In Russia, AI Method Analyzes Complex Oil and Gas Formations
The new technology enables on-site identification of mineral composition without relying on laboratory testing.
Researchers at Skoltech and SPbGU (Saint Petersburg State University) have developed a machine-learning algorithm that determines the mineral composition of rock formations without costly laboratory analysis. The technology is designed for unconventional reservoirs, including the Bazhenov Formation in Western Siberia.
The approach relies on standard well-logging data and information about the thermal properties of rocks. Using these inputs, the model builds a detailed mineralogical profile of a well along its entire length. This eliminates the need for point-based and expensive methods such as core analysis or downhole spectroscopy.
The system is based on a gradient boosting algorithm embedded in a chain of regressors. The architecture accounts for relationships between minerals, improving prediction accuracy. The model calculates mass and volumetric fractions of clay, calcite, dolomite, pyrite, quartz, feldspar, mica, siderite, as well as total organic carbon content.
To validate the results, researchers applied the predicted parameters to a theoretical thermal conductivity model. The calculations showed a high probability of agreement with experimental measurements.
Digital Drilling
The study’s author, Batyrkhan Gainitdinov, a postgraduate student in Skoltech’s Neftegazovoye Delo (Oil and Gas Engineering) program, said that the key challenge in working with unconventional reservoirs such as the Bazhenov Formation is their high heterogeneity and complex mineral composition.
According to the project’s supervisor, Dmitry Koroteev, the methodology can be used for real-time data interpretation during drilling, identifying promising intervals in complex reservoirs, and optimizing enhanced oil recovery techniques. Ultimately, this can reduce the economic costs of exploration and field development.
