Study And Application Of Fuzzy Cluster Analysis On Hydrological Forecasting

Flood forecasting is based on the flood formation and movement rules, using the past and real-time hydrological and meteorological information. The flood development of the future within a certain period is forecasted and analyzed. Flood forecasting is an important non-engineering measures for disaster prevention and play an important role as advisors and ears in the national economy.Using mathematical methods to describe and simulate the hydrological cycle process, the concept of the hydrologic model come into being. Hydrology model arise by research hydrological phenomena and the laws of hydrology, and in the continuous development and improvement. The present model widely used at most is conceptual hydrological model, which is of some physical basis, good applicability, empirical, simple model structure and good practicality. The influence factor of hydrological forecast results is mainly hydrology, meteorology, geography and geology and other uncertainties factors. These factors have a certain randomness and fuzziness. It is the focus of this thesis how to statistically analyze these uncertainties factors by using fuzzy cluster analysis and forecast real-time flood using regional mature runoff-conflux conceptual model.This thesis is based on the fuzzy clustering analysis and the basic principles of mathematics and according to influence factor to the results of basin hydrological forecast. The historical flood data samples of Cao'e basin is fuzzily classified using fuzzy cluster analysis based on the weighted characteristics, then the different types of historical samples are respectively calibrated on runoff-conflux parameters. Combining hydrology theory and forecast model Operational Mechanism and using improved features of the technology-weighted, the numerical and attributes characteristics which affect the results of forecast will weighed and chose, inapparent characteristics will be taken out, and the remarkable characteristics is weighed. Fuzzy logic rules is easy to adjust, understand and reflect the views of experts, so it is preferable to apply in the actual project. Through hydrological models and Fuzzy Cluster Analysis coupling, it's better to fuzzy cluster analysis according to its influence characteristics in historical flood, and parameters of hydrological models are respectively calibrated under the clustering results. Finally, The types of real-time flood is identified by using fuzzy recognition technology , and its optimal parameter group can improve forecast accuracy of the real-time flood.A better result is received by applying above-mentioned coupling model on Cao'e basin. Compared with solid-state parameters no using Fuzzy Cluster Analysis, the forecasting precision is greatly increased. In this coupling model, not only numerical characteristics but also attributes characteristics is taken into account and the views of experts is reflected. Applied in a basin which has a better historical data series, the forecasting precision of coupling model can be increased. It has been proved that the coupling model is better flexibility and applicability, moreover, its method is simple and physical conception is distinct. This model deserves more research and development.