摘要
为优化神经网络模型的应用效果,研究了基于神经网络的降雨-径流模型,根据Copula熵法确定预报因子,并与传统的线性相关法进行比较分析,采用BP、RBF、GRNN三种神经网络建立降雨-径流模型,应用均方根误差、合格率、确定性系数三个指标为模型选取评价准则。通过对金沙江流域的径流预报,发现基于Copula熵法的BP模型预报结果更接近实测值,精度更高。
To optimize the applied effect of neural network model,this paper studies rainfall-runoff model based neural network.Compared with the traditional linear correlation method,Copula entropy method is used to determine prediction factors.BP（Back Propagation） neural network,RBF（Radical Basis Function） neural network and GRNN（Generalized Regression NN） neural network are adopted to establish the rainfall-runoff model.And then,the root-mean-square error（RMSE）,the qualification rate（QR） and the deterministic coefficient（DC） are taken as the evaluation criteria for selection of model.Forecasting runoff of Jinshajiang Basin indicates that the forecasting results of BP model based Copula entropy method is closer to the observed value and it has higher precision.
出处
《水电能源科学》
北大核心
2013年第6期21-25,共5页
Water Resources and Power
基金
国家自然科学基金资助重点项目(51239004)
华中科技大学研究生创新创业基金资助项目(HF-11-05-2013)
华中科技大学国家级大学生科技创新训练基金资助项目
关键词
神经网络模型选取
水文预报
预报因子选择
Copula熵
selection of neural network model
hydrological forecasting
selection of prediction factors
Copula entropy
