论文部分内容阅读
针对郁江流域灾害性洪水发生频繁,选取南宁水文(2)站历史观测中具有代表性的24场超73 m的洪水水位流量资料进行水位(H)—流量(Q)关系曲线拟合研究,分别采用简单趋势线、多项式及BP神经网络三种不同类型方法来进行拟合。结果表明,三种模拟方法均能满足精度要求,其中三阶正交多项式拟合精度最佳,均方残差达到0.000 001,BP神经网络表现最差,但模型本身不存在明显的优劣之分。因此对于不同流域应该选用何种模型模拟,需结合流域自然地理特征及洪水成因规律进行具体分析。
According to the frequent occurrence of catastrophic floods in the Yujiang River Basin, 24 representative waterlogging data of 24 ultra-73 m floods in the historical observation of Nanning Hydrological Station (2) were used to fit the curve of water level (H) -flow (Q) Using simple trend line, polynomial and BP neural network three different types of methods to fit. The results show that all the three simulation methods can meet the precision requirements. Among them, the fitting accuracy of the third-order orthogonal polynomial is the best, the mean square residual error reaches 0.000 001, BP neural network has the worst performance, but the model itself does not have obvious advantages and disadvantages Minute. Therefore, it is necessary to carry out a detailed analysis of the model of the basin which should be selected for different types of watersheds, combining with the natural geographical features of watersheds and the causes of floods.