This paper delves into the realm of short-term electricity demand forecasting, with a particular emphasis on the integration of an intramonth cycle into the forecasting model. This is a novel approach, as traditional seasonality methods have primarily focused on modeling the intraday, intraweek, and intrayear seasonal cycles of electricity load data for one-day ahead forecasting. To accommodate the intramonth seasonal cycle, a new mathematical modeling scheme is developed. This scheme is applicable to several models, including the ARMA model, HWT exponential smoothing, and the IC exponential smoothing model. In addition to the intramonth cycle, this paper also explores the incorporation of weather forecasts into the electricity demand forecasting. A mathematical model is established for the weather forecasts and the associated forecasting errors. The output of this weather model is then fed into our electricity demand forecasting model. It is shown that this fourfold seasonal method, which includes the intramonth cycle and weather forecasts, outperforms the traditional triple seasonal method. Furthermore, the inclusion of weather forecasts significantly enhances the forecasting accuracy of electricity demand. This research thus provides valuable insights into improving short-term electricity demand forecasting.

Electricity demand; Modeling; Forecasting; Weather forecasts