利用电缆HFCT高速采样的系统暂态扰动辨识

System Transient Disturbance Identification Using Cable HFCT High Speed Sampling

分散分布于电力系统各处的暂态扰动不易被变电站监测设备捕捉,因此提出充分利用电缆附件处装设的高频电流传感器(high frequency current transformer,HFCT)作为高速信号采集源,从而实现暂态扰动信号辨识。首先利用真实的HFCT样品进行暂态扰动信号采样实验,论证了HFCT采集暂态扰动信号的可行性和有效性。然后通过深入分析暂态扰动信号经HFCT滤波后高频突出、信号微弱的具体特征,提出一种基于多重时频特征矩阵的暂态扰动辨识方法。该方法利用参数优化的变分模态分解,将高频微弱波形信号解构为不同中心频率的多重分量,从中提取出波形特征矩阵。接着对不同频率分量做进一步Wigner-Ville分布时频分析,得到时频图谱并提取各分量的时频图谱特征矩阵。最后,融合波形特征矩阵与时频图谱特征矩阵构造出多重时频特征矩阵,将其作为长短期记忆网络的输入,实现暂态扰动信号的分类辨识。实验数据测试表明,该文提出的多重时频特征矩阵能充分挖掘高频微弱信号的局部细节,适用于经HFCT滤波后的暂态扰动信号分类辨识,且抗噪性能较好,为暂态扰动信号分类辨识提供了新的监测思路与方法。

The transient disturbance signals scattered throughout the power system are not easily captured by the substa-tion monitoring equipment.Therefore,high frequency current transformers(HFCTs)installed in the cable accessories were used as the source of high-speed signal acquisition to realize the identification of transient disturbance signals in this paper.Firstly,transient disturbance signal sampling experiments were conducted using real HFCT samples,demonstrating the feasibility and effectiveness of HFCT in collecting transient disturbance signals.Then,by in-depth analysis of the spe-cific characteristics of the transient disturbance signal with prominent high frequency and weak signal after being filtered by the HFCT,a transient disturbance identification method based on multiple time-frequency feature matrices was pro-posed.A parameter-optimized variational modal decomposition was applied to deconstruct the high-frequency weak waveform signal into multiple components with different center frequencies,in which waveform characteristic matrix could be extracted.Besides,Wigner-Ville distribution time-frequency analysis was performed on different frequency components to obtain the time-frequency spectrum and to extract the time-frequency spectrum feature matrix of each component.Finally,the waveform feature matrix and the time-frequency spectrum feature matrix were combined to con-struct a multiple time-frequency feature matrix,which was input into the long short-term memory network for transient disturbance signal classification and identification.Tests using experimental data illustrate that the multiple time-frequency feature matrix proposed in this paper can fully mine the local details of high-frequency weak signals,and it is suitable for the classification and identification of transient disturbance signals filtered by HFCT with dramatic an-ti-noise performance.The method proposed in this paper provides a novel monitoring idea and method for the classification and identification of transient disturbance signals.