Automatic techniques and machine learning to improve stratigraphic interpretation

Abstract

Stratigraphic interpretation workflows pursuits a better understanding of the reservoirs. Seismic data provides important information through detailed analysis of its response and extracted attributes to identify the geometry and extension of stratigraphic features and possible changes in lithology or fluid content. The challenge in this case, is to find a combination of methods and best practices that allow the interpreters to explore the seismic data in an effective manner and unveil hidden patterns from the information. Machine learning techniques applied to seismic interpretation have been successful in this regard and very useful in assisting with limitations involving big volumes of data, automation, and data classification. In this sense, the stratigraphic interpretation becomes a more automatic and reliable process.In this work, we aimed to show how advanced visualization techniques, with a proper combination of attributes and an efficient machine learning method, as Self Growing Neural Network, can improve the stratigraphic interpretation process, generate accurate results and provide information for uncertainty analysis.