文档名：基于变分模态分解和改进灰狼算法优化深度置信网络的自动转换开关故障识别
摘要：自动转换开关(ATSE)是保证系统连续供电的设备,对其进行健康监测和故障诊断对系统的稳定运行具有重要意义.为了实现对ATSE的非侵入式故障识别,该文提出一种基于电流信号变分模态分解(VMD)的特征提取和改进灰狼算法(IGWO)优化深度置信网络(DBN)相结合的故障诊断方法.该方法首先利用样本熵确定VMD分解次数并对故障电流进行分解;其次对分解后得到的本征模态函数进行小波包能量的提取,并利用IGWO对DBN网络结构参数进行优化;最后通过DBN将电流能量特征与ATSE的故障类型建立起映射关系从而完成最终的故障识别.所提IGWO采用了分段调节与非线性递减的衰减因子相结合的策略,以平衡算法全局搜索和局部搜索能力;并采用莱维飞行更新探狼的移动位置,来避免算法陷入早熟收敛.实验结果表明,该算法不仅能显著提高前期对参数寻优的训练速度,后续泛化实验的故障分类准确率也有98.78%的良好表现.
Abstract：AutomaticTransferSwitchingEquipment(ATSE)isadevicethatensuresthecontinuouspowersupplyofthesystem.FaultdiagnosisofATSEisofgreatsignificanceforthecontinuousoperationofthesystem.ThecurrentdataofsolenoidcoilscanbeusedtoidentifytheelectricalandmechanicalfaultsofATSE.However,thenoiseinthecurrentandthenetworkstructureparametersoftheintelligentfaultdiagnosismethodaredifficulttoconfirm.Therefore,anATSEfaultdiagnosismethodisproposedbasedonvariationalmodedecomposition(VMD)featureextractionandoptimizeddepthbeliefnetwork(DBN).Inthismethod,WaveletPacketEnergy(WPE)isextractedasafeaturevectorfromthemodecomponentsdecomposedbyVDM,andthenetworkstructureparametersofDBNaresetbyanimprovedgreyWolfalgorithm(IGWO)tocompleteATSEfaultidentification.Firstly,thefaultcurrentsignalisdecomposedbyVMD,anditssampleentropiesareobtainedforeachmodecomponentafterdecomposition.ThedecompositionnumbercorrespondingtothelowestsampleentropyvalueistakenasthefinaldecompositionnumberofVMD.Secondly,principalcomponentanalysisisusedtoselecttheWPEofeachmodecomponentafterdecomposition,andtheformedfinalfeaturevectorspaceisinputtoDBN.Meanwhile,toavoidprematureconvergenceandlocaloptimumproblemsduringfaultclassification,IGWOisusedtooptimizethenetworkstructureparametersofDBNtorecognizevariousfaultsofATSE.ATSEfaultsimulationsaredesignedandcarriedoutinthispaper.AfterdeterminingthedecompositionfrequencyofVMD,modaldecompositiononthecoilcurrentisperformed.ThesuperiorityofVMDdecompositionisdemonstratedbytheabsenceofmodalaliasingintheintrinsicmodefunctionsatdifferentfrequencies.ThefeaturevectorsareinputintotheoptimizedDBNalgorithmforfaultclassificationexperiments.Theexperimentalresultsareasfollows:(1)ThedifferenceinclassificationaccuracybetweentheoptimizedDBNandthetrainingsetisonly0.4%,indicatingnoover-fittingprobleminthemodifiedmodel.(2)TheoptimizedDBNachievesaclassificationaccuracyof98.78%forfourcommonfaults.Comparedwiththenon-optimizedmethods,theproposedmethodhasthehighestaccuracyandthebeststabilityinfaultdiagnosisofATSE.Thefollowingconclusionscanbedrawn:(1)Afterdeterminingthedecompositiontimebasedonthesampleentropy,theVMDcaneffectivelyavoidthemodeoverlap,thusrealizingtheextractionoffaultfeaturesfromATSEcurrentsignals.(2)Comparedwiththeshallowneuralnetwork,DBNhasapowerfulmappingabilityandcanaccuratelycharacterizethecomplexmappingrelationshipbetweentheoriginalcurrentsignalandtheATSEfaulttype.(3)TheoptimizedDBNhasnoover-fittingphenomenausingIGWO,andtheclassificationaccuracyofATSEfaultsisimproved.
作者：刘帼巾 刘达明 缪建华 杨雨泽 王乐康 刘琦 Author：LiuGuojin LiuDaming MiaoJianhua YangYuze WangLekang LiuQi
作者单位：省部共建电工装备可靠性与智能化国家重点实验室(河北工业大学)天津300130施耐德万高(天津)电气设备有限公司天津300384
刊名：电工技术学报
Journal：TransactionsofChinaElectrotechnicalSociety
年，卷(期)：2024, 39(4)
分类号：TM564.3
关键词：优化灰狼算法 深度置信网络 自动转换开关 故障识别
Keywords：ImprovedWolfoptimizer deepbeliefnetwork automatictransferswitchingequipment faultdiagnosis
机标分类号：TP391.41TM732TN919.81
在线出版日期：2024年3月5日
基金项目：河北省自然科学基金，河北省省级科技计划资助项目，河北省省级科技计划资助项目基于变分模态分解和改进灰狼算法优化深度置信网络的自动转换开关故障识别[
期刊论文] 电工技术学报--2024, 39(4)刘帼巾 刘达明 缪建华 杨雨泽 王乐康 刘琦自动转换开关(ATSE)是保证系统连续供电的设备,对其进行健康监测和故障诊断对系统的稳定运行具有重要意义.为了实现对ATSE的非侵入式故障识别,该文提出一种基于电流信号变分模态分解(VMD)的特征提取和改进灰狼算法(IG...参考文献和引证文献
参考文献
引证文献
本文读者也读过
相似文献
相关博文
基于变分模态分解和改进灰狼算法优化深度置信网络的自动转换开关故障识别 Fault Identification of Automatic Transfer Switching Equipment Based on VMD-WPE and IGWO Optimized DBN
基于变分模态分解和改进灰狼算法优化深度置信网络的自动转换开关故障识别.pdf
- 文件大小:
- 1.22 MB
- 下载次数:
- 60
-
高速下载
|
|
