Modern Flood Forecast Study

China water conservancy is changing from engineering style to resource style. The runoff forecast system, as an important element of the resource style water conservancy , not only is a flood control non-engineering measure, but also can be applied to put water resources to rational use, which becomes more and more attractive by the world recently. The correct runoff forecast in time can yield good economic and social returns.The hydrology system have some characters of complicated open macro system to a certain extent, the complexity and indeterminacy of the hydrology system becomes more and more standing out with the hydrology scale changing from the microscopical to the midscopical, even to the macroscopical, which present spatial variability, differ between the unit and the system, the model re-parameterization and the different scale model coupling. Since a mass of the deterministic and probabilistic information and knowledge must be handled to research the open system, and the single qualitative explanation or the single model can not deal with them entirely, a method integrating the qualitative with the quantitative should be employed. That is to say, complicated classify or hierarchical structure or the interactivity between the higher grade the lower grade should be analyzed synthetically through nonlinear model or the deterministic and probabilistic information and knowledge should be separated and then analyzed respectively. The newly mathematical tools such as the artificial neural networks, the gene algorithm, the wavelet analysis and chaos theory have provided effective means. To increase forecast precision and length, this paper couple those theories to research the nonlinear hydrological problems.(1) BP-GA model application to the flood forecast. In this paper, the following about this has been done:㏕he final BP networks topological structure. Some preceding rain factors were list, then stepwise regression algorithm was employed to select the obvious factors from the list as the input of the BP networks. And the trial-and-error method is employed to define the number of the hidden layer nodes.(2)Defined the objective function. Combining absolute error function with relative error function, and setting it as the objective function of BP networks application to the flood forecast.(3)The span analysis of the genetic operators.8 schemes were provided to analyze the relation of the genetic operators such as population size, cross probability, mutation probability. At last, a optimum group of genetic operators was selected.(2) The model(NNBP) combined the wavelet soft-threshold de-noising with BP-GA model application to the flood forecast. The noise were often attached to the real hydrology time series yield by some unknown factors. To reduce the influence of them to the BP networks, the wavelet soft-threshold de-noising method was employed before BP networks began to train.(3) The collective model(HGCM)based on the individual forecast model of the high frequency and the low frequency application to the flood forecast. The real runoff time series was divided into the high frequency item and the low frequency item with the help of the wavelet analysis first, then the two items were modeled by chaos theory and the stepwise regression algorithm, at last the output of the two models were added together.. (4) The model (BPCM) combined the BP-GA algorithm and the chaos theory application to the flood forecast. A positive Lyapunov exponent was got to show the residual series was chaotic in this paper. The volterra filter based on the self-organizing method was employed to forecast, at last the output of the two models were added together.