Approbation of the approach with identifying production and accumulation cycles based on dynamic level tracking in production wells using machine learning methods

UDK: 622.276.5:658.018.2

DOI: 10.24887/0028-2448-2025-10-44-48

Key words: machine learning, neural network, automatic restart, work frequency management, artificial well, dynamic level, echogram

Authors: E.I. Sagdeev (RN-BashNIPIneft LLC, RF, Ufa); I.I. Zakiryanov (RN-BashNIPIneft LLC, RF, Ufa); Sh.Kh. Ishkina (RN-BashNIPIneft LLC, RF, Ufa); R.M. Amekachev (RN-BashNIPIneft LLC, RF, Ufa); A.Ya. Davletbaev1,2 (RN-BashNIPIneft LLC, RF, Ufa; Ufa University of Science and Technology, RF, Ufa); V.P. Miroshnichenko (RN-Yuganskneftegaz LLC, RF, Nefteyugansk); G.A. Shchutsky (RN-Yuganskneftegaz LLC, RF, Nefteyugansk); A.S. Sukmanov (RN-Yuganskneftegaz LLC, RF, Nefteyugansk); A.V. Sergeychev (Rosneft Oil Company, RF, Moscow)

The paper presents an approach for the automatic determination of work-and-accumulation cycle durations in artificial production wells operating in automatic restart regimes. An algorithm based on convolutional neural network that uses field data on time-varying dynamic fluid level (pump intake pressure) in the well’s annular space was developed. The authors trained the neural network using data from operating regimes of wells in Western Siberia. The implemented algorithm demonstrated high effectiveness in determining the actual cycle durations of work and accumulation cycles. It also outperforms traditional rule-based and autocorrelation methods. During the testing of the algorithm on real field data, it identified moments of well-start and well-stop with sufficient accuracy for practical use. This enables engineers to monitor the current operating regimes of an artificial well in real time. The authors plan to use this algorithm for simulation of oil production by detecting current operating regime, adjusting work/stop durations and further optimizing the cycles. Engineers can also apply this approach to automate monitoring of operating regimes in artificial wells and to detect deviations from planned regimes in a timely manner. In addition, the proposed method scales easily and can be adapted to different conditions of production and oil field, making it applicable both to new and existing wells.

References

1. Pashali A.A., Khalfin R.S., Sil’nov D.V. et al., On the optimization of the periodic mode of well production, which is operated by submergible electric pumps in Rosneft Oil Company (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2021, no. 4, pp. 92-96, DOI: https://doi.org/10.24887/0028-2448-2021-4-92-96

2. Pityuk Yu.A., Kunafin A.F., Bayramgalin A.R. et al., Identification of unplanned shutdowns for buildup tests in real time (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2020, no. 2, pp. 32–35, DOI: https://doi.org/10.24887/0028-2448-2020-2-32-35

3. Ishkina Sh.Kh., Zakir’yanov I.I., Sagdeev E.I. et al., Approbation of the machine learning based approach to acoustic liquid level determination (In Russ.), Ekspozitsiya neft’ i gaz, 2024, no. 5, pp. 51–56, DOI: https://doi.org/10.24412/2076-6785-2024-5-51-56

4. Pashali A.A., Sil’nov D.V., Topol’nikov A.S. et al., Bringing the oil wells equipped by elictrical submersible and sucker rod pumps on to stable production based on complex approach using machine learning and digital twins (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2021, no. 7, pp. 112–117, DOI: https://doi.org/10.24887/0028-2448-2021-7-112-117.

5. Lovrić M., Milanović M., Stamenković M., Algoritmic methods for segmentation of time series: An overview, Journal of Contemporary Economic and Business Issues, 2014, V. 1, no. 1, pp. 31–53.

6. Perslev M. et al., U-time: A fully convolutional network for time series segmentation applied to sleep staging, Advances in Neural Information Processing Systems, 2019, V. 32.

7. Yuncong Yu et al., Segmentation of multivariate time series with convolutional neural networks, Proceedings of the International Conference on Calibration-Methods and Automotive Data Analytics, Berlin, Deutschland, 21.05.2019 – 22.05.2019.

8. Certificate of registration of a computer
program RU 2025661932. Programmnoe obespechenie “EchoTools” (EchoTools
software), Authors: Davletbaev A.Ya