改进EEMD算法在高压并联电抗器声信号去噪中的应用  

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

作　　者：王果 雷武[2] 闵永智 万保权[1] 李宝鹏 王毅斌 WANG Guo;LEI Wu;MIN Yongzhi;WAN Baoquan;LI Baopeng;WANG Yibin(State Key Laboratory of Power Grid Environmental Protection,China Electric Power Research Institute,Wuhan 430074,China;School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]中国电力科学研究院有限公司电网环境保护国家重点实验室,湖北武汉430074 [2]兰州交通大学自动化与电气工程学院,甘肃兰州730070

出　　处：《电力系统保护与控制》2023年第24期164-174,共11页Power System Protection and Control

基　　金：2022年度电网环境保护国家重点实验室开放基金项目资助(GYW51202201459)。

摘　　要：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(ensemble empirical mode decomposition,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(intrinsicmode function, 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(empirical mode decomposition, 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。The acoustic signals generated during the operation of high-voltage shunt reactors are a critical basis for accurately determining the reactor's operational status.However,collecting these acoustic signals on-site can be subject to interference from various external noises.The measuring instruments often fail to effectively pre-process these signals,resulting in an inaccurate assessment of the reactor's operating condition.This paper presents an enhanced ensemble empirical mode decomposition(EEMD)acoustic signal denoising approach for high-voltage shunt reactors,one which relies on multi-sensor data fusion and the selection of a minimum lower frequency limit for termination.Initially,a consistent data fusion algorithm is used to correlate and filter the fault sensors,discarding any invalid sensors and determining the active sensor group.Subsequently,the minimum lower limit frequency for each sensor signal is chosen as the screening termination criterion for the intrinsic mode function(IMF)through spectral analysis,and the EEMD decomposition is conducted.The correlation coefficient method is then employed to extract the effective IMF components.Finally,the extracted IMF components are superimposed and reconstructed to obtain the denoised signal.Experimental and measured signals demonstrate that the method can achieve signal denoising accurately.A comparison with the traditional empirical mode decomposition(EMD)method and the standard EEMD denoising technique verifies the practical application effectiveness and practicability of the proposed algorithm.This work is supported by the Open Fund of State Key Laboratory of Power Grid Environmental Protection(No.GYW51202201459).

关 键 词：高压并联电抗器 声信号去噪 集合经验模态分解 频率截止 多传感器融合 

分 类 号：TM472[电气工程—电器]