Research On Hydrologic Model In Binjiang Basin

Flood problem is a global problem that has been increasing at a worrisome pace in recent years. With 70 years developing, hydrologic model is always the most important nonstructural measure for flood control. And in recent years, the artificial neural network model and the distributed model have become the most popular hydrologic models. In this paper, these models have been researched in Binjiang Basin, and the main research results and initiatives are listed as follows:1) The neural network water level forecasting model has been built.Early in 1980's the neural network has been applied in hydrological forecasting, while there is little research in which the watershed outlet level is predicted only by rainfall input. So the neural network water level forecasting model has been built in this paper.a) The scale factor training algorithm has been brought forward to accelerate the training of BP network.b) Optimized initial weights by genetic algorithm (GA) the BP network can avoid the local searching problems, and enhance the forecast precision.c) With different physical meaning and different range, the rainfall data and water level data have been processed by scale coefficient with different standard. Especially, the water level must be subtracted a base level in order to increase the proportion of the level amplitude in level data. By these processes, the simulation ability of the model will be improved.d) When there are sparse rain gage stations in the watershed, in order to make use of limited information, the point rainfall data have been adopted as the input of model.e) With the combination of feedforward network and recursion network the real-time water level forecast model has been built, i.e., the feedforward network used for model calibration and the recursion network used for level forecast. And the study demonstrates that the approach adopted herein produces reasonably satisfactory results.