APPLICATION OF ARTIFICIAL NEURAL NETWORKS TO HISTORICAL DATA-ANALYSISFOR SHORT-TERM ELECTRIC-LOAD FORECASTING

The paper illustrates two different Artificial Neural Networks (ANN) a rchitectures for electric Short-Term Load Forecasting (STLF). Two mult i-layer perceptron ANN using the back-propagation learning algorithm h ave been implemented which provide different, although complementary, forecasting approaches (static and dynamic). In order to test the pote ntialities of the architectures implemented, the ANN have been applied to the Short-Term Forecasting of Italian hourly electric load. The im portance of this load (peak demands up to about 38 000 MW) requires to ols for STLF which must be as more accurate and precise as possible. T his fact has imposed the adoption of some algorithmic enhancements to the basic back-propagation algorithm formulation. Since an adequate fo rmulation of the influence exerted on hourly electric load by the main meteorological and climatic factors is nor known at present, the data set used for ANN training phase has concerned only historical series of electric hourly demand. The paper illustrates the two ANN architect ures as well as the computational platforms used for implementation, F inally, some results obtained from the application of the two ANN to t he short-term forecasting of Italian electric load relevant to three d ifferent weeks of the year 1993 are comparatively reported.