Seasonal Artificial Neural Network Forecasters

O. Alsayegh, B. Alfeeli, and O. Almatar (Kuwait)


Annual power demand, artificial neural networks, daily power demand, short-term load forecasting.


This paper presents the development of an artificial neural network (ANN) based short-term load forecasting system for the power control center of the Ministry of Electricity and Water (MEW), Kuwait. The proposed seasonal ANNs (SANNs) consist of 12 independent networks. Every three networks are assigned to a season, namely, winter, transition from winter to summer, summer, and transition from summer to winter. Each network (of the three networks) is trained with weather-related data, historical electric load-related data, and social event-related data of particular hour time duration. The hour time durations include 00 to 08, 09 to 15, and 16 to 24. In using the data from the calendar years 1998 and 1999 as a test case, the absolute average error for day-ahead forecasting is reduced from 5.24% to 1.33% by applying SANNs compared with the MEW regression-based forecasting system.

Important Links:

Go Back