A method and system for online intelligent monitoring of high-voltage cable anti-theft return line

Abstract

The invention relates to the technical field of data processing, and in particular to an online intelligent monitoring method and system for an anti-theft return line of a high-voltage cable. The method comprises: collecting historical cable echo signals and cable status labels, extracting wavelet coefficients of the historical cable echo signals by using wavelet transform technology, calculating the correlation between each wavelet coefficient and all labels, constructing a wavelet coefficient matrix, calculating the similarity between each column of wavelet coefficients and each other column of wavelet coefficients, comprehensively evaluating the importance of the wavelet coefficient according to the correlation between each wavelet coefficient and all labels, and the similarity between the wavelet coefficient in the column where the wavelet coefficient is located and the wavelet coefficient in other columns, constructing a classification prediction model to predict the probability of a label corresponding to the wavelet coefficient, integrating the output of the model according to the importance of the wavelet coefficient, obtaining the probability of cable theft, dynamically adjusting the monitoring frequency according to the probability of cable theft, and having strong adaptability and being able to accurately monitor the cable.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an online intelligent monitoring method and system for an anti-theft return line of a high-voltage cable. Background Art

[0002] As a key infrastructure for power transmission, high-voltage cables occupy a pivotal position in the modern power supply system, and their safe and stable operation is of great significance. However, cables are often accessed by unauthorized personnel, and the means of access are constantly upgraded and diversified, making it difficult for traditional manual inspection methods to deal with them. These unauthorized personnel often use various covert means to commit crimes, such as accessing cables in complex terrain or remote areas, making it difficult for manual inspections to achieve full coverage and timely detection.

[0003] With the rapid development of science and technology, intelligent monitoring technology has emerged. The rise of Internet of Things technology enables sensors to be deployed on cable lines to realize real-time data collection and transmission. The advancement of sensor technology has improved the accuracy and sensitivity of monitoring various cable status parameters. The maturity of big data analysis technology has made it possible to process massive monitoring data, and can mine valuable information from complex data, thereby realizing accurate judgment of cable status. For example, the Chinese patent document with the existing authorization announcement number CN111507310B discloses a method for identifying signals of artificial cable touching operations in an optical cable channel based on Φ-OTDR. It connects Φ-OTDR to the optical cable to be tested, obtains the distribution of vibration data along the optical cable, and then uses a series of means such as signal extraction, image processing, and pattern recognition to process data features, and finally identifies whether the optical cable has been maliciously damaged or stolen.

[0004] However, the above monitoring method relies on specific Φ-OTDR equipment and has poor adaptability. At the same time, its signal processing process is relatively complicated and requires the help of artificial simulation of cable touching operations. Since the simulation scene is often difficult to fully match the actual situation of the actual cable being stolen or damaged, this will inevitably lead to the template signal being unable to fully and accurately reflect the various complex situations that may actually occur, which will ultimately affect the accuracy and reliability of cable theft monitoring.

[0005] In summary, given the severity of the current situation where cables are frequently accessed by unauthorized persons and the limitations of existing monitoring technologies, there is an urgent need to develop a more accurate and adaptable monitoring method to effectively address the problem of cable theft prevention and ensure the safe and stable operation of the power supply system. Summary of the invention

[0006] In order to solve the problem that the traditional monitoring technology has poor adaptability, cannot fully and accurately reflect various complex situations, and affects the accuracy and reliability of monitoring, the present invention proposes an online intelligent monitoring method for high-voltage cable anti-theft return line, including:

[0007] Collecting historical cable echo signals for multiple times, and determining corresponding cable status labels for each collected historical cable echo signal, wherein the labels include disconnection reflection, connector reflection, and grounding reflection;

[0008] Decomposing each historical cable echo signal by using wavelet transform technology to extract multiple wavelet coefficients of each historical cable echo signal;

[0009] Calculate the correlation between each wavelet coefficient and all labels;

[0010] The wavelet coefficient matrix is ​​constructed by taking the number of times of collecting historical cable echo signals as rows and the multiple wavelet coefficients of the historical cable echo signals collected each time as columns, and then calculating the similarity between each column of wavelet coefficients in the wavelet coefficient matrix and each other column of wavelet coefficients;

[0011] The importance of each wavelet coefficient is comprehensively evaluated based on the correlation between each wavelet coefficient and all labels, as well as the similarity between the wavelet coefficient in the column where the wavelet coefficient is located and the wavelet coefficient in each other column;

[0012] A classification prediction model is constructed, and each wavelet coefficient is used as the input of the model. The output of the model is the probability of the label corresponding to the wavelet coefficient. According to the importance of each wavelet coefficient, the output of the model is integrated to obtain the probability of cable theft. The monitoring frequency is dynamically adjusted according to the probability of cable theft.

[0013] The above technical solution can cover the signal characteristics of cables in various states such as normal operation, possible interference and theft by collecting historical cable echo signals at different times and under different working conditions for many times. Determine an accurate label for each signal so that the data has clear classification information, which is convenient for subsequent algorithms to identify the difference of signals under different states. By calculating the similarity, we can understand the degree of correlation between each wavelet coefficient and different cable states, so as to find out the characteristics of wavelet coefficients that are highly correlated with the cable theft state. Constructing a wavelet coefficient matrix can organize and analyze the wavelet coefficients of all collected historical cable echo signals as a whole. Calculating the similarity between columns can reveal the correlation and redundancy between different wavelet coefficients, further explore the internal relationship between wavelet coefficients, and comprehensively consider the relationship between wavelet coefficients and labels and the relationship between wavelet coefficients, so as to more comprehensively and objectively evaluate the importance of each wavelet coefficient in cable state classification. It not only attaches importance to the ability of wavelet coefficients to distinguish the cable status itself (correlation with the label), but also considers its uniqueness and representativeness in the entire wavelet coefficient set (similarity with other columns of wavelet coefficients). The construction of a classification prediction model can utilize the relationship between the characteristics of wavelet coefficients learned from historical data and the cable status, and make real-time predictions and judgments on newly collected cable echo signals. It can quantitatively evaluate the possibility of cables being in different states (especially stolen states) and obtain a more accurate probability of cable theft. This solution has good adaptability, can accurately reflect the state of the cable, and improves the accuracy and reliability of monitoring.

[0014] Preferably, the method for collecting historical cable echo signals is:

[0015] Using the low-voltage pulse echo positioning method, a coaxial line is installed at the center of the high-voltage cable return line conductor, a low-voltage pulse signal is applied, and the cable echo signal reflected back due to the reflection phenomenon caused by the impedance mismatch point of the cable line is collected.

[0016] Preferably, the calculation formula for the correlation between each wavelet coefficient and all labels is:

[0017]

[0018] In the formula, It is The correlation between wavelet coefficients and all labels, is the normalization function, To collect the total number of historical cable echo signals, is the total number of tags, It is All the tags under The wavelet coefficients are in the radius The degree of aggregation within the range of is the normalization function;

[0019] The method to obtain is:

[0020] Calculate the All historical cable echo signals under the label The mean of the wavelet coefficients, with the mean as the center, Draw a circle with radius The first of all historical cable echo signals under the label The total number of wavelet coefficients that fall within the circle.

[0021] The above technical solution is based on All the tags under The wavelet coefficients are in the radius The degree of aggregation within the range is used as a measure of the The wavelet coefficients and It is the key factor to identify the similarity of the labels, and can mine the wavelet coefficient features that are closely related to the labels within a specific range, reflecting the distribution law of the signal characteristics of the cable under different states.

[0022] Preferably, another method for obtaining the degree of aggregation is:

[0023] For All historical cable echo signals under the label The wavelet coefficients are clustered based on density to obtain multiple clusters. The cluster with the largest statistical scale contains The number of wavelet coefficients is taken as All the tags under The wavelet coefficients are in the radius The degree of aggregation within the range .

[0024] Preferably, the calculation formula for the similarity between each column of wavelet coefficients and each other column of wavelet coefficients in the wavelet coefficient matrix is:

[0025]

[0026] In the formula, is the first The wavelet coefficients and The similarity between the wavelet coefficients of the columns, It is The wavelet coefficients and The covariance between the wavelet coefficients of the columns, It is The standard deviation of the wavelet coefficients, It is The standard deviation of the wavelet coefficients.

[0027] The above scheme can reflect the degree of correlation between different wavelet coefficient features by calculating the similarity between different columns in the wavelet coefficient matrix. These wavelet coefficients are obtained after wavelet transform of the cable echo signal. They describe the status information of the cable from different angles. High similarity means that these features have a strong correlation in reflecting the cable status.

[0028] Preferably, the importance of each wavelet coefficient satisfies the following relationship:

[0029]

[0030] In the formula, It is The importance of wavelet coefficients, It is The correlation between wavelet coefficients and all labels, is the first The wavelet coefficients and The similarity between the wavelet coefficients of the columns, It is The total number of similarities between the wavelet coefficients of a column and all other wavelet coefficients of a column, is a normalization function.

[0031] The above technical solution provides a comprehensive and systematic method to measure the value of wavelet coefficients in cable status classification. It not only considers the direct correlation degree of the wavelet coefficient itself to the cable status (represented by the label), but also takes into account the relationship between the wavelet coefficient and other wavelet coefficients, avoiding the one-sidedness of judging the importance based on a single factor.

[0032] Preferably, the method of integrating the output of the model to obtain the probability of cable theft includes:

[0033] The output of the model is integrated according to the following formula:

[0034]

[0035] in, is the probability of cable theft, It is The prediction results of the classification model corresponding to the wavelet coefficients are: It is The importance of wavelet coefficients, is the total number of wavelet coefficients.

[0036] The above technical solution achieves the comprehensive integration of the information contained in multiple wavelet coefficients by weighted summing the classification model prediction results corresponding to each wavelet coefficient according to its importance. Considering that different wavelet coefficients have different importance in reflecting the status of cable theft, this method can make full use of the advantages of each feature and avoid the limitations of single feature judgment, so as to more comprehensively and accurately evaluate the possibility of cable theft.

[0037] Preferably, the method of integrating the output of the model to obtain the probability of cable theft also includes:

[0038] The output of the model is integrated according to the following formula:

[0039]

[0040] in, is the probability of cable theft, It is The prediction results of the classification model corresponding to the wavelet coefficients are: It is The importance of wavelet coefficients, It is The prediction accuracy of the classification model corresponding to the wavelet coefficients, is the total number of wavelet coefficients;

[0041] The method for obtaining the prediction accuracy of the classification model is:

[0042] A test set is selected from all the cable echo signals collected historically, and the prediction accuracy is obtained by dividing the number of correct predictions in the test set by the total number of test sets.

[0043] The above technical solution can dynamically adjust the influence of each classification model in the integration by multiplying the importance of wavelet coefficients with the model prediction accuracy as part of the weight. For those wavelet coefficients that are both important and can be accurately predicted by the model, greater weights will be given, allowing them to play a more critical role in calculating the probability of cable theft. For features with high importance but unstable model predictions, their weights will be reduced accordingly to ensure that the final result is more reliable.

[0044] Preferably, the method for dynamically adjusting the monitoring frequency according to the probability of cable theft is:

[0045] Set the initial normal monitoring frequency to every Monitor the cable once every minute;

[0046] When the cable theft probability is greater than the preset first threshold When Monitor once every minute;

[0047] When the cable theft probability is at the preset first threshold and below, and greater than or equal to the preset second threshold When Monitor once every minute;

[0048] When the cable theft probability is less than the preset second threshold and is greater than or equal to a preset third threshold P3 When Monitor once every minute;

[0049] When the cable theft probability is less than the preset third threshold When Monitor once every minute; .

[0050] The above technical solution reduces the number of invalid monitoring times by reasonably adjusting the monitoring frequency, improves the overall work efficiency of the monitoring system, can respond quickly to high-risk situations, and promptly discover and deal with problems. For low-risk situations, it avoids the waste of time and resources caused by excessive monitoring, allowing the monitoring system to focus more on monitoring key situations and improve the effectiveness of monitoring.

[0051] The present invention also provides an online intelligent monitoring system for an anti-theft return line of a high-voltage cable. The intelligent monitoring system comprises a memory and a processor. A computer program is stored in the memory. The processor executes the computer program to implement the steps of any one of the intelligent monitoring methods.

[0052] The present invention has the following effects:

[0053] The present invention uses various technical means to comprehensively analyze the characteristics of cable echo signals, accurately evaluate the importance of wavelet coefficients, build an effective model and dynamically adjust the monitoring frequency, thereby achieving accurate judgment of the cable status, having strong adaptability and improving the accuracy and reliability of monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0055] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0057] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0058] Reference Figure 1 The present invention provides an online intelligent monitoring method for high-voltage cable anti-theft return line, comprising the steps of: -step :

[0059] : Get historical cable echo signals and their labels.

[0060] In the field of cable security monitoring, especially for cable theft monitoring, this solution aims to obtain historical cable echo signals and their labels through a specific low-voltage pulse echo positioning method, and build an effective cable theft prevention monitoring system based on this.

[0061] First, a brief description of the monitoring principle and device settings is given:

[0062] Low-voltage pulse echo location method: The low-voltage pulse echo location method is used to locate the fault of the return line. The core is to install a coaxial line at the center of the return line conductor of the high-voltage cable, and use the characteristics of the low-voltage pulse signal propagating in the coaxial line. When the pulse signal is transmitted in the cable line, it will be reflected when it encounters a point where the impedance does not match, forming an echo signal.

[0063] Signal propagation and reflection mechanism: Low-voltage pulse signals propagate along the coaxial line. During the propagation process, impedance mismatch occurs due to various conditions in the cable line. In addition to extreme conditions such as conductor disconnection (open circuit), short circuit, and grounding fault that may be caused by cable theft, impedance unevenness may also occur at cable joints and where the cable passes through metal pipes due to factors such as material properties and structural changes. These impedance unevenness points will cause the pulse signal to reflect, forming an echo signal that is received by the monitoring equipment. Based on the echo signal, it can be determined whether the cable has been stolen.

[0064] According to the above monitoring principle and device, the historical echo signal is collected. In actual operation, by applying a low-voltage pulse signal on the coaxial line, then receiving the reflected echo signal, recording and storing it, the historical cable echo data is formed.

[0065] Next, the echo type label is given based on the collected historical cable echo signals and the actual situation. When the echo signal characteristics are consistent with the reflection characteristics at the joint, it is marked as joint reflection; if it is consistent with the reflection characteristics of the broken line, it is marked as broken line reflection; if it is the reflection characteristics corresponding to the ground fault, it is marked as ground reflection, etc.

[0066] For the disconnection reflection label: When a cable is stolen, it is very likely to be cut, resulting in disconnection. In this case, the echo signal generated is marked as disconnection reflection, which is a label that can better indicate that the cable has been stolen.

[0067] For connector reflection labels: This label indicates that the echo signal is caused by reflections generated at the cable connector. This usually means that there is an impedance mismatch at the connector. This is usually caused by the manufacturing process of the cable connector, material differences, or aging after long-term use. It has no direct relationship with cable theft, but more reflects changes in the physical properties of the normal connection part of the cable or normal maintenance issues.

[0068] For the Ground Reflection Label: This label indicates the reflective signature of a ground fault. A ground fault may be caused by damage to the cable insulation, which creates an electrical path between the cable core and the earth (or grounding device). This may be caused by long-term exposure to environmental factors (such as moisture, corrosion, etc.), or the quality of the cable itself causing the insulation to deteriorate. It is not necessarily a sign of cable theft. However, this label may also appear if an unauthorized person damages the cable insulation during the theft process, causing a ground fault.

[0069] All historical cable echo signals collected in this step and the corresponding labels can provide a classification basis for subsequent AI-based data analysis, so as to train the model to accurately identify different cable states, especially to distinguish whether the cable is in an abnormal state caused by theft.

[0070] : Multiple wavelet coefficients of each historical cable echo signal are extracted using wavelet transform technology.

[0071] In the process of analyzing and processing the cable echo signal, it is necessary to extract the characteristics of the cable echo signal. Wavelet transform is a powerful signal processing tool. When applied to the cable echo signal, it can decompose the signal into components of different frequencies, thereby obtaining wavelet coefficients such as high-frequency coefficients and low-frequency coefficients. These wavelet coefficients contain rich information of the cable echo signal at different frequency scales, such as the amplitude change of the signal, the frequency distribution characteristics, etc. By analyzing these wavelet coefficients, many characteristics of the cable echo signal can be understood. Therefore, the wavelet transform technology is used to extract multiple wavelet coefficients of each historical cable echo signal, and these wavelet coefficients are used to reflect the characteristics of the cable echo signal.

[0072] In addition, when the signal is subjected to wavelet transform, since the level and scale of the wavelet transform for signal decomposition are predetermined, the number of wavelet coefficients obtained for each historical cable echo signal is the same. For example, 10 historical cable echo signals are collected, and 15 wavelet coefficients are extracted from each historical cable echo signal.

[0073] : Calculate the correlation between each wavelet coefficient and all labels.

[0074] Since wavelet transform is a mathematical tool that decomposes a signal into different frequency components, when performing wavelet transform on the echo signal of the cable, the high-frequency coefficients and low-frequency coefficients obtained can take any real values ​​within a certain numerical range. For example, the high-frequency coefficients may be real numbers such as 1.23, 2.78, etc., and these values ​​change continuously, so the wavelet coefficients are continuous variables.

[0075] Since the label represents a certain state or classification of the cable echo signal, for example, in cable fault monitoring, the signal is either a label corresponding to the normal state or a label corresponding to a certain fault, and there will be no ambiguous labels between the two states. It is like a switch, with only a limited number of fixed positions (states), not a continuously changing quantity, but a discrete variable.

[0076] Therefore, when calculating the correlation, the conventional method of calculating similarity cannot be used. The calculation formula provided here is:

[0077]

[0078] In the formula, It is The correlation between wavelet coefficients and all labels, To collect the total number of historical cable echo signals, is the total number of tags, It is All the tags under The wavelet coefficients are in the radius The degree of aggregation within the range.

[0079] There are two ways to obtain it:

[0080] The first method is to find all the The historical cable echo signals of a tag (such as the tag of the broken line reflection) are then obtained. wavelet coefficients. Since the same wavelet transform algorithm is used, the number of wavelet coefficients extracted from each historical cable echo signal is the same. The mean of the wavelet coefficients, with this mean as the center, Draw a circle with radius is the default value, which is set to 1 here. Finally, count the number of How many of the wavelet coefficients fall within this circle? The number of wavelet coefficients is .

[0081] For example, The label is the broken line reflection. There are 10 historical cable echo signals under this label, and each historical cable echo signal has 10 wavelet coefficients. =2, the first of the 10 historical cable echo signals is obtained. wavelet coefficients, then the number of the second wavelet coefficients is 10. Then calculate the mean of these second wavelet coefficients, take the mean as the center of the circle, and Draw a circle with radius , and count the number of the second wavelet coefficients to find that 5 of them fall within the circle, then The value of is 5.

[0082] The second method is to find all the The historical cable echo signals of a tag (such as the tag of the broken line reflection) are then obtained. wavelet coefficients, for the All the tags under wavelet coefficients, and use density-based clustering algorithm to cluster these The wavelet coefficients are clustered. This clustering algorithm will The spatial distribution density of the wavelet coefficients divides them into different clusters. After the clustering is completed, find the largest cluster (including the The cluster with the most wavelet coefficients), count the number of The number of wavelet coefficients is taken as .

[0083] For example, suppose the The label is the cable break reflection. There are 5 historical cable echo signals under this label. , get the fifth wavelet coefficients of the five historical cable echo signals, a total of 75, after clustering, get three clusters, the number of the fifth wavelet coefficients contained in each cluster is 20, 10, 45, then take 45 as .

[0084] : Construct a wavelet coefficient matrix, and then calculate the similarity between each column of wavelet coefficients in the wavelet coefficient matrix and each other column of wavelet coefficients.

[0085] The wavelet coefficients of all historical cable echo signals collected multiple times are used to construct a wavelet coefficient matrix. Since the number of wavelet coefficients (wavelet coefficients are extracted through wavelet transform) of each historical cable echo signal collected is the same, the wavelet coefficient matrix is ​​constructed with the number of times the historical cable echo signal is collected as rows and the multiple wavelet coefficients of each historical cable echo signal collected as columns. By analyzing the similarity between the wavelet coefficients in each column of the matrix, the similarity of the characteristics of different frequency bands in the multiple collection processes can be understood, and then the stability or change of the cable status can be inferred. For example, 8 historical cable echo signals are collected, and 5 wavelet coefficients are extracted from each historical cable echo signal, then a wavelet coefficient matrix with 8 rows and 5 columns is constructed.

[0086] Next, the calculation formula for the similarity between each column of wavelet coefficients and each other column of wavelet coefficients in the wavelet coefficient matrix is:

[0087]

[0088] In the formula, is the first The wavelet coefficients and The similarity between the wavelet coefficients of the columns, It is The wavelet coefficients and The covariance between the wavelet coefficients of the columns, It is The standard deviation of the wavelet coefficients, It is The standard deviation of the wavelet coefficients.

[0089] In this formula, the covariance measures the Column and Whether the overall change trend of the wavelet coefficients of the two columns is consistent, that is, whether they increase or decrease at the same time, but the value of covariance is affected by the value range (standard deviation) of the variable itself. It is difficult to accurately compare the similarities between different pairs of columns using covariance alone. Therefore, by dividing by the standard deviation of each of the two columns of wavelet coefficients and , which can eliminate the influence caused by the different value ranges of the two columns of wavelet coefficients, making the calculated It is possible to measure the similarity between wavelet coefficients in different columns under a relatively unified standard, which is more comparable.

[0090] when When the value is close to 1, it means that the similarity between the two columns of wavelet coefficients is very high. This means that in the process of collecting historical cable echo signals for many times, the change trends of the frequency band characteristics represented by these two columns of wavelet coefficients are very similar. From the perspective of cable status analysis, it may mean that the characteristics of the cable related to these frequency bands (such as reflection characteristics caused by local defects, etc.) are relatively stable at different collection times and have no obvious changes.

[0091] when When the value is close to 0, it indicates that the similarity between the two columns of wavelet coefficients is very low, that is, the frequency band characteristics they represent have different changing trends in multiple acquisitions. This may suggest that the characteristics of the cable in the corresponding frequency band have large fluctuations, which may be due to the influence of external interference, aging or local damage on the cable, resulting in obvious changes in the characteristics of the reflected signal in these frequency bands at different times.

[0092] In summary, by constructing the wavelet coefficient matrix and calculating the inter-column similarity using the given formula, it is possible to effectively extract features from the historical cable echo signals and analyze their similarity at different acquisition times, providing valuable information for cable status monitoring and fault diagnosis.

[0093] :According to the correlation between each wavelet coefficient and all labels, as well as the similarity between the wavelet coefficient in the column where the wavelet coefficient is located and the wavelet coefficient in each other column, the importance of the wavelet coefficient is comprehensively evaluated.

[0094] If the similarity between a wavelet coefficient and a label is greater, it means that the wavelet coefficient is more important for the division of the cable status. If the similarity between a wavelet coefficient and other wavelet coefficients is greater, it means that the wavelet coefficient can be replaced by other wavelet coefficients and the importance of the wavelet coefficient is smaller.

[0095] Therefore, the importance of each wavelet coefficient is evaluated according to the following formula:

[0096]

[0097] In the formula, It is The importance of wavelet coefficients, It is The correlation between wavelet coefficients and all labels, is the first The wavelet coefficients and The similarity between the wavelet coefficients of the columns, It is The total number of similarities between the wavelet coefficients of the column and the wavelet coefficients of all other columns, that is, the total number of similarities between the other columns Column, No. The wavelet coefficients and this The column will get Similarity, This is a normalization operation so that the importance of different wavelet coefficients can be compared on the same scale.

[0098] In this formula, Measured the The degree of correlation between the wavelet coefficients and the labels representing the cable status. For example, in the cable status monitoring scenario, the labels may indicate whether the cable is normal, whether there is a certain type of fault, and other different states. This correlation reflects the correlation between the first The potential ability of a wavelet coefficient to distinguish different cable states. If the value is large, it means that the wavelet coefficient has a strong correlation with a certain feature of the cable state and may play an important role in determining the cable state (such as normal or faulty).

[0099] In this formula, It is used to describe the similarity of signal features represented by different columns of wavelet coefficients at different acquisition times. High similarity indicates that the corresponding frequency band feature change trends are similar, suggesting that the relevant characteristics of the cable are relatively stable in these frequency bands. Low similarity indicates that the feature change trends are very different, which may be related to factors such as cable status changes or interference.

[0100] In this formula, Given the A comprehensive measure of the similarity between the wavelet coefficients of a column and other columns as a whole. By averaging, we can have a more comprehensive understanding of the relative position of the wavelet coefficients of this column in the entire wavelet coefficient matrix, that is, its average performance in similarity with other columns.

[0101] In this formula, The relevance of the wavelet coefficients to the cable state division and its similarity to other columns in the wavelet coefficient matrix are comprehensively considered. The numerator reflects its potential ability to distinguish the cable state, and the denominator reflects the possibility of it being replaced by other columns (high average similarity means it is easier to be replaced), so this ratio can reflect the first The relative importance of the wavelet coefficients after considering these two factors.

[0102] when The bigger and The smaller the time, the The wavelet coefficient has a strong correlation with the label and a low average similarity with other wavelet coefficients, which shows that the wavelet coefficient plays an important role in the cable status classification because it can distinguish the cable status well (high correlation) and is not easily replaced by other wavelet coefficients (low average similarity). The larger the value of .

[0103] when The smaller and The larger the The wavelet coefficient has a weak correlation with the label and a high average similarity with other wavelet coefficients, which means that the wavelet coefficient is not strongly associated with the cable status characteristics (low correlation) and is easily replaced by other wavelet coefficients (high average similarity). The contribution of this wavelet coefficient to the cable status division is small. Therefore, the importance of this wavelet coefficient The smaller the value of .

[0104] In summary, by calculating the importance of all wavelet coefficients, we can observe the comprehensive performance of different wavelet coefficients in the division of cable states and the relationship with other wavelet coefficients. If the importance of some wavelet coefficients changes significantly in a certain period or a certain cable sample (such as the originally important wavelet coefficients become unimportant, or the originally unimportant wavelet coefficients become important), this may imply that the characteristics of the cable have changed, such as the emergence of new faults.

[0105] : Construct a classification prediction model of wavelet coefficients, integrate the output results of the classification prediction model according to the importance of each wavelet coefficient, and obtain the probability of cable theft.

[0106] Classification models are constructed for different wavelet coefficients and labels, such as decision trees, support vector machines (SVM), neural networks, etc. Taking the decision tree as an example, it has the advantages of being easy to understand and explain. Each wavelet coefficient is used as input, and the cable status label corresponding to the wavelet coefficient is used as output. The decision tree model is trained for different wavelet coefficients. For example, for the first wavelet coefficient, the value of the first wavelet coefficient in all collected samples is used as input, and the corresponding cable status label is used as output to train a decision tree model; the same is true for the second wavelet coefficient, and multiple decision tree models are trained in sequence. During the training process, the decision tree will divide different cable status categories by constructing a tree structure according to the input wavelet coefficient value, because the labels of joint reflection, grounding reflection and broken line reflection can represent the three situations of cable joint fault, grounding fault and cable theft respectively.

[0107] After the model training is completed, for the newly collected cable echo signal, the wavelet coefficients are first extracted and then input into the corresponding trained models to obtain the prediction results of each model, including the cable joint failure probability, grounding failure probability and theft probability.

[0108] Next, the output results of the model are integrated to finally determine the probability of cable theft. Two integration methods are provided here. The first integration method focuses on integrating the prediction results of each classification model according to the importance of the wavelet coefficients. The second integration method further combines the prediction accuracy of the classification model on the basis of considering the importance of the wavelet coefficients, thereby more comprehensively determining the probability of cable theft.

[0109] Specifically, the first integration method satisfies the following relationship:

[0110]

[0111] In the formula, is the probability of cable theft, is the total number of wavelet coefficients, It is The prediction result of the classification model corresponding to the wavelet coefficient, that is, The classification model corresponding to each wavelet coefficient gives the prediction result for the newly collected cable echo signal. The prediction results are specifically the probability of joint failure, grounding failure and theft. It is a judgment output of the possibility of the current state of the cable based on the analysis and learning of the input wavelet coefficients by each classification model. It is one of the basic data sources for finally determining the probability of cable theft. It is The importance of the wavelet coefficients reflects the The relative importance of the wavelet coefficients in the entire analysis system for the classification of cable status, etc. The wavelet coefficients with high importance should occupy a more important position in the final comprehensive judgment such as the probability of cable theft, because they are considered to play a more critical role in distinguishing different cable status (including whether it is stolen).

[0112] In this formula, It is the sum of the importance of all wavelet coefficients, reflecting the total indicator of the overall importance of all wavelet coefficients. This part is used as the denominator for normalization processing, so that the importance of different wavelet coefficients can be comprehensively considered on the same scale when calculating the probability of cable theft, ensuring that the final result will not be deviated due to factors such as the value range of the importance of individual wavelet coefficients, thereby ensuring the relative fairness and comprehensiveness of the results.

[0113] In this formula, As the The normalized weight of the importance of the wavelet coefficients, which accurately reflects the The relative weight that each wavelet coefficient should occupy when comprehensively considering all wavelet coefficients is used to reasonably distribute the influence of the prediction results of each classification model in the subsequent calculation of the probability of cable theft.

[0114] In summary, the formula multiplies the prediction result of the classification model corresponding to each wavelet coefficient by the normalized weight of the importance of the wavelet coefficient, and then accumulates and sums to obtain the probability of cable theft. This means that when determining the probability of cable theft, the importance difference of each wavelet coefficient is fully considered. Instead of simply averaging the prediction results of each classification model, the contribution of different classification models is treated differently according to the importance of the wavelet coefficient. For example, if the importance of a wavelet coefficient is relatively high, the prediction result of the classification model corresponding to it will account for a larger proportion in the calculation of the probability of cable theft, that is, the classification model has a greater impact on the final result; conversely, the classification model corresponding to the wavelet coefficient with lower importance has a relatively small impact on the final result. Moreover, through normalization processing, no matter what the value range of the importance of each wavelet coefficient is, the final calculated probability of cable theft can be within a reasonable and relatively stable range, and there will be no extreme value or unreasonable fluctuation in the calculation result of the probability of cable theft due to the excessive or too small importance value of a wavelet coefficient, making the result more credible and interpretable.

[0115] Specifically, the second integration method satisfies the following relationship:

[0116]

[0117] In the formula, is the probability of cable theft, is the total number of wavelet coefficients, It is The prediction results of the classification model corresponding to the wavelet coefficients are: It is The importance of wavelet coefficients, It is The prediction accuracy of the classification model corresponding to the wavelet coefficients.

[0118] for , which is obtained by the following method: select a test set from all the cable echo signals collected historically, divide the number of correct predictions in the test set by the total number of test sets to get the first The prediction accuracy of the classification model corresponding to the wavelet coefficients measures the prediction accuracy of the classification model on historical data. And the prediction accuracy is an important indicator for evaluating the performance of a classification model, which reflects the ability of the model to accurately predict the cable status based on the given wavelet coefficients. When determining the probability of cable theft, considering the prediction accuracy of the classification model can more reasonably distribute the influence of the prediction results of each classification model, because the prediction results of the classification model with high accuracy should occupy a more important position in the final result.

[0119] In this formula, The importance values ​​of all wavelet coefficients are multiplied by the corresponding classification model prediction accuracy and then added up to obtain a total indicator that comprehensively considers the importance of wavelet coefficients and the prediction accuracy of the classification model. It is used as the denominator for normalization processing, which is similar to the role of the denominator in the first integration method. Different wavelet coefficients and their corresponding classification models can be comprehensively considered on the same scale when calculating the probability of cable theft, while taking into account the two important factors of the importance of wavelet coefficients and the prediction accuracy of the classification model.

[0120] In this formula, is through the The product of the importance of the wavelet coefficient and the prediction accuracy is divided by the sum of the products of the importance of all wavelet coefficients and the prediction accuracy of the corresponding classification model to obtain the relative importance weight of the wavelet coefficient when considering the importance and accuracy. Its value range is between 0 and 1. This weight accurately reflects the The relative proportion that each wavelet coefficient and its corresponding classification model should occupy when comprehensively considering all wavelet coefficients and their corresponding classification models, so as to reasonably allocate the influence of the prediction results of each classification model in the subsequent calculation of the probability of cable theft and determine the probability of cable theft more reasonably.

[0121] In summary, the second integration method not only considers the importance of wavelet coefficients, but also the prediction accuracy of the corresponding classification model when determining the probability of cable theft, making the final calculated probability of cable theft more reliable. Compared with the first integration method, it can better screen out the prediction results of classification models that are important for cable status classification (high importance) and can accurately predict cable status (high accuracy), reducing the errors that may be caused by simply considering one of the factors, importance or accuracy, thereby improving the accuracy and reliability of the judgment of the probability of cable theft.

[0122] If the importance of a wavelet coefficient is high and the prediction accuracy of its corresponding classification model is also high, then the prediction result of its corresponding classification model will account for a larger proportion in calculating the probability of cable theft, that is, the classification model has a greater impact on the final result; conversely, a classification model with low importance and low prediction accuracy has a relatively small impact on the final result.

[0123] : Dynamically adjust the monitoring frequency based on the probability of cable theft.

[0124] Set the initial normal monitoring frequency to every The cable is monitored once every minute. The experience value is 10, and the first threshold (high risk threshold) is preset. 0.7, the second threshold (medium-high risk threshold) The first threshold is 0.4, the third threshold (medium-low risk threshold) is 0.15, and the fourth threshold (low risk threshold) is 0.05.

[0125] For the cable to be judged, when the probability of cable theft is greater than 0.7, it means that the cable is at a high risk of theft. At this time, some obvious abnormal signal characteristics may have appeared, such as the waveform changes in the cable echo signal that are highly similar to the typical theft mode, or multiple key wavelet coefficient features showing abnormal values ​​at the same time. These signs indicate that the cable is very likely to be obtained or seriously damaged by unauthorized personnel. Immediate monitoring needs to be strengthened to obtain the latest status information of the cable as quickly as possible. Therefore, it is set The monitoring interval of minutes can ensure that the cable is monitored multiple times in a short period of time, and possible signs of theft can be quickly discovered, such as signal mutations caused by damage to the cable sheath, instantaneous abnormal current in the return line, etc., to buy precious time for timely anti-theft measures (such as calling the police, dispatching personnel to the scene, etc.).

[0126] When the probability of cable theft is in the interval [0.4, 0.7], it means that the cable has a certain degree of theft risk. Some local abnormal signals have been detected, but no obvious theft pattern has been formed. For example, a wavelet coefficient feature has deviated from the normal range to a certain extent, but it is not enough to determine that the cable has been stolen. In this case, You can closely monitor cable status changes within minutes to detect potential problems in a timely manner.

[0127] When the probability of cable theft is in the range of [0.15, 0.4), it indicates that the risk of theft is relatively low. Some weak abnormal signals are detected, which may be caused by environmental interference or normal aging of cables. However, the potential risk of theft cannot be completely ruled out. The monitoring frequency of minutes is moderate, which can continuously observe the cable status and ensure safety without excessive consumption of resources.

[0128] When the probability of cable theft is less than 0.15, it means that the cable echo signal is basically stable, and the wavelet coefficient characteristics are consistent with the normal operation state, indicating that the possibility of cable theft is extremely small. Set the monitoring frequency to The monitoring is carried out once every minute. The 30-minute monitoring interval is sufficient to minimize the number of monitoring times and reduce monitoring costs while ensuring the safety of the cable.

[0129] The present invention also provides an online intelligent monitoring system for an anti-theft return line of a high-voltage cable. The intelligent monitoring system comprises a memory and a processor. A computer program is stored in the memory. The processor executes the computer program to implement the steps of the intelligent monitoring method.

[0130] In the description of this specification, “plurality” means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0131] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for online intelligent monitoring of high-voltage cable anti-theft return line, characterized in that: include: Collecting historical cable echo signals for multiple times, and determining corresponding cable status labels for each collected historical cable echo signal, wherein the labels include disconnection reflection, connector reflection, and grounding reflection; Decomposing each historical cable echo signal by using wavelet transform technology to extract multiple wavelet coefficients of each historical cable echo signal; Calculate the correlation between each wavelet coefficient and all labels; The wavelet coefficient matrix is ​​constructed by taking the number of times of collecting historical cable echo signals as rows and the multiple wavelet coefficients of the historical cable echo signals collected each time as columns, and then calculating the similarity between each column of wavelet coefficients in the wavelet coefficient matrix and each other column of wavelet coefficients; The importance of each wavelet coefficient is comprehensively evaluated based on the correlation between each wavelet coefficient and all labels, as well as the similarity between the wavelet coefficient in the column where the wavelet coefficient is located and the wavelet coefficient in each other column; A classification prediction model is constructed, and each wavelet coefficient is used as the input of the model. The output of the model is the probability of the label corresponding to the wavelet coefficient. According to the importance of each wavelet coefficient, the output of the model is integrated to obtain the probability of cable theft. The monitoring frequency is dynamically adjusted according to the probability of cable theft.

2. The method for online intelligent monitoring of high-voltage cable anti-theft return line according to claim 1 is characterized in that: The method for collecting historical cable echo signals is: Using the low-voltage pulse echo positioning method, a coaxial line is installed at the center of the high-voltage cable return line conductor, a low-voltage pulse signal is applied, and the cable echo signal reflected back due to the reflection phenomenon caused by the impedance mismatch point of the cable line is collected.

3. The method for online intelligent monitoring of high-voltage cable anti-theft return line according to claim 1 is characterized in that: The correlation between each wavelet coefficient and all labels is calculated as: ； In the formula, It is The correlation between wavelet coefficients and all labels, is the normalization function, To collect the total number of historical cable echo signals, is the total number of tags, It is All the tags under The wavelet coefficients are in the radius The degree of aggregation within the range; The method to obtain is: Calculate the The first of all historical cable echo signals under the label The mean of the wavelet coefficients, with the mean as the center, Draw a circle with radius The first of all historical cable echo signals under the label The total number of wavelet coefficients that fall within the circle.

4. The method for online intelligent monitoring of high-voltage cable anti-theft return line according to claim 3 is characterized in that: Another method of obtaining the degree of aggregation is: For The first of all historical cable echo signals under the label The wavelet coefficients are clustered based on density to obtain multiple clusters. The cluster with the largest statistical scale contains The number of wavelet coefficients is taken as All the tags under The wavelet coefficients are in the radius The degree of aggregation within the range .

5. The method for online intelligent monitoring of high-voltage cable anti-theft return line according to claim 1 is characterized in that: The calculation formula for the similarity between each column of wavelet coefficients in the wavelet coefficient matrix and each other column of wavelet coefficients is: ； In the formula, is the first The wavelet coefficients and The similarity between the wavelet coefficients of the columns, It is The wavelet coefficients and The covariance between the wavelet coefficients of the columns, It is The standard deviation of the wavelet coefficients, It is The standard deviation of the wavelet coefficients.

6. The method for online intelligent monitoring of high-voltage cable anti-theft return line according to claim 1 is characterized in that: The importance of each wavelet coefficient satisfies the following relationship: ； In the formula, It is The importance of wavelet coefficients, It is The correlation between wavelet coefficients and all labels, is the first The wavelet coefficients and The similarity between the wavelet coefficients of the columns, It is The total number of similarities between the wavelet coefficients of a column and all other wavelet coefficients of a column, is a normalization function.

7. The method for online intelligent monitoring of high-voltage cable anti-theft return line according to claim 1 is characterized in that: Methods for integrating the output of the model to obtain the probability of cable theft include: The output of the model is integrated according to the following formula: ; in, is the probability of cable theft, It is The prediction results of the classification model corresponding to the wavelet coefficients are: It is The importance of wavelet coefficients, is the total number of wavelet coefficients.

8. The method for online intelligent monitoring of high-voltage cable anti-theft return line according to claim 1 is characterized in that: The method of integrating the output of the model to obtain the probability of cable theft also includes: The output of the model is integrated according to the following formula: ; in, is the probability of cable theft, It is The prediction results of the classification model corresponding to the wavelet coefficients are: It is The importance of wavelet coefficients, It is The prediction accuracy of the classification model corresponding to the wavelet coefficients, is the total number of wavelet coefficients; The method for obtaining the prediction accuracy of the classification model is: A test set is selected from all the cable echo signals collected historically, and the prediction accuracy is obtained by dividing the number of correct predictions in the test set by the total number of test sets.

9. The method for online intelligent monitoring of high-voltage cable anti-theft return line according to claim 1 is characterized in that: The method of dynamically adjusting the monitoring frequency according to the probability of cable theft is: Set the initial normal monitoring frequency to every Monitor the cable once every minute; When the cable theft probability is greater than the preset first threshold When Monitor once every minute; When the cable theft probability is at the preset first threshold and below, and greater than or equal to the preset second threshold When Monitor once every minute; When the cable theft probability is less than the preset second threshold and is greater than or equal to a preset third threshold When Monitor once every minute; When the cable theft probability is less than the preset third threshold When Monitor once every minute; .

10. An online intelligent monitoring system for high-voltage cable anti-theft return line, characterized in that: The intelligent monitoring system comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the steps of the intelligent monitoring method according to any one of claims 1 to 9.

Images