The Pematang Group Sand Analysis Using Growing Neural Network Machine Learning

Abstract

The Central Sumatera Basin (CSB) is one of Indonesia's giant mature basins and attractive hydrocarbon producers. P.T. Badan Operasi Bersama - Bumi Siak Pusako (PT.BOB-BSP) company has several 3D seismic surveys in CSB that are of old vintage but still provide important data for seismic reservoir analysis in exploring lacustrine sediments of the Pematang Group, one of the most challenging reservoirs that produce hydrocarbons. In order to optimize seismic data for reservoir characterization of the Pematang Group, adequate seismic interpretation tools are needed, especially with the help of the latest technology to find optimal locations for well placement and help increase oil production by exploring and exploiting hydrocarbon in this area. Some of the latest technologies that can be applied to find new insights from old seismic data are using the Growing Neural Gas (GNG) algorithm clustering techniques and volumes visualization technology. GNG can be used for clustering to organize a collection of k-dimensional vectors into groups whose members share similar features for 3D seismic volumes of data, significantly clustering the quality of reservoir sand and anomalies in lacustrine sediments of the Pematang Group. This study aims to use GNG clustering techniques for reservoir characterization in 3D seismic surveys using data from the CSB. The workflow consists of three main phases: preparation, execution, and delivery. The first phase consists of collecting, cleaning, and preprocessing seismic, wells and interpretation data. Once the preparation is considered satisfactory, it will be followed by the execution, starting with data processing, clustering, and validation. GNG clustering methods were deployed to build a facies volume based on three seismic attributes inside the interval of the Pematang Group. After the facies model was satisfactory, a facies map was extracted from the facies volume along the Pematang Zone to generate a facies distribution map. The last step is to validate the facies extraction map of the facies model with the existing well log data and check the accuracy based on the well log result. Once the validated results are satisfying, the workflow status will change from execution to the delivery phase to create the final project conclusion and presentation. This study demonstrated excellent prediction to image or model seismic facies as facies volumes. This facies model allows for classifying and identifying the Pematang sand distribution for reservoir characterization analysis. It provides new insights and optimal locations for exploration drilling targets more effectively, increases drilling success and reduces cost and risk for optimizing exploration hydrocarbon strategy in Pematang Group.