Prognoza gubitaka u prijenosu električne energije primjenom strojnog učenja

master's thesis

• Title (english): Forecasting Electricity Transmission Losses Using Machine Learning
• Author: Franko Pandžić
• Mentor: Zoran Kalafatić (mentor)
• Committee member: Zoran Kalafatić (predsjednik povjerenstva)
• Committee member: Tomislav Hrkać (član povjerenstva)
• Committee member: Ninoslav Holjevac (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-07, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Prognoza gubitaka snage u prijenosu električne energije od iznimne je važnosti operaterima elektroenergetskih sustava. Naime, zbog tromosti sustava važne odluke o upravljanju moraju se donositi satima ili čak danima unaprijed. Neoptimalne odluke dovode do velikih regulativnih troškova, stoga je znanje o budućim gubicima snage neophodno.
Gubitke snage na dalekovodu Melina-Velebit najbolje opisuju druge elektroenergetske veličine; uključujući tok snage voda, proizvodnje svih elektrana i razmjene energije (uvoz i izvoz u druge susjedne države). Meteorološke pojave također utječu na gubitke snage, međutim, informacije o istima nisu korištene u ovom radu zbog nedovoljnog opsega dostupnih podataka.
Prije učenja, podacima su se uklonile bilo kakve moguće prepoznatljive nepravilnosti te su interpolirana nedostajuća mjerenja.
Kod učenja razmatrani su, ne brojeći baseline modele, model slučajnih šuma, model pojačavanja gradijenta te ponavljajuća neuronska mreža s LSTM ćelijama. Najbolje rezultate po MAE, MAPE i RMSE generirao je model pojačavanja gradijenta dok je LSTM neuronska mreža lošije prognozirala čak i od baseline modela.

Abstract (english)

The forecast of power losses in the transmission of electricity is of utmost importance to power system operators. Namely, due to the inertia of the system, important management decisions must be made hours or even days in advance. Non-optimal decisions lead to high regulatory costs therefore the information of future power losses is essential.
Power losses in Melina-Velebit transmission line are best described by other power quantities; including transmission line power flows, combined power plant production and energy exchange (imports and exports of power to other neighboring countries). Meteorological phenomena also affect power losses, however, information about them was not used for model training in this paper due to insufficient available data.
Before model training, any possible identifiable irregularities were removed from the available data with all the missing measurements being interpolated.
For model training, not counting baseline models, a random forests model, a gradient boosting model and a repeating neural network with LSTM cells were considered. The best results by MAE, MAPE and RMSE were generated by the gradient boosting model while the LSTM neural network performed worse even than the baseline models.

• Keywords: gubitak snage; vremenska serija; strojno učenje; neuronska mreža; pojačavanje gradijenta; slučajne šume
• Keywords (english): power loss; time series; machine learning; neural network; gradient boosting; random forests
• Language: croatian
• URN:NBN: urn:nbn:hr:168:608902
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 15:46:35