基于声阵列与SHO-VMD-FastICA的750kV电抗器声纹分离方法研究及应用

WANG Guo; LI Baopeng; MIN Yongzhi; HE Yigang; HE Jianshan; HUO Yichen

doi:10.13336/j.1003-6520.hve.20241442

English Version Journal Login

您当前的位置：

首页 >

文章列表页 >

基于声阵列与SHO-VMD-FastICA的750kV电抗器声纹分离方法研究及应用

更新时间：2026-09-16

- 基于声阵列与SHO-VMD-FastICA的750kV电抗器声纹分离方法研究及应用
- Vol. 51, Issue 9, Pages: 4588-4598(2025)
- 作者机构：
  
  1. 兰州交通大学自动化与电气工程学院,兰州,730070
  2. 武汉大学电气与自动化学院,武汉,430072
- 作者简介：
- 基金信息：
- DOI：10.13336/j.1003-6520.hve.20241442
  CLC：
- Received：30 September 2025，
  
  Published：2025
- 稿件说明：
移动端阅览
WANG Guo, LI Baopeng, MIN Yongzhi, et al. 基于声阵列与SHO-VMD-FastICA的750kV电抗器声纹分离方法研究及应用[J]. 2025, 51(9): 4588-4598.
DOI：

WANG Guo, LI Baopeng, MIN Yongzhi, et al. 基于声阵列与SHO-VMD-FastICA的750kV电抗器声纹分离方法研究及应用[J]. 2025, 51(9): 4588-4598. DOI： 10.13336/j.1003-6520.hve.20241442.

摘要

750 kV电抗器实测声纹信号中包含高低频段的干扰噪声，严重影响电抗器声纹识别精度，需要进行电抗器本体声纹分离。首先，采用声阵列采集750 kV电抗器声纹信号，并通过数据一致性检验算法对各阵元进行甄别，筛选有效阵元数据构建观测信号矩阵；然后，利用海马优化算法对变分模态分解算法的分解层数K与惩罚因子α进行寻优，进而根据最优参数组合[K，α]分解观测信号，并选取1 kHz以内的模态分量重构观测信号，实现高频干扰噪声的滤除；最后，采用快速独立成分分析法从多维降噪观测信号中分离出750 kV电抗器本体声纹信号。采用该方法分离所得750 kV电抗器本体声纹信号与对照信号的相似系数为0.954 2，信噪比为10.548 2 dB，与SCA算法相比分别提升6.63%和2.282 7 dB，表明该方法具有更高的适用性和准确性。

Abstract

The measured voiceprint signals of 750 kV reactor contain interference noises in different frequency bands

which affect the accuracy of voiceprint recognition of the reactors

and it is necessary to separate the voiceprint from the collected signals. This study employed an acoustic array to collect voiceprint signals of 750 kV reactor

optimized the decomposition layer number K and penalty factor α of the variational mode decomposition algorithm using the Sea-Horese optimization algorithm

removed high-frequency interference

and used the fast independent component analysis algorithm to separate the voiceprint of the reactor body. The correlation coefficient between the voiceprint signal of the 750 kV reactor body separated by the method in this article and the comparison signal is 0.954 2

and the SNR is 10.548 2 dB. Compared with the SCA algorithm

it improves by 6.63% and 2.282 7 dB respectively

indicating that the method in this article has higher applicability and accuracy.

关键词

Keywords

references

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

基于LSTM-Transformer-BiLSTM的串并行覆冰预测方法

Mechanical Fault Diagnosis of On-load Tap-changer Based on ANHGA-VMD and Coupled Hidden Markov Model

Prediction Model of Dissolved Gas in Transformer Oil Based on Variational Modal Decomposition and Recurrent Neural Network with Gated Recurrent Unit

Mechanical Fault Diagnosis Method of Circuit Breaker Based on Blind Source Separation and NSVDD Under Noisy Background

Short-term Power Load Forecasting Based on Extreme Gradient Boosting and Temporal Convolutional Network

Related Author

ZHAO Hongfeng

ZHANG Jiansheng

CAO Zhuangzhuang

HE Shan

DONG Xinsheng

ZOU Delei

HUANG Wenquan

PENG Changqing

Related Institution

School of Electrical Engineering, Xinjiang University

Electric Power Science Research Institute of State Grid Xinjiang Electric Power Company

Engineering Research Center of Ministry of Education for Renewable Energy Power Generation and Grid Connection Control (Xinjiang University)

Kuitun Electric Power Supply Company, State Grid Xinjiang Electric Power Company, Kuitun

Kashi Electric Power Supply Company, State Grid Xinjiang Electric Power Company

AI问答

⁰