A New Model for Short-Term Power System Load Forecasting Using Wavelet Transform Fuzzy RBF Neural Network

Power load changes periodically. And the effects of climatic (precipitation, relative humidity, temperature, wind speed) on the load should be fuzzy. In order to solve the problem, this chapter presents a method combining wavelet transform, fuzzy set concept, and neural networks for short-term load forecasting. Through the wavelet transform, the load sequence decomposes into subsequences consisting of different wavelet coefficients. On the other side, by the fuzzy neural network, the samples of five meteorological factors influencing power load are transformed into fuzzy input with the subsequences, and then, the suitable RBF neural networks for the forecasting are selected. Finally, the load forecasting sequence is obtained by the reconstruction of the forecasted results from the subsequences. The simulation results demonstrate the proposed method possesses validity and practicability with a mean absolute error below 1.5 %.