Par Marie Bruguet, Ronan Le Saout et Arthur Thomas.
Uniquement disponible en anglais.
This paper addresses the challenge of accurately adjusting energy consumption data for weather and seasonal variations. In this paper, we revisit the concept of Heating Degree Days (HDD) by proposing a new General Weather Indicator (GWI). This indicator is a vector of the optimal combination of heating and cooling day variables across multiple weather variables, including temperature, wind, sunlight duration, rain, and cloudiness. To construct the GWI, we propose a novel econometric approach which uses K-means to identify thresholds and LASSO penalisation for the selection of variables. An empirical analysis of electricity and natural gas consumption in France reveals that there are nuanced sectoral responses to weather, highlighting the importance of integrating wind and sunlight duration along with temperature. The reliability of the GWI was confirmed by robustness checks across spatial and temporal scales, which demonstrated its suitability for use in different geographical areas and at different data frequencies. Using the traditional base temperature of 17°C instead of the optimal GWI for energy adjustment, we found an underestimated consumption response to temperature of up to 7.3%, or 0.80 TWh per month, equivalent to more than one and a half nuclear reactors, highlighting the importance of accurate weather elasticity estimates.