Application of machine learning models for intelligent management of oil transportation efficiency

UDK: 681.518:622.692.4
DOI: 10.24887/0028-2448-2021-12-136-139
Key words: anti-turbulent additive, hydraulic efficiency of oil pipelines, intelligent control, machine learning, gradient boosting, neural networks
Authors: D.A. Cherentsov (Industrial University of Tyumen, RF, Tyumen), A.U. Yakupov (Industrial University of Tyumen, RF, Tyumen), K.S. Voronin (Industrial University of Tyumen, RF, Tyumen), Yu.D. Zemenkov (Industrial University of Tyumen, RF, Tyumen), E.L. Chizhevskaya (Industrial University of Tyumen, RF, Tyumen)

Effective management of a modern industrial company engaged in oil transportation should be based on scientifically grounded, most accurate assessments of the state of the production system in real time. Increasing the throughput of oil pipelines using anti-turbulent additives is by far the most economical and effective way. However, the use of modern models, dependencies and research results to determine pressure losses in oil pipelines when using anti-turbulent additives at real facilities lead to significant deviations from the actual values. To solve this problem, it is proposed to use simulation based on machine learning models. Analysis of the initial data made it possible to determine the features and target variables for training the models. The most popular models for solving the problem of multivariate regression (search for a function of n-variables) are considered as machine learning models, such as: linear regression, decision trees, random forest, gradient boosting, artificial neural networks and ensembles of models. The model was trained in Python using popular machine learning libraries: sklearn, keras, pytorch, catboost, etc. Using cross-validation, the hyperparameters for each of the considered model were determined, which provide the best quality metrics. When comparing the forecasting results with the actual data not participating in the training process, one of the models showed a satisfactory error in comparison with the rest of the models. The possibility of increasing the forecast accuracy of machine learning models for new data by retraining existing models is also considered. Simulation modeling based on machine learning models can be effectively used as a method for assessing the required amount of anti-turbulent additives to ensure the required hydraulic efficiency of oil pipelines. It is proposed to use artificial neural networks (ANN) as a recommended model for use. This is due to the fact that the type of the objective function is not known in advance, and when training the ANN, the process of finding a function that most correctly describes the target dependence takes place. The use of simulation modeling as a tool for intelligent control makes it possible to successfully evaluate the effect of an anti-turbulent additive on the hydraulic efficiency of oil pipelines, thus significantly reducing the operating costs of pumping and thereby increasing the efficiency of the enterprise.

References

1. Gareev M.M., Al'mukhametova D.A., Akhmetvalieva G.F., Substantiation of methods for predicting the efficiency of pumping oil and petroleum products using an anti-turbulent additive through pipelines of different diameters (In Russ.), Transport i khranenie nefteproduktov i uglevodorodnogo syr'ya, 2018, no. 2, pp. 10–15.

2. Golunov N.N., Influence of small drag reducing agents on hydraulic efficiency and an interface volume by the batching technology (In Russ.), Territoriya neftegaz, 2018, no. 6, pp. 92–97.

3. Golunov N.N., Lur'e M.V., The construction of the turbulence phenomenological theory in a liquid with drug reducing additives (In Russ.), Nauka i tehnologii truboprovodnogo transporta nefti i nefteproduktov = Science & Technologies: Oil and Oil Products Pipeline Transportation, 2020, V. 10, no. 2, pp. 148–156, DOI: 10.28999/2541-9595-2020-10-2-148-156.

4. Zholobov V.V., Sinel'nikov S.V., Ignatenkova A.I., The prospects of dra for reducing the energy consumption of thermal stations in a "hot" pumping (In Russ.), Nauka i tehnologii truboprovodnogo transporta nefti i nefteproduktov = Science & Technologies: Oil and Oil Products Pipeline Transportation, 2019, no. 3, pp. 256–265, DOI: 10.28999/2541-9595-2019-9-3-256-265.

5. Korshak A.A., The versatility of the generalized formulas of L.S. Leibenson (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2019, no. 5, pp. 105–108, DOI: 10.24887/0028-2448-2019-5-105-108.

6. Nechval' A.M., Muratova V.I., Chen' Yan, Analysis of various factors affecting hydraulic reduction efficiency by drag reducing additives (In Russ.), Problemy sbora, podgotovki i transporta nefti i nefteproduktov, 2019, no. 2(118), pp. 142–152, DOI: 10.17122/ntj-oil-2019-2-142-152.

7. Nikolaev A.K., Zaripova N.A., Matveeva Yu.G., Effectiveness testing of the anti-turbulent suspension additive m-flowtreat for pressure oil pipelines (In Russ.), Territoriya Neftegaz, 2019, no. 1-2, pp. 102–110.

8. Moiseenko T.A., Muratova V.I., Nechval' A.M., Farukhshina R.R., Determination of the optimal concentration of an anti-turbulence additive using differential calculus and mathematical analysis (In Russ.), Transport i khranenie nefteproduktov i uglevodorodnogo syr'ya, 2020, no. 3, pp. 35–39, DOI: 10.24411/0131-4270-2020-10307.

9. Revel'-Muroz P.A. et al., Estimation of the oil pumping technology effectiveness with drag reduction agents (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2020, no. 1, pp. 90–95, DOI: 10.24887/0028-2448-2020-1-90-95.

10. Cherentsov D.A., Zaraev V.F., Mareeva A.Yu. et al., Applying machine learning to predict the effect of an anti-turbulence additive on the hydraulic efficiency of oil pipelines (In Russ.), Territoriya Neftegaz, 2021, no. 3–4, pp. 14–22.

11. Chen' Yan, Nechval' A.M., Muratova V.I., Pen Yan, Numerical simulation method predicting the distribution of velocity in the process of reducing turbulent resistance adding drag reducing additives in the tube (In Russ.), Transport i khranenie nefteproduktov i uglevodorodnogo syr'ya, 2019, no. 2, pp. 9–13, https://doi.org/10.24411/0131-4270-2019-10202

12. Muratova V.I., Valeev A.R., Chen' Yan et al., Comparative analysis of the effectiveness of anti-turbulence additives in laboratory conditions (In Russ.), Transport i khranenie nefteproduktov i uglevodorodnogo syr'ya, 2020, no. 4, pp. 18–23, DOI: 10.24411/0131-4270-2020-10403

13. CatBoost is a high-performance open source library for gradient boosting on decision trees, URL: https://catboost.ai/

14. URL: https://habr.com/ru/company/ods/blog/322626/

15. URL: https://habr.com/ru/company/ods/blog/424781/


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

20.09.2022
19.09.2022
15.09.2022