Study On The Runoff Change Characteristic And Forecast Method Of Muling River Basin

Hei Longjiang Province is along the east of Eurasia in mid-latitude, on the west of the Pacific coast, approaching cold Siberia on the north side and across the temperate and boreal through the north to south. Hei Longjiang Province is in the nonsoon region and the humid and semi-humid areas, which result to the rich rainfall. So the rational development and utilization of water resource is particularly important. River runoff system, as an important part of the water resource, is influenced by the climate, topography and human activities and so on, showing non-linear and stochastic characteristics. Therefore, analyze the changes in the characteristics of water resources Is of great significance of the prediction of the future development.This paper takes the daily rainfall data of 30 years from the upstream and downstream hydrological stations on the Muling River to the analyze and predict research. This paper relies on the Natural Science Foundation of Heilongjiang Province and mainly uses the theory and the mode of wavelet analysis, artificial neural network theory and the runoff fills model of Xinanjiang. And also use the time series and the real data after the field survey, combining data analysis and the theory. As to the data of the hydrological station of LIshu and Hubei gate, this paper mainly considers the method of wavelet analyze, Mann-Kendall and sliding analysis to analyze the changing characteristics of runoff. And it combined with the wavelet analysis and ANN to improve the reasonableness of the prediction of the future runoff of Muling River. The main achievements of the research are as follows.(1) First, in the macroscopic view, analyze the characteristics of runoff series from several aspects, including the annual runoff of the four stations from upstream to downstream in Muling River, seasonal changes of the runoff and the runoff characteristic in different ages. The amount of runoff in summer and fall is far more than that in spring and winter, and the least amount appears in January and February, which is frozen-based. The annual runoff mainly from the summer flood, but since 2000, the upstream station of Muling River has little flood peak. Mann-Kendall method found the mutation is more likely appears near 1979 and 1979 is the mutation point. Besides, sliding T-test confirms the accuracy of the year of mutation point.(2) It can globally observe the division of the rainfall by evaporation, surface runoff, subsurface runoff and interflow by applying the Xinanjiang model. Based on the characteristics of the soil in northeast, topography and climate, select the appropriate model parameters, valuation and sensitivity factors of Muling River. WM, WUM and WLM are fixed parameters; CI. CS and CG are sensitive factors and need meet CI<CS<CG. Find the value of parameter K after running the model first time, and then get the result after running the model second time with the value of K.(3) Wavelet neural network has been widely used in recent years, this model improved the traditional wavelet network to get a better goodness of fit and accelerate the convergence rate. The best results appears when the value of hidden layer is 3 after the testing and inspection.