ДОСЛІДЖЕННЯ ЕФЕКТИВНОСТІ НЕЙРОННОЇ МЕРЕЖІ PROPHET ДЛЯ ВИЗНАЧЕННЯ ПАРАМЕТРІВ ВІТРОВОГО ПОТОКУ

• Datum: 2025
• Hauptverfasser: Molyboh , О., Baltovskyy , Y.
• Format: Artikel
• Sprache: Ukrainisch
• Veröffentlicht: Institute of Renewable Energy National Academy of Sciences of Ukraine 2025
• Schlagworte: renewable energy; measurement of wind flow parameters; wind energy potential; mean wind speed; MCP approach; Prophet neural network; forecasting; and wind power density.
• Online Zugang: https://ve.org.ua/index.php/journal/article/view/585
• Назва журналу: Vidnovluvana energetika

The aim of this article is to assess the feasibility of using the Prophet neural network together with selected Python libraries for the correlation analysis and mid-term forecasting of wind potential in the Azov region, as well as to compare the obtained results with several linear regression methods. In this study, wind-speed measurements from meteorological masts were correlated with ERA5 reanalysis data at an altitude of 100 m, and the corresponding slope coefficients and intercepts were determined. Synthetic time series were generated from the reanalysis data for the 2009–2019 period and used as regressors in training the neural network. Based on these series, the neural network produced hourly wind-speed forecasts for 2020–2022, from which the mean wind speeds, Weibull distribution shape and scale parameters, and wind-power density at 100 m were calculated. The mid-term forecasts (1–2 years ahead) were validated against measurements from a second meteorological mast located nearby. For the long-term forecast, the results were compared with wind-potential assessments obtained using the WindPRO software.