Forecasting Natural Gas Demand Using Meteorological Data: Neural Network Method

The need for prediction and patterns of gas consumption especially in the cold seasons is essential for consumption management and policy planning decision making. In residential and commercial uses which account for the bulk of gas consumption in the country the effects of meteorological variables have the highest impact on consumption.  In the present research four variables include daily average temperature , average daily relative humidity , sunny hours per day , and average wind speed  were used to predict consumption in the short term. The results for the three cities  include Ilam, Ivan and mehran have been obtained in the form of quadratic polynomial equations and according to the above variables. The results for three models based on the normalized root mean square error have are 0.21, 0.112 and 0.123. Also the rerult of coefficent of determination for three cities are 0.8356, 0.87060.7936, respectively with are favorable.