ENERGY SOURCES, vol.26, no.5, pp.463-472, 2004 (SCI-Expanded)
In Turkey, the major purpose of natural gas demand in residences is heating, where the residential gas usage is approximately 18% of total gas consumption. The studies on forecasting gas demand for future periods have a great importance because natural gas is an imported energy source. This article describes an approach to obtain appropriate models for forecasting residential monthly natural gas consumption. The method relies on dividing a year into two seasons as heating and non-heating periods and estimating individual autoregressive time series models for each period instead of attempting to capture the seasonal patterns in a single model. In the models, the dynamic relationships between natural gas consumption with time and degree-days measured by weather temperature variations are investigated. This method is applied to one of the five cities supplied by natural gas for residential use in Turkey. A comparison of the forecast accuracy of proposed procedure with that of a single model approach applied to the full set of observations is provided. The results reveal that to use separate models for each period reduces the forecast errors significantly whet? the major purpose of natural gas demand is space heating.