Hydrocarbon Pre-Drilling Prediction based on Seismic using Unsupervised Learning: Example from Malaysian Basin Field

During the pre-drilling phase of hydrocarbon exploration, only seismic data is available for study to determine the presence of hydrocarbon. Currently, to identify potential hydrocarbons, amplitude interpretation often used. However, because of the complexity of structures and lithologies, there is a great deal of ambiguity or difficulty in interpreting the amplitude resulting failed well drilling is a typical occurrence that has been documented throughout the world. The amplitude traps become the most significant variables in influencing the inability to estimate the hydrocarbon location accurately. The proposed method aims to address the issue of hydrocarbon discovery during the pre-drilling stage. Self-Organizing Maps has been offered as an unsupervised learning technique for examining the objective in this proposed method. The algorithm uses a combination of pre-stack seismic features to distinguish between non-hydrocarbon and hydrocarbon prospective areas. Self-Organizing Maps can detect potential anomalies that point to the presence of hydrocarbon without training to detect potential hydrocarbon, which makes it suited for this phase because no unlabeled data or prior information is available. The suggested method is evaluated using Marmousi2 model and both well and seismic data from many fields in the Malaysian basin, and it is demonstrating excellent performance and a promising result.