Application of Seismic Facies Clustering Intelligent Recognition Technology in Yuman Area

In Western Manjiaer depression of Tarim Basin, multi-stage composite platform margin is developed, and extensive beach facies reservoir is developed in Ordovician, which is one of the key areas of platform basin exploration in recent years. The reef-bank reservoir in the lower section of Yingshan Formation in Yuman area has the characteristics of high-energy facies belt with high paleogeomorphology and banded distribution, and has the material basis for the development of large-scale and high-quality reservoirs. This study takes the beach body characterization as the core, and uses multi-attribute identification of beach facies reservoirs for qualitative characterization. The lower section of Yingshan Formation and the Penglaiba Formation of Ordovician components are divided into four types: beaded, blank, chaotic, and strong reflection. Combined with multi-stage facies-controlled inversion, the quantitative judgment of beach body recognition is carried out, that is, based on the seismic reflection characteristics, the convolution’s deep learning method is used to generate label samples from manual labeling in an interactive way, and then the label is affinely extended to multiple labels by software. At the same time, the structural characteristics of seismic data are extracted by using multiple convolution kernels, and the structural characteristics are compared with manual labels. The error is calculated, and the classification is determined according to the error, and the intelligent seismic facies recognition is realized. Finally, the intelligent recognition of seismic facies clustering is realized, and the method of reef-bank reservoir characterization is formed by this method. The above method can provide support for the study of reservoir forming conditions and favorable area optimization in this area.