基于改进频移经验模态分解的低频振荡参数提取

对于规模越来越大的复杂电力系统来说,采用基于量测数据的低频振荡研究方法日益受到重视。经验模态分解(EMD)方法的分解过程具有自适应且适于分析非平稳信号,在低频振荡参数提取方面应用较多,但EMD方法存在模态混叠等现象。当信号中2个单频分量的频率在2倍频内时,频移经验模态分解(FS-EMD)可将2个分量分解开。但当信号中有多个单频分量的频率在2倍频内时,FS-EMD就无法分解。为了提高EMD的频率分辨率并使分解方法具有通用性,文中提出了改进的频移经验模态分解(RFS-EMD)算法。此方法增大了信号中组成分量的频率比,且保证频率不翻转,使之可循环使用RFS-EMD算法分解复杂信号。该方法在应用于电力系统低频振荡模态参数的提取时,能较好地提取多个2倍频范围内的低频振荡模态分量的频率、幅值、相位及阻尼比等参数。数值仿真和实例分析均表明了该方法的有效性。

Extraction of Low Frequency Oscillation Parameters Based on Refined Frequency Shift Empirical Mode Decomposition

For increasing the complex power grid, low-frequency oscillation research methods based on measurement are more and more attention. Among these methods, the fitting result of prony method is sensitive to noise, and the order of the model is difficult to be selected. The wavelet function is not easy to determine. EMD is a multi-resolution signal-processing method, it can be used non-stationary signals analysis. EMD is more and more used in extraction of the low-frequency oscillation parameters, but it has disadvantages such as model mixing. As more than two individual components in a signal with frequencies within an octave is indecomposable by frequency shift empirical mode decomposition (FS-EMD) method, to improve the frequency resolution and let FS-EMD more general, a refined frequency shift EMD(RFS-EMD) is presented. The proposed method is able to enlarge the component frequency ratios, and to ensure that the frequency does not turn over, so that the complex signal can be decomposed through repeating call this method. In its application to extracting model parameters of low frequency oscillation in power system, this method proves fairly effective in extracting model parameters in a signal with several frequencies within an octave respectively, such as, frequency, amplitude, phase and damping ratio, etc, as shown by numerical simulation and case analysis results.