Energy consumption estimation with diferansiyel polynominal neural network technique

Electricity energy is one of the commodities in terms that mirrors the level of welfare and modernization of the countries. Electric energy is the type of energy that must be consumed when it is produced without storing it. For this reason, the prediction accuracy of the demand for consumption is important to meet the supply. The use of electrical energy and the need for it is increasing day by day with the developments such as the acceleration of population growth, developing technology and rapid industrialization. In addition, as electricity energy is one of the important elements of competition among countries, countries are working on obtaining more accurate estimates by developing correct estimation system. In this study, Turkey's electric energy consumption has been estimated with the artificial neural networks technique, which are prominent in application prevalence and differential polynomial neural networks which a new kind of neural networks technique. Export, import, population, installed power and gross domestic product are important factors affecting electricity consumption. Therefore, by considering as independent variables were used as inputs for the 1965-2016 model years in Turkey's electricity consumption estimated.As a result of the comparisons, the mean absolute percentage error of the results obtained from the differential polynomial neural network method was 4.32% lower. The statistical methods used in the analysis of the results have revealed that the differential polynomial neural network performed highly accurate estimates. Keywords: Differential polynomial neural networks, artificial neural networks, electric energy, consumption forecast