基于数据挖掘的河网糙率直接反演方法

Direct inversion method of river network roughness based on data mining

糙率是河网水力计算中的重要参数,目前多采用优化方法反演求得。基于数据挖掘的思想,本文提出一种新的河网糙率直接反演方法。该方法以河段测量值为输入,糙率为输出,应用BP神经网络进行直接反演。在一个9河段组成的河网算例中,运用这种方法得到的各河段反演糙率最大误差不超过1.5%,并用Bayesian方法对反演结果的唯一性和最优性进行了评价。

The roughness usually obtained based on is a very important parameter in hydraulic calculation of river network, the value is conventional optimization inversion In this paper, a new method based on data mining is presented to inverse the river network roughness directly. The BP neural network is introduced in direct inversion with measured data of flow as input value and river roughness as output value. In a computing example of river network composed of nine channels by using the direct inversion method, the maximal error is less then 1.5 percent. Meanwhile, the Bayesian method is used to evaluate the multiplicity and optimality of inversion results.