You are on page 1of 14

SPE

SPE 15297

BOAST-DRC:
— —Black Oil and Col?densate Reservoir Simulation
on an IBM-PC
by J.R. Fanchi,Marathon oil co.
SPE Member

Copyright 1986, Society of Petroleum Engineers

TINS paper was prepared for presentahon at the Sympoatum on Pelroleum Industry Application of Microcomputers 01 the Soc!ety o :>elroleum Engineers
held m SilverCreek, CO ;.me 1.S-20, 1986

Th!a paper waa selected for presentation by an SPE Program CommMee following review of information contained m an abstract submitted by lhe
author(s) Conlenla of the paper, as preaenlad, have no! been reviewed by the SGCwly of Petroleum Engineers a,ld are subject 10 cormcbon by the
authur(a). The material, as preaenled. does not neceaaarily reflecl any position of Ihe Sa-5ety of Petroleum Engineers, !1sofhcers. or members. Papers
presented al SPE meetmgs are sublecl lo publication review by Editorial Oommiffees of the society of Petroleum Engineers Perm8saion 10 copy IS
restricted to an absmacl of not more than m words. Illustrations may not fm copied. The abstrac! should contain conspicuous acknowledgmem of wh Ire
and by whom the papar is presented. Wme Publication Manager, SPE, P.O. Box S33S8S, Richardson, TX 76083-2835 Telex, 730889, SPEDAL
-—

ABSTRACT Run times for the IBM-PC XT version


range from 0.3 secondlgrid block/tfme step
BOAST-DRC is a three-dimensional,three- for 1-D problems to 0.7 second/grid
phase simulator designed for microcomputer block/tires?step for 3-D problems. Run times
applications. It may be used for either are very sensitive to the number of itera-
black-oil or retrograde condensate reservoir tfons taken by either the iterative matrix
simulation. The program formulation, md solver or the pressure solution procedure.
solutiow to problems encountered in prepar- In at least one case inwhfch a direct matrix
ing a microcomputer versfon are presented. solver was used, problem run time was limited
Results for several test problems, including by the speed at which the output could be
the first (black-oil) and third (condensate) printed rather than 5Y the calculations.
SPE Comparative Solution Projects, are shown. Formulation of BOAST4M!C is presented in
Section II. Important differences between
I. WT’ROOUCTIOH BOAST-IIRCand the DOE version are highlighted
there. Section 111 provides a general dis-
BOAST-DRC is an improved version of the cussion of fluid properties and variable sat-
Implicit Pressure - Explicit Saturation uration pressure calculations. Several work-
(IMPES) simulator BO~ST published by the U,S. saving initialization options are explained
Departmentof Energy (DOE). BOAST-DRC sifw in Section IV. Details of well model calcu-
lates isothermal Darcy flow in up to three lat+ons are pr$sented in Section V. Analytic
dimensions. It assumes that reservoir fluids aquifer models #nd a few +mPortant m+scel-
can be described by up to three fluid phases laneous topics3- are discussed in Section
(oil, gas, and water) with physical proper- VI.Several conversion problems were encoun-
ties that depend on pressure only. Gas is tered in translating the Burroughs B7900
allowed to dissolve in both the oil and water mainframe version to a version suitable for
phases. Vaporized oil (wet gas) can be use on an IBM-PC. The most noteworthy are
modelled when the retrograde condensate op- discussed in Section VII. Results of some
tion is invoked. The retrograde condensate example problems are sufmnarizedin Section
op’?on provides the user with a tool to study VIII. Conclusions are presented in Section
t$ effects of lean gas injection or cycling IX.
of Che performance of a condensate reservoir.
BOAST-DRC is designed for use on an 11. FORMULATION
IBM-PC XT or AT. The PC version of BOAST-DRC
has all the features of the mainframe ver- The formulation of BOAST-DRC differs
sion, including a direct solution method for from the 00E version of BOASTby inclusion of
moderately sized problems. Analytic aquifer a vaporized oil term in the flow equations.
models provide a storage efficfent means of Flow equations for the three phzses are col-
studying aquifer fnflux in either black oil lected in Table 1. Additional definitions of
or condensate systems. variables and fluid properties ar~ presented
——-.— in Tables 2 and 3, respectively. A compar-
ison of Tables 1 through 3 wit+ the formula-
References and illustrations at end of paper. tion in Reference 1 shows that all new tef’fffs
.-
1Q7
2 ROAST-DRC: Black Oii and Condensate Reservoir Simulation on an IBM-PC XT SpE 15297

in BOAST-DRC include the l!quid content tfons. In BOAST it is. unimportant which
factor VS. Note that gas compr?ssibllity saturation is c?lculateclfirst. Consequen-
includes a nev term proportional to the p~es- tly, BOAST calculates the oil phase satura-
sure derivative of the liquid content, which tion first and the water phase saturation
is treated as a function of pressure only. second.
Convective tralsport coefficients, Un7ike ROAST, “’-- BOAST-DRC oil phase
including gravity and capillary pressure con- flow equation includes both the oil and gas
tributions, are calculated as described in phase saturations in fts accumulation term.
Reference 1 with the followfng exception. This means that only one eq;lationis avail-
Solubilities in the transmissibility terms able for finding two variab;es if the oil
are single point upstream weighted. This phase flow equation is solvej first. To
differs fr~ the mfdpoint weighting used in avo:d this problem, water phase saturation is
the DOE version. The change was made to calculated first from the water phase flow
improve front tracking for lean gas injection equation. Once the water phase saturation is
problems. It also improves material balance. known, gas phase saturation in the oll phase
Another change that was made to improve mat- flow equation accumulation term can be repla-
erial balance was to use a reducsd set of ced using the phase saturation constraint in
flow equations when less than three phases Table 2, Rearranging the accumulation term
are present in a grid block. These changes lets us solve the oil phase flow equation for
have little affect on liquid daminated sys- oil phase saturation. Gas phase saturation
terns.They are important primarily for gas then follows from the saturation constraint
dominated systems. equation and the updated values of water- and
oil-phase saturations. Thfs procedure is
$~~1~ pR~E~RE used with the chain rule expansion treatment
of black oil and condensate problems.
BOAST-DRC is dn IMPES simulator. Auxil-
iary relations and the three flow equations Saturation Cal@latlons Fora Mass Conserving
are combined algebraically to yield a pres- Expansfon Procedure
sure equation. The procedure for construc-
ting a pressure equation, and the order in Amass conserving expansion procedure is
which saturations are subsequently computed, optionally available for solvfng bubble point
depends on the method used to expand the problems. The primary variables in the mass
accumulation ‘termsof the flow equations. conserving expansfon procedure are oil phase
Two methods of expanding the accumula- pressure, water phase saturation, and gas
tion terms are available as options in phase saturation. If the grid block is
BOAST-DRC. The accumulation terms can be undersaturated, varfable substitution i$ used
expanded using a chain rule expansion, as in to replace the gas phase saturation with the
the original DOE version of BOAST, or by bubblz pofnt pressure. The ffrst Iteration
using a mass conserving expansion (also known serfto determine a pressure dist.”ibution
;$+!, This distribution is used to deter-
as consistent expansion) with variable sub-
stitution. A comparison of these expansions mine convective transport between blocks.
is given in Appendix A. Most authors rec- All three primary variables are then compu-
onnnend using a mass conserving exposion. ted. The water equation provides a value for
Implicit coefficients in the mass conserving the water phase saturation. All three flow
expansion are correctly calculated by iter- equations are combined to yield a new oil
sting on the pressure solution. More compu- phase pressure P*. The remaining primary
ter work is necessary, but good material variable (gas saturation or bubble pofnt
balances are achievable. As shown in Appen- pressure) is calculated from ~oth the oil and
rlixA, there are circumstances in which the gas equations. Convergence is achieved when
mass conserving expansion and the chain rule P* equals pk+l within a user-specified tol-
expansion are comparable. Consequently, the erance (usually 0.1 psi).
chain rule expansion In the DOE version of
BOAST can be readily applied to many real- MATRIX SOLUTIONf4ETlkOOS
fstfc problems. The first four examples in
Volume 11 of the BOAST documentation justify BOAST-ORC has several solution methods
this assertion, as do examples shown later fn that represent two basic techniques of solv-
this paper. If the chain rule expansion is fng a linear algebraic problem: (1) direct
applicable, the implicit pressure distrfbu- solution (TANK, BAND, 04, DHT); and (2) line
tion can be computed without iterating.This successive over-relaxation iterative methods
considerably reduces the computer work (LSORX, LSORY, LSORZ). Reference 1 describes
needed, and is especially helpful for practi- most of these methods. OHT is a storage
cal microcomputer applications. efficient band algorithm.6 The X, Y, and Z
appended to the LSOR acronym indicates the
Saturation Calculations For a Chain Rule direction fn which the trfdiagonal aigorfthm
Expansfon Procedure is alfgned.
The TANK method is used when BOAST-DRC
The 00E version of BOAST uses the up- fs run with a single grid block. When a
dated pressures from the pressure equation in sfngle grfd block is defined, BOAST-DRC func-
the oil- and water-phase flow equations to tions as a sophisticated material balance
find the updated oil- and water-phase satura- program.

i-
SPE 15297 J. R. Fanchi 3

The LSORnethcds are most appropriate in dew point resulting from, for example, injec-
the 1-D case because they are coded to func- ting dry gas into a wet gas block. This
tion as tridiagonal algorithms if a 1-D grid effect is accnunted for by variable substitu-
is defined. In the 1-D case, a single pass tion as follows.
is made through the aPf)tYJ@ate LSOR subrou- The oil saturation calculated from the
tine. Pressures returned from the tridiagon- oil equation cd” be negative if enough dry
al algorithm called by the LSOR subroutine gas invades a wet gas block, If a negative
are not modified by the LSOR acceleration oil saturation is calculated, the block is an
parameter U. undersaturated gas block. The oil phase flow
Problems in 2-O or 3-D are solved using equation shown in Table 1 simplifies to
either 04, DHT, or the LSOR method which is
aligned in the directiol: of maximum flux. l-SW
For example, consider a 3-D model with com-
munication bettieenlayers. The large cross-
sectional area which is perpendicular to the
direction of flow Ctnerally results in the
L =—a
0 at
[( )1
.$—
Brs”
9
(1)

because tne oil saturation is zero. Equation


greatest transmissibility aligned in the (1) is solved for liquid content r . The
Z-direction. In such cases, the LSIIRmethod pressure that corresponds to the cafculated
of choice is LSORZ. For a horizontal 2-D liquid content on the saturated liquid con-
problem, the LSOR method o~ choice will be tent curve (Figure 2) is the new dew point
either LSORX or LSORY. For numerically pressure of the block. Oil saturation in the
difficult 2-D or 3-D problems, a direct tlock is set to zero.
solver like the DHT method is used, The OHT If oil saturation is positive, then a
method is particu?arl:”helpful for the PC two-phase hydrocarbon system is assumed to
version of BOAST-DRC because the BAND and 04 exist. In this case the dew point pressure
direc’ solution methods demand tcomuch stor- is set to the updated grid block pressure. A
age for efficient use on a PC. liquid content is obtained at the new pres-
sure value and a new oil saturation is found
III. FLUID PROPERTIES from the oil flow equation.
6ENERAL
Iv. INITIALIZATIONGPTIONS
The principal oil PVT data needed by
BOAST-DRC are oil formation volume factor, Initial pressure and saturation distri-
volubility of gas in oil, and oil viscosity, butions can bs input to BOAST-DRC block-by-
which are assumed to be functions of pressure block. Some work saving algorithms for
only. Illustrative curves are shown in initializing pressure and saturation distri-
Figure 1. Undersaturated curves are shown as butions are also available.
solid lines and saturated curves as dashed
lines for a bubble point pb. The user PRESSURE INITIALIZATION
specifies the saturated curves and the slopes
of the straight lines representing the under- Improper pressure initialization can
saturated curves for the initial bubble result in differences in phase potentials
point. that cause fluid migration in the reservoir
Water PVT data are input and used in a model even though no operating wells are pre-
manner that is directly analogous, to the sent. BOAST-ORC uses the pressure initiali-
undersaturated treatment of oil PVT data. zation algorithm described in Reference 1 as
Gas PVT data are depicted in Figure 2. a means of achieving an equilibrated pressure
The possibility of a dew point pressure pd distribution.
is taken into consideration for the liquid
content curve rs. All gas PVT data are SATURATION
INITIALIZATION
treated on a dry gas basis in the sfmulator.
This is quite important for condensate Several saturation initialization op-
systems. tions are available in BOAST-ORC.The coordin-
ate convention is shown in Figure 3.
BUBBLE POINT TRACKIN6FOR BLACK OIL SYSTEMS
Vertical Equflibrim
The bubble point tracking algorithm in
the DOE version of BOAST does not work well Data for the vertical equilibrium option
for modelling lean gas injection into an include the water-oil and gas-oil contacts
undersaturated oil reservoir. BOAST will WOC and GOC, the oil and water saturations
tend to overestimate pressure drawdown in an for a two-phase water-oil system (S01 and
oil block that is being depleted. The error SWOI, respectively), the gas and water satur-
is avoided in BOAST-DRC because variable sub- ations for a two-phase gas-water system (SGI
stitution is used with the mass conserving and SkiGI,respectively), and a residual oil
expansion option. This point is illustrated saturation (SOR). In addition, block thick-
in the examples. ness (DZ) and elevation (EL) must be known.
If the transition zone between phases is
OEN P01NTTRACKIN6FORCONDEI!SATE SYSTEMS negligibly small re~ative to the thickness of
the block, there are six possible configura-
Analogous to the bubble point tracking tions the block may be in (Figure 4). They
situation is the possibility of a change in represent a system in which ideal vertical

)
equtllbrtum between phases has been establi-
shed. kw/Bw
~ ‘q — (5)
wk gk Ag/Bg k ‘
Capillary Pressure Initialization [1

The input capillary pressure curves can


be used to determine transition zone satura-
ko/Bo

.1
tions for either an oil-water or gas-water
system. q +(r)
ok ‘qgk -k S k ‘gk
[ ‘ (6)
Oil-Water System:
If the block midpoint elevation (EL) is Note the appearance o+ the liquid content
greater than or equal to WOC, the block con- tenn in the equation for oil rate.
tains only water. If not, an estimate of
capillary pressure is calculated using Pressure Controlled Hell for a Condensate
Syste,
Pcow = (Ow - po) (woc - EL)/144 . (*)
For a given flowing bottom hole pressure
Densities are calculated for a pressure (PWF), phase rates by layer are:
that is the average of the block pressure and
the user ir~ut pressure at HOC. A water sat- Aw
uration is calculated from the capillary = (PI) ~ AP ; (7)
pressure estimate in Equation (2) and the in- ‘Wk
put capillary pressure curve for an oil-water [1
systern. Oil saturation is one minus the
water saturation,.
‘gfl~pl)kA~~+Rso~] /~sw)kqwk}+
Gas-Water System:
If the block midpoint elevation is less
than or equal to GOC, the block contains only
water. If not, an estimate of capillary
presssure is calculated using
[l-(RS&(rSJk], ‘8)
P = (Pw- Pg) (GOC - EL)/144 . ~3)
ego

Densities are calculated for a pressure


k
[1
that is the average of the block pressure and
the user input pressure at GOC. A water sat- q = (PI) + Ap+r (9)
uration fs calculated from the capillary ok
ok
() skqgk”
pressure curve for an oil-water system. In
this case, it is assumed that there is no The pressure difference is the difference
capillary pressure difference between the gas between the grid block pressure at the pre-
phase and the nonexistent oil phase. Thus sent time and the use! specified wellbore
the capillary pressure curve for an oil-water flowing pressure (PWF); thus
system is equivalent to a capillary pressure
curve for a gas-water system. Gas saturation
is one minus the water saturation. AP=Pn-PWF . (lo)

v. NELL REPRESENTATION KEN IHJE’TIONUELL OPTIONS

References 1 and 7 describe several Dry Gas ‘!ateSpecified


methods of representing wells in BOAST.
BOAST-DRC contains all of the explicit well For a dry gas rate qg, dry 9as injec-
control options described in Reference 1. tion rate by layer is
The code has been modified to allow more than
one well to exist in a grid block. Addi- ‘Wx) ‘o+hw+xg
[( )1 k
tional options have been added for use with q ‘q
condensate reservoir models. These new gk g ~
k=l[(WI) (Ao+Aw+hg)]k “(11)
options are described here.

NEM PRODUCTION HELL OPTIONS Wet Gas Rate Specified

Dry Gas Rate Specified If a wet gas rate is specified, dry gas
injection rate is given by Equation (1.). In
Given the dry gas rate qg, the 9aS, addition, an oil injection rate resulting
water and oil rates for layer k are: from reinfection ~f vaporized oil has the
from
(PI) Ag/Bg ~
[ 1 qok =
qfl . (4) ‘S
() k ‘gk . (12)
gk. PIAB
g \=,[( ) gjgk 1 ‘
+An
SPE 15297 J; R. Fanchi 5

VI. SELECTED TOPICS Crntrol of numerical dispersion by altering


the time-step size in an IMPES simulator
ANALYTIC AQUIFER MODELS requires balancing competing factors. Maxi-
mum time-step size is determined by satura-
Several analytic aquifer models have tion and pressure change requirements in
been implemented in BOAST-ORC. They include addition to Equation 14. The rotation effect
the pot, steady-state, and Carter-Tracy aqui- associated with numerical dispersions can be
fer models. Tbeje models are described b! reduced by decreasing the time-step size. In
Aziz and Settari. An analytic correlation general, the most practical way to evaluate
is used to determine the Carte#-Tracy influ- the effect of numerical dispersion on simula-
ence function and its derivative with respect tion results u.1.lgBOAST-DRC is to run a test
to dimensionless time. case with a more refined grid (more grid
blocks). This provides a basis for compari-
THREE-PHASE RELATIVE PERMEABILITY son of sim’llationruns.

Stone’s three-phase relative permeabil- VII. BOAST-DRC ON AN IBM-PC


ity model, as modified by Oietrich and
Bond@p,3 is available as an option. One of the advantages of a program the
size of BOAST-DRC is tk!atit is small enough
MATERIAL
BALANCE CALCULATION to be put on a microcomputer. BOAST-ORC iS
approximately 6000 lines of FORTRAN 77 code.
An instantaneousmaterial balance calcu- In converting it to run on an IBM-PC XT or
lation is performed as a check on the accur- AT, several problems were encour,tered.
acy of finite difference calculations. The Before copying BOAST-ORC from a main-
volumes in place of oil, gas, and water at frame (Burroughs B79013) to the IBM-PC, the
the beginning of the time-step are compared program was divided into several files, none
to their respective volumes in place at the of which exceeded roughly 1000 lines of code.
end of the time-step with injected and produ- All COHMON blocks were put in separate files
ced fluids being accounted for. and included in the source files using INCLU-
In addition to the instantan~ous materi- DE statements. The arrays in COMMON and
al balance calculation, which is reported in DIMENSION statements were expressed in terms
a one-line time-step sunsnary,a cumulative of variable names which were defined by PARA-
material balance calculation is made and re- METER statements. This last item made redi-
ported in a user-requested Suimnary Report. mensioning the program relatively easy.
The cumulative material balance compares mat- Compilation of the program showed that
erial present at the end of a time-step with variables wnich were defined by REAL or
material present at the beginning of the run. INTEGER statements had to be defined as such
If the cumulative material balances are in every subroutine in which they appeared.
a percent or less, program accuracy is usual- These definitions were required even if the
ly acceptable for many problems. By con- variable only appeared in the subroutine as
trast, large cumulative material balance an element of a COMMON block.
Indicates that the IMPES formulation is hav- Unlike some machines, IBM does not rcu-
ing problems. This situation can ordinarily tinely initialize all program variables to
be corrected by adjusting calculation param- zero. Consequently an uninitialized variable
eters such as time-step size or solution tol- can assume a nonsense value. To get around
erances. this problem without reducing the portability
of the code, a subroutine was added that
NUMERICAL DISPERSION specifically set all COMMON block variables
to zero. The subroutine is called at the
An estimate of the magnitude of numeri- beginning of program execution.
cal dispersion in BOAST-ORC is A batch file was written to provide a /
means of recompiling and/or redimensioning
(13) the program with minimum effort. By definin9
a file containing all PARAMETER statements, ({
it is possible to redimension the program
where v is the frontal advance rate, Ax Is simply by editing the PARAMETER file, then
the grid block size in the direction of fron- recompiling the entire source code. Compila-
tal advance, At is the time-step size, PorQs- tion of the program takes approximately one
ity is a fraction, and Dnum is the numeri- hour on an IBM-PC XT, and about 30 minutes on
cal dispersion. It should be noted that an IBM-PC AT. As a matter of routine, the
reducing time-step size will increase numeri- 8087 microprocessor chip and associated
cal dispersion because BOAST-DRC is an IMPES mathematical library doutines are included in
simulator.q Adding grid blocks is a reliable the compilation. \-
way to reduce numerical dispersion if time-
step size is adjusted so that Dnum of Equa- VIII. EWLE PRWLEMS
tion (13) does not become negative. An opti-
mum time-step size can be estimated from Four examples are presented below as
BOAST-ORC test problems. These problems
. (14) illustrate the capabilities and limitations
of BOAST-DRC.

141
6 BOAST-DRC: Black Oil and Condensate Reservoir Simulation on an IBM-PC XT SPE 15297

LINEAilBUCKLEY-LEVERE~ UATERFLODD limitations of the chain rule expansion


Optfon in BOAST-DRC. An initially undersat-
A linear Buckley-Leverett waterflood urated, 59” API oil is produced from a corner
provides an analytical test of the saturation block f,lthe lowermo%t layer of a three-layer
calculation of a simulator. Fluid and grid grid. Lean gas is injected into the upper
properties for this example are given as layer at the opposite corner.
Example 1, Volume II of Reference 1. Results Table 7 shows results for the case when
for a grid with 40 grid blocks are shown in the bubble point is allowed to vary and the
Figure 5. Also shown in Figure 5 are results chain rule expansion option is used. This
for a three-dimensional, three-phase reser- option is unable to adequately model lean gas
voir simulator called NBOS. NBOS is based on injection, as indicated by a cumulative gas
a two-pseudocomponent representation of hy- material balance error in excess of 10%. The
drocarbon PVT data. The two-pseudocomponent more sophisticated variable substitution al-
representation allows rigorous treatment of gorithm with mass conserving expansion option
either variable bubble point or variable dew alleviates the problem, as shown in Table 8.
point problems. A detailed description of Early gas breakthrough at the explicit pro
NBOS is provided in Reference 9. duction well block accounts for the relative-
Differences between NllOS and BOAST-DRC ly minor differences between BOASi-bt?.and
are negliqlble. Both simulators exhibit an Odeh’s participants. The ov~ral! ROAST-DRC
expected smearing of the front due to numerf- perfomnance for gas injection into a bubble
cal dispersion. Results for a grid with 10 point reservoir is acceptable if the mass
grid blocks are shown in Table 4. Water pro- conserving expansion option is used.
duction at 360 days is 32 STB/D for NBOS and
33 STB/D for BOAST-DRC. Agreement between GAS CYCLINGOF A RETROGRADE CONDENSATE
the simulators is excellent for this case. RESERVOIR
Run time on an IBM-PC XT was approxi-
mately 6 minutes. This time is due to print- The condensate option in BOAST-DRC was
er speed more than computer processing. valldated by running t~~ third SPE compara-
Relatively short run times are typical for tive solution problem. One two-component
tank and 1-D models. sfmulator and several multicomponent simula-
tors were originally used to model the cyc-
3-o PRIMARY DEPLETION FOLLOWED ling of lean gas through a 3-O, retrograde
BY HATERFLOOD condensate reservoir. Two operating scen-
arios were considered: (1) constant sales
The second example is a more realistic rate to blowdown, and (2) gas sales deferral.
application than the first. It Involves prf- The volumes of injected gas in both cases are
mary depletion at a user-speciffed oil pro- the same, but injection rate schedules
ductfon rate of an fnftfally undersaturated differ. Case 2 has a larger gas injectfon
reservofr. The oil has an API gravity of 59 rate than Case 1 for the ffrst five years;
and an fnitial gas volubility of 1270 SCF/ and a smaller gas !njectlon rate for years
STB. The reservofr pressure fs allowed to go sfx through ten. Blowdown (production with-
below bubble point and a gas cap forms fn the out fnjection) occurs in years 11 through 15
upper of two layers. Four rate controlled for both cases. Fluld property data for
water fnjectfon wells are drilled and com- BOAST-DRC and NBOS5 are constructed from a
pleted in the lower layer of the reservofr. two-component material balance match of con-
The ensufng pattern is a five spot wfth the stant volume depl~~ion data sfmilar to that
producer at the center. The producer fs described by Coats .
placed on bottom-hole pressure control and is Results of the runs are shown in Figures
allowed to produce until water breakthrough. 6 through 9. Run tfme is approximately 20
Bubble point pressure is allowed to vary in hrs. on an IBM-PC XT. A direct matrfx solver
this problem. More details defining the was used because the LSOR methods took too
problem are presented in Table 5. many iterations for convergence. NBOS and
Results for both the chain rule expan- BOAST-DRC results differ somewhat because
sion option and the mass conserving expansfon each program calculates gas compressibility
option are summarized in Table 6. This exam- usfng dffferent physical assumptions. NBOS
ple shows that efther expansion can be used calculates gas compressibility for a wet gas,
to solve the problem. The slight material whfle BOAST-ORC calculates a dry gas compres-
balance improvement with the mass conserving sibflfty. The actual gas compressibility
expansion fs hardly worth the four-fold will be a value that ranges between wet and
increasz in computer run time required by the dry gas values. Thus, the two-component
iteratfve method. The chain rule expansfon approach used here is approximate.
generally gfves acceptable results for llquid Overall results for the two-component
domfnated systems like waterfloodfng . black- sfmulators are comparable to the results of
011 reservoir. the more sophfstfcated, and more tfme-consum-
fng, compositional simulators used in Refer-
6AS INJECTION INTO ANUNOERSAWMT’ED ence 11. Incremental production by gas sales
RESERVOIR deferral obtained wfth NBOS at the end of 15
years of production is about 144 MSTB; BOAST-
Odeh10 has pu~l!shed a black.oil simula- ORC had a :inal incremental production of 178
tor comparison problem that demonstrates the MST8. These values are withfn the range of

1A9
S?E 15297 J. R. Fanchi 7

incremental oil reported in Reference 11. 6. Thurnau, D. H.: “BANDSOLVE”, Comm. ACM,
Note that both NBOS and BOAST-DRC revaporize Vol. 6, August 1963, pg. 441;~ Algo-
condensate more rapidly than equation-of- rithm 195.
state based compositional simulators, ~2This
is typ{cal of two-component simulators. 7. Tsau, J. S., Hill, A. D., and
Cumulative material balance err.w for Sepehrnoori, K.: “Modeling o? Permea-
BOAST-DRC was less than 1%. NBOS had a neg- bility Reducing Vertical Conformance
ligible material balance error because it Treatments”, paper SPE 13569, presented
employs a mass conserving expansion. BOAST- at the International Symposium on Oil-
ORC is faster than NBOS because the chain field and Geothemnal Chemistry, Phoenix,
rule expansion requires only a single pres- April 9-11, 1985.
sure distribution calculation. When appli-
cable, the faster chain rule expansion formu- 8. Aziz, K. tind Settari, A.; Petroleum
lation is an advantage for IBM-PC applica- Reservoir Simulation, Elsevier Applied
tions. The price for the speed advantage is science Publishers, Condon, 1979.
a material balance error that must be care-
fully monitored. 9. Shank, G. D. and Vestal, C. R.: “Prac-
tical Techniques in Two-Pseurlocomponent
IX. CONCLUSIONS Black Oil Simulation”, paper SPE 15156,
presented at the SPF Rocky Mountain
1. Practical reservoir simulation problems Regional Meeting, Billings, May 19-21,
can be solved using a microcomputer. 1986.

2. The single pass chain rule expansion 10. Odeh, A. S.: “Comparison of Solutions to
method can be reliably used to solve a Three-Dimensional Black-Oil Reservoir
many types of reservoir simulation prob- Simulation Problem”, J. Pet. Tech., Jan-
lems as long as cumulative material bal- uary 1981, pp. 13-25.
ances are monitored carefully.
11. Kenyon, D. E. and Behie, A.: “Third SPE
3. The conventional mass conserving expan- Comparative Solution Project: Gas
sion method should be used for modelling Cycling of Retrory’adeCondensate Reser-
gas injection into a black-oil reser- voirs”, paper S?E 12278, presented at
voir. the Reservoir Simulation Symposium, San
Francisco, November 15-18, 1983.
4. BOAST-ORC helps fill the gap between
simple material balance calculations and 12. Coats, K. H.: “Simulation of Gas Con-
sophisticated mainframe simulation stu- densate Reservoir Performance”, J. Pet.
dies. Tech., October 1985, pp. 1870-18~

REFERENCES NOMENCLATURE

1. Fanchi, J. R., Harpole, K. J., and Gas formation volume factor


Bujnowski, S. W.: BOAST: U. S. DOE :9
Research Report DOE/-3-3. U. S. Oil formation volume factor
Department two volumes, B: Water formation volume factor
Bartlesville~ Okla~~~~~’1982.
Gas compressibility
C9
2. Fanchi, J. R.: “Analytical Representa- Co Oil compressibility
tion of the Van Everdingen-HurstAquifer
Influence Functions for Reservoir Simu- Rock compressibility
Cr
lation”, Sot. of Pet. Eng. J., June c Water compressibility
1985, pp.-b. w
Onum Numerical dispersion
3. Dietrich, J. K., and Bondor, p. L.:
“Three-Phase Relative Permeability EL Block mid-point elevation
Models,” paper SPE 6044, presented at GOC Gas-oil contact
the SPE 51st Annual Fall Meeting, New
Orleans, Oct. 3-6, 1976. K Permeability

K Permeability dyadic (usually a


4. Lantz, R. B.: “Quantitative Evaluation
diagonal tensor in applications)
of Numerical Diffusion (Truncation
Error),” SOc. of Pet. Eng. J., September k Gas relative permeability
rg
lg71, pp. 315-320; Trans. A3ME, Volume
k Oil relative permeability
251. ro
k Water relative permeability
5. Fanchi, J. R.: “MultidimensionalNumer- rw
ical Dispersion,” Sot. of Pet. Eng. J., L Left hand side of gas equation,
February 1983, pp.17fTl-lsI. !3 Table 1

*4Q .
H ROAST-ORC: Black Oil and Condensate Reservoir Simulation on an IBM-PC XT SPE 15297

L Left hand side of oil equation, Po Oil phase density


o Table 1
Pose Oil phase density at standard
L Left hand side of water equation, conditions
w TabTe 1 Water ~hase density
Pw
Pcow Oil-water capillary pressure Water phase density at standard
Pwsc
Pcgo Gas-oil capillary pressure conditions
Pg Gas phase pressure Gas phase viscos;,y
Po Oil phase pressure Oil phase viscosity
Pw Water phase pressure Water phase viscosity
P Pressure Time-step size
PGOC Pressure at gas-’lilconta-+ Grid block size
PI Productivity ind~.x
Pwoc Pressure at wat~r-oil contact
ACKWULEOGEUENTS
Gas rate
‘cl I t!lankD. E. Kenyon and H. Kazemi for
q~ Oil rate
useful discussions about the implementation
Qw Water rate of a condensate option and a mass conserving
expansion option.
Qg Table 2

Q. Table 2
Liquid content of condensate APPENOIXA: EXPANSION W ACCUMULATION TERM
‘s

R Gas volubility in oil Accumulation terms have the form At


so
R Gas volubility in wate- (ah). A consistent, mass conserving finite
Sw difference expansion is
Gas phase saturation
‘9
so Oil phase saturation
Sw Water phase saturation
t Time
v Speed of frontal advance
Zquatian (Al) can be rearranged to give
u: Well injectivity
Water-oil contact
n+l bn+l - an bn
woc At (ah)= a (A.2)
z vertical direction increasincI At
positively downward
which is the proper expansion. The chain
Porosity rule expansion of At (ah) used in BOAST-DRC
Gas phase potential is
Oil phase potential

‘9
A.
Water phase potential
= krglvg; gas phase mobil ty
= kro/~o; oil phase mobi? ty
‘t(ab’=an[~l+bn[-
.
h.
In general, Eq. (A.3) is not equa? to
u = kr~~w; water phase mob’lity If either a:~ or bn crc slowly
Eq. (A.2).
varying functions, or the time-step size At
Pg Gas phase density
is small, then Eqs. (A.2) and (A.3) are equi-
Pgsc Gas phase density at standard valent. When Eq. (A.3) is not a good approx-
conditions imation, material balance error occurs.
wE J5?97

TABLE1. MassConservation
F1ow Equations TABLE2. Associated
Variable
Definitions

Saturation
Constraint: l=so+sw+sq
ho ?.. Q.
oil: L= 7 *K* V@O+V. fe~r v*-_
o o Pose Rates: Q. = qo + ‘Sqg
9s9
Qg = Oq + Rsoqo+ Rswqw
so s
= ~[+; + ‘s2 )]
q
‘g;J !3.2
PhasePotentf
als: @=p
00 -m
iv qw
Hater: L = v. F!*v @—_ Pwz
w q w PWsc @w= p-p
o cow - m

a
$Sw pgz
. ‘5” PO+P
5 8; 9 Cgo -m
[1
?. k A Qg Capfllary
Pressures: pcow= p. - Pw

1
o w
Grs: Lg = v*Ko:
“ v+R vQo + R —w -..
so ~ sw Bw w
f’q. :
[ 99 Pcgo‘ ~g - Pw

s so Sw
. a+ g+R +R
5 ~ so ~ Sw q
[( )1

TABLE3. PhaseOenslti”s
andCompresslbfl
fties TA8LE4. Buckl
eY-Leverett
LfnearHaterfl
oodat 360Days

1
1
Oensittes: P =—p +R
o B. Osc so ‘gsc
[

1
P=— [
w Bw ‘WSC
+R
Sw ‘gsc
1 1 5117 .270 .730 5116 ,27’) ,7;,

. ,,! :
1 2 5048 .301 .699 5047 .3[,1

‘g
=—p
Bg [
gsc
+rp
s Osc
1 3 4958 .311 .689 4957 .!:i ,, :..,

4 4864 .327 .673 4863 .3,!7 ,..,1;

1 a+
Coapressfbflf
tfes: c ,!,r),
. r ‘F% 5 4763 .349 ,651 4762 ,349

6 4652 .379 .621 4651 .J7fl .6?1


~ aB B. ars
c
!3
.._

~
[
___9
6q apo

aB
_
.lg apo

09 aReo
1 7
8
4522
4361
,433

.507
.567

.493
4521

4360
.433

.507
.56)
.4[)3

c
o
=-0.

[ B. apo

~ aB
.
B. apo

Bg aR5w
1 9
10
4173
4054
.650 .350
.736 .264
4172

4054
.650

.736
.350

.264

c
w
=-
[
_2-_—
Bw apo Bw apo
1 al Balance = 1.0000
Cumul atfve Of1 Materf
Cumulative GasMaterialBalance = 0.9995
Cumulat{~e MaterMaterfalBalance= 1,0000
TABLE5. InputDatafor3-DNaterfl
ood Ex~l e

NX=NY=5; NZ=2
DX = DY = 500 ft.
TABLE6. Resultsfor a 3-D WJte+lood
InitialSo = 0.88
Initial~= 0.12
UellhoreRadius= 0.25ft.
Skin= 0.0
Year Pressure Gas Saturation Productil Hell Rates
Block [5, 5, 2) Block (5,5,2) Oi1 Gas
GRIO OATA (psia) (fr) (STB/O) (Mscffo)
1

Layer DZ Node Oepth Kx K Kz Initial P CR:* MCE** CRE NCE CRE MCE CRE MCE
o
(ft.) (ft.) (red.) (id .) (red.) (fr.
) (psi)
@.5 3416 3425 .058 .058 1125 1155 1882 1942

1 20 8335 20 20 5 .12 4317 1.0 3231 3236 .067 .067 566 578 1327 1372

2 30 8360 100 100 5 .28 4384 1.5 3233 3234 .048 .049 607 609 1677 1714

2.0 3223 3221 .035 .035 620 613 1717 1755

2.5 3227 3224 .025 .025 663 651 1424 1455

HELL DATA 3.0 3271 3265 .020 .020 824 807 1175 1160

Hel 1 ~id Locatfon BHFP* Max Rate Years Active 3.5 3306 3303 .021 .021 935 925 1257 1250
JK (psi) (STB/0) (Start/Stop)
4.0 3340 3329 .021 .021 1052 1015 1303 1266

P-1 33 1-2 3000 2000oil 0/5 4.5 3365 3363 .021 .021 1126 1121 1350 1343

1-1 112 - 500 water** 1/5 5.0 3375 3375 .021 .021 1136 1137 1346 1349

1-2 152 - 500 water** 1/5

I-3 512 - 500 water** 1f5 * Chain Rule Expansion


*Wass Conserving Expansion
I-4 552 - 500 water** 1/5
Cumulative Material Balances CRE MCE

of1 0.99981 0.99975

G&s 0.99145 0.99667

* Bottom hole flowing pressure of pressure-controlled, rate constrained k~ter 0.99914 0.99979
oil production well.
●*Negative rate denotes injection B7900 Cormuter Run Time (CPU seconds) 99 440 w

l?el~tive Permeability and PVT Oata from Example 4 Of Reference 1. IBM-PC XT Computer RunTime(CPUseconds) 3463 --- m

.
SE :! 52.97

TA8LE7. LeanGas Injection


Intoa Black-Oil
Reservoir:ChainRuleExpansion
Option

Time Produced Oil Produced GOR Gas Sat in Pressure (psia) in Pressure (psia) in
(Yrs.) Rate (t4STlJ/D ) (Hscf/STB) Block (10,1 O,3*) Block (10,10,3) Block (1,1.1)**

0.0 20.00 1270 0.000 4800 4783

1,0 20.00 1302 .022 4022 5623

2.0 20.00 1356 .031 3867 5?55

3.0 20.00 3928 .150 3727 5065

4.0 9.71 11200 .210 3406 4524

5.0 7.57 15913 .7:8 3288 4292

6.0 7.01 17138 .232 3228 4182

7.0 6.33 !9425 .238 3182 4076

8.0 5.63 21880 .244 3106 3954

9.0 5.17 23383 .247 3036 3840

10.0 4.81 24462 .250 2968 3741

*Production Wel 1 B1 ock


**Injection Heal B~o~k

Cumulative ofl material balance = 0.9454


Cumulative gas material bala~ce < 0.9
Cumulative water material blance = 0.9998

TABLE8. LeanGas Injection


Into a Black-Oil Reservoir: Mass Conserving Expansion Option

Time Produced Oi1 ~roduced GOR Gas Sat in P(psia) in P(psia) in


(Yrs. ) Rate (MSTB/0) [Mscf/STB) Block (10,10,3) Block (10,10,3) Block (1,1,1)

0.0 20.0 1270 0.000 4800 4783

1.0 20,0 1284 0.000 4438 6182

2.0 20.0 1270 0.000 5154 6725

3.0 20.0 1320 0,000 6022 7179

3.8 20.0 7820 -. .. -.


4.0 17.9 8566 0.186 4433 6506

5.0 13.0 10688 0.204 3952 5527

6.0 10.1 12979 0.216 3641 4912

7.0 8.3 15364 0.225 3423 4523

8.0 7.0 18124 0.234 3294 4282

9.0 6.5 18692 0.236 3199 4130

10.0 5.8 20240 0.240 3093 3997

Oil Mat Bal = 0.99983


Gas Hat flal = 0.99966
Mater Mat 8al = 0.98614
.

— Saturated
---- Undersalurated ,

B.

/--; ------.--_:

/;
I

Pressure_ pb P max”

&q
‘sO Pressure--D pb P max

1 +
Pressure~ pb pwax 9

Fig.
1-Example of oil PVT date.

/
I
Pressure - Pressure --
Pressure+

FIo.
2—Example of gne PVT ds!a,

—Datum

-----
Y (+)

)’
x
1
f
z
El .
-----
i50-c
-----
woc
-.

Fig. 3-Preseure Initlellzation contacts. 9


~ 15297
.
For
Oepths Increasing Oownward
Calculate: Block BOTTOM = EL + 0,5 ● OZ
Block THICK = OZ
Block TOP = BOTTOM - THICK
Uater zone thickness WTHICK = BOTTOM - WOC
Gas zone thicknes GTHICK = GOC - TOP

Case 1 7 3 4 ——
5 6

GOC rop TOP ‘— Goc rop _ GOC


TOP ;Oc---1’, TOP woc _
TOP
f
Goc
1

:1
IfOc woc _
f
BOTTOM 30T ----r. BOT BOT BOT BOT
woc woc Goc ~
woc
—— —

Sg= 0,0 f GTHICK GTHICK f=l - WTHICK so= 0.0 so= Sg’ 0,0
f=l - —
‘= THICK THICK THICK
f ~. _WTHICK .!
s = sot - SOI*f = 0.0 S = SWGI s = 1,0
o THICK o 9 w w

SW. sw~l Sg= fg ●SG[ ~= (1-f)*sGI .. I-so[*f S9= SGI

so=(l-fg-fw)*soI = 1-s.-s9 soI*f


w 0=
SW. 1-s -s
09

----------- ----------------- ------------ -------------- . --------- ------- -----

If so< Sor lf so< Sor If So< Sor


, , *
then then then
so= 0.0 so= 0,0 so= 0,0

fg ●sG[
s= S = 1. O-SUG1 s = 1.0
9 ~w) 9 w

Sti= 1 - 59 SW=SWGI

Fig. 4-Saturation Inltlalization by vertical equilibrium.

0 BOAST-DRC and NBOS


x
are indistinguishable

~- Analytical
Solution
1.o-

0.8r
c
0
%:-. * x nsxxx
‘$ o,~ 360 days
3
. ~120 days K ~’

ii al s
(n
~ 0.4 - y x
I
$02 10 x
L-LL--a--__4 _s--..s —

o.o~
0.0 1.0
DimensionlessLength
Injector Producer

Fig. S-ftucklay-Leverett waterflood dieplaoement.


100o- 1 I I 1 I I I I I

~ BOAST-DRCfCASEl >. A-
2700
e~ BOAST-DRCICASE2 >/
“\, ~.—.o NBOSICASE 1
\. 2400 //“ .>--’
800 -
“\ I / > s- ---S--3------
A.
“\.
700 – \ 21O+J
P’ .“~:.”~”-”2”
O“\.\
/f ““;~
‘\ 1800
600 – / “’”/”
n m ///’ /./
>500 – !jj 1500 // ,
z
c1 #::P”
1200
400 – ‘.
<
.. .
300 - 900 ,/”’
.
~ BOAST-DRCICASE 1
200 – 600
~ A BOAST- DRcl CAS.E 2
b
o—-o NBOS!CASE 1
100 300
1~ x----x NWSICASE 2
t 1

00
1 1.5
I [
3 4.5
I I
6
I
7.5 9
I 1 I
12
~, 0
o
I
1.5 3
I
4.5
I 1
6
I
7.5
I
9
I
10.5
I
12
I
13.5 15
I

YEARS OF PRODUCTION YEARS OF PRODUCTION

Fb. 6-Stock.tmkol! rntc. Fb. 7-&ack-Unk dlpmducul.

360 ~ BOAST-DRC
t h~ NBOS 1
320

0.42
t 1 280

240

:~4,~~d,,,i
[

; 20+3
> ~+y
160
—++—+ — -.s-+-4
//
120
/f
60
/’
40 /
//<.,’:’,: ‘

o
o 1,5 3 4.5 6 7.5 9 10.5 12 13.5 15 0 1.5 3 4.5 6 7.5 9 10.5 12 13.5 15
YEARSOF PRODUCTION YEARSOF PRODUCTION

W. @-ml 8Mumli011In Pradudm Wdl Block (7.7,4),


Fig.9-lncmmOmal cum-.mpnntuctkm *z-clscl].

**”D

You might also like