20.500.12556/RUL-170968 Napovedovanje proizvodnje električne energije v vetrnih elektrarnah na osnovi nevronskih mrež Application of Neural Networks for Predicting Wind Farm Energy Generation Obnovljivi viri energije zavzemajo vse pomembnejši delež v elektroenergetskem sistemu, kar povečuje potrebo po zanesljivem napovedovanju njihove proizvodnje. V diplomskem delu smo se osredotočili na razvoj modela za napovedovanje proizvodnje električne energije vetrnega polja. Zanesljive napovedi so namreč ključnega pomena za učinkovito načrtovanje obratovanja elektrarn, optimizacijo proizvodnih zmogljivosti ter uravnoteženje elektroenergetskega sistema. V tej nalogi je prikazan razvoj modela, ki je na podlagi zgodovinskih podatkov o proizvodnji električne energije in vremenskih razmerah sposoben napovedovati prihodnjo proizvodnjo iz vetrnih virov. Posebna pozornost je namenjena pripravi in obdelavi podatkovne zbirke, saj kakovost vhodnih podatkov pomembno vpliva na učinkovitost napovednega modela. Učna množica temelji na prosto dostopnih arhivih vremenskih podatkov in izmerjeni električni moči, ki jo je vetrno polje oddajalo v omrežje. Modeli so razviti v programskem jeziku Python, pri čemer je uporabljena odprtokodna knjižnica TensorFlow, ki omogoča gradnjo in učenje kompleksnih nevronskih mrež. Na osnovi izsledkov naloge je razvidno, da zaradi kompleksnosti vremenskih vzorcev najprimernejši napovedni model temelji na nevronski mreži. Renewable energy sources are taking on an increasingly important role in the power system, which raises the need for reliable forecasting of their production. This thesis focuses on the development of a model for forecasting electricity generation from a wind farm. Reliable forecasts are crucial for effective power plant operation planning, optimization of production capacities, and balancing the power system. This work presents the development of a model capable of forecasting future wind power generation based on historical electricity production data and weather conditions. Special attention is given to the preparation and processing of the dataset, as the quality of input data significantly affects the performance of the forecasting model. The training dataset was based on publicly available archives of weather data and the measured electrical power output delivered by the wind farm to the grid. The models were developed in the Python programming language, using the open-source TensorFlow library, which enables the construction and training of complex neural networks.Based on the findings of the thesis, it is evident that, due to the complexity of weather patterns, the most suitable forecasting model is based on a neural network. vetrne elektrarne nevronske mreže linearna regresija napovedovanje značilke wind farm neural networks forecasting linear regression features true false false Slovenski jezik Angleški jezik Diplomsko delo/naloga 2025-07-24 07:30:18 2025-07-24 07:30:23 2025-08-01 13:57:45 0000-00-00 00:00:00 2025 0 0 XI, 32 f. 2025 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 621.311.245:004.032.26(043.2) 271886 244620035 39455.pdf 39455.pdf 1 7CFBC2C1E8C8183EA5356182585C22D5 a8cf39b82c4ac6cd86533a9264336122116172c7faf15d49ce8d87c73764b02a 4786d6d7-684f-11f0-9328-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=213271 Fakulteta za strojništvo 0 0 0