Study Of Annual Runoff And Forecast In The Tuwei River Basin

The change of runoff has the function of predominance to the variety of the whole hydrology water resources system,It also has profound influence for us how to make use of district of the water resources with the reasonable effectively develop.Along with the development of energy industry and the construction of water conservancy facilities,people more and more pay attention to the variation of runoff and prediction research for the basin of Tuwei river.This will be also for the water resources reasonable development, and water conservancy project construction and social economic development is important and far-reaching significance.This text makes use of the annual runoff data of the Gaojiachuan station in the Tuwei river basin as the research object,using the mathematical statistics method, wavelet analysis method, through an analytical calculation to carry on the research of system to the variety regulation of the runoff characteristic of the Tuwei river basin.And based on the threshold regression model and neural network model to forecast the runoff variety.The main results as followed:(1)The runoff of the Tuwei river basin has a decreasing trend.By using moving average method and the rank correlation test to analysis the qualitative and quantitative trends of the runoff for the Gaojiachuan station, the statistics showed that annual runoff of Tuweihe river has a significant downward trend.(2)By the calculations of unevenly distributed through the year, complete adjustment coefficient, concentration ,the results show that runoff is more evenly distributed during the year.Valley of the runoff regulation capacity increases.(3)Runoff of the Gaojiachuan station in Tuweihe basin is analysised by the wavelet transform, .It's found that the annual runoff process primarily in 2 years, 8 years and 19 years change cycle, which is about 19 years time scale as the first cycle.Also found that the current annual runoff in the late dry season, water had turned on the rise.(4)The logging step length(d) and every threshold interval self-related model stages (nj) are determined by using self-related techniques, the threshold interval amount(l) and the optimizing range of the threshold values(r(1)~r(l-1)) are calculated by Dot Figure which is contained in D.D.C. method, then 0.618 method is adopted in optimizing the threshold values, and at last self-related models are set up for the data series in different threshold intervals respectively. the results show that the model of forecast accuracy was B.(5)Using the neural network toolbox of the Matlab software, two years before the runoff as an input factor, then runoff as the output factor, a three-layered neural network model.Through the forecast annual runoff and found neural network model built by fitting 86% pass rate, forecast accuracy of class.It shows that the neural network model for forecasting the annual runoff in the basin have higher precision, It is a very effective method of annual runoff forecast.Finally, the neural network model prediction results with the comprehensive comparison, we find that neural network model for the study is given watershed runoff prediction can be more accurate forecasting results.