Вы находитесь на странице: 1из 10

✎

SPE 6050

An Iterative Sequence for Phase-Equilibria Calculations

Incorporating the Redlich-Kwong Equation of State
D. D. FUS!W-L
J. L. YANOSIK
MEMBERS 3PE-AIME
.1 AMOCO PRODUCTION
TuLSA, OKLA,
cO,

ABSTRACT as phase densities and viscosities) are obtained

Phase equilibria equations that incorporate the The Redlich-Kwong 1 equation of state is used
Qedlich.Kwvng equation of staie are nonlinear and, widely in the petroleum industry for phase equilibria
tl.rere~ore, must be solved by an iterutiue method. calculations. The phase equilibria equations th,~t
The rnetbod of successive substitutions commonly incorporate this equation of state are nonlinear.
is used. This method, however, almost always As a result, they must be solved by an iterative
diverges near the critical region /or bubble point, method, The method of successive substitutions
dew point, and two-phase calculations. iterative commonly is ~sed. This method, however, almost
methods that converge for these calculations are always diverges for bubble point, dew point, and
presented. These iterative methods are tailed two-phase calculations near the critical region.
“minimum variable Newton -Rapbson” (MVNR) This region is extremely important when studying
methods because they try to minimize the number of multiple-contact, miscible oil recovery methods
variubles for which sirnultaneotis iteration is required involving CO z or rich-gas injection because the
and use the Newton- Rapbson method /or the path of the over-all fluid mixture passes through
correc~ion step. Procedures are given /or obtaining this region. The method of successive substitutions
starting values /or the first iteration and several aiso will diverge for some fluid mixtures near their
example pro biems are discussed. saturation (bubble point or dew point) pressure at
conditions removed from the critical region.
INTRODUCTION This paper presents a reliable iterative sequence
Reservoir performance predictions for gas that can be used ‘to predict phase equilibria of
condensate and volatile oil reservoirs require a multiple-contact, miscible oil recovery methods.
knowledge of the vapor-liquid phase equilibria of The method includes sequences for calculation of
the reservoir fluids. A similar knowledge also is the saturation pressure and phase equilibria in the
two-phase region.
requ{red when studying multiple-contact, miscible
oil recovery methods that involve injection of These MVNR methods rely on minimization of
hydrocarbons and/or carbon dioxide. Such knowledge the number of unknowns for which simultaneous
is obtained experimentally or calculated from iteration is required and use the Newton-Raphson
physical properties of the components of the method for the correction step. Minimization is
subject co the constraint that all additional unknowns
reservoir fluid system. Calculation is desirable
can be calculated by using simple linear equations
because experimental determination is both laborious
and expensive. or, at most, an iteration method applied to one
A common basis for calculation of vapor-liquid equation in one unknown.
phase equilibria is the single-stage separation unit. .MVNRis compared with the method of successive
This unit represents a PVT cell in which a fluid substitutions for a two-phase fluid mixture at
mixture of known over-all composition is equilibrated various pressures for a fixed temperature. MVNR
also is compared with the method of successive
at the temperature and pressure of interest. Liquid
s~bscitutions for saturation-envelope calculations
and vapor compositions and moles of liquid and
near the critical region.
vapor per mole of fluid mixture are determined.
Reliable estimates of other fluid properties (such
DESCRIPTION OF PHYSICAL SYSTEM
Original manuscript received in Society of Petrolsum &rg[neers The singIe-stage separation unit is the basis for
office Aug. 2, 1976. Paper arxeptzd for puhllcation July 12,
1977. Revised manuscript received March 22, 1978. Paper (SPE the phase equilibria calculations discussed in this
6050) first presented at the SPE-AIME 51x Annual Fall study. This unit represents a PVT cell in which a
Technicaf Conference and %hlbftiorr, heId in New Orlearm,
Oct. 3-6, 1976. fluid mixture of known over-all composition is
0037 -9999 /78/000S-6050\$00. 2S equilibrated at the temperature and pressure of
@ 1978 Society of Petroleum Engineers of AIME interest. Liquid and vapor compositions and moles
JUNE, 197S 173
,

of liquid and moles of vapor per mole of fluid which shows that Zi moles of component i, when
mixture are determined, equilibrated, will be distributed as Lx i moles of
Fluid mixtures considered areeither single-phase component i in the Iiquid phase and Vyi moles of
and saturated or two-phase (liquid md vapor) at component i in the vapor phase. From Eqs, 1 and 2,
the system temperature and pressure. Fig, I presents ~i, Yi ~ and Zi are mole fractions and, by definition,
,
a simplified phase diagram for a given fluid
mixture as a function of temperature and pressure.
The bubble-point and dew-point loci are shown and
their intersection is the critical region. The curve
() ;Z ~ , must be unity,
i=l
These equations also require a restrictive equation
on the liquid phase compositions of
described by these loci is called the saturation
envelope, The region encompassed by the srtturacion
envelope is the two-phase region, Phase equilibria
of fluid mixtures that are within the envelope ~cil
!Xi=l’” ”””””.””.’(sa)
are described by the equations for a single. stage
separation unit, The saturation pressure equations or a restrictive equation on the vapor phase
that are simplifications of the single-stage, compositions,
separation equations describe che phase equilibria
of fluid mixtures on the saturation envelope.
The single-stage, separation and saturation- ~yi=l . . . . . . . . . ..o(3b)
pressure equations are presented and then folIowed ia~
by pzeseneat~on of MVNR for calculation of phase Both Eqs. 3a and 3b cannot be used because, given
equilibria described by these equations. one, the second may be derived hy combining the
given equarion with Eqs. 1 and 2,
SINGLE-STAGE SEPARATION EQUATIONS The component fugacity in each phase is
introduced co develop a criterion for thermodynamic
The single-stage separation unit can be described
equilibria. Physically, the fugacity of a component
with four types of equations, an ouer.all material
in one, phase, with respect to the fugacity of that
balance, component material balances, a restrictive
component in a second phase, is a measure of the
equation on the phase compositions, and tbermody.
potential for transfer of that cc,mponent between
namic phase equilibria equations. The overall
phases. The phase with the lower component
material balance can be written as
fugacity accepts the component from the second
L+v= l, . . . . . . . . . . . ..(1) phase. Equal fugacic;es of a component in the two
phases results in a zero net transfer. A zero transfer
wht :e L is the moles of liquid per mole of fluid for all components implies a fluid system that is in
mixture and V is the moles of vapor per mole of thermodynamic equilib~ium. Therefore, the thermody-
fluid mixture. The component material balance for namic phase equilibria equation can be expressed
each component within the fluid mixture can be by
writren as
=f ~v;~~i:n..........(4
LXi + VYi = Zi ; l~i:n, . . . .“ . .(2) f iL

REDLICH-KWONG EQUATION OF STATE

The component fugacity is a function of the partial
molal volume of the component. Because the partial
molal volume is difficult to establish experimentally,
it is necessary to relate the partial molal volume
and the component figacity to easily determined
quantities — for example, temperattire, pressure,
VAPOR phase composition, and the components in the fluid
system. This is accomplished with a combination of
the thermodynamic definition of fugacity and an
equation of state — in this case, the Redlich-Kwcmg
equation of state as presented by Zudkevitch and
Joffe.2 Expressed in terms of the liquid phase
compressibility factor, this equation of state is

z~-z~+zL[~-BL(BL+l)l

o m 2(KI 300 400 500 ‘%BL= *’”””””*”’”””””(5)

TEMPERATURE,
‘f and the equation of state, expressed in terms of
FIG. 1 — EXAMPLE PHASE DIAGRAM. the vapor-phase compressibility factor, is

174 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

3 n f
Zv-z:+zv [~” Bv(Bv+Vl iT. ,
—.. . . . . . . . . (11)
P~ = 1
i-l \$iv
-%Bv=O” ””r”
”””’””””(6)
where the component fugacity ~oefficient in the
The derivations of Eqs. 5 and 6 incorporate the vapor phsse is defined as
actual gas law, pv = ZRT, which is valid f,>r the
liquid and vapor phases. f J tr
Following the work of Zudkevitch and Joffe, the
component fugacity in the liquid phase now ca c be
espressed as The thermodynamic phase-equilibria equcztion
~xi exp tbiL {zL-l) ] ; ~<i<n , (Eq. 4) states that, for all components, the fugacity
of each compnent in the liquid phase is equal to
‘iL = B1, Ui -- the fugacity Qf that compment in the vapor phase
(ZL - BL) 1 + ~~ when equilibrium is attained. Using these equations
[] wi:h Eqs. 11 and 12 shows that the vapor-phase
{
. . . . . . . ...*. .* .**. ● ✎ .(7) compositions must equal ucity at equilibrium for
the system pressure, p, to equal the saturation
Similarly, the component fugacity in the vapor phase
pressure, p ~. During the iteration sequence for
is given by the relation
calculation of the saturation pressure, however, the
vapor-phase compositions are not forced to equal
PYi eXP [biv (Zv-l) ]
— ; Ici<n , unity. In general, the sum of these compositions is
‘iv ‘c
(Zv - Bv; l+r

BV‘i - – less than unity if the estimated pressure is low,
and greater than unity if the pressure is high.
[1 v Consequently, the restrictive equation on the vapor-
. . . . . . . . ...* . . . . . . ● ✎ .(8) phase compositions is not incluti.?d in the set of
saturation pressure equations.
The equation parameters are defined in Appendix A
and the Nomenclature. These are Ilmctions of SATURATED VAPOR
pressure, temperature, phase composition, and the A saturated vapor exists for a given temperature
component parameters, Q gi and ~bi. The componei~t at the pressure at which an infinitesimal amount of
parameters are assumed independent of ccxnposition liquid first appears. This pressure is referred to as
and pressure. MVNR methods presented in this the dew-puint pressure. Using logic similar to that
paper can be adapted readily to any equation of for developing the saturated liquid equations,
state. equtitions can be developed for a saturated vapor.

SATURATION PRESSURE EQUATIONS DESCRIPTION OF MVNR METHODS

A saturated fluid state may be described To introduce MVNR methods, consider the
mathematically by Eqs. 4 through 8, a saturation single-stage separation equations, “Eqs. 1 through
pressure equation, and simplifications of the 4, 7, and 8. These’ equations constitute a set of
over-all and component material balances. (4n + 2) equations that are to be solved
simultaneously to determine 1., V, and n liquid-phase
SATURATED” LIQUID
compositions, xi, ~ vapor-phase compositions) yi>
A saturated liquid exists; for a given temperature, z liquid-phase fugacities, /iL, and Y vapor-phase
at the pressure at which an infinitesimal amount of
fugacities, /iV . The solution must be iterative
vapor first appears. This presstire is referred to as
because the equations are nonlinear.
the bubble-point pressure. Based on this definition,
There are numerous methods for organizing the
the over-all material balance for a saturated liquid iterative sequence. At one extreme, it would be
is simply
possible to correct all (4n + 2) unknowns at each
L=l~o-.....”o..**** .(9) iteration. A more efficient method? however, is
called minimum variable Newton-Raphson (MVNR),
and the component material balances are which reduces the size of the correction step by
eliminating as many unknowns as possible from the
‘i ‘zi; l~i:n”.”.*””””’ ’10) correction steps This is done by dividing the
The restrictive on the liquid-phase
equiztion unknown variables into two groups. The first group
compositions is redundant because, by definition, contains the iteration (independent) variables that
the fluid mixture compositions, zi, add up to unity. are the unknowns to be corrected. The second group
The saturation (bubble-point) pressure is related to contains the dependent variables. One equation
the component /ugacities in the liquid phase and must be used to define each dependent variable,
the component fugacity coe//icier2ts in the vapor and the collection of these equations is referred to
phase by as the de/ining equations. The remaining equations
(:qual in number to the number of iteration variables)

JUNE, 1~ 175
.

are called the error eqrfatioirs, ‘: he iteration iteration.

sequence for MVNR then will be a four-step process. Dew. Point Vapor, The saturation pressure
The four steps are (1) select the iteration variables equations for a dew-point fluid also constitute a set
and assume values for these variables- (2) use the of (3n + 1) equations that are to be solved simul-
de/ining” equations to calcuIate the dependent taneously to determine the dew-point pressure, p~;
variables, (3) use the error equations to calculate ~ lifuid-phase compositions, xi; n liquid-phase
the error, and (4) use the Newton-Raphson method fugacities, /iL; and ~ vapor-phase figacities, /iV.
to correct the iteration variables. This process is Similar to MVNR for bubble-point liquids, the
repeated until the error calculated in Step 3 is less number of iteration variables is n + 1; the pressure
than a small error tolerance, f, which corresponds is p, and ~ liquid-phase compositions is xi. MVNR
to a converged solution. using this set of iteration variables is called the
The methods have been defined as a four-step the p-,Y iteration.
process. The development of MVNR is presented Initial Estimates of Iteration Variubles. Initial
step-by-step for two-phase calculations and for estimates of p and yi for the p-Y iteration and of p
saturation. pressure determination. and xi for the p-X iteration are required. Our
experience has shown that component K values
SELECTION OF ITERATION VARIABLES
calculated by the empirical relationq
~
Tzuo.Phase Fluids exp{5.37(1-t-wi) (l-l/Tri) ]
Ki m ; I<i<nf
Two sets of iteration variables that satisfy the --
MVNR criteria are (1) L and (n - 1) liquid-phase ‘ri
. . . . . . . . . . . . . . . . . . ,0 (13)
compositions, xi, and (2) V and (rz - 1) vapor-phase
cornoositions, yi. MVNR that uses the first set will can be used co flash the over-all fluid mixture at
be referred to as the L-X iteration, and MVNR that the system temperature and an estimated saturation
uses the second set will be ieferred to as rhe V-Y pressure. The flash calculations result in liquid
iteration. The L-X iteration is used for a fluid and vapor compositions. Depending on the iteration
system that is predominai~tly vapr and the V-Y used (p.X or p. Y), the appropriate phase composi-
iteration is wed for a fluid system that is tions are used as initial estimates of the iteration
predominantly liquid. The section entitled “Example variables, xi or yi. The other phase composition is
Problems” demonstrates the necessity and equal to the composition of the fluid mixture (Eq.
mechanics for using two iterations — that is, L-X lo).
and V-Y iterations.
Initial estimates of the iteration variables are DEFINING EQUATIONS
required. Generally, the phase distribution as a The second step of MVNR is to. solve the defining
function of pressure between the saturation pressure, equations for the dependent variables.
p~, and a given pressure, p, must be calculated.
Therefore, the procedure is to calculate the V.Y Ilcrf/tiun Steps
saturation pressure followed by two-phase
calculations for the given pressurb iange. Assuming 1. Values for Ihe iteration variables V, yz,
the saturation pressure is known, the following Y3,..., Yn, are available.
method is used to obtain initial estimates for the 2. The total hydrocarbon mole bc{lcinc-e (Eq. I) is
L-X and V-Y iterations. For the first two-phase used to calculate L,
calculation using an L-X iteration (dew-point fluid), 3. The restrictive eqz~ution on the vapor-phase
the initial values are f- = O and the n-1 liquid-phase compositions (Eq. 3b) is used to calculate the
compositions, xi, determined by the dew-point composition of the first component in the vapor
calculation. For the V-Y iteration (bubble-point phase, yl.
fluid), the initial values of the iteration variables 4. The component rnoie balcince (Eq. 2) is used n
areV=Oandtherz- 1 vapor-phase compositions, times to calculate xl, X2, . . . . %n.
yj, determined by the bubble-point calculation. For 5. The hydrocarbon-phase compressibility factors,
each successive pressure, the initial estimates of ZL and ZV, are calculated by solving the Redlich-
the iteration variables are the values of these Kwong equation of state (Eqs. 5 and 6).
variables calculated at the previous pressure. 6. Eq. 7 is used n times to calculate the
component fugacities, /if’, in the liquid phase. Eq.
Saturation Pressure 8 is used n times to calculate the component
fugacities, /iv, in the vapor phase.
Bubblepoint Liquid. Eqs. 4, 7, 8, and 11
constitute a set of (% + 1) equations that are to be L-X It erat ion
~olved simultaneously to determine the bubble-point
pressure, p ~, n vapor-phase compositions, Yj, z The steps outlined for the I’-Y iteration are
liquid-~hase f@acities, /iL, and ~. vapor-phase applicable to the L-X iteration if I!., V, ~i} and yi
fugacities, /iv. The number of iteration variables are interchanged. In addition, Eq. 3b is replaced
required by MVNR is n + 1; the pressure is p; and g with Eq, 3a.
vapor-phase compositions is yi. MVNR using this
set of iteration variables is called the p-Y
176 - SOCIETY OF PETROLEUM ENGINEERS JOURNAL

p.Y and p-X Iterations V-Y Iteration

Steps 5 and 6 of the V-Y iteration outline the Consider the V-Y iteration for a two-phase fluid
defining equations for these iterations. Fewer that is predominantly vapor. If ~ is defined as a
defining equations are required by these iterations
thpn required by the L-X and V-Y iterations because vector, (v, yz, y3t. ..s Yn ) T and ~ as the error
of the lesser total number of unknowns. vector, (+1, 42\$ s.. ~ q5n)T, then a linear correction
algorithm can be written as
ERROR DETERMINATION
The error equations are the equations remaining Jq{*q+l -~qJ = -~,...... .(17)
after selection of the defining equations.
where the superscript q indicates the quantity is
LoX and V-Y Iterations evaluated using known values from iteration. q. This
set of simultaneous linear equations is solved for
A review of Paragraphs 1 and 2 in Section 3
(V-Y and L.X iteration~~, and Eqs. 1 through 8 the elements of ~q+l, which are the values of the
shows that Eq. 4 has not been included as a defining iteration variables used in ,the next iteration.
equation. There are n of these equations and there Qua&atic convergence of Eq, 17 near the solution
are n iteration variables. is obtained when 9 is the Jacobian matrix, defined
~~ error equations, which represent the residual as the matrix in which each element is the partial
ei the thermodynamic phase equilibria equation (Eq. derivative of one of the errors with respect to one
4) can be written as iteration variable when all other iteration variables
are constant,
-f ~V ; lc_i.~n . . . . .(14) The elements of \$ for the V-Y integration are
q = ‘iL
defined by the following equations.
These equations are evaluated. If the Euclidean
norm of the @i’S, Ni
—;l<i<n, . . . . . . ..(l8)
Ji3. ‘w --
n.
wi
— * l<i<n, 2~:n . . . . “(I9)
Jij, “ay ~ ’--
is less than the error tolerance, the iterative
sequence ends. If not, the iteration variables are The analytical expressions for the derivatives in
corrected by the methods outlined in Section D. Eqs. 18 and 19 can be obtained by differentiating
the defining equations (Eqs. 1, 2, and 3) with
p-Y and p-X Iterations respect to the iteration variables (see Appendix B).
The resulting equations then are used during the
For the p-Y iteration, there are n + 1 iteration
implicit differentiation of the error equations. The
variables. Eq. 14 represents n error equations, and
elements of g are
the remaining error equation is the residual of Eq.
11, which :an be written as
jil.~.-~
w~ n afiL
‘x);
—(Yj-j
1

j=2 axj
(15)
‘n+l = ‘s -P’”””””
i=l ‘J’iv l<~<no.....a.(zfj)
——

iteration is

Analytical expressions for d/iL/d%i and d~v /dy.

THE CORRECTION STEP
can be developed by differentiating Eqs. 7 and b
The purpose of the correction step is to obtain with respect to xi md yj ~ respectively.
improved estimates of the iteration variables that, The elements of the Jacobian matrix, g, for
in turn, will reduce the residuals calculated by the L-X, p-Y, and p-X iterations are given in Appendix
appropriate error equations. The correction step c.
reduces to a matrix problem with an order directly
proportional to the number of components, n, for the EXAMPLE PROBLEMS
L-X and V-Y iterations and proportional to n + 1
for the p-X and p-Y iterations. C02 -HYDROCARBON SYSTEM

The prediction of reservoir performance of

JUNE. 1978 In
.
. ,

multiple-contact, miscible oil recovery methods TABLE 1- ‘WCLI DEAN NORMS FOR Ml XTURE A
with either CC)2 or rich-gas injection requires an USING MVNR

understandir. g of the transition zone that exists Iteration Numl ?! Euclidean Norm

between the in-place oil and the injected gas, It o 5.94

1 1.45
commonly is accepted that near-critical fluid
2 0.2S3
mixtures exist within a small region of this 3 0,110
transition zor,e. The experimental program required 4 0,0851
to obtain phase-equilibria data for this region is 5 0.0360
6 0.0022
laborious and ~!xpensive. The alternate approach is
7 8 x 10-8
to minimize tile experimental program and tidjust
the parameters of .m equation state so that it can
TABLE 2- RESERVOIR OIL COMPOSITION
be used for the phase-equilibria calculations.
Component Mole Fraction
The first problem considers the calculation of
the saturation envelope for the C02-nC4 -CIO fluid N2 0,00808
c, 0.E899
system at 1,600 psia and 1600F using the Redlich- C02 0,01601
Kwong equation of state. Comparison of the cfJ 0.12114
envelopes calculated by MVNR methods with the c~ o.075a8
method of successive substitutions is made. Icd 0.01076
Fig. 2 p,.sents the saturation envelope as nC4 0.03065
calculated by the method of successive substitu- ic~ 0.00840
IC5 0,01074
tions. The saturated (bubble-point) liquid and the c~ 0.01272
saturated (dew-point) vapor curves should continue 180 BP* 0.01718
to close on each other as the over-all mixture 230 BP 0.01764
becomes richer in nC4 and the intersection should 280 BP 0.01418
be the critical mixture at these conditions, Instead, 330 BP 0.01115
380 BP 0.07065
the curves move away from each other near the
critical region. This implies that in this region, *BP represents bollln~”po!nt fractions of the C7+.

either the equation of state errs or the method of

successive slabstitutions fails to converge. The The results clearly demonstrate that phase
investigation of this behavior was the principal equilibria as calculated by the equation of state
reason for developing MVNR methods. are physically realistic, which lends confidence to
Fig. 3 presents the saturation envelope for this adjusting the parameters so that the equation of
simple fluid system at 1,600 psi and 1600F as state matches experimental data.
calculated by MVNR methods. For comparison, the
calculated saturation curves presqnted in Fig. 2 MULTICOMPONENT RESERVOIR OIL
also are included in Fig. 3.’ The saturation curves Saturation-pressure and phase-equilibria calcula-
calculated by MVNR merhods continue to close on tions in the two-phase region using MVNR are
each other as the over-all mixture becomes richer demonstrated in the second example problem.
in nC4. In other words, MVNR converged for Reservoir fluid composition is given in Table 2 and
near-critical mixtures. The Euclidean norms for the temperature is 235°F,
Point A of Fig. 3 are given in Table 1. These Fig. 4 presents the calculated phase distribution
norms show that MVNR converged quadratically between the calculated saturation pressure (4,952
near the solution. psia) and 4,SOO psia. The program required 10

/A’\
---EQUILIBRIUM TIE LINES C*2

—SUCCESSIVE
SUBSTITUTIONS
‘MVNR SATURATED VAPOR
-“-- TIE LINES
SATURATED VAPOR

Q
9

7S%c% C02
1s%co* 75%C*2 25%clD
2S%C10 25%C4 75 - :: C4
f
‘SATURATED LIQUID
FIG. 2 - CALCULATED PHASE EQUILIBRIA BY FIG, 3 - COMPARISON OF SOLUTION METHODS AND
SUCCESSIVE SUBSTITUTION, ccl ~ - n~, - c ~~ CALCULATED PHASE-EQUILIBRIA FOR CO~ - nC, -
FLUID SYSTEM AT 1,600 PSIA AND 160%. C ~o FLUID SYSTEM AT 1,600 PSIA AND 160%.
m SOCIETY OF PETSOLEUM ENGINEERS JOURNAL
.

iterations to calculate the saturation pressure. The MVNR Methods. ” For two-phase fluids, either the
Euclidean norm as a function of the iteration number L-X or V-Y iteration is used. The LoX iteration is
is presented in Table, 3. Quadratic convergence used for a fluid that is predominantly vapor (L < V),
was obtained near the soIuticm. Two-phase calcula- while the V-Y iteration is used for a fluid mixture
tions for the results presented in Fig. 4 each that is predominantly liquid (L > V). Therefore, the
required four or five iterations. The Euclidean norm V-Y iteration was used :s calculate the results in
as a function of iteration number for the calcular.ions Fig, 4. To demonstrate the effect of improper
at 4,950 psia is presented in Table 3, Quadratic selection of iteration variabIes, the phase equilibria
convergence was obtained near the solution. MV1-R in Fig. 4 was calculated using only the L ..Y iteration.
methods required a total of 52 iterations to calculate (The calculations began at 4,500 psia, not at the
the saturation pressure and the ph~se equilibria at saturation pressure.) The results were identical
10 pressures (Fig. 4). to those in Fig. 4, up to and including 4,850 psia.
The method of successive substitwions also was The L-X iteration required two additional iterations
used to calculate the phase equilibria presented in ac 4,900 psia, but did converge. At 4,950 psia, the
Fig. 4. The program allowed 100 iterations per L-X iteration diverged, The decision to switch
solution. At all pressures, the Euclidean norm iteration variables based on the “predominant
exceeded the sFecified tolerance of 10+ after 100 phase” is arbitrary as demonstrated by these
iterations. However, the calculated mole fraction of results because L was much larger than V before
liquid, L, as a function of pressure, agreed within convergence occurred. However, this method o:
0.001 of the values calculated by MVNR methods. selecting iteration variabIes is successful.
Table 3 presents the Euciidean norm for the first 10
iterations of the method of suc~.essive substitutions CONCLUSIONS
at 4,gs0 psia and the error nonm after 100 iterations.
Note that the norm was reduced to only 0,07. The Iteration sequences, referred to as minimum
method of successive substitutions required a total variable Newton-Raphson (MVNR) methods, were
of 1,000 iterations to calculate the phase equilibria deveIoped for calculating the phase e(,uilibria in
p~esented in Fig. 4, This did not include a the two-phase region and the saturation pressure of
saturation-pressure calculation. a multicomponenc fluid mixture. The phase-equilibria
For this example, MVNR methods required equations used for these calculations incorporate
approximately 10 ‘5 CPU second/component/iteration the RedIich-Kwong equation of state. This study
and the total CPU time, compared with the t;me presents results for a multicomponent fluid mixture
required by the method of successive substitutions, that demon strate the foIlowing conclusions.
was reduced by a factor of seven. MVNR methods 1. MVNR methods can be used to calculate phase
actually performed more calculations because the equilibria of C02 and rich-gas, multiple-contact,
saturation pressure and the two-phase calculations miscible oil recovery methods because these
were calculated. This clearly demonstrates the methods converge for near-critical fIuid mixtures.
computational advantage of MVNR methods. The method of successive substitutions diverges
The convergence of MVNR methods depends on for these mixtures.
the proper choice of iteration variables as 2. MVNR methods exhibit quadratic convergence
discussed in the section entitled,. c‘Description of for phase-equilibria calculations in the two-phase
region. The success of these calculations depends
on the proper selection of iteration variables, A
successful selection procedure is given.
3. MVNR methods converge quadratically for
49W-
TABLE 3- EUCLIDEAN NCRMS, RESERVOIR
OIL PHASE-EQUILIBRIA CALCULATIONS
= Method of SUCCeSelve
~ 4m — —. MVN R Methods substitutions
. Iteramtg Saturation Two-Phsse Two-Phase
a —. pressure at
— 4,950 psla
st 4,950 psla
2 o 240.6 3.28 3.3
~ 4700- .. *
1 4s.3 0,90 3.0s
& 2 5.67 0,13 2.91
3 1,71 0,0025 2,76
4 1.32 6. X 10-7 2.62
4600“ 5 0.81 2.48
6 0,43 2.36
7 0.136 2.24
8 0.0126 2,14
451Xl 1 9 9. x 10+’ 2,03
.5 .6 .7 .8 ,9 1.0 10 4. x 10-9 1.94
. .
L, MOU FRACTION LIQUID .
FIG. 4 — LIQUID MOLE FRACTION VS PRESSURE/ .
RESERVOIR-OIL SYSTEM. 100 0:07

.

saturation-pressure calculations. \$)a,flb = generalized component parameters —

4. MVNRmcthods are more economical, basedon function of the reduced temperature and
computer cost, than the method of successive Pitzer acentric factor for a component
substitutions.
SUBSCRIPTS
NOMENCLATURE i = component number
j = component number
a = temperature, pressure, and composition-
dependent component parameter L = liquid phase

A= temperature, pressure, and composition. s = saturation pressure

dependent phase parameter V = vapor phase
b = temperature, pressure, and composicion- SUPERSCRIPTS
dependent component parameter
9 = iteration number
B= temperature, pressure, and composition-
T = transpose of a vector
dependent phase parameter
Cii = interaction parameter for component i with REFERENCES
component j
1. Redlich, O. and Kwong, J.’ “N. S.: Cbern, Reu, 44
/iL = fugacity of component i in the liquid (1949) 233.
phase, psia
2, Zudkevitch, David and Joffe, Joseph: “Correlation
ijv = fugacity of component i in the vapor and Prediction of Vapor-Liquid Equilibria with the
psia Redlich-Kwong Equation of State,’\$ AICM? JOFW,
(Jan. 1970) Vol. 16, No. 1, 112-119.
elements of Jacobian correction matrix, ~
3. Soave, Giorgio: “Equilibriu.n Constante from a
L Jacobian comection matrix Modified Redlich-Kwong Equation of State, ~’ Chem.
Ki = equilibrium ratio, Ki = yi/%i Rng, Sci. (1972) VOX.27, 1197-1203.
L= mole fraction of liquid phase, mol liquid/ 4. Bruno, j. A., Yanosik, J, L., and Tierney, J. W.:
~~Diatillat~on calculations with Nonideal Mixtures,”
mol fluid Extractive and Aziotropic Distillation, Advances in
n = number of components in the fluid system Chemistry Series 11S, ACS, Washington, D. C, (1972),
,5. WUeon, Grant, M.: 8’A Modified Redlich-Kwong
P = pressure, psia --
Equation of State, Application to General Physical
P, = reduced pressuse Data Calculat~ons,*\$ paper No. lSC presented at the
R. gas constant AIChE 65th National Meeti(tg, Cleveland, Ohio, May
4-7, 1969.
Z= iteration variable vector
T= system temperature, OR
APPENDIX A
T, = reduced temperature
‘i = expontnt in Eq. 7, AL (hiL - biL)/BL REDLICH-KWONG EQUATION OF STATE
v = phase molar volume
v. mole fraction of vapor phqse, mol vapor/ The Redlich*Kwong equation of state (Eqs. 5
mol fluid and 6) and the parameters for this equation are
presented as follows.
Wi = exponent in Jlq. 8, A v (2aiv - biv) /Bv
The liquid phase,
‘i = mole fraction of component i in the liquid
phase, mol i/mol liquid ~-
23 z: + zL[~ - BL(BL + 1)1
Yj = mole fraction of component i in the vapor
phase, mol i/mol vapor -~BL=~, . . . . . . . . . .. (A-l)
‘i = mole fraction of component i in the fiuid,
mol i/mol fluid Vapor phase,
z. phase compressibility facror
aij = temperature a’ci pressure-dependent com- z: - z: + zv[~ - BV(BV + 1)1
ponent paramtter
pi = temperature and p:essure-dependent
“%%= 0’”””””””””””(A-2)
com-
onent parameter
8..al = Kronecher 8 function; ~ij = O for i #j, and parameters are
8jj=lfOri=j

;=
= error tolerance for the Euclidean norm
error equation for MVNR ~ =-- ! !
i-l j=l
Xpjuij’ “ “ “ “ “ “ “(A-3)
+= error vector for MVNR
\$“ componint fugacit y coefficient
a= Pitzet acentric factor
~= ! !
ifsll j=l
yp’j~ij’ “ “ “ “ “ “ “(A-4)
la SOCIETY OF PETROLEUM ENGINEERS JOURNAL
and

aY~
=-l; 2~~n . . . . . . . .. (B-3)
F
j
Differentiating the component material balances
with respect to the independent liquid-phase
compositions gives

axi
‘Is ; zli,j~n . . . . (B-4)
~ --.1 ij

Differentiating these equations with respect to L

‘v =?
i-l
YiBi’”” ”””””””(A-*) gives

f3io, ., . . . . . . . .. (A-9)
biL = ~ APPENDIX C

@i ELEMENTS OF \$ L-X, p-Y, ANDP-X

(A-1O)
ITERATIONS
biv=l\$” ””’”’o ”””””””.
L-X ITERATION

ai j = aji Wi n afiv
1 — (Yj-xj ) ;
1/2 ~il”~=-ij=2aYj

= (1 -

. . . . . . . ..*.*
Cij)
{1‘aipri
~
.. ‘ri
““”””*
‘aj
~2.5
rj
‘rj

“ (A-n)
I<i<n
——. . .. ”””
.(C-l)

QP~
= —..
bir . . . . . . .. .(A-12)
‘i T
ri .

The parameters aij and ~i are independent of com- 1

of these parameters with P-Y ITERATION .,
position so derivatives
respect tO %j and yj are zero. The component
parameters Qai and Qbi are independent of pressure wi = - afiv
— Ici, j&n, . . (C-3)
and composition so derivatives of these parameters
with respect to xjt y~\$ or P are Zero. The cOmPOnent
~i,j = E j ay’–
j

a\$i afiL afiv ; l<i<n

~i, ni-1 = ~ = ap ~p – –
APPENDIX B
. . . . . . . . . . . . . . .**.~*~ . (c-4)
DERIVATIVES OF DEFINING EQUATIONS
From the overall mat erial balance, a\$n+l
~n+l, j = ayj
~=-1 . . . . . . . . . . . . ..(B-l)
w
Because y2, y3, . . . , yn are iteration variables, the
restrictiveequation on the liquid-phase composition
is used to obtain derivatives of Yi with respect tO
Yj as

. . . . . . . ...*. ● ..**” ““” . (G5)

ayi - ~ . (B-2)
~ ij ; 2~i’j~n’ “ “ “ “ “ “
-.
1s1
JUNE, 1978
.
. ●

,,. ,
:. .,, .:
. !’
,. .81\$*+1
~n+l,n+l = 3p

f
PY~ ‘iL afiv
— .8fiL + ~L.- .—
..9+ f ‘%vpx~ af~L
f iv

. . . . . . . . . . .
ap P ‘iv

. . . . . . . . .
8P

(c%)
‘j L
. . . .,.
II —-

. . . . . . . . .
f:L axj

● . .

l~~n

. (c-9)
*

P-X, ITERATION

1’ a\$ni-l
w~ afiL
‘n+l, ni-l = ~p
l~i,j~n, . . (C-7)
ji,j = q” 3X4

ji,n+l=~=~-
Wi afiL afiv

3P
; I<i<rl
. .
,
=1-
n
l—-—
i=l fiv
II
(
PXi

. . . . . . . . . . . . . . .
afiV

ap
+ ‘iv
P

fiv
‘iL
afiL
T

. . . (c-lo)
“.
}

. . . . . . . . . . . . . . . . . . . . (C-8)
***