Neural network prediction of reservoir properties of the reservoir according to seismic data on the example of clinoform deposits of Western Siberia

As part of the tasks for designing the development of promising areas in the oil and gas industry, a certain problem was identified with the inability to apply classical methods of geological modeling applying seismic-facial neural network prediction of reServoir propertieS of the reServoir according to SeiSmic data on the example of clinoform depoSitS of weStern Siberia analysis for the quantitative assessment of productive reservoir properties.

Consequently, attempts were made to test machine learning techniques and algorithms implemented in the specialized software IP_Seismic (IPLAB LLC).

Purpose. Main goal of the work was the development of projects for introducing new wells, calculating economic feasibility and profitability of financial expenditures based on a 3D geological model of the field.

Methods. The work used machine learning algorithms based on full-featured Kolmogorov functions implemented in the separate software IP_Seismic, as well as 3D geological modeling techniques in Petrel software.

Results. The outcome of the work resulted in the adjustment of the development program and the implementation of a certain number of business cases. Various typical examples of using machine-learning techniques are presented in this article, based on what additional wells for drilling, risk reduction programs, and the implementation of an alternative development system were proposed.

Conclusion. This work presents typical ways of finding promising drilling zones with neural network forecasting. The advanced mathematical basis laid down in specialized software, which was subsequently utilized in preparing the solution, is tested. Based on the developed methodology, examples of successful application of the approach for adjusting and designing operational drilling programs are presented, which can be replicated in other development projects.

1. Kobrunov A., Priezzhev I. Hybrid combination genetic algorithm and controlled gradient method to train a neural network. Geophysics. — 2016. — vol. 81. — № 4. — Pp. 1–9.

2. Kolmogorov A.N. On the representation of continuous functions of several variables by superpositions of continuous functions of a smaller number of variables // Reports of the USSR Academy of Sciences. — 1957. — № 14(5). — Pp. 953–956.

3. Priezzhev I., Shmaryan L., Bejarano G. Non-linear multi trace seismic inversion using neural network and genetic algorithm — “Genetic Inversion” // Annual Meeting St Petersburg, EAGE, Extended Abstracts. — 2008.

4. Tikhonov A.N., Arsenin V.Y. Solutions of ill-posed problems / V.H. Winston and Sons, Washington D.C., 1977.

5. Priezzhev I.I. New age, Kolmogorov full functional neural network usage for nonlinear predictive seismic inversion. / EAGE, Saint Petersburg, 2020 // Geosciences: Converting Knowledge into Resources. Saint Petersburg, Russia, 6–9 April. — 2020.

6. Priezdev I.I. Neural networks of a new generation based on Kolmogorov’s theorem and their application for predictive inversion constructions // GeoEurasia, Moscow, February 3–5. — 2020.