Fault Location Method of Power Equipment Based on MF-DNN Algorithm Using Acoustic Signal
曾彦超 1郑耀华 1金仲铂 1李玟璇 1李梓玮 1陈西 1袁超2
扫码查看
点击上方二维码区域，可以放大扫码查看
作者信息
1. 南方电网广东肇庆供电局,广东 肇庆 526060
2. 湖南大学 电气与信息工程学院,湖南 长沙 410012
折叠
摘要
电力设备故障声学定位面临严重的背景噪声干扰,导致波达方向(direction of arrival,DOA)估计精度和实时性不佳,故障定位性能受限.对此,提出一种基于匹配滤波器和深度神经网络(match filter-deep neural networks,MF-DNN)的DOA估计技术,在低信噪比情况下,实现基于声学检测的高速电力设备故障声学定位.该方法通过匹配滤波器提高波达信号的信噪比,提取上三角元素特征降低了特征维度,采用自适应矩估计算法优化网络更新策略,降低了网络规模,最终实现低信噪比条件下对特征目标的高速DOA估计.实验结果表明,该方法在低信噪比下的估计速度和估计精度均优于传统算法.
Abstract
When using acoustic signals to locate faults in the power equipment,the system often experiences severe background noise interference,resulting in low accuracy of direction of arrival(DOA)estimation and slow localization speed.Therefore,this paper proposes a DOA estimation technique based on matched filters and deep neural networks(MF-DNN).When the signal to noise ratio(SNR)of acoustic signals is very low,this method achieves high-speed acoustic localization of power equipment faults.The method improves SNR of the arrival signal through a matched filter module and reduces the feature dimension by extracting the features of the upper triangle elements.It optimizes the network update strategy through the adaptive moment estimation(Adam)algorithm,reduces the network size,and ultimately achieves high-speed DOA estimation of feature targets under low SNR conditions.The experimental results show that the estimation speed and accuracy of this method are superior to traditional algorithms under low SNR acoustic signals.
关键词
故障定位/波达方向估计/匹配滤波器/深度神经网络/低信噪比
Key words
fault location/direction of arrival/DOA/matched filter/deep neural network(DNN)/low signal to noise ratio
维普
万方数据