Reservoir lithofacies analysis using 3D seismic data in dissimilarity space

Abstract

Seismic data interpretation is one of the most important steps in exploration seismology. Seismic facies analysis (SFA) with emphasis on lithofacies can be used to extract more information about structures and geology, which results in seismic interpretation enhancement. Facies analysis is based on unsupervised and supervised classification using seismic attributes. In this paper, supervised classification by a support vector machine using well logs and seismic attributes is applied. Dissimilarity as a new measuring space is employed, after which classification is carried out. Often, SFA is carried out in a feature space in which each dimension stands as a seismic attribute. Different facies show lots of class overlap in the feature space; hence, high classification error values are reported. Therefore, decreasing class overlap before classification is a necessary step to be targeted. To achieve this goal, a dissimilarity space is initially created. As a result of the definition of the new space, the class overlap between objects (seismic samples) is reduced and hence the classification can be done reliably. This strategy causes an increase in the accuracy of classification, and a more trustworthy lithofacies analysis is attained. For applying this method, 3D seismic data from an oil field in Iran were selected and the results obtained by a support vector classifier (SVC) in dissimilarity space are presented, discussed and compared with the SVC applied in conventional feature space. © 2013 Sinopec Geophysical Research Institute.