الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 16 |  |  |
| | | from | 16 | | |
Abstract
The electric load demand has a dynamic nature and influences by many factors, such as weather, economic and social activities and different load components (residential, industrial, commercial etc.). By analysis of only historical load data, it is difficult to obtain accurate forecasting for load demand. An Information system for forecasting processes based on unsupervised, supervised neural networks is proposed. The unsupervised learning process IS performed using Kohonen’s Neural Network (KNN) for clustering input space into affinity number of classes. For each class, the supervised learning process is performed using Feed forwared Neural Network (FNN). KNN is used for data classification to identify the day classes/types which are essential for forecasting processes. The unsupervised process performs the role of front-end data compression. The historical database that contains the data of the attributes of the forecasting process that cover two years is developed using the recorded actual data collected from Libya energy system. FNN is used to learn the relationship among past, current, future daily load and weather patterns that obtained from the historical database. All input patterns information are stored in distributed form among the various connection weights. The Comparison of the actual values with forecasted values of FNN with and without KNN is made to demonstrate that the forecasting accuracy with KNN is very encouraging.