According to randomness and periodicity of power load

a new method for short-term load forecasting is proposed based on 2-D wavelet transform and least square support vector machine（LS-SVM）.At first

the load series matrix is decomposed into 4 components based on 2-D wavelet transform

then every component is reconstructed and according to the characteristics of the series

the different LS-SVM models are established to forecast load

finally the ultimate forecasting result is obtained after the forecasted load is superposed.The real forecasting result shows that the proposed method is much more accurate and validate

and it also can decrease the risk of the forecasting

so it is an effective method for short-term load forecasting.