Compositional Modeling

With an Equation of State

L.X. Nghiem, SPE. Computer Modelling Group
D.K. Fang, * SPE. Computer Modelling Group
K. Aziz, SPE. Computer Modelling Group

Abstract
This paper describes an implicit-pressure explicit- simultaneous solution of II h (v + I) nonlinear equations,
composition and explicit-saturation compositional where II" is the number of grid blocks and v is the
model. The model uses an equation of state (Peng- number of components in the hydrocarbon system. The
Robinson) for phase equilibrium and density calcula- explicit treatment of the transmissibilities limits the size
tions. Interfacial tension effects are considered also. The of the allowable time step. On the other hand, Coats pro-
formulation of the pressure equation yields a symmetric posed a fully implicit equation-of-state compositional
and diagonally dominant matrix that allows the use of the model. His fOffi1Ulation requires the simultaneous solu-
iterative conjugate gradient method for large systems. tion of 11,,(21' + 4) equations. Although the stability of
Simulation of laboratory CO 2 displacements shows Coats' model is better than that of Fussell and Fussell's
good agreement between ealculated and experimental model, the computational cost of Coats' implicit model
results. The int1uence of interfacial tension is in- may become prohibitive for systems containing a large
vestigated. Physically reasonable results also have been' number of grid blocks and components.
obtained for hypothetical areal and cross-sectional This paper presents an implicit-pressure, explicit-
problems. composition and saturation equation-of-state composi-
tional model that is a variation of that proposed by
Introduction Kazemi et al. 10 The fOffi1Ulation of the pressure equa-
The design of high-pressure-gas, enriched-gas, or tion yields a symmetric and diagonally dominant matrix
COrinjection schemes requires an accurate prediction that allows the use of the iterative conjugate gradient
of the vapor-liquid equilibrium between the oil-in-place method for large systems. The diagonal dominance is
and the injected t1uid. In recent years, vapor-liquid also a desirable feature for the numerical stability of
equilibrium calculations have been enhanced by the in- direct elimination methods.
troduction of many two-constant equations of state that An efficient method for t1ash calculation II allows the
can be applied to both the vapor and liquid phases. 1-3 use of the model in the vicinity of the critical point and
The application of these equations of state to petroleum avoids the computation of the saturation pressure in
reservoir t1uids was made possible by the ability to deteffilining the single-phase region.
evaluate the parameters of these equations from proper- A new model for relative peffileabilities that is depend-
ties of the heavy fraction-i .e., density, average boiling ent on interfacial tension is proposed and the sensitivity
point, and molecular weight 4 -6 _which can be measured of the recovery to interfacial tension is discussed.
easily. A well model that allows the simulation of constant
Two compositional models that utilize an equation of volume, constant injection/production wells and
state for phase equilibrium and properties calculation are multiblock well completions is provided.
described by Fussell and Fusse1l 7 . 8 and Coats. 9 When The Peng-Robinson equation of state 3 is used in all
both gas and oil are present in every grid block, the for- examples, although the solution method is general and
mulation of Fussell and Fussell requires the applicable for any other equation of state.

Mathematical Model
• Now with Shell Canada Resources Ltd
Flow Equations
0197· 7520/81/0012·9306$00 25
Copyright 1981 Society of Petroleum Engineers of AIME Assuming Darcy's law is valid, the material balance on

DECEMBER 1981 687

 the water phase and v hydrocarbon components yields PgS g
the following v + I difference equations. V= , ........................ (8)
PoS o + PgS g
The fugacities, Imo and Img, and molar densities, Po
and P g' are computed at pressure Po'

Constraint Equations
The definition of mole fractions and saturations gives the
and additional constraint equations:

Il [Tllyll
o mo (Ilpll0 + 1 - "0111
101-UJ
An) + TilgYll1g
II 2..: zm = 2..:Ymo = 2..:Ymg = I, ............. (9)
III m 111

L + V= I, .............................. (10)
and
So +Sg +S", = I ......................... (11)

Each grid block will have its own set of algebraic

equations (Eqs. 5 through 11). The coupling of the
variables between adjacent grid blocks is through the
Flow Eqs. 1 and 2.
m = I ... v . .......................... (2)
Solution Method
The molar balance equation of the hydrocarbon system
is obtained by summing Eq. 2 over v hydrocarbon Outline
components: Eqs. 1 and 2 and 5 through 11 represent a system of
n h (3v+ 6) equations in n h (3v + 6) unknowns (p o.i'
Il[T;:(IlP~+ 1 - 'Y;;IlD) + T:J(llp~+ 1 zm.i' Y1l1o.i' Ymg.i' L i , Vi, SIt'.i' SO.i' and Sg.i;
i= 1 ... nh, m= I ... v), where nh and v are the
number of grid blocks and the number of components in
the oil/gas system, respectively.
The equations are solved using an iterative-sequential
method. Eqs. I and 4 are combined to fonn the pressure
equation, which is linearized and solved with either a
direct or iterative method. Once P);'+ I), the (f+ l)th
iterate of p;:
+ I, is obtained, Eq. I is solved for Slt+ I)
and Eq. 2 is solved for zli; + I). Flash calculations then
are perfonned to obtain ),(1'+
1110
I) ,,(1'+ I) V(I'+ I) L(I'+ I)
'JIll,R' , ,

where p)}'+ I), and P~I'+ I). Finally, Eq. 7 or 8 and Eq. II are
used to obtain S)}"+ I) and S~I'+ I). The procedure is
qh = 2..:qll1' ............................. (4) repeated until convergence is achieved.

Pressure Equation
Equilibrium Equations
The pressure equation is obtained by multiplying Eq.
It is assumed that interphase thennodynamic exchange in by a constant parameter 8 and adding the product to Eq.
the reservoir is rapid compared with fluid flow so that the 4:
gas and oil are in phase equilibrium. It also is assumed
that mass transfer between the water and hydrocarbon Il [8TII.( ll + 1 _ UApllCH'O
\t llp () IlD) + Til() (Apfl
1l
- 'V .
1\1 U () + 1
phases is negligible.
The thennodynamic equilibrium conditions are given
by the equality of the component fugacities in the oil and
gas phases-i.e., - I'll
o IlD) + Tilg (Ilpll() + I +LlApl1cog II AD)]
- 'YgLl

11110 =Il11g' m =I ... v . ..................... (5)

V
From a material balance on the oil and gas phases, the +8qlt. +qh - ~(<f>I1+IQ/l+1 - <f>lI ex ll) =0,
following algebraic equations are obtained. Ilt
Zm = LYlI10 + VYlI1g' m = 1 ... v, ............. (6) .................................. (12)

L = PoS o ........................ (7) where

PoS o + PgS g ,
and ex = 8p\\.S", + PoSo + PgS g . ........•....... (13)

688 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 Since the units of Eq. I are moles of H 2 0 per time and the derivative of a, saturations and compoSitions are
those of Eq. 2 are moles of hydrocarbon per time, it is assumed constant while all the pressure-dependent terms
important to scale these equations appropriately before are differentiated properly. Saturations and compositions
adding them together. Better convergence can be a- are updated explicitly in a and oalop 0 after each itera-
chieved by introducing the scaling factor, fJ, which con- tion. The iterative process is intermediate between sim-
verts moles of H 2 0 into equivalent moles of ple substitutions and Newton's iteration.
hydrocarbon. The porosity and molar densities of water, oil, and gas
In the program, 0 is evaluated at time zero and at a grid are related to Po by the following equations.
block specified by the user using the following equation.
¢ = ¢*[I + cJ(Po-p/')]. ................... (18)
_ PoSo + PgSg I
0- Pw(So + Sg} [ = O· ................... (14) PI\' =PI;'[1 +c",(po-P,;')]. ................ (19)
Typical values for molar densities in reservoir
eng~neering ~roblems ~re: P ~ = 6 x 10 3 3 g mol/m ~ , Pj = zP; T,j=o,g . ...................... (20)
P g - 12x 10 g mol/m 'PH' - 55.6 x 10 g mol/m .
}
Using Eq. 14, 0 usually falls in the interval (0.1, 0.2).
Hence, the derivative of the accumulation term in Eq.
For practical purposes, any value of 0 in that range can
16b is given by
be used instead of computing it from Eq. 14.
o¢a ) (!)
Pressure Equation Solution oPo- I. = cj 1>*a(l)
(- I
+ '+'1
-i-.(!) (OS' *
11'( wP w

Let Fi denote the left side of Eq. 12 for the ith grid block
and letp(f+0,1
I) be the fth . iterate ofp"-+:
(},I
I. The (f+ I)th :.l
UP o
:.l
uP g )
(n
iterate can be obtained from the following Newton-like +So--+Sg-- , ...... (21)
iteration. oPo oPo i
where
OPj = _1_(I_ Po OZi),
j = o,g . ........ (22)
oPo R T Zj Zi oPo
= -F(n
I , I'-1
- ... nh, .............. . (IS)
Since op;lop 0 is positive, it follows that o¢alop 0 is
also positive. Furthermore, it is shown later that
where J ik is the ikth element of an approximation to the oq ".lop 0 and oq ,,lop 0 are negative or zero, depending
Jacobian described below. The sum in Eq. 15 is over the on the type of well. Examination of Eq. 16 shows that
ith grid block and its adjacent blocks. TheJ ik 's are com- the Jacobian is symmetric and strictly diagonally domi-
puted as follows. nant. The formulation of the pressure equation in Ref. 10
of (i, does not yield a symmetric matrix nor does it guarantee
= ( __ 1_)
= (OT". + To + Tg) + k)/2 ;li diagonal dominance.
oPo k The pressure equation is solved using either direct
.................................... (l6a)
elimination or the iterative conjugate gradient method
with incomplete Choleski factorization ICCG(O) as pro-
and posed by Meijerink and van der Vorst. 12

Composition Equation
Composition is computed explicitly at every iteration
after piJ'+ I) has been detemlined. Let

................. (23)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (l6b)
and
where the subscript (i + k)/2 denotes the interface be-
tween grid blocks i and k.
In ~q. 13, P () is a function of p 0 and Y 1110' and P g is a
functIOn of p 0 and YlIlg' The actual derivative of a with An explicit composition equation is obtained by re-
respect to p will involve the temlS oS ".lop 0' oS,/op 0'
(! arranging Eq. 2:
OS~ lop 0' oy /nO lop 0' and oy II1g lop 0' which are com-
plicated to compute and will yield partial derivative
terms in th,e off-diagonal elements. An approximation to
zU+
III
1) = rl .1 [Til""
o. 1110
.1<1>(1'+
()
I) + Tilg.\,"mg .1<1>(('+
g
I)]

(oalop o}}U is obtained by neglecting these terms-i.e.,

o
a)
(i')
_ (
:.l
up"
:.l
up 0
:.l
up~ )
(!)
( -- - OS".-- + So-- + S~--
oPo I oPo oPo 'oPo I

.............................. (17)
In solving nonlinear equations, an exact Jacobian is -'-
·.1t
[~¢(i'+ 1) (p 0 S0 + P g Sg ) (i'+ I)] . (25)
not required for convergence. In the approximation of

DECEMBER 1981 689

 An approximation to the denominator in Eq. 25 is ob- Therefore, it is arbitrarily named either "oil" or "gas"
tained from Eq. 4: depending on whether the value of V obtained from flash
calculations is zero or one.
~-/-(I'+I)( S + S )(1'+1)
~t ~ Po 0 Pg g
Flow Coefficients and Capillary Pressures
This section describes the methods used for evaluating
the flow coefficients (T"Ylllo' TgYlIIg' and Til') and the
= ~[TI1~<I>(I'+
() () A~(I'+ I)J
I) + TilgU'¥g capillary pressures.
By definition,
A k k rj
T j = - - - P j ' j = o,g,w, ................ (32)
V" -/-11 (p",,+P
S S ) . M iJ.j .
+qh+-~ gg
11
. . . . . . . . . (26)
~t where A is the area of the grid block perpendicular to the
Composition as computed previously is guaranteed to direction of the grid block length, M.
.satisfy A and ~f are evaluated at the interface between grid
blocks and absolute permeability k is evaluated using the
'\'~(I'+I) - I harmonic average. Relative permeability k r' viscosity iJ.,
LJ4..-m -.
111 and density P are evaluated at the upstream block.
The method is equivalent to solving Eq. 2 for the number Under immiscible conditions, gas/oil relative
of moles of m and Eq. 3 for the total number of moles of permeability curves usually show considerable curvature
hydrocarbon, and obtaining zll1 by taking the ratio. and there are residual oil and gas saturations below
which the respective phases are immobile. As the in-
Saturation Equation jected fluid and the oil-in-place become gradually misci-
ble through multiple contact, the interfacial tension be-
Saturations also are computed explicitly at every itera-
tween the oil and gas decreases toward zero. This will
tion. From Eq. I:
make the relative permeability curves approach straight
~T::.~<I>ft+I)+ql1.+ V"¢11 11511
lines without any residual oil and gas saturations.
(I' 1 l\ At P II \\" The following relative permeability model that ex-
S\\"+ )= --------~'""~_ _ __
hibits the described behavior at low interfacial tension is
used. Let k rg and k rag be the relative permeabilities in
the oil/gas system measured under immiscible condi-
tions. Relative permeabilities k rg and k rag at interfacial
................................. (27) tension a are computed using the following equations.
Oil and gas saturations for the (e + I )th iteration are
computed by performing a flash calculation on ziiz + I)
m = I . . . II atp)[+ I) to obtain ~vmo
(1'+ I) , ),(1'+
mg
I) ' (p(c+
)
I)
,
P,~'+ I), V(I'+ I), and LU'+ I). If the results of the flash
yield a two-phase system, the oil and gas saturations can .(~)I+h,r,
................... (33a)
be computed from Eqs. 7, 8, and II: I - S wr

s)1'+ I) = [(I - S",)Lp g (+ I) ••.......•..... (28) k rog = ( I-e" r)k- rog + e "rk mill"
-{l -{l

Lpg + Vp"
and
.(~)I+h"r,
.................. (33b)
.................. (29) I - S I1'r

If a single-phase system is predicted, then, depending

on the nature of the phase, a
r= *'
a
................................ (34)
S<f+I)[orS(I'+I)]
(} . g = I-S<C+
w.
1) . . . .. . . . . . . . . . (30)
a g >0, ................................ (35a)
and and
a O >0, ................................ (35b)
S~+I)[or S~+I)] =0 ...................... (3\)

An efficient flash-calculation method described by where a ()' a g' b ()' an~ b g ar~ ad~ustab\e parameters. a*
Nghiem and Aziz II is used. It provides convergence in is a specified interfacial tensIOn mtroduced for conven-
the vicinity of the critical point and avoids the computa- ience to make r, a g , b g , ao' and b o dimensionless.
tion of the saturation pressure to determine the single- Common values used for a* are in the range of I to 5
phase region. dyne/cm. Eq. 33 is based on the requirement that the
The attribution of the term "oil" or "gas" to a single- relative permeabilities to oil or gas in the presence of ir-
phase fluid around the critical point is meaningless. reducible water saturation are equal.
690 SOCIETY OF PETROLEUM ENGINEERS JOURNAL
 Note that for values of a such that e ~(/,r = 0 and Eqs. 39 and 40 apply to both oil and gas compositions.
e ~(/"r =.0, k rg' and k rog are equal to f rg' and frog' If only one hydrocarbon phase is present in grid block
respectively. As a decreases toward zero, k rM and k ror; i-I, the global composition, z/II' is used instead ofY/II in
approach the respective straight lines k roil\' [S/(I-S "'r») these equations.
and kroi".(Sj(l-S"T»)'
The three-phase oil relative permeability is estimated Well Models
usingAziz and Settari's version 13 of Stone's equation: This section describes the formulation of rate- or
pressure-constrained injection/production wells, and
k ro -- k roil\' [( --.-
k roll' + k n ,· ) ( --.-
k rog + krg ) multiblock well completions. The rates are positive for
k TO/'" k rol\" injection wells and negative for production wells by
-(k nt + k rg ) J. ..................... (36)
conventions.
The molar flow rates for water and for component In
into or from the ith grid block containing a well are
!he values of f rg , frog' k n," and k roll' are computed
uSing either tabulated data or the following equations. q". = p(:.Q(: . ............................. (41)
-
k rg = k roi,,'
( Sg-Sgr )'"
. . .............. (37a)
and
I-S gr-S",/,
q /II -_ ) ,111Il() plIQI1
0
IIQII
0 + Y /IIg Pg
11
g, ............... (42)

where Qj is the volumetric flow rate of phase

f. =k. (I-Sg-Sorg-S".r)"'" jU = o,g,w).
rog ro/'" I-S -S . . ........ (37b)
org wr For constflnt-rate injection wells, Qj'S are specified
and the compositions and densities in Eqs. 41 and 42 are
those of the injection fluid at P:~.i·
= (SII,-S"T )"" For constant-rate production wells, the compositions
k nt· krl\'ro ............. (37c)
I-S"T-SOnt and densities are those of the ith grid block.
For constant-pressure injection wells, the volumetric
flow rates (Qj's) are determined from
I-S",-Son,' )"''''
kro" = k rIJ/ ,,· ( I-S -S . . . . . . . . . . . . (37d) Qj = Iwj(Phh-P:U I), PM >p:;t I, .......... (43)
\tor orH'
with
T~e capillary pressure, P cog' at the ~as/oil interfacial
tensIOn, a, IS related to the capillary P CO!; (immiscible 271'khfAI
IWi = , ..................... (44)
condition) by the following equation. . r q
P cog -_ (I -e ~cr)-P('og,····.················ (38) In( C e ) + s
rlt'
where c is an adjustable positive number.
The gas/oil interfacial tension is calculated using the where C is a factor that takes into account the geomet~
Macleod-Sugden correlation 14 and the viscosities of oil of the grid block and the location of the well within it I
and gas are computed from a modified Jossi, Stiel, and and AI is the total mobility of the fluids in the grid
Thodos correlation. 14 block. 17
The compositions Y mo and Y m" which multiply To For constant-pressure production wells, the flow rate
and Tg in Eq. 2, also can be evaYuated at the upstream ofphasej is given by Eq. 43 withPhh less thanp~U~ I and
block. The use of this single-point upstream formulation AI in Eq. 44 is replaced by Ai' the mobility of phase j in
for composition may cause excessive numerical disper- the grid block. .
sion. To reduce the numerical dispersion, a two-point All the compositions and properties in Eqs. 41 through
upstream weighting for composition, similar to the 44 are computed at time level n. Hence, it can be shown
development of Todd et at. 15 on relative permeabilities, easily that aq ,JapZ + I = aq h lapZ + I = 0 for constant
is introduced. rate wells and that
Let i-I and i-2 be the two upstream blocks of grid
aq", '
block i. The two-point upstream composition at the inter- an+' = -/wwPw ....................... (45a)
face i-'/2 is obtained by extrapolating the composition in 'Po
grid block i-I and i-2 to that interface: and
Mi~1 aqh
Ym.i~1/2 = Ym.i~1 + flf. + flf. (Ym.i~I-Ym.i~2)' an+' = -/woPo-lwgPg ................. (45b)
/~I /~2 'Po
................................ (39) for constant bottom hole pressure wells.
where Mi~ I is the length of grid block i-I. To avoid In many field-simulation problems, the well is com-
overshoot and undershoot problems, the following con- pleted through several grid blocks and the total rate of
straints are introduced. the multilayer well is specified. It is then necessary to
determine the bottomhole pressure and the pressure
Ym.i~'I2 ~max(Ym.i~I' Ym.i) ................ (40a) along the portion of the well open to flow from the reser-
and voir. The bottomhole pressure is the pressure that yields
the specified flow rate. A schematic of a multiblock
Ym.i~'I2 ~min (Ym.i~I' Ym.i)' ............... (40b) completion well is shown in Fig. 1 where, for conven-

DECEMBER 1981 691

 TABLE 1-SATURATION PRESSURE AND SWELLING DATA

Recombined
Component, mol% Reservoir Fluid Recombined Reservoir Fluid + CO 2
CO 2 1.64 18.45 37.03 56.79 77.74
N2 2.18 1.81 1.40 0.96 0.49
C1 28.99 24.03 18.55 12.73 6.57
C2 7.97 6.61 5.10 3.50 1.80
C3 7.15 5.93 4.58 3.14 1.62
iC 4 1.21 1.00 0.77 0.53 0.27
nC 4 3.54 2.94 2.27 1.56 0.80
iC 5 2.00 1.66 1.28 0.88 0.45
nC 5 2.12 1.76 1.36 0.93 0.48
C 6+ 43.20 35.81 27.66 18.98 9.78
Properties
C 6+ density at 15.6 °C, g/mL 0.8358 0.8358 0.8358 0.8358 0.8358
C 6+ molecular weight 178.47 178.47 178.47 178.47 178.47
Bubble-point pressure, psia (kPa) 2,300 2,705 2,945 3,515 6,000
(15 858) (18 650) (20 305) (24 235) (41 368)
Reservoir temperature, OF (0C) 206 (96.6) 206 (96.6) 206 (96.6) 206 (96.6) 206 (96.6)
Solution GOR' 746 (132.9) 1,079 (192.1) 1,336 (238.0) 2,662 (474.1) 5,836
(1039.4)
Swelling factor' * 1.000 1.103 1.133 1.487 2083
Calculated Data
Bubble-point pressure, psia (kPa) 2,303 2,584 3,000 3,754 6,157
(15 880) (17 814) (20 681) (25 884) (42 449)
Solution GOR 665 (118.5) 966 (172.0) 1,489 (265.2) 2,567 (457.2) 5,903
(1051.3)
Swelling factor 1.000 1.084 1.227 1.503 2.170
~Cubic feet of gas at 14.7 psia and 60 OF per barrel of oil at 147 psia and 60 OF.
Cubic centimeters of gas at 1 aIm and 15.6 °C per cubic centimeter of 011 at 1 aIm and 15.6 °C
* *Barrels of oil + CO 2 at saturation pressure and temperature per barrel of all at saturation pressure and temperature.

TABLE 2-RECOMBINED RESERVOIR OIL COMPOSITION

Model Actual Composition

Component Component (mol%)
CO 2 CO 2 1.64
N2 N2 2.18
C1 C1 28.99
C2 C2 7.97
C3 C3 7.15
nC 4 iC 4 ± nC 4 4.75
nC 5 iC 5 ± nC 5 4.12
6
CC15 J 19.40
11.90
J 43.20
C 22 11.90

3
t
3 C6
Interaction Coefficients
for C 6+ Fractions
0.105 0.120 0.090
C 15 0.120 0.120 0.110
C 22 0.150 0.120 0.145

2
t
2
~WELL
ELEMENT

1
/ t1
i
GRID BLOCK

Fig. 1-Multiblock well completion.

692 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 ience, the grid block and corresponding well elements That above derivation involves rate constrained multi-
are numbered from one to n e . block well completions. If the completion is pressure
The volumetric flow rate of phase} into the ith well constrained-i.e., P hh, I is specified-the derivation is
element is much simpler. The pressure in each well element can be
obtained by using Eqs. 47 through 49 and the derivatives
Qj.i = III"j.i (p bh,i-P:U I) of the molar flow rates with respect to p;U I can be com-
with i= I ... n e , }=o,g,w ........... (46) puted from Eq. 53 by setting ap h/z.1 lap;U I equal to
zero.
and
From Eq. 54, afbh,lla~IJ I is always less than one.
Pbh.i=Pbh,I-H i , ........................ (47) Hence , aq H' lap"n,t+ and aq,1
lap',;.+1' I are always less than
or equal to zero depending on the type of well. This
where Hi is the hydrostatic head of the fluid column at i guarantees that the Jacobian in Eq. 16 is always
with respect to Well Element I: diagonally dominant.
HI =0, ............................... (48a)
Examples
;-1
A few applications of the model that include one- and
Hi = 2: Pr,k+ ,/og(Dk-I-D k ), ........... (48b)
,= I
two-dimensional examples of CO 2 displacements are
described in this section. The C0 7 /reservoir oil system
and a
of Sigmund et al. 18 is used with different representa-
tion for the C 6 + fraction, which yields a better match of
2: (Pj Qj ) k + '/0 the phase-behavior data. Composition and physical prop-
Pr,k+ '/0 = ---"'------, where} = o,g,lV. erties of the recombined reservoir fluid with CO 2 are
given in Table 1. The phase behavior is predicted using
2: (Qj)k+ ,/, the Peng-Robinson equation of state with components
shown in Table 2 .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (49)
In that treatment, frictional and kinetic energy effects
are neglected in the wellbore. Assuming constant- Slim-Tube Displacement
pressure drawdown for all well elements-i.e., In this example, the effect of interfacia,I tension on
+ I
P/Ihh.i /I + I -
-Po.i + I
- P11hh,k 11 +
-Po., I
were *
h ' k 'I, .....
(50) simulation results is examined. The slim-tube tests were
conducted in a 0.375-in.-OD X 0.305-in.-JD stainless
Eq. 49 yields steel tube, 58.79 ft long, packed with 140-mesh beads
that had a liquid pemleability of about 5.07 darcy. The
2: (Pj I wj ) k + ';'
pore volume of the tube was about 21.85 cu in., which
Pr,k + '/0 = ---"'-------~ ............. (51) corresponds to a porosity of 42.4 %. The injection rate
2: (I II) ) k + ,/, was 0.976 cu in/hr.
The following relative penneability curves were used
}

Let Qj be the specified total flow rate. It can be shown in the simulation.
that -
krg=S~-,
7

2: 2:i IlIj.i (P:}] I + Hi) + Qj f = (S (I -0 . 2 ):2 5

;= I /() 0.8
P h/z, I
2: 2: III).i
I j
; =

................................. (52)
k ~
= (1_,,-aI3.31 )k~
• ~
+ (,-an.31 S g'

The derivatives of the molar flow rates with respect to

p'~)~/ then are given by
k ro = (l_e-aI3.31)f ro + e-a1331 S 0'

aPO,I
a~l~ I = I"'".i P::',i ( :P::'~':
Po"
-I) ....... ,. (53a)
where (J is in dyne/cm. The value of 3.3 I corresponds to
and the interfacial tension in dyne/cm between the vapor and
~
a ,,+ I
- (I
-
'I.
lI'O.i PO ,1
+1 Il'g,IPg,1
. 11 . ) ( ap"
aph/z,1 -I)
+ I '
liquid phases obtained by flashing the recombined reser-
voir oil at 1,885.5 psia and 205.9 OF. The capillary
PO,I 0,1
pressure is assumed negligible. The slim tube was
, , .... , , .... , ..... , ........... (53 b) represented by a 25 grid-block system.
Figs. 2a and 2b show the experimental and computed
where, from Eq. 52,
oil recovery (calculated from the standard volume of

ap h/z,i
2: III),i original oil in place) at two different pressures, 3,000
and 3,840 psia and a temperature of 205.9 OF. Results
_,~"c:..j_ _ _ _ _ with} = (J,g, 11'. . . (54)
obtained with fixed relative pemleability curves that are
ap"+
n,l
I
independent of interfacial tension (i.e., kro = fro'
k rg = f rg ) are shown abo.

DECEMBER 1981 693

 100

100 80
(q)
000
0
0
80 0
I- 60
>-
er Z
w W
> u
8w 60 0::
W
o Gas saturation
er Cl. 40 o C02 mole percent
f'> C 1 mole percent
l-
~ 40 o EXPERIMENTAL
U I:> CALC. - INTERFACIAL TENSION
er
W 0 CALC. - FIXED REL. PERM. CURVES 20
c-

O
25 50 75 100 125 150
PORE VOLUME INJECTED (%) 2.5

E
0
100 0
"-
VI
(b) Q) 2.0
c:
>.
'0
80
>- Z
er Q 1.5
w (f)
>
0 60 Z
(}
LiJ W
er I-
1.0
I- ...J
Z o EXPERIMENTAL
ex
w 40
u
er
II CALC. - INTERFACIAL TENSI.oN
U
w o CALC. - FIXED REL. PERM. CURVES ex
U-
c- o:: 0.5
20 w
I-
Z
0 0
0 )!5 50 75 100 125 150 0 5 10 15 25
PORE VOLUME INJECTED (%) GRID BLOCK NO.

Fig. 2-0il recovery curve at (a) 20 684 kPa and (b) 23 993 Fig. 3-Saturation, composition, and interfacial tension pro-
kPa. files at 20 684 kPa and 61.9% PV injected.

TABLE 3-0ATA FOR AREAL RUN TABLE 4-0ATA FOR CROSS-SECTIONAL RUN

Dimensions, ft (m) Dimensions, ft (m)

Length 2,639 (804.3) Length 6,597 (2010.860)
Width 2,639 (804.3) Width 330 (100.543)
Thickness 15 (4.6) Thickness 60 (18.288)
Permeability, darcy 0.030 Rock compressibility, kPa- 1 o
Porosity, fraction 0.020 Initial average reservoir pressure,
Rock compressibility, kPa- 1 o psia (kPa) 2,466 (17 000)
Initial reservoir pressure, Temperature, OF (0C) 206 (96.6)
psia (kPa) 2,466(17 000) Initial water saturation, % o
Temperature, OF (0C) 206 (96.6) Relative permeability data
Initial water saturation, fraction 0.4 k ro/w = k rwro 1.0
Relative permeability data (Eq. 37) Sorg 0.2
k rofW = k rwro 1.0 Sgr = Sorw = Swr 0.0
Sorg 0.3 e og 2.0
Sgr 0.0 eg 1.5
Sorw 0.1 e ow = e w 1.0
Swr 0.2
e og 2.0 Permeability
eg = e ow 1.5 (darcy) Porosity
ew 2.5 Layer
~ kv (fraction)
1 (bottom) 0.020 0.002 0.17
2 0.040 0.004 0.25
3 0.030 0.003 0.20
4 0.010 0.001 0.15

694 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 100

80

~ 60
Z
ILl
U
a:: o Gas saturation
ILl
a. 40 o CO 2 mole percent
" C1 mole percent

20

2.5
E
0
"-
f/l
CD 2.0
c:
>.
"0

Z
52 1.5
(/)
Z
ILl
~
..J
1.0
<t
(3
~ 0.5
a::
ILl
~
~
0.0
0 10 15 20 25 ~INJECTOR ~PRODUCER
GRID BLOCK NO.

Fig. 4-Saturation, composition, and interfacial tension pro- Fig. 5-Gas saturation contours at 3,582 days.
files at 23 993 kPa and 66.7% PV injected.

Higher recovery is obtained when interfacial tension is for this example. These fixed curves should be thought
considered. At 1.29 PY injected, the additional of as average curves that account for the effects of
recoveries are 3.46 and 1.59% at 3,000 and 3,480 psia, porous medium heterogeneities and changes in inter-
respectively. Fig. 3 shows the computed profiles of CO 2 facial tension and mobility ratios with respect to pressure
and C 1 concentration, gas saturation, and interfacial ten- and composition.
sion at 3,000 psia and 61.9% PY injected. One also can The program took five iterations per time step and the
observe a leading methane bank (grid block 21), fol- average computation time on the Honeywell DPS 2 with
lowed by a miscible region (grid blocks 18 through 20). Multics operating system is 0.471 seconds per time step
It is interesting to note that the interfacial tension per grid block. Direct elimination was used to solve the
decreases sharply as the miscible region is approached pressure equation. About 50 time steps were required to
and becomes almost zero just behind the miscible region inject 1.36 PY of CO 2 ,
(grid blocks 16 and 17). Fig. 4 shows the computed pro-
files for the same variables at 3,480 psia and 66.7% PY CO 2 Areal Displacement
injected; note a shorter transition zone (the gas saturation Half of a 5-spot pattern of a CO 2 lDJection process was
decreases more abruptly) and the absence of a leading simulated using an 8 x 8 areal grid system that covers an
C 1 bank. area of 160 acres. Pertinent data for the simulation are
The influence of interfacial tension on the recovery given in Table 3. The bottomhole pressures for injectors
depends on the size of the region of very low interfacial and producers are 3,000 and 2320.6 psia, respectively.
tension 19 (less than 10- 1 dyne/cm). That region is larger The interfacial tension and capillary effects are assumed
in the displacement at 3,000 psia, which is a multiple- negligible. Fig. 5 shows the gas saturation contour at
contact miscible process, than the one at 3,480 psia, 3,582 days. Simulation results show an enrichment zone
which is close to a first contact miscible process. This where the densities of oil and gas gradually approach
explains why the computed additional recovery due to each other. However, no miscible bank was fornled and
interfacial tension effects is higher for the former the process was immiscible.
displacement (3.46% compared with 1.59% at 1.29 PY The program took 6.5 iterations per time step and the
injected). However, these effects are quite small and average execution time on the Honeywell DPS 2 was
fixed relative permeability curves could have been used 0.757 seconds per time step per grid block when direct
DECEMBER 1981 695
 TABLE 5-PROGRAM PERFORMANCE FOR SOME TYPICAL
AREAL RUNS WITH THREE COMPONENTS

Grid System
10x 10 13x13 17x17 25x 25
Average execution time per
grid block per time step,
CPU seconds*
Direct elimination 0.125 (0.022) 0.112 (0.020) 0.121 (0.022) 0.153 (0.027)
ICCG (0) 0.134 (0.024) 0.122 (0.022) 0.115 (0.021) 0.119 (0.021)
Average number of
iterations per time step 6.2 5.8 6.2 6.2
Average number of
iterations in ICCG (0) routine 12.2 15.1 18.4 25.4
• ExecutIon times are for the Honeywell DPS 2 wIth Multics operatrng system. Figures In parentheses represent the eqUivalent execu-
tion time for the CDC 6600 based on our benchmark runs. .

tomhole pressure of 2,290.6 psia was maintained in the

producing multi block completion well. The capillary
pressure and interfacial tension are assumed negligible.
Fig. 6 shows the gas saturation and CO 2 and C I
005
global mole fractions contours at 3,778 days (33.65%
PY injected). One can observe a leading methane-rich
gas bank in the third and fourth layers. The simulation
results show the fonnation of a near miscible region
GAS SATURATION
following the methane bank. Gas breaks through after
6,588 days (66.8% PY injected), and the recovery at
8,692 days (1.04 PY injected) is 71.8%.
Direct elimination was used to solve the pressure equa-
tion. The program took six iterations per time step, and
the average execution time on the Honeywell DPS 2 was
0.658 seconds per time step per grid block. Ninety-nine
time steps were required to inject 1.04 PY of CO 2 ,

C1 GLOBAL COMPOSITION Computational Experience

The computation times reported earlier are for the
Honeywell DPS 2 with Multics operating system. On the
basis of some benchmark runs, the execution time for the
Honeywell DPS 2 is five and six-tenths times that of
CDC 6600.
To compare the computing time requirements of our
model with the ones of Fussell and Fussell 7 and Coats, 9
a series of areal runs was made using the ternary system
CO 2 GLOBAL COMPOSITION
C I -nC 4 -nC 10' The execution times for both direct
elimination and ICCG(O) along with other pertinent in-
fonnation are reported in Table 5. The numbers show
Fig. 6-Gas saturation and composition contours at 3,778 that the execution time is of the same order as the model
days. of Fussell and Fussell. The use of iterative conjugate
gradient method takes more time than direct elimination
elimination was used to solve the pressure equation. for an areal system with a bandwidth less than 35.
Fifty-four time steps were required to inject 89.9 % PY
of CO, . The execution time when the ICCG(O) was used Conclusions
was 1.03 times higher and about 12 iterations were need- This paper presents an implicit-pressure explicit-
ed in the conjugate gradient routine to obtain the solution composition and explicit-saturation method for the
of the pressure equation. This shows that for a bandwidth simulation of multidimensional compositional problems.
of 17, direct elimination is still faster than ICCG(O). Although the convergence near the solution is not
quadratic as in the model by Fussell and Fussell,7 it is
Cross-Sectional Displacement simple to program and its perfonnance is very
Cross-sectional runs were perfonned for CO 2 IllJection satisfactory .
into a stratified reservoir. A 20 x 4 grid system is used to Most of the important mechanisms of high-pressure
represent the reservoir. Pertinent data for this study are gas, enriched-gas, and CO 2 floods have been taken into
given in Table 4. CO 2 was injected at a pressure of account. The iterative scheme can be modified to include
3,000 psia. During the first 45 days, the oil was pro- the solubility of some components such as CO 2 , S02,
duced at a constant rate of 251.6 RB/D. After that, a bot- and H 2 S into the water phase using Henry's law.

696 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 One desirable feature of the formulation is the sym- /"1'<1 equivalent radius in well model
metric and diagonally dominant pressure matrix that r \I' well radius
allows the use of ICCG(O) methods for large problems. R universal gas constant
The use of the ICCG(O) method yields better execution
skin factor
time than direct elimination for bandwidths greater than
saturation of phase} (j = (J,g, w)
35. A two-point upstream fommlation for composition
reduces the numerical dispersion and an efficient tlash- residual saturation of phase j
calculation method permits the use of the model around residual oil saturation in gaslliquid system
the critical point. The drawback of the formulation is the at irreducible water saturation
time-step size limitation due to the explicit treatment of SOnt· residual oil saturation in water/oil system
compositions and transmissibilities. f time
The model has been tested against slim-tube ex- T temperature
perimental data, and satisfactory results were obtained. TI transmissibility of phase j (j = (J.g. w)
Areal and cross-sectional runs for hypothetical field V mole fraction of vapor phase in hydrocarbon
problems yield physically reasonable results with all the system
typical enrichment phenomena in the transition zone and
block volume
the gravity segregation effects. '
mole fraction of component 11/ in phase}
Nomenclature (j = (J,g)
. . .·m global mole fraction of component 111 in
A area perpendicular to flow
hydrocarbon system
Cr rock compressibility
z compressibility factor
c ll · water compressibility
8PII,SII' + PoSo + PgS~
e shape factor in well model gPj. gradient of phase)
D depth (measured positive down)
difference operator
e - exponential function
(POS iI + PgS~)/[PII,(SiI + S~)J
f= well fraction k,/Jl.J, mobility of phase)
fill} fugacity of component m in phase j
(j = (J,g)
~ 'A j • total mobility
g gravity acceleration j
h thickness
11'I viscosity of phase.i
H head in multiblock well completion model
/J - number of hydrocarbon components
1'1".) well index for phase j (j = (J,g, IV)
PI molar density of phase .i
J Jacobian matrix
PI mass density of phase j
k permeability
relative permeability to phase j (j = (J,g, w)
a = interfacial tension
porosity
relative permeability to oil or to gas at ir-
potential of phase.i (j = (J,g, 1\')
reducible water saturation
kroK relative permeability to oil in gas/liquid
Superscripts
system at irreducible water saturation
(I') iteration level
k roll · relative permeability to oil in oil/water
system 11 old time level
relative permeability to water at residual oil n+l new time level
krll'ro
saturation
Subscripts
horizontal permeability
vertical permeability bh bottomhole
mole fraction of liquid phase in hydrocar- c critical
bon system g gas
number of well elements in a multiblock i,k grid block indices
well completion j phase index
number of grid blocks 111 component index
nh
(J oil
p} pressure of phase j (j = (J,g, IV)
P COK oil/gas capillary pressure total
water/oil capillary pressure IV water
Pc\\'o
molar injection/production rate of X,Y,Z coordinate directions
hydrocarbon
molar injection/production rate of compo- Difference Notation
qm
nent m I::.Tl::.p = I::. x Trl::.px + I::.\,T\,l::.p,. + 1::.; T;l::.p;
molar injection/production rate of water
and
volumetric injection/production rate of
phase j (j = o,g, IV)
DECEMBER 1981 697
 where T',i + '12 is the x direction transmissibility between II. Nghiell1. L.X. and Aziz. K.: "A Robust Iterative Method tilr
grid block i and i + I, Flash Calculations Using the SO<!\T-Redlich-Kwong or the Peng-
Robinson Equation of State." paper SPE 8285 presented at the
SPE 54th Annual Meeting. Las Vegas. Sept. 23-26. 1979,
Acknowledgments 12. Meijerink. J.A. and van der Vorst. H.A.: "An Iterative Solution
This research is supported by the Alberta/Canada Energy Method jilr Linear Systems of which the Coefficient Matrix is a
Resources Research Fund administered by the Dept. of SYll1metric M-Matrix." Malh. C(lIIII'. (Jan. 1977) 31. 148-162.
13. Aziz. K. and Settari. A.: Pl'lro/l'UIII Res{'/'\'{iir Silllu/mi(!lI. Ap-
Energy and Natural Resources of the Province of Alber- plied Science Publishers. London (1979) 36.
ta. We thank P,K. W, Vinsome for his helpful comments 14. Reid. R.C .. Prautsnitz. J.M .. and Sherwood. T.K.: The Pml'{'/'-
and suggestions on this work and A, Behie for program- lies o(Gasl's "lid Liqllids. third edition. McGraw·Hillinc .. New
ming the conjugate gradient routine, The phase-behavior York City (1977) 426.614.
15. Todd. M.R .. O·Dell. P.M .. and Hirasaki. GJ.: "Methods ti)r In-
package used in the simulator is a modified version of
creased Accuracy in Numerical Reservoir Simulators." SoC, Pel.
the Hyprotech tluid properties program (HYPROP). Ellg.1. (Dec. 1972) 12.515-530.
16. Au. A,D.K. Behie. A .. Rubin. B .. and Vinsoll1e. P.K.W.:
"Techniques tilr Fully Implicit Reservoir Simulation." paper SPE
References 9302 presented at the SPE 55th Annual Meeting. Dallas. Sept.
21-24. 1980.
I. Zudkevitch. D. and Jofrc. J.: "Correlation and Prediction of 17. Chappelcar. J.E. and Williamson. A.S,: "Reprcsenting Wells in
Vapor· Liquid Equilibriull1 with the Redlil'h-K\\ong Equation of Numcrical Reservoir Simulation-Thcory and Implementation."
State." AICh!:' 1. (Jan. 1970) 16. 112-119. paper SPE 7697 presented at the SPE Firth Symposium on Rcser-
2. Soave. G.: "Equilibriulll Constanh froll1 a Modified Redlich- voir Silllulation. Denvcr. Feb. 1-2, 1979.
Kwong Equation of State." Chelll. !-;lIg. Sci. (1977) 27. Ig, Sigmund. P.M .. Aziz. K .. Lee. J.T .. Nghicm. L.X .. and Mehra.
1197-1203. R.: "Laboratory CO e Floods and Thcir Computer Simulation."
3. Pengo D.Y. and Robinson. D.B.: "A New Two-Constant Equa- papcr presented at the 10th World Pet. Cong .. Bucharcst. Scpt.
tion of State." Illd. Ellg. C/zOII, FUlld. (1976) 15. 59-64. 1979.
4, Bishnoi. P.R .. Heidell1ann. R.A .. and Shah. M.K.: "Calculation 19. Bardon. C. and Longeron. D.G.: "Influence of Very Low Inter-
of Thermodynall1ic Properties of Bitull1en Systell1s." paper lacial Tensions on Relative Permeability." SoC, Pel. t;lIg. 1. (Oct.
presented at the CIM Canada-Venezuela Oil Sands Symposiull1. 1980) 20.391-401.
Edll1onton. Alta .. May 30-June 4. 1977.
5. Yarborough. L.: "Application of a Generalized Equation of State
SI Metric Conversion Factors
to Petroleull1 Reservoir Fluids." paper presented at the I&EC ,
SYll1posiull1 on Equations of State in Engineering and Research. acre x 4.046 873 E +03 m-
176th ACS Natl. Meeting. Miall1i Beach. FL. Sept. 10-15. 1978.
atm X 1,013 250* E + 02 kPa
6. KaU. D.L. and Firoozabadi. A.: "Predil'ting Phase Behavior of
Condensate/Crude-Oil Systems Using Methane Interaction Coeffi- bbl X 1,589 873 E - 01 m"
cients." J, Per. Tech. (Nov. 1978) 30. 1649-1655. cu in. x 1.638 706 E+OI cm"
x 2,831 685 E - 02 3
7. Fu"ell. L.T. and Fu"cJl. D.O.: "An Iterative Technique for cu ft 111
Compositional Reservoir Models." Soc. Pel. f;lIg, 1. (Aug. 1979) dyne x 1.0* E - 02 mN
19. 211-220.
8. Fussell. D.O. and Yanosik. J.L.: "An Iterative Sequence tilr
g mol x 1.0* E - 03 kmol
OF (OF - 32)/1.8 °C
Phase Equilibrium Calculations Incorporating the Redlich-Kwong
Equation of State." Soc, Pel. Ellg. 1. (June 1978)18. 173·182. psia x 6,894 757 E+OO kPa
9. Coats. K.H.: .. An Equation of State Compositional Model." Soc. '("(lIl\l'f'Ulll t;Kl11r t\ l'\d<...'! SPEJ
Pel. !:'ng. J. (Oct. 1980) 363-376.
10. Kazell1i. H .. Vestal. C.R .. and Shank. G.D.: "An Efficient Original manuscript received in SOCiety of Petroleum Engineers office July 21, 1980.
Paper accepted for publication Apnl 21, 1981. Revised manuscript received Sept. 23.
Multicomponent Numerical Simulator." SoC, Pet. Eng. 1. (Oct. 1981. Paper (SPE 9306) first presented at the SPE 551h Annual Technical Con·
1978) 18. 355-368. ference and Exhlbilion held In Dallas Sept. 21-24, 1980

698 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

Discussion of Compositional Modeling SPE 10894

With an Equation of State

Jahan Mansoori, SPE, Phillips Petroleum Co.

The following addresses "Compositional Modeling not valid and the Newton method may lead to a diverg-
With an Equation of State," L.X. Nghiem, D.K. Fong, ing, or oscillatory, or slow-converging system of equa-
and K. Aziz (Dec. 1981 SPEJ, Pages 687-698). tions. A new formulation has been presented to over-
The compositional formulation as presented calls for come this difficulty.
the solution of the pressure equation given by A situation where an unstable behavior first was
observed was during the simulation of pressure decline
of an undersaturated reservoir in which a hydrocarbon
~ rOT w n(~p 0 n+ I -~Pcwo n -)'w n ~h) gas phase was generated as the reservoir pressure passed
through the bubble point. Since variation in water satura-
+To n(~p 0 n+1 -'V
'0
n~h)+T n(~p
g 0
n+1 +~pn
cog tion is negligible compared with that of hydrocarbon
phases, Eq. 2 can be differentiated with respect to
pressure to give

V
___b_(¢n+lan+l _¢na n ) =0 .............. (1)
~t

The Newton method adopted for the solution requires

an approximation to the term a(¢a)/ap 0 where ¢ is the where the term in parentheses represents the total
grid-block porosity and a is defined as hydrocarbon mass in a grid block. From a material
balance on oil and gas phases, the vapor phase fraction
can be derived as

The following approximation to aa/ap 0 is presented

by the authors, in which the terms aSwlap 0' aSolap 0' ......................... (5)
aSglapo' aYmolapo, and aYmglapo are neglected
because of their complicated computation:
U sing the phase saturation constraint

Eq. 5 yields
Although this equation leads to a diagonally dominant
Jacobian matrix using SRK or PR equation of state, ex-
perience has shown that under certain reservoir flow con-
ditions where the time variations of hydrocarbon phase
saturations are considerably more pronounced than the
respective changes in their physical properties, Eq. 3 is
The total hydrocarbon mass in each grid block,
0197·7520/82/0041·0894$00.25 therefore, can be derived from Eq. 5 as

202 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 0
~g V~ 0 (I - S w )
~oSo +~gSg =
v ~g+V(~o-~g)
z
"
~o~g(I-Sw) 0
............ (6) ......
E-<
a:
=>
00
WN
Since S IV is treated as a constant, Eq. 6 is differentiated
W
to give cr
=>
(J)
(J)
W
cr
a.... ~ present work
'-' 0

~-
a:
.................................. (7) =>
Q
......
(J)
or W
cr
0
a d
-(~oSo +~gSg) 0 2 ~ 6 8 W 12
apo NUMBER Of ITERATIONS
~(,'~g 2 + V(~o 2~g' -~g 2~0')- V'~o~g(~o -~g)
Fig. 1-Numerical behavior of Newton method for the two ap-
[~ g + V( ~ () - ~ g)] 2
proximations of J(q,a)/JP o'

'(l-Sw), ............................ (8)

where primes indicate the derivative with respect to

pressure. methods of approximations. The method presented here
This equation also provides the correct values for the reached the convergence criteria within about six itera-
two limiting cases of So=O (only gas phase present) or tions, while the other technique indicated an oscillatory
S g =0 (only oil phase present) using V= V' =0 for the behavior.
former and V= 1, V' =0 for the latter. To study the effect
of this approximation on the diagonal dominance of the
Jacobian matrix, it can be shown easily that, under the Nomenclature
conditions of h = depth (measured positive down)
Po = oil phase pressure
~o>~g>O ............................... (9) P cog = oil-gas capillary pressure
P cwo = water-oil capillary pressure
~g, > ~ 0' > 0 ............................ (10)
q He = molar injection/production rate of
V'<O, ................................. (II)
hydrocarbon
q w = molar injection/production rate of water
Eq. 8 yields a positive value that contributes to the Sj = saturation of phase j (j=o,g,w)
stability of Newton's method. Some or all of these con- t = time
ditions may not be satisfied for retrograde condensation Tj = transmissibility of phase j (j=o,g,w)
process. The implications of this situation have yet to be V = mole fraction of vapor phase in hydrocarbon
explored. For computational purposes, V' in Eq; 8 may system
be approximated by Vb = block volume
Y mg = mole fraction of component m in phase j
V' = V !l,P
---V Iz,P-Ll.P
--- (j=o,g)
/).p Z m = overall mole fraction of component m
a = (J~wSw+~gSg+~oSo
z
where and P are the latest calculated values of overall 11 = contribution of dispersion effect
(J = (~oSo+~gSg)/[L(So+Sg)]
composition and pressure, respectively, for each grid
block and /)'P is a reasonably selected pressure incre- ~j = molar density of phase j
ment. The phase behavior subroutine is called each time cf> = porosity
to calculate V' during the simulation. Fig. 1 presents the /). = difference operator
residual-of-pressure equation, which is a measure of
convergence vs. the number of iterations for the two SPEJ

APRIL 1982 203

Author's Reply to Discussion of SPE 10903

Compositional Modeling
With an Equation of State
Long X. Nghiem, SPE, Computer Modelling Group

We have observed the same convergence problems improve the convergence of the algorithm described in
described by Mansoori with certain fluids under certain our paper by including aVI ap 0 in the derivative aod ap o.
reservoir flow conditions. They usually occur during This method requires one additional flash calculation per
pressure decline across the bubble point of an oil reser- grid block per iteration to compute aVI ap o. Since no
voir or the dewpoint of a gas condensate reservoir. data were provided, we were unable to compare Man-
Under certain circumstances, during the iteration of soori's approach with ours.
the pressure equation, the gas released from an oil while Coats I has reported anomalous pressure response even
pressure declines below the bubble point affects the ap- for fully implicit simulators.
proximate Jacobian described in our paper in such a way
that the pressure increases after the subsequent iteration, Nomenclature
and all or most of the gas goes back into solution. This nh number of grid blocks
creates a vicious circle and yields oscillations. The same p o,k oil pressure of kth grid block
reasoning applies to certain gas condensate reservoirs as V vapor mole fraction
the pressure declines below the dewpoint. ex defined in Eq. 13 of our paper
A solution to this problem is to use a damping factor in
the update of the pressure solution-i.e.,
op k change in pressure
w damping factor
(f) iteration level

where op If) is the change in pressure obtained by solving Reference

our Eq. 15, and w(f) is a damping factor (O<w(f) ::51) I. Coats, K.H.: "Reservoir Simulation: A General Model Formula-
that allows a smooth convergence to the solution. Note tion and Associated Physical/Numerical Sources of Instability,"
Boundary and Interior Layers-Computational and Asymptotic
that w can vary with the iteration level (f). An ap- Methods, J.J.H. Miller (ed.), Boole Press, Dublin (1980).
propriate selection of w(f) will prevent oscillations.
Mansoori's Discussion provides another method to 0197·7520/82/0041·0903$00.25

204 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 Errata

"Author's Reply to Discussion of Compositional

Modeling With an Equation of State, " April SPEJ, Page
204, listed only one author. Khalid Aziz, Computer
Modelling Group, coauthored the Reply. The editors
regret this omission and thank the other author, Long
Nghiem, for pointing it out.
Also in the April issue, an error is noted by author
lohan Mansoori in his "Discussion of Compositional
Modeling With an Equation of State." The last equation
(unnumbered) in Col. I, Page 203, should read

The editors regret these errors and thank the authors

for pointing them out.
556 AUGUST 1982 SOCIETY OF PETROLEUM ENGINEERS JOURNAL