Short-term forecasting of electricity prices: Do we need a different model for each hour?
This empirical paper is a continuation of our earlier work on time series forecasting of day-ahead electricity prices. Given the controversy in the literature whether to use one large model across all hours or 24 separate models, we study if the model structure (and not only the coefficients) should change for different periods of the day. We find that leaving out the statistically insignificant factors leads to, on average, better point forecasts.
|Date of creation:||2008|
|Publication status:||Published in MET (Medium Econometrische Toepassingen) 16.2 (2008) 8-13.|
|Contact details of provider:|| Postal: Wybrzeze Wyspianskiego 27, 50-370 Wroclaw|
Web page: http://prac.im.pwr.wroc.pl/~hugo
More information through EDIRC