摘要
水面宽阔河段、洪泛区和潮汐河口段的水文及水动力学计算模拟多采用二维模型,然而因模型复杂的非线性映射关系常导致参数率定效率和精度不高,甚至出现“异参同效”问题。为此,提出一种应用BP神经网络自动率定二维模型糙率参数的方法。以海南昌化江入海口段为计算实例进行数值检验,结果表明:编写的Python程序能够准确率定HEC-RAS二维模型实例各糙率分区的糙率参数,各糙率分区的率定糙率值均在参数取值范围内,且均在观测糙率值±0.011内,未出现“异参同效”现象;水位率定计算误差满足设定允许误差要求,设定允许误差0.20 m、0.15 m和0.10 m的平均率定计算误差小于0.10 m,设定允许误差0.05 m的平均率定计算误差接近0 m;设定允许误差为0.15 m时,率定程序的运算效率最高。选用海南201409号台风“威马逊”时期的昌化江实测资料进行验证,证明该方法有效可靠,能够实现HEC-RAS软件二维模型糙率率定功能,可推广应用于其他模型参数率定。
2-D model is always adopted for most of the hydrological and hydrodynamic simulations and calculations made on the river reaches with wide water surface,floodplains and tidal estuaries,but the complicated nonlinear mapping relation of the model often leads to the problem of lower efficiency and accuracy of the parameter calibration and even the problem of“the different parameters with the same effects”.Therefore,a method of automatically calibrating the roughness parameter for 2-D model with BP neural network is put forward herein.By taking the estuary section of the Changhua River in Hainan Province as the calculation case,a numerical verification is made,and the result shows that the roughness parameters of all the roughness sub-zones of the HEC-RAS 2-D model for the actual case can be accurately calibrated with the written Python program and the calibrated roughness values of all the roughness sub-zones are within the parameter value-taking ranges,while all of them are±0.011 within the ranges of the observed roughness values without the phenomenon of“the different parameters with the same effects”.The error from the calculation of water level calibration can meet the requirement of the allowable specification error,while the mean calibration calculation error for the allowable specification errors of 0.20 m,0.15 m and 0.10 m are less than 0.10 m and the mean calibration calculation error for the allowable specification error of 0.05 m is close to 0 m.When the allowable specification error is 0.15m,the computational efficiency of the calibration program is the highest.The measured data of Changhuajiang River during the Super Typhoon Rammasun(No.201409)in Hainan Province are chosen for the verification,thus the method is verified to be effective and reliable,and then can realize the function of roughness calibration for the HEC-RAS 2-D model,which can be popularized and applied to the parameter calibration for the other relevant models.
作者
夏铭辉
秦景
牛文龙
雷添杰
XIA Minghui;QIN Jing;NIU Wenlong;LEI Tianjie(Beijing IWHR Corporation,Beijing 100048,China;China Institute of Water Resources and Hydropower Research,Beijing 100044,China)
出处
《水利水电技术》
北大核心
2020年第5期38-46,共9页
Water Resources and Hydropower Engineering
基金
国家重点研发计划(2018YFC1508203)。
