Forecasting electrical load using a multi-time-scale approach

This paper describes the application of a multi-time-scale technique to the modelling and forecasting of short-term electrical load. The multi-time-scale technique is based on adjusting the underlying short sampling period forecast time series with specific target points and possible aggregated demand. This allows not only improvement of the short sampling period forecast, but also focuses on weighting the accuracy of the forecast at certain critical points e.g. the overnight minimum and daily peaks. Various model types may be utilised at the upper level (forecating the aggregated consumption and target points at daily level), including intelligent models such as neural and fuzzy models, but the base model is currently restricted to a linear form. Results for the Irish national electrical grid demonstrate the effectiveness of the technique.