改进EEMD算法在高压并联电抗器声信号去噪中的应用

高压并联电抗器运行过程中产生的声信号是准确判定电抗器运行状态的重要依据，在对电抗器声信号现场采集时易受到多种外界噪声的干扰，测量仪器无法有效进行预处理，导致对电抗器运行状态的评估发生误判。提出了一种基于多传感器融合及最小下限频率截止的改进集合经验模态分解(ensembleempiricalmodedecomposition,EEMD)高压并联电抗器声信号去噪方法。首先，利用一致性数据融合算法对各声纹传感器进行关联和甄别，剔除失效传感器，确定有效传感器组。其次，选取有效传感器组中的最小下限频率作为固有模态函数(intrinsicmodefunction,IMF)的筛选截止条件并进行集合经验模态分解。然后利用相关系数法提取有效的IMF分量。最后对有效IMF分量叠加重构，得到去噪声信号。模拟实验和实测结果表明，该方法具有较好的去噪效果。通过与传统经验模态分解法(empiricalmodedecomposition,EMD)、标准EEMD去噪技术的比较，验证了该方法在实际应用过程中的有效性和实用性。

Application of an improved EEMD algorithm in high voltage shunt reactor sound signal denoising

Theacousticsignalsgeneratedduringtheoperationofhigh-voltageshuntreactorsareacriticalbasisforaccuratelydeterminingthereactor''soperationalstatus.However,collectingtheseacousticsignalson-sitecanbesubjecttointerferencefromvariousexternalnoises.Themeasuringinstrumentsoftenfailtoeffectivelypre-processthesesignals,resultinginaninaccurateassessmentofthereactor''soperatingcondition.Thispaperpresentsanenhancedensembleempiricalmodedecomposition(EEMD)acousticsignaldenoisingapproachforhigh-voltageshuntreactors,onewhichreliesonmulti-sensordatafusionandtheselectionofaminimumlowerfrequencylimitfortermination.Initially,aconsistentdatafusionalgorithmisusedtocorrelateandfilterthefaultsensors,discardinganyinvalidsensorsanddeterminingtheactivesensorgroup.Subsequently,theminimumlowerlimitfrequencyforeachsensorsignalischosenasthescreeningterminationcriterionfortheintrinsicmodefunction(IMF)throughspectralanalysis,andtheEEMDdecompositionisconducted.ThecorrelationcoefficientmethodisthenemployedtoextracttheeffectiveIMFcomponents.Finally,theextractedIMFcomponentsaresuperimposedandreconstructedtoobtainthedenoisedsignal.Experimentalandmeasuredsignalsdemonstratethatthemethodcanachievesignaldenoisingaccurately.Acomparisonwiththetraditionalempiricalmodedecomposition(EMD)methodandthestandardEEMDdenoisingtechniqueverifiesthepracticalapplicationeffectivenessandpracticabilityoftheproposedalgorithm.