A cable defect monitoring method and system based on harmonic data

Abstract

The application provides a cable defect monitoring method and system based on harmonic data, and the method comprises the following steps: acquiring cable harmonic data; inputting the cable harmonic data into a first model to obtain a first array; inputting the cable harmonic data into a second model to obtain a second array; performing calculation based on the first array and the second array to obtain a third array; taking the defect type corresponding to the maximum value in the third array as the result output, so as to realize the monitoring of the cable defect. The method and system provided by the application combine the clustering model and the classification model to monitor the defect type, can not only more accurately judge the defect existing in the distribution cable, but also can monitor other cable defect types except water tree aging, quantitatively realize the one-to-one correspondence between the cable harmonic data and the defect, make the defect identification result more reliable, and provide a basis for timely repairing the defect and ensuring the reliability of the power system.

Description

Technical Field

[0001] This invention relates to the field of cable monitoring technology, specifically to a cable defect monitoring method and system based on harmonic data. Background Technology

[0002] Cables are crucial components for power transmission and distribution, playing a vital role in power systems and forming an indispensable part of modern society. However, long-term operation and environmental factors can lead to various defects in cables, such as insulation wear, aging, and corrosion of the metal conductors. Failure to detect and repair these defects promptly will reduce the reliability of the power system, potentially causing malfunctions or even serious safety accidents.

[0003] Harmonic feature detection technology, as an emerging method for cable defect detection, is based on the analysis of the correlation between cable harmonic signals and defects. When a cable has defects, such as insulation wear or aging, the defect will cause nonlinear changes in the current and voltage in the cable, resulting in the generation of harmonic signals. These harmonic signals appear in the current and voltage waveforms of the cable system with different frequency components. By analyzing the spectrum and amplitude changes of the harmonic signals, cable defects can be accurately located and analyzed.

[0004] Currently, existing technologies have only studied the correlation between the third harmonic signal of a cable and water tree aging. More specifically, the components of the third harmonic signal of a cable are closely related to the length of the water tree, the breakdown strength, etc. In other words, it is only possible to determine whether water tree aging is present based on the harmonic signal, as well as the length of the water tree and the breakdown strength.

[0005] However, existing technologies do not study the harmonic signals corresponding to other cable defects, such as water treeing, residual insulation tape on the crimped connector surface, semi-conductive suspension on the insulation surface, metallic suspension on the insulation surface, and metal tip discharge. In other words, it is impossible to identify cable defect types other than water treeing and aging based on harmonic signals. Summary of the Invention

[0006] This invention provides a cable defect monitoring method and system based on harmonic data, which solves the problem that existing technologies cannot identify cable defect types other than water treeing and aging based on harmonic signals.

[0007] On one hand, the present invention provides a cable defect monitoring method based on harmonic data, the method comprising: Acquire cable harmonic data; The cable harmonic data is input into the first model to obtain the first array; The cable harmonic data is input into the second model to obtain the second array; A third array is obtained by performing calculations based on the first and second arrays. The defect type corresponding to the largest value in the third array is output as the result to achieve cable defect monitoring.

[0008] Furthermore, the first array is a distance array between the cable harmonic data and the cluster center of each defect in the first model; the second array is a probability array of the cable harmonic data belonging to each defect type in the second model.

[0009] Furthermore, the third array is the ratio of the values ​​in the second array to the corresponding values ​​in the first array.

[0010] Furthermore, the cable harmonic data includes: fundamental wave data, multiple harmonic data, actual current intensity, and actual measurement location.

[0011] Furthermore, the construction of both the first and second models includes obtaining a sample dataset; The acquisition of the sample dataset includes: Obtain a test harmonic dataset; wherein, the test harmonic dataset includes multiple defect datasets, one of which corresponds to a cable defect type or a cable defect type; each defect dataset includes multiple harmonic component data, and each harmonic component data corresponds to a test current intensity and a test measurement location; A sample dataset is obtained based on multiple defect datasets, as well as the test current intensity and test measurement location corresponding to the harmonic component data.

[0012] Furthermore, the construction of the first model is based on the Kmeas clustering model, including: The sample dataset was re-divided using the Kmeas clustering model to obtain the first clustering result; The first model is obtained by removing noisy data from the first clustering result.

[0013] Furthermore, the second model is constructed based on the XGBoost multi-class classification model, including: Defect tags are obtained based on the type of cable defect and multiple cable defect types, or based on the type of defect cluster centers in the first model; and the defect tags are then subjected to one-hot encoding to obtain sample tags. The filtered sample dataset is divided into a training set and a validation set according to a set ratio; The XGBoost multi-class classification model is trained using the training set, and the parameters of the XGBoost multi-class classification model are adjusted using the validation set until the model converges, thereby obtaining the second model.

[0014] Furthermore, the types of cable defects include one or more of the following: water tree defects, residual insulation tape defects on the surface of the crimped tube, semiconductive suspension defects on the insulation surface, metallic suspension defects on the insulation surface, and metal tip discharge defects.

[0015] Furthermore, the step of re-dividing the sample dataset includes: Each defect dataset is randomly selected to obtain multiple cluster centers corresponding to the defect type; and all unselected sample points in the defect dataset are marked as test points. Calculate the distance between the current test point and each of the cluster centers, and classify the current test point based on the distance; Based on the classification calculation of the classified sample points and test points, the cluster centroid vector value is obtained; The cluster centroid vector value is compared with the cluster center. If the cluster center changes, the distance between the test point and each cluster center is recalculated. The process continues until the cluster center no longer changes, and the first clustering result is obtained.

[0016] On the other hand, the present invention also provides a cable defect monitoring system based on harmonic data, the system comprising at least a plurality of data acquisition modules and a plurality of data processing modules, performing the steps of any of the methods described above.

[0017] In general, the technical solution conceived in this invention can achieve the following beneficial effects compared with the prior art: This invention provides a cable defect monitoring method and system based on harmonic data. On the one hand, it combines clustering model and classification model to identify defects in power distribution cables. Finally, it combines clustering model and classification model to further process the results, thereby more accurately judging the defects existing in the power distribution cables and making the defect identification results more reliable.

[0018] Secondly, by artificially creating common defects in power distribution cables and applying different currents to different locations on single-core cables, N independent tests are conducted to collect harmonic signals at the cable terminals, obtaining complete harmonic variation data of the cable defect evolution process. Then, cosine similarity calculations are performed on the data of different types of defects to achieve clustering, and the data is filtered after comparison with the original data to obtain stable data and a robust clustering model. Further judgment is made based on the distance from the clustering model output to each defect, not only obtaining more accurate cable defect type results but also monitoring other cable defect types besides water tree aging. This quantitatively realizes a one-to-one correspondence between cable harmonic data and defects, providing a basis for timely defect repair and ensuring the reliability of the power system. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the process flow of a cable defect monitoring method and system based on harmonic data provided by the present invention; Figure 2 This is a schematic diagram of the process for re-dividing the sample dataset of a cable defect monitoring method and system based on harmonic data provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Furthermore, the technical features involved in the various embodiments described below can be combined with each other as long as they do not conflict with each other.

[0022] It should be noted that, in the description of the embodiments of the present invention, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, or system. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, or system that includes said element.

[0023] On the one hand, this invention provides a cable defect monitoring method based on harmonic data, such as... Figure 1 As shown, the method includes: Step 101: Obtain cable harmonic data.

[0024] Specifically, a power quality analyzer is used to collect primary cable harmonic data, which includes: fundamental wave data, multiple harmonic data, actual current intensity, and actual measurement location.

[0025] Since multiple harmonic data are obtained based on fundamental frequency data, and the higher the order of the harmonic data, the weaker its relationship with the fundamental frequency data, harmonic data of the 2nd to 11th orders are preferred. As an embodiment of the present invention, multiple harmonic data... The 2nd to 11th harmonics are represented as Therefore, the harmonic data of a cable Represented as: ;in, Indicates fundamental frequency data, Represents multiple harmonic data. Indicates the actual current intensity; Indicates the actual measurement location.

[0026] Step 102: Input the cable harmonic data into the first model to obtain the first array; input the cable harmonic data into the second model to obtain the second array.

[0027] The first array represents the distance between the cable harmonic data and the cluster centers of each defect type in the first model; the second array represents the probability array of the cable harmonic data belonging to each defect type in the second model. In other words, the first model is called... Second Model Predict the presence and type of defects in cables.

[0028] Specifically, cable harmonic data Input into the first model, calculate The distances to each defect cluster center in the first model are used to obtain the first array, denoted as: ;in For the cable harmonic data that needs to be tested to defects distance, The smaller the value, the more likely the cable is defective. The greater the chance, the higher the probability.

[0029] At the same time, cable harmonic data The input is fed into the second model, and the second array is calculated, which is: ;in, The cable harmonic data that needs to be tested is defective. The probability, The larger the value, the more likely the cable is to be defective. The greater the chance, the better.

[0030] It should be noted that the cable harmonic data contains the same content as each sample data in the sample dataset used in the first and second models. Specifically, if each sample data includes the test fundamental wave data, multiple test harmonic data, test current intensity, and test measurement location, then the cable harmonic data also includes the fundamental wave data, multiple harmonic data, actual current intensity, and actual measurement location; and the harmonic data has the same order as the test harmonic data.

[0031] The first model is built based on the Kmeas clustering model, which means that the sample dataset is input into the Kmeas clustering model for training and optimization; the second model is built based on the XGBoost multi-class classification model, which means that the sample dataset is input into the XGBoost multi-class classification model for training and optimization.

[0032] The construction of both the first and second models involves obtaining a sample dataset; the acquisition of the sample dataset includes: First, the experimental harmonic dataset is obtained. Specifically, the cable undergoes corresponding defect treatment to simulate various cable defects. Harmonic data is collected by applying different intensities of voltage to the treated cable and the defect-free cable at different measurement locations. In other words, different levels of voltage are applied to the cable using a variable frequency series resonant withstand voltage test device to deepen the defect development and obtain the experimental harmonic dataset.

[0033] It should be noted that different current intensities were applied to cables with various types of defects, as well as to cables without defects, and harmonic data were collected at different locations. The test harmonic dataset includes multiple defect datasets, with one defect dataset corresponding to either a cable defect type or a specific cable defect type. Each defect dataset includes multiple harmonic component data, with each harmonic component data corresponding to a test current intensity and a test measurement location.

[0034] That is, the defect dataset includes Each defect dataset includes [various types]. Data for harmonic components. Among them, , Indicates the number of types of test current intensities; Indicates the number of types of test measurement locations; Indicates the number of independent trials; and, , , , , All are positive integers.

[0035] As an embodiment of the present invention, five typical defects were created: water tree defect, residual insulation tape defect on the surface of the crimped tube, semiconductive suspension defect on the insulation surface, metal suspension defect on the insulation surface, and metal tip discharge defect.

[0036] More specifically, these five typical methods of creating defects include: Grooves are cut into the cable joint insulation, then water is injected into the grooves during the joint waterproofing process. Finally, the joint is submerged in water to simulate water tree defects, thus obtaining the first type of defect. ; When fabricating cable joints, a second type of defect is obtained by simulating residual defects in the insulation tape on the crimped tube surface by wrapping insulating adhesive around the surface of the crimped tube. ; By simulating the semi-conductive suspension defect on the insulation surface by leaving two inverted triangular semi-conductive shielding layers on the main insulation surface of the cable joint, a third type of defect was obtained. ; During joint fabrication, metal debris was sprinkled onto the main insulation of the intermediate joint to simulate metal suspension defects on the insulation surface, resulting in the fourth type of defect. ; When making cable joint crimping tubes, a small-diameter copper wire is placed in the crimping tube for crimping, leaving about five centimeters of the copper wire wrapped together with the main insulation to simulate a metal tip discharge defect, thus obtaining the fifth type of defect. .

[0037] Different current intensities were applied to the five different types of defects mentioned above, as well as to defect-free cables (a total of six types). and in different locations Harmonic data was acquired using the harmonic data acquisition unit of a power quality analyzer. The test current intensities were set to three types: A, B, and C, denoted as follows: , , The test measurement locations are set in two ways: one is the location containing the armor. and positions without armor .

[0038] Specifically, for Category I defective cables Defect dataset The acquisition is as follows: Apply current In the armored position The harmonic current signal of the cable is measured, and the harmonic component data sampled by the power quality analyzer is calculated. Apply current In the armored position The harmonic current signal of the cable is measured, and the harmonic component data sampled by the power quality analyzer is calculated. Apply current In the armored position The harmonic current signal of the cable is measured, and the harmonic component data sampled by the power quality analyzer is calculated. Apply current In the position without armor The harmonic current signal of the cable is measured, and the harmonic component data sampled by the power quality analyzer is calculated. Apply current and in the position without armor The harmonic current signal of the cable is measured, and the harmonic component data sampled by the power quality analyzer is calculated. Apply current and in the position without armor The harmonic current signal of the cable is measured, and the harmonic component data sampled by the power quality analyzer is calculated. .

[0039] In this way, the above data collection operation is repeated at different times. Next, the first type of defect was obtained. Defect dataset from sub-independent trials .

[0040] Similarly, we obtain the first... Types of defects Defect dataset from sub-independent trials And defect-free cables Defect dataset from sub-independent trials ;in, The serial number indicating the defect type. .

[0041] In this way, multiple defect datasets are obtained, each corresponding to either no cable defect type or one cable defect type, which is also the experimental harmonic dataset. .

[0042] Next, a sample dataset was obtained based on multiple defect datasets, as well as the test current intensity and test measurement location corresponding to the harmonic component data.

[0043] For the first type of defect dataset ,include One data point. That is to say, The sample size for each trial is [number]. Data for each harmonic component; each harmonic component data includes 11 columns of features, namely fundamental frequency data. and 2nd to 11th harmonic data ( , (is a positive integer).

[0044] It should be noted that the cable harmonic data contains the same information as each harmonic component data (sample data) in the sample dataset here, and the harmonic order collected in the test is based on the harmonic order in the sample data.

[0045] Fundamental wave data were obtained based on the first type of defect dataset. and multiple harmonic data The test current intensity and test measurement location, which correspond to the harmonic component data, are then concatenated sequentially after the harmonic component data to obtain the sample dataset.

[0046] As an embodiment of the present invention, the current intensity , or and measurement location or The two fields are concatenated in the fundamental frequency data. and multiple harmonic data Next, we obtained the first type of sample dataset. , One of the samples for: Each sample includes 13 columns of features, namely the fundamental frequency data. 2nd to 11th harmonic data Test current intensity and test measurement location .

[0047] Similarly, other types of defect datasets are as follows: Thus, the sample dataset is obtained. ;in, Indicates the number of defect types.

[0048] The sample dataset is fed into the Kmeas clustering model for training and optimization to obtain the first model. Specifically, the construction of the first model is based on the Kmeas clustering model, including: firstly, using the Kmeas clustering model to re-divide the sample dataset to obtain the first clustering result; then, removing noisy data from the first clustering result to obtain the first model.

[0049] Sample dataset Equivalent to initial clustering To ensure the validity of the collected data, it is necessary to remove noisy data. The mean clustering model further divides the sample datasets of various defects. The dataset was re-divided, with the number of data points in the sample dataset being [number missing]. One, of which ,in, , , , All are positive integers.

[0050] It should be noted that for a given sample dataset Two requirements need to be met.

[0051] First, the sum of squared distances from each sample point to the center of its cluster needs to be minimized after clustering. That is, for all... There are several types: ;in, Indicates the first A cluster set, Indicates the cluster center; Indicates the first Clusters The number of samples included in it.

[0052] Second, the selection of cluster centers should make the criterion function Extremely small. That is to say, making The extreme value is the smallest. ,therefore, Therefore, ,Right now The cluster center of a class should be selected as the mean of the samples in that class.

[0053] Furthermore, such as Figure 2 As shown, the steps for re-dividing the sample dataset include: For each defect dataset, one sample point is randomly selected to obtain multiple cluster centers corresponding to the defect type; and all unselected sample points in all defect datasets are marked as test points. For example, [the following is a separate, unrelated sentence:] Each cluster center is represented as: , where is denoted as a sample point and the other points are denoted as test points, and the number in parentheses is the sequence number of the iteration operation.

[0054] Calculate the distance between the current test point and each cluster center, and classify the current test point based on the distance. That is, calculate the distance between each test point in the sample dataset and each cluster center. The distance to each cluster center is used to determine the type of the cluster center that is closest to the nearest cluster center. That is, if... ,but ;in This is the sequence number of the iteration operation. This represents the number of cluster centers.

[0055] Based on the classification calculations performed on the categorized sample points and test points, the cluster centroid vector values ​​are obtained. Then, for each category of the categorized sample set, the average value is calculated and used as the new cluster centroid vector value. ;in, is the number of samples in the j-th class.

[0056] The cluster centroid vector values ​​are compared with the cluster centers. If the cluster centers change, the distance between the test point and each cluster center is recalculated. This process is repeated until the cluster centers no longer change, yielding the first clustering result. In other words, the current clustering result is obtained. The values ​​of the centroid vectors of each cluster are respectively compared with the previous one. The cluster centroids are compared. If a cluster centroid changes, the process continues to converge, calculating new cluster centroids, until the cluster centroids no longer change, resulting in a new clustering result. .

[0057] Furthermore, the new clustering results Each defect cluster Compared with the initial cluster Each defect cluster Compare them.

[0058] for Each sample point in View the corresponding middle Does it exist; if If it does not exist, it means that the point is a noise point, and it should be removed from [the list of points]. Remove from the middle and update simultaneously. The cluster centers are determined, and this process continues until all points have been filtered out to obtain the clustering results. That is, the first model after data processing. ,at this time The samples in the middle are smaller than 10 after filtering. , recorded as .

[0059] The sample dataset is fed into the XGBoost multi-class classification model for training and optimization to obtain the second model.

[0060] Specifically, the second model is built based on the XGBoost multi-class classification model, including: First, defect labels are obtained based on the types of cable defects and multiple cable defect types, or based on the type of defect cluster centers in the first model; then, one-hot encoding is performed on the defect labels to obtain sample labels.

[0061] Based on the absence of cable defect types and multiple cable defect types, or the defect types in the first model, the features of the second model are constructed to obtain defect labels. That is, defect type [ ]; for defect labels One-hot encoding yields: ; Where K represents the type of defect.

[0062] As one embodiment of the present invention, five types of defects plus no defects, totaling six data points, are used to label the defects. .

[0063] Then, the filtered sample dataset is divided into training and validation sets according to a set ratio. It should be noted that the filtering method used in the filtered sample dataset is the same as that in the first model, which is to remove noise points from the samples.

[0064] For example, each type in the excluded sample dataset has three measured currents. and two measurement locations ,common Data points. That is, sample input data. Size is Sample Labels Size is The filtered sample dataset is For the sample set The dataset is divided into two parts, and 80% of the samples from each defect are selected to form the training set. The remaining 20% ​​form the validation set. .

[0065] Finally, the XGBoost multi-class classification model is trained using the training set and its parameters are adjusted using the validation set until convergence, thus obtaining the second model. In other words, the XGBoost multi-class classification model is used to train the sample set... During training, the XGBoost multi-class classification model employs multiple base learners. Each base learner is relatively simple to avoid overfitting. The next learner learns the results from the previous base learners. and actual value The residuals are continuously reduced by learning from multiple learners to decrease the difference between the model value and the actual value.

[0066] Specifically, ;in, Indicates the preceding The results of each base learner, indicated in the lower right corner. Indicates the first Each sample number, superscript Indicates the base learner index. Indicates the first Individual base learners, Indicates the first Each sample is analyzed. The results from all trees are then summed to obtain the model's prediction for a single sample. After learning from the training set using a classification model, the model's performance is validated on the validation set, and parameters are tuned until convergence. The model parameters are then saved to obtain the second model. .

[0067] Step 103: Calculate based on the first and second arrays to obtain the third array. The third array is the ratio of the values ​​in the second array to the corresponding values ​​in the first array.

[0068] In order to synthesize the first model Compared with the second model The increased accuracy makes the model more robust, especially for the first array. Each value and the corresponding second array Each value in Perform division and normalization, i.e.: ;in, It is a normalization function.

[0069] Step 104: Output the defect type corresponding to the largest value in the third array as the result to achieve cable defect monitoring. That is, to... The defect category i corresponding to the largest value in the data is taken as the defect of the newly acquired cable.

[0070] On the other hand, the present invention also provides a cable defect monitoring system based on harmonic data. The system includes at least multiple data acquisition modules and multiple data processing modules, performing the steps of any of the methods described above. The technical features of the system are consistent with those of the method, and will not be repeated here.

[0071] In summary, this invention artificially creates common defects in power distribution cables and applies different currents to single-core cables using a current-temperature rise system. Then, it collects harmonic signals at the cable terminals through N independent tests using a power quality analyzer, obtaining complete harmonic variation data of the cable defect evolution process. K-means clustering is used to calculate cosine similarity for different types of defect data to achieve clustering, and the data is filtered after comparison with the original data to obtain stable data and a robust clustering model. Then, features and samples are constructed, and an XGBoost multi-classification model is used to train the samples, resulting in a stable defect classification model. Finally, the clustering model and classification model are combined for further processing of the results to more accurately determine the defects present in the power distribution cables. This invention addresses the current pain point of not conducting in-depth research on the correlation between cable aging and other defects and cable harmonics, quantitatively realizing a one-to-one correspondence between cable harmonic data and defects, providing a basis for timely defect repair and ensuring the reliability of the power system.

[0072] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0073] It should be understood that the embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0074] The technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0075] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0076] The above description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of embodiments of this disclosure upon considering the specification and practicing the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

[0077] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0078] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A cable defect monitoring method based on harmonic data, characterized in that, The method includes: Acquire cable harmonic data; The cable harmonic data is input into the first model to obtain the first array; the first array is the distance array between the cable harmonic data and the cluster center of each defect in the first model; The cable harmonic data is input into the second model to obtain the second array; the second array is a probability array of the cable harmonic data belonging to various defect types in the second model; The third array is obtained by calculating the ratio of the values ​​in the second array to the corresponding values ​​in the first array, and then the third array is normalized. The defect type corresponding to the maximum value in the normalized third array is output as the result to achieve cable defect monitoring. The construction of both the first model and the second model includes obtaining a sample dataset; The construction of the first model is based on the Kmeas clustering model, including: re-dividing the sample dataset using the Kmeas clustering model to obtain a first clustering result; removing noisy data from the first clustering result to obtain the first model; The second model is constructed based on the XGBoost multi-class classification model, including: obtaining defect labels based on the type of defect cluster centers in the first model; performing one-hot encoding on the defect labels to obtain sample labels; dividing the filtered sample dataset into a training set and a validation set according to a set ratio; training the XGBoost multi-class classification model using the training set, and adjusting the parameters of the XGBoost multi-class classification model using the validation set until the model converges, thereby obtaining the second model.

2. The cable defect monitoring method based on harmonic data as described in claim 1, characterized in that, The cable harmonic data includes: fundamental frequency data, multiple harmonic data, actual current intensity, and actual measurement location.

3. The cable defect monitoring method based on harmonic data as described in claim 1, characterized in that, The acquisition of the sample dataset includes: Obtain a test harmonic dataset; wherein, the test harmonic dataset includes multiple defect datasets, one of which corresponds to a cable defect type or a cable defect type; each defect dataset includes multiple harmonic component data, and each harmonic component data corresponds to a test current intensity and a test measurement location; A sample dataset is obtained based on multiple defect datasets, as well as the test current intensity and test measurement location corresponding to the harmonic component data.

4. The cable defect monitoring method based on harmonic data as described in claim 3, characterized in that, The second model is constructed based on the XGBoost multi-class classification model, including: Defect tags are obtained based on the absence of cable defect types and multiple cable defect types, and the defect tags are subjected to unique hot coding to obtain sample tags.

5. The cable defect monitoring method based on harmonic data as described in claim 1, characterized in that, The types of cable defects include one or more of the following: water tree defects, residual insulation tape defects on the surface of the crimped tube, semi-conductive floating defects on the insulation surface, metal floating defects on the insulation surface, and metal tip discharge defects.

6. The cable defect monitoring method based on harmonic data as described in claim 1, characterized in that, The steps for re-dividing the sample dataset include: Each defect dataset is randomly selected to obtain multiple cluster centers corresponding to the defect type; and all unselected sample points in the defect dataset are marked as test points. Calculate the distance between the current test point and each of the cluster centers, and classify the current test point based on the distance; Based on the classification calculation of the classified sample points and test points, the cluster centroid vector value is obtained; The cluster centroid vector value is compared with the cluster center. If the cluster center changes, the distance between the test point and each cluster center is recalculated. The process continues until the cluster center no longer changes, and the first clustering result is obtained.

7. A cable defect monitoring system based on harmonic data, characterized in that, The system includes at least multiple data acquisition modules and multiple data processing modules, and performs the steps of the method according to any one of claims 1 to 6.

Images