Paper Title
Time Series Decomposition of Natural Gas Consumption

Abstract
Natural gas takes place one of the most important topics of energy sector in Turkey. This consumption is divided into subcategories related to the fields. In this research, residential natural gas consumption is studied among these areas. It is known that natural gas usage is affected from seasons and/or cycles. In this research, curve fitting and seasonal affect included decomposition methods are used for demand forecasting. Study fundamentally based on accomplishing daily predictions of the year 2014, on two different datasets between years 2011-2014 and 2011-2013. Unlike few data predictions done in the literature, this study covers excessive data predictions for 365 days. For the first round, data set between 2011-2014 years are used for daily prediction of 2014. Here, additive trend seasonal decomposition method gave the lowest MAPE of 16%. For the second dataset of 2011-2013 years, 26% MAPE is obtained by seasonal additive decomposition method. Trend and seasonal affect included additive decomposition method gave 29% MAPE for this dataset. Additionally, both multiplicative with trend and trend seasonal decomposition methods gave ratio of 27% MAPE. Besides exponential growth trend, linear and quadratic trends tend to fall. Index Terms�Trend models, time series, decomposition, forecasting, natural gas, consumption, cycling, short term.