Method for on-line identification of vibration signal of GIS (Geographic Information System)

Abstract

The invention discloses a vibration signal online identification method for a GIS, and the method comprises the steps: firstly, monitoring GIS electromagnetic vibration signals including a contact vibration signal and a housing vibration signal, partial discharge vibration signals including a gas vibration signal and a basin-type insulator vibration signal, and connection part vibration signals including a wiring terminal vibration signal and a flange signal; and then noise reduction is carried out on the electromagnetic vibration signals, the partial discharge vibration signals and the connection part vibration signals by adopting a super-directivity beam forming algorithm, and online identification is carried out on the vibration signals by adopting a DWT-MFCC characteristic variance algorithm, so that online identification of the GIS vibration signals is realized, the operation state of the GIS is monitored in real time, and the reliability of the GIS is improved. And a data basis is provided for real-time analysis of the operation state of the GIS.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and particularly to a method for on-line identification of vibration signals for GIS. Background Art

[0002] The monitoring of the operating state of GIS is a research field that power grid enterprises attach great importance to. If the different positions of the equipment can be on-line identified through known vibration signals, it will not only be beneficial to the selection of equipment operation and maintenance strategies but also guide the power grid planning. As the most important equipment in the substation, the separation and on-line identification of GIS vibration signals are beneficial to the safe and stable operation of GIS, contribute to extending the service life of the substation, and improve the safety and economy of power grid operation.

[0003] Through on-line signal acquisition, signal denoising, and signal on-line identification, the real-time monitoring of the operating conditions of GIS during operation can be further realized. The operating state of GIS directly affects the safe and stable operation of GIS. Therefore, correct and timely monitoring of the operating state can not only predict GIS accidents in advance but also avoid huge economic losses caused by such accidents. Therefore, carrying out research on the vibration characteristics and contact state monitoring method of GIS is of great significance for ensuring the safe and reliable operation of the power system.

[0004] At present, the monitoring of the operating state of GIS is all offline detection methods. By collecting detection signals, the offline monitoring of the operating state of GIS is realized, and the operating state of GIS cannot be monitored in real time, nor can the operating state of GIS be analyzed in real time at all times.

[0005] Therefore, the existing technology still needs to be improved. Summary of the Invention

[0006] In view of the deficiencies of the above-mentioned prior art, the purpose of the present invention is to provide a method for on-line identification of vibration signals for GIS, aiming to solve the problem that the monitoring of the operating state of GIS in the prior art cannot monitor the operating state of GIS in real time and cannot analyze the operating state of GIS in real time at all times.

[0007] To achieve the above purpose, the present invention adopts the following technical solutions:

[0008] A method for on-line identification of vibration signals for GIS, comprising:

[0009] Step 1: Measure the multi-vibration sources of the GIS. Measure the electromagnetic vibration signals, partial discharge vibration signals, and connection component vibration signals of the GIS through piezoelectric sensors. The electromagnetic vibration signals include shell vibration signals and contact vibration signals. The partial discharge vibration signals include gas vibration signals and basin insulator vibration signals. The connection component vibration signals include terminal vibration signals and flange vibration signals. Establish a multi-vibration source model based on the electromagnetic vibration signals, partial discharge vibration signals, and connection component vibration signals.

[0010] Step 2: Perform noise reduction on the superdirective beamforming algorithm based on the multi-vibration source model established in Step 1. Combine the signals received by multiple sensors, and perform weighted and phase adjustment on them to form a superdirective beam to enhance the signals in the target direction and suppress the signals in other directions, and obtain the multi-vibration source signals after noise reduction.

[0011] Step 3: Identify based on the multi-vibration source signals after noise reduction obtained in Step 2 through the DWT-MFCC identification algorithm.

[0012] Further, in Step 1, when measuring the multi-vibration sources of the GIS, the obtained electromagnetic vibration signal matrix of the GIS is [C1, C2, C3, C4, C5, C6, C7, C8, C9,], the obtained partial discharge vibration signal matrix of the GIS is [A1, A2, A3, A4, A5, A6], and the obtained connection component vibration signal matrix of the GIS is

[0013] [Q1, Q2, Q3, Q4, Q5, Q6]. Establish a multi-vibration source model through the above vibration signal matrices.

[0014] Further, among the electromagnetic vibration signals [C1, C2, C3, C4, C5, C6, C7, C8, C9,], C1 is a piezoelectric vibration sensor built in at the left half position of the GIS tank to measure the vibration intensity at the half position of the GIS electromagnetic vibration signal; C2 is a piezoelectric vibration sensor built in at the middle half position of the GIS tank to measure the vibration intensity at the half position of the GIS electromagnetic vibration signal; C3 is a piezoelectric vibration sensor built in at the right half position of the GIS tank to measure the vibration intensity at the right half position of the GIS tank; C4 is a piezoelectric vibration sensor built in at the one-third position of the GIS electromagnetic vibration signal to measure the vibration intensity at the one-third position of the GIS electromagnetic vibration signal; C5 is a piezoelectric vibration sensor built in at the middle one-third position of the GIS tank to measure the vibration intensity at the one-third position of the GIS electromagnetic vibration signal; C6 is a piezoelectric vibration sensor built in at the one-third position of the GIS electromagnetic vibration signal to measure the vibration intensity at the right one-third position of the GIS tank; C7 is a piezoelectric vibration sensor built in at the left top position of the GIS tank to measure the vibration intensity at the top of the GIS electromagnetic vibration signal; C8 is a piezoelectric vibration sensor built in at the middle top position of the GIS tank to measure the vibration intensity at the top of the GIS electromagnetic vibration signal; C9 is a piezoelectric vibration sensor built in at the right top position of the GIS tank to measure the vibration intensity of the GIS electromagnetic vibration signal.

[0015] Further, among the partial discharge vibration signals [A1, A2, A3, A4, A5, A6], A1 is a piezoelectric vibration sensor built in at the one-third position of the upper half surface of the contact of the GIS to measure the vibration intensity at the one-third position of the GIS partial discharge vibration signal; A2 is a piezoelectric vibration sensor built in at the two-thirds position of the upper half surface of the contact of the GIS to measure the vibration intensity at the two-thirds position of the GIS partial discharge vibration signal; A3 is a piezoelectric vibration sensor built in at the top position of the upper half surface of the contact of the GIS to measure the vibration intensity of the GIS partial discharge vibration signal; A4 is a piezoelectric vibration sensor built in at the one-third position of the lower half surface of the contact of the GIS to measure the vibration intensity at the one-third position of the GIS partial discharge vibration signal; A5 is a piezoelectric vibration sensor built in at the two-thirds position of the lower half surface of the contact of the GIS to measure the vibration intensity at the two-thirds position of the GIS partial discharge vibration signal; A6 is a piezoelectric vibration sensor built in at the bottom position of the contact of the GIS to measure the vibration intensity of the GIS partial discharge vibration signal.

[0016] Further, among the vibration signals [Q1, Q2, Q3, Q4, Q5, Q6] of the link component, Q1 is a piezoelectric vibration sensor built into the lower position of the GIS terminal, which measures the vibration intensity of the vibration signal of the GIS connection component; Q2 is a piezoelectric vibration sensor built into the middle position of the GIS terminal, which measures the vibration intensity of the vibration signal of the GIS connection component; Q3 is a piezoelectric vibration sensor built into the upper position of the GIS terminal, which measures the vibration intensity of the vibration signal of the GIS connection component; Q4 is a piezoelectric vibration sensor built into the upper part of the GIS contact, which measures the vibration intensity of the vibration signal of the GIS connection component; Q5 is a piezoelectric vibration sensor built into the lower part of the GIS contact, which measures the vibration intensity of the vibration signal of the GIS connection component; Q6 is a piezoelectric vibration sensor built into both sides of the GIS contact, which measures the vibration intensity of the vibration signal of the GIS connection component.

[0017] Further, in step 2, noise reduction is performed on the detected multiple vibration sources, and the specific steps are as follows:

[0018] After the processing in step 1, the GIS multi-vibration source signal matrix (A) is obtained, and its dimension is (m×n), where (m) is the number of sensors and (n) is the number of sampling points; first, noise removal and filtering preprocessing are performed on this matrix to obtain the preprocessed signal matrix (A pre );

[0019] Step 2.1, weighting:

[0020] A weight matrix (W) is introduced, and its dimension is (m×1), representing the weight of each sensor. We multiply the signal matrix (A pre ) by the weight matrix (W) to obtain the weighted signal matrix (A weighted ):

[0021] [A weighted = A pre ·W];

[0022] Step 2.2, phase adjustment:

[0023] A phase adjustment matrix (Φ) is introduced, and its dimension is also (m×1), representing the phase adjustment of each sensor. Then, we multiply the weighted signal matrix (A weighted ) by the phase adjustment matrix (Φ) to obtain the finally formed superdirective beam signal matrix (A beamformed ):

[0024]

[0025] Step 2.3, mathematical derivation:

[0026] Assume (a i) represents the signal vector collected by the (i)-th sensor, with a length of (n×1), i.e., (a i = [a i1 , a i2 ,..., a in ) T ), for the weighted signal matrix (A weighted ), its j-th column represents the signal vector of the j-th sampling point after weighting, which is expressed as:

[0027] [A weighted = [a 1 ·w 1 , a 2 ·w 2 , …, a m ·w m )

[0028] where, (w i ) represents the weight of the (i)-th sensor;

[0029] For the noise-reduced signal matrix (A beamformed ) after phase adjustment, its j-th column represents the formed beam signal vector, which is expressed as:

[0030]

[0031] where, (φ i ) represents the phase adjustment of the (i)-th sensor.

[0032] Furthermore, in step 3, the electromagnetic vibration signal, partial discharge vibration signal, and connecting component vibration signal obtained by noise reduction in step 2 are subjected to signal identification. By performing wavelet transform, the signal is decomposed into different frequency components; then, the Mel-frequency cepstral coefficients of each frequency band are calculated to describe the spectral characteristics of the signal. The MFCC coefficients of each vibration signal are formed into a feature vector variance K and input into the multi-vibration source signal after noise reduction. When the feature vector variance K of the electromagnetic vibration signal ≤ 0.5, it is identified as the housing vibration signal, and when K > 0.5, it is the contact vibration signal; when the feature vector variance of the partial discharge vibration signal 1 ≤ K, it is the gas vibration signal, and K < 1 is the basin insulator vibration signal; for the connecting component vibration signal, when the feature vector variance K ≤ 0.01, it is the flange vibration signal, and K > 0.01 is the terminal vibration signal.

[0033] Furthermore, in step 3, the DWT-MFCC identification algorithm is used for identification, and the specific steps are as follows:

[0034] Step 3.1, Signal matrix:

[0035] The vibration signal matrix of GIS is (A):

[0036] [A = [a 1 , a 2 ,..., α n

[0037] where (a i ) is the sample value of the vibration signal;

[0038] Step 3.2, Discrete Wavelet Transform:

[0039] Perform discrete wavelet transform on the signal matrix (A), select an appropriate wavelet basis to extract the approximation coefficients and detail coefficients of the signal. The formula of DWT is as follows:

[0040] [(cA, cD) = pywt.dwt(A, 'db1′)]

[0041] where (cA) is the approximation coefficient and (cD) is the detail coefficient;

[0042] Step 3.3, Mel Frequency Cepstral Coefficients:

[0043] By processing the coefficients generated by DWT, Mel frequency cepstral coefficients can be obtained, including:

[0044] Spectrum calculation: Calculate the Fourier transform to obtain the spectrum of the signal;

[0045] Filter bank: Pass the spectrum through the Mel filter bank to obtain the corresponding energy;

[0046] Discrete Cosine Transform: Apply DCT to the output of the filter bank to obtain MFCC:

[0047] [MFCC = [cA, cD]]

[0048] Step 3.4, Calculate the Feature Variance:

[0049] The feature variance can measure the degree of dispersion of the signal features. After feature extraction, given the MFCC matrix as (MFCC):

[0050]

[0051] where, is the mean of MFCC and K is the variance.

[0052] The technical solution adopted by the present invention has the following beneficial effects:

[0053] ​In the present invention, first, the GIS electromagnetic vibration signals, including the contact vibration signal and the housing vibration signal, the partial discharge vibration signals, including the gas vibration signal and the basin insulator vibration signal, and the connection component vibration signals, including the terminal vibration signal and the flange signal, are monitored. Then, the superdirective beamforming algorithm is used to reduce the noise of the electromagnetic vibration signals, the partial discharge vibration signals, and the connection component vibration signals, and the DWT-MFCC feature variance algorithm is used to perform online identification of the vibration signals, thereby realizing the online identification of the GIS vibration signals and real-time monitoring of the operating state of the GIS, providing a data basis for real-time analysis of the operating state of the GIS. Description of the Drawings

[0054] Figure 1 It is a multi-vibration source signal diagram of a method for online identification of vibration signals for GIS provided by the present invention;

[0055] Figure 2 It is a GIS vibration signal identification model diagram of a method for online identification of vibration signals for GIS provided by the present invention. Detailed Embodiment

[0056] To make the objectives, technical solutions and effects of the present invention clearer and more definite, the following further describes the present invention in detail with reference to the accompanying drawings and by way of examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0057] A method for online identification of vibration signals for GIS is as Figure 2 shown and includes:

[0058] Step 1: Measure the multi-vibration sources of the GIS, measure the electromagnetic vibration signals, partial discharge vibration signals, and connection component vibration signals of the GIS through piezoelectric sensors, where the electromagnetic vibration signals include the housing vibration signal and the contact vibration signal, the partial discharge vibration signals include the gas vibration signal and the basin insulator vibration signal, and the connection component vibration signals include the terminal vibration signal and the flange vibration signal, and establish a multi-vibration source model based on the electromagnetic vibration signals, partial discharge vibration signals, and connection component vibration signals;

[0059] Among them, when measuring the multi-vibration sources of the GIS, the obtained GIS electromagnetic vibration signal matrix is

[0060] [C1, C2, C3, C4, C5, C6, C7, C8, C9,], the obtained GIS partial discharge vibration signal matrix is [A1, A2, A3, A4, A5, A6], and the obtained GIS link component vibration signal matrix is

[0061] [Q1, Q2, Q3, Q4, Q5, Q6].

[0062] In the electromagnetic vibration signals [C1, C2, C3, C4, C5, C6, C7, C8, C9], C1 is a piezoelectric vibration sensor installed at the left half position of the GIS tank to measure the vibration intensity at the half position of the GIS electromagnetic vibration signal; C2 is a piezoelectric vibration sensor installed at the middle half position of the GIS tank to measure the vibration intensity at the half vibration position of the GIS electromagnetic vibration signal; C3 is a piezoelectric vibration sensor installed at the right half position of the GIS tank to measure the vibration intensity at the right half position of the GIS tank; C4 is a piezoelectric vibration sensor installed at the one-third position of the GIS electromagnetic vibration signal to measure the vibration intensity at the one-third position of the GIS electromagnetic vibration signal; C5 is a piezoelectric vibration sensor installed at the middle one-third position of the GIS tank to measure the vibration intensity at the one-third position of the GIS electromagnetic vibration signal; C6 is a piezoelectric vibration sensor installed at the one-third position of the GIS electromagnetic vibration signal to measure the vibration intensity at the right one-third position of the GIS tank; C7 is a piezoelectric vibration sensor installed at the left top position of the GIS tank to measure the vibration intensity at the top of the GIS electromagnetic vibration signal; C8 is a piezoelectric vibration sensor installed at the middle top position of the GIS tank to measure the vibration intensity at the top of the GIS electromagnetic vibration signal; C9 is a piezoelectric vibration sensor installed at the right top position of the GIS tank to measure the vibration intensity of the GIS electromagnetic vibration signal.

[0063] In the partial discharge vibration signals [A1, A2, A3, A4, A5, A6], A1 is a piezoelectric vibration sensor installed at the one-third position of the upper half surface of the GIS contact to measure the vibration intensity at the one-third position of the GIS partial discharge vibration signal; A2 is a piezoelectric vibration sensor installed at the two-thirds position of the upper half surface of the GIS contact to measure the vibration intensity at the two-thirds position of the GIS partial discharge vibration signal; A3 is a piezoelectric vibration sensor installed at the top position of the upper half surface of the GIS contact to measure the vibration intensity of the GIS partial discharge vibration signal; A4 is a piezoelectric vibration sensor installed at the one-third position of the lower half surface of the GIS contact to measure the vibration intensity at the one-third position of the GIS partial discharge vibration signal; A5 is a piezoelectric vibration sensor installed at the two-thirds position of the lower half surface of the GIS contact to measure the vibration intensity at the two-thirds position of the GIS partial discharge vibration signal; A6 is a piezoelectric vibration sensor installed at the bottom position of the GIS contact to measure the vibration intensity of the GIS partial discharge vibration signal.

[0064] Among the vibration signals of the link component [Q1, Q2, Q3, Q4, Q5, Q6], Q1 is a piezoelectric vibration sensor built in the lower position of the GIS terminal, which measures the vibration intensity of the vibration signal of the GIS connection component; Q2 is a piezoelectric vibration sensor built in the middle position of the GIS terminal, which measures the vibration intensity of the vibration signal of the GIS connection component; Q3 is a piezoelectric vibration sensor built in the upper position of the GIS terminal, which measures the vibration intensity of the vibration signal of the GIS connection component; Q4 is a piezoelectric vibration sensor built in the upper part of the GIS contact, which measures the vibration intensity of the vibration signal of the GIS connection component; Q5 is a piezoelectric vibration sensor built in the lower part of the GIS contact, which measures the vibration intensity of the vibration signal of the GIS connection component; Q6 is a piezoelectric vibration sensor built on both sides of the GIS contact, which measures the vibration intensity of the vibration signal of the GIS connection component.

[0065] Step 2: Based on the multi-vibration source model established in Step 1, perform noise reduction using the superdirective beamforming algorithm. By combining the signals received by multiple sensors and performing weighted and phase adjustments on them, a superdirective beam is formed to enhance the signal in the target direction and suppress the signals in other directions, obtaining the multi-vibration source signal after noise reduction.

[0066] The specific steps for noise reduction of the detected multi-vibration sources are as follows:

[0067] After the processing in Step 1, the GIS multi-vibration source signal matrix (A) is obtained, with its dimension being (m×n), where (m) is the number of sensors and (n) is the number of sampling points. First, perform noise removal and filtering preprocessing on this matrix to obtain the preprocessed signal matrix (A pre );

[0068] Step 2.1: Weighting:

[0069] Introduce a weight matrix (W) with its dimension being (m×1), representing the weight of each sensor. Multiply the signal matrix (A pre ) by the weight matrix (W) to obtain the weighted signal matrix (A weighted ):

[0070] [A weighted = A pre ·W];

[0071] Step 2.2: Phase adjustment:

[0072] Introduce a phase adjustment matrix (Φ) with its dimension also being (m×1), representing the phase adjustment of each sensor. Then, multiply the weighted signal matrix (A weighted ) by the phase adjustment matrix (Φ) to obtain the finally formed superdirective beam signal matrix (A beamformed ):

[0073]

[0074] Step 2.3, Mathematical derivation:

[0075] Assume that (a i ) represents the signal vector collected by the (i)-th sensor, and its length is (n×1), that is, (a i = [a i1 , a i2 ,..., a in ) T ) For the weighted signal matrix (A weighted ), its j-th column represents the signal vector of the j-th sampling point after weighting, which is expressed as:

[0076] [A weighted = [a 1 ·w 1 , α 2 ·w 2 ,..., α m ·w m )

[0077] where, (w i ) represents the weight of the (i)-th sensor;

[0078] For the noise-reduced signal matrix (A beamformed ) after phase adjustment, its j-th column represents the formed beam signal vector, which is expressed as:

[0079]

[0080] where, (φ i ) represents the phase adjustment of the (i)-th sensor.

[0081] Step 3, Based on the multi-vibration source signal after noise reduction obtained in Step 2, perform identification through the DWT-MFCC identification algorithm. Identify the electromagnetic vibration signal, partial discharge vibration signal, and connecting component vibration signal obtained by noise reduction in Step 2. Through wavelet transform, decompose the signal into different frequency components; then, calculate the Mel frequency cepstral coefficients of each frequency band to describe the spectral characteristics of the signal. Form the feature vector variance K of the MFCC coefficients of each vibration signal and input it into the multi-vibration source signal after noise reduction. When the feature vector variance K of the electromagnetic vibration signal ≤ 0.5, it is identified as the housing vibration signal. When K > 0.5, it is the contact vibration signal; when the feature vector variance of the partial discharge vibration signal 1 ≤ K, it is the gas vibration signal, and K < 1 is the basin insulator vibration signal; for the connecting component vibration signal, when the feature vector variance K ≤ 0.01, it is the flange vibration signal, and K > 0.01 is the terminal vibration signal.

[0082] Identification is performed using the DWT-MFCC identification algorithm. The specific steps are as follows:

[0083] Step 3.1, Signal matrix:

[0084] The vibration signal matrix of GIS is (A):

[0085] [A = [α 1 , a 2 ,..., a n

[0086] where (a i ) is the sample value of the vibration signal;

[0087] Step 3.2, Discrete wavelet transform:

[0088] Perform discrete wavelet transform on the signal matrix (A), select an appropriate wavelet basis to extract the approximation coefficients and detail coefficients of the signal. The formula for DWT is as follows:

[0089] [(cA, cD) = pywt.dwt(A, "db1'))]

[0090] where (cA) is the approximation coefficient and (cD) is the detail coefficient;

[0091] Step 3.3, Mel frequency cepstral coefficients:

[0092] By processing the coefficients generated by DWT, Mel frequency cepstral coefficients can be obtained, including:

[0093] Spectrum calculation: Calculate the Fourier transform to obtain the spectrum of the signal;

[0094] Filter bank: Pass the spectrum through the Mel filter bank to obtain the corresponding energy;

[0095] Discrete cosine transform: Apply DCT to the output of the filter bank to obtain MFCC:

[0096] [MFCC = [cA, cD]]

[0097] Step 3.4, Calculate the feature variance:

[0098] The feature variance can measure the degree of dispersion of the signal features. After feature extraction, the given MFCC matrix is (MFCC):

[0099]

[0100] where, is the mean of MFCC and K is the variance.

[0101] ​Specifically, take the GIS electromagnetic vibration signal under normal operation of 252 kVA and 72.5 kV GIS as an example to establish a method for online identification of multiple vibration sources:

[0102] Step 1

[0103] As Figure 1 shown, measure the GIS electromagnetic vibration signal through a piezoelectric sensor to construct the electromagnetic vibration signal [C1, C2, C3, C4, C5, C6, C7, C8, C9]. C1 is a piezoelectric vibration sensor installed at the left half position of the GIS tank to measure the vibration intensity at the half position of the GIS electromagnetic vibration signal; C2 is a piezoelectric vibration sensor installed at the middle half position of the GIS tank to measure the vibration intensity at the half vibration of the GIS electromagnetic vibration signal; C3 is a piezoelectric vibration sensor installed at the right half position of the GIS tank to measure the vibration intensity at the right half position of the GIS tank; C4 is a piezoelectric vibration sensor installed at the one-third position of the GIS electromagnetic vibration signal to measure the vibration intensity at the one-third vibration of the GIS electromagnetic vibration signal; C5 is a piezoelectric vibration sensor installed at the middle one-third position of the GIS tank to measure the vibration intensity at the one-third position of the GIS electromagnetic vibration signal; C6 is a piezoelectric vibration sensor installed at the one-third position of the GIS electromagnetic vibration signal to measure the vibration intensity at the right one-third of the GIS tank; C7 is a piezoelectric vibration sensor installed at the left top position of the GIS tank to measure the vibration intensity at the top of the GIS electromagnetic vibration signal; C8 is a piezoelectric vibration sensor installed at the middle top position of the GIS tank to measure the vibration intensity at the top of the GIS electromagnetic vibration signal; C9 is a piezoelectric vibration sensor installed at the right top position of the GIS tank to measure the vibration intensity of the GIS electromagnetic vibration signal, and the electromagnetic vibration signal matrix A = [-0.95105652, -0.80901699, -0.58778525, -0.30901699, 0, 0.30901699, 0.58778525, 0.80901699] is obtained;

[0104] Measure the GIS electromagnetic vibration signal through a piezoelectric sensor to construct the electromagnetic vibration signal

[0105] [C1, C2, C3, C4, C5, C6, C7, C8, C9,], where C1 is a piezoelectric vibration sensor installed at the left half position of the GIS tank to measure the vibration intensity at the midpoint of the GIS electromagnetic vibration signal; C2 is a piezoelectric vibration sensor installed at the middle half position of the GIS tank to measure the vibration intensity at the midpoint of the GIS electromagnetic vibration signal; C3 is a piezoelectric vibration sensor installed at the right half position of the GIS tank to measure the vibration intensity at the right half of the GIS tank; C4 is a piezoelectric vibration sensor installed at the one-third position of the GIS electromagnetic vibration signal to measure the vibration intensity at the one-third of the GIS electromagnetic vibration signal; C5 is a piezoelectric vibration sensor installed at the middle one-third position of the GIS tank to measure the vibration intensity at the one-third of the GIS electromagnetic vibration signal; C6 is a piezoelectric vibration sensor installed at the one-third position of the GIS electromagnetic vibration signal to measure the vibration intensity at the right one-third of the GIS tank; C7 is a piezoelectric vibration sensor installed at the left top position of the GIS tank to measure the vibration intensity at the top of the GIS electromagnetic vibration signal; C8 is a piezoelectric vibration sensor installed at the middle top position of the GIS tank to measure the vibration intensity at the top of the GIS electromagnetic vibration signal; C9 is a piezoelectric vibration sensor installed at the right top position of the GIS tank to measure the vibration intensity of the GIS electromagnetic vibration signal, and the electromagnetic vibration signal matrix B = [1.09185956, 1.08979323, 1.05173937,

[0106] 0.978533763, 0.836516304, 0.660931038, 0.459420093, 0.242359754, 0.00329446102].

[0107] Step 2

[0108] The electromagnetic vibration signal matrix A of the GIS multi-vibration source signal is obtained as [-0.95105652,

[0109] -0.80901699, -0.58778525, -0.30901699, 0, 0.30901699, 0.58778525, 0.80901699]. First, we preprocess this matrix, including removing noise, filtering, etc., to obtain the preprocessed signal matrix (A pre ).

[0110] Introduce a weight matrix (W) with a dimension of (m×1) to represent the weight of each sensor. We multiply the signal matrix (A pre ) by the weight matrix (W) to obtain the weighted signal matrix (A weighted ):

[0111] [Aweighted = A pre ·W]

[0112] Introduce a phase adjustment matrix (Φ), which also has the dimension of (m×1) and represents the phase adjustment of each sensor. Then, we multiply the weighted signal matrix (A weighted ) by the phase adjustment matrix (Φ) to obtain the finally formed superdirective beam signal matrix (A beamformed ):

[0113]

[0114] Assume that (a i ) represents the signal vector collected by the (i)-th sensor, and its length is (n×1), that is, (a i = [a i1 , a i2 ,..., a in T ). For the weighted signal matrix (A weighted ), its j-th column represents the signal vector of the j-th sampling point after weighting, which can be expressed as:

[0115] [A weighted = [α 1 ·w 1 , a 2 ·w 2 ,..., a m ·w m

[0116] where (w i ) represents the weight of the (i)-th sensor.

[0117] For the signal matrix (A beamformed ) after phase adjustment, its j-th column represents the formed beam noise reduction signal vector, which can be expressed as:

[0118] Abeamformed = [0.30901699 0.58778525 0.80901699 0.95105652 1 0.95105652 0.80901699 0.58778525 0.30901699];

[0119] ​​The GIS multi-vibration source signal matrix B = [1.09185956, 1.08979323, 1.05173937, 0.978533763, 0.836516304, 0.660931038, 0.459420093, 0.242359754, 0.00329446102] is obtained. First, we preprocess this matrix, including noise removal, filtering, etc., to obtain the preprocessed signal matrix (B pre ).

[0120] 2. Superdirective beamforming

[0121] 2.1 Weighting

[0122] A weight matrix (W) with dimension (m×1) is introduced, representing the weight of each sensor. We multiply the signal matrix (B pre ) by the weight matrix (W) to obtain the weighted signal matrix (B weighted ) as follows:

[0123] [B weighted = B pre ·W]

[0124] 2.2 Phase adjustment

[0125] A phase adjustment matrix (Φ) with dimension (m×1) is introduced, representing the phase adjustment of each sensor. Then, we multiply the weighted signal matrix (B weighted ) by the phase adjustment matrix (Φ) to obtain the finally formed superdirective beam signal matrix (B beamformed ) as follows:

[0126]

[0127] 2.3 Mathematical derivation

[0128] Assume that (b i ) represents the signal vector collected by the (i)-th sensor, with length (n×1), i.e., (b i = [b i1 , b i2 ,..., b in T ). For the weighted signal matrix (B weighted ), its j-th column represents the signal vector of the j-th sampling point after weighting and can be expressed as:

[0129] [B weighted = [b 1 ·w 1 , b 2 ·w 2 ,..., b​m ·w m

[0130] Among them, (w i ) represents the weight of the (i)-th sensor.

[0131] For the phase-adjusted signal matrix (B beamformed ), the (j)-th column thereof represents the formed beam signal vector, which can be expressed as:

[0132]

[0133] Among them, (φ i ) represents the phase adjustment of the (i)-th sensor.

[0134] 3. Target signal extraction

[0135] From the beamformed signal matrix (B beamformed ), the vibration signal in the target direction can be extracted through signal processing techniques such as peak detection and energy detection.

[0136] Bbeamformed = [0.154813621 0.306301880 0.451563796e 0.5878354910.712694171 0.823944633 0.919621344 0.998026728 1.05764637]

[0137] Step 3, give the mathematical definition of variance in the DWT-MFCC algorithm:

[0138] Variance represents the degree of dispersion of a set of data and is the average of the sum of the squares of the differences between each data and its mean. For a one-dimensional data set X, its variance Var(X) can be calculated using the following formula:

[0139]

[0140] Among them, N is the size of the data set, x _i is the i-th data point, is the mean of the data set.

[0141] Perform DWT-MFCC feature extraction on the contact vibration signal and the enclosure vibration signal of the GIS, and then calculate their variances to better distinguish these two vibration signals.

[0142] First, give the electromagnetic vibration signal matrix A and the electromagnetic vibration signal matrix B after noise reduction of the GIS:

[0143] ​A = [0.30901699 0.58778525 0.80901699 0.95105652 1 0.95105652 0.80901699 0.58778525 0.30901699]

[0144] B = [0.154813621 0.306301880 0.451563796e 0.587835491 0.712694171 0.823944633 0.919621344 0.998026728 1.05764637]

[0145] Perform wavelet transform on these two signal matrices, then calculate MFCC features and extract variances.

[0146] The specific steps are as follows:

[0147] Perform wavelet transform on the electromagnetic vibration signal A:

[0148] First, we perform wavelet transform on the signal matrix A using the wavelet basis function.

[0149] The approximation coefficients and detail coefficients of the wavelet transform can be calculated by the following formula:

[0150] [cA, cD = pywt.dwt(A, "db1')]

[0151] where cA are the approximation coefficients and cD are the detail coefficients.

[0152] The calculated approximation coefficients and detail coefficients are:

[0153] cA = [0.94889467, 1.19320989, 1.25641864, 0.51579731, -0.68765619, -1.3748388]

[0154] cD = [-0.35491984, -0.14719551, 0.03933924, -0.07907399, -0.02745301, 0.09206]

[0155] Calculate MFCC features:

[0156] MFCC features are usually calculated through steps such as the Mel filter bank, logarithmic compression, and discrete cosine transform. Here, for simplicity of calculation, assume that we directly use the coefficients after DWT as MFCC features.

[0157] MFCC w=[0.94889467, 1.19320989, 1.25641864, 0.51579731, -0.6876561

[0158] 9, -1.374838, -0.35491984, -0.147195, 0.03933924, -0.07907399, -0.027451, 0.092060]

[0159] Calculate the variance of MFCC features:

[0160] The calculation formula for variance is:

[0161]

[0162] where N is the dimension of MFCC features, MFCCwi is the i-th dimensional feature, is the mean of MFCC features.

[0163] The variance of the MFCC features of the obtained contact vibration signal is:

[0164] [Var(MFCC w ) = 0.6591689523]

[0165] Repeat the above steps to process the electromagnetic vibration signal B:

[0166] Perform wavelet transform on the electromagnetic vibration signal B:

[0167] [cA, cD = pywt.dwt(X i , 'db1)]

[0168] where cA is the approximation coefficient and cD is the detail coefficient.

[0169] The obtained approximation coefficient and detail coefficient are:

[0170] cA = [0.75467416, 1.20446166, 0.76006532, -0.06007371, -0.7916624, -1.05764637]

[0171] cD = [-0.3879299, -0.12685089, 0.15305291, 0.05743962, -0.08838762, 0.03882792]

[0172] Calculate MFCC features:

[0173] MFCCi = [0.75467, 1.20446, 0.76006 - 0.06007, -0.7916, -1.05764 0.3879, 0.12685, 0.15305, 0.05743 - 0.08838, 0.03882]

[0174] Calculate the variance of the MFCC features:

[0175]

[0176] where N is the dimension of the MFCC features, MFCCii is the i-th dimensional feature, and \(\bar{MFCCi}\) is the mean of the MFCC features.

[0177]

[0178] The variance of the MFCC features of the calculated enclosure vibration signal is:

[0179] [Var(MFCC i ) = 0.4430910698]

[0180] Calculated the variances of the MFCC features of the electromagnetic vibration signals A and B:

[0181] The variance of the MFCC features of the electromagnetic vibration signal A is K = (Var(MFCC w ) = 0.6591689523),

[0182] The variance of the MFCC features of the electromagnetic vibration signal B is K = (Var(MFCC i ) = 0.4430910698.

[0183] According to the feature variances, when K ≤ 0.5, the electromagnetic vibration signal is an enclosure vibration signal, and when K > 0.5, it is a contact vibration signal. Therefore, the electromagnetic vibration signal A is a contact vibration signal, and the electromagnetic vibration signal B is an enclosure vibration signal.

[0184] When the vibration source signals are partial discharge vibration signals and connection component vibration signals, the measured signals such as gas vibration signals, basin insulator vibration signals, flange vibration signals, and terminal vibration signals can also achieve real-time condition monitoring of the GIS tank body.

[0185] Other embodiments of the present invention will be readily apparent to those skilled in the art upon consideration of the specification and practice of the solutions disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include known common knowledge or conventional technical means in the technical field not disclosed in this disclosure. The specification and examples are only regarded as exemplary, and the true scope and spirit of the present invention are indicated by the claims.

Claims

1. A method for online identification of vibration signals for GIS, characterized in that: include: Step 1, measuring multiple vibration sources of the GIS, measuring the electromagnetic vibration signal, partial discharge vibration signal, and connection component vibration signal of the GIS through a piezoelectric sensor, wherein the electromagnetic vibration signal includes a housing vibration signal and a contact vibration signal, the partial discharge vibration signal includes a gas vibration signal and a basin insulator vibration signal, and the connection component vibration signal includes a terminal vibration signal and a flange vibration signal, and establishing a multi-vibration source model based on the electromagnetic vibration signal, the partial discharge vibration signal, and the connection component vibration signal; Step 2: Based on the multi-vibration source model established in step 1, a super-directional beamforming algorithm is used to reduce noise. By combining the signals received by multiple sensors and weighting and phase-adjusting them, a super-directional beam is formed to enhance the signal in the target direction and suppress the signals in other directions, thereby obtaining the multi-vibration source signal after noise reduction. Step 3: Based on the noise-reduced multi-vibration source signals obtained in step 2, identification is performed using the DWT-MFCC identification algorithm.

2. The method for online identification of vibration signals for GIS according to claim 1, characterized in that: In step 1, the multiple vibration sources of GIS are measured, and the electromagnetic vibration signal matrix of GIS is obtained as [C1, C2, C3, C4, C5, C6, C7, C8, C9], the partial discharge vibration signal matrix of GIS is obtained as [A1, A2, A3, A4, A5, A6], and the vibration signal matrix of the link components of GIS is obtained as [Q1, Q2, Q3, Q4, Q5, Q6]. A multi-vibration source model is established through the above vibration signal matrix.

3. The method for online identification of vibration signals for GIS according to claim 2, characterized in that: In the electromagnetic vibration signal [C1, C2, C3, C4, C5, C6, C7, C8, C9], C1 is a piezoelectric vibration sensor built-in at the left half of the GIS tank body, which tests the vibration intensity at half of the GIS electromagnetic vibration signal; C2 is a piezoelectric vibration sensor built-in at the middle half of the GIS tank body, which tests the vibration intensity at half of the GIS electromagnetic vibration signal; C3 is a piezoelectric vibration sensor built-in at the right half of the GIS tank body, which tests the vibration intensity at half of the right half of the GIS tank body; C4 is a piezoelectric vibration sensor built-in at one-third of the GIS electromagnetic vibration signal, which tests the vibration intensity at one-third of the GIS electromagnetic vibration signal. ; C5 is a piezoelectric vibration sensor built into the middle one-third of the GIS tank body to test the vibration intensity of the one-third of the GIS electromagnetic vibration signal; C6 is a piezoelectric vibration sensor built into the one-third of the GIS electromagnetic vibration signal to test the vibration intensity of the right one-third of the GIS tank body; C7 is a piezoelectric vibration sensor built into the top position of the left side of the GIS tank body to test the vibration intensity of the top of the GIS electromagnetic vibration signal; C8 is a piezoelectric vibration sensor built into the middle top position of the GIS tank body to test the vibration intensity of the top of the GIS electromagnetic vibration signal; C9 is a piezoelectric vibration sensor built into the top position of the right side of the GIS tank body to test the vibration intensity of the GIS electromagnetic vibration signal.

4. The method for online identification of vibration signals for GIS according to claim 2, characterized in that: In the local discharge vibration signal [A1, A2, A3, A4, A5, A6], A1 is a piezoelectric vibration sensor built-in at one-third of the upper half of the contact of the GIS, which tests the vibration intensity of the one-third position of the local discharge vibration signal of the GIS; A2 is a piezoelectric vibration sensor built-in at two-thirds of the upper half of the contact of the GIS, which tests the vibration intensity of the two-thirds position of the local discharge vibration signal of the GIS; A3 is a piezoelectric vibration sensor built-in at the top of the upper half of the contact of the GIS, which tests the vibration intensity of the local discharge vibration signal of the GIS; A4 is a piezoelectric vibration sensor built-in at one-third of the lower half of the contact of the GIS, which tests the vibration intensity of the one-third position of the local discharge vibration signal of the GIS; A5 is a piezoelectric vibration sensor built-in at two-thirds of the lower half of the contact of the GIS, which tests the vibration intensity of the two-thirds position of the local discharge vibration signal of the GIS; A6 is a piezoelectric vibration sensor built-in at the bottom of the GIS contact, which tests the vibration intensity of the local discharge vibration signal of the GIS.

5. The method for online identification of vibration signals for GIS according to claim 2, characterized in that: Among the vibration signals of the link components [Q1, Q2, Q3, Q4, Q5, Q6], Q1 is a piezoelectric vibration sensor built into the lower part of the GIS terminal block to test the vibration intensity of the vibration signal of the GIS connection component; Q2 is a piezoelectric vibration sensor built into the middle part of the GIS terminal block to test the vibration intensity of the vibration signal of the GIS connection component; Q3 is a piezoelectric vibration sensor built into the upper part of the GIS terminal block to test the vibration intensity of the vibration signal of the GIS connection component; Q4 is a piezoelectric vibration sensor built into the upper part of the GIS contact to test the vibration intensity of the vibration signal of the GIS connection component; Q5 is a piezoelectric vibration sensor built into the lower part of the GIS contact to test the vibration intensity of the vibration signal of the GIS connection component; Q6 is a piezoelectric vibration sensor built into the two sides of the GIS contact to test the vibration intensity of the vibration signal of the GIS connection component.

6. The method for online identification of vibration signals for GIS according to claim 1, characterized in that: In step 2, noise reduction is performed on the detected multiple vibration sources. The specific steps are as follows: After processing in step 1, the GIS multi-vibration source signal matrix (A) is obtained, and its dimension is (m×n), where (m) is the number of sensors and (n) is the number of sampling points. First, this matrix is ​​preprocessed by removing noise and filtering to obtain the preprocessed signal matrix (A pre ); Step 2.1, weighting: A weight matrix (W) with a dimension of (m×1) is introduced to represent the weight of each sensor, and the signal matrix (A pre ) is multiplied by the weight matrix (W) to obtain the weighted signal matrix (A weighted ): [A weighted =A pre ·W]; Step 2.2, phase adjustment: A phase adjustment matrix (Φ) is introduced, whose dimension is also (m×1), which represents the phase adjustment of each sensor. Then, the weighted signal matrix (A weighted ) is multiplied by the phase adjustment matrix (Φ) to obtain the final super-directional beam signal matrix (A beamformed ): Step 2.3, mathematical derivation: Assume (a i ) represents the signal vector collected by the (i)th sensor, whose length is (n×1), that is, (a i =[a i1 , a i2 , ..., a in ] T ), for the weighted signal matrix (A weighted ), the (j)th column represents the signal vector of the (j)th sampling point after weighting, expressed as: <h2 style=";text-align:left;direction:ltr">[A<h2 style=";text-align:left;direction:ltr"> weighted <h2 style=";text-align:left;direction:ltr"> (a1 w1, a2 w2,..., w)<h2 style=";text-align:left;direction:ltr"> m <h2 style=";text-align:left;direction:ltr"> ·w<h2 style=";text-align:left;direction:ltr"> m <h2 style=";text-align:left;direction:ltr"> ]] Among them, (w i ) represents the weight of the (i)th sensor; For the noise reduction signal matrix after phase adjustment (A beamformed ), the (j)th column represents the formed beam signal vector, expressed as: Among them, (φ i ) represents the phase adjustment of the (i)th sensor.

7. The method for online identification of vibration signals for GIS according to claim 1, characterized in that: In the step 3, the electromagnetic vibration signal, partial discharge vibration signal and connecting component vibration signal obtained by noise reduction in step 2 are subjected to signal identification, and the signal is decomposed into different frequency components by performing wavelet transform; then, the Mel frequency cepstrum coefficient of each frequency band is calculated to describe the spectral characteristics of the signal, and the MFCC coefficients of each vibration signal are combined into a characteristic vector variance K, and the multi-vibration source signal after noise reduction is input, and when the characteristic vector variance K≤0.5 of the electromagnetic vibration signal is identified as a shell vibration signal, and when K>0.5 is a contact vibration signal; when the characteristic vector variance of the local discharge vibration signal is 1≤K, it is a gas vibration signal, and K<1 is a basin insulator vibration signal; when the characteristic vector variance K≤0.01 of the connecting component vibration signal is a flange vibration signal, and K>0.01 is a terminal vibration signal.

8. The method for online identification of vibration signals for GIS according to claim 7, characterized in that: In step 3, the DWT-MFCC recognition algorithm is used for recognition, and the specific steps are as follows: Step 3.1, signal matrix: The vibration signal matrix of GIS is (A): <h2 style=";text-align:left;direction:ltr">[A = [a1,a2,...,a<h2 style=";text-align:left;direction:ltr"> n <h2 style=";text-align:left;direction:ltr"> ]] Among them (a i ) is the sample value of the vibration signal; Step 3.2, discrete wavelet transform: Perform discrete wavelet transform on the signal matrix (A) and select a suitable wavelet basis to extract the approximate coefficients and detail coefficients of the signal. The formula of DWT is as follows: [(cA,cD)=pywt.dw(A,′db1′)] Where (cA) is the approximation coefficient and (cD) is the detail coefficient; Step 3.3, Mel frequency cepstral coefficients: By processing the coefficients generated by DWT, the Mel frequency cepstrum coefficients can be obtained, including: Spectrum calculation: Calculate Fourier transform to obtain the spectrum of the signal; Filter bank: Pass the spectrum through the Mel filter bank to obtain the corresponding energy; Discrete Cosine Transform: Apply DCT to the output of the filter bank to get the MFCCs: [MFCC = [cA, cD]] Step 3.4, calculate the feature variance: The feature variance can measure the discreteness of the signal features. After feature extraction, the MFCC matrix is ​​given as (MFCC): in, is the mean of MFCC and K is the variance.

Images