The goal of this dissertation is to present a quantitative study in time series of weekly electrical charge demand at the southeast region, particulary at Rio de Janeiro and São Paulo. In this work will be analysed the last 7 years, from january 1991 to november of 1997. The next time series were study: LIGHT, CERJ, CESP, CPFL and ELETROPAULO. Aimming to test the model against real data the concept of sample data was utilized in this dissertation. Another concept used in this work was outperformance. Outperformance is a Bayesian concept that involves the combination of two or more techniques in order to enchance the forecasting results. Artificial neural network and Box and Jenkins method are combined in this work. It is also interesting to notice that weight elimination, which is a new ANN technique, proved to be faster then classical back- propagation and yielded better results.