The Relation between Gas Demand and Temperature: A Study in Statistical Demand Forecasting

Abstract

The paper describes the methods of relating gas demand to temperature which have previously been used for forecasting gas demand in the North Western Area, and introduces a new approach which involves the use of exponential weighting as applied to the average temperatures occurring in the 4 weeks prior to the forecast. An analysis of previous gas demand experience (in relation to actual temperatures throughout) showed that the new method gave a much improved fit to the data, and that this accuracy was maintained even in prolonged periods of abnormal weather conditions. The new method also threw more light on the effect on gas demand of other meteorological factors apart from temperature, and showed up significant effects which had not been detected when using other methods. The method can be used for forecasting on both a daily and a weekly basis, and is considered to represent a significant step forward in the technique of gas demand forecasting.