Comparative analyzing the technology of predicting reservoir properties according to seismic data based on linear and nonlinear prediction algorithms

Abstract

The article deals with the results of dynamic interpretation of 3D CDPM seismic data in combination with logging data. The authors were faced with the task of comparing the results of pre-stack seismic inversion and neural learning technology in the conditions of the Arykty gas condensate field. The reason for this was the low seismic knowledge of the studied area, the fairly good predicted prospects for oil and gas content in the region, and the need to study the criteria for selecting a particular dynamic interpretation procedure. In this regard, an attempt was made to consider the structure and oil and gas content of the Arykty gas condensate field from the point of view of a comparative analysis between pre-stack inversion, which is currently actively used in the geological exploration industry within the framework of a standard interpretation graph, and neural machine learning. Justification for the relevance of the studies performed is availability of new wells and updated logging data, the ability to update and to compare the results of synchronous pre-stack inversion, and the ability to test neural learning algorithms. The conducted studies make it possible to identify criteria for the preferential use of machine learning in the conditions of the Arykty field, to take a fresh look at the features of the internal structure of the rocks that make up the productive part of the section, and to demonstrate physically the advantages of machine learning results in comparison with pre-stack seismic inversion.