Start-up for energy companies and data centers based on a neural network developed by INRTU graduate Maksim Pugachev

Maxim Pugachev, a graduate of INRTU, has developed a methodology for forecasting power consumption of power systems based on machine learning. The project is realized within the ‘Startup as a diploma’ program. Energy companies and data centers can become the potential customers.

Yevgeny Boloev, Associate Professor of the Department of Power Plants, Networks and Systems, is the head of the graduate’s qualification project. Nikolay Konovalov, the head of the Department of Physics and the head of ‘Carbon’ Student Scientific Society was a scientific advisor of the project.

According to Maxim Pugachev, electric load forecasting is a relevant direction of research in the electric power industry:

“This is necessary for optimal operation and balance in energy systems. Increasing the accuracy of forecasts will save energy resources and increase the profit of energy enterprises and data centers. The Irkutsk Regional Dispatch Office is responsible for sustainable energy supply in our region, where classical forecasting methods are actively used. We propose to improve the existing solutions with artificial intelligence”.

During the preparation of the thesis project Maxim Pugachev analyzed the climatic features of the region. He studied how temperature changes and solar activity affect the power supply. Also, the student considered the Bitcoin rate influence the demand of electricity consumption in the Irkutsk region.

“Cryptocurrency mining activity is high in the region. This is due to the favorable electricity pricing. Therefore, it is important to take Bitcoin rate into account in order to make forecasts more accurate,” - added Maksim Pugachev.

The software product developed by the student is able to give hourly forecasts taking into account the set parameters, to identify peak hours, when electricity cost reaches its maximum. The latter will be particularly relevant for data centers that spend a lot of money on electricity.

“For the time being, our neural network is the best at identifying peak hours, accuracy over 90 percent. We intend to upgrade the software, to provide it with a user-friendly interface. Users will be able to enter their own parameters to increase accuracy, obtain ready forecasts,” - Maxim Pugachev shared his plans for the development of the startup.

The State Examination Commission noted the relevance and quality of the thesis project, grading it as excellent.

Reference:

The ‘Startup as a diploma’ program has been implemented in INRTU since 2020. Graduates with business ideas can create a unique product or technology, receive recommendations from experienced entrepreneurs and find investors. Successful defense of a start-up gives three extra points for INRTU master’s admission. In 2023, 25 INRTU students studying engineering and humanitarian professions took part in the ‘Startup as a Diploma’ program.

Photos by Arseny Chekmariov