R.S. Khisamov, A.P. Bachkov, S.E. Voitovich, E.G. Grunis, R.A. Alekseev
Artificial Intelligence as important tool of modern geologist
Key words: petroleum geology; seismic exploration; machine learning; C++ Programming Language
For citation: Khisamov R.S., Bachkov A.P., Voitovich S.E., Grunis E.G., Alekseev R.A. Artificial Intelligence as important tool of modern geologist. Geologiya nefti i gaza. 2021;(2):37–45. DOI: 10.31087/0016-7894-2021-2-37-45. In Russ.
The paper presents the methodologies developed by the scientific and production center “Neuroseism” of TGRU department of Public Сompany “Tatneft”, based on the use of high-level programming languages to implement new approaches for interpretation of seismic exploration data using artificial intelligence technology based on neural network. Designed by TGRU department of Public Сompany “Tatneft” and protected by two patents of the Russian Federation, the neural net technology makes forecast of oil-objects based on the solution of problems by artificial intelligence, allowing to extract more information from seismic data. “Neuroseism” system is a learning multi-layer neural network. Examples for network training are reflected seismic waves recorded from reservoirs in areas of confirmed oil deposits. The configured and trained neural network is further used in the analysis of seismic profiles and 3D seismic cubes on the exploration area. Based on the results, forecast maps of the oil potential of productive deposits are constructed, on the basis of which recommendations for conducting exploration work are issued. During 2014–2018 a new modification technology “Neuroseism” have been developed, dubbed “Neuroseism-Foreground”, allowing adaptation and optimization of this technology to predict the oil distribution in Frasnian-Famennian carbonate complex. The program “Neuroseism-Foreground” performs an automated search for the best training sample of a seismic signal based on self-testing. The program is designed to identify or clarify the prospects for the oil content of domanic deposits, allows you to significantly reduce the risks when drilling exploration and production wells by allocating areas similar in production potential to the areas where the forecast’s training wells are located with industrially exploited deposits in domanic sediments.