Study on sedimentary facies based on unsupervised neural network seismic attribute clustering

Abstract

The classification of seismic facies based on unsupervised neural network self-organizing analysis (SOMA) is a comprehensive attribute clustering method. The key to the application of this method is to optimize seismic attributes, determine the number of clustering types, and analyze the relationship between seismic facies and sedimentary facies. Under the guidance of seismic sedimentology theory, we use the SOMA (self-organizing analysis) technology for cluster analysis of attributes, carry out seismic- sedimentary facies analysis by combining basic geological data, and select four seismic attributes such as RMS amplitude, information entropy, chaotic Li and fractal correlation dimension for cluster analysis. Taking the Cretaceous Suhongtu Formation in the Aitgele sag as a case, and using the method, we found such sedimentary facies as fan delta, braided river delta, shallow shore lake and deep lake. Traditional seismic-sedimentary facies analysis can judge the type of seismic facies by artificially observing seismic reflection. In contrast, our technology can reduce the unreliability of sedimentary facies analysis in areas with less well data. It provides a new basis for sedimentary facies analysis for oil and gas exploration. Also it is a practical, objective and accurate technical means.