Load forecasting based on wavelet analysis combined with the fuzzy support vector kernel regression method

In the analysis of power system with respect to the load forecast, the currently used methods appeared to be insufficient. Based on this, the wavelet analysis (WA) combined with the fuzzy support vector kernel regression method was proposed by considering the characteristics of the load power in load forecast. To start with, wavelet transform was employed to acquire the wavelet decomposition of power load sequence, including the low-frequency profile sequence and high-frequency detail sequence. Then, the fuzzy support vector kernel regression method was introduced to get forecasts on sub-sequences, respectively. Finally, the prediction on the final sequences was reconstructed as a result of prediction, which was compared with the fractal prediction. The results showed that in the case of a small sample number, the prediction method could prevent the kernel function method from over-learning, and further improve the forecast accuracy. It indicates that it is possible for the method to be used for online operation of power load forecasting.