城镇化平原河网水动力参数动态反演

Dynamic Inversion of Hydrodynamic Parameters of Plain River Network

随着中国平原河网地区城镇化建设的不断推进,洪水实时模拟和预测预警成为了城镇地区雨洪管理的主要研究内容和重要技术支撑,其中水动力参数动态反演是洪水实时模拟和预测预警的难点之一。作者利用糙率先验知识,探索了城镇化平原区河网水动力模型重要参数糙率的动态反演技术。首先,基于糙率空间变化的连续性和缓变性,提出了以糙率空间分布平滑性作为先验知识,建立了表征糙率空间变化的平滑度矩阵。然后,将糙率空间分布平滑性作为软约束条件,并结合适用于非线性动态系统实时校正的扩展卡尔曼滤波算法,建立了一个稳健的城镇化平原河网糙率参数动态反演模型,从而进行河网水动力模型重要参数糙率的动态反演。通过一个四级河道组成的河网的应用分析,系统检验了糙率空间分布平滑度权重系数、糙率初始估值和测站个数对河道糙率动态反演效果的影响。结果表明:1)通过调整糙率空间分布平滑度权重系数,可以有效控制糙率动态反演的稳定性和空间反演特性;2)靠近监测点的河道糙率反演值趋于真实值,远离监测点的河道糙率反演值趋于空间分布平滑;3)靠近监测点的河道糙率反演值受初始估值影响较小,远离监测点的河道糙率反演值受初始估值影响较大;4)糙率初始估值越精确,监测点个数越多,则动态反演开始阶段的波动调整幅度越小,糙率系统误差越小,反演效果越好。所提出的模型综合了河道糙率先验知识和糙率反演的特征,可有效用于城镇化平原河网糙率参数的动态反演。

With the continuous development of urbanization plain of China,the real-time flood simulation and forecasting has become the main research content and important technical support for flood management in urbanization plain.The dynamic inversion of hydrodynamic parameters is one of the difficulties in real-time flood simulation and forecasting.Based on prior knowledge of rive roughness,this paper explored an efficient dynamic inversion technique for estimating river roughness parameters.Firstly,based on the continuity and slowness of spatial variation of river roughness,a prior knowledge that used smoothness matrix representing spatial variation of roughness was proposed.Then,taking the prior knowledge as a soft constraint condition,and combining with the extended Kalman filter algorithm for real-time correction of nonlinear dynamic system,a robust dynamic inversion model of river network roughness parameters in urbanization plain was developed.Through an example analysis,the influences of weight of roughness smoothness item,initial roughness estimation and station number on dynamic inversion results were systematically examined.The results show that:1)the stability and spatial inversion characteristics of river roughness parameters can be effectively controlled by adjusting weight of roughness smoothness item;2)the inversed roughness values of channels near stations tend to be true,and the ones of channels far from stations tend to be smooth in spatial distribution;3)the inversed roughness values of channels near stations are less affected by the initial roughness estimation,while the ones of channels far from stations are more affected by the initial roughness estimation;4)the more accurate the initial roughness estimation and the more the number of stations,the smaller the systematic error of inversed roughness,the smaller the amplitude of fluctuation adjustment at the beginning of dynamic inversion,and the better the inversion effect.The proposed model can be effectively applied to the dynamic inversion of river roughness parameters in urbanization plain.