基于随机矩阵的变电站多源异常声音信号的定向方法

Method of Locating Multi-source Abnormal Acoustic Signals of Substation Equipment Based on Random Matrix

变电站的安全稳定运行至关重要,通过分析变电站中电力设备的可听声特征变化可以判断设备是否处于不正常运行状态,对异常声源进行主动定位可以快速确定故障区段,进而提高异常声音的检测效率。在实际应用中,由于准确性和实时性的要求常采用大阵列小快拍系统,造成当多个目标声源角度相近时,传统的声音定位方法不能对其有效分辨。根据随机矩阵理论提出基于均匀平面矩阵的声源定位方法,利用Stieltjes变换校正样本协方差矩阵的特征值和特征向量,然后通过空间谱的谱峰搜索对异常声源定位,可以有效提高分辨相邻声源的成功率。研究结果表明,所提方法可以使用更少的快拍数分辨相邻声源,且分辨成功率得到有效提升。

The safe and stable operation of substation is very important.Whether the power equipment is at abnormal operation condition can be judged by way of analyzing the variation of the audible sound characteristics of the equipment in substation.The active location of abnormal sound source can quickly determine the fault section,and thus improve the detection efficiency of abnormal sound.In practical application,the small snapshot system with large array is often used due to the requirements of accuracy and real-time,which makes the traditional sound location methods not effectively distinguish multiple targets when the sound source angles are close.According to the random matrix theory,a sound source location method based on uniform plane matrix is proposed.The eigenvalues and eigenvectors of the sample covariance matrix are corrected by Stieltjes transform,and then the abnormal sound source is located through the spectral peak search of spatial spectrum,which can effectively improve the success rate of distinguishing adjacent sound sources.The research results show that the proposed method can use less snapshots to distinguish adjacent sound sources,and the resolution success rate is effectively improved.