基于深度学习的电力设备异常声音检测技术分析.pdfVIP

S标准与检测电声技术

tandardsTestingudioEngineering

：基于深度学习的电力设备异常声音检测技术分析[J].电声技术，2025，49（2）：190

文献引用格式：沈翔宇：193.

SHENXY.Analysisofabnormalsounddetectiontechnoloelectricpowerequipmentbasedondeeplearning[J].Audio

Engineering,49(2):190-193.

中图分类号：TP183文献标识码：ADOI:10.16311/j.audioe.2025.02.059

基于深度学习的电力设备异常声音检测技术分析

沈翔宇

（江苏省供电公司苏州供电分公司，江苏苏州201800)

摘要：针对传统电力设备异常声音检测方法存在的主观性强、效率低等问题，提一种基于深度学习的异常声音检测技术。该

技术通过声学特征增强预处理、改进型深度卷积神经网络特征提取及集成学习多模态决策融合3个环节，智能识别电力设备

异常声音。实验结果表明，该技术在不同负载工况下均表现出优异性能，检测精度在90%以上，可为电力设备状态监测提供

新思路。

关键词：电力设备；异常声音检测；深度学习

AnalysisofAbnormalSoundDetectionTechnologyofElectricPowerEquipmentBasedonDeepLearning

SHENXiangyu

(JiangsuProvincePowerSupplyCompanySuzhouPowerSupplyBranch,Suzhou201800,China)

Abstract:Inordertosolvethesubjectiveandlow-efficiencyproblemsintraditionalelectricalpowerequipmentabnormalsound

detectionmethods,anovelabnormalsounddetectiontechnologybasedonDeepLearningisproposed.Thistechnologyincludesthree

stages:soundfeatureenhancementpreprocessing,improveddeepconvolutionalneuralnetworkfeatureextraction,andmulti-modal

decisionfusionthroughintegratedlearning.Itintelligentlyidentifiesabnormalsoundsinelectricalpowerequipment.Experimental

resultsshowthatthetechnologyhasexcellentperformanceunderdifferentloadconditions,withdetectionaccuracyabove90%.Itcan

provideanewideaforelectricalpowerequipmentconditionmonitoring.

Keywo