基于K-means聚类和同步挤压小波变换的次同步振荡检测

Sub-synchronous oscillation detection based on K-means clustering and frequency synchrosqueezing wavelet transforms

电力系统次同步振荡具有非平稳、非线性特性,现有检测方法难以捕获振荡特征和变化趋势,为此提出K-means结合SWT的振荡检测方法,并将该方法引入次同步振荡谐波检测分析中。首先,利用SWT较强的抗模态混叠能力和抗噪性,在噪声环境下清晰直观表征信号振荡模态。同时,在SWT中运用频域切片,提取电力信号中的多重振荡模态,进行重构和参数辨识。考虑到SWT将小波系数挤压至中心频率,采用K-means聚类方法准确求出重构前振荡信号中心频率,并自动选择信号重构频域区间。最后,通过仿真算例验证该方法的有效性。

Power system has non-stationary and nonlinear characteristics of the sub-synchronous oscillation(SSO),it is difficult for existing detection methods to capture the oscillation characteristics and the changing trend.In this paper,an oscillation detection approach,which combines K-means clustering and synchrosqueezed wavelet transform(SWT),is proposed to achieve the harmonic detection and analysis of subsynchronous oscillation.The anti-modal aliasing ability and anti-noise ability of the SWT are utilized to clearly and intuitively show the oscillation modes of the signals with noise.The frequency domain slicing is employed in the SWT to extract multiple oscillation modes for the reconstruction and the parameter identification.Considering that the SWT will squeeze the wavelet coefficients to the central frequency,the K-means clustering method is applied to calculate the central frequency of the oscillating signal.At the same time,the frequency interval of the signal can be automatically selected for reconstruction.Finally,the simulations are conducted to examine the effectiveness of the proposed method.