Prediction of liquid accumulation in field gas pipelines based on machine learning

About the authors:

Pavel A. Krylov, Master’s student, basic department of LLC “TNNC”, University of Tyumen, Tyumen, Russia
paul.kryloff@yandex.ru

Nail G. Musakaev, Dr. Sci. (Phys.-Math.), Professor, Professor of the Department of Applied and Technical Physics, School of Natural Science, University of Tyumen, Tyumen, Russia; Chief Researcher, Tyumen Branch of the Khristianovich Institute of Theoretical and Applied Mechanics of the Siberian Branch of the Russian Academy of Sciences, Tyumen, Russia; musakaev68@yandex.ru, https://orcid.org/0000-0002-8589-9793

Abstract:

Liquid accumulation in field gas pipelines is a common problem that disrupts flow stability. In the process of operation, under the influence of terrain and working conditions, there is a gradual deposition of water and gas condensate from upstream to downstream sections of the pipeline, which in turn reduces the efficiency of transportation, causes an increase in pressure losses, pressure pulsations, promotes corrosion and hydrate formation processes.

Due to the complexity of multiphase flow, the mechanism of fluid accumulation is still controversial. Currently, most of the techniques that predict accumulation are semi-empirical and do not have sufficient accuracy. Advances in machine learning and artificial intelligence technologies provide a wide range of possibilities for analyzing, identifying potential dependencies and predicting data behavior.

The aim of this paper is to obtain a multifactor prediction model for liquid accumulation in field gas pipelines with high generalization and prediction ability.

Based on the statistical data of pipeline operation in oil and gas condensate fields of Western and Eastern Siberia, a data set was created to calculate the parameters in the dynamic simulator of unsteady multiphase flows required for machine learning. After data preprocessing, the model was trained using teacher learning algorithms and further compared, including methods: logistic regression (LR), linear discriminant analysis (LDA), K-nearest neighbors (KNN), decision tree (CART), naive Bayesian classifier (NB), linear support vectors machines (LSVC), support vectors machines (SVC), bagging (BG), random forest (RF), extreme randomized trees classifier (ET), adaptive boosting (AB), gradient boosting (GB), extreme gradient boosting (XGB), and multilayer perceptron (MLP), the most optimal of which were found to be the «decision tree» and «K-nearest neighbors» algorithms. These models were optimized using «cross-validation» methods, then trained on training data and tested.

Many different combinations of non-accumulation pipeline operation were established, and the degree of importance of various parameters on the accumulation process was determined. The developed model can be a useful tool for analyzing and localizing the fluid accumulation process, providing a more simplified and comprehensive prediction than other models.

References:

Aliev, Z.S., Andreev, S.A., Vlasenko, A.P., Korotayev, Yu.P. (1978). Technological mode of operation of gas wells. Nedra. Moscow. [In Russian]

Akhikyan, A.I., Danilyuk, S.S. (2024). Adaptive random forest and its application. Bulletin of Science, 1 (6). 1393–1402. [In Russian]

Borisevich, Yu.P., Golovanova, Yu.V., Krasnova, G.Z. (2022). Determination of condensate-hazardous places on gas pipelines, Natural and Technical Sciences, (10), 197–200. [In Russian]

Buznikov, N.A., Istomin, V.A., Mitnitsky, R.A. (2016). Influence of the accumulated liquid in the field pipeline on the movement of hydrate formation inhibitor. Vesti gazovoy nauki, (26), 112–116. [In Russian]

Vatuzov, S.M., Vanchugov, I.M., Rezanov, K.S., Avtomonov, P.Yu., Tanasenko, M.S., Shestakov, R.A. (2023).To the question of corrosion of field pipelines. Problems of gathering, preparation and transportation of oil and petroleum products, (5), 106–122. [In Russian]

Gefan, G.D., Ivanov, V.B. (2012). Method of support vectors and an alternative simple linear classifier. Information technologies and problems of mathematical modeling of complex systems, (10), 84–94. [In Russian]

Zhangirov, T.R., Perkov, A.S., Ivanova, S.A., Liss, A.A., Grigorieva, N.Y., Chistyakova, L.V. (2019). Comparison of the efficiency of solving the classification problem by methods of linear discriminant analysis and artificial neural networks. Izvestiya SPbGETU LETI, (5). 64–73. [In Russian]

Ilyichev, V.Yu., Zhukova, Yu.M., Shamov, I.V. (2021).Using the technology of gradient bousting to create approximation models. Notes of the scientist, (12-1), 62–67. [In Russian]

Klapchuk, O.V., Yelin, N.N. (1979).True concentrations of liquid and gas in gas pipelines of the field - gas processing plant system. Gas Industry, (3), 18–28. [In Russian]

Korgun, D.A. (2019). Ensemble methods of classification in machine learning. In Mathematical methods of control: Collection of scientific papers (pp. 32–39). Tver State University [In Russian]

Krasnov, A.N. (2018). Features of gas-collection system operation under liquid accumulation. Petroleum Engineering, 16 (4), 118–126. [In Russian]

Kutateladze, S.S., Nakoryakov, V.E. (1984). Heat and mass transfer and waves in gas-liquid systems. Nauka. Novosibirsk. [In Russian]

Lurie, M.V. (2011). Gas-liquid slug flow in horizontal pipeline. Oil Industry, (7), 122–124. [In Russian]

Mitrofanova, A.S., Komlev, G.V. (2019). Learning perceptron. Trends in the development of science and education. (49-12), 69–71. [In Russian]

Nurullaev, Z.V., Ochilov, A.U.U. (2017). Study of pressure pulsation in field pipelines. Nauchny Aspect, (4-1), 146–148. [In Russian]

Odisharia, G.E., Tochigin, A.A. (1998). Applied hydrodynamics of gas-liquid mixtures. Gazprom VNIIGAZ. [In Russian]

Pylov, P.A., Maytak, R.V., Dyagileva, A.V., Salycheva, A.D. (2024). Methods for restoring nonparametric regression under conditions of unbalanced data. Infra-Engineering. Vologda. [In Russian]

Rodionov, A.V., Ischenko, K.L. (2024). Investigation of the influence of the k-nearest neighbors algorithm parameters on the model quality metrics. System Analysis and Mathematical Modeling, 6 (2), 251–262. [In Russian]

Saburov, V.S. (2024). Bayesian classifier in machine learning. Step in Science, (1). 78–81. [In Russian]

Usachev, P. (2018).Solving the classification problem on the basis of decision tree algorithms. In Modern technologies in science and education - STNO-2018: Proceedings of the International Scientific and Technical Forum (vol.3, pp. 73–76). Ryazan State Radio Engineering University. [In Russian]

Tskhovrebov, A.S. (2024). Application of machine learning in oil and gas industry. In Actual issues of oil and gas reservoir systems research: Abstracts of the V International Scientific and Practical Conference (p.40). Gazprom VNIIGAZ LLC. [In Russian]

Rastogi, A., Fan, Y. (2020). Experimental and modeling study of onset of liquid accumulation. Journal of natural gas science and engineering, 73, 103064.