Estimation of natural gas consumption of power plants in Turkey via support vector regression

The aim of this study is to forecast the natural gas consumptions of power plants in Turkey via support vector machines regression method. With this aim the data set is obtained from Turkey's Energy Market Regulatory Authority and Energy Affairs General Directorate between the years 2013-2018. In this study, first of all, the place in Turkish market, ratio within the primary energy sources, production, consumption, import and export values of natural gas, as a power supply is examined. Because of the differences in measurements of these values, the related data set is standardized before the statistical analysis. While, the consumption in energy plants (thousand cm3) is considered as a dependent variable, industrial consumption (thousand cm3), city consumption (million cm3), production (million cm3), import (million cm3) and export (million cm3) values are considered as independent variables. All types of kernel functions (Linear, Polynomial, Radial Basis Function (RBF) and Sigmoid) in Support Vector Regression are tested. RBF is chosen as the forecasting kernel function because of having the minimum Mean Square Error (MSE). Then, support vectors, weights and decision constants are determined. By multiplying weights with support vectors and adding the bias, the final model is obtained. By the help of final model, estimates of natural gas consumption of power plants in Turkey, for May-December 2018 are obtained.