SPE-SAS-698

AN ACCURATE PREDICTION OF CO2 MINIMUM MISCIBILITY PRESSURE

(MMP) USING ALTERNATING CONDITIONAL EXPECTATION ALGORITHM
(ACE)
Osamah Alomair, SPE, Adel Malallah, SPE, Adel Elsharkawy, SPE, and Maqsood Iqbal, SPE, Kuwait University
Copyright 2011, Society of Petroleum Engineers
This paper was prepared for presentation at the 2011 SPE Saudi Arabia Section Technical Symposium and Exhibition held in AlKhobar, Saudi Arabia, 1518 May 2011.
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, as presented, does not necessarily reflect any position of the Society of Petroleum
Engineers, its officers, or members. Papers presented at the SPE meetings are subject to publication review by Editorial Committee of Society of Petroleum Engineers. 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 where and whom the paper was presented. Write Liberian, SPE, P.O. Box 833836,
Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract
Miscible gas injection nowadays becomes an imperative enhanced oil recovery (EOR) approach for increasing oil
recovery. Due to the massive cost associated with this approach a high degree of accuracy is required for
predicting the outcome of the process. Such accuracy includes, the preliminary screening parameters for gas
miscible displacement; the minimum miscibility pressure (MMP) and the availability of the gas.
All conventional and stat-of-the-art MMP measurement methods are either time consuming or decidedly cost
demanding processes. Therefore, in order to address the immediate industry demands a nonparametric approach
(ACE) is employed in this study to estimate an important parameter MMP. ACE algorithm correlates optimal
transforms of a set of predictors with an optimal response transform. Finally, the proposed model has produced a
maximum linear effect between these transformed variables. More than 100 MMP data points are considered both
from the relevant published literature and the experimental work. Five MMP measurements for Kuwaiti Oil are
included as a part of the testing data. The proposed model is validated using detailed statistical analysis and it
reveals that the results are more reliable than the existing correlations for pure CO 2 injection to enhance oil
recovery. In addition to its accuracy, the ACE approach is more powerful, quick and can handle a huge data.
Introduction
The minimum miscibility pressure (MMP) is defined as the lowest pressure at which we have a distinct point of
maximum curvature when recovery of oil at 1.2 PV gas injected is plotted verse pressure (Johnson and Pollin;
1981). This pressure can be located graphically by the intersection of two lines that define both an immiscible and

[Paper Number]

miscible performance regimes on a plot of recovery versus pressure, or recovery versus composition. MMP is one
of the most important factors in the selection of candidate reservoirs for gas injection at which miscible recovery
takes place.

This minimum dynamic miscibility pressure (MDMP) or simply minimum miscibility pressure, is

dependent upon the composition of the injected gas, the reservoir temperature, and the characteristics of the in
place fluid. This pressure of miscibility is independent of the nature of the porous media or of the velocity of
displacement.
Experimental Techniques for Measuring MMP
There are several experimental methods that can be used for measuring MMP for an oil-solvent system.
Traditionally, slim tube studies were conducted for this purpose (Yellig et al., 1980, Huang et al., 1994). However,
the faster and more precise rising bubble apparatus (RBA) method is becoming more common for measuring MMP
(Christiansen and Haines, 1987). This technique was first introduced by researchers at Marathon oil in 1984
(Huang et al., 1984). A comparison of the two measurement techniques has been discussed by many investigators
(Novosad et al., 1989, Elsharkway et. al., 1992, Huang and Dyer, 1993, and Srivastava et al., 1994). A rapid
experimental method for measuring MMP at low temperatures (below about 50 oC) based on the measurement of
density of the injection-gas-rich upper phase in contact with stock tank oil as a function of pressure has been
reported by Harmon and Grigg; 1986. On the other hand, flow experiments offer the most reliable method to
determine the pressure required for miscibility with N 2, CO2 (both pure and impure), and hydrocarbon gas (Glaso,
1985). Because of the complexity of the of the interactions between crude oil and gas in a flowing system the
onset of the miscibility can only be determined by comparing the relative displacement efficiencies of controlled
flow experiments (Glaso,1987). Deffrenne et. al. (1961) presented a miscible displacement system and made the
use of slim tube apparatus for enriched or vaporizing gas drives. He measured and compared the MMP of oil using
different solvents like methane, nitrogen and various other gases (Firoozabadi and Aziz, 1986).
P-X (Pressure-Composition) diagram method needs a number of careful lab experiments using PVT cell.
Increasing amount of solvents are mixed with the oil, and Bubble Point Pressure (BPP), Dew Point Pressure
(DPP) and gas liberation curves are obtained for each mixed ratio. When the BPP and DPP curves no longer
rises with the increasing amounts of solvent added, it is assumed that FCM has been achieved. As obvious,
these tests needs accurate monitoring, and are expensive.
Correlations for Predicting MMP
Correlations for predicting MMP have been proposed by a number of investigators, and are important tools for
rapid and accurate MMP calculation. Enich et al. (1988) pointed out that, ideally, any correlation should account for
each parameter known to affect the MMP, should be based on thermodynamic or physical principles that affect the
miscibility of fluids, and should be directly related to the multiple contact miscibility process. For an initial and quick
estimate, operators use correlation currently available in the literature. For screening purposes, they gave a fair
first guess depending on the data used. Moreover, they are inexpensive and can be detained by simple hand
calculation. However, the success of the correlations is usually limited to the composition range in which these
correlations were developed. The CO2 MMP correlations fall into two categories: the pure and impure CO 2; while
the other category treats MMPs of other gases.
In 1960, Benham et al. presented empirical curves that can estimate miscibility conditions for reservoir oils that are

[Paper Number]

displaced by rich gas, followed by some proposed equations that have been derived for predicting MMP. These
equations are a result of curve fitting Benham et al.s data (Glaso, 1985).
Glaso presented a generalized correlation for predicting the MMP required for multi-contact miscible displacement
of reservoir fluids by hydrocarbon, CO 2 or N2 gas. Glaso also showed that for hydrocarbon systems, paraffinicity
has an effect on MMP (Glaso, 1985). Several methods have been proposed to predict the CO 2 dynamic miscibility
pressure of reservoir fluids from easily obtainable field data. The most widely discussed of these methods are; the
National Petroleum Council correlation, the Gulf Universities Research Consortium correlation, the method
developed by Holm and Josendal (1974) and the method developed by Yellig and Metcalfe (1980). However, none
of these correlations gives adequate emphasis to oil properties and composition and all fail to accurately predict
the miscibility pressure for variety of crude oil types. This led to investigate other techniques for predicting MMP
such as regression analysis and artificial intelligent methods.
MMP Regression Analysis
The regression analysis addresses the effect of one or more independent variables (predictors or covariates) on a
dependent variable (response). The initial stages of data analysis often involve exploratory analysis. Unfortunately
traditional multiple regression techniques are limited, since they usually require a priori assumptions about the
functional forms that relate the response and predictor variables. When the relationship between the response and
the predictor variable is unknown or inexact, linear parametric regression can yield erroneous and even misleading
results.

This is the primary motivation for the use non-parametric regression techniques, which make few

assumptions about the regression surface (Friedman and Stuetzle, 1981).

The objective of fully exploring and explaining the effect of covariates on a response variable in regression analysis
is facilitated by properly transforming the independent variables. There are number of parametric transformations
for continuous variables in regression analysis. Estimating the optimal transformation is the primary motivation for
the use of non-parametric regression techniques, which make few assumptions about the regression surface
(Breiman and Friedman, 1985). Non-parametric regression techniques are based on successive refinements by
attempting to define the regression surfaces in an iterative fashion while remaining data driven as opposed to
model driven. These non-parametric regression methods can be broadly classified into those which do not
transform the response variable (such as Generalised additive models) and those which do (such as Alternating
Conditional Expectations, ACE).

Moreover, the ACE algorithm can handle variables other than continuous

predictors such as categorical (ordered or unordered), integer and indicator variables (Wang and Murphy, 2004).
The present approach to estimate MMP is guided by the view that statistical methods for dealing with data that
exhibit strong linear associations are well developed; consequently, many non-standard problems are best
addressed by transforming the data to achieve increased linear association. The analysis given here also serves
to illustrate the exploratory use of the ACE algorithm to suggest expressions, and the use of R 2 from the ACE
transformed variables as a benchmark.

The ACE-Transformed variables exhibit substantially greater linear

association than the untransformed variables. One of the principal benefits of the ACE algorithm is that it provides
a theoretical standard against which more analytically appealing transformation can be judged (Veaux, et al.,
1989). The power of the ACE approach lies in its ability to recover the functional forms of variables and to uncover
complicated relationships (Wang and Murphy, 2004).

[Paper Number]

The advantage of the non-parametric regression is easy to use and can quickly provides results that reveal the
dominant independent variables and relative characteristics of the relationships (Wu et al., 2000). It can be applied
both bivariate and multivariate cases and it yields maximum correlations in transformed space (Malallah et al.,
2005). A modification of ACE algorithm with graphical (GRACE) interface was later proposed by Xue et al. (1997).
Alternating Conditional Expectation (ACE)
This study uses an algorithm (ACE) of Brieman and Friedman (1985) for estimating the transformations of a
response and a set of predictor variables in multiple regression problems in enhanced oil recovery. The name
alternating conditional expectations refers to the algorithm used to compute optimum transforms (viz. those that
minimize the summation of squares of the error). The mathematical expectation is the mean of the distribution of
a population and is denoted by or E(Z), where Z is the variable that describes the experiment. is mostly used
mostly in uni-variate statistics. The word conditional in ACE is meant to indicate that the means of Z/Q variables
(i.e., the conditional variables) are determined. The conditions in ACE are the values of the dependent variable or
those of an independent variable. Conditional expectations can be expressed as:

i X i

i X i

Xi

and

(1)

The proposed nonparametric approach can be applied easily for estimating the optimal transformation of different
gas injection data to obtain maximum correlation between observed variables. An ACE regression model can be
expressed as:
p

Y i X i
i 1

(2)
where is a function of the response (dependent) variable Y, i are functions of the predictors (independent)
variables Xi , i =1,2,3, p. Thus, the ACE model replaces the problem of estimating a linear function of a pdimensional variable X = (X1, X2, X3, X4,Xp) by estimating p separate one-dimensional functions, i, and using
an iterative method. These transformations are achieved by minimizing the unexplained variance of a linear
relationship between the transformed response variable and the sum of transformed predictor variables.
For a given dataset consisting of a response variable Y and predictor variables X 1, X2, X3, X4, Xp, the ACE
algorithm starts out by defining arbitrary measurable mean transformations (Y), 1(X1), 2(X2), 3(X3), p(Xp).
The error variance (2) that is not explained by a regression of the transformed dependent variable on the sum of
transformed independent variables is (under the constraint, E(2(Y) =1)

, 1 , 2 , 3 ,... p E Y i X i
i 1


2

(3)
The minimization of (2) with respect to 1(X1), 2(X2), 3(X3), p(Xp) and (Y) is carried out through a series of
singlefunction minimizations, resulting in the following equations:

[Paper Number]

i X i E Y j X j
j i

Xi

(4)

i 1

X
i 1

(5)

The two basic mathematical operations involved here is conditional expectation and iterative minimization, hence
the name alternating conditional expectations. The final i (Xi), i=1, 2, 3 p, and (Y) after the minimization are
estimates of the optimal transformations i* (Xi), i=1, 2, 3 p, and *(Y). In the transformed space, the response
and predictor variables are related as follows:
p

Y i X i e
i 1

(6)
Where e* (misfit) is the error not captured by the use of the ACE transformations and is assumed to have a normal
distribution with zero mean. These optimal ACE transformations are derived solely from the given data and do not
require a priori assumptions of any functional form for the response or predictor variables and thus provide a
powerful tool for exploratory data analysis. The dependent variable for any data point is calculated as:

X
i 1

(7)

The calculation involves n forward transformations of X1, X2, X3, X4, Xp to 1(X1), 2(X2), 3(X3), p(Xp) and
a backward transformation:

X
i 1

(8)
Investigating the Factors Affecting MMP
Generally, MMP increases steadily with increasing temperature, and oils with higher density and molecular weight
have a higher MMP. It has been reported that even small impurities, can significantly affect the miscibility pressure

[Paper Number]

(Glaso, 1987). Alston et. al. (1985) documented the fact that the achievement of miscibility is strongly related to
reservoir temperature and oil composition, particularly C 5+ molecular weight. Holm and Josendal (1974) found that
MMP was only affected by the type of hydrocarbons present in the range C 5 to C30 fractions of the crude oil. Yellig
and Metcalfe (1980) found little significance of C 7+ properties of the oil on the CO 2 MMP. Alston et. al. (1985) have
shown that the reservoir oil volatile and intermediate fractions can significantly affect the MMP when their ratios
depart from unity (Glaso, 1985). This also explained the effects of both solution gas (live oil systems) and impurity
of CO2 sources (Alston, et. al., 1985). James et al (1981) presented an empirical correlation which predicted the
MMP for a wide variety of live oils and stock oils with both pure and diluted CO 2. This correlation, requiring only the
oil gravity, molecular weight, reservoir temperature and injection gas composition, showed substantially better
agreement with experiment. Many correlations relating the MMP to the physical properties of the oil and the
displacing gas have been proposed to facilitate screening procedures and to gain insight into the miscible
displacement process (Alston et. al., 1985; Orr and Silva, 1987; Rathmell et al.; 1971)
To study the effect of these parameters on tested data, several sensitivity analyses were conducted. Figure 1
show the relationship between the independent variables and MMP for the data used in this study. The correlation
coefficient of each independent variable is shown. It is clear that the temperature is the most dominant factor.
Methane has the same proportional effect where as C 6 and C7+ has adverse effect on MMP.

Figure 1: Effect of different independent variables on MMP

MMP
(psig)

T (F)

MC7+

MC5+

C7+

C6

C5

C4

C3

C2

C1

N2

CO2

H2S

Table1: Data range used for the input variables

11.23

9.49

16.00

73.61

262

286

265

3705

0.28

0.25

0.79

5.70

132

151

71

1101

76.43
2.56

18.40

16.44
0.001

0.89

24.00
0.001

23.16

17.56
0.001

0.01

Max

Min

[Paper Number]

Development of ACE MMP Model

As discussed earlier, MMP is a function of temperature, crude oil composition and composition of the solvent. To
understand the in-situ crude oil composition impact on MMP, the functional form of MMP Model is:

MMP f HCCOMP , NHCCOMP , T , MC5 , MC7

(9)

HCCOMP =

Mole Fraction of hydrocarbons (C1, C2, C3, C4, C5, C6, and C7+)

NHCCOMP

Mole Fraction of non-hydrocarbons (H2S, CO2, N2)

Temperature

MC5+

Molecular weight of Pentane Plus

MC7+

Molecular weight of Heptane Plus

The data set used in this study consisted of 113 MMP measurements (pure CO 2) taken from worldwide gas
injection projects from the published literature. The ranges of variables and MMPs used for this study are shown in
Table 1.

The collected data cover a wide range of API gravities (13 58 oAPI), reservoir pressures and

temperatures. The data were divided into two sets. The training set consisted of 96 points and a testing set of 17,
which were randomly selected from the total set of data. Out of 17 testing points, 5 were taken from MMP
measurements of Kuwait oil fields. All the Kuwaiti crude oil samples were thoroughly studied in Kuwait University
PVT Lab. Both, the detailed compositional analyses and minimum miscibility pressure measurements were
experimentally determined. The ACE algorithm provides a nonparametric optimization of the dependent (MMP) and
independent variables (HCCOMP, NHCCOMP, T, MC5+, MC7+); it does not provide a computational model for these
variables. However, the optimal data transforms can be fitted by simple polynomials that can be used to predict
the dependent variable. The default polynomial is of degree two but for any improvement the degree can be
increased. To find the maximum/optimal correlation, the ACE algorithm has the capability of using the independent
and dependent variables in their actual space or in the logarithmic space. After testing all possible combinations of
the independent and dependent variables in either logarithmic or actual space, the following suit of polynomials is
considered;

The sum of all these optimal transformed independent variables is;

Temp_Tr= -7.9731E-06x2 + 2.1085E-02x - 2.8306E+00
MMP= 2.5923E+01 SumTr2 + 6.5136E+02 SumTr + 2.0097E+03

MMP

ln_ H2S_Tr= 3.0640E-04x4 + 2.7747E-03x3 + 6.3899E-03x2 - 4.0849E-03x - 7.9337E-02

H2S

Temp

ln_ CO2_Tr= 9.4029E-06x6 - 3.2025E-05x5 - 1.0969E-03x4 - 8.1253E-04x3 + 2.0166E-02x2 + 1.3278E-02x - 3.2906E-02

CO2

Mw7+_Tr= 1.5634E-05x2 - 1.5754E-03x - 4.4180E-01

ln_ C7+Tr= -2.4459E-02x2 + 3.8895E-02x + 2.1657E-01

C7+

Mw7+

ln_ C6_Tr= -5.4686E-02x5 + 2.6628E-01x4 - 3.6496E-01x3 + 7.6998E-02x2 - 8.3604E-04x + 1.1194E-01

C6

Mw5+_Tr= 1.9056E-08x4 - 1.5287E-05x3 + 4.5136E-03x2 - 5.7789E-01x + 2.6867E+01

ln_ C5_Tr= -1.9674E-04x4 - 2.7402E-02x3 + 1.8143E-03x2 + 1.4054E-01x - 7.2174E-02

C5

Mw5+

ln_ C4_Tr= -5.1912E-02x6 + 2.7982E-01x5 - 3.7272E-01x4 - 2.4664E-01x3 + 7.1682E-01x2 - 2.7774E-01x - 1.3485E-02

C4

ln_ N2_Tr= -3.3252E-04x4 + 6.1601E-04x3 + 8.4462E-03x2 - 2.5561E-02x - 8.4381E-03

ln_ C3_Tr= 1.6381E-02x4 - 1.3052E-01x3 + 2.6607E-01x2 - 1.2019E-01x - 4.3356E-02

C3

N2

ln_ C2_Tr= 5.5795E-04x6 - 6.7876E-03x5 + 1.0678E-02x4 + 1.2248E-01x3 - 4.5724E-01x2 + 5.2793E-01x - 2.0734E-01

ln_ C1_Tr= 1.3912E-02x2 + 3.2202E-02x - 1.8611E-01

C1

C2

Optimal Transform Polynomials

Variable

8
[Paper Number]

[Paper Number]

13

SUM pi
i 1

(10)
The predicted / calculated MMP will be:

MMP pO1 SUM

(11)

The best combination that yields the highest correlation coefficient (R 2), the lowest average absolute relative error
(AARE), the lowest average relative error (ARE), and the lowest standard deviation (SD) is:

MMP a2 SUM 2 a1SUM 1 a0

(12)
or

MMP 25.923 SUM 2 651.360 SUM 1 2009.7

(13)

Data Analysis and Results Discussion

In the proposed nonparametric ACE model for MMP estimation, there are 13 predictors. All these are tested both in
real and logarithmic space and the maximum optimal correlation coefficients (more than 90 %) are obtained. All
mole fractions (hydrocarbon and non-hydrocarbon) are found more optimal to their respective transforms in the log
space whereas temperature and both plus fraction molecular weights are having maximal correlation effect in the
real space.
Hydrocarbon Optimal Transforms
A suite of Figure 2 (a, b, c, d, e, f and g) is ACE defined optimal transforms in the logarithmic space. Different kinds
of trends are observed.
Light hydrocarbon gases (C1 and C2) show optimal transforms. ACE defined ethane transform is in logarithmic
space with a polynomial of degree six. Methane is showing a pattern.
Intermediate hydrocarbons (C3, C4, and C5) transforms are relatively better defined in the logarithmic space. All the
polynomials are of higher degrees. There is a pattern is observed in butane and pentane.
Heavy end hydrocarbons (C6 and C7+) transforms are determined with polynomials of 5 & 2 respectively.

10

Figure 2 (a)

Figure 2 (b)

Figure 2 (c)

[Paper Number]

[Paper Number]

Figure 2 (d)

Figure 2 (e)

11

12

[Paper Number]

Figure 2 (f)

Figure 2 (g)
Non-hydrocarbon Optimal Transforms
Non-hydrocarbon gases (CO2, H2S and N2) optimal transforms, determined by ACE are very well defined in Figure
3 (a, b, and c) in the logarithmic space. There are some patterns observed of all the three gases.

[Paper Number]

Figure 3 (a)

Figure 3 (b)

13

14

[Paper Number]

Figure 3 (c)
Real Space Optimal Transforms
Both plus fraction molecular weights (Mw5+ and Mw7+) and temperature transforms are defined in the real space by
ACE. Heptane plus ACE transform is showing some trend. Temperature transform is defined in real space optimally
and a second degree polynomial determined by ACE is showing a good trend in Figure 4(c).

Figure 4(a)

[Paper Number]

15

Figure 4 (b)

Figure 4 (c)
Figure 5 shows the optimal transform of the dependent (response) variable. This is also tested in both (real and
logarithmic) spaces and finally it is defined in the real space. Coefficients of this fit polynomial will be incorporated
in the final SUM equation to estimate MMP.

16

[Paper Number]

Figure 5: Optimal Transformation of MMP by ACE.

Finally, all these predictors transforms both from real and logarithmic space are added up and correlate with
transform of MMP. In figure 6, an excellent correlation is obtained with R 2 value of 0.98014. This proves the
incredible power of ACE algorithm.

Figure 6: (MMP_Tr vs. Sum of Optimal Transformation of independent variables) by ACE

(Optimal Regression)
All transformed independent variables (predictors) and the response (MMP) are found numerically. The best linear
regression between them is shown in the last figure-6, and the equation of this match is;
MMP= 25.923 SUMTr2 + 651.360 SUMTr1 + 2009.7
Table 2 presents the statistical analysis between the results predicted by ACE model and that by published
correlations. It is inferred from the analysis that ACE is a powerful tool and shows high accuracy as compared to

[Paper Number]

17

other correlations.
Table 2: Statistical Analysis for MMP (Training Data)
ARE

SD

(%)

RMSE

4.68

-0.66

139.20

Alston (1985)

14.40

-1.94

Cronqst (1978)

12.97

Yell-Met (1980)

Model

AARE (%)

ACE

SSE (%)

R
SSE

SSR

1.86E+06

0.956

0.907

518.96

2.59E+07

0.394

1.921

-11.26

484.53

2.25E+07

0.471

2.069

15.58

14.89

391.89

1.47E+07

0.654

1.088

Orr-Jen (1984)

15.36

2.63

394.70

1.50E+07

0.649

1.936

Glaso (1985)

7.99

5.39

248.03

5.91E+06

0.861

0.967

Validation of ACE Model for MMP

To check the validity/credibility of ACE model and to check its predictive capability for MMP, all the derived
polynomials of variables (both predictors and Response) were examined using testing data of 17 points. Five of
these are the experimental measurements made for Kuwaiti oil fields.

Figure 7: Cross-plot of MMP Testing data.

There is a good match between the experimentally measured and ACE estimated values for MMP. It is shown in
figure 7. This proves the validity of our proposed ACE model. A detailed statistical analysis with different statistical
tools is explained in table 3 shown below.
Table 3: Statistical Analysis for MMP (Testing Data)

18

[Paper Number]

AARE

ARE

SD

(%)

(%)

RMSE

ACE

2.84

10.22

3324

Alston (1985)

-0.03

22.65

Cronqst (1978)

-13.75

Yell-Met (1980)

Model

SSE (%)

R
SSE

SSR

1.88E+10

0.818

0.783

12394

2.61E+11

0.628

2.583

15.56

11572

2.28E+11

0.858

2.326

15.93

15.93

9359

1.49E+11

0.842

1.054

Orr-Jen (1984)

3.84

15.06

9426

1.51E+11

0.781

1.897

Glaso (1985)

2.92

8.54

5923

5.96E+10

0.833

1.387

For comparison purposes, the same testing data was also applied to the available (pure CO 2) MMP correlations.
The overall performance of ACE for predicting MMP values is better and convincing.
Conclusions
A nonparametric model to predict MMP is developed based on 96 data points. The proposed ACE model is shown
to be more accurate than the existing conventional regression correlations. This model is able to predict MMP for
pure CO2 as a function of thirteen independent variables (all possible factors affecting MMP). The model has
certain advantages:

The approach solves the general problem of establishing the linearity assumption required in regression
analysis, so that the relationship between response and independent variables can be best described and
existence of non-linear relationship can be explored and uncovered.

An examination of these results can give the data analyst insight into the relationships between these
variables, and suggest if transformations are required.

The ACE plot is very useful for understanding complicated relationships and it is an indispensable tool for
effective use of the ACE results.

It provides a straightforward method for identifying functional relationships between dependent and
independent variables.

Although ACE provides a largely automated approach to estimating optimal transformations, it does not mean that
the ACE results should be trusted blindly and used dogmatically, additional information and experience of the data
analyst remain important. It should be emphasised that the success of the ACE algorithm, like other modern
statistical methods, relies on the quality of the data and underlying association between the response and
independent variables.
Acknowledgement
The authors would like to thank Kuwait University for supporting this project through the Research Grant GE 01/07.
Nomenclature
AARE = average absolute relative error, %
ARE = average relative error, %

[Paper Number]

19

C1 = methane mole fraction

C2 = ethane mole fraction
C3 = propane mole fraction
C4 = butane mole fraction
C5 = pentane mole fraction
C6 = hexane mole fraction
C7+ = heptane plus mole fraction
CO2 = carbon dioxide mole fraction
E = mathematical expectation
e* = misfit (error)
f = function
HCCOMP = a group, mole fractions of hydrocarbon composition
H2S = hydrogen sulphide mole fraction
Ln = natural log
MC5+ = Molecular wt of Pentane Plus
MC7+ = Molecular wt of Heptane Plus
MMP = minimum miscibility pressure (psig)
NHCCOMP = a group, mole fractions of non-hydrocarbon composition
N2 = nitrogen mole fraction
R = correlation coefficient, %
RE = relative error
R2 = ratio of data variability
RMSE = Root Mean Square Error
SD = standard deviation of the errors
SSE = sum of squares of the errors
SSR = regression sum of squares
SST = total sum of squares
T = temperature (F)
Tr = transform
References
Adel Malallah et al.: Accurate Estimation of the World Crude Oil Properties Using Graphical Alternating Conditional
Expectations,
Alcocer, C. F. and Menzie, D. E.: EOR by Nitrogen Injection: The Effect of Increased Temperature and the amount
of gas in solution, paper SPE 12709 presented at SPE/DOE 1984 EOR Symposium, Tulsa, April 15-18.
Alston, R. B., Kokolis, O. P. and James, C. F. (1985). CO 2 Minimum Miscibility Pressure: A Correlation for Impure
CO2. SPE 11959.

20

[Paper Number]

Ballard, J. R. and Smith, L. R.: Reservoir Engineering, design of a low pressure Rich Gas Miscible flood Project,
J. Pet. Tech. (May 1972) 599-605.
Brieman, L., and Friedman, J. H.: Estimating optimal transformations for multiple regressions and correlation, j.
Am. Stat. Assoc. 1985, 580-619.
C. H. Wu et al.: Non-Parametric regression and neural-network infill drilling recovery models for carbonate
reservoirs, Computer and Geosciences 26 (2000) 975-987.
Christian, L. D. et al. : Planning a Tertiary Oil-Recovery Project for Jay/LEC Field Unit, JPT (Aug. 1981) 1535-44.
Christiansen, R. L. and Haines, H. K. (1987). Rapid Measurement of Minimum Miscibility Pressure with the RisingBubble Apparatus. SPE 13114.
Cook, A. B., Walter, C. J., and Spencer, G. C.: Realistic K-values of C 7+ Hydrocarbons for Calculating Oil
Vaporization during Gas Cycling at High Pressure. J. Pet. Technol. 1969, 901-915.
Deffrenne, P. et al.: The determination of pressures of miscibility, paper SPE 116 presented at the 1961 SPE
Annual Meeting, Dallas, Oct. 8-16.
Dulao Wang and Michael Murphy: Estimating Optimal Transformations for multiple Regression Using the ACE
Algorithm, Journal of data Science 2 (2004), 329-346.
Elsharkay, A. M., Poettmann, F. H. and Christiansen, R. L.: Measuring Minimum Miscibility Pressure: Slim-Tube or
Rising Bubble Method, paper SPE/DOE 24114, presented at the 1992 eighth Symposium on EOR, Tulsa, April 2224.
Enich et al.: A thermodynamic correlation for the MMP in CO 2 flooding of petroleum reservoirs, SPERE, February,
81-92. 1988.
Firoozabadi, A. and Aziz, K. (1986). Analysis and Correlation of Nitrogen and Lean-Gas Miscibility Pressure. SPE
13669.
Gasem, K.A.M., Dickson, K.B., Shaver, R.D., and Robinson Jr., R.L.: Experimental phase densities and interfacial
tensions for a CO2/Synthetic-oil and a CO2/reservoir-oil system, SPE Reserv. Eng. Aug. 1993, 170174.
Glaso, O. (1985). Generalized Minimum Miscibility Pressure Correlation. SPE 12893.
Glaso, O. (1987). Miscible Displacement: Recovery Tests with Nitrogen. SPE 17378.

[Paper Number]

21

Harmon, R. A. and Grigg, R. B. (1986). Vapor Density Measurement for Estimating Minimum Miscibility Pressure.
SPE 15403.
Hier, L. Miscibility Variations in Compositionally Grading Petroleum Reservoirs, Ph.D. Dissertation, The
Norwegian University of Science and Technology, 1997.
Holm, L. W. and Josendal, V. A., Mechanisms of Oil Displacement by Carbon Dioxide, (1974), SPE 4736.
Huang, S. S. and Dyer. S. B.: Miscible Displacement in the Weyburn Reservoir: A Laboratory Study, JCPT (1993)
32, No. 7, 42-5.
Huang, S.S. et al.: A Laboratory Miscible Displacement Study for the Recovery of Saskatchewans Crude Oil, J.
Cdn. Pet. Tech.,April 1994, 33- 43.
Jaubert, J. N., Arras, L., Neau, E., and Avaullee, L. Properly Defining the Classical Vaporizing and Condensing
Mechanisms When a Gas Is Injected Into a Crude Oil. Ind. Eng. Chem. Res. 1998, 37, 4860.
Jensen, F. and Michelsen, M. L. Calculation of First Contact and Multiple Contact Minimum Miscibility Pressure. In
Situ 1990, 1-17.
Jessen, K., and Michelsen, M. L. Minimum Miscibility Pressure Calculation. IFP seminar: Production of reservoir
fluids in frontier conditions, Rueil-Malmaison, Dec 4-5, 1997.
Johnson, J.P. and Pollin, J.S. (1981). Measurement and Correlation of CO 2 Miscibility Pressures SPE/DOE 9790.
Koch, H. A. Jr. and Hutchinson, C. A. Jr.:Miscible Displacements of Reservoir Oil using Flue Gas, JPT (Jan.
1958) 7-19; Trans., AIME, 213.
Kuo, S. S. Prediction of Miscibility for the Enriched-Gas Drive Process. SPE 1985, 14152, 1-12.
Lee, J. I. and Reitzel, G. A.: High pressure, Dry Gas Miscible Flood-Brazeau River Nisku Oil Pools, JPT (Nov.
1982) 2503-2509.
Marcelo Blanco et al.: Modelling of an environmental parameter by use of the alternating conditional expectation
method, Chemometrics and Intelligent Laboratory Systems 46 (1999) 31-39.
McNeese, C. R.: High Pressure Gas Process and the use of Flue gas, paper presented at the 1963 ACS Spring
Meeting, Los Angeles, March 31-April 4.
Meltzer, B. D., Hurdle, J. M., and Cassingham, R. W.: An efficient Gas Displacement Project Rayleigh Field,
Mississippi, JPT (May 1965) 509-14.

22

[Paper Number]

Metcalfe, R. S., Fussel, D. D., and Shelton, J. L.:A Multi-cell Equilibrium Separation Model for the Study of Multiple
Contact Miscibility in Rich Gas Drives. SPEJ 1973, 147-155.
Neau, E., Avaullee, L., and Jaubert, J. N.: A New Algorithm for Enhanced Oil Recovery Calculations. Fluid Phase
Equilib. 1996, 117, 265-272.
Novosad, Z., Sibbald, L. R. and Costain, T. G.: Design of miscible solvents for a rich gas drive-comparison of slim
tube tests with rising bubble tests, paper 89-40-3, presented at 1989 CIM 40 th Annual Technical Meeting, Banff,
May 28-31.
Orr Jr., F.M., and Jensen, C.M., 1984.: Interpretation of pressure composition phase diagrams for CO 2/crude-oil
systems. Soc. Pet. Eng. J. 10, 485497.
Orr, F. M. and Silva, M. K. Effect of Oil Composition on Minimum Miscibility Pressure-Part 2: Correlation, Soc. Pet.
Eng. Res. Eng. 1987, 27 (6), 479-491.
Pedersen, K. S., Fjellerup J., and Thomassen, P.: Studies of Gas Injection into Oil Reservoirs by a Cell to Cell
Simulation Model. SPE 1986, 15599, 1-8.
Peterson, A. V.: Optimal Recovery Experiments with N2 and CO2, Pet. Eng. Intl. (Nov. 1978) 40-50.
Rathmell, J.J., Stalkup, F. I., and Hassinger, R. C.:A laboratory investigation of miscible displacement by CO 2,
paper SPE 3483 presented at the SPE 1971 Annual Fall Meeting, New Orleans, Oct. 3-6.
Richard d. De Veaux et al.: ACE Guide Transformation method for estimation of the coefficient of soil-water
receptivity, Technometrics, Feb 1989, Vol. 31, No. 1.
Rushing, M. D. et al.:Nitrogen may be used for Miscible Displacement in Oil

Reservoirs, JPT (Dec. 1978)

1715-16.
Srivastava, R.K. et al.: A comparison of minimum miscibility pressure determinations for Weyburn CO 2 solvent
design, paper no-94-50 presented at the 45th Annual Technical Meeting, Calgary, June 12-15, 1994.
Stalkup, F. I. Displacement Behavior of the Condensing/ Vaporizing Gas Drive Process. SPE 1987, 16715, 171182.
Wang, Y., and Orr, F. M.: Analytical Calculation of Minimum Miscibility Pressure. Fluid Phase Equilib. 1997, 139,
101-124.
Whitson, C. and Michelsen, M. L. The Negative Flash. Fluid Phase Equilib. 1989, 53, 51-7.

[Paper Number]

23

Whitson, C. H. Trends in Modeling Reservoir Fluids. Seventh International Conference on Fluid Properties and
Phase Equilibria for Chemical Process Design, Snowmass, CO, June 18- 23, 1995.
Xue, G., Datta-Gupta, A., Valko P. and Blasingame T.:Optimal Transformations for Multiple Regression: Application
to Permeability Estimation from Well Logs, SPE Formation Evaluation, (June 1997) 85-94.
Yellig, W. F., and Metcalfe, R. S. (1980) .: Determination and Prediction of CO 2 Minimum Miscibility Pressures.
SPE 7477
Zhou, DD., and Orr, F. M. Jr.: An analysis of Rising Bubble experiments to determine MMPs, paper SPE 30786
presented at the 1995 SPE Annual Technical Conference and Exhibition, Dallas, Oct. 22-25.
Zick, A. A. A Combined Condensing/Vaporizing Mechanism in the Displacement of Oil by Enriched Gases. SPE
1986, 15493, 1-11.