One day ahead load forecasting by recurrent neural networks
No Thumbnail Available
Date
1997
Journal Title
Journal ISSN
Volume Title
Publisher
C R L PUBLISHING LTD
Abstract
In recent years, many applications of neural network methodologies to power system problems have been reported. Among them, short term load forecasting has been one of the most popular. Multilayer perceptron networks have constituted the preferred architecture, achieving successful results. However this network model generally fails to deal with the temporal characteristics of the load signal, being more suitable for static pattern recognition tasks. Dynamic or recurrent networks have shown better capabilities for time signals modeling and forecasting. This paper presents the application of a recurrent network model, which uses a very limited amount of data, to the load forecasting problem. Particularly, the Elman, recurrent model was applied to the 24 hour ahead load forecasting for the Chilean Central Interconnected System (SIC). The load values are considered as a time series, taking advantage of the temporal processing capabilities of this neural network model.
Description
Keywords
load forecasting, recurrent neural networks, SYSTEMS