基于Copula熵的神经网络径流预报模型预报因子选择

Determination of input variabes for artificial neural networks for flood forecasting using Copula entropy method

采用神经网络进行水文预报的关键问题之一是预报因子(输入变量)的选择,目前国内尚缺有效、系统的理论方法,国外主要是采用偏互信息(Patial mutual information,PMI)法。本文针对偏互信息计算方法的缺陷,引入Copula熵的概念,推导Copula熵与互信息的关系,提出采用Copula熵计算PMI;并借助模拟试验检验了所提方法的合理性;最后,将该方法应用到三峡水库的水文预报中,并与现行方法进行了比较分析。结果表明,本文所提方法不仅具有理论基础,而且结果合理可信。

One of the key steps in artificial neural networks （ANN） forecasting is the determination of significant input variables. A partial mutual information （PMI） method was used to characterize the dependence of a potential model between its input and output variables. We also developed a copula entropy method for effective calculation of mutual information （MI） and PMI, and verified its accuracy and performance using numerical tests. This forecasting technique has been applied to a real-world case study of the Three Gorges reservoir （TGR）, and results show that the proposed method is useful and effective for identification of suitable inputs of flood forecasting model.