A power system load forecasting method using wavelet neural network with a process of decomposition-forecasting-reconstruction and error analysis based on SPSS is presented in this paper. First of all, the load sequence is decomposed by wavelet transform into each scale wavelet coefficients of navigation. In this step, choosing an appropriate wavelet function decomposition of load is needed. In this paper, by comparing the signal-to-noise ratio (SNR) and the mean square error (MSE) of the different wavelet functions for load after processing; It is concluded that the most suitable wavelet function for the load sequence in this paper is db4 wavelet function. The scale of wavelet coefficients is obtained by load wavelet decomposition. In the process of wavelet coefficient of processing, the db4 wavelet function is used to decompose the original sequence in 3 scales; High frequency and low frequency wavelet coefficient is got through setting threshold. Secondly, these wavelet coefficients are used as the training sample of the input to the nonlinear regression neural network for processing, and then the forecasting result is obtained by the wavelet reconstruction. Finally, the actual and forecasting values are compared by SPSS with a comprehensive statistical charting capability, which is able to draw beautiful charts and is easy to edit.