Multiple RBF-NN model for electrical load prediction based on anti-aliasing wavelet transform

Electrical load prediction is very important for scheduling and administration of electricity power system. In order to improve the short-term electrical load forecasting accuracy, a method based on multiple RBF neural networks is proposed in this paper. The wavelet transform method is firstly used to decompose the electrical load into different parts, corresponding to different influencing factors. In addition, an anti-aliasing method is designed to eliminate the frequency aliasing during wavelet decomposition and reconstruction. Then a compound model with three sub RBF neural networks is constructed, each for a part of the decomposed electrical load. Utilizing the historical data, the model is trained and prediction is carried out. Simulation results show that the proposed multiple RBF neural network method can achieve good performance, the mean absolute percentage error decreases from 3.86 by ANN method to 1.7 by the proposed method.