The traditional hydrological time series prediction models cannot get good prediction accuracy and they generally do not consider the actual noise. A new kind of hydrological forecasting model (Wavelet De-noise and Rank Set Pair Analysis forecasting model

WD-RSPA forecasting model) was built based on wavelet de-noising and set pair analysis

and was applied to predict the annual runoff of Makou Station and the annual precipitation of Shenzhen Station. The results show that when the dimension T=4

the prediction error of Makou Station with coif3RSPA model is 11.97%; and when T=6

the prediction error of Shenzhen Station with db5-RSPA model is 17.73%. Compared to the results predicted by AR(1) model and RSPA model

the results of WD-RSPA model are closer to the observed data

indicating that it is a practicable method to predict the hydrological time series.