Application of machine learning method for planning of well drilling in producing oil formations

UDK: 622.276.1/.4.001.57
DOI: 10.24887/0028-2448-2021-7-23-27
Key words: machine learning methods, Kohonen neural networks, Bayesian neural networks, structural changes, Arps model, regression analysis models, panel data models, production enhancement operations, feasibility study, field development scenario, project well, LAZURIT proxy model
Authors: B.G. Ganiev (Tatneft PJSC, RF, Almetyevsk), А.V. Nasybullin (TatNIPIneft, RF, Bugulma), Ram.Z. Sattarov (TatNIPIneft, RF, Bugulma), V.S. Timofeev (Novosibirsk State Technical University, RF, Novosibirsk), А.V. Faddeenkov (Novosibirsk State Technical University, RF, Novosibirsk), А.Yu. Timofeeva (Novosibirsk State Technical University, RF, Novosibirsk)

The paper describes algorithms for predicting oil production rates based on Arps model whose parameters are estimated in terms of well production performance and the algorithm based on panel data models with a trend component described by Arps model. An algorithm for selection of input factors based on Bayesian neural networks has been proposed and implemented. An algorithm for construction of piecewise multiple regression models to estimate the Arps constant and to predict oil production rates based on application of Kohonen networks and structural change analysis approaches has been proposed and implemented. This method for prediction of oil production rate and production decline considers numerous factors that influence production performance. A module for prediction of oil production performance of project wells has been developed. The host application is written in Python 3.6 programming language. Computational algorithms of model building are implemented in R programming language. The authors describe the principle of operation of software module for forecasting oil production performance of project wells.

The described method has been field tested at production sites of Tatneft Company. Machine learning has been done using selected geological and production data from producing wells in Kynovian and Pashian reservoirs for a group of production areas of Romashkinskoye field. Machine learning enabled selection of project wells and estimation of production and economic performance. A comparative analysis of existing methods for prediction of input oil production rates and annual production decline rates of project wells for the selected group of areas has been performed. Resultant data suggest applicability of machine learning method for prediction of production performance of project wells in oil fields. This is particularly important for mature fields, which provide sufficient accumulated statistical data required to apply machine learning methods.

References

1.  Latifullin F.M., Sattarov Ram.Z., Sharifullina M.A., Application of lazurit workstation software package for geological and reservoir modeling and well intervention planning for Tatneft’s production assets (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2017, no. 6, pp. 40–43, DOI: 10.24887/0028-2448-2017-6-40-43.

2. Nasybullin A.V., Sattarov Ram.Z., LatifullinF.M. et al., Creation of software tool for long-term investment planning with a view to the effective development of oil fields (In Russ.), Neftjanoe hozjajstvo = Oil Industry, 2019, no. 12, pp. 128–131, DOI: 10.24887/0028-2448-2019-12-128-131.

3. Khisamov R.S., Ganiev B.G., Galimov I.F. et al., Computer-aided generation of development scenarios for mature oil field (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2020, no. 7, pp. 20–25, DOI: 10.24887/0028-2448-2020-7-22-25.

4. Lysenko V.D., Proektirovanie razrabotki neftyanykh mestorozhdeniy (Oil field development design), Moscow: Nedra Publ., 1987, 247 p.

5. Silaev K.O., Silaeva A.N., Methods for the analysis of production decline curves used in the fuel and energy sector (In Russ.), Ekonomika i sotsium, 2016, no. 9(28), pp. 543–546, URL: https://readera.org/ekonomika-socium/2016-9-28.

6. Gujarati D.N., Basic econometrics, McGraw-Hill, 1995, 1003 p.

7. Poirier D.J., The econometrics of structural change, Amsterdam: North-Holland, 1976, 206 p.

8. Kohonen T., Self-organizing maps, New York, 2001, 501 p.

9. Rousseeuw P.J., Tutorial to robust statistics, Journal of chemometrics, 1991, V. 5, no. 1, pp. 1–20.

10. Huber P.J., Robust statistics, New York: John Wiley & Sons, 1981.

11. Faddeenkov A.V., Khaylenko E.A., Estimation of regression model parameters by the trimmed maximum likelihood method (In Russ.), Nauchnyy vestnik Novosibirskogo gosudarstvennogo tekhnicheskogo universiteta, 2016, V. 65, no. 4, pp. 135–145.

12. Denisov V.I., Timofeev V.S., Sign method: advantages, problems, algorithms (In Russ.), Nauchnyy vestnik Novosibirskogo gosudarstvennogo tekhnicheskogo universiteta, 2001, no. 1 (10), pp. 21–35.

13. Timofeev V.S., Vostretsova E.A., Robust estimation of regression model parameters based on ideas of least-square method (In Russ.), Nauchnyy vestnik Novosibirskogo gosudarstvennogo tekhnicheskogo universiteta, 2007, no. 2 (27), pp. 57–67.

14. Timofeev V.S., The characteristic function in parameter estimation problem for regression model (In Russ.), Nauchnyy vestnik Novosibirskogo gosudarstvennogo tekhnicheskogo universiteta, 2010, no. 2 (39), pp. 43–52.

The paper describes algorithms for predicting oil production rates based on Arps model whose parameters are estimated in terms of well production performance and the algorithm based on panel data models with a trend component described by Arps model. An algorithm for selection of input factors based on Bayesian neural networks has been proposed and implemented. An algorithm for construction of piecewise multiple regression models to estimate the Arps constant and to predict oil production rates based on application of Kohonen networks and structural change analysis approaches has been proposed and implemented. This method for prediction of oil production rate and production decline considers numerous factors that influence production performance. A module for prediction of oil production performance of project wells has been developed. The host application is written in Python 3.6 programming language. Computational algorithms of model building are implemented in R programming language. The authors describe the principle of operation of software module for forecasting oil production performance of project wells.

The described method has been field tested at production sites of Tatneft Company. Machine learning has been done using selected geological and production data from producing wells in Kynovian and Pashian reservoirs for a group of production areas of Romashkinskoye field. Machine learning enabled selection of project wells and estimation of production and economic performance. A comparative analysis of existing methods for prediction of input oil production rates and annual production decline rates of project wells for the selected group of areas has been performed. Resultant data suggest applicability of machine learning method for prediction of production performance of project wells in oil fields. This is particularly important for mature fields, which provide sufficient accumulated statistical data required to apply machine learning methods.

References

1.  Latifullin F.M., Sattarov Ram.Z., Sharifullina M.A., Application of lazurit workstation software package for geological and reservoir modeling and well intervention planning for Tatneft’s production assets (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2017, no. 6, pp. 40–43, DOI: 10.24887/0028-2448-2017-6-40-43.

2. Nasybullin A.V., Sattarov Ram.Z., LatifullinF.M. et al., Creation of software tool for long-term investment planning with a view to the effective development of oil fields (In Russ.), Neftjanoe hozjajstvo = Oil Industry, 2019, no. 12, pp. 128–131, DOI: 10.24887/0028-2448-2019-12-128-131.

3. Khisamov R.S., Ganiev B.G., Galimov I.F. et al., Computer-aided generation of development scenarios for mature oil field (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2020, no. 7, pp. 20–25, DOI: 10.24887/0028-2448-2020-7-22-25.

4. Lysenko V.D., Proektirovanie razrabotki neftyanykh mestorozhdeniy (Oil field development design), Moscow: Nedra Publ., 1987, 247 p.

5. Silaev K.O., Silaeva A.N., Methods for the analysis of production decline curves used in the fuel and energy sector (In Russ.), Ekonomika i sotsium, 2016, no. 9(28), pp. 543–546, URL: https://readera.org/ekonomika-socium/2016-9-28.

6. Gujarati D.N., Basic econometrics, McGraw-Hill, 1995, 1003 p.

7. Poirier D.J., The econometrics of structural change, Amsterdam: North-Holland, 1976, 206 p.

8. Kohonen T., Self-organizing maps, New York, 2001, 501 p.

9. Rousseeuw P.J., Tutorial to robust statistics, Journal of chemometrics, 1991, V. 5, no. 1, pp. 1–20.

10. Huber P.J., Robust statistics, New York: John Wiley & Sons, 1981.

11. Faddeenkov A.V., Khaylenko E.A., Estimation of regression model parameters by the trimmed maximum likelihood method (In Russ.), Nauchnyy vestnik Novosibirskogo gosudarstvennogo tekhnicheskogo universiteta, 2016, V. 65, no. 4, pp. 135–145.

12. Denisov V.I., Timofeev V.S., Sign method: advantages, problems, algorithms (In Russ.), Nauchnyy vestnik Novosibirskogo gosudarstvennogo tekhnicheskogo universiteta, 2001, no. 1 (10), pp. 21–35.

13. Timofeev V.S., Vostretsova E.A., Robust estimation of regression model parameters based on ideas of least-square method (In Russ.), Nauchnyy vestnik Novosibirskogo gosudarstvennogo tekhnicheskogo universiteta, 2007, no. 2 (27), pp. 57–67.

14. Timofeev V.S., The characteristic function in parameter estimation problem for regression model (In Russ.), Nauchnyy vestnik Novosibirskogo gosudarstvennogo tekhnicheskogo universiteta, 2010, no. 2 (39), pp. 43–52.



Attention!
To buy the complete text of article (a format - PDF) or to read the material which is in open access only the authorized visitors of the website can. .

Mobile applications

Read our magazine on mobile devices

Загрузить в Google play

Press Releases

24.11.2021
23.11.2021
02.11.2021
Конкурс на соискание молодежной премии имени академика И.М. Губкина