948 Ind. Eng. Chem. Process Des. Dev.

1985, 2 4 , 948-954
Literature Cited
Bromely, L. A. A I CM J . 1973, 19, 313.
Chen. C. C.: Britt. H. I. : Boston. J . F.: Evans, L. B. AIChE J . 1982, 28. 588.
Robinson, D. R.; J encks, W. P. J . Am. Chem. SOC. 1965, 87, 2462.
Rudakov, E. S.; Lutsyk, A. I . Russ. J . Phys. Chem. 1979, 53, 731.
Setschenows .1- J . z . . Stoechiom. Verwandfschaftsl 1889* 4 ,
I I I .
Gamble, D. S. can. J.. Chem. 1967; 45, 2805.
auque, w. F.; hnung, E. w.; Kunzler, J, E,; Rubin, T, R, J , A ~ , them,
Soc. 1980, 82, 62.
Gordon, J . E. The Organic Chemlstry of Electrolytes; Wiley-Interscience:
New York, 1975; Chapter 1.
Hanson, C.; Ismail, H. A. M. J . Appi. Chem. Biotechnol. 1975, 25, 319.
ed.; Wlley: New York, 1969; Val. 19, p 441.
Librovich, N. 8.; Zarakhani. N. G.; Vlnnik, M. I. Russ. J . Phys. Chem. 1971,
45, 1257.
Libr, M, Reaction ~chenisms in Sulfwic Acid and Saong Acid
Solutions; Academlc Press: New York, 1971; p 19.
Moeller, T. Inorganic Chemistry; Wlley: New York, 1952: p 532.
Munro, D. C. Appl . Specwosc. 1968, 22, 199.
PRzer, K. S.; Roy, R. N.; Sylvester, L. F. J . Am. Chem. SOC. 1977, 99,
Wagman, D. D.; Evans, W. H.; Parker, V. 8.: Schumm, R. H.; Halow, I.;
Bailey, S. M.; Churney, K. L.; Nutail, R. J . J . Phys. Chem. Ref. Data
1982, 1 1 , 1.
Young, T. F.; Maranviile, L. F.; Smith, H. M. I n The Structure of Electrolytic
Solutions; Hamer, W. J ., Ed.; Wiley: New York, 1959; Chapter 4.
Zemaitis, J . F. I n Thermodynamics of Aqueous Solutions with Industrial
Applications, ACS Symposium Series No. 133; Newman, S., Ed.; Ameri-
can Chemical Soclety: Washington, D.C.. 1980; Chapter 10.
Zemaitis, J . F. Chem. Solve., Inc., Morristown, NJ , personal communication,
1982.
Received f or review March 7 , 1983
Accepted October 25, 1984
Kirk, R, E,; mmr, D. F, ~n ~ ~ ~ ~ ~ l o ~ e d ~ of Chemical Technow-, 2nd
Zarakhani, N. G.; Vinnik, M. I. RUSS. J. PhYS. Chem. 1963, 37, 260.
AQ3n
Ridd&J. A.; Bunger, W. 6. Organic Solvents, Physical Properties and
~et hods of Purification, 3rd ed.; Wiley-Interscience: New york. 1970; p Presented at the AIChE Annual Meeting, LOS Angeles, CA, Nov
78. 1982. Sessions on Computers in Process Design and Analysis.
Thermodynamics of Petroleum Mixtures Containing Heavy
Hydrocarbons. 3. Efficient Flash Calculation Procedures Using the
SRK Equation of State
Karen Schou Pedersen, t S Per Thomassen, and Aage Fredenslund
Instftuttet for Kemiteknik, Danmarks Teknlske H oj skol e, DK-2800 Lyngby, Denmark, and STA TOIL, Den Norske Stats
Oljeselskap as , N-400 7 Stavanger, Norway
Two methods for drastically reducing oil/gas flash calculation computing times using the SRK equation of state
are described. (1) The characterization procedure given in parts 1 and 2 of this series has been extended with
a procedure for grouping hydrocarbon fractions. The predictions of the phase behavior using a total of only six
hydrocarbon fractions (C1-C30+) are as accurate as when 40 hydrocarbon fractions are used. (2) The above-
mentioned characterization procedure uses binary interaction coefficients (kt values) equal to zero for all hydro-
carbon-hydrocarbon interactions. In naturally occurring oil and gas systems, the contents of non-hydrocarbons
(mainly N, and COP) are often below 10 mol % . For such mixtures it isfound that using kt =0 for all interactions
(also with the non-hydrocarbons) has virtually no effect onthe calculated results. Explicit useof the assumption
of zero ki / values leads to substantial savings in the flash calculation computer time.
Introduction
The first two articles of this series (Pedersen et al.,
1984a,b) presented a characterization method for heavy
hydrocarbons that, when applied with the SRK equation
of state, permits accurate predictions of the phase behavior
of the phase equilibria in gas condensates and heavy oils.
About 40 hydrocarbon fractions (pseudocomponents) were
used to represent a petroleum mixture. However, in many
practical applications, such as oil reservoir computer sim-
ulation studies and two-phase flow calculations, it is not
possible to include such a large number of components in
the calculations. In this work, methods for reducing the
number of pseudocomponents are investigated.
Not only computer storage facilities but also computing
time may be limiting factors. Michelsen and Heidemann
(1981) have shown that the computing time needed to
calculate a critical point by using a cubic, two-constant
I nstituttet for Kemiteknik.
*Present address: Calsep Aps, Lyngby Hovedgade 29, DK-2800
8 STATOIL.
Lyngby, Denmark.
equation of state can be considerably reduced if all the
binary interaction coefficients ( ki j values) are equated to
zero. Similar computational advantages exist in two-phase
flash calculations (Michelsen, 1983). The consequences
of using kij values equal to zero are examined with respect
to computing time and accuracy for oil/gas systems.
Data
In order to develop the procedure for reducing the
number of pseudocomponents and to study the effect of
that and the effect of equating the kij values to zero, it is
necessary to have available reliable analytical, volumetric,
and phase equilibrium data for many different gas con-
densates and heavy oils. Such data and procedures for
obtaining the data are given by Pedersen et al. (1984a,b).
Two additional analytical techniques have recently been
introduced. By use of vacuum distillation at a pressure
of 130 N/m2, the end point of the TBP distillation has
been extended approximately from CZ0 to Cs0. Enough
carbon number fraction is collected to permit measure-
ments of specific gravity and molecular weight. In addi-
tion, the PNA analysis has been improved. A liquid-liquid
chromatographic method which on a preparative scale
0196-4305/85/1124-0948$01.50/0 0 1985 American Chemical Society
Ind. Eng. Chem. Process Des. Dev., Vol. 24, No. 4, 1985 949
splits each carbon number fraction into a paraffinic (P)
+naphthenic (N) fraction and an aromatic (A) fraction
is used. Refractive indices are used to determine the
relative amounts of P and N. (For further details see
Pedersen et al., 1984a.)
Three analytical data sets on North Sea reservoir fluids
are listed in Table VII. Sample A is a typical gas con-
densate for which TBP data are available up to C30+.
Sample B is a heavy oil and sample C a heavy gas con-
densate. The Cg-Cm fractions of sample C have been split
in a P +N and an A fraction.
The data given in Table VI1may be used to characterize
the reservoir fluids as recommended by Pedersen et al.
( 1984a).
Recommended Characterization Procedure for the
SRK Equation of State
The components of a reservoir fluid may be classified
as follows: defined components for which there exists a
quantitative GC analysis and for which T,, P,, and w are
known; TBP fractions covering the true boiling point
fractions between C7 and ca. Cs0, each containing many
different components. Careful analysis has shown that for
the North Sea oils, the best results were obtained by using
the Cavett (1964) relations for T, and P,
T, =768.071 +1.7134T50 - (0.10834 X 10-2)TW2 +
(0.3889 x 10-6)T503 - (0.89213 X 10-2)T50API +
(0.53095 x 10*)T5,2API +(0.32712 x 10-7)T502AP12 (1)
log P, =2.829 +(0.9412 X 10-3)T5~-
(0.30475X 10-5)T502 +(0.15141 X 10-8)T503 - (0.20876 X
104)T5,API +(0.11048 x 10-7)T502API +
(0.1395 X 10-9)T502AP12 - (0.4827 X 10-7)T50AP12 (2)
and the Lee-Kesler relations (Kesler and Lee, 1976) for
w =[In PB~ - 5.92714 +6.09648/T~, +
0:
1.28862 hT B , - 0.169347T~,e]/[15.2518 -
15.6875/T~, - 13.4721 In TBr -k 0.43577T~,6] (3)
(for TB, <0.8)
w =-7.904 +0.1352K - 0.007465@ +8.359T~, +
(1.408 - 0.001063K)TBr (4)
(for TBr >0.8).
In eq 1 and 2, API =141.5/SG - 131.5, where SG is the
60 OF/60 O F specific gravity. The critical temperature, T,,
and the midvolume boiling point of the fraction, T50, are
given in OFand P, is psia. In eq 3 and 4, TBr is the reduced
boiling point, TB/T,, and PB, is the reduced pressure,
PB/Pc, where P B is the pressure at which the boiling point,
TB, has been measured. K is the Watson characterization
factor that equals TB1/3/sG, where TBis in O R and SG is
as above.
The results are improved if the PNA distribution of each
boiling point fraction is taken into consideration as sug-
gested by Erbar (1977).
TBP Residue. In the case of gas condensate, the molar
composition of the TBP residue is estimated by assuming
a logarithmic relationship between the molar content, zN,
of a fraction and the corresponding carbon number, CN,
for C N L 7.
(5)
A, and B1 are constants determined by a least-squares fit
to the experimental TBP data for the C,-C, fractions. Cm
is the heaviest carbon number fraction considered, and
molar fractions below 5 X lo4 are neglected. If the
measured TBP residue is larger than the one calculated
C N =A1 +B1 In ZN
Table I. Input to Flash Program for Sample A (Gas
Condensate)
components mol % wt % T,, K P,, atm w
NZ 0.64 0.63 126.2 33.5 0.040
co2 9.16 4.19 304.2 72.8 0.225
C1 68.80 38.83 190.6 45.4 0.008
c2-c3 13.54 16.85 335.7 45.2 0.123
c4-c6 3.94 9.20 450.7 34.3 0.225
c&10 2.30 8.57 565.0 31.6 0.408
cll-c15 0.94 5.13 653.3 25.1 0.656
c16-c48 0.68 6.60 766.2 14.8 1.085
Table 11. Input to Flash Program for Sample B (Heavy
Oil)
comDonents mol % wt % T,., K P,. atm w
N2 0.34 0.10
COZ 0.84 0.41
C1 49.13 8.69
c2-c3 10.76 4.26
c4-c6 7.34 5.69
CTc15 19.69 29.87
cl6-c26 7.80 23.56
C27--C76 4.10 26.31
126.2 33.5 0.040
304.2 72.8 0.225
190.6 45.4 0.008
338.2 45.0 0.126
464.6 33.2 0.242
620.3 25.3 0.628
759.4 14.9 1.039
939.0 9.8 1.488
from the obtained A, and B1 values, A, and B1 are changed
accordingly as explained in part 1 of this series.
For the heavy oil systems, a different characterization
procedure is used.
The TBP distillation curve is extrapolated to 100 wt %
distilled off by fitting a fifth degree polynomial to the
experimental TBP data plus the following two artificially
generated data points:
(6)
- 10-0.2127T 1.103SG-0.6495
T80wt% - 50wt %
- 101.083T 0.7097~~0.6717
TlciJwt%- 5Gwt% (7)
Tsowt%, Tam%, and TlmW% are the temperatures in K at
which 50, 80, and 100 wt % have been distilled off. SG
is the specific gravity of the TBP sample that roughly
corresponds to the C7+fraction of the total mixture.
The specific gravities of the subfractions of the TBP
residue are obtained by assuming a logarithmic depen-
dence of SG against carbon number.
For the boiling points, the values of Katz and Firooza-
badi (1978) are used up to C45. For gas condensates, the
relation
TB (K) =97.58MW0.3323SG0.0460g (8)
is used for the fractions heavier than C45. For the heavy
oils, the weight-based TBP curve is related to carbon
number fractions assuming 6K boiling point intervals per
carbon number for the C45+fractions.
The characterization procedure described above yields
equation of state input parameters for typically 80 com-
ponents or pseudocomponents. Obviously, for extensive
process calculations it is necessary to drastically reduce
this number.
Reduction of the Number of Components
A procedure has been developed whereby all the hy-
drocarbon segments of the C7+fractions are given ap-
proximately equal importance. This is accomplished via
a grouping of carbon number fractions on weight basis.
Typical groupings are shown in Tables I and 11, from which
it may be seen that Nz, COz, and CH4 (C,) are given sep-
arate status, ethane and propane are grouped together
(C2-C3), and so are c4s, ci s, and c6s. The c7+fraction
is divided into three (or more) groups, which on a weight
basis are of approximately equal size.
The group critical properties are computed as weight-
mean values:
950 Ind. Eng. Chem. Process Des. Dev., Vol. 24, No. 4, 1985
T,i = x;MWjT,j/&jMW; (9)
J J
Pci =Cx;MWjPC;/CxjMW;
J J
wi =Cx;MWjwj/Cx;MWj (11)
J I
Index i is a group of carbon number fractions containing,
e.g., C1.(methane) only, or Cz and CB, or C4, C5, and C6, etc.
Index J runs from the lowest to the highest carbon number
within a group i. For the C,+carbon number fractions,
Pcj, Tcj, and wj in eq 9-11 are computed by using eq 1-4.
Tables I and I1 give examples of input parameters to
SRK-flash calculations using six hydrocarbon fractions
corresponding to, respectively, sample A and B of Table
VII. Such inputs plus overall compositions, temperatures,
and pressures are all that is required for the flash calcu-
lations.
Flash Calculations Using the SRK Equation of
State with Interaction Coefficients Equal to Zero
written as follows (Soave, 1972):
The Soave-Redlich-Kwong equation of state may be
RT a
p=---
U(U +b) u - b
For mixtures, the constants a and b, are given by the
mixing rules
a =CCYgjaij (13)
I J
b =CYibi
i
Indices i and j indicate summation over all components
and/or groups of carbon number fractions. Usually, aij for
i #j is given by
aij =(aiiajj)li2(1 - ki j ) (15)
where the binary interaction coefficients ki j are small,
typically zero for hydrocarbon-hydrocarbon interactions
and in the range 0-0.15 for others. The pure-component
parameters aii and bi are given by
aii =(0.42748R2Tci2/Pci) X [l +mi(l - (T/Tci)'/2)]2
(16)
bi =0.07780RTci/Pci (17)
In eq 16, mi is given in terms of the acentric factor, wi.
mi =0.480 +1.574wi - 0.176wi2 (18)
If the characterization procedure described in the pre-
vious two sections is used in connection with the SRK
equation, all hydrocarbon-hydrocarbon kij values (eq 15)
are equal to zero. If in addition ki; values for the non-
hydrocarbon-hydrocarbon and non-hydrocarbon-non-
hydrocarbon interactions may be assumed equal to zero.
Michelsens's (1983) simplified flash calculation procedure
described briefly below may be applied. Weshall later see
that k , =0 for all i - j interactions is a good approximation
for reservoir fluids even if they contain up to 10 mol %
N2 plus COz.
When k, =0, for all i and j , the mixing rule of eq 13 may
be written
a =a2 (19)
(y =Cyi(yi, (y. I = 11 (20)
with
1
Table 111. Deviations between Measured and Calculated
Saturation Points for 15 North Sea Reservoir Fluids"
3CH 6CH 10CH 20CH 40CH
fractns fractns fractns fractns fractns
AAD 18.1 5.6 5.0 5.2 5.0
BIAS -0.1 -3.2 -2.2 -2.5 -2.3
" CH: hydrocarbon; AAD: average absolute percent deviation;
BIAS: percent bias.
Table IV. Relative Computer Time for Flash Calculations
on Petroleum Mixtures"
3CH 6CH 10CH 20CH 40CH
program fractns fractns fractns fractns fractns
kij Z 0 0.08 0.16 0.25 0.46 1.00
kij =0 0.05 0.06 0.08 0.09 0.15
"CH: hydrocarbon; ki j #0: flash program with some of the
binary interaction coefficients different fromzero; ki, =0: flash
program with all binary interaction coefficients equal to zero.
With all kij =0 the fugacity coefficient for component
i is found to be given by
In +i ="( b RT - 1)- In (&(u - b)) -
Equation 21 is valid for both vapor and liquid phases. For
a vapor phase, vapor-phase properties uv, bv, and av are
used, and similarly uL, bL, and aL are used for the liquid
phase.
The equilibrium ratio, Ki =yi/xi, is given by
Ki =+?/c$? (22)
From eq 21 it may be seen that
Ki =Ki(T,P,aV,aL,bV,bL) (all i ) (23)
The volume, u, does not appear in eq 23 as u is given
implicitly by a, b, T, and P (see eq 12).
As may be readily shown (Michelsen 1983), a and b for
the feed (F), vapor (V), and liquid (L) are interrelated.
ffF =Pf f V +(1 - 6) aL (24)
b~=6bv +(1 - 6)bL (25)
P is the fraction of moles in the feed that appears in the
vapor phase. aF and bF are known.
Thus, in a flash calculation where T and P are specified,
Ki is a function of only three variables:
Ki =Ki(aV,bv,P) (all i) (26)
This means that, irrespective of the number of components
i, T, P, flash calculation procedures involving only three
variables can now be formulated. For multicomponent
mixtures such as reservoir fluids, this observation may be
used to improve drastically the efficiency of flash algor-
ithms. One such algorithm, i.e., the one used in this work,
is described in detail by Michelsen (1983).
Results
We examine separately the effect of reducing the num-
ber of components and of equating all ki j values to zero.
Reduction of the Number of Components. Dew and
bubble point calculations have been carried out for 15
reservoir fluids (the samples for which molar compositions
are given in the supplements to parts 1 and 2 of this series
and samples A-C of this work). The samples cover pe-
troleum mixtures ranging from light gas condensates to
heavy oils. Table I11 lists the deviations between exper-
imental and calculated saturation points using 3,6, 10, 20,
Ind. Eng. Chem. Process Des. Dev., Vol. 24, No. 4, 1985 951
:>
P
- Ex p er i men t
-.- 8 CH- f r a c t . ( k i j # O)
8 CH- f r l c t . ( k i j - 0 )
-
-
9
100 200 300 400
Pr essur e ( a t m)
Figure 1. Constant volume depletion results for sample A at T =
121 OC. Vapor and liquid densities have been determined by the
method of PBneloux et al. (1982).
100 200 300 400 500
Pr es s u r e ( at m)
Figure 2. Constant mass expansion results for sample C at T =129
"C. The SRK equation is used for the vapor densities and the
Alani-Kennedy equation (1960) for the liquid densities.
and 40 groups of carbon number fractions. The relative
flash calculation computer time is shown in Table IV. The
computer time (cpu s) needed to perform one flash cal-
culation for a mixture of eight components is given in
Table VI1 for five different computers.
Tables V and VI show measured and calculated phase
compositions of a gas condensate (Table VII, sample A)
and a heavy oil (Table VII, sample B) using 6 and 40
groups of hydrocarbon fractions.
Figures 1-4 present results of simulations of three PVT
experiments described in part 2 of these series using 6 and
40 hydrocarbon fractions. The pseudocomponents defined
for the total initial mixture are maintained for both the
gas and liquid phases for any further process calculations
on each separate phase.
Figure 1 shows measured and predicted liquid dropout
curves of a constant volume depletion process on sample
A. In this case the differences between the results for 6
and 40 CH fractions are so small that only one curve is
shown. Figure 2 shows results for a constant mass ex-
pansion process on sample C. Figures 3 and 4 present
952 Ind. Eng. Chem. Process Des. Dev., Vol. 24, No. 4, 1985
Table VII. Analytical Data for Samdes A. B. and C"
PNA distribution, mol %
sp grav
comDonents mol % (15 "Ci15 "C) MW P N A
0.64
9.16
68.80
8.43
5.11
0.81
1.45
0.52
0.53
0.63
0.83
0.95
0.52
0.26
0.20
0.17
0.16
0.15
0.11
0.086
0.078
0.068
0.050
0.046
0.035
0.025
0.034
0.023
0.017
0.018
0.014
0.012
0.013
0.047
0.34
0.84
49.23
6.32
4.46
0.86
2.18
0.93
1.33
2.06
3.33
4.06
2.76
1.33
1.79
1.70
1.81
1.46
1.49
1.08
1.13
0.99
0.88
Sample A (Gas Condensate)*
0.741
0.780
0.807
0.819
0.810
0.828
0.849
0.857
0.868
0.872
0.859
0.854
0.866
0.873
0.876
0.876
0.875
0.877
0.876
0.878
0.882
0.886
0.889
0.908
Sample B (Heavy Oil)*
96
107
121
134
147
161
175
190
206
222
237
251
263
339 G o +
0.7395
0.7518
0.7756
0.7930
0.7902
0.8060
0.8203
0.8311
0.8446
0.8515
0.8542
0.8561
0.8663
99
106
120
139
146
160
174
194
205
218
234
248
265
0.50
0.45
0.48
0.47
0.56
0.55
0.54
0.49
0.52
0.55
0.57
0.70 Cia+
0.56
0.58
0.58
0.57
0.62
0.59
0.57
0.57
0.53
0.53
0.56
0.58
0.57
0.42
0.38
0.27
0.30
0.27
0.24
0.22
0.27
0.20
0.19
0.20
0.11
0.39
0.33
0.30
0.33
0.31
0.33
0.32
0.31
0.32
0.31
0.30
0.28
0.28
0.08
0.17
0.25
0.23
0.17
0.21
0.24
0.24
0.28
0.26
0.23
0.19
0.05
0.09
0.12
0.10
0.07
0.08
0.11
0.12
0.15
0.16
0.14
0.14
0.15
c 19
c20+ 7.64 0.9350 465 0.56 0.20 0.24
sp grav (15 "C/15 "C)
. _
PNA distribution mol 70
total
components mol % fractn P +N A MW P N A
Sample C (Heavy Gas Condensate)
N2 0.12
CO2 2.49
C1 76.43
c2 7.46
c3 3.12
i-C, 0.59
i-C5 0.50
n-C5 0.59
c6 0.79
n-C, 1.21
Ind. Eng. Chem. Process Des. Dev., Vol. 24, No. 4, 1985 953
Table VI1 (Continued)
sp grav (15 OC/15 "C)
PNA distribution mol %
total
components mol % fractn P +N A MW P N A
C, 0.95 0.726 95 0.564 0.361 0.076
1.08
0.78
0.592
0.467
0.345
0.375
0.304
0.237
0.208
0.220
0.169
0.140
0.833
0.747
0.769
0.781
0.778
0.785
0.802
0.815
0.817
0.824
0.825
0.831
0.841
0.873
0.738
0.756
0.765
0.777
0.785
0.792
0.877
0.805
0.807
0.812
0.820
0.827
0.862
0.874
0.883
0.898
0.953
0.934
0.987
0.951
0.949
0.968
0.975
0.962
106
116
133
152
164
179
193
209
218
239
250
264
377
0.113
0.483
0.530
0.681
0.757
0.709
0.635
0.729
0.624
0.668
0.675
0.652
0.519
0.611
0.311
0.275
0.193
0.123
0.183
0.209
0.168
0.232
0.185
0.192
0.190
0.320
0.277
0.206
0.195
0.126
0.120
0.108
0.156
0.103
0.144
0.147
0.133
0.158
0.161
"P, paraffinic compounds; N, naphthenic compounds; A, aromatic compounds. bThe PNA analyses of samples A and B are only based on
measurements of refractive index.
Table VIII. TBP Data for Samde B"
u
5 ; 1.4
> .
El
5 1.3
-
0 >
..........
Ex p e r i me n t
4 0 CH- t r ac t i o n .
6 CH- f r ac t i o n s
- -
1.1
l '* I
1 1 , I
400 300 200 100 0
Pr es s u r e ( a t m )
Figure 3. Differential liberation results for sample B at T =93.3".
The SRK equation is used for the vapor densities and the Alani-
Kennedy equation (1960) for the liquid densities.
- .-
0 Ex p a r i me n 1
......... 40 CH- t r ac t i o n 1
6 CH- l r a c t i o n i
I I I I
4 00 300 200 100 0
Figure 4. Differential liberation results for sample B at T =93.3
"C. The oil density has been determined by using the Alani-Ken-
nedy equation (1960).
relative and absolute liquid volumes of sample B during
a differential liberation process.
Calculation with All Binary Interaction Coeffi-
cients Equal to Zero. As already indicated, wenormally
use ki j values equal to zero for interactions between hy-
drocarbons and nonzero values for interactions between
Pr es s u r e ( a t m )
temp, cum wt temp, cum wt temp, cum wt
"C % distilled "C % distilled "C % distilled
100 10.54 218 28.43 288 43.10
127 14.54 237 32.48 304 46.52
152 19.17 255 36.13 318 49.67
176 21.55 272 40.06 332 52.67
197 24.92
"specific gravity of TBP sample =0.8499 (15 "C/15 "C).
non-hydrocarbons and between hydrocarbons and non-
hydrocarbons. The influence on the predicted saturation
points of setting all ki j values equal to zero has been
checked for mixtures containing COz and N2 If COP +N2
constitutes less than 1.5 mol % of the total mixture, the
calculated saturation points deviate less than 170 from
those calculated with kij #0. For sample A, a gas con-
densate that contains about 10 mol % non-hydrocarbons,
the effect of kij =0 on the calculated dew points is 4.0%.
Relative computer times for flash calculations with the
kij =0 version are given in Table IV. Calculated phase
compositions of samples A and B at separator conditions
using the kj. =0 version are shown in Tables V and VI.
The simdations of PVT experiments mentioned in the
previous section have been repeated with the kij =0 ver-
sion. The liquid dropout curve obtained for sample A by
using six hydrocarbon fractions is shown in Figure 1. The
results of the constant mass expansion calculation on
sample C and those of the differential liberation calculation
on sample B are not shown because they are almost in-
distinguishable from the results with k, #0.
Discussion
I t is evident from Table IV that much computer time
may be saved in oil/gas flash calculations if the kij values
(see eq 15) are equated to zero for all i-j interactions. The
calculations performed on gas condensates and heavy oils
show that the kij =0 approximation has little influence
on the results for North Sea reservoir fluids, which seldom
contain more than 10 mol % CO, +N,. If the content of
COz and N, is less than 1.5 mol % , the influences on the
predicted phase behavior and saturation points are neg-
ligible; only the equilibrium ratio for CO,, KCo2, is affected
(see Table VI). If the content of COz +Nz approaches 10
mol 70, the ki, =0 and k , #0 versions of the programs
give saturation pressures that differ by 4%. This is the
case for sample A of Table VII. Only minor differences
are in this case observed in the two-phase region, except
for the calculated CO, concentrations.
954 Ind. Eng. Chem. Process Des. Dev., Vol. 24, No. 4, 1985
Table IX. Computer Time (cpu 8 ) Needed for One
One-Phase and One Two-Phase Flash Calculation on
Different ComputersD (Eight Component Mixture)
program one-phase two-phase
comwter version region region
IBM 3081 k,j f 0
k , =0
Burrough B-7800 k, , #0
k,, =0
VAX l l /750 k,, f 0
UNIVAC ll00/62 k , #0
k , =0
HP9020 k L J #
0.021
0.012
0.080
0.030
0.203
0.279
0.340
0.150
0.029
0.010
0.125
0.060
0.326
0.374
0.530
0.250
The two-phase calculation corresponds to the data and results
The above results indicate that one may take advantage
of the k,, =0 approximation in many applications where
flash calculations are needed for petroleum mixtures. This
conclusion does not apply to situations where mixtures
with appreciable contents of non-hydrocarbons are present,
e.g., COz injection studies.
Also, reducing the number of components (or pseudo-
components) greatly reduces the computer time needed
for flash calculations (see Table IV). The results obtained
show that a total of only six groups of carbon number
fractions gives saturation points and PVT properties that
are sufficiently accurate for many practical purposes. As
shown in Tables V and VI, the only discrepancy is the
content of heavy components in the vapor phase.
The liquid dropout curves predicted for samples A and
C deviate somewhat from the measured ones as seen in
Figures 1 and 2. Very accurate predictions are obtained
for the dew point of sample A, but there are some problems
with the liquid volumes. As the liquid phase only con-
stitutes a few volume percent of the total volume, our
results are satisfactory for most practical purposes. Sample
C is a heavy gas condensate for which wecalculate the
following critical properties: T, =43.8 "C; P, =416 atm.
The reservoir temperature is 129 "C, which means that
the mixture at the dew point in Figure 2 is nearly critical.
The near-critical condition is also reflected in the calcu-
lated phase compositions, which slightly below the dew
point were nearly identical. Problems with predictions of
densities are to be expected in the critical region, and we
attribute most of the discrepancies shown in Figure 2
hereto. I t is worth noting that no computational problems
were encountered in these flash calculations.
Conclusions
A simplified SRK calculation procedure (Michelsen,
1983) enables considerable savings in computer time for
mixtures for which all k, , values can be set to zero. The
computer time is only a weak function of the number of
components, and the storage requirements are drastically
reduced. Irrespective of the number of components, only
three variables need to be stored between the iterations.
For mixtures of hydrocarbons, COz, and N2 in which the
COz +N2 content does not exceed 10 mol % , all k, values
can be set to zero without significant influence on the
predicted phase behavior.
I n addition, a new procedure has been developed that
allows reservoir fluids to be represented by a total of six
hydrocarbon pseudocomponents only. I t has been shown
of Table V.
that six pseudocomponents are sufficient to represent
densities and saturation points as well as what was pre-
viously reported in parts 1 and 2 of this series using 40
hydrocarbon fractions.
Acknowledgment
Wethank Arne Hole of Statoil and M. L. Michelsen of
Instituttet for Kemiteknik for their continued interest in
this work.
Nomenclature
a
A1
API
b
R,
CN
k ,
K
K,
m
MW
P
R
SG
T
T B
T50
T50 wt %
XI
Y1
21
CY
P
41
Br
F
i, j
L
V
W
C
SRK parameter defined in eq 12
parameter defined in eq 5
API specific gravity (=141.5/SG-131.5)
SRK parameter defined in eq 12
parameter defined in eq 5
carbon number
binary interaction coefficient
Watson characterization factor (TB'/3/sG)
equilibrium factor, component i
SRK parameter defined in eq 18
molecular weight
pressure
gas constant
specific gravity
temperature
normal boiling point
midvolume boiling point
midweight boiling point
liquid-phase mole fraction of component i
vapor-phase mole fraction of component i
feed mole fraction of component i
Greek Symbols
parameter defined in eq 20
vapor mole fraction
acentric factor
fugacity coefficient of component i
Subscri pt s
reduced boiling point
critical property
feed
component or carbon number group indices
liquid
vapor
Registry No. C02, 124-38-9; N2, 7127-31-9.
Literature Cited
Alani, H. G.; Kennedy, H. T. Pet . Trans., Am. Inst. Mi n. Eng. 1960, 219,
200.
Cavett, R. H. "Physical Data for Distillation Calculations - Vapor-Liquid
Equiiibrla", presented at the 27th Midyear Meeting, API Division of Refin-
ing, San Francisco, May 15, 1964.
Erbar, J . H. Gas Processors Association, Tulsa, OK, 1977, Research Report
13.
Katz, D. L.; Firoozabadi, A. Pet . Techno/. 1978, 20, 1649.
Kesler, M. G.; Lee, B. I . Hydrocarbon Process. 1978, 55, 153.
Michelsen, M. L.; Heidemann, R. A. AI ChEJ. 1981, 27, 521.
Michelsen, M. L. "Simplified Flash Calculations for Cubic Equations of State",
SEP 8313; Instituttet for Kemiteknik: Lyngby. Denmark, 1983.
Pedersen, K. S.; Thomassen, P.; Fredenslund, Aa. a. Ind. Eng. Chem. Pro-
cess Des . Dev. 1984, 23, 163-170. b. Ind. Eng. Chem. Process Des.
Dev., 1984, 23, 566-573.
PEneloux, A,; FrBze. P. Nuid Phase Equilib. 1982, 8 . 7.
Soave, G. Chem. Eng. Sci . 1972, 27, 1197.
Received for reuiew October 3, 1983
Revised manuscript received October 2, 1984
Accepted October 23, 1984