Reservoir fluid

Montadher kareem owayed

University of Baghdad

College of Engineering

Higher Diploma
Report about A Systematic and Consistent
Approach To Determine Binary
Interaction Coefficients for the
Peng-Robinson Equation of State
Summary:-
A new technique to determine binary interaction coefficients for a cubic equation of state (EOS) is
introduced. The technique is physically consistent and easy to apply when an EOS is tuned to
laboratory-measured PVT data. The method's simplicity provides a rational way of evaluating the
sensitivity of binary interaction coefficients to fluid composition and temperature. Laboratory-
measured expansion data are demonstrated to be crucial for proper determination of binary
interaction coefficients for near-critical fluids. The ability of the technique has been demonstrated
by matching laboratory-measured constant-composition-expansion curves for six fluid
compositions, including black-oil, volatile-oil, rich-gas-condensate, and lean-gas-condensate
systems. In addition, the near-critical fluid behavior has been evaluated by matching constant-
composition-expansion data for a fluid at four temperatures in the near-critical range, two on
either side of the critical temperature. The literature reports that the Peng-Robinson EOS (PR-EOS)
cannot predict the retrograde phase behavior of the rich-gas-condensate fluids applied in this
study. This paper shows that this is because insufficient and/or improper binary interaction
coefficients were used in the fluid characterization. The method enables tuning of near-critical
fluid behavior in addition to normal bubble point and dew point system behavior. Use of the
proposed procedure to assign binary interaction coefficients may allow matching of renascent
condensation for a near-critical bubble point system.

Introduction:-
Use of a cubic EOS to predict phase behavior became standard after the invention of the digital
computer. The EOS provides continuity and smoothness of PVT-dependent fluid properties. The
first cubic
EOS, expressing the continuity from gaseous to liquid states, was derived by van der Waals in
1873.' van der Waals' work formed the foundation for similar EOS's such as the Redlich-Kwong,2
the Soave-Redlich-Kwong,3 and the PR-EOS.4 None of the EOS's accurately predict the phase
behavior of a complex hydrocarbon mixture at reservoir conditions unless fine tuning of the EOS
parameters has taken place. 5,6 During the tuning process, the critical properties (Pc and Tc) of the
high-boiling-point fractions are normally adjusted within reasonable ranges to obtain a good
density match. In addition, binary interaction coefficients may be assigned to reflect the attractive
forces between light and
heavy molecules. Many correlations estimate the critical properties of the high-boiling-point
fractions in a hydrocarbon mixture. Whitson6 studied the effect of using various critical property
correlations in EOS predictions. A universally adequate method of determining binary interaction
coefficients has yet to be proposed. Katz and Firoozabadi 7 assigned interaction coefficients
between methane and heavier hydrocarbon components based on minimizing the error in the
calculated bubble point pressure of the binary mixture of methane with the heavier component in
question. In this fashion, a linear relationship was found between the methane binary interaction
coefficients and the density of the heavier components at standard conditions. As this study
shows, however, the binary interaction coefficients at reservoir pressures are strongly dependent
on fluid composition. Additionally, it seems that temperature has some effect as well. Hence,
binary interaction coefficients found on the basis of binary mixtures alone are not adequate when
dealing with complex multicomponent mixtures at elevated temperatures and pressures. The
latest innovation in EOS tuning is that of nonlinear regression, which minimizes the error between
EOS-calculated and laboratory observed data. 5 This procedure, however, often does not preserve
the physical significance of the regression parameters. The final calculated numerical values of the
regression parameters, even though they generally give a reasonable match to laboratory data,
appear to be independent of molecular weight. In addition, the computing time required for
nonlinear regression may be significant. Tuning an EOS for a near-critical fluid has generally been
considered difficult or even, in some cases, impossible. Firoozabadi et al. 8 tuned the PR-EOS to
match exactly the dewpoint for Kilgren's9 rich-gas-condensate laboratory data. In spite of this, the
PR-EOS predicted a bubblepoint system when the fluid was flashed. Whitson6 reported a similar
phenomenon when he used both the PR-EOS and Yarborough's version of the modified Zudkevich-
JoffeRedlich-Kwong lO EOS on the rich-gas-condensate data reported by Jacoby. 11 In this case,
the laboratory data indicated dewpoint systems, whereas the "tuned" PR-EOS predicted
bubblepoint
systems. This study shows that the apparent inability of the PREOS to predict the retrograde
behavior of the Jacoby gas condensate properJy6 is caused by the use of insufficient and/or
improper binary interaction coefficients. It is believed that the similar phenomenon observed by
Firoozabadi et al. 8 can be explained in the same way. The purposes of this paper are (1) to
propose a simple, systematic,
and physically consistent way to determine binary interaction coefficients for the PR-EOS; (2) to
demonstrate that the PR-EOS is capable of predicting the retrograde behavior of near-critical gas
condensate fluids; and (3) to examine the sensitivity of the binary interaction coefficients to
mixture composition and/or temperature. Jacoby et al." presented data on recombined samples of
separator oil and gas from a retrograde gas-condensate reservoir. The recombination was
performed for GOR's between 2,000 and 25,000 scf/STB [360 and 4460 std m3/stock-tank m3] and
represented eight related synthetic reservoir fluid mixtures (Fluids S-1 through S-8). Constant-
composition-expansion data were available for Fluids S-l through S-6; hence, only these six fluids
were examined in this study. Compositional data were available up to a normal boiling point of
475°F [246°C] corresponding to a C 13+ fraction. In addition, the liquid density at standard
conditions, 14.7 psia and 60°F [101 kPa and 15.5°C], and the molecular weight of each boiling
point fraction were given. Table 1 contains the available data for
the six fluids evaluated and adjusted acentric factors, pseudocritical pressures, and pseudocritical
temperatures for the high-boiling-point fractions F6 through F 13+
Method Description
The PR-EOS has the following form 4:
P=Pr+P
a, ....................................... (1)
where the repulsion pressure term is given by
Pr=RT/(V-b) .................................... (2)

and the attraction pressure term is given by

Pa = -al[V(V+b)+b(V-b)] . ........................ (3)
The parameter bin Eq. 2 is related to the molecular volume. The parameter a in Eq. 3 is related to
the energy of intermolecular interaction. Binary interaction coefficients are used to model the
intermolecular interaction through empirical adjustment of the a factor in the attraction pressure
term. Peng and Robinson4 found that the interaction coefficients were dependent on the
difference in molecular size of the components in a binary system. The method proposed in this
paper is to assign hydrocarbon/hydrocarbon binary interaction coefficients based on the following
hypotheses.
1. The interaction between two hydrocarbon components increases as the relative difference
between their molecular weights
increases-i.e. ,
kij <k;(j+ I) ................••..•.••.•............. (4)
and k(i+I)j<kij . .................................... (5)
2. Hydrocarbon components with the same molecular weight have
a binary interaction coefficient of zero-i.e.,
ki;=0. ........................................... (6)
3. The binary interaction coefficient matrix is symmetric-i.e.,
kij=kj; . .......................................... (7)
4. The interaction coefficient between ethane and the heaviest
fraction is proportional to the binary interaction coefficient between
methane and the heaviest fraction, etc.-i.e.,
k(i+ l)Fm+1 =Ri X kiFm+, ............................. (8)
where Ri >= 1. . ...................................... (9)
5. The binary interaction coefficient between Hydrocarbon Components i and j is a
function of the molecular weight of Component
i. j and Fm+ and the binary interaction coefficient between the light
hydrocarbon molecule and the heaviest fraction, k;Fm+ -i.e.,
kij=f(M;, Mj , M Fm +, kiFm+)' ...................... (10)
The methodology presented in this paper is based on a linear relationship between
kij and the nth power of Mj . Mathematically, the
above hypotheses are combined in the following equations 12 :
kij=mi(Mj)^n +d;, ................................ (II)
where i=C 1, C2 , C3 · .. , j=F6 , F7 , F8 … . .Fm+' n≠0, k(Cl +J)j
=kC2j '
mi=k;Fm+/[(MFm+)^n-(Mi)^n], ...................... (12)
and di = -(Mi)^n Xm; . ............................... (13)
Eq. 11 represents a straight-line plot of (Mj)n against kij on a linear basis. This is
shown schematically in Fig. 1, which illustrates the relationship between the various
binary interaction coefficients.
For n=O, the equation to determine interaction coefficients
becomes (to a first-order approximation)
kij =kiFm+ [In(Mj/Mi)/ln(MFm+1Mi)] . ................. (14)
The following describes the steps involved in the tuning process
of the binary interaction coefficients.
Step 1. Choose the numerical value of the exponent n.
Step 2. Choose the number of nonzero binary interaction coefficients.
Step 3. Choose the ratios Rcl' Rc2 , etc.
Step 4. Estimate the numerical value of the binary interaction
coefficient between methane and the heaviest fraction, kC1Fmt'
Step 5. Calculate the constants mi and di in Eq. 11 (for n*u).
Step 6. Calculate the remaining binary interaction coefficients
with the chosen function.
Step 7. Predict the saturation pressure by use of the EOS.
Step 8. Check that the predicted saturation pressure (bubblepoint
or dewpoint) matches the actual saturation pressure. If there is
agreement, then continue. If there is disagreement, then go back
to Step 4 and estimate a new value for the methane/heavy-fraction
binary interaction coefficient. A simple trial-and-error procedure
will make the process converge rapidly, with the saturation pressure
being matched to any required accuracy.
Step 9. Simulate the expansion experiment for which laboratory
data are available (constant-composition expansion, differential
liberation, constant-volume depletion).
Step 10. Check the agreement between the laboratory data and
the EOS prediction. If there is a reasonable agreement, then the
procedure is complete. If there is disagreement, then go back to
Step 1, 2, or 3 and choose a new binary interaction coefficient relationship.
The advantage of the proposed method is that it enables assignment of binary
interaction coefficients between an extensive
number' of components in a systematic and consistent way while varying only a few parameters,
such as (1) the value of the power n, (2) the number of nonzero binary interaction coefficients, and
(3) the numerical value of the ratios Rc1 , RC2 ....
The Appendix gives practical example of how to generate the binary interaction coefficients for
Fluid S-3.
Discussion of Results
Fig. 2 shows the PR-EOS constant-composition-expansion predictions for Synthetic Fluid Mixtures
Sol through S-6. The binary interaction coefficients have been assigned by use of the binary
functions and the Ri values between methane, ethane, propane, and
butane binaries as indicated in Fig. 2. It is evident that the binary
interaction coefficients are strongly dependent on the fluid composition. In this respect, it is
worthwhile to note the following.
1. For a black-oil system (Fluid S-l), the most essential binary
interaction coefficients are between methane and the highest-boiling point fractions. All other
binaries may be set to zero
2. For near-critical fluids (Fluids S-2 through S-4), it is crucial to include binary interaction
coefficients between more of the light components and all of the intermediate and heavy fractions.
3. For lean-gas condensates (Fluids S-5 and S-6), only methane binaries are of importance.
Fig. 3 illustrates that the binary interaction coefficients to some extent are temperature dependent.
A proper match of the laboratory data in this near-critical temperature range requires assignment
of
nonzero binary interaction coefficients between the light components (methane, ethane, propane,
isobutene, and normal butane) and the intermediate and heavier fractions. The temperature
dependence, however, does not seem to be quite as systematic as the compositional dependence.
The binary interaction coefficients that give the optimum match for the rich-gas condensate (Fluid
S-3) at the near-critical temperature, 163°F [73°C], are listed in Table 2. Figs. 4 through 6 show the
effect (on the rich-gas-condensate match) of varying either the power n, the number of binary
interaction coefficients with a numerical value of zero, or the ratios Rc1 , Rc2 , and Rc3 . For two
sets of binary interaction coefficients (Curve 3 on Fig. 5 and Curve 1 on Fig. 6), the PR-EOS predicts
renascent condensation. ("At
a constant temperature below the critical ... the quantity of the liquid phase decreases, then
increases in an abnormal manner, and thereafter decreases normally." 13) Note that in this case all
physical
conditions (e.g., temperature) and parameters (e.g., fluid composition and component critical
properties) are kept constant. The only variable is the relationship between the binary interaction
coefficients-i.e., the binary function. This means that the EOS predicted phase behavior of a fluid
can be altered significantly simply by varying the relationship between the binary interaction
coefficients.
Whitson 6
included
phase-
behavior
predictions for
the Jacoby
Synthetic Fluid
Mixtures S-1
through S-7.
Even though
numerous
attempts were
made to match
the
experimental
results,
Whitson pre-
dicted a critical fluid composition (GOR) of 5,000 to 7,500 scf/STB [890 to 1340 std m3/stock-tank
m3] and critical temperature of 260°F [127°C], which are considerably different from those
measured experimentally (critical GOR between 3,500 and 4,000 scf/STB [630 and 720 std
m3/stock-tank m3] and critical temperature between 132 and 163°F [56 and 73°C)). In his study,
Whitson apparently applied only one nonzero binary interaction coefficient (using a value for that
between methane and the C6+ fraction of 0.0536). It is evident from Figs. 4 through 6 that binary
interaction coefficients for the rich-gas condensate, Fluid S-3, at 163°F [73°C] can
be determined properly only if laboratory-measured expansion data are available. Figs. 4 through
6 also illustrate that the saturation pressure for a fluid may be matched exactly, independent of
the binary function used. Therefore, tuning the EOS to match measured properties at the
saturation pressure is not sufficient (at least in the case of near-critical fluids).

Conclusions
1. The binary interaction coefficients required in the PR-EOS for an acceptable match to
laboratory-measured data consistently are strongly sensitive to mixture composition.
2. The interaction coefficients also appear to be influenced by temperature. The temperature
dependence, however, seems to be less pronounced and more random than the compositional
dependence.
3. Use of improper and/or insufficient binary interaction coefficients in an EOS may result in the
prediction of a bubble point system even though laboratory measurements show a rich-gas-
condensate
system.
4. Laboratory-measured expansion data (Le., constant composition expansion, differential
liberation, or constant-volume depletion) are crucial to determine the proper binary interaction
coefficients for near-critical fluid conditions. Without expansion data, there is no control on the
EOS-predicted phase behavior during pressure depletion.
5. Tuning an EOS to match only the saturation pressure does not give an unambiguous
determination of the binary interaction coefficients. A large number of combinations of binary
interaction coefficients can be found to give a match of the saturation pressure.
6. A power function relating the binary interaction coefficients to the molecular weight of the
components was found to be an excellent and systematic way to determine binary interaction
coefficients for the PR-EOS.
Nomenclature
a = EOS parameter, psia(ft3/lbm mol)2
[kPa(m3/kmol)2]
b = EOS parameter, ft3/lbm mol [m3/kmol]
di = intercept constant in power function for
Component i, dimensionless
F6 = heavy fraction (corresponding to C6)
F7 = heavy fraction (corresponding to C7)
mi = slope constant in power function for Component i,
(Ibm mol/lbm)n [(kmol/kg)n]
kij = binary interaction coefficient between Components
i and j, dimensionless
MFm+ = molecular weight of the heaviest fraction Fm+,
Ibm/Ibm mol [kg/kmol]
Mi = molecular weight of Component i, Ibm/Ibm mol
[kg/kmol]
M
j = molecular weight of Component j, Ibmllbm mol
[kg/kmol]
n = exponent in power function, dimensionless
P = pressure, psia [kPa]
Pa = attraction pressure, psia [kPa]
Pc = critical (fi"essure, psia [kPa]
Pr = repulsion pressure, psia [kPa]
R = universal gas constant, 10.73 (psia-ft3)/(lbm molOR) [8.3145 (kPa·m3)/(kmol·K)]
RCI = ratio between kCZFm+ and kCIFm +' dimensionless
Ri = ratio between k(i+ I)Fm+ and kiFm+' dimensionless
T = absolute temperature, oR [K]
Tc = critical temperature, oR [K]
V = molar volume, ft 3/lbm mol [m 3/kmol]
VL = liquid volume, fraction
w = Pitzer's acentric factor
Acknowledgments
I thank Dansk Olie- & Gasproduktion A/S for their financial support, Scientific Software Intercomp
for providing their PVT program, and the Petroleum Engineering Dept. at the Colorado School of
Mines for the use of computing resources.
Appendix-Determination of Binary Interaction
Coefficients for Fluid 5-3 at 163°F [73°C]
Binary interaction coefficients for Fluid S-3 at 163 OF [73°C] were
determined according to the 10 steps outlined in the text. Physical
parameters are given for all heavy fraction components in Table 1.
Step 1. Numerical power value: n=5.
Step 2. Nonzero hydrocarbon binary interaction coefficients:
kij*O, where i=C 1, Cz, C3, iC4 , nC 4 andj=F6' F7 ··· .F13+.
Default PVT program values chosen for Nz and CO2 binary interaction coefficients (see Table 2).
Step 3. Ratios Ri : RCI =0.8, Rcz =0.8, and RC3 =0.8.
Step 4. Estimate methanelF13+ fraction binary interaction
coefficient: kCI - FJ3 + =0.2310. Hence,
kcz -FJ3+ = (0.8)(0.2310) =0.1848,
kC3 -FI3+ =(0.8)(0.1848)=0.1478,
k iC4 -FJ3+ =knC4'-FJ3+ =kC4 -FI3+'
and kC4 -Fl3+ =(0.8)(0.1478) =0.1183.
Step 5. Calculate mi and di :
mCI =0.2310/[(216)5 -(16.043)5]
=4.9130xlO- 13 (Ibm molllbm)5
[4.913xlO- 13 (kmollkg)5],
d
CI = -(16.043)5 x4.9130x 10- 13 = -5.2212 X 10-7 ,
mC2 =0.1848/[(216)5 -(30.070)5]
=3.9306x 10- 13 (Ibm molllbm)5
[3.93 x 10 -13 (kmol/kg)5],
d
C2 = -(30.070)5 x3.9306x 10- 13 = -9.6632 X 10-6 ,
mC3 =0.1478/[(216)5 -(44.097)5]
=3.1446xlO- 13 (Ibm molllbm)5
[3.145 x 10 -13 (kmol/kg)5],
d
C3 = -(44.097)5 x3.1446x 10 -13 = -5.2433 x 10 -5,
mC4 =0.1183/[(216)5 -(58.124)5]
=2.5196 x 10 -13 (Ibm molllbm)5
[2.52 x 10 -3 (kmollkg)5],
and d
C4 = -(58.124)5 x2.5196 x 10 -13 = -1.6715 X 10-4 .
Step 6. Calculate remaining binary interaction coefficients:
kCI -FlO =4.9130 x 10 -13(131)5 -5.2212 X 10-7 =0.0190,
kC2 -FlO =3.9306x 10- 13 (131)5 -9.6632 x 10-6 =0.0152,
kC3 -FlO =3.1446 X 10- 13 (131)5 -5.2433 X 10-5 =0.0121,
kiC4-FIO =2.5196x 10 -13(131)5 -1.6715 x 10 -4 =0.0096,
and knC4-FIO =2.5196x 10 -13(131)5 -1.6715 x 10 -4 =0.0096.
The remaining binary interaction coefficients are calculated in a
similar manner. The symmetric matrix is given in Table 2.
Step 7. Predict the saturation pressure with the PR-EOS:
PR-EOS: Pdew =3,940 psia [27.166 MPa]
Laboratory: Pdew=3,941 psia [27.173 MPa].
Step 8. Compare calculated and predicted saturation pressure:
excellent match. Continue.
Step 9. Simulate constant-composition expansion (see Table A-I).
Step 10. Compare PR-EOS expansion simulation and the
laboratory-measured data: reasonable agreement. The binary interaction coefficient matrix has
been determined properly.
51 Metric Conversion Factors
°F (°F-32)/1.8 =°C
Psi*6.894 757 E+00=kpa
In "A Systematic and Consistent Approach To Determine Binary Interaction Coefficients for the
Peng-Robinson Equation of State" (Nov. 1989 SPERE, Pages 488-94), Slot-Petersen proposes a
technique for determining binary interaction coefficients (BIC's) for the Peng-Robinson equation of
state (PR-EOS).1 He applies this technique to represent phase and volumetric behavior of the
Jacoby synthetic mixtures2 denoted Fluids S-l through S-6. The match to the measured data for
these mixtures obtained by Slot-Petersen is quite good, as shown in his Figs. 2 and 3. The fluid
characterization included a single set of C6+ pseudo component properties .used for all mixtures,
but different sets of BIC's for each specific mixture. For Fluid S-3, different sets of BIC's were
determined at each specific temperature. From these results, he concludes that the composition
dependence on BIC's is more important than the temperature dependence. The Jacoby synthetic
mixtures were recombined in various ratios of companion separator oil and gas samples produced
from a gas-condensate reservoir. The result is a family of seven related mixtures composed of the
same components. Therefore, the seven
mixtures should be characterized with a universal set of C6+ pseudo component properties and
BIc's.
Slot-Petersen's characterization is inconsistent from a practical
point of view. For example, consider a compositional reservoir
simulation study of a reservoir exhibiting a compositional
gradient-e.g., with Fluid S-6 at the top and Fluid S-l at the bottom. It would be impossible to model
such a continuous compositional variation with the different EOS characterizations proposed

by Slot-Petersen. All mixtures must be characterized with a single

set of pseudo component properties and BIC's.
S~reide3 obtained a consistent PR-EOS characterization for
Fluids Sol through S-6. The hexanes-plus (C6+) fraction was split
into a pseudocomponent lumped from several pure hexane components, denoted C6s,4 and five
C7+ pseudocomponents on the basis of the C7+ characterization methods proposed by Whitson et
al. 5 and S~reide. 3 C7+ pseudocomponent properties were slightly modified to match measured
data simultaneously using an efficient nonlinear regression technique. Plus-fraction compositions
for
all mixtures are given in Table D-l, and the final C7+ characterization is shown in Table D-2. Non
hydrocarbon hydrocarbon BIC's
and pure-component volume translation (VT) parameters are given in Table D-3.
Good agreement between measured and calculated saturation pressures is obtained for all
mixtures when the PR-EOS characterization is used, as Table D-4 shows. Calculated liquid volume
fractions for Fluids S-l through S-6 at l63.5°F [73.1 0c] are compared with measured data in Fig. D-l.
The obtained match is excellent for
Fluids S-l through S-4, with a slight underestimation for Fluids
S-5 and S-6. Similar results are shown in Fig. D-2 for Fluid S-3
at four temperatures between 96.5 and 207.3of [35.8 and 97.40c].
The critical temperature is estimated somewhat high, with liquid
volumes overestimated at 96.5 and 131°F [35.8 and 55°C], but
the match is good considering the wide range in composition, pressure, and temperature. Unlike
the Slot-Petersen characterization,
this characterization is consistent and could be applied easily with
an EOS-based compositional reservoir simulator.

Nomenclature
b = repulsive EOS constant
c = VT parameter
M = molecular weight, Ibm/mol [g/mol]
Pc = critical pressure, psia [kPa]
Ps = saturation pressure, psia [kPa]
s = c/b, dimensionless VT parameter
Tb = normal boiling point, OR [K]
Tc = critical temperature, OR [K]
'Y = specific gravity (water=1)
w = acentric factor
$1 Metric Conversion Factors

°F (°F-32/1.8) =°C
Ibm*4.535 924 =kg
Psi *6.894 757 =Kpa
°R °R/1.8 =K
S~reide and Whitson contend that my characterization of the Jacoby synthetic fluid mixtures 1 is
inconsistent. They also state that
S~reide2 obtained a consistent Peng-Robinson equation of state
(PR-EOS)3 characterization for the six mixtures.
Consistency is defmed as keeping to the same principles. 4 In my
paper, binary interaction coefficients (BIC's) were determined on
the basis of specific hypotheses about intermolecular interaction. The sets ofBIe's given in my
paper are all compatible and in agreement with these governing principles; hence, the procedure
is fully consistent.
The Jacoby synthetic mixtures are recombined from companion separator oil and gas samples.
S~reide and Whitson argue that characterization of all the fluid mixtures should therefore use a
universal set of heavy-fraction pseudo-component properties and BIC's to have a consistent fluid
characterization.
I agree with S~reide and Whitson that a universal set of heavy fraction critical properties should be
used under these circumstances. This was actually done (see Table 1 of my paper). In fact, the
possibility of using a universal critical property characterization for all mixture compositions was
one of the main reasons the Jacoby mixtures were used. They form a representative fluid system
for evaluating possible composition and/or temperature effects of the BIC's. Another main reason
for using the Jacoby synthetic fluid mixtures was that the reported laboratory data included a full
heavy end fractional distillation analysis. A perturbation technique minimizing the necessary
adjustment of the critical properties calculated from correlations for the high-boiling-point
fractions was used to fine tune the EOS to predict confidently the high-boiling point laboratory-
measured specific gravities. 5 BIC's are used in cubic EOS's to model intermolecular interaction.
Fig. R-I is a representation of the relative molecular distribution in three Jacoby synthetic fluid
systems, Fluids S-1 (black oil), S-3 (rich gas condensate), and S-6 (lean gas condensate). In these
systems, all individual-component physical properties are the same. The only difference is the
overall composition-Le., the relative distribution between light, intermediate, and heavy
components.

Because of interference from the surrounding molecules, the intermolecular interaction between
specific heavy and light molecules
(as modeled by a BIC) is not
necessarily the same in these
three
systems. Therefore, from a
physical point of view, BIC's are
not
dictated to be independent of
fluid composition
Fig. R-2 shows the different BIC relationships between methane
and the high-boiling-point fractions I used to match the constant composition expansion (CCE) of
the six mixtures (Fluids S-l through S-6) at 163°F [72.8°C]. The compositional dependence seems
obvious as discussed in my paper. Note that the methane BIC's for the near-critical Fluids S-2
(volatile oil) and S-3 (rich gas condensate) are identical. The only difference between the BIC's for
these two fluid mixtures is that the ethane, propane, and butane BIC's used to characterize Fluid S-
2 are reduced by some 25% compared with those for Fluid S-3 (see Fig. 2). Similarly, Fig. R-3 shows
the methane BIC relationships used to match the CCE of Fluid S-3 at different temperatures. In this
case, the BlC's seem more closely related because they fall within a fairly narrow range and follow
a similar trend. Again, slight differences exist in the ethane, propane, and butane binaries (see Fig.
3 of my paper). S~reide and Whitson show that S~reide2 obtained a PR-EOS characterization of the
six mixtures with a universal set of heavy fraction pseudo component properties and a single set of
BIC's. For comparison, these BIC's are shown in Fig. R-2. The CCE matches obtained by S~reide and
Whitson are shown in their Figs. D-l and D-2. Table R-l compares my calculated saturation
pressures with those obtained by S~reide.
My characterization of the Jacoby synthetic mixtures differs from S~reide's in several ways. First,
when characterizing the high boiling-point fractions (corresponding to C6 through C 13+), I paid
due attention to the measured fractional distillation analysis and had confidence in the given
compositional split between the distillation cuts and their respective measured molecular weights
and specific
gravities. Hence, I did not perform any artificial lumping and/or splitting of the eight heavy
fractions. Second, I used a perturbation technique to minimize the required adjustment of the
critical properties for the PR-EOS to predict liquid densities at standard conditions properly before
evaluating the influence of the BIC's on predicted phase and volumetric behavior. Third, my fluid
characterization was performed manually. S~reide2 characterized the fluids by (1) using a PR-EOS
with volume translation, (2) lumping and splitting the eight C6+ laboratory-measured high-boiling-
point fractions into six heavy pseudo components by use of a statistical molar distribution
function, and (3) using nonlinear regression.
The lumping and splitting technique applied by S~reide results in a molar distribution of the heavy
ends different from that obtained from measured fractional distillation analysis (see Table 1 and
Table D-l). S~reide first lumped the seven distillation cuts with a normal boiling point greater than
200°F [93.3 0c] into an overall C7+ fraction. Next, a probabilistic gamma distribution function
was used to model the molar distribution of the lumped C7+ fraction as a continuous function of
molecular weight. The gamma distribution parameters IX (characterizing the form of the
distribution function) and 11 (minimum molecular weight of any component within the lumped
C7+ fraction) were determined as 1.518 and 90, respectively. Finally, the continuous gamma
distribution function was discretized into five pseudo components (see Table D-2) by a generalized
discretization method that used Gaussian quadrature Methane high-boiling-point BIC's were
determined from a modified Chueh-Prausnitz correlation. After S~ide had gone through all these
steps, the EOS predicted a too-high mixture critical temperature, as indicated by bubblepoint
predictions instead of dew points for Fluid S-3 at 163.5 and 207.3of [73.1 and 97.4°C] (see Table R-
l). S~reide then used nonlinear regression in an attempt to force fit to measured data the EOS
predicted saturation pressures and CCE liquid volume fractions for· all six fluid mixtures
simultaneously. Regression parameters consisted of heavy-pseudo component critical and boiling-
point properties and methane BIC's. In addition, S~reide manually adjusted the gamma distribution
parameter ex for Fluids S-3, S-5, and S-6 (by three different factors-i.e., 0.95, 1.01, and 1.05) to
"match experimental phase and volumetric data... 2 I believe that by handling the molar
distribution in this way, S~reide introduces inconsistency in his fluid characterization. S~reide and
Whitson state that it is impossible to apply my fluid characterization in a compositional reservoir
simulator to model a reservoir with a compositional gradient. It is true that simulation of reservoirs
exhibiting a compositional gradient may be impractical with the currently available compositional
reservoir simulator models and composition-dependent BIC's. This deficiency might be overcome if
BIC relationships could be related to composition through the mole fraction of a controlling
component-e.g., methane concentration. Nevertheless, I disagree with S~eide and Whitson's
conclusion that BIC's must be independent of fluid composition because of the limitations of
currently available computer models. I obtained extraordinarily good matches by my
characterization technique and documented in my paper that the apparent inability of the PR-EOS
to predict the retrograde behavior of the Jacoby fluid mixtures properly reported by Whitson6
could be a result of inappropriate BIC's in his characterization at that time.264
S~reide and Whitson's present characterization, even after nonlinear regression, still has problems
with the CCE liquid dropout for Fluid S-3 at different temperatures, and also for Fluids S-5 and
8-6 at 163.soF. S~reide and Whitson used compositionally independent BIC's to obtain a fair match
after some mathematical manipulation, including nonlinear regression and alteration and
extrapolation ofthe molar distribution of the high-boiling-point cuts. I obtained an excellent match
through a relatively simple manual tuning process with compositionally dependent BIC's and
without altering composition, specific gravities, and molecular weights found from the trueboiling-
point distillation.
Reference:-
1. Jacoby, R.H., Koeller, R.C., and Berry, V.J.: "Effect of Composition
and Temperature on Phase Behavior and Depletion Performance of Rich
Gas-Condensate Systems," Trans., AIME (1959) 216, 406-11.
2. S~reide, I.: "Improved Phase Behavior Predictions of Petroleum Reservoir Fluids From a Cubic
Equation of State," PhD dissertation, Norwegian Inst. of Technology, Trondheim (1989).
3. Peng, D. and Robinson, D.B.: "A New Two-Constant Equation of
State," Ind. & Eng. Chern. Fund. (1976) IS, No. I, 59-64.
4. Hornby, A.S.: "Oxford Advanced Leamer's Dictionary of Current English," Oxford U. Press, New
York City (1974).
5. Slot-Petersen, C.: "A Comparison of the Recoveries Obtainable by Depletion, Gas Recycling and
Waterflooding of a Retrograde Gas Condensate Reservoir," MS thesis, Colorado School of Mines,
Golden (1986).
6. Whitson, C.H.: "Effect of C7+ Properties on Equation-of-State Predictions," SPE/ (Dec. 1984) 685-
96.
7. Slot-Petersen, C.: "A Systematic and Consistent Approach To Determine Binary Interaction
Coefficients for the Peng-Robinson Equation of
State," SPERE (Nov. 1989) 488-94.
References
I. van der Waals, J.D.: "On the Continuity of the Liquid and Gaseous
State," PhD dissertation, Sigthoff, Leiden, Holland (1873).
2. Redlich, O. and Kwong, J.N.S.: "On the Thennodynamics of Solutions.
V. An Equation of State. Fugacities of Gaseous Solutions," Chem. Rev.
(1949) 44, 233-44.
3. Soave, G.: "Equilibrium Constants From a Modified Redlich-Kwong
Equation of State," Chem. Eng. Sci. (1972) 27, No.6, 1197-1203.
4. Peng, D. and Robinson, D.B.: "A New Two-Constant Equation of
State," Ind. & Eng. Chem. Fundamentals (1976) 15, No.1, 59-64.
5. Coats, K.H. and Smart, G.T.: "Application ofa Regression-Based EOS
PVT Program to Laboratory Data," SPERE (May 1986) 277-99; Trans.,
AIME,281.
6. Whitson, C.H.: "Effect of C 7+ Properties on Equation-of-State Predictions," SPEl (Dec. 1984)
685-96.
7. Katz, D.L. and Firoozabadi, A.: "Predicting Phase Behavior of Condensate/Crude-Oil Systems
Using Methane Interaction Coefficients,"
JPT (Nov. 1978) 1649-55; Trans., AIME, 265.
8. Firoozabadi, A., Hekim, Y., and Katz, D.L.: "Reservoir Depletion
Calculations for Gas Condensates Using Extended Analyses in the PengRobinson Equation of
State," Cdn. J. Chem. Eng. (Oct. 1978) 56,
610-15.
9. Kilgren, K.H.: "Phase Behavior of a High-Pressure Condensate
Reservoir Fluid," JPT(Aug. 1966) 1001-05; Trans., AIME, 237.
10. Yarborough, L.: "Application of a Generalized Equation of State to
Petroleum Reservoir Fluids," Proc. , ACS Symposium on Equation of
State in Engineering and Research, K.C. Chao and R.L. Robinson Jr.
(eds.), American Chemical Soc., Washington, DC (1979) 385-435, Ser.
182.
II. Jacoby, R.H., Koeller, R.C., and Berry, V.J.: "Effect of Composition
and Temperature on Phase Behavior and Depletion Perfonnance of Rich
Gas-Condensate Systems," Trans., AIME (1959) 216, 406-11.
12. Slot-Petersen, C.: "A Comparison of the Recoveries Obtainable by
Depletion, Gas Recycling and Waterflooding of a Retrograde Gas Condensate Reservoir," MS
thesis, Colorado School of Mines, Golden
(1986).
13. Katz, D.L-etal.: "Handbook of Natural Gas Engineering," McGrawHill Book Co. Inc., New York
City (1959).