magistrsko delo
Matic Tajnik (Author), Drago Bokal (Mentor), Tanja Gologranc (Co-mentor)

Abstract

Magistrsko delo obravnava primerjavo pristopov različnih tehnik k napovedovanju porabe električne energije. Delo je razdeljeno na pet poglavij. V prvem poglavju so predstavljene tehnike modeliranja, ki so potrebne za razumevanje opravljenih analiz in nadaljnjih primerjav, to so: večstopenjska linearna regresija, metoda podpornih vektorjev, naključni gozd in mehka logika. Pregledu metod modeliranja sledi poglavje, kjer so predstavljeni indeksi kakovosti modelov. Razdeljeni so v pet podpoglavij: napake, determinacijski koeficient, popravljen determinacijski koeficient, statistični F-test in informacijski kriteriji. V tretjem poglavju so podrobno predstavljeni in razčlenjeni integrirani avtoregresijski modeli premikajoče sredine (ARIMA). Naprej je predstavljena avtokorelacija in njene funkcije, sledi definicija stacionarnosti in diferenciranja časovne vrste, predstavljeni so sezonski ARIMA modeli, na koncu sledijo koraki Box-Jenkins metodologije za izgradnjo ARIMA modelov. V četrtem poglavju je povzeta uporaba taksonomije, izdelana je razširitev taksonomije napovedovanja v elektrogospodarstvu, predstavljena je obdelana literatura in prikaz taksonomskih enot, ki so bile vsebovane v njej. Poleg taksonomskih enot so za obravnavano literaturo predstavljeni grafi primerjav tehnik modeliranja. V zadnjem poglavju so predstavljeni izračuni in primerjava rezultatov natančnosti modelov za napovedovanje. Najprej je predstavljena lastna časovna vrsta, sledi konstrukcija ARIMA modela po Box-Jenkins metodologiji in kasneje še modelov AutoARIMA (funkcija, ki samostojno določi parametre modela), multiple linearne regresije (MLR) in metode podpornih vektorjev (SVM). Na koncu poglavja so prikazane analize primerjav med modeli glede na dolžino in odmik učnega obdobja. Primerjani so tudi modeli za 12 urno napovedovanje.

Keywords

napovedovanje;linearna regresija;naključni gozd;podporni vektorji;ARIMA modeli;taksonomija;mehka logika;informacijski kriteriji;magistrska dela;

Data

Language: Slovenian
Year of publishing:
Typology: 2.09 - Master's Thesis
Organization: UM FNM - Faculty of Natural Sciences and Mathematics
Publisher: [M. Tajnik]
UDC: 519.21(043.2)
COBISS: 22918152 Link will open in a new window
Views: 2222
Downloads: 289
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Autoregressive integrated moving average models for forecasting electricity consumption
Secondary abstract: This master thesis investigates the comparison of approaches to forecasting electricity consumption. The thesis is divided into five chapters. In the first chapter, we present the modelling techniques which are necessary for understanding the analyses and further comparisons, these are: multiple linear regression, support vector machines, random forest and fuzzy logic model. Review of the modelling methods is followed by a section which presents the indices of quality. They are divided into five sub-sections: errors, coefficient of determination, adjusted coefficient of determination, statistical F-test and information criteria. In the third chapter, the autoregressive integrated moving average (ARIMA) models are presented in detail. Next, autocorrelation and other functions are presented, definition and differentiation of stationarity of the time series, seasonal ARIMA models are discussed, and lastly, the steps of Box-Jenkins methodology building ARIMA models are listed. The fourth chapter summarizes the use of taxonomies, the extension of taxonomy prediction in the electricity sector is made, the literature is presented and the taxonomic units, which were contained in it, are shown. In addition to the taxonomic units, the graphs that show the comparison of the modelling techniques are presented. In the last chapter, calculations and the comparison of results of the precision of forecasting models are made. First, time series is presented, followed by the construction of the ARIMA model and later on, the models AutoARIMA, multiple linear regression (MLR) and support vector machines (SVM). At the end of the chapter there is the analysis of comparisons between models depending on the length of the learning period and delay. A comparison is also made between models for 12-hour prediction.
Secondary keywords: forecasting;multiple linear regression;support vector machine;ARIMA models;taxonomy;fuzzy logic;information criteria;master theses;
URN: URN:SI:UM:
Type (COBISS): Master's thesis/paper
Thesis comment: Univ. v Mariboru, Fak. za naravoslovje in matematiko, Oddelek za matematiko in računalništvo
Pages: XVI, 95 f.
ID: 9247736