The emerging issues of natural carbon dioxide (CO2) production alongside natural gas in Southeast Asian producing fields, presents challenges to well and platform facilities integrity and impedes the commercial feasibility of greenfield development.
Forecasting CO2 presence at pre-drill sites is becoming more difficult, even with sophisticated techniques like advanced quantitative inversion, Amplitude versus Offset (AVO) analysis, advanced seismic attribute analysis, and Petroleum Systems Modeling (PSM). These methods function largely independently, hindering collaborative data utilisation and insight extraction. In seismic approaches, a seismic halo, which appears as a continuous flat spot, is sometimes seen as a potential CO2 indicator. However, this feature does not consistently correlate with CO2 presence or concentration, which varies significantly from 20 to 80 percent in various fields.
In order to address this challenge an innovative approach of integrating various technical data into a Data Driven model has been adopted. This abstract focuses on the Machine Learning (ML) approach taken to predict the CO2 risk map classification from a set of seismic attributes.
