Study On The Application Of Streamflow Forecasting Methods In Dry Period

With the development of socialism modernization, hydrologic forecasting becomes in great request. Direction of drought resisting and flood protection, construction, management an operation of hydraulic project, as well as construction of national defense all require hydrologic department to provide runoff prediction with long lead times and high veracity.Methods of streamflow forecasting fall into two categories: process-driven methods and data-driven methods. The process-driven methods conceive a streamflow process as the output of a watershed system in the view of system theory, and mathematically approximate the internal physical processes of the watershed system that govern the streamflow process based on some understanding of those physical processes. By contrast, data-driven methods are fundamentally black-box methods, which mathematically identify the connection between the inputs and the outputs, without considering the internal physical mechanism of watershed system.This thesis firstly summarizes the significance and present status of the streamflow forecasting. Three of data-driven methods are introduced: Stepwise regression analysis, BP ANN and multivariate time series model. Factors that affect Heishiguan, Lushi, Luanchuan gauge's daily streamflow process are analyzed. Take appropriate factors into consideration, and determine structure and parameters of the models.By the methods of recursion, streamflow process is forecasted in low-water period of 1996-2001, with lead-times of several days. The results are compared and analyzed.The models this thesis constructs are to be perfected to be used in more regions.