|Abstract: ||Reservoir characterization refers to the process of inferring information about
reservoir properties from seismic data. Obtaining accurate information about
the reservoir properties such as porosity, lithology, and permeability is an
essential objective in seismic exploration, especially in new areas that lack well
control. This thesis contributes to the integrated analysis of 3-D seismic data
and well logs for a square study area in the eastern province of Saudi Arabia,
allowing improved understanding, interpretation and characterization of an
upper Jurassic carbonate reservoir. The thesis focuses on the analysis aspect of
the 3-D post-stack for seismic reservoir characterization through the
interpretive use of seismic attributes using different approaches.
The thesis can be divided into two key stages. First, a pre-processing stage
covering the quality-control of the seismic data sets, calculation of seismic
attributes, flattening of the 3-D seismic cube along target horizons, and
calibration between seismic data and well-logs. The instantaneous attributes
(amplitude, phase and frequency) of seismic data can be calculated and used,
along with relative acoustic impedance, as the main seismic attributes to
elucidate reservoir characteristics and to reduce exploration risk.
Secondly, a main analysis stage develops and tests different effective
techniques for analyzing seismic data and conducting reservoir
characterization. Five main tools have been developed in-house through
MATLAB coding to obtain accurate spatial mapping of the reservoir most
important properties that can be used for modelling and simulation which
provide better understanding of the reservoir under investigation. This
particular choice of tools should work properly for post-stack data.
The following summarises and highlights the main contributions of the thesis.
First, is to enhance the predictive performance of the conventional multiple
linear regression method through coupling information from cluster analysis.
Then, I introduce the ‘grey system theory’, which was originally developed in
China and has seen little application in geophysics, as a new tool for
hydrocarbon exploration; I propose its use for detecting hydrocarbon anomalies
associated with the carbonate reservoir. Next, I implement a Kohonen self-organizing
map (SOM) neural network for clustering the reservoir
heterogeneity (main lithofacies), and enhance the method by feeding it multiple
attributes as an input. Furthermore, I estimate reservoir porosity and
permeability by implementing a supervised back-propagation neural network.
Finally, a hybrid approach that combines an artificial neural network and a
fuzzy interface is developed for estimating well lithology from well logs.
Different informative results were drawn from this study which can be
summarised as follow:
The result indicates that the upper part of the ZOI is more porous than the
lower part. The reservoir porosity is ranging from 5% to around 28% within the
ZOI with an average porosity of approximately 15%. In addition, the reservoir
permeability shows ranging values from less than 500md to 2500md. The zone
of interest (ZOI), in general, is divided into three distinct subzones ranging in
their reservoir quality. This study indicates that the upper zone, middle zone,
and lower zone of the ZOI are featured by (medium porosity / high
permeability), (high porosity / low permeability), and (low porosity / medium
permeability), respectively. The mapping result of the reservoir lithofacies
spatial distribution indicates that there are at least nine major lithofacies
deposits. Wackestone, packstone, grainstone, and mudstone are four types of
the main lithofacies within the study area.
The main conclusions drawn from this study can be summarised as follow:
(a) The main aim of this study was achieved by estimating the reservoir
porosity and permeability, as well as, clustering the reservoir lithology
into the main lithofacies through ‘multiple linear regression’ and
‘artificial neural networks’ methods which proved (after validation) to
be a powerful technique for characterizing reservoirs, especially the
(b) The grey system theory has been introduced to the reservoir study
field and ‘grey attribute’ is proposed to highlight hydrocarbon
accumulations after finding good correlation with the producing wells
in the area.
(c) An innovative implementation of ART2 neural network has been
proposed to estimate the intra-well lithology by a hybrid-system that
combines the neural network classification with the fuzzy interface for
a better result. The final result indicated that the zone of interest
(ZOI) is dominated by grainy packstone, wackestone/packstone, and
muddy wackestone for the top, middle, and bottom subzones,
Different regional maps have been generated for the reservoir main properties
(porosity and permeability), lithofacies, and hydrocarbon accumulation.
Validation of the result has been performed taken as a measure of the method
performance and accuracy. The correlation coefficient was used to represent
the success ratio. For example, the success ratio for predicting the reservoir
porosity were 79% and 85% for the improved multiple linear regression
method and back propagation neural network method, respectively. The result
of each method has contributed substantially to achieve the main objectives of
this study not only in obtaining better understanding of the reservoir spatial
distribution for future planned drilling in the area, but also offering new input
for remodelling the reservoir and updating the simulation.|