Prediction of potential oil production rate of wells in extra-viscous oil fields using machine learning methods

UDK: 004.896:622.276.346

DOI: 10.24887/0028-2448-2025-7-61-66

Key words: machine learning, prediction of production rate, extra-viscous oil, CatBoost, reservoir simulation modeling, clustering, potential oil production rate of wells, steam-assisted gravity drainage (SAGD), cyclic steam stimulations

Authors: R.М. Amerkhanov (Almetyevsk State Technological University «Petroleum Higher School», RF, Almetyevsk; TatNIPIneft, RF, Almetyevsk); М.I. Amerkhanov (Tatneft-Dobycha JV, RF, Almetyevsk); А.А. Diyakonov (Almetyevsk State Technological University «Petroleum Higher School», RF, Almetyevsk)

The paper considers the development and application of machine learning methods for prediction of potential oil production rate of wells in extra-viscous oil (EVO) fields. This study is essential due to uncertainty associated with traditional prediction methods. The authors propose an innovative approach based on machine learning for automation and improved accuracy of the prediction process through analysis of historical and simulated data. The study is based on 567 wells data, including reservoir properties, production performance and parameters of geological and reservoir simulation models. Data preprocessing stage involved outlier removal, imputation of missing values and wells clustering using k-means algorithm. CatBoostRegressor algorithm for oil production rates prediction, with coefficient of determination R² = 0,785, resulted in the best output. Further analysis of feature importance and SHAP analysis confirmed physical validity of the model with identification of key factors (net pay thickness, oil saturated rock volume). The practical value of the research includes creation of a web interface enabling a convenient application of the model by reservoir engineers and geologists. This approach ensures real-time prediction of potential oil production rates to optimize EVO fields development strategies. Future researches entail integration of the model with reservoir simulations and extension of the training dataset to account for field production performance. Research findings are indicative of substantial increase in prediction accuracy (up to 33 %) compared to traditional methods, which confirms the efficiency of machine learning methods for prediction of potential oil production rate of wells in EVO fields of the Republic of Tatarstan.

References

1. Zakharov YA.V. et al., Opredeleniye optimal′nogo rezhima osvoyeniya parnykh gorizontal′nykh skvazhin kak odnogo iz vazhnykh etapov realizatsii tekhnologii parogravitatsionnogo drenirovaniya (Determination of the optimal mode of development of paired horizontal wells as one of the important stages of implementation of steam-assisted gravity drainage technology), Collected papers “Osobennosti razvedki i razrabotki mestorozhdeniy netraditsionnykh uglevodorodov” (Features of exploration and development of non-traditional hydrocarbon deposits), Proceedings of International scientific and practical conference, Kazan, 2–3 September 2015, Kazan: Ikhlas Publ., 2015, pp. 157–160.

2. Butler R.M., Steam-assisted gravity drainage: Concept, development, performance and future, Journal of Canadian Petroleum Technology, 1994, V. 33, no. 2,

pp. 44–50, DOI: https://doi.org/10.2118/94-02-05

3. Biglov R.R., Zaripov A.T., Shaykhutdinov D.K., Novyy podkhod k postroyeniyu geologicheskikh modeley zalezhey sverkhvyazkoy nefti (SVN) dlya usloviy sheshminskogo gorizonta Respubliki Tatarstan (A new approach to constructing geological models of extra-viscous oil (EVO) deposits for the conditions of the Sheshminsky horizon of the Republic of Tatarstan), Collected papers “Innovatsii v razvedke i razrabotke neftyanykh i gazovykh mestorozhdeniy” (Innovations in oil and gas exploration and development), Proceedings of International scientific and practical conference dedicated to the 100th anniversary of V.D. Shashin’s birth, 7–8 September 2016, Kazan, Part 1, Kazan′: Ikhlas Publ., 2016, pp. 258–260.

4. Butler R.M., Gravity drainage to horizontal wells, Journal of Canadian Petroleum Technology, 1992, V. 31, no. 4, pp. 31–37, DOI: https://doi.org/10.2118/92-04-02

5. Shaykhutdinov D.K., Sovershenstvovanie sistemy razrabotki zalezhey sverkhvyazkoy nefti Respubliki Tatarstan v usloviyakh vysokoy neodnorodnosti neftenasyshchennogo plasta (Improving the system for the development of super-viscous oil deposits in the Republic of Tatarstan in conditions of high heterogeneity of the oil-saturated formation): thesis of candidate of technical science, Bugul’ma, 2018.

6. Amerkhanov R.M., Gilyazov A.Kh., D’yakonov A.A. et al., Optimization of production well operation through combination of engineering approach, computer programming and machine learning methods (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2024, no. 8, pp. 94–99, DOI: http://doi.org/10.24887/0028-2448-2024-8-94-99

7. Prokhorenkova L. et al., CatBoost: unbiased boosting with categorical features, NeurIPS 2018: 32nd Conference on Neural Information Processing Systems, Dec 2-8, 2018, Montréal, Canada 2018, DOI: https://doi.org/10.48550/arXiv.1706.09516