中文摘要：

通过实验对比的方法,验证了数据串行时间延迟处理对于提高BP神经网络对时间序列问题模拟精度的有效性,并将其应用于黄河下游夹河滩至高村段河道洪水预报中.预报检验结果表明,方法的预报精度优于马斯京根法.在洪水预报的实际应用中,采用N=2的延迟线不但使训练样本具有时序效应,而且能将洪水时段水位（流量）变幅的概念赋予网络,使训练样本蕴含了更多的信息,能有效提高模型的预报精度.

英文摘要：

The validity of data serial time-delay treatment is verified, which can improve simulation precision of the BP network on temporal series by the way of experimental comparison. The model is then used to forecast the river flood between Jiahetan and Gaocun of the lower reaches of the Yellow River. The result showed that the forecasting precision of BP network model which is accomplished by the above method is superior to that of the Muskingum method. In the practical application of flood forecasting, the delay line may adopt N=2. Not only do the training samples have time order effect, but also can give the variation of flood level （discharge） to the network in such kind of network. The training samples contain more information. It has a great significance to improve forecasting precision of the intelligent on the flood process.