An evaluation of methods for very short-term load forecasting using minute-by-minute British data
This paper uses minute-by-minute British electricity demand observations to evaluate methods for prediction between 10 and 30Â minutes ahead. Such very short lead times are important for the real-time scheduling of electricity generation. We consider methods designed to capture both the intraday and the intraweek seasonal cycles in the data, including ARIMA modelling, an adaptation of Holt-Winters' exponential smoothing, and a recently proposed exponential smoothing method that focuses on the evolution of the intraday cycle. We also consider methods that do not attempt to model the seasonality, as well as an approach based on weather forecasts. For very short-term prediction, the best results were achieved using the Holt-Winters' adaptation and the new intraday cycle exponential smoothing method. Looking beyond the very short-term, we found that combining the method based on weather forecasts with the Holt-Winters' adaptation resulted in forecasts that outperformed all other methods beyond about an hour ahead.
Year of publication: |
2008
|
---|---|
Authors: | Taylor, James W. |
Published in: |
International Journal of Forecasting. - Elsevier, ISSN 0169-2070. - Vol. 24.2008, 4, p. 645-658
|
Publisher: |
Elsevier |
Keywords: | Electricity demand Very short-term Exponential smoothing Combining |
Saved in:
Online Resource
Saved in favorites
Similar items by person
-
The 101 [one hundred and one] best-performing companies in America
Paul, Ronald N., (1986)
-
Exponentially weighted information criteria for selecting among forecasting models
Taylor, James W., (2008)
-
Coherent Probabilistic Forecasts for Hierarchical Time Series.
Ben Taieb, Souhaib, (2017)
- More ...