Classification of the gas flow metering results using machine learning algorithms

The article discusses the problem of classifying the results of the technological process of gas flow metering for the presence of deviations in them. The classification was carried out using an LSTM neural network trained on labeled data, compared with other machine learning algorithms, and combined with a regression LSTM neural network. The use of this algorithm, for the purpose of predictive maintenance, early identification and prevention of the causes of certain types of deviations, is aimed at increasing the efficiency of gas transportation system management

Keywords: gas transportation network, gas balance, classification, machine learning algorithms, LSTM neural networks.

References

  1. Report «Gas distribution and gas sales on the domestic market of the Russian Federation». https://groznyrg.ru/pdf/dokladGustov.pdf.
  2. A. R. Salikov. Imbalance in gas distribution networks//Gas of Russia. 2015. № 4. P. 36-41.
  3. G. A. Khvorov, S. I. Kozlov, G. S. Akopova, A. A. Evstifeev. Reducing losses of natural gas during transportation through main gas pipelines of OJSC Gazprom//Gas industry. 2013. № 12. P. 66-69.
  4. Website of PJSC Gazprom. https://gasandmoney.ru/novosti/gazprom-budet-vychislyat-nebalans-gaza-pri-pomoshhi-big-data.
  5. Digital economy of the Russian Federation. https://digital.gov.ru/ru/activity/directions/858.
  6. S. Haykin. Neural networks. Complete course. M.: Williams, 2018. 1104 p.
  7. V. P. Dyakonov, V. V. Kruglov. MATLAB 6.5 SP1/7/7 SP1/7 SP2 + Simulink 5/6. Artificial intelligence and bioinformation tools. M.: Solon-press, 2006.
  8. A. K. Petrova. Detection of anomalies in the technological processes of the gas transmission system using networks with a long short-term memory (LSTM)//Conference paper «Information processes and systems' design and quality assurance». St. Petersburg state electrotechnical university «LETI». 2022. P. 95-98.

Authors