利用傅里叶谐波分析法的时序数据周期迭代辨识算法

Iterative identification algorithm for periodic component of time series data using Fourier analysis method

针对现有的时序数据其周期辨识算法存在着辨识精度低及计算成本高的问题,在傅里叶谐波分析法的基础上,提出了一种具有基频迭代机制的周期辨识算法。首先,利用方差分析法从原始序列中析出其周期长度的整型估算值;然后,以任意小的频率间隔在估算值区间内进行傅里叶谐波的迭代拟合;最后,基于最小拟合残差和的准则来确定最优的周期成分。实验表明,该算法不仅具有良好的计算效能,而且还能精确地辨识出与序列样本长度无关的周期成分。

In view of the existing time series data,the algorithm of periodic identification has the problems of low identification accuracy and high computational cost,on the basis of Fourier harmonic analysis,this paper proposed a periodic identification algorithm with fundamental frequency iteration mechanism. First of all,it used the method of analysis of variance from the original sequence in the precipitation period length integer estimation. Then,to the frequency interval arbitrarily small in the estimated value of iterative fitting,interval Fourier harmonic. Finally,to determine the optimal cycle components based on the criterion of minimum fitting error. The experimental results show that the algorithm not only has good computational efficiency,but also can identify the periodic components independent of the sample length accurately.