Analysis of the Natural Inflow Short-term Forecasting Models Efficiency to the HPP Site for the Day-ahead Based on Regression Methods and Machine Learning Methods
доклад, тезисы доклада, статья из сборника материалов конференций
Библиографическое описание:Analysis of the Natural Inflow Short-term Forecasting Models Efficiency to the HPP Site for the Day-ahead Based on Regression Methods and Machine Learning Methods : доклад, тезисы доклада / Sergey V. Mitrofanov, Pavel V. Matrenin, Uriy A. Sekretarev. - [S. l. : s. n.], 2023. - Текст : непосредственный // 24th International Conference of Young Professionals in Electron Devices and Materials / IEEE 24th International Conference of Young Professionals in Electron Devices and Materials (EDM) (2023 ; 29.06 - 03.07 ; Novosibirsk). - Novosibirsk, 2023. - 24. - P. 1130-1134. - The reported study was supported by Russian Science Foundation, research project No. 22-79-00181. - ISBN 9798350336870, DOI 10.1109/EDM58354.2023.10225023.
Аннотация:This article considers the problem of short-term forecasting of natural inflow a day ahead to the site of a hydroelectric power station. The publications review on the research topic, carried out in the first part of the article, substantiates the relevance of using machine learning methods for solving the problem of predicting HPP modes. The second part of the article compares the accuracy of models based on linear and polynomial regression with models based on ensemble methods of machine learning, in particular, decision trees, random forest, gradient boosting of decision trees, single decision tree stacking and random forest. The site of the Sayano-Shushenskaya hydroelectric power plant is considered as an object of research. The calculation results presented in the article were obtained on the basis of retrospective information on the natural inflow to the HPP and meteorological factors for 4 years of observations with a discretization step of a day.
