“Gazprom Neft” has experienced a new technology based on machine learning to study mountain portacatena Alexandrov | 02.11.2020
“Gazprom Neft” has successfully tested a new technology based on machine learning and digital view for the study of rocks in the fields of Western Siberia.
The project has proved that it is possible in 7 times to speed up the laboratory analysis of the full-sized core for a decision on continued program of his research.
Core sample — a rock sample of cylindrical shape, is extracted from the borehole during drilling; the only direct source of information on mineral resources, which can be removed
to the surface. On his research based key production decisions in the development of oil and gas fields. Still layered core description
was prepared solely by the expert geologist. This process was very difficult to automate: characteristics of rocks is determined by many are not always obvious signs, many
the fragments of layers is extremely diverse, and often an expert decision when their analysis has been based on subjective experience.
The cores recovered from wells and “digitized” using a special photography. The accumulated array of images is analyzed by an automatic system that recognises the photo
lithological layers, determine their name, integrity, and other characteristics.
To teach the mathematical model the center of the development and monetization of data “Gazprom oil” was developed classifier lithological layers — the principles by which one or the other
element of the cross section of the core needs to be correlated with certain characteristics of the geological formation. A dataset of more than 17 thousand photographs of Kern in the day
and ultraviolet light was prepared in partnership with the geological faculty of Moscow state University. M. V. Lomonosov.
In combination with other developments “Gazprom oil” the new system, according to preliminary estimates, will allow to save annually on laboratory studies of about 85 million rubles.
the oil and gas, Machine learning , computer vision, Digitization