Scientific work
Project Description
Forecasting in the field of power engineering and electric power industry, is an integral part of
power system management processes. At the same time, stable and accurate forecast values of power
consumption are very important both for an individual enterprise and for the power industry due to the
increase of network capacity, as well as maintaining the balance of supply and demand in the power
market. At the same time, there is a problem of capacity reservation for regional power supply to
consumers.
To solve this problem, we propose an intelligent decision support system for advance (early)
prediction of power consumption based on neural network analysis of retrospective data for the last 4
years: electric meter code, date, power consumption, atmospheric pressure, wind speed, precipitation,
air temperature, holidays and weekends.
Thus, the more accurate the forecast, the smaller percentage of energy capacity reserve is
needed. Consequently, there is less financial cost for their reservation, which is especially useful in the
current economic realities.
Objective of the project
The goal of the project is to calculate the predicted values of electricity consumption for a
certain number of days ahead. For mathematical calculations of the predicted values we developed our
own machine learning library using the Java programming language. The main tool is a recurrent
neural network developed from scratch, and for its training we used a modified method of back
propagation of error, the main difference from the "classical" one is the addition of a weight coefficient
of matrix sensitivity, a function of gradient dependence on time and its own deviation metric. This
makes it possible to achieve an average prediction accuracy in the range of 90-100%.
At the same time, the main difficulty in advance prediction is the sharp increase in power
consumption over a short period of time (usually 1 day), caused by various reasons. Our neural
network model takes this into account and allows to give more accurate results.
Project Scientific Team
Palchevsky Evgeny Vladimirovich – Project Manager, Senior Lecturer, Department of Data Analysis
and Machine Learning, Financial University under the Government of the Russian Federation.
Antonov Vyacheslav Viktorovich – scientific supervisor of the project, Doctor of Technical Sciences,
Professor, Head of the Department of Automated Control Systems of the Ufa University
of Science and Technology.
Rodionova Lyudmila Evgenievna – PhD, Associate Professor, Department of Automated Control Systems,
Ufa University of Science and Technology.
Kromina Lyudmila Aleksandrovna – PhD, Associate Professor, Department of Automated Control Systems,
Ufa University of Science and Technology.
Fakhrullina Almira Raisovna – PhD, Associate Professor, Department of Automated Control Systems,
Ufa University of Science and Technology.
Breikin Tim – PhD and Senior Lecturer in Mathematics and Engineering at
Sheffield Hallam University, UK.
Publications on the subject of the project
1. Palchevsky, E.V. A system based on an artificial neural network of the second generation for
decision support in especially significant situations / E.V. Palchevsky, V.V. Antonov, R.R.
Enikeev, T. Breikin // Journal of Hydrology. Vol. 616. – Elsevier, 2023. – pp. 128844. (Scopus: Q1, WoS: Q1)
Link to publication
2. Palchevsky E.V. Visualization of flood zones based on time series forecasting
and GIS technologies / E.V. Palchevsky, V.V. Antonov, L.E. Rodionova,
L.A. Kromina, A.R. Fakhrullina // Computer Optics. 2023. (Scopus: Q2, WoS: Q3)
The link to the publication will appear after the publication of the article.
3. Palchevsky, E.V. Threats complex distributed systems parrying based on their development
prognostication / E.V. Palchevsky, O.I. Khristodulo, S.V. Pavlov // Advances in Social Science,
Education and Humanities Research. Vol. 483. – Atlantis Press, 2020. – С. 191-194. (Web of Science)
Link to publication
4. Palchevsky, E.V. Threat prediction in complex distributed systems using artificial neural
network technology / E.V. Palchevsky, O.I. Khristodulo, S.V. Pavlov // CEUR Workshop Proceedings.
Vol. 2763. - CEUR Workshop, 2020. - C. 289-284. (Scopus)
Link to publication
5. Palchevsky, E.V. Intelligent data analysis for forecasting threats in complex distributed
systems / E.V. Palchevsky, O.I. Khristodulo, S.V. Pavlov, A.M. Kalimgulov // CEUR Workshop
Proceedings. Vol. 2744. - CEUR Workshop, 2020. - C. 285-296. (Scopus)
Link to publication
6. Palchevsky, E.V. Decision support system based on application of the second generation
neural network / E.V. Palchevsky, V.V. Antonov // Programmnaya Ingeneria. Vol. 13. No. 6.
- New technologies, 2022. - С.301-308. (RSCI)
Link to publication
7. Palchevsky, E.V. Forecasting based on an artificial neural network of the second generation
for decision support in especially significant situations / E.V. Palchevsky, V.V. Antonov,
R.R. Enikeev // Software & Systems. Volume 35. Issue 3. – Publisher: ZAO NII "Tsentrprogramsistem",
Tver, 2022. – pp. 488-503. (RSCI)
Link to publication