An Approach for Characterization and
Lumping of Plus Fractions of Heavy Oil
I. Rodriguez and A.A. Hamouda, University of Stavanger, Norway
Summary
Heavy-oil fluids contain large concentrations of high-molecular-
weight components, including a large content of the plus fractions,
such as C7+.
Different approaches have been developed to characterize the
petroleum plus fractions to improve prediction of the pseudocom-
ponents properties by equations of state (EOSs).
A method is developed in this work to split the plus fraction into
single carbon numbers (SCN), generating the mole fraction and the
respective molecular weight. The developed method is based on the
relationships between three-parameter gamma (TPG) distribution,
experimental mole fraction, molecular weight, and SCN data obtained
from the literature and industrial contacts. TPG is used to fit the trend
of the compositional analysis. The characterized mole distribution as
a function of SCNs is generated by integrating the TPG between the
limiting molecular weights (LMw). The limiting molecular weights are
determined simultaneously during the integration process by fitting the
characterized and experimental mole fractions.
The developed method is easy to use. In addition, the approach
is not dependent on the assumption that only normal carbon num-
bers exist in the composition resulting on fixed molecular weights
for each single carbon number.
There are several correlations generated to predict physico-
chemical properties as a function of SCNs. Those correlations have
been originally developed to work with light oil. Our approach is
combined with some of the correlations and is tested for heavy-oil
samples to identify the ranges in which they can be applied. Two
lumping schemes are used to group the SCNs into pseudocompo-
nents. The properties for each pseudo-component in this work are
used to predict pressure/volume/temperature (PVT) data, constant
volume depletion, using the Peng-Robinson EOS (PR-EOS), and
the PVTP™ commercial simulator.
Introduction
EOSs have been extensively used to generate PVT and phase-
behavior calculations of petroleum fluids. The PR-EOS (Peng
and Robinson 1976) was originally developed to predict PVT and
phase behavior for pure substances with known critical properties.
The use of an EOS in multicomponent mixtures, like hydrocarbon
mixtures, is possible by the application of mixing rules (Danesh
1998; Ahmed 2007). The application of mixing rules implies that
properties such as Pc, Tc, and acentric factor  should be known
for all the components of the mixture, and then grouped together
to perform calculations (Danesh 1998; Ahmed 2007).
The constituents of a hydrocarbon system are classified in
two categories: the well-defined components and the undefined
petroleum fractions, which are those heavy compounds lumped
together and identified as the plus-fraction [i.e., C7+ (Ahmed
2007)]. Several samples of heavy oil indicate that the plus fraction
C10+ contains a molar fraction close to 70% and that the fraction
C80+ is also representative with values close to 2% (Pedersen et al.
2004). The importance of characterizing the plus fraction arises
when the modeled fluid has high molecular weight and high den-
sity (heavy oil).
Copyright © 2010 Society of Petroleum Engineers
This paper (SPE 117446) was accepted for presentation at the International Thermal
Operations and Heavy Oil Symposium, Calgary, 20–23 October 2008, and revised for
publication. Original manuscript received for review 12 August 2008. Revised manuscript
received for review 1 February 2009. Paper peer approved 18 February 2009.
April 2010 SPE Reservoir Evaluation & Engineering
Characterization of plus fraction usually consists of three parts:
(1) splitting the fraction into a certain number of component groups
called SCNs; (2) estimation of the physicochemical properties of
the SCN; and (3) lumping of the generated SCNs (Pedersen et al.
2006).
Several methods have been developed to estimate the mole
distribution of the compounds in the plus fraction (Katz 1983;
Lohrenz et al. 1964; Pedersen et al. 1982; Ahmed et al. 1985;
Whitson 1983). A common feature among these methods is that
components with the same number of carbon atoms are lumped
in groups called SCNs with a single predefined value for proper-
ties like Pc, Tc, , Tb, M, and density . Whitson (1983) proposed
a method using the TPG distribution in order to estimate the mol
fraction as a function of the molecular weight. The molecular
weights used are the generalized values presented for each SCN
(Katz and Firoozabadi 1978). Whitson (1983) noted some discrep-
ancies of the molecular weight between the plots and the tables
presented by Katz et al. (1978).
SCN groups contain hundreds of isomers/components with the
same number of carbon atoms. Because of the uncertainty of the
isomers/components present in each SCN group, it is not possible
to set a unique value on the molecular weight. The aim of this work
is to find a new approach to characterize the undetermined fraction
by splitting the Cn+ fraction into a representative number of SCNs,
calculating their mole fractions and molecular weights.
Generalized correlations have been developed to generate criti-
cal properties for the SCN including those by Sancet (2007), Twu
(1984), Ahmed (1985), Kesler-Lee (1976), Riazi-Daubert (1987),
and Edmister (1958). Some of these correlations are tested with
heavy-oil samples in this work.
Lumping schemes such as those of Whitson (1983) and Lee et al.
(1981) are also tested.
Theory
Whitson’s Method To Estimate the Mole Fraction Distribution.
Whitson (1983) used TPG distribution to estimate the mole fraction
of the SCNs within the Cn+ fraction. TPG is defined (Whitson 1983)
as a function of the molecular weight (M) by Eq. 1:
⎛ M −⎞
( M −  )−1 exp ⎜ −
⎝  ⎟⎠
p( M ) = , . . . . . . . . . . . . . . . . . . . (1)
  ( )

where  is the minimum molecular weight present in the Cn+
fraction,  is used to fit the shape of the distribution, and  is the
gamma function. Whitson recommended  = 92, the molecular
weight of toluene, as a good estimation of , if C7+ is the plus
fraction; for other plus fractions, Eq. 2 is used as approximation.
 = 14 n − 6 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
and
M Cn + − 
= , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)

where MCn+ is the molecular weight of the Cn+ fraction.
The mole fraction of an SCN is then generated by calculation
of the cumulative frequency of occurrence between the limits Mi−1
and Mi multiplied by the mole fraction of the Cn+ fraction.
283
 Mi
z ( i ) = zCn + ∫ p ( x )dx ,
Mi −1
where i is the SCN and Mi is the molecular weight of the SCN usu-
ally defined as the molecular weight of the normal component.
Correlations Used To Estimate Physicochemical Properties of
the SCN. Several correlations have been developed to estimate
the physical properties of petroleum fractions and are in principal
function of the specific gravity (), the boiling point temperature
(Tb), and the molecular weight (M). The studied correlations are
the correlations proposed by: Sancet (2007), Twu (1984), Ahmed
(1985), Kesler-Lee (1976), and Edmister (1958); their formulations
are shown in Appendix A.
Lumping Methods. The generation of an appropriate mole dis-
tribution that represents the plus fraction usually requires splitting
the plus fraction into large number of components (SCN) with
their respective mole fractions and physical properties. The use of
a large number of components in a reservoir simulation is time con-
suming. Lumping is defined by Ahmed (2007) as the reduction of
the number of components used in EOS calculations for reservoir
fluids. Pedersen et al. (2006) states that lumping consists of
• Deciding what carbon number fractions are to be lumped into
same pseudocomponent.
• Deciding the mixing rules that will average Tc, Pc,  of the
individual carbon number fractions to one Tc, Pc,  to represent
the lumped pseudocomponents.
Two lumping methods are tested in this paper, Whitson’s lump-
ing scheme and Lee’s lumping scheme. The formulations for the
lumping methods tested are shown in Appendix B.
Developed Method in This Work
The developed method here is based on a modification of Whitson’s
approach. The approach is based on the fact that SCN groups may
represent hundreds of different compounds with the same number
of carbon atoms. These compounds can be classified in different
types: paraffins, naphthenes, and aromatics. Uncertainty of the
types and distribution of compounds present in each SCN indicates
that the average value of molecular weight of an SCN is different
from one sample to another. The modification of Whitson’s method
includes a simple procedure to determine the value of the fitting
parameter  and a new definition of the limits used to calculate
the frequency of occurrence for each SCN.
Determination of ␣.  is the fitting parameter, which defines the
shape of the mole fraction distribution. The left skewed mole distri-
bution characteristic of heavy-oil samples means that the value of the
fitting parameter  is higher than 1 for a characterization with TPG
Mole Fraction and TPG Distribution
Fig. 1—Best trend fit generated between the three-parameter function with ␣ = 1.7 and experimental mole fraction of Sample 2.
The maximum values are found at Mw = 263, TPG = 0.001659, and experimental mole fraction = 0.0307.
284
and Whitson’s method. Different values of  higher than 1 are tested
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4) in order to find the best trend fit between TPG and the experimental
mole distribution. The best trend fit is found when the molecular
weight of the maximum mole fraction of the experimental data [gas
chromotography (GC)] and the molecular weight of the maximum
value of the TPG have the same value as shown in Fig. 1.
New Definition of the Limits for the Cumulative Frequency of
Occurrence of SCN. The new limits used to calculate the cumula-
tive frequency of occurrence are defined by the best fit between
TPG and the experimental mole fraction (zexp) from GC analysis.
Characterization of a Cn+ fraction is based on the relation between
the TPG and zexp of the previous carbon numbers (e.g., C7 through
C29, if C30+ fraction is characterized). At least two previous carbon
numbers and their respective mole fractions must be known to
use this approach. The relationship between TPG and zexp is found
by iteration and presented as a function of the limiting molecular
weight (LMw) vs. carbon number.
Once the TPG has been generated, it is divided into SCNs with
the new definition of the limits by the introduction of the limiting
molecular weight (LMw), which is the highest molecular weight
included in the SCN; Eq. 4 is then transformed to
LMw
z char (C7 ) = zC7+ ∫ p ( M ) dM , . . . . . . . . . . . . . . . . . . . . . . . . (5)
Mwmin
for the first carbon number characterized; for example C7 is used if
the relation between TPG and zexp is based on carbon numbers from
C7 to Cn−1. The use of the C7+ fraction is arbitrary; any other plus
fraction can be used, however it is recommended to generate as
many values of LMw function of SCN as possible. The integration
should have dM small enough to better fit the distribution curve.
M wmin , in Eq. 5, is the minimum molecular weight present in the
C7+ fraction. Whitson (1983) presented the following approxima-
tion to calculate the minimum molecular weight: M wmin = 14 ∗ n − 6,
where n is the carbon number; giving M wmin = 92 for n = 7.  is
defined as the highest molecular weight present in the previous
carbon number, C6, since the probability of occurrence in C7 is
zero.  may be expressed as:  = M wmin − dM .
LMw is calculated by best fit between zchar(C7) and zexp(C7) with
the following iterative process:
1. zchar(C7) is calculated with the first estimation of the limiting
molecular weight: LMw = M wmin + dM
2. If zchar(C7) is smaller than zexp(C7), then LMw = LMw + dM , the
limiting molecular weight is updated for each iteration.
3. The process is repeated until zchar(C7) ≥ zexp(C7)
4. The last LMw used in the process is assigned to C7.
The process to characterize carbon numbers higher than seven
is analogous to the iterative process presented above. The inferior
1.E–01
Exp. Mole frac = 0.0307
TPG distribution with ∝ = 1.7
1.E–02 Experimental mole fraction, Sample 2
Mw = 263
TPG max = 0.001659
1.E–03
1.E–04
1.E–05
0 500 1000 1500 2000 2500
Mw
April 2010 SPE Reservoir Evaluation & Engineering
 1.E–01

Mole Fraction and TPG Distribution

1.E–02

1.E–03

1.E–04 TPG distribution with ∝ = 1.7

Experimental mole fraction, Sample 2
Estimated mole fraction, Sample 2
1.E–05
0 200 400 600 800 1000
Mw

Fig. 2—TPG distribution, experimental, and estimated mole fractions of SCN for Sample 2. Experimental and estimated mole frac-
tions of SCN show similar values; molecular weight of SCN defined by the method proposed is slightly different from generalized
values of molecular weight presented in literature.

1000
Limiting Molecular Weight, LMw

900
800
700
600
500
400
300
200 LMw of Sample 2
100 Linear extrapolation of LMw

0
0 10 20 30 40 50
Cn

Fig. 3—LMw as a function of SCN for Sample 2. Linear extrapolation of LMw to generate SCN of the C36+ fraction.

limit for SCN(i) becomes then LMw ( i − 1) + dM , for i = 8,9,…Cn−1.
The limiting molecular weight (LMw ) as a function of the carbon
number can be generated for carbon numbers between 7 and Cn−1.
The process is illustrated in Fig. 2.
The approach proposed requires defining the limiting molecu-
lar weights LMw that will be used to split the Cn+ fraction into
SCNs. The function LMw vs. SCN, calculated previously, is
extrapolated in order to adjust the molecular weight of the plus
fractions. Fig. 3 presents a linear extrapolation of this function
to determine the limiting molecular weights of the SCN present
in the Cn+ fraction.
After the limiting molecular weight function has been extrapo-
lated, the mole fraction vs. carbon number distribution is generated
and presented in Fig. 4.
Results and Discussion
The uncertainty associated with the molecular weight assigned to
each single carbon number was a motivation to develop this method.

0.04
Experimental mole fraction
Estimated mole fraction
0.03
Mole Fraction

0.02

0.01

0.00
0 50 100 150 200 250
Cn

Fig. 4—Mole fraction estimated with the method proposed for Sample 2.

April 2010 SPE Reservoir Evaluation & Engineering 285

 0.06
Experimental mole fraction
0.05 Estimated mole fraction

Mole Fraction
0.04

0.03

0.02

0.01

0.00
0 10 20 30 40 50 60 70 80 90 100
Cn

Fig. 5—Experimental and estimated mole fraction for Sample 1.

TABLE 1—SENSITIVITY OF ΔM IN Mw CALCULATIONS FOR TABLE 2—PROPERTIES OF THE USED SAMPLES

SAMPLE 1
GC ΔM = 0.01 ΔM = 0.1 ΔM = 1.0 Sample API Gravity C7 mole% MC7+

Mw 454.54 453.9881 453.4949 453.7935 2 10 79.07 527.9

mole% 100 99.95 99.91 99.93 3 14 85.60 438.8
last Cn 99 98 94 75 4 18 98.47 321.8
5 27.3 58.45 441.8
6 21.7 67.80 495.9
The developed method has been tested for data of six samples
obtained from the literature and by personal communications.
smaller the M, the better the fit between the characterized and
Sample 1. The method in this paper is illustrated for data from experimental mole fraction, according to Eq. 5 (i.e., when Sample
Sample 1, which consists of an extended compositional analysis 1 is characterized with M = 1.0, and M = 0.01 the last Cn are
to C100+. The provided data presented as mole fraction for each 75 and 98, respectively).
carbon number. The experimental and estimated mole fractions
are shown in Fig. 5. Verification of the Method. Five additional samples of heavy
The developed method is sensitive to the interval of discretiza- oil are used to validate the method presented in this paper. The
tion for the molecular weight ( M), as expected. Table 1 shows a properties of Samples 2 through 6 are listed in Table 2. The data
sensitivity study of M as a function of the molecular weight, the for Samples 1, 3, 5, and 6 are taken from operated fields in Asia
mole fraction, and the number of SCN generated for Sample 1 in and South America. Samples 2 and 4 are from Krejbjerg and Ped-
comparison to the experimental data. The shape parameter () for ersen (2006). Table 2 shows the API gravity, the mole percent, and
sample 1 is calculated by the best-trend-fit procedure explained in molecular weight of C7+ fraction.
the section Developed Method in This Work and has a value of Two parameters are kept constant for all the studied samples;
 = 2.3 for this composition. Fig. 6 shows the molecular weight the first is the carbon number to be characterized n = 7 and the
of each SCN for Sample 1. The molecular weights are calculated minimum molecular weight present in the fraction. Table 3 pres-
by the developed approach and are presented as three functions ents the generated mole fractions, molecular weights for the plus
of molecular weight vs. SCN for values of M shown in Table 1. fractions for Samples 2 through 6, the calculated shape parameter
As shown, SCN values are affected by the magnitude of M. The (), and the last carbon number.

1400

1200
Molecular Weight

1000

800

600

400 Δ M = 0.01
Δ M = 0.1
200 ΔM=1

0
0 10 20 30 40 50 60 70 80 90 100
Cn

Fig. 6—Sensitivity of the discretization interval of the molecular weight for Sample 1.

286 April 2010 SPE Reservoir Evaluation & Engineering

 TABLE 3—RESULTS FOR SAMPLES 2 THROUGH 6
Shape Last SCN
Sample Experimental Estimated n for Cn+ Fraction Parameter α Generated

mole frac C7+ 0.7907 0.7907

MC7+ 527.93 527.69
2 36 1.7 211
mole frac Cn+ 0.2517 0.2505
MCn+ 1038.1 918.58
mole frac C7+ 0.8560 0.8560
MC7+ 438.79 438.51
3 30 1.4 158
mole frac Cn+ 0.2816 0.2801
MCn+ 849.02 770.48
mole frac C7+ 0.9847 0.9847
MC7+ 321.83 321.73
4 41 1.4 313
mole frac Cn+ 0.1503 0.1478
MCn+ 761 684.56
mole frac C7+ 0.5845 0.5845
MC7+ 441.77 441.43
5 30 1.5 189
mole frac Cn+ 0.2835 0.2823
MCn+ 655 658.29
mole frac C7+ 0.6780 0.6780
MC7+ 495.95 495.64
6 30 1.1 318
mole frac Cn+ 0.3863 0.3848
MCn+ 696 717.05

The calculated molecular weights for carbon numbers between C7
and Cn by the proposed method as a function of carbon number are
shown in Fig. 7. Fig. 7 also compares the obtained results with the
other correlations. The molecular weight of a specific SCN differs
from one sample to another, as expected. The variation of the molecu-
lar weight of SCN is small for low carbon numbers; for example,
the molecular weight of C7 ranges from 95.86 (Sample 6) to 102.16
(Sample 5), which is approximately 6% calculated from the studied
samples, while for C29, molecular weight ranges from 326.33 (Sample
6) to 505.79 (Sample 2), which is approximately 35%.
Estimation of Critical Properties for SCNs. The physical proper-
ties for each SCN are generated with correlations that are primary
function of molecular weight, boiling temperature, specific gravity,
and carbon number. These correlations are originally developed to
represent properties of Cn fractions with a low number of carbon
atoms, n.
Prediction of physical properties for heavy-oil samples (with
high carbon number) demonstrates the range in which different
correlations could be used.
In general, Samples 2 through 6 show similar results for the
limitation of the different correlations. Therefore, only the results
from Sample 2 are shown to illustrate the observations obtained
by the correlations for the following physical properties: critical
pressure, critical temperature, boiling temperature, acentric factor,
and specific gravity.
Critical Pressure. Several methods have been studied. Critical
pressure is calculated with four different correlations. Table 4
shows the properties needed to calculate critical pressure with
the correlations studied. The methods used for calculation of the
properties are also included in Table 4.
The calculated values of critical pressure are presented as a
function of carbon number and shown in Fig. 8. Several regions
are identified in Fig. 8 to determine the applicability carbon num-
ber range.
Region I can be defined as the carbon numbers lower than 20;
in this region, it is possible to see that all the correlations follow
a similar trend. However, Kesler and Lee and Ahmed correlations
deviate significantly from the other correlations at carbon numbers
7 and 20, respectively.
Region II includes carbon numbers between 20 and 60; this
region shows clearly that the values predicted by the studied corre-
lations follow a similar trend except for the correlation by Ahmed,
as shown in Fig. 8. The values predicted by the studied correlations
show significant differences, which makes it difficult to identify the
correct critical pressure without experimental verification.

1400
Mw Katz and Firoozabadi
1200 Mw Whitson SPE
Mw Ahmed correlation
Mw Sample 1
Molecular Weight

1000 Mw Sample 2
Mw Sample 3
Mw Sample 4
800 Mw Sample 5
Mw Sample 6
600

400

200

0
0 10 20 30 40 50 60 70 80 90 100

Fig. 7—Molecular weight calculated as a function of the SCN by this method and other correlations.

April 2010 SPE Reservoir Evaluation & Engineering 287

 TABLE 4—DEPENDENCY OF THE CORRELATIONS FOR TABLE 5—DEPENDENCY OF THE CORRELATIONS FOR
CRITICAL PRESSURE CRITICAL TEMPERATURE
Origin of the Origin of the
Correlation Functional Variable Functional Variable Correlation Functional Variable Functional Variable

Sancet Mw Method developed Sancet Mw Method developed

Ahmed Cn – Ahmed Cn –
Kesler-Lee SG and Tb Logarithmic and Twu Lee-Kesler SG and Tb Logarithmic and Twu
correlations correlations
Twu Tb and Tc Twu correlation Twu Tb Twu correlation

Finally in Region III, critical pressures predicted for carbon
numbers higher than 60 are approaching asymptotes. None of the
three correlations in Region III could be identified to be superior
to the others.
Critical Temperature. Critical temperature is obtained as a
function of the carbon number with four correlations. The correla-
tions and the origin of the functional variables used to calculate
the critical temperature are shown in Table 5.
All of the correlations used to predict critical temperature have
practically the same results for carbon numbers lower than 40, as
shown in Fig. 9. Correlation by Ahmed does not seem to be valid
for carbon numbers higher than 40 or 60. The Sancet and Kesler
and Lee correlations predict practically equal values for all the
carbon numbers studied, giving an advantage to Sancet because
it is just a function of the molecular weight. The Twu correlation
underpredicts when compared to Sancet and Kesler and Lee for
carbon numbers above 40.
Boiling Point Temperature. Table 6 shows the used boiling-
point-temperature correlations and the used functional variables
for each correlation method.
The Sancet, Twu, and Ahmed correlations are used to calculate
boiling point temperature. Fig. 10 shows the physical property as
a function of the carbon number. Boiling-point-temperature cor-
relations show similar predictions for carbon numbers lower than
30. The Sancet and Twu correlations show a similar trend when
the boiling point temperature is calculated for a carbon number
higher than 30. The Twu correlation underpredicts the physical
property if compared to Sancet. The values calculated with the

500 35
Region II Sancet
Ahmed 30
Critical Pressure, psia

400

Critical Pressure, bara

Kesler-Lee
25
Twu
300
20
Region III
200 15

10
100
5
Region I
0 0
0 10 20 30 40 50 60 70 80 90 100
Cn

Fig. 8—Critical pressure calculated with different correlations of SCNs for Sample 2.

2400

1200
2000
Critical Temperature, K
Critical Temperature, R

1000
1600
800
1200
600
800 Sancet
400
Ahmed
400 Kesler-Lee
200
Twu
0 0
0 10 20 30 40 50 60 70 80 90 100
Cn

Fig. 9—Critical temperature calculated with different correlations of SCNs for Sample 2.

288 April 2010 SPE Reservoir Evaluation & Engineering

 TABLE 6—DEPENDENCY OF THE CORRELATIONS FOR TABLE 7—DEPENDENCY OF THE CORRELATIONS FOR
BOILING-POINT TEMPERATURE ACENTRIC FACTOR
Origin of the Origin of the
Correlation Functional Variable Functional Variable Correlation Functional Variables Functional Variable

Sancet Mw Method developed Edmister Pc, Tc, and Tb Sancet

Ahmed Cn – Edmister Pc, Tc, and Tb Ahmed
Twu Mw Method developed Ahmed Cn –
Kesler-Lee Pc, Tc, and Tb Lee-Kesler, Twu
Mw and SG Method developed
Ahmed correlation indicate that boiling-point temperatures for logarithmic
carbon numbers higher than 60 do not follow the same trend as
for the other two correlations.
Acentric Factor. Table 7 shows the correlations used to calcu- specific gravity decrease, which does not seem to be logical, so a
late the acentric factor for each SCN. Correlations for acentric fac- logarithmic correlation of the specific gravity vs. carbon number
tor usually depend on some physical properties; this dependence is is generated from different literature values.
shown together with the origin of functional variable in Table 7. The logarithmic correlation is:
All the correlations used present almost similar predictions for
carbon numbers lower than 20, as shown in Fig. 11. For carbon SG = 0.1129 ln ( Cn ) + 0.5203 . . . . . . . . . . . . . . . . . . . . . . . . . (6)
numbers between 20 and 40, it is observed that Edmister cor-
relation with critical properties from the Ahmed and Kesler-Lee Summary of the Observed Limits of the Correlations Used To
correlations show similar predictions. Carbon numbers higher than Generate Physical Properties. The observations identified for each
40 may be predicted by the Kesler-Lee correlation. correlation and physical property are used to determine the ranges
Specific Gravity. Specific gravity is needed to generate critical where the correlations may be used to predict a specific property,
properties with Kesler-Lee. The only correlation available for cal- as shown in Table 8.
culating specific gravity as a function of SCN, to our knowledge,
is Ahmed’s correlation. The function presented by Ahmed exhib- Lumping of SCNs Into Pseudocomponents. Two lumping
its a maximum value at Cn = 70, at a higher carbon number the schemes, from Whitson and Lee, are studied with Samples 2

2400
Boiling Point Temperature, R

Boiling Point Temperature, K

1200
2000
1000
1600
800
1200
600

800
400
Sancet
400 Ahmed 200
Twu
0 0
0 10 20 30 40 50 60 70 80 90 100
Cn

Fig. 10—Boiling-point temperature calculated with different correlations for SCNs of Sample 2.

3 Edmister with Sancet's Pc, Tc, Tb

Acentric Factor

Edmister with Ahmed's Pc, Tc, Tb

Ahmed correlation
Kesler-Lee correlation
2

0
0 10 20 30 40 50 60 70 80 90 100
Cn

Fig. 11—Acentric factor calculated with different correlations for SCNs of Sample 2.

April 2010 SPE Reservoir Evaluation & Engineering 289

 TABLE 8—SUMMARY OF THE APPLICABLE RANGES FOR THE DIFFERENT
CORRELATIONS

Range of Carbon Numbers and Correlations Valid

Critical Critical Boiling Acentric Specific

Correlation Pressure Temp Point Temp Factor Gravity

Sancet 7–Inf 7–Inf 7–Inf – –

Ahmed 7–60* 7–40 7–40 7–20 7–60
Kesler-Lee 7–Inf* 7–Inf – 7–Inf –
Twu 7–Inf** 7–Inf 7–Inf – –
Edmister/Sancet – – – 7–20 –
†
Edmister/Ahmed – – – 7–60 –
Logarithmic – – – – 7–Inf
* Correlation shows significant difference around carbon number 7 and 20 in comparison with other studied correlations.
** Correlation is valid while the property Tb used is smaller than Tc: this occurs at approximately carbon number 150,
which is the limit for Twu correlation of critical pressure.
†
Correlation might be valid up to carbon number 50.

through 6. Sample 1 is not included in the lumping analysis mij

because of a lack of data, such as the molecular weight, which is
necessary to characterize the fraction and generate reliable predic-
tions using equations of state.
The number of pseudocomponents and the limiting molecular
weight are calculated using Whitson’s method, Eqs. B-1 and B-2,
respectively. The mole fraction of each pseudocomponent is calcu-
lated simply by adding the mole fractions of all the SCN within the
pseudocomponents’ carbon number limits as defined by Whitson’s
method.
Lee’s lumping scheme is also used to group the SCNs into
a defined number of pseudocomponents. The physicochemical
properties used in this method are critical temperature and critical
pressure generated with Sancet correlations, and the acentric factor
() generated with Kesler-Lee correlation.
The absolute rate of change is calculated for all of the properties
at each boiling-point temperature. The absolute rate of change is
introduced here as the absolute value of mij, which is defined by
Lee’s lumping scheme as the slope of the physicochemical proper-
ties with respect to the boiling point temperature. Absolute values
of mij are used, otherwise the obtained critical pressure has negative
slopes. The rates of changes are defined as follows:
⎡ ( Pc )i ⎤ ⎡ ( Tc )i ⎤ ⎡ ( )i ⎤
mi1 = abs ⎢ ⎥ , mi 2 = abs ⎢ ⎥ and mi 3 = abs ⎢ ⎥,
⎢⎣ ( b )i ⎥⎦
T ⎢⎣ ( b )i ⎥⎦
T ⎢⎣ ( Tb )i ⎥⎦
. . . . . . . . . . . . . . . . . . . . . . . . (7)
m ij = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)
mij max i =1,...,n f
The weighted sum of the normalized absolute rate of change Mi
of the three properties considered: Pc, Tc, and  is calculated by:
3
∑m ij
j =1
Mi = .
3
Lee’s lumping scheme suggests that SCNs with similar values
of the normalized absolute rate of change ( Mi ) should be lumped
together in a pseudocomponent. For this work, similar values of Mi
are defined as values with a maximum deviation of 0.1; for exam-
ple the first pseudocomponent lumps all SCNs with a value of Mi
higher than 0.8. SCNs with a value of Mi between 0.7 and 0.8 are
lumped in Pseudocomponent 2, and so on. Fig. 13 shows the values
of Mi for Sample 2. Mi is plotted vs. the boiling temperature (Tb)
and the carbon number (Cn). Mi vs. Tb function helps to identify
the limiting Tb used to lump SCNs based on Lee’s lumping scheme.
Mi vs. Cn shows the carbon numbers that are lumped together; for
example, the first SCN in the last pseudocomponent SCN = 56
corresponds to the boiling-point temperature Tb = 1721.33 [R], as
shown in Fig. 13.
The criterion used here to lump the SCNs is the increment of
0.1 unit of the absolute rate of change ( Mi ) in the properties of
the lumped SCN. However, any higher increment unit could be
selected to reduce the number of the pseudocomponents as long

1.2

1.0
Specific Gravity

0.8

0.6

0.4

0.2 Ahmed correlation

Logarithmic correlation

0.0
0 10 20 30 40 50 60 70 80 90 100
Cn

Fig. 12—SG using Ahmed’s correlation and the logarithmic correlation of the SG data from Ahmed’s correlation and other several
samples from literature.

290 April 2010 SPE Reservoir Evaluation & Engineering

 Tb, R
657.8 857.8 1057.8 1257.8 1457.8 1657.8 1857.8 2057.8 2257.8 2457.8
1.0
Ps-comp 1
0.9
Ps-comp 2
0.8
Ps-comp 3
0.7
Mi vs Cn Ps-comp 4
0.6 Mi vs Tb
Mi Ps-comp 5
0.5
Ps-comp 6
0.4
Ps-comp 7
0.3
0.2
0.1
0.0
7 27 47 67 87 107 127 147 167 187 207
Cn

Fig. 13—Weighted sum of absolute rate of change (M i ) as function of Tb and Cn for Sample 2.

3, 5, and 6. PVT properties of Samples 1 and 4 are not generated

TABLE 9—RELATIONSHIP BETWEEN NUMBER OF SCN
AND NUMBER OF PSEUDOCOMPONENTS FOR WHITSON
because experimental data is not available for comparison. The
AND LEE METHODS properties calculated are formation volume factor (Bo), gas solubil-
ity ratio (Rs), and density () as functions of pressure, at reservoir
Number of Pseudocomponents temperature. Saturation pressure is also calculated.
Sample Number of SCN Whitson Lee Table 10 compares the experimental and predicted values
of saturation pressure. The error produced in the prediction of
2 205 9 7 saturation pressure is different for all the samples, from 2% for
3 152 8 7 Sample 3 to 49% for Sample 2. The predictions presented have
4 307 9 7 been calculated using values of critical properties estimated by
the characterization process. No matching has been performed
5 183 8 7
enforced in this calculation, which would have improved the
6 312 9 7 recorded deviation from the experimental data.
Predictions of formation volume factor (Bo) show the following
as the accuracy/sensitivity change of the properties of the lumped errors with respect to the experimental values: 0.5, 5.3, 2.8, and
components are acceptable. 0.5% for Samples 2, 3, 5, and 6, respectively. Predicted solubil-
The number of pseudocomponents generated with the lump- ity ratio (Rs) shows differences with respect to the experimental
ing methods developed by Whitson and Lee is compared with values of 21.1, 47.7 , 24.5, and 13.3% for Samples 2, 3, 5, and
the number of SCNs characterized by the method proposed in 6, respectively. The error calculated between experimental and
this paper. Table 9 shows the relationship between the number of simulated density () is 1.2, 27.6, 23, and 10%, for Samples 2, 3,
pseudocomponents generated vs. the number of SCNs for all the 5, and 6, respectively.
studied samples. Figs. 14 and 15 show the upper limiting carbon The properties Bo, Rs, and  have been generated with PR-EOS.
number for each of the generated pseudocomponents with the The critical properties used with the PR-EOS model are generated
Whitson and Lee approaches, respectively. for each SCN with the following correlations: Pc, Tc, and Tb with
Sancet correlations; acentric factor () with Kesler-Lee correla-
Predictions of PVT Properties With PR–EOS. A commercial tion; molecular weight with the method proposed in the section
simulator, PVTP, is used to predict PVT properties of Samples 2, Developed Method in This Work; and SG with the logarithmic

10
9
8 Sample 2
Pseudocomponent

7 Sample 3
Sample 4
6 Sample 5
5 Sample 6

4
3
2
1
0
0 50 100 150 200 250 300 350
Cn

Fig. 14—Number of pseudocomponents generated by Whitson’s lumping scheme and their correspondent upper SCNs.

April 2010 SPE Reservoir Evaluation & Engineering 291

 10
Sample 2
9 Sample 3
8 Sample 4

Pseudocomponent
Sample 5
7 Sample 6
6
5
4
3
2
1
0
0 50 100 150 200 250 300 350
Cn

Fig. 15—Number of pseudocomponents generated by Lee’s lumping scheme and their correspondent SCNs.

extrapolation of Ahmed’s correlation. The SCNs are lumped in
pseudocomponents with Whitson’s method, and the properties for
the pseudocomponents calculated with Hong’s mixing rules.
The properties calculated using the PR-EOS are strongly
dependent on the critical properties defined for each component
of the mixture and the mixing rules used. The correlations used to
generate critical properties of the samples studied are originally
developed for light oil.
The estimated deviation from experimental data varies for the
estimated saturation pressure from approximately 49 to 2%, which
may be sample specific rather than the method used; however,
experimental work is needed for developing critical properties
correlations.
Conclusions
We developed a method to characterize plus fractions of heavy oil.
The reviewed heavy-oil fluids show a left-skewed distribution of
the mole fraction as a function of carbon number.
We based the developed method on the fact that molecular
weight is not uniquely related to carbon numbers because of the
hidden exponential increase of number of isomers/components
with increasing the carbon numbers. In the method, TPG distri-
bution from Whitson’s approach is used to characterize the plus
fraction and generate molecular-weight/carbon-number function
for each sample by fitting the TPG distribution to the experimental
mole fraction.
The developed method was applied to five different heavy-oil
samples. The predicted molecular weight is shown to be close to
the generalized molecular weight associated with carbon number;
however, these differ from one sample to another, which, perhaps,
reflects the difference in the molecular weight for each SCN.
Whitson’s and Lee’s methods are used to lump SCNs gener-
ated for Samples 2 through 5 into pseudocomponents. Whitson’s
lumping scheme approach produced nine pseudocomponents for
Samples 2, 4, and 6, while for Samples 3 and 5, the number of
the pseudocomponents generated by Whitson’s approach is eight.
Even though the number of components are reduced significantly
with Whitson’s lumping scheme, the use of eight or nine pseudo-
components to define the C7+ fraction, in addition to the number
of components needed to describe the well defined fractions: C1 to
C6, requires long computation time to run compositional reservoir
simulation.
The properties generated with PR-EOS show differences with
the experimental data; the error varies from approximately 2 to
47% for predictions of saturation pressure at reservoir tempera-
ture. Predicted Bo values show an error between 0.5 and 5.3%, Rs
between 13 and 47%; and density  between 1.2 and 27%. The
critical properties used in PR-EOS are generated with correlations
developed to predict critical properties for light oil. However, the
deviation from the experimental data seems to be related to sample
specific rather than to the method used.
Nomenclature
Cn = carbon number
ki,l = binary interaction coefficient between component i and l
Kw = Watson characterization factor
LMw = limiting molecular weight, lbm/lbm mol, kg/kg mol
mij = slope of the property j vs. the boiling point at component i
M = molecular weight, lbm/lbm mol, kg/kg mol
Mi = averaged change of physicochemical properties
Mp = molecular weight of normal paraffins, lbm/lbm mol, kg/
kg mol
n = number of carbon atoms of the first SCN in the plus frac-
tion
nf = number of SCNs in the plus fraction
np = number of physicochemical properties
N = number of carbon atoms of the last SCN characterized
Ng = number of groups
Pc = critical pressure, psia, bara
Pcp = critical pressure of normal paraffins, psia, bara
T = temperature °F, °C
Tb = boiling point temperature, °R, °K

TABLE 10—SATURATION PRESSURE CALCULATED WITH PR-EOS AT RESERVOIR

TEMPERATURE

Reservoir Calculated Experimental

Sample Temperature (°F) Psat (psia) Psat (psia) Error (%)

2 125.6 527.704 1037 49.1

3 205.6 301.569 295 2.2
5 145.76 1103.61 1375 19.7
6 151.88 870.39 1060 17.9

292 April 2010 SPE Reservoir Evaluation & Engineering

 Tc = critical temperature, °R, °K Pedersen, K.S., Thomassen, P., and Fredenslund, A.A. 1982. Phase Equi-
Tcp = critical temperature of normal paraffins, °R, °K libria and Separation Process. Report SEP 8207, Institute of Kemitek-
wi* = weight average fraction of component i nik, Denmark Tekniske Hojskole, Lyngby, Denmark.
z(i) = mole fraction of the carbon number i Peng, D.-Y. and Robinson D.B. 1976. A new two-constant equation of state.
Ind. Eng. Chem. Fund. 15 (1): 59–64. doi: 10.1021/i160057a011.
zchar(i) = characterized mole fraction of carbon number i
Riazi, M.R. and Daubert, T.E. 1987. Characterization parameters for
zexp(i) = experimental mole fraction of carbon number i petroleum fractions. Ind. Eng. Chem. Res. 26 (4): 755–759. doi:
zi* = normalized mole fraction of component i 10.1021/ie00064a023.
 = shape parameter of the three parameter gamma function Twu, C.H. 1984. An Internally Consistent Correlation for Predicting the
M = discretization interval of molecular weight, lbm/lbm mol, Critical Properties and Molecular Weight of Petroleum and Coal-tar
kg/kg mol Liquids. Fluid Phase Equilibria 16: 137–150.
 = specific gravity defined with pw=100 gm/cm3 Whitson, C.H. 1983. Characterizing Hydrocarbon Plus Fractions. SPE J.
 = gamma function 23 (4): 683–693. SPE-12233-PA. doi: 10.2118/12233-PA.
 = minimum molecular weight in the three-parameter gamma
function, lbm/lbm mol, kg/kg mol Appendix A—Correlations Used To
 = acentric factor Calculate Critical Properties
 = density, lb/ft3, gm/cm3 Sancet Correlations. Sancet (2007) presented a set of correla-
tions to predict values of critical properties Tc and Pc for the C7+
Subscripts fraction as function of the molecular weight. Sancet stated that
c = critical the correlations proposed give excellent liquid density predictions
when there are high-molecular-weight hydrocarbons present in the
i,l = component number
fluid. The correlations are:
j = physicochemical property
L = lumped fraction
Pc [ psia ] = 82.82 + 653 ∗ e −0.007427 M , . . . . . . . . . . . . . . . . . . (A-1)
p = normal paraffins
Tc [ R ] = −778.5 + 383.5 ln ( M − 4.075) , . . . . . . . . . . . . . . (A-2)
Acknowledgment
The authors wish to acknowledge StatoilHydro for its financial and
support to the University of Stavanger.
Tb [ R ] = 194 + 0.001241( Tc [ R ])
1.869
. . . . . . . . . . . . . . . . . . (A-3)
References
Ahmed, T. 2007. Equations of State and PVT Analysis, 59–61, 99–107,
112–127. Houston: Gulf Publishing Company. Twu Correlations. Twu (1984) presented correlations to calcu-
Ahmed, T., Cady, G., and Story, A. 1985. A Generalized Correlation for late critical pressure (Pcp) and critical temperature (Tcp) of normal
Characterizing the Hydrocarbon Heavy Fractions. Paper SPE 14266 paraffins. The critical properties Tcp and Pcp are calculated as
presented at the SPE Annual Technical Conference and Exhibition, Las functions of the boiling point temperature (Tb) with the following
Vegas, Nevada, USA, 22–25 September. doi: 10.2118/14266-MS. expressions.
Danesh, A. 1998. PVT and Phase Behavior of Petroleum Reservoir Fluids,
No. 47, 130, 364–367. Amsterdam: Developments in Petroleum Sci- −1
ence, Elsevier Science B.V. ⎡ 0.533272 + 0.191017 ∗ 10 −3 Tb ⎤
Edmister, W.C. 1958. Applied Hydrocarbon Thermodynamics Part 4.
⎢ ⎥
⎢ −7 2 ⎥
Compressibility Factors and Equations of State. Petroleum Refiner ⎢ +0.779681 ∗ 10 Tb ... ⎥
37: 173–179. Tcp = Tb ⎢ ⎥ , . . . . . (A-4)
Hong, K.C. 1982. Lumped-Component Characterization of Crude Oils ⎢ 0.959468 ∗ 10 2 ⎥
for Compositional Simulation. Paper SPE 10691 presented at the ⎢ −0.284376 ∗ 10 −10 Tb3 + 13 ⎥
SPE/DOE Enhanced Oil Recovery Symposium, Tulsa, 4–7 April. doi: ⎢ ⎛ Tb ⎞ ⎥
10.2118/10691-MS.
⎢⎣ ⎝⎜ 100 ⎠⎟ ⎥⎦
Katz, D.L. 1983. Overview of Phase Behavior of Oil and Gas Production.
2
J. Pet Tech 35 (6): 1205–1214. SPE-9995-PA. doi: 10.2118/9995-PA. ⎛ 1
⎞
Katz, D.L. and Firoozabadi, A. 1978. Predicting Phase Behavior of Con- Pcp = ⎜ 3. 83354 + 1. 19629
2
+ 34.8888
⎟ , . . . . . . . . . . (A-5)
densate/Crude-Oil Systems Using Methane Interaction Coefficients. ⎜⎝ +36.1952
2 + 104.193
4 ⎟⎠
J. Pet Tech 30 (11): 1649–1655. SPE-6721-PA. doi: 10.2118/6721-PA.
Kay, W.B. 1936. Gases and Vapors At High Temperature and Pres-
sure—Density of Hydrocarbon. Ind. Eng. Chem. 28 (9): 1014–1019. where
doi: 10.1021/ie50321a008.
Kesler, M.G. and Lee, B.I. 1976. Improve Prediction of Enthalpy of Frac- Tb
tions. Hydrocarbon Processing 55: 153–158.
= 1− . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-6)
Tcp
Krejbjerg, K. and Pedersen, K.S. 2006. Controlling VLLE Equilibrium with
a Cubic EoS in Heavy Oil Modeling. Paper 2006-052 presented at the
Canadian International Petroleum Conference. Calgary, 13–15 June. The molecular weight of paraffins is given by the implicit
Lee, S.T., Jacoby, R.H., and Chen, W.E. 1981. Experimental and Theoretical expression presented in Eq. A-7.
Studies on the Fluid Properties Required for Simulation of Thermal Proc-
esses. SPE J. 21 (5): 535–550. SPE-8393-PA. doi: 10.2118/8393-PA. ⎡ ⎤
Lohrenz, J., Bray, B.G., and Clark, C.R. 1964. Calculating Viscosities ⎢5.71419 + 2.71579 ln M ⎥
of Reservoir Fluids From Their Compositions. J. Pet Tech 16 (10): ⎢ p ⎥
⎢ ⎥
( )
2
1171–1176; Trans., AIME, 231. SPE-915-PA. doi: 10.2118/915-PA. Tb = exp ⎢ −0.286590 ln M p ⎥
Pedersen, K.S. and Christensen, P.L. 2006. Phase Behavior of Petroleum
⎢ 39.8544 0.122488 ⎥ , . . . . . . . . . . . . . . . . (A-7)
Reservoir Fluids, 86–88, 92. Boca Raton, Florida, USA: CRC Press. ⎢− − ⎥
( )
2
Pedersen, K.S., Milter, J., and Sørensen, H. 2004. Cubic Equations of State ⎢⎣ ln M p ln M p ⎥⎦
Applied to HT/HP and Highly Aromatic Fluids. SPE J. 9 (2): 186–192.
( )
2
SPE-88364-PA. doi: 10.2118/88364-PA. −24.7522 ln M p + 35.155 ln M p

April 2010 SPE Reservoir Evaluation & Engineering 293

 TABLE A -1—VALUES OF COEFFICIENTS An FOR EQ. A-9

A1 A2 A3 A4 A5

M –131.11375 24.96156 –0.34079022 2.4941184 E 3 468.32575

Tb , °R 434.38878 50.125279 –0.9027283 7.0280657 E 3 –601.85651
0.86714949 3.4143408 E 8 –2.839627 E 5 2.4943308 E 8 –1.1627984
–0.50862704 –8.700211 E 2 –1.8484814 E 5 1.4663890 E 5 1.8518106
Tc , °R 915.53747 41.421337 –0.7586859 5.8675351 E 3 –1.3028779 E 10
Pc (psia) 275.56275 –12.522269 0.29926384 –2.8452129 E 3 1.7117226 E 3
vc (ft /lbm)
3
5.223458 E 2 7.8709139 E 4 –1.9324432 E 5 1.7547264 E 7 4.4017952 E 2

which can be solved iteratively using the following initial guess: The Watson characterization factor Kw is an overall character-
istic of the hydrocarbon fractions. Kw is defined:
Tb
Mp = . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-8) 1
10.44 − 0.0052Tb
Kw =
(1.8Tb )3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-14)

Ahmed Correlations. Ahmed et al. (1985) proposed a set of cor-
relations to calculate several properties of the SCN groups. The Using Riazi-Daubert correlations, it is possible to relate the
properties: M, Tb, , , Tc, Pc, and vc can be calculated as function Watson characterization factor and the molecular weight with the
of the carbon number with the following polynomial expression: following expression:

A5 K w = 4.5579 M 0.15178 −0.84573. . . . . . . . . . . . . . . . . . . . . . . (A-15)

 = A1 + A2 (Cn ) + A3 (Cn ) + A4 (Cn ) +
2 3
, . . . . . . . . . . . (A-9)
Cn

where  is physical property; Cn is carbon group number (i.e., Edmister’s Correlation. Edmister (1958) presented the following
SCN); and the coefficients An for n = 1, 2,…5 are presented in correlation to estimate the acentric factor () of pure liquids and
Table A-1. petroleum fractions:

Kesler-Lee Correlations. Kesler and Lee (1976) correlated Tc, ⎡ ⎛ p ⎞⎤

Pc, and  as function of the boiling point temperature (Tb) and the 3 ⎢ log ⎜ c ⎟ ⎥
⎝ 14.70 ⎠ ⎦
SG (). The correlations are: = ⎣ − 1, . . . . . . . . . . . . . . . . . . . . . . . . . (A-16)
⎡ Tc ⎤
7 ⎢ − 1⎥
Tc = 341.7 + 811 ⎣ Tb ⎦

+ ( 0.4244 + 0.1174 ) Tb +
( 0.4669 − 3.2623 ) ∗105 , where  is acentric factor; pc is critical pressure, psia; Tc is critical
Tb temperature, R; and Tb is normal boiling-point temperature, R.
. . . . . . . . . . . . . . . . . . . . . (A-10)
Appendix B—Lumping Method Used
0.0566 ⎛ 2.2898 0.11857 ⎞ Whitson’s Lumping Scheme. Whitson (1983) proposed a method
ln Pc = 8.3634 − − ⎜ 0.24244 + +
 2 ⎟⎠
for estimating the number of multiple carbon numbers (MCNs)
 ⎝ 
groups that are required for good description of a Cn+ fraction. The
⎛ 3.648 0.47227 ⎞ number of groups (Ng) is a function of the number of carbon atoms
∗10 −3 Tb + ⎜ 1.4685 + + ,
⎝   2 ⎟⎠ of the last SCN generated by the characterization process (N) and
the number of atoms of the first SCN within the Cn+ fraction (n).
⎛ 1.6977 ⎞
∗10 −7 Tb2 − ⎜ 0.42019 + ∗ 10 −3 Tb3
⎝  2 ⎟⎠ N g = Int ⎡⎣1 + 3.3 log ( N − n ) ⎤⎦ . . . . . . . . . . . . . . . . . . . . . . . (B-1)
. . . . . . . . . . . . . . . . . . . . . (A-11) The molecular weights separating each MCN group are deter-
mined by the following relation:
⎛ 6.09648 ⎞
ln Pbr − 5.92714 + + 1.28862 ln Tbr − 0.1669347Tbr6 ⎟
⎝⎜
1 Ng
T ⎠ ⎛M ⎞
= br
, M I = M Cn ⎜ N + ⎟ , . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-2)
⎛ 15.6875 ⎞ ⎝ M Cn ⎠
⎜⎝ 15.2518 − T − 13.4721 lnn Tbr + 0.4357Tbr6 ⎟
⎠
br
where MN+ is molecular weight of the SCN N; M Cn is molecu-
. . . . . . . . . . . . . . . . . . . . . (A-12) lar weight of the SCN n, which usually is taken as C7; and I =
1,2,3…., Ng.
for Tbr ≥ 0.8.
Lee’s Lumping Scheme. Lee et al. (1979) proposed a simple
 = −7.904 + 0.1352 K w − 0.007465K w2 + 8.359Tbr procedure to lump oil fractions into pseudocomponents. They
stated that the SCN with close physicochemical properties can be
+
(1.408 − 0.01063K w ) , . . . . (A-13) accurately represented by a single fraction. The closeness of these
Tbr properties is determined by the slopes of curves when the proper-
ties are plotted against the weighted average boiling point of each
Tb fraction, which is treated as the boiling point temperature Tb. The
for Tbr ≥ 0.8 where Tbr = ; Pb is the pressure at which Tb is weighted sums of the slopes of these curves are used as criteria
Tc
measured, usually 14.69 psia; Kw is the Watson characterization for lumping the oil fractions in the following steps:
P • Plot the available physicochemical properties of each oil frac-
factor; and Pbr = b . tion vs. the correspondent weight-averaged boiling point.
Pc
294 April 2010 SPE Reservoir Evaluation & Engineering
 • Calculate numerically the slope mij of the curves plotted in L L

Step 1 at each weight-averaged boiling point, where mij is the slope Binary interaction coefficient kkL = 1 − ∑ ∑ i* *j (1 − kij )
i ∈L j ∈L
of the property curve vs. boiling point; i = 1,2,…, nf (number of
SCN in the plus fraction); and j = 1, 2, …, np (number of physi- where kkL is binary interaction coefficient between the wkth compo-
cochemical properties). nent and the lumped fraction.
• Compute the normalized absolute slope m ij The subscription L denotes the lumped fraction.
Kay’s Mixing Rules. The mixing rules proposed by Kay (1936)
mij are used in Lee’s lumping scheme to regroup SCN fractions. These
m ij = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-3) mixing rules define the normalized mole fraction of a component,
mij max i =1,...,n f
i, within the set of the lumped fraction, that is i ∈ L, as:

• Compute the weighted sum of slopes M i for each fraction by zi

zi* = L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-6)
nP
∑z
∑m
i
ij i ∈L
j =1
Mi = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-4) The mixing rules for the different properties are:
nP
L
Pseudocritical pressure PcL = ∑ zi* Pci
where M i represents the averaged change of physicochemical i ∈L
properties of the oil fractions along the boiling-point axis. L
Group those oil fractions with similar M i . Pseudocritical temperature TcL = ∑ zi*Tci
i ∈L
Mixing Rules for Lumping of SCN. Hong’s and Kay’s mixing L
wi* MiVci
rules are adapted from Ahmed’s (2007) approach. Pseudocritical volume VcL = ∑
Hong’s Mixing Rules. Hong (1982) used the weight fraction i ∈L ML
average, wi*, for the mixing parameter in characterizing the C7+ L
fractions. The normalized weight fraction of a component, i, within Pseudoacentric factor  L = ∑ zi*i
the set of the lumped fraction, that is, i ∈ L, is defined as: i ∈L

zi M i Pseudomolecular weight M L = ∑ zi* Mi

w =
*
i L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-5) i ∈L

∑z M i i
i ∈L Ivan Rodriguez is a reservoir engineer for Weatherford Petroleum
Consultants in Stavanger. E-mail: ivan.rodriguez@wftpc.com.
He has worked for Geoconsult as a reservoir engineer (2008)
The mixing rules for the different properties are:
and Petroecuador as Performance Management Engineer for
L
E&P (2009). Rodriguez holds a BSc degree in petroleum engi-
Pseudocritical pressure PcL = ∑ wi* Pci neering from National Polytechnic School, Quito-Ecuador and
i ∈L
an MSc degree in reservoir engineering from the U. of Stavanger.
L Aly A. Hamouda is a professor at the U. of Stavanger. He is a
Pseudocritical temperature TcL = ∑ wi*Tci member of the Norwegian Academy of Science and in the
i ∈L Editorial Board for publications of book series in oilfield chem-
L istry book by Budapest Academy of science. Hamouda holds
Pseudocritical volume VcL = ∑ wi*Vci a PhD degree from Imperial College Science and Technology,
i ∈L
London U. He worked for 15 years with Phillips Petroleum and
worked at ConocoPhillips research Center in Oklahoma in
L
1984 in research related to formation damage. Hamouda is
Pseudoacentric factor  L = ∑ wi*i currently conducting a wide range of research activities within
i ∈L
the areas of heavy oil recovery, SAGD simulation and experi-
L mental, PVT for heavy oil CO2 recovery for asphaltenic oil, and
Pseudomolecular weight M L = ∑ i* Mi CO2 sequestration.
i ∈L

April 2010 SPE Reservoir Evaluation & Engineering 295