Sei sulla pagina 1di 14

SPE 93478

Use of Genetic Algorithm To Predict Minimum Miscibility Pressure (MMP) Between Flue
Gases and Oil in Design of Flue Gas Injection Project
M.K. Emera, SPE, and H.K. Sarma, SPE, Australian School of Petroleum, U. of Adelaide

Copyright 2005, Society of Petroleum Engineers Inc.


Introduction
This paper was prepared for presentation at the 14th SPE Middle East Oil & Gas Show and CO2 miscible flooding is among the more widely applied
Conference held in Bahrain International Exhibition Centre, Bahrain, 12–15 March 2005.
non-thermal EOR techniques. Among gas injection processes,
This paper was selected for presentation by an SPE Program Committee following review of
information contained in a proposal submitted by the author(s). Contents of the paper, as
CO2 is preferred to hydrocarbon gases (HC) because of its
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to lower cost, high displacement efficiency, and the potential for
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at concomitant environmental benefits through its disposal in
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
petroleum reservoir. Key factors that affect CO2 flooding are
for commercial purposes without the written consent of the Society of Petroleum Engineers is reservoir temperature, oil characteristics, reservoir pressure
prohibited. Permission to reproduce in print is restricted to a proposal of not more than 300
words; illustrations may not be copied. The proposal must contain conspicuous and the purity of injected CO2 itself. Field case histories from
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
CO2 floods in Permian Basin, West Texas suggest that CO2
purity should not be viewed as too rigid constraint, as the use
Abstract of low purity CO2 stream could also be economic and effective
There is an increasing global awareness of the detrimental in enhancing oil recovery. In fact, certain impurities, such as
effects of industrial flue gases on the environment. As a H2S and SOx, could contribute towards attaining CO2-oil
consequence, much emphasis is being placed to harness flue miscibility at lower pressure, while the presence of C1 and N2
gases that contain high CO2 concentrations and sequester them increases the MMP. However, from an operational
in suitable geological formations. A plausible means of perspective, it is often the remaining low percentages of non-
sequestrating flue gases is to inject them into petroleum CO2 gases that are more difficult and costly to remove,
reservoirs while also enhancing oil recovery. An a priori requiring expensive gas separation facilities. Safety and
understanding of the pressure at which various flue gas compression cost considerations, also, justify near-miscible
components, notably CO2, become miscible with reservoir CO2 flood applications for some reservoirs. Therefore, the
fluids is critical to the design and implementation of a flue gas potential of injecting flue gases containing both CO2 and non-
injection project. CO2 components (H2S, N2, SOx, O2, and C1-C4) could be an
A new genetic algorithm (GA)-based correlation has been attractive option, provided the flue gas composition does not
developed to estimate the flue gas-oil MMP. In developing affect the process performance adversely and their overall
this correlation, the GA software developed in our earlier work impact on miscibility with the oil, separation/purification at
has been modified to account for various component gases in the surface, and subsequent re-injection is evaluated and well
the flue gas stream. The correlation estimates the MMP as a understood.
function of the injected gas solvency in the oil. The solvency, The objective of this study is to develop a more reliable
in turn, is related to critical properties of the injected gas. The flue gas-oil MMP correlation using the GA approach and to
correlation has been successfully validated against published compare its efficiency against other commonly used
experimental data and several correlations in the literature. It correlations as listed in Table 1. The software designed in our
yielded the best match with an average error of 4.7% and a earlier work (1) to develop a MMP correlation for pure CO2 and
standard deviation of 6.3%, followed by Sebastian et al. oil has been modified to account for flue gases, which also
correlation with 13.1% error and 22.0% standard deviation and contain non-CO2 components. The GA software used in this
Alston et al. correlation with 14.1% error and 43.2% standard study has been presented in the flowchart provided in Fig. 1.
deviation. This figure also presents the stopping criterion under which
An advantage of the GA-based correlation over other the fitness of the solution is decided and accepted.
correlations is that it can be used for gas mixtures with higher
N2 concentrations (tested up to 20 mole%) and with non-CO2 Factors Affecting the Flue Gas-Oil MMP
component concentrations of H2S, N2, SOx, O2, and C1-C4 up The factors that affect flue gas-oil MMP are reservoir
to 78 mole% with a higher accuracy. Equally important, it temperature, oil characteristics, and injected gas composition.
could be a useful tool when experimental data are not For reservoir temperature, there exists a distinct
available and/or when developing an optimal and economical proportionality between reservoir temperature and MMP
laboratory program to estimate the flue gas-oil MMP. because MMP increases as the reservoir temperature
increases (2). The MMP also increases with high molecular
2 SPE 93478

weight (MW) oil and oils containing higher concentrations of uses the weight-fraction mixing rule in determining
C1 and N2. However, the presence of intermediates (C2-C4) pseudocritical pressure and temperature. A multiplying factor
decreases the MMP. (MFi) was used to adjust the injected gas pseudocritical
The existence of non-CO2 components (e.g., H2S, SOx, and temperature to present the best fit among the pseudocritical
C2-C4) whose critical temperatures are higher than that of CO2 properties (pressure and temperature) and MMP. This factor
(31°C) causes an improvement in the solubility of flue gas in was used, with different values (based on the gas component),
the reservoir oil (3, 4). This results in an increased injected gas for the components: SO2, H2S, C2, C1, and N2. The GA-based
pseudocritical temperature and a lower MMP. On the other flue gas-oil MMP correlation and the MFi values are presented
hand, the existence of the components (e.g., N2, O2, and C1) as follows:
with lower critical temperatures causes a reduction in the
solubility of flue gas in the reservoir oil and has the opposite Pr, flue gas 1.8TCW + 32 1.8TCW + 32 2
effect. = 3.406 + 5.786 × ( ) - 23.0 × ( )
Pr, CO2 1.8TC, CO2 + 32 1.8TC, CO2 + 32
Wilson (5) stated that the pseudocritical temperature of the
injected gas affects MMP and it could be used as a parameter 1.8TCW + 32 3 1.8TCW + 32 4
in a miscibility correlation. Likewise, Rutherford (6) found, + 20.48 × ( ) - 5.7 × ( )
1.8TC, CO2 + 32 1.8TC, CO2 + 32
empirically, that the HC gas-oil MMP in HC miscible floods
was a function of the pseudocritical temperature of the (1)
injected gas at a constant pressure. Jacobson (7), also, suggested
a similar scheme of using the pseudocritical temperature as a where,
correlation parameter for acid gases (CO2 with H2S)-oil MMP MMP flue gas
prediction. However, instead of using actual values, apparent Pr, flue gas =
P CW
critical temperatures were used for non-HC components as
correlation parameters. Alston et al. (2) followed a similar n
PCW = ∑ w i PCi
approach to correlate flue gas-oil MMP using the i=1
pseudocritical temperature of the injected gas, where apparent
MMPCO2
critical temperatures for C2 and H2S components were also Pr, CO2 =
used to determine the pseudocritical temperature using the PC,CO2
weight-fraction mixing rule. They found that the weight-
fraction mixing rule gave better results than the mole-fraction n
TCW = ∑ MFi w T
method. Similarly, Kovarik (8) presented a correlation that is i=1 i ci
also based on the pseudocritical temperature. In addition to the
weight-fraction mixing rule, he used the mole-fraction rule to Values of MFi are as below:
determine the pseudocritical temperature and found that the Components MFi (multiplying factor)
two methods presented similar results. Moreover, Sebastian et SO2 0.3
al. (9), also, used the mole-fraction mixing rule to determine the H 2S 0.59
injected gas pseudocritical temperature in developing their CO2 1.0
flue gas-oil MMP correlation. They also used an apparent C2 1.1
critical temperature (51.67°C) for H2S. Dong (3) presented a C1 1.6
similar approach to that of Sebastian et al. but instead of using N2 1.9
apparent critical temperatures; he used a factor with non-CO2 All other injected gas
components (H2S, SO2, O2, N2, and C1) in determining the components 1.0
injected gas pseudocritical temperature to represent the
strength of these components in changing the apparent critical GA-based Flue Gas-Oil MMP Correlation Test
temperature of the injected flue gas relative to pure CO2. The GA-based flue gas-oil correlation was tested against
Besides the pseudocritical temperature, Eakin and the literature data and other correlations in the literature
Mitch (10), also, used the injected gas pseudocritical pressure in (Table 2 and Fig. 2 present the literature data used in this
developing their correlation, as they stated that the change in study and the GA-based correlation prediction results). It is
the rising bubbles in the rising bubble apparatus (RBA) and found that GA-based correlation yields the best prediction
also the low value of the interfacial tension between the among all other tested correlations (2,3,8,9,10). Our correlation
injected gas and crude oil would occur only near the critical yielded an average error of 4.7% and standard deviation (of
point. calculated MMP/experimental MMP) equal to 6.3%. On the
other hand, Alston et al. correlation presented an average error
GA-based Flue Gas-Oil MMP Correlation equal to 14.1% and standard deviation equal to 43.2%. Also,
The GA-based correlation to predict flue gas-oil MMP is Sebastian et al. correlation presented an average error equal to
based on the injected flue gas pseudocritical properties 13.1% and standard deviation equal to 22.0%. Table 3
(pseudocritical temperature and pressure), in addition to the presents a comparison among our GA-based correlation
CO2 pseudocritical properties, and CO2-oil MMP (could be accuracy and that of the two correlations with a closer match
determined experimentally or through the CO2-oil MMP to ours (e.g., Alston et al. and Sebastian et al.). This
correlations that available in the literature). This correlation comparison is also presented in Fig. 3 and Fig. 4. Table 3,
SPE 93478 3

also, extends the comparison to other tested correlations, and accuracies in data gathered from the literature and
notably those of Kovarik, Dong, and Eakin-Mitch correlations. experiments (due to use of different experimental apparatuses,
These correlations yielded lower accuracy. different miscibility criteria, and experimental
It is possible to note from Table 3 that the maximum error procedures/errors including larger incremental pressure steps
of Alston et al. correlation prediction is 311.6% (for the to determine MMP), the GA-based flue gas-oil MMP was
injected flue gas that is composed of 80 mole% of CO2 and 20 successful in yielding the best prediction of the flue gas-oil
mole% of N2). This high error occurs because Alston et al. MMP among all other correlations tested during the course of
correlation should not be used for flue gas streams containing this study.
more than 8 mole% of N2. However, if this particular data
point (as identified within the circle in Fig. 3) is excluded Conclusions
from our accuracy test, the GA-based correlation still yields a A GA-based correlation has been developed to predict the
better accuracy with an average error of 4.8% and standard flue gas-oil MMP and has been successfully validated using
deviation equal to 6.34% compared to that of Alston et al. data available in the public domain. Based on preceding
correlation, which results in average error of 9.4% and discussions, the following conclusions are drawn:
standard deviation equal to 18.2%. Sebastian et al. correlation
gives an average error of 12.7% with a standard deviation 1. The flue gas-oil MMP is a function of the solvency of
equal to 21.45%. As shown in Table 3, the other tested the injected flue gas into reservoir oil. The injected
correlations yield even higher errors. gas solvency into oil depends on the critical
properties (pseudocritical temperature and pressure)
Discussion of the injected gas and the GA-based correlation
The GA-based correlation to predict flue gas-oil MMP accounts for these properties for both injected flue
could be superior to other correlations commonly used. This gas and pure CO2.
correlation is based on the solvency of the injected gas into 2. The use of the multiplying factor improves the
reservoir oil. As injected gas solvency into oil depends on the relationship between MMP and pseudocritical
critical properties (pseudocritical temperature and pressure) of properties of the injected gas.
the injected gas, so the GA-based correlation is a function of 3. The GA-based correlation presented an average error
the pseudocritical temperatures and pressures of the injected of 4.7% with a standard deviation (of calculated
flue gases and pure CO2 besides CO2-oil MMP. A multiplying MMP/experimental MMP) equal to 6.3%. On the
factor was used in developing this correlation to present a other hand, Alston et al. correlation presented an
better relationship between MMP and the pseudocritical average error of 14.1% and standard deviation equal
properties of the injected gas. The difference between this to 43.2%, while Sebastian et al. correlation presented
relationship before and after using this multiplying factor in an average error of 13.1% and standard deviation
the correlation development is explained in Fig. 5 and Fig. 6, equal to 22.0%. The errors with other tested
respectively. With the exception of C2, the multiplying factor correlations were much higher.
is qualitatively proportional to the equilibrium constants of the 4. GA-based correlation could be used for higher N2
injected gas non-CO2 components (e.g. SO2, H2S, C1, and N2). content (tested up to 20 mole% based on the available
From Table 3, it is evident that the GA-based correlation data) and higher non-CO2 component fractions
presents better prediction accuracy with different types of non- (tested up to 78 mole%).
CO2 components that may co-exist with CO2 in the flue gas 5. When data from standard or similar experimental
stream (e.g., H2S, N2, SOx, O2, and C1-C4). This correlation procedures are used, the accuracy of the GA-based
was tested for the existence of up to 78 mole% of non-CO2 correlation could be further enhanced. In such
components and up to 20 mole% of N2 in the injected gas and situations, it could be used with even greater
presented high prediction accuracy. On the other hand, confidence and with lower error.
although Alston et al. correlation presented adequate results 6. In the absence of any measured site-specific MMP
for the existence of different types of non-CO2 components data, GA-based correlation could be used as an
(e.g., C1- C4 and H2S) in the injected gas, its performance for effective and convenient predictive tool to
flue gas streams that contain N2 was, however, inadequate; guesstimate the MMP for initial design calculations.
especially for injected gas streams containing more than 8 In addition, it could also contribute towards designing
mole% of N2 (this is one of its limitations). Furthermore, a more efficient and economical experimental
Alston et al. correlation is unable to deal with the existence of programs.
SO2 in the flue gas. For Sebastian et al. correlation, its
performance for flue gases that contain N2 and/or SO2 was not Nomenclatures
adequate. Also, the existence of high fractions of non-CO2 ρMMP CO2 density at MMP, g/cm3
components (more than 55 mole%) in the injected flue gas API Oil gravity °API
presented lower prediction accuracy. For other tested Ci Carbon number
correlations, their errors were much higher than the foregoing Exp. MMP Experimental minimum miscibility pressure,
correlations (GA-based, Alston et al., and Sebastian et al. MPa
correlations). F Weighting composition parameter
Notwithstanding above general observations and the SFi Dong (1999) factor representing the strength
caution that must be applied with regard to the uncertainties
4 SPE 93478

of species i in changing the apparent critical TR Reservoir temperature, °C


temperature of the mixture relative to the TRes. Reservoir temperature, °K
critical temperature of pure CO2 wi Weight fraction of component i, fraction
Fit (i) Average fitness of chromosome i, where the wic2+ Normalized weighting fraction of component i
chromosome has many fitness values based on in the C2+ fractions of crude oil
the number of data available (j) xi Mole fraction of gas component i
MFi Multiplying factor of the injected gas xCO2 Mole percentage of CO2 in the injection gas, %
component i y Mole fraction of diluted component
FR Mole percent of C2 to C6 in the reservoir fluid, yC2+ Ethane plus HC mole fraction in the injected
% gas, fraction
I Oil characterization index yi Mole fraction of the gas component i in the
Ki Normalized partition coefficient for carbon injected gas (yCO2, yH2S, yN2, and yC1) fraction
number i
M Average molecular weight of the oil
Minj. Molecular weight of the injected gas 1 n ABS(MMPexp. − MMPcal.)
Average error = ∑( ) × 100%
MMPcal. Calculated minimum miscibility pressure, MPa n i=1 MMPexp.
MMPcal. (i,j) Calculated value of CO2-oil MMP of
chromosome i based on the variables’ values
of the data number j 2
n ⎡ ABS(MMPexp. − MMPcal. ) ⎤
MMPexp. Experimental minimum miscibility pressure, ∑ ⎢( ) × 100⎥
MPa i=1⎢ MMPexp. ⎥
MMPexp. (i,j) Experimental value of CO2-oil MMP of STD = ±
⎣ ⎦ %
chromosome i based on the variables’ values n −1
of the data number j
MMPflue gas Flue gas-oil MMP, MPa Acknowledgement
MMPCO2 CO2-oil MMP, MPa The authors would like to thank Santos Limited for its
MWC7+ C7+molecular weight support to research on CO2 EOR process within the Center for
n Number of non-CO2 components & GA Improved Petroleum Recovery at Australian School of
population size Petroleum, University of Adelaide. The first author is a
nn Number of available data points recipient of the Santos Post-Graduate-Scholarship. The
P (m) Mutation probability authors also thank Prof. Sam Huang of Saskatchewan
P (c) Crossover probability Research Council (SRC), Canada for providing some of the
PC, CO2 CO2 critical pressure, MPa data used in this study.
PC, inj. Injected gas critical pressure, MPa
PCi Critical pressure of the gas component i, MPa References
Pcm Mole average pseudocritical pressure, MPa 1. Emera, M.K. and Sarma, H.K.: “Use of Genetic
PCW Weight average pseudocritical pressure, MPa Algorithm to Estimate CO2-Oil Minimum Miscibility
Pfit(i,j) Fitness function of data number j of Pressure–A Key Parameter in Design of CO2
chromosome i, fraction Miscible Flood”, Journal of Petroleum Science and
Pr, flue gas Reduced flue gas minimum miscibility Engineering, In Press, Article no. 1232, 2004.
pressure, fraction 2. Alston, R.B., Kokolis, G.P. and James, C.F.: “CO2
Pr, CO2 Reduced CO2 minimum miscibility pressure, Minimum Miscibility Pressure: A correlation for
fraction Impure CO2 Streams and Live Oil Systems”, SPEJ,
STD Standard deviation, % pp268-274, April, 1985.
Tac Mole average pseudocritical temperature with 3. Dong, M.: “Potential of Greenhouse Gas Storage and
using factor SFi, °K Utilization through Enhanced Oil Recovery -Task 3:
TC, CO2 Critical temperature of pure CO2 gas, °C Minimum Miscibility Pressure (MMP) Studies”,
TC, inj. Injected gas critical temperature, °K Final Report (SRC Publication No. P-110-468-C-
Tci Critical temperature of the gas component i, 99), Sep. 1999.
°C 4. Zhang, P.Y., Huang, S., Sayegh, S., Zhou, X.L.:
TCi Critical temperature, °K “Effect of CO2 impurities on gas injection EOR
Tcm Mole average pseudocritical temperature, °C process”, SPE paper 89477 presented at the
SPE/DOE Fourteenth Symposium on Improved Oil
TCM Mole average pseudocritical temperature, °K
Recovery held in Tulsa, OK, USA, April 17-21,
TCW Weight average pseudocritical temperature
2004.
with using the Multiplying factor (MFi), °C
5. Wilson, J.F.: “Miscible Displacement-Flow Behavior
Tcw Weight average pseudocritical temperature, °C and Phase Relationships for a Partially Depleted
Tpc Pseudocritical temperature (may be weight Reservoir”, Trans., AIME Vol. 219, 223, 1960.
average or mole average), °C
Tr Reduced reservoir temperature, °C
SPE 93478 5

6. Rutherford, W.M.: ”Miscibility Relationships in the 16. Dicharry, R.M., Perryman, T.L., and Ronquille, J.D.:
Displacement of Oil by Light Hydrocarbons”, SPEJ, “Evaluation and Design of a CO2 Miscible Flood
pp340-346, Dec. 1962. Project-SACROC Unit, Kelly-Snyder Field”, JPT,
7. Jacobson, H.A.: “Acid Gases and their Contribution pp1309-1318, Nov. 1973.
to Miscibility”, J. Can. Pet. Tech., 56, April 1972. 17. Metcalfe, R.S.: “Effects of Impurities on Minimum
8. Kovarik, F.S.: “A Minimum Miscibility Pressure Miscibility Pressures and Minimum Enrichment
Study Using Impure CO2 and West Texas Oil Levels for CO2 and Rich-Gas Displacements”, SPEJ,
Systems: Data Base, Correlations, and Compositional pp219- 225, April 1982.
Simulation”, SPE paper 14689 presented at the SPE 18. Holm, L.W. and Josendal, V.A.: “Mechanisms of Oil
Production Technology Symposium held in Lubbock, Displacement by Carbon Dioxide”, JPT, pp1427-
Texas, November 11-12, 1985. 1436, Dec. 1974.
9. Sebastian, H.M., Wenger, R.S., and Renner, T.A.: 19. Zick, A.A.: “A Combined Condensing/Vaporizing
“Correlation of Minimum Miscibility Pressure for Mechanism in the Displacement of Oil by Enriched
Impure CO2 Streams”, JPT, pp2076-2082, Nov. 1985. Gases”, SPE paper 15493, presented at the 61st
10. Eakin, B.E. and Mitch, F.J.: “Measurement and Annual Conference and Exhibition of the SPE held in
Correlation of miscibility Pressures of Reservoir New Orleans, LA, October 5-8, 1986.
Oils”, SPE paper 18065 presented at the 63rd Annual 20. Cardenas, R.L., Alston, R.B., Nute, A.J., Kokolis,
Technical Conference and Exhibition of SPE held in G.P.: “Laboratory Design of a Gravity-Stable
Houston, TX, October 2-5, 1988. Miscible CO2 Process”, JPT, 111-118, Jan. 1984.
11. Johnson, J.P. and Pollin, J.S.: “Measurement and 21. Dong, M., Huang, S., Dyer, S.B., Mourits, F.M.: “A
Correlation of CO2 Miscibility Pressures”, SPE paper Comparison of CO2 minimum miscibility pressure
9790, presented at the 1981 SPE/DOE Joint determinations for Weyburn crude oil”, Journal of
Symposium on Enhanced Oil Recovery held in Tulsa, Petroleum Science and Engineering, 31, pp13-22,
Oklahoma, April 5-8, 1981. 2001.
12. Orr, F.M. Jr. and Silva, M.K.: “Effect of Oil 22. Dong, M, Huang S.S., Srivastava, R.: “A laboratory
Composition on Minimum Miscibility Pressure-Part study on near-miscible CO2 injection in Steelman
2: Correlation”, SPEJ, pp479-491, Nov. 1987. reservoir”, JCPT, Vol. 40, No.2, pp53-61, Feb. 2001.
13. Enick, R.M., Holder, G.D., Morsi, B.I.: “A 23. Stalkup Jr., F.I.: “Miscible Displacement” SPE
Thermodynamic Correlation for the Minimum Monograph Series, pp141, 1984.
Miscibility Pressure in CO2 Flooding of Petroleum
Reservoirs”, SPERE, pp81-92, Feb. 1988. SI Metric Conversion Factors
14. Yuan, H., Johns, R.T., Egwuenu, A.M., Dindoruk, B.: °API 141.5/(131.5+°API) = g/cm3
“Improved MMP Correlations for CO2 Floods Using ft3 × 2.831 685 E-02 = m3
Analytical Gas Flooding Theory”, SPE paper 89359 °F (°F-32)/1.8 = °C
presented at the SPE/DOE Fourteenth Symposium on psia × 6.894 757 E-00 = kPa
Improved Oil Recovery held in Tulsa, OK, USA, lb × 2.20462 =kg
April 17-21, 2004.
15. Graue, D.J. and Zana, E.T.: “Study of a possible CO2
Flood in Rangely Field”, JPT, pp1312-1318, July
1981.
6 SPE 93478

Table 1 Commonly used flue gas-oil MMP correlations.

Author Correlation Remarks


2
MMP − PC,inj. = α inj. (TRes. − TC,inj. ) + I(β( − M inj )
−2 −4 2
I = −11.73 + 6.313 × 10 × M − 1.954 × 10 × M +
−7 3 −5
2.502 × 10 × M + (0.1362 + 1.138 × 10 × M)API − Limitations:
−5 2 -Temperature range from 26.85 to 136.85 °C,
7.222 × 10 × API - Used for Less than 10 mole% of non-CO2
Johnson & Pollin where, β = 0.285 and for pure CO2, αinj.=0.13 MPa/°K, components in the injection flue gas,
(1981)(11) 3 -Used for C1 and/or N2 as non-CO2
10 y components only.
-For N2 impurity α inj. = 0.0722 (1.8 + ),
TRes. − TC,inj.
2
10 y
- For C1 impurity α inj. = 0.0722 (1.8 + ).
TRes. − TC,inj.

MMPflue gas ( 1.935×87.8 ) -H2S and C2 critical temperatures are modified


87.8 1.8Tcw +32
=( ) ,where: to 51.67°C.
Alston et al. (1985)(2) MMPCO2 1.8Tcw + 32 -If MMP < Pb, the Pb is taken as MMP
Limitation:
-Used only for not more than 8 mole% of N2.
Tcw = ∑ w i × Tci ,

MMPflue gas
−2
= 1.0 − 2.13 × 10
(TcM − 304.2) +
-H2S critical temperature is modified 325°K
MMPCO2
(51.67°C)
−4 −7
(9)
Sebastian et al. (1985) 2 3
2.51× 10 (TcM − 304.2) − 2.35 × 10 (TcM − 304.2) Limitation:
- Used for non-CO2 ratios up to 55 mole%.
n
CM = i∑ x × TCi
where, T
=1 i
-Kovarik (8) found that the weight average
MMPflue gas = 0.2814 × (548 − (1.8Tpc + 492)) + MMPCO2 fraction properties correlated the data similar
to the mole average fraction properties.
Kovarik (1985)(8) where, -Limitation:
n n This correlation is used only for C1, as a non-
Tpc = ∑ w i × Tci or Tpc = ∑ x × T CO2 component, with less than 20 mole%
i =1 i =1 i ci
ratios.
ρ MMP = −0.524 × F + 1.189 , when: F p 1.467
-MMP could be obtained at the reservoir
ρ MMP = 0.42 , when: F f 1.467 , temperature and ρMMP.
-This correlation is used for CO2 and flue gas-
where: oil MMP prediction.
37 Limitations:
F = ∑ K i × w ic2+ -This correlation does not take into
Orr and Silva (1987)(12) 2 consideration the presence of C1 and all other
non-hydrocarbons in the oil composition and
and log(K ) = 0.7611 − 0.04175 × C can be used only when the carbon distribution
i i
from C2 to C37 is available.
wi -Uncertain for TR>87.78°C for oils containing
and w ic2+ = significant amount of dissolved gas.
37
∑ wi If MMP < Pb, the Pb is taken as MMP.
2
-Molecular weight of C5+ is modeled as a
single alkane of equivalent molecular weight.
A graphical correlation that is a function of the reservoir temperature and molecular Limitations:
weight of C5+ with a graphical corrections for light gas (C1 and N2) and intermediates
-35°C≤TR≤115°C
Enick et al. (1988)(13) (C2 to C4) that exist in the oil and another correction factor for the non-CO2
-156≤MWc5+ ≤256
components in injected flue gas which is based on the mole fraction of these
components that may exist in the injected gas streams. -7 MPa ≤MMP≤30 MPa
-Can be used up to 25 mole% of non-CO2
components in the flue gas.
SPE 93478 7

Table 1 Commonly used flue gas-oil MMP correlations…Contd.

Author Correlation Remarks


1
0.06912 0.005955(M WC7 + )
lnP r, flue gas = (0.1697 − )y C1 (MW C7 + ) 2 + (2.3855 − )y C2 + +
Tr Tr
3
1
0.01023 0.0005899( MW C7 + ) 2
(0.1776 − )y N2 (MW C7 + ) 2 + (0.01221(M WC7 + ) − )y CO2 + -Eakin and Mitch (10) generally considered
Tr Tr MMP to correspond to the critical point of the
101.429 0.00375(MW )
C7 + )y flue gas-oil mixture, where the rapid change in
( + H2S
MW C7 + Tr the bubbles in the (RBA), low values of
Eakin and Mitch interfacial tension between gas and liquid, and
(1988)(10) where, first contact miscibility may be achieved.
MMPflue gas n Limitation:
Pr = , Pcm = ∑ x i PCi -This correlation does not take into
Pcm i=1 consideration the presence of some non-CO2
components, e.g., SO2 and O2.
1.8TR + 492 n
Tr = , Tcm = ∑ x × T
1.8Tcm + 492 i −1 i ci

-Dong (3) used a factor (SFi) representing the


strength of species i in changing the apparent
MMPflue gas Tac 4 critical temperature of the flue gas relative to
=( ) the critical temperature of CO2 (this factor is
MMPCO2 304.2 consisted with component K-values).
Dong (1999)(3) -The values of SFi for the different
where, components are as follows:
n -For H2S, SFi = 0.7 - For SO2, SFi = 0.5
Tac = ∑ SFi x i TCi -For C1, SFi = 2.5 - For O2, SFi = 5.0
i=1 -For N2, SFi = 7.5 - For CO2, SFi = 1.0
- For other non-CO2 components, SFi = 1.0
-This correlation was developed based on an
analytical theory for MMP calculation from an
For flue gas-oil MMP (CO2 with methane): equations-of-state (EOS) to predict MMP
(used for the flue gas that composed of CO2
MMPflue gas and C1 only). This correlation is based on
= 1.0 + m(x CO2 − 100) MWC7+, reservoir temperature, mole
MMPCO2 percentage of C2-C6 in the oil, and mole
percentage of CO2 in the flue gas streams.
Yuan et al. (2004)(14) where, For flue gas-oil MMP:
m = a 1 + a 2 MWC7 + + a 3 FR + (a 4 + a 5 MWC7 + + -a1= -6.5996E-02 -a2= -1.5246E-04
-a3= 1.3807E-03 -a4= 6.2384E-04
FR 2 2 -a5= -6.7725E-07 -a6= -2.7344E-02
a6 )TR + (a 7 + a 8 MWC7 + + a 9 MWC7 + + a 10 FR )TR -a7= -2.6953E-06 -a8= 1.7279E-08
2
MWC7 + -a9= -3.1436E-11 -a10= -1.9566E-08
Limitations:
-Used for injection streams contain up to 40
mole% of C1

Table 2 Literature experimental data and GA-based correlation prediction of the flue gas-oil MMP.

Actual flue
Flue gas Composition CO2-oil MMP, GA-based GA-based
Reference TCW, °C gas-oil MMP,
(Mole, %) MPa MMP, MPa error, %
MPa
Ref. 2 92.5% of CO2, 7.5% of C1 26.69 9.48 10.35 10.77 4.1
Ref. 2 90% of CO2, 10% of C1 25.10 11.13 13.10 13.49 2.9
92.25% of CO2, 7.75% of n-
Ref. 2 43.15 24.15 19.69 18.05 8.3
C4
Ref. 2 95% of CO2, 5% of n-C4 38.92 23.45 18.62 18.51 0.6
Ref. 2 90.5% of CO2, 9.5% of C3 37.28 23.45 18.62 18.76 0.8
87.5% of CO2, 6.3% of N2,
Ref. 2 29.20 25.14 23.17 24.49 5.7
6.2 % of n-C4
86.4% of CO2, 10.7% of C1,
Ref. 2 29.76 24.28 23.10 23.81 3.0
2.9% of n-C4
95% of CO2, 4.9% of C1,
Ref. 15 28.07 15.52 16.83 16.71 0.7
0.1% of N2
Ref. 16 90% of CO2, 10% of C1 25.10 11.04 12.76 13.37 4.8
8 SPE 93478

Table 2 Literature experimental data and GA-based correlation prediction of the flue gas-oil MMP…Contd.

Actual flue
Flue gas Composition CO2-oil MMP, GA-based GA-based
Reference TCW, °C gas-oil MMP,
(Mole, %) MPa MMP, MPa error, %
MPa
Ref. 17 75% of CO2, 25% of H2S 35.27 8.28 7.53 7.48 0.7
Ref. 17 50% of CO2, 50% of H2S 40.02 8.28 6.55 7.28 11.0
Ref. 17 90% of CO2, 10% of C1 25.10 8.28 11.04 10.03 9.1
45% of CO2, 10% of C1, 45%
Ref. 17 33.01 8.28 8.83 8.12 8.0
of H2S
60% of CO2, 20% of C1, 20%
Ref. 17 21.47 8.28 14.07 12.10 14.0
of H2S
67.5% of CO2, 10% of C1,
Ref. 17 28.82 11.72 12.41 12.66 2.0
22.5% of H2S
45% of CO2, 10% of C1, 45%
Ref. 17 33.01 11.72 10.38 11.51 10.9
of H2S
60% of CO2, 20% of C1, 20%
Ref. 17 21.47 11.72 17.24 17.14 0.6
of H2S
Ref. 17 90% of CO2, 10% of C2 31.47 11.04 10.07 10.33 2.6
Ref. 17 90% of CO2, 10% of C3 37.61 11.04 9.31 8.77 5.9
Ref. 17 90% of CO2, 10% of C4 44.35 11.04 7.90 8.05 1.9
Ref. 17 80% of CO2, 20% of C2 31.92 11.04 9.66 9.92 2.7
Ref. 17 80% of CO2, 20% of C3 44.17 11.04 7.93 7.88 0.6
Ref. 17 90% of CO2, 10% of C2 31.47 13.45 13.04 12.59 3.4
Ref. 17 90% of CO2, 10% of C3 37.61 13.45 11.04 10.68 3.2
Ref. 17 90% of CO2, 10% of C4 44.35 13.45 8.97 9.81 9.4
Ref. 17 80% of CO2, 20% of C2 31.92 13.45 12.88 12.08 6.1
Ref. 17 80% of CO2, 20% of C3 44.17 13.45 10.50 9.61 8.5
Ref. 17 80% of CO2, 20% of C1 18.35 8.28 14.83 13.49 9.0
68% of CO2, 22% of H2S,
Ref. 17 28.74 8.28 10.28 8.96 12.8
10% of C1
40% of CO2, 40% of H2S,
Ref. 17 24.99 8.28 12.06 10.72 11.1
20% of C1
Ref. 17 75% of CO2, 25% of H2S 35.27 11.72 10.35 10.59 2.4
Ref. 17 50% of CO2, 50% of H2S 40.02 11.72 8.97 10.31 14.9
Ref. 17 90% of CO2, 10% of C1 25.10 11.72 15.17 14.20 6.4
Ref. 17 80% of CO2, 20% of C1 18.35 11.72 18.74 19.11 2.0
55% of CO2, 25% of H2S,
Ref. 17 22.18 11.72 16.45 16.74 1.8
20% of C1
Ref. 18 90% of CO2, 10% of C1 25.10 13.10 16.21 15.87 2.1
22.18% of CO2, 23.49% of C1,
Ref. 19 23.5% of C2, 27.4 % of C3, 44.66 30.19 16.55 16.77 1.3
3.38% of C4
79.2% of CO2, 8.8% of N2,
Ref. 20 33.70 25.14 23.17 20.19 12.9
12% of n-C4
94.1% of CO2, 3.1% of N2,
Ref. 21 23.64 12.80 14.50 16.57 14.3
2.8% of C1
Ref. 21 90.1% of CO2, 9.9% of C1 24.95 12.80 16.01 15.61 2.5
89.8% of CO2, 5.1% of N2,
Ref. 21 18.00 12.80 20.51 21.34 4.1
5.1% of C1
Ref. 3 90% of CO2, 10% of H2S 32.70 12.01 11.60 11.28 2.8
Ref. 3 80% of CO2, 20% of H2S 34.38 12.01 11.40 10.97 3.8
Ref. 3 70% of CO2, 30% of H2S 36.19 12.01 10.40 10.74 3.3
Ref. 3 90% of CO2, 10% of SO2 31.58 12.01 10.50 11.58 10.2
Ref. 3 85% of CO2, 15% of N2 1.62 12.01 33.01 36.54 10.7
65% of CO2, 15% of N2, 20%
Ref. 3 5.05 12.01 34.01 33.81 0.6
of SO2
60% of CO2, 10% of N2, 30%
Ref. 3 15.49 12.01 23.00 22.95 0.2
of SO2
90% of CO2, 5% of O2, 5% of
Ref. 3 15.73 12.01 23.00 22.13 3.8
N2
SPE 93478 9

Table 2 Literature experimental data and GA-based correlation prediction of the flue gas-oil MMP…Contd.

Actual flue
Flue gas Composition CO2-oil MMP, GA-based GA-based
Reference TCW, °C gas-oil MMP,
(Mole, %) MPa MMP, MPa error, %
MPa
80% of CO2, 5% of O2, 5% of
Ref. 3 17.01 12.01 22.50 21.12 6.1
N2, 10% of SO2
Ref. 22 94.4% of CO2, 5.6% of C2 31.29 16.50 16.01 15.67 2.1
Ref. 22 70.8% of CO2, 29.2% of C2 32.37 16.50 14.50 14.24 1.8
Ref. 22 60% of CO2, 40% of C2 32.92 16.50 13.60 13.54 0.5
Ref. 22 46.3% of CO2, 53.7% of C2 33.69 16.50 11.90 12.62 6.0
Ref. 22 97% of CO2, 3% of C3 33.02 16.50 14.80 14.91 0.7
Ref. 22 94.9% of CO2, 5.1% of C3 34.40 16.50 14.30 14.25 0.3
Ref. 22 88.4% of CO2, 11.6% of C3 38.66 16.50 13.00 12.82 1.4
Ref. 22 84.1% of CO2, 15.9% of C3 41.48 16.50 12.40 12.23 1.4
Ref. 22 80.4% of CO2, 19.6% of C3 43.91 16.50 11.40 11.82 3.7
Ref. 22 75.7% of CO2, 24.3% of C3 46.99 16.50 11.00 11.31 2.8
Ref. 22 68.3% of CO2, 31.7% of C3 51.85 16.50 10.30 10.05 2.5
Ref. 22 63.4% of CO2, 36.6% of C3 55.07 16.50 9.10 8.56 6.0
Ref. 23 80% of CO2, 20% of C1 18.35 8.28 13.79 13.49 2.2
Ref. 23 80% of CO2, 20% of N2 -9.26 8.28 28.97 29.27 1.0
Average Error, % 4.7

Table 3 Comparison among the GA-based correlation prediction accuracy and that of other correlations to predict flue gas-oil MMP.

Sebastian Alston Alston Eakin &


Flue gas Actual GA-based GA- Sebastian Dong Dong Kovarik Kovarik Eakin &
et al. et al. et al. Mitch
Reference composition MMP, MMP, based et al. MMP, error, MMP (mole) Mitch
MMP MMP, error, MMP,
(Mole, %) MPa MPa error, % error, % MPa % (mole), MPa error, % error, %
MPa MPa % MPa
92.5% of CO2,
Ref. 2 10.35 10.77 4.1 11.38 10.0 10.96 5.9 11.20 8.3 13.9 34.6 - -
7.5% of C1
90% of CO2,
Ref. 2 13.10 13.49 2.9 14.19 8.3 13.60 3.8 13.88 5.9 17 29.8 - -
10% of C1
92.25% of
Ref. 2 CO2, 7.75% of 19.69 18.05 8.3 19.86 0.8 17.13 13.0 27.27 38.5 19.5 0.8 - -
n-C4
95% of CO2,
Ref. 2 18.62 18.51 0.6 20.64 10.8 18.27 1.9 25.37 36.2 20.5 10.1 - -
5% of n-C4
90.5% of CO2,
Ref. 2 18.62 18.76 0.8 20.57 10.4 19.05 2.3 25.43 36.6 20.4 9.6 - -
9.5% of C3
87.5% of CO2,
Ref. 2 6.3% of N2, 23.17 24.49 5.7 27.22 17.5 22.68 2.1 45.16 94.9 27.1 17.1 - -
6.2 % of n-C4
86.4% of CO2,
Ref. 2 10.7% of C1, 23.10 23.81 3.0 29.21 26.4 24.07 4.2 32.08 38.8 28.8 24.6 - -
2.9% of n-C4
95% of CO2,
Ref. 15 4.9% of C1, 16.83 16.71 0.7 17.55 4.3 17.06 1.4 17.46 3.7 18.6 10.2 20.4 21.2
0.1% of N2
90% of CO2,
Ref. 16 12.76 13.37 4.8 14.07 10.2 13.48 5.7 13.76 7.8 16.9 32.6 18.8 47.1
10% of C1
75% of CO2,
Ref. 17 7.53 7.48 0.7 7.42 1.4 7.13 5.4 7.16 4.9 -0.3 104.1 17.5 131.7
25% of H2S
50% of CO2,
Ref. 17 6.55 7.28 11.0 6.68 1.9 6.26 4.5 6.17 5.9 -9 237.7 20.9 218.3
50% of H2S
90% of CO2,
Ref. 17 11.04 10.03 9.1 10.55 4.4 10.11 8.4 10.32 6.5 14.2 28.3 20.4 84.5
10% of C1
45% of CO2,
Ref. 17 10% of C1, 8.83 8.12 8.0 8.65 2.0 7.25 17.9 8.04 8.9 -1.5 117.3 20 126.9
45% of H2S
10 SPE 93478

Table 3 Comparison among the GA-based correlation prediction accuracy and that of other correlations to predict flue gas-oil
MMP …Contd.
Sebastian Alston Alston Eakin &
Flue gas Actual GA-based GA- Sebastian Dong Dong Kovarik Kovarik Eakin &
et al. et al. et al. Mitch
Reference composition MMP, MMP, based et al. MMP, error, MMP (mole) Mitch
MMP MMP, error, MMP,
(Mole, %) MPa MPa error, % error, % MPa % (mole), MPa error, % error, %
MPa MPa % MPa
60% of CO2,
Ref. 17 20% of C1, 14.07 12.10 14.0 12.29 12.7 11.13 20.9 11.47 18.5 12.9 8 21.8 55.2
20% of H2S
67.5% of CO2,
Ref. 17 10% of C1, 12.41 12.66 2.0 13.54 9.0 11.99 3.4 12.93 4.2 9.8 21.4 21.1 70.3
22.5% of H2S
45% of CO2,
Ref. 17 10% of C1, 10.38 11.51 10.9 12.26 18.1 10.27 1.0 11.39 9.8 1.9 81.5 21.7 108.6
45% of H2S
60% of CO2,
Ref. 17 20% of C1, 17.24 17.14 0.6 17.41 0.9 15.77 8.6 16.25 5.8 16.4 5 23.1 33.9
20% of H2S
90% of CO2,
Ref. 17 10.07 10.33 2.6 11.01 9.4 10.45 3.7 11.05 9.7 11.1 10.3 17.6 75.2
10% of C2
90% of CO2,
Ref. 17 9.31 8.77 5.9 9.61 3.2 8.89 4.6 12.02 29.1 7.8 15.9 17.8 91.4
10% of C3
90% of CO2,
Ref. 17 7.90 8.05 1.9 8.59 8.8 7.36 6.8 12.62 59.8 5.9 25.3 16.5 108.9
10% of C4
80% of CO2,
Ref. 17 9.66 9.92 2.7 10.99 13.8 9.88 2.3 11.06 14.6 11.1 14.5 18.2 88.2
20% of C2
80% of CO2,
Ref. 17 7.93 7.88 0.6 8.42 6.2 7.67 3.3 13.07 64.7 4.5 43.1 16.9 113.3
20% of C3
90% of CO2,
Ref. 17 13.04 12.59 3.4 13.42 3.0 12.73 2.3 13.47 3.3 13.5 3.7 18.3 40.8
10% of C2
90% of CO2,
Ref. 17 11.04 10.68 3.2 11.71 6.2 10.83 1.9 14.65 32.7 10.2 7.1 17.5 58.2
10% of C3
90% of CO2,
Ref. 17 8.97 9.81 9.4 10.47 16.8 8.97 0.0 15.38 71.6 8.3 7.3 17.8 99.1
10% of C4
80% of CO2,
Ref. 17 12.88 12.08 6.1 13.39 4.0 12.04 6.5 13.48 4.7 13.5 4.6 19.6 51.9
20% of C2
80% of CO2,
Ref. 17 10.50 9.61 8.5 10.27 2.2 9.34 11.0 15.92 51.7 6.9 34 18.3 73.9
20% of C3
80% of CO2,
Ref. 17 14.83 13.49 9.0 13.38 9.8 13.93 6.1 12.71 14.3 19.9 34.3 18.4 24.3
20% of C1
68% of CO2,
Ref. 17 22% of H2S, 10.28 8.96 12.8 9.58 6.8 8.50 17.3 9.15 10.9 6.5 36.9 18.4 79.1
10% of C1
40% of CO2,
Ref. 17 40% of H2S, 12.06 10.72 11.1 11.27 6.5 9.14 24.2 10.32 14.4 6 50.5 20.3 68.5
20% of C1
75% of CO2,
Ref. 17 10.35 10.59 2.4 10.52 1.6 10.13 2.1 10.15 1.9 3.1 69.7 19 84.1
25% of H2S
50% of CO2,
Ref. 17 8.97 10.31 14.9 9.46 5.5 8.87 1.1 8.74 2.5 -5.6 162.2 20.5 129
50% of H2S
90% of CO2,
Ref. 17 15.17 14.20 6.4 14.95 1.5 14.33 5.6 14.62 3.7 17.6 16 21.1 39
10% of C1
80% of CO2,
Ref. 17 18.74 19.11 2.0 18.95 1.1 19.73 5.3 18.01 3.9 23.4 24.6 24.1 28.5
20% of C1
55% of CO2,
Ref. 17 25% of H2S, 16.45 16.74 1.8 17.04 3.6 14.94 9.2 15.83 3.8 14.6 11 21.2 28.6
20% of C1
90% of CO2,
Ref. 18 16.21 15.87 2.1 16.70 3.1 16.01 1.2 16.34 0.8 19 17.1 18.2 12.3
10% of C1
22.18% of
CO2, 23.49%
of C1, 23.5%
Ref. 19 16.55 16.77 1.3 33.13 100.2 18.45 11.5 63.56 284.0 32.8 98.2 21.1 27.2
of C2, 27.45%
of C3, 3.38%
C4
79.2% of CO2,
Ref. 20 8.8% of N2, 23.17 20.19 12.9 27.22 17.5 18.99 18.0 58.21 151.2 25.8 11.5 - -
12% of n-C4
94.1% of CO2,
Ref. 21 3.1% of N2, 14.50 16.57 14.3 15.42 6.3 15.86 9.3 17.51 20.7 17.3 19.5 14.2 1.8
2.8% of C1
SPE 93478 11

Table 3 Comparison among the GA-based correlation prediction accuracy and that of other correlations to predict flue gas-oil
MMP …Contd.
Sebastian Alston Alston Eakin &
Flue gas Actual GA-based GA- Sebastian Dong Dong Kovarik Kovarik Eakin &
et al. et al. et al. Mitch
Reference composition MMP, MMP, based et al. MMP, error, MMP (mole) Mitch
MMP MMP, error, MMP,
(Mole, %) MPa MPa error, % error, % MPa % (mole), MPa error, % error, %
MPa MPa % MPa
90.1% of CO2,
Ref. 21 16.01 15.61 2.5 16.28 1.7 15.73 1.7 15.93 0.5 18.6 16.4 18.2 13.6
9.9% of C1
89.8% of CO2,
Ref. 21 5.1% of N2, 20.51 21.34 4.1 17.58 14.3 19.75 3.7 21.37 4.2 20.5 0.2 21.2 3.3
5.1% of C1
90% of CO2,
Ref. 3 11.60 11.28 2.8 11.49 0.9 11.26 2.9 11.34 2.2 8.6 25.5 17.9 54.7
10% of H2S
80% of CO2,
Ref. 3 11.40 10.97 3.8 11.00 3.5 10.63 6.7 10.70 6.1 5.2 54.7 18 57.5
20% of H2S
70% of CO2,
Ref. 3 10.40 10.74 3.3 10.54 1.4 10.05 3.4 10.09 3.0 1.7 83.9 19.6 88.3
30% of H2S
90% of CO2,
Ref. 3) 10.50 11.58 10.2 9.25 12.0 7.78 25.9 10.67 1.5 5.7 45.6 16.2 54.6
10% of SO2
85% of CO2,
Ref. 3 33.01 36.54 10.7 21.04 36.3 47.55 44.0 36.09 9.3 25.7 22.3 26.8 18.7
15% of N2
65% of CO2,
Ref. 3 15% of N2, 34.01 33.81 0.6 12.37 63.6 7.57 77.7 30.11 11.5 12.8 62.3 18.2 46.4
20% of SO2
60% of CO2,
Ref. 3 10% of N2, 23.00 22.95 0.2 8.05 65.0 6.35 72.4 19.14 16.8 1.9 91.7 14.6 36.5
30% of SO2
90% of CO2,
Ref. 3 5% of O2, 5% 23.00 22.13 3.8 17.02 26.0 23.72 3.1 23.49 2.1 20.4 11.2 18.6 19.3
of N2
80% of CO2,
5% of O2, 5%
Ref. 3 22.50 21.12 6.1 13.00 42.2 9.50 57.8 21.25 5.6 14 37.7 15.76 30
of N2, 10% of
SO2
94.4% of CO2,
Ref. 22 16.01 15.67 2.1 16.48 2.9 15.94 0.4 16.51 3.2 16.6 3.7 18.6 16.4
5.6% of C2
70.8% of CO2,
Ref. 22 14.50 14.24 1.8 16.39 13.0 14.09 2.9 16.56 14.2 16.5 13.6 20.5 41.7
29.2% of C2
60% of CO2,
Ref. 22 13.60 13.54 0.5 16.36 20.3 13.32 2.1 16.58 21.9 16.4 20.7 21.7 59.8
40% of C2
46.3% of CO2,
Ref. 22 11.90 12.62 6.0 16.31 37.0 12.48 4.8 16.61 39.6 16.3 37.3 23.5 97.7
53.7% of C2
97% of CO2,
Ref. 22 14.80 14.91 0.7 15.82 6.9 15.30 3.4 16.93 14.3 15.6 5.5 17 15
3% of C3
94.9% of CO2,
Ref. 22 14.30 14.25 0.3 15.37 7.5 14.61 2.1 17.24 20.5 14.9 4.3 16.4 14.6
5.1% of C3
88.4% of CO2,
Ref. 22 13.00 12.82 1.4 14.06 8.1 12.92 0.6 18.21 40.0 12.8 1.8 15.7 20.7
11.6% of C3
84.1% of CO2,
Ref. 22 12.40 12.23 1.4 13.28 7.0 12.10 2.5 18.88 52.2 11.3 8.6 19.1 54
15.9% of C3
80.4% of CO2,
Ref. 22 11.40 11.82 3.7 12.66 11.0 11.52 1.0 19.47 70.7 10.1 11.4 19.1 67.2
19.6% of C3
75.7% of CO2,
Ref. 22 11.00 11.31 2.8 11.93 8.4 10.91 0.8 20.24 83.9 8.5 22.4 18.7 70
24.3% of C3
68.3% of CO2,
Ref. 22 10.30 10.05 2.5 10.95 6.3 10.19 1.1 21.49 108.6 6.1 41 18.6 80.8
31.7% of C3
63.4% of CO2,
Ref. 22 9.10 8.56 6.0 10.39 14.2 9.81 7.8 22.36 145.6 4.5 51.1 18.5 103.4
36.6% of C3
80% of CO2,
Ref. 23 13.79 13.49 2.2 13.38 3.0 13.80 0.1 12.71 7.8 19.9 44.3 - -
20% of C1
80% of CO2,
Ref. 23 28.97 29.27 1.0 17.27 40.4 119.22 311.6 33.87 16.9 26.4 8.8 - -
20% of N2

Average error for all the data points, % o 4.7 13.1 14.1 29.6 34.6 60.8
STD for all the data points, % 6.3 22.0 43.2 55.0 54.0 74.0
Average error for all the data points except
4.8 12.7 9.4 29.8 35.0 60.8
the last point, %
STD for all the data points except the last
6.34 21.45 18.2 55.0 54.0 74.0
point, %
12 SPE 93478

Initial population of randomly


real-coded chromosomes (with number
of genes based on the model parameters)
with population size = 100

Evaluate each chromosome (Evaluation Function)


PFit (i,j) = 5000 / (5000 +(⏐MMPcal.(i,j) –MMPexp.(i,j)⏐))
nn
Fit (i ) = (∑ PFit (i, j )) / nn
j =1

Select two parent chromosomes


(Roulette Wheel Parent Selection)

Produce new offspring chromosomes (two children)


(Crossover (One-Point Crossover, P(c) =100% of) &
Mutation P(m) =100% of)
For mutation, New Value = 0.93×Old Value + 0.9× Random Value

Evaluate each offspring (Evaluation Function)


PFit (i,j) = 5000 / (5000 +(⏐MMPcal.(i,j) –MMPexp.(i,j)⏐))
nn
Fit (i) = (∑ Fit (i, j )) / nn
j =1

Return the two chromosomes that have


the maximum fitness of the two parents and
the two children chromosomes to the population

Stopping Criterion The Best


No (When the difference between Yes Chromosome
the best chromosome fitness and in the Population
the average population fitness presents the Solution
≤ 10-6)

Fig. 1. Flowchart of the genetic algorithm software design.


SPE 93478 13

50
40
45
GA-based
35 40 prediction,
MPa
35
GA-based Predicted MMP, MPa

Predicted MMP, MPa


30
30
Alston et al.
25 prediction,
25
MPa
20 20

15 Sebastian et
15 al.
10 prediction,
10 MPa
5

5 0
0 10 20 30 40 50
0 Exp. MMP, MPa
0 5 10 15 20 25 30 35 40
Exp. MMP, MPa Fig. 4. Comparison among GA-based, Alston et al. (2), and
Sebastian et al. (9) correlations within 95% accuracy for all the
data except the point that is identified within the circle in Fig.
Fig. 2. GA-based flue gas-oil MMP correlation prediction 3 (composed of: 80 mole% of CO2 and 20 mole% of N2).
results within 90 % accuracy.
Ratio between reduced MMP (flue gas/CO2),

3.5
120
3
GA-based
100 2.5
prediction,
fraction

MPa 2
Predicted MMP, MPa

80
1.5
Alston et al.
60
prediction, 1
MPa
0.5
40
Sebastian et 0
al. 250 270 290 310 330 350
20 prediction,
MPa Weight average pseudocritical temperature without using
multiplying factor (MFi), °K
0
0 20 40 60 80 100 120
Fig. 5. Relationship between reduced MMP ratio (reduced flue
Exp. MMP, MPa
gas-oil MMP to reduced CO2-oil MMP) and the weight
average pseudocritical temperature before using the
Fig. 3. Comparison among GA-based, Alston et al. (2)
, and multiplying factor (MFi).
Sebastian et al. (9) correlations within 95% accuracy.
14 SPE 93478

3.5

2.5

1.5

0.5

0
250 270 290 310 330 350
Weight average pseudocritical temperature with
using multiplying factor (M Fi), °K

Fig. 6. Relationship between reduced MMP ratio (reduced flue


gas-oil MMP to reduced CO2-oil MMP) and the weight
average pseudocritical temperature after using the multiplying
factor (MFi).

Potrebbero piacerti anche