Improvement of electric load forecasting reusing past instances and contextual based adaptation

Case based reasoning methodology has been used for improving electric load forecasting. Classical extrapolation computed by curve fitting algorithms is adjusted according to short term variations previously registered. Retrieved registers are adapted according to socioeconomic knowledge available related to the prevision point. This strategy has been integrated in a software framework, named FUTURA, in order to operate with real data bases for electric load forecasting in Peru. Improving results are contrasted with other strategies.