ОБРАБОТКА ИНФОРМАЦИИ И АНАЛИЗ ДАННЫХ

Comparative analysis of statistical models for the task of natural gas composition analysis

I. A. Brokarev^a, S. V. Vaskovskii<sup>b

a</sup> National University of Oil and Gas «Gubkin University», Moscow, 119991, Leninskiy prospect, 65, bl.1
^b V. A. Trapeznikov Institute of Control Sciences of Russian Academy of Sciences, Moscow, 117997, Profsoyuznaya street, 65

Аннотация: A large number of statistical methods are being developed to solve the problem of natural gas composition analysis. Statistical models are used in these methods for determination of natural gas composition by its known physical parameters. The choice of a statistical model for the method under discussion is a difficult task. No general algorithm has been found for selecting a model for a specific task. Basic statistical models, that are often used in practice, are studied in the article. The comparative analysis of the models is carried out according to a number of important criteria for solving the discussed problem. As a result, it is concluded that the neural network model is the most effective model for the natural gas composition analysis. Recommendations are given on choosing a statistical model in the tasks of natural gas quality analysis that are similar to the problem under consideration.

Ключевые слова: machine learning, statistical models, neural network analysis, composition analysis, natural gas.

Язык публикации: английский

DOI: 10.14357/20718632200104