Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

SPE 64520

Plus Fraction Characterization and PVT Data Regression for Reservoir Fluids near
Critical Conditions
Julian Y. Zuo, SPE, and Dan Zhang, DB Robinson Research Ltd.

Copyright 2000, Society of Petroleum Engineers Inc.


are then chosen as adjustable parameters to fit the
This paper was prepared for presentation at the SPE Asia Pacific Oil and Gas Conference and experimental PVT data. The lower and upper bounds can be
Exhibition held in Brisbane, Australia, 16–18 October 2000.
set for each adjustable parameter. Finally, regressions are
This paper was selected for presentation by an SPE Program Committee following review of performed to match the measured saturation pressure,
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to volumetric and viscosity data of reservoir fluids separately or
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at simultaneously.
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
The developed characterization and regression techniques
for commercial purposes without the written consent of the Society of Petroleum Engineers is have been tested for oils and gas condensates near critical
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous conditions and under ordinary conditions as well as for
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. multiple samples that taken from a reservoir with a strongly
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
compositional gradient. The results show that the implemented
methods provide a reliable and efficient tool for representing
Abstract PVT properties and phase behavior of reservoir fluids.
PVT information of reservoir fluids is of great importance to
reservoir engineers. In order to describe phase behavior and Introduction
PVT properties using equations of state (EOS), a Reservoir fluids are composed of thousands of different
characterization procedure is required for true boiling point hydrocarbons and some non-hydrocarbons, such as N2, CO2
(TBP) and plus fractions of reservoir fluids. Although and H2S. The components of a reservoir fluid may be
reasonable prediction accuracy can be obtained, the EOS classified into three categories: (1). Defined components
parameters need to be tuned to improve the agreement whose critical temperature, critical pressure and acentric factor
between the measured properties and the calculated results. (Tc, Pc and ω) are well known; (2). True boiling point (TBP)
In this work, several characterization methods have briefly or single carbon number (SCN) fractions whose density and
been reviewed and the results of two widely used molecular weight are either measured or estimated; and (3) A
characterization procedures have been compared with the TBP residue or plus fraction which is too heavy to be
experimental data. The characterization methods can be separated by use of a TBP distillation procedure.
applied to both single and multiple sample systems. For single As we know, equations of state (EOS) need critical
sample systems, either an exponential or a three-parameter temperatures, critical pressures and acentric factors of
gamma distribution function was used to split the plus fraction components to estimate the EOS parameters. Therefore, a
into subfractions. Then the TBP fractions and split characterization procedure is required for TBP and plus
subfractions were lumped into groups (pseudo-components) fractions of reservoir fluids.
by means of molar averaging, mass averaging, Qauss- A number of characterization procedures have been
Laguerre quadrature or arbitrary selection. For multiple available in the open literature. Among them, the widely used
sample systems, a method has been developed to generate the characterization methods in the petroleum industry are those
same number of pseudo-components for each sample and the proposed by Pedersen et al.1,2 and Whitson3-6. Pedersen et al.1
same physical properties of each pseudo-component while the proposed a characterization procedure for TBP and plus
mole fractions of pseudo-components reflect the overall fractions. An exponential distribution function was applied to
distribution of C7+ mixtures. split the plus fraction into subfractions. The fractions heavier
In addition, a regression technique has been developed to than C6 were lumped into pseudo-components with
fit the PVT data and phase behavior for single or multiple approximately equal weight fraction of each pseudo-
samples of reservoir fluids. A modified Levenberg-Marquardt component (mass averaging). Later, Pedersen et al.2 extended
nonlinear least-square algorithm has been utilized in this work. this method to gas condensate mixtures and developed the
A sensitivity analysis is performed to determine which
property correlations to estimate Tc, Pc and ω based on a great
parameters are more sensitive. The more sensitive parameters
number of the measured PVT data of reservoir fluids.
2 J. Y. ZUO AND D. ZHANG SPE 64520

Whitson3-5 proposed a characterization method for C7+ An Exponential Distribution Function. Pedersen et al.1,2
fractions. A three-parameter gamma distribution function was proposed that the molar composition of the plus fraction is
used to represent the molar distribution of TBP and plus estimated by an exponential distribution function:
fractions. Whitson et al.6 then extended this procedure to
multiple related equilibrium fluid systems. They used the ln zi = A + BM i (1)
Gauss-Laguerre quadrature method to lump the TBP and plus
fractions into pseudo-components. where zi and Mi stand for the mole fraction and the molecular
The advantage of the characterization procedures weight of subfraction i, respectively. Coefficients A and B are
mentioned above is that the experimental PVT data are not determined from the measured or estimated weight fraction
necessarily required. The minimum input is the compositions and molecular weight of the plus fraction. The molecular
of reservoir fluids and the densities and molecular weights of weight of a given carbon number (Ni) split subfraction can be
the plus fractions. However, it is well known that cubic computed by:
equations of state cannot represent phase behavior of reservoir
fluids accurately, especially for gas condensates and for near M i = 14 N i − 4 (2)
critical reservoir fluids. As a result, the parameters in
equations of state need to be tuned to improve the Once A and B are determined, the mole fraction and
representation of PVT and physical properties of reservoir molecular weight can be attributed to each subfraction of the
fluids. plus fraction. The density (ρ) of the subfractions of the plus
In this work, the results calculated by using the fraction at atmospheric pressure are estimated by:
characterization methods of Pedersen et al.2 and Whitson3-6
have been compared with the experimental data. In addition, a ρ i = C + D ln N i (3)
regression technique has been developed to match PVT data
and phase behavior for single or multiple samples of reservoir where coefficients C and D are determined from the measured
fluids. The implemented characterization and regression or estimated densities of the plus fraction and that of the last
techniques have been tested for oils and gas condensates near defined or TBP (SCN) fraction. For a given plus fraction,
critical conditions and under ordinary conditions as well as for equations (1)-(3) guarantee that the mole fraction, molecular
multiple samples that were taken from a reservoir with a weight and density are consistent with the measured or
strong compositional gradient. Satisfactory results have been estimated data for the plus fraction.
obtained. Once Mi and ρi are known for TBP fractions and
subfractions, the property correlations (such as Pedersen et
Characterization of TBP and Plus Fractions al.2, Winn11, Cavett12, Kesler and Lee13, Twu14, and Riazi and
Since the early 1980s, continuous distribution models have Daubert15) can be used to estimate Tci, Pci and ωi.
been applied to describe the molar distribution of C7+ The number of components and fractions (including
fractions. Cotterman et al.7 and Cotterman and Prausnitz8 subfractions) in a mixture may be greater than 80; for instance,
applied continuous thermodynamics to compute vapor-liquid the plus fraction is split up to C80. Lumping methods are
equilibria of continuous and semicontinuous systems. required to group the TBP fractions and split subfractions into
However, the gamma distribution functions were maintained several pseudo-components. In the original paper of Pedersen
in the phase equilibrium calculations. Thus it is difficult to et al.1,2, only the mass averaging was employed. In this work,
implement this method into compositional reservoir the TBP fractions and subfractions can be lumped into pseudo-
simulators. As pointed out by the authors, six to ten quadrature components by means of molar averaging, mass averaging, or
points (pseudo-components) are usually sufficient to arbitrary selection.
accurately describe a continuous molar distribution. However, the characterization method of Pedersen et al.1,2
Behrens and Sandler9 used a semicontinuous cannot be suitable for multiple sample systems because each
thermodynamic description to model C7+ fractions. The sample has its own coefficients A, B, C and D. In the case of a
Gaussian quadrature method was applied to choose an optimal multiple sample system, we first specify the “main” fluid
discretization of C7+ fractions. Recently, Manafi et al.10 sample among multiple samples. In general, the sample that
extended continuous thermodynamics to represent phase dominates the simulation process is chosen as the main fluid
behavior of petroleum fluids. sample. The main fluid sample is then characterized using the
As mentioned in the previous section, the widely used above characterization procedure. Based on the results of the
characterization methods in the petroleum industry are those main fluid sample, the TBP fractions and subfractions of the
proposed by Pedersen et al.1,2 and Whitson3-6. Therefore, the plus fraction of the remaining samples are lumped by keeping
two distribution functions have been utilized to split plus the similar molecular weight to that of the pseudo-components
fractions in this work: an exponential and a three-parameter of the main fluid sample. Each pseudo-component has the
gamma distribution function. same properties as those of the main fluid sample.
SPE 64520PLUS FRACTION CHARACTERIZATION AND PVT DATA REGRESSION FOR RESERVOIR FLUIDS NEAR CRITICAL CONDITIONS 3

A Three-Parameter Gamma Distribution Function. interaction coefficient (kij), dimensionless volume shift
Whitson et al.3,4,5 proposed a three-parameter gamma parameter (Vt), and power (θ) in the binary interaction
distribution function to describe the relation between mole coefficient correlation between hydrocarbons:
fractions (zi) and molecular weights (Mi) of TBP and plus
fractions: θ
2Vci1 / 6 Vcj1 / 6
k ij = 1 − ÿ (9)
Vci1 / 3 + Vcj1 / 3 ÿ
M −η
(M − η )α −1 exp ÿÿ
β
p( M ) = (4)
β α Γ (α )
where Vc denotes the critical molar volume. The binary
interaction coefficient between the specified components i and
j will overwrite the value calculated by equation (9). The
where Γ is the gamma function. α gives a measure of the dimensionless volume shift parameter is defined as:
shape of the distribution. η is the lowest molecular weight in
the plus fraction and β stands for a normalization condition: ~ c
Vi t = i (10)
M Cn + − η = αβ (5) bi

Mcn+ is the molecular weight of the plus fraction. where c and b stand for the volume shift parameter and the
covolume of the EOS. For light components (C1-C6, N2, CO2
The cumulative distribution function is the integral of p(x)
and H2S), Vt’s are estimated by the correlation of Jhaveri and
from η to x:
Youngren17, while for heavy components, Vt’s are determined
by matching density of each component at atmospheric
∞ yα + 1
P( x ,α ) = p(x )dx = e − y
x pressure because the correlation of Jhaveri and Youngren is
(6)
j =0 Γ (α + j + 1)
η
not applicable to heavy components.
In addition, Vt’s are also correlated by the following
where y=(x-η)/β. expression:
The mole fraction is given by:
~
Vi t = d 1 + d 2Vi t (11)
z i = P(M i ,α ) − P(M i −1 ,α ) (7)
where d1 and d2 are two adjustable parameters (the default
and the molecular weight is expressed as: values are d1=0 and d2=1).
Four viscosity models18-21 are used to calculate viscosities
αβ
Mi = η + [P(M i ,α + 1) − P(M i −1 ,α + 1)] (8)
of reservoir fluids. When viscosity is computed, the following
equation is employed to calculate critical temperatures in the
zi
models of Pedersen and Fredenslund19, Aasberg-Petersen et
The TBP fractions and subfractions can then be lumped al.20 and Guo et al.21:
into pseudo-components by means of molar averaging, mass
averaging, Qauss-Laguerre quadrature or arbitrary selection. Tcivis = d ivis Tci (12)
The density of each pseudo-component is estimated by the
method of Whitson3,4. Once the molecular weight and density For the Lohrenz-Bray-Clark viscosity model18, the critical
of each pseudo-component are known, the property molar volume is computed by:
correlations (such as Pedersen et al.2, Winn11, Cavett12, Kesler
and Lee13, Twu14, and Riazi and Daubert15) are applied to Vcivis = divisVci (13)
estimate Tci, Pci and ωi.
For multiple sample systems, the characterization The coefficients (dvis) can be tuned to fit experimental
procedure with the gamma distribution function extended by viscosity data. The default values are equal to unity.
Whitson et al.6 is used in this work.
PVT Properties to Be Fitted. The following measured
Regression Technique data may be fitted by tuning the EOS parameters: the
In general, cubic EOS cannot accurately predict volumetric volumetric and physical property data from saturation
properties and phase behavior of reservoir fluids. Therefore, (dew/bubble point) pressure test, constant composition
the EOS parameters are tuned to fit the measured PVT data. expansion (CCE) test, constant volume depletion (CVD) test,
differential liberation (DL) test, separator (SPE) test, swelling
Tunable Parameters. For a given cubic EOS (such as the test and P-T flash test. Each lab test may be performed several
PR EOS16), the tunable parameters are: critical temperature times at different temperatures for different reservoir fluid
(Tc), critical pressure (Pc), acentric factor (ω), binary samples.
4 J. Y. ZUO AND D. ZHANG SPE 64520

It should be noted that volume shift parameters do not Table 1 illustrates the results predicted by the exponential
influence saturation pressure calculations and viscosity (referred to as Method I) and the gamma distribution function
coefficients do not affect saturation pressure and volumetric (referred to as Method II). The average deviation of the
data calculations. Therefore, saturation pressure data, predicted saturation pressures by Method I is 4.88% and that
volumetric data and viscosity data can be fitted separately or by Method II is 13.21%. Method I is much better than Method
simultaneously. II. Coats and Smart22 used the PR EOS to predict saturation
pressures for several oils and gases with an average deviation
Regression Procedure. The EOS parameter tuning is a of 14%. If we utilized Methods I and II and Coats and Smart’s
nonlinear minimization problem. The objective function is method to predict the saturation pressures for the same
expressed as: systems only taken from Coats and Smart22 as shown in Table
1, the deviations are 4.9%, 6.3% and 13.6% by Methods I and
ÿ 2
eiexp − eical (x ) ÿ
II and Coats and Smart’s method, respectively.
ÿ M exp
F (x ) = wi (14) Fig. 1 compares the predicted saturation pressure using
i =1 eexp ÿ
i Methods I and II with the experimental data for Mixture 4 of
Pedersen et al.1 near critical conditions. Method I predicts two
where x is the vector of adjustable parameters with N (number dew points and two bubble points at the four experimental
of adjustable parameters) dimension. Mexp, w, eexp and ecal temperatures, which are consistent with the measured data.
denote the number of experimental data points, the weighing While Method II under-predicts saturation pressures and
factor, the experimental and calculated values, respectively. predicts all the bubble points at the four experimental
The lower bound Xmin and the upper bound Xmax are used temperatures (all four open circles are on the left side of the
to scale the adjustable parameters X. critical point), which conflicts with the experimental
observation.
X j − X min Table 2 lists the predicted saturation densities by
xj =
j
(15) application of Methods I and II and the measured data for
X max − X min
j j
several reservoir fluids reported by Coats and Smart22. The
saturation pressures were matched by tuning the power (θ) in
Thus, xj always lies between zero and unity. Eq. (9) since pressures affect density calculations. The
The goal of the EOS parameter tuning is to find the values deviation of the predicted densities is less than 1%.
of adjustable parameters so that the objective function reaches In general, Method I gives better predictions of PVT
its minimum. The partial derivatives (gradients) of the properties and phase behavior of reservoir fluids than Method
objective function, with respect to the adjustable parameters, II.
represent the sensitivity of the objective function to the
parameters. The parameters with larger absolute values of Examples of the EOS Tuning
gradients are more sensitive; therefore, they are good As mentioned above, a cubic EOS with a suitable
candidates for being adjustable variables. A modified characterization procedure may not represent PVT properties
Levenberg-Marquardt nonlinear least-square algorithm is and phase behavior of reservoir fluids accurately. The EOS
utilized in this work. parameters are usually tuned to fit the experimental data. This
A typical regression procedure is as follows: section gives several examples of the EOS parameter tuning
(1). Select tunable parameters (initial values are obtained for different kinds of reservoir fluids. The Peng-Robinson
from characterization) and set lower and upper bounds for EOS16 and the property (Tc, Pc and ω) correlations of Pedersen
each parameter; et al.1 were employed in the following calculations.
(2). Input experimental data and specify a weighing factor
for each data point; Near Critical Gas Condensate. The composition of the near
(3). Perform a sensitivity analysis and sort the tunable critical gas condensate is given in Table 3. The experimental
parameters in decreasing order of ∂F/∂xj; data were measured by the laboratory of DB Robinson
(4) Select the more sensitive parameters as adjustable Research Ltd. The original composition was analyzed to C30+.
parameters and carry out a nonlinear regression. Method I was used to characterize the TBP and plus
fractions of this near critical gas condensate. The C7+ fraction
Comparison of Two Characterization Procedures was lumped into 15 pseudo-components with approximately
The two characterization procedures mentioned above have equal weight percentage of each pseudo-component. The
been applied to predict saturation pressures of reservoir fluids. experimental data to be matched are the dew point, the liquid
The Peng-Robinson EOS16 was used in all calculations. C7+ volume percentage and the relative volume in the CCE test
fractions were lumped into 5 pseudo-components with mass and the liquid dropout in the CVD test. In totally, 12
averaging for the exponential distribution function and with parameters were tuned and the adjustable parameters before
the Gaussian quadrature for the gamma distribution function, and after tuning are given in Table 4.
respectively. The property correlations of Pedersen et al.1 were
employed to estimate Tc, Pc and ω of each pseudo-component.
SPE 64520PLUS FRACTION CHARACTERIZATION AND PVT DATA REGRESSION FOR RESERVOIR FLUIDS NEAR CRITICAL CONDITIONS 5

Fig. 2 shows the phase envelope of the near critical gas


condensate. The dashed and the solid lines are the saturation Gas Condensate. The composition of the gas condensate is
(bubble/dew) point curves before and after the EOS parameter also given in Table 3. The original composition was analyzed
tuning, respectively. The open circle and square represent the to C11+.
critical points predicted by use of the PR EOS before and after Method I was applied to characterize the TBP and plus
tuning, respectively. It can be seen that the dew point fractions of this gas condensate. The C7+ fraction was lumped
prediction is in good agreement with the measured point into 15 pseudo-components by means of mass averaging. The
denoted by an open diamond symbol. experimental data to be matched are: the dew point, the liquid
A comparison between the calculated results and the volume percentage and the relative volume in the CCE test
experimental data of the liquid volume percentage and relative and the liquid dropout in the CVD test. The adjustable
volume as a function of pressures in the CCE test at 281.9 oF parameters before and after the EOS tuning are tabulated in
is given in Fig. 3. The liquid volume percentage increases Table 6.
steeply as pressures drop slightly near the dew point. As we Fig. 9 shows the phase envelope of the gas condensate.
can see, the calculated results, after the EOS parameter tuning, The dashed and the solid lines are the dew point curves before
agree extremely well with the experimental data. and after the parameter tuning, respectively. The solid circles
Fig. 4 shows the calculated and experimental results of the represent the experimental data. The dew point calculations
liquid dropout in the CVD test, which are in good agreement. are in good agreement with the measured data after the EOS
tuning. There is no critical point for this gas condensate.
Near Critical Oil. The composition of the near critical oil is A comparison between the calculated results and the
also given in Table 3. The original composition was analyzed experimental data of the liquid volume percentage and relative
to C30+. volume as a function of pressures in the CCE test is shown in
Method I was used to characterize the TBP and plus Fig. 10. The PR EOS without tuning (dashed curves) over-
fractions of the near critical oil. The C7+ fraction was lumped predicts the relative volume over the pressure range, and much
into 6 pseudo-components in terms of mass averaging. The over-predicts the liquid volume percentage at higher pressures,
experimental data to be matched are: the bubble point, the but under-predicts it at lower pressures. The calculated results
liquid volume percentage, the oil density and the relative after the parameter tuning are in good agreement with the
volume in the CCE test, and the liquid density, the oil FVF experimental data, especially for liquid volume percentage
and the solution GOR in the DL test. The tunable parameters around the dew point, which is usually much over-predicted
before and after the EOS tuning are listed in Table 5. by a cubic EOS.
Fig. 5 shows the phase envelope of the near critical oil.
The dashed and the solid lines are the saturation point Multiple Samples of Riemens et al.25 Riemens et al.25
(bubble/dew) curves before and after the EOS parameter published PVT properties of two fluid samples taken from
tuning, respectively. The open circle and square represent the wells producing from the same geological formation in Birba
critical points predicted by use of the PR EOS before and after Field. One is the oil sample and the other is the gas condensate
tuning, respectively. The bubble point prediction agrees well sample.
with the measured data (diamond) after the EOS tuning. The method of Whitson et al.6 was employed to
A comparison between the calculated results and the characterize the C7+ fractions of the multiple samples in Birba
experimental data of the liquid volume percentage and relative Field. The C7+ fractions were lumped into 3 pseudo-
volume as a function of pressures in the CCE test is depicted components with the Gaussian quadrature method. The
in Fig. 6. The liquid volume percentage decreases steeply as experimental data to be matched are the saturation pressures
pressures drop slightly near the bubble point. The calculated and GOR. In total, 4 parameters were tuned and the tunable
results, after the EOS parameter tuning, match extremely well parameters before and after tuning are given in Table 7. The
with the experimental data. calculated results are tabulated in Table 8. The calculations
Fig. 7 compares the calculated results of the liquid density after the EOS tuning are in good agreement with the measured
with the experimental data, as a function of pressures, in the data.
CCE and DL tests. The PR EOS, without tuning, somewhat
over-predicts the liquid density while the calculated results, Multiple Samples of Hoffmann et al.26 Hoffmann et al.26
after the EOS tuning, are in good agreement with the reported PVT properties of two oil and gas cap fluid samples.
experimental data. The method of Whitson et al.6 was used to characterize the
A comparison of the calculated results of the oil formation C7+ fractions of the multiple samples and the C7+ fractions
volume factor (Bo, FVF) and solution GOR with the were lumped into 2 pseudo-components with the Gaussian
experimental data in the DL test is shown in Fig. 8. The PR quadrature method. The experimental data to be matched are:
EOS, without tuning, under-predicts the oil FVF and solution the saturation pressures, GOR and liquid densities. In total, 6
GOR at pressures a little lower than the bubble point as the parameters were tuned, and the tunable parameters, before and
predicted bubble point pressure is slightly higher than the after the EOS tuning, are given in Table 9. The calculated
measured point. The agreement is good, between the results are given in Table 10. The calculations after the EOS
calculated and the experimental data, after the EOS tuning. tuning agree well with the measured data.
6 J. Y. ZUO AND D. ZHANG SPE 64520

Oil 2 of Coats and Smart22. Coats and Smart22 reported PVT Nomenclature
properties of the very volatile Oil 2. In the previous section, A = coefficient in Eq. (1)
we utilized Method I with 5 pseudo-components for B = coefficient in Eq. (1)
representing the C7+ fraction. The predicted bubble point b = covolume of the PR EOS, ft3/lb mole
pressure was 3879.9 psia, which is considerably below the C = coefficient in Eq. (3)
measured 4474.7 psia. After the power (θ) in Eq. (9) was c = volume shift parameter, ft3/lb mole
adjusted to match the observed bubble point pressure, Oil 2 D = coefficient in Eq. (3)
behaves as a gas condensate at the reservoir temperature of Dev% = (1/Mexp)Σj|1-ecal/eexp|j×100
176 OF. This is obviously discrepant with the experimental dvis = viscosity coefficient in Eqs. (12) and (13)
observation. d1 = coefficient in Eq. (11)
Therefore, we applied two pseudo-components to represent d2 = coefficient in Eq. (11)
the C7+ fraction. The predicted bubble point pressure is closer e = property
to the measured value. Then power (θ) in Eq. (9) was tuned to F(x) = objective function
match the observed bubble point pressure. In this case, Oil 2 kij = binary interaction coefficient
behaves as oil at the reservoir temperature of 176 OF. Next, the M = molecular weight, lb/lb mole
EOS was employed to predict PVT properties of Oil 2. The Mexp = number of data points
overall relative deviation is 1.93%. Satisfactory results were Mi = molecular weight of species i, lb/lb mole
obtained. N = number of adjustable parameters
This example indicates that the number of pseudo- Ni = carbon number of species i
components chosen in the characterization of TBP and plus P = pressure, psi
fractions is sometimes crucial. The number of pseudo- P(x, α) = cumulative distribution function, defined in Eq. (6)
components may result in an incorrect representation of the p(x) = distribution or probability density function
reservoir fluid, especially for reservoir fluid in the vicinity of Pc = critical pressure, psi
the critical point. Usually, 2-6 pseudo-components for oil and Tc = critical temperature, oF
10-15 pseudo-components for gas condensate are sufficient to Vc = molar critical volume, ft3/lb mole
represent C7+ fractions. Vt = dimensionless volume shift parameter
w = weighing factor
Other Systems of Coats and Smart22. In addition, we x = integral variable in Eq. (6)
compared Method I with the method of Coats and Smart22. In x, X = adjustable parameter
our calculations, C7+ fractions were lumped into 5 pseudo- y = (x-η)/β
components. Only power (θ) in Eq. (9) was tuned to match zi = mole fraction of species i
saturation pressures. The other properties were predicted by α = parameter of the gamma distribution function
the PR EOS. The overall deviation of the predicted PVT β= parameter of the gamma distribution function
properties using Method I for Gas 4, Gas 5, Oil 3, Oil 4, Oil 6 η = lowest molecular weight in the plus fraction
and Oil 7 is around 3%, which is comparable to the results θ = power in the binary interaction coefficient
regressed by Coats and Smart22. correlation between hydrocarbons, in Eq. (9)
ρ = density, lb/ft3
Conclusions ω = acentric factor
In this work, several characterization methods have briefly
been reviewed and the results obtained by using two widely Superscripts
used characterization procedures have been compared with the cal = calculated value
experimental data. The characterization method of Pedersen et exp = experimental value
al.2 usually gives better predictions than that of Whitson et max = maximum value (upper bound)
al.5,6 without the EOS beong tuned. min = minimum value (lower bound)
In addition, a regression technique has been developed to
fit PVT data and phase behavior for single or multiple samples Subscripts
of reservoir fluids. The implemented characterization and c = value at critical conditions
regression techniques have been tested for oils and gas i,j = species i, j
condensates at near critical conditions and under ordinary s = value at saturation conditions
conditions as well as for multiple samples. The results indicate
that the implemented methods provide a reliable and efficient
tool for representing PVT properties and phase behavior of References
reservoir fluids. 1. Pedersen, K.S., Thomassen, P. and Fredenslund, Aa.:
“Characterization of Gas Condensate Mixtures,” Presented in
Chorn, L.G. and Mansoori, G.A.: C7+ Fraction
Characterization, Taylor & Francis New York Inc., New York
SPE 64520PLUS FRACTION CHARACTERIZATION AND PVT DATA REGRESSION FOR RESERVOIR FLUIDS NEAR CRITICAL CONDITIONS 7

(1989) 137. 22. Coats, K.H. and Smart, G.T.: “Application of a Regression-
2. Pedersen, K.S., Fredenslund, Aa. and Thomassen, P.: Based EOS PVT Program to Laboratory Data,” SPE Reservoir
Properties of Oils and Natural Gases, Gulf Publishing Co., Eng. (May 1986) 277.
Houston, Texas (1989). 23. Agarwal, R.K., Li, Y.-K. and Nghiem, L.: “A Regression
3. Whitson, C.H.: “Characterizing Hydrocarbon Plus Fractions,” Technique with Dynamic Parameter Selection for Phase-
Soc. Pet. Eng. J. (August 1983) 683. Behavior Matching,” SPE Reservoir Eng. (February 1990) 115.
4. Whitson, C.H.: “Effect of C7+ Properties on Equation of State 24. Christesen, P.L.: “Regression to Experimental PVT Data,” J.
Predictions,” Soc. Pet. Eng. J. (December 1984) 685. Can. Pet. Tech. (1999) 38, 1.
5. Whitson, C.H., Anderson, T.F. and Soreide, I.: “Application of 25. Riemens, W.G., Shulte, A.M. and de Jong, L.N.J.: “Birba Field
the Gamma Distribution Model to Molecular Weight and PVT Variations Along the Hydrocarbon Colum and
Boiling Point Data for Petroleum Fractions,” Chem. Eng. Confirmatory Field Tests,” J. Pet. Tech. (January 1988) 83.
Comm. (1990) 96, 259. 26. Hoffmann, A.E., Crump, J.S. and Hocott, C.R.: “Equilibrium
6. Whitson, C.H., Anderson, T.F. and Soreide, I.: “C7+ Constants for a Gas-Condensate System,” Pet. Trans. AIME
Characterization of Related Equilibrium Fluids Using the (1953) 198, 1.
Gamma Distribution,” Presented in Chorn, L.G. and Mansoori,
G.A.: C7+ Fraction Characterization, Taylor & Francis New
York Inc., New York (1989) 35. SI Metric Conversion Factors
7. Cotterman, R.L., Bender, R. and Prausnitz, J.M.: “Phase o
Equilibria for Mixtures Including Very Many Components
F (oF+459.67)/1.8 =K
Development and Application of Continuous Thermodynamics psi x 6.894 757 E+00 = kPa
for Chemical Process Design,” Ind. Eng. Chem. Proc. Des. Dev. lb/ft3 x 1.601 846 E+01 = kg/m3
(1985) 24, 194. scf/stb x 1.781 073 E+00 = std m3/st m3
8. Cotterman, R.L. and Prausnitz, J.M.: “Flash Calculations for *Conversion factor is exact
Continuous or semicontinuous Mixtures Using an Equation of
State,” Ind. Eng. Chem. Proc. Des. Dev. (1985) 24, 434.
9. Behrens, R.A. and Sandler, S.I.: “The Use of Semicontinuous
Description to Model the C7+ Fraction in Equation of State
Calculations,” SPE Reservoir Eng. (August 1988) 1041.
10. Manafi, H., Mansoori, G.A. and Ghotbi, S.: “Phase Behavior
Prediction of Petroleum Fluids with Minimum Characterization
Data,” J. Pet. Sci. Eng. (1999) 22, 67.
11. Winn, F.W.: “Physical Properties by Nomogram,” Pet. Refiner
(1957) 36, 157.
12. Cavett, R.H.: “Physical Data for Distillation Calculation,
Vapor-Liquid Equilibria,” Presented at the 27th Midyear
Meeting, API Division of Refining, San Francisco, CA, May
15, 1964.
13. Kesler, M.G. and Lee, B.I.: “Improve Prediction of Enthalpy of
Fractions,” Hydrocarbon Processing (1976) 55, 153.
14. Twu, C.H.: “An internally Consistent Correlation for Predicting
the Critical Properties and Molecular Weights of Petroleum and
Coal-Tar Liquids,” Fluid Phase Equilibria (1984) 16, 137.
15. Riazi, M.R. and Daubert, T.E.: “Prediction of the Composition
of Petroleum Fractions,” Ind. Eng. Chem. Process Des. Dev.
(1980) 19, 289.
16. Peng, D.-Y. and Robinson, D.B.: “A New Two-Constant
Equation of State,” Ind. Eng. Chem. Fundam. (1976) 15, 59.
17. Jhaveri, B.S. and Youngren, G.K.: “Three Parameter
Modification of the Peng-Robinson Equation of State to
Improve Volumetric Predictions,” SPE Reservoir Eng. (August
1988) 1033.
18. Lohrenz, J., Bray, B.G. and Clark, C.R.: “Calculating
Viscosities of Reservoir Fluids from Their Compositions,” J.
Petr. Tech. (October 1964) 1171.
19. Pedersen, K. S. and Fredenslund, Aa.: “An Improved
Corresponding States Model for the Prediction of Oil and Gas
Viscosities and Thermal Conductivities,” Chem. Eng. Sci.
(1987) 42, 182.
20. Aasberg-Petersen, K., Knudsen, K. and Fredenslund, Aa.:
“Prediction of the Viscosities of Hydrocarbon Mixtures,” Fluid
Phase Equilibria (1991) 70, 293.
21. Guo, X.-Q., Wang, L.-S., Rong, S.-X. and Guo, T.-M.:
“Viscosity Models Based on Equations of State for
Hydrocarbon Liquids and Gases,” Fluid Phase Equilibria
(1997) 139,405.
8 J. Y. ZUO AND D. ZHANG SPE 64520

TABLE 1 COMPARISON OF THE PREDICTED SATURATION


PRESSURES WITH EXPERIMENTAL DATA USING TWO TABLE 3 COMPOSITIONS OF RESERVOIR FLUIDS
CHARACTERIZATION PROCEDURES
Near Critical Gas Near Critical Oil Gas Condensate
Sample T Exp. Method I Dev Method II Dev CO2 1.812 1.013 0.68
O
F psia psia % psia % N2 0.000 0.613 3.43
Mix 11 205.88 4090.0 3703.0 9.46 3753.0 8.24 C1 76.96 55.850 76.11
Mix 2 1 246.02 5772.5 5350.7 7.31 3085.0 46.56 C2 7.172 12.858 9.07
Mix 3 1 289.40 6526.7 6065.1 7.07 3643.2 44.18 C3 2.853 6.704 2.71
Mix 4 1 287.96 5665.2 5662.6 0.05 5078.6 10.35 C4 1.333 5.009 1.30
Mix 4 1 303.98 5666.3 5667.0 0.01 5065.8 10.60 C5 0.285 2.883 0.56
Mix 4 1 325.94 5594.1 5655.5 1.10 5030.9 10.07 C6 0.187 1.893 0.43
Mix 4 1 338.00 5568.0 5640.7 1.31 5003.4 10.14 C7+ 9.398 13.177 5.71
Mix 5 1 281.00 4787.7 4764.4 0.49 4402.5 8.05
Oil 122 180.00 2534.7 2459.4 2.97 2464.6 2.77
Oil 2 22 176.00 4474.7 3879.9 13.29 3709.8 17.09
Oil 3 22 179.00 2611.7 2652.9 1.58 2566.1 1.75
Oil 4 22 250.00 2561.7 2435.6 4.92 2416.3 5.68
Oil 6 22 234.00 2760.7 2563.4 7.15 2531.7 8.29
Oil 7 22 131.00 1708.7 1674.7 1.99 1674.9 1.98
Gas 2 22 190.00 4464.7 4426.0 0.87 4172.8 6.54
TABLE 4 TUNED PARAMETERS FOR THE NEAR CRITICAL GAS
Gas 4 22 240.00 3374.7 3651.4 8.20 3520.7 4.33 CONDENSATE
Gas 5 22 267.00 4856.7 4770.8 1.77 4637.2 4.52
Mix 12 311.00 5627.5 5019.1 10.81 3416.1 39.30 Parameter Initial Value Final Value
Mix 2 2 199.04 3885.6 3810.9 1.92 1912.8 50.77 θ 0.0000 -0.1089
Mix 3 2 160.88 3466.4 3406.0 1.74 3438.4 0.81 d1 0.0000 -0.1062
Mix 4 2 285.80 4935.6 4795.0 2.85 4614.7 6.50 d2 1.0000 1.023975
Mix 9 2 180.50 4075.6 3969.9 2.59 3935.3 3.44 Tc of Pseudo 2, oF 686.066 719.186
Mix 5 2 264.20 6729.8 6187.8 8.05 3942.0 41.42 Tc of Pseudo 3, oF 726.782 759.362
Mix 6 2 199.94 3981.3 3697.8 7.12 3721.4 6.53 Tc of Pseudo 5, oF 809.258 865.436
Mix 7 2 200.12 3839.2 3399.0 11.47 3310.1 13.78 Tc of Pseudo 6, oF 831.956 897.404
Mix 8 2 200.12 3839.2 3536.2 7.89 3758.9 2.09 Tc of Pseudo 7, oF 850.658 905.72
Mix 10 2 89.96 6712.4 5969.3 11.07 6645.5 1.00 Tc of Pseudo 8, oF 869.576 915.17
Mix 10 2 154.04 5824.7 5601.7 3.83 6191.5 6.30 Tc of Pseudo 12, oF 1014.332 1004.36
Mix 10 2 257.72 4401.9 4453.6 1.17 4929.7 11.99 Tc of Pseudo 13, oF 1072.814 1005.656
Mix 10 2 336.92 3330.1 3123.1 6.22 3699.7 11.10 Tc of Pseudo 15, oF 1479.668 1478.75
Overall 4.88 13.21

TABLE 2 COMPARISON OF THE PREDICTED SATURATION TABLE 5 TUNED PARAMETERS FOR THE NEAR CRITICAL OIL
DENSITIES WITH EXPERIMENTAL DATA USING TWO Parameter Initial Value Final Value
CHARACTERIZATION PROCEDURES
θ 0.0000 0.00416
Sample T Exp. Method I Dev. Method II Dev. d1 0.0000 -0.06377
O
F lb/ft2 lb/ft2 % lb/ft2 % d2 1.0000 0.659552
Oil 122 180 47.96 47.86 0.21 47.77 0.40 Tc of Pseudo 1, oF 550.544 515.516
Oil 222 176 33.01 33.48 1.40 33.30 0.88 Tc of Pseudo 2, oF 621.986 591.746
Oil 322 179 44.17 44.05 0.27 44.70 1.20 Tc of Pseudo 3, oF 696.398 660.182
Oil 422 250 40.34 40.20 0.35 40.09 0.62 Tc of Pseudo 4, oF 802.706 764.600
Oil 622 234 38.01 38.39 0.99 38.26 0.66 Tc of Pseudo 5, oF 923.576 898.304
Oil 722 131 44.48 44.37 0.25 44.31 0.38 Tc of Pseudo 6, oF 1199.930 1184.936
Gas 222 190 29.54 29.54 0.00 29.89 1.18 Vt of Pseudo 1 -0.00855 -0.02335
Gas 522 267 19.15 18.72 2.30 18.80 1.83 Vt of Pseudo 4 0.01922 -0.01835
Overall 0.72 0.89
SPE 64520PLUS FRACTION CHARACTERIZATION AND PVT DATA REGRESSION FOR RESERVOIR FLUIDS NEAR CRITICAL CONDITIONS 9

TABLE 6 TUNED PARAMETERS FOR THE GAS CONDENSATE TABLE 9 TUNED PARAMETERS FOR THE FLUID SAMPLES OF
HOFFMANN ET AL.23
Parameter Initial Value Final Value
θ 0.0000 0.2948 Parameter Initial Value Final Value
Tc of Pseudo 1, oF 511.815 545.612 θ 0.0000 -0.36257
Tc of Pseudo 2, oF 553.686 601.974 d1 0.0000 -0.26879
Tc of Pseudo 4, oF 628.384 685.46 d2 1.0000 0.933407
Tc of Pseudo 5, oF 686.409 752.328 Pc of Pseudo 1, psi 345.687 366.338
Tc of Pseudo 7, oF 763.062 801.021 Tc of Pseudo 1, oF 722.184 670.476
Tc of Pseudo 8, oF 809.261 825.462 Vt of Pseudo 2 -0.5709 -0.2000
Tc of Pseudo 9, oF 852.614 834.408
Tc of Pseudo 10, oF 893.716 868.25
Tc of Pseudo 11, oF 941.954 887.274
Tc of Pseudo 12, oF 1005.816 913.537
Tc of Pseudo 13, oF 1075.751 922.45
Tc of Pseudo 14, oF 1169.195 935.357
Tc of Pseudo 15, oF 1384.106 1274.4 TABLE 10 RESULTS FOR THE FLUID SAMPLES OF HOFFMANN
Pc of Pseudo 11, psi 202.15 181.98 ET AL.26
Pc of Pseudo 12, psi 194.22 174.046
Oil Gas
Pc of Pseudo 13, psi 188.025 174.35
Property Exp. Before After Exp. Before After
Pc of Pseudo 14, psi 182.724 168.665
Pc of Pseudo 15, psi 177.794 181.24 Ps, psia 3837 4236 3854 3839.7 4650.3 3825.5
GOR, scf/stb 910 947 869
FVF, scf/stb 1.475 1.466 1.430
ρs,lb/ft3 40.98 44.09 39.81
ρs,lb/ft3 40.80 45.05 40.75
TABLE 7 TUNED PARAMETERS FOR THE RESERVOIR FLUID ρs,lb/ft3 40.81 46.23 41.87
SAMPLES OF RIEMENS ET AL.25 ρs,lb/ft3 42.15 47.44 42.97
ρs,lb/ft3 44.69 48.72 44.08
Parameter Initial Value Final Value
ρs,lb/ft3 45.00 50.16 45.25
θ 0.0000 0.01888
Dev, % 5.04 2.68 12.20 1.40
Tc of Pseudo 3, oF 1272.93 1291.38
d1 0.0000 -0.3437
d2 1.0000 -1.4500

TABLE 8 RESULTS FOR THE RESERVOIR FLUID SAMPLES OF


RIEMENS ET AL.25
Birba-2 Birba South-1
Property Exp. Before After Exp. Before After
Ps, psia 6883 6610 6883 6844 6568 6844
ρs, lb/ft3 41.89 42.13 41.30 30.59 27.02 25.90
GOR, scf/stb 1465 1521 1465 6176 7894 6176
FVF, scf/stb 1.65 1.65 1.63 4.80 3.94
Dev, % 2.09 0.66 14.50 5.11
10 J. Y. ZUO AND D. ZHANG SPE 64520

6000 80 5
Exp.
After Tuning
5000
Before Tuning 4
60 Exp. (Vr)

Relative Volume
Vliquid/Vtotal %
4000 After Tuning (Vr)
Pressure, psia

Before Tuning (Vr) 3

3000 40

Exp. (Pedersen et al.) 2


2000 Method I
20
CP (Method I) 1
1000 Method II
CP (Method II)
0 0
0 0 2000 4000 6000 8000 10000
-150 50 250 450 650 850
Pressure, psia
Temperature, F

Fig. 1—Comparison of the Predicted Saturation Pressures Fig. 3—Liquid Volume Percentage and Relative Volume as
1 a Function of Pressure in the CCE Test for the Near
with the Experimental Data for Mixture 4 of Pedersen et al. o
Near Critical Conditions. Critical Gas Condensate at 281.9 F.

10000 60

8000
Liquid Dropout % in CVD

40
Pressure, psia

6000

4000
Exp. Dew Point 20 Exp.
CP (Before Tuning) After Tuning
CP (After Tuning) Before Tuning
2000
Before Tuning
After Tuning
0
0
0 2000 4000 6000 8000 10000
-400 -200 0 200 400 600 800 1000
Temperature, F Pressure, psia

Fig. 2—Phase Envelope of the Near Critical Gas


Condensate. Fig. 4—Liquid Dropout as a Function of Pressure in the
o
CVD Test for the Near Critical Gas Condensate at 281.9 F.
SPE 64520PLUS FRACTION CHARACTERIZATION AND PVT DATA REGRESSION FOR RESERVOIR FLUIDS NEAR CRITICAL CONDITIONS 11

5000 0.7 0.8


Exp.
Before Tuning
4000

Liquid Density(CCE), g/cc


After Tuning 0.6

Liquid Density(DL), g/cc


0.6 Exp. (DL)
Pressure, psia

3000 Before Tuning (DL)


After Tuning (DL) 0.4

Exp. BP
2000
CP (Before Tuning) 0.5
CP (After Tuning) 0.2
1000 Before Tuning
After Tuning
0.4 0
0 0 2000 4000 6000 8000 10000
-300 0 300 600 900 Pressure, psia
Temperature, F

Fig. 7—Liquid Density as a Function of Pressure in the


Fig. 5—Phase Envelope of the Near Critical Oil. o
CCE and DL Tests for the Near Critical Oil at 231 F.

100 4 8000 5
Exp.
Exp. Before Tuning
80 Before Tuning After Tuning 4
3 6000
After Tuning Exp. (Bo)
Solution GOR, scf/stb

Oil FVF(Bo), bbl/stbbl


Relative Volume
Vliquid/Vtotal %

Exp. (Vr) Before Tuning (Bo)


60 Before Tuning (Vr) 3
After Tuning (Bo)
After Tuning (Vr) 2 4000
40 2

1 2000
20 1

0 0 0 0
0 2000 4000 6000 8000 10000 0 1000 2000 3000 4000 5000
Pressure, psia Pressure, psia

Fig. 6—Liquid Volume Percentage and Relative Volume as Fig. 8—Solution GOR and Oil Formation Volume Factor as
a Function of Pressure in the CCE Test for the Near a Function of Pressure in the DL Test for the Near Critical
o o
Critical Oil at 231 F. Oil at 231 F.
12 J. Y. ZUO AND D. ZHANG SPE 64520

10000

8000
Pressure, psia

6000

4000
After Tuning
Before Tuning
Exp.
2000

0
0 200 400 600 800
Temperature, F

Fig. 9—Phase Envelope of the Gas Condensate.

30 2.5
After Tuning (Vol.%)
Before Tuning (Vol.%)
Exp. (Vol.%)
2.0
After Tuning (Rel. Vol.)
Before Tuning (Rel. Vol.)
20
Relative Volume
Liquid Volume%

Exp. (Rel. Vol.)


1.5

1.0
10

0.5

0 0.0
2000 4000 6000 8000 10000
Pressure, psia

Fig. 10—Liquid Volume Percentage and Relative Volume


as a Function of Pressure in the CCE Test for the Gas
o
Condensate at 270.5 F.

You might also like