Studying the applicability of artificial intelligence models based on machine learning algorithms for the selection of hydraulic fracturing fluid
UDC: 622.276.66.002.34:622.234.573
DOI: -
Authors:
SHEPELEV I.A.
1,
BUDKEVICH R.L.1,
GAYFULLIN T.L.1,
ZAKIROV R.R.1,
ALENKIN I.A.1
1 Almetyevsk State Technological University "Higher school of Oil", Almetyevsk, Russia
Keywords: destructor, stapler, rheology, dynamic viscosity, artificial intelligence, rupture fluid, machine learning
Annotation:
Hydraulic fracturing is a complex and multi-stage process that requires precise planning and execution. The introduction of modern technologies and methods can significantly improve the efficiency of hydraulic fracturing and minimize its negative effects. At the same time, the process of optimizing the composition of hydraulic fracturing fluid is quite a big problem, since its properties are influenced by a large number of factors, such as the anionic and cationic composition of water, its hydrogen index, temperature and others. That’s why the specialists in the laboratory should select the formulation of hydraulic fracturing fluid for each specific case. This is a labor-intensive and not always effective method, since in conditions of limited time it is not always possible to determine the most optimal formulation, and due to the large number of factors affecting the liquid functional properties, it is impossible to develop a systematic approach. This problem made the authors of the article study algorithms based on the artificial intellect (AI) model, trained on experimental data obtained in the process of selecting the composition of hydraulic fracturing fluid. Models testing proved the fact that correct interpretation of inlet testing results, which are known in advance, is possible only for a limited set of points, whereas for the other ones the error was clearly obvious. The results of MAPE and R2 calculation showed that the predictive accuracy is currently lower than the required one. So, development of a more complex AI algorithm that can be used for training on a limited set of data is required. To solve the problems associated with a limited set of data, a technique was developed for synthesizing a mathematical model using a number of methods based on additive nonlinear regression, LASSO- and Ridge-regressions. The technique has proven to be an effective method when there appear problems with data volume and a large number of control parameters.
Bibliography:
1. Osnovy tekhnologii gidravlicheskogo razryva plastov: ucheb. posobie / V.G. Salimov, Sh.F. Takhautdinov, A.V. Nasybullin, O.V. Salimov. – Kazan’: "Fen" AN RT, 2021. – S. 3.
2. Nasybullin A.V., Salimov O.V., Sakhabutdinov O.V. Vliyanie vyazkosti tekhnologicheskoy zhidkosti na geometriyu treshchin gidrorazryva // Oborudovanie i tekhnologii dlya neftegazovogo kompleksa. – 2017. – № 4. – S. 29–34.
3. Fattakhov I.G., Garifullina Z.A., Gizatullin I.R. Povyshenie effektivnosti gidrorazryva plasta pri podache v potok zhidkogo destruktora // Problemy razrabotki neftyanykh i gazovykh mestorozhdeniy s trudnoizvlekaemymi zapasami: sb. tr. Vseross. nauch.-tekhn. konf. – Ufa: Izd-vo UGNTU, 2020. – S. 151–155.
4. Antonov G.B., Ivanov S.I. Lineynaya regressiya kak odin iz metodov statisticheskogo issledovaniya // Izv. Velikolukskoy GSKhA. Obshchestvennye nauki. – 2021. – S. 64–75.
5. Lineynaya regressiya (Linear regression): sayt. – URL: https://wiki.loginom.ru/articles/linear-regression.html
6. Kaftannikov I.L., Parasich A.V. Osobennosti primeneniya derev’ev resheniy v zadachakh klassifikatsii // Vest. YuUrGU. Ser. Komp’yuternye tekhnologii, upravlenie, radioelektronika. – 2015. – T. 15, № 3. – S. 26–32.
7. Salakhutdinova K.I., Lebedev I.S., Krivtsova I.E. Algoritm gradientnogo bustinga derev’ev resheniy v zadache identifikatsii programmnogo obespecheniya // Nauch.-tekhn. vestn. informatsionnykh tekhnologiy, mekhaniki i optiki. – 2018. – T. 18, № 6. – S. 1016–1022. – doi: 10.17586/2226-1494-2018-18-6-1016-1022
8. Mnogosloynyy perseptron i algoritm obratnogo rasprostraneniya oshibki: sayt. – URL: https://www.mql5.com/ru/articles/8908
9. Kugaevskikh A.V., Muromtsev D.I., Kirsanova O.V. Klassicheskie metody mashinnogo obucheniya. – SPb.: Universitet ITMO, 2022. – 53 s.
10. MAPE – srednyaya absolyutnaya oshibka – praktika primeneniya. – 2022. – URL: https://4analytics.ru/metodi-analiza/mape-srednyaya-absolyutnaya-oshibka-praktika-primeneniya.html
Back to issue