SPE-172843-MS

Fluid Characterization and EoS Modelling of PVT Experiments

Amao Temitope Olusegun, University of Ibadan; Amao Abiodun Matthew, King Saud University, KSA

Copyright 2014, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Nigeria Annual International Conference and Exhibition held in Lagos, Nigeria, 05– 07 August 2014.

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
Equations of state (EoS) are used to simulate laboratory pressure, volume and temperature (PVT)
experiments. Each PVT experiment mimics a form of phase behavior progression for the reservoir fluid
from the reservoir and ultimately to the surface facilities. The use of a characterized fluid sample in
compositional reservoir simulation study is precluded by a comparison and regression of EoS simulations
to PVT laboratory data.
In this study, EoS phase behavior software was applied to match laboratory PVT data of several
reservoir fluid samples from offshore Niger Delta. The paper presents the fluid characterization routines,
simulations, and regression capabilities of the software which was used for the determination of EoS
parameters needed to match experimental data.
Four EoS models are available through the generalized Martin’s equation in the software used. This
enabled the use of Redlich-Kwong (RK), Soave-Redlich-Kwong (SRK), Peng-Robinson (PR) and
Zudkevitch-Joffe (ZJ) EoS. Two 3-parameter extensions of the PR EoS are also available, one based on
a Peneloux et al. volume shift, the other being an implementation of the Schmidt-Wenzel (SW)
two-parameter PR EoS. The SRK EoS similarly has a three-parameter extension. This gives a total of
seven different EoS models. The predictive capability, strength and weaknesses of these models were
subsequently compared and analyzed.
Effect of the degree of input fluid characterization and pseudodization on EoS PVT experimental
modeling was investigated i.e. characterization to C7⫹, C12⫹, C20⫹, C30⫹, on the simulation output
for each of the EoS models available in the software. The results showed varying degree of simulated
experimental data congruence depending on the PVT experiment being modeled. With regression, both
the PR and ZJ EoS gave comparable and generally good agreement with laboratory data. Without
regression or significant tuning of EoS parameters, neither EoS adequately predicts observed reservoir
fluid PVT behavior. This work further elucidates the succinct effect of fluid models used in compositional
reservoir simulations which forms the basis for different production scenario and realizations.

Introduction
Equations of state have been used to model fluid and predict the thermodynamic state of the system. EoS
are the only way fluid composition, their phase changes and behavior can be represented in numerical
models of petroleum engineering processes. These encompass fluid models in production engineering to
2 SPE-172843-MS

Figure 1—Sample 1 DL-GOR Figure 2—Sample 2 DL-GOR

Figure 3—Sample 3 DL-GOR Figure 4 —Sample 1 DL Oil Relative Volume

Figure 5—Sample 2 DL Oil Relative Volume Figure 6 —Sample 3 DL Oil Relative Volume
SPE-172843-MS 3

highly complex models used in compositional res-

ervoir simulation, enhanced oil recovery (EOR)
processes like miscible gas injection and chemical
EOR. Cubic equations of state in the spirit of Van
der Waals equation especially have been very useful
in process modeling in all facets of engineering. The
first EoS that took cognizance of the fact that mol-
ecules exists and showed why gases do not behave
in an ideal way was the van der Waals EoS. The
initial conception of EoS was for single component
systems, however, van der Waals also came up with
mixing rules that made EoS more versatile and
effected their application to mixtures or multicom-
ponent systems.
Figure 7—Sample 1 DL-Liquid Density
In the history of EoS, several hundred EoS have
been developed and research is still ongoing to find
the perfect EoS, however, the oil industry has stuck with two main cubic EoS which are the Soave Redlich
Kwong (SRK) and Peng Robinson (PR) EoS for most applications. Cubic EoS have found wide
applicability because of their simplicity, relative ease of computation and seeming versatility. Also, their
usage is based on just knowing the critical properties (Tc, Pc, vc, acentric factor) of the components of any
mixture or fluid system. It should however be noted that EoS are based on empirical data fitting and
laboratory observations, therefore they are not perfect neither are they absolute truths.
As narrated earlier, the use of cubic EoS is premised on knowing the critical properties of the
constituent components in a fluid system. However, hydrocarbon fluids are a mixture of several
components that spans the light hydrocarbon gases, to non-hydrocarbons and also the heavy hydrocarbon
fractions, depending on the reservoir fluid composition and components which may be in the hundreds.
Therefore their representation in numerical models is usually limited as it is not yet possible to integrate
all the components in the fluid composition in the modeling process. The practice in the industry is to use
limited composition in the modeling and then compare the results with experimental data. This therefore
poses the first challenge in using cubic EoS for reservoir fluid modeling; how do we characterize and
adequately represent the fluid’s component’s critical properties in the EoS when we are limited by the
degree of component characterization. Therefore, the first study question is; what degree of fluid
characterization is good enough despite inherent limitations in adequately representing fluid system?
The present practice of petroleum engineering in all its facets has migrated from the use of charts and
monographs; we have progressed from reliance on correlations and fitted equations with their simplifying
assumptions and inherent non-uniqueness. The practice of petroleum engineering as it is today is
inherently dependent on technology. Reservoir simulation, production and facilities modeling, integrated
field development models and all other related processes must have a fluid component in the model. One
methodology developed for this is to represent the fluid system with the use of an EoS. However, these
models are usually limited in the number of components they can use to represent the fluid system. This
results in the use of “pseudocomponents”, which are aggregations of more than one component of the fluid
system. This adds more variables into the EoS conundrum, and it’s like an inverse of the first challenge.
Therefore, the second challenge is how do we combine components in such a way that the essence of the
fluid system is conserved, how do we make something not explicit enough smaller and still consistently
represent the fluid system?
Before the fluid model can be used for example in reservoir simulation, the simulations and predictions
from the cubic EoS are compared with PVT experimental data. This ultimately leads to a process called
“tuning” (regression), where the predictions are forced to match the experimental data by varying some
4 SPE-172843-MS

Figure 8 —Sample 2 DL-Liquid Density

properties of the EoS and the components. Several

studies had been done on the dangers of tuning and
its effects on the consistency of the tuned fluid in
subsequent process modeling in conditions outside
the tuned data. This is in similitude with history
matching, hence before the model can be worthy of
usage, it must predict what is known from PVT
experiments. However, because of the previous
challenges raised about the fluid system and its
characterization, there is usually a “mismatch”. This
variation is often attributed to uncertainties inherent
in the estimation of molecular weight and critical
properties of the plus fractions (undefined fractions) Figure 9 —Sample 3 DL-Liquid Density
of petroleum reservoir fluids, binary interaction pa-
rameters etc. Tuning therefore involves adjusting
some of the input parameters of the EoS to make it match the experimental data. The third challenge
therefore is how do we tune, if we must tune, and regress EoS parameters to achieve a fit between
experimental and simulated data and what does these all imply in the developed fluid model?

Study Methodology
In answering the questions posed in the previous section, the following methodology was employed in this
study. The first part of this work sought to proffer an answer to the first question by investigating how
the degree of fluid characterization determines how well EoS are able to model PVT experimental data.
The compositional analysis and PVT reports for three fluid samples from reservoirs offshore Niger Delta
were obtained for this study. The fluid samples were bottom-hole samples for which preliminary analysis
and data validation checks had been performed. These PVT reports served as the source of the fluid
sample compositions and PVT experimental data. The PVT reports contained essential information on the
fluid samples and the laboratory experiments. All the samples used in this study were oil samples and each
sample had extended compositional analysis up to C30⫹ fraction. Industry standard PVT software was
used for this part of the study.
The PVT software is cubic EoS based with several simulation capabilities, this include PVT experi-
ments simulation, fluid characterization for reservoir and process simulation etc. The first step in using the
SPE-172843-MS 5

Figure 10 —Sample 1 CCE Relative Volume

Figure 11—Sample 1 DL-Liquid Density Figure 12—Sample 3 CCE-Relative Volume

PVT program is the definition of the components that make up the fluid sample and their respective mole
fraction, molecular weight and plus fraction (Cn⫹) density etc. The program performs phase behavior and
equilibrium calculations relating to saturation pressure, density and viscosity calculations, simulation of
pressure depletion experiments such as Constant Composition Expansion (CCE), Constant Volume
Depletion (CVD) and Differential Liberation (DL) and other pertinent PVT simulations based on the
defined fluid system and selected EoS model. Four EoS models are available through the generalized
Martin’s equation in the PVT software. This allows the use of Redlich-Kwong (RK), Soave-Redlich-
Kwong (SRK), Peng-Robinson (PR) and Zudkevitch-Joffe (ZJ) EoS. Two 3-parameter extensions of the
PR EoS are also available, one based on a Peneloux et al. volume shift, the other being an implementation
of the Schmidt-Wenzel (SW) two-parameter PR EoS. The SRK EOS similarly has a three-parameter
extension. This gives a total of seven different EoS models for performing phase-equilibrium and property
calculations. Furthermore, calculations pertaining to splitting the plus fraction, component pseudodization
and non-linear regression of EoS parameters to match laboratory data are possible with the software.
The characterized fluid samples used in this study are presented in Table 1. Three samples called
Sample 1, Sample 2 and Sample 3 where used for the study. The extended characterization is presented
in the left part of the table while the properties of the plus fraction characterization are presented on the
right.
6 SPE-172843-MS

Table 1—Extended and Plus Fraction Composition of Reservoir Fluid The results of the simulations with the four de-
Samples Used for the Study
grees of characterization are presented in the next
Components Sample 1 Sample 2 Sample 3
section. It should be noted that the characterizations
N2 0.10 0.07 0.07 used in the first part of this study were the default as
CO2 0.29 0.26 0.53 presented in the PVT report, no splitting or lumping
H2S 0.00 0.00 0.00
scheme was carried out on the fluid system at this
C1 44.20 50.63 50.02
C2 4.00 5.18 5.52 stage.
C3 0.94 2.01 2.34
i-C4 0.35 0.67 0.64 Results for the Degree of Fluid
n-C4 0.26 0.82 0.89 Characterization
i-C5 0.21 0.63 0.59
n-C5 0.22 0.41 0.44 Figures 1 to 12 showed the results of experimental
C6 0.32 1.31 1.10 simulations with the different degree of character-
C7 0.62 2.18 2.00 ization of the fluid samples. The results showed a
C8 1.17 2.31 2.30
marked deviation between the experimental data
C9 1.36 2.07 2.29
C10 1.95 2.14 2.12
and predictions for the DL data. The only the CCE
C11 2.00 2.07 2.07 relative volume data showed a close agreement with
C12 2.54 2.07 2.01 experimental data as shown in the figures 10 and 12.
C13 4.65 2.04 2.01 In the next section, the effect of pseudodization
C14 5.84 2.94 2.91
and grouping to the simulation of experimental data
C15 6.13 2.15 2.32
C16 3.31 2.23 2.04 by cubic EoS is investigated.
C17 1.94 1.68 1.65
C18 1.91 1.69 1.48 Pseudodization and Grouping
C19 1.73 1.37 1.30 The approach required splitting the plus fraction
C20 1.42 1.20 1.17
C21 1.22 1.00 1.07
into Single Carbon Number group (SCN), followed
C22 1.05 0.90 1.00 by groping the SCN groups into multiple carbon
C23 0.99 0.84 0.94 number groups (MCN). The splitting technique
C24 0.92 0.75 0.88 used is based on the Whitson three-parameter prob-
C25 0.86 0.67 0.83
ability density function (PDF) which relates mole
C26 0.77 0.62 0.77
C27 0.42 0.55 0.72
fraction to molecular weight. The software’s Multi
C28 0.37 0.52 0.67 Feed Semi-Continuous Thermodynamic (SCT) split
C29 0.24 0.48 0.60 was the splitting technique of choice. Upon split-
C30⫹ 5.70 3.54 2.71 ting, the extended components were grouped into
Plus Fractions
three pseudocomponents tagged (FRC1, FRC2 and
Mole % Sample 1 Sample 2 Sample 3
FRC3). The Kessler-Lee correlation was chosen for
C7⫹ 49.11 38.01 37.86
C10⫹ - 31.45 31.27
the pseudocomponent critical properties and acen-
C12⫹ 42.01 - - tric factor. Tables 2, 3 & 4 shows the details of the
C20⫹ 13.96 11.07 11.36 multi feed SCT split for samples 1, 2 & 3 respec-
C30⫹ 5.70 3.54 2.71 tively. The mole weight of the heaviest pseudocom-
Molar Mass ponent (FRC3) was adjusted to ensure that more of
C7⫹ 276.52 245 239 the mole fraction is in the heaviest component. The
C10⫹ - 284 266 phase diagram of the extended fluid composition
C12⫹ 301.22 - -
C20⫹ 496.64 419 394
showed minimal deviation (at the reservoir temper-
C30⫹ 750 615 592 ature) from that of the original fluid composition
Density (g/cc) indicating a good split as shown in Figure 1 for
C7⫹ 0.8898 0.9044 0.9047 sample 13.
C10⫹ - 0.9155 0.9159 The extended composition was thereafter used to
C12⫹ 0.8996 - - match the laboratory determined saturation pressure
C20⫹ 0.9593 1.0176 1.0193
C30⫹ 1.01 1.2331 1.3921
SPE-172843-MS 7

Table 2—SAMPLE 1 MULTI FEED SCT SPLIT

Component Mole Percent Weight Fraction Mole Weight Specific Gravity

CO2 0.29 0.087659 28.01 0.56119

N2 0.1 0.01924 44.01 0.59535
C1 44.2 4.8703 16.04 0.53564
C2 4 0.82612 30.07 0.56559
C3 0.94 0.2847 44.1 0.59554
IC4 0.35 0.13972 58.12 0.62548
NC4 0.26 0.1038 58.12 0.62548
IC5 0.21 0.10407 72.15 0.65543
NC5 0.22 0.10902 72.15 0.65543
C6 0.32 0.18462 86.2 0.69178
FRC1 9.8171 7.7689 115.22 0.78096
FRC2 18.919 27.428 211.09 0.84768
FRC3 20.374 58.074 415 0.92894

Table 3—SAMPLE 2 MULTI FEED SCT SPLIT

Component Mole Percent Weight Fraction Mole Weight Specific Gravity

N2 0.07 0.018398 28.01 0.56119

CO2 0.26 0.10737 44.01 0.59535
C1 50.63 7.6205 16.04 0.53564
C2 5.18 1.4616 30.07 0.56559
C3 2.01 0.83177 44.1 0.59554
I-C4 0.67 0.3654 58.12 0.62548
N-C4 0.82 0.44721 58.12 0.62548
I-C5 0.63 0.42653 72.15 0.65543
N-C5 0.41 0.27758 72.15 0.65543
C6 1.31 1.0596 86.2 0.69178
FRC1 9.8466 10.744 116.28 0.80769
FRC2 16.052 31.182 207.01 0.8733
FRC3 12.111 45.458 400 0.95477

Table 4 —SAMPLE 3 MULTI FEED SCT SPLIT

Component Mole Percent Weight Fraction Mole Weight Specific Gravity

N2 0.07 0.018847 28.01 0.56119

CO2 0.53 0.22421 44.01 0.59535
C1 50.02 7.7121 16.04 0.53564
C2 5.52 1.5955 30.07 0.56559
C3 2.34 0.99193 44.1 0.59554
I-C4 0.64 0.35755 58.12 0.62548
N-C4 0.89 0.49721 58.12 0.62548
I-C5 0.59 0.40918 72.15 0.65543
N-C5 0.44 0.30515 72.15 0.65543
C6 1.1 0.91143 86.2 0.69178
FRC1 10.521 11.773 116.41 0.81064
FRC2 16.301 32.551 207.74 0.87677
FRC3 11.038 42.653 402 0.95875

by carrying out a special regression on the molecular weight of the characterized components in the split.
The parameters regressed on are SCT splitting parameter, SCT plus fraction weight and SCT Watson
Factor. This caused the saturation pressure calculated by the EoS to assume an improved match with the
8 SPE-172843-MS

Figure 13—Superimosed phase diagrams of the original fluid composition and the splited composition for Sample 1

Table 5—Pseudodization Methodology

Scheme A Scheme B Scheme C Scheme D

Component Group Component Group Component Group Component Group

CO2 1 X1⫹ CO2 2 X2⫹ CO2 1 X1⫹ CO2 1 X1⫹

N2 2 X2⫹ C2 2 X2⫹ C2 1 N2 1
C1 3 C1⫹ N2 1 X1⫹ N2 1 C1 1
C2 4 C2⫹ C1 1 X1⫹ C1 1 C2 1
C3 4 C3 3 C3⫹ C3 2 C3⫹ C3 2 C3⫹
IC4 5 C4⫹ IC4 3 IC4 2 IC4 2
NC4 5 NC4 3 NC4 2 NC4 2
IC5 5 IC5 4 C5⫹ IC5 2 IC5 2
NC5 5 NC5 4 NC5 2 NC5 2
C6 5 C6 4 C6 2 C6 2
FRC1 6 C8⫹ FRC1 5 C8⫹ FRC1 3 C8⫹ FRC1 3 C8⫹
FRC2 7 C16⫹ FRC2 6 C16⫹ FRC2 4 C16⫹ FRC2 4 C16⫹
FRC3 8 C31⫹ FRC3 7 C31⫹ FRC3 5 C31⫹ FRC3 4

experimental saturation pressure. The aim was not to obtain a perfect fit because of the expected distortion
from component pseudodization.
The initial fluid description for each of the sample was lumped into eight components. Subsequently,
the number pseudocomponents were reduced to 7, 5 and 4 respectively.
Table 5 shows the pseudodization methodology used in the study. Scheme A has 8 pseudocomponents
while Scheme B, C and D had 7, 5 and 4 pseudocomponents respectively. Tables 6, 7, 8 and 9 show the
pseudocomponents mole percent, weight fraction, mole weight and specific gravity for Sample 1 under
each of the schemes aforementioned. Similar tables were obtainable for Samples 2 and 3.
The next step was to match the laboratory PVT data by simulating the laboratory experiments using the
three-parameter Peng-Robinson EoS. It is very important that the EoS generated volumetric data fit the
SPE-172843-MS 9

Table 6 —SAMPLE 1 GROUPED INTO 8 COMPONENTS

Component Group Mole Percent Weight Percent Mole Weight Specific Gravity

CO2 1 X1⫹ 0.29 0.080181

N2 2 X2⫹ 0.1 0.017599
C1 3 C1⫹ 44.2 4.4548
C2 4 C2⫹ 4.94 1.016 32.739 0.55447
C3 4
IC4 5 C4⫹ 1.36 0.58652 68.647 0.61362
NC4 5
IC5 5
NC5 5
C6 5
FRC1 6 C8⫹ 7.0668 5.1152 115.22 0.77451
FRC2 7 C16⫹ 16.302 21.619 211.09 0.84068
FRC3 8 C31⫹ 25.741 67.111 415 0.92127

Table 7—SAMPLE 1 GROUPED INTO 7 COMPONENTS

Components Group Mole Percent Weight Percent Mole Weight Specific Gravity

CO2 2 X2⫹ 4.29 0.91378 31.012 0.56348

C2 2 X2⫹
N2 1 X1⫹ 44.3 4.8896 16.07 0.42586
C1 1 X1⫹
C3 3 C3⫹ 1.55 0.52822 49.617 0.57585
IC4 3
NC4 3
IC5 4 C5⫹ 0.75 0.39771 77.207 0.65205
NC5 4
C6 4
FRC1 5 C8⫹ 9.8171 7.7689 115.22 0.781
FRC2 6 C16⫹ 18.919 27.428 211.09 0.8477
FRC3 7 C31⫹ 20.374 58.074 415 0.9289

Table 8 —SAMPLE 1 GROUPED INTO 5 COMPONENTS

Components Group Mole Percent Weight Percent Mole Weight Specific Gravity

CO2 1 X1⫹ 48.59 5.8033 17.389 0.43801

C2 1
N2 1
C1 1
C3 2 C3⫹ 2.3 0.92593 58.614 0.6007
IC4 2
NC4 2
IC5 2
NC5 2
C6 2
FRC1 3 C8⫹ 9.8171 7.7689 115.22 0.781
FRC2 4 C16⫹ 18.919 27.428 211.09 0.8477
FRC3 5 C31⫹ 20.374 58.074 415 0.9289

laboratory observations before using the EoS for compositional reservoir simulation. Each Sample’s
grouped composition was used as input fluid definition in the PVT software.
The results of the simulation runs based on the pseudodized fluid characterization are presented in the
next section.
10 SPE-172843-MS

Table 9 —SAMPLE 1 GROUPED INTO 4 COMPONENTS

Component Group Mole Percent Weight Percent Mole Weight Specific Gravity

CO2 1 X1⫹ 48.59 5.8033 17.389 0.43801

C2 1
N2 1
C1 1
C3 2 C3⫹ 2.3 0.92593 58.614 0.6007
IC4 2
NC4 2
IC5 2
NC5 2
C6 2
FRC1 3 C8⫹ 9.8171 7.7689 115.22 0.781
FRC2 4 C16⫹ 39.293 85.502 316.82 0.8898
FRC3 4

Table 10 —Regression Parameters

Results for Pseudodization Regression Sample 1 Sample 2 Sample 3

Tuning and Regression Regression 1 Tcrit:FRC3 Tcrit:C1 Tcrit:C1

Acf:FRC3
Regression is aimed at adjusting the EoS model Omega_A:FRC3
parameters to get an improved match of the exper- Omega_B:FRC2
imental data. The foremost step in the regression Omega_B:FRC3
was to determine the parameter set that would be Regression 2 Acf:FRC3Tcrit:C1 Sshift:FRC2
Omega_A:FRC2
Sshift:FRC2 Sshift:FRC3
used. A rule of thumb in selecting regression vari-
Omega_A:FRC3
Sshift:FRC3
ables is to neglect EoS parameters of components Omega_B:FRC2
that do not affect the value of the regression data. Omega_B:FRC3
This is often the case with components that have Sshift:FRC2
very little compositional dominance in the fluid Sshift:FRC3

composition. Furthermore, phase behaviour studies

have identified the plus fraction (Cn⫹) as a key
element in attaining agreement between simulation
results and laboratory results. This is because the plus fraction contains a mixture of components so its
properties may not have been accurately determined in the PVT lab and also from correlations. This makes
the plus fraction an ideal candidate for regression to fit the equation of state to the experimental data. The
sensitivity analysis tool of the PVT software was employed in determining the EoS parameter(s) that
affects the difference between the observed and simulated values the most. It is desirable to have a few
sensitive parameters as against numerous insensitive ones. Table 10 shows the EoS parameters chosen as
regression variable for the two regression procedures.
If a bad fit is observed, the regression is rejected, the EoS regression variables are readjusted and the
regression is done again until an acceptable fit is obtained. Increasing the regression weight of observa-
tions having the largest differences between calculated and observed data can help in the regression
process but this has to be done prudently.

Results for Tuning and Regression

Analysis and Discussion of Results
Input fluid characterization to C7⫹ gave a better prediction of the laboratory volumetric and phase
behavior data in figures 1-12 compared to the extended compositions. Fluid samples characterized to
SPE-172843-MS 11

Figure 14 —Sample 1 DL-GOR Figure 15—Sample 1 DL-Oil Relative Volume

Figure 16 —Sample 1 DL Liquid density Figure 17—Sample 2 DL-GOR

Figure 18 —Sample 2 DL Oil Relative Volume Figure 20 —Sample 2 DL-Oil Liquid Density

C30⫹ had the largest average deviation from laboratory data. Therefore, from this study, extended degree
of fluid characterization did not give a better prediction of experimental data.
12 SPE-172843-MS

Figure 21—Sample 1 CCE Relative Volume Figure 22—CCE Liquid Density

Figure 23—Sample 2 CCE Relative Volume Figure 24 —Sample 2 CCE Liquid Density

Figure 25—Sample 3 DL-GOR Figure 26 —Sample 2 CCE Liquid Density

It is important to understand the terms used in presenting the results for the pseudodization and
regression. The word “predicted” refers to EoS results obtained with the default three-parameter Peng-
Robinson EoS model. “Adjusted” describes the EoS results calculated after special regression had been
done to match the bubble point pressure. While “regressed” pertains to EoS results calculated after
regression on EoS parameters to obtain an improved match with laboratory PVT data. After adjusting the
SPE-172843-MS 13

Figure 27—Sample 3 DL-GOR

Figure 28 —Sample 3 CCE-Liquid Density

Figure 29 —Sample 3 DL-GOR Figure 30 —Sample 3 DL-Oil Relative Volume

Figure 31—Sample 3 DL-Liquid Density Figure 32—Sample 3-CCE Relative Volume

molecular weight of the FRC-1, 2&3 pseudos, the deviation in predicted and laboratory observed
saturation pressure for the three samples reduced to a maximum difference of 0.24%. The effect of
molecular weight adjustment also led to improved predictions of the other experiments that were
simulated. The approach to component played a key role in predictions obtained for each of the
pseudodization schemes. It has been reported in literatures that component pseudodization interferes
negatively with the EoS predictions and subsequent match with experimental data. The results presented
14 SPE-172843-MS

Figure 33—Sample 3-CCE Liquid Density Figure 34 —Sample 2 DL-GOR

Figure 35—Sample 2 DL-Oil Relative Volume Figure 36 —Sample 2 DL-Liquid Density

Figure 37—Sample 2 CCE Relative Figure 38 —Sample 2 CCE-Liquid Density

in Figures 14-28 showed that the pseudodization schemes gave reasonable results with the laboratory
experimental data.
The EoS parameters of methane and the FRC-pseudos were the most sensitive during the tuning
procedure. The match of DL experiment GOR was difficult. However, the regression parameters
presented in table 10 were capable of maintaining the regression results within the 5-10% deviation range
which is permissible for PVT studies.
SPE-172843-MS 15

Figure 39 —Sample 1 DL-GOR Figure 40 —Sample 1 DL-Liquid Density

Figure 41—Sample 1 CCE Relative Volume Figure 42—Sample 1 CCE Liquid Density

Figure 43—Sample 1 DL-GOR Figure 44 —Sample 3 DL-Liquid Density

Consequent to the effective grouping schemes, the tuning procedure that followed was fairly direct and
straightforward. Very few EoS parameters were tweaked and this is a very much desirable fear bearing
in mind that having too many regression parameters could actually result in very unrealistic EoS
predictions. The results in figures 33 and 34 show the remarkable closeness of scheme 4 prediction (with
no regression) to the “predicted” result and the ensuing regression.
16 SPE-172843-MS

Conclusions
This study has further elucidated the understanding of phase behavior and EoS PVT modeling. The three
questions of degree of characterization, pseudodization and tuning or regression posed as the study
objectives have been investigated.
Therefore, based on this study and the results of the simulations, the following conclusions can be
drawn; An extended degree of fluid characterization did not give a better prediction of experimental data;
this was seen in the first part of the work, when different degrees of characterizations were used in
simulating PVT experimental data.
Pseudodization works and components can be lumped into different groups as long as some funda-
mental lumping and splitting tenets are adhered to. In line with this, this study shows that the manner of
splitting the plus fraction and the subsequent component pseudodization is very important to obtaining a
good match between EoS results and laboratory data.
Tuning may not always be necessary in some well grouped pseudocomponent fluid models, and when
tuning is necessary tune respecting the fundamentals of molecular behavior and fluid chemistry.
The difficulty with the DL data matching experimental data for most of the simulation may be due to
the fundamental scheme of DL experiments which involved the removal of some species from the system
(cell) during the PVT experiments.

Bibliography
1. Adetunji Lukemon, Thermodynamically Equivalent Pseudocomponents Validated For Composi-
tional Reservoir Models. PhD. Dissertation, Texas Tech University, USA
2. Ahmed,T. H., Hydrocarbon Phase Behaviour, Gulf Publishing Company, Vol. 7, (1989)
3. Ahmed,T.: Equation of state and PVT analysis, Gulf Publishing Co., Houston (2007)
4. Al-MeshariA. and William D. McCain: “New Strategic Method to Tune Equation-of-state for
compositional simulation,” Paper SPE 106332.
5. Al-MeshariA. et alet al.: “Tuning an Equation of State-The Critical Importance of Correctly
grouping composition into Pseudocomponents,” paper 96416 SPE presented at SPE ATCE,
Dallas, Texas, October 2009.
6. Ding-Yu Peng and Donald B. Robinson, “A New Two-Constant Equation of State”, Ind. Eng.
Chem., Fundam., Volume 15, No 1, 1976
7. GaganisV. et alet al.: “Monitoring PVT Properties Derivatives Ensures Physically Tuned EoS
Behaviour over the entire operating conditions range,” paper 94211 SPE presented at Europe/
EAGE Annual Conference, Madrid, June 2005.
8. Ingolf-Soreide, Improved Phase Behaviour Prediction of Petroleum Reservoir Fluids from a
Cubic Equation of State. PhD. Dissertation, NTNU, Norway.
9. Karen Schou Pedersen and Peter L. Christensen, Phase Behavior of Petroleum Reservoir Fluids,
Taylor Francis Group LLC, 2007
10. Otto Redlich and J.N.S. Kwong, “On the Thermodynamics of Solutions, An Equation of State,
Fugacities of Solutions, Paper presented at the Symposium on Thermodynamics and Molecular
Structure of Solutions, American Chemical Society, Portland, Oregon, 1948
11. PedersenK.S. and ChristensenL.P.: “Phase Behaviour of Petroleum Reservoir Fluids”, Taylor &
Francis Group, Boca Raton (2007)
12. PedersenK.S.: “On the Dangers of Tuning Equation of State Parameters,” Paper SPE 14487.
13. RafaelA. and McCainW.: “An Efficient Tuning Strategy to calibrate cubic EOS for compositional
simulation,” paper 77382 SPE presented at SPE ATCE, San Antonio, Texas, September-October
2002.
SPE-172843-MS 17

14. Soave,G., Equilibrium Constants from a Modified Redlich-Kwong Equation of State, Chem. Eng.
Sci. 27, 1197–1203, 1972.
15. Wang,P. et alet al.: “Proper Use of Equation of State for Compositional Reservoir Simulation”,
JPT (July 2001) 74.
16. Whitson,C.H et alet al.: Phase Behaviour, Monograph series, SPE Richardson, TX (2000)