Cable partial discharge defect identification method based on support vector machine

Abstract

The invention relates to a cable partial discharge defect identification method based on a support vector machine, and belongs to the technical field of electrics. The problems that an existing method is low in recognition rate and prone to interference are solved. The method comprises the following steps: firstly, carrying out interference suppression on an acquired partial discharge high-frequency signal, wherein the interference suppression comprises narrowband interference suppression by subspace reconstruction and frequency slice wavelet transform white noise suppression based on ensemble empirical mode decomposition; then, carrying out support vector multi-classification by utilizing an optimized support vector data description algorithm, and constructing a plurality of classifiers; and finally, calculating the generalized distance from a test sample to each classifier, and making a decision, thereby achieving the recognition of the partial discharge defect. According to the method, the partial discharge defect recognition rate is effectively improved, the anti-interference capability is high, and a reliable technical means is provided for power cable insulation state detection.

Description

Technical Field

[0001] The invention belongs to the field of electrical technology and relates to a cable partial discharge defect recognition method based on a support vector machine. Background Art

[0002] The operating environment of power cables is generally harsh, such as some low tunnels and wet cable trenches. In such harsh environments, the cable is exposed to moisture and water seepage, which further increases the possibility of its insulation performance deterioration. In addition, human factors are also factors that cannot be ignored in causing cable failures. In short, the causes of cable failures can be summarized as follows: (1) natural aging of insulation; (2) damage by external forces; (3) the influence of insulation defects and design and manufacturing processes; (4) overvoltage. For example, during the manufacturing of cables, due to factors such as poor process quality, impurities and bubbles may be left in the cable insulation, or some micro defects that are invisible to the naked eye may be caused by the carelessness of installation and maintenance personnel. Under the action of high voltage, the electric field distribution of these cable insulation defects will be severely distorted and significantly enhanced. When the dielectric strength of the insulating medium is exceeded, discharge will occur in the local insulation area, but it will not cause breakdown immediately. That is, local discharge will occur inside the cable, but with the cumulative effect, the insulation at the defect will further deteriorate, which will eventually promote the failure of the main insulation of the cable terminal and cause breakdown, which will cause serious cable transmission line operation failures.

[0003] Partial discharge defect identification is one of the important means of detecting the insulation status of power cables. It has a strong ability to detect the initial defects in cable insulation. Different partial discharge defects are interrelated with the quality of the insulation status of power cables. The degree of deterioration of the insulation material can be preliminarily determined by the detected cable partial discharge defect category. The implementation of partial discharge defect detection on power cables enables the staff to timely understand and discover the potential hidden dangers of power cables, and grasp the insulation status of power cables in different operating periods, so as to take corresponding measures, realize more reasonable and planned maintenance, reduce the economic losses caused by power cable failures, and achieve the purpose of reliable power supply. Therefore, it is extremely important to identify partial discharge defects in cables.

[0004] In partial discharge defect recognition, pattern recognition technology can classify and identify different discharge types by adjusting various parameters. The construction of partial discharge pattern classifier is one of the core issues of pattern recognition.

[0005] The most commonly used classifiers are support vector machine classifiers, distance-based pattern classifiers, and neural network classifiers. (1) Support Vector Machine (SVM) classifier: Support vector machine optimizes learning accuracy and generalization ability in a limited number of training samples to achieve the best performance, and is therefore widely used in various fields. (2) Distance-based pattern classifier: Distance-based pattern classifier determines which type the sample to be identified belongs to based on the size of the interval between the sample to be identified and the training sample. The smaller the interval between the two, the higher the degree of fit. (3) Neural network classifier: A neural network classifier has a large number of neurons, and each neuron is connected by weights.

[0006] The existing methods for identifying cable partial discharge defects still have the following problems and difficulties. (1) There are many sources of interference in the power system, including other equipment, electromagnetic interference, and the operating noise of the equipment itself; (2) Processing the large amount of data generated during high-frequency signal acquisition and identifying partial discharge signals requires more efficient algorithms to reduce the false alarm rate and ensure high-accuracy identification; (3) Cable partial discharge signals have complex nonlinear characteristics. When selecting kernel functions, it is difficult to determine the most suitable kernel function type and corresponding parameters.

[0007] Aiming at the problem of partial discharge signal extraction and defect identification of power cables, this paper uses a combination of random pulse interference suppression, periodic narrowband interference suppression and white noise interference suppression to suppress the collected high-frequency partial discharge signal to improve the accuracy of partial discharge signals. By studying the basic algorithm of support vector machine, a classifier of four defect discharge types is constructed. First, a classifier of one type of defect discharge type is selected, and the discharge characteristic parameters of this type of defect discharge are obtained according to the data of this type of training samples, and these training samples are sent to the support vector machine for learning as target samples. The Gaussian radial kernel function and one-to-many decision method are used to train, learn and classify the classifier. Then, the discharge characteristic parameters of the remaining three types of discharge defects and the radius of the standard hypersphere and optimized hypersphere of each type of defect discharge are obtained respectively, and the maximum classification interval is obtained. All test sample sets are input into the classifier, and the generalized distance of each test sample set to the four types of defect discharge is calculated, and classification and identification are performed through the decision rule. Summary of the invention

[0008] In view of this, the purpose of the present invention is to provide a cable partial discharge defect identification method based on support vector machine. The interference suppression of the collected partial discharge high-frequency signal is studied, and the algorithm of support vector machine is used to select and extract feature data of cable partial discharge data to realize partial discharge defect classification and identification.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] A cable partial discharge defect recognition method based on support vector machine includes the following steps:

[0011] Step 1: performing interference suppression on the collected partial discharge high-frequency signal, wherein the interference suppression includes:

[0012] Subspace reconstruction suppresses narrowband interference;

[0013] Frequency slicing wavelet transform white noise suppression based on ensemble empirical mode decomposition (EEMD);

[0014] Step 2: Use the optimized support vector data description (SVDD) algorithm to perform support vector multi-classification;

[0015] Step 3: Train the classifier and use the decision-making method to identify partial discharge defects.

[0016] Further, the subspace reconstruction to suppress narrowband interference includes:

[0017] The original noisy signal is constructed into a HANKEL matrix and decomposed to obtain singular values;

[0018] Separate the local discharge subspace containing white noise interference and the narrowband interference subspace matrix from the distribution characteristics of singular values;

[0019] Reconstruct the narrowband interference and obtain the partial discharge signal containing white noise interference.

[0020] Further, the frequency slicing wavelet transform white noise suppression based on EEMD decomposition includes:

[0021] Perform EEMD decomposition on the local discharge subspace to obtain multiple intrinsic mode functions (IMFs) and residual functions;

[0022] Perform frequency slicing wavelet time-frequency conversion on each IMF;

[0023] The signal of each processed IMF is reconstructed to obtain the partial discharge signal after white noise suppression.

[0024] Further, the optimized support vector data description SVDD algorithm includes:

[0025] The binary tree multi-classification algorithm is used to divide multiple types of defect discharges into multiple binary classification problems;

[0026] For each binary classification problem, the SVDD algorithm is used to build a classifier;

[0027] Calculate the generalized distance from the test sample to each classifier;

[0028] Classification and recognition are performed based on generalized distance.

[0029] Furthermore, the support vector machine algorithm adopts a Gaussian radial basis kernel function.

[0030] Further, the decision-making method includes:

[0031] Calculate the generalized distance from the test sample to each classifier;

[0032] Classify the test sample into the defect discharge type represented by the closest classifier.

[0033] Furthermore, the interference suppression step is performed between step two and step three.

[0034] Furthermore, the interference suppression step is performed between step one and step two.

[0035] Furthermore, the partial discharge defect identification includes identifying at least four defective discharge types.

[0036] Furthermore, the partial discharge defect identification includes identifying at least three defective discharge types.

[0037] The beneficial effects of the present invention are:

[0038] (1) By combining random pulse interference suppression, periodic narrowband interference suppression and white noise interference suppression, various interferences in the collected partial discharge high-frequency signal can be effectively removed, the accuracy of the signal can be improved, and a reliable basis can be provided for subsequent defect identification.

[0039] (2) The optimized support vector data description (SVDD) algorithm is used to perform support vector multi-classification and construct multiple classifiers, which can effectively distinguish different types of partial discharge defects and significantly improve the defect recognition rate.

[0040] (3) By suppressing interference of local discharge high-frequency signals and optimizing the SVDD algorithm, the present invention can effectively resist the influence of various interference factors and maintain a high recognition accuracy even in complex environments.

[0041] (4) The present invention is capable of identifying multiple types of partial discharge defects, including but not limited to four types, and provides a comprehensive detection method for power cable insulation status detection.

[0042] (5) The present invention can efficiently process local discharge high-frequency signals and perform defect identification, which helps to improve detection efficiency and reduce detection costs.

[0043] (6) By timely identifying cable partial discharge defects, cable failures can be effectively prevented and the safe and stable operation of the power system can be ensured.

[0044] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0046] Figure 1 It is a flow chart of the present invention;

[0047] Figure 2 is the interference suppression flow chart;

[0048] Figure 3 It is a pattern recognition flow chart;

[0049] Figure 4 Schematic diagram of the optimal classification line; (a) is a schematic diagram showing that the distance between the split dotted line and the solid line is large; (b) is a schematic diagram showing that the distance between the split dotted line and the solid line is small;

[0050] Figure 5 Schematic diagram of improved SVDD. DETAILED DESCRIPTION

[0051] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0052] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0053] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0054] like Figure 1 As shown, the present invention includes the following parts:

[0055] The interference suppression of partial discharge high-frequency signals is achieved through random pulse interference suppression, periodic narrow-band interference suppression and white noise interference suppression;

[0056] Support vector multi-classification is performed using the optimized support vector data description (SVDD) algorithm;

[0057] Train the classifier and use the decision method to identify partial discharge defects.

[0058] 1. Interference suppression method in partial discharge signal

[0059] According to the characteristics of various interference signals, random pulse interference is usually strong, low frequency and easy to identify. Periodic narrowband interference has a high degree of recognition in the frequency domain. White noise interference is a broadband interference random signal, which has similar characteristics to partial discharge in the frequency domain. Therefore, the following process can be used step by step in actual interference suppression, as follows: Figure 2 shown.

[0060] 1.1 Subspace reconstruction to suppress narrowband interference

[0061] Subspace reconstruction to suppress narrowband interference is to construct a HANKEL matrix from the original noisy signal and decompose it to get singular values. From the distribution characteristics of the singular values, the local discharge subspace containing white noise interference and the narrowband interference subspace matrix can be separated, and then the narrowband interference is reconstructed to obtain the local discharge signal containing white noise interference. The method of constructing the HANKEL matrix is ​​as follows: Assuming that the target signal sequence is F, it can be expressed as:

[0062] F={f(1)f(2)f(3)f(4)f(5)…f(N)}(1)

[0063] Where N is the total number of sampling points, and the HANKEL matrix is ​​constructed from the target signal F:

[0064]

[0065] In the formula, p+q=N+1, and the value range of q is

[0066] According to the obtained HANKEL matrix, not only can the narrowband interference be removed, but also the narrowband interference frequency can be calculated, which will be calculated separately below. The narrowband interference frequency is determined using a matrix bundle algorithm, which represents the target signal as a linear combination of M exponents.

[0067]

[0068] Where X(kT s ) represents a periodic narrowband interference signal, n(kT s ) represents the partial discharge signal with white noise interference, R i is the complex amplitude of periodic narrowband interference, α i is the attenuation parameter, ω i is the angular frequency; M is the target partial discharge signal order; T s is the sampling time interval of the oscilloscope. Decompose the constructed HANKEL matrix.

[0069] H=USV T (4)

[0070] Based on the decomposition results, sort the singular values ​​and select M larger singular values ​​σ in the diagonal matrix S. i The column vectors in the matrix S1, and the corresponding singular phasors in V are composed of V′=(ν1,ν2,ν3,…,ν M ), that is, V1 is V′ minus the first row, and V2 is V′ minus the last row.

[0071]

[0072] Assume that the generalized eigenvalue of Y2-λY1 is z i To express it, the narrowband interference frequency can be obtained by the following formula.

[0073]

[0074] The HANKEL matrix is ​​used to remove narrowband interference. First, the first 12 singular values ​​are determined as the narrowband interference subspace according to the size of the singular values, and the rest are the combined subspace of partial discharge and white noise interference. The inverse process of singular value decomposition is performed on the remaining subspace, that is, the denoised narrowband interference signal. Since the inverse process is not strictly HANKEL, the average value of the obtained signal at the same time can be taken as the signal value at that time point.

[0075] 1.2 Frequency Slicing Wavelet Transformation White Noise Suppression Based on EEMD Decomposition

[0076] The empirical mode decomposition (EMD) method is suitable for analyzing and processing non-stationary and nonlinear signals, but this method has problems and shortcomings, mainly: ① The IMF (intrinsic mode function) obtained by EMD decomposition has mode aliasing; ② The terminal effect affects the decomposition effect. In order to suppress the mode aliasing phenomenon of EMD, Handrin et al. in France used EMD to perform statistics on the results of white noise decomposition and proposed an improved EMD method based on noise-assisted analysis, namely, Ensemble Empirical Mode Decomposition (EEMD).

[0077] The essence of the EEMD method is a multiple empirical mode decomposition of superimposed Gaussian white noise. It uses the statistical characteristics of Gaussian white noise with uniform frequency distribution, and changes the extreme point characteristics of the signal by adding different white noises of the same amplitude each time. Then, the corresponding MF obtained by multiple EMDs is averaged to offset the added white noise, thereby effectively suppressing the generation of modal aliasing. The EEMD decomposition steps are as follows:

[0078] (1) Set the overall average number M;

[0079] (2) A white noise n with a standard normal distribution i (t) is added to the original signal x(t) to produce a new signal: x i (t) = x(t) + n i (t), where n i (t) represents the ith additive white noise sequence, x i (t) represents the additional noise signal of the i-th trial i=1,2,…M;

[0080] (3) For the obtained noisy signal x i (t) Perform EMD decomposition respectively to obtain the form of their respective IMF sums:

[0081]

[0082] In the formula, ci,j (t) is the jth IMF obtained by decomposing after adding white noise for the i-th time, r i,j (t) is the residual function, representing the average trend of the signal, and J is the number of IMFs;

[0083] (4) Repeat steps (2) and (3) M times, adding white noise signals with different amplitudes each time to obtain the set of IMFs:

[0084] c 1,j (t)c 2,j (t)…c M,j (t),j=1,2,…J(8)

[0085] (5) Based on the principle that the statistical mean of unrelated sequences is zero, the corresponding IMFs are averaged to obtain the final IMF after EEMD decomposition, namely:

[0086]

[0087] In the formula, c j (t) is the jth IMF decomposed by EEMD, i = 1, 2, ... M, j = 1, 2, ... J. Denoising is performed by performing frequency slicing wavelet time-frequency conversion based on EEMD decomposition, and finally the processing results of each component are reconstructed to obtain the partial discharge signal after suppressing white noise.

[0088] 2. Partial discharge defect pattern recognition method based on support vector machine

[0089] The pattern recognition system consists of four parts: data collection, data processing, feature selection, and classification decision. Figure 3 shown.

[0090] 2.1 Support Vector Machine Algorithm Principle

[0091] The fundamental idea of ​​support vector machines is to find an optimal classification line, which is extended to high-dimensional space as the optimal classification surface. The optimal classification line mentioned above is to accurately separate the two types and minimize the probability of misclassification. The best case is to correctly classify all the categories so that the classification interval reaches the maximum value. Figure 4 As shown in the figure, the solid line in the middle is called the splitting hyperplane, and the points passed by the two dotted lines are the support vectors. From the figure, we can see that the distance between the splitting dotted line and the solid line in (a) is relatively large, while the distance between the splitting dotted line and the solid line in (b) is relatively small. What the support vector machine is looking for is to maximize the interval.

[0092] There are many nonlinear problems that need to be solved in pattern recognition. SVM first maps the input space to a high-dimensional space by defining the nonlinear mapping of the kernel function, and solves the optimal boundary surface in this high-dimensional space. In the process of solving the optimal boundary hyperplane, it is necessary to adopt the idea of ​​structural risk minimization (SR) principle to improve the ability of machine learning. By adjusting the parameters of the kernel function, the empirical risk and the confidence range are optimized at the same time, so as to obtain the minimum real risk.

[0093] The binary classification principle of SVM is as follows: By introducing the input space R n The transformation Φ to the high-dimensional Hilbert space transforms the data set (x i ,y i ),i=1,…,n,x i ∈R d ,y i ∈{1,-1} is mapped to φ((x1),y1),…,φ((x i ),y i ), the original problem is as follows:

[0094]

[0095] sty i ((ω·x i )+b)≥1-ξ i ,i=1,2,…,n(11)

[0096] ξ i ≥0,i=1,2,…,n(12)

[0097] Where C>0 is called the penalty factor, which indicates the degree of penalty for misclassified samples. The larger the value of C, the greater the degree of penalty. ω is the normal vector of the optimal classification hyperplane, b is the offset parameter, and ξ i is the relaxation factor of the i-th sample. For the kernel function K(x i ,y i ), by introducing Lagrange multipliers and optimality conditions, the original problem can be transformed into a dual problem:

[0098]

[0099] C≥a i ≥0, i=1,…,n (15)

[0100] Where: a i is the Lagrange multiplier.

[0101] If K is a positive definite kernel function, then the dual problem has a solution, and its optimal solution is Then we can get ω * b * :

[0102]

[0103] Among them: The coefficient αi only has the corresponding data sample point (x i ,y j ) can be non-zero only if the equality of equation (11) holds. Thus, the optimal classification decision function can be obtained:

[0104]

[0105] 2.2 Selection and Adjustment of Kernel Function

[0106] In the actual detection of partial discharge, due to the differences in different types of discharge defect models and starting discharge voltages, the sample sets of the original space of feature quantities corresponding to different types of defects will appear nonlinear, high-dimensional and complex. Conventional SVM models cannot handle such problems. In order to solve such problems and effectively diagnose different types of defect discharge, kernel functions are introduced to solve the problem of high-dimensional nonlinearity. The function K(x i ,x j ) can transform the original low-dimensional space to which the partial discharge sample x belongs to into another high-dimensional space, and find the corresponding minimum hypersphere in the high-dimensional space, so that the original nonlinear sample classification problem can be transformed into a high-dimensional linear classification problem. And the kernel function appears in the form of the scalar product of the mapping function, so the form of the mapping function is not required, and there is no need for its explicit form, which directly solves the nonlinear problem in the input space process. Select the kernel function K(x i ,x j ) replaces the inner product operation in the decision function (x i ·x j ). We can get:

[0107]

[0108] The above formula can be optimized to calculate the support vector of the classification hypersphere and the final decision function. Then the distance from the sample to be tested to the center of the sphere is calculated, and the samples of different types of defect discharge can be classified.

[0109] The required kernel function can be determined according to the smoothness of the classifier's decision function.

[0110] Gaussian radial basis kernel function:

[0111]

[0112] The Gaussian radial kernel function can handle infinite-dimensional feature space. Therefore, for partial discharge samples of different types of defect discharge, although they are nonlinear, they can be linearized in the feature space under this kernel function. And in terms of parameter selection, only one parameter value needs to be determined. Therefore, this paper chooses this kernel function to optimize the sample feature data.

[0113] 2.3 Support Vector Optimization Algorithm

[0114] According to the charge movement characteristics of partial discharge and the dispersion of partial discharge samples, the SVDD algorithm has randomness in the training of discharge defect type samples, resulting in inconsistent SV hyperspheres. This causes similar samples to appear on both sides of the classification sphere, resulting in misclassification when identifying and classifying partial discharge types. In order to solve this problem, we modify the SVDD model and use the OR-SVDD algorithm suitable for cable partial discharge defect type identification. Figure 5 shown.

[0115] In order to solve the multi-class separable problem, this paper adopts a binary tree multi-classification algorithm, taking one class of defect discharge as one class, and the remaining k-1 (k>2) classes of defect discharge as one class; then select the k-1 class of defect discharge in the second class as a separate class, and the remaining k-2 classes of defect discharge are divided into one class, and classify them in turn.

[0116] Subsequently, a classifier was constructed for the four types of partial discharge defects to diagnose and identify different feature parameters. The maximum classification interval was calculated based on the selected kernel function. All the test sample sets were then input into the classifier, and the generalized distance from each test sample set to the four types of defect discharges was calculated. Classification and identification were performed using the decision rule.

[0117] The above-mentioned partial discharge defect recognition method based on support vector machine first realizes the interference suppression of partial discharge high-frequency signals by combining three methods: random pulse interference suppression, periodic narrowband interference suppression and white noise interference suppression. Then, the support vector multi-classification is performed by using the optimized support vector data description (SVDD) algorithm; finally, the classifier is trained and the decision method is used to perform partial discharge defect recognition.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A cable partial discharge defect recognition method based on support vector machine, characterized in that: The following steps are involved: Step 1: performing interference suppression on the collected partial discharge high-frequency signal, wherein the interference suppression includes: Subspace reconstruction suppresses narrowband interference; Frequency slicing wavelet transform white noise suppression based on ensemble empirical mode decomposition (EEMD); Step 2: Use the optimized support vector data description (SVDD) algorithm to perform support vector multi-classification; Step 3: Train the classifier and use the decision-making method to identify partial discharge defects.

2. The cable partial discharge defect identification method based on support vector machine according to claim 1 is characterized in that: The subspace reconstruction to suppress narrowband interference includes: The original noisy signal is constructed into a HANKEL matrix and decomposed to obtain singular values; Separate the local discharge subspace containing white noise interference and the narrowband interference subspace matrix from the distribution characteristics of singular values; Reconstruct the narrowband interference and obtain the partial discharge signal containing white noise interference.

3. The cable partial discharge defect identification method based on support vector machine according to claim 1 is characterized in that: The frequency slicing wavelet transform white noise suppression based on EEMD decomposition includes: Perform EEMD decomposition on the local discharge subspace to obtain multiple intrinsic mode functions (IMFs) and residual functions; Perform frequency slicing wavelet time-frequency conversion on each IMF; The signal of each processed IMF is reconstructed to obtain the partial discharge signal after white noise suppression.

4. The cable partial discharge defect identification method based on support vector machine according to claim 1 is characterized in that: The optimized support vector data description (SVDD) algorithm includes: The binary tree multi-classification algorithm is used to divide multiple types of defect discharges into multiple binary classification problems; For each binary classification problem, the SVDD algorithm is used to build a classifier; Calculate the generalized distance from the test sample to each classifier; Classification and recognition are performed based on generalized distance.

5. The cable partial discharge defect identification method based on support vector machine according to claim 1 is characterized in that: The support vector machine algorithm adopts a Gaussian radial basis kernel function.

6. The cable partial discharge defect identification method based on support vector machine according to claim 1 is characterized in that: The decision-making method includes: Calculate the generalized distance from the test sample to each classifier; Classify the test sample into the defect discharge type represented by the closest classifier.

7. The cable partial discharge defect identification method based on support vector machine according to claim 1 is characterized in that: The interference suppression step is performed between step two and step three.

8. The cable partial discharge defect identification method based on support vector machine according to claim 1 is characterized in that: The interference suppression step is performed between step one and step two.

9. The cable partial discharge defect identification method based on support vector machine according to claim 1, characterized in that: The partial discharge defect identification includes identifying at least four defect discharge types.

10. The cable partial discharge defect identification method based on support vector machine according to claim 1, characterized in that: The partial discharge defect identification includes identifying at least three defect discharge types.

Images