Vous êtes sur la page 1sur 7

Fluid Characterization for Miscible EOR

Projects and CO2 Sequestration


Kristian Jessen,* Stanford University, and Erling H. Stenby, Technical University of Denmark

Summary seconds even for fluid descriptions of 10 components or more.


Accurate performance prediction of miscible enhanced-oil- Later, Jessen et al. (2001) used this approach to generate approxi-
recovery (EOR) projects or CO2 sequestration in depleted oil and mate solutions to the dispersion-free, 1D-displacement problem for
gas reservoirs relies in part on the ability of an equation-of-state multicomponent gas-injection processes.
(EOS) model to adequately represent the properties of a wide Analytical and numerical methods for predicting the perfor-
range of mixtures of the resident fluid and the injected fluid(s). The mance of a gas-injection process depend on an EOS to predict the
mixtures that form when gas displaces oil in a porous medium will, phase behavior of the mixtures that form in the course of a dis-
in many cases, differ significantly from compositions created in placement process. The role of the phase behavior in relation to
swelling tests and other standard pressure/volume/temperature numerical diffusion in compositional reservoir simulation has been
(PVT) experiments. Multicontact experiments (e.g., slimtube dis- pointed out previously by Stalkup (1990) and by Stalkup et al.
placements) are often used to condition an EOS model before (1990). Recently, Jessen et al. (2004) proposed a method to quan-
application in performance evaluation of miscible displacements. tify the interplay of the phase behavior and numerical diffusion in
However, no clear understanding exists of the impact on the resultant a finite-difference simulation of a gas-injection process. By ana-
accuracy of the selected characterization procedure when the fluid lyzing the phase behavior of the injection-gas/reservoir-fluid sys-
description is subsequently included in reservoir simulation. tem, a measure of the impact, referred to as the dispersive distance,
In this paper, we present a detailed analysis of the quality of can be calculated. The dispersive distance is useful when designing
two different characterization procedures over a broad range of and interpreting large-scale compositional reservoir simulations.
reservoir fluids (13 samples) for which experimental swelling- In this work, the analytical theory of gas-injection processes is
test and slimtube-displacement data are available. We explore the used to analyze an extensive database of PVT measurements and
impact of including swelling-test and slimtube experiments in slimtube experiments published by Jaubert et al. (2002a). Jaubert
the data reduction and demonstrate that for some gas/oil sys- et al. (2002b) applied the characterization procedure of Avaulleé
tems, swelling tests do not contribute to a more accurate prediction et al. (2001) combined with the Peng-Robinson (PR) EOS (Peng
of multicontact miscibility. Finally, we report on the impact that and Robinson 1976) to analyze the database. The main question
use of EOS models based on different characterization proce- addressed in Jaubert et al. (2002b) was whether experimental deter-
dures can have on recovery predictions from dynamic 1D displace- mination of the MMP by slimtube experiments is really needed if
ment calculations. other, less-expensive multicontact tests are available. They concluded
on the basis of their analysis that unless pure CO2 is injected into a
reservoir, slimtube experiments may not be needed to accurately pre-
dict the development of miscibility in gas/oil displacements. How-
Introduction ever, to our best knowledge, most commercial software does not offer
During the past few decades, a significant effort has been invested the characterization procedure of Avaulleé et al. (2001).
in the studies and development of improved-oil-recovery pro- The purpose of the present work was to investigate the influ-
cesses. From a technical point of view, gas injection can be a very ence on EOR performance predictions exerted by the choice of
efficient method for improving the oil production, particularly in commonly available characterization methods for the heavy frac-
the case when miscibility develops during the displacement pro- tion of the reservoir fluid and the choice of EOS. This leads to a
cess. The lowest pressure at which a gas should be injected into the considerable number of combinations. Some options are described
reservoir to obtain the multicontact miscible displacement—the in Chorn and Mansoori (1989). In the present study, two com-
minimum miscibility pressure (MMP)—has consequently attained monly applied combinations of characterization procedure and
a very important status in EOR studies. EOS are investigated.
Various methods for measuring and calculating the MMP have
been proposed in the literature. Many of these are based on sim- Experimental Database
plifications such as the ternary representation of the compositional In the open literature, the amount of experimental MMP data
space. This method fails to honor the existence of a combined where the reservoir-fluid compositions have been described in de-
mechanism controlling the development of miscibility in real res- tail is very limited. To this end, Jaubert et al. (2002a) published a
ervoir fluids. Zick (1986) and Stalkup (1987) described the exist- database (13 reservoir oils, each with one or two injection gases)
ence of the condensing/vaporizing mechanism. They showed that including the detailed compositional data, standard PVT experiments,
the development of miscibility (MMP) in multicomponent gas- swelling tests, and MMP determinations from slimtube experi-
displacement processes could, independent of the mechanism con- ments. That database is used extensively in this comparative study.
trolling the development of miscibility, be predicted accurately by Jaubert et al. (2002a) provides a detailed compositional analy-
1D compositional simulations. A semianalytical method for pre- sis of the heptanes-plus fractions (up to C20+) from fractional
dicting the MMP was later presented by Wang and Orr (1997), true-boiling-point (TBP) distillations of the reported crude oils.
who played an important role in the development and application The database includes amount, molecular weight, and specific
of the analytical theory of gas-injection processes. Jessen et al. gravity for each of the TBP cuts (C7 to C19) along with the prop-
(1998) subsequently developed an efficient algorithm for perform- erties and amounts of the TBP residuum (C20+) at the end of the
ing these calculations, reducing the MMP calculation time to a few TBP distillation. For additional detail on the experimental methods
and standards for characterizing the heptanes-plus fraction by TBP
distillation, the reader is referred to Whitson and Brule (2000).
* Now with the Mork Family Department of Chemical Engineering and Materials Science at
the University of Southern California, Los Angeles. Figs. 1 and 2 report the variability of the mole fraction and
specific gravity, respectively, of the various TBP cuts as a function
Copyright © 2007 Society of Petroleum Engineers
of their molecular weight. The properties and amounts of TBP
This paper (SPE 97192) was accepted for presentation at the 2005 SPE Annual Technical residue for each reservoir fluid in the database are shown in
Conference and Exhibition, Dallas, 9–12 October, and revised for publication. Original
manuscript received for review 24 October 2006. Revised manuscript received 11 February Fig. 3. From Figs. 1 through 3, it is clear that the fluids included
2007. Paper peer approved 17 February 2007. in the database of Jaubert et al. range from light to medium oils

482 October 2007 SPE Reservoir Evaluation & Engineering


Fig. 1—Mole% vs. molecular weight of various TBP cuts. Fig. 2—Density at 1 atm (14.7 psi) and 15°C (59°F) vs. molecular
weight of various TBP cuts.
that are typically good candidates for miscible-gas-injection
projects (Taber et al. 1997). parameters is typically performed to improve on the accuracy by
Experimental MMP values are reported for injection of multicom- regression of experimental data. In this work, we investigate the
ponent gases (lean to rich) as well as a single data point for injection performance of the SRK and PR methods at four stages:
of pure CO2. For two of the reservoir fluids in the database, MMP 1. Pure prediction based on the default characterizations
values are reported for two different injection gas compositions. 2. After regression of standard PVT measurements such as con-
stant-mass expansion (CME) and differential-depletion experi-
Characterization Procedures ments (DDEs)
Two different characterization procedures were applied to generate 3. After regression of CME, DDE, and swelling test(s)
EOS input from the compositional-analysis data described in the 4. After regression of CME, DDE, swelling test, and MMP
previous section: Stages 1–3 allow for an evaluation of the impact of the indi-
• The characterization method by Pedersen et al. (1989), with vidual standard PVT experiment and the benefit of including
parameters from Knudsen and Stenby (1996), combined with the swelling-test data in the regression on the accuracy of the predicted
Soave-Redlich-Kwong (SRK) EOS (Soave 1972). In the following MMP. This can be a useful step in assessing the uncertainty of
sections, we refer to this combination as the SRK method. EOS prediction in cases where a limited set of experimental data
• The characterization method suggested by Whitson et al. is available.
(1989) combined with the modified version of the PR EOS (Peng At Stage 4 of the tuning process, we include the experimental
and Robinson 1976; Robinson and Peng 1978). In the following MMP in the regression. At this point, the true strength of the
sections, we refer to this combination as the PR method. analytical theory of gas-injection processes becomes evident. If
The two methods for assigning properties to the C7+ fraction of traditional multicell approaches or slimtube simulations for pre-
a reservoir fluid differ mainly in the selection of distribution func- diction of MMP are used in the regression of experimental MMP
tion relating molecular weight to mole fractions. In the SRK data, the regression process can easily become very CPU intensive.
method, an exponentially decaying function describes the amount Instead, we use the key-tie-line algorithm of Jessen et al. (1998)
of fluid with respect to the molecular weight of single-carbon- for prediction of the MMP as implemented in our Miscible Injec-
number groups (i.e., a discrete representation), whereas the PR tion PVT software (MI-PVT). The efficiency of the key-tie-line
method relies on a gamma distribution of variable shape to relate approach makes the inclusion of MMP data in the parameter-
molecular weight to mole fractions in a continuous approach. An- estimation process very inexpensive, even when more than one
other significant difference between the two methods is the selec- MMP data point must be matched. In our implementation, the CPU
tion of pseudocomponents representing the TBP cuts and the TBP requirement for parameter estimation including MMP is, in gen-
residue. In the SRK method, the heavy end (C7+) of the reservoir eral, no more than a few minutes.
fluid is divided into a number of pseudocomponents representing
equal mass fractions, whereas the pseudocomponents in the PR
method are commonly selected on the basis of quadrature points in
a Gaussian integration rule (use of equal mass fractions is also
possible). Both methods are used widely in the petroleum industry
and serve as well-documented reference points in the following
sections. For more details on the characterization procedures, the
reader is referred to Pedersen et al. (1989), Whitson et al. (1989),
and Whitson and Brule (2000).
All fluid characterizations were performed using a total of 15
components (defined and pseudocomponents). Because H2S is
present only in minor amounts throughout the database, CO2 and
H2S are combined in all fluid descriptions, allowing for a total of
five pseudocomponents to represent the C7+ fraction.
Tuning Approach
When a given combination of characterization procedure and EOS
is used to generate a fluid description, the predictive capability in
the original form may not be sufficient to represent all experimen-
tal data with appropriate accuracy. Consequently, tuning of model Fig. 3—Properties and amounts of TBP residue.

October 2007 SPE Reservoir Evaluation & Engineering 483


The possible options for selecting regression parameters when the accuracy of the characterization procedures at the different
tuning a 15-component fluid description are essentially endless. A stages, we use the average-relative-error percent of
commonly used tuning parameter is the molecular weight of the 1. Saturation pressures from CME/DDE and swelling test(s): ⌬Psat
TBP residue (Mw+). Because of experimental uncertainties, this 2. Overall fluid densities from CME and CME during swelling
quantity is usually not measured with more than 10% accuracy. test(s) and liquid-phase densities from DDEs: ⌬␳fluid
Consequently, Mw+ can be adjusted before a recharacterization if, 3. Relative volumes from CME experiments: ⌬Vrel
for example, the predicted saturation pressure is in error. However, 4. MMP from slimtube experiments: ⌬MMP
it often occurs that such an adjustment is insufficient to ensure Tables 1 and 2 summarize the results of characterization and
appropriate overall accuracy and that more parameters must be tuning for the investigated fluids JF1 through JF13. JF8 is omitted
adjusted. To reduce the degrees of freedom when tuning the two from the table for formatting purposes without changing the over-
EOS models to the experimental data, we restrict our tuning ap- all picture. The number of experimental data points included in
proach to involve columns of pseudocomponents. That is, for ex- each of the average relative errors is listed in the tables (in paren-
ample, we adjust the column of critical temperatures (Tc) for all theses) and varies from two data points (Psat and MMP only) for
pseudocomponents by a common factor. By this approach, we JF8 to 457 data points for JF5.
preserve the trends of the original fluid characterizations. At To illustrate the accuracy of the EOS models at Stage 4 in the data
Stages 2 and 3, we adjust only Tc and the critical pressures (pc) of regression, Figs. 4 through 7 compare the calculated (SRK-method
the pseudocomponents. However, in some cases, more adjustable characterization) and experimental values for JF12. The accuracy
variables are required to obtain the desired accuracy of the model. for JF12 is representative for all other fluids in the database.
In such cases, we include the acentric factor (␻) and the volume At each stage of the regression procedure, we observe the fol-
shift parameter (S) of the pseudocomponents in the regression. lowing predictive behavior of the two characterization approaches:
Throughout this work, no binary-interaction coefficients are in-
cluded in the regression of experimental data. This results in a
maximum of four adjustable parameters. Stage 1
A standard Marquard-Levenberg algorithm was used for the • The default PR-method characterization predicts, with few
data regression. The experimental data (pressures, densities, and exceptions, the experimental saturation pressures more accurately
relative volumes) were assigned different weights in the param- than the default SRK-method characterization.
eter-estimation process to reflect the relative importance in the • The default SRK-method characterization predicts, in general,
context of miscible displacements. In this work, the weights were the MMP more accurately than the default PR characterization.
fixed at 50, 10, and 1 for pressures, densities, and relative volumes, • Fluid densities and relative volumes are predicted with simi-
respectively. For each fluid system, the experimental data were lar accuracy by the two methods.
inspected for internal consistency before parameter estimation, and
no significant discrepancies were observed.
Stage 2
• Saturation pressures of the original reservoir fluids, as well as
Characterization Results the saturation pressures reported for the swelling tests, are pre-
The 13 reservoir oils in the work of Jaubert et al. (2002a) were dicted equally well by the two methods, with JF13 as the only
characterized with methods PR and SRK. They are referred to in exception. We note that swelling-test data are not included at this
the following section as Jaubert Fluid 1 through 13 (JF1 through stage. For JF13, the injection gas used in the swelling test is pure
JF13). For each of the 13 fluids and each of the characterization CO2. This reservoir fluid is the heaviest of the fluids in the data-
procedures, the predicted values were compared with the experi- base and causes the PR method some problems because of the high
mental values at each of the four stages listed above. To compare values of binary-interaction coefficients between CH4 and the

484 October 2007 SPE Reservoir Evaluation & Engineering


pseudocomponents (up to 0.13). The problem arises because the Stage 4
PR-method characterization prematurely predicts a liquid/liquid • The predicted CME, DDE, swelling test, and MMP at Stage
split (LLE), forming a hydrocarbon-rich liquid phase and a CO2- 4 are, by both methods, in satisfactory agreement with the experi-
rich liquid phase. This, in turn, results in an overprediction of mental results for all the investigated reservoir fluids. All experi-
saturation pressures at higher solvent (CO2) concentrations. The mental MMP values can be calculated within the accuracy of slim-
problem can be mitigated by reducing the binary-interaction coef- tube experiments (≈5%).
ficients manually and repeating the regression of standard PVT • No significant differences in predictive capabilities of the
data (CME and DDE). The SRK method does not make use of two characterization methods are observed.
interaction coefficients between CH4 and pseudocomponents and
hence is less likely to predict LLE inaccurately. Dispersive Distance and Miscibility With
• For a majority of the investigated reservoir fluids, regression Pure CO2
of Tc and pc to CME/DDE data does not improve on the accuracy
of the predicted MMP. As described in the introduction, the analytical theory of gas-
injection processes makes it possible to evaluate the interplay of
phase behavior and numerical diffusion of a given fluid system
Stage 3 (Jessen et al. 2004). The dispersive distance is defined as the
• Including swelling-test experiments in the regression does distance between the point in composition space that becomes a
not, on average, improve significantly on the prediction of the critical point at the MMP and the dilution line (in composition
MMP for both characterization methods.

Fig. 5—Experimental and calculated overall fluid densities


Fig. 4—Experimental and calculated saturation pressure at vari- from CME experiments with JF12 at temperatures from 299.85
ous temperatures for JF12. The critical point is predicted. to 394.45 K.

October 2007 SPE Reservoir Evaluation & Engineering 485


Fig. 6—Comparison of calculated and experimental behavior of
an isothermal swelling test (JF12): saturation pressures and Fig. 7—Calculated and experimental overall fluid densities from
saturated liquid densities. CME during swelling test: JF12 (Fig. 6). “Beta” refers to the mole
fraction of solvent in the overall mixture.

space) that is defined by the oil and the injection-gas compositions.


The larger the dispersive distance, the more sensitive a displace- both the MMP and the dispersive distance, the two methods are
ment calculation will be to the gridblock size used in single-point- found to be in good agreement. By comparing the MMP for in-
upwind, finite-difference-based compositional simulation (numeri- jection of pure CO2 with the lean-rich-gas MMP from the data-
cal diffusion). Numerical artifacts in coarse-grid simulations can base, it is clear that all the investigated reservoir fluids could, in
easily render the performance prediction of multicontact miscible terms of the local displacement efficiency, be produced efficiently
floods unreliable. To investigate if the two characterization meth- by a miscible flood in concert with CO2 sequestration.
ods will display different sensitivity when the fluid descriptions
are ultimately included in reservoir simulations, the dispersive Flow Predictions
distance has been calculated for the 13 fluid systems and is re- As a final comparison of the investigated characterization meth-
ported in Table 3. In general, the dispersive distances for the SRK ods, we consider two 1D displacement calculations where gas is
and the PR methods are in excellent agreement, with the PR injected to displace reservoir fluid JF10 at near-miscible condi-
method marginally higher for the gas/oil systems reported by Jaub- tions. To isolate the effects of phase behavior from the transport
ert et al. (2002a). Only one example was found where the calcu- properties, the f-theory viscosity model of Quinones-Cisneros
lated values are very different. This occurs for JF8, where the et al. (2004) was used to generate a fictive set of viscosity data with
calculated dispersive distance of the PR method is almost an order the PR method. The SRK characterization was then tuned to the
of magnitude greater than that of the SRK method. However, only fictive data set to ensure that, at least, the viscosities of the original oil
two data points (psat and MMP) are used in the regression, and were calculated identically by the two fluid descriptions. Note that
consequently, the overall quality of the resultant EOS models can- the regression of viscosity data by the f-theory does not change the
not be examined in great detail. results of PVT calculations. The fictive viscosity data from the
As an additional test of the two characterization methods, the PR-method characterization are compared with the prediction of
MMP and dispersive distance for injection of pure CO2 were cal- the SRK-method characterization after regression in Fig. 8.
culated for all reservoir fluids. The results are given in Table 3 and The MMP for JF10 and associated injection gas (G10) is re-
are compared with the dry/lean-gas MMPs from the database. For ported to be 287 bar (Table 3). To compare the predicted displace-

486 October 2007 SPE Reservoir Evaluation & Engineering


Fig. 8—Oil viscosities calculated by the SRK- and PR-method Fig. 9—Saturation profiles of near-miscible-displacement cal-
fluid descriptions used in the displacement calculations for culations for JF10. Calculated MMP is 269 bar with SRK and
reservoir fluid JF10. 268 bar with PR relative to the experimental value of 287 bar
(∼6% error).

ment efficiency when using the PR and SRK methods, near-


miscible dispersion-free 1D calculations were performed at 250 In summary, the conclusions of this study are as follows:
and 260 bar. The dispersion-free displacement calculations were 1. Without regression of experimental PVT data, the characteriza-
performed by the approach of Jessen et al. (2001). The saturation tion procedure of Whitson et al. (1989) in general predicts the
profiles after 0.5 pore volumes injected (PVI) at 260 bar and after saturation pressure of the investigated reservoir fluids more ac-
0.75 PVI at 250 bar are shown in Fig. 9. The saturation profiles curately than the approach of Pedersen et al. (1989). The oppo-
predicted by the two methods are in excellent agreement for both site trend is observed for default predictions of the MMP.
displacement calculations. In particular, the leading edge of the 2. Regression of standard PVT experiments (CME and DDE) does
displacement that sets the breakthrough time is predicted to be not ensure an accurate calculation of the MMP.
almost identical for the two characterization procedures in both 3. Including swelling-test data in the parameter estimation im-
calculation examples. Recovery predictions for the two displace- proves only slightly the accuracy of the calculated MMP.
ments are compared in Fig. 10. The calculated recovery curves 4. When the experimental MMP data are included in the parameter
using the SRK- and the PR-method fluid descriptions are in good estimation, no significant difference is observed between the
agreement, suggesting that marginal differences should be ex- investigated characterization methods.
pected between a reservoir simulation using the PR- and the SKR- 5. The key-tie-line approach for calculation of the MMP is highly
fluid description. suitable for implementation in regression software because of
The fact that two fundamentally different approaches for as- the low CPU requirement per MMP calculation.
signing properties to the pseudocomponents representing the C7+ 6. The dispersive distance calculated by the two methods is found
fraction of the reservoir fluid result in almost identical prediction to be in good agreement and does not suggest any significant
of the displacement behavior suggests that the regression of ex- difference in the sensitivity to numerical diffusion in finite-
perimental data was performed in a consistent fashion. Further- difference-based compositional simulation.
more, the observation demonstrates that the predictive capabilities 7. Predictions of the MMP and the dispersive distance for injec-
of the two characterization methods are close to identical when tion of pure CO2 based on the final fluid descriptions are in
care is taken to accurately represent appropriate experimental data excellent agreement.
(MMP). A similar agreement should not be expected for injection 8. With good agreement in the predicted MMP, predictions of local
of pure CO2 without additional tuning to appropriate experimental displacement efficiency and, consequently, the predicted break-
data. This is because the MMP values calculated by the two meth- through time and oil recovery from 1D dispersion-free simula-
ods for pure CO2 (Table 3) differ more (10%) than the MMP
values predicted by SRK and PR for the injection gas composition
that was included in the regression (0.4%).

Discussion and Conclusions


Based on an extensive database of detailed fluid descriptions with
experimental PVT and MMP data, the two most commonly used
methods for reservoir-fluid modeling have been compared. The
characterization procedures of Whitson et al. (1989) and Pedersen
et al. (1989) each combined with the cubic EOS with which they are
most often used, (i.e., the PR EOS and SRK EOS, respectively).
The results of matching experimental PVT, swelling-test, and
slimtube data, based on the investigated fluid-characterization
methods and the selected tuning strategies, suggest that the con-
clusion of Jaubert et al. (2002b) may not be valid in general. We
find that experimental determination of the MMP is indeed a valu-
able data point when building an EOS model to be used in EOR
performance evaluation. For only a few of the investigated fluid
systems, the accuracy of the calculated MMP based on regression Fig. 10—1D recovery of oil in place (OIP) by near-miscible
to standard PVT experiments and swelling tests alone will be displacements of JF10. Calculated MMP is 269 bar with SRK
appropriate for prediction of the local displacement efficiency and and 268 bar with PR relative to the experimental value of 287
design of injection-gas compositions for potential EOR projects. bar (∼6% error).

October 2007 SPE Reservoir Evaluation & Engineering 487


tion at near-miscible conditions are almost identical for the two Pedersen, K.S., Fredenslund, A., and Thomassen, P. 1989. Contributions in
fluid models. Petroleum Geology & Engineering 5. In Properties of Oils and Natural
The above conclusions are based on the given database, and it Gases, Vol. 5. Houston: Gulf Publishing Company.
should be noticed that the investigated fluids cover the range of Peng, D-Y. and Robinson, D.B. 1976. A New Two-Constant Equation of
medium-to-light oils and rich-to-light hydrocarbon gases. The con- State. Industrial & Engineering Chemistry Fundamentals 15 (1): 59–64.
clusions may be different for heavy oils and high-pressure/high- Quinones-Cisneros, S.E, Dalberg, A., and Stenby, E.H. 2004. PVT Char-
temperature fluids. acterization and Viscosity Modeling and Prediction of Crude Oils.
Petroleum Science and Technology 22 (9): 1309–1326.
Nomenclature Robinson, D.B., and Peng, D.-Y. 1978. The Characterization of the Hep-
Mw ⳱ molecular weight tanes and Heavier Fractions. GPA Research Report 28, Tulsa.
Pc ⳱ critical pressure Soave, G. 1972. Equilibrium Constants from a Modified Redlich-Kwong
S ⳱ volume shift parameter Equation of State. Chem. Eng. Sci. 27 (6): 1197–1203.
Tc ⳱ critical temperature Stalkup, F.I. 1987. Displacement Behavior of the Condensing/Vaporizing
Gas Drive Process. Paper SPE 16715 presented at the SPE Annual
␻ ⳱ acentric factor
Technical Conference and Exhibition, Dallas, 27–30 September. DOI:
⌬MMP ⳱ average relative error in predicted MMP (%) 10.2118/16715-MS.
⌬Psat ⳱ average relative error in predicted saturation Stalkup, F.I. 1990. Effects of Gas Enrichment and Numerical Dispersion
pressures (%) on Enriched-Gas-Drive Predictions. SPERE 5 (4): 647–655. SPE-
⌬Vrel ⳱ average relative error in predicted relative volumes 18060-PA. DOI: 10.2118/18060-PA.
(%) Stalkup, F.I., Lo, L.L., and Dean, R.H. 1990. Sensitivity to Gridding of
⌬␳fluid ⳱ average relative error in predicted fluid densities (%) Miscible Flood Predictions Made With Upstream Differenced Simula-
tors. Paper SPE 20178 presented at the SPE/DOE Enhanced Oil Re-
Acknowledgments covery Symposium, Tulsa, 22–25 April. DOI: 10.2118/20178–MS.
The authors are grateful to Jean-Noël Jaubert for providing us with Taber, J.J., Martin, F.D., and Seright, R.S. 1997. EOR Screening Criteria
an electronic version of the experimental data sets. The first author Revisited—Part 1: Introduction to Screening Criteria and Enhanced
also would like to acknowledge financial support from the mem- Recovery Field Projects. SPERE 12 (4): 189–198. SPE-35385-PA.
bers of the Stanford University Petroleum Research Institute for DOI: 10.2118/35385-PA.
Gas Injection. Wang, Y. and Orr, F.M. Jr. 1997. Analytical Calculation of Minimum
Miscibility Pressure. Fluid Phase Equilibria 139 (1–2): 101–124.
References Whitson, C.H., Anderson, T.F., and Søreide, I. 1989. C7+ Characterization
Avaulleé, L., Duchet-Suchaux, P., Durandeau, M., and Jaubert, J.N. 2001. of Related Equilibrium Fluids Using the Gamma Distribution. Ad-
A New Approach in Correlating the Thermodynamic Oil Properties. vances in Thermodynamics, vol.1—C7+ Characterization. eds. Chorn
J. Pet. Sci. Eng. 30 (1): 43–65. and Mansoori, 35–56. New York City: Taylor & Francis.
Chorn, L.G. and Mansoori, G.A. 1989. C7+ Fraction Characterization. In Ad- Whitson, C.H. and Brule, M.R. 2000. Phase Behavior. Monograph Series,
vances in Thermodynamics, vol. 1. New York City: Taylor & Francis. SPE, Richardson, Texas, 20.
Jaubert, J.N., Avaullee, L., and Souvay, J.F. 2002a. A Crude Oil Data Bank Zick, A.A. 1986. A Combined Condensing/Vaporizing Mechanism in the
Containing More Than 5,000 PVT and Gas Injection Data. J. Pet. Sci. Displacement of Oil by Enriched Gases. Paper SPE 15493 presented at
Eng. 34 (1–4): 65–107. the SPE Annual Technical Conference and Exhibition, New Orleans,
Jaubert, J.N., Avaullee, L., and Pierre, C. 2002b. Is It Still Necessary to 5–8 October. DOI: 10.2118/15493-MS.
Measure the Minimum Miscibility Pressure? Industrial Engineering
Chemistry Research 41 (2): 303–310.
Kristian Jessen is an assistant professor in the Mork Family De-
Jessen, K., Michelsen, M.L., and Stenby, E.H. 1998. Global Approach for partment of Chemical Engineering and Materials Science at
Calculation of Minimum Miscibility Pressure. Fluid Phase Equilibria the University of Southern California. He holds MS and PhD de-
153 (2): 251–263. grees in chemical engineering from the Technical University of
Jessen, K., Wang, Y., Ermakov, P., Zhu, J., and Orr, F.M. Jr. 2001. Fast Denmark. He is a cofounder of the company Tie-Line Technol-
Approximate Solutions for 1D Multicomponent Gas Injection Prob- ogy, specializing in software/algorithms for modeling phase
lems. SPEJ 6 (4): 442–451. SPE-74700-PA. DOI: 10.2118/74700-PA. behavior and miscibility in relation to EOR. Erling H. Stenby is a
Jessen, K., Stenby, E.H., and Orr, F.M., Jr. 2004. Interplay of Phase Be- professor of applied thermodynamics at the Department of
havior and Numerical Dispersion in Finite Difference Compositional Chemical Engineering, Technical University of Denmark. Since
1994, he has been director of the Center for Phase Equilibria
Simulation. SPEJ 9 (2): 193–201. SPE-88362-PA. DOI: 10.2118/
and Separation Processes (IVC-SEP). He has been a visiting
88362-PA. professor in France and China and a consultant for companies
Knudsen, K. and Stenby, E.H. 1996. PVT Behavior of Reservoir Fluids in several countries. In 2001, he cofounded the company Tie-
Over a Wide Range of Temperatures. Internal report, Technical Uni- Line Technology, specializing in modeling phase behavior and
versity of Denmark: Department of Chemical Engineering, Lyngby. miscibility in relation to gas injection processes.

488 October 2007 SPE Reservoir Evaluation & Engineering

Vous aimerez peut-être aussi