GIS partial discharge positioning method based on multichannel signal processing

Abstract

The invention provides a GIS partial discharge positioning method based on multichannel signal processing, and relates to the field of fault diagnosis of high-voltage electrical equipment, and the method comprises the steps: S1, collecting multichannel electromagnetic wave signals generated by partial discharge through a plurality of ultrahigh frequency sensors distributed in GIS equipment; s2, preprocessing the collected electromagnetic wave signal, including noise suppression and time difference calculation; s3, based on the time difference and amplitude information of the multi-channel electromagnetic wave signals, calculating the position of a partial discharge source by adopting a fusion algorithm and combining a double-end positioning formula; and S4, displaying the position and the discharge intensity of the partial discharge source through the monitoring host, and generating a related diagnosis report. According to the invention, precise positioning of the partial discharge source in the GIS equipment can be realized, so that the positioning precision and the anti-interference capability are improved, and safe operation and maintenance of the GIS equipment are ensured.

Description

Technical Field

[0001] The invention relates to the field of fault diagnosis of high-voltage electrical equipment, and in particular to a GIS partial discharge positioning method based on multi-channel signal processing. Background Art

[0002] GIS (gas insulated switchgear) is widely used in high-voltage power transmission and distribution systems due to its advantages of small footprint and high operational reliability, especially in scenarios such as urban substations and important load centers that require high input power and high stability. However, during the long-term operation of GIS equipment, due to the combined effects of various factors such as electricity, heat, machinery and environment, the insulation material will gradually age or be damaged, leading to the occurrence of partial discharge (PD). Partial discharge is an early manifestation of insulation failure inside GIS equipment. Its continued development will further expand local insulation defects such as dendrites or carbonization, which may eventually cause serious insulation breakdown or even overall equipment failure. Therefore, the effective detection and positioning of partial discharge is a key technology to ensure the safe operation of GIS equipment and extend its service life.

[0003] Partial discharge generates ultra-high frequency (UHF) electromagnetic wave signals, which can be collected and analyzed by sensors. However, due to the complex electromagnetic environment inside GIS equipment, partial discharge signals are easily interfered by noise, making signal recognition difficult. In addition, existing partial discharge positioning methods are insufficient in accuracy and anti-interference ability, making it difficult to meet the operating requirements of high-voltage GIS equipment. Therefore, how to improve the recognition and positioning accuracy of partial discharge signals is a technical problem that needs to be solved urgently. Summary of the invention

[0004] In view of this, the present invention proposes a GIS local discharge positioning method based on multi-channel signal processing, which collects electromagnetic wave signals generated by local discharge through multiple UHF sensors distributed inside the GIS equipment, uses an improved signal processing algorithm to perform noise suppression and time difference calculation, combines the time difference and amplitude information of the multi-channel signal, adopts a fusion algorithm and a double-end positioning formula, and realizes the precise positioning of the local discharge source inside the GIS equipment, thereby improving the positioning accuracy and anti-interference ability, and ensuring the safe operation and maintenance of the GIS equipment.

[0005] The technical solution of the present invention is implemented as follows: The present invention provides a GIS partial discharge positioning method based on multi-channel signal processing, comprising:

[0006] S1, collect multi-channel electromagnetic wave signals generated by partial discharge through multiple UHF sensors distributed inside the GIS equipment;

[0007] S2, preprocessing the collected electromagnetic wave signals, including noise suppression and time difference calculation;

[0008] S3, based on the time difference and amplitude information of multi-channel electromagnetic wave signals, a fusion algorithm is used in combination with a double-end positioning formula to calculate the location of the local discharge source;

[0009] S4. The location and discharge intensity of the local discharge source are displayed on the monitoring host, and a relevant diagnostic report is generated.

[0010] On the basis of the above technical solution, preferably, in step S2, noise suppression adopts a wavelet denoising algorithm based on a feature library, and the process is as follows:

[0011] Perform wavelet decomposition on the collected electromagnetic wave signals to obtain multi-scale detail coefficients and approximate coefficients;

[0012] According to the feature library of partial discharge signals, the threshold of wavelet decomposition is dynamically adjusted. The threshold calculation formula is:

[0013]

[0014] In the formula, λ a is the threshold of the a-th layer, σ a is the standard deviation of the detail coefficient of the ath layer, L is the signal length; K a is the adjustment factor; SNR a is the signal-to-noise ratio of the detail coefficient of the a-th layer, Among them, s a is the signal component, n a is the noise component, E[·] represents the mathematical expectation; Kurt a is the kurtosis of the detail coefficient of the a-th layer, Among them, c a is the detail coefficient of layer a, μ a is the mean value of the detail coefficient of layer a;

[0015] The detail coefficient is processed by soft threshold, and the processed detail coefficient d a The calculation formula is:

[0016]

[0017] In the formula, sgn(c a ) is a symbolic function;

[0018] The processed detail coefficients and approximate coefficients are used for wavelet reconstruction to obtain the denoised signal.

[0019] On the basis of the above technical solution, preferably, the feature library of the partial discharge signal is constructed by the following steps:

[0020] Select clear signal samples from the partial discharge history data recorded by GIS equipment. The signal samples include a variety of typical partial discharge types.

[0021] Perform time-frequency transformation on signal samples and extract time-frequency domain features, including amplitude, spectrum distribution and main frequency band power; further extract nonlinear features of signal samples, including fractal dimension, sample entropy and multi-scale entropy;

[0022] Normalize the extracted signal features to eliminate differences in amplitude and spectrum range;

[0023] Perform multimodal weighted fusion on the extracted time-frequency domain features and nonlinear features to generate a multimodal feature vector;

[0024] Based on the constructed multimodal feature vector, the self-attention convolutional neural network is used to perform deep feature extraction;

[0025] The density-based dynamic clustering algorithm DBSCAN-ADV is used to cluster the depth features. According to the density distribution of frequency bands and amplitudes, the depth features are classified and the noise points with low density are removed. DBSCAN-ADV introduces a dynamic neighborhood radius adjustment mechanism and sets the adaptive neighborhood parameter ∈ according to the frequency band characteristics of GIS partial discharge signals:

[0026]

[0027] In the formula, F c is the multimodal feature vector, M is the number of collected signal samples;

[0028] The classified feature data are stored as standard feature templates to form a feature library of partial discharge signals.

[0029] On the basis of the above technical solution, preferably, during the operation of the GIS device, the new signal collected in real time is compared with the standard feature template in the feature library. If the feature of the new signal exceeds the template error threshold, the feature library is automatically updated to generate a new standard feature template.

[0030] On the basis of the above technical solution, preferably, in step S2, the time difference calculation is performed based on the electromagnetic wave signal after noise suppression processing, and the steps are as follows:

[0031] The N UHF sensors installed in the GIS equipment are numbered {C1, C2, ..., C N}, N UHF sensors are combined in pairs to form sensor pairs, forming a total of For sensor pairs;

[0032] The cross-correlation function is used to calculate each pair of sensors (C i ,Cj )The received signal (x i ,y j ) Similarity under different time delays, the cross-correlation function calculation formula is as follows:

[0033]

[0034] In the formula, is the signal x when the time delay is τ i and i similarity; [0,T] represents the time range of signal acquisition; x i (t) and y j (t) are the sensor pairs (C i ,C j ) received original time domain signal; X i (f) is x i Fourier transform of (t); for y j The complex conjugate of the Fourier transform of (t); For x i (t) and y j The phase difference function of (t+τ) represents the phase difference between the two signals under the time delay τ; w1(t) is the amplitude weight function; w2(f) is the frequency weight function; w3(t) is the phase weight function;

[0035] The time delay corresponding to the maximum similarity value is taken as the time difference of the pair of sensors.

[0036] Based on the above technical solution, preferably, step S3 includes:

[0037] S31, for each pair of sensors (C i ,C j ), extract the time difference Δt ij , amplitude characteristics M ij and arrival angle A ij , where the amplitude characteristic M ij By pairing the sensor (C i ,C j ) is normalized and the amplitude average is calculated to obtain the arrival angle A ij Based on the phase difference and spatial geometric relationship of the signal, the signal arrival sensor pair (C i ,C j ) to obtain the angle information;

[0038] S32, according to the sensor pair (C i ,C j ) Extracted time difference Δt ij , amplitude characteristics Mij and arrival angle A ij , calculate the comprehensive characteristic weight ω ij ;

[0039] S33, calculate the sensor pair (C based on the double-end positioning formula i ,C j )’s single pair positioning result L M,ij ;

[0040] S34, repeating steps S31-S33, calculating the comprehensive characteristic weights of all sensor pairs and the positioning results of a single pair;

[0041] S35. Perform weighted fusion on all single-pair positioning results using comprehensive characteristic weights to calculate the final location of the local discharge source.

[0042] Based on the above technical solution, preferably, in step S32, the comprehensive characteristic weight ω ij The calculation formula is:

[0043]

[0044] In the formula, Δt ij For the sensor pair (C i ,C j ) time difference; is the average time difference of all sensor pairs; T s is the standardized parameter of time difference; M ij For the sensor pair (C i ,C j )’s amplitude characteristics; σ M,ij For the sensor pair (C i ,C j ) is the amplitude standard deviation; σ0 is the reference value of the amplitude standard deviation; A ij For the sensor pair (C i ,C j )’s arrival angle; k represents the kth sensor pair; is the number of sensor pairs; α, β, γ are weight coefficients, and α+β+γ=1.

[0045] Based on the above technical solution, preferably, a single pair of positioning results L M,ij The calculation formula is:

[0046]

[0047] Where, L ij For the sensor pair (C i ,C j ) distance; V is the propagation speed of electromagnetic waves in GIS;

[0048] The final calculation formula for the location of the partial discharge source is:

[0049]

[0050] Where, L M is the location of the final partial discharge source.

[0051] On the basis of the above technical solution, preferably, the UHF sensor includes:

[0052] Built-in sensor, installed inside the GIS tank, adopts a disc sensor structure;

[0053] The external sensor is installed on the pot-type insulator outside the GIS tank to receive the electromagnetic wave signal radiated from the inside of the GIS to the outside;

[0054] An environmental noise sensor, used for collecting environmental noise signals;

[0055] The frequency range of the UHF sensor is 300MHz to 1500MHz, and each sensor is equipped with an independent acquisition channel, using a 50MHz sampling rate and a 12-bit ADC.

[0056] On the basis of the above technical solution, preferably, in step S4, the monitoring host generates a PRPS three-dimensional map and a PRPD two-dimensional map in real time to display the phase, amplitude and frequency characteristics of the local discharge signal; records the occurrence time, location and intensity of the local discharge event, and forms trend analysis and multi-level alarm information to generate relevant diagnostic reports; has three synchronization modes: built-in PT synchronization, AC220V synchronization and external FM synchronization, with a frequency measurement accuracy of 0.02Hz and a frequency range of 49.0-51.0Hz; and transmits monitoring data and diagnostic reports to the server via optical fiber communication.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] (1) The present invention provides a GIS partial discharge positioning method based on multi-channel signal processing. The electromagnetic wave signals generated by partial discharge are collected by multiple ultra-high frequency sensors distributed inside the GIS equipment. The improved signal processing algorithm is used to suppress noise and calculate the time difference. The time difference and amplitude information of the multi-channel signals are combined, and the fusion algorithm and the double-end positioning formula are used to achieve accurate positioning of the partial discharge source inside the GIS equipment, thereby significantly improving the positioning accuracy and anti-interference ability, and ensuring the safe operation and maintenance of the GIS equipment.

[0059] (2) By using a wavelet denoising algorithm based on a feature library, the collected electromagnetic wave signal is decomposed and reconstructed by wavelet, which effectively suppresses noise interference and improves the recognition accuracy of partial discharge signals. The algorithm dynamically adjusts the threshold of wavelet decomposition, which can adapt to the noise level in different signal environments and ensure the stability and reliability of signal processing;

[0060] (3) Construct a feature library of partial discharge signals, and achieve deep feature extraction and classification of partial discharge signals through time-frequency transformation and nonlinear feature extraction, combined with self-attention convolutional neural network and dynamic clustering algorithm. The feature library can be dynamically updated to ensure accurate recognition and classification of new signals;

[0061] (4) The cross-correlation function is used to calculate the time difference of the sensor pair, and combined with the double-end positioning formula, the location of the local discharge source is accurately calculated. The single-pair positioning results of all sensor pairs are weighted fused by comprehensive characteristic weights, which further improves the accuracy and robustness of the positioning results. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0063] Figure 1 The figure is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] like Figure 1 As shown, the present invention provides a GIS partial discharge positioning method based on multi-channel signal processing, comprising:

[0066] S1, collect multi-channel electromagnetic wave signals generated by partial discharge through multiple UHF sensors distributed inside the GIS equipment;

[0067] S2, preprocessing the collected electromagnetic wave signals, including noise suppression and time difference calculation;

[0068] S3, based on the time difference and amplitude information of multi-channel electromagnetic wave signals, a fusion algorithm is used in combination with a double-end positioning formula to calculate the location of the local discharge source;

[0069] S4. The location and discharge intensity of the local discharge source are displayed on the monitoring host, and a relevant diagnostic report is generated.

[0070] Specifically, in one embodiment of the present invention, the UHF sensor includes:

[0071] Built-in sensor, installed inside the GIS tank, adopts a disc sensor structure;

[0072] The external sensor is installed on the pot-type insulator outside the GIS tank to receive the electromagnetic wave signal radiated from the inside of the GIS to the outside;

[0073] An environmental noise sensor, used for collecting environmental noise signals;

[0074] The frequency range of the UHF sensor is 300MHz to 1500MHz, and each sensor is equipped with an independent acquisition channel, using a 50MHz sampling rate and a 12-bit ADC.

[0075] In this embodiment, the UHF sensors are divided into three types: UHF built-in sensors, UHF external sensors, and environmental noise sensors. The built-in sensors are installed at key locations inside the GIS tank, usually close to areas where partial discharge may occur, such as insulators or conductor connections. This sensor adopts a disc structure and can efficiently capture high-frequency electromagnetic wave signals generated by partial discharge. The external sensor is installed at the pot-type insulator outside the GIS tank to receive electromagnetic wave signals radiated from the inside of the GIS. The environmental noise sensor is arranged around the equipment to collect external environmental noise signals.

[0076] During the signal acquisition process, each UHF sensor is equipped with an independent acquisition channel. The design of these acquisition channels ensures the integrity and accuracy of the signal. The frequency range of the sensor is 300MHz to 1500MHz, which can cover the main frequency band of partial discharge signals. The sampling rate is set to 50MHz, combined with a 12-bit ADC (analog-to-digital converter), which can accurately capture the amplitude and time characteristics of the signal.

[0077] During the acquisition process, the UHF sensor converts the high-frequency electromagnetic wave signal generated by partial discharge into an electrical signal, and performs preliminary processing through the signal conditioning module. This includes signal amplification, filtering and anti-interference processing to ensure signal quality. At the same time, the noise signal collected by the environmental noise sensor is also recorded synchronously.

[0078] Specifically, in one embodiment of the present invention, the preprocessing process of step S2 includes noise suppression and time difference calculation.

[0079] Noise suppression is the first step in signal preprocessing, and its purpose is to extract the effective signal of partial discharge from the collected multi-channel electromagnetic wave signals while minimizing the influence of environmental noise and other interference signals.

[0080] In this embodiment, noise suppression adopts a wavelet denoising algorithm based on a feature library, and the process is as follows:

[0081] The collected electromagnetic wave signal is subjected to wavelet decomposition to obtain multi-scale detail coefficients and approximate coefficients; wavelet decomposition can decompose the signal into components in different frequency ranges, thereby effectively separating the partial discharge signal and the noise signal.

[0082] According to the feature library of partial discharge signals, the threshold of wavelet decomposition is dynamically adjusted. The threshold calculation formula is:

[0083]

[0084] In the formula, λ a is the threshold of the a-th layer, σ a is the standard deviation of the detail coefficient of the ath layer, L is the signal length; K a is the adjustment factor; SNR a is the signal-to-noise ratio of the detail coefficient of the a-th layer, Among them, s a is the signal component, n a is the noise component, E[·] represents the mathematical expectation; Kurt a is the kurtosis of the detail coefficient of the a-th layer, Among them, c a is the detail coefficient of layer a, μ a is the mean value of the detail coefficient of layer a;

[0085] The detail coefficient is processed by soft threshold, and the processed detail coefficient d a The calculation formula is:

[0086]

[0087] In the formula, sgn(c a ) is a symbolic function;

[0088] The processed detail coefficients and approximate coefficients are used for wavelet reconstruction to obtain the denoised signal.

[0089] In this embodiment, considering the non-stationarity and impulse of GIS partial discharge signal, the signal-to-noise ratio and signal kurtosis are introduced to dynamically adjust the threshold. The GIS partial discharge signal may have different signal-to-noise ratios at different scales. By introducing the signal-to-noise ratio, the threshold can be adjusted more specifically, and a higher threshold can be used at a scale with a higher signal-to-noise ratio to better retain signal details. Kurtosis reflects the sharpness of the signal waveform. Partial discharge signals usually show the characteristics of strong impulse and high kurtosis, while the kurtosis of noise is relatively low. Kurtosis can effectively distinguish between signals and noise and improve the accuracy of threshold calculation. By setting the adjustment coefficient, the threshold can be appropriately fine-tuned according to actual application requirements.

[0090] The feature library of partial discharge signals is constructed by the following steps:

[0091] Select clear signal samples from the partial discharge history data recorded by GIS equipment. The signal samples include a variety of typical partial discharge types, including internal discharge, surface discharge, suspended discharge, etc. Select signals with high signal-to-noise ratio and less interference to ensure the accuracy of the samples. The samples should be able to reflect the partial discharge characteristics of GIS equipment under different operating conditions.

[0092] Perform time-frequency transformation on signal samples and extract time-frequency domain features, including amplitude, spectrum distribution and main frequency band power; further extract nonlinear characteristics of signal samples, including fractal dimension, sample entropy and multi-scale entropy.

[0093] The process of time-frequency domain feature extraction is: perform short-time Fourier transform on the signal, generate a time-frequency spectrum, extract the signal's spectrum distribution and main frequency band power. Calculate the signal's amplitude characteristics, including peak amplitude, mean amplitude, and amplitude change rate. Extract the signal's main frequency, bandwidth, and frequency center. Calculate the signal's energy distribution in different frequency bands and identify the main frequency bands of partial discharge signals. The time-frequency domain features can reflect the spectrum characteristics and time-domain variation patterns of partial discharge signals.

[0094] The process of nonlinear feature extraction is: calculate the fractal dimension of the signal to reflect the geometric complexity of the signal. Calculate the sample entropy of the signal to quantify the randomness and complexity of the signal. Calculate the entropy value of the signal at different time scales to analyze the multi-scale dynamic characteristics of the signal. Nonlinear features can supplement the deficiencies of time-frequency domain features and provide a more comprehensive description of partial discharge signals.

[0095] The extracted signal features are normalized to eliminate the differences in amplitude and spectrum range; the normalization formula is:

[0096]

[0097] In the formula, X is the original eigenvalue, X min , Xmax The normalization process can ensure that different features have the same dimension in the subsequent fusion and classification process, avoiding the influence of the difference in the range of feature values ​​on the results.

[0098] The extracted time-frequency domain features and nonlinear features are multimodally weighted and fused to generate a multimodal feature vector. Weights are assigned according to the importance of the features, for example, higher weights are assigned to the spectrum distribution and the power of the main frequency band. The weighted time-frequency domain features and nonlinear features are concatenated to generate a multimodal feature vector. The multimodal feature vector can comprehensively reflect the multidimensional characteristics of the partial discharge signal and provide input for deep feature extraction.

[0099] Based on the constructed multimodal feature vector, the self-attention convolutional neural network is used to extract deep features; the steps are as follows:

[0100] Input layer: The multimodal feature vector is input into the self-attention convolutional neural network.

[0101] Convolutional layer: extract local features of the signal through multi-layer convolution operations.

[0102] Self-attention mechanism: A self-attention mechanism is introduced in the convolutional layer to capture the long-distance dependencies between features.

[0103] Output layer: Generates a deep feature vector as a high-dimensional feature representation of the signal.

[0104] Deep feature extraction can further explore the potential characteristics of the signal and provide higher accuracy for feature classification.

[0105] The density-based dynamic clustering algorithm DBSCAN-ADV is used to cluster the depth features. According to the density distribution of frequency bands and amplitudes, the depth features are classified and the noise points with low density are removed. DBSCAN-ADV introduces a dynamic neighborhood radius adjustment mechanism and sets the adaptive neighborhood parameter ∈ according to the frequency band characteristics of GIS partial discharge signals:

[0106]

[0107] In the formula, F c is the multimodal feature vector, M is the number of collected signal samples;

[0108] The classified feature data are stored as standard feature templates to form a feature library of partial discharge signals.

[0109] During the operation of GIS equipment, the new signal collected in real time is compared with the standard feature template in the feature library. If the feature of the new signal exceeds the template error threshold, the feature library is automatically updated to generate a new standard feature template. The dynamic update mechanism can ensure the real-time and adaptability of the feature library.

[0110] The time difference calculation is based on the electromagnetic wave signal after noise suppression processing. Its purpose is to provide basic data for locating the local discharge source by analyzing the time difference between signals received by different sensors. The steps are as follows:

[0111] The N UHF sensors installed in the GIS equipment are numbered {C1, C2, ..., C N}, N UHF sensors are combined in pairs to form sensor pairs, forming a total of For sensor pairs;

[0112] The cross-correlation function is used to calculate each pair of sensors (C i ,C j )The received signal (x i ,y j ) Similarity under different time delays, the cross-correlation function calculation formula is as follows:

[0113]

[0114] In the formula, is the signal x when the time delay is τ i and i similarity; [0,T] represents the time range of signal acquisition; x i (t) and y j (t) are the sensor pairs (C i ,C j ) received original time domain signal; X i (f) is x i Fourier transform of (t); for y j The complex conjugate of the Fourier transform of (t); For x i (t) and y j The phase difference function of (t+τ) represents the phase difference between the two signals under the time delay τ; w1(t) is the amplitude weight function; w2(f) is the frequency weight function; w3(t) is the phase weight function;

[0115] The time delay corresponding to the maximum similarity value is taken as the time difference of the pair of sensors.

[0116] In this embodiment, the cross-correlation function is the matching degree of two signals under different time delays. Its goal is to find the time offset with the greatest similarity between the two signals by performing time shift operations on the signals. i (t) and y i (t):

[0117] When x i (t) and y j When (t+τ) reaches the maximum “match” at a delay of τ, it means that the two signals have the highest similarity at this time point.

[0118] Specifically, the amplitude weight function w1(t) is calculated as follows:

[0119]

[0120] The calculation process is: Calculate the signal x i (t) and the delayed signal y j The sum of the amplitudes of (t+τ) is normalized to form an amplitude weight function.

[0121] The frequency weight function w2(f) is calculated based on the power spectral density:

[0122]

[0123] Among them, P xy (f) is the cross power spectrum density of the signal, for P xy (f) Perform normalization processing to form a frequency weight function.

[0124] The phase weight function w3(t) is calculated as:

[0125]

[0126] When the phase difference When the phase difference is small, the cosine value is close to 1 and the weight is large; When it is larger, the cosine value is close to 0 and the weight is smaller.

[0127] Phase difference function The calculation process is:

[0128] For signal x i (t) and y j (t+τ) performs Hilbert transform:

[0129]

[0130] Wherein, H[·] represents the Hilbert transform, and j is the imaginary unit.

[0131] Get the instantaneous phase of a signal:

[0132]

[0133]

[0134] Calculate the phase difference function:

[0135]

[0136] In order to avoid the error caused by phase jump, the phase can be unwrapped to ensure the continuity of the phase difference.

[0137] The calculation process of the cross-correlation function is: first calculate w1(t), w2(f), w3(t); calculate the time domain partial integral Frequency domain partial integration Through Fourier transform, the frequency domain correlation is converted to the time domain, and the phase part is integrated Add the partial results together By looking for The maximum value of the optimal time delay τ max , which is the time difference received by the sensor pair.

[0138] Specifically, in one embodiment of the present invention, step S3 includes:

[0139] S31, for each pair of sensors (C i ,C j ), extract the time difference Δt ij , amplitude characteristics M ij and arrival angle A ij , where the amplitude characteristic M ij By pairing the sensor (C i ,C j ) is normalized and the amplitude average is calculated to obtain the arrival angle A ij Based on the phase difference and spatial geometric relationship of the signal, the signal arrival sensor pair (C i ,C j ) to obtain the angle information;

[0140] S32, according to the sensor pair (C i ,C j ) Extracted time difference Δt ij , amplitude characteristics M ij and arrival angle A ij , calculate the comprehensive characteristic weight ω ij ;

[0141] S33, calculate the sensor pair (C based on the double-end positioning formulai ,C j )’s single pair positioning result L M,ij ;

[0142] S34, repeating steps S31-S33, calculating the comprehensive characteristic weights of all sensor pairs and the positioning results of a single pair;

[0143] S35. Perform weighted fusion on all single-pair positioning results using comprehensive characteristic weights to calculate the final location of the local discharge source.

[0144] In step S32, the comprehensive characteristic weight ω ij The calculation formula is:

[0145]

[0146] In the formula, Δt ij For the sensor pair (C i ,C j ) time difference; is the average time difference of all sensor pairs; T s is the standardized parameter of time difference; M ij For the sensor pair (C i ,C j )’s amplitude characteristics; σ M,ij For the sensor pair (C i ,C j ) is the amplitude standard deviation; σ0 is the reference value of the amplitude standard deviation; A ij For the sensor pair (C i ,C j )’s arrival angle; k represents the kth sensor pair; is the number of sensor pairs; α, β, γ are weight coefficients, and α+β+γ=1.

[0147] Single pair positioning result L M,ij The calculation formula is:

[0148]

[0149] Where, L ij For the sensor pair (C i ,C j ) distance; V is the propagation speed of electromagnetic waves in GIS;

[0150] The final calculation formula for the location of the partial discharge source is:

[0151]

[0152] Where, L M is the location of the final partial discharge source.

[0153] In this embodiment, the process of step S3 is specifically as follows:

[0154] 1. Extract time difference, amplitude characteristics and arrival angle

[0155] Time difference Δt ij Using the cross-correlation function Calculate the time delay of a signal, The time delay corresponding to the maximum value of is recorded as the time difference Δt ij .

[0156] For sensor pair (C i ,C j ) is normalized and the amplitude average is calculated:

[0157]

[0158] Among them, M ij Characterizes the energy strength of the signal. The larger the amplitude, the stronger the signal. Through normalization processing, the amplitude differences between different sensor pairs are eliminated.

[0159] Based on the phase difference of the signal and the spatial geometric relationship of the sensor, the signal arrival at the sensor pair (C i ,C j ) angle:

[0160]

[0161] Where V is the propagation speed of electromagnetic waves in GIS, D ij For the sensor pair (C i ,C j ). The angle information is used to evaluate whether the signals come from the same discharge source.

[0162] 2. Calculate the comprehensive feature weight

[0163] Calculate the degree of deviation of the time difference:

[0164]

[0165] Normalize it:

[0166]

[0167] The time difference reflects the delay of signal propagation and is the core parameter of positioning. To measure the credibility of the time difference, the smaller the deviation, the greater the weight.

[0168] Compute the normalized value of the magnitude characteristic:

[0169]

[0170] Combined with the effect of amplitude standard deviation:

[0171]

[0172] The amplitude characteristic reflects the energy strength of the signal. The larger the amplitude, the more reliable the signal. M,ij To evaluate the stability of the amplitude, the smaller the standard deviation, the greater the weight.

[0173] Calculate the cosine of an angle:

[0174] cos(A ij ) Normalization processing:

[0175]

[0176] The arrival angle reflects the spatial propagation direction of the signal. The closer the angles are, the more likely the signals are from the same source. ij ) to measure the credibility of the angle.

[0177] This embodiment can effectively reduce the error caused by the insufficiency of a single feature by comprehensively considering the time difference, amplitude and angle. The design of the comprehensive characteristic weight formula takes into account the characteristics of GIS partial discharge signals, which have weak amplitude, high frequency and strong randomness. Therefore, the combination of amplitude characteristics and time difference in the formula can effectively capture signal characteristics. The introduction of the arrival angle combines the spatial geometric relationship of the sensor and enhances the spatial resolution capability of positioning.

[0178] 3. Calculate single pair positioning results

[0179]

[0180] This formula is a double-end positioning formula, which uses the time difference and the geometric position of the sensors to calculate the single-pair positioning results of each pair of sensors.

[0181] 4. Calculate the single pair positioning results of all sensor pairs

[0182] 5. Weighted fusion to obtain the final positioning result

[0183]

[0184] The weighted fusion method effectively integrates the information of multiple sensor pairs and reduces the impact of single pair positioning errors.

[0185] Specifically, in one embodiment of the present invention, the monitoring host generates a PRPS three-dimensional map and a PRPD two-dimensional map in real time to display the phase, amplitude and frequency characteristics of the local discharge signal; records the occurrence time, location and intensity of the local discharge event, and forms trend analysis and multi-level alarm information to generate relevant diagnostic reports; has three synchronization modes: built-in PT synchronization, AC220V synchronization and external frequency modulation synchronization, with a frequency measurement accuracy of 0.02Hz and a frequency range of 49.0-51.0Hz; and transmits monitoring data and diagnostic reports to the server via optical fiber communication.

[0186] In this embodiment, step S4 mainly displays the location and discharge intensity of the local discharge source through the monitoring host and generates a relevant diagnostic report. Specifically, it includes the following aspects:

[0187] 1. Display the location and discharge intensity of the local discharge source:

[0188] Based on the local discharge source location and discharge intensity calculated in step S3, the local discharge source location and discharge intensity are displayed in real time on the monitoring host. The specific location and intensity of the local discharge source are intuitively displayed through visualization means.

[0189] 2. Generate relevant diagnostic reports:

[0190] Record the time, location, intensity and other information of partial discharge events. Generate trend analysis and multi-level alarm information to help users understand the development trend and severity of partial discharge. Output diagnostic reports as a reference for equipment operation status.

[0191] 3. Visual display:

[0192] The monitoring host generates PRPS (Phase Resolved Partial Discharge) three-dimensional graphs and PRPD (Phase Resolved Partial Discharge) two-dimensional graphs to display the phase, amplitude and frequency characteristics of partial discharge signals. These graphs can help users analyze the characteristics and patterns of partial discharge.

[0193] 4. Data transmission and synchronization:

[0194] The monitoring data and diagnostic reports are transmitted to the server via optical fiber communication, which facilitates remote monitoring and data storage. It supports multiple synchronization methods to ensure the consistency of the data time base.

[0195] Specifically, the specific implementation process of step S4 is as follows:

[0196] 1. Data Visualization

[0197] The position display is based on the local discharge source position L calculated in step S3. M, the location of the discharge source is displayed graphically on the interface of the monitoring host. If it is a two-dimensional or three-dimensional positioning result, the location of the discharge source can be marked on the plane map or three-dimensional model of the GIS equipment.

[0198] According to the amplitude characteristic of the discharge signal M ij , display the discharge intensity in numerical or graphical form. The discharge intensity can be visually indicated by a mark with a color or size change.

[0199] The PRPS spectrum shows the three-dimensional distribution of the phase, amplitude and frequency of the partial discharge signal. The PRPD spectrum shows the two-dimensional distribution of the phase and amplitude of the partial discharge signal. These spectra help users identify the discharge type and mode by analyzing the periodic characteristics of the partial discharge signal.

[0200] 2. Diagnostic report generation

[0201] Record the time, location and intensity of each partial discharge event. Through trend analysis, determine the changing trend of partial discharge (such as whether the discharge intensity is gradually increasing).

[0202] According to the discharge intensity and trend, different levels of alarms (such as low, medium and high alarms) can be set. When the discharge intensity exceeds the set threshold, the alarm is triggered and the user is notified.

[0203] The report includes the time, location, intensity, trend analysis and alarm information of partial discharge events. The report can be output in electronic or printed form as a basis for equipment maintenance.

[0204] 3. Data transmission and storage

[0205] The monitoring data and diagnostic reports are transmitted to the server via optical fiber communication, which is convenient for remote monitoring and data storage. It supports PT synchronization, AC220V synchronization and external frequency modulation synchronization to ensure the consistency of the time base of the data. The synchronization accuracy is 0.02Hz and the frequency range is 49.0-51.0Hz.

[0206] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A GIS partial discharge location method based on multi-channel signal processing, characterized in that: include: S1, collect multi-channel electromagnetic wave signals generated by partial discharge through multiple UHF sensors distributed inside the GIS equipment; S2, preprocessing the collected electromagnetic wave signals, including noise suppression and time difference calculation; S3, based on the time difference and amplitude information of multi-channel electromagnetic wave signals, a fusion algorithm is used in combination with a double-end positioning formula to calculate the location of the local discharge source; S4. The location and discharge intensity of the local discharge source are displayed on the monitoring host, and a relevant diagnostic report is generated.

2. A GIS partial discharge location method based on multi-channel signal processing as claimed in claim 1, characterized in that: In step S2, noise suppression uses a wavelet denoising algorithm based on a feature library, and the process is as follows: Perform wavelet decomposition on the collected electromagnetic wave signals to obtain multi-scale detail coefficients and approximate coefficients; According to the feature library of partial discharge signals, the threshold of wavelet decomposition is dynamically adjusted. The threshold calculation formula is: In the formula, λ a is the threshold of the a-th layer, σ a is the standard deviation of the detail coefficient of the ath layer, L is the signal length; K a is the adjustment factor; SNR a is the signal-to-noise ratio of the detail coefficient of the a-th layer, Among them, S a is the signal component, n a is the noise component, E[·] represents the mathematical expectation; Kurt a is the kurtosis of the detail coefficient of the a-th layer, Among them, c a is the detail coefficient of layer a, μ a is the mean value of the detail coefficient of layer a; The detail coefficient is processed by soft threshold, and the processed detail coefficient d a The calculation formula is: In the formula, sgn(c a ) is a symbolic function; The processed detail coefficients and approximate coefficients are used for wavelet reconstruction to obtain the denoised signal.

3. A GIS partial discharge location method based on multi-channel signal processing as claimed in claim 2, characterized in that: The feature library of partial discharge signals is constructed through the following steps: Select clear signal samples from the partial discharge history data recorded by GIS equipment. The signal samples include a variety of typical partial discharge types. Perform time-frequency transformation on signal samples and extract time-frequency domain features, including amplitude, spectrum distribution and main frequency band power; further extract nonlinear features of signal samples, including fractal dimension, sample entropy and multi-scale entropy; Normalize the extracted signal features to eliminate differences in amplitude and spectrum range; Perform multimodal weighted fusion on the extracted time-frequency domain features and nonlinear features to generate a multimodal feature vector; Based on the constructed multimodal feature vector, the self-attention convolutional neural network is used to perform deep feature extraction; The density-based dynamic clustering algorithm DBSCAN-ADV is used to cluster the depth features. According to the density distribution of frequency bands and amplitudes, the depth features are classified and the noise points with low density are removed. DBSCAN-ADV introduces a dynamic neighborhood radius adjustment mechanism and sets the adaptive neighborhood parameter ∈ according to the frequency band characteristics of GIS partial discharge signals: In the formula, F c is the multimodal feature vector, M is the number of collected signal samples; The classified feature data are stored as standard feature templates to form a feature library of partial discharge signals.

4. A GIS partial discharge location method based on multi-channel signal processing as claimed in claim 3, characterized in that: During the operation of the GIS equipment, the new signal collected in real time is compared with the standard feature template in the feature library. If the feature of the new signal exceeds the template error threshold, the feature library is automatically updated to generate a new standard feature template.

5. A GIS partial discharge location method based on multi-channel signal processing as claimed in claim 2, characterized in that: In step S2, the time difference calculation is performed based on the electromagnetic wave signal after noise suppression processing, and the steps are as follows: The N UHF sensors installed in the GIS equipment are numbered {C1, C2, ..., C N }, N UHF sensors are combined in pairs to form sensor pairs, forming a total of For sensor pairs; The cross-correlation function is used to calculate each pair of sensors (C i ,C j )The received signal (x i ,y j ) Similarity under different time delays, the cross-correlation function calculation formula is as follows: In the formula, is the signal x when the time delay is τ i and i similarity; [0,T] represents the time range of signal acquisition; x i (t) and y j (t) are the sensor pairs (C i ,C j ) received original time domain signal; X i (f) is x i Fourier transform of (t); for y j The complex conjugate of the Fourier transform of (t); For x i (t) and y j The phase difference function of (t+τ) represents the phase difference between the two signals under the time delay τ; w1(t) is the amplitude weight function; w2(f) is the frequency weight function; w3(t) is the phase weight function; The time delay corresponding to the maximum similarity value is taken as the time difference of the pair of sensors.

6. A GIS partial discharge location method based on multi-channel signal processing as claimed in claim 5, characterized in that: Step S3 includes: S31, for each pair of sensors (C i ,C j ), extract the time difference Δt ij , amplitude characteristics M ij and arrival angle A ij , where the amplitude characteristic M ij By pairing the sensor (C i ,C j ) is normalized and the amplitude average is calculated to obtain the arrival angle A ij Based on the phase difference and spatial geometric relationship of the signal, the signal arrival sensor pair (C i ,C j ) to obtain the angle information; S32, according to the sensor pair (C i ,C j ) Extracted time difference Δt ij , amplitude characteristics M ij and arrival angle A ij , calculate the comprehensive characteristic weight ω ij ; S33, calculate the sensor pair (C based on the double-end positioning formula i ,C j )’s single pair positioning result L M,ij ; S34, repeating steps S31-S33, calculating the comprehensive characteristic weights of all sensor pairs and the positioning results of a single pair; S35. Perform weighted fusion on all single-pair positioning results using comprehensive characteristic weights to calculate the final location of the local discharge source.

7. A GIS partial discharge location method based on multi-channel signal processing as claimed in claim 6, characterized in that: In step S32, the comprehensive characteristic weight ω ij The calculation formula is: In the formula, Δt ij For the sensor pair (C i ,C j ) time difference; is the average time difference of all sensor pairs; T s is the standardized parameter of time difference; M ij For the sensor pair (C i ,C j )’s amplitude characteristics; σ M,ij For the sensor pair (C i ,C j ) is the amplitude standard deviation; σ0 is the reference value of the amplitude standard deviation; A ij For the sensor pair (C i ,C j )’s arrival angle; k represents the kth sensor pair; is the number of sensor pairs; α, β, γ are weight coefficients, and α+β+γ=1.

8. A GIS partial discharge location method based on multi-channel signal processing as claimed in claim 6, characterized in that: Single pair positioning result L M,ij The calculation formula is: Where, L ij For the sensor pair (C i ,C j ) distance; V is the propagation speed of electromagnetic waves in GIS; The final calculation formula for the location of the partial discharge source is: Where, L M is the location of the final partial discharge source.

9. A GIS partial discharge location method based on multi-channel signal processing as claimed in claim 1, characterized in that: UHF sensors include: Built-in sensor, installed inside the GIS tank, adopts a disc sensor structure; The external sensor is installed on the pot-type insulator outside the GIS tank to receive the electromagnetic wave signal radiated from the inside of the GIS to the outside; An environmental noise sensor, used for collecting environmental noise signals; The frequency range of the UHF sensor is 300MHz to 1500MHz, and each sensor is equipped with an independent acquisition channel, using a 50MHz sampling rate and a 12-bit ADC.

10. A GIS partial discharge location method based on multi-channel signal processing as claimed in claim 1, characterized in that: In step S4, the monitoring host generates a PRPS three-dimensional map and a PRPD two-dimensional map in real time to display the phase, amplitude and frequency characteristics of the partial discharge signal; Record the time, location and intensity of partial discharge events, and form trend analysis and multi-level alarm information to generate relevant diagnostic reports; built-in PT synchronization, AC220V synchronization and external frequency modulation synchronization, three synchronization modes, frequency measurement accuracy is 0.02Hz, frequency range is 49.0-51.0Hz; transmit monitoring data and diagnostic reports to the server via optical fiber communication.

Images