基于ICEEMDAN-SVD与IBES-TWSVM的OTDR事件识别算法

打开文本图片集

关键词：光时域反射计；ICEEMDAN；SVD；样本熵；信号降噪；IBES；TWSVM；事件识别中图分类号：TN929.1-34 文献标识码：A 文章编号：1004-373X（2025）17-0068-09

OTDR event recognition algorithm based on ICEEMDAN-SVD and IBES-TWSVM

SHI Haoming，CHEN Jun （CollgeofPhysicsand InformationEngineering，Fuzhou University，Fuzhou 35o1o8，China）

Abstract：Inresponsetotheloweventrecognitionaccuracyrateandlengthyprocesing timeinthe traditionalopticalcable faultanalysisandidentificationduetothefactsthattheOTDR（opticaltimedomainrefletometer）signalsareinterferedbynoise anddependentonartificial experience，anOTDReventrecognitionalgorithmbasedonthecombinationof ICEEMDAN-SVDand IBES-TWSVMis proposed.Firstly，ICEEMDAN（improvedcomplete ensembleempirical mode decompositionwithadaptivenoise） isutilizedtodecomposetheOTDRsignalsintomultipleIMFs （intrinsicmodefunctions），andthesampleentropyisemployedto distinguish betweenusefulandnoisecomponentswithintheIMFs.Next，theusefulcomponentsareretained，whilethenoise componentsaresubjectedtoSVD（singularvaluedecomposition）toextractanyresidualusefulsignals.Subsequently，theuseful signals frombothprocesingsarereconstructedtoobtaindenoisedOTDRsignalsFinally，animprovedBES （baldeaglesearch） algorithmisused tooptimizethepenaltyfactorandkernelfunctionparametersofTWSVM（twinsupprtvectormachine）.The optimizedTWSVMisthen trainedand testedbyafeaturedatasettoachieveOTDReventrecognition.Experimentalresults demonstrate thatwhenthenumberof the trainingsamplesis73Oandthatofthetestsamplesis312，thealgorithmachieves an eventrecognitionrateofapproximately95.52%inopticalcablelines，witharecognitiondurationof7.12seconds.Theproposed algorithmoutperformsthe traditionalOTDReventrecognitionalgorithmsin termsofboth eventrecognitionacuracyrateand recognition speed.

KeyWords： OTDR; ICEEMDAN; SVD; sample entropy; signal denoising; IBES; TWSVM; event recognition

0 引言

近年来，随着电力系统的快速发展，电力光缆凭借其承载信息量大、传递速度快和远距离传输稳定等优势，已成为电力通信系统的主要通信手段。（剩余11102字）