Sparse blind separation-based tubular power equipment partial discharge positioning detection method
Through sparse blind separation technology and signal complexity screening, combined with cross-correlation delay estimation, the problems of insufficient noise suppression capability and large positioning errors in tubular power equipment are solved, and high-precision fault point detection is achieved.
Patent Information
- Application Number
- CN202510450885.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-11
- Publication Date
- 2025-07-25
AI Technical Summary
In the prior art, in detecting tubular power equipment, especially equipment containing T-type pipe joints and elbows, the noise suppression ability is limited, the positioning error is large, and it is difficult to accurately detect the fault point.
Using a sparse blind separation method, two vibration sensors are used to collect signals and pre-process them. Blind separation is performed by sparse component analysis method, combined with sample entropy and classifier to screen the signal complexity, and finally positioning detection is performed through cross-correlation delay estimation.
It improves the positioning accuracy of tubular power equipment, effectively suppresses noise source signals, reduces sensor usage, reduces cost, and improves detection accuracy and consistency.
Smart Images
Figure CN120370105A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of partial discharge location of power equipment, and particularly to a method for detecting partial discharge location of tubular power equipment based on sparse blind separation. Background Art
[0002] Accurately detecting the partial discharge point of a faulty power equipment is an important prerequisite for ensuring the orderly construction and safe operation of urban power equipment.
[0003] During the process of partial discharge location detection, the noise sources can be classified into internal noise sources and external noise sources. External noise mainly comes from traffic noise, pedestrian walking noise, construction noise, etc. The internal noise of tubular power equipment mainly comes from noise sources such as T-shaped pipe joints and elbows. When partial discharge passes through a bent pipe, the curvature of the bent pipe has an important influence on the sound field characteristics of partial discharge. When a faulty tubular power equipment has T-shaped pipe joints and elbows, it greatly increases the detection difficulty. Traditional partial discharge location methods have limited detection capabilities for power equipment with T-shaped pipe joints and bent pipes. Currently, the mainstream methods for detecting and locating partial discharge of tubular power equipment are: neural network learning method, time delay estimation method, wavelet analysis method for partial discharge signals of tubular power equipment, etc. The neural network learning method refers to using more than one acoustic or pressure sensor to pick up the partial discharge signals of tubular power equipment, and using neural network algorithms such as BP learning method and Elman learning method to analyze the partial discharge signals to complete the location detection work; while the currently mainstream wavelet analysis method for partial discharge sound signals of tubular power equipment has limited location capabilities for power equipment with T-shaped pipe joints and bent pipes, because the wavelet analysis method has a strong dependence on the selection of wavelet basis functions, and different wavelet basis functions result in different positioning errors; the time delay estimation method has limited ability to suppress internal and external noises of tubular power equipment, and due to the existence of internal noise in tubular power equipment, the correlation peak of the partial discharge signal obtained by performing cross-correlation analysis is easily submerged by the noise correlation peak, resulting in a relatively large positioning error. Although the above-mentioned various methods and theories are innovative to a certain extent, there are still deficiencies and defects in the location detection of partial discharge tubular power equipment with T-shaped pipe joints and elbows, such as limited noise suppression ability and large positioning error. Summary of the Invention
[0004] In view of the above deficiencies in the prior art, the present invention provides a method for detecting partial discharge location of tubular power equipment based on sparse blind separation, which is used to solve the problems of limited noise suppression ability and large positioning error in the location detection of partial discharge tubular power equipment with T-shaped pipe joints and elbows in the prior art.
[0005] In order to achieve the above invention purpose, the technical solution adopted by the present invention is as follows:
[0006] A method for detecting partial discharge location of tubular power equipment based on sparse blind separation, comprising the following steps:
[0007] S1. Collect the noisy partial discharge source signals at both ends of the tubular power equipment using two vibration sensors, and perform preprocessing to obtain the preprocessed noisy partial discharge source signals. At the same time, use the sparse component analysis method to perform blind separation on the preprocessed noisy partial discharge source signals to obtain the blindly separated noisy partial discharge source signals;
[0008] Among them, the blindly separated noisy partial discharge source signals include the internal noise source signals and partial discharge source signals of the tubular power equipment;
[0009] S2. Calculate the complexity of the blindly separated noisy partial discharge source signals using sample entropy to obtain the complexity value of the noisy partial discharge source signals;
[0010] Among them, the complexity value of the noisy partial discharge source signals includes the complexity value of the internal noise source signals and the complexity value of the partial discharge source signals of the tubular power equipment;
[0011] S3. Input the complexity value of the noisy partial discharge source signals into the trained classifier for signal separation to obtain the separated partial discharge source signals and the internal noise source signals of the tubular power equipment;
[0012] S4. Perform cross-correlation time delay estimation on the separated partial discharge source signals to generate the positioning detection results of the tubular power equipment.
[0013] The present invention has the following beneficial effects:
[0014] 1. The partial discharge positioning detection method for tubular power equipment based on sparse blind separation proposed by the present invention uses the sparse component analysis technology to separate the noisy partial discharge source signals, making it more rigorous in mathematical principle, and solving the problems of under-decomposition and over-decomposition existing in the existing technologies such as empirical mode decomposition. At the same time, it effectively processes the separation of noisy partial discharge signals under underdetermined conditions, reduces the usage amount of vibration sensors, and saves costs;
[0015] 2. When screening the partial discharge source signals and internal noise source signals, the signal complexity value and the classifier are combined to complete the screening of the partial discharge source signals. Compared with the existing calculation methods such as approximate entropy in the prior art, the data length generated by combining the signal complexity value and the classifier in the present invention has little influence on the calculation result of the signal complexity, and at the same time has better consistency, lower requirements for the integrity of the data, and improves the positioning detection accuracy and suppresses the noise source signals;
[0016] 3. Before making a time delay estimation for the local discharge source signal, a definition of a cross-correlation quality factor Q is introduced. The higher the Q value, the more obvious the cross-correlation peak. Experience shows that the cross-correlation quality factor below a certain empirical value is not readable. The coherence function proposed in the present invention is a frequency domain coherence, which is defined as the quotient of the square of the cross-power spectrum density and the auto-power spectrum density. Therefore, after the cross-correlation function is processed using the method proposed in the present invention, compared with the direct cross-correlation, the cross-correlation quality factor is greater than the empirical value, which greatly increases the readability of the cross-correlation peak. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 It is a flow chart of the method for locating and detecting partial discharge of tubular power equipment based on sparse blind separation proposed by the present invention;
[0018] Figure 2 A schematic diagram of the positions of the vibration sensors arranged at both ends of the tubular power equipment in the embodiment;
[0019] Figure 3 Schematic diagram of power spectrum of noisy local discharge source signal in the embodiment;
[0020] Figure 4 It is a schematic diagram of the calculation results of the random complexity of the partial discharge source signal and the noise source signal in the embodiment. DETAILED DESCRIPTION
[0021] The specific implementation modes of the present invention are described below to facilitate those skilled in the art to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.
[0022] like Figure 1 As shown, the method for localizing and detecting partial discharge of tubular power equipment based on sparse blind separation includes the following steps S1-S4:
[0023] S1. Use two vibration sensors to collect noisy partial discharge source signals at both ends of the tubular power equipment of the T-type pipe head and the bend pipe head respectively, and preprocess them to obtain preprocessed noisy partial discharge source signals. At the same time, use sparse component analysis to blindly separate the preprocessed noisy partial discharge source signals to obtain blindly separated noisy partial discharge source signals; wherein the blindly separated noisy partial discharge source signals include internal noise source signals and partial discharge source signals of the tubular power equipment.
[0024] In this embodiment, two tubular power equipment both containing partial discharge and respectively having T-shaped pipe joints or elbows are taken as the research objects, and two vibration sensors are used to respectively collect the noisy partial discharge source signals of the tubular power equipment with T-shaped pipe joints or elbows. Specifically, as Figure 2 shown, where Figure 2 (a) is a schematic diagram of the process of collecting the noisy partial discharge source signals of the tubular power equipment with a T-shaped pipe joint, Figure 2 (b) is a schematic diagram of the process of collecting the noisy partial discharge source signals of the tubular power equipment with an elbow. That is, the vibration sensor with a magnetic base is adsorbed on the tubular power equipment, and the magnetic base should not be screwed too tightly to avoid the failure to respond to the partial discharge signal in a timely manner due to the too tight screwing of the sensor and the magnetic base. In addition, after using two sensors to collect the noisy partial discharge source signals at both ends of the tubular power equipment, all the noisy partial discharge source signals are preprocessed including pre-whitening, frequency domain transformation and normalization to generate the preprocessed noisy partial discharge source signals, so as to use them as the input observation signals of the blind separation algorithm. Since there are only a few dominant spectral peaks in the entire frequency domain of the noisy partial discharge signal, which satisfies the sparse property, the blind separation algorithm adopts a separation method based on sparse component analysis (SCA). The power spectrum of the noisy partial discharge source signal is as Figure 3 shown; therefore, after all the observation signals are separated by the sparse component analysis method, a series of source output signals are obtained. These source output signals include both partial discharge source signals and internal noise source signals of the tubular power equipment, so as to calculate the complexity in the subsequent steps.
[0025] Specifically, step S1 specifically includes S11 - S16:
[0026] S11. Arrange two vibration sensors at both ends of the tubular power equipment to collect the noisy partial discharge source signals, that is:
[0027]
[0028] where X and Y respectively represent the matrices of the noisy partial discharge source signals collected by the two vibration sensors, and X1, X2, X i , X m respectively represent the 1st, 2nd, ith and mth groups of noisy partial discharge source signals collected by the first vibration sensor, and Y1, Y2, Y i , Y m respectively represent the 1st, 2nd, ith and mth groups of noisy partial discharge source signals collected by the second vibration sensor.
[0029] In this embodiment, the first vibration sensor and the second vibration sensor respectively collect m groups of noisy partial discharge source signals, and the m groups of noisy partial discharge source signals collected by each vibration sensor form a matrix of noisy partial discharge source signals; in addition, based on the propagation characteristics of the partial discharge source signals, the following model can be established, that is:
[0030]
[0031] S12. Perform pre-whitening processing on the noisy partial discharge source signal to obtain a pre-whitened noisy partial discharge source signal.
[0032] In this embodiment, the purpose of performing pre-whitening processing on the noisy partial discharge source signal is to maximize the mutual independence of the source mixed signals in the noisy partial discharge source signal, thereby improving the effectiveness of the blind separation algorithm in subsequent steps.
[0033] S13. Perform frequency-domain transformation on the pre-whitened noisy partial discharge source signal, extract the mixed signals according to the correlations in different frequency bands, and obtain a first mixed signal matrix and a second mixed signal matrix, that is:
[0034]
[0035] wherein, X′ and Y′ respectively represent the first mixed signal matrix and the second mixed signal matrix, and X′1, X′2, X′ j , X′ n respectively represent the first, second, jth, and nth groups of noisy partial discharge source signals in the first mixed signal matrix, and Y′1, Y′2, Y′ j , Y′ n respectively represent the first, second, jth, and nth groups of noisy partial discharge source signals in the second mixed signal matrix.
[0036] In this embodiment, the purpose of the frequency-domain transformation is to extract the mixed signals according to the correlations in different frequency bands, so as to find the corresponding mixing matrix and transfer function based on the received pre-whitened noisy partial discharge source signal, thereby making the sparse coefficients as sparse as possible.
[0037] S14. Perform normalization processing on the first mixed signal matrix and the second mixed signal matrix to obtain a normalized first mixed signal matrix and a normalized second mixed signal matrix, that is:
[0038]
[0039] wherein, A and B respectively represent the normalized first mixed signal matrix and the normalized second mixed signal matrix, and A1, A2, A j , A n respectively represent the first, second, jth, and nth groups of noisy partial discharge source signals in the normalized first mixed signal matrix, and B1, B2, B j , B n respectively represent the first, second, jth, and nth groups of noisy partial discharge source signals in the...
[0040] S15. Extract the first source signal matrix using the sparse component analysis method based on the first mixed signal matrix and the normalized first mixed signal matrix, i.e.:
[0041] s = [s1, s2, …, s n
[0042] where s represents the first source signal matrix, and s1, s2, s n represent the 1st, 2nd, …, nth first source signals in the first source signal matrix respectively.
[0043] S16. Extract the second source signal matrix using the sparse component analysis method based on the second mixed signal matrix and the normalized second mixed signal matrix, i.e.:
[0044] S = [S1, S2, …, S n
[0045] where S represents the second source signal matrix, and S1, S2, S n represent the 1st, 2nd, …, nth second source signals in the second source signal matrix respectively.
[0046] The first source signal matrix and the second source signal matrix together constitute the noisy partial discharge source signal for blind separation.
[0047] In summary, this step uses a popular technique in current blind separation techniques, namely the sparse component analysis technique, when separating the observed signal (noisy partial discharge source signal). In the partial discharge detection of tubular power equipment, the internal noise source and the partial discharge source signal are superimposed. It can be considered that a series of noises existing in T-joints, elbows, and other joints of tubular power equipment are additive noises, which are theoretically independent and incoherent with the partial discharge signal. However, in practice, even after whitening the signal, this internal noise source cannot be strictly independent and uncorrelated with the partial discharge source, and in most cases, the observed signal to be separated is underdetermined. Therefore, compared with empirical mode decomposition techniques such as empirical mode decomposition (EMD) and ensemble empirical mode decomposition (EEMD), the separation method of sparse component analysis adopted in the present invention is more rigorous in mathematical principle and solves the problems of under-decomposition and over-decomposition in empirical mode decomposition. At the same time, compared with traditional blind separation techniques such as independent component analysis (ICA) and principal component analysis (PCA), the separation method proposed in the present invention can effectively handle the separation of noisy partial discharge signals under underdetermined conditions, reduce the usage of sensors, and save costs.
[0048] S2. Calculate the complexity of the noisy partial discharge source signal for blind separation using sample entropy to obtain the complexity value of the noisy partial discharge source signal; wherein, the complexity value of the noisy partial discharge source signal includes the complexity value of the internal noise source signal of the tubular power equipment and the complexity value of the partial discharge source signal.
[0049] In this embodiment, the sample entropy is used to calculate the complexity of the noisy partial discharge source signal for blind separation, so that the calculated complexity value can be used as the input of the classifier in the subsequent steps. Among them, the complexity value of the noisy partial discharge source signal is as Figure 4 shown.
[0050] S3. Input the complexity value of the noisy partial discharge source signal into the trained classifier for signal separation to obtain the separated partial discharge source signal and the internal noise source signal of the tubular power equipment.
[0051] In this embodiment, since the signal randomness (complexity) of the partial discharge source signal itself is much greater than the internal noise of the tubular power equipment such as the T-joint noise and elbow noise, the classifier can distinguish the partial discharge source signal and the internal noise source signal of the tubular power equipment according to this training criterion.
[0052] Specifically, when the trained classifier in step S3 is trained, based on the criterion that the complexity of the partial discharge source signal is much greater than the internal noise source signal of the tubular power equipment, the classifier is trained by assigning different labels by judging whether it is a partial discharge source signal, generating an empirical threshold, and taking the source signal exceeding the empirical threshold as the partial discharge source signal.
[0053] Specifically, step S3 specifically includes:
[0054] Input the complexity value of the noisy partial discharge source signal into the trained classifier for signal separation. By taking the source signal exceeding the empirical threshold in the complexity value of the noisy partial discharge source signal as the partial discharge source signal, the separated partial discharge source signal and the internal noise source signal of the tubular power equipment are obtained.
[0055] Among them, the separated partial discharge source signal includes a first partial discharge source signal matrix and a second partial discharge source signal matrix, and each partial discharge source signal in the first partial discharge source signal matrix corresponds to each partial discharge source signal in the second partial discharge source signal matrix, that is:
[0056]
[0057] Among them, A′ represents the first partial discharge source signal matrix, B′ represents the second partial discharge source signal matrix, A′1, A′2, A′ a , A′ k respectively represent the 1st group, 2nd group, ath group, and kth group of partial discharge source signals in the first partial discharge source signal matrix, and B′1, B′2, B′ a , B′ k respectively represent the 1st group, 2nd group, ath group, and kth group of partial discharge source signals in the second partial discharge source signal matrix.
[0058] Among them, the a-th group of partial discharge source signals A' in the first partial discharge source signal matrix a corresponds to the a-th group of partial discharge source signals B' in the second partial discharge source signal matrix a respectively.
[0059] In summary, in steps S2 - S3, when screening partial discharge source signals and internal noise source signals, the signal complexity value and the classifier are combined to complete the screening task. The results show that the complexity values of partial discharge source signals are basically greater than 1, while the complexity values of internal noise source signals are generally less than 1; these data reflect the degree of random complexity of the signals. The larger the value, the greater the randomness of the signals. Therefore, compared with calculation methods such as approximate entropy, the data length has little influence on the calculation result of the signal complexity in the present invention. At the same time, the method proposed in the present invention has better consistency and lower requirements for data integrity.
[0060] S4. Perform cross-correlation time-delay estimation on the separated partial discharge source signals to generate a positioning detection result for tubular power equipment.
[0061] In this embodiment, cross-correlation time-delay estimation is performed on the partial discharge source signals obtained by the classifier. Here, a correlation peak enhancement technique is adopted to make the partial discharge correlation peak more obvious. According to the characteristics of the tubular power equipment material, the empirical sound speed can be used to complete the positioning of the partial discharge point of the tubular power equipment. Among them, a correlation peak enhancement technique is a band-pass filtering technique based on the coherence function, and the coherence function is the frequency-domain coherence determined by the signal auto-power spectrum and cross-power spectrum.
[0062] Specifically, step S4 specifically includes S41 - S43:
[0063] S41. Respectively perform cross-correlation time-delay estimation on each partial discharge source signal in the first partial discharge source signal matrix of the separated partial discharge source signals and each partial discharge source signal in the corresponding second partial discharge source signal matrix by using the peak enhancement method of band-pass filtering based on the coherence function to obtain several groups of different time-delay values.
[0064] S42. Calculate the average value of several groups of different time-delay values;
[0065] S43. According to the average value of several groups of different time-delay values and the tubular power equipment material, select the empirical sound speed to determine the positioning position of the partial discharge point, that is:
[0066]
[0067] Among them, L represents the propagation path length from the partial discharge point positioning position to a certain sensor, L1 represents the total length of the distance between two sensors, v represents the empirical sound speed, τ represents the average value of several groups of different time delay values, B represents the volume elastic modulus parameter of the fluid medium inside the pipe of the tubular power equipment, ρ represents the density parameter of the fluid medium inside the pipe of the tubular power equipment, E represents the Young's elastic modulus of the ductile iron material, and h represents the pipe thickness of the tubular power equipment.
[0068] In this embodiment, after automatically sending the complexity value into the classifier, if a support vector machine is used for screening and classification work, the effective partial discharge source signals are screened out, and the screening process ends to obtain the partial discharge source signals, A′1, A′2, …, A′ a , …, A′ k and B′1, B′2, …, B′ a , …, B′ k , where A′1 corresponds to B′1, A′2 corresponds to B′2, A′ a and B′ a correspond, A′ k and B′ k correspond, and the cross-correlation time delay estimation is performed on these corresponding partial discharge source signals. Among them, the time delay estimation uses the peak enhancement technology based on the coherence function band-pass filtering, and several groups of different time delay values can be obtained. Take their arithmetic mean and denote it as τ; at the same time, according to the different materials of the tubular power equipment, the empirical sound speed v is selected to determine the partial discharge point position.
[0069] In summary, before performing the time delay estimation on the partial discharge source signal in this step, the definition of a cross-correlation quality factor Q is introduced. The higher the Q value, the more obvious the cross-correlation peak. Its experience shows that people think that the cross-correlation quality factor below a certain empirical value is not readable. And the coherence function proposed in the present invention is a frequency domain coherence, which is defined as the quotient of the square of the cross-power spectral density and the auto-power spectral density; therefore, after processing the cross-correlation function using the method proposed in the present invention, compared with the direct cross-correlation, the cross-correlation quality factors are all greater than this empirical value, greatly increasing the readability of the cross-correlation peak.
[0070] In the present invention, specific embodiments are applied to elaborate the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.
[0071] Those of ordinary skill in the art will realize that the embodiments described herein are provided to assist the reader in understanding the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations that do not depart from the essence of the present invention based on these technical revelations disclosed in the present invention, and these deformations and combinations are still within the scope of protection of the present invention.
Claims
1. A partial discharge location detection method for tubular power equipment based on sparse blind separation, characterized in that It includes the following steps: S1. Use two vibration sensors to collect the noisy partial discharge source signals at both ends of the tubular power equipment, and perform preprocessing to obtain the preprocessed noisy partial discharge source signals. At the same time, use the sparse component analysis method to perform blind separation on the preprocessed noisy partial discharge source signals to obtain the blindly separated noisy partial discharge source signals; Among them, the blindly separated noisy partial discharge source signals include the internal noise source signals and partial discharge source signals of the tubular power equipment; S2. Use sample entropy to calculate the complexity of the blindly separated noisy partial discharge source signals to obtain the complexity values of the noisy partial discharge source signals; Among them, the complexity values of the noisy partial discharge source signals include the complexity values of the internal noise source signals and partial discharge source signals of the tubular power equipment; S3. Input the complexity values of the noisy partial discharge source signals into the trained classifier for signal separation to obtain the separated partial discharge source signals and the internal noise source signals of the tubular power equipment; S4. Perform cross-correlation time delay estimation on the separated partial discharge source signals to generate the positioning detection results of the tubular power equipment.
2. The partial discharge location detection method for tubular power equipment based on sparse blind separation according to claim 1, characterized in that, Step S1 specifically includes: S11. Arrange two vibration sensors at both ends of the tubular power equipment to collect the noisy partial discharge source signals, that is: Among them, X and Y represent the noisy partial discharge source signal matrix collected by the two vibration sensors, X1, X2, X i , X m They represent the first, second, i-th and m-th groups of noisy partial discharge source signals collected by the first vibration sensor, Y1, Y2, Y i , Y m They respectively represent the first, second, i-th and m-th groups of noisy partial discharge source signals collected by the second vibration sensor; S12. Perform pre-whitening processing on the noisy partial discharge source signals to obtain the pre-whitened noisy partial discharge source signals; S13. Perform frequency domain transformation on the pre-whitened noisy partial discharge source signals, and extract the mixed signals according to the correlation in different frequency bands to obtain the first mixed signal matrix and the second mixed signal matrix, that is: Among them, X′ and Y′ respectively represent the first mixed signal matrix and the second mixed signal matrix, and X′1, X′2, X′ j , X′ n respectively represent the first, second, jth, and nth groups of noisy partial discharge source signals in the first mixed signal matrix, and Y′1, Y′2, Y′ j , Y′ n respectively represent the first, second, jth, and nth groups of noisy partial discharge source signals in the second mixed signal matrix; S14. Perform normalization processing on the first mixed signal matrix and the second mixed signal matrix to obtain the normalized first mixed signal matrix and the second mixed signal matrix, that is: where A and B respectively represent the normalized first mixed signal matrix and the normalized second mixed signal matrix, and A1, A2, A j , A n respectively represent the noisy partial discharge source signals in the 1st, 2nd, jth, and nth groups in the normalized first mixed signal matrix, and B1, B2, B j , B n respectively represent the noisy partial discharge source signals in the 1st, 2nd, jth, and nth groups; S15. According to the first mixed signal matrix and the normalized first mixed signal matrix, use the sparse component analysis method to extract the first source signal matrix, that is: s = [s1, s2, …, s n where s represents the first source signal matrix, and s1, s2, s n respectively represent the 1st, 2nd, and nth first source signals in the first source signal matrix; S16. According to the second mixed signal matrix and the normalized second mixed signal matrix, use the sparse component analysis method to extract the second source signal matrix, that is: S = [S1, S2, …, S n where S represents the second source signal matrix, and S1, S2, S n represent the first, second, and nth second source signals in the second source signal matrix, respectively; Among them, the first source signal matrix and the second source signal matrix jointly form the blindly separated noisy partial discharge source signals.
3. The partial discharge location detection method for tubular power equipment based on sparse blind separation according to claim 2, wherein, When training the trained classifier in step S3, based on the criterion that the complexity of the partial discharge source signals is much greater than that of the internal noise source signals of the tubular power equipment, the classifier is trained by assigning different labels by judging whether it is a partial discharge source signal, generating an empirical threshold, and taking the source signals exceeding the empirical threshold as partial discharge source signals.
4. The method for detecting the partial discharge location of a tubular power equipment based on sparse blind separation according to claim 3, wherein, Step S3 specifically includes: Input the complexity values of the noisy partial discharge source signals into the trained classifier for signal separation. By taking the source signals exceeding the empirical threshold in the complexity values of the noisy partial discharge source signals as partial discharge source signals, the separated partial discharge source signals and the internal noise source signals of the tubular power equipment are obtained; Among them, the separated partial discharge source signals include the first partial discharge source signal matrix and the second partial discharge source signal matrix, and each partial discharge source signal in the first partial discharge source signal matrix corresponds to each partial discharge source signal in the second partial discharge source signal matrix, that is: Among them, A′ represents the first-stage source emission signal matrix, B′ represents the second-stage source emission signal matrix, A′1, A′2, A′ a , A′ k respectively represent the first group, the second group, the a-th group, and the k-th group of partial discharge source signals in the first-stage source emission signal matrix, and B′1, B′2, B′ a , B′ k respectively represent the first group, the second group, the a-th group, and the k-th group of partial discharge source signals in the second-stage source emission signal matrix; Among them, the a-th group of partial discharge source signals A' in the first partial discharge source signal matrix a corresponds to the a-th group of partial discharge source signals B' in the second partial discharge source signal matrix a respectively.
5. The method for partial discharge location detection of tubular power equipment based on sparse blind separation according to claim 4, wherein, Step S4 specifically includes: S41. Respectively perform cross-correlation time-delay estimation on the partial discharge source signals in the first partial discharge source signal matrix and the corresponding partial discharge source signals in the second partial discharge source signal matrix in the separated partial discharge source signals by using the peak enhancement method based on coherent function band-pass filtering to obtain several groups of different time-delay values; S42. Calculate the average value of several groups of different time-delay values; S43. Select the empirical sound speed according to the average value of several groups of different time-delay values and the tubular power equipment material to determine the partial discharge point positioning position, that is: where L represents the propagation path length from the partial discharge point positioning position to a certain sensor, L1 represents the total length of the distance between two sensors, v represents the empirical sound speed, τ represents the average value of several groups of different time-delay values, B represents the volume elastic modulus parameter of the fluid medium inside the pipe of the tubular power equipment, ρ represents the density parameter of the fluid medium inside the pipe of the tubular power equipment, E represents the Young's elastic modulus of the ductile iron material, and h represents the pipe thickness of the tubular power equipment.