Yan 2015

Fluid Phase Equilibria 386 (2015) 96124
Contents lists available at ScienceDirect
Fluid Phase Equilibria

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / fl u i d
PVT modeling of reservoir uids using PC-SAFT EoS and Soave-BWR EoS
Wei Yan 1, Farhad Varzandeh 2 , Erling H. Stenby *
Center for Energy Resources Engineering (CERE), Department of Chemistry, Technical University of Denmark, DK-2800 Lyngby, Denmark
A R T I C L E I N F O
A B S T R A C T
Article history:
Received 7 July 2014
Received in revised form 19 November 2014
Accepted 25 November 2014
Available online xxx
Cubic equations of state, such as the SoaveRedlichKwong (SRK) and the PengRobinson (PR) EoS, are
still the mostly used models in PVT modeling of reservoir uids, and almost the exclusively used models
in compositional reservoir simulations. Nevertheless, it is promising that recently developed non-cubic
EoS models, such as the Perturbed Chain Statistical Associating Fluid Theory (PC-SAFT) EoS and the Soave
modied BenedictWebbRubin (Soave-BWR) EoS, may partly replace the roles of these classical cubic
models in the upstream oil industry. Here, we attempt to make a comparative study of non-cubic models
(PC-SAFT and Soave-BWR) and cubic models (SRK and PR) in several important aspects related to PVT
modeling of reservoir uids, including density description for typical pure components in reservoir
uids, description of binary VLE, prediction of multicomponent phase envelopes, and PVT calculation of
reservoir uids. Extensive data are used in the comparison and the four models are treated as equally as
possible. We adopt the method of Pedersen et al. as the framework for heptanes plus characterization and
the same correlations for estimating the critical properties for SRK, PR and Soave-BWR. For PC-SAFT, new
correlations for estimating its model parameters in heptanes plus are developed. The results reveal that
the non-cubic models are clearly advantageous in density calculation of pure components. For binary VLE
and multicomponent phase envelopes, the results are similar for the four models. For PVT prediction, the
non-cubic models show advantages in some high pressure high temperature (HPHT) uids but no clear
advantages in general, indicating the necessity for further improvement of the characterization
procedure.
2014 Elsevier B.V. All rights reserved.
Keywords:
Equation of state
Reservoir uids
C7+ characterization
PC-SAFT
Soave-BWR
1. Introduction
In the oil and gas industry, there is an obvious contrast between
the upstream and the downstream with regards to adopting newly
developed equations of state (EoS). In the downstream, these new
EoS models are usually quickly accepted while in the upstream, the
most widely used EoS models are still the cubic equations of state
(CEoS), specically the SoaveRedlichKwong (SRK) EoS [1] and
the PengRobinson (PR) EoS [2]. These two CEoSs are the most
used models in PVT modeling of reservoir uids and almost the
exclusively used models in compositional reservoir simulations.
There are many reasons for the phenomenon: in contrast to the
large variety of chemicals in the downstream, hydrocarbons are
the major components of concern in the upstream; CEOSs are
generally believed to provide enough accuracy in routine PVT
modeling; as a main user of PVT modeling results, compositional
* Corresponding author. Tel.: +45 4525 2012; fax: +45 4588 3136.
E-mail addresses: weya@kemi.dtu.dk (W. Yan), farvar@kemi.dtu.dk
(F. Varzandeh), ehst@kemi.dtu.dk (E.H. Stenby).
1
Tel.: +45 4525 2914; fax: +45 4588 2258.
2
Tel.: +45 4525 2033.
0378-3812/$ see front matter 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.uid.2014.11.022
reservoir simulation has much bigger uncertainties in its geology

and uid ow parts, making the accuracy of PVT modeling a less
important issue in many situations; last but not the least,
compositional reservoir simulations are already time-consuming
compared with black oil reservoir simulations, using a more
complicated thermodynamic model will further increase the
computation time and therefore a strong justication for doing
that is needed. Despite all the above reasons for sticking to the
traditional models for PVT modeling, the upstream industry does
have the needs to try more complex thermodynamic models as
more exploration and production activities are carried out at
deeper reservoirs, which means higher pressures and higher
temperatures. High pressure and high temperature (HPHT)
reservoirs are risky to develop and require substantial investment,
especially if they are offshore. To make the exploration and
production viable, it is preferred to make more careful evaluation
based on more accurate PVT modeling of HPHT reservoir uids. In
this area, using non-cubic EoSs which provide a better density
description can be advantageous.
We have focused on two non-cubic EoSs in this study: the
Perturbed Chain Statistical Associating Fluid Theory (PC-SAFT) EoS
[3,4], which has a solid basis of statistical mechanics, and the Soave
modied BenedictWebbRubin (Soave-BWR) EoS [5], which gives
W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124
97
Table 1
Applicable ranges for the reference EoS models for various components.
accurate density description even around the critical point. It

should be mentioned that PC-SAFT has been applied to reservoir
uid systems [619], especially for asphaltene precipitation
modeling [615]. The focus of this study is a bit different and it
is mainly on the performance of PC-SAFT and Soave-BWR in basic
PVT modeling, including a comparison with the classical CEoS
models SRK and PR. Compared with PC-SAFT, Soave-BWR has
received much less attention since it was proposed [5]. We select
the model here mainly because it represents a different class of
EoSs, namely, the virial-type EoS and also because its density
accuracy for many components in reservoir uids is even better
than PC-SAFT.
In the following sections, we will rst briey review the PCSAFT and Soave-BWR models. A comprehensive comparison will
then be presented between two non-cubic models, PC-SAFT and
Soave-BWR, and two cubic models, SRK and PR, in description of
pure component densities, calculation of binary VLE, prediction of
multicomponent phase envelope, and nally PVT modeling of
reservoir uids. In connection with PVT modeling, characterization
of heptanes plus (C7+) is required and a characterization method
for PC-SAFT with newly developed correlations for its model
parameters will be introduced.
In addition to covering various aspects related to the basic PVT
modeling, this comparative study has included a large amount of
data, either experimental ones collected from the open literature
and condential reports, or synthetic ones generated by the most
reliable reference EoS models. To make a fair comparison between
the four selected models, we have attempted to treat them in the
same manner. For example, all the optimal binary interaction
parameters are regressed for the four models and then used in the
subsequent predictions for multicomponent phase envelopes and
PVT modeling. In the PVT modeling part, the parameters for the
pseudo components in C7+ are generated by essentially the same
characterization procedure with differences only in the correlations for model parameters, and no parameter tuning using the
experimental PVT data has been performed for a particular model.
In the comparative study, particular treatments in favor of a
specic model have been avoided so that the results can reveal the
advantages as well as limitations of the compared models with as
little bias as possible.
Components
Tmin (K)
Tmax (K)
Pmax (bar)
rmax (mol/L)
Tr,min
N2
CO2
H2S
CH4
C2H6
C3H8
nC4H10
nC5H12
nC6H14
nC7H16
nC8H18
nC9H20
nC10H22
nC11H24
nC12H26
iC4H10
iC5H12
iC6H14
Benzene
Toluene
63.15
216.59
187.70
90.69
90.37
85.53
134.90
143.47
177.83
182.55
216.37
219.70
243.50
247.54
263.60
113.73
112.65
119.60
278.67
178.00
2000
2000
760
625
675
650
575
600
600
600
600
600
675
700
700
575
500
550
750
700
22000
8000
1700
10000
9000
10000
2000
1000
1000
1000
1000
8000
8000
5000
7000
350
10000
10000
5000
5000
53.15
37.24
29.12
40.07
22.42
20.60
13.86
11.20
8.85
7.75
6.69
6.06
5.41
4.97
4.53
12.90
13.30
9.38
11.45
10.58
0.50
0.71
0.50
0.48
0.30
0.23
0.32
0.31
0.35
0.34
0.38
0.37
0.39
0.39
0.40
0.28
0.24
0.24
0.50
0.30
Finally, it is obvious that non-cubic EoS models are less matured

in PVT modeling than cubic models. Even for PC-SAFT which has
been frequently used in recent asphaltene precipitation modeling
[615], there are few systematic studies, at least to our knowledge,
about its comparison with SRK and PR in basic PVT modeling. In our
opinion, the basic PVT modeling issue should be addressed before
the non-cubic models can be really accepted for routine use in the
upstream of the oil and gas industry. With a better understanding
of the strengths and weaknesses of the non-cubic models, it is
easier for the engineers to decide whether and when to use these
models and where to improve these models. We notice that some
recent cubic EoS models like PPR78 [2022] and UMR-PRU [2325]
are predictive in the sense that all the interaction parameters can
be calculated using the group-contribution concept. Neither PCSAFT nor Soave-BWR has developed to such a stage in terms of PVT
modeling. Our comparison here uses just the original SRK and PR
where the temperature independent interaction parameters are
regressed from the binary data.
Table 2
AAD% and maximum absolute deviation in the calculated density for different EoSs within 273500 K and 02000 bar without volume transition.
Components
SRK
PR
PC-SAFT
Soave-BWR
CPA
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
N2
CO2
H2S
CH4
C2H6
C3H8
nC4H10
nC5H12
nC6H14
nC7H16
nC8H18
nC9H20
nC10H22
nC11H24
nC12H26
iC4H10
iC5H12
iC6H14
Benzene
Toluene
2.24
4.88
3.04
1.29
2.07
3.28
5.21
7.67
10.16
12.37
14.46
16.54
18.33
18.38
21.48
4.23
5.85
7.84
8.70
11.08
7.01
21.00
16.48
16.47
19.91
20.60
20.07
22.71
23.14
20.09
22.26
24.60
26.56
26.70
29.60
21.24
26.58
21.31
15.92
17.90
9.35
5.04
7.61
8.78
8.52
7.42
5.51
3.62
3.00
3.37
4.84
7.02
9.02
9.08
12.54
6.48
5.00
3.66
3.36
2.74
13.62
13.28
9.51
12.43
11.85
12.01
12.82
13.60
14.44
11.05
13.46
16.06
18.25
18.40
21.63
13.44
18.97
13.82
6.84
8.57
3.23
1.23
0.47
1.35
2.07
1.71
1.61
1.22
1.42
1.57
1.77
2.20
2.05
2.09
2.01
2.23
1.56
1.93
0.73
1.04
4.91
20.90
6.72
6.11
5.73
3.86
8.38
15.71
6.91
5.85
6.93
7.92
8.27
9.31
10.14
6.55
8.37
8.37
3.05
4.88
0.57
0.74
0.38
0.88
0.64
1.44
1.81
1.70
1.35
1.29
1.04
0.85
0.78
2.61
0.63
2.17
2.78
3.17
1.03
0.79
2.63
3.81
3.45
4.08
4.15
4.57
3.46
4.73
3.39
3.51
3.22
3.38
3.44
4.74
2.84
4.48
4.72
5.26
3.12
1.62
3.94
3.18
1.87
2.62
3.24
4.45
5.50
2.86
2.92
3.00
3.24
3.82
4.41
5.32
5.86
3.69
2.73
2.63
5.84
14.77
35.97
6.54
10.47
8.25
17.60
19.13
8.82
9.51
10.51
11.86
13.16
14.51
15.19
7.65
7.60
8.10
Average
8.95
21.01
6.30
13.70
1.67
7.94
1.33
3.73
3.63
12.53
No CPA model parameters for i-pentane and i-hexane are available and these two components are excluded in the average for CPA.
98
Fig. 1. AAD% in the calculated density for different EoSs within 279500 K and 02000 bar without volume transition.
Fig. 2. Contour map of deviation in the calculated densities for methane by different EoSs: (a) SRK; (b) PR; (c) PC-SAFT; and (d) Soave-BWR. The relative deviations (%) are
labeled on the contour lines. The green and the red circles indicate the conditions for the minimum and the maximum deviations, respectively. (For interpretation of the
references to color in this gure legend, the reader is referred to the web version of this article.)
99
Table 3
AAD% and maximum absolute deviations in the calculated saturated liquid density for different EoSs without volume transition.
Components
SRK
PR
PC-SAFT
Soave-BWR
CPA
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
N2
CO2
H2S
CH4
C2H6
C3H8
nC4H10
nC5H12
nC6H14
nC7H16
nC8H18
nC9H20
nC10H22
nC11H24
nC12H26
iC4H10
iC5H12
iC6H14
Benzene
Toluene
4.41
12.66
6.50
4.94
6.92
8.36
9.72
11.63
13.07
14.70
15.84
17.56
18.37
17.78
20.10
9.10
10.06
11.32
11.66
13.27
20.02
24.91
19.37
21.05
22.82
23.82
25.02
27.02
25.49
25.62
25.78
30.51
30.18
32.34
33.38
24.11
26.54
24.75
25.81
27.06
9.58
4.39
7.11
9.04
6.96
5.71
4.66
3.32
2.94
3.79
5.06
7.03
7.94
7.30
9.93
5.05
4.48
3.38
4.04
2.85
13.34
16.94
10.69
12.78
14.48
15.52
16.78
18.98
16.97
17.40
17.62
22.80
22.39
24.66
26.02
15.83
18.46
16.51
17.72
19.02
2.30
2.58
0.68
0.91
0.84
0.71
0.90
1.75
1.28
1.64
1.48
1.12
1.27
1.26
1.52
0.59
1.42
1.27
1.26
1.42
3.22
14.47
3.05
5.55
2.84
4.34
6.79
7.35
12.40
13.64
16.77
12.11
10.75
12.08
9.47
5.51
3.47
9.43
9.27
9.55
2.16
2.45
1.19
1.84
0.44
1.20
1.54
1.33
1.33
1.54
1.61
1.54
0.72
2.18
1.03
1.56
1.17
2.64
1.73
2.11
1.31
3.20
3.93
3.07
1.61
0.74
1.12
1.04
4.72
6.83
8.39
3.94
6.77
6.77
4.51
1.54
1.50
4.57
1.23
1.43
3.04
2.19
2.09
2.97
2.18
2.43
4.12
1.73
2.03
1.87
1.68
2.15
1.75
2.31
2.36
2.00
1.90
1.61
21.00
30.67
16.70
19.98
21.48
31.31
29.81
30.87
40.40
29.32
25.74
40.75
21.92
37.02
39.82
18.57
20.23
28.15
Average
11.90
25.78
5.73
17.74
1.31
8.60
1.56
3.41
2.25
27.99
No CPA model parameters for i-pentane and i-hexane are available and these two components are excluded in the average for CPA.
2. Model description
2.1. The PC-SAFT equation of state
The PC-SAFT EoS proposed by Gross and Sadowski [3,4] is one of
the successful modications of the original SAFT EoS [26]. Unlike
SAFT, PC-SAFT uses the mixture of hard-sphere chains as the
reference system and then introduces the dispersive attractions.
PC-SAFT is proposed to model asymmetric and highly non-ideal
systems. PC-SAFT can be expressed in terms of the reduced
~:
Helmholtz energy a
~
a
A
~ id a
~ hc a
~ disp a
~ assoc
a
NkT
(1)
~ id is the ideal gas contribution, a

~ hc is the contribution of the
where a
~ disp is the dispersion
hard-sphere chain reference system, a
contribution arising from the square well attractive potential
~ assoc is the association contribution based on Wertheims
and a
theory [4,26]. When PC-SAFT is applied to a system only consisting
~ assoc
of non-associating components, as the case in this study, the a
term in Eq. (1) disappears. Without the association term, the
expression used in Eq. (1) is comparable to a conventional cubic
~ id a
~ hc corresponding to the repulsive contribution in a
EoS, with a
~ disp corresponding to the attractive contribution.
cubic EOS and a
However, the repulsive and attractive contributions in a cubic EOS
~ hc and a
~ disp in the PC-SAFT are theoretically
are empirical while a
~ hc in PC-SAFT takes into
derived from statistical mechanics. a
account both the hard sphere contribution and the chain
contribution:
Fig. 3. AAD% in the calculated saturated liquid density for different EoSs without volume transition.
100
Table 4
Trends in the temperature volume translation parameters: the parameters that are
not monotonic decreasing with temperature are marked with N while the others
are kept blank.
Components
N2
CO2
H2S
CH4
C2H6
C3H8
nC4H10
nC5H12
nC6H14
nC7H16
nC8H18
nC9H20
nC10H22
nC11H24
nC12H26
iC4H10
iC5H12
iC6H14
Benzene
Toluene
SRK
N
N
N
PR
PC-SAFT
Soave-BWR
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
~ hc m a
~ hs a
~ chain m a
~ hs
a
and di is the temperature-dependent segment diameter of

component i [27]
h

e i
(6)
di s i 1 0:12exp 3 i
kT
The radial distribution function at contact ghs
ij in Eq. (2) is given
by
ghs
ij
~ disp 2prI1 h; mm2 es 3 prmC 1 h; mI2 h; mm2 e2 s 3

a
(2)
m2 e2 s 3
where m is the mean segment number in the mixture:

X
xi mi
The hard-sphere term is given by the mixture version of the

CarnahanStarling EoS for hard-spheres
!
#
"
3
1 3z1 z2
3z32
z
hs
2
~
(4)
a
z0 ln1 z3
z0 1 z3 z3 1 z3 2
z23
s ij
with zn dened by
6
2
XX
eij
xi xj mi mj
s 3ij
kT
i
j
(10)
with the conventional BerthelotLorentz combining rules for eij

and s ij :
p
eij eii ejj 1 kij
(11)
(3)
p X
zn r xi mi dni n 2 f0; 1; 2; 3g
(8)
where the integrals I1 and I2, and the compressibility expression C1

are functions of m and the packing fraction h (or z3). The van der
Waals one-uid mixing rules are applied to the dispersion term,
resulting in

XX
eij 3
m2 es 3
xi xj mi mj
s ij
(9)
kT
i
j
(7)
~ disp is modeled using a second order

The dispersive term a
perturbation theory on chain molecules rather than hard spheres,
~ disp is given
which makes it different from other versions of SAFT. a
by
N
N
N
N
N
N
X
xi mi 1lnghs
ii

2
di dj
di dj
1
2z2
2z22
2
1 z3
di dj 1 z3
di dj 1 z3 2
s i s j
2
(12)
Despite the complex form of the PC-SAFT EoS, there are only three
model parameters for a non-associating component, the chain
length m, the segment diameter s and the segment energy e.
(5)
Table 5
AAD% and maximum deviations in the calculated density for different EoSs within 273500 K and 1000 bar with volume transition.
Components
SRK
PR
PC-SAFT
Soave-BWR
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
N2
CO2
H2S
CH4
C2H6
C3H8
nC4H10
nC5H12
nC6H14
nC7H16
nC8H18
nC9H20
nC10H22
nC11H24
nC12H26
iC4H10
iC5H12
iC6H14
Benzene
Toluene
0.84
1.68
1.09
1.09
1.14
1.05
0.96
0.90
0.87
0.83
0.83
0.83
0.81
0.86
0.82
1.01
0.92
0.89
0.56
0.58
8.02
19.64
15.70
16.82
19.78
20.25
19.45
21.40
21.34
15.65
12.11
9.85
8.17
6.00
5.56
20.87
26.10
20.44
11.47
9.60
0.72
1.13
0.86
0.77
1.05
1.08
1.06
1.03
1.09
1.04
1.08
1.08
1.07
1.17
1.11
1.11
1.08
1.08
0.70
0.74
5.13
16.06
11.28
12.93
15.73
15.82
15.82
16.42
16.66
10.43
7.00
5.17
3.69
3.24
3.09
16.79
21.61
16.56
6.90
5.18
0.23
0.59
0.38
0.35
0.42
0.56
0.59
0.72
0.66
0.79
0.79
0.82
0.91
0.91
1.00
0.62
0.75
0.77
0.56
0.65
3.37
21.05
6.74
6.49
4.77
5.05
8.72
16.03
7.20
3.44
2.86
3.10
3.35
3.88
4.04
7.33
8.90
8.82
3.14
3.02
0.40
0.23
0.18
0.33
0.31
0.42
0.49
0.58
0.50
0.57
0.53
0.52
0.51
0.34
0.46
0.55
0.56
0.54
0.42
0.45
2.30
3.83
3.79
4.21
4.28
5.21
3.64
5.62
2.47
2.37
1.90
2.14
1.98
1.14
1.59
4.56
5.28
3.44
1.23
1.66
Average
0.93
15.41
1.00
11.27
0.65
6.57
0.45
3.13
101
Table 6
AAD% and maximum deviations in the calculated saturated liquid density for different EoSs with volume transition.
Components
SRK
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
AAD%
Max Dev.
N2
CO2
H2S
CH4
C2H6
C3H8
nC4H10
nC5H12
nC6H14
nC7H16
nC8H18
nC9H20
nC10H22
nC11H24
nC12H26
iC4H10
iC5H12
iC6H14
Benzene
Toluene
4.96
9.60
4.62
5.25
4.98
5.24
5.46
5.84
5.82
6.09
6.29
6.71
6.77
6.38
6.67
5.40
5.65
5.76
7.16
6.01
20.56
23.79
18.74
21.34
22.69
23.53
24.43
26.06
23.92
23.79
23.52
28.15
27.54
29.86
30.27
23.76
26.03
23.93
24.85
25.59
3.86
5.73
2.81
3.76
3.44
3.49
3.61
3.82
3.91
3.95
4.03
4.42
4.31
4.21
4.17
3.56
3.74
3.77
4.18
3.81
15.69
19.27
14.11
16.67
18.02
18.77
19.58
21.23
18.93
18.67
18.36
23.17
22.41
24.73
25.34
18.93
21.23
19.01
20.12
20.79
0.96
1.74
0.71
0.95
1.33
1.85
2.04
2.44
2.21
2.65
2.52
2.62
2.63
2.54
2.53
2.10
2.54
2.66
1.67
1.97
4.77
14.76
3.00
5.92
2.36
3.09
5.52
6.30
11.00
12.35
15.43
10.32
9.21
10.65
8.10
3.89
2.76
7.84
8.33
8.51
0.65
0.97
2.26
2.02
0.92
1.03
1.97
2.18
2.01
2.32
2.25
2.32
1.31
0.72
1.21
2.19
2.17
2.17
1.01
2.05
1.43
2.83
4.25
2.84
1.42
1.41
1.46
1.84
3.65
5.81
7.53
2.92
5.68
5.68
3.53
1.71
1.73
2.66
1.23
1.66
Average
6.03
24.62
3.93
19.75
2.03
7.71
1.69
3.06
PR
PC-SAFT
von Solms et al. [28] simplied the original PC-SAFT EoS by

assuming that all the segments in the mixture have the same mean
diameter d, which gives a mixture volume fraction identical to that
of the actual mixture. The mean diameter d is given by
0X
11=3
n
xi mi di
B
C
B i
C
dB X
C
@
xm A
i
(13)
~ hs and ghs terms to

This modication simplies the a
~ hs
a
4h 3h2
1 h2
(14)
ghs
Soave-BWR
1 h=2
1 h3
(15)
The simplied PC-SAFT EoS is identical to the original PC-SAFT EoS

for pure components and equally accurate for mixtures [28]. The
main advantage of the simplied version is that it somewhat
reduces the computation times for non-associating systems and
markedly for associating systems. The simplied version of PCSAFT is used in our calculation below.
2.2. The Soave-BWR equation of state
The BenedictWebbRubin (BWR) equation of state [29]
belongs to the so-called virial type equations of state. Despite
Fig. 4. AAD% in the calculated density for different EoSs within 279500 K and 1000 bar with volume transition.
102
Fig. 5. AAD% in the calculated saturated liquid density for different EoSs with volume transition.
Table 7
AAD (%) in the calculated high pressure liquid densities for heavy normal alkanes.
SRK
PR
PC-SAFT
Soave-BWR
nC7
nC9
nC11
nC13
nC17
nC20
Average
11.58
2.97
1.95
2.14
14.45
4.70
1.74
1.57
17.13
7.58
1.80
1.99
19.68
10.45
1.60
1.85
26.80
18.42
1.94
3.15
29.32
21.26
2.47
3.93
19.83
10.90
1.92
2.44
Fig. 6. Change of s for non n-alkanes with the relative difference in specic gravity SG. s 0 and SG0 are for the n-alkanes at the same boiling point temperatures.
103
Fig. 7. Change of e for non n-alkanes with the relative difference in specic gravity SG. e0 and SG0 are for the n-alkanes at the same boiling point temperatures.
its empirical nature, it provides a highly accurate density

description than many other types of EoS models. The original
BWR takes the following functional form:
z
P
1 Br C r2 Dr5 Er2 1 F r2 expF r2
RT r
(16)
where r is the density, and B, C, D, E and F are the ve model

parameters. A recent modication of the BWR EoS is given by Soave
in 1995 [30]:
z
P
1 Br C r2 Dr4 Er2 1 F r2 expF r2
RT r
(17)
where the exponent ve of the density in the original equation was

changed to four. In 1999, Soave [5] further simplied the equation
by dropping the C r2 term:
P
1 Br Dr4 Er2 1 F r2 expF r2
RT r
(18)
Although the number of terms in the equation is reduced, with

proper parameterization, the new version of Soave-BWR, i.e.,
Eq. (18), turned out to be better than the old version, i.e., Eq. (17).
The Soave-BWR EoS used here refers to the 1999 version.
There are four parameters in Soave-BWR, B, D, E and F. Their
values at the critical point are linked to a new set of notations
dened by
b BT c rc
(19)
d DT c r4c
(20)
Fig. 8. Change of m for non n-alkanes with the relative difference in specic gravity SG. m0 and SG0 are for the n-alkanes at the same boiling point temperatures.
104
Table 8
Deviations in bubble point pressures and vapor phase compositions predicted by SRK, PR, PC-SAFT and Soave-BWR (experimental data from [49,50]).
System
Temp. range
SRK
PR
PC-SAFT
Soave-BWR
COMP1
COMP2
Tmin (K)
Tmax (K)
DP/P (%)
DY1 (mol%)
DP/P (%)
DY1 (mol%)
DP/P (%)
DY1 (mol%)
DP/P (%)
DY1 (mol%)
C1
CO2
H2S
C2
C3
iC4
C4
iC5
C5
C6
C7
C8
C9
C10
C12
C14
C16
C20
C24
C32
C36
153
193
130
90
110
115
344
176
183
200
223
223
244
257
321
289
305
321
343
346
301
367
283
361
378
411
444
450
444
511
423
423
583
318
433
704
371
388
343
347
11.20
15.76
1.69
5.70
3.36
5.27
9.72
5.47
13.94
10.68
13.79
13.60
10.35
15.62
18.28
14.41
12.61
9.79
20.47
27.20
3.13
4.54
0.49
0.87
1.18
0.98
4.30
1.10
0.29
0.48
0.49
0.31
0.42
0.79
11.16
16.00
1.40
6.12
3.83
6.30
9.77
6.78
15.71
11.80
15.05
15.07
11.37
16.61
18.85
15.02
11.25
15.43
16.10
23.74
3.09
4.66
0.31
0.98
1.38
1.06
4.34
0.99
0.32
0.45
0.49
0.32
0.41
1.08
8.35
14.31
2.03
4.00
3.92
5.81
10.63
6.83
10.13
8.92
12.28
12.39
10.83
12.48
19.16
14.54
14.59
16.21
12.19
2.60
2.09
3.63
0.51
0.65
1.00
0.85
3.89
0.90
0.30
0.30
0.43
0.26
0.44
0.63
11.41
15.30
1.42
3.51
3.20
3.88
9.75
3.97
5.95
11.86
10.05
14.68
17.72
26.76
17.59
21.67
18.59
15.79
3.46
5.11
0.35
0.65
1.04
0.76
3.08
1.42
0.55
0.67
0.21
0.28
1.21
4.84
N2
CO2
H2S
C1
C2
C3
iC4
C4
iC5
C5
C6
C7
C8
C9
C10
C12
C16
218
228
78
111
120
255
250
278
277
311
305
322
344
311
344
463
301
344
184
290
353
394
411
377
378
444
497
344
344
411
367
703
3.97
17.53
6.97
9.60
12.33
17.04
14.41
14.70
14.47
21.66
17.91
26.68
27.52
19.02
21.15
16.32
0.71
6.05
1.76
1.43
1.41
2.67
2.18
2.67
1.66
1.61
4.54
0.11
1.06
2.74
17.51
7.48
6.85
13.19
18.88
14.70
16.85
16.51
22.55
18.05
27.06
28.50
22.28
23.44
16.11
0.56
6.24
2.05
1.68
1.48
3.18
2.29
3.10
2.02
1.73
5.08
0.13
1.13
3.50
15.73
8.12
7.87
13.03
16.83
14.85
20.01
18.73
24.47
17.81
27.39
24.53
24.15
19.41
1.50
7.25
1.96
1.91
1.77
3.83
2.44
4.74
2.96
1.83
5.48
0.17
1.06
5.42
17.43
7.63
8.03
12.57
11.14
9.58
5.57
7.14
10.71
7.09
7.77
4.48
15.39
10.41
13.49
1.45
5.67
1.99
1.32
1.46
1.91
2.58
1.34
1.01
1.49
1.86
0.09
3.41
CO2
H2S
C2
C3
iC4
C4
iC5
C5
C6
C7
C8
C9
C10
C12
C13
C14
C15
C16
C19
C20
C21
C22
C24
C28
C32
C36
225
207
233
273
228
278
253
273
311
216
343
236
254
255
269
270
463
293
300
302
315
373
348
336
373
366
302
361
398
419
453
463
393
502
466
343
594
267
339
311
305
664
304
373
338
373
573
423
573
423
11.71
15.46
15.37
13.71
14.88
16.27
12.15
15.70
17.44
15.30
20.02
19.36
10.54
11.62
12.61
0.55
16.06
0.13
10.96
10.24
22.93
9.17
12.95
11.78
7.76
3.96
6.09
4.07
3.30
3.49
3.19
3.49
1.44
2.41
1.23
1.00
1.79
1.81
0.07
0.02
12.34
15.83
15.24
13.61
14.75
16.31
12.27
15.69
17.64
15.38
19.28
18.75
10.77
11.95
14.26
1.13
14.62
8.67
13.78
13.81
21.48
7.74
11.59
10.95
9.18
4.01
6.08
3.96
3.14
3.28
3.12
3.52
1.28
2.19
1.09
0.84
1.61
1.68
0.12
0.03
12.99
14.49
15.01
12.62
13.96
15.76
12.00
17.06
16.84
15.29
17.77
18.47
7.78
13.03
15.07
2.82
16.64
3.24
3.52
3.80
29.67
18.39
23.94
21.17
23.41
4.19
5.67
3.82
2.76
3.23
2.74
3.06
1.47
2.16
1.36
1.48
1.81
1.43
0.14
0.03
12.54
15.72
14.02
9.79
12.48
12.16
10.39
14.53
7.90
11.63
8.77
7.73
5.40
4.96
2.81
3.50
11.99
5.51
7.09
7.19
8.99
11.84
12.44
15.66
27.59
3.95
5.95
3.53
1.99
2.32
1.82
2.01
0.72
0.68
0.40
0.08
0.49
1.05
0.01
0.00
H2S
C2
C3
C4
iC4
C5
iC5
C6
C7
C10
C15
200
217
367
278
278
323
323
311
278
423
283
367
418
398
444
413
423
478
444
423
12.54
10.78
5.97
6.36
12.04
8.94
12.40
15.25
11.16
8.35
5.77
4.54
2.19
2.44
1.87
3.13
1.50
1.80
0.17
0.12
13.10
10.81
5.73
6.02
11.81
9.03
11.56
14.90
9.27
7.93
6.06
4.41
2.06
2.31
1.89
3.00
1.46
1.62
0.27
0.23
13.35
10.92
5.49
6.11
11.40
8.94
11.84
13.25
13.91
9.58
5.97
4.13
2.46
2.19
1.81
3.05
1.20
0.96
0.21
0.04
12.97
8.93
3.75
2.79
4.13
4.20
2.79
3.64
14.42
6.09
3.89
1.29
1.56
1.13
1.86
0.43
0.88
0.18
13.08
2.08
13.48
2.11
13.26
2.12
10.11
1.79
Average
105
Table 9
Regressed kij values for the four EOS models and their deviations in calculated bubble point pressures and vapor phase compositions (experimental data from [49,50]).
System
SRK
PR
PC-SAFT
Soave-BWR
COMP1
COMP2
kij
DP/P (%)
DY1 (mol%)
kij
DP/P (%)
DY1 (mol%)
kij
DP/P (%)
DY1 (mol%)
kij
DP/P (%)
DY1 (mol%)
C1
CO2
H2S
C2
C3
iC4
C4
iC5
C5
C6
C7
C8
C9
C10
C12
C14
C16
C20
C24
C32
C36
0.0993
0.0879
0.0030
0.0180
0.0054
0.0100
0.0267
0.0213
0.0355
0.0280
0.0410
0.0449
0.0411
0.0442
0.0703
0.0586
0.0534
0.0298
0.0728
0.0868
1.99
4.09
1.55
3.77
3.06
4.26
8.84
2.36
3.77
7.45
3.97
2.68
4.29
7.11
1.43
5.34
13.00
7.20
2.96
8.87
0.78
1.80
0.45
0.76
1.13
0.89
4.02
1.00
0.20
0.43
0.36
0.24
0.31
0.65
0.0985
0.0859
0.0017
0.0188
0.0108
0.0168
0.0253
0.0270
0.0417
0.0335
0.0451
0.0485
0.0409
0.0500
0.0693
0.0561
0.0541
0.0500
0.0559
0.0763
1.94
4.29
1.35
2.77
2.75
3.78
8.94
2.24
3.90
7.54
4.15
3.03
4.30
6.88
1.35
5.39
13.61
12.83
2.86
7.25
0.75
1.69
0.32
0.79
1.28
0.87
4.07
0.81
0.17
0.36
0.35
0.28
0.26
0.86
0.0328
0.0411
0.0058
0.0034
0.0061
0.0041
0.0236
0.0121
0.0114
0.0103
0.0159
0.0151
0.0172
0.0103
0.0329
0.0189
0.0172
0.0223
0.0160
0.0040
2.31
3.95
1.47
4.30
2.90
5.27
8.71
3.28
5.58
6.56
5.93
4.85
6.51
5.88
4.96
4.11
3.69
3.99
3.25
1.07
1.30
1.66
0.39
0.70
0.93
0.80
3.39
0.77
0.17
0.29
0.35
0.29
0.33
0.79
0.0804
0.0692
0.0002
0.0024
0.0079
0.0044
0.0142
0.0076
0.0048
0.0149
0.0149
0.0184
0.0311
0.0315
0.0425
0.0441
0.0405
0.0110
0.0866
0.0400
1.97
3.93
1.41
2.99
2.62
4.19
8.72
3.05
5.71
9.25
5.17
6.64
7.06
12.22
5.89
9.60
11.42
12.80
12.89
0.84
1.66
0.36
0.65
1.01
0.78
3.16
1.25
0.43
0.58
0.26
0.24
0.82
1.51
N2
CO2
H2S
C1
C2
C3
iC4
C4
iC5
C5
C6
C7
C8
C9
C10
C12
C16
0.0196
0.1802
0.0307
0.0392
0.0838
0.1021
0.0974
0.0962
0.0934
0.1496
0.1402
0.1834
0.1896
0.1079
0.1612
0.2504
2.27
5.20
2.34
4.17
8.55
4.64
4.95
4.42
4.59
7.11
8.16
4.75
2.58
5.32
8.35
5.05
0.70
1.42
1.27
0.71
0.67
1.40
2.22
1.52
0.96
1.57
2.18
0.10
1.02
0.0067
0.1852
0.0326
0.0504
0.0944
0.1065
0.1037
0.1022
0.1000
0.1548
0.1419
0.1821
0.1885
0.1172
0.1654
0.2110
2.61
5.51
2.13
4.04
8.56
4.41
4.71
4.14
4.58
6.26
6.90
4.31
2.35
4.97
8.40
3.02
0.70
1.27
1.15
0.54
0.50
1.23
1.78
1.50
0.80
1.21
2.08
0.10
0.77
0.0046
0.1082
0.0258
0.0472
0.0669
0.0734
0.0711
0.0727
0.0665
0.0853
0.0751
0.0877
0.0852
0.0540
0.0683
0.0921
2.93
9.76
2.12
3.30
8.04
4.72
4.46
4.09
4.97
5.03
5.91
4.46
2.48
6.39
8.06
1.77
1.47
1.15
1.14
0.48
0.32
1.21
1.64
1.64
0.99
0.75
1.88
0.15
0.35
0.0163
0.1041
0.0246
0.0333
0.0447
0.0361
0.0341
0.0148
0.0145
0.0280
0.0086
0.0143
0.0048
0.0414
0.0180
0.0778
2.55
9.07
2.09
2.69
7.86
3.47
3.90
3.16
4.30
5.26
5.85
4.74
3.80
4.64
10.46
7.82
1.61
1.48
1.14
0.84
0.81
1.44
3.14
1.08
0.98
1.74
1.92
0.08
1.80
CO2
H2S
C2
C3
iC4
C4
iC5
C5
C6
C7
C8
C9
C10
C12
C13
C14
C15
C16
C19
C20
C21
C22
C24
C28
C32
C36
0.1014
0.1315
0.1443
0.1269
0.1395
0.1321
0.1274
0.1320
0.1096
0.1238
0.1151
0.1237
0.1019
0.0953
0.1086
0.1084
0.1178
0.1113
0.1051
0.1011
0.0936
0.0560
0.0421
0.0284
0.0145
1.25
2.46
3.12
2.62
3.13
3.15
4.37
4.59
4.93
6.31
1.86
4.72
1.50
3.23
1.51
0.79
4.08
2.19
3.12
3.35
5.70
5.66
9.10
9.79
6.88
0.99
2.17
0.83
1.05
1.20
1.26
1.45
0.56
1.23
0.12
0.11
0.58
0.53
0.06
0.02
0.0999
0.1272
0.1372
0.1181
0.1315
0.1219
0.1198
0.1236
0.0977
0.1155
0.1054
0.1115
0.0948
0.0967
0.0959
0.0979
0.1028
0.0978
0.0912
0.0875
0.0791
0.0333
0.0307
0.0154
0.0264
1.34
2.48
3.15
2.37
2.95
2.83
3.99
4.43
5.09
6.57
1.75
4.07
1.09
2.52
1.27
0.85
4.13
2.64
3.43
3.51
6.03
7.70
9.19
10.21
6.91
0.97
2.13
0.89
1.16
1.29
1.35
1.49
0.64
1.38
0.10
0.26
0.68
0.59
0.10
0.03
0.0627
0.0662
0.0751
0.0631
0.0672
0.0671
0.0665
0.0726
0.0527
0.0699
0.0600
0.0662
0.0687
0.0725
0.0718
0.0700
0.0599
0.0625
0.0612
0.0590
0.0561
0.0517
0.0497
0.0471
0.0424
1.16
2.76
2.98
2.23
2.79
1.65
3.60
3.63
6.21
6.90
1.96
4.25
0.58
3.21
3.58
1.27
5.51
4.50
4.87
4.80
4.82
3.36
3.83
4.33
2.90
0.87
2.47
0.55
0.89
1.08
1.01
1.01
0.61
1.11
0.15
0.09
0.55
0.24
0.06
0.05
0.0733
0.0937
0.0802
0.0602
0.0675
0.0550
0.0541
0.0538
0.0209
0.0389
0.0205
0.0177
0.0219
0.0179
0.0049
0.0103
0.0113
0.0071
0.0060
0.0086
0.0018
0.0293
0.0205
0.0305
0.0567
1.53
2.48
3.07
1.90
2.52
1.53
3.64
3.18
6.57
5.52
2.43
5.45
0.98
2.52
2.54
1.43
11.38
4.69
6.05
5.43
9.15
15.89
11.39
11.12
11.19
0.88
2.11
0.53
0.76
1.14
0.57
0.67
0.50
0.68
0.16
0.03
0.53
0.88
0.02
0.00
H2S
C2
C3
C4
iC4
C5
iC5
C6
C7
C10
C15
0.0870
0.0838
0.0979
0.0594
0.0692
0.0781
0.0691
0.0790
0.0447
0.0067
1.26
1.87
1.85
2.34
2.84
3.93
2.18
5.04
3.94
7.88
1.27
1.82
1.44
1.63
1.66
1.93
0.60
1.08
0.19
0.13
0.0834
0.0791
0.0894
0.0542
0.0625
0.0707
0.0599
0.0691
0.0352
0.0110
1.37
1.88
1.86
2.11
2.50
3.96
2.07
4.60
3.75
8.22
1.56
1.81
1.47
1.69
1.73
2.00
0.73
1.19
0.27
0.21
0.0581
0.0538
0.0504
0.0384
0.0454
0.0474
0.0404
0.0475
0.0387
0.0287
1.31
2.09
1.83
1.99
1.69
3.24
3.06
4.57
3.28
3.48
1.37
1.59
0.70
1.10
1.26
1.56
0.39
0.85
0.15
0.10
0.0605
0.0486
0.0364
0.0160
0.0152
0.0182
0.0016
0.0043
0.0336
0.0000
1.51
1.95
1.84
1.97
1.50
3.38
2.73
3.65
4.31
1.62
1.73
0.87
1.29
1.22
1.47
0.44
0.90
0.13
4.41
1.01
4.38
1.00
3.96
0.87
5.30
0.99
Average
106
Table 10
Statistics for the four EoS models in binary VLE calculation: sensitivities and average
kij.
SRK
Average sensitivity in DP/P (%)
Average sensitivity in DY1 (mol%)
jkij j
PR
PC-SAFT
m in Eqs. (25)(27) is a function of v and a simple proportionality

relationship is used for it:
Soave-BWR
m mv
with an empirical value 1.2 given to m. A linear mixing rule is used

for Zc:
109
1.4
115
12.6
231
22.5
161
17.2
0.086
0.082
0.048
0.031
(29)
X
xi Z ci T ci =Pci
i
Z cm X
xi T ci =Pci
e ET c r2c
(21)
f F r2c
(22)
The above four values b, d, e and f can be determined

from the
three
2
critical constraints Pc Z c rc RT c and @P=@rT c @ P=@r2
0,
Tc
plus an empirical constraint f = 0.77. The temperature dependence
of B, D, E, and F were determined by correlating partly synthetic
and partly experimental data, including vapor pressures for Tr
between 0.4 and 1 and v between 0 and 0.9, saturated liquid
densities for Tr greater than 0.4 and compressed liquid densities at
Pr = 10, and gas-phase densities of C1C4 alkanes. The resulting
temperature functions for B, D and E are expressed in terms of the
reduced temperature and the acentric factor, while F is treated as
temperature independent.
Soave developed the mixing rules for Tc, Pc and v based on the
mixing rules used for the classical CEoS models like SRK and PR. The
developed mixing rules calculate the mixture Tc, Pc and v in such a
way that the corresponding mixing parameters a and b in SRK or PR
are as close as possible to their values calculated by the van der
Waals mixing rules. The nal mixing rules for Tc, Pc and v are
S1
T cm p p2
S2 S3
Pcm
(23)
T cm
S3
(24)
s
S2
mm
S3
S1
(25)
XX
T ci T cj
xi xj 1 kij p1 mi 1 mj
Pci Pcj
i
j
s
XX
T ci T cj
S2
xi xj 1 kij
mm
Pci Pcj i j
i
j
S3
(26)
(27)
X T ci
xi
Pci
i
(28)
Table 11
Interaction parameters for CH4C7+, N2C7+, CO2C7+, and H2SC7+ for SRK, PR, PCSAFT and Soave-BWR.
EoS
N2C7+
CO2C7+
H2SC7+
CH4C7+
SRK
PR
PC-SAFT
Soave-BWR
0.1079
0.1172
0.0540
0.0414
0.1237
0.1115
0.0662
0.0103
0.0447
0.0352
0.0387
0.0000
0.0510
0.0522
0.0103
0.0441
(30)
In summary, there are four parameters for each component in

Soave-BWR, Tc, Pc, v and Zc. For n-alkanes, a good approximation of
Zc is given by
Z c 0:2908 0:099v 0:04v2
(31)
which is also used by Soave in generating the synthetic saturated

liquid densities [1,30]. Soave also pointed out that for some
compounds, it is important to use Zc as an independent parameter
to reproduce both vapor pressures and liquid densities accurately.
However, our experiences show that at least for phase equilibrium
of highly asymmetric binary pairs in reservoir uids, setting Zc as
an independent parameter does not offer much advantage, and in
fact sometimes adverse effects. Therefore, it is decided to use
Eq. (31) to calculate Zc, which reduces the number of model
parameters to just three.
3. Density calculation of pure components
We have selected 23 typical components in reservoir uids to
compare PC-SAFT, Soave-BWR, SRK and PR in density calculation.
These components are split into two groups:
Group I: N2, CO2, H2S, C1, C2, C3, nine n-alkanes from nC4 to nC12,
three iso-alkanes (iC4, iC5 and iC6), benzene and toluene.
Table 12
An overview of the phase envelopes tested in this study.
Gas
Source
C1
C6+
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
[52]
[52]
[52]
[52]
[52]
[53]
[53]
[53]
[53]
[53]
[55]
[55]
[55]
[55]
[55]
[55]
[56]
[56]
[57]
[57]
[57]
[57]
[57]
[58]
[59]
[60]
[60]
[61]
[62]
[62]
98.94
90.42
96.62
88.19
83.35
89.96
88.76
86.48
96.47
89.00
91.00
81.40
95.90
95.00
94.50
94.30
69.11
84.45
93.51
84.28
96.61
94.09
93.60
81.18
85.13
81.00
60.00
90.99
82.32
82.05
0.04
0.01
0.03
0.16
0.09
0.00
0.00
0.00
0.00
0.00
0.00
2.30
0.00
0.00
0.00
0.27
0.11
0.05
0.00
0.00
0.00
0.00
0.00
0.00
0.00
5.00
9.00
0.00
1.98
1.99
Remark
C1/C2/C4 ternary system

Large deviation in dew point branch

High CO2 concentration (26%)
N2/C1/C2/C3 four-component system

Group II: nC7, nC9, nC11, nC13, nC17 and nC20.For the components
in Group I, high accuracy reference EoS models [31] can be used to
generate smooth and evenly spaced synthetic density data. We
have compared these EoS models for uid density over a wide
temperature and pressure range and saturated liquid density.
The comparison in this group covers light as well as heavy
components. For the components in Group II,liquid density data up
to extremely high pressures [32] are available for model
evaluation. The components in this group are relatively heavier
than those in Group II whereas it should be noted that nC7, nC9 and
nC11 appear in both groups. The method for density comparison is
inspired by the study of de Hemptinne and Ungerer [33] where
they evaluated ve equations of state for the determination of
volumetric properties of pure hydrocarbon components. In
addition to using different models in the comparison here, we
also update the reference equations of state and include more
components. The saturated liquid densities are also included in the
comparison.
107
3.1. Comparison for Group I

For uid density, we have selected 273500 K and 02000 bar
as the temperature and pressure range of interest. This temperature and pressure range essentially covers most of the conditions in
reservoir processes. Table 1 summarizes the applicable ranges of
the reference EoS models used in this study. Since the applicable
ranges of some reference EoS models cannot cover the whole range
of 273500 K and 02000 bar, we distinguish between two types of
absolute average deviations in the comparison: AAD% is simply
calculated in 273500 K and 02000 bar without considering the
applicable ranges of the reference EoS models, and AAD0% is
calculated with the data points only within the applicable ranges of
the reference models. Our results indicate that AAD% and AAD0%
are very close. We only show AAD% below and the results for
AAD0% can be found in the supplementary material. For saturated
density, comparison is made in the reduced temperature (Tr) range
from 0.35 to 1.0. For some components, Tr = 0.35 is lower than the
Fig. 9. Phase envelopes for Gas 5 with (a) zero kijs and (b) the optimal kijs.
108
lowest applicable reduced temperatures (Tr,min) of their reference

EoS models (see Table 1). In such cases, we use Tr,min instead.
Table 2 and Fig. 1 show the deviations in uid density within
273500 K and 02000 bar calculated by different models without
volume translation. Both AAD% and maximum absolute deviation
(%) are reported in Table 2. On average, Soave-BWR and PC-SAFT
perform similarly, corresponding to a deviation around 1.5%.
Soave-BWR is slightly better than PC-SAFT. PR increases the
deviation to 6% and SRK gives the highest deviation around 9%. The
AAD% for different components are typically smaller than 2% for
Soave-BWR and PC-SAFT with Soave-BWR a bit better for
components not heavier than C2. The AAD% for SRK are relatively
small for components not heavier than C2, and actually comparable
to those for PC-SAFT. But the AAD% increases quickly for heavier
components, reaching around 10% for nC6 and around 20% for nC12.
On the other hand, PR gives larger deviations for light components
but smaller deviations for heavier ones than SRK.
For methane, we further show the contour maps of the

deviations in a wider range (150500 K and 02000 bar) in Fig. 2.
Soave-BWR gives the smallest average deviation and also very
small minimum and maximum deviations. PC-SAFT and SRK have
similar average deviations, which are slightly higher than that of
Soave-BWR. However, SRK gives much larger negative deviations,
i.e., under predicting the densities, at temperatures and pressures
close to the critical point. The low average density deviation of SRK
is largely attributed to its good performance for the methane gas
phase. Compared with SRK, PC-SAFT gives smaller maximum
values in density deviations. PR gives the worst performance for
methane among the four, mainly due to its poor density results for
the methane gas phase.
Table 3 and Fig. 3 show the deviations in saturated liquid
density calculated by different models without volume translation.
Soave-BWR and PC-SAFT again give around 1.5% in AAD%, much
smaller than 6% for PR and 12% for SRK. SRK is poorer than PR for
most components except for N2 and C1, where it is clearly better,

and for H2S and C2, where it is nearly the same.
Unlike PC-SAFT, both SRK and PR reproduce the experimental
critical points. If we relax this requirement and obtain their model
parameters by tting saturated density and vapor pressures, the
results can be improved. The Cubic-Plus-Association (CPA) EoS
[34,35] for non-associating compounds is actually SRK with model
parameters tted to vapor pressure and saturated liquid density
data. By use of the CPA model parameters from Ref. [36], we have
obtained the deviations for CPA as listed in Tables 2 and 3. CPA gives
much better results but the deviations, especially the maximum
absolute deviations, are still larger than those for non-cubic
models. This indicates that in addition to parameter determination,
the functional form of an EoS is another important factor for
density description. It should also be noted that PC-SAFT does not
reproduce experimental critical temperatures and pressures,
which leads to larger density deviations around critical points.
Actually, we observe that the maximum absolute deviations for PCSAFT usually appear around the critical temperature and pressure
109
for many components. However, Soave-BWR does not have such a

problem.
Volume translation can be used to improve density calculation
especially for liquid phases. The volume translation parameter is
normally treated as temperature independent [37] since temperature dependent volume translation can lead to negative heat
capacity at extreme high pressures [38] and sometimes crossing P
v isotherms [39]. Here, we adopt temperature dependent volume
translation in order to improve the density calculation results,
especially those for CEoSs, as much as we can. The temperature
dependent volume translation parameters are calculated at
different temperatures. At each temperature, the volume translation parameter is determined by matching the density from the
reference EoS at 1000 bar. In this manner, better density results at
high pressures can be expected. This temperature dependent
volume translation still suffers negative heat capacity at innite
pressures [38] but not necessarily the problem of crossing
isotherms [39]. Table 4 shows the trends in the volume translation
parameters for different components and models. If the volume
110
translation parameter is monotonically decreasing with temperature, the problem of crossing isotherms will not happen;
otherwise, it will. It can be seen that for SRK and PR, the trends
are ne for most hydrocarbons. CO2 and H2S give the problematic
trends for all the four models. For Soave-BWR and PC-SAFT, the
trends are problematic with many hydrocarbons. But on the other
hand, Soave-BWR and PC-SAFT already give very accurate density
results and the translation is not really necessary. It is possible to
force the volume translation parameter to follow a certain trend
while minimizing the density calculation deviation but no such
effort is made here.
The results with temperature dependent volume translation
are presented in Tables 5 and 6 and Figs. 4 and 5. For uid density
over a wide temperature and pressure range, the temperaturedependent volume translation greatly improves the results for
SRK and PR but the effect for Soave-BWR and PC-SAFT is modest,
which is not surprising since the room for further improvement is
very small. However, it should be noted that the maximum
absolute deviations for SRK and PR are much larger than those for
PC-SAFT and Soave-BWR, showing again the importance of the
functional form of an EoS. For saturated liquid density, the
improvement is obvious for SRK and modest for PR; for SoaveBWR and PC-SAFT, the results actually become slightly worse.
Compared to Soave-BWR and PC-SAFT, SRK and PR give larger
deviations in saturated liquid density even with temperature
dependent volume translation. This is because volume translation
simply shifts the Pv isotherm but does not change its slope. If the
improvement of density is focused on one region (high pressures
in this case), there is no guarantee that density in other regions can
be improved to a similar extent. Temperature dependent volume
translation can be designed to match the saturated liquid density
but this usually leads to the problem of crossing Pv isotherms
[39]. In short, if we want accurate density description over a wide
temperature and pressure range, Soave-BWR and PC-SAFT are
better than SRK an PR even if temperature dependent volume
translation is applied to the latter two.
3.2. Comparison for Group II

Table 7 summarizes the results for heavy normal alkanes. The
pressure ranges from 0 to 5000 bar and the temperature ranges
from 303 to 573 K (nC20 from 373 K). PC-SAFT almost always gives
the smallest deviations for those heavy n-alkanes. Soave-BWR is
the second best and gives just slightly higher deviations. SRK and
PR are much poorer, giving 20% and 11% deviations, respectively.
The comparison of density calculation for these four EoS models
shows that Soave-BWR and PC-SAFT have clear advantages over
the classical CEoSs for both light and heavy components in
reservoir uids. These two models have therefore great potential
for PVT modeling where high accuracy in density is required.
4. Heptanes plus characterization
Characterization of the heptanes plus (C7+) fraction is a
necessary step to apply any EoS to calculate the phase equilibrium
of petroleum reservoir uids. The input for characterization is the
111
limited information about the C7+ fraction, usually expressed in

terms of a number of cuts dened by boiling point (Tb) ranges. The
specic gravity SG and molecular weight MW for each cut or for the
whole C7+ fraction are typically the only additional information
known. The characterization procedure is to provide the necessary
information for EoS calculation by generating a detailed molar
composition for single carbon number (SCN) components in the C7
+ fraction, estimating the model parameters (usually critical
properties) of the SCN components, and grouping the many SCN
components into an appropriate number of pseudo components.
4.1. Characterization for SRK, PR and Soave-BWR
The method proposed by Pedersen et al. [40,41] is used here as
the framework for C7+ characterization. The method consists of
three steps: (1) calculating the mole fraction of SCN components in
the last plus fraction using exponential decay; (2) estimating the
critical properties Tc, Pc and v with a set of correlations developed
by themselves [40]; and (3) lumping the SCN components into a
112
few pseudo components with approximate equal mass, where the

lumped components properties are calculated as the mass average
of the SCN components properties. Another popular characterization method in PVT modeling is the method proposed by Whitson
et al. [42]. It shares the same three steps although the detailed
choices are different. A major difference is that Whitson et al. uses
the gamma distribution for the molar distribution in C7+. This
seems to be superior to the method of Pedersen et al. However, it
should be noted that the exponential distribution in the method of
Pedersen et al. is applied to the last plus fraction instead of C7+.
Since many oil analyses can provide composition information to
carbon number much higher than C7, e.g., the commonly employed
simulated distillation using the ASTM 2887 protocol can cover
boiling point range up to 1000 F or around C43, the molar
distribution estimated by the method of Pedersen et al. is generally
good enough unless the SCN components are over lumped. In this
study, the method of Pedersen et al. was employed for SRK, PR, and
Soave-BWR with just one modication: its correlations for Tc, Pc
and v are replaced by the Twu correlations [43] for Tc and Pc and
the LeeKesler/KeslerLee correlations [44,45] for v. The same set
of correlations is used in the method of Whitson et al. The reason
for not using the original correlations proposed by Pedersen et al.
[40], which are developed by tting PVT data directly, is that our
experiences show they are poor in reproducing the critical
properties of individual heavy hydrocarbon components in C7+.
The Twu correlations and the LeeKesler/KeslerLee correlations,
in contrast, give a much better estimate.
4.2. Characterization for PC-SAFT
Characterization for PC-SAFT also adopts the rst and the third
steps in the method of Pedersen et al. The second step, estimation
of model parameters, has to be replaced by a set of newly
developed correlations for the PC-SAFT model parameters m, s and
e. Correlations for the PC-SAFT model parameters for n-alkanes or
other hydrocarbons of a certain homologous series are available in
the literature [28,46]. Those correlations are usually expressed as

functions of molecular weight MW. However, estimating the PCSAFT model parameters for a SCN component is a bit different. A
SCN component is a narrow boiling point range cut without a
denite chemical composition or a xed molecular weight. The
component is characterized by its average boiling point Tb and
specic gravity SG. The variation in SG for a SCN component of a
certain Tb reects the variation in aromaticity. Higher SG indicates
a higher content of aromatic or cyclic hydrocarbons in the fraction.
The developed correlations should be able to estimate m, s and e
based on Tb and SG instead of MW.
To develop such a set of correlations, a database of m, s and e is
needed for relevant hydrocarbons. 358 hydrocarbons in the DIPPR
database [47], including 30 n-alkanes (up to nC36), are used for this
purpose. The values of m, s and e are regressed using the
synthetic experimental vapor pressure and saturated liquid
density data in the reduced temperature range from 0.5 to 0.9,
which are calculated using the correlations in DIPPR. The regressed
113
m, s and e for some compounds are excluded in the development of

correlations, including (1) the compounds which are too light, such
as methane; (2) the compounds without a vapor pressure
correlation or a liquid density correlation valid over the required
temperature range; and (3) the compounds whose parameters are
signicantly off the trends. The nal set of hydrocarbons includes
29 n-alkanes and 318 other hydrocarbons.
A two-step perturbation method, which is used by Twu [43] to
develop the correlations for critical properties, is employed here to
develop the correlations for the PC-SAFT model parameters. In the
rst step, the properties of the n-alkane at the Tb of the SCN
component are calculated. The molecular weight of the n-alkane at
this Tb, MW0, can be estimated using Twus correlations [43]. The
PC-SAFT model parameters for n-alkanes, m0, e0 and s 0, are then
calculated by the following linear correlations:
m0 0:02644MW0 0:83500
(32)
114
m0 e0
6:90845MW0 139:30870
k
(33)
m0 s 30 1:71638MW0 19:19189
(34)
The above equations differ slightly from those in [28,46] due to the
difference in the selected n-alkanes and temperature ranges. In the
second step, the properties of the SCN component is estimated by
using the difference in specic gravity SG SG0 as the perturbation
parameter, where SG0 is the specic gravity of the n-alkane
calculated by Soaves correlation [48]:

SG0 1:8T b 1=3 11:7372 3:336 103 T b 976:3T 1
b
1
3:257 105 T 2
b
(35)
Establishing simple but accurate correlations between the difference in m, e and s and the difference in SG is far from a simple task.
No general trends with sufciently small scattering can be found
between m m0 , e e0 =k, or s s 0 and DSG SG SG0 .
However, it seems that s =s 0 varies in a relatively narrow range,
as indicated in Fig. 6. Therefore, as a rst approximation, it is
assumed that:
s s0
(36)
By xing s equal to s 0, the values of m and e are regressed again

and the following equation found to give the best t for e as shown
in Fig. 7.
e e0 1:2357DSG 1
(37)
With Eqs. (36) and (37) as constraints, the m values for other
hydrocarbons in DIPPR are retted. The new m values are shown in
Fig. 8 and the following correlation is suggested:
m m0 1:3192DSG 1:8218DSG 1
2
(38)
Eqs. (37) and (38) provide the basis for determining the nal
correlations for e and m. In the nal step, the coefcients in these
two expressions are further tuned to t the vapor pressures and
saturated liquid densities for all the 318 other hydrocarbons.
Different weights have been assigned to different chemical classes
in these hydrocarbons: cyclohydrocarbons and aromatics are set to
5, iso-alkanes to 2, and other unsaturated hydrocarbons to 1. The
nal correlations are suggested for e and m:
e e0 1:1303391DSG 1
(39)
m m0 1:0460471DSG2 1:6209973DSG 1
(40)
and s is still calculated by Eq. (36). It should be noted that Eq. (40)
has a minimum value at DSG 0:7748204. To keep the monotonicity, the ratio of m/m0 will be xed at this minimum for DSG
larger than that. We can also assume that e and m are functions of
both DSG and SG and develop the following set of correlations:
e e0

115
0:59822690
0:55100489
DSG 1
SG
m0

1 2:6934054 0:57209122
DSG
SG
(41)
(42)
This set of correlations gives slightly better vapor pressures and

saturated liquid densities in the nal regression step. However, our
test shows its PVT calculation results are not as good as the former
simpler expressions. Therefore, only the correlations (39) and (40)
are used in the later PVT calculation.
5. Regression of binary interaction parameters
To apply PC-SAFT and Soave-BWR to reservoir uids, optimal
binary interaction parameters should be regressed. The major
components in reservoir uids include N2, CO2, H2S, and various
hydrocarbons. The interaction parameters kijs between two
116
hydrocarbon components can usually be set to zero unless one of

them is methane. The kijs between non-hydrocarbons and
hydrocarbons are usually not zero. In this section, regression of
kijs is focused on the pairs containing at least one of the following
four components: N2, CO2, H2S, and CH4.
The binary VLE data are taken from the DECHEMA Chemical
Data Series VI [49] and The Polish Academy of Science-TRC Floppy
Book Series [50]. Although optimal kijs for SRK and PR are already
regressed in DECHEMA, these values are regressed again with the
complete data from both DECHEMA and TRC series so that a
comparison with Soave-BWR and PC-SAFT can be easily made. It
should also be noted that kijs are treated as temperature
independent in the regression. In some advanced cubic models
[2025], temperature dependent kijs are used for a better phase
equilibrium description. However, we stick to temperature
independent ones here for an easy and fair comparison between
the cubic and the non-cubic models. The binary pairs studied here
are shown in Tables 8 and 9 and the temperature ranges for those
binary data can be found in Table 8. First, pure prediction

calculation with zero kijs is made and the results from SRK, PR, PCSAFT and Soave-BWR are compared in Table 8. Bubble point
pressure calculation is used and the comparison is therefore on the
calculated bubble point pressure and vapor phase composition.
The deviations shown in the table are absolute average deviations.
SRK, PR and PC-SAFT give similar deviations in the predicted
bubble point pressures and vapor phase compositions, while
Soave-BWR gives the smallest deviations in the predictive
calculation. Table 9 provides the regression results including the
regressed kijs. The regression is made by minimizing the
deviations in bubble point pressures, as suggested in DECHEMA
[49]. After regression, SRK and PR show similar average deviations
in DP/P (%) (around 4.4%) and in DY1 (mol%) (around 1%). PC-SAFT
gives slightly better results, around 4.0% in DP/P (%) and 0.9% in
DY1 (mol%). Soave-BWR gives a similar deviation (1%) in DY1 (mol
%) but the largest deviation (5.3%) in DP/P (%) among the four
models. The comparison shows that the four models are
comparable in binary VLE calculation. The non-cubic models do

not show clear advantages although PC-SAFT gives slightly lower
average deviations than SRK and PR. The model that gives the best
prediction, i.e., Soave-BWR, is not necessarily the one that can give
the best description after parameter tuning. In fact, it is slightly
worse than the others. It should also be noted that although the kij
value usually increases with the carbon number, as in the case for
SRK, PR and PC-SAFT, the trend does not seem to hold for SoaveBWR. For example, the kij for a CH4-hydrocarbon pair up to C3 is
almost zero for Soave-BWR but its value for iC4 and heavier
hydrocarbons becomes negative the value for C32 decreases to
0.087. This unusual trend will affect application of Soave-BWR to
oil systems since generalization of the kij with heavy hydrocarbon
can be difcult. To be fair, Soave-BWR has not been intensively
studied as the SAFT-family EoS models. Compared with the
BenedictWebbRubinStarling (BWRS) EoS [51] evaluated in the
DECHEMA [49], Soave-BWR is a dramatic improvement both in VLE
117
accuracy and simplicity. There is, however, still a room to further

improve this model.
Table 10 presents some statistics of the four models in the
binary VLE calculation. Two properties are investigated here, one is
the average kij and the other is the sensitivity of the VLE results w.r.
t. kij. The average kij is calculated by averaging the absolute values
of the optimal kijs for all the binary pairs. Its magnitude reects
how far the optimal interaction parameter is to the default value
(kij = 0). If a model has a smaller average kij, it means that only a
small adjustment is needed to tune it from its default state to its
optimal state. Among the four models, SRK and PR give similar
average kij values (around 0.08), PC-SAFT gives an obviously
smaller value (around 0.05) and Soave-BWR gives the smallest
(0.03). SRK and PR seem to be inferior to the other two models if
this kij criterion is used. And Soave-BWR seems to be the best since
the smallest tuning from its default value is needed. The sensitivity
w.r.t. kij is calculated as follows: if we change kij from its default
118
value (kij = 0) to its optimal value, the deviations in calculated

bubble point pressure and vapor phase composition will decrease.
The absolute values of these decreases divided by the change in kij
are dened as the sensitivities, i.e.,
Sensitivity in
DP=Poptimal DP=Pdefault
DP
j
j
P
kij;optimal kij;default
DY1optimal DY1default
j
Sensitivity in DY1 j
kij;optimal kij;default
(43)
(44)
In Table 10, the sensitivities for DP/P (%) and DY1 (mol%) averaged
for all the binary pairs are reported. For both bubble point pressure
and vapor phase composition, SRK and PR are less sensitive to kij
than PC-SAFT and Soave-BWR. According to the average sensitivities, if we change kij by 0.01, SRK and PR will give around 1% change
in DP/P (%), as compared with 2.3% for PC-SAFT and 1.6% for SoaveBWR. This indicates that PC-SAFT is roughly twice as sensitive as
SRK and PR, and Soave-BWR is around 50% more sensitive than SRK
and PR. It should be mentioned that the change of DP/P with kij is
not linear and its slope is in principle different at different kij
values. But these average sensitivities calculated using Eqs. (43)
and (44) still provide a rough sensitivity index for these models.
The relatively high sensitivities for PC-SAFT and Soave-BWR can
become a disadvantage when applying them to ill-dened uids
since a more sensitive model is more vulnerable to the uncertainty
in parameters, such as kij.
Finally, in order to apply the four models to PVT calculation, we
need to generalize the kij values for components above C7+. Table 11
gives the generalized kij values for N2C7+, CO2C7+, H2SC7+, and
CH4C7+. They are generated based on the optimal kij values with
heavy components up to C20 in Table 9. The kij values in Table 11 are
either calculated by averaging the kij values from C7 to C20 or set to
the value for a certain binary pair. There is some arbitrariness in
this process. In principle, we can tune the kij values within a certain
range through regression of PVT data. However, we would like to

limit parameter tuning as much as possible in this comparative
study and thus have not done that.
6. PVT modeling using PC-SAFT and Soave-BWR
Two types of tests are performed in this section. First, PC-SAFT
and Soave-BWR are applied to predicting phase envelopes of
synthetic gases. Then, the two non-cubic EoS models are applied to
predicting PVT calculation of reservoir uids containing ill-dened
plus fraction. Only the second test needs the characterization
described in Section 4. All the calculations here are predictive and
no parameter tuning to the experimental data is tried. For both
tests, comparison with SRK and PR is made.
6.1. Phase envelopes for synthetic gases
PC-SAFT and Soave-BWR are used to predict the phase
envelopes of 30 synthetic natural gases from eleven different
sources [5262]. Table 12 lists the basic information about those
gases. Only a few gases, such as Gases 26 and 27, have a signicant
amount of components heavier than n-pentane. A few gases are
ternary systems, including Gases 10, 25, 26 and 27.
Prediction calculations are made both with zero kijs and the
optimal kijs given in Section 5. Comparison with SRK and PR are
also made. Some selected results are presented in Figs. 920. In
each gure, the prediction with zero kijs is given in (a) and the one
with the optimal kijs in (b). From the calculation results, including
Figs. 920 and those not presented here, the following observations are obtained:
119
(6) Predictions for Gases 12, 15 and 16 from source [55] show very
large deviations in the dew point branches, probably due to big

errors in the experiment.
(7) Gases 26 and 27, measured by Urlic et al. [60], are highly
asymmetric ternary systems, maybe the most challenging ones
among the tested systems. None of the EoSs gives satisfactory
prediction over the whole temperature and pressure range
(Figs. 18 and 19). PR seems to give the best overall performance
for this system. Interestingly, for Gas 26, PR with zero kijs
seems to give a better resultusing non-zero kijs improves the
dew point prediction but signicantly overshoots the bubble
point branch. For this system, the four EoS models cannot be
divided into the two groups as mentioned in (4). Each EoS gives
its unique prediction, especially for the low molecular gas Gas
26. Compared with the other models, Soave-BWR gives a very
different prediction. If zero kijs are used, Soave-BWR predicts
much larger envelopes than the others. When non-zero kijs are
used, Soave-BWR predicts much atter envelopes than with
zero kijs but still large deviations in the dew point branches.
6.2. Reservoir uids

The PVT data for 35 reservoir uids from our internal PVT
database [64] and Jaubert et al.s work [65] are used in the test.
Table 13 gives an overview of these uids including their CH4 and
C7+ contents. The ability to predict the saturation pressures and
phase densities is checked and the results are summarized in
Table 14. The tested phase densities include the single phase
densities above the saturation pressure for both oils and gases, and
the liquid densities below the saturation pressure for oils.
(1) The prediction results are generally satisfactory for systems
(2)
(3)
(4)
(5)
containing components up to C5, both for zero kijs and the

optimal kijs. Using the optimal kijs gives very little differences
for systems with components only up to C5 (see Figs. 912, 14
17 and 20).
Gas 17 is the only system with a very high content of CO2
(26 mol%). For that system, using non-zero kijs is essential to
get the correct phase envelope (Fig. 13). PR seems to be a bit
better than the others for Gas 17 while it is also a matter how kij
is selected.
Predicting the bubble point branch is relatively easy, as can be
seen for Gas 24 in Fig. 16. Using non zero kijs can still improve
the prediction for this system. The difference in predictions
from different EoSs usually happens in the dew point branch.
The four EoS models seem to fall into two groups in terms of
their predicted phase envelopes: SRK, PC-SAFT and Soave-BWR
generally predict a bit larger phase envelope, whereas PR give a
bit smaller phase envelope (Figs. 911, 14, 15, 17 and 20). This
seems to be the case both for zero kijs and non-zero kijs. In
general, SRK, PC-SAFT, and Soave-BWR give predictions closer
to the experimental data than PR. This is in agreement with
Alfradique and Castiers observation that PC-SAFT gives closer
predictions of phase envelopes than PR [63]. But it should also
be noted that PC-SAFT seems to be similar to another CEoS
(SRK) in phase envelope prediction.
As shown in Figs. 10, 11 and 15, for all the four EoS models, large
deviations are observed in the dew point pressure branch
above the cricondenbar for experimental data from several
sources where the dew points were measured by the chilled
mirror method [5254,57]. The experimental measurements
tend to show larger envelopes than the predicted ones. It is not
clear whether it is due to some inherent deciency in the
thermodynamic models, or due to systematic errors in the
experiments.
Table 13
An overview of the reservoir uid systems tested.
Fluid no.
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
Fluid
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
C1 (%)
C7+ (%)
T range (K)
49
50
50
50
32
32
23
24
46
54
27
36
31
32
45
41
42
39
23
42
36
35
40
75
64
63
63
28
39
22
33
48
47
47
45
31
31
40
44
48
61
44
46
29
21
43
34
37
35
50
54
55
51
73
54
54
34
31
6
12
11
11
68
54
57
66
25
27
29
26
366.45
337.25
337.85
337.25
346.15
333.15
374.85
372.05
387.35
388.15
394.25
383.15
393.15
393.15
322.65
344.95
346.05
346.15
329.75
319.55
339.85
427.6
418.1
323.15469.15
423.15455.15
423.15460.15
323.15460.15
345.93
318.15
331.15
329.15
303.25417.85
394.25
373.75
299.85394.25
120
Table 14
Summary of AAD(%) in predicted saturation pressures and reservoir uid densities (single phase densities and liquid densities in the two-phase region).
Saturation pressure
Fluid 1
Fluid 2
Fluid 3
Fluid 4
Fluid 5
Fluid 6
Fluid 7
Fluid 8
Fluid 9
Fluid 10
Fluid 11
Fluid 12
Fluid 13
Fluid 14
Fluid 15
Fluid 16
Fluid 17
Fluid 18
Fluid 19
Fluid 20
Fluid 21
Fluid 22
Fluid 23
Fluid 24
Fluid 25
Fluid 26
Fluid 27
Fluid 28
Fluid 29
Fluid 30
Fluid 31
Fluid 32
Fluid 33
Fluid 34
Fluid 35
Average
Density
SRK
PR
PC-SAFT
Soave-BWR
SRK
PR
PC-SAFT
Soave-BWR
3.84
7.71
5.22
3.98
2.71
3.97
1.17
15.41
10.58
8.46
5.67
0.00
11.50
7.87
0.43
4.91
9.46
1.22
0.05
1.83
4.43
5.62
6.92
7.58
5.84
8.82
10.49
21.74
7.25
16.14
19.88
8.93
1.03
5.41
5.23
6.89
6.35
10.73
8.54
7.95
0.71
0.16
2.94
16.89
12.49
10.93
7.05
1.34
12.22
8.66
5.35
0.52
4.56
2.57
3.97
3.29
0.31
5.94
7.75
2.30
0.37
3.00
4.93
16.36
1.63
11.85
12.17
10.99
2.88
7.55
7.84
6.37
6.14
0.11
0.54
2.08
0.14
3.14
6.03
1.88
11.67
17.84
9.45
8.66
0.58
5.72
10.30
5.17
10.47
3.26
6.39
12.28
4.85
0.90
0.37
17.62
20.18
22.05
27.41
18.18
16.07
28.74
24.32
22.61
21.68
14.29
25.41
11.04
6.25
9.00
8.29
10.69
3.59
13.65
3.91
11.92
1.43
9.56
1.12
5.04
7.34
2.97
20.99
9.18
5.58
10.33
12.76
21.55
9.72
3.30
3.34
9.26
8.09
11.38
15.30
0.28
17.72
0.88
10.73
7.03
7.51
0.60
10.64
8.31
13.25
13.51
14.11
13.85
13.32
11.94
19.18
16.72
13.91
11.02
12.60
15.40
11.62
10.53
9.90
7.66
7.45
11.02
9.45
12.48
11.77
9.51
11.39
8.60
12.84
1.97
18.58
14.02
13.11
15.56
12.21
3.26
3.42
4.11
3.89
2.93
1.65
9.82
7.07
4.37
1.52
2.53
5.68
1.44
0.26
0.53
3.22
3.60
0.70
1.12
2.56
1.84
1.01
1.34
1.99
2.61
9.33
9.63
4.53
3.36
6.07
3.51
2.56
1.14
1.65
1.39
1.41
2.44
3.18
0.37
3.50
7.21
1.92
0.87
3.92
2.93
3.53
4.35
6.31
4.30
3.75
2.83
3.97
4.51
0.74
5.29
3.32
8.61
2.01
4.99
3.50
1.99
3.28
2.09
3.23
4.27
6.07
4.01
10.93
3.75
0.84
3.63
2.66
1.82
3.49
9.13
11.82
13.27
11.99
14.49
11.42
12.00
2.12
1.37
19.57
18.62
13.96
11.98
26.12
8.86
3.49
2.57
4.75
8.14
Figs. 21 and 22 show the results for two HPHT oils. For Fluid
22 at 427.6 K, PC-SAFT and Soave-BWR give the closest prediction
of the bubble point pressure while SRK and PR under predict the
bubble point pressure slightly. PC-SAFT, Soave-BWR and PR all give
the closest density prediction while SRK under predicts the
densities. For Fluid 23 at 418.1 K, the similar results are observed.
Soave-BWR and PC-SAFT again give the closest prediction of the
bubble point pressure. SRK and PR under predict the bubble point
pressure and SRK under predicts the liquid densities. Although PCSAFT and Soave-BWR gives excellent prediction for these two HPHT
oils, their predictions for other uids are not always better than
SRK and PR.
The most challenging systems for PC-SAFT and Soave-BWR are
the gas condensate systems, including Fluids 2427Fluid 27 is a
near critical system, showing gas condensate behavior at the high
temperatures and volatile oil behavior at the low temperatures.
PC-SAFT actually gives a pretty good prediction, similar to that of
PR, for the compressibility factors of Fluid 24 at high pressure
single phase regions (Fig. 23). However, the saturation pressures
predicted by PC-SAFT and Soave-BWR for those four uids are
actually poorer (Table 14 and Fig. 24(a)).
On Average, PC-SAFT and PR give the smallest deviations in
density, while PC-SAFT is slightly better than PR, and SRK is the
poorest in density prediction and undershoots the densities for
almost all the uids. The advantage of PC-SAFT in density modeling
seems to be somewhat kept with the newly developed characterization method. The densities tested here are mainly liquid
densities of oils and PR happens to predict reasonable liquid
density if the hydrocarbons are not too heavy (Table 14).
The accuracy in saturation pressure prediction is not signicantly different for PR and SRK. PR performs a bit better than the
other three while PC-SAFT seems to be the worst in saturation
pressure prediction. In contrast to the best performance for the
densities of well-dened components, Soave-BWR is much poorer
than PC-SAFT and PR for the liquid densities of reservoir uids,
although it is still better than SRK. The poor performance of SoaveBWR may be due to the fact that its density calculation is highly
Fig. 21. Liquid density for Fluid 22 in differential liberation at 427.6 K.
Fig. 22. Liquid density for Fluid 23 in differential liberation at 418.1 K.
dependent on the critical compressibility factors which are

difcult to estimate accurately for oils.
Finally, it is tried to improve the saturation pressure results of
PC-SAFT for Fluids 2427. As suggested by Soave [48], the normal
boiling point and the estimated boiling point at 10 mmHg are used
to tune the model parameters, Pc and v in the case of Soaves study,
121
and m and e in the case of PC-SAFT. The tuning is performed for

each SCN component in the second step of the characterization
method of Pedersen et al. For PC-SAFT, s estimated by Eq. (36) is
kept unchanged. This additional constraint just closes the system
of equations so that three model parameters can be determined.
Similar tuning is performed for SRK, PR and Soave-BWR for
comparison. For those three EoS models, Tc is xed, and Pc and v
are tuned to match the boiling points. The phase envelopes for
Fluid 24 with the new characterization are shown in Fig. 24(b).
Compared with Fig. 24(a), the prediction for PC-SAFT is improved;
however, the predictions for SRK and PR become poorer and that
for Soave-BWR becomes even worse. Similar trends can be found
for Fluids 2527. Tuning the parameters to t the boiling points is
theoretically sound although it has not improved the model
predictions for all the EoS models for these tested cases. In this
study, we emphasize comparison of cubic and non-cubic models in
PVT modeling without extensive tuning. In the future, other tuning
approaches including regression calculations should be further
investigated.
Table 15 summarizes the global deviations for the various
properties calculated by the four EoS models. The deviations with
differences smaller than 1% from each other are marked in bold
letters. They can be taken as equivalent calculations since the
results are similar. According to this summary, it appears that PR
has the best overall performance except for saturated liquid
density and liquid density for heavy n-alkanes. It should be noted
that the comparable AAD% in density with PR is obtained by use of
temperature dependent volume translation, and even in this case,
Fig. 23. Gas phase compressibility of Fluid 24 at different temperatures: (a) 323 K; (b) 373 K; and (c) 469 K.
122
Fig. 24. Phase envelopes for Fluid 24 using characterization (a) without matching the boiling points for SCN components and (b) with matching the boiling points for SCN
components.
Table 15
Summary of the deviations for the various properties studied in this work.
Properties
SRK
PR
PC-SAFT Soave-BWR
Pure compounds
Densitya (%)
0.93 1.00
Saturated liquid densitya (%)
6.03
3.93
Densityb (%)
8.95
6.30
Saturated liquid densityb (%)
11.90
5.73
Liquid density for heavy n-alkanesb (%) 19.83 10.90
0.65
2.03
1.67
1.31
1.92
0.45
1.69
1.33
1.56
2.44
Binary mixturesc
Bubble point pressure (%)
Vapour phase composition (mol%)
3.96
0.87
5.30
0.99
Reservoir uids
Saturation pressure (%)
Densityb (%)
a
b
c
4.41
1.01
6.89
12.21
Density calculated with volume translation.

Density calculalted without volume translation.
Calculated with the optimal kij values.
4.38
1.00
6.37 11.04
3.51
3.28
8.31
8.14
its maximum absolute deviations are still much higher than those
of PC-SAFT and Soave-BWR (see Table 5). PC-SAFT and Soave-BWR
are advantageous for density calculation, similar to SRK and PR for
binary VLE but poorer in saturation pressures of reservoir uids.
The power of cubic models lies in their simplicity and lower
computation time with respect to non-cubic models. PVT modeling
is the focus of this study and the computation speed issues will be
addressed in our subsequent paper.
7. Conclusions
A comprehensive evaluation of PC-SAFT and Soave-BWR, in
comparison with SRK and PR, has been made with regards to their
potential in PVT modeling of reservoir uids. Both PC-SAFT and
Soave-BWR are much better than SRK and PR in density calculation
of the well-dened light and heavy components in reservoir uids
over a wide temperature and pressure range, which forms a major
motivation for extending the two non-cubic models to reservoir

uids.
The characterization method of Pedersen et al. [40,41] is used
as the framework for characterization, with specic modications
for different EoS models: for PC-SAFT, new correlations for
estimating m, s and e have been developed; for Soave-BWR, SRK
and PR, Twus correlations for Tc and Pc, and the LeeKesler/Kesler
Lee correlations for v are selected. The optimal kij values for all the
four EoSs have been regressed from extensive binary VLE data
[49,50]. The binary VLE calculation also shows that PC-SAFT and
Soave-BWR are similar to SRK and PR in correlating the important
binary pairs in reservoir uids. Although Soave-BWR gives a better
prediction than the other three when zero kij values are used, its
regression results for highly asymmetric systems seem to be
poorer. Soave-BWR and PC-SAFT give smaller average kij values
than SKR and PR. On the other hand, they are more sensitive to the
change in kij.
Phase envelope prediction of synthetic gases shows that the
four EoS models are similar for not too asymmetric synthetic gases,
with or without the optimal kij values. SRK, PC-SAFT and SoaveBWR seems to be slightly better than PR. For highly asymmetric
synthetic gases [60], the predictions from the four models
somewhat differ and Soave-BWR is usually very different from
the others. Regarding the PVT modeling of reservoir uids with illdened plus fractions, PC-SAFT and Soave-BWR shows some
excellent predictions for HPHT oils but on average they are no
better than SRK and PR in predicting saturation pressures, and no
better than PR in predicting liquid densities. For gas condensate
systems, PC-SAFT and Soave-BWR give relatively poor predictions.
In summary, PC-SAFT and Soave-BWR have big potentials in
applications related to reservoir uids, including PVT modeling.
Compared with SRK and PR, their advantages seem to lie in better
density description rather than better VLE description. Obviously,
there are challenges to be solved for both non-cubic models. The
characterization method for PC-SAFT used in this study, including
the newly developed correlations for m, e, and s , follows the
framework widely used in the upstream industry. This has
provided a consistent basis in the comparison with SRK and PR.
But it is clear that the current characterization method for PC-SAFT
needs further improvement. The easiest way would be a new set of
parameter correlations and a more complex way could involve
modifying other steps in characterization. No matter which way is
used, a natural question would be whether the same methodology
can be applied for the competing CEoS models, such as SRK and PR,
to further improve their behavior. For Soave-BWR, its description
for highly asymmetric systems is not satisfactory which may partly
explain its different behavior from the other EoS models in some
PVT calculations. This requires more study on its mixing rule as
well as on how to parameterize the heavy components for this
model.
This study is focused on the general aspects of PVT modeling
and has not addressed several particular problems with non-cubic
models. In general, we need more precaution when using noncubic models. One potential problem with non-cubic models is
multiple liquid density roots at low temperatures. Such a problem
for PC-SAFT has been analyzed by Privat et al. [66] and Polishuk
et al. [67]. Polishuk et al. also pointed out that retting the
universal constants as performed by Liang et al. [68] could be a
solution to the problem. For Soave-BWR, we also noticed a similar
problem at very low temperatures. Another problem with PC-SAFT
is that the model does not reproduce the critical properties of pure
components exactly. This can cause deviations when the system
conditions are close to the critical point of a major component in
the system. This problem, however, does not exist for Soave-BWR
which reproduces the critical properties exactly.
123
Acknowledgement
This study was carried out under the NextOil project funded by
the Danish National Advanced Technology Foundation, Maersk Oil
and DONG Energy.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.uid.2014.11.022.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
G.S. Soave, Chem. Eng. Sci. 27 (1972) 11971203.

D.Y. Peng, D.B. Robinson, Ind. Eng. Chem. Fundam. 15 (1976) 5964.
J. Gross, G. Sadowski, Ind. Eng. Chem. Res. 40 (2001) 12441260.
J. Gross, G. Sadowski, Ind. Eng. Chem. Res. 41 (2002) 10841093.
G.S. Soave, Fluid Phase Equilib. 164 (1999) 157172.
D.L. Gonzalez, P.D. Ting, G.J. Hirasaki, W.G. Chapman, Energy Fuels 19 (2005)
12301234.
D.L. Gonzalez, G.J. Hirasaki, J. Creek, W.G. Chapman, Energy Fuels 21 (2007)
12311242.
D.L. Gonzalez, F.M. Vargas, G.J. Hirasaki, W.G. Chapman, Energy Fuels 22 (2008)
757762.
F.M. Vargas, D.L. Gonzalez, J.L. Creek, J. Wang, J. Buckley, G.J. Hirasaki, W.G.
Chapman, Energy Fuels 23 (2009) 11471154.
F.M. Vargas, D.L. Gonzalez, G.J. Hirasaki, W.G. Chapman, Energy Fuels 23 (2009)
11401146.
D. Gonzalez, A.K.M. Jamaluddin, T. Solbakken, G.J. Hirasaki, W.G. Chapman, SPE
110833 presented at the 2007 SPE ATCE held in Anaheim, Calibornia.
K. Pedersen, C.H. Srensen, SPE 110483 presented at the 2007 SPE ATCE held in
Anaheim, California, USA, 1114 November, 2007.
S.R. Panuganti, F.M. Vargas, D.L. Gonzalez, A.S. Kurup, W.G. Chapman, Fuel 93
(2012) 658669.
S. Punnapala, F.M. Vargas, Fuel 108 (2013) 417429.
S.R. Panuganti, M. Tavakkoli, F.M. Vargas, D.L. Gonzalez, W.G. Chapman, Fluid
Phase Equilib. 359 (2013) 216.
L. Sun, H. Zhao, C. McCabe, AIChE J. 53 (2007) 720731.
H. Zhao, P. Morgado, A. Gil-Villegas, C. McCabe, J. Phys. Chem. B 110 (2006)
2408324092.
K.S. Pedersen, S. Leekumjorn, K. Krejbjerg, J. Azeem, SPE 162346 presented at
the Abu Dhabi International Petroleum Exhibition and Conference, Abu Dhabi,
UAE, 1114 November, 2012.
S. Leekumjorn, K. Krejbjerg, Fluid Phase Equilib. 359 (2013) 1723.
J.N. Jaubert, F. Mutelet, Fluid Phase Equilib. 224 (2004) 285304.
S. Vitu, R. Privat, J.N. Jaubert, J. Supercrit. Fluids 45 (2008) 126.
J.N. Jaubert, R. Privat, F. Mutelet, AIChE J. 56 (2010) 32253235.
E. Voutsas, K. Magoulas, D. Tassios, Ind. Eng. Chem. Res. 43 (2004) 62386246.

[24] E. Voutsas, V. Louli, C. Boukouvalas, K. Magoulas, D. Tassios, Fluid Phase Equilib.
241 (2006) 216228.
[25] V. Louli, G. Pappa, C. Boukouvalas, S. Skouras, E. Solbraa, K.O. Christensen,
E. Voutsas, Fluid Phase Equilib. 334 (2012) 19.
[26] W.G. Chapman, K.E. Gubbins, G. Jackson, M. Radosz, Ind. Eng. Chem. Res. 29
(1990) 17091721.
[27] W.G. Chapman, G. Jackson, K.E. Gubbins, Mol. Phys. 65 (1988) 1057.
[28] N. von Solms, M.L. Michelsen, G.M. Kontogeorgis, Ind. Eng. Chem. Res. 42
(2003) 10981105.
[29] M. Benedict, G.B. Webb, L.C. Rubin, J. Chem. Phys. 8 (1940) 334.
[30] G.S. Soave, Ind. Eng. Chem. Res. 34 (1995) 39813994.
[31] E.W. Lemmon, M.L. Huber, M.O. McLinden, REFPROP Reference Fluid
Thermodynamic and Transport Properties. NIST Standard Reference Database
23, Version 9.0, 2010.
[32] A.K. Doolittle, J. Chem. Eng. Data 9 (1964) 275279.
[33] J.C. de Hemptinne, P. Ungerer, Fluid Phase Equilib. 106 (1995) 81109.
[34] G.M. Kontogeorgis, E.C. Voutsas, I.V. Yakoumis, D.P. Tassios, Ind. Eng. Chem.
Res. 35 (1996) 43104318.
[35] G.M. Kontogeorgis, I.V. Yakoumis, H. Meijer, E.M. Hendriks, T. Moorwood, Fluid
Phase Equilib. 158160 (1999) 201209.
[36] The Cubic-Plus-Association EoS: Parameters for Pure Compounds and
Interaction Parameters. Technical Report, Technical University of Denmark,
2012.
[37] A. Peneloux, R. Evelyne, Fluid Phase Equilib. 8 (1982) 723.
[38] P.H. Salim, M.A. Trebble, Fluid Phase Equilib. 65 (1991) 5971.
[39] O. Pfohl, Fluid Phase Equilib. 163 (1999) 157159.
[40] K.S. Pedersen, P. Thomassen, Aa. Fredenslund, Advances in thermodynamics,
in: L.G. Chorn, G.A. Mansoori (Eds.), C7+ Fraction Characterization, vol. 1, Taylor
& Francis, New York, 1989, pp. 137152.
[41] K.S. Pedersen, Aa. Fredenslund, P. Thomassen, Properties of Oils and Natural
Gases, Gulf Publishing Company, Huston, 1989.
124
[42] C.H. Whitson, T.F. Anderson, I. Sreide, Advances in thermodynamics, in:

L.G. Chorn, G.A. Mansoori (Eds.), C7+ Fraction Characterization, vol. 1, Taylor &
Francis, New York, 1989, pp. 3556.
[43] C.H. Twu, Fluid Phase Equilib. 16 (1984) 137150.
[44] B.I. Lee, M.G. Kesler, AIChE J. 21 (1975) 510527.
[45] M.G. Kesler, B.I. Lee, Hydrocarb. Process. 55 (1976) 153158.
[46] A. Tihic, G.M. Kontogeorgis, N. von Solms, M.L. Michelsen, Fluid Phase Equilib.
248 (2006) 2943.
[47] T.E. Daubert, R.P. Danner, Physical and Thermodynamic Properties of Pure
Compounds: Data Compilation, Hemisphere, New York, 2001.
[48] G. Soave, Fluid Phase Equilib. 143 (1998) 2939.
[49] H. Knapp, R. Doring, L. Oellrich, U. Plocker, J.M. Prausnitz, Chemical data series,
VaporLiquid Equilibria for Mixtures of Low Boiling Substances, vol. VI,
DECHEMA, Frankfurt, 1982.
ski, A. Skrzecz, TRC Data Bases for Chemistry and Engineering
[50] A. Maczyn
Floppy Book on VaporLiquid Equilibrium Data. Binary Systems Version 19981, Thermodynamics Data Center, Institute of Physical
Chemistry and Institute of Coal Chemistry of the Polish Academy of
Sciences.
[51] K.E. Starling, Fluid Thermodynamic Properties of Light Petroleum Systems,
Gulf Publishing Company, 1973.
[52] S. Avila, S.T. Blanco, I. Velasco, E. Rauzy, S. Otin, Ind. Eng. Chem. Res. 41 (2002)
37143721.
[53] S. Avila, S.T. Blanco, I. Velasco, E. Rauzy, S. Otin, Energy Fuels 16 (2002)
928934.
[54] S.T. Blanco, S. Avila, I. Velasco, E. Rauzy, S. Otn, Fluid Phase Equilib. 171 (2000)
233242.
[55] M. Gonzalez, A. Lee, J. Chem. Eng. Data 13 (1968) 172176.

[56] C. Jarne, S. Avila, S.T. Blanco, E. Rauzy, S. Otin, I. Velasco, Ind. Eng. Chem. Res. 43
(2004) 209217.
[57] . Mrch, Kh. Nasrifar, O. Bolland, E. Solbraa, A.O. Fredheim, L.H. Gjertsen, Fluid
Phase Equilib. 239 (2006) 138145.
[58] J. Oscarson, B. Saxey, Measurement of Total Fraction Condensed and Phase
Boundary for a Simulated Natural GasGas Processors Association (GPA)
Research Report RR-56, Brigham Young University, Utah, USA, 1982.
[59] J.S. Parikh, R.F. Bukacek, L. Graham, S. Leipziger, J. Chem. Eng. Data 29 (1984)
301303.
[60] L.E. Urlic, L.J. Florusse, E.J.M. Straver, S. Degrange, C.J. Peters, Transp. Porous
Media 52 (2003) 141157.
[61] J. Zhou, P. Patil, S. Ejaz, M. Atilhan, J. Holste, K. Hall, J. Chem. Thermodyn. 38
(2006) 14891494.
[62] F. Gozalpour, A. Danesh, A.C. Todd, D.H. Tehrani, B. Tohidi, Fluid Phase Equilib.
206 (2003) 95104.
[63] M.F. Alfradique, M. Castier, Calculation of phase equilibrium of natural gases
with the PengRobinson and PC-SAFT equations of state, Oil Gas Sci. Technol.
62 (2007) 707714.
[64] Internal PVT Database at Center for Energy Resources Engineering, Technical
University of Denmark, 2014.
[65] J.N. Jaubert, L. Avaullee, J.F. Souvay, J. Pet. Sci. Eng. 34 (2002) 65107.
[66] R. Privat, R. Gani, J.N. Jaubert, Fluid Phase Equilib. 295 (2010) 7692.
[67] I. Polishuk, R. Privat, J.N. Jaubert, Ind. Eng. Chem. Res. 52 (2013) 1387513885.
[68] X. Liang, B. Maribo-Mogensen, K. Thomsen, W. Yan, G.M. Kontogeorgis, Ind.
Eng. Chem. Res. 51 (2012) 1490314914.

Yan 2015

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Yan 2015

Transféré par

Droits d'auteur :

Formats disponibles

Fluid Phase Equilibria 386 (2015) 96124

Contents lists available at ScienceDirect

Fluid Phase Equilibria

reservoir simulation has much bigger uncertainties in its geology

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

accurate density description even around the critical point. It

Finally, it is obvious that non-cubic EoS models are less matured

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

~ id is the ideal gas contribution, a

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

and di is the temperature-dependent segment diameter of

~ disp 2prI1 h; mm2 es 3 prmC 1 h; mI2 h; mm2 e2 s 3

where m is the mean segment number in the mixture:

The hard-sphere term is given by the mixture version of the

with the conventional BerthelotLorentz combining rules for eij

where the integrals I1 and I2, and the compressibility expression C1

~ disp is modeled using a second order

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

von Solms et al. [28] simplied the original PC-SAFT EoS by

~ hs and ghs terms to

The simplied PC-SAFT EoS is identical to the original PC-SAFT EoS

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

its empirical nature, it provides a highly accurate density

where r is the density, and B, C, D, E and F are the ve model

where the exponent ve of the density in the original equation was

Although the number of terms in the equation is reduced, with

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

m in Eqs. (25)(27) is a function of v and a simple proportionality

with an empirical value 1.2 given to m. A linear mixing rule is used

The above four values b, d, e and f can be determined

In summary, there are four parameters for each component in

which is also used by Soave in generating the synthetic saturated

C1/C2/C4 ternary system

Large deviation in dew point branch

N2/C1/C2/C3 four-component system

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

3.1. Comparison for Group I

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

lowest applicable reduced temperatures (Tr,min) of their reference

For methane, we further show the contour maps of the

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

most components except for N2 and C1, where it is clearly better,

for many components. However, Soave-BWR does not have such a

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

3.2. Comparison for Group II

limited information about the C7+ fraction, usually expressed in

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

few pseudo components with approximate equal mass, where the

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

the literature [28,46]. Those correlations are usually expressed as

m, s and e for some compounds are excluded in the development of

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

By xing s equal to s 0, the values of m and e are regressed again

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

This set of correlations gives slightly better vapor pressures and

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124

hydrocarbon components can usually be set to zero unless one of

binary data can be found in Table 8. First, pure prediction

W. Yan et al. / Fluid Phase Equilibria 386 (2015) 96124