月降水序列混沌特性识别及预测

Chaos Characteristics Identification and Prediction of Monthly Average Precipitation

以泾河流域西峰镇站1951-2001年月降水时间序列为基础，利用互信息函数法求时间序列的时间延迟，引入伪邻近法求该序列的嵌入维数，计算结果显示其嵌入维数在[2，12]之间变化，再利用G-P算法计算该序列的关联维数，结果表明序列具有明显的混沌特性。用Volterra级数一步预测方法对混沌时间序列进行预测，实验表明，在训练数为220、样本数为392时预测结果较优。

Based on the monthly average precipitation at Xi-Feng town from 1951 to 1999, this paper gets the delay time using mutual information method, and this paper also introduces Cao method to solve the embedding dimension using Cao method. The result of delay time is 3,and the result of embedding dimension is [2,12]. Based on the delay time and the embedding dimension, this paper gets the correlation dimension using classical GP method, the result is 2. 1548. It demonstrates that the monthly average precipitation series have chaos characteristics. Then, we use Volterra Series step forecast method to predict the series. This paper gets the better result under the training number 220 and the sample number 392, and the correlation coefficient is up to 0. 6367.