A new method of decline curve forecasting for project wells on the base of machine learning algorithms
https://doi.org/10.7868/S2587739920040102
Abstract
The article describes new decline curves (DC) forecasting method for project wells. The method is based on the integration of manual grouping of DC and machine learning (ML) algorithms appliance. ML allows finding hidden connections between features and the output. Article includes the decline curves analysis of two well completion types: horizontal and slanted wells, which illustrates that horizontal wells are more effective than slanted.
About the Authors
S. I. Gabitova
Gazpromneft NTC LLC
Russian Federation
Russian Federation
Saint-Petersburg
S. A. Davletbakova
Gazpromneft NTC LLC
Russian Federation
Russian Federation
Saint-Petersburg
V. Yu. Klimov
Gazpromneft NTC LLC
Russian Federation
Russian Federation
Saint-Petersburg
S. V. Shuvaev
Gazpromneft NTC LLC
Russian Federation
Russian Federation
Saint-Petersburg
I. Ya. Edelman
Salym Petroleum Development N.V.
Russian Federation
Russian Federation
Moscow
S. Shmidt
Salym Petroleum Development N.V.
Russian Federation
Russian Federation
Moscow
References
1. Arps J.J. Analysis of Decline Curves. Transactions of the AIME. 1945, vol. 160, no. 1, pp. 228–247. doi: 10.2118/945228-G
2. Levshin A.G. et al. Planirovanie I organizatsia experimenta [Planning and organizing of experiment]. Moscow, RGAU-MSKhA, 2015. 65 p.
3. Geron A. Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow. O’Reilly Media Inc., 2018. 718 p.
4. Shalev-Shwartz Sh., Ben-David Sh. Understanding Machine Learning Algorithms: From Theory to Algorithms. Cambridge, Cambridge University Press, 2014. 449 p
Review
For citations:
Gabitova S.I., Davletbakova S.A., Klimov V.Yu., Shuvaev S.V., Edelman I.Ya., Shmidt S. A new method of decline curve forecasting for project wells on the base of machine learning algorithms. PROneft. Professionally about Oil. 2020;(4):69-74. (In Russ.) https://doi.org/10.7868/S2587739920040102
Views:
338
Email this article 