An intelligent hybrid scheme for time-series forecasting of electric load

This paper presents an intelligent hybrid scheme for short-term electric load forecasting using multilayered perceptrons. The hybrid neural network uses the membership values of the linguistic properties of the past load and weather parameters and the output of the network is defined as the fuzzy class membership values of the forecasted load. A hybrid learning algorithm consisting of unsupervised and supervised learning phases is used for training of the feedforward neural network. In the unsupervised learning phase optimal fuzzy membership values of input/output variables are obtained along with the optimal fuzzy logic rules. Kalman filter is used for the supervised learning phase. Extensive tests have been performed on a two-year utility data for the generation of peak and average load profiles in 24 and 168 hours ahead time frame. Results for typical winter and summer months are given to confirm the effectiveness of the hybrid scheme in comparison to standard ANN approach using backpropagation algorithm. Key words:Hybrid learning scheme, fuzzy logic, ANN-based architecture. Kalman filters, load forecasting. I