CN114861722A - Partial discharge narrow-band interference suppression method based on time-frequency spectrogram separation - Google Patents
Partial discharge narrow-band interference suppression method based on time-frequency spectrogram separationDownload PDFInfo
- Publication number
- CN114861722A CN114861722ACN202210472721.7ACN202210472721ACN114861722ACN 114861722 ACN114861722 ACN 114861722ACN 202210472721 ACN202210472721 ACN 202210472721ACN 114861722 ACN114861722 ACN 114861722A
- Authority
- CN
- China
- Prior art keywords
- signal
- time
- noise
- matrix
- frequency
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
- G01R31/1227—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
- G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- General Engineering & Computer Science (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
Translated fromChinese
Description
Translated fromChinese技术领域technical field
本发明涉及高压电力设备绝缘状态诊断技术，具体是基于时频谱图分离的局部放电窄带干扰抑制方法。The invention relates to an insulation state diagnosis technology of high-voltage power equipment, in particular to a partial discharge narrow-band interference suppression method based on time-spectrogram separation.
背景技术Background technique
高压电力设备的绝缘状态可通过监测局部放电(Partial Discharge,PD)信号进行诊断，通常现场采集的PD信号会受到电磁干扰影响，导致难以直接开展PD信号的分析及特征提取。PD信号面临的电磁干扰主要为脉冲型干扰、白噪声干扰及周期性窄带干扰，其中，周期性窄带干扰拥有持续时间长、能量强和随机性大的显著特点，导致PD信号容易被周期性窄带干扰完全淹没。因此，为了利用PD信号对高压电力设备的绝缘状态诊断时提升精确性，对周期性窄带干扰的抑制尤其重要。The insulation state of high-voltage power equipment can be diagnosed by monitoring partial discharge (PD) signals. Usually, PD signals collected on site are affected by electromagnetic interference, which makes it difficult to directly analyze and extract PD signals. The electromagnetic interference faced by PD signals is mainly pulsed interference, white noise interference and periodic narrowband interference. Among them, periodic narrowband interference has the characteristics of long duration, strong energy and large randomness, which makes PD signals easy to be blocked by periodic narrowband interference. The distractions are completely overwhelmed. Therefore, in order to improve the accuracy of diagnosing the insulation state of high-voltage power equipment using PD signals, the suppression of periodic narrowband interference is particularly important.
目前，国内外学者对PD信号中周期性窄带干扰的抑制技术开展了大量研究。樊高辉、刘尚合、刘卫东及王雷于2017年04月06日在《高电压技术》发表的名称为“FFT谱最小熵解卷积滤波抑制放电信号中的周期性窄带干扰”的文献，其在快速傅里叶变换(fastFourier transform，FFT)算法的基础上，引入了解卷积滤波法和经典阈值法用于提取窄带干扰的频点，然后将窄带干扰对应频点的FFT系数进行压缩，达到削弱窄带干扰能量的目的。虽然该方法可以有效削弱窄带干扰的频域能量，但是由于FFT算法存在频谱泄露等缺陷和阈值法存在阈值选取困难的问题，所以该方法最终的降噪效果不够理想。马星河和张登奎于2021年08月17日在《电工技术学报》发表的名称为“基于改进经验小波变换的高压电缆局部放电噪声抑制研究”的文献，其借助小波分解算法的优异时频分析能力，能一定程度上削弱窄带干扰的能量，但是该类方法难以选择适宜的小波基函数、小波分解层数和小波分解阈值。魏海增、马宏忠、黄涛及黄烜城于2019年05朋15日在《电力系统及其自动化学报》发表的名称为“基于EMD的ICA降噪方法在电厂开关柜局部放电信号中的应用”的文献，其使用经验模态分解算法对PD信号中窄带干扰进行分离，但是该方法的稳定性较差，容易出现端点效应、模态混叠等问题。徐永干、姜杰、唐昆明、张太勤、罗建及谢敏于2019年09月17日在《电网技术》发表的名称为“基于Hankel矩阵和奇异值分解的局部放电窄带干扰抑制方法”的文献，毕潇文、钟俊、张大堃、周电波及阮莹于2021年03月25日在《电网技术》发表的名称为“基于改进奇异值与经验小波分解的局放去噪算法”的文献，以及杨晓丽、黄宏光、舒勤、张大堃及周电波于2020年11月25日在《高电压技术》发表的名称为“基于SVD和低秩RBF神经网络的局部放电信号提取方法”的文献，这三篇文献均提出了自适应奇异值分解法抑制染噪PD信号中窄带干扰，但是该方法存在奇异值阈值选取困难的问题，难以有效抑制小幅值的窄带干扰。宋立业、蒲霄祥及李希桐于2021年11月23日在《电工电能新技术》发表的名称为“基于广义S变换和随机子空间的局放窄带干扰抑制方法”的文献，其利用广义S变换算法得到染噪PD信号的时频谱图，然后直接在该时频谱图中确定窄带干扰和PD信号对应的时频区域子矩阵，以提取窄带干扰的特征参数，实现窄带干扰抑制，但是在窄带干扰强能量特点的影响下，PD信号的时频特征可能会被窄带干扰所掩盖，导致该类方法的可靠性下降。At present, scholars at home and abroad have carried out a lot of research on the suppression technology of periodic narrowband interference in PD signals. Fan Gaohui, Liu Shanghe, Liu Weidong and Wang Lei published a paper titled "FFT Spectral Minimum Entropy Deconvolution Filtering to Suppress Periodic Narrow Band Interference in Discharge Signals" in "High Voltage Technology" on April 6, 2017. On the basis of the Fourier transform (fast Fourier transform, FFT) algorithm, the deconvolution filtering method and the classical threshold method are introduced to extract the frequency points of the narrowband interference, and then the FFT coefficients of the corresponding frequency points of the narrowband interference are compressed to weaken the narrowband interference. The purpose of interfering with energy. Although this method can effectively weaken the frequency domain energy of narrowband interference, the final noise reduction effect of this method is not ideal due to the defects such as spectrum leakage in the FFT algorithm and the difficulty of threshold selection in the threshold method. Ma Xinghe and Zhang Dengkui published a paper titled "Research on Partial Discharge Noise Suppression of High Voltage Cables Based on Improved Empirical Wavelet Transform" published in "Journal of Electrotechnical Technology" on August 17, 2021, which relies on the excellent time-frequency analysis ability of wavelet decomposition algorithm , which can weaken the energy of narrowband interference to a certain extent, but it is difficult to select suitable wavelet basis functions, wavelet decomposition layers and wavelet decomposition thresholds for this type of method. Wei Haizeng, Ma Hongzhong, Huang Tao and Huang Xuancheng published a document titled "Application of EMD-based ICA Noise Reduction Method in Partial Discharge Signals of Switchgear in Power Plants" on May 15, 2019 in the Journal of Electric Power Systems and Automation. The empirical mode decomposition algorithm is used to separate the narrowband interference in the PD signal, but the stability of this method is poor, and problems such as end effect and modal aliasing are prone to occur. Xu Yongqian, Jiang Jie, Tang Kunming, Zhang Taiqin, Luo Jian and Xie Min published in "Power Grid Technology" on September 17, 2019, the paper entitled "Partial discharge narrowband interference suppression method based on Hankel matrix and singular value decomposition", Bi Xiaowen, Zhong Jun, Zhang Dakun, Zhou Dianbo and Ruan Ying published a paper titled "Partial Discharge Denoising Algorithm Based on Improved Singular Value and Empirical Wavelet Decomposition" in "Power Grid Technology" on March 25, 2021, and Yang Xiaoli , Huang Hongguang, Shu Qin, Zhang Dakun and Zhou Dianbo published in "High Voltage Technology" on November 25, 2020, a document entitled "Partial discharge signal extraction method based on SVD and low-rank RBF neural network", these three All the literatures proposed the adaptive singular value decomposition method to suppress the narrowband interference in the noisy PD signal, but this method has the problem of difficult selection of the singular value threshold, and it is difficult to effectively suppress the narrowband interference of small amplitude. Song Liye, Pu Xiaoxiang and Li Xitong published a paper titled "Partial Discharge Narrowband Interference Suppression Method Based on Generalized S-Transform and Random Subspace" in "New Technology of Electrical Engineering" on November 23, 2021, which uses the generalized S-transform algorithm Obtain the time-spectrogram of the noisy PD signal, and then directly determine the time-frequency region sub-matrix corresponding to the narrowband interference and the PD signal in the time-spectrogram to extract the characteristic parameters of the narrowband interference and achieve narrowband interference suppression, but when the narrowband interference is strong Under the influence of energy characteristics, the time-frequency characteristics of PD signals may be masked by narrow-band interference, resulting in a decrease in the reliability of this type of method.
发明内容SUMMARY OF THE INVENTION
本发明的目的在于解决现有技术抑制染噪PD信号中窄带干扰的不足，提供了一种基于时频谱图分离的局部放电窄带干扰抑制方法，其能分离出染噪PD信号时频谱图中窄带干扰和PD信号，提升窄带干扰影响下PD信号时频分析的准确率，从而有效抑制染噪PD信号中窄带干扰。The purpose of the present invention is to solve the deficiencies of the prior art in suppressing the narrowband interference in the noise-dyed PD signal, and provides a partial discharge narrowband interference suppression method based on the separation of the time-spectrogram, which can separate the narrowband in the frequency-spectrogram of the noise-dyed PD signal. Interference and PD signals, improve the accuracy of time-frequency analysis of PD signals under the influence of narrowband interference, thereby effectively suppressing narrowband interference in noisy PD signals.
本发明的目的主要通过以下技术方案实现：The object of the present invention is mainly realized through the following technical solutions:
基于时频谱图分离的局部放电窄带干扰抑制方法，包括：The partial discharge narrowband interference suppression method based on time-spectrogram separation, including:
S1、采样获取染噪PD信号；S1. Sampling to obtain a noise-dyed PD signal;
S2、采用广义S变换对染噪PD信号的离散信号进行时频变换处理，得到二维时频复数矩阵，其中，二维时频复数矩阵中横轴为时间采样点，纵轴为频率采样点；S2. Use generalized S transform to perform time-frequency transformation on the discrete signal of the noise-dyed PD signal, and obtain a two-dimensional time-frequency complex number matrix, wherein the horizontal axis in the two-dimensional time-frequency complex number matrix is the time sampling point, and the vertical axis is the frequency sampling point. ;
S3、对二维时频复数矩阵求模得到关于染噪PD信号的时频谱图；S3, modulo the two-dimensional time-frequency complex matrix to obtain a time-spectrogram of the noise-dyed PD signal;
S4、采用数学形态学获取染噪PD信号时频谱图中各行分量的平滑分量和非平滑分量，将分离出的平滑分量和非平滑分量作为行分量构成新的矩阵，分离出窄带干扰和原始PD信号的时频谱图，以确定窄带干扰的数目与PD脉冲的时间区域；S4. Use mathematical morphology to obtain the smooth components and non-smooth components of each line component in the spectrogram when the noise-dyed PD signal is obtained, use the separated smooth components and non-smooth components as line components to form a new matrix, and separate the narrowband interference and the original PD. The time-spectrogram of the signal to determine the number of narrowband interferers and the time region of the PD pulse;
S5、将采样得到的染噪PD信号根据时间轴划定为含有PD信号的信号帧和仅含噪声的噪声帧，采用总体最小二乘不变旋转矢量技术分析噪声帧内时域数据，以提取窄带干扰的时域特征量；S5. Delineate the sampled noise-dyed PD signal into a signal frame containing PD signal and a noise frame containing only noise according to the time axis, and use the overall least squares invariant rotation vector technology to analyze the time domain data in the noise frame to extract Time-domain characteristic quantity of narrowband interference;
S6、重构出窄带干扰的时域波形，从染噪PD信号的时域波形中去除窄带干扰的时域波形，得到降噪后PD信号。S6, reconstruct the time domain waveform of the narrowband interference, remove the time domain waveform of the narrowband interference from the time domain waveform of the noise-dyed PD signal, and obtain the PD signal after noise reduction.
本发明针对高压电力设备局部放电信号受周期性窄带干扰影响大的问题，提出了一种基于时频谱图分离的局部放电窄带干扰抑制方法，其结合广义S变换和数学形态学，从染噪PD信号中分离出原始PD信号和窄带干扰信号的时频谱图，从而在时频域内准确辨识出窄带干扰和原始PD信号，避免强能量的窄带干扰掩盖原始PD信号时频特征的问题，提升窄带干扰影响下PD信号时频分析的准确率。本发明再利用总体最小二乘不变旋转矢量技术准确估计窄带干扰的特征参数，以抑制染噪PD信号中窄带干扰。Aiming at the problem that the partial discharge signal of high-voltage power equipment is greatly affected by periodic narrowband interference, the invention proposes a partial discharge narrowband interference suppression method based on time-spectrogram separation. The time-spectrogram of the original PD signal and the narrow-band interference signal is separated from the signal, so that the narrow-band interference and the original PD signal can be accurately identified in the time-frequency domain, avoiding the problem that the strong-energy narrow-band interference masks the time-frequency characteristics of the original PD signal, and improving the narrow-band interference The accuracy of time-frequency analysis of PD signals under the influence. The invention further utilizes the overall least squares invariant rotation vector technology to accurately estimate the characteristic parameters of the narrowband interference, so as to suppress the narrowband interference in the noise-dyed PD signal.
进一步的，所述步骤S2中采用广义S变换对染噪PD信号的离散信号进行时频变换处理包括以下步骤：Further, in the step S2, using generalized S transform to perform time-frequency transform processing on the discrete signal of the noise-dyed PD signal includes the following steps:
染噪PD信号x(t)的广义S变换结果G(τ,f,λ)被定义为The generalized S-transform result G(τ, f, λ) of the noisy PD signal x(t) is defined as
式中，t和τ为两组时间变量；f是频率；λ是调节因子；w(t-τ,f,λ)是高斯窗函数，对应的表达式为：In the formula, t and τ are two sets of time variables; f is the frequency; λ is the adjustment factor; w(t-τ, f, λ) is the Gaussian window function, and the corresponding expression is:
令x(n)为染噪PD信号x的离散信号，同时定义f＝n/(NT)、τ＝iT，其中T是采样周期，N是x(n)的数据总量，得到x(n)的广义S变换结果为：Let x(n) be the discrete signal of the noise-dyed PD signal x, and define f=n/(NT), τ=iT, where T is the sampling period, N is the total amount of data of x(n), get x(n ), the generalized S transform result is:
式中，i、m、n是x(n)的广义S变换结果中三组变量，被定义为0，1，…，N-1；In the formula, i, m, n are three groups of variables in the generalized S transform result of x(n), which are defined as 0, 1, ..., N-1;
利用x(n)的广义S变换结果对x(n)开展时频变换处理后得到二维的时频复数矩阵。Using the generalized S transform result of x(n) to carry out time-frequency transform processing on x(n), a two-dimensional time-frequency complex number matrix is obtained.
进一步的，所述步骤S4中采用数学形态学获取染噪PD信号时频谱图中各行分量的平滑分量和非平滑分量包括以下步骤：Further, in the step S4, the smooth component and the non-smooth component of each line component in the spectrogram when using mathematical morphology to obtain the noise-dyed PD signal includes the following steps:
所述染噪PD信号时频谱图中各行分量Gt(nGt)的膨胀变换结果公式为：The dilation transformation result formula of each row component Gt (nGt ) in the spectrogram when the noise PD signal is dyed is:
腐蚀变换结果公式为：The result formula of corrosion transformation is:
式中，g(ig)是结构元素；⊕是膨胀运算符；⊙是腐蚀运算符；DGt是Gt(nGt)的定义域；Dg是g(ig)的定义域，nGt是数组Gt中某个索引，ig是数组g中某个索引；where g(ig ) is the structuring element; ⊕ is the dilation operator; ⊙ is the erosion operator; DGt is the domain of Gt (nGt ); Dg is the domain of g (ig ), nGt is an index in the array Gt , ig is an index in the array g;
通过膨胀变换结果公式和腐蚀变换结果公式的级联组合，得到数学形态学中开运算和闭运算分别为：Through the cascading combination of the dilation transformation result formula and the erosion transformation result formula, the opening operation and closing operation in mathematical morphology are obtained as:
(Gt·g)(nGt)＝((Gt⊕g)⊙g)(nGt)(Gt ·g)(nGt )=((Gt ⊕g)⊙g)(nGt )
式中，
将开、闭运算进行级联组合，分别得到形态开-闭滤波器运行结果[Foc(Gt)](nGt)和形态闭-开滤波器运行结果[Fco(Gt)](nGt)为:The open and closed operations are cascaded to obtain the operation result of the morphological open-closed filter [Foc (Gt )](nGt ) and the operation result of the morphological closed-open filter [Fco (Gt )]( nGt ) is:
采用形态开-闭滤波器和形态闭-开滤波器的混合运算，得到信号Gt(nGt)中的平滑分量[F1(Gt)](nGt)和非平滑分量[F2(Gt)](nGt)分别为：Using the mixed operation of the morphological open-closed filter and the morphological closed-open filter, the smooth component [F1 (Gt) ](n Gt) and the non-smooth component [F2 ( Gt )](nGt ) are:
[F1(Gt)](nGt)＝([Foc(Gt)](nGt)+[Fco(Gt)](nGt))/2[F1 (Gt )](nGt )=([Foc (Gt )](nGt )+[Fco (Gt )](nGt ))/2
[F2(Gt)](nGt)＝Gt(nGt)-[F1(Gt)](nGt)。[F2 (Gt )](nGt )=Gt (nGt )−[F1 (Gt )](nGt ).
进一步的，所述步骤S5中采用总体最小二乘不变旋转矢量技术分析噪声帧内时域数据包括以下步骤：Further, in the step S5, using the overall least squares invariant rotation vector technique to analyze the time-domain data in the noise frame includes the following steps:
S51、将噪声帧内时域数据信号y(ny)构成Hankle矩阵H为：S51, the time-domain data signal y(ny ) in the noise frame is formed into a Hankle matrix H as:
式中，Ny是信号y(ny)的数据长度，L是H的列数；In the formula, Ny is the data length of the signal y(ny ), and L is the number of columns of H;
S52、将H开展奇异值分解，得到S52, carry out singular value decomposition of H to obtain
H＝UΣVTH=UΣVT
式中，T是取矩阵的共轭转置；U和V分别是左、右正交矩阵；Σ是对角矩阵，其中的对角元素是矩阵H的奇异值；In the formula,T is the conjugate transpose of the matrix; U and V are the left and right orthogonal matrices, respectively; Σ is the diagonal matrix, and the diagonal elements are the singular values of the matrix H;
S53、分离出矩阵U中第1行到第L行，第1列到第P列的子矩阵为矩阵U1；同时分离出矩阵U中第2行到第L+1行，第1列到第P列的子矩阵为U2，以此得到U1和U2分别为：S53, separate out the 1st row to the Lth row in the matrix U, and the submatrix from the 1st column to the Pth column is the matrix U1 ; The submatrix of the P-th column is U2 , so that U1 and U2 are obtained as:
U1＝U[1:L,1:P]U1 =U[1:L,1:P]
U2＝U[2:L+1,1:P]U2 =U[2:L+1,1:P]
式中，P是分离后时频图谱确定的窄带干扰个数的2倍；In the formula, P is twice the number of narrowband interferences determined by the time-frequency spectrum after separation;
S54、构建矩阵Z＝[U1 U2]，然后对Z开展奇异值分解，得到S54, construct matrix Z=[U1 U2 ], and then carry out singular value decomposition on Z to obtain
式中：
S55、利用V12和V22构建矩阵ψ为S55. Use V12 and V22 to construct a matrix ψ as
ψ＝-V12V22-1ψ=-V12 V22-1
S56、对ψ开展特征值分解，得到对应的特征值为μk(k＝1,2,…,P)，从而估计得到各窄带干扰分量的频率
式中，arg(*)是计算复数的相角；In the formula, arg(*) is the phase angle of the complex number;
S57、利用最小二乘法计算各窄带干扰分量的幅值
Y＝μcY=μc
Y＝[y(0)，y(1)，…，y(Ny-1)]TY=[y(0), y(1), ..., y(Ny -1)]T
式中，c是中间变量矩阵，ck是c中元素。In the formula, c is the intermediate variable matrix, and ck is the element in c.
进一步的，所述信号帧内窄带干扰相位
式中，ΔT是信号帧和噪声帧的时延。where ΔT is the time delay between the signal frame and the noise frame.
综上所述，本发明与现有技术相比具有以下有益效果：本发明能有效分离出染噪PD信号时频谱图中窄带干扰和PD信号，具备良好的窄带干扰抑制效果，能有效恢复局部放电信号的波形特征，进而能提升窄带干扰影响下PD信号时频分析的准确率。To sum up, compared with the prior art, the present invention has the following beneficial effects: the present invention can effectively separate the narrowband interference and the PD signal in the frequency spectrum of the noisy PD signal, has a good narrowband interference suppression effect, and can effectively restore local The waveform characteristics of the discharge signal can further improve the accuracy of the time-frequency analysis of the PD signal under the influence of narrowband interference.
附图说明Description of drawings
此处所说明的附图用来提供对本发明实施例的进一步理解，构成本申请的一部分，并不构成对本发明实施例的限定。在附图中：The accompanying drawings described herein are used to provide further understanding of the embodiments of the present invention, and constitute a part of the present application, and do not constitute limitations to the embodiments of the present invention. In the attached image:
图1为PD信号的仿真波形图；Fig. 1 is the simulation waveform diagram of PD signal;
图2为仿真PD波形的时频谱图；Fig. 2 is the time-frequency spectrum diagram of the simulated PD waveform;
图3为染噪PD信号时频谱图中频率能量和时间的关系图；Fig. 3 is the relation diagram of frequency energy and time in the spectrogram when the noise PD signal is dyed;
图4为分离后窄带干扰和原始PD信号的时频谱图；Fig. 4 is the time-spectrogram diagram of narrowband interference and original PD signal after separation;
图5为仿真PD波形的窄带干扰抑制结果图；Fig. 5 is the narrowband interference suppression result graph of the simulated PD waveform;
图6为混合干扰下的染噪PD波形图；Fig. 6 is the PD waveform diagram of dyed noise under mixed interference;
图7为混合干扰下染噪PD信号的时频谱图分离结果图；Fig. 7 is the time-spectrogram separation result diagram of the noise-dyed PD signal under mixed interference;
图8为混合干扰下染噪PD波形的降噪结果图；Fig. 8 is the noise reduction result graph of the noise-dyed PD waveform under mixed interference;
图9为实测的染噪PD信号图；Fig. 9 is the measured PD signal graph of dyed noise;
图10为实测PD信号的时频谱图；Fig. 10 is the time-spectrogram of the measured PD signal;
图11为实测PD信号的时频谱图分离结果图；Fig. 11 is the time-spectrogram separation result diagram of the measured PD signal;
图12为实测PD信号的窄带干扰抑制结果图；FIG. 12 is a graph of the narrowband interference suppression result of the measured PD signal;
图13为本发明一个具体实施例的流程图。FIG. 13 is a flowchart of a specific embodiment of the present invention.
具体实施方式Detailed ways
为使本发明的目的、技术方案和优点更加清楚明白，下面结合实施例和附图，对本发明作进一步的详细说明，本发明的示意性实施方式及其说明仅用于解释本发明，并不作为对本发明的限定。In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and the accompanying drawings. as a limitation of the present invention.
实施例：Example:
如图13所示，基于时频谱图分离的局部放电窄带干扰抑制方法，包括：S1、采样获取染噪PD信号；S2、采用广义S变换对染噪PD信号的离散信号进行时频变换处理，得到二维时频复数矩阵，其中，二维时频复数矩阵中横轴为时间采样点，纵轴为频率采样点；S3、对二维时频复数矩阵求模得到关于染噪PD信号的时频谱图；S4、采用数学形态学获取染噪PD信号时频谱图中各行分量的平滑分量和非平滑分量，将分离出的平滑分量和非平滑分量作为行分量构成新的矩阵，分离出窄带干扰和原始PD信号的时频谱图，以确定窄带干扰的数目与PD脉冲的时间区域；S5、将采样得到的染噪PD信号根据时间轴划定为含有PD信号的信号帧和仅含噪声的噪声帧，采用总体最小二乘不变旋转矢量技术分析噪声帧内时域数据，以提取窄带干扰的时域特征量；S6、重构出窄带干扰的时域波形，从染噪PD信号的时域波形中去除窄带干扰的时域波形，得到降噪后PD信号。As shown in Figure 13, the partial discharge narrow-band interference suppression method based on time-spectrogram separation includes: S1, sampling to obtain the noisy PD signal; S2, using generalized S transform to perform time-frequency transform processing on the discrete signal of the noisy PD signal, A two-dimensional time-frequency complex number matrix is obtained, wherein, in the two-dimensional time-frequency complex number matrix, the horizontal axis is the time sampling point, and the vertical axis is the frequency sampling point; S3, modulo the two-dimensional time-frequency complex number matrix to obtain the time of the noise-dyed PD signal. Spectrogram; S4. When using mathematical morphology to obtain smooth components and non-smooth components of each line component in the spectrogram, the separated smooth components and non-smooth components are used as line components to form a new matrix, and the narrowband interference is separated out and the time-spectrogram of the original PD signal to determine the number of narrow-band interference and the time area of the PD pulse; S5. Delineate the sampled noise-dyed PD signal into a signal frame containing the PD signal and a noise-only noise according to the time axis frame, using the overall least squares invariant rotation vector technology to analyze the time domain data in the noise frame to extract the time domain feature quantity of the narrowband interference; S6, reconstruct the time domain waveform of the narrowband interference, from the time domain of the noisy PD signal The time domain waveform with the narrowband interference removed from the waveform is obtained to obtain the PD signal after noise reduction.
本实施例中，信号x(t)的广义S变换结果G(τ,f,λ)被定义为：In this embodiment, the generalized S-transform result G(τ, f, λ) of the signal x(t) is defined as:
式中，t和τ为两组时间变量；f是频率；λ是调节因子；w(t-τ,f,λ)是高斯窗函数，对应的表达式为：In the formula, t and τ are two sets of time variables; f is the frequency; λ is the adjustment factor; w(t-τ, f, λ) is the Gaussian window function, and the corresponding expression is:
由于式(2)中窗函数的宽度和分析频率成反比，所以广义S变换在低频区域能获得更高的频率分辨率，而在高频区域能获得更高的时间分辨率，利于PD信号这类非平稳信号的时频分析。同时可以通过设定合适的λ，对广义S变换的最终时频变换结果进行人为调节。如果将λ设定在(0,1)范围内，那么可以增加式(2)中窗函数的宽度，达到提升频域分辨率的目的；如果将λ设定在(1,+∞)范围内，那么可以减小式(2)中窗函数的宽度，达到提升时域分辨率的目的。Since the width of the window function in Equation (2) is inversely proportional to the analysis frequency, the generalized S transform can obtain higher frequency resolution in the low frequency region, and can obtain higher time resolution in the high frequency region, which is beneficial to the PD signal. Time-frequency analysis of non-stationary-like signals. At the same time, the final time-frequency transform result of the generalized S transform can be adjusted artificially by setting an appropriate λ. If λ is set in the range of (0, 1), the width of the window function in equation (2) can be increased to achieve the purpose of improving the frequency domain resolution; if λ is set in the range of (1, +∞) , then the width of the window function in equation (2) can be reduced to achieve the purpose of improving the temporal resolution.
在实际使用广义S变换时，令x(n)是信号x的离散信号，同时定义f＝n/(NT)、τ＝iT，其中T是采样周期，N是x(n)的数据总量，得到x(n)的广义S变换结果为：When the generalized S transform is actually used, let x(n) be the discrete signal of the signal x, and define f=n/(NT), τ=iT, where T is the sampling period, and N is the total amount of data of x(n). , the generalized S-transform result of x(n) is obtained as:
式中，i、m、n是x(n)的广义S变换结果中三组变量，被定义为0，1，…，N-1。In the formula, i, m, n are the three groups of variables in the generalized S transform result of x(n), which are defined as 0, 1, ..., N-1.
在利用式(3)对x(n)开展时频变换处理后，可以得到一个二维的时频复数矩阵，该矩阵的横轴是时间采样点，该矩阵的纵轴是频率采样点。对该二维时频复数矩阵进行求模处理，可以得到关于染噪PD信号的时频谱图。After using formula (3) to carry out the time-frequency transform processing on x(n), a two-dimensional time-frequency complex matrix can be obtained, the horizontal axis of the matrix is the time sampling point, and the vertical axis of the matrix is the frequency sampling point. By performing modulo processing on the two-dimensional time-frequency complex matrix, a time-spectrogram of the noise-dyed PD signal can be obtained.
本实施例采用的广义S变换方法是一种基于S变换的改进时频变换方法，该方法在S变换方法的窗函数中加入了调节因子λ，不仅继承了S变换对非平稳信号的优异时频分析能力，同时增加了时频变换结果的人为可控性。本实施例采用的广义S变换方法同时具备小波变换方法的多分辨率分析和短时傅里叶变换方法的单频率分量分析的优点，并且不需要人为选择小波函数或窗函数的参数，减小了人为因素干扰。The generalized S-transform method used in this embodiment is an improved time-frequency transform method based on S-transform. This method adds an adjustment factor λ to the window function of the S-transform method, which not only inherits the excellent time-frequency transformation of S-transform for non-stationary signals frequency analysis capability, while increasing the human controllability of the time-frequency transformation results. The generalized S-transform method adopted in this embodiment has the advantages of multi-resolution analysis of the wavelet transform method and single-frequency component analysis of the short-time Fourier transform method, and does not need to artificially select the parameters of the wavelet function or the window function, reducing the Human factor interference.
数学形态学中存在两种基本的变换形式：腐蚀和膨胀。染噪PD信号时频谱图中各行分量Gt(nGt)的膨胀和腐蚀变换结果分别表示为：There are two basic forms of transformation in mathematical morphology: erosion and dilation. The dilation and erosion transformation results of each line component Gt (nGt ) in the spectrogram when the PD signal is stained are expressed as:
式中，g(ig)是结构元素；⊕是膨胀运算符；⊙是腐蚀运算符；DGt是Gt(nGt)的定义域；Dg是g(ig)的定义域，nGt是数组Gt中某个索引，ig是数组g中某个索引。where g(ig ) is the structuring element; ⊕ is the dilation operator; ⊙ is the erosion operator; DGt is the domain of Gt (nGt ); Dg is the domain of g (ig ), nGt is an index in array Gt , and ig is an index in array g.
通过式(4)和式(5)的级联组合，可以得到数学形态学中开运算和闭运算分别为：Through the cascade combination of Equation (4) and Equation (5), the open operation and closed operation in mathematical morphology can be obtained as:
(Gt·g)(nGt)＝((Gt⊕g)⊙g)(nGt) (7)(Gt ·g)(nGt )=((Gt ⊕g)⊙g)(nGt ) (7)
式中，
利用开、闭运算的组合可以对目标信号进行削峰平谷，从而分离出目标信号的平滑分量和非平滑分量。将式(6)和式(7)的开、闭运算进行级联组合，分别得到形态开-闭滤波器运行结果[Foc(Gt)](nGt)和形态闭-开滤波器运行结果[Fco(Gt)](nGt)为:The target signal can be clipped and flattened by the combination of open and closed operations, so as to separate the smooth component and the non-smooth component of the target signal. Combining the open and closed operations of equations (6) and (7) in cascade, the operation results of the morphological open-closed filter [Foc (Gt )](nGt ) and the operation of the morphological closed-open filter are obtained respectively. The result [Fco (Gt )](nGt ) is:
在利用数学形态学算法对目标信号开展分量分离时，为了减小统计偏倚现象的影响，本实施例采用形态开-闭滤波器和形态闭-开滤波器的混合运算，得到信号Gt(nGt)中的平滑分量[F1(Gt)](nGt)和非平滑分量[F2(Gt)](nGt)分别为：In order to reduce the influence of the statistical bias phenomenon when using the mathematical morphology algorithm to separate the components of the target signal, the present embodiment adopts the mixed operation of the morphological open-closed filter and the morphological closed-open filter to obtain the signal Gt (n The smoothed component [F1 (Gt )](nGt ) and the non-smoothed component [F2 (Gt )](nGt ) inGt ) are:
[F1(Gt)](nGt)＝([Foc(Gt)](nGt)+[Fco(Gt)](nGt))/2 (10)[F1 (Gt )](nGt )=([Foc (Gt )](nGt )+[Fco (Gt )](nGt ))/2 (10)
[F2(Gt)](nGt)＝Gt(nGt)-[F1(Gt)](nGt) (11)。[F2 (Gt )](nGt )=Gt (nGt )−[F1 (Gt )](nGt ) (11).
本实施例基于数学形态学进行信号特征提取，利用合适的结构元素去采集目标信号的特征信息，从而刻画出目标信号的结构特征。In this embodiment, signal feature extraction is performed based on mathematical morphology, and appropriate structural elements are used to collect feature information of the target signal, so as to describe the structural features of the target signal.
本实施例采用总体最小二乘不变旋转矢量技术(total least squares-estimation of signal parameters via rotational invariance techniques,TLE-ESPRIT)计算窄带干扰的特征参数，其具体计算流程如下：In this embodiment, the total least squares-estimation of signal parameters via rotational invariance techniques (TLE-ESPRIT) is used to calculate the characteristic parameters of narrowband interference, and the specific calculation process is as follows:
(1)将噪声帧内时域数据信号y(ny)构成Hankle矩阵H为：(1) Construct the Hankle matrix H of the time-domain data signal y(ny ) in the noise frame as:
式中，Ny是信号y(ny)的数据长度，L是H的列数，L的优选值为Ny/3。In the formula, Ny is the data length of the signal y(ny ), L is the number of columns of H, and the preferred value of L is Ny /3.
(2)将H开展奇异值分解，得到(2) Carry out singular value decomposition of H to get
H＝UΣVT (13)H=UΣVT (13)
式中，T是取矩阵的共轭转置；U和V分别是左、右正交矩阵；Σ是对角矩阵，其中的对角元素是矩阵H的奇异值。In the formula,T is the conjugate transpose of the matrix; U and V are the left and right orthogonal matrices, respectively; Σ is the diagonal matrix, and the diagonal elements are the singular values of the matrix H.
(3)分离出矩阵U中第1行到第L行，第1列到第P列的子矩阵为矩阵U1；同时分离出矩阵U中第2行到第L+1行，第1列到第P列的子矩阵为U2，以此得到U1和U2分别为：(3) The submatrices from the 1st row to the Lth row and the 1st column to the Pth column in the matrix U are separated to be the matrix U1 ; at the same time, the 2nd row to the L+1th row and the 1st column in the matrix U are separated. The submatrix to the P-th column is U2 , so that U1 and U2 are obtained as:
U1＝U[1:L,1:P] (14)U1 =U[1:L,1:P] (14)
U2＝U[2:L+1,1:P] (15)U2 =U[2:L+1,1:P] (15)
式中，P是分离后时频图谱确定的窄带干扰个数的2倍。In the formula, P is twice the number of narrowband interferences determined by the time-frequency spectrum after separation.
(4)构建矩阵Z＝[U1 U2]，然后对Z开展奇异值分解，得到(4) Construct the matrix Z=[U1 U2 ], and then carry out singular value decomposition on Z to obtain
式中：
(5)利用V12和V22构建矩阵ψ为(5) Use V12 and V22 to construct a matrix ψ as
ψ＝-V12V22-1 (18)ψ=-V12 V22-1 (18)
(6)对ψ开展特征值分解，得到对应的特征值为μk(k＝1,2,…,P)，从而估计得到各窄带干扰分量的频率
式中，arg(*)是计算复数的相角。where arg(*) is the phase angle of the complex number.
(7)利用最小二乘法计算各窄带干扰分量的幅值
Y＝μc (20)Y=μc (20)
Y＝[y(0)，y(1)，…，y(Ny-1)]T (21)Y=[y(0), y(1), ..., y(Ny -1)]T (21)
式中，c是中间变量矩阵，ck是c中元素。In the formula, c is the intermediate variable matrix, and ck is the element in c.
PD脉冲信号通常会呈现衰减振荡的波形特征，因此采用式(25)所示的单指数衰减振荡函数s1(t)和式(26)所示的双指数衰减振荡函数s2(t)对真实的局部放电信号进行模拟。The PD pulse signal usually exhibits the waveform characteristics of damping oscillation, so the single exponential damping oscillation function s1 (t) shown in equation (25) and the double exponential damping oscillation function s2 (t) shown in equation (26) are used to pair Real PD signals are simulated.
式中，a1和a2是仿真PD信号的幅值；τ1和τ2是仿真PD信号的衰减常数；fs1和fs2是仿真PD信号的振荡频率。In the formula, a1 and a2 are the amplitudes of the simulated PD signal; τ1 and τ2 are the attenuation constants of the simulated PD signal; fs1 and fs2 are the oscillation frequencies of the simulated PD signal.
周期性窄带干扰snoise(t)的表达式为：The expression of periodic narrowband interference snoise (t) is:
式中，Ak是窄带干扰的幅值；fk是窄带干扰的频率；
利用表1中参数构造4组仿真PD脉冲，如图1(a)所示。由于现场窄带干扰频带具有一定宽度，因此本实施例设计了多组不同频带的窄带干扰进行模拟测试。利用表2中参数构造5组窄带干扰，并将窄带干扰叠加在图1(a)中原始的PD仿真波形上，得到染噪的PD仿真波形如图1(b)所示，图1中波形的采样时长和采样频率分别设置为120μs和40MHz。值得说明的是，对比表1和表2可知，PD信号脉冲在脉冲序号1、3和窄带干扰的干扰序号3的频带存在重合现象，以模拟实际情况中局部放电信号频带与窄带干扰频带重合的情况。从图1中可以看出，在叠加了窄带干扰之后，原始PD信号几乎完全被窄带干扰淹没，因此无法直接从时域中将原始PD信号和窄带干扰进行分离，导致PD信号的识别和分析受到限制。Four groups of simulated PD pulses are constructed using the parameters in Table 1, as shown in Figure 1(a). Since the on-site narrow-band interference frequency band has a certain width, in this embodiment, multiple groups of narrow-band interference with different frequency bands are designed for simulation testing. Use the parameters in Table 2 to construct 5 groups of narrow-band interference, and superimpose the narrow-band interference on the original PD simulation waveform in Figure 1(a) to obtain the noise-dyed PD simulation waveform as shown in Figure 1(b), the waveform in Figure 1 The sampling duration and sampling frequency are set to 120μs and 40MHz, respectively. It is worth noting that, comparing Table 1 and Table 2, it can be seen that the frequency bands of the PD signal pulses in
表1仿真PD信号的参数表Table 1 Parameter table of simulated PD signal
表2仿真窄带干扰的参数表Table 2 Parameter table of simulated narrowband interference
PD脉冲信号通常会呈现衰减振荡的波形特征，因此可以采用式(25)所示的单指数衰减振荡函数s1(t)和式(26)所示的双指数衰减振荡函数s2(t)对真实的局部放电信号进行模拟。The PD pulse signal usually exhibits the waveform characteristics of damping oscillation, so the single exponential damping oscillation function s1 (t) shown in equation (25) and the double exponential damping oscillation function s2 (t) shown in equation (26) can be used Simulation of real PD signals.
利用广义S变换方法分别对图1(a)中原始的PD仿真波形和图1(b)中染噪的PD仿真波形进行时频分析，得到两者对应的时频谱图如图2所示。需要说明的是，本实施例将式(3)中λ优选设置为0.3。Using the generalized S transform method, the original PD simulation waveform in Fig. 1(a) and the noise-dyed PD simulation waveform in Fig. 1(b) are used for time-frequency analysis respectively, and the corresponding time-spectrograms of the two are obtained as shown in Fig. 2. It should be noted that, in this embodiment, λ in formula (3) is preferably set to 0.3.
在图2的时频谱图中，原始PD信号和窄带干扰呈现出明显不同的特点。原始PD信号在时间轴方向上的跨度较小，同时该信号在频率轴方向上的跨度较大；窄带干扰在频率轴方向上的跨度较小，同时该信号在时间轴方向上的跨度较大。综上所述，可以利用原始PD信号和窄带干扰在时频谱图中的不同特征，对两者进行分离。进一步对图2(b)中染噪PD信号的时频谱图开展分析，可以发现：当窄带干扰和原始PD信号的时频能量存在混叠时，由于窄带干扰的能量较强，所以窄带干扰会掩盖原始PD信号的时频特征，难以直接在时频谱图中区分窄带干扰和PD信号。若直接在时频谱图中区分窄带干扰和PD信号，具有很大的局限性，难以准确划分出窄带干扰和PD信号对应的时频子区域。In the time-spectrogram of Figure 2, the original PD signal and the narrowband interference exhibit significantly different characteristics. The span of the original PD signal in the direction of the time axis is small, and the span of the signal in the direction of the frequency axis is large; the span of the narrowband interference in the direction of the frequency axis is small, and the span of the signal in the direction of the time axis is large. . To sum up, the original PD signal and the narrowband interference can be separated by using their different characteristics in the time-spectrogram. Further analysis of the time-frequency spectrum of the noise-dyed PD signal in Figure 2(b) shows that when the time-frequency energy of the narrowband interference and the original PD signal are aliased, the narrowband interference will have strong energy due to the strong energy of the narrowband interference. Masking the time-frequency characteristics of the original PD signal makes it difficult to directly distinguish narrow-band interference and PD signals in the time-spectrogram. If the narrowband interference and the PD signal are directly distinguished in the time-spectrogram, it has great limitations, and it is difficult to accurately divide the time-frequency sub-regions corresponding to the narrowband interference and the PD signal.
对图2(b)中染噪PD信号时频谱图的每行数据开展单独分析，即单独分析各频率能量和时间的变化关系曲线，得到2组典型的变化关系曲线如图3所示。其中图3(a)的频率能量仅由原始PD信号构成，图3(b)的频率能量由原始PD信号和窄带干扰共同构成。这里需要说明的是，由于染噪PD信号的首端和末端存在数据的截断效应，所以染噪PD信号的广义S变换模矩阵在时间序列的首端和末端都存在畸变，为了准确分析染噪PD信号广义S变换模矩阵的特征，本实施例在广义S变换模矩阵时间序列的首端和末端都删除一小部分采样点。The data of each row of the spectrogram when the noise PD signal is dyed in Figure 2(b) are separately analyzed, that is, the change relationship curve of each frequency energy and time is analyzed separately, and two sets of typical change relationship curves are obtained as shown in Figure 3. The frequency energy in Fig. 3(a) is only composed of the original PD signal, and the frequency energy in Fig. 3(b) is composed of the original PD signal and narrowband interference. It should be noted here that due to the truncation effect of the data at the head and end of the noise-dyed PD signal, the generalized S-transform mode matrix of the noise-dyed PD signal is distorted at the head and end of the time series. The characteristics of the generalized S-transform mode matrix of the PD signal, in this embodiment, a small part of sampling points are deleted at the beginning and the end of the generalized S-transform mode matrix time series.
分析图3可以看出，在各频率能量和时间的变化关系曲线中，窄带干扰会呈现为平滑分量，而原始PD信号是一个“丘状”结构，呈现为非平滑分量，本实施例利用该特点将两者进行区分，以分离出染噪PD信号时频谱图中窄带干扰与原始PD信号。本实施例采用数学形态学方法分离信号中平滑分量和非平滑分量，进而分离染噪PD信号时频谱图中窄带干扰和PD信号。It can be seen from the analysis of Fig. 3 that in the relationship curve between the energy and time of each frequency, the narrowband interference will appear as a smooth component, while the original PD signal is a "hill" structure and appears as a non-smooth component, which is used in this embodiment. The feature distinguishes the two to separate the narrowband interference and the original PD signal in the spectrogram when the PD signal is stained with noise. In this embodiment, the mathematical morphology method is used to separate the smooth component and the non-smooth component in the signal, so as to separate the narrow-band interference and the PD signal in the frequency spectrum of the noise-dyed PD signal.
通常局部放电的持续时间小于1μs，在测试回路中电容和电感的影响下，实际测试的PD波形的持续时间通常会大于1μs。为了有效分离染噪PD信号时频谱图中原始PD信号和窄带干扰，本实施例中结构元素选择为时间长度为5μs的扁平结构元素。Usually the duration of partial discharge is less than 1μs. Under the influence of capacitance and inductance in the test loop, the duration of the PD waveform actually tested is usually greater than 1μs. In order to effectively separate the original PD signal and the narrowband interference in the spectrogram when the PD signal is stained with noise, the structural element in this embodiment is selected as a flat structural element with a time length of 5 μs.
依照本实施例，首先将染噪PD信号时频谱图中各行分量作为目标信号，利用数学形态学方法分离目标信号的平滑分量和非平滑分量，然后将分离出的平滑分量和非平滑分量作为行分量构成新的矩阵，以此分离出窄带干扰和原始PD信号的时频谱图，从而提取窄带干扰和原始PD信号的特征量。为了方便观察，本实施例对分离出的原始PD信号时频谱图进行了取模处理。According to this embodiment, firstly, each line component in the spectrogram of the noise PD signal is used as the target signal, and the smooth component and the non-smooth component of the target signal are separated by mathematical morphology, and then the separated smooth component and non-smooth component are used as the line. The components form a new matrix to separate the time-spectrogram of the narrowband interference and the original PD signal, thereby extracting the feature quantities of the narrowband interference and the original PD signal. For the convenience of observation, in this embodiment, modulo processing is performed on the time-frequency spectrum of the separated original PD signal.
利用数学形态学方法对图2(b)中染噪PD信号的时频谱图进行处理，分离出窄带干扰和原始PD信号的时频谱图分别如图4(a)和图4(b)所示。首先分析图4(a)中分离出的窄带干扰时频谱图，依照窄带干扰的时频特性，可以确定仿真的染噪PD信号中存在5组窄带干扰。接着分析图4(b)中分离出的原始PD信号时频谱图，由于部分PD信号和窄带干扰在染噪PD信号时频谱图中发生了时频能量的混叠，所以图4(b)中部分PD信号(仿真的局放脉冲1和局放脉冲3)的时频能量分布与图2(a)中原始PD信号的时频能量分布产生了差异，但是仍然具备原始PD信号的时频特性，因此依旧可以按照PD信号的时频特性，确定PD脉冲的时间区域。The time-spectrogram of the noisy PD signal in Fig. 2(b) is processed by mathematical morphology method, and the time-spectrograms of the narrowband interference and the original PD signal are separated as shown in Fig. 4(a) and Fig. 4(b) respectively. . First, analyze the time-spectrogram of the narrowband interference separated in Figure 4(a). According to the time-frequency characteristics of the narrowband interference, it can be determined that there are 5 groups of narrowband interference in the simulated noise-dyed PD signal. Then analyze the spectrogram of the original PD signal separated in Fig. 4(b), because some PD signals and narrowband interference have time-frequency energy aliasing in the spectrogram when the noise PD signal is dyed, so in Fig. 4(b) The time-frequency energy distribution of some PD signals (simulated
当采样的时间窗口较长时，PD信号仅会占据较小的时间，更多的时间是由窄带干扰等噪声所占据，所以可以将采样得到的染噪PD信号根据时间轴划定为含有PD信号的信号帧和仅含窄带干扰等噪声的噪声帧。When the sampling time window is long, the PD signal will only occupy a small amount of time, and more time is occupied by noise such as narrowband interference, so the sampled noise-dyed PD signal can be classified as containing PD according to the time axis. The signal frame of the signal and the noise frame containing only noise such as narrowband interference.
由于染噪PD信号在噪声帧内不含有PD信号，因此可以利用TLE-ESPRIT方法分析噪声帧内时域数据，以准确提取窄带干扰的时域特征量。需要说明的是，利用噪声帧内时域数据计算得到的相位仅对应噪声帧内的窄带干扰，为计算信号帧中相位以重构窄带干扰，需要将式(23)中
式中，ΔT是信号帧和噪声帧的时延。where ΔT is the time delay between the signal frame and the noise frame.
通过TLE-ESPRIT提取窄带干扰的时域特征量后，然后利用式(28)可以重构出窄带干扰的时域波形，最后从染噪PD信号的时域波形中去除窄带干扰的时域波形，得到降噪后PD信号。After extracting the time-domain feature of the narrowband interference by TLE-ESPRIT, the time-domain waveform of the narrowband interference can be reconstructed by using equation (28), and finally the time-domain waveform of the narrowband interference can be removed from the time-domain waveform of the noisy PD signal, The PD signal after noise reduction is obtained.
对图1(b)所示含有窄带干扰的染噪PD信号利用本实施例进行降噪，得到降噪结果如图5(a)所示。为了便于比较，本实施例引入FFT谱最小熵解卷积滤波方法和自适应奇异值分解方法对仿真染噪PD信号进行窄带干扰降噪，分别得到对应的降噪结果如图5(b)和图5(c)所示。The noise-dyed PD signal containing narrowband interference shown in FIG. 1(b) is denoised by using this embodiment, and the denoising result is obtained as shown in FIG. 5(a). For the convenience of comparison, the FFT spectrum minimum entropy deconvolution filtering method and the adaptive singular value decomposition method are introduced in this embodiment to perform narrowband interference noise reduction on the simulated noisy PD signal, and the corresponding noise reduction results are obtained as shown in Figure 5(b) and shown in Figure 5(c).
由图5可知，对于FFT谱最小熵解卷积滤波方法而言，由于傅里叶变换算法存在频谱泄露的特点，导致该方法的最终降噪结果存在明显的“边缘效应”(边缘处的噪声较大)，同时该方法中阈值法无法提取小幅值窄带干扰的频点，因此该方法难以抑制小幅值的窄带干扰，导致整个降噪波形中残留较多的窄带干扰能量。对于自适应奇异值分解方法而言，由于PD波形和小幅值窄带干扰的奇异值数值非常接近，甚至会导致奇异值的能量产生相互的混叠，所以最终的降噪结果存在窄带干扰抑制不干净的情况。利用本实施例对染噪的仿真PD波形进行降噪时，既可以最大程度地削弱窄带干扰的能量，同时也能保留PD信号的波形特征。As can be seen from Figure 5, for the FFT spectral minimum entropy deconvolution filtering method, due to the characteristics of spectral leakage in the Fourier transform algorithm, the final noise reduction result of this method has obvious "edge effect" (noise at the edge. At the same time, the threshold method in this method cannot extract the frequency points of small-amplitude narrow-band interference, so this method is difficult to suppress small-amplitude narrow-band interference, resulting in more residual narrow-band interference energy in the entire noise reduction waveform. For the adaptive singular value decomposition method, since the singular value values of the PD waveform and the small-amplitude narrow-band interference are very close, even the energy of the singular values will be aliased with each other, so the final noise reduction result has a narrow-band interference suppression. clean condition. When the noise-stained simulated PD waveform is denoised by using this embodiment, the energy of the narrowband interference can be weakened to the greatest extent, and the waveform characteristics of the PD signal can also be preserved at the same time.
为了对图5中各方法的实际降噪结果开展量化分析，本实施例采用了信噪比(signal to noise ratio，SNR)、波形相似系数(normalized correlation coefficient，NCC)和均方误差(mean square error，MSE)三组特征参数对仿真PD波形的降噪效果进行评估，计算得到图5中各方法降噪结果的评价参数如表3所示。对比表3中的各项参数可以看出，本实施例的SNR和NNC值最大，同时MSE值最小，说明本实施例对窄带干扰的降噪效果最好，对原始PD信号的还原效果最佳。In order to carry out quantitative analysis on the actual noise reduction results of each method in FIG. 5 , this embodiment adopts a signal to noise ratio (SNR), a waveform similarity coefficient (normalized correlation coefficient, NCC), and a mean square error (mean square error). error, MSE) three sets of characteristic parameters to evaluate the noise reduction effect of the simulated PD waveform, and the evaluation parameters of the noise reduction results of each method in Fig. Comparing the parameters in Table 3, it can be seen that the SNR and NNC values of this embodiment are the largest, while the MSE value is the smallest, indicating that this embodiment has the best noise reduction effect on narrowband interference and the best restoration effect on the original PD signal. .
表3降噪结果的评价参数表Table 3 Evaluation parameter table of noise reduction results
通常情况下，实际采集的PD信号会同时含有白噪声干扰和窄带干扰，为了验证本实施例的适用性，先对图1(a)的仿真原始PD波形添加2dB的白噪声干扰，再加入仿真的窄带干扰，得到对应混合干扰下的PD波形如图6所示，此时计算得到染噪PD信号的信噪比为-26.5915dB。Usually, the actually collected PD signal will contain both white noise interference and narrowband interference. In order to verify the applicability of this embodiment, 2dB white noise interference is added to the simulated original PD waveform in Figure 1(a) first, and then the simulation is added. Figure 6 shows the PD waveform under the corresponding mixed interference, and the signal-to-noise ratio of the noise-dyed PD signal is calculated to be -26.5915dB.
利用本实施例分离图6中染噪PD波形的时频谱图，得到分离后窄带干扰和PD信号的时频谱图分别如图7(a)和图7(b)所示。从图7中可以看出，在染噪PD信号存在白噪声干扰的情况下，本实施例仍旧可以有效分离出窄带干扰和PD信号的时频谱图，其中分离出的窄带干扰时频谱图无异常变化，可以准确判断窄带干扰数量；分离出的PD信号时频谱图中会存在白噪声能量，但是PD信号的时频特征仍然清晰可见，可以准确划定PD信号的时频区域，确定染噪PD信号的信号帧和噪声帧。Using this embodiment to separate the time-spectrograms of the noise-dyed PD waveforms in FIG. 6 , the time-spectrograms of the separated narrowband interference and PD signals are obtained as shown in FIG. 7( a ) and FIG. 7( b ), respectively. As can be seen from FIG. 7 , in the case where the noise-dyed PD signal has white noise interference, the present embodiment can still effectively separate the time-spectrogram of the narrow-band interference and the PD signal, and the time-spectrogram of the separated narrow-band interference is not abnormal. changes, the amount of narrowband interference can be accurately judged; there will be white noise energy in the spectrum of the separated PD signal, but the time-frequency characteristics of the PD signal are still clearly visible, which can accurately delineate the time-frequency region of the PD signal and determine the PD Signal frame and noise frame of the signal.
利用本实施例对图6中染噪PD信号的窄带干扰进行抑制，得到对应的降噪结果如图8所示。从图8中可以看出，此时已经可以明显观察到PD信号的波形，计算得到此时的信噪比已经降为1.4571dB。上述结果证明了：在存在白噪声干扰的情况下，本实施例仍旧可以有效抑制染噪PD信号中窄带干扰，从而提高染噪PD信号的信噪比。The narrowband interference of the noisy PD signal in FIG. 6 is suppressed by using this embodiment, and the corresponding noise reduction result is obtained as shown in FIG. 8 . As can be seen from Figure 8, the waveform of the PD signal can be clearly observed at this time, and the calculated signal-to-noise ratio at this time has been reduced to 1.4571dB. The above results prove that in the presence of white noise interference, this embodiment can still effectively suppress narrowband interference in the noise-dyed PD signal, thereby improving the signal-to-noise ratio of the noise-dyed PD signal.
为了验证本实施例抑制实测染噪PD信号中窄带干扰的有效性，在实验室中搭建10kV交联聚乙烯电力电缆的工频局放测试系统，局放缺陷设置为电缆终端处存在半导电层突起缺陷，采样频率设置为100MHz，采样点数设置为4000。由于实验室近似于无噪环境，人为向实测PD信号中叠加5组窄带干扰，幅值分别设置为20、8、25、35、25mV，频率分别设置为2.51、5.36、10.32、15.33、22.42MHz，相位分别设置为45°、60°、30°、30°和90°，得到实测的染噪PD信号如图9所示。从图9中可以看出，在窄带干扰的影响下，原始PD信号已经严重被污染，难以进行特征分析。In order to verify the effectiveness of this example in suppressing the narrow-band interference in the measured dye-noise PD signal, a power frequency partial discharge test system of 10kV XLPE power cable was built in the laboratory, and the partial discharge defect was set as the existence of a semiconductive layer at the cable terminal For protrusion defects, the sampling frequency is set to 100MHz, and the number of sampling points is set to 4000. Since the laboratory is close to a noise-free environment, 5 groups of narrowband interference were artificially added to the measured PD signal, the amplitudes were set to 20, 8, 25, 35, and 25mV, and the frequencies were set to 2.51, 5.36, 10.32, 15.33, and 22.42MHz. , and the phases are set to 45°, 60°, 30°, 30°, and 90°, respectively, and the measured noise-dyed PD signal is shown in Figure 9. As can be seen from Figure 9, under the influence of narrowband interference, the original PD signal has been seriously polluted, making it difficult to perform feature analysis.
利用广义S变换对图9中实测的染噪PD信号进行时频分析，得到对应的时频谱图如图10所示，从图10中可以看出，由于窄带干扰的能量较强，所以在实测染噪PD信号的时频谱图中，PD信号的时频特征已经完全被掩盖，直接在该时频谱图中划定窄带干扰和PD信号的方法难以实施，具备较大的局限性。Use generalized S transform to perform time-frequency analysis on the measured noise-dyed PD signal in Fig. 9, and obtain the corresponding time-spectrogram as shown in Fig. 10. It can be seen from Fig. 10 that due to the strong energy of narrow-band interference, the measured In the time-spectrogram of the noisy PD signal, the time-frequency characteristics of the PD signal have been completely covered up. The method of directly delineating narrowband interference and PD signals in the time-spectrogram is difficult to implement and has great limitations.
利用本实施例对图10实测PD信号的时频谱图进行分离，得到分离后窄带干扰和PD信号对应的时频谱图如图11所示，从图11中可以看出，此时窄带干扰和PD信号的时频谱图得到了有效的分离，利用分离出的窄带干扰时频谱图可以确定窄带干扰个数为5，同时利用分离出的PD信号时频谱图可以划定信号帧和噪声帧，以确保利用TLS-ESPRIT准确提取窄带干扰特征参数，重构窄带干扰，得到最终的降噪结果如图12(a)所示。Using this embodiment to separate the time-spectrogram of the PD signal measured in Figure 10, the time-spectrogram corresponding to the narrowband interference and the PD signal after separation is obtained as shown in Figure 11. It can be seen from Figure 11 that at this time, the narrowband interference and PD The time-spectrogram of the signal has been effectively separated. Using the separated time-spectrogram of the narrow-band interference, the number of narrow-band interference can be determined to be 5. At the same time, the time-spectrogram of the separated PD signal can be used to delineate the signal frame and the noise frame. Use TLS-ESPRIT to accurately extract the narrowband interference feature parameters, reconstruct the narrowband interference, and obtain the final noise reduction result as shown in Figure 12(a).
为了说明本实施例的优越性，再分别利用FFT谱最小熵解卷积滤波法和自适应奇异值分解法对图9中实测的染噪PD信号进行降噪处理，分别得到降噪结果如图12(b)和图12(c)所示。从图12中可以看出，利用FFT谱最小熵解卷积滤波法进行降噪时，降噪结果中残余的窄带干扰能量较强，尤其是在信号的边缘部分；利用自适应奇异值分解法进行降噪时，虽然大部分窄带干扰能够被有效抑制，但是会残留部分的小幅值窄带干扰；而采用本实施例进行降噪时，降噪结果中窄带干扰的残余能量较小，同时PD信号的波形特征得到了保留。In order to illustrate the superiority of this embodiment, the deconvolution filtering method with minimum entropy of FFT spectrum and the adaptive singular value decomposition method are used to denoise the PD signal measured in Fig. 9, respectively, and the denoising results are obtained as shown in Fig. 12(b) and Fig. 12(c). It can be seen from Figure 12 that when noise reduction is performed using the FFT spectral minimum entropy deconvolution filtering method, the residual narrowband interference energy in the noise reduction result is strong, especially at the edge of the signal; using the adaptive singular value decomposition method When noise reduction is performed, although most of the narrowband interference can be effectively suppressed, some small-amplitude narrowband interference will remain; however, when noise reduction is performed using this embodiment, the residual energy of the narrowband interference in the noise reduction result is small, and the PD The waveform characteristics of the signal are preserved.
因为难以采集到完全没有噪声的PD信号，所以不能采用SNR、NCC、MSE等参数对实测PD信号的降噪效果进行量化分析，本实施例引入噪声抑制比ρ抑对图12中降噪效果进行评估。该参数可以表现降噪结果中PD信号的凸显程度，该值越大，说明降噪结果更好。ρ抑的具体定义为：Because it is difficult to collect a PD signal without noise at all, parameters such as SNR, NCC, and MSE cannot be used toquantify the noise reduction effect of the measured PD signal. Evaluate. This parameter can represent the prominence of the PD signal in the noise reduction result. The larger the value, the better the noise reduction result. The specific definition ofρ is:
式中，σ1是降噪前信号的标准差，σ2是降噪后信号的标准差。In the formula, σ1 is the standard deviation of the signal before noise reduction, and σ2 is the standard deviation of the signal after noise reduction.
计算得到本实施例的方法、FFT谱最小熵解卷积滤波法和自适应奇异值分解法的噪声抑制比结果分别为30.7559、7.3017、16.4680，通过对比可以发现，本实施例的ρ抑值更大，说明本实施例可以有效抑制窄带干扰，以恢复PD信号的波形特征。值得说明的是，由图11可知，在本实施例实测的染噪PD信号中，PD信号和窄带干扰存在明显的频带重叠，而本实施例依然可以取得较好的窄带干扰抑制效果。The noise suppression ratio results of the method of this embodiment, the FFT spectrum minimum entropy deconvolution filtering method and the adaptive singular value decomposition method are 30.7559, 7.3017, and 16.4680, respectively. By comparison, it can be found that the ρsuppression value of this embodiment is higher. large, indicating that this embodiment can effectively suppress narrowband interference, so as to restore the waveform characteristics of the PD signal. It is worth noting that, as can be seen from FIG. 11 , in the noise-dyed PD signal measured in this embodiment, the PD signal and narrowband interference have obvious frequency band overlap, and this embodiment can still achieve a better narrowband interference suppression effect.
本实施例提出了一种基于时频谱图分离的局部放电窄带干扰抑制方法，该方法可以有效降低染噪PD信号中窄带干扰的能量，恢复PD信号的波形特征。本实施例的具体结论如下：This embodiment proposes a partial discharge narrowband interference suppression method based on time-spectrogram separation, which can effectively reduce the energy of the narrowband interference in the noise-dyed PD signal and restore the waveform characteristics of the PD signal. The specific conclusions of this embodiment are as follows:
(1)通过分析染噪PD信号的时频谱图，明确当窄带干扰的能量较强时，窄带干扰会掩盖原始PD信号的时频特征，难以直接在染噪PD信号的时频谱图中辨识原始PD信号和窄带干扰。(1) By analyzing the time-spectrogram of the noise-dyed PD signal, it is clear that when the energy of the narrow-band interference is strong, the narrow-band interference will mask the time-frequency characteristics of the original PD signal, and it is difficult to directly identify the original PD signal in the time-spectrogram of the noise-dyed PD signal. PD signals and narrowband interference.
(2)结合广义S变换和数学形态学，本实施例所提方法可以从染噪PD信号中分离出原始PD信号和窄带干扰信号的时频谱图，从而在时频域内准确辨识出窄带干扰和原始PD信号。(2) Combined with generalized S transform and mathematical morphology, the method proposed in this embodiment can separate the time-spectrogram of the original PD signal and the narrowband interference signal from the noisy PD signal, so as to accurately identify the narrowband interference and the narrowband interference in the time-frequency domain. Raw PD signal.
(3)在分离出原始PD信号和窄带干扰信号的时频谱图的基础上，利用TLS-ESPRIT算法可以准确估计窄带干扰的特征参数，实现窄带干扰的抑制。(3) On the basis of separating the time-spectrogram of the original PD signal and the narrowband interference signal, the TLS-ESPRIT algorithm can accurately estimate the characteristic parameters of the narrowband interference and realize the suppression of the narrowband interference.
(4)仿真和实测染噪PD信号的降噪结果表明：对比传统的FFT谱最小熵解卷积滤波法和自适应奇异值分解法，本实施例具备更好的窄带干扰抑制效果，有利于小幅值窄带干扰的抑制。(4) The noise reduction results of the simulated and measured PD signals with dyed noise show that: compared with the traditional FFT spectral minimum entropy deconvolution filtering method and the adaptive singular value decomposition method, this embodiment has better narrowband interference suppression effect, which is beneficial to Suppression of small amplitude narrowband interference.
利用本实施例的方法、FFT谱最小熵解卷积滤波方法和自适应奇异值分解方法对仿真和实测的染噪PD信号开展降噪处理，通过对比降噪结果，验证了本实施例可以有效抑制染噪PD信号中窄带干扰，并且降噪结果中残余的窄带干扰能量较小。Using the method of this embodiment, the FFT spectrum minimum entropy deconvolution filtering method and the adaptive singular value decomposition method to carry out noise reduction processing on the simulated and measured noise-dyed PD signals, and by comparing the noise reduction results, it is verified that this embodiment can effectively The narrowband interference in the noisy PD signal is suppressed, and the residual narrowband interference energy in the noise reduction result is small.
以上所述的具体实施方式，对本发明的目的、技术方案和有益效果进行了进一步详细说明，所应理解的是，以上所述仅为本发明的具体实施方式而已，并不用于限定本发明的保护范围，凡在本发明的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。The specific embodiments described above further describe the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims (5)
Translated fromChinese
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210472721.7ACN114861722B (en) | 2022-04-29 | 2022-04-29 | Partial discharge narrow-band interference suppression method based on time-frequency spectrogram separation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210472721.7ACN114861722B (en) | 2022-04-29 | 2022-04-29 | Partial discharge narrow-band interference suppression method based on time-frequency spectrogram separation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114861722Atrue CN114861722A (en) | 2022-08-05 |
| CN114861722B CN114861722B (en) | 2023-04-11 |
Family
ID=82635419
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210472721.7AActiveCN114861722B (en) | 2022-04-29 | 2022-04-29 | Partial discharge narrow-band interference suppression method based on time-frequency spectrogram separation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114861722B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115980528A (en)* | 2023-02-14 | 2023-04-18 | 湖南湘能智能配电设备有限公司 | Ring main unit noise signal rapid diagnosis and elimination method and device |
| CN116486836A (en)* | 2023-03-29 | 2023-07-25 | 昆明理工大学 | A SNR Estimation Method for Transmitted Speech |
| CN119782873A (en)* | 2024-11-27 | 2025-04-08 | 国网江苏省电力有限公司连云港供电分公司 | A method and system for intelligent simulation test of primary and secondary fusion centralized DTU equipment |
| CN119916147A (en)* | 2024-12-19 | 2025-05-02 | 国网安徽省电力有限公司电力科学研究院 | Active suppression method and system for background noise of partial discharge ultra-high frequency signal |
| CN119959711A (en)* | 2025-04-10 | 2025-05-09 | 国网山西省电力公司电力科学研究院 | Spectral signal feature extraction method for GIS equipment and discharge fault diagnosis method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090196524A1 (en)* | 2008-02-05 | 2009-08-06 | Dts Digital Images, Inc. | System and method for sharpening of digital images |
| CN103245832A (en)* | 2013-05-16 | 2013-08-14 | 湖南大学 | Harmonic time frequency characteristic parameter estimating method based on fast S conversion and analysis meter |
| CN111308285A (en)* | 2020-03-03 | 2020-06-19 | 西南交通大学 | Narrow-band interference noise reduction method |
| CN111460367A (en)* | 2020-03-20 | 2020-07-28 | 淮阴工学院 | Algorithm for solving unbalanced data leakage of halogen conveying pipeline based on S transformation/WGAN |
| CN113777449A (en)* | 2021-08-31 | 2021-12-10 | 云南电网有限责任公司昆明供电局 | Narrowband interference suppression method for cable partial discharge based on improved SVD algorithm |
- 2022
- 2022-04-29CNCN202210472721.7Apatent/CN114861722B/enactiveActive
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090196524A1 (en)* | 2008-02-05 | 2009-08-06 | Dts Digital Images, Inc. | System and method for sharpening of digital images |
| CN103245832A (en)* | 2013-05-16 | 2013-08-14 | 湖南大学 | Harmonic time frequency characteristic parameter estimating method based on fast S conversion and analysis meter |
| CN111308285A (en)* | 2020-03-03 | 2020-06-19 | 西南交通大学 | Narrow-band interference noise reduction method |
| CN111460367A (en)* | 2020-03-20 | 2020-07-28 | 淮阴工学院 | Algorithm for solving unbalanced data leakage of halogen conveying pipeline based on S transformation/WGAN |
| CN113777449A (en)* | 2021-08-31 | 2021-12-10 | 云南电网有限责任公司昆明供电局 | Narrowband interference suppression method for cable partial discharge based on improved SVD algorithm |
Non-Patent Citations (3)
| Title |
|---|
| 宋立业等: "基于广义S 变换和随机子空间的局放窄带干扰抑制方法", 《电工电能新技术》* |
| 徐永干等: "基于Hankel 矩阵和奇异值分解的", 《电网技术》* |
| 王文浩等: "金属氧化物避雷器泄漏电流在线监测数据自适应滤波算法", 《高压电器》* |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115980528A (en)* | 2023-02-14 | 2023-04-18 | 湖南湘能智能配电设备有限公司 | Ring main unit noise signal rapid diagnosis and elimination method and device |
| CN115980528B (en)* | 2023-02-14 | 2024-03-26 | 湖南湘能智能配电设备有限公司 | Ring main unit noise signal rapid diagnosis and elimination method and device |
| CN116486836A (en)* | 2023-03-29 | 2023-07-25 | 昆明理工大学 | A SNR Estimation Method for Transmitted Speech |
| CN119782873A (en)* | 2024-11-27 | 2025-04-08 | 国网江苏省电力有限公司连云港供电分公司 | A method and system for intelligent simulation test of primary and secondary fusion centralized DTU equipment |
| CN119916147A (en)* | 2024-12-19 | 2025-05-02 | 国网安徽省电力有限公司电力科学研究院 | Active suppression method and system for background noise of partial discharge ultra-high frequency signal |
| CN119959711A (en)* | 2025-04-10 | 2025-05-09 | 国网山西省电力公司电力科学研究院 | Spectral signal feature extraction method for GIS equipment and discharge fault diagnosis method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114861722B (en) | 2023-04-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114861722A (en) | Partial discharge narrow-band interference suppression method based on time-frequency spectrogram separation | |
| Liu et al. | Synchrosqueezing extracting transform and its application in bearing fault diagnosis under non-stationary conditions | |
| Pachori et al. | Cross-terms reduction in the Wigner–Ville distribution using tunable-Q wavelet transform | |
| CN107957566B (en) | Magnetic resonance depth measurement method for extracting signal based on frequency selection singular spectrum analysis | |
| CN108287073B (en) | Resonance band selection method based on singular value component frequency domain spectrum | |
| CN102323518A (en) | A Partial Discharge Signal Recognition Method Based on Spectral Kurtosis | |
| CN111308285A (en) | Narrow-band interference noise reduction method | |
| Ma et al. | Noise reduction for desert seismic data using spectral kurtosis adaptive bandpass filter | |
| Ray et al. | A new threshold function for de-noising partial discharge signal based on wavelet transform | |
| Hong et al. | Noise reduction method of ship radiated noise with ensemble empirical mode decomposition of adaptive noise | |
| Cao et al. | A method for extracting weak impact signal in NPP based on adaptive Morlet wavelet transform and kurtosis | |
| Liu et al. | A novel wavelet selection scheme for partial discharge signal detection under low SNR condition | |
| Chan et al. | Hybrid method on signal de‐noising and representation for online partial discharge monitoring of power transformers at substations | |
| US7676046B1 (en) | Method of removing noise and interference from signal | |
| George et al. | Time localised band filtering using modified S-transform | |
| Boufenar et al. | Time-frequency analysis techniques review and their application on roller bearings prognostics | |
| Ray et al. | Extract Partial Discharge signal using wavelet for on-line measurement | |
| CN117520916A (en) | Cable partial discharge signal denoising method | |
| Yang et al. | A novel adaptive detection method for generic power quality waveform abnormality | |
| Li | Research on extraction of partial discharge signals based on wavelet analysis | |
| Zhang et al. | Simultaneous denoising and preserving of seismic signals by multiscale time-frequency peak filtering | |
| Rao et al. | Detection and localization of narrow‐band transient signals using the Wigner distribution | |
| Dong et al. | Distribution Cable Partial Discharge Hybrid Noise Suppression Method based on Adaptive Dual Complementary Convolutional Network | |
| Wu et al. | A partial discharge denoising method based on ICEEMDAN-adaptive threshold | |
| Ma et al. | Voltage Estimation Method for Distribution Network Based on Wiener Filter |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |