The Wujiang River Basin Runoff Evolution Law Analysis And Runoff Forecasting Model Research

As one of the main hydrological factors of the water cycle, river runoff is affected by human activities, watershed climate and geographical factors, and its change is a non-linear dynamical system. In this paper, according to use of Hongjiadu and Goupitan station of 54 years hydrological data, runoff evolution law under the influence of human activities is analysised. And do more accurate and reasonable projecting of future changes in river runoff to provide an important basis for the optimal allocation of river basin water resources and rational development and utilization. The main research content and results achieved of this dissertation are outlined as follows:(1) According to analyse the runoff characteristics of Hongjiadu and Goupitan station, we can see that during runoff annual distribution of flood and dry are clear, and the difference between flood season and dry season of runoff is large. In the age of 1970s,1980s, variation of runoff during the year is more stable, however, after 90 years it was once again tended to be uneven. From the change processes of all characteristic index can be seen, the two stations characteristic index of the appearing year of maximum and minimum are not fully consistent, that is the variation of runoff at Hongjiadu station and Goupitan station are also different.(2) Mann-Kendall Rank correlation test and Spearman Rank correlation test are adopted to analysis the tendency of annual runoff and monthly runoff series at Hongjiadu station and Goupitan station. The results show that annual runoff time series of Hongjiadu station and Goupitan station all exist unconspicuous decreasing tendency. November's runoff series decline at Hongjiadu station achieve a significant level, and March and July's performance showed insignificant upward trend, in which the non-flood season in November-following in January of runoff exist more obvious downward trend. Goupitan in January, February, March, April, June and July runoff series show that have no significant upward trend. (3) Power spectrum analysis and maximum entropy spectral analysis method are used to analysis and calculate the collected data of annual runoff. Comparing the calculated results, the maximum entropy spectral analysis results show that Hongjiadu station has significant periods of 14,3.1 years and Goupitan station has significant periods of 11.2,7,3.1 years. But power spectral analysis results show that Hongjiadu station only has 3.2 years significant periods and Goupitan has no significant cycle. It reflects the maximum entropy spectral analysis method has the characteristics of high-resolution.. (4) The stepwise regression model, divided period stationary autoregressive model and stepwise regression model based on rainfall factors are adopted to forecast Hongjiadu station of 1-5d daily runoff flow. It is able to provide a scientific basis for the future for the development of hydroelectric power generation plans. And the gray Dynamic prediction model is used to research Hongjiadu and Goupitan stations for the long-term runoff prediction. Compare with annual runoff predicting results of the stepwise regression model, the models predicting accuracy is higher and that models have practical value.