Title :
Load Behavior Changes after Holidays on Thursdays
Author :
de Luca, L.A.D. ; de Oliveira, C.M. ; Wazlawick, R.S.
Author_Institution :
Grad. Program in Comput. Sci., Fed. Univ. of Santa Catarina, Florianopolis
Abstract :
This article shows that the effect of holiday on Thursday affects significantly the load behavior on the next Friday due to the Brazilian culture of joining the holiday with the weekend. A statistic test shows that there are differences between the Friday following the holidays and other Fridays. Also, using a legacy forecast system with usual MAPE (mean absolute percentage error) rates at 2.5%, forecasting tests for these special Fridays that follows the holiday reach about 10% MAPE. This error rate shows that this fact should be considered while performing load forecasting. Preliminary forecasting tests using an artificial neural network to adjust the legacy forecast result output for these special Fridays reduce the 10% MAPE to the same 2.5% MAPE level obtained by the legacy system on common days. It demonstrates that this is a possible solution that can be also applied to this specific forecast problem or to similarones.
Keywords :
load forecasting; neural nets; power system analysis computing; statistical testing; Brazilian culture; Thursday; artificial neural network; forecasting tests; holiday effect; legacy forecast system; load behavior; load forecasting; mean absolute percentage error rates; Computer science; Conferences; Demand forecasting; Load forecasting; Neural networks; Power system planning; Power system security; Support vector machines; System testing; Weather forecasting; Holiday; Load Forecast; Long weekend;
Conference_Titel :
Computational Science and Engineering Workshops, 2008. CSEWORKSHOPS '08. 11th IEEE International Conference on
Conference_Location :
San Paulo
Print_ISBN :
978-0-7695-3257-8
DOI :
10.1109/CSEW.2008.26