High-frequency working condition partial discharge signal denoising method based on CFOA-VMD combined multi-scale wavelet threshold

Abstract

The invention relates to the technical field of high-frequency working condition partial discharge signal de-noising, in particular to a high-frequency working condition partial discharge signal de-noising method based on CFOA-VMD combined with a multi-scale wavelet threshold, which comprises the following steps: acquiring an oil-paper insulation partial discharge signal, and importing the partial discharge signal as an input into a fishing optimization algorithm, solving by adopting a variational mode decomposition algorithm, and outputting an optimal parameter; decomposing the partial discharge signal based on the optimal parameter to obtain an IMF component; calculating a kurtosis value of each IMF component, and screening effective signal components based on the kurtosis values; performing multi-scale adaptive wavelet threshold processing on the effective signal component; and performing wavelet reconstruction on the processed effective signal component to obtain a denoised partial discharge signal. The signal noise reduction time can be shortened, and the signal precision can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of high-frequency condition partial discharge signal denoising, and particularly to a high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined multi-scale wavelet thresholding. Background Art

[0002] High-frequency transformers are the core components for electrical isolation and voltage transformation in power electronic transformers. Connected to high-power power electronic converters, they operate under repetitive pulse square-wave voltages with short rise times, large amplitudes, and high frequencies (400 Hz - 20 kHz) for a long time. Frequent high-frequency pulses cause problems such as partial discharge, severe heating, and premature insulation failure of the insulation. Existing research shows that high-frequency partial discharge is the main cause of insulation failure in high-frequency transformers. At the same time, existing experimental studies have found that the power supply signal under high-frequency conditions is mixed with the partial discharge signal, and the true partial discharge signal has been submerged by the high-frequency power supply signal, resulting in blurred detected partial discharge signals and affecting the effective identification and positioning of discharge activities. Therefore, how to accurately denoise the partial discharge signal of oil-paper insulation under high-frequency conditions is a key scientific issue to ensure the safe and stable operation of high-frequency transformers.

[0003] The partial discharge signal of oil-paper insulation under high-frequency conditions is usually interfered by high-frequency power supply noise. During the partial discharge detection process, the noise generated by the high-frequency power supply persists. The frequency band range of the high-frequency power supply background noise is usually relatively wide. Common low-frequency noise is in the range of 10 kHz to 100 kHz, mainly caused by switching devices in the power supply; intermediate-frequency noise is in the range of 100 kHz to 1 MHz, mainly caused by the noise generated by power supply regulation; high-frequency noise is in the range of 1 MHz to dozens of MHz, and the noise in this frequency band may be generated by electromagnetic interference in the power supply circuit or harmonics of the switching power supply. Before conducting the oil-paper insulation partial discharge experiment, the high-frequency power supply background noise was detected first, and the detected background noise was analyzed by Fourier transform. It was found that the main noise with large amplitudes was concentrated around 23.5 MHz, 47.5 MHz, and 71 MHz, and there would always be noise with small amplitudes in the remaining frequency ranges, which affected the subsequent research on partial discharge signals. Therefore, it is necessary to accurately denoise the partial discharge signal of oil-paper insulation under high-frequency conditions. Summary of the Invention

[0004] The purpose of the present invention is to provide a high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined multi-scale wavelet thresholding, to solve the problem of the mixing of high-frequency condition power supply signals and partial discharge signals, and the problems in the existing technology that the decomposition layer number and penalty factor of the VMD algorithm need to be determined manually according to experience, and the optimization time of other intelligent optimization algorithms is too long when optimizing the decomposition layer number and penalty factor, so as to shorten the PD signal noise reduction time and improve the PD signal accuracy.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold, comprising:

[0007] Obtain an oil-paper insulation partial discharge signal, import the partial discharge signal as an input into a fishing optimization algorithm, and use a variational mode decomposition algorithm for solution to output optimal parameters;

[0008] Decompose the partial discharge signal based on the optimal parameters to obtain IMF components;

[0009] Calculate the kurtosis value of each IMF component, and screen out effective signal components based on the kurtosis value;

[0010] Perform multi-scale adaptive wavelet threshold processing on the effective signal components;

[0011] Perform wavelet reconstruction on the processed effective signal components to obtain a denoised partial discharge signal.

[0012] Optionally, the partial discharge signal includes any one of a partial discharge simulation signal and a partial discharge test signal, wherein the partial discharge simulation signal is obtained by simulating through constructing a high-frequency condition oil-paper insulation partial discharge simulation signal model, and the partial discharge test signal is obtained by conducting tests through building an oil-paper insulation partial discharge test platform under high-frequency pulses.

[0013] Optionally, obtaining the partial discharge simulation signal by simulating through constructing a high-frequency condition oil-paper insulation partial discharge simulation signal model includes:

[0014] Construct a single exponential decay model and a double exponential decay oscillation model;

[0015] Set the parameters of the simulation signal, and add periodic narrowband interference and white noise, and perform simulation based on the single exponential decay model and the double exponential decay oscillation model to obtain an oil-paper insulation partial discharge simulation signal.

[0016] Optionally, the single exponential decay model is:

[0017]

[0018] The double exponential decay oscillation model is:

[0019]

[0020] Among them, f(t) is the expression of the single-exponential decay model, X(t) is the expression of the double-exponential decay oscillation model, A and B are the amplitudes of the partial discharge signals of oil-paper insulation under high-frequency conditions, τ is the decay constant of the signal, t0 is the starting time of the discharge pulse of the signal, t is the current time of the discharge pulse of the signal, and f c is the oscillation frequency of the signal.

[0021] Optionally, obtaining the partial discharge test signal through building a partial discharge test platform for oil-paper insulation under high-frequency pulses includes:

[0022] Using a high-frequency pulse power supply to simulate a high-frequency square-wave pulse voltage, a high-voltage probe to measure the voltage test signal, and an ultra-high-frequency current sensor to collect the partial current signal, and inputting the partial discharge current signal and the voltage test signal into a high-frequency oscilloscope to build the test platform;

[0023] Before the test, detecting the background noise generated by the high-frequency power supply and performing Fourier analysis on the background noise to obtain the noise concentration frequency band of the background noise;

[0024] In the test platform, using a needle-plate electrode model to simulate partial discharge defects according to the noise concentration frequency band, preprocessing the insulating paper specimen and placing it on the test platform, and performing a partial discharge test of oil-paper insulation under high-frequency pulses using the step-up voltage method to obtain the partial discharge test signal.

[0025] Optionally, taking the partial discharge signal as an input and importing it into the fishing optimization algorithm, and using the variational mode decomposition algorithm to solve, the output optimal parameters include:

[0026] Taking the partial discharge signal as an input and importing it into the fishing optimization algorithm, and setting the initial parameters of the fishing optimization algorithm;

[0027] Setting the number of decomposition modes K and the penalty factor α in the variational mode decomposition algorithm as the optimization objectives, setting the minimum envelope entropy as the fitness function of the fishing optimization algorithm, and performing the solution to obtain the optimal parameters of the variational mode decomposition.

[0028] Optionally, the minimum envelope entropy is:

[0029] Fitness=min(Ep i );

[0030]

[0031] Among them, Fitness is the minimum envelope entropy expression, Ep iis the envelope entropy of each IMF component obtained after the partial discharge signal of oil-paper insulation is decomposed by the variational mode decomposition algorithm. a(j) is the envelope signal obtained after the IMF component undergoes the Hilbert transform, and p j is the normalized form of a(j), and N is the number of envelope signals.

[0032] Optionally, the method for calculating the kurtosis value of each of the IMF components is as follows:

[0033]

[0034] where Q is the kurtosis value of the IMF component, μ is the arithmetic mean of the signal amplitude, and E(x - μ) 4 is the fourth-order mathematical expectation, and σ is the standard deviation.

[0035] The beneficial effects of the present invention are as follows:

[0036] The high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold proposed by the present invention can not only remove high-frequency power supply background noise signals, but also solve the problem of too long optimization time for other intelligent optimization algorithms when optimizing the decomposition layer number and penalty factor, thereby shortening the PD signal denoising time, improving the PD signal accuracy, making the denoised signal have a better signal-to-noise ratio, a smaller root mean square error, and a waveform similarity coefficient. Description of the Drawings

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0038] Figure 1 is the flowchart of the high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold according to the embodiment of the present invention;

[0039] Figure 2 is the simulated signal diagram of the simulated high-frequency condition oil-paper insulation partial discharge signal according to the embodiment of the present invention;

[0040] Figure 3 is the iteration diagram of the CFOA-VMD optimal parameters of the simulated signal according to the embodiment of the present invention;

[0041] Figure 4 is the time-domain diagram of each decomposition component of the simulated signal according to the embodiment of the present invention;

[0042] Figure 5 is the frequency-domain diagram of each decomposition component of the simulated signal according to the embodiment of the present invention;

[0043] Figure 6 The denoised signal diagram of the analog signal after being processed by this method in the embodiment of the present invention;

[0044] Figure 7 The frequency domain diagram of the analog signal and the denoised signal in the embodiment of the present invention;

[0045] Figure 8 The time domain diagram and frequency domain diagram of the background noise of the high-frequency power supply in the embodiment of the present invention;

[0046] Figure 9 The measured partial discharge signal diagram of oil-paper insulation under high-frequency conditions in the embodiment of the present invention;

[0047] Figure 10 The CFOA-VMD optimal parameter iteration diagram of the measured partial discharge signal in the embodiment of the present invention;

[0048] Figure 11 The time domain diagram of each decomposed component of the measured partial discharge signal in the embodiment of the present invention;

[0049] Figure 12 The frequency domain diagram of each decomposed component of the measured partial discharge signal in the embodiment of the present invention;

[0050] Figure 13 The denoised signal diagram of the measured partial discharge signal after being processed by this method in the embodiment of the present invention;

[0051] Figure 14 The frequency domain diagram of the measured partial discharge signal and the denoised signal in the embodiment of the present invention. Detailed implementation manners

[0052] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0053] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0054] Embodiment 1:

[0055] This embodiment provides a denoising method for high-frequency condition partial discharge signals based on CFOA-VMD combined with multi-scale wavelet threshold, as Figure 1 shown, including:

[0056] Obtain the partial discharge signal of oil-paper insulation, import the partial discharge signal as input into the fishing optimization algorithm, and use the variational mode decomposition algorithm for solution to output the optimal parameters;

[0057] Decompose the partial discharge signal based on the optimal parameters to obtain IMF components;

[0058] Calculate the kurtosis value of each IMF component, and screen out the effective signal components based on the kurtosis value;

[0059] Perform multi-scale adaptive wavelet threshold processing on the effective signal components;

[0060] Perform wavelet reconstruction on the processed effective signal components to obtain the denoised partial discharge signal.

[0061] Specifically, this embodiment can not only remove the high-frequency power supply background noise signal, but also solve the problem of too long optimization time of other intelligent optimization algorithms when optimizing the decomposition layer number and penalty factor, thereby shortening the PD signal denoising time, improving the PD signal accuracy, making the denoised signal have a better signal-to-noise ratio, smaller root mean square error and waveform similarity coefficient.

[0062] Further, the partial discharge signal includes any one of a partial discharge simulation signal and a partial discharge test signal. Among them, the partial discharge simulation signal is obtained by simulating through constructing a high-frequency condition oil-paper insulation partial discharge simulation signal model, and the partial discharge test signal is obtained by conducting tests through building an oil-paper insulation partial discharge test platform under high-frequency pulses.

[0063] Further, obtaining the partial discharge simulation signal by simulating through constructing a high-frequency condition oil-paper insulation partial discharge simulation signal model includes:

[0064] Construct a single exponential decay model and a double exponential decay oscillation model;

[0065] Set the parameters of the simulation signal, and add periodic narrowband interference and white noise, and perform simulation based on the single exponential decay model and the double exponential decay oscillation model to obtain the partial discharge simulation signal of oil-paper insulation.

[0066] Further, the single exponential decay model is:

[0067]

[0068] The double exponential decay oscillation model is:

[0069]

[0070] Among them, f(t) is the expression of the single exponential decay model, X(t) is the expression of the double exponential decay oscillation model, A and B are the amplitudes of the partial discharge signals of oil-paper insulation under high-frequency conditions, τ is the decay constant of the signal, t0 is the starting moment of the discharge pulse of the signal, t is the current moment of the discharge pulse of the signal, and f c is the oscillation frequency of the signal.

[0071] Furthermore, obtaining the partial discharge test signal through building a partial discharge test platform for oil-paper insulation under high-frequency pulses for testing includes:

[0072] Using a high-frequency pulse power supply to simulate a high-frequency square wave pulse voltage, a high-voltage probe to measure the voltage test signal, and an ultra-high-frequency current sensor to collect the partial current signal, and inputting the partial discharge current signal and the voltage test signal into a high-frequency oscilloscope to build the test platform;

[0073] Before the test, detect the background noise generated by the high-frequency power supply, and perform Fourier analysis on the background noise to obtain the noise concentration frequency band of the background noise;

[0074] In the test platform, use a needle-plate electrode model to simulate partial discharge defects according to the noise concentration frequency band, pre-treat the insulating paper specimen and place it on the test platform, and use the step-up voltage method to conduct a partial discharge test of oil-paper insulation under high-frequency pulses to obtain the partial discharge test signal.

[0075] Furthermore, import the partial discharge signal as an input into the fishing optimization algorithm, and use the variational mode decomposition algorithm to solve it. The output optimal parameters include:

[0076] Import the partial discharge signal as an input into the fishing optimization algorithm, and set the initial parameters of the fishing optimization algorithm;

[0077] Set the number of decomposition modes K and the penalty factor α in the variational mode decomposition algorithm as the optimization objectives, set the minimum envelope entropy as the fitness function of the fishing optimization algorithm, and perform the solution to obtain the optimal parameters of the variational mode decomposition.

[0078] Furthermore, the minimum envelope entropy is:

[0079] Fitness=minc(Ep i );

[0080]

[0081] Among them, Fitness is the expression of the minimum envelope entropy, and Ep iis the envelope entropy of each IMF component obtained after the partial discharge signal of oil-paper insulation is decomposed by the variational mode decomposition algorithm. a(j) is the envelope signal obtained after the IMF component undergoes the Hilbert transform, and p j is the normalized form of a(j), and N is the number of envelope signals.

[0082] Furthermore, the method for calculating the kurtosis value of each of the IMF components is as follows:

[0083]

[0084] where Q is the kurtosis value of the IMF component, μ is the arithmetic mean of the signal amplitude, E(x - μ) 4 is the fourth-order mathematical expectation, and σ is the standard deviation.

[0085] Example 2:

[0086] The following combines Figures 2 - 7 , taking the case of a partial discharge simulation signal as the partial discharge signal, to specifically illustrate this method, including the following steps:

[0087] Step 1, simulate the partial discharge simulation signal model of oil-paper insulation under high-frequency conditions;

[0088] 1.1) Construct a single exponential decay model:

[0089]

[0090] 1.2) Construct a double exponential decay oscillation model:

[0091]

[0092] where f(t) is the expression of the single exponential decay model, X(t) is the expression of the double exponential decay oscillation model, A and B are the amplitudes of the partial discharge signal of oil-paper insulation under high-frequency conditions, τ is the decay constant of the signal, t0 is the starting moment of the discharge pulse of the signal, t is the current moment of the discharge pulse of the signal, and f c is the oscillation frequency of the signal.

[0093] 1.3) Set the amplitude, decay constant, starting moment of the discharge pulse, and oscillation frequency of the simulated signal;

[0094] 1.4) Add periodic narrowband interference and white noise to generate a simulated signal of partial discharge of oil-paper insulation, as Figure 2 shown.

[0095] Step 2: Set the initial parameters of the CFOA optimization algorithm. Set the number of decomposition modes K and the penalty factor α in the VMD decomposition algorithm as the optimization objectives, and set the minimum envelope entropy as the fitness function of the CFOA optimization algorithm. Import the partial discharge signal of oil-paper insulation in Step 1 into the CFOA optimization algorithm, and according to the fitness function, output the optimal parameter combination of variational mode decomposition, namely the number of decomposition modes K and the penalty factor α.

[0096] 2.1) Set the search ranges of the number of decomposition modes K and the penalty factor α in the VMD decomposition algorithm to [3, 15] and [1000, 3000] respectively;

[0097] 2.2) Set the minimum envelope entropy as the fitness function. The formula for the minimum envelope entropy is as follows:

[0098] Fitness=min(Ep i );

[0099] where Fitness is the expression of the minimum envelope entropy, and Ep i is the envelope entropy of each IMF component obtained after the partial discharge signal of oil-paper insulation is decomposed by the VMD algorithm. The envelope entropy of the IMF component is expressed as:

[0100]

[0101] where a(j) is the envelope signal obtained after the IMF component undergoes Hilbert transform; p j is the normalized form of a(j), and N is the number of envelope signals;

[0102] 2.3) After optimization by the CFOA algorithm, when the number of iterations reaches 28 for the partial discharge simulation signal, the fitness function is the lowest, as Figure 3 shown. Output the parameters: the decomposition layer number K = 11, and the penalty factor α = 1479.8583.

[0103] Step 3: According to the output optimal parameter combination, namely the number of decomposition modes K and the penalty factor α, set the parameters of VMD, and import the partial discharge signal in Step 1 into the VMD algorithm for decomposition to obtain multiple IMF components;

[0104] Set the VMD parameters according to the parameters optimized by CFOA in Step 2: the decomposition layer number K = 11 and the penalty factor α = 1479.8583, and decompose the simulated partial discharge signal of oil-paper insulation in Step 1 to obtain 11 IMF components, as Figure 4 shown; then calculate the spectrogram of each IMF component, as Figure 5 shown.

[0105] Step 4: For each IMF component obtained in Step 3, calculate the kurtosis value of each IMF component to determine the effective signal component and the noise component;

[0106] 4.1) Calculate the kurtosis value of the IMF component. The calculation method is as follows:

[0107]

[0108] In the formula, Q is the kurtosis value of the IMF component, μ is the arithmetic mean of the signal amplitude, E(x - μ) 4 is the fourth-order mathematical expectation, and σ is the standard deviation;

[0109] According to the calculation, the kurtosis values from IMF1 to IMF11 are: 1.7444, 1.7468, 1.8697, 89.4797, 53.7964, 88.6244, 7.2983, 2.9663, 3.2996, 2.9004, 2.9124.

[0110] 4.2) Since the partial discharge signal is a kind of instantaneous mutation signal, its kurtosis value is often much greater than 3. Therefore, the IMF components with kurtosis values greater than 3 are regarded as effective signal components, and the IMF components with kurtosis values less than 3 are regarded as noise components;

[0111] Specifically, IMF4, IMF5, IMF6, IMF7, IMF9 are effective signal components, and IMF1, IMF2, IMF3, IMF8, IMF10, IMF11 are noise components.

[0112] Step 5: Use the multi-scale adaptive wavelet threshold function to perform wavelet threshold denoising on the effective signal components screened in Step 4;

[0113] 5.1) Perform wavelet decomposition on the effective signal components screened in Step 4, select the wavelet basis as db6 and the decomposition level as 4;

[0114] 5.2) Calculate the multi-scale adaptive threshold for each layer and select soft threshold processing;

[0115] 5.3) Threshold function selection:

[0116]

[0117] where σ is the noise standard deviation estimated based on the detail coefficients, and N is the signal length;

[0118] 5.4) Reconstruct the signal after threshold processing.

[0119] Step 6: Reconstruct the effective components after multi-scale adaptive threshold processing in Step 5 to obtain the denoised high-frequency working condition oil-paper insulation partial discharge signal, asFigure 6 as shown; the spectrograms of the oil-paper insulation analog signal and the denoised signal are as Figure 7 shown.

[0120] Example 3:

[0121] Next, in combination with Figures 8 - 14 , taking the partial discharge test signal as the partial discharge signal, this method will be specifically described.

[0122] Step 1: Build a partial discharge test platform for oil-paper insulation under high-frequency pulses to obtain partial discharge signals;

[0123] 1.1) To simulate the high-frequency square-wave pulse voltage borne by oil-paper insulation under high-frequency working conditions, a high-frequency pulse power supply is selected, which can output a bipolar high-frequency square-wave pulse voltage signal with a peak-to-peak value of 0 - 40 kV and a frequency of 0 - 20 kHz;

[0124] 1.2) A high-voltage probe measures the output voltage waveform, with a measurable voltage range of 0 - 40 kV, a frequency bandwidth of 20 MHz, and an attenuation ratio of 5000:1;

[0125] 1.3) The ultra-high-frequency current sensor method is used to collect the current signals of partial discharge. The HFCT used has a power frequency bandwidth of 10 - 500 MHz, which is actually extended to 1000 MHz during the actual test;

[0126] 1.4) The collected high-frequency discharge signals and voltage test signals are sent to a high-frequency oscilloscope with a bandwidth of 4 GHz and a maximum sampling rate of 20 GS / s for real-time observation and recording of voltage values;

[0127] 1.5) Detect the background noise of the high-frequency power supply. The detection of the background noise of the high-frequency power supply is as Figure 8 shown;

[0128] 1.6) Perform Fourier transform on the background noise to analyze the main noise regions in the background noise. The main interferences in the background noise are mainly concentrated around 23.5 MHz, 47.5 MHz, and 71 MHz, and there will always be noises with smaller amplitudes in the remaining frequency ranges;

[0129] 1.7) Select a needle-plate electrode model, and simulate partial discharge defects according to the concentrated frequency bands of background noise interference in 1.6). The needle electrode uses a medical No. 6 needle, the diameter of the plate electrode is 75 mm, and the gap between the needle and the insulating paper is 0 mm;

[0130] 1.8) Pretreatment of the specimen: Remove moisture, air, impurities, etc. from the oil using an oil filter; Place the cut insulating paper in a vacuum drying oven, heat it to 105 °C, and keep it for 48 h; Keep it for 24 h under the conditions of 80 Pa and 85 °C; Immerse the dried insulating paper in the filtered oil and keep it for 48 h under the conditions of 80 Pa and 85 °C;

[0131] 1.9) Adopt the stepped voltage boosting method to conduct the partial discharge test of oil-paper insulation under high-frequency pulses. The partial discharge signals of oil-paper insulation collected are as Figure 8 shown.

[0132] Step 2: Set the initial parameters of the CFOA optimization algorithm. Set the number of decomposition modes K and the penalty factor α in the VMD decomposition algorithm as the optimization objectives, and set the minimum envelope entropy as the fitness function of the CFOA optimization algorithm. Import the partial discharge signals of oil-paper insulation in Step 1 into the CFOA optimization algorithm respectively, and output the optimal parameter combination of variational mode decomposition, namely the number of decomposition modes K and the penalty factor α, according to the fitness function;

[0133] 2.1) Set the search ranges of the number of decomposition modes K and the penalty factor α in the VMD decomposition algorithm to [3, 15] and [1000, 3000] respectively;

[0134] 2.2) Set the minimum envelope entropy as the fitness function. The formula for the minimum envelope entropy is as follows:

[0135] Fitness = min(Ep i );

[0136] where Fitness is the expression of the minimum envelope entropy, and Ep i is the envelope entropy of each IMF component obtained after the partial discharge signal of oil-paper insulation is decomposed by the VMD algorithm. The envelope entropy of the IMF component is expressed as:

[0137]

[0138] where a(j) is the envelope signal obtained after the IMF component undergoes the Hilbert transform; p j is the normalized form of a(j), and N is the number of envelope signals;

[0139] 2.3) After being optimized by the CFOA algorithm, when the number of iterations reaches 53 for the partial discharge simulation signal, the fitness function is the lowest, as Figure 9 shown. Output the parameter decomposition layer number K = 12 and the penalty factor α = 707.5134.

[0140] Step 3: Decompose the mode number K and the penalty factor α according to the output optimal parameter combination, set the parameters of VMD, and import the partial discharge signal in Step 1 into the VMD algorithm for decomposition to obtain multiple IMF components;

[0141] Decompose the number of layers K = 12 and the penalty factor α = 707.5134 of VMD according to the parameters optimized by CFOA in Step 2, set the VMD parameters, and decompose the simulated signal of partial discharge in oil-paper insulation simulated in Step 1 to obtain 12 IMF components, as Figure 10 shown; then calculate the spectrogram of each IMF component, as Figure 11 shown.

[0142] Step 4: For each IMF component obtained in Step 3, calculate the kurtosis value of each IMF component to judge the effective signal component and the noise component;

[0143] 4.1) Calculate the kurtosis value of the IMF component, and the calculation method is as follows:

[0144]

[0145] In the formula, Q is the kurtosis value of the IMF component, μ is the arithmetic mean of the signal amplitude, and E(x - μ) 4 is the fourth-order mathematical expectation, and σ is the standard deviation;

[0146] According to the calculation, the kurtosis values from IMF1 to IMF12 are: 26.2023, 185.5831, 107.5909, 132.1940, 124.7973, 111.5784, 134.4352, 120.4346, 187.2896, 195.4728, 159.5737, 165.1486.

[0147] 4.2) Since the partial discharge signal is a kind of instantaneous mutation signal, its kurtosis value is often much greater than 3. Therefore, the IMF components with kurtosis values greater than 3 are regarded as effective signal components, and the IMF components with kurtosis values less than 3 are regarded as noise components.

[0148] Step 5: Use the multi-scale adaptive wavelet threshold function to perform wavelet threshold denoising on the effective signal components screened in Step 5;

[0149] 5.1) Perform wavelet decomposition on the effective signal components screened in Step 4, select the wavelet basis as db6 and the decomposition layer number as 4;

[0150] 5.2) Perform multi-scale adaptive threshold calculation on each layer and select soft threshold processing;

[0151] 5.3) Threshold function selection:

[0152]

[0153] Among them, σ is the standard deviation of the noise estimated based on the detail coefficient, and N is the signal length;

[0154] 5.4) Reconstruct the signal after threshold processing.

[0155] Step 6, reconstruct the effective components after multi-scale adaptive threshold processing in Step 5 to obtain the denoised high-frequency operating condition oil-paper insulation partial discharge signal, as Figure 12 shown; the spectrograms of the oil-paper insulation simulation signal and the denoised signal are as Figure 13 shown.

[0156] Comparison and verification:

[0157] For the processing of the oil-paper insulation partial discharge simulation signal, this method is compared with various methods, as shown in Table 1.

[0158] Table 1

[0159]

[0160] Through the comparison of the results, it shows that this method has a higher signal-to-noise ratio, a lower average waveform distortion error, and a waveform similarity coefficient closer to 1 compared with the previous methods, and the denoising effect is better; and when using GWO for optimization, the optimization time is up to more than 500 seconds, while the CFOA in this method only takes more than 30 seconds for optimization, greatly shortening the time.

[0161] For the actually measured partial discharge signal, since the partial discharge signal data without noise cannot be obtained, the denoising effect cannot be quantitatively evaluated by calculating the signal-to-noise ratio, average waveform distortion error, and waveform similarity coefficient. The method of calculating the noise suppression ratio is used to quantitatively evaluate the denoising effect, and the noise suppression ratio for the actually measured signal is 0.16023 dB. Through the comparison, it shows that this method has a higher signal-to-noise ratio, a lower average waveform distortion error, and a waveform similarity coefficient closer to 1 compared with the previous methods, and the denoising effect is better.

[0162] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention should fall within the protection scope determined by the claims of the present invention.

Claims

1. A denoising method for high-frequency condition partial discharge signals based on CFOA-VMD combined with multi-scale wavelet threshold, characterized in that Including: Obtain the partial discharge signal of oil-paper insulation, import the partial discharge signal as an input into the fish swarm optimization algorithm, and use the variational mode decomposition algorithm to solve it, outputting the optimal parameters; Decompose the partial discharge signal based on the optimal parameters to obtain IMF components; Calculate the kurtosis value of each IMF component, and screen out the effective signal components based on the kurtosis value; Perform multi-scale adaptive wavelet threshold processing on the effective signal components; Perform wavelet reconstruction on the processed effective signal components to obtain the denoised partial discharge signal.

2. The high-frequency operating condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold according to claim 1, wherein The partial discharge signal includes any one of a partial discharge simulation signal and a partial discharge test signal. Among them, the partial discharge simulation signal is obtained by simulating through constructing a high-frequency condition oil-paper insulation partial discharge simulation signal model, and the partial discharge test signal is obtained by conducting tests on a high-frequency pulse oil-paper insulation partial discharge test platform.

3. The high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold according to claim 2, wherein Obtaining the partial discharge simulation signal by simulating through constructing a high-frequency condition oil-paper insulation partial discharge simulation signal model includes: Construct a single exponential decay model and a double exponential decay oscillation model; Set the parameters of the simulation signal, and add periodic narrowband interference and white noise, and simulate based on the single exponential decay model and the double exponential decay oscillation model to obtain the partial discharge simulation signal of oil-paper insulation.

4. The high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold according to claim 3, characterized in that, The single exponential decay model is: The double exponential decay oscillation model is: Among them, f(t) is the expression of the single exponential decay model, X(t) is the expression of the double exponential decay oscillation model, A and B are the amplitudes of the partial discharge signals of oil-paper insulation under high-frequency conditions, τ is the decay constant of the signal, t0 is the starting time of the discharge pulse of the signal, t is the current time of the discharge pulse of the signal, and f c is the oscillation frequency of the signal.

5. The high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold according to claim 2, wherein Obtaining the partial discharge test signal by conducting tests on a high-frequency pulse oil-paper insulation partial discharge test platform includes: Use a high-frequency pulse power supply to simulate a high-frequency square wave pulse voltage, a high-voltage probe to measure the voltage test signal, and an ultra-high frequency current sensor to collect the local current signal, and input the partial discharge current signal and the voltage test signal into a high-frequency oscilloscope to build the test platform; Before the test, detect the background noise generated by the high-frequency power supply, and perform Fourier analysis on the background noise to obtain the noise concentration frequency band of the background noise; In the test platform, use a needle-plate electrode model to simulate partial discharge defects according to the noise concentration frequency band, pre-treat the insulating paper specimen and place it on the test platform, and use the step-up voltage method to conduct a high-frequency pulse oil-paper insulation partial discharge test to obtain the partial discharge test signal.

6. The high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold according to claim 1, characterized in that, Importing the partial discharge signal as an input into the fish swarm optimization algorithm and using the variational mode decomposition algorithm to solve and output the optimal parameters includes: Import the partial discharge signal as an input into the fish swarm optimization algorithm, and set the initial parameters of the fish swarm optimization algorithm; Set the decomposition mode number K and the penalty factor α in the variational mode decomposition algorithm as the optimization objectives, set the minimum envelope entropy as the fitness function of the fish swarm optimization algorithm, and solve to obtain the optimal parameters of the variational mode decomposition.

7. The high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold according to claim 6, characterized in that The minimum envelope entropy is: Fitness=min(Ep i ); Among them, Fitness is the minimum envelope entropy expression, and Ep i is the envelope entropy of each IMF component obtained after the partial discharge signal of oil-paper insulation is decomposed by the variational mode decomposition algorithm. a(j) is the envelope signal obtained after the IMF component undergoes the Hilbert transform, and p j is the normalized form of a(j), and N is the number of envelope signals.

8. The high-frequency condition partial discharge signal denoising method based on CFOA-VMD combined with multi-scale wavelet threshold according to claim 1, wherein The method for calculating the kurtosis value of each IMF component is: Among them, Q is the kurtosis value of the IMF component, μ is the arithmetic mean of the signal amplitude, and E(x - μ) 4 is the fourth-order mathematical expectation, and σ is the standard deviation.

Images