D.N. Zolnikov
DOI: https://doi.org/10.25689/NP.2024.4.217-232
Abstract
The article considers an approach to increasing the success of oil reservoir transfer and integration activities through the use of machine learning and neural networks. Currently, the success rate of activities on the object of research leaves 50-60 %. During the geological and commercial analysis, 65 parameters were identified that affect the effectiveness of the activities. The training sample included 880 actual activities carried out for the translation and introduction of layers. The neural network model turned out to be the most accurate tool for predicting the success of events – the accuracy of forecasting on the validation sample was more than 80 %. The model was tested on a separate sample, including 50 activities for the transfer and incorporation of layers in 2023. According to the forecast results, 41 out of 50 forecasts turned out to be correct, which is 82 %. An increase in additional oil production from the wells of the test set is possible by 33,9 %. This approach will increase the accuracy of successful forecasts for events by 20-30 % compared to the existing approach, which helps to reduce the number of unsuccessful events and increase additional oil production.
References
2. Timonov A.V. A systematic approach to the selection of geological and technical measures to regulate the development of oil fields: dis. ...candidate of Technical Sciences: 25.00.17. – Ufa., 2010. – 151 p. (in Russian)
3. Koltun A.A. Efficiency assessment and optimal planning of geological and technical measures at oil fields of dis. ...candidate of Technical Sciences: 05.13.01. – Moscow, 2005. – 112 p. (in Russian)
4. Grigorieva M.P. Geological justification of the choice of wells for transfer to unexplored formations on the example of the Fainskoye field // Scientific and Technical Bulletin of JSC NK ROSNEFT. 2012. No. 1. pp. 7-9. (in Russian)
5. Limanovskaya O.V. Fundamentals of machine learning: a textbook. Ministry of Science and Higher Education of the Russian Federation – Yekaterinburg: Ural University Press, 2020. – 88 p. (in Russian)
6. Scholle F. Deep learning in Python. 2nd intern. edition: textbook. St. Petersburg: St. Petersburg, 2023. – 576 p. (in Russian)
Authors
D.N. Zolnikov, leading specialist URRM Samotlorneftegaz, Tyumen Oil Research Center LLC, Industrial University of Tyumen
42, Maxim Gorky Str., Tyumen, 625048, Russian Federation
E-mail: DN_Zolnikov2@tnnc.rosneft.ru
For citation:
D.N. Zolnikov Perspektivy primeneniya mashinnogo obucheniya i neyronnykh setey dlya prognozirovaniya uspeshnosti meropriyatiy po perevodu i priobshcheniyu neftyanykh plastov [Prospects for the use of machine learning and neural networks to predict the success of oil reservoir transfer and integration activities]. Neftyanaya Provintsiya, No. 4(40), 2024. pp. 217-232. DOI https://doi.org/10.25689/NP.2024.4.217-232. EDN TLKBPJ (in Russian)