APPLICATION OF MACHINE LEARNING TECHNIQUES FOR DETAILED FACIES MODELLING OF LOWER GORU FORMATION BY USING THE WELL LOG DATA OF SAWAN GAS FIELD, CENTRAL INDUS BASIN, PAKISTAN

Abstract

The advancement in computational technology revolutionized the exploration geoscience. Machine learning tools can be used for the analysis of large datasets and to predict the relationships present among the data. Information about subsurface rock formations is gathered during the exploration of hydrocarbons. The subsurface data is quite large and difficult to handle. So, machine learning techniques like K-means clustering and self-organizing maps can be used to find manage subsurface data. The Lower Goru Formation is a prolific reservoir of Middle Indus Basin and is known for its lithological heterogeneity. The lithology of the Lower Goru Formation comprises of alternating layers of sandstone and shale. The sandstone of the Lower Goru Formation is further divided into D, C, B and A interval. Efficient identification of lithology is the key step of reservoir characterization. Present study focuses the petrophysical analysis of the Lower Goru Formation followed by the facies classification by using K-means clustering and self-organizing maps. The study was performed by using the well log data of Sawan-02, Sawan-03 and Sawan-07 wells. The petrophysical analysis reveals that the C interval of the Lower Goru Formation is a good reservoir having the average volume of shale 15%, effective porosity 12% and water saturation 20%. K-means clustering is used to classify the data into ten clusters which were then consolidated into five groups by using hierarchical clustering. These groups of data were used for facie classification. It is evident from K-means clustering the C interval of the Lower Goru Formation mainly consists of reservoir facies. The self-organizing map were used to display the multidimensional well data into two-dimensional facies map. The result obtained from selforganizing maps confirm the results of K-means clustering. So, these methods can be efficiently used for the lithological identification of reservoir formation which will reduce the risk associated with exploration of hydrocarbons.