基于IEWT和噪能转移SR-MLS反演识别技术的低频振荡信号分析

电力系统低频振荡信号是典型的多分量混噪信号,特征提取较困难。为此,采用基于顺序统计滤波原理(order statistics filter,OSF)平顶上包络的改进经验小波变换算法(improved empirical wavelet transform,IEWT)和随机共振—移动最小二乘(resonance-moving least squares,SR-MLS)反演识别技术相结合的方法实现对振荡信号的特征分析。IEWT结合了小波分析的完备理论性和经验模态分解的自适应性,通过构造一系列正交小波滤波器组对信号进行分解。首先,根据OSF最大值滤波器原理得到频谱的有效平顶上包络,进而确定EWT的边界并对原始振荡信号进行抗噪主导模态分离,然后结合SR-MLS反演识别技术,在残余噪声的帮助下增强振荡特征并有效提取。最后,在自合成模拟信号仿真、IEEE 16机68节点系统仿真以及电网实测数据3个算例仿真下通过与经典Prony法、VMD-Hilbert法对比,表明了所提方法的可行性及有效性。

Analysis of Low Frequency Oscillatory Signals by IEWT and Energy Transfer SR-MLS Inversion Recognition Techniques

The low frequency oscillatory signal of power system is a typical multi-component mixed noise signal, difficult to extract features. In this paper, characteristic analysis of oscillatory signal is performed by combining improved empirical wavelet transform(IEWT) algorithm based on order statistics filter(OSF) flat top envelope and stochastic resonance-moving least squares(SR-MLS) inversion recognition technology. IEWT combines the complete theory of wavelet analysis and adaptability of empirical mode decomposition, and decomposes the signal by constructing a series of orthogonal wavelet filter banks. Firstly, according to the principle of OSF maximum filter, an effective flat-top envelope of the spectrum is obtained, and then the IEWT boundary is determined and anti-noise dominant mode separation of the original oscillatory signal is carried out. Secondly, combined with SR-MLS inversion identification technology, the oscillatory features are enhanced and extracted effectively with the help of residual noise. Finally, feasibility and effectiveness of the proposed method are shown by comparing with classical Prony method and variational mode decomposition(VMD) and Hilbert method in three examples: self-synthesis analog signal simulation, IEEE 16-machine 68-bus system simulation and actual measured data of a power grid in China.