利用夹角余弦和聚类分析的电离层TEC混沌预测
Chaotic Properties and Prediction of Ionospheric TotalElectron Content Based on Cosine and Cluster Analysis
-
摘要: 目的 利用120°E、45°N上空的2008年年积日101～150d时间段内共600个电离层格网TEC数据,分析了该点上空电离层 TEC参数的混沌特性,发现其关联维数为2.263 2,嵌入维数 m=5,最大 Lyapunov指数为0.083 3,该 TEC时间序列具有混沌的特征,存在混沌现象。利用加权一阶局域法对 TEC时间序列进行预测时,提出了利用夹角余弦和聚类分析方法对相似相点进行选择的方法,结果表明,在5维相空间中,该方法除在第4分向量略不及欧氏距离和夹角余弦方法外,其余4个分向量均优于后两种方法。利用该方法选择的相似相点进行一阶局域预测时,得到的标准差STD(0.618TECU)和 RMS(0.623TECU)均小于欧氏距离和夹角余弦得到的STD和 RMS,说明该方法可以准确地搜索到与基准点相关性更强的相似相点,预测精度更高。Abstract: Objective The paper analyzes the chaotic properties of ionospheric total electron content based on the600TEC data from 101～150din 2008over 120°E and 45°N.Calculation results show that the TECtime series is chaotic when correlation dimension is 2.263 2,embedding is 5and largest Lyapunov ex-ponent is 0.083 3.A new method to choose the more similar phase points is advanced based on cosineand cluster analysis while the weighted one-rank local-region forecasting method is used sucessfuly toforecast the time series of TEC.The search results show that the paper’s method can hunt more simi-lar phase points than the euclidean distance and cosine in 5dimensional space.Forecasting resultsshow that the STD(0.618TECU)and RMS(0.623TECU)based on cosine and cluster analysis areless than the other two methods.Therefore,the method can choose by rule and line similar phasepoints and improve forecasting precision.
下载:
