The article considers the development of an adequate model for predicting electric submersible pump (ESP) failures. The input data for the modeling is dynamic parameters of ESP assembled in 76 production wells located in one oil field. The obtained dynamics were corrected by eliminating anomalous observations and filling the gaps with subsequent aggregation into dynamic series with equal observation periods. Then dynamic series were formed with a time step equal to a day. Chronological average and standard deviations average were created for each aggregated indicator and for each value of the series. These deviations were calculated as a difference of chronological average for the previous day and for 7 days. Additional difference was determined between average for current day and average for 7, 14 and 30 days, as well as a standard deviation for the same period. The transformed telemetry data were binarized by dividing them by variables above or below the critical cutoff threshold associated with the risk of ESP shutdown and determined by ROC analysis. The generalized multivariate Gsslasso Cox multivariate model (Bayesian hierarchical Cox model) was formed based on the pre-selected statistically significant risk predictors. The obtained predictions were compared with the actual number of unscheduled ESP stops. The main risk factors predictors are the high gas factor (GOR) and the factor of prolonged pump operation in the up-thrust region (high flow rate, low head).
References
1. Tanzharikov P.A., Nurman A.D., Sultan E.S., Methods of control and parameters of reliability management of technical systems in the oil and gas industry (In Russ.), Voprosy nauki, 2023, no. 2, pp. 136–142.
2. Xing Zhang, Ranran Wei. Zhicai Wu et al., Risk assessment and reliability analysis of oil pump unit based on D-S evidence theory, Energies, 2023, V. 16,
DOI: http://doi.org/10.3390/en16134887
3. Khabibullin R.A., Shabonas A.R., Gurbatov N.S., Timonov A.V., Prediction of ESPS failure using ML at Western Siberia oilfields with large number of wells,
SPE-201881-MS, 2020, DOI: https://doi.org/10.2118/201881-MS
4. Shabonas A.R., ESP operation mode optimization to increase run to failure time (In Russ.), Neftepromyslovoe delo, 2021, no. 8(632), pp. 30–36,
DOI: http://doi.org/10.33285/0207-2351-2021-8(632)-30-36
5. Serebryannikov A.A., Development of predictive model to reduce the emergency of the electric centrifugal pump unit (In Russ.), Innovatsionnye tekhnologii: teoriya, instrumenty, praktika, 2020, V. 1, pp. 264–268.
6. Skobelkin A.S., Yamaliev V.U., Application of artificial neural networks to predict the state of an electric centrifugal pump installation (In Russ.), Vestnik nauki, 2019, V. 2, no. 6(15), pp. 411–414.
7. Jiarui Chen, Wei Li, Peihao Yang et al., Prediction and classification of faults in electric submersible pumps, AIP Advances, 2022, no. 12,
DOI: http://dx.doi.org/10.1063/5.0065792
8. Chernikov V.S., ESP reliability analysis and failure pre diction technique (In Russ.), Territoriya Neftegaz, 2011, no. 5, pp. 36–39.
9. Al-Ballam Sh., Karami H., Devegowda D., A data-based reliability analysis of ESP failures in oil production wells, Journal of Energy and Power Technology, 2022,
V. 4(4), р. 36, DOI: http://doi.org/10.21926/jept.2204036
10. R Fattakhov I.G., Kadyrov R.R., Ziyatdinov A.M. et al., Testing electric centrifugal pump operation in the "right" zone (In Russ.), Sovremennye problemy nauki i obrazovaniya, 2015, no. 1–2, рр. 75-79.
11. Rukin M.V., Molchanova V.A., Urazakov K.R., Method for determining the mean time between failures of ESP units (In Russ.), Izvestiya Tomskogo politekhnicheskogo universiteta. Inzhiniring georesursov, 2022, V. 333, no. 12, pp. 219-229, DOI: http://doi.org/10.18799/24131830/2022/12/3792
