收稿日期:2019
-
05
-
23ꎮ
基金项目:福建省自然科学基金资助项目(2016J01734)ꎮ
作者简介:金保明(1970—)ꎬ男ꎬ高级工程师ꎬ博士ꎬjbm720@ 126.comꎮ
引文格式:金保明ꎬ李斐斐ꎬ周传争ꎬ等.山区流域 RBF 神经网络洪水预报方法[ J].南昌大学学报(工科版)ꎬ2019ꎬ41(4):
371
-
376.
文章编号:1006
-
0456(2019)04
-
0371
-
06
山区流域 RBF 神经网络洪水预报方法
金保明
1
ꎬ李斐斐
1
ꎬ周传争
2
ꎬ王伟
1
ꎬ卢光毅
1
(1.福州大学土木工程学院ꎬ福建 福州 350116ꎻ2.浙江省钱塘江管理局勘测设计院ꎬ浙江 杭州 310016 )
摘要:建立以高斯核函数为径向基函数的神经网络模型ꎬ以崇阳溪山区流域为例进行分析ꎮ 利用泰森多边形
法将流域划分为 6 个子流域ꎻ选取流域 1997—2014 年中 15 场洪水过程作为训练样本ꎬ以流域内 6 个雨量站各时段
雨量资料和武夷山水文站前期流量过程作为输入ꎬ以流域出口断面武夷山水文站相应流量过程作为输出ꎬ运用自
组织正交最小二乘法确定径向基函数的中心ꎬ采用伪逆规则求解其权值ꎬ建立 RBF 神经网络洪水预报模型ꎻ利用余
下的 6 场洪水过程对模型进行检验ꎮ 结果表明:各场次洪水流量过程平均相对误差和洪峰流量相对误差绝大部分
在 10%以内ꎬ确定性系数均大于 0.9ꎬ预报精度符合要求ꎬ可以为防汛部门预测洪水提供依据ꎮ
关键词:径向基函数ꎻ神经网络ꎻ洪水预报ꎻ山区流域
中图分类号:TV124ꎻP338 文献标志码:A
Research on flood forecasting model of radial basis
function neural network in mountainous basin
JIN Baoming
1
ꎬLI Feifei
1
ꎬZHOU Chuanzheng
2
ꎬWANG Wei
1
ꎬLU Guangyi
1
(1.College of Civil EngineeringꎬFuzhou UniversityꎬFuzhou 350116ꎬChinaꎻ
2.Survey and Design Institute of Qiantang River Administration of Zhejiang ProvinceꎬHangzhou 310016ꎬChina)
Abstract:A neural network model with Gauss kernel function as radial basis function was establishedꎬand the
mountainous watershed of Chongyang River was taken as an example in this paper.Firstlyꎬthe watershed was divided
into six sub ̄watersheds by Thiessen polygon method.And then 15 flood processes in the basin from 1997 to 2014
were selected as training samplesꎬthe rainfall data of six rainfall stations in the basin and the pre ̄discharge process
of Wuyishan Hydrological Station were taken as inputꎬand corresponding discharge process of Wuyishan Hydrologi ̄
cal Station at the outlet section of the basin was taken as output.Self ̄organizing orthogonal least squares method was
used to determine the center of radial basis functionꎬthe pseudo ̄inverse rule was used to solve its weightꎬand then
RBF neural network flood forecasting model was established.Finallyꎬthe remaining six flood processes were used to
test the model.The results showed that the average relative error of each flood discharge process and the relative er ̄
ror of peak discharge were mostly less than 10%ꎬthe deterministic coefficient was greater than 0.9ꎬand the predic ̄
tion accuracy met the requirementsꎬwhich could provide a basis for flood control departments to predict flood.
Key Words:radial basis functionꎻneural networkꎻflood forecastingꎻmountainous basin
山区流域降雨强度大ꎬ降雨集中ꎬ形成的洪水具
有暴涨暴落的特点ꎬ经常引发洪涝灾害ꎬ难于预防和
预报ꎬ往往给当地带来巨大的损失ꎮ 山区中小流域
洪水预报显得至关重要ꎮ 水文预报是流域主要的防
洪非工程措施之一ꎬ做好山区中小流域的实时水文
预报工作ꎬ增长洪水预报预见期ꎬ及时制定相应的防
汛指挥决策ꎬ积极主动采取防御措施ꎬ最大程度减少
洪涝灾害的影响ꎬ减少经济损失ꎬ对促进国民经济的
第 41 卷第 4 期
2019 年 12 月
南昌大学学报(工科版)
Journal of Nanchang University(Engineering & Technology)
Vol.41 No.4
Dec.2019
评论0