CN110244202A - Based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method - Google Patents
Based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method Download PDFInfo
- Publication number
- CN110244202A CN110244202A CN201910543829.9A CN201910543829A CN110244202A CN 110244202 A CN110244202 A CN 110244202A CN 201910543829 A CN201910543829 A CN 201910543829A CN 110244202 A CN110244202 A CN 110244202A
- Authority
- CN
- China
- Prior art keywords
- wavelet
- coefficient matrix
- conversion coefficient
- synchronous compression
- signal
- 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
Classifications
-
- 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
- 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
- G01R31/1263—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 of solid or fluid materials, e.g. insulation films, bulk material; of semiconductors or LV electronic components or parts; of cable, line or wire insulation
-
- 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
- G01R31/14—Circuits therefor, e.g. for generating test voltages, sensing circuits
-
- 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
- G01R31/16—Construction of testing vessels; Electrodes therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
Abstract
Based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method, including wavelet transformation is carried out to simulation local signal and obtains matrix of wavelet coefficients；Higher order statistic analysis is carried out to wavelet conversion coefficient matrix, the preliminary inhibition of noise is carried out to wavelet conversion coefficient matrix using kurtosis threshold value criterion, obtains corrected wavelet conversion coefficient matrix；Synchronous compression wavelet conversion coefficient matrix is obtained using corrected wavelet conversion coefficient matrix, the noise threshold level of synchronous compression wavelet conversion coefficient matrix is obtained using Generalized Cross Validation algorithm；The inhibition for carrying out residual noise to synchronous compression wavelet conversion coefficient matrix using shearing thresholding algorithm using noise threshold level, obtains the lesser synchronous compression wavelet conversion coefficient matrix of noise jamming；Compressed wavelet inverse transformation is synchronized to the lesser synchronous compression wavelet conversion coefficient matrix of noise jamming, obtains the one-dimensional time-domain signal of more pure shelf depreciation.The method of the present invention has denoising precision high, and operation time short advantage is suitable for occasions such as partial discharge of transformer denoisings.
Description
Technical field
It is specifically a kind of to be based on synchronous compression wavelet transformed domain transformation the present invention relates to partial discharge of transformer detection field
Device Denoising of Partial Discharge.
Background technique
Shelf depreciation (Partial Discharge, PD) is a kind of high localized microcosmic electric discharge phenomena.Due to insulation
Layer runs down, and PD will lead to insulating layer and be destroyed to make transformer that catastrophic failure occur completely.When shelf depreciation is one-dimensional
Important carrier of the domain signal as research partial discharge of transformer phenomenon, is mainly done by narrow-band noise and white noise signal
It disturbs.And this interference will affect being normally carried out for the follow-up works such as Partial Discharge Detection, pattern-recognition, therefore interference signal is carried out
Inhibition is just particularly important.
The resolution ratio of traditional time frequency analysis algorithm is low, readable poor, and denoising precision does not reach requirement.Traditional wavelet threshold
Denoising Algorithm is fixed on to be denoised using soft or hard -threshold.But wavelet shrinkage coefficient Shi Bulian will lead to using hard -threshold scheme
It is continuous, and reduced together using soft-threshold denoising scheme it will cause uncorrelated wavelet coefficient, so as to cause the signal letter after denoising
It makes an uproar relatively low.
Summary of the invention
In order to solve the above technical problems, the present invention provides one kind based on synchronous compression wavelet transformed domain partial discharge of transformer
Denoising method, this method have denoising precision high, and operation time short advantage is suitable for fields such as partial discharge of transformer denoisings
It closes.
The technical scheme adopted by the invention is as follows:
Based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method, comprising the following steps:
Step 1: wavelet transformation being carried out to the one-dimensional time-domain signal of the partial discharge of transformer of acquisition, obtains one-dimensional time-domain signal
Wavelet conversion coefficient matrix；
Step 2: higher order statistic analysis being carried out to wavelet conversion coefficient matrix, using kurtosis threshold value criterion, small echo is become
The preliminary inhibition that coefficient matrix carries out noise is changed, corrected wavelet conversion coefficient matrix is obtained；
Step 3: utilizing corrected wavelet conversion coefficient matrix, obtain synchronous compression wavelet conversion coefficient matrix, use
Generalized Cross Validation algorithm obtains the noise threshold level of synchronous compression wavelet transformation SS-CWT coefficient matrix；
Step 4: noise threshold level is utilized, using shearing thresholding algorithm, to synchronous compression wavelet transformation SS-CWT coefficient
Matrix carries out the inhibition of residual noise, obtains the lesser synchronous compression wavelet conversion coefficient matrix of noise jamming；
Step 5: to the lesser synchronous compression wavelet conversion coefficient matrix of noise jamming, synchronizing compressed wavelet inversion
It changes, obtains the one-dimensional time-domain signal of more pure shelf depreciation.
Of the invention a kind of based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method, beneficial effect is:
1: time frequency analysis resolution ratio is higher: traditional Time-Frequency Analysis Method, such as: Short Time Fourier Transform, S-transformation, due to calculating
The limitation of method itself, scale do not have that wavelet transformation is flexible, and the time frequency analysis figure of wavelet transformation is readable poor.Synchronous compression small echo
Transformation overcomes the defect of wavelet transformation, enables the clearer differentiation purified signal of the algorithm and noise.
2: two kinds of algorithm joints guarantee that denoising precision is higher: when conventional threshold values Denoising Algorithm determines noise-removed threshold value level, depositing
Inhibited in the excessive problem of calculation amount, and just for single noise.Higher order statistic analysis can tentatively make an uproar to narrowband
Sound and white noise are inhibited.Generalized Cross Validation algorithm can determine noise-removed threshold value level with less calculation amount, shear threshold
Value is higher than conventional threshold values denoising scheme signal-to-noise ratio.Two kinds of algorithm joint denoisings, can remove the two of interference local discharge signal
Kind noise, while the signal distortion after denoising is low.
3: the present invention is same in conjunction with traditional wavelet Algorithm constitution by high-resolution time frequency analysis algorithm --- synchronous compression
It walks compressed wavelet and converts algorithm.The algorithm improves time-frequency on the basis of inheriting wavelet transform dimension flexibly this advantage
The resolution ratio of analysis ensure that the precision of denoising.Noise-removed threshold value level is solved using Generalized Cross Validation algorithm simultaneously.
Using in conjunction with soft, hard -threshold advantage denoising scheme --- shearing threshold value improves the signal-to-noise ratio of signal after denoising.
Detailed description of the invention
Fig. 1 (a) is that shelf depreciation of the present invention simulates one-dimensional time-domain signal；
Fig. 1 (b) is the SS-CWT time frequency analysis figure that shelf depreciation of the present invention simulates one-dimensional time-domain signal；
Fig. 1 (c) is that shelf depreciation of the present invention simulation contaminates one-dimensional time-domain signal of making an uproar；
Fig. 1 (d) is the SS-CWT time frequency analysis figure that shelf depreciation of the present invention simulation contaminates one-dimensional time-domain signal of making an uproar；
Fig. 1 (e) is the Short Time Fourier Transform time frequency analysis figure that shelf depreciation of the present invention simulation contaminates one-dimensional time-domain signal of making an uproar；
Fig. 1 (f) is the Wavelet Transform Time Frequency Analysis figure that shelf depreciation of the present invention simulation contaminates one-dimensional time-domain signal of making an uproar.
Fig. 2 (a) is that shelf depreciation of the present invention simulation dye makes an uproar one-dimensional time-domain signal after higher order statistic analysis denoises
One-dimensional time-domain signal；
Fig. 2 (b) is the present invention after higher order statistic analysis denoises, the SS-CWT time frequency analysis figure of time-domain signal；
Fig. 2 (c) is that shelf depreciation of the present invention simulation dye makes an uproar one-dimensional time-domain signal by higher order statistic analysis and broad sense friendship
After pitching threshold denoising, one-dimensional time-domain signal and original clean signal contrast figure；
Fig. 2 (d) is the present invention after higher order statistic analysis and generalized crossover threshold denoising, the SS-CWT of time-domain signal
Time frequency analysis figure；
Fig. 2 (e) for higher order statistic analysis of the present invention and the removed noise signal of generalized crossover thresholding algorithm SS-CWT
Time frequency analysis figure.
Fig. 3 (a) is the present invention after higher order statistic analysis denoises, the intersection small echo spectrogram of time-domain signal；
Fig. 3 (b) is the present invention after higher order statistic analysis and generalized crossover threshold denoising, and the intersection of time-domain signal is small
Wave spectrogram.Fig. 4 (a) is the shelf depreciation laboratory simulations platform that the present invention constructs；
The shelf depreciation power-frequency voltage that Fig. 4 (b) obtains for the present invention from the shelf depreciation laboratory simulations platform constructed
Time-domain signal figure；
The shelf depreciation electric current time domain that Fig. 4 (c) obtains for the present invention from the shelf depreciation laboratory simulations platform constructed
Signal graph；
Fig. 4 (d) is one section of shelf depreciation time domain impulsive signals that the present invention is intercepted from laboratory shelf depreciation current signal
Figure.
Fig. 4 (e) is one section of shelf depreciation time domain impulsive signals that the present invention is intercepted from laboratory shelf depreciation current signal
SS-CWT time frequency analysis figure；
Fig. 4 (f) is that narrow-band noise and white noise is added to one section of shelf depreciation time domain impulsive signals of interception in the present invention
Time-domain signal figure；
Fig. 4 (g) is the SS-CWT time frequency analysis after the present invention makes an uproar to one section of shelf depreciation time domain impulsive signals dye of interception
Figure；
Fig. 4 (h) is after the present invention makes an uproar to one section of shelf depreciation time domain impulsive signals dye of interception, using high-order statistic point
Time-domain signal figure after analysis and the joint denoising of Generalized Cross Validation algorithm；
Fig. 4 (i) is the present invention to carry out higher order statistic analysis and generalized crossover and tests to dye laboratory local discharge signal of making an uproar
After demonstrate,proving algorithm joint denoising, the SS-CWT time frequency analysis figure of time-domain signal.
Specific embodiment
Based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method, comprising the following steps:
Step 1: wavelet transformation being carried out to the one-dimensional time-domain signal of the partial discharge of transformer of acquisition, obtains one-dimensional time-domain signal
Wavelet conversion coefficient matrix；
Step 2: higher order statistic analysis being carried out to wavelet conversion coefficient matrix, using kurtosis threshold value criterion, small echo is become
The preliminary inhibition that coefficient matrix carries out noise is changed, corrected wavelet conversion coefficient matrix is obtained；
Step 3: utilizing corrected wavelet conversion coefficient matrix, obtain synchronous compression wavelet conversion coefficient matrix, use
Generalized Cross Validation algorithm obtains the noise threshold level of synchronous compression wavelet transformation SS-CWT coefficient matrix；
Step 4: noise threshold level is utilized, using shearing thresholding algorithm, to synchronous compression wavelet transformation SS-CWT coefficient
Matrix carries out the inhibition of residual noise, obtains the lesser synchronous compression wavelet conversion coefficient matrix of noise jamming；
Step 5: to the lesser synchronous compression wavelet conversion coefficient matrix of noise jamming, synchronizing compressed wavelet inversion
It changes, obtains the one-dimensional time-domain signal of more pure shelf depreciation.
In step 1, the wavelet transformation is continuous wavelet transform (Continuous Wavelet Transform, CWT),
Its expression formula are as follows:
Wherein: * represents complex conjugate operation.<s, ψ>be time-domain signal s and wavelet mother function ψ operator, WsAs small echo becomes
The coefficient matrix changed, a are scale factors, and τ is time shift method.
Wavelet conversion coefficient matrix is the output result W of wavelet transformations(a, τ) relies on wavelet mother function to time domain
Signal is converted.
Step 2, higher order statistic analysis is the preliminary denoising to signal, comprising the following steps:
Step 2.1: calculate the kurtosis value of wavelet conversion coefficient matrix:
Wherein: σWsFor the standard deviation of wavelet coefficient, μWsFor the mean value of matrix W s wavelet coefficient, N is sampling number, kurts
For the kurtosis value of the wavelet conversion coefficient of signal s, WsFor the coefficient matrix of wavelet transformation；μ is mean operation, is asked operand
Average value.σ is standard difference operation, seeks standard deviation to operand.
Step 2.2: construction kurtosis threshold value criterion:
α is confidence level, and N is signal sampling points, kurtsFor the kurtosis value of the wavelet conversion coefficient of signal s.
The preliminary inhibition of shown noise, refers to: higher order statistic analysis algorithm is the statistics spy from matrix of wavelet coefficients
Sign sets about being calculated, and there is no carry out noise suppressed, the denoising result figure of Fig. 2 (a) and 2 (b) using its time and frequency domain characteristics
It can reflect, which is rough to dye noise cancellation signal progress noise suppressed, but there are still small part noises not to remove.
What each element in wavelet conversion coefficient matrix that wavelet transformation is obtained was generated with above-mentioned kurtosis threshold value criterion
Kurtosis threshold value kurtsIt compares, if the element in the matrix is less than kurtosis threshold value, the element is in wavelet conversion coefficient matrix
In value be 0, value that is on the contrary then retaining the element is constant.Each element in wavelet conversion coefficient matrix is carried out
Stating the matrix constituted after kurtosis threshold denoising is corrected wavelet conversion coefficient matrix.
Step 3, the processing of synchronous compression threshold denoising is the denoising again to dye noise cancellation signal, comprising the following steps:
Step 3.1: calculate synchronous compression wavelet conversion coefficient matrix:
Wherein: TyFor synchronous compression wavelet conversion coefficient matrix, fs is the sample frequency of signal, and N is the sampled point of signal
Number, ωlIt is its l ∈ [1, N] of first of discrete frequency, △ ω is the difference of discrete frequency.akFor k-th of wavelet coefficient WyScale because
Son.
WyFor corrected wavelet conversion coefficient matrix described in above-mentioned steps 2.2.
△ω-1For the inverse of two neighboring discrete frequency difference, i.e., inverse of a matrix operation is carried out to the difference of discrete frequency；M is to ask
With judge formula, the scale factor a and time shift method τ for only meeting the judgement formula can just calculate corresponding synchronous compression small echo and become
Change coefficient matrix Ty；τ is the time shift method of wavelet conversion coefficient matrix, is an important parameter of wavelet transformation；△akIt is adjacent
The difference of scale factor；ω is discrete frequency；△ ω is the difference of adjacent discrete frequency；ωlFor first of discrete frequency；ωl-1It is
L-1 discrete frequency；ak-1For -1 scale factor of kth.
w(ak, τ) and it is candidate instantaneous frequency, the frequency expression are as follows:
Wherein: Wy(a, τ) is Continuous Wavelet Transform Coefficients matrix, while being also corrected described in above-mentioned steps 2.2
Wavelet conversion coefficient matrix, a are scale factors, and △ a is the difference of scale factor, and τ is time shift method,For derivative operation, wyAs
Candidate instantaneous frequency.I is imaginary unit.
Step 3.2: it is horizontal to calculate noise-removed threshold value:
Wherein: GCV is Generalized Cross Validation algorithm coefficient,It is using the threshold coefficient of synchronous compression domain threshold value λ, N0It is
The coefficient number of threshold level λ is not up in synchronous compression wavelet conversion coefficient matrix, N sampling number, λ is going of finally acquiring
It makes an uproar threshold level.
Synchronous compression wavelet transformation SS-CWT coefficient matrix: required by the synchronous compression wavelet transformation that as formula (4) are carried out
The result T obtainedy.The noise threshold of signal y can be accurately found by Generalized Cross Validation algorithm.
In step 4, the expression formula of Threshold Denoising Method is sheared are as follows:
Wherein: λ is the noise-removed threshold value level for passing through cross validation algorithm and obtaining, TyFor synchronous compression wavelet conversion coefficient
Matrix, α are constant, ηyTo carry out the matrix after shearing threshold denoising to synchronous compression wavelet conversion coefficient matrix.
|Ty| to seek matrix caused by signed magnitude arithmetic(al) to each element in synchronous compression wavelet conversion coefficient matrix；
|Ty|αFor the synchronous compression wavelet conversion coefficient matrix to the operation that takes absolute value | Ty| carry out power-function arithmetic, the power-function arithmetic
Index be α；λ is to utilize the noise threshold of the obtained signal y of Generalized Cross Validation algorithm in above-mentioned steps 3.2；λ α is pair
Noise threshold λ carries out power-function arithmetic, and the index of the power-function arithmetic is α.
The lesser synchronous compression wavelet conversion coefficient matrix of noise jamming is ηy.Formula (7) is to shear threshold denoising to calculate
The statement of method, meaning are as follows: the synchronous compression wavelet conversion coefficient matrix that will be taken absolute value | Ty| in each element and step
The noise threshold of 3.2 obtained signal y is compared.It is right | Ty| in the different values of each element transported accordingly
It calculates, last obtained matrix ηyAs noise jamming it is smaller after synchronous compression wavelet conversion coefficient matrix, the matrix when
Frequency is indicated as shown in Fig. 2 (d), to by corrected wavelet conversion coefficient matrix caused by higher order statistic analysis algorithm
After the denoising for carrying out next step, the noise jamming of the synchronous compression wavelet conversion coefficient matrix is largely filtered out.
In step 5, synchronous compression wavelet conversion coefficient matrix can be converted to one-dimensional time domain by synchronous compression wavelet inverse transformation
Signal, expression formula are as follows:
Wherein: Re is the operation of real part value, ωlFor discrete frequency, t is the time of sampled signal, and s (t) is sampled
Pure one-dimensional time-domain signal of the local discharge signal after denoising,For constant CψInverse operation value, Ty(ωl, t) and it is to pass through
Time t and discrete frequency ωlCalculated synchronous compression wavelet transform matrix is the calculated result of synchronous compression wavelet transformation,
ωlIt is the important parameter for synchronizing compressed wavelet transformation with t.
CψExpression formula are as follows:
Wherein: ξ is the value of wavelet mother function ψ, and the value of ξ is in [0 ,+∞], CψFor small wave constant,
ξ-1It is the inverse of wavelet mother function ψ value, i.e., all values of independent variable all carry out derivative operation in wavelet mother function；
ψ*(ξ) is the complex conjugate operation of wavelet mother function output, i.e., when wavelet mother function independent variable takes ξ, to corresponding
Wavelet mother function dependent variable carries out complex conjugate operation.
Fig. 1 (a) is that shelf depreciation of the present invention simulates one-dimensional time-domain signal；As can be seen that simulation shelf depreciation time domain letter
It number include two kinds of pulses, which belongs to damped oscillation type signal, more meet the temporal signatures of practical local discharge signal.
Fig. 1 (b) is the SS-CWT time frequency analysis figure that shelf depreciation of the present invention simulates one-dimensional time-domain signal；It can be seen that SS-
CWT time frequency analysis figure clearly shows the time-frequency band where two pulse signals, the discontinuous characteristic of the two time-frequency bands
It is shown in figure.
Fig. 1 (c) is that shelf depreciation of the present invention simulation contaminates one-dimensional time-domain signal of making an uproar；As can be seen that the dye is made an uproar, shelf depreciation is believed
Number, it joined two kinds of main noise types that Partial Discharge Detection is influenced in engineering practice: narrow-band noise and white noise.And it is added
The noise signal entered covers entire time domain, and the spatial structure of original Partial Discharge has been destroyed.
Fig. 1 (d) is the SS-CWT time frequency analysis figure that shelf depreciation of the present invention simulation contaminates one-dimensional time-domain signal of making an uproar；It can see
Out, SS-CWT time frequency analysis figure clearly shows that the time-frequency band that noise is distributed, white noise and narrow-band noise are respectively different
The projection that characteristic will also recognize that is in time frequency analysis figure.
Fig. 1 (e) is the Short Time Fourier Transform time frequency analysis figure that shelf depreciation of the present invention simulation contaminates one-dimensional time-domain signal of making an uproar；
As can be seen that the resolution ratio of the time frequency analysis figure of Short Time Fourier Transform is too low, the noise signal less than 2MHz can not be opened up
Reveal and, and time-frequency band where local discharge signal is also smudgy.
Fig. 1 (f) is the Wavelet Transform Time Frequency Analysis figure that shelf depreciation of the present invention simulation contaminates one-dimensional time-domain signal of making an uproar.It can see
Out, although the time frequency analysis figure resolution ratio of wavelet transformation is slightly above Short Time Fourier Transform time frequency analysis figure, the time frequency analysis
Figure is readable poor, and the performance of frequency is not enough.
Fig. 2 (a) is that shelf depreciation of the present invention simulation dye makes an uproar one-dimensional time-domain signal after higher order statistic analysis denoises
One-dimensional time-domain signal；Although inhibiting as can be seen that higher order statistic analysis is only rough to noise, see from denoising result
From the point of view of, which denoises effect
Fig. 2 (b) is the present invention after higher order statistic analysis denoises, the SS-CWT time frequency analysis figure of time-domain signal；It can
To find out, the white noise full of entire time-frequency band is largely reduced, and the narrow-band noise near signal frequency is also a large amount of accordingly
It filters out, the time-frequency characteristics of original clean signal are obvious.
Fig. 2 (c) is that shelf depreciation of the present invention simulation dye makes an uproar one-dimensional time-domain signal by higher order statistic analysis and broad sense friendship
After pitching verification algorithm denoising, one-dimensional time-domain signal and original clean signal contrast figure；As can be seen that two kinds of algorithm joints of the present invention
Denoising works well, and original signal waveform time domain waveform feature is obvious, and waveform distortion is low, with original clean signal contrast phase
It is high like degree.
Fig. 2 (d) is the present invention after higher order statistic analysis and generalized crossover threshold denoising, the SS-CWT of time-domain signal
Time frequency analysis figure；As can be seen that the signal time-frequency characteristics after two kinds of algorithm joint denoisings of the present invention are complete and clear, but still can
To see that some time-frequency bands there are some faint white noises, show as nattier blue irregular curve and region in time-frequency figure.
Fig. 2 (e) for higher order statistic analysis of the present invention and the removed noise signal of generalized crossover thresholding algorithm SS-CWT
Time frequency analysis figure；As can be seen that the white Gaussian noise of construction is distributed in entire time-frequency band, shown as in time-frequency figure light blue
Irregular curve and region.And the narrow-band noise constructed is stablized in some frequency separation, and a Duan Shen is shown as in time-frequency figure
Blue distortion and continuous sine curve.
Fig. 3 (a) is the present invention after higher order statistic analysis denoises, the intersection small echo spectrogram of time-domain signal；It can see
Out, the signal after tentatively denoising and signal energy of the original clean signal in some time-frequency bands are more similar, are intersecting
Yellow area is shown as in small echo spectrogram, the more similar then yellow of signal energy is deeper.Signal and original clean after preliminary denoising
Phase angle difference between signal is smaller in some time-frequency bands, and the direction for black arrow, phase angle are cashed in intersecting small echo spectrogram
Difference is smaller, then arrow is directed toward the level that more levels off to.Comprehensive analysis, preliminary Denoising Algorithm has certain noise removal capability, but denoises effect
Fruit is limited.Fig. 3 (b) is the present invention after higher order statistic analysis and generalized crossover threshold denoising, and the intersection of time-domain signal is small
Wave spectrogram；As can be seen that only having fraction regional signal energy similar compared to preliminary denoising, by two algorithm joint denoisings
It afterwards, then is that the buff area of large area is shown as in intersecting small echo spectrogram there are the signal energy at more time-frequency band is similar
Domain.Equally, the relative phase angle difference of the signal after two kinds of algorithm joint denoisings and original clean signal is in more time-frequency band
Interior lower, cashing in intersecting small echo spectrogram is that the direction of the similar region black arrow of signal energy largely levels off to water
It is flat.Comprehensive analysis, the present invention mention two kinds of algorithms and carry out that joint denoising effect is outstanding, and denoising precision height, waveform distortion is low.
Fig. 4 (a) is the shelf depreciation laboratory simulations platform that the present invention constructs；As can be seen that the part that the present invention constructs
Discharge test room emulation platform is by testing transformer, divider, slide rheostat, partial discharge model, current transformer, oscillography
Device and computer are constituted.Partial discharge model is made of insulation lubricating cup, insulating paper and copper electrode, the model energy approximate simulation transformer
Partial discharge phenomenon under paper oil insulation environment.The oscillograph highest sample rate of selection can reach 5GS/s, therefore can guarantee the platform
The Partial Discharge signal accuracy of record.
The shelf depreciation power-frequency voltage that Fig. 4 (b) obtains for the present invention from the shelf depreciation laboratory simulations platform constructed
Time-domain signal figure；As can be seen that the power-frequency voltage that experiment porch is applied to partial discharge model, sampled point reaches 2,000 ten thousand.
Therefore the Partial Discharge signal integrity that the platform is recorded, the progress of enough subsequent analysis work.
The shelf depreciation electric current time domain that Fig. 4 (c) obtains for the present invention from the shelf depreciation laboratory simulations platform constructed
Signal graph；As can be seen that the shelf depreciation current waveform that the experiment porch in the present invention is recorded includes 2,000 ten thousand sampled points,
Partial discharge phenomenon is obvious.The waveform contains a partial discharge pulse more than ten, and the region that red block is chosen is shelf depreciation
Impulse waveform feature is one the most apparent.
Fig. 4 (d) is one section of shelf depreciation time domain impulsive signals that the present invention is intercepted from laboratory shelf depreciation current signal
Figure；As can be seen that partial discharge pulse's signal of interception includes 600 sampled points, after signal amplitude normalization, pulse signal is made an uproar
Sound is faint, and Partial Discharge feature is obvious.
Fig. 4 (e) is one section of shelf depreciation time domain impulsive signals that the present invention is intercepted from laboratory shelf depreciation current signal
SS-CWT time frequency analysis figure；As can be seen that the obtained partial discharge pulse's waveform of experiment porch in entire time-frequency band still
There are a large amount of white noises, but the energy of these white noises is small, and nattier blue irregular lines are shown as in time frequency analysis figure.
And original signal pulse, near 1 the μ s and 16MHz in time frequency analysis figure, color is navy blue.
Fig. 4 (f) is that narrow-band noise and white noise is added to one section of shelf depreciation time domain impulsive signals of interception in the present invention
Time-domain signal figure；A large amount of white noises and narrowband noise signals are by partial discharge pulse's signal characteristic added by as can be seen that
It covers.
Fig. 4 (g) is the SS-CWT time frequency analysis after the present invention makes an uproar to one section of shelf depreciation time domain impulsive signals dye of interception
Figure；As can be seen that irregular light blue region is increased in time frequency analysis figure, the noise energy of low frequency region is wanted before also making an uproar compared with dye
Greatly, it is deeper that blue region color is shown as in time frequency analysis figure.Comprehensive analysis, in the environment of strong noise, SS-CWT time-frequency
Parser remains to distinguish purified signal component and noise component(s) with biggish resolution ratio.
Fig. 4 (h) is after the present invention makes an uproar to one section of shelf depreciation time domain impulsive signals dye of interception, using high-order statistic point
Time-domain signal figure after analysis and the joint denoising of Generalized Cross Validation algorithm；As can be seen that the present invention mentions what two kinds of algorithms combined
Signal denoising research preferably can carry out noise suppressed to local discharge pulse signal.The time domain waveform of signal only remains after denoising
The lower apparent pulse of feature and a small amount of faint noise.
Fig. 4 (i), which is the present invention, tests using higher order statistic analysis and generalized crossover dye laboratory local discharge signal of making an uproar
After demonstrate,proving algorithm joint denoising, the SS-CWT time frequency analysis figure of time-domain signal.Contained by local discharge signal after can be seen that denoising
Noise energy is lower, and it is shallower that the irregular curve color full of entire time-frequency band is shown as in time-frequency figure.Comprehensive analysis, this hair
Bright mentioned algorithm Denoising Study equally has stronger noise removal capability to laboratory local discharge signal.
Table 1 compares algorithm
| Serial number | 1 | 2 | 3 |
| Method | S-transformation+singular value decomposition | Improved Protrugram+ wavelet transformation | VMD+ wavelet transformation |
As shown in Table 1, method 1 presses down narrow-band noise using the time-frequency matrix that generalized S-transform is generated using the transformation
System, white noise is handled using singular value decomposition, and the Denoising Algorithm that the document proposes is than traditional one-dimensional time-domain signal
Denoising method is more effective, and wave distortion is smaller, preferably remains the feature of signal.Method 2 combines improved Protrugram
Algorithm and wavelet algorithm carry out joint denoising to narrow-band noise and white noise, compared to spectrum kurtosis and simple mathematical morphology filter
There is lower computation complexity.Method 3 carries out inhibition and the Traditional Wavelet of narrow band signal using variation mode decomposition algorithm
Convert the inhibition that white noise is carried out to signal.Variation mode can preferably separate the frequency band where narrow band signal, and Traditional Wavelet becomes
Transducing carries out white noise inhibition to compared with Low SNR signal.
The analysis of 2 denoising result of table
From table 2 it can be seen that for RMSE index, the present invention puts up the best performance, using Generalized Cross Validation as the denoising of core
Research ensures the minimum of root-mean-square error.But for NCC index, the performance of two pulses is not as good as method 3, this is because originally
The suppressing method of invention contains two processes, and after pretreated initial denoising, Generalized Cross Validation algorithm is again to making an uproar
Sonic system number will lead to a degree of " overdamping " when being decayed, leading to denoising waveform, there are small distortions.The present invention is wide
Using trimming threshold value, the signal distortion after denoising is low for adopted cross validation algorithm, waveform rising edge and failing edge and original clean
Signal waveform matching, therefore in the comparison of VTP, the method for the present invention is slightly poorer than the method that method 3 proposes.Present invention denoising belongs to connection
Hop algorithm removes two kinds of common noises of shelf depreciation jointly, and noise removal capability is better than general algorithm, therefore also shows in RNL index
It is excellent.
3 relative time evaluation index of table
| Method | 1 | 2 | 3 | The present invention |
| Relative time | 2.12 | 1.83 | 1.27 | 1.00 |
From table 3 it can be seen that algorithm proposed by the present invention is most fast comparing the operation time of its excess-three kind algorithm.Therefore
Consolidated statement 2 and table 3 are it is found that the global de-noising ability of the method for the present invention is better than three kinds of compared algorithms.
Claims (6)
1. being based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method, it is characterised in that the following steps are included:
Step 1: wavelet transformation being carried out to the one-dimensional time-domain signal of the partial discharge of transformer of acquisition, obtains the small of one-dimensional time-domain signal
Wave conversion coefficient matrix；
Step 2: higher order statistic analysis being carried out to wavelet conversion coefficient matrix, using kurtosis threshold value criterion, to wavelet transformation system
Matrix number carries out the preliminary inhibition of noise, obtains corrected wavelet conversion coefficient matrix；
Step 3: utilizing corrected wavelet conversion coefficient matrix, synchronous compression wavelet conversion coefficient matrix is obtained, using broad sense
Cross validation algorithm obtains the noise threshold level of synchronous compression wavelet transformation SS-CWT coefficient matrix；
Step 4: noise threshold level is utilized, using shearing thresholding algorithm, to synchronous compression wavelet transformation SS-CWT coefficient matrix
The inhibition for carrying out residual noise, obtains the lesser synchronous compression wavelet conversion coefficient matrix of noise jamming；
Step 5: to the lesser synchronous compression wavelet conversion coefficient matrix of noise jamming, synchronizing compressed wavelet inverse transformation, obtain
To the more pure one-dimensional time-domain signal of shelf depreciation.
2. being based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method according to claim 1, feature exists
In: in step 1, the wavelet transformation is continuous wavelet transform (Continuous Wavelet Transform, CWT), table
Up to formula are as follows:
Wherein: * represents complex conjugate operation；<s, ψ>be time-domain signal s and wavelet mother function ψ operator, WsAs wavelet transformation
Coefficient matrix, a are scale factors, and τ is time shift method；
Wavelet conversion coefficient matrix is the output result W of wavelet transformations(a, τ), rely on wavelet mother function to time-domain signal into
Row transformation.
3. being based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method according to claim 1, feature exists
In:
Step 2, higher order statistic analysis is the preliminary denoising to signal, comprising the following steps:
Step 2.1: calculate the kurtosis value of wavelet conversion coefficient matrix:
Wherein: σWsFor the standard deviation of wavelet coefficient, μWsFor the mean value of matrix W s wavelet coefficient, N is sampling number, kurtsFor letter
The kurtosis value of the wavelet conversion coefficient of number s, WsFor the coefficient matrix of wavelet transformation；μ is mean operation, is averaging to operand
Value；σ is standard difference operation, seeks standard deviation to operand；
Step 2.2: construction kurtosis threshold value criterion:
α is confidence level, and N is signal sampling points, kurtsFor the kurtosis value of the wavelet conversion coefficient of signal s；
The kurtosis of each element in wavelet conversion coefficient matrix that wavelet transformation is obtained and the generation of above-mentioned kurtosis threshold value criterion
Threshold value kurtsIt compares, if the element in the matrix is less than kurtosis threshold value, the element is in wavelet conversion coefficient matrix
Value is 0, and value that is on the contrary then retaining the element is constant；Each element in wavelet conversion coefficient matrix is subjected to above-mentioned peak
The matrix constituted after degree threshold denoising is corrected wavelet conversion coefficient matrix.
4. being based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method according to claim 1, feature exists
In:
Step 3, the processing of synchronous compression threshold denoising is the denoising again to dye noise cancellation signal, comprising the following steps:
Step 3.1: calculate synchronous compression wavelet conversion coefficient matrix:
Wherein: TyFor synchronous compression wavelet conversion coefficient matrix, fs is the sample frequency of signal, and N is the sampling number of signal, ωl
It is its l ∈ [1, N] of first of discrete frequency, △ ω is the difference of discrete frequency；akFor k-th of wavelet coefficient WyScale factor；
WyFor corrected wavelet conversion coefficient matrix described in above-mentioned steps 2.2；
△ω-1For the inverse of two neighboring discrete frequency difference, i.e., inverse of a matrix operation is carried out to the difference of discrete frequency；M is that summation is sentenced
Disconnected formula, the scale factor a and time shift method τ for only meeting the judgement formula can just calculate corresponding synchronous compression wavelet transformation system
Matrix number Ty；τ is the time shift method of wavelet conversion coefficient matrix, is an important parameter of wavelet transformation；△akFor adjacent scale
The difference of the factor；ω is discrete frequency；△ ω is the difference of adjacent discrete frequency；ωlFor first of discrete frequency；ωl-1It is l-1
Discrete frequency；ak-1For -1 scale factor of kth；
w(ak, τ) and it is candidate instantaneous frequency, the frequency expression are as follows:
Wherein: Wy(a, τ) is Continuous Wavelet Transform Coefficients matrix, while being also that corrected small echo described in above-mentioned steps 2.2 becomes
Coefficient matrix is changed, a is scale factor, and △ a is the difference of scale factor, and τ is time shift method,For derivative operation, wyAs candidate wink
When frequency；I is imaginary unit；
Step 3.2: it is horizontal to calculate noise-removed threshold value:
Wherein: GCV is Generalized Cross Validation algorithm coefficient,It is using the threshold coefficient of synchronous compression domain threshold value λ, N0It is synchronous
The coefficient number of threshold level λ is not up in compressed wavelet transform coefficient matrix, N sampling number, λ is the denoising threshold finally acquired
Value is horizontal；
Synchronous compression wavelet transformation SS-CWT coefficient matrix is the obtained knot of synchronous compression wavelet transformation that formula (4) is carried out
Fruit Ty；The noise threshold of signal y can be accurately found by Generalized Cross Validation algorithm.
5. being based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method according to claim 1, feature exists
In:
In step 4, the expression formula of Threshold Denoising Method is sheared are as follows:
Wherein: λ is the noise-removed threshold value level for passing through cross validation algorithm and obtaining, TyFor synchronous compression wavelet conversion coefficient matrix,
α is constant, ηyTo carry out the matrix after shearing threshold denoising to synchronous compression wavelet conversion coefficient matrix；
|Ty| to seek matrix caused by signed magnitude arithmetic(al) to each element in synchronous compression wavelet conversion coefficient matrix；|Ty|α
For the synchronous compression wavelet conversion coefficient matrix to the operation that takes absolute value | Ty| carry out power-function arithmetic, the finger of the power-function arithmetic
Number is α；λ is to utilize the noise threshold of the obtained signal y of Generalized Cross Validation algorithm in above-mentioned steps 3.2；λαFor to noise
Threshold value λ carries out power-function arithmetic, and the index of the power-function arithmetic is α；
The lesser synchronous compression wavelet conversion coefficient matrix of noise jamming is ηy；Formula (7) is to shear the table of Threshold Filter Algorithms
State, meaning are as follows: the synchronous compression wavelet conversion coefficient matrix that will be taken absolute value | Ty| in each element and step 3.2 required by
The noise threshold of the signal y obtained is compared；It is right | Ty| in the different values of each element carry out corresponding operation, last institute
The matrix η acquiredyAs noise jamming it is smaller after synchronous compression wavelet conversion coefficient matrix.
6. being based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method according to claim 1, feature exists
In: in step 5, synchronous compression wavelet conversion coefficient matrix can be converted to one-dimensional time-domain signal by synchronous compression wavelet inverse transformation,
Its expression formula are as follows:
Wherein: Re is the operation of real part value, ωlFor discrete frequency, t is the time of sampled signal, and s (t) is put for the part sampled
Pure one-dimensional time-domain signal of the electric signal after denoising,For constant CψInverse operation value, Ty(ωl, t) and it is to pass through time t
With discrete frequency ωlCalculated synchronous compression wavelet transform matrix is the calculated result of synchronous compression wavelet transformation, ωlWith t
It is the important parameter for synchronizing compressed wavelet transformation；
CψExpression formula are as follows:
Wherein: ξ is the value of wavelet mother function ψ, and the value of ξ is in [0 ,+∞], CψFor small wave constant, ξ-1It is wavelet mother function ψ
The inverse of value, i.e., all values of independent variable all carry out derivative operation in wavelet mother function；ψ*(ξ) is wavelet mother function output
Complex conjugate operation carries out complex conjugate operation to corresponding wavelet mother function dependent variable that is, when wavelet mother function independent variable takes ξ.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910543829.9A CN110244202B (en) | 2019-06-21 | 2019-06-21 | Transformer partial discharge denoising method based on synchronous compression wavelet transform domain |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910543829.9A CN110244202B (en) | 2019-06-21 | 2019-06-21 | Transformer partial discharge denoising method based on synchronous compression wavelet transform domain |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110244202A true CN110244202A (en) | 2019-09-17 |
| CN110244202B CN110244202B (en) | 2021-07-20 |
Family
ID=67888818
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910543829.9A Active CN110244202B (en) | 2019-06-21 | 2019-06-21 | Transformer partial discharge denoising method based on synchronous compression wavelet transform domain |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110244202B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111368713A (en) * | 2020-03-02 | 2020-07-03 | 西南交通大学 | Vehicle network system harmonic time-frequency analysis method based on synchronous compression wavelet transform |
| CN111516548A (en) * | 2020-04-23 | 2020-08-11 | 华南理工大学 | Cloud platform-based charging pile system for realizing power battery fault diagnosis |
| CN112633225A (en) * | 2020-12-31 | 2021-04-09 | 矿冶科技集团有限公司 | Mining microseismic signal filtering method |
| CN112964355A (en) * | 2020-12-08 | 2021-06-15 | 国电南京自动化股份有限公司 | Instantaneous frequency estimation method based on spline frequency modulation wavelet-synchronous compression algorithm |
| CN114779028A (en) * | 2022-06-13 | 2022-07-22 | 北京京能能源技术研究有限责任公司 | Generator partial discharge online monitoring device and monitoring method |
| CN115728612A (en) * | 2022-12-01 | 2023-03-03 | 广州广电计量检测股份有限公司 | Transformer discharge fault diagnosis method and device |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103576060A (en) * | 2013-10-11 | 2014-02-12 | 华南理工大学 | Partial discharge signal denoising method based on wavelet adaptive threshold |
| CN103903232A (en) * | 2014-04-10 | 2014-07-02 | 北京工业大学 | Method for conducting image denoising and enhancing in wavelet domain through evolutionary programming |
| CN103902844A (en) * | 2014-04-24 | 2014-07-02 | 国家电网公司 | Transformer vibration signal de-noising method based on EEMD kurtosis threshold value |
| CN104849757A (en) * | 2015-04-30 | 2015-08-19 | 长江大学 | System and method for eliminating random noise in seismic signals |
| CN105223482A (en) * | 2015-10-22 | 2016-01-06 | 西安交通大学 | The wavelet decomposition two-value denoising method of partial-discharge ultrahigh-frequency signal waveform |
| CN107067385A (en) * | 2017-01-23 | 2017-08-18 | 上海兴芯微电子科技有限公司 | A kind of image enchancing method and device |
| CN107356843A (en) * | 2017-04-17 | 2017-11-17 | 武汉科技大学 | The partial discharge of transformer method for diagnosing faults of small echo is synchronously extruded based on gradient threshold |
| CN108154081A (en) * | 2016-11-30 | 2018-06-12 | 东北林业大学 | Based on instantaneous frequency stability SWT logistics equipment vibration signal noise-reduction methods |
| CN108983051A (en) * | 2018-07-25 | 2018-12-11 | 武汉科技大学 | Based on the synchronous shelf depreciation kind identification method for squeezing wavelet transformation |
| CN109782347A (en) * | 2019-01-18 | 2019-05-21 | 南京邮电大学 | A kind of seismic travel time inversion imaging method based on wavelet analysis |
-
2019
- 2019-06-21 CN CN201910543829.9A patent/CN110244202B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103576060A (en) * | 2013-10-11 | 2014-02-12 | 华南理工大学 | Partial discharge signal denoising method based on wavelet adaptive threshold |
| CN103903232A (en) * | 2014-04-10 | 2014-07-02 | 北京工业大学 | Method for conducting image denoising and enhancing in wavelet domain through evolutionary programming |
| CN103902844A (en) * | 2014-04-24 | 2014-07-02 | 国家电网公司 | Transformer vibration signal de-noising method based on EEMD kurtosis threshold value |
| CN104849757A (en) * | 2015-04-30 | 2015-08-19 | 长江大学 | System and method for eliminating random noise in seismic signals |
| CN105223482A (en) * | 2015-10-22 | 2016-01-06 | 西安交通大学 | The wavelet decomposition two-value denoising method of partial-discharge ultrahigh-frequency signal waveform |
| CN108154081A (en) * | 2016-11-30 | 2018-06-12 | 东北林业大学 | Based on instantaneous frequency stability SWT logistics equipment vibration signal noise-reduction methods |
| CN107067385A (en) * | 2017-01-23 | 2017-08-18 | 上海兴芯微电子科技有限公司 | A kind of image enchancing method and device |
| CN107356843A (en) * | 2017-04-17 | 2017-11-17 | 武汉科技大学 | The partial discharge of transformer method for diagnosing faults of small echo is synchronously extruded based on gradient threshold |
| CN108983051A (en) * | 2018-07-25 | 2018-12-11 | 武汉科技大学 | Based on the synchronous shelf depreciation kind identification method for squeezing wavelet transformation |
| CN109782347A (en) * | 2019-01-18 | 2019-05-21 | 南京邮电大学 | A kind of seismic travel time inversion imaging method based on wavelet analysis |
Non-Patent Citations (3)
| Title |
|---|
| 卢才武，等: "基于SS-CWT和Yoon-V自适应阈值的弱振幅微震分频降噪方法", 《地球物理学进展》 * |
| 尹欣，等: "《摩擦焊及其检测技术》", 30 June 2012 * |
| 肖昌明，等: "基于广义交叉阈值同步压缩小波的超声信号特征提取方法", 《机械强度》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111368713A (en) * | 2020-03-02 | 2020-07-03 | 西南交通大学 | Vehicle network system harmonic time-frequency analysis method based on synchronous compression wavelet transform |
| CN111368713B (en) * | 2020-03-02 | 2022-06-28 | 西南交通大学 | Vehicle network system harmonic time-frequency analysis method based on synchronous compression wavelet transform |
| CN111516548A (en) * | 2020-04-23 | 2020-08-11 | 华南理工大学 | Cloud platform-based charging pile system for realizing power battery fault diagnosis |
| CN111516548B (en) * | 2020-04-23 | 2021-11-23 | 华南理工大学 | Cloud platform-based charging pile system for realizing power battery fault diagnosis |
| CN112964355A (en) * | 2020-12-08 | 2021-06-15 | 国电南京自动化股份有限公司 | Instantaneous frequency estimation method based on spline frequency modulation wavelet-synchronous compression algorithm |
| CN112964355B (en) * | 2020-12-08 | 2023-04-25 | 国电南京自动化股份有限公司 | Instantaneous frequency estimation method based on spline frequency modulation wavelet-synchronous compression algorithm |
| CN112633225A (en) * | 2020-12-31 | 2021-04-09 | 矿冶科技集团有限公司 | Mining microseismic signal filtering method |
| CN112633225B (en) * | 2020-12-31 | 2023-07-18 | 矿冶科技集团有限公司 | Mining microseism signal filtering method |
| CN114779028A (en) * | 2022-06-13 | 2022-07-22 | 北京京能能源技术研究有限责任公司 | Generator partial discharge online monitoring device and monitoring method |
| CN115728612A (en) * | 2022-12-01 | 2023-03-03 | 广州广电计量检测股份有限公司 | Transformer discharge fault diagnosis method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110244202B (en) | 2021-07-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110244202A (en) | Based on synchronous compression wavelet transformed domain partial discharge of transformer denoising method | |
| Zhu et al. | Seismic signal denoising and decomposition using deep neural networks | |
| Jian et al. | On the denoising method of prestack seismic data in wavelet domain | |
| Ikram et al. | Digital audio forensics using background noise | |
| RU2570592C1 (en) | Method of detecting and analysing acoustic emission signals | |
| CN110967599A (en) | Electric energy quality disturbance detection and positioning algorithm | |
| CN109557429A (en) | Based on the GIS partial discharge fault detection method for improving wavelet threshold denoising | |
| CN106771598B (en) | A kind of Adaptive spectra kurtosis signal processing method | |
| CN109932624B (en) | Cable partial discharge period narrow-band interference denoising method based on Gaussian scale space | |
| Hussein et al. | Denoising of acoustic partial discharge signals corrupted with random noise | |
| CN104849590A (en) | Method for detecting weak pulse signals under mixed noise interference | |
| Le et al. | Weak LFM signal dectection based on wavelet transform modulus maxima denoising and other techniques | |
| Bin-Habtoor et al. | Removal speckle noise from medical image using image processing techniques | |
| Sharma et al. | ECG denoising using weiner filter and adaptive least mean square algorithm | |
| CN109630908B (en) | Pipeline leakage positioning method capable of reducing noise for multiple times | |
| Wang et al. | A method for enhancement and automated extraction and tracing of Odontoceti whistle signals base on time-frequency spectrogram | |
| Conru et al. | Time-frequency detection using Gabor filter bank and Viterbi based grouping algorithm | |
| CN107248869B (en) | A kind of multicomponent linear frequency-modulated signals denoising method based on Lyu's distribution | |
| ÖZAYDIN et al. | Speech enhancement using maximal overlap discrete wavelet transform | |
| CN103915102B (en) | Method for noise abatement of LFM underwater sound multi-path signals | |
| Fan et al. | Interference mitigation for synthetic aperture radar using deep learning | |
| Cicone et al. | Jot: a variational signal decomposition into jump, oscillation and trend | |
| CN110780162A (en) | Method for extracting partial discharge signal of primary and secondary fusion power distribution switch and detection device | |
| Boufenar et al. | Time-frequency analysis techniques review and their application on roller bearings prognostics | |
| Faisal et al. | Suppression of false-terms in wigner-ville distribution using time and frequency windowing |
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 |