . . . . . . . . . . . . . (A-3) f whete Ap = pressure dmp across matrix block, LB = distance over which Ap acts, LC = characteristic length for matrix/fracture flow, A*D = urbs~ D15.6146, BID, D = diffusion coefficient.

Gicuiation of Fracture Properties


InTsim fractute ~tmeabilities are a function of ftacture width squared. 1 Expressing Win centimeters and kf in millidtucies yields kf=8.44(109)W2. . . . . . . . . . . . . . . . . . . . . ..(A-4)

APPENDIX

Semi-Implicit Formulation
A semi-implicit formulation, which uses the implicit matrix/fractute flow described in the text and an implicit fmctttte solution for presswe and gas saturation or samration pressure followed by a sequential water calculation, also is included in the model. This formulation is adequate for many two-phase water/oil problems and cart be used for some time-phase studies. This forFEBRUARY 1983

Effective fmctute permeabilities can be derived 18 by assuming linear flow through the fmcture system parallel to four faces of a cubic matrix block of length L. The expression for effective permeability for this system can be written as K.A m+KfAf A (A-5) ---

K,=

The term A represents the area included by one face of the matrix block plus the minor area corresponding to one-half fracture width surrounding the matrix block.
53

For W< <L, =


AL
and

S1 Metric Conversion Factors


. . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-6) bbl X 1.589873 Cp x 1.0* CU ft X 2.831 685 dyne x 1.0* ft X 3.048* in. x 2.54* Ibm x 4.535924 psi X 6.894 757 scf x 2.863 WI
Convemon

Af

2W

K, =Km +KfA/A

=K~+l.69(1010).

W3
L

. . . . . . . . . . . . . . . (A-7)

E01 E-03 E-02 E-02 E-01 E+OO E01 E+OO E-02

= = = = = = = = =

m3 Paes m3 mN m cm kg kpa std m3

Fracture porosity for this system is equal to

factof exam m

SPEJ
m SOCKAY Petfo@um of Engmwm OWOAug.18, 1SS0. *E AMU@ T*n~ cm-

3W
df=y.
L

Origmdmmuacrvt recai~

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-8)

NOV. 3, PWMr accemedorPubliiuon f June23, 1SS3. !Owsed F manuscnpt remivad

1ss3. PEW C3PE93W fimt pmented at the +* fwmw mrl ExhtMon w inOallaa seLx. 1-24. 2

54

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

You might also like