论文部分内容阅读
讨论了神经网络在降雨径流模拟中的运用现状,并利用改进的BP网络模型,实现了对复杂流域降雨径流的预报。在构建网络模型时,对模型进行了改进:基于各雨量站对径流模拟的影响不同的特点,采用两轴法计算流域平均日雨量;并把汛期和枯水期分开来建模,使降雨径流模型的预报更为准确。使用这种新方法建立了松花江干流流域的降雨径流预报模型,并与其他方法进行对比,实验表明,采用新方法预报精度较高,能较好地反映计算流域的降雨径流规律。
The status quo of the application of neural network in rainfall-runoff simulation is discussed. By using the improved BP network model, the prediction of rainfall runoff in complex watershed is realized. The model was improved when constructing the network model. Based on the different effects of each rainfall station on the runoff simulation, two-axis method was used to calculate the mean daily rainfall in the river basin. The flood season and the dry season were modeled separately to make the rainfall runoff model Forecast is more accurate. Using this new method, a rainfall runoff forecasting model was established for the mainstream of the Songhua River and compared with other methods. Experiments show that the forecasting accuracy of the new method is high, which can well reflect the law of rainfall runoff calculation.