CONSTRUCTION OF A SPATIAL DISTRIBUTION MODEL OF WIND ENERGY CHARACTERISTICS

The aim of this article is to develop a model for the spatial distribution of wind energy characteristics across the territory of Ukraine. The subject of the study includes datasets of wind speed values, as well as methods of data correlation, validation, and interpolation. Research results. Based on NASA reanalysis datasets and measurement results from 70 meteorological stations in Ukraine, a dataset of paired wind speed data corresponding to the same location but obtained through different methods was created. Through a comparative analysis of regression task results, evaluated using machine learning models trained on the dataset, the Random Forest model was selected as the most accurate (based on RMSE, R², and Pearson correlation coefficient) for predicting wind speedc deviations in NASA reanalysis data to bring them closer to actual values. The Pearson correlation coefficient improved by 0.07 in the worst case and by 0.66 in the best case. Using the Random Forest model’s predictions, corrections were made to all wind speed values in the NASA reanalysis data. The accuracy of the corrected data was also confirmed by the study of trend dynamics over the course of a year using three different data sources: meteorological station measurements, NASA reanalysis data, and corrected NASA reanalysis data. Using the universal kriging method, the corrected wind speed values were interpolated at grid nodes across the entire territory of Ukraine. The accuracy of the interpolation results was validated using the cross-validation method. Conclusion. Based on these results, a GIS-based tool was created, enabling the determination of reliable wind energy characteristics at any given point across Ukraine. The proposed GIS can primarily be used for the design of wind power plants and for selecting optimal locations for their deployment.