You are on page 1of 12

SPE 126649

Prediction of Porosity and Permeability of Oil and Gas Reservoirs using


Hybrid Computational Intelligence Models
Fatai Anifowose and Abdulazeez Abdulraheem, SPE, Center for Petroleum and Minerals, King Fahd University of
Petroleum and Minerals, Dhahran 31261, Saudi Arabia

Copyright 2010, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE North Africa Technical Conference and Exhibition held in Cairo, Egypt, 1417 February 2010.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been reviewed
by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or
members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is
restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
This work utilizes the capabilities of Data Mining and Computational Intelligence in the prediction of two important
petroleum reservoir characteristics, viz., porosity and permeability, based on the hybridization of three existing Artificial
Intelligence techniques: Fuzzy Logic, Support Vector Machines, and Functional Networks, using several real-life well-log data.
Two hybrid models have been built. In both, Functional Networks was used to select the best input variables for training
directly from data using its functional approximation capability. In the first model, the selected input variables were passed to
Type-2 Fuzzy Logic System to handle uncertainties, if any exists. These are then passed to Support Vector Machines for training
and making final predictions using the test data. In the second model, the input variables selected by Functional Networks were
passed to Support Vector Machines to transform them to a higher dimension, called a feature space, and then to Type-2 Fuzzy
Logic to handle uncertainties, if any exists, extract inference rules and make final predictions using the test data. The results
showed that the hybrid models perform better with higher correlation coefficients than the individual techniques when used alone
for the same sets of data. This is a further confirmation of the theory that hybrids perform better than their individual components
and possess the best qualities of each of the components. In terms of execution time, the hybrid models took less time for both
training and testing than the Type-2 Fuzzy Logic, but more time than Functional Networks and Support Vector Machines. This
could be the price for having a better and more robust model.
The process of combining multiple computational intelligence techniques to build a single hybrid model has become
increasingly popular. As reported in the literature, the performance indices of these hybrid models have proved to be better than
their individual components when used alone.
Hybrid models are extremely useful in the field of reservoir characterization in petroleum engineering which requires a
high quality information and accurate prediction for efficient exploration and management of oil and gas resources.

Keywords: hybrid computational intelligence, petroleum reservoir characteristics, fuzzy logic, support vector machines,
functional networks.

1. Introduction
Petroleum reservoir characterization is a process for quantitatively describing various reservoir properties in spatial
variability by using available field data. It plays a crucial role in modern reservoir management: making sound reservoir decisions
and improving the reliability of the reservoir predictions. The ultimate goal is a reservoir model with realistic tolerance for
imprecision and uncertainty. Porosity and permeability are two fundamental reservoir properties which relate to the amount of
fluid contained in a reservoir and its ability to flow. These properties make significant impacts on petroleum field operations and
reservoir management [Lim 2005].
A good number of studies have been carried out on the use of various Computational Intelligence (CI) schemes, such as
Multilayer Perceptrons, Radial Basis Function, Functional Networks, Support Vector Machines, Artificial Neural Networks and
many others, to predict the characteristics of oil and gas reservoirs such as depth, temperature, pressure, volume, drive mechanism,
structure and seal, well spacing, well-bore integrity, porosity, and permeability [Giovanni and Vincenzo 2005].
2 SPE 126649

The combination of two or more Computational Intelligence schemes as a single model is called Hybrid Computational
Intelligence (HCI) and is becoming increasingly popular. This increased popularity lies in the extensive success of hybrid systems
in many real-world complex problems. A key prerequisite for the merging of technologies is the existence of a "common
denominator" to build upon [Giovanni and Vincenzo 2005]. In this study, part of the "common denominator" for Fuzzy Logic
(FL), Support Vector Machines (SVM) and Functional Networks (FN) is the inference procedures they deploy and their excellent
predictive capabilities. We attempt to combine the individual capabilities of FL, SVM and FN in an HCI scheme, to predict two
characteristics of oil and gas reservoirs, namely porosity and permeability, with better performance indices.
Our motivations for this work include the quest for higher performance accuracy in the prediction of oil and gas reservoir
properties, the recently increasing popularity of hybrid intelligent systems, the reported success of these systems in many real-
world complex problems, the need to complement the weaknesses of one algorithm with the strengths of others and hence to
combine the cooperative and competitive characteristics of the individual techniques and the existing theoretical and experimental
justifications [Munakata 2008, Peddabachigari et al. 2007, Phillips-Wren et al. 2008] that hybrids produce more accurate results
than the individual techniques used separately.
The rest of this paper is organized as follows: Section 2 reviews the literature on how Artificial Intelligence has been used
in Petroleum Engineering. Section 3 presents a brief overview of FL, SVM, FN and Hybrid Systems. Section 4 describes the
datasets, tools, the experimental design and the criteria of quality measurement used to evaluate the validity of this work. Section 5
presents the results of the experimental study and a detailed discussion while conclusion is presented in Section 6.

2. Literature Survey
The application of the capabilities of Artificial Intelligence (AI) has been widely appreciated in petroleum engineering, as
well as in other fields. Some of the areas of petroleum technology in which AI has been successfully used include seismic pattern
recognition; porosity and permeability predictions; identification of sandstone lithofacies; drill bit diagnosis; and analysis and
improvement of oil and gas well production [AbdulKadir et al. 2005, Ali 1994, Goda and Behrenbruch 2007]. Porosity and
permeability measurements are frequently made on plugs extracted from the core of wells drilled for oil and gas exploration. The
data are valuable for linking permeability to porosity, and these both serve as standard indicators of reservoir quality in the oil and
gas industry.
Porosity is the percentage of voids and open space in a rock or sedimentary deposit. The greater the porosity of a rock, the
greater with be its ability to hold water and other earth materials, such as oil. Porosity is very important in evaluating the potential
volume of hydrocarbons contained in a reservoir. Permeability is the ease with which fluid is transmitted through a rock's pore
space. Although a rock may be very porous, it is not necessarily very permeable. Permeability is a measure of how interconnected
are the individual pore spaces in a rock or sediment. Permeability is a key parameter associated with the characterization of any
hydrocarbon reservoir. In fact, it is not possible to have accurate solutions to many petroleum engineering problems without
having accurate permeability value [Schlumberger 2007].
[Abdelkader et al. 2005] produced 2-D fracture intensity and fracture network maps in a large block of field using
Artificial Neural Network (ANN) and FL. The results showed that the proposed approach is a practical methodology to map the
fracture network. [Maqsood and Adwait 2000] used ANN to predict permeability from petrographic data while using FL to screen
and rank the predictor variables with respect to the target variable. The result demonstrated the generalizing capability of ANN. A
similar study was done by [Mohsen et al. 2007] when they proposed a new method for the auto-design of ANN based on Genetic
Algorithm.
The traditional FL, now referred to as Type-1 FL has featured in a number of research efforts, especially in reservoir
characterization. Type-2 FL has also featured in many recently published articles in various fields and especially in reservoir
properties modeling such as in [Abdelkader et al. 2005]. One of the earliest references to the application of FL in the petroleum
industry was by [Fang and Chen 1997] who presented a fuzzy modeling for predicting porosity and permeability from the
compositional and textural characteristics of sandstones. They found that the fuzzy modeling is assumption-free, tolerant of
outliers, and capable of making both linguistic and numeric predictions based on qualitative and quantitative data.
SVMs have been used extensively in many areas, including oil and gas [Jian and Wenfen 2006, Taboada et al. 2007], with
very promising results. [Taboada et al. 2007] used different kinds of SVMs: SVM classification (multi-class one-against-all),
ordinal SVM and SVM regression, and they found that the SVMs are perfectly comparable to kriging (a statistical model) and
have better control of outliers. Like the other AI techniques described in the previous sections, FN have also featured in a number
of research studies. [Castillo et al. 2001] gave a comprehensive demonstration of the application of FN in Statistics and
Engineering.
Most hybrids found in the literature are usually fused with Neural Networks and one other technique. This is due to their
popularity and age-long use in the computational intelligence field. Other hybrid systems include different combinations of several
techniques such as SVM, FL, Decision Trees, Extreme Learning Machines, Genetic Algorithm, Hidden Markov Model and Radial
Basis Functions, as applied in various fields including reservoir characterization.
SPE 126649 3

A fuzzy Linear Programming Support Vector Machine (LP-SVM) was used for multiclass classification problems by
[Abe 2004], demonstrating the superiority of the hybrid over the conventional SVMs. [Deyi et al. 2005] developed a hybrid
genetic programming and fuzzy/neural network inference system to estimate the permeability of reservoirs. When compared with
contemporary estimation approaches, the hybrid yielded more consistent and robust estimated results. [Chikhi and Batouche 2004]
combined a neural method with Radial Basis Functions to enhance the classification of lithofacies of certain wells. This was
followed by a similar hybridization effort by [Salim 2006] that combined ANN with Hidden Markov Models to obtain the
lithological identification of the same wells. The results obtained by the hybrid were close to those obtained by the fuzzy Adaptive
Resonance Theory (ART) approach applied to the same borehole with the same well logs. A hybridization of SVMs and the
Interval Type-2 FL System (FLS) was performed by [Chen et al. 2007] to better handle uncertainties existing in real classification
data. A method of fuzzy-c-means clustering by using genetic algorithm in a hybrid was proposed in [Wang 2008]. Experimental
results showed that the method can search global optimum partly to make the clustering analysis more rational. Another interesting
hybrid system was described in [Habibi et al. 2008].

3. Fuzzy Logic, Support Vector Machines, Functional Networks and Hybrids


3.1 Fuzzy Logic
Type-2 FLS was introduced as an extension of the concept of Type-1 FLS. Type-2 FLS has membership grades that are
themselves fuzzy. For each value of a primary variable (e.g., pressure and temperature), the membership is a function (not just a
point value). The secondary Membership Function (MF) has its domain in the (0, 1) interval. Hence, the MF of a Type-2 FLS is
three dimensional, and it is the newly introduced third dimension that provides new degrees of design freedom for handling
uncertainties. Figures 1 and 2 below show respectively the structure of a Type-2 FLS and Gaussian MF with uncertain mean.
Further descriptions of the concepts of Fuzzy Logic Systems, including Type-2 Fuzzy Logic System and their applications can be
found in [Castillo and Melin 2008, Karnik and Mendel 1999, Mendel 2003, Wang 2008].

3.2 Support Vector Machines


Support Vector Machines are a set of related supervised learning methods used for classification and regression. They
belong to a family of generalized Linear Classifiers. They can also be considered as a special case of Tikhonov Regularization.
SVMs map input vectors to a higher dimensional space, which is called a feature space, where a maximal separating hyperplane is
constructed [Burges 1998, Littman 2000]. This is shown in figure 3 below. More references on the structure of SVMs and their
applications can be found in [Abe 2004, Taboada et al. 2007, Xing et al. 2008].

3.3 Functional Networks


Functional Networks are extensions of Neural Networks which consist of different layers of neurons connected by links.
Each computing unit or neuron performs a simple calculation: a scalar typically monotone function f of a weighted sum of inputs.
The function f, associated with the neurons, is fixed and the weights are learned from data using some well-known algorithms such
as the least-square fitting. A functional network consists of a layer of containing the input data; a layer of output units containing
the output data; one or several layers of neurons or computing units which evaluate a set of input values, coming from the input
units, and which give a set of output values to the output units. The computing units are connected to each other, in the sense that
the output from one unit can serve as part of the input to another neuron. Once the input values are given, the output is determined
by the neuron type, which can be defined by a function [Castillo et al. 2001]. Figure 4 shows the structure of a Functional Network
and its simplification.

3.4 Hybrids
An approach resulting from the combination of two or more approaches is called a hybrid. It has also been defined as an
approach that combines different theoretical backgrounds and algorithms such as data mining and soft computing methodologies.
The main idea behind hybridization is to complement the weaknesses of one technique with the strength of other techniques. Since
no single technique is good for everything and in all situations, there is the need to combine the individual capabilities of each
technique to obtain a more versatile and robust technique. A number of hybrids and their applications have been mentioned in
section 2.

4. Data, Experimental Design and Criteria for Evaluation


4.1 Description of Data
Well logs for porosity and permeability from six wells were used for the design, testing and validation of this work. The
three well logs for porosity were obtained from a drilling site in the Northern Marion Platform of North America (site 1) and the
three for permeability from a drilling site in the Middle East (site 2). The datasets from site 1 have six predictor variables for
porosity, while the dataset from site 2 has eight predictor variables for permeability. These are shown in tables 1 and 2.
4 SPE 126649

Fig. 1: The structure of a Type-2 FLS Fig. 2: Gaussian MF with uncertain mean

Regression in Primal Space


Y

Training Dataset

Fig. 3: Mapping input vectors to a higher dimensional Fig. 4: Structure of Functional Networks and its
space in SVM Simplification

Predictors for Porosity Predictors for Permeability Full Meaning


1 Core 1 GR Gamma Ray Log
2 Top Interval 2 PHIE Porosity Log
3 Grain Density 3 RHOB Density Log
4 Grain Volume 4 SWT Water Saturation
5 Length 5 RT Deep Resistivity
6 Diameter 6 MSFL Microspherically Focused Log
7 NPHI Neutron Porosity Log
8 CALI Caliper Log
Table 1: Predictor Variables
for Site 1 well log for Table 2: Predictor Variables for Site 2 well log for
Porosity Permeability

The available data for each of the wells is divided into training and test data by using a Stratified Sampling technique.
Using this technique, 70% of the entire data goes for training, and the remaining 30% goes for testing. To further ensure fairness
and integrity of the results obtained, several iterations were made, and the average of the runs was obtained. Table 3 shows the two
well logs with their sizes and divisions into training and test sets.

4.2 Experimental Design


The methodology in this work is based on the standard Computational Intelligence approach to hybridization of Artificial
Intelligence techniques using Fuzzy Logic (FL), Support Vector Machines (SVM) and Functional Networks (FN). We have used
SPE 126649 5

the Least Squares-SVM (LS-SVM) for the implementation of SVM, Iterative Least-Squares Functional Networks for FN, and the
Interval Type-2 proposed by [Karnik and Mendel 1999, Mendel 2003] for the Type-2 Fuzzy component. The use of Interval Type-
2 FLS was preferred over the general Type-2 version because, according to [Mendel 2003], the former is too complicated as the
process of calculating the meet operations for each fired rule and the procedures for type reduction are prohibitive, time-
consuming and memory-intensive. Hence, Interval Type-2 FLS is more practical for implementation than the general Type-2.
The hybrid models were designed to benefit immensely from the strength of the individual techniques, and to complement
the weaknesses of one technique with the advantages of the others, and hence to combine the cooperative and competitive
characteristics of the individual techniques. The reasons for choosing each of the components of the hybrid in this work are
presented in table 4.

Site 1 (Porosity) Site 2 (Permeability)


Wells 1 2 3 1 2 3
Data Size 415 285 23 355 477 387
Training (70%) 291 200 16 249 334 271

Testing (30%) 124 85 7 106 143 116

Table 3: Division of datasets into Training and Testing

Technique Strength Weakness

Type-2 FLS - Ability to determine an exact membership functions for a - Complexity of implementation.
fuzzy set. - Too much time spent in tuning the
- Ability to handle uncertainties. parameters used for inferencing
- Ability to extract rules directly from input data [Karnik and during the training process.
Mendel 1999, Mendel 2003]. - Does not obtain good performance
when the number of training data is
small [Mendel 2003].
SVM - Ease of training. - It is weak in the sense that it needs
- No local optima, unlike in neural networks. a good kernel function [Littman
- It scales relatively well to high dimensional data. 2000].
- Ability to explicitly control the tradeoff between
complexity and error.
- It is also known to have the capability of using small
training dataset [Jian and Wenfen 2006].

FN - There is no need to include weights associated with - In some cases, the function obtained
links. from the learning process is not
- Functional approximation. invertible. In this case, the existence
- Ability to select the best among functions by minimizing of a single optimal value is not
the sum of square errors [Castillo et al. 2001]. guaranteed [Castillo et al. 2000].

Table 4: Strengths and weaknesses of FL, SVM and FN

4.2.1 Framework of the Hybrid Models


To achieve our aim, two different hybrid models were built: FN-Fuzzy Logic-SVM (FFS) and FN-SVM-Fuzzy Logic
(FSF). FN was used as the base for each of the models. This is due to its functional approximation capability and its ability to
select the best variables for the system directly from the training data.
The following subsections describe the framework design of the two hybrid models and the optimized parameters of each
of the techniques before they were used in the hybrids.

4.2.1.1 FN-Fuzzy Logic-SVM (FFS)


This model is composed of three major blocks containing, respectively: Functional Networks (FN), Type-2 Fuzzy Logic
and Support Vector Machines (SVM). The FN block, using its least-squares fitting algorithm, was used to select the best variables
from the input data. The dimensionality of the input data is automatically handled by the FN block that plays the role of a best-
variable selector in the model. The best variables are extracted from the input data and then divided into training and test sets using
6 SPE 126649

the Stratified Sampling approach as explained in section 4.1. The training sample is passed to the Fuzzy Logic block where
uncertainties are removed. The training data, with uncertainties removed, is then used to train the SVM block. Finally, the test data
is passed through the trained SVM block to perform the regression task. The role performed by the Fuzzy Logic block in this
model is to ensure that in case an input data containing uncertainties is used, such uncertainties would have been removed before
the data is passed to the SVM block for training. In this way, only clean data is allowed to enter the SVM block which performs
the prediction task after the training process. This is an attempt to complement the performance of the hybrid with the ability of
Type-2 Fuzzy Logic to handle uncertainties. However, in the absence of uncertainties in the input data, then the performance of the
Type-2 FLS block is reduced to that of Type-1 FLS [Karnik and Mendel 1999, Mendel 2003]. Figure 5 shows the design
framework of this model.
It should be noted that Type-2 Fuzzy Logic System has been shown in several works such as in [Castillo and Melin 2008,
Castillo et al. 2000, Dongrui and Mendel 2008, Li et al. 2007, Wang 2008, Zarandi et al. 2007] to have the ability to remove
uncertainties using its extension to a third dimension. It would be futile to attempt to revalidate this already established fact in the
literature. Hence, the focus of this work is the successful combination of the individual techniques to implement hybrid models
that demonstrate the combined capabilities of each technique and their role in the hybrid models.

4.2.1.2 FN - SVM - Fuzzy Logic (FSF)


The architecture of this model is similar to the FFS. The only difference is that the Type-2 Fuzzy Logic and SVM blocks
in the FFS are transposed to obtain a new architecture. The training sample from FN is passed to the SVM block. Part of the
training process in SVM involves the transformation of the input data to a higher dimensional space, called the feature space
[Littman 2000], where an optimal hyperplane is constructed. The SVM block transforms the training set to the feature space and
passes it to the Type-2 Fuzzy Logic block for training. Training in Type-2 Fuzzy Logic System involves fuzzification and
transformation to a higher dimension (the third dimension) where uncertainties are easily handled and inference rules are extracted
from the transformed data. The final predictions are made by passing the test sample through the trained FL block in order to
evaluate the performance of the overall hybrid model. The role performed by the SVM block in this model is to ensure that, in case
the volume of the training data is small, the hybrid will still be able to cope. Figure 6 shows the design framework of this model.

4.2.2 Model Implementation, Configuration and Validation


In order to learn the Functional Network, there was the need to do a model selection to choose the best Functional
Network model using the Minimum Description Length (MDL) principle. It is also used to compare models with different
parameters, because it has a penalty term for overfitting. Accordingly, the best Functional Network model for a given problem
corresponds to the one with the smallest MDL value. This was calculated using the Backward-Forward method. A detailed
description of this method can be found in [Castillo et al. 2001]. The general behavior of SVM is described as: Given a set of data
points,
G = {( x i , d i )}i
n

x d
where i is the input vector, i is the desired target value and n is the total number of data points. SVMs approximate
the function with three distinct characteristics: estimating the regression in a set of linear functions, defining the regression
estimation as the problem of risk minimization with respect to the -insensitive loss function and minimizing the risk by using the
Structural Risk Minimization (SRM) principle.
Training data
Least Square with uncertainties
Fitting removed.

Training Fuzzy
Dataset
Stratified SVM
Input FN Predicted
Data Sampling Output
Testing
Dataset
Data with best SVM trained with
input variables clean data from
Fuzzy Logic.

Fig. 5: Design framework of FFS hybrid model.


SPE 126649 7

Training data transformed


Least Square to higher dimensional
Fitting feature space.

Training
SVM
Dataset
Input Stratified
FN Sampling Fuzzy Predicted
Data Output
Testing
Dataset
Data with best Fuzzy Logic
input variables trained with
transformed data
from SVM

Fig. 6: Design framework of FSF hybrid model.

The linear function is formulated in the high dimensional feature space in the form:

y = f (x ) = w (x ) + b
where ( x ) is the high dimensional feature space, which is non-linearly mapped from the input space x. The
coefficients w and b are estimated according to the process of risk minimization. The goal of this risk function is to find a function
that has, at most, deviation from the actual values in all the training data points. In this work, the kernel used is polynomial,
which is of the form:
K (x , y ) = (x T y + 1)d
and the error, designated by , was set to 0.2. The other optimized parameters are: C= 450; lambda=1e-7; epsilon=0.2;
kerneloption=0.30; kernel='poly'; and verbose=1.
Interval Type-2 Fuzzy Sets with Gaussian Membership Functions are the most widely used Type-2 Fuzzy Sets because
they are simple to use and at present, it is very difficult to justify the use of any other kind. When the Type-2 Fuzzy Sets are
Interval Type-2 Fuzzy Sets, all secondary grades (flags) are equal to 1 [Karnik and Mendel 1999, Mendel 2003].
For the Type-2 Fuzzy Logic System, rules were extracted directly from the input data using the Back-propagation
(Steepest Descent) method. This option was preferred over the extraction of rules from a subset of the data, because this ensures all
data points are completely represented in the inference engine. This leads to more time spent in implementation, but better
accuracy and efficiency is guaranteed. To be able to use Type-2 Fuzzy Logic System effectively, the standard steps as suggested
by [Mendel 2003] were followed. More details of the structure and implementation of this are also available in [Karnik and
Mendel 1999, Mendel 2003]. The optimized parameters of the Type-2 Fuzzy block are: sn2 = sn1 (this is the standard deviation of
the input data) and alpha4 = alpha = 0.1 (this is the learning parameter).
The individual techniques were run separately. Then their relevant components were combined in the hybrids and also run
with the same input dataset. In order to ensure most fairness in the results, several iterations were made and the average values of
the results were taken. This is necessary due to the behavior of the Stratified Sampling approach used to divide the input data into
training and testing sets. Since the input data is randomized during the stratification processes, hence, slightly different results are
obtained for each run of the experiments.

4.3 Criteria for Performance Evaluation


In order to establish a valid evaluation of this work, we have used Correlation Coefficient (CC), Root Mean-Squared
Error (RMSE), and Execution Time (ET) as criteria for measuring the performance. The CC measures the statistical correlation
between the predicted and actual values. A value of 1 means perfect statistical correlation and a 0 means there is no correlation
at all.

The formula is:


( x x ')( y y ')
( x x ') ( y y ')
2 2
8 SPE 126649

where x and y are the actual and predicted values, while x and y are the mean of the actual and predicted values.
The RMSE is one of the most commonly used measures of success for numeric prediction, computed by taking the square
root of the average of the squared differences between each predicted value xn and its corresponding actual value yn. It gives the
error value the same dimensionality as the actual and predicted values.

(x 1 y 1 ) 2 + (x 2 y 2 ) 2 + ... + (x n y n ) 2
The formula is: where n is the size of data.
n

The ET is simply the total CPU time taken (measured in seconds) for a model to run from the beginning to the end of the
desired process, and it is computed as: T 2 T 1 , where T2 is the CPU time at the end of the process and T1 is the CPU time at the
beginning.

5. Experimental Results and Discussion


5.1 Experimental Results
We implemented and validated the individual techniques as well as the two hybrid models in the prediction of porosity
and permeability by using the training and testing data described in section 4.1. Several iterations were made, and the average of
results was taken. Tables 5 7 present the result of the prediction of porosity for the 3 wells in site 1, showing the comparative
correlation coefficients, root mean squared errors and the execution times for the five techniques as applied to the three porosity
wells.

Site 1, Well 1 (Porosity)


Model Correlation Coefficient RMSE Execution Time (s)
Training Testing Training Testing Training Testing
SVM 0.820348 0.94502 7.54717 2.80199 15.458333 0.000001
FunNet 0.835855 0.957491 6.17685 4.15040 0.098958 0.0000001
Fuzzy Logic 0.840697 0.816021 6.85104 7.52180 155.395833 55.937500
Hybrid-FFS 0.920608 0.969179 7.06432 3.72873 60.979167 9.093750
Hybrid FSF 0.916415 0.957003 6.61915 5.67851 61.562500 8.890625

Table 5: Comparative Results of the Porosity Prediction Site 1, Well 1

Site 1, Well 2 (Porosity)


Model Correlation Coefficient RMSE Execution Time (s)
Training Testing Training Testing Training Testing
SVM 0.806304 0.775851 6.75723 7.17699 7.354167 0.000001
FunNet 0.803918 0.775465 6.77203 7.18608 0.093750 0.0000001
Fuzzy Logic 0.798557 0.723302 7.31978 7.69430 71.526042 25.817708
Hybrid - FFS 0.82147 0.80758 9.18052 7.09860 28.442708 4.411458
Hybrid FSF 0.8147 0.813319 9.09662 9.45112 31.567708 4.296875

Table 6: Comparative Results of the Porosity Prediction Site 1, Well 2

Site 1, Well 3 (Porosity)

Model Correlation Coefficient RMSE Execution Time (s)


Training Testing Training Testing Training Testing

SVM 0.980730 0.898397 1.64195 2.95456 0.315625 0.000001


FunNet 0.824680 0.803017 4.75388 6.26507 0.062500 0.0000001
Fuzzy Logic 0.910924 0.599756 4.26323 7.04103 0.518750 0.129688
Hybrid-FFS 0.93055 0.914020 6.37047 3.34832 0.493750 0.042188
Hybrid FSF 0.920961 0.928830 6.81454 8.32623 0.471094 0.051042

Table 7: Comparative Results of the Porosity Prediction Site 1, Well 3

These are presented in figures 7 to 9, showing the comparative correlation coefficients and execution times (in seconds)
of the five techniques as applied to the three wells for porosity.
SPE 126649 9

Figure 7: Porosity Training and Testing Correlation Coefficients for all techniques and on all Wells.

Figure 8: Porosity Training and Testing Execution Times for all techniques and on Wells 1 and 2.

Figure 9: Porosity Training and Testing Execution Times for all techniques and on Well 3.

Similarly, tables 8 - 10 present the result of the prediction of Permeability for the 3 wells in site 2 showing the
comparative correlation coefficients, root mean squared errors and the execution times (in seconds) for the five techniques as
applied to the three permeability wells.

Site 2, Well 1 (Permeability)


Model Correlation Coefficient RMSE Execution Time (s)
Training Testing Testing Testing Training Testing
SVM 0.854109 0.832785 0.64724 0.68445 10.052083 0.000001
FunNet 0.852724 0.833204 0.64785 0.68113 0.140625 0.0000001
FL 0.869565 0.844982 0.62187 0.66946 187.494792 68.828125
Hybrid - FFS 0.872843 0.858610 1.10615 0.66580 43.333333 6.231250
Hybrid FSF 0.827372 0.901920 0.72800 0.64289 44.630208 6.531250

Table 8: Comparative Result of the Permeability Prediction for Site 2, Well 1


10 SPE 126649

Site 2, Well 2 (Permeability)

Model Correlation Coefficient RMSE Execution Time (s)


Training Testing Testing Testing Training Testing
SVM 0.864894 0.881033 0.63803 0.61685 18.661458 0.005208
FunNet 0.868102 0.877959 0.62799 0.62760 0.182292 0.000001
Fuzzy Logic 0.860039 0.866123 0.65863 0.67214 355.026042 128.739583
Hybrid FFS 0.88142 0.895119 1.11178 0.60509 79.062500 12.463542
Hybrid FSF 0.88144 0.903620 0.73102 0.72689 87.697917 14.140625

Table 9: Comparative Result of the Permeability Prediction for Site 2, Well 2

Site 2, Well 3 (Permeability)

Model Correlation Coefficient RMSE Execution Time (s)


Training Testing Testing Testing Training Testing

SVM 0.800820 0.767553 0.73625 0.76481 12.802083 0.010417


FunNet 0.787706 0.730623 0.74712 0.76140 0.177083 0.000001
Fuzzy Logic 0.753439 0.741249 0.83188 0.86444 241.822917 89.036458
Hybrid FFS 0.768380 0.789960 1.10064 0.73154 51.885417 7.791667
Hybrid FSF 0.769102 0.801785 0.78078 0.78106 59.447917 11.098958

Table 10: Comparative Result of the Permeability Prediction for Site 2, Well 3

These are also presented in Figures 10 and 11, showing the comparative correlation coefficients and execution times (in
seconds) of the five techniques as applied to the three wells for permeability.

Figure 10: Permeability Training and Testing Correlation Coefficients for all techniques and on all Wells.

Figure 11: Permeability Training and Testing Execution Times for the all techniques and on all Wells.
SPE 126649 11

5.2 Discussion of Results


In the prediction of porosity and permeability using the FFS and FSF hybrid models, the results showed that the hybrid
models performed better than, or competitively equal to, the three individual techniques used separately, in terms of their
correlation coefficient. A special characteristic was observed in the result of site 1 well 3 for Porosity (table 7), where SVM
demonstrated its ability to withstand a shortage of training data but FL demonstrated otherwise. Still, the hybrid models performed
better than the individual components due to the cooperative spirit that has been built into them.
In terms of execution time, the results showed that Functional Networks are the fastest in terms of both training and
testing, followed by SVM. Type-2 Fuzzy Logic took the most time for both training and testing, due to its complexity as described
in section 4.2 and Table 4. The hybrid models proved to be faster than the Fuzzy Logic component, but not for SVM and FN. This
is the price for obtaining better models in terms reliability and robustness.
Since the root mean squared errors are generally low for all the models, the little variations would not effectively
demonstrate the true performance of these models. Hence, more weight is given to correlation coefficient and execution time for
the evaluation of the models.

6. Conclusion
Two design frameworks for the hybridization of Fuzzy Logic, Support Vector Machines and Functional Networks have
been implemented and presented. They were tested with two well logs containing six porosity and permeability datasets. The
results showed that the hybrids performed better than the individual techniques used separately for all the datasets in the prediction
of porosity and permeability. With any novel data, the hybrid models will be able to select the best set of input variables to use,
extract inference rules, and handle uncertainties that might be present, directly from the data, and perform the required prediction.
The hybrid models were also seen to be very reliable, robust and effective, based on the reports in the literature and the results in
this study.
This work focuses only on the successful combination of these individual techniques to implement hybrid models that
demonstrate the combined capabilities of each technique, rather than attempting to repeat previous work on the ability of Type-2
Fuzzy Logic System to handle various degrees of uncertainties in input datasets. Also, the Type-2 FLS has some limitations as
presented in [Karnik and Mendel 1999, Mendel 2003], and our frameworks were built on these limitations.
Comparing the performance of these hybrid models with similar existing models was difficult either because such works
as in [Mohsen et al. 2007, Jian H. and Wenfen 2006, Chikhi and Batouche 2004, Deyi et al. 2005] did not clearly present
numerical performance indices or did not publish their datasets. Also, SVM, FN and Type-2 FL had sparsely been used in the
prediction of porosity and permeability of oil and gas reservoirs. Interestingly, our hybrid models with a maximum R2 of 0.90
outperformed the Artificial Neural Network model of [Maqsood and Adwait 2000] with an R2 of 0.70 in the prediction of
permeability. However, the models did not use the same dataset.
Motivated by the success of this work, part of our future work will be directed towards modifying these hybrid models in
order to solve classification and pattern recognition problems in the oil and gas industry such as Lithofacie and History Matching.

Acknowledgments
We would like to thank King Fahd University of Petroleum and Minerals (KFUPM); the Center for Petroleum and Minerals in
KFUPM; Department of Information and Computer Science in KFUPM and the University Research Institute for providing the
computing facilities and a conducive research atmosphere.
References
1. AbdulKadir O., Tiab D. and Mazouzi A., Application of Artificial Intelligence to Characterize Naturally Fractured Zones in
Hassi Messaoud Oil Field, Algeria, Journal of Petroleum Science & Engineering, 49 (2005) 122-141.
2. Abe S., "Fuzzy LP-SVMs for Multiclass Problems", Proceedings of the European Symposium on Artificial Neural Networks,
Belgium (2004) 429-434.
3. Ali J. K., Neural Networks: A New Tool for the Petroleum Industry?, Proceedings of Society of Petroleum Engineers
European Petroleum Computer Conference, Aberdeen, U.K. (1994) 217-231.
4. Castillo E., Gutirrez J., Cobo A. and Castillo C., "A Minimax Method for Learning Functional Networks", Neural Processing
Letters, 11 (2000) 3949.
5. Castillo O. and Melin P., Intelligent Systems with Interval Type-2 Fuzzy Logic, International Journal of Innovative
Computing, Information and Control, Vol. 4, No.4 (2008) 771-784.
6. Castillo E., Gutirrez J., Hadi S., and Lacruz B., Some Applications of Functional Networks in Statistics and Engineering,
Technometrics, Vol. 43 (2001) 1024.
7. Chen X., Li Y., Harrison R., and Zhang Y.Q., "Type-2 Fuzzy Logic Based Classifier Fusion for Support Vector Machines",
Applied Soft Computing Journal, doi:10.1016/j.asoc.2007.02.019 (2007).
12 SPE 126649

8. Chikhi S. and Batouche M., "Probabilistic Neural Method Combined with Radial-Bias Functions Applied to Reservoir
Characterization in the Algerian Triassic Province", Journal of Geophysics and Engineering, Vol. 1 (2004) 134-142.
9. Deyi X., Dave W., Tina Y. and San R., "Permeability Estimation Using a Hybrid Genetic Programming and Fuzzy/Neural
Inference Approach", Society of Petroleum Engineers Annual Technical Conference and Exhibition, Dallas, Texas, U.S.A.
(2005).
10. Dongrui W. and Mendel J.M., A Vector Similarity Measure for Linguistic Approximation: Interval Type-2 and Type-1 Fuzzy
Sets, Information Sciences 178, (2008) 381402.
11. Fang J. H. and Chen H. C., "Fuzzy Modeling and the Prediction of Porosity and Permeability from the Compositional and
Textural Attributes of Sandstone", Journal of Petroleum Geology, Vol. 20(2) (1997) 185-204.
12. Giovanni A. and Vincenzo L., Using FML and Fuzzy Technology in Adaptive Ambient Intelligence Environments,
International Journal of Computational Intelligence Research, Vol.1, No.2 (2005) 171182.
13. Goda H.M., and Behrenbruch H.R., "Use of Artificial Intelligence Techniques for Predicting Irreducible Water Saturation in
Australian Hydrocarbons Basins", Society of Petroleum Engineers Asia Pacific Oil & Gas Conference and Exhibition,
Jakarta, Indonesia (2007).
14. Habibi J., Moshiri B. and Sedigh A. K., Contractive Predictive Control of Mixed Logical Dynamical Hybrid Systems,
International Journal of Innovative Computing, Information and Control, Vol. 4, No.6 (2008) 1283-1298.
15. Jian H. and Wenfen H., "Novel Approach to Predict Potentiality of Enhanced Oil Recovery", Proceedings of the Society of
Petroleum Engineers Intelligent Energy Conference and Exhibition, Amsterdam, The Netherlands (2006).
16. Karnik N. and Mendel J., Type-2 Fuzzy Logic Systems, IEEE Transactions on Fuzzy Systems, Vol. 7, No. 6 (1999).
17. Li Y., Sun X., Hua J. and Gong C., Modeling Redundant Structure in Ecosystem by Type-2 Fuzzy Logic System, Ecological
Modeling, Article in Press (2007).
18. Lim J., Reservoir Properties Determination using Fuzzy Logic and Neural Networks from Well Data in Offshore Korea,
Journal of Petroleum Science and Engineering 49, (2005) 182 192.
19. Littman W., Introduction to Support Vector Machines, Lecture Notes in Computer Science (LNCS), Vol. 536 (2000).
20. Maqsood A. and Adwait C., Using Artificial Intelligence to Predict Permeability from Petrographic Data, Computers &
Geosciences 26 (2000) 915 - 925.
21. Mendel J., Type-2 Fuzzy Sets: Some Questions and Answers, IEEE Connections, Newsletter of the IEEE Neural Networks
Society, Vol. 1 (2003) 10-13.
22. Mohsen S., Morteza A. and Ali Y.V, Design of Neural Networks using Genetic Algorithm for the Permeability Estimation of
the Reservoir, Journal of Petroleum Science and Engineering, Vol. 59 (2007) 97105.
23. Munakata T., "Fundamentals of the New Artificial Intelligence Neural, Evolutionary, Fuzzy and More", 2nd Edition, Springer
Science+Business Media (2008).
24. Peddabachigari S., Abraham A., Grosan C. and Thomas J., Modeling Intrusion Detection System using Hybrid Intelligent
Systems, Journal of Network and Computer Applications, Vol. 30 (2007) 114132.
25. Phillips-Wren G., Ichalkaranje N. and Jain C., Intelligent Decision Making: An AI-Based Approach, Studies in
Computational Intelligence, Volume 97, Springer-Verlag Berlin Heidelberg (2008) 88-355.
26. Salim C., "A Fuzzy ART versus Hybrid NN-HMM Methods for Lithology Identification in the Triasic Province", 2nd
Information and Communication Technologies (ICTTA '06), Vol.1 (2006) 1884-1887.
27. Schlumberger Excellence in Educational Development, Science Lab Project for Porosity,
http://www.seed.slb.com/en/scictr/lab/porosity/index.htm (2007).
28. Taboada J., Matas J.M., Ordez C. and Garca P.J., Creating a Quality Map of a Slate Deposit using Support Vector
Machines, Elsevier Journal of Computational and Applied Mathematics, Vol. 204 (2007) 84-94.
29. Wang Y., Fuzzy Clustering Analysis by Using Genetic Algorithm, Innovative Computing, Information and Control Express
Letters, Vol.2, No.4 (2008) 331-337.
30. Xing Y., Wu X. and Xu Z., Multiclass Least Squares Auto-correlation Wavelet Support Vector Machines, International
Journal of Innovative Computing, Information and Control Express Letters, Vol. 2, No. 4 (2008) 345-350.
31. Zarandi M. F., Rezaee B., Turksen I. B. and Neshat E., A Type-2 Fuzzy Rule-Based Expert System Model for Stock Price
Analysis, Expert Systems with Applications, doi: 10.1016/j.eswa (2007) 9-34.

You might also like