Study On Short-Term Load Forecasting Method Based On Data Mining

For a multifactor power load prediction problem and typical training sample selection, a new method for Short-Term Load Forecasting (STLF) based on data mining is put forward. First of all, through adjusting amplitude of wavelet modulus maxima and processing the wavelet decomposed detail signal by soft threshold based on wavelet analysis and singularity theory, fault data in original loads are eliminated. Then, through wavelet transform, the processed load sequence is decomposed into different frequency parts. For each part, information entropy and principal component analysis are integrated to reduce load influential factors; dynamic clustering analysis is used to automatically determine hidden nodes and training set; ant colony optimization algorithm is employed to optimize the network parameters initialized by dynamic clustering and least square method. Finally, the eventual forecasted results are obtained through wavelet restructure. The testing results of STLF in actual power network show that the proposed method possesses higher forecasting accuracy and better adaptability.