The influence of meteorological parameters on Italian electric hourly load: the selection of variables of the ANN training set for short term load forecasting

Artificial neural networks can be successfully used for short-term load forecasting and it is well known that better forecasting performances can be obtained taking into account the weather influence on electric load. An extensive analysis has been conducted for selecting the monitoring sites most representative of the Italian general weather conditions. The selection activity has been based on the computation of correlation functions. The analysis of the results thus obtained has permitted the identification of correlated and noncorrelated sites for each meteorological variable. The variables identified have been used for integrating the training set of an available ANN in order to test its forecasting performance. The paper reports on research activity aimed at carrying out an adequate model permitting correct representation of weather influence on Italian electric hourly load