基于模极大值小波域和TLS-ESPRIT的振荡瞬态识别方法

对于电能质量扰动检测和定位中振荡瞬态的检测、识别，目前普遍采用的是时频特征矢量提取和智能模式识别方法，此类方法无法准确提取电能质量振荡瞬态信号不同频率分量的组成。结合模极大值小波域和总体最小二乘法旋转不变技术的信号参数估计(total least squares-estimation of signal parameters via rotational invariance techniques，TLS-ESPRIT)可以很好地实现振荡信号的检测与识别。对于输入信号，首先采用模极大值小波域检测振荡发生的起始时刻和终止时刻，然后利用振荡时间间隔内的信号建立观测空间矩阵，通过奇异值分解和总体最小二乘法实现特征值截尾，将采样信号观测空间分解为信号子空间和噪声子空间，得到振荡信号每个构成频率分量的相应参数。仿真结果证实了所提出方法的可行性。

Identification of Oscillatory Transients Based on Modulus Maximum Wavelet Domain and TLS-ESPRIT

At present the time domain eigenvector extraction and intelligent pattern recognition are widely applied to oscillatory transients detection and recognition for power quality disturbances detection and location, however, these methods cannot precisely extract the composition of different frequency components in power quality oscillatory transient signals. The detection and recognition of oscillatory signals can be well implemented by combining modulus maximum wavelet domain with total least squares-estimation of signal parameters via rotational invariance techniques (TLS-ESPRIT). For input signals, firstly the initial time and the end time of the oscillation are detected by modulus maximum wavelet domain; then using the signals within the time interval of the oscillation the observation space matrix is built, and by means of singular value decomposition and total least squares method the truncation of eigenvalue is implemented; and then the observation space of sampled signals is resolved into signal subspace and noise subspace, thus the corresponding parameters of each frequency component that composes the oscillatory signals are obtained. Simulation results verify the feasibility of the proposed method.