Analysis of the relationship between load profile and weather condition

In order to make short-term load forecasting more accurate, the weather sensitive component of total load is identified through a new approach with an aim to investigate the relationship between temperature variation and load variation. After applying Empirical Mode Decomposition (EMD), the load variation can be separated into several parts and the one named residue with the lowest frequency shows a high degree of correlation with temperature variation in regression analysis. The correlation coefficient of winter days however changes in value and polarity over three time periods through a day, which is probably caused by human behaviors. The final regression results show that the residue of load variation, rather than real load and load variation, could represent the weather sensitive part of load well with a correlation coefficient around 0.9.