I. INTRODUCTION
Load forecasting is very important to the operation of electrical systems for economic and security reasons 131 if accurate assessment of system load power with a lead of at least 24 hours and then have to coordinate energy economy by building. Enough of the pre-supply load Security analysis system used production load to assess the state of the system prior to its commitments in the future may be able to also load forecasting system is an important analysis tool that is used. To restore power, many authors have already delivered using artificial neural networks (ANNs), load forecasting problem [1,2,4] However, most of this area is considered a short-term IEEE 0-7803-. 2972-4/96$5.00@1996 Load forecasting (STLJ) whlch is unpredictable load by borrowing up to 24 hours of weather variables. (M temperature unly) was used in most of these applications, the ICA (Research Use C0mputatic) NAL intelligence PUC- Rio, Brazil), we used the back propagation neural network [5] In a multi-step process to predict. loaded with up to 744 hours forecasts are made by feeding a neural network with a load ago. Hours of the day, we have to mesh nervous four different predictive load of dfferent of the week does not use temperature data due to the nature of the consumers supplied by the electric company in the study, our objective is to show that the correct productions by. tlvs and how to determine the limit of time as we add productions the following analysis is done to identify structural loads previous series neural networks suitable in part 3, we present system topology. chosen nervous and the four present case studies results and conclusions of the work will be presented in five sections.