«Automation and Informatization of the fuel and energy complex»

Specific features of data analysis when building a model for predicting the volume of the final product output in chemical production

UDC: 004.896+517.977.58
DOI: -

Authors:

¹ National University of Oil and Gas "Gubkin University", Moscow, Russia
² Digital Technologies and Platforms, Moscow, Russia

Keywords: prediction, chemical production, correlation analysis, Lasso-regression, selection of significant features, machine learning methods

Annotation:

The authors of the article study specific features of building models based on machine learning methods for predicting the yield of the final product in the industrial production of nitric acid. The proposed approach can be used to build models, based on the collected volume of real data, for various production processes. Conducting an active experiment in production conditions to identify such models is not required, which allows getting a significant economic effect. When building models based on machine learning algorithms, specific features of the equipment configuration at a specific enterprise are taken into account, and the system response time to control actions is determined. The authors also propose a procedure of determining the time interval between a change of input parameters and the corresponding response of the target variable. The procedure is based on the use of Lasso-regression, its result is confirmed by an analysis of the importance of features when building a model using the Random Forest algorithm.

Bibliography:

1. Il'in A.P., Kunin A.V., Il'in A.A. Proizvodstvo azotnoy kisloty: ucheb. posobie. – Ivanovo: Ivanovskiy gos. khim.-tekhnolog. un-t, 2011. – 269 s.
2. Nurutdinov N.N., Gershkovich Yu.B. Metodika identifikatsii dinamicheskikh modeley s peremennym transportnym zapazdyvaniem // Avtomatizatsiya i informatizatsiya TEK. – 2023. – № 6(599). – S. 22–29. – DOI: 10.33285/2782-604X-2023-6(599)-22-29
3. Rabotnikov M.A., Shumikhin A.G. Algoritmy uprezhdayushchego upravleniya s primeneniem adaptivnoy dinamicheskoy modeli protsessa // Khimiya. Ekologiya. Urbanistika. – 2023. – T. 3. – S. 281–284.
4. Amirkhanov I.R., Kaleeva E.S., Fatun D.A. Sistema optimizatsii v real'nom vremeni: metody i realizatsiya // Avtomatizatsiya i informatizatsiya TEK. – 2023. – № 2(595). – S. 13–19. – DOI: 10.33285/2782-604X-2023-2(595)-13-19
5. Hastie T., Tibshirani R., Friedman J. The Elements of Statistical Learning. Data Mining, Inference, and Prediction. – 2nd Edition. – New York, NY: Springer, 2009. – 767 p. – DOI: 10.1007/978-0-387-84858-7
6. Ladha L., Deepa T. Feature Selection Methods and Algorithms // Int. J. on Computer Science and Engineering. – 2011. – Vol. 3, No. 5. – P. 1787–1797.
7. Sorzano C.O.S., Vargas J., Montano A.P. A survey of dimensionality reduction techniques. – 2014. – DOI: 10.48550/arXiv.1403.2877
8. Tibshirani R. Regression shrinkage and selection via the LASSO // J. of the Royal Statistical Society. Series B (Methodological). – 1996. – Vol. 58, Issue 1. – P. 267–288. – DOI: 10.1111/j.2517-6161.1996.tb02080.x
9. Kiseleva V.A., Kuzin A.A., Shul'pina P.D. Otbor priznakov v mashinnom obuchenii // Telekommunikatsii i inform. tekhnologii. – 2023. – T. 10, № 2. – S. 21–29.
10. Metodika vybora vkhodnykh priznakov dlya algoritmov mashinnogo obucheniya / V.V. Tagunov, K.Ya. Kudryavtsev, A.I. Petrova, T.I. Voznenko // Vestn. Nats. issled. yadernogo un-ta "MIFI". – 2022. – T. 11, № 1. – S. 51–58. – DOI: 10.56304/S2304487X22010114
11. Breiman L. Random forests // Machine learning. – 2001. – Vol. 45, Issue 1. – P. 5–32. – DOI: 10.1023/A:1010933404324
12. Kochueva O., Akhmetzianov R. Surrogate Models for the Compressibility Factor of Natural Gas // Distributed Computer and Communication Networks: Control, Computation, Communications: 26th Int. Conf., DCCN 2023, Moscow, Sept. 25–29, 2023. – Cham: Springer, 2024. – P. 516–526. – DOI: 10.1007/978-3-031-50482-2_40