基于神经网络理论的河道水情预报模型

河道水流运动过程特别是洪水演进过程是一个复杂的非线性动力学过程，鉴于神经网络具有很强的处理大规模复杂非线性动力学系统的能力，本文将神经网络理论用于河道水情预报的研究，以期识别水流运动变化过程与其影响因子之间的复杂非线性关系，为河道水情预报提供了一条新的途径。在此基础上建立了螺山站洪水预报的非线性动力学模型，通过分析研究得出近年来特别是1998年长江中游出现的小流量高水位现象与螺山汉口河段累计淤积有关，并得到螺山站水位变化与河床淤积之间的定量关系。

Model for flood prediction based on neural network theory

Flood evolvement exhibits a complicated non-linear dynamical process. Seeing that neural network possesses the capabilities of dealing with complex non-linear dynamical systems, this paper demonstrates how it can be used in flood prediction as a new approach involving the non-linear relationship between flood evolvement and its influence factors such as discharge, channel deformation, and so on. Hence the neural network approach was applied to the flood prediction of the Yangtse River at the Luoshan station. Our preliminary results suggest that phenomenon of small discharge with high level in the middle reaches of the Yangtse River recently, especially in 1998, is related to the downstream aggregation. And quantitative relations between water level variation at the Luoshan station and downstream aggregation were obtained