Recognition of Rainstorm Field in Flood Discharge of High Dam Based on Neural Network

Abstract

Atomization in flood discharge of high dam is a serious and complicated problem which belongs to the research field of water-air and air-water two phase flow. The motion of atomized flow is restricted by water head, discharge and operation scheme and affected by environmental wind and terrain. Numerical simulation is the main method of forecasting the range of atomization. Neural network as a new method in this study has the identities of distribution and nonlinearity which very adapt to simulate the behavior of atomized flow. In design and operation process of a hydro project rainstorm field of the atomization is the main range ensuring the security of the project. In this paper a pattern recognition model of rainstorm field based on neural network is constructed and trained by prototype data. Finally the rainstorm fields of two hydro projects in designing are computed and the results are in comparison with those of mechanics model. The comparison result shows the neural network model can predict the rainstorm field quickly with acceptable accuracy.