Application of time series models for heating degree day forecasting

Abstract

Abstract This study aims at constructing short-term forecast models by analyzing the patterns of the heating degree day (HDD). In this context, two different time series analyses, namely the decomposition and Box–Jenkins methods, were conducted. The monthly HDD data in France between 1974 and 2017 were used for analyses. The multiplicative model and 79 SARIMA models were constructed by the decomposition and Box–Jenkins method, respectively. The performance of the SARIMA models was assessed by the adjusted R2 value, residual sum of squares, the Akaike Information Criteria, the Schwarz Information Criteria, and the analysis of the residuals. Moreover, the mean absolute percentage error, mean absolute deviation, and mean squared deviation values were calculated to evaluate the performance of both methods. The results show that the decomposition method yields more acceptable forecasts than the Box–Jenkins method for supporting short-term forecasting of the HDD.