Primjena metoda strojnog učenja u prognoziranju proizvodnje električne energije iz vjetroelektrana

undergraduate thesis

• Title (english): Machine learning application for wind farm production forecasting
• Author: Fran Hruza
• Mentor: Ninoslav Holjevac (mentor)
• Committee member: Ninoslav Holjevac (predsjednik povjerenstva)
• Committee member: Matija Zidar (član povjerenstva)
• Committee member: Tomislav Baškarad (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-07-07, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Rad daje pregled modela koji se danas koriste za predviđanje proizvodnje VE. Integriranje OIE, od kojih su među najzastupljenijim VE, zahtjeva pouzdana predviđanja električne energije u različitim vremenskim rezolucijama, od satne do godišnje prognoze. Izgradnjom jednostavnog LSTM modela neuronske mreže moguće je dobiti predviđanja čija se točnost može usporediti s komercijalnim alatima poput WPPT3 modela. Postupak izgradnje mreže započinje analizom podataka dobivenih iz SCADA sustava. Specifično, u radu su korišteni podatci VE Bruška. Analiziranjem podataka dobiven je pogled na vremenske uvjete u kojima VE radi i utemeljena je korelacija vremenskih čimbenika s proizvodnjom VE. Nakon analize, izgrađen je LSTM model koji je treniran i testiran na povijesnim podatcima kako bi na temelju temperature, atmosferskog pritiska, brzine i smjera vjetra mogao predvidjeti proizvodnju električne energije VE. Nakon treniranja i testiranja modela, temelju podataka dobivenih iz WRF modela za predviđanje vremenske prognoze dobivena su predviđanja koja su bila uspoređena sa stvarnom proizvodnjom u istom vremenskom periodu i predviđanjima iz WPPT3 modela te je utvrđeno da model ima vrlo dobre performanse usporedive s WPPT3 modelom.

Abstract (english)

The paper gives an overview of models that are used in today's predictions of wind farm power production. Integrating renewable energy sources, a significant percentage of which are made up of wind farms, demands reliable forecast of energy production in various time resolutions, from hourly to yearly forecasts. By building a simple LSTM neural network model, it is possible to achieve forecast whose accuracy is comparable to commercial tools like the WPPT3 model. The process of building the model starts with analysing data acquired from the SCADA system. Specifically, data from the Bruška wind farm were used. By analysing the data, the weather environment in which the wind farm operates, as well as the correlation between weather factors and wind farm power production were established. After analysing the data, an LSTM model was trained and tested on historical data to be able to predict the power production of wind farms based on temperature, air pressure, wind direction and wind speed. After training and testing, the model was used to predict power production based on weather data acquired from the WRF model for weather forecasting. The predictions were compared with actual production and predictions made by the WPPT3 model in the same time period. It was determined that the constructed model has very good performance, comperable with the WPPT3 model.

• Keywords: Vjetorelektrane; LSTM; strojno učenje; rekurzivne neuronske mreže; predviđanje proizvodnje elektricne energije; analiza podataka
• Keywords (english): Wind farms; wind power forecasting; LSTM; recurrent neural networks; data analysis
• Language: croatian
• URN:NBN: urn:nbn:hr:168:152328
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-10-11 09:03:43