The paper presents an innovative approach to optimization of production well operation through combination of engineering methods, computer programming, and machine learning. The authors highlight the importance of estimating maximum allowable bottomhole pressure with account of saturation pressure, gas content, and reservoir stress state. Python-based program, integral with corporate information system of an oil company, was developed for automation of candidate well selection for bottomhole pressure optimization. Data collection and preparation involve generation and processing of spreadsheets, creation of new parameters, and data integration into a single database. Geological risks are also considered using data from updated reservoir simulation models of the company's fields. Well production potential estimation algorithm is divided into two blocks: increasing the production rates through optimization of existing downhole pumping equipment and its replacement with more efficient equipment. Models considering the dynamic fluid level, current bottomhole pressure, and operating parameters of pumping equipment help determine the optimal operating parameters of surface drive and downhole pumping equipment. Moreover, machine learning models for solution of multiple regression problem forecast changes in water cut behavior in production wells, once the production increases. An essential element of the research is creation of a web application for easy access to data and model predictions. Implementation of this interface accelerates and simplifies access to data required for analysis and decision-making, thereby significantly reducing the time and resources. Thus, a comprehensive approach that combines engineering methods, computer programming, and machine learning significantly improves the efficiency and enables automation of bottomhole pressure optimization in production wells to result in increased oil production at minimum costs.
References
1. Balandin L.N., Gribennikov O.A., Sviridova I.A., Current state of work of producing wells depending on BHP (In Russ.), Bulatovskie chteniya, 2017, V. 2, pp. 65–69.
2. Nazarova L.N., Razrabotka neftegazovykh mestorozhdenii s trudnoizvlekaemymi zapasami (Development of oil and gas fields with hard-to-recover reserves), Moscow: Publ. of Gubkin University, 2011, 156 p.
3. Mishchenko I.T., Skvazhinnaya dobycha nefti (Oil production), Moscow: Neft’ igaz Publ., 2003, 816 p.
4. Doskazieva G.Sh., Kuangaliev Z.A., Imangalieva G.E., Optimization of the wellsfields on «Dossormunaigas» (In Russ.), 2020, no. 3.1, pp. 30-34, URL: https://7universum.com/pdf/tech/3.1(72.1)/3.1(72.1).pdf
5. Nazarenko M.Yu., Zolotukhin A.B., Application of machine learning for probabilistic production forecasting and ultimately recoverable reserves estimation (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2020, no. 9, pp. 109–113, DOI: http://doi.org/10.24887/0028-2448-2020-9-109-113
6. Topol’nikov A.S., Machine learning for artificial lift (In Russ.), Neftegaz.ru, 2021, no. 5, pp. 14-19.
7. Zotkin O.V. et al., The new approach to improvement oil reservoir proxy model predictions using machine learning algorithms (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2019, no. 12, pp. 60–63, DOI: http://doi.org/10.24887/0028-2448-2019-12-60-63
8. 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
9. Chen T., Guestrin C., Xgboost: A scalable tree boosting system, ACM SIGKDD: Proceedings of the 22nd International Conference on Knowledge Discovery and Data Mining, Aug 2016, 2016, pp. 785-794, DOI: https://doi.org/10.1145/2939672.2939785.
10. Ke G. et al., LightGBM: A highly efficient gradient boosting decision tree, Advances in Neural Information Processing Systems 30 (NIPS 2017), Proceedings of 31st Conference on Neural Information Processing Systems, Long Beach, CA, USA, Dec 2017, 2017, 10 p.
11. Altmann A. et al., Permutation importance: a corrected feature importance measure, Bioinformatics, 2010,V. 26, Issue 10, pp. 1340-1347,
DOI: https://doi.org/10.1093/bioinformatics/btq134
