A review of applicable mathematical/statistical methods for anticipation and design of a model for anticipating natural gas consumption

Pridruživanjem Republike Hrvatske Europskoj Uniji započeta je liberalizacija tržišta plina. Liberalizacijom tržišta stvaraju se uvjeti za povećanje konkurencije i ukidanje državnih monopola, ali isto tako ostvaruje se smanjivanje nabave cijene plina kao i snižavanje transportnih troškova. Uvođenjem financijske odgovornosti za neodgovorno i neplanirano ponašanje na tržištu, svim sudionicima postaje izrazito bitna mogućnost što točnijeg predviđanja očekivane potrošnje. Predviđanje potrošnje prirodnog plina predstavlja problem predviđanja vrijednosti neke varijable u ovisnosti o raznim faktorima koji na njeno ponašanje utječu svaki u određenom omjeru. Zadatak diplomskog rada bio je istražiti matematičke modele predviđanja koji bi bili primjenjivi za izgradnju sustava predviđanja potrošnje prirodnog plina te izvršiti demonstraciju rada odabranih modela na realnom setu podataka. Kreirana su četiri linearna modela predviđanja i jedan nelinearan model. Modeli su trenirani i testirani na stvarnim podacima, potom su uspoređeni. Doneseni su zaključci o prednostima i nedostatcima svakog pojedinog modela.Croatia's accession to the European Union initiated the liberalization of the gas market. Market liberalization makes conditions for increased competition and the abolition of state monopolies, but also reduces purchase prices and reduces transport costs. By introducing financial responsibility for irresponsible and unplanned behavior on the market, the possibility of accurate forecasting of consumption becomes extremely important to all participants. Prediction of natural gas consumption indicates the problem of predicting the value of a variable depending on the various factors affecting its behavior in each particular proportion. The task of this graduate thesis was to investigate the mathematical forecasting models that were applied for the construction of a system that predicts natural gas consumption and to demonstrate the work of selected models on a real data set. Three regression model predictions and one nonlinear model were created. The models were trained and tested on the actual data that was up and the conclusions about the advantages and disadvantages of each particular model