基于小波预处理技术的低频振荡Prony分析

Prony analysis of low frequency oscillation based on wavelet pretreatment technology

提出了一种利用改进的小波软阈值去噪技术,首先对电力系统低频振荡数据进行预处理,然后采用Prony算法提取低频振荡信号特征的分析方法。在分析Prony算法原理的基础上,分析了参数选择对算法的分析速度与精度有较大影响,提出了Prony算法主要参数的选择策略,即信号抽样频率应大于信号最高频率的2倍,以避免频谱混叠;信号时间长度应包含2个周期最低频率的振荡,以提高参数估计精度;模型初始阶数应远大于信号中实际包含的指数项个数,以使最优子集分量逼近观察到的数据。仿真和动模实验结果表明,基于小波预处理技术的Prony算法具有分辨率高、拟合效果好的优点,能满足电力系统低频振荡特征分析的需要。

An analysis method is presented,which uses improved wavelet soft- threshold denoising technology to preprocess the data of low frequency oscillations,and then Prony algorithm to extract the oscillation character. Based on the basic principle of Prony algorithm,the greater influence of parameter selection on the velocity and accuracy is analyzed,and the selection strategies of main parameters are presented. To avoid the frequency spectrum alias,the sampling frequency should exceed two times of the highest frequency of signal. To improve the estimative accuracy,the analysis time should cover two cycles of the lowest frequency of signal. To approach the observed data through the optimal subsets, the initial rank of the model should outclass the number of the exponential parts of signal. Simulations and dynamic experiments show that,Prony algorithm based on wavelet pretreatment technology has high resolution and good fitting performance,meeting requests of character analysis for low frequency oscillations in power systems.