年径流时间序列的混沌分析

在介绍重构相空间技术的主要定量指标(关联维数D2和柯尔莫奇诺夫(Kolmogorov)熵)的基础上，针对长江上游绵阳地区和岷江上游紫坪铺水文站的年径流时间序列，探讨了不同嵌入维m下其关联维数的变化规律。得到绵阳地区年径流时间序列的饱和关联维D2=4.11，最低嵌入维m=8，Kolmogorov熵k= 0.303；紫坪铺年径流时间序列的饱和关联维D2=2.58，最低嵌入维m=5，Kolmogorov熵k= 0.302；并且得到两个年径流序列的最大预测年限为4a，为年径流预测提供了较为科学的依据。

Chaos Analysis of Annual Runoff Time Sequence

Based on introducing the main quantitative indexes of correlation dimension D2 and Kolmogorov entropy k in rebuilding time sequence imbedding space, the change rules of correlation dimension De at dofferemt built-in dimension m are discussed through analyses of the annual runoff time sequence in Mianyang region of upstream Yangtze River and the annual runoff time sequence in Zipingpu Hydrological Station of upstream Mingjiang River, For the annual runoff in Mianyang region, the saturation correlation dimension, minimum built-in dimension, and Kolmogorov entropy are calculated, which areD2 = 4. 11, m= 8 and K=0. 303, respectively; For the annual runoff in Zipingpu Hydrological Station, the results areD2 =2.58, m= 5 and K=0. 302, respectively. And it is obtained that the maximum forecasting length for the two annual runoff time sequences should be 4 years. The two time series chaotic analyses provide a scientific gist for annual runoff forecasting.