NEURAL-NETWORK WITH FUZZY SET-BASED CLASSIFICATION FOR SHORT-TERM LOAD FORECASTING

Electric power utilities require forecast of system demand or electric al load for one to seven days ahead. This paper studies a short-term e lectric load forecasting technique using a multi-layered feedforward A rtificial Neural Network (ANN) and a fuzzy set-based classification al gorithm. The hourly data is subdivided into various class of weather c onditions using the fuzzy set representation of weather variables and then the ANN's are trained and used to perform the load forecasting up to 120 hours ahead with a remarkable accuracy.