Cluster analysis of 3D seismic data for oil and gas exploration

Abstract

Seismic attributes are information extracted from seismic data and constitute important tools to estimate the geological structure of a place, helping the understanding of the subsurface and reducing the uncertainties on interpretations. This comprehension is crucial to tasks such as lithology prediction and reservoir characterization. Seismic attributes are generated by transforming data from a seismic line (two dimensional data) or a seismic volume (three dimensional data). This work presents a study of clustering algorithms to these attributes and the techniques employed follow two distinct approaches: a self organizing map to perform crisp clustering and fuzzy c-means to perform partial clustering. The evaluations of the partitions are performed with the PBM index which indicates the best number of groups. Data from a Brazilian oil field is used to test the algorithms.