Low-voltage line insulation anomaly detection method and system based on intelligent fusion terminal

Abstract

The present application relates to the field of data processing, and more particularly to a low-voltage line insulation anomaly detection method and system based on an intelligent fusion terminal, the method comprising: collecting temperature data and current data of each line as an insulation state feature vector of each line; selecting an optimal cluster number based on the insulation state feature vector of each line; calculating the deterioration speed of each line based on the trend and speed of the harmonic distortion rate of each line over time, calculating the mean of the deterioration speeds of all lines in each cluster as the average deterioration degree of each cluster, sorting the average deterioration degrees of each cluster based on their sizes, and selecting the best split point to divide the clusters into a healthy line group and an insulation anomaly line group; and calculating the insulation risk index of each line in the healthy line group based on the distance of each line in the healthy line group from the best split point and dynamically adjusting the monitoring time period of the line.

Description

Technical Field

[0001] This invention relates to the field of data processing. In particular, it relates to a method and system for detecting insulation anomalies in low-voltage lines based on an intelligent fusion terminal. Background Technology

[0002] In modern power distribution networks, the deterioration of insulation performance of low-voltage lines is one of the main causes of power safety accidents. According to statistics, the number of electric shock accidents and fire accidents caused by insulation faults of low-voltage lines remains high every year nationwide, causing not only huge economic losses, but also seriously threatening the safety of people's lives and property.

[0003] Traditional insulation testing methods mainly rely on periodic manual inspections or threshold alarms based on a single parameter. These methods suffer from significant drawbacks, including poor timeliness, high false alarm rates, and a lack of predictability. Specifically: 1. Poor timeliness: Manual inspections typically occur every month or even longer, making it difficult to detect sudden insulation degradation in a timely manner. This is especially true under adverse weather conditions, where line insulation performance can deteriorate rapidly, and manual inspections cannot provide real-time monitoring. 2. High false alarm rate: The single-parameter threshold method is susceptible to interference from load fluctuations, ambient temperature, and the start-up and shutdown of high-power appliances. For example, the start-up of household appliances such as air conditioners and water heaters can cause a sudden increase in current, leading to false alarms and severely impacting maintenance personnel's trust in the system and response efficiency. 3. Lack of predictability: Existing technologies can only detect severe insulation faults that have already occurred and cannot provide early warnings for progressive insulation degradation. By the time the system issues an alarm, the line is often already in a severely faulty state, missing the optimal time for intervention.

[0004] Currently, although some studies use abnormal data for insulation assessment, their focus is on current anomalies caused by insulation abnormalities. However, current parameters are easily affected by environmental interference, and the start-up of high-power appliances usually causes a sudden increase in current. Moreover, line degradation is a long-term dynamic process. Current anomalies can only reflect whether the insulation performance of the line is abnormal, and cannot provide real-time monitoring and early warning for lines that are still in normal condition but are deteriorating. In addition, existing technologies generally use fixed thresholds for binary classification, which simply separates the continuous insulation degradation process into two discrete states: healthy and abnormal. This leads to the neglect of early degradation signals, the inability to capture the gradual change process of insulation status, and the difficulty in achieving true early warning. Summary of the Invention

[0005] To address the problem of insufficient early warning capability caused by the existing technology simply dividing the continuous insulation degradation process into two discrete states, the present invention provides solutions in the following aspects.

[0006] In the first aspect, the low-voltage line insulation anomaly detection method based on intelligent fusion terminal includes: collecting and preprocessing temperature and current data of each line to obtain insulation state feature vectors of each line including temperature data and harmonic distortion rate; clustering based on the insulation state feature vectors of each line to select the optimal number of clusters; calculating the degradation rate of each line based on the trend and speed of harmonic distortion rate change over time, calculating the average degradation rate of all lines in each cluster as the average degradation degree of each cluster, sorting the clusters based on the magnitude of the average degradation degree of each cluster, calculating the optimal segmentation point of each cluster after sorting based on a binary classification algorithm, dividing the sorted clusters into healthy line groups and insulation anomaly line groups based on the optimal segmentation point; calculating the insulation risk index of each line in the healthy line group based on the distance between each line in the healthy line group and the optimal segmentation point, and dynamically adjusting the monitoring time period of the line based on the magnitude of the insulation risk index of each line.

[0007] This method uses temperature-harmonic distortion rate as the dual feature of insulation status feature vector for new lines, which improves the accuracy of line degradation segmentation. Based on the insulation status feature vector of each line, clustering is performed to select the optimal number of clusters. Lines with the same degree of degradation are further grouped into the same cluster. By sorting each cluster and selecting the best split point, the accuracy of segmenting healthy and abnormal lines is improved, avoiding errors caused by subjectively set thresholds. Based on the insulation risk index, healthy lines are dynamically monitored, which further improves the timeliness of monitoring and avoids waste of resources.

[0008] Preferably, the degradation rate is calculated as follows: a time window is selected, and the trend of harmonic distortion rate of different lines changing with time within the time window is calculated as the degradation trend of the line. The degradation trend is divided by the length of the time window to obtain the degradation rate of the line.

[0009] Preferably, the method for calculating the degradation trend is as follows: collect the harmonic distortion rate of the line at each moment within the time window, compare the rise and fall of the harmonic distortion rate at adjacent moments in the line, and accumulate all the rise and fall states of the time window based on the sign function to obtain the degradation trend of the line within the time window.

[0010] By treating the "cumulative value of rising and falling signs" as the degradation trend and dividing it by the window length to obtain the degradation rate, the system can not only sensitively capture early gradual degradation, but also quantify the rate of degradation with a unified dimension. Real-time and quantitative assessment of the health status of the line can be achieved with zero threshold and zero manual parameters, significantly reducing false alarms and missed alarms and reducing the workload of operation and maintenance.

[0011] Preferably, the size of the preset base time window is used, and the degradation factor is calculated by subtracting the degradation rate of the line in the current time window from 1. The product of the degradation factor and the base time window is then used as the size of the next time window.

[0012] Preferably, the optimal split point is calculated as follows: preset the split points for each cluster, calculate the mean of the average degradation degree of all healthy lines based on the split points as the degradation degree of healthy lines, calculate the mean of the average degradation rate of all abnormal lines as the degradation degree of abnormal lines, traverse all clusters as split points, calculate the inter-class variance corresponding to each split point based on the degradation degree of healthy lines and the degradation degree of abnormal lines, and select the split point that maximizes the inter-class variance as the optimal split point.

[0013] By automatically calculating the inter-class variance based on the "average degradation rate of the healthy / abnormal groups", the system can find the optimal segmentation point that maximizes the difference between groups without manually setting thresholds, achieving accurate and robust division of healthy and abnormal lines, significantly reducing the false positive rate and improving the reliability of early identification.

[0014] Preferably, the minimum degradation rate threshold is calculated based on the mean and variance of the degradation rate of healthy lines in historical data. When the cluster with the largest average degradation degree is less than the minimum degradation rate threshold, all lines are determined to be healthy lines.

[0015] By introducing a minimum degradation rate threshold, false alarms can be automatically suppressed in extreme scenarios where all data are from healthy lines, ensuring that the system will not falsely report anomalies due to noise or data drift, thereby further improving the robustness and reliability of the detection.

[0016] Preferably, clusters are numbered based on the distance between them and the dividing point, and the numbering of each cluster is normalized to obtain the danger level of each cluster. The danger level of each cluster is then subtracted from 1 to obtain the insulation risk index of each cluster in the healthy line group.

[0017] Preferably, the monitoring time period is calculated as follows: a health threshold for the insulation risk index is preset, a basic monitoring period and a minimum monitoring period are set. When the insulation risk index is greater than the preset health threshold, the basic monitoring period is used as the monitoring time period for the cluster. When the insulation risk index is less than the preset health threshold, the insulation risk index is multiplied by the basic monitoring period and added to the minimum monitoring period to obtain the monitoring time period for the cluster.

[0018] Secondly, a low-voltage line insulation anomaly detection system based on an intelligent fusion terminal includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the low-voltage line insulation anomaly detection method based on an intelligent fusion terminal described in any one of the claims is implemented.

[0019] The present invention has the following effects: 1. This method uses temperature-harmonic distortion rate as the dual feature of insulation state of new lines, which improves the accuracy of line degradation segmentation. Based on the insulation state feature vector of each line, clustering is performed to select the optimal number of clusters. Lines with the same degree of degradation are further grouped into the same cluster. By sorting each cluster and selecting the best split point, the accuracy of segmenting healthy and abnormal lines is improved, and errors caused by subjectively set thresholds are avoided. Based on the insulation risk index, healthy lines are dynamically monitored, which further improves the timeliness of monitoring and avoids waste of resources.

[0020] 2. By treating the "cumulative value of rising and falling signs" as the degradation trend and dividing it by the window length to obtain the degradation rate, the system can not only sensitively capture early gradual degradation, but also quantify the rate of deterioration with a unified dimension. It can achieve real-time and quantitative assessment of line health status with zero threshold and zero manual parameters, significantly reducing false alarms and missed alarms and reducing maintenance workload. By automatically calculating the inter-class variance through the "average degradation rate of healthy / abnormal groups", the system can find the optimal segmentation point that maximizes the difference between groups without manually setting thresholds, achieving accurate and robust classification of healthy and abnormal lines, significantly reducing the false judgment rate and improving the reliability of early identification.

[0021] 3. By introducing a minimum degradation rate threshold, false alarms can be automatically suppressed in extreme scenarios where all data are from healthy lines, ensuring that the system will not falsely report anomalies due to noise or data drift, thereby further improving the robustness and reliability of the detection. Attached Figure Description

[0022] Figure 1 This is a flowchart of steps S1-S4 in the low-voltage line insulation anomaly detection method based on intelligent fusion terminal of the present invention.

[0023] Figure 2 This is a flowchart illustrating the structure of the low-voltage line insulation anomaly detection system based on an intelligent fusion terminal, as described in this invention. Detailed Implementation

[0024] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0025] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0026] Reference Figure 1 The low-voltage line insulation anomaly detection method based on intelligent fusion terminal includes steps S1-S4, as follows: S1: Collect and preprocess the temperature and current data of each line to obtain the insulation state feature vector of each line, including temperature data and harmonic distortion rate.

[0027] Infrared thermometers are used to collect temperature data of the circuit. The infrared thermometer is installed around the low-voltage line, and it periodically collects the surface temperature of the line and the ambient temperature. The difference between the surface temperature and the ambient temperature is calculated as the line's temperature data. Lines with good insulation performance will not experience leakage due to the complete insulation material covering them; therefore, the surface temperature of the line will be close to the ambient temperature, with the difference approaching zero. However, lines with abnormal insulation performance may experience leakage due to localized insulation damage, causing a localized rise in the surface temperature. Therefore, the difference between the surface temperature of the line and the ambient temperature is an important indicator for judging the insulation effectiveness of the line.

[0028] The current signals of each line are collected using the same sampling period as the temperature data. The current signals are decomposed using Fast Fourier Transform to obtain the fundamental frequency and each harmonic. For each time point, the harmonic distortion rate is calculated based on each harmonic.

[0029] The collected temperature data and harmonic distortion rate are standardized and normalized preprocessed to eliminate the dimensional differences between different features, and the insulation feature vector of the line at each time step is obtained. The insulation feature vector is a two-dimensional feature vector composed of temperature data and harmonic distortion rate.

[0030] S2: Cluster the insulation state feature vectors of each line to select the optimal number of clusters.

[0031] After obtaining the processed feature data of the line, it is necessary to classify the line deterioration based on these features to distinguish between normal lines and deteriorated lines, so as to realize the monitoring of line insulation anomalies. However, since the degree of line deterioration varies, it is difficult to adapt to the fluctuation of the degree of line deterioration if the line is simply divided into two categories or fixed categories, and the classification effect is not very good. Therefore, it is necessary to dynamically classify the line according to the insulation feature vector.

[0032] The optimal number of clusters for each line can be calculated using either the K-means clustering algorithm or the silhouette coefficient calculation method. Specifically, the operation of clustering the lines based on their insulation feature vectors using the silhouette coefficient calculation method is as follows: The number of initial cluster centers is preset. First, a sample is randomly selected as the first cluster center. For other unselected samples, the squared distance from the sample point to the nearest selected cluster center is calculated and divided by the sum of the squared distances from all sample points to the nearest selected cluster center to obtain the squared distance weighted probability of the sample point, which is used as the probability of the sample point being selected as the next initial center, until the preset number of cluster centers are selected.

[0033] Set a range for the number of clusters, iterate through different numbers of clusters, calculate the profile coefficient corresponding to each number of clusters, find the number of clusters that maximizes the profile coefficient as the optimal number of clusters, calculate the distance from each sample point to the initial center of each cluster, assign the sample to the cluster closest to it, calculate the mean of the insulation state feature vector of the sample under each cluster, and use it as the new cluster center. Iterate the above operations until the maximum number of iterations is reached or the cluster centers converge, and finally the line degradation degree within the clusters is similar.

[0034] S3: Based on the trend and speed of the harmonic distortion rate of each line changing over time, calculate the degradation rate of each line, calculate the average degradation rate of all lines in each cluster as the average degradation degree of each cluster, sort the clusters according to the magnitude of the average degradation degree of each cluster, calculate the optimal split point of each cluster after sorting based on the binary classification algorithm, and divide the sorted clusters into healthy line groups and insulation abnormal line groups based on the optimal split point.

[0035] A dynamic time window is set up to collect the harmonic distortion rate of each line at different times within the time window. The trend of the harmonic distortion rate of different lines changing with time is calculated as the line degradation trend. Specifically, the rise and fall of the harmonic distortion rate at adjacent times in the line is compared, and all rise and fall states of the time window are accumulated based on the sign function as the line degradation trend within the time window. Specifically, if the harmonic distortion rate at a time increases compared to the previous time, the degradation trend is incremented by 1; if the harmonic distortion rate at a time decreases compared to the previous time, the degradation trend is decremented by 1; if the harmonic distortion rate at a time does not change compared to the previous time, the degradation trend remains unchanged.

[0036] The degradation rate of the line is obtained by dividing the final degradation trend of the time window by the length of the time window. The degradation trend reflects the trend of the line harmonic distortion rate changing over time, and the degradation rate reflects the speed at which the degradation trend of the line changes over time.

[0037] The size of subsequent time windows is dynamically adjusted based on the degradation rate of the line in the current time window. When the degradation rate in the current window increases, the size of subsequent time windows should be reduced. The degradation factor is calculated by subtracting the degradation rate of the line in the current time window from 1. The size of the base time window is preset, and the product of the degradation factor and the base time window is calculated as the size of the next time window.

[0038] The formula for calculating the size of the subsequent time window is: ; in, For the first The line is in the time window The rate of degradation within, This indicates the size of the base window. The length of the base window can be selected as a longer period, such as one month or one week. This determines the size of the subsequent time window.

[0039] The time window always ends at the current moment, with the starting point dynamically changing according to the size of the time window. The average degradation rate of all lines at the current moment is calculated. Based on the clustering results in step S2, the mean degradation rate of all lines within each cluster is calculated as the average degradation degree of that cluster. The average degradation degrees of each cluster are sorted by magnitude. Preset split points for each cluster; these split points are used to separate healthy and abnormal lines. The mean average degradation degree of all healthy lines is calculated based on these split points, and the mean average degradation rate of all abnormal lines is calculated as the degradation degree of the abnormal lines. All clusters are traversed as split points. Based on the degradation degrees of healthy and abnormal lines, the inter-class variance corresponding to each split point is calculated. The split point that maximizes the inter-class variance is selected as the optimal split point. Healthy and abnormal lines are then determined based on the optimal split point. The method of dividing all data into two classes based on split points and calculating the inter-class variance based on the means of the two classes is existing technology and will not be elaborated further here.

[0040] Based on the mean and variance of the degradation rate of healthy lines in historical data, a minimum degradation rate threshold is preset. When the cluster with the largest average degradation degree is less than the minimum degradation rate threshold, all lines are determined to be healthy lines.

[0041] The formula for calculating the minimum degradation rate threshold is: ;in, This represents the minimum degradation rate threshold. This represents the average rate of degradation of healthy lines in historical data. This represents the variance of the rate of degradation of healthy lines in historical data.

[0042] S4: Based on the distance between each line in the healthy line group and the optimal split point, calculate the insulation risk index of each line in the healthy line group, and dynamically adjust the monitoring time cycle of the line based on the magnitude of the insulation risk index of each line.

[0043] After completing the anomaly assessment of all lines, the insulation status of the lines in the healthy line group is currently normal. However, some lines have a higher probability of evolving into abnormal lines. Therefore, it is necessary to conduct irregular insulation anomaly assessments on the healthy lines to monitor dynamic changes and provide early warnings of anomalies. Clusters closer to the split point within the healthy line group have a greater risk of anomalies and require more frequent monitoring and early warnings.

[0044] Clusters are numbered based on their distance from the split point. Clusters closer to the split point exhibit greater variation. The cluster numbers are then normalized to obtain the risk level of each cluster. The risk level of each cluster is subtracted from 1 to obtain the insulation risk index of each cluster in the healthy line group.

[0045] A health threshold for the preset insulation risk index is established, along with a basic monitoring period and a minimum monitoring period. When the insulation risk index is greater than the preset health threshold, the basic monitoring period is used as the monitoring time period for that cluster. When the insulation risk index is less than the preset health threshold, the basic monitoring period is multiplied by the insulation risk index as a coefficient and added to the minimum monitoring period to obtain the monitoring time period for the cluster. In this embodiment, the preset health threshold can be 0.25, and the size of the basic and minimum monitoring periods can be flexibly set based on the environment in which the line is located.

[0046] Reference Figure 2 The system includes a processor and a memory. The memory stores computer program instructions. When the computer program instructions are executed by the processor, they implement the low-voltage line insulation anomaly detection method based on an intelligent fusion terminal according to the first aspect of the present invention.

[0047] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0048] The low-voltage line insulation anomaly detection method and system based on intelligent fusion terminal provided by this invention successfully solves the problem of insufficient early warning capability caused by the traditional method simply splitting the continuous insulation degradation process into two discrete states. It improves the accuracy of line anomaly judgment and achieves optimized allocation of monitoring resources through dynamic monitoring mechanism, ensuring the timeliness of monitoring.

[0049] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for detecting insulation anomalies in low-voltage lines based on intelligent fusion terminals, characterized in that, include: Temperature and current data of each line are collected and preprocessed to obtain insulation state feature vectors of each line, including temperature data and harmonic distortion rate. Clustering is performed based on the insulation state feature vectors of each line to select the optimal number of clusters; Based on the trend and rate of change of harmonic distortion rate over time for each line, the degradation rate of each line is calculated. The degradation rate is calculated as follows: a time window is selected, and the trend of harmonic distortion rate over time for different lines within the time window is calculated as the degradation trend of the line. The degradation trend is then divided by the length of the time window to obtain the degradation rate of the line. Specifically, the degradation trend is calculated by collecting the harmonic distortion rate of the line at each moment within the time window, comparing the rise and fall of the harmonic distortion rate at adjacent moments, and accumulating all rise and fall states within the time window based on the sign function, which is then used as the degradation trend of the line within the time window. The average degradation rate of all lines in each cluster is calculated as the average degradation degree of each cluster. The clusters are sorted according to the average degradation degree of each cluster. The optimal split point of each cluster after sorting is calculated based on the binary classification algorithm. Based on the optimal split point, the sorted clusters are divided into healthy line groups and insulation abnormal line groups. Based on the distance between each line in the healthy line group and the optimal split point, the insulation risk index of each line in the healthy line group is calculated, and the monitoring time cycle of the line is dynamically adjusted based on the magnitude of the insulation risk index of each line.

2. The method for detecting insulation anomalies in low-voltage lines based on an intelligent fusion terminal according to claim 1, characterized in that, The size of the preset base time window is used as the degradation factor, which is the deterioration rate of the line in the current time window minus 1. The product of the degradation factor and the base time window is used as the size of the next time window.

3. The method for detecting insulation anomalies in low-voltage lines based on an intelligent fusion terminal according to claim 1, characterized in that, The optimal split point is calculated as follows: Predetermine the split points for each cluster, calculate the mean of the average degradation degree of all healthy lines based on the split points as the degradation degree of healthy lines, calculate the mean of the average degradation rate of all abnormal lines as the degradation degree of abnormal lines, traverse all clusters as split points, calculate the inter-class variance corresponding to each split point based on the degradation degree of healthy lines and the degradation degree of abnormal lines, and select the split point that maximizes the inter-class variance as the optimal split point.

4. The method for detecting insulation anomalies in low-voltage lines based on an intelligent fusion terminal according to claim 3, characterized in that, The minimum degradation rate threshold is calculated based on the mean and variance of the degradation rate of healthy lines in historical data. When the cluster with the largest average degradation degree is less than the minimum degradation rate threshold, all lines are determined to be healthy lines.

5. The method for detecting insulation anomalies in low-voltage lines based on an intelligent fusion terminal according to claim 1, characterized in that, The insulation risk index is calculated as follows: Clusters are numbered based on their distance from the split point, and the numbers of each cluster are normalized to obtain the danger level of each cluster. The danger level of each cluster is then subtracted from 1 to obtain the insulation risk index of each cluster in the healthy line group.

6. The method for detecting insulation anomalies in low-voltage lines based on an intelligent fusion terminal according to claim 1, characterized in that, The monitoring time period is calculated as follows: A health threshold for the insulation risk index is preset, and a basic monitoring period and a minimum monitoring period are set. When the insulation risk index is greater than the preset health threshold, the basic monitoring period is used as the monitoring time period for the cluster. When the insulation risk index is less than the preset health threshold, the insulation risk index is multiplied by the basic monitoring period and added to the minimum monitoring period to obtain the monitoring time period for the cluster.

7. The method for detecting insulation anomalies in low-voltage lines based on an intelligent fusion terminal according to claim 1, characterized in that, The optimal number of clusters is selected as follows: The system presets the initial number of cluster centers and the initial cluster centers, sets the range of the number of clusters, iterates through different numbers of clusters, calculates the silhouette coefficient corresponding to each number of clusters, and finds the number of clusters that maximizes the silhouette coefficient as the optimal number of clusters.

8. A low-voltage line insulation anomaly detection system based on an intelligent fusion terminal, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the low-voltage line insulation anomaly detection method based on an intelligent fusion terminal according to any one of claims 1-7.

Images