20.500.12556/DKUM-92762 Razvoj modela za napovedovanje odjema toplote v sistemu daljinskega ogrevanja z uporabo umetne inteligence magistrsko delo Development of a Model for Heat Demand Prediction in a District Heating System Using Artificial Intelligence Magistrsko delo obravnava razvoj različnih modelov globokih nevronskih mrež za napovedovanje odjema toplote v sistemu daljinskega ogrevanja. Razviti so bili trije napovedni modeli, pri čemer ima vsak svoje prednosti in slabosti. Predstavljen je celoten postopek razvoja – od zbiranja in urejanja podatkov, izbire ustreznih slojev in arhitekture nevronskih mrež, do določitve hiperparametrov učenja. Za vsako kombinacijo modelov so prikazane tudi metrične napake. Poleg tega so ocenjeni prihranki oziroma dodatni stroški ob uporabi posameznih napovedi. Rezultati so pokazali, da razviti modeli v nekaterih primerih podajo boljšo napoved od trenutno uporabljenega sistema napovedovanja in z njimi lahko prispevamo k optimizaciji stroškov. V primerih, kjer zaostajajo, pa nam razviti modeli ponujajo dobro izhodišče za nadaljnji razvoj in izboljšave. The master's thesis addresses the development of various deep neural network models for predicting heat demand in a district heating system. Three predictive models were developed, each with their own advantages and disadvantages. The entire development process is presented – from data collection and preprocessing, selection of appropriate layers and neural network architectures, to the determination of learning hyperparameters. For each model combination, metric errors are also presented. In addition, estimated savings or additional costs associated with using each prediction are evaluated. The results showed that in some cases, the developed models provide better predictions than the currently used forecasting system and can contribute to cost optimization. In cases where they underperform, the developed models offer a solid foundation for further development and improvement. daljinski sistem ogrevanja globoko učenje LSTM programsko napovedovanje District heating system deep learning LSTM software prediction true true false [P. Jakopiček] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2025-05-12 13:01:17 2025-05-27 13:48:37 2025-06-18 03:08:45 0000-00-00 00:00:00 2025 0 Maribor 0 Maribor 1 spletni vir (1 datoteka PDF (XIV, 94 f.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 [004.8.032.26:519.216]:697.34(043.2) 239581443 MAG_Jakopicek_Patrik_2025.pdf MAG_Jakopicek_Patrik_2025.pdf 1 00815AB8267BBC2C54E3C01A35B59367 b043d30b2f5fc453134be498ad1228c8e675ee7a5637930a78fd8f579dfd5ca0 1302f5d5-3494-11f0-ad3e-00155d000105 https://dk.um.si/Dokument.php?lang=slv&id=191094 Fakulteta za strojništvo 0 0 0