A method for constructing partial discharge characteristic matrix based on multi-physics information

Abstract

The present invention relates to the field of electrical engineering, and in particular to a method for constructing a partial discharge characteristic matrix based on multi-physics information. The method comprises: obtaining multi-physics information of partial discharge through multiple means and constructing an original matrix of the multi-physics information; analyzing and extracting features from various signal characteristics based on the original matrix to construct a characteristic matrix of the multi-physics information; and performing algorithmic evaluation on the multi-physics information features, selecting significant features based on the evaluation scores, and constructing a simplified matrix of the multi-physics information features. The present invention utilizes the relevant characteristics of multiple types of physical signals associated with partial discharge to construct a partial discharge characteristic matrix based on multi-physics information. This method solves the problems of sensitivity differences and detection blind spots for different discharge types using a single detection method. The method simultaneously extracts and simplifies the multi-physics information features, reducing data redundancy while retaining feature validity.

Description

Technical Field

[0001] The present invention relates to the field of electrical engineering, and in particular to a method for constructing a partial discharge characteristic matrix based on multi-physical information. Background Art

[0002] Gas-insulated switchgear (GIS) achieves electrical insulation through high-pressure electronegative gas. Its advantages include a small footprint, low maintenance costs, and strong environmental adaptability, making it widely used in power systems. However, during GIS assembly or long-term operation, internal defects such as sharp burrs and material air gaps can cause strong electric fields to cause partial discharge (PD). If not detected promptly and effectively, the development of PD can further degrade insulation, shorten equipment life, and even lead to serious power accidents. Therefore, timely detection of PD and identification of discharge type facilitates early warning and provides support to operations and maintenance personnel, ensuring safe and stable equipment operation.

[0003] Current partial discharge detection methods utilize various physical phenomena associated with discharge. High-frequency current methods detect high-frequency current signals generated by charge movement in space; ultra-high frequency methods detect electromagnetic wave signals generated by discharge using antennas in specific frequency bands; and ultrasonic methods utilize the piezoelectric effect of materials to detect ultrasonic signals generated by discharge. These methods are widely used for online detection of partial discharge in power equipment. However, due to the complex field environment, each method has limitations in its application. Furthermore, while the commonly used multi-physics information combined detection method can compensate for the shortcomings of a single detection method, the detection and analysis of multi-physics information can easily lead to data redundancy. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the defects of the background technology and provide a method for constructing a local discharge characteristic matrix based on multi-physical information. A variety of detection methods are combined to form a joint detection method of multi-physical information to solve problems such as sensitivity differences and detection blind spots of different discharge types caused by a single detection method. At the same time, by extracting the characteristics of multi-physical information, data simplification is achieved and data redundancy is reduced.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A method for constructing a partial discharge characteristic matrix based on multi-physics information, the method comprising the following steps:

[0007] S1: Obtain multi-physics information of partial discharge through various means and construct the original matrix of multi-physics information;

[0008] S2: Based on the original matrix, analyze and extract features of various signals to construct a feature matrix of multi-physics information;

[0009] S3: Perform algorithmic evaluation on multi-physics information features, select significant features based on the evaluation scores, and construct a simplified matrix of multi-physics information features.

[0010] Furthermore, the step S1 is specifically as follows:

[0011] S1.1: Multi-physics information acquisition: Acquire high-frequency current, ultra-high frequency, and ultrasonic signals generated by partial discharge. These three types of physical signals are synchronously acquired along with device voltage signals through pulse triggering. The acquisition time is 10 seconds.

[0012] S1.2: Pulse and voltage amplitude information acquisition: Using the classic threshold method, the start and end times of the three types of physical signal pulses, as well as the pulse amplitude information and applied voltage amplitude information between the start and end times, are acquired based on the thresholds.

[0013] S1.3: Construct the original matrix of multi-physical information, record the time, pulse amplitude information and external voltage amplitude information of all pulses within the acquisition time of 10s, construct the pulse amplitude information original matrix and voltage original matrix based on the pulse amplitude information and external voltage amplitude information of high-frequency current, ultra-high frequency and ultrasound, and finally construct the original matrix of multi-physical information.

[0014] Furthermore, in step S1.2, according to the threshold value I t Get the start time t0 and end time t of three types of physical signal pulses end The calculation formula is:

[0015] I t =1.1I n,max (1)

[0016] Among them, I n,max is the maximum noise value of the physical signal when there is no partial discharge;

[0017] The physical signal pulse start time t0 is determined based on the following:

[0018]

[0019] Select I i Time t i is the pulse starting time t0;

[0020] Physical signal pulse end time t end The judgment is based on the following:

[0021]

[0022] Select Ii Time t i is the pulse end time t end ;

[0023] Determine the pulse start time t0 and end time t end After that, record t0 and t end The pulse amplitude information I(t0~t end ) and the external voltage amplitude information U(t0~t end ).

[0024] Furthermore, the construction of the original matrix of multi-physics information in step S1.3 is as follows:

[0025] For pulse amplitude information, the original vectors of high-frequency current (HFCT), ultra-high frequency (UHF), and ultrasonic (AE) pulse amplitude information are as follows:

[0026]

[0027] Where, I HFCT (i,1) is the original vector of high-frequency current pulse amplitude information; I UHF (i,1) is the original vector of UHF pulse amplitude information; I AE (i,1) is the original vector of ultrasonic pulse amplitude information; I HFCT (t0~t end ) is t0~t end High-frequency current pulse amplitude information; I UHF (t0~t end ) is t0~t end UHF pulse amplitude information; I AE (t0~t end ) is t0~t end Ultrasonic pulse amplitude information;

[0028] Construct the original matrix I of pulse amplitude information:

[0029] I=(I HCT , I UHF , I AE ) (5)

[0030] The original vectors of each signal voltage are as follows:

[0031]

[0032] Where U HFCT (i,1) is the original voltage vector of the high-frequency current signal; U UHF (i,1) is the original voltage vector of the UHF signal; U AE (i,1) is the original vector of ultrasonic signal voltage; UHFCT (t0~t end ) is t0~t end The information of the applied voltage amplitude of the high-frequency current; U UHT (t0~t end ) is t0~t end UHF external applied voltage amplitude information; U AE (t0~t end ) is t0~t end Ultrasonic applied voltage amplitude information;

[0033] Construct the voltage raw matrix U:

[0034] U=(U HFCT , U UHF , U AE ) (7)

[0035] Construct the original matrix A0 of multi-physics information:

[0036] A0=(I,U) (8)

[0037] Furthermore, the step S2 is specifically as follows:

[0038] S2.1: Feature extraction: Based on the characteristics of high-frequency current, ultra-high frequency, and ultrasonic signals, three features are extracted: average pulse intensity, pulse repetition rate, and average pulse energy. Three features are extracted from ultrasonic signals: average pulse intensity, pulse repetition rate, and average pulse dispersion.

[0039] S2.2: Multi-physical characteristic matrix construction, constructing characteristic vectors according to the eigenvalues of the three types of signals, extracting multiple groups of characteristic vectors based on multiple groups of data to form a characteristic matrix.

[0040] Furthermore, the characteristic value calculation formulas of the three types of signals are as follows:

[0041] The formula for calculating the average pulse intensity is:

[0042]

[0043] Where I(i,1) is the pulse information collected by a single type of physical signal within the acquisition time range, and N is the total number of pulses;

[0044] The pulse repetition rate calculation formula is:

[0045]

[0046] Where, T is the acquisition time;

[0047] In addition, for high-frequency current and ultra-high-frequency signals, the pulse energy average value feature can also be extracted. The single pulse energy calculation formula is as follows:

[0048]

[0049] The formula for the average pulse energy is:

[0050]

[0051] For ultrasonic signals, the pulse dispersion average feature can be extracted. The dispersion calculation formula of a single cluster ultrasonic signal is as follows:

[0052] H i =-∑P(I)logP(I) (13)

[0053] The calculation formula for the average value of pulse dispersion is as follows:

[0054]

[0055] Furthermore, the three types of signals can each obtain three eigenvalues, for a total of nine eigenvalues. The eigenvectors constructed based on the eigenvalues are shown below:

[0056]

[0057] a=(a HFCT , a UHF , a AE ) (16)

[0058] Where a HFCT is the characteristic vector of high-frequency current; a UHF is the characteristic vector of UHF; a AE is the ultrasonic eigenvector; a is the characteristic matrix of a single set of signal data;

[0059] When multiple sets of data are collected to form multiple original matrices, a feature matrix A can be formed after feature extraction. C :

[0060]

[0061] Furthermore, based on the Pearson correlation coefficient between features, the algorithm evaluation of multi-physics information features is performed, and significant features are selected according to the evaluation scores to construct a simplified matrix of multi-physics information. The formula for calculating the Pearson correlation coefficient between features is as follows:

[0062]

[0063] Where X and Y are any two different features. A coefficient value of 1 means that X and Y can be well described by the straight line equation. All data points fall well on the straight line, and X increases as Y increases. A coefficient value of -1 means that all data points fall on the straight line, and X decreases as Y increases. A coefficient value of 0 means that there is no linear relationship between the two variables.

[0064] According to the relevant calculation results, the multi-physical information feature simplified matrix A is constructed S :

[0065] A S =(C1, C2…C n ) (19)

[0066] Where C i is the feature vector.

[0067] Furthermore, the Pearson coefficient is limited to the range of (-0.5, 0.5), retaining the features with weak correlation, and the feature vector C i The values between them are (-0.5, 0.5).

[0068] The beneficial effects of the present invention are: proposing a method for constructing a partial discharge characteristic matrix based on multi-physical information, realizing the complementary advantages of the joint detection of partial discharge by multi-physical means and the full utilization of the physical information associated with partial discharge, avoiding the problems of single detection sensitivity differences and detection blind spots, and reducing data redundancy while retaining feature validity through targeted signal feature extraction and simplification. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0070] Figure 1 Schematic diagram of the overall process of the method for constructing a partial discharge characteristic matrix based on multi-physics information in the present invention;

[0071] Figure 2 This is the key technical process of the construction method in the present invention. DETAILED DESCRIPTION

[0072] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0073] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0074] like Figure 1 A method for constructing a partial discharge characteristic matrix based on multi-physics information is shown, and the method includes the following steps:

[0075] S1: Obtain multi-physics information of partial discharge through various means and construct the original matrix of multi-physics information;

[0076] S2: Based on the original matrix, analyze and extract features of various signals to construct a feature matrix of multi-physics information;

[0077] S3: Perform algorithmic evaluation on multi-physics information features, select significant features based on the evaluation scores, and construct a simplified matrix of multi-physics information features.

[0078] In the present invention, by constructing a multi-physical information original matrix, constructing a feature matrix based on the original matrix, and constructing a feature reduction matrix, high-frequency current detection, ultra-high frequency detection, and ultrasonic detection methods are combined to cover as many types of local discharge defects as possible, effectively avoiding the problems of high-frequency current method and ultra-high frequency method being susceptible to on-site electromagnetic interference, and ultrasonic method having large signal attenuation and being susceptible to mechanical vibration interference, making up for the detection blind spots and sensitivity differences under a single method, reducing missed judgments and misjudgments, and using algorithm evaluation methods to extract and simplify multi-physical information features, reducing data redundancy, reducing calculation and analysis time, and improving efficiency.

[0079] like Figure 2 As shown in Figure 2, in order to obtain the multi-physics information of partial discharge to construct the original matrix, the specific steps are as follows:

[0080] S1.1: Multi-physics information acquisition: Rogowski coils, ultra-high frequency sensors, and ultrasonic sensors are used to acquire high-frequency current, ultra-high frequency, and ultrasonic signals generated by partial discharge. Pulse triggering is used to synchronously acquire these three types of physical signals and device voltage signals. The acquisition time is 10 seconds.

[0081] S1.2: Pulse and voltage amplitude information acquisition: Using the classic threshold method, the start and end times of the three types of physical signal pulses, as well as the pulse amplitude information and applied voltage amplitude information between the start and end times, are acquired based on the thresholds.

[0082] S1.3: Construct the original matrix of multi-physical information, record the time, pulse amplitude information and external voltage amplitude information of all pulses within the acquisition time of 10s, construct the pulse amplitude information original matrix and voltage original matrix based on the pulse amplitude information and external voltage amplitude information of high-frequency current, ultra-high frequency and ultrasound, and finally construct the original matrix of multi-physical information.

[0083] In the actual implementation process, high-frequency current, ultra-high frequency and ultrasonic signals are obtained through Rogowski coils, ultra-high frequency sensors and ultrasonic sensors respectively.

[0084] Specifically, the threshold I t Get the start time t0 and end time t of three types of physical signal pulses end The calculation formula is:

[0085] I t =1.1I n,max (1)

[0086] Among them, I n,max is the maximum noise value of the physical signal when there is no partial discharge;

[0087] The physical signal pulse start time t0 is determined based on the following:

[0088]

[0089] Select I i Time t i is the pulse starting time t0;

[0090] Physical signal pulse end time t end The judgment is based on the following:

[0091]

[0092] Select I i Time t i is the pulse end time t end ;

[0093] Determine the pulse start time t0 and end time t end After that, record t0 and t end The pulse amplitude information I(t0~t end ) and the external voltage amplitude information U(t0~t end ).

[0094] According to the obtained pulse amplitude information of the three types of physical signals and the external voltage amplitude information, the physical signal pulse amplitude original vector and the voltage original vector are obtained;

[0095] The original vector of the physical signal pulse amplitude information is as follows:

[0096]

[0097] Where, I HFCT (i,1) is the original vector of high-frequency current pulse amplitude information; I UHF (i,1) is the original vector of UHF pulse amplitude information; I AE (i,1) is the original vector of ultrasonic pulse amplitude information; I HFCT (t0~t end ) is t0~t end High-frequency current pulse amplitude information; I UHF (t0~t end ) is t0~t end UHF pulse amplitude information; I AE (t0~t end ) is t0~t end Ultrasonic pulse amplitude information;

[0098] Construct the original matrix I of pulse amplitude information:

[0099] I=(I HFCT , I UHF , I AE ) (5)

[0100] The original vectors of each signal voltage are as follows:

[0101]

[0102] Where U HFCT (i,1) is the original voltage vector of the high-frequency current signal; U UHF (i,1) is the original voltage vector of the UHF signal; U AE (i,1) is the original vector of ultrasonic signal voltage; U HFCT (t0~t end ) is t0~t end The information of the applied voltage amplitude of the high-frequency current; U UHF (t0~t end ) is t0~t end UHF external applied voltage amplitude information; U AE (t0~t end ) is t0~t end Ultrasonic applied voltage amplitude information;

[0103] Construct the voltage original matrix U:

[0104] U=(U HFCT , U UHF , U AE ) (7)

[0105] Construct the original matrix A0 of multi-physics information:

[0106] A0=(I,U) (8)

[0107] Based on the obtained original matrix of multi-physics information, eigenvalue calculation is performed to construct the multi-physics information characteristic matrix. The specific steps are as follows:

[0108] S2.1: Feature extraction: Based on the characteristics of high-frequency current, ultra-high frequency, and ultrasonic signals, three features are extracted: average pulse intensity, pulse repetition rate, and average pulse energy. Three features are extracted from ultrasonic signals: average pulse intensity, pulse repetition rate, and average pulse dispersion.

[0109] S2.2: Multi-physical characteristic matrix construction, constructing characteristic vectors according to the eigenvalues of the three types of signals, extracting multiple groups of characteristic vectors based on multiple groups of data to form a characteristic matrix.

[0110] The calculation formulas for the characteristic values of the three types of physical signals are as follows:

[0111] The formula for calculating the average pulse intensity is:

[0112]

[0113] Where I(i,1) is the pulse information collected by a single type of physical signal within the acquisition time range, and N is the total number of pulses;

[0114] The pulse repetition rate calculation formula is:

[0115]

[0116] Where, T is the acquisition time;

[0117] In addition, for high-frequency current and ultra-high-frequency signals, the pulse energy average value feature can also be extracted. The single pulse energy calculation formula is as follows:

[0118]

[0119] The formula for the average pulse energy is:

[0120]

[0121] For ultrasonic signals, the pulse dispersion average feature can be extracted. The dispersion calculation formula of a single cluster ultrasonic signal is as follows:

[0122] H i =-∑P(I)logP(I) (13)

[0123] The calculation formula for the average value of pulse dispersion is as follows:

[0124]

[0125] Three types of physical signals can each obtain three eigenvalues, for a total of nine eigenvalues. The eigenvectors constructed based on the eigenvalues are shown below:

[0126]

[0127] a=(a HFCT , a UHF , a AE ) (16)

[0128] Where a HFCT is the characteristic vector of high-frequency current; a UHF is the characteristic vector of UHF; a AE is the ultrasonic eigenvector; a is the characteristic matrix of a single set of signal data;

[0129] When multiple sets of data are collected to form multiple original matrices, a feature matrix A can be formed after feature extraction. C :

[0130]

[0131] Based on the Pearson correlation coefficient between features, the algorithm evaluation of multi-physics information features is performed, and significant features are selected according to the evaluation scores to construct a simplified matrix of multi-physics information. The formula for calculating the Pearson correlation coefficient between features is as follows:

[0132]

[0133] Where X and Y are any two different features. A coefficient value of 1 means that X and Y can be well described by the straight line equation. All data points fall well on the straight line, and X increases as Y increases. A coefficient value of -1 means that all data points fall on the straight line, and X decreases as Y increases. A coefficient value of 0 means that there is no linear relationship between the two variables.

[0134] According to the relevant calculation results, the multi-physical information feature simplified matrix A is constructed S :

[0135] A S =(C1, C2…C n ) (19)

[0136] Where C i is the feature vector.

[0137] The Pearson coefficient is limited to the range of (-0.5, 0.5), and the eigenvector C iBy limiting the range of the Pearson coefficient between features, the features with weak correlation are screened out, thereby streamlining the number of features and greatly reducing the calculation and analysis time.

[0138] Those skilled in the art will appreciate that the present invention is not limited to the foregoing embodiments. The foregoing embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a partial discharge characteristic matrix based on multi-physics information, characterized in that: The method comprises the following steps: S1: Obtain multi-physics information of partial discharge through various means and construct the original matrix of multi-physics information; S2: Based on the original matrix, analyze and extract features of various signals to construct a feature matrix of multi-physics information; S3: Perform algorithm evaluation on multi-physics information features, select significant features based on the evaluation scores, and construct a simplified matrix of multi-physics information features; The step S1 is specifically as follows: S1.1: Multi-physics information acquisition: Acquire high-frequency current, ultra-high frequency, and ultrasonic signals generated by partial discharge. These three types of physical signals are synchronously acquired along with device voltage signals through pulse triggering. The acquisition time is 10 seconds. S1.2: Pulse and voltage amplitude information acquisition: Using the classic threshold method, the start and end times of the three types of physical signal pulses, as well as the pulse amplitude information and applied voltage amplitude information between the start and end times, are acquired based on the thresholds. S1.3: Constructing the original matrix of multi-physics information. Record the time, pulse amplitude information, and applied voltage amplitude information of all pulses within the acquisition time of 10 seconds. Based on the pulse amplitude information and applied voltage amplitude information of high-frequency current, ultra-high frequency, and ultrasound, construct the pulse amplitude information original matrix and voltage original matrix, and finally construct the original matrix of multi-physics information. The step S2 is specifically as follows: S2.1: Feature extraction: Based on the characteristics of high-frequency current, ultra-high frequency, and ultrasonic signals, three features are extracted: average pulse intensity, pulse repetition rate, and average pulse energy. Three features are extracted from ultrasonic signals: average pulse intensity, pulse repetition rate, and average pulse dispersion. S2.2: Multi-physics feature matrix construction, constructing feature vectors based on the eigenvalues of the three types of physical signals, extracting multiple sets of feature vectors based on multiple sets of data to form a feature matrix; Based on the Pearson correlation coefficient between features, the algorithm evaluation of multi-physics information features is performed, and significant features are selected according to the evaluation scores to construct a simplified matrix of multi-physics information. The formula for calculating the Pearson correlation coefficient between features is as follows: Where X and Y are any two different features. A coefficient value of 1 means that X and Y can be well described by the straight line equation. All data points fall well on the straight line, and X increases as Y increases. A coefficient value of -1 means that all data points fall on the straight line, and X decreases as Y increases. A coefficient value of 0 means that there is no linear relationship between the two variables. According to the relevant calculation results, the multi-physical information feature simplified matrix A is constructed S : A S =(C1,C2…C n ) (19) Where C i is the feature vector.

2. The method for constructing a partial discharge characteristic matrix based on multi-physics information according to claim 1, characterized in that: In step S1.2, according to the threshold value I t Get the start time t0 and end time t of three types of physical signal pulses end The calculation formula is: IN t =1.1I n,max (1) Among them, I n,max is the maximum noise value of the physical signal when there is no partial discharge; The physical signal pulse start time t0 is determined based on the following: Select I i Time t i is the pulse starting time t0; Physical signal pulse end time t end The judgment is based on the following: Select I i Time t i is the pulse end time t end ; Determine the pulse start time t0 and end time t end After that, record t0 and t end The pulse amplitude information I(t0~t end ) and the external voltage amplitude information U(t0~t end ).

3. The method for constructing a partial discharge characteristic matrix based on multi-physics information according to claim 2, characterized in that: The construction of the original matrix of multi-physics information in step S1.3 is as follows: For pulse amplitude information, the original vectors of high-frequency current HFCT, ultra-high frequency UHF, and ultrasonic AE pulse amplitude information are as follows: Where, I HFCT (i,1) is the original vector of high-frequency current pulse amplitude information; I UHF (i,1) is the original vector of UHF pulse amplitude information; I AE (i,1) is the original vector of ultrasonic pulse amplitude information; I HFCT (t0~t end ) is t0~t end High-frequency current pulse amplitude information; I UHF (t0~t end ) is t0~t end UHF pulse amplitude information; I AE (t0~t end ) is t0~t end Ultrasonic pulse amplitude information; Construct the original matrix I of pulse amplitude information: I=(U HFCT ,I UHF ,I AE ) (5) The original vectors of each signal voltage are as follows: Where U HFCT (i,1) is the original voltage vector of the high-frequency current signal; U UHF (i,1) is the original voltage vector of the UHF signal; U AE (i,1) is the original vector of ultrasonic signal voltage; U HFCT (t0~t end ) is t0~t end The information of the applied voltage amplitude of the high-frequency current; U UHF (t0~t end ) is t0~t end UHF external applied voltage amplitude information; U AE (t0~t end ) is t0~t end Ultrasonic applied voltage amplitude information; Construct the voltage raw matrix U: U=(U HFCT ,IN UHF ,IN AE ) (7) Construct the original matrix A0 of multi-physics information: A0=(I,U) (8).

4. The method for constructing a partial discharge characteristic matrix based on multi-physics information according to claim 3, characterized in that: The calculation formulas for the characteristic values of the three types of physical signals are as follows: The formula for calculating the average pulse intensity is: Where I(i,1) is the pulse information collected by a single type of physical signal within the acquisition time range, and N is the total number of pulses; The pulse repetition rate calculation formula is: Where, T is the acquisition time; In addition, for high-frequency current and ultra-high-frequency signals, the pulse energy average value feature can also be extracted. The single pulse energy calculation formula is as follows: The formula for the average pulse energy is: For ultrasonic signals, the pulse dispersion average feature can be extracted. The dispersion calculation formula of a single cluster ultrasonic signal is as follows: H i =-∑P(I)logP(I) (13) The calculation formula for the average value of pulse dispersion is as follows: 。 5. The method for constructing a partial discharge characteristic matrix based on multi-physics information according to claim 4, characterized in that: The three types of physical signals can each obtain three eigenvalues, for a total of nine eigenvalues. The eigenvectors constructed based on the eigenvalues are shown below: a=(a HFCT ,a UHF ,a AE ) (16) Where a HFCT is the characteristic vector of high-frequency current; a UHF is the characteristic vector of UHF; a AE is the ultrasonic eigenvector; a is the characteristic matrix of a single set of signal data; When multiple sets of data are collected to form multiple original matrices, a feature matrix A can be formed after feature extraction. C : 。 6. The method for constructing a partial discharge characteristic matrix based on multi-physics information according to claim 5, characterized in that: The Pearson correlation coefficient is limited to the range of (-0.5, 0.5), retaining the weak correlation features, the eigenvector C i The values between them are (-0.5, 0.5).

Images