基于人工神经网络的洪水水位预报模型

Artificial neural network model for flood water level forecasting

本文利用人工神经网络技术,以确定性系数为目标函数,建立以上游和本站水位资料预报本站未来若干时段洪水水位的预报模型,以探讨神经网络技术在水文预报中的应用。研究成果为提高网络训练速度和预报可能产生的超历史洪水情况,给出了输入输出层的数据规范化的处理方法。选择珠江三角洲河网地区水位站资料,对预报模型进行检验,结果表明在合理选择输入层单元数据和预见期的条件下,可以取得很好的预报成果。

The artificial neural network technology is applied to establish the model for forecasting the flood water level based on the data of upstream hydrological station and local station. The deterministic coefficient in forecasting norm for hydrology is taken as the objective function. For improving the network training speed and forecasting the probable high water level exceeding the highest water level in history, a standardized method is given to treat the data of input layer and output layer. The proposed forecasting model is applied to analyze the hydrological data of two stations located in Beijiang River, Zhujiang Delta. The result shows that by reasonable selection of original data for input layer element and forecasting period, satisfactory forecast precision can be obtained