-
oa Sweet Spot Interpretation from Multiple Attributes: Machine Learning and Neural Networks Technologies
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, First EAGE/AMGP/AMGE Latin-American Seminar in Unconventional Resources, Nov 2017, Volume 2017, p.1 - 4
Abstract
The “sweet spot” interpretation of shale gas routinely involves advanced visualization techniques, and generation of numerous seismic data types and attributes. Commonly used seismic attributes include the total organic carbon content (TOC), pore pressure, stresses, rock elasticity, brittleness and fracture development. To derive even more useful information from the multiple attributes and provide a easily tool for the characteristic analysis of the target shale reservoir, current visualization techniques, machine learning and neural networks technologies, including self-organizing map (SOM), K-means clustering, principal component analysis (PCA) and two-dimensional HSV color map have been all introduced to reveal the geologic features that are not previously identified or easily interpreted from the numerous seismic attributes. Through the attribute patterns generated by SOM, the unsupervised (K-means clustering), supervised (human-computer interaction monitoring clustering) classifications can be applied for flexibly explaining the characteristics of “sweet spot”. Integrating with the PCA and HSV color map techniques, the distribution in the attributes selected are intuitively reflected and described.