Multi-attribute Based Seismic Facies Analysis and Classification for One of the Iranian Offshore Oil fields

The use of seismic data to better constrain the reservoir model between wells has become an important goal for seismic interpretation. We propose a methodology for deriving soft geologic information from seismic data and discuss its application through a case study for one of the offshore Iranian oil fields. The methodology is based on seismic facies analysis and classification.Seismic classification is a breakthrough technology to determine distribution of pore fluid and lithology from multiple seismic attribute volumes. In this study, different surface based, volume based, grid based and VRS (volume reflection spectrum) attributes are extracted from the top of the reservoir or entire reservoir interval. In order to reduce the geological uncertainty and redundancy in seismic attributes, careful study of different seismic attributes have been done to choose among all available attributes, those which are most appropriate for the classification. Following the attribute extraction, evaluation and selection, seismic facies classification is applied in both unsupervised and supervised approaches. Unsupervised method which is implemented in this study is based on K-mean algorithm which is a statistical method of clustering. Supervised classification method is based on Multi-layer Perceptron (error back propagation algorithm) which is a neural network paradigm. For both approaches, classifications with different number of classes and different combinations of seismic attributes are run. Results of two approaches confirm each other and allow identification of the main reservoir (seismic) facies and heterogeneities. Seismic facies in this study can be related to lateral variations of porosity over the reservoir interval. The generated class grid (map) provides thelocation of prospective targets and the associated probability maps provide quantitative estimation of risk.

Keywords:

seismic attribute , seismic facies classification , supervised , unsupervised