You are on page 1of 17

SPE-175877-MS

EOS Tuning – Comparison between Several Valid Approaches and New


Recommendations
Mahmoud T. Ali, Mansoura Petroleum Company; Ahmed H. El-Banbi, Cairo University

Copyright 2015, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE North Africa Technical Conference and Exhibition held in Cairo, Egypt, 14 –16 September 2015.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is 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 acknowledgment of SPE copyright.

Abstract
Gas condensates and volatile oil reservoirs are simulated using compositional simulation or modified
black-oil (MBO) approaches to account for composition changes in the reservoir1. The current approach
in the industry is to use a tuned EOS model to represent the fluid phase behavior and volumetric
properties. However; there is no consensus on how to tune an EOS model, how many pseudo-components
to use, or what weight factors to apply to experimental data when tuning an EOS model for use in
compositional and MBO reservoir simulation studies.
Regarding the EOS tuning parameters, for example, some investigators prefer to regress on the binary
interaction parameters (BIP) and the (⍀’s) parameters of the EOS. Others have chosen the volume shift,
BIP’s, and critical parameters as matching parameters. Also, there is a newer approach where some
investigators have chosen to vary the plus fraction molecular weight followed by the volume shift
parameters.
In this study, we have constructed an EOS model program to perform the five commonly used tuning
approaches. We have investigated the different tuning approaches using a large data base of PVT samples
spanning the range of black-oils through volatile oils and gas condensates (including rich gas condensates
and lean wet gases). We have compared the different tuning approaches based on two criteria: (1) the
quality of match after applying the tuning procedure, as indicated by the absolute average error (AAE),
and (2) the percentage deviation in the tuning parameters from their original values.
The tuning of many reservoir fluids using the different approaches has shown that some approaches
were more superior to others and usually yielded excellent matches. While the other tuning approaches
gave moderate to good matches. Based on this work, we were also able to give practical recommendations
on tuning approaches that yield consistent EOS models.

Introduction
Cubic Equation of state (EOS) models are used extensively in the petroleum industry. They are mostly
used for (1) reservoir modeling with a compositional simulator; (2) generating black-oil (BO) or modified
black-oil (MBO) PVT formulations for reservoir simulators; or (3) Production and process engineering
calculations.
2 SPE-175877-MS

In general, the prediction capabilities of the cubic EOS, without proper tuning, are questionable
particularly for multi-component petroleum mixtures. In other words, phase behavior models based on
these equations may predict highly erroneous results, particularly for near critical fluids2–4. Currently, the
industry approach to improve the prediction capabilities of an EOS model is to tune it against experi-
mental data generated at specific conditions of pressures and temperatures in the PVT laboratory.
Although the industry has no consensus on single standard method of tuning, the various approaches carry
some similarities. The basic idea in tuning an EOS hinges on adjusting some uncertain values of EOS
input parameters to minimize the difference between the predicted and measured values.
In this study, we have constructed an EOS model program to perform the five commonly used tuning
approaches5. We have studied the different tuning approaches using 30 fluid samples with a wide range
of C7⫹ mole% (3.5– 61) spanning the range of black-oils through volatile oils and gas condensates. For
consistency purposes, the final number of components after tuning with any of the five tuning approaches
are about 7 pseudo-components (C1, C2-C3, C4-C6, two plus fractions, and non-hydrocarbon components).
We have compared between the different tuning approaches based on the quality of match after applying
the tuning procedure, as indicated by the absolute average error (AAE) while taking into considerations
the percentage deviation in the tuning parameters from their original values. In addition, we have
presented the prediction results of PR6, 7 EOS using the industry recommended C7⫹ characterization and
SRK8, 9 EOS using the industry recommended C7⫹ characterization without tuning. The industry
recommended C7⫹ characterizations will be summarized in the next section.
The results have provided guidelines regarding the appropriate tuning approaches that yield consistent
EOS models, monotonic properties, and excellent match between the calculated and observed data.

Industry Recommended C7ⴙ Characterizations


Whitson et al.3 recommended the following C7⫹ characterization procedure for cubic EOS’s.
1. Use the Twu10 or Lee-Kesler11 critical property correlation for Tc and pc (we used Lee-Kesler11
correlation).
2. Choose the acentric factor to match Tb; alternatively, use the Kesler-Lee12 correlation (we used
Kesler-Lee12 correlation).
3. Determine volume-translation (c) values to match specific gravities; alternatively, use Peneloux et
al.13 correlation for the SRK EOS or Jhaveri and Youngren14 correlation for the PR EOS.
4. When measured TBP data are not available, a mathematical split should be made with either (1)
the gamma distribution with Gaussian-quadrature or equal-mass fractions or (2) an exponential
distribution (e.g. Pedersen’s4 distribution).
5. Specific gravities and boiling temperatures should be estimated using Søreide15 correlations.
6. For the PR EOS, authors recommended the non-hydrocarbon BIP’s proposed by Nagy and
Shirkovskiy16, and the modified Chueh-Prausnitz17 equation for C1 through C7⫹ pairs.
As an alternative, Whitson et al.3 recommended the use of Pedersen et al.4 characterization procedure
with the SRK EOS. The following steps summarize Pedersen’s3, 4 fluid characterization procedure.
1. Split using Pedersen’s splitting scheme.
2. Calculate critical properties of all single carbon number (SCN) fractions using Pedersen’s
correlations.
3. Group C7⫹ into 3 to 12 fractions using equal-weight fractions in each group; use weight-average
mixing rules.
4. Calculate volume-translation parameters for C7⫹ fractions to match specific gravities; pure
component (c) values are taken from Peneloux et al.13
SPE-175877-MS 3

Tuning Approaches
Many investigators2–5, 18–23 have studied different approaches for the EOS models tuning in order to
obtain a consistent model with high prediction capabilities for use in describing the reservoir fluid phase
behavior and flow and volumetric properties under different pressure and temperature conditions. In our
study, we have only investigated the five commonly used approaches proposed by Coats and Smart19
(1986), Whitson and Brule3 (2000), Christensen21 (1999), Aguilar and McCain22 (2002), and Al-Mehsari
and McCain23 (2005) on a large database of PVT samples. Hereafter, we will discuss each tuning method
in some detail.
Coats and Smart Approach19
Coats and Smart19 (1986) used a methodology based on splitting the plus fraction using Whitson’s
method. They then used systematic regression logic as follows: regress on the binary interaction
parameters between methane and heaviest plus fraction to match the saturation pressure and then ⍀a and
⍀b of the methane and heaviest plus fraction are used as regression parameters to match the different PVT
observations. The Saturation pressure is given a weight factor of 40 while most others are given weight
of 1.
The authors also provided limits for the regression parameters such as (–1 to 0.9) for BIP’s, (0.1 to 1.3)
for ⍀a, and (0.02 to 1.25) for ⍀b.
Whitson and Brule Approach2
In the SPE phase behavior monograph2 (2000), the authors discussed the regression process in a general
manner and they concluded that using Coats and Smart method of regression can provide even better
results if used with the volume translation. They showed that using the Tc and Pc instead of the ⍀a and
⍀b in Coats and Smart work can yield the same results. In Appendix C in the monograph they obtained
a good match by regressing on the Tc’s and Pc’s of all the plus fractions and the BIP’s between the
methane and the plus fractions.
The authors did not provide any limits for the change in the regression parameters. For our work here,
we applied Coats and Smart19 method limits for the BIP’s and ⫾15% limit for the Tc’s and Pc’s.
Christensen Approach21
Christensen21 (1999), on the contrary to the other methods, used the SRK EOS and characterized the C7⫹
as proposed by Pedersen et al. In his study, he presented a procedure consisting of three steps for
regression of EOS parameters to experimental PVT data. The first step is to regress on the C7⫹ molecular
weight in the range of ⫾10%, then to regress on the volume translation parameters in the range of ⫾100%,
and finally to regress on the two most sensitive parameters in Pedersen’s correlation in the range of
⫾20%. In our application of this methods here, instead of using the two most sensitive parameters, we
picked the first two parameters in the arrangement of regression parameters presented in PVTsim Software
Method documentation24.
Aguilar and McCain Approach22
Aguilar and McCain22 (2002) chose to regress first on the C7⫹ Molecular weight in the range of ⫾20%
and to split the C7⫹ into SCN to C45⫹ to match the saturation pressure, then to lump the components into
two pseudo components and regress on the critical properties of the heaviest pseudo component to
preserve the saturation pressure match. Finally, they recommended regression on the volume translation
parameters to achieve the other observation match.
The authors did not provide limits for second (regression on critical properties for the two pseudo-
components) and third (volume translation) steps parameters. In our application of the method in this
work, we did not impose limits on the second step regression while we applied the Christensen21 approach
limits for the volume translation.
4 SPE-175877-MS

Al-Meshari and MCcain Approach23


Al-Meshari and MCcain23 (2005) proposed some modifications to Aguilar and MCcain method. They
modified the splitting procedure itself, but the main difference in the regression was that acentric factor
was used instead of the critical properties of the heaviest pseudo component to preserve the saturation
pressure match in the second step of Aguilar and MCcain22 regression procedure. The third step
(regression on volume shift) was kept as in the Aguilar and McCain22 method.
In our work here, we could not reproduce the results of Al-Meshari and McCain23 method for the gas
samples given in their reference (while we could reproduce the results of their oil samples). We also failed
to contact the main author to investigate our results. Therefore, we presented the results of this method
only for our oil samples for fair comparison with the other methods.
Fluid Database
In our study, the five common tuning approaches have been tested on 30 fluids with a wide range of C7⫹
mole% (3.5 – 61%). The database covers gas condensates (both rich gas condensates and lean wet gases),
volatile oils, and black oils including near critical fluids as shown in Fig. 1.

Figure 1—C7ⴙ mole% of fluids used in this study

We have studied the different tuning approaches on 11 gas and 19 oil samples. The fluids were
collected from different sources including research papers, Ph. D. dissertation, and PVT lab reports.
Absolute Average Error (AAE) is used throughout the study to quantify comparisons between
measured and calculated data. The following equation shows how the AAE is computed.
(1)

Where n⫽ number of points in an observation, number of observations, or number of experiments


Results
The testing of the five common tuning approaches on our 11 gases and 19 oils have shown interesting
results about how to tune the EOS models to be used in reservoir studies. In the next section, we will
SPE-175877-MS 5

present in details the results of one gas sample and one oil sample to show the differences in tuning
capabilities between the five approaches. Following this section, we will present the results we obtained
from studying all the 30 fluids.
Results for Gas-7
This is a typical gas condensate fluid that contains 7.2 mole% C7⫹. The hydrocarbon system exhibits a
dew point pressure of 5,054 psia at 229°F. The fluid composition is presented in Table 1. The fluid was
subjected to CCE and CVD experiments.

Table 1—Gas-7 Original Composition

The five approaches have been tested on this fluid to minimize the difference between the experimental
and calculated results using only 7 components as shown in Table 2.

Table 2—The Number of Components After Tuning

The results of the CCE and CVD experiments observations using only 7 components are presented in
Figs. 2 and 3 and Tables 3 and 4. We have also shown the PR EOS and SRK EOS results using the
industry recommended C7⫹ characterization with the original composition without tuning (WT) in
addition to the five tuning approaches prediction results.
6 SPE-175877-MS

Figure 2—Gas-7 predicted CCE experiment observations without tuning and after tuning with the different approaches

Figure 3—Gas-7 predicted CVD experiment observations without tuning and after tuning with the different approaches
SPE-175877-MS 7

Table 3—Gas-7 Predicted Saturation Pressure Without Tuning and After Tuning with the Different Approaches

Table 4 —Gas-7 Absolute Avergae Errors for Different Experiments Observations

The results for this fluid show that without tuning, PR EOS using the industry recommended C7⫹
characterization has better prediction capabilities than SRK EOS using the industry recommended C7⫹
characterization. However, without tuning, we are still far from having an acceptable match between the
experimental and predicted results.
All the tuning approaches gave excellent match (AAE’s of 0%) for the dew-point pressure except for
Christensen approach which gave an AAE’s of 5.3%.
For the CCE and CVD experiments, the four tuning approaches gave a good to excellent match for
relative volume and compressibility factors observations. The two liquid drop out observations have
shown a superior prediction capability for the EOS when Whitson and Brule tuning approach was used.
We concluded from the results that this fluid cannot be matched without tuning, and Whitson and Brule
tuning approach should be used as it gave the lowest AAE (2.2%) with 7 pseudo-components.
8 SPE-175877-MS

Results for Oil-8


This is a typical volatile oil fluid that contains 21.3 mole% C7⫹. The hydrocarbon system exhibits a
bubble point pressure of 2,667 psia at 220°F. The fluid composition is presented in Table 5. The fluid was
subjected to CCE and DL experiments.

Table 5—Oil-8 Original Composition

The five tuning approaches have been tested on this fluid in an attempt to minimize the difference
between the experimental and predicted fluid properties using only 7 components (Table 2).
The results of the CCE and DL experiments observations using only 7 components are presented in
Figs. 4 and 5 and Tables 6 and 7. Similar to Gas-7 results, we have shown the PR EOS and SRK EOS
using the industry recommended C7⫹ characterization prediction without tuning (WT) predictions along
with the five tuning approaches prediction results.
SPE-175877-MS 9

Figure 4 —Oil-8 predicted CCE experiment relative volume observation without tuning and after tuning with the different approaches

Figure 5—Oil-8 predicted DL experiment observations without tuning and after tuning with the different approaches
10 SPE-175877-MS

Table 6 —oil-8 Predicted Saturation Pressure Without Tuning and After Tuning with the Different Approaches

Table 7—Oil-8 Absolute Avergae Errors for Different Experiments Observations

The results have shown that without tuning, SRK EOS using the industry recommended C7⫹
characterization has better prediction capabilities than PR EOS using the industry recommended C7⫹
characterization. Also, both characterizations gave a moderate to good match between the experimental
and predicted results.
Al-Meshari and McCain and Aguilar and McCain tuning approaches matched the bubble point pressure
perfectly, while Whitson and Brule gave the higher AAE of 7%. The five tuning approaches gave
excellent matches for the CCE relative volume observation. For the DL experiment, all the tuning
approaches except Coats and Smart gave a good to excellent matches for oil formation volume factor,
gas-oil ratio, gas Z-factor, and liquid density observations. Al-Meshari and McCain and Aguilar and
McCain tuning approaches gave the best matches among the five tuning approaches.
SPE-175877-MS 11

We concluded from these results that this fluid can have a moderate to good match without tuning, but
Al-Meshari and McCain and Aguilar and McCain tuning approaches give an excellent match with AAE
as low as 3% for 7 pseudo-components.
The same procedure was followed in studying all 30 fluids in our database. The results of our
investigation showed that Al-Meshari and McCain and Aguilar and McCain tuning approaches consis-
tently perfectly match saturation pressures as shown in Table-1. For the CCE, CVD, and DL experiments;
Whitson and Brule tuning approach showed superior prediction performance over the other approaches for
the gases, while Al-Meshari and McCain approach gave the minimum AAE between predicted and
experimental measurements for the oils. Al-Meshari and McCain and Aguilar and McCain approaches
always yielded the best matches for the saturation pressure. Al-Meshari and McCain approach always
required less modification in the C7⫹ Mw than Aguilar and McCain approach.
The results for all 30 fluids CCE, DL, and CVD prediction errors are shown in Tables 8 and 9. We had
only five PVT samples with separator experiments data. Our investigation shows that Whitson and Brule
approach generally provide lowest error compared to the other approaches for the separator test as shown
in Table 10. The overall AAE for each fluid observations is summarized in Table 11.

Table 8 —Fluid Database Errors in Saturation Pressure and CCE Experiment


12 SPE-175877-MS

Table 9 —Fluid Database Errors in CVD and DL Experiments

Table 10 —Fluid Database Errors in Separator Experiments


SPE-175877-MS 13

Table 11—Fluid Database Runs Errors

Discussion
In this paper, we compared the five most commonly EOS tuning approaches to a large database of fluids
spanning all fluid types. The comparison results show that Whitson and Brule tuning approach is
recommended for gases (Fig. 6) while Al-Meshari and McCain approach is recommended for oils (Fig.
7). The results of Table 11 show that Al-Meshari and McCain approach is the best for our database for
both black oil and volatile oil samples.
14 SPE-175877-MS

Figure 6 —Gas samples absolute average error

Figure 7—Oil samples absolute average error

In the tuning (and characterization) approaches investigated during this study, the monotonicity of the
EOS parameters was maintained and modifications were made within the recommended ranges suggested
by the authors of the five approaches. This guarantees that the resultant EOS models are stable and
SPE-175877-MS 15

increase the model chances of good prediction capabilities outside the range of pressures and temperatures
provided by the laboratory observations.
We investigated these EOS tuning approaches under the assumption that we would like to generate
good EOS model with reasonable number of components (around 7 pseudo-components) for composi-
tional reservoir simulation. We have not thoroughly investigated if our results will hold true when we use
higher number of pseudo-components to describe the fluid behavior of these 30 fluids. However, it is
intuitively expected that the use of higher number of pseudo-components will give more degrees of
freedom to obtain better matches. The true challenge for the EOS is to match the laboratory observations
with fewer components, and be able to predict the phase behavior and PVT properties at reservoir and
surface temperatures and pressures. This is usually sufficient for most applications of EOS in the
petroleum industry.

Conclusions
The main objective of this study was to test the effectivness of the five commonly used EOS tuning
approaches proposed by Coats and Smart (1986), Whitson and Brule (2000), Christensen (1999), Aguilar
and McCain (2002), and Al-Mehsari and McCain (2005) on different reservoir fluids. We tested those
approaches on 30 fluids spanning a wide range of phase behavoir and fluid properties. We have
successfully reached valuable guidelines regarding which approach to be used in tuning the EOS model
with low number of components (about 7) for use primarily in compositional reservoir simulation models.
Based on the work performed in this investigation and results, we can conclude the following:
1. It is usually enough to split the C7⫹ fraction into two pseudo-components, if the right tuning
approach is applied.
2. Lumping C2-C3 and C4-C5 along with splitting the C7⫹ fraction into two pseudo-components can
yield an EOS model with good to excellent prediction capabilities, if the right tuning approach is
applied.
3. The five commonly used tuning approaches always yield consistent EOS models, with monotonic
properties.
4. Al-Meshari and McCain and Aguilar and McCain approaches are often capable of predicting
almost perfect saturation pressure. Al-Meshari and McCain approach requires less modification in
the C7⫹ Mw than Aguilar and McCain approach.
5. We recommend the use of Whitson and Brule tuning approach for gas samples as it consistently
yielded good to excellent matches between observed and predicted results in all the experiments
for our database.
6. We recommend the use of Al-Meshari and McCain tuning approach for oil samples (both black
and volatile oils) as it provided good to excellent match between observed and predicted results
for almost all experiments in the fluids tested here.

Nomenclature
c ⫽ volume translation
n ⫽ number of points in an observation, number of observations, or number of experiments
Pc ⫽ critical pressure, psi
Tc ⫽ critical temperature, °F
⍀a - ⍀b ⫽ constants in cubic EOS’s

Abbreviations
AAE ⫽ Absolute Average Error
abs ⫽ absolute value of a number
16 SPE-175877-MS

BO ⫽ Black Oil
BIP ⫽ Binary Interaction Parameters
CCE ⫽ Constant Composition Expansion
CVD ⫽ Constant Volume Depletion
DL ⫽ Differential Liberation
EOS ⫽ Equation of State
MBO ⫽ Modified Black Oil
SCN ⫽ Single Carbon Number
TBP ⫽ True Boiling Point

References
1– El-Banbi, Ahmed H., Forrest, J. K., Fan, L., and McCain, W. D., Jr.: “Producing Rich-Gas-
Condensate Reservoirs--Case History and Comparison Between Compositional and Modified
Black-Oil Approaches,” paper SPE 58988 presented at the SPE Fourth International Petroleum
Conference and Exhibition, Villahermosa, Mexico. Feb. 1–3, 2000.
2– Danesh, A., PVT and Fluid Behaviour of Petroleum Reservoir Fluids, ELSEVIER, Amesterdam,
1998.
3– Whitson,C. H. and Brule,M. R., Phase Behavior, Monograph Vol. 20 SPE HENRY L. DOHERTY
SERIES, Richardson, Texas, 2000.
4 – Pedersen, K. S. and Christensen, P. L., Phase Behavior of Petroleum Reservoir Fluids. Taylor &
Francis Group, LLC, Boca Raton, 2007.
5– Ali, Mahmoud T.: “A New Approach For Tuning The Equation of State of Petroleum Reservoir
Fluids”, Msc. Thesis, Cairo University, Cairo, Egypt (2013).
6 – Peng, D. Y. and Robinson, D. B.: “A New-Constant Equation of State,” Ind. & Eng. Chem. (1976)
15, No. 1, 59.
7– Robinson, D. B., Peng, D. Y., and Ng, H. -Y.: “Capabilities of the Peng-Robinson
8 – Redlich, O. and Kwong, J. N. S.: “On the Thermodynamics of Solutions. V. An Equation of State.
Fugacities of Gaseous Solutions,” Chem. Rev. (1949) 44, 233.
9 – Soave, G.: “Equilibrium Constants from a Modified Redlich-Kwong Equation of State,” Chem.
Eng. Sci. (1972) 27, No. 6, 1197.
10 – 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) No. 16,
137.
11– Lee, B. I. and Kesler, M. G. “A Generalized Thermodynamic Correlation Based on Three-
Parameter Corresponding States,” AIChE J. (1975) 21, 510.
12– Kesler, M. G. and Lee, B. I.: “Improve Predictions of Enthalpy of Fractions,” Hydro. Proc.
(March 1976) 55, 153.
13– Peneloux, A., Rauzy, E., and Freze, R.: “A Consistent Correction for Redlich-Kwong-Soave
Volumes,” Fluid Phase Equilibria (1982) 8, 7.
14 – Jhaveri, B. S. and Youngren, G. K.: “Three-Parameter Modification of the Peng-Robinson
Equation of State To Improve Volumetric Predictions,” SPERE (August 1988) 1033.
15– Søreide, I.: “Improved Phase Behavior Predictions of Petroleum Reservoir Fluids From a Cubic
Equation of State”, Dr.Ing. Dissertation, Norwegian Inst. of Technology, Trondheim, Norway
(1989).
16 – Nagy, Z. and Shirkovskiy, A. I: “Mathematical Simulation of Natural Gas Condensation
Processes Using the Peng-Robinson Equation of State,” paper SPE 10982 presented at the 1982
SPE Annual Technical Conference and Exhibition, New Orleans, 26 –29 September.
SPE-175877-MS 17

17– Chueh, P. L. and Prausnitz, J. M.: “Calculation of High-Pressure Vapor–Liquid Equilibria,” Ind.
Eng. Chem. (1968) 60, No. 13.
18 – Pedersen, K. S., Thomassen, P., and Fredenslund, A.: “On the Dangers of Tuning Equation of
State Parameters”, paper SPE 14487 available from SPE, Richardson, Texas (1985).
19 – Coats, K. H. and Smart, G. T.: “Application of a Regression-Based EOS PVT Program to
Laboratory Data”, SPERE (May 1986) 277–299.
20 – Agarwal, R. K., LI, Y. -K. and Nghiem, L, “A regression technique with dynamic parameter
selection for phase behaviour matching”, SPERE (1990) 115–119.
21– Christensen, P. L., Regression to experimental PVT data, J. Can. Petroleum Technol. 38, 1–9,
1999.
22– Aguilar, R. A. and McCain, W. D., Jr.:” An Efficient Tuning Strategy to Calibrate Cubic EOS
for Compositional Simulation,” paper SPE 77382, presented at the SPE ATCE, San Antonio,
Texas, USA, Oct. 2002.
23– Al-Meshari, A.: “New Strategic Method To Tune Equation-Of-State To Match Experimental
Data For Compositional Simulation”, Ph. D. Dissertation, Texas A&M University, College
Station, Texas, USA (2004).
24 – PVTsim 18 Method Documentation, Version 16, Calsep A/S, 2007.

You might also like