基于遗传熵谱估计的年径流周期识别

Genetic entropy spectral estimation method and its application to annual runoff periodic identification

为识别年径流量序列的隐含周期,提出基于加速遗传算法的熵谱估计算法,与传统的方差谱和Burg谱相比,该方法由熵谱分析的4个等价条件构建多目标函数,并以加速遗传算法作为优化算法,谱估计结果不依赖于初始值的选取,对数据长度、信噪比和初相位有较强的适应性。在三川河流域后大成站1956-2000年径流量序列周期识别中的应用结果表明,在95%的置信检验水平下,序列中存在着12.29年和2.67年的显著隐含周期,为三川河流域年径流的变化规律和变化的阶段性研究提供了一条新的定量研究手段。

An improved entropy spectral estimation method, Genetic Entropy Spectral estimation （GES）, is proposed to identify the implicit periods in annual runoff time series. The method is based on the accelerating genetic algorithm （AGA）, which is mainly used to optimize the parameters of maximtun entropy spectral analysis method （MESA）, and minimize the four equivalent conditions of MESA. Compared to the traditional variation spectral method and Burg spectral method, the entropy estimation results based on the improved method is not depend on the selection of initial value, further more, the method has high adaptability for data length, signal noise ratio and initial phases. Taking Houdacheng station in the Sanchuanhe River basin as a case, an annual runoff series from 1956 to 2000 is studied with the method. And results show that there are two prominent periods of 12.29 years and 2.67 years in the time series with 95% confidence level. GES method can provide a new approach for variation law and phases analysis study of runoff series.