magistrsko delo
Igor Jaušovec (Avtor), Iztok Kramberger (Mentor)

Povzetek

V magistrskem delu smo predstavili modele napovedovanja odjema zemeljskega plina. Te smo v osnovi delili na linearne in nelinearne ter nevronske mreže. Izvedli smo več modelov napovedi ločeno po tipih odjemalcev. S primerjavo modelov smo ugotovili zanesljivost modela. Vsem modelom izračuna napovedi smo preko zajema znanih vhodnih in izhodnih podatkov odjema zemeljskega plina in vremenskih spremenljivk na učnem vzorcu določili uteži in nato validirali izračunano napoved odjema na osnovi znanih vhodnih podatkov. Šele z validiranim modelom smo lahko s pomočjo napovedanih vremenskih spremenljivk napovedali odjem. Napovedan odjem smo po potrebi s korekcijo po dnevu v tednu popravili in ovrednotili uspeh napovedi z znanimi odjemi. Pri izvedbi modela napovedi smo se zavedali, da pojavi v naravi niso linearno odvisni, vendar lahko za vsak pojav v naravi opišemo kot linearno odvisen, če gledamo dovolj majhno območje. Modele napovedi smo preizkusili pri kratkoročnem napovedovanju porabe zemeljskega plina. V nalogi smo si zastavili cilj, da se razvije takšna metoda za napovedovanje, ki bo neodvisno od napake napovedi vremena zanesljiva v vseh dneh v mesecu znotraj stimulirane dovoljene tolerance. Takšno smo tudi izvedli tako, da deluje skoraj popolnoma avtomatično.

Ključne besede

regresijska analiza;napovedovanje;zemeljski plin;magistrske naloge;

Podatki

Jezik: Slovenski jezik
Leto izida:
Tipologija: 2.09 - Magistrsko delo
Organizacija: UM FERI - Fakulteta za elektrotehniko, računalništvo in informatiko
Založnik: [I. Jaušovec]
UDK: 004.414:665.612(043)
COBISS: 19695894 Povezava se bo odprla v novem oknu
Št. ogledov: 881
Št. prenosov: 148
Ocena: 0 (0 glasov)
Metapodatki: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Ostali podatki

Sekundarni jezik: Angleški jezik
Sekundarni naslov: NATURAL GAS CONSUMPTION FORECAST MODELING
Sekundarni povzetek: In this master’s thesis we present the models of forecasting natural gas off-take. These are basically divided into linear and nonlinear and neural networks. We carried out several models of forecasts separately by type of customer. By comparing the models we found the reliability of the model. For all models of forecast calculations we used the acquisition of known input and output data of the natural gas off-take and weather variables on the learning sample to determine the weights and then validated the calculated off-take forecast on the basis of known input data. Only the validated model allowed us to predict the off-take through forecasted weather variables. Where appropriate, by correction of the day in week, we corrected the forecasted off-take and evaluated the success of the predictions with known off-takes. In the implementation of the model of prediction we were aware that phenomena in nature are not linearly dependent, however, any phenomenon in nature can be described as a linearly dependent when looking at small enough area. Prediction models were tested in short-term forecasting of natural gas off-take. In the thesis we have set ourselves the goal to develop such a method for prediction, which will, independent of the error in forecast, be reliable in all days of the month within stimulated allowed tolerance. This was also performed, so that it operates almost completely automatically.
Sekundarne ključne besede: regression analysis;forecast;natural gas;
URN: URN:SI:UM:
Vrsta dela (COBISS): Magistrsko delo
Komentar na gradivo: Univ. v Mariboru, Fak. za elekrotehniko, računalništvo in informatiko
Strani: V, 106 str.
ID: 9132436