A Trous algorithm of wavelet transform was used to decompose wavelet signal, and the cross-correlation analysis was used to analyze the sequence of each wavelet transform, and then the mathematical model correspond with wavelet transform sequence was established, finally wavelet random coupling model was obtained by wavelet reconstruction algorithm. Then, according to the rainfall data in crop growth period of Farm Chahayang from 1956 to 2008, the wavelet random coupling model was established to fit the model prediction test. The results showed that the prediction and fitting accuracy of the model was high, the model could reflect the rainfall variation regulation in the region, and it was a practical prediction model. It was very important for us to determine reasonably irrigation schedule and to use efficiency coefficient of precipitation resource.