Abnormal electricity utilization real-time monitoring and positioning method, system and device based on LMU and storage medium

Abstract

The invention discloses an LMU-based abnormal electricity utilization real-time monitoring and positioning method, system and device and a storage medium, and belongs to the technical field of electric power system operation monitoring, and the method comprises the steps: carrying out the probability sorting, setting the differentiated collection frequency, collecting the voltage, current and electric quantity data of a target ammeter, forming a data set for abnormal recognition, and carrying out the abnormal recognition of the target ammeter; executing zero-live wire ratio and correlation analysis and correlation analysis of electric quantity and line loss data, and identifying a preliminary abnormal electric meter; and adjusting an acquisition strategy, re-calculating a correlation index, dividing a confidence level according to a result change, and determining whether abnormal positioning is completed or not. And when the abnormality is not positioned, clustering analysis is performed on the data of the electric meters in the transformer area, and the installation position of the line monitoring unit is determined. According to the method, the problems that abnormal features are covered by noise, the positioning range is too large and the manual troubleshooting efficiency is low in a traditional fixed acquisition and centralized analysis mode are solved, and the abnormal range can be quickly converged in the scene of large ammeter scale and strong load fluctuation.

Description

Technical Field

[0001] This invention relates to the field of power system operation monitoring technology, specifically to a method, system, device, and storage medium for real-time monitoring and location of abnormal power consumption based on an LMU (Local Measurement Unit). Background Technology

[0002] Abnormal electricity usage, such as electricity theft, leakage, and overloading, not only threatens the safe and stable operation of the power system but also seriously disrupts market order, causing huge economic losses to the country and society. Therefore, it is particularly important to study efficient and accurate abnormal electricity usage detection solutions.

[0003] Traditional methods for detecting abnormal electricity usage often rely on manual judgment and experience-based analysis, which suffers from problems such as lack of specificity, long processing times, and high false positive rates. Therefore, it is necessary to introduce new technologies and methods to improve the efficiency and accuracy of abnormal electricity usage detection.

[0004] In the era of big data, data and information across various industries are experiencing explosive growth. As a fundamental industry, the power sector possesses massive amounts of electricity consumption data. This data contains a wealth of information, and through big data analysis and mining techniques, it is possible to reveal patterns in user electricity consumption and promptly detect abnormal electricity usage issues. However, traditional statistical methods struggle to deeply mine the information behind the data, leading to frequent "data disasters" and "data waste." Therefore, it is necessary to utilize advanced data analysis technologies to improve data utilization and provide strong support for the detection of abnormal electricity usage.

[0005] The development of IoT technology has provided new ideas for power system operation monitoring. Through sensor networks, the operating status of electrical equipment, such as power consumption, voltage, and current, can be monitored in real time, and the data can be transmitted to monitoring equipment in real time. Combined with artificial intelligence (AI) technology, intelligent identification and fault prediction of electrical equipment can be performed, allowing for the early detection of potential safety hazards. Furthermore, AI technology can analyze and process massive amounts of electricity consumption data, building abnormal electricity consumption detection models and improving the accuracy and efficiency of detection.

[0006] With the continuous maturation of machine learning technology, its application in abnormal electricity consumption detection is becoming increasingly widespread. By collecting electricity consumption data from households or businesses, and using machine learning algorithms for data preprocessing, feature extraction, and model training, efficient abnormal electricity consumption detection models can be constructed. These models can automatically analyze user electricity consumption characteristics, identify abnormal electricity consumption patterns, and provide strong support for the management and decision-making of power companies.

[0007] However, since most low-voltage distribution areas with abnormal power consumption are large in scale, have many devices, and have complex physical topologies, it is difficult to achieve full analysis and processing of power consumption data and accurate location of abnormal power consumption using existing equipment. Therefore, this invention proposes a method to narrow down the search range by installing monitoring devices (LMUs) at smaller branches, thereby improving the efficiency and accuracy of abnormal power consumption detection. Summary of the Invention

[0008] To address the aforementioned technical issues, this invention proposes a real-time monitoring and location method for abnormal electricity consumption based on LMU (Local Measurement Unit), comprising: extracting features based on historical voltage data of each meter in the distribution area, calculating the correlation between meters, generating an anomaly probability ranking, and setting differentiated collection frequencies. Based on the collection frequency, voltage, current, and power data of the target meter are collected to form a dataset for anomaly identification. Perform calculations on the ratio of live and neutral currents, correlation analysis on live and neutral currents, and correlation analysis on electricity consumption and line loss in the dataset to identify preliminary abnormal meters. For the initially abnormal electricity meters, adjust the data collection strategy and recalculate various indicators. Based on the changes in the results, classify the confidence level and determine whether the anomaly location has been completed. When the anomaly cannot be located, cluster analysis is performed on the meter data within the transformer area to determine the installation location of the line monitoring unit.

[0009] As a preferred embodiment of the LMU-based real-time monitoring and location method for abnormal electricity consumption described in this invention, the step of extracting features based on historical voltage data of each meter in the distribution area, calculating the correlation between meters, generating anomaly probability ranking, and setting differentiated collection frequencies includes: For a single meter voltage data sequence ,definition The abnormal features of the electricity meter extracted based on voltage data are expressed as follows: ; in, For the convolution result, voltage sequence Using convolution kernels Convolution operation, For convolution kernel, , For the first Voltage value at each point For the first The convolution kernel weights at each point, The kernel length is [length]. The Pearson correlation coefficient is calculated based on the abnormal characteristic data of the electricity meter extracted from the voltage data. The expression is as follows: ; in, It is the Pearson correlation coefficient between meter A and meter B. and These are electricity meters A and electricity meter B After convolution, the i-th data point in the sequence, It is the number of data points in the voltage data sequence. and These are electricity meters A and electricity meter B The mean of the voltage data sequence after convolution; The convolution results of all electricity meters are used to calculate the Pearson correlation coefficient with other electricity meters. The average correlation between each electricity meter and the two electricity meters with the lowest correlation in the distribution area is determined. The electricity meters are sorted in ascending order according to the average correlation to generate a list of abnormal electricity consumption probability for each electricity meter. The electricity meters with the highest ranking are identified as high-priority objects, and the collection frequency of high-priority objects is increased.

[0010] The beneficial effects of this preferred technical solution are as follows: By employing a method based on voltage data convolution processing and correlation analysis, meters within the distribution area that exhibit significantly different fluctuation trends from other meters can be identified in advance and given higher priority during data acquisition. This avoids the problem of missing short-term anomalies or sporadic fluctuations that is easily encountered in traditional fixed-frequency acquisition, reduces reliance on full-data acquisition, improves acquisition efficiency, and also reduces communication pressure on the distribution area side.

[0011] As a preferred embodiment of the LMU-based real-time monitoring and location method for abnormal electricity consumption described in this invention, the step of collecting voltage, current, and electricity consumption data of the target meter according to the acquisition frequency to form an abnormal identification data set includes, Under the set differentiated sampling frequency, periodic sampling operations are performed on the target electricity meter to obtain time series data of voltage, current, and electricity consumption; The original sampled values ​​of each meter within the sampling period are associated with the corresponding feature extraction results to construct a structured data set containing timestamps, data fields, and feature fields.

[0012] As a preferred embodiment of the LMU-based real-time monitoring and location method for abnormal electricity consumption described in this invention, the step of calculating the ratio of live and neutral currents, performing correlation analysis of live and neutral currents, and analyzing the correlation between electricity consumption and line loss to identify preliminary abnormal meters includes: Calculate the mean ratio of neutral current to live current of the target meter within a set time window, and the Pearson correlation coefficient between the neutral current and the live current. Determine whether the mean falls within the first set range and whether the Pearson correlation coefficient is less than the first threshold. If either condition is met, the live and neutral wires are abnormal, and the meter with the initial abnormality is locked. The power consumption data of the target meter and the line loss data of the corresponding transformer area are subjected to third-order Bessel filtering, that is, the filtered data sequence is obtained by substituting into the difference equation; the Pearson correlation coefficient between the filtered power consumption data and the line loss data is calculated. If the Pearson correlation coefficient is greater than the first threshold, it is determined that the line loss is abnormal and the meter with the initial abnormality is identified.

[0013] As a preferred embodiment of the LMU-based real-time monitoring and location method for abnormal power consumption described in this invention, wherein: The expression for the difference equation is: ; in, For a sequence of length L, the first... The result after filtering the point sequence For a sequence of length L, the first... Input data sequence at points, and These are the coefficients that determine the characteristics of the filter. ; 、 、 、 The first The point, the first The point, the first The point, the first One point.

[0014] As a preferred embodiment of the LMU-based real-time monitoring and location method for abnormal electricity consumption described in this invention, the step of adjusting the data acquisition strategy for the initial abnormal electricity meter and recalculating various indicators, classifying confidence levels based on changes in results, and determining whether abnormal location has been completed includes: For meters that have been initially identified as abnormal, voltage, current and power data are collected again by increasing the corresponding collection frequency and optimizing the collection sequence. The mean value of the ratio of neutral and live wire current, the correlation between neutral and live wire current, and the correlation index between electricity consumption and line loss were recalculated for the re-collected data using the same anomaly judgment method as the initial judgment. The recalculated indicators are compared with the initial judgment results. If the average value of the ratio of live and neutral current is within the set range and the correlation index of live and neutral current is greater than or equal to the first threshold, or the correlation index of power consumption and line loss is greater than or equal to the first threshold, it is marked as a high-confidence anomaly, and the anomaly location result is directly output. If the mean of the ratio of live and neutral currents is not within the set range and the correlation index of live and neutral currents decreases and is less than or equal to the first threshold, or the correlation index of power consumption and line loss decreases and is less than the first threshold, then it is marked as a low confidence anomaly and no anomaly location result is output. If the mean of the neutral-live wire current ratio is not within the set range and the neutral-live wire current correlation index is less than the first threshold and the power consumption and line loss correlation index is greater than or equal to the first threshold, or if the mean of the neutral-live wire current ratio is not within the set range and the neutral-live wire current correlation index is greater than or equal to the first threshold and the power consumption and line loss correlation index is less than the first threshold, or if the mean of the neutral-live wire current ratio is not within the set range and the neutral-live wire current correlation index increases and is less than or equal to the first threshold, or if the power consumption and line loss correlation index increases and is less than the first threshold, then it is marked as a medium confidence anomaly. No anomaly location result is output. A new round of voltage, current and power consumption data is collected repeatedly. The mean of the neutral-live wire current ratio, the neutral-live wire current correlation, and the power consumption and line loss correlation index are calculated. The judgment is iteratively judged until the high confidence anomaly or low confidence anomaly is met.

[0015] As a preferred embodiment of the LMU-based real-time monitoring and location method for abnormal power consumption described in this invention, the step of performing cluster analysis on the meter data within the transformer area to determine the installation location of the line monitoring unit when the abnormality cannot be located includes: When a new round of voltage, current, and power data is collected and the various indicators are recalculated, but the precise location of the abnormal meter has not yet been output, it is necessary to perform convolution smoothing on the voltage data of all meters in the transformer area and calculate the Pearson correlation coefficient between the meters. Based on the correlation results, the DBSCAN clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) is used to complete the meter clustering. According to the clustering results, a line monitoring unit is installed at the line bifurcation point corresponding to each meter cluster. The expression for convolution smoothing is: ; in, To smooth the convolution result, the convolution kernel is defined as follows: , The kernel length is 1. voltage sequence Using convolution kernels Convolution operation, For the first Voltage value at each point For convolution smoothing The convolution kernel weights at each point.

[0016] The beneficial effects of this preferred technical solution are as follows: When abnormal meters cannot be directly located, clustering meters with similar voltage fluctuation trends and installing monitoring equipment at the corresponding line branch nodes helps to control the location range within a smaller branch. In practical field applications, this reduces false alarm areas and lowers the workload of subsequent investigations, especially in areas with a large number of meters and significant data fluctuations, demonstrating strong practicality.

[0017] This invention provides a real-time monitoring and location system for abnormal power consumption based on LMU. This invention solves the problems of existing technologies, such as difficulty in locating abnormal power consumption, low investigation efficiency, and reliance on manual experience, in scenarios with a large number of electricity meters in the distribution area, large data fluctuations, or unclear abnormal characteristics.

[0018] As a preferred embodiment of the LMU-based real-time monitoring and location system for abnormal power consumption described in this invention, it is characterized by comprising: a feature extraction module, an anomaly identification module, an analysis module, an anomaly location module, and an output module. The feature extraction module is based on the historical voltage data of each meter in the transformer area, extracts features to calculate the correlation between meters, generates an anomaly probability ranking, and sets differentiated collection frequencies. The anomaly identification module collects voltage, current, and power data from the target meter based on the collection frequency, forming a data set for anomaly identification. The analysis module performs zero-live-wire ratio calculation and correlation analysis on the data set, as well as correlation analysis between electricity consumption and line loss data, to identify preliminary abnormal meters; The anomaly location module adjusts the data collection strategy for the initial abnormal electricity meter and recalculates various indicators, classifies the confidence level based on the changes in the results, and determines whether the anomaly location has been completed. The output module performs cluster analysis on the meter data within the transformer area to determine the installation location of the line monitoring unit when the anomaly cannot be located.

[0019] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the LMU-based real-time monitoring and location method for abnormal power consumption.

[0020] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the LMU-based real-time monitoring and location method for abnormal power consumption.

[0021] The beneficial effects of this invention are as follows: This invention proposes a scheme that uses different acquisition strategies for different electricity meters based on different weights. By differentiating the acquisition strategies for different electricity meters, it effectively solves the problems of limited data acquisition range, insufficient algorithm accuracy, and poor algorithm adaptability in existing algorithms. At the same time, it proposes to decentralize the data processing and analysis process to the edge side, thereby dispersing the pressure of data processing and analysis, and efficiently and accurately extracting information related to abnormal electricity consumption. This significantly improves the problems of insufficient data processing capabilities, poor real-time monitoring, and slow response speed in existing algorithms.

[0022] Meanwhile, by embedding the algorithm within edge devices to perform large-scale analysis of historical electricity consumption behavior, this invention reduces the workload of manual analysis and judgment and the professional requirements of personnel handling the situation, and greatly improves the automation, digitalization, and intelligence of abnormal electricity consumption monitoring, analysis, and processing. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 The above is a flowchart of an LMU-based real-time monitoring and location method for abnormal power consumption, provided as an embodiment of the present invention. Detailed Implementation

[0025] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0026] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for real-time monitoring and location of abnormal power consumption based on an LMU, including: Based on the historical voltage data of each meter in the distribution area, features are extracted to calculate the correlation between meters, anomaly probability ranking is generated, and differentiated collection frequencies are set.

[0027] Based on the collection frequency, voltage, current, and power data of the target meter are collected to form a data set for anomaly identification.

[0028] The average value of the ratio of live and neutral currents, the correlation analysis of live and neutral currents, and the correlation analysis of electricity consumption and line loss are performed on the dataset to identify preliminary abnormal meters.

[0029] For the initially abnormal electricity meters, adjust the data collection strategy and recalculate various indicators. Based on the changes in the results, classify the confidence level and determine whether the anomaly location has been completed.

[0030] When the anomaly cannot be located, cluster analysis is performed on the meter data within the transformer area to determine the installation location of the line monitoring unit.

[0031] In the existing process of detecting abnormal electricity consumption in distribution substations, conventional methods mainly rely on fixed-frequency data collection and centralized unified analysis. However, in substations with a large number of meters, significant load variations, and complex electricity consumption behaviors, problems often arise such as data fluctuations masking anomalies, unstable feature identification, and the inability to accurately and promptly locate abnormal meters. Furthermore, for situations where meter data cannot be directly used for location, manual segmented investigation is often necessary, resulting in long investigation cycles, heavy workloads, and a high risk of omissions.

[0032] This invention improves the data acquisition efficiency and targeting of high-risk meters by constructing anomaly ranking based on voltage characteristic differences and dynamically adjusting the collection frequency through analysis and judgment. In the judgment stage, a dual-path identification mechanism is introduced, which combines the correlation analysis of live and neutral wire current and power consumption-line loss, to enhance the ability to identify different types of anomalies. When the anomaly has not yet converged, line monitoring units are deployed based on the cluster analysis results to divide the transformer area into controllable sub-segments, and then local iterative positioning is performed.

[0033] The above solution solves the problems of coarse anomaly identification granularity, experience-dependent location path, and low overall investigation efficiency in traditional methods. In the test scenario, it can realize the analysis and location of power consumption anomalies in large-scale transformer areas, has strong on-site adaptability, and significantly improves the investigation efficiency.

[0034] Example 2, the second embodiment of the present invention, provides a method for real-time monitoring and location of abnormal power consumption based on LMU.

[0035] Based on the historical voltage data of each meter in the distribution area, features are extracted to calculate the correlation between meters, anomaly probability ranking is generated, and differentiated collection frequencies are set.

[0036] The current common data collection strategy is to collect data at a fixed frequency according to the order of network access. However, the data delay of abnormal meters cannot be controlled during real-time data collection, making it impossible to detect abnormal meters in a timely manner. At the same time, this strategy is also prone to overlooking abnormal problems that occur within the collection interval. Therefore, this invention proposes a method to increase the collection frequency and locate abnormalities by sorting the meters according to their abnormality probability. This method prioritizes monitoring the meters most likely to be abnormal, thereby detecting problems in a timely manner and reducing the probability of misjudgment.

[0037] By performing differential processing on the data, abnormal data characteristics caused by the electricity consumption behavior of a single or a few meters within the distribution area can be located.

[0038] For a single meter voltage data sequence ,definition The convolution used to extract abnormal features from the electricity meter is as follows: ; in, For the convolution result, voltage sequence Using convolution kernels Convolution operation, For convolution kernel, , For the first Voltage value at each point For the first The convolution kernel weights at each point, is the kernel length.

[0039] A further point to clarify is that the voltage of each electricity meter fluctuates based on the supply voltage of the distribution area. Since the voltage fluctuations caused by the user's own electricity consumption are much smaller than the changes in the supply voltage, directly analyzing the original voltage anomalies would mask them. Therefore, this invention extracts high-frequency voltage features through convolution, effectively identifying voltage changes caused by the user's own electricity consumption, and thus accurately capturing abnormal electricity consumption characteristics of the meters.

[0040] The Pearson correlation coefficient is calculated based on the abnormal characteristic data of the electricity meter extracted from the voltage data. The expression is as follows: ; in, It is an electricity meter A and electricity meter B The Pearson correlation coefficient between them and These are electricity meters A and electricity meter B After convolution, the first... i Data points, It is the number of data points in the voltage data sequence. and These are electricity meters A and electricity meter B The mean of the voltage data sequence after convolution.

[0041] The correlation of all meters is calculated. By differencing and convolving the data, the main changes in voltage data are caused by the electricity consumption characteristics of each meter itself. Based on experience in abnormal electricity consumption analysis, the greater the difference between the electricity consumption characteristics of a meter and those of other meters in the distribution area, the more likely that meter is an abnormal electricity consumption meter. Therefore, meters are sorted in ascending order according to the average correlation value of the two meters with the lowest correlation to the rest of the distribution area, resulting in a collection order from high to low probability of abnormal electricity consumption. The collection frequency should be increased for meters in the top 10% of those with the highest probability of abnormal electricity consumption to facilitate subsequent anomaly detection.

[0042] Based on the collection frequency, voltage, current, and power data of the target meter are collected to form a data set for anomaly identification.

[0043] Under the set differentiated sampling frequency, periodic sampling operations are performed on the target meter to obtain time series data of voltage, current, and power consumption.

[0044] The original sampled values ​​of each meter within the sampling period are associated with the corresponding feature extraction results to construct a structured data set containing timestamps, data fields, and feature fields.

[0045] The average value of the ratio of live and neutral currents, the correlation analysis of live and neutral currents, and the correlation analysis of electricity consumption and line loss are performed on the dataset to identify preliminary abnormal meters.

[0046] There are two methods to identify the initial abnormal meters. One method is based on the average ratio of live and neutral current and the comparison of correlation indicators of live and neutral current. The other method is to compare the correlation between the target meter's electricity data and the line loss data of the transformer area.

[0047] Calculate the mean ratio of neutral current to live current of the target meter within a set time window, and the Pearson correlation coefficient between the neutral current and the live current.

[0048] Determine whether the mean falls within the first set range and whether the Pearson correlation coefficient is higher than the first threshold. If either condition is met, the live and neutral wires are abnormal, and the meter with the initial abnormality is locked.

[0049] In the implementation of this application: when determining the correlation between the average value of the neutral-live current ratio and the neutral-live current, a sliding window judgment of data is required for 4 consecutive hours. If the average value of the neutral-live current ratio is not in the range of 0.8 to 1.2, or the correlation between the neutral-live current and the current is greater than 0.8, the neutral-live current is considered abnormal, and the preliminary abnormal meter is locked.

[0050] In this invention, the first set range and the first threshold are empirical values ​​obtained from normal experiments, which are [0.8, 1.2] and 0.8 respectively.

[0051] The electricity consumption data of the target meter and the line loss data of the transformer area are subjected to third-order Bessel filtering to obtain the filtered data sequence. The Pearson correlation coefficient between the filtered electricity consumption data and the line loss data is calculated. If the Pearson correlation coefficient is greater than the first threshold, the line loss is determined to be abnormal, and the abnormal meter is initially identified.

[0052] It should be further explained that calculating the correlation between electricity meter usage and line loss is one way to identify abnormal electricity consumption within a distribution area; meters with a high correlation may be abnormal. During the correlation calculation process, since the line loss data includes fixed losses and metering errors, this invention uses Bessel hierarchical filtering to remove noise during the data acquisition process to avoid data shifts caused by noise processing and to better preserve the waveform of abnormal signals.

[0053] A Bessel filter is an analog filter, and its representation is as follows: ; in, It is the representation of the transfer function in the Laplace domain. It is a complex frequency variable. This is an optional gain constant, typically set to 1 during normalization. It is a Béssel polynomial, and the coefficients of a Béssel polynomial are obtained by its specific recursive relation or by looking up a table.

[0054] In the filtering process, this invention selects a third-order Bessel polynomial for filtering, as follows: ; in, It is a third-order Bessel polynomial.

[0055] At the same time, the analog filter is converted into a digital filter using a bilinear transform, changing the frequency from... Domain (continuous-time system) mapped to The domain (discrete-time system) is as follows: ; in, It is the sampling period. It is a numeric field variable. It is a complex frequency variable. A digital field variable that is sampled with a delay of one cycle.

[0056] Since the required cutoff frequencies vary during the calculation process, a frequency scaling factor is added to set the cutoff frequency: ; in, For the scaled complex frequency variable, It is the angular cutoff frequency. This is the cutoff frequency.

[0057] Substituting and expanding, we can obtain exist The third-order transfer function over the domain (discrete-time system) is: ; in, 、 、 These represent digital field variables sampled with delays of one period, two periods, and three periods, respectively. and These are the coefficients that determine the characteristics of the filter. This reflects the filter's feedforward and feedback weights on the signal, and is determined by the sampling period and cutoff frequency.

[0058] The final form, converted to a difference equation, is: ; in, For the first The result after filtering the point sequence For the first Input data sequence at points, and These are the coefficients that determine the characteristics of the filter. ， 、 、 、 The first The point, the first The point, the first The point, the first One point.

[0059] Based on the filtered data, the Pearson correlation coefficient between the electricity consumption of each meter in the distribution area and the line loss is calculated. When the correlation coefficient is higher than 0.8, the meter is considered to be abnormal, and the meter with the preliminary abnormality is identified.

[0060] For the initially abnormal electricity meters, adjust the data collection strategy and recalculate various indicators. Based on the changes in the results, classify the confidence level and determine whether the anomaly location has been completed.

[0061] For meters that have been initially identified as abnormal, voltage, current and power data are collected again by increasing the corresponding collection frequency and optimizing the collection sequence.

[0062] The recalculated indicators are compared with the initial judgment results. If the average value of the ratio of live and neutral current is within the set range and the correlation index of live and neutral current is greater than or equal to the first threshold, or the correlation index of power consumption and line loss is greater than or equal to the first threshold, it is marked as a high-confidence anomaly, and the anomaly location result is directly output. If the mean of the ratio of live and neutral currents is not within the set range and the correlation index of live and neutral currents decreases and is less than or equal to the first threshold, or the correlation index of power consumption and line loss decreases and is less than the first threshold, then it is marked as a low confidence anomaly and no anomaly location result is output. If the mean of the neutral-live wire current ratio is not within the set range and the neutral-live wire current correlation index is less than the first threshold and the power consumption and line loss correlation index is greater than or equal to the first threshold, or if the mean of the neutral-live wire current ratio is not within the set range and the neutral-live wire current correlation index is greater than or equal to the first threshold and the power consumption and line loss correlation index is less than the first threshold, or if the mean of the neutral-live wire current ratio is not within the set range and the neutral-live wire current correlation index increases and is less than or equal to the first threshold, or if the power consumption and line loss correlation index increases and is less than the first threshold, then it is marked as a medium confidence anomaly. No anomaly location result is output. A new round of voltage, current and power consumption data is collected repeatedly. The mean of the neutral-live wire current ratio, the neutral-live wire current correlation, and the power consumption and line loss correlation index are calculated. The judgment is iteratively judged until the high confidence anomaly or low confidence anomaly is met.

[0063] In the implementation of this application: for the initially abnormal electricity meter that has been locked, it is necessary to reduce noise again by adjusting the collection sequence and collection frequency.

[0064] For meters with initial abnormalities identified by the lock-in, increasing the sampling frequency enhances the accuracy and timeliness of abnormality judgment within the continuous sliding window time range, thus avoiding missed or false judgments caused by sudden abnormalities.

[0065] A further point to clarify is that by dynamically and flexibly adjusting the sampling frequency and sliding window, some discrete and intermittent abnormal power consumption caused by traditional fixed time interval sampling can be detected in a timely and accurate manner, while also effectively avoiding misjudgments caused by occasional anomalies.

[0066] For meters with abnormal line loss correlation, the impact of noise caused by time deviation on the calculation can be reduced by advancing the collection order of abnormal meters, placing the collection order of meters with larger power consumption in the middle, and the collection order of meters with smaller power consumption at the end.

[0067] After recalculating all indicators, anomaly judgment is performed again. If the average value of the ratio of live and neutral current is within the set range of 0.8 to 1.2 and the correlation index of live and neutral current is greater than or equal to the first threshold of 0.8, or the correlation index of power consumption and line loss is greater than or equal to the first threshold of 0.8, then it is marked as a high-confidence anomaly, and the anomaly location result is directly output. If the average value of the ratio of live and neutral currents is not within the set range of 0.8 to 1.2 and the correlation index of live and neutral currents decreases and is less than or equal to the first threshold of 0.8, or the correlation index of power consumption and line loss decreases and is less than the first threshold of 0.8, then it is marked as a low confidence anomaly and no anomaly location result is output. In other cases, the anomaly is marked as medium confidence and no anomaly location result is output. A new round of voltage, current and power data can be collected repeatedly to calculate the mean of the ratio of live and neutral currents, the correlation between live and neutral currents and the correlation between power and line loss. The judgment is iterated until the anomaly is satisfied with either high confidence or low confidence.

[0068] If it is ultimately determined that there is still an anomaly in the transformer area, and there is no high-confidence abnormal meter location under the transformer area, then an LMU will be installed to perform iterative precise location operations.

[0069] Further points to be made: by simply adding a small number of sensors, more precise iterations can be achieved, avoiding the manual costs of meticulous problem-solving. At the same time, the automatically calculated installation locations also greatly reduce the time and labor costs of manual on-site investigations before installation.

[0070] When the anomaly cannot be located, cluster analysis is performed on the meter data within the transformer area to determine the installation location of the line monitoring unit.

[0071] In the implementation of this application: For large-scale power distribution areas, the excessive number of meters leads to excessive data noise. Adjustment methods such as convolution, filtering, and adjusting the acquisition order cannot make the required features of the data frame greater than the noise, ultimately failing to accurately locate the anomaly. Therefore, by analyzing and processing the large feature signals of power consumption in the power distribution area, and adding LMUs to the corresponding large signal nodes, the branch intervals are reduced to small power distribution areas, reducing the problem of excessive noise caused by the excessive number of meters, gradually narrowing the scope of problem investigation, and more accurately locating the anomaly.

[0072] To extract the overall characteristics of the large signal in the transformer area, it is necessary to smooth the collected voltage data to reduce the interference of occasional noise in the transformer area on the clustering. This can be achieved by smoothing the collected data from each meter through convolution.

[0073] For a single meter voltage data sequence ,definition The convolution for the clustered signals extracted based on voltage data is as follows: ; in, To smooth the convolution result, the convolution kernel is defined as follows: , The kernel length is 1. voltage sequence Using convolution kernels Convolution operation, For the first Voltage value at each point For convolution smoothing The convolution kernel weights at each point.

[0074] The Pearson correlation coefficient is calculated based on the extracted clustered signal feature data: ; in, It is the Pearson correlation coefficient between meter A and meter B. and These are the i-th data points in the sequence after the convolution of meter A and meter B, respectively. It is the number of data points in the voltage data sequence. and These are the mean values ​​of the voltage data sequences after convolution of meter A and meter B, respectively.

[0075] DBSCAN clustering was used to determine the location of the electricity meter installation.

[0076] Determine the scan radius and the minimum number of samples required to form a dense region.

[0077] Since the choice of scanning radius and minimum sample size has a great impact on the clustering results, the scanning radius is currently set to the 90th percentile of the correlation distance between meters in the transformer substation area. The minimum sample size is set to the number of meters in the meter box if there is data on the number of meters in the meter box in the transformer substation area, and if there is no relevant data, the number 2 is used as the minimum sample size.

[0078] The clustering process is as follows: For each point, check whether the number of points in its scan radius neighborhood has reached the minimum number of samples. If so, create a new cluster and add the point to the cluster.

[0079] Then, recursively add points within the scan radius neighborhood of each point in the cluster to the cluster until no new points can be added.

[0080] The final cluster of meters after clustering is a set of meters with highly correlated electrical signals. According to Kirchhoff's laws, the longer the common line of meters with similar voltage drop fluctuations, the more the LMU can be installed at the corresponding branch point of the line of each meter cluster. This can reduce the range of anomalies and reduce the impact of too many meters and other noise factors.

[0081] By repeating the above steps in each smaller area of ​​the narrowed-down system, the problem can be accurately located.

[0082] Example 3 is the third embodiment of the present invention, which differs from the previous two embodiments in that: If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0083] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0084] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0085] It should be understood that various parts of the present invention can be implemented using a combination of hardware, software, and firmware. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0086] Example 4 is the fourth embodiment of the present invention. This embodiment provides an LMU-based real-time monitoring and location system for abnormal power consumption, including: a feature extraction module, an anomaly identification module, an analysis module, an anomaly location module, and an output module.

[0087] The feature extraction module extracts features based on the historical voltage data of each meter in the transformer area, calculates the correlation between meters, generates anomaly probability ranking, and sets differentiated collection frequencies.

[0088] The anomaly identification module collects voltage, current, and power data from the target meter based on the collection frequency, forming a data set for anomaly identification.

[0089] The analysis module performs correlation analysis on the ratio of live to neutral wires and the correlation analysis on the power consumption and line loss data of the data set to identify preliminary abnormal meters.

[0090] The anomaly location module adjusts the data collection strategy for the initially abnormal electricity meters and recalculates the correlation indicators. Based on the changes in the results, it classifies the confidence level and determines whether the anomaly location has been completed.

[0091] The output module performs cluster analysis on the meter data within the transformer area to determine the installation location of the line monitoring unit when the anomaly cannot be located.

[0092] Example 5, the fifth embodiment of the present invention, provides a method for real-time monitoring and location of abnormal power consumption based on LMU. To verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0093] This method is mainly used to analyze the intrinsic correlation between two time series data. By filtering out noise interference, it can analyze a more fundamental correlation.

[0094] Taking the relationship between electricity meter readings and line loss in the distribution area as an example, data from 96 collection points per day (one point every 15 minutes) were used for analysis.

[0095] The filtering method uses a third-order Bessel filter, which is known for preserving waveforms in signal processing, and the correlation is calculated using the Pearson correlation coefficient.

[0096] For ease of demonstration and calculation, five data points are selected as sample data for subsequent calculations: assumed This is the raw data of electricity consumption from the meter. Data on line loss in the transformer area: =[9.8, 12.5, 15.5, 13.6, 11.3]; =[0.55, 0.58, 0.86, 0.65, 0.51]; The most prominent feature of Bessel filters is their maximum flat group delay (linear phase response). This means that when processing signals, they can preserve the waveform of the original signal to the greatest extent possible without overshooting or ringing. They are well-suited for processing time-series data that need to maintain their shape, such as charge curves.

[0097] The implementation of digital filtering relies on difference equations, which are usually derived from analog filter prototypes through the "bilinear transform method".

[0098] Prototype transfer function of a third-order Bessel low-pass filter.

[0099] The transfer function of a standardized third-order Bessel low-pass filter for: ; Where s is a complex variable of the Laplace transform.

[0100] To apply this analog filter to a discrete digital signal (96 data points), a bilinear transform is used, with the following substitution rule: ; in, It is the sampling period. This is the unit delay operator in the Z-transform. In the current embodiment, there are 96 points per day, so the sampling period is... Minutes. To simplify calculations, the frequency is usually normalized.

[0101] After complex algebraic substitutions and simplifications, the transfer function of the digital filter can be obtained. : ; in, and These are the Z-transforms of the input signal (raw data) and the output signal (filtered data), respectively.

[0102] Performing an inverse Z-transform on the above equation yields the difference equation used for practical calculations: ; The current formula indicates that the filtered data at the current time... It consists of raw data from the current and several past moments ( ) and filtered data from the past few moments ( It was decided jointly.

[0103] In order to obtain specific and The coefficient needs to have a cutoff frequency set. The cutoff frequency determines the strength of the filter; the smaller the value, the more fluctuations are filtered out, and the smoother the curve. It is a normalized frequency between 0 and 1, with 1 corresponding to the Nyquist frequency (half the sampling rate).

[0104] Select cutoff frequency as The coefficients of the difference equation for a third-order Bessel filter at that cutoff frequency can be easily obtained using scientific computing libraries (such as Python's SciPy).

[0105] The obtained coefficients (examples) are as follows: ; ; Therefore, the difference formula is as follows: ; Before the calculation begins (i.e.) Set all historical data to zero: , .

[0106] Calculation Sample Selection Perform the calculation.

[0107] Calculate the output sequence The process is as follows: Step 1: Calculation (n=0).

[0108] formula: .

[0109] Substitute: , , , , , All are 0; ; ; result: .

[0110] Step 2: Calculation (n=1).

[0111] formula: ; Substitute: and ; ; ;

[0112] result: .

[0113] Step 3: Calculation (n=2).

[0114] formula: ; Substitute: ; ;

[0115] result: .

[0116] Step 4: Calculation (n=3).

[0117] formula: ; Substitute: ; ;

[0118] result: .

[0119] Step 5: Calculation (n=4).

[0120] formula: ; Substitute: ; ;

[0121] result: .

[0122] Similarly, the actual calculation process involves iteratively calculating multiple days' worth of data collected daily from 96 data points to obtain a smooth, filtered data sequence. The same filtering process is then applied to the transformer substation line loss data.

[0123] After obtaining two sets of filtered and smoothed data (one set being the meter reading), The other group is the transformer area line loss. This invention uses the Pearson correlation coefficient to measure the linear relationship between them.

[0124] The Pearson correlation coefficient r is a value between -1 and 1.

[0125] r=1 indicates a perfectly positive linear correlation. r=-1 indicates a perfectly negative linear correlation. r=0 indicates no linear correlation.

[0126] The formula for calculating the Pearson correlation coefficient is as follows: ; in, It is the number of data points ( ), and They are the first The filtered power value and filtered line loss value at each time point.

[0127] Calculate the Pearson correlation coefficient The steps are as follows: Prepare data: List two sets of filtered data pairs. .

[0128] Calculate the sum of each term: This is the sum of all filtered battery values. This is the sum of all filtered line loss values. This is the sum of squares of all filtered battery values. The sum of squares of all filtered line loss values. It is the sum of the products of the power consumption value and the line loss value at each moment.

[0129] Substitute into the formula: Substitute the above calculation results into the formula for the Pearson correlation coefficient to calculate the final r value.

[0130] The five data points from both original sets of data, after undergoing third-order Bessel filtering, are as follows: Filtered power =[0.00764, 0.05232, 0.18179, 0.43969, 0.84596]; Filtering background area line loss =[0.00043, 0.00284, 0.00966, 0.02308, 0.04308]; here .

[0131] Refer to Table 1 to calculate the sum of each item.

[0132] Table 1 Data Display Table

[0133] Calculate the correlation: ; in, This is the correlation calculated after differencing.

[0134] molecular: ; Denominator: Part of ; Part of ; The denominator as a whole: ; ; Correlation calculation before filtering and comparison of data before and after filtering.

[0135] The five data points of the original two sets of data are as follows: =[9.8, 12.5, 15.5, 13.6, 11.3]; =[0.55, 0.58, 0.86, 0.65, 0.51]; Refer to Table 2 to calculate the sum of each item.

[0136] Table 2 Data Display Table

[0137] Calculate the correlation: ; in, Calculate the correlation of the raw data.

[0138] molecular: ; Denominator: Part of: ; Part of ; The denominator as a whole: ; ; Data correlation after smoothing.

[0139] Filtered power =[0.00764, 0.05232, 0.18179, 0.43969, 0.84596]; Filtering background area line loss =[0.00043, 0.00284, 0.00966, 0.02308, 0.04308]; Calculation results (this is from previously completed calculations): ; in, This is the result after filtering.

[0140] Refer to Table 3 for comparison and analysis.

[0141] Table 3 Comparative Analysis Table

[0142] The filtering operation significantly increased the correlation coefficient between power consumption and line loss from 0.890 to 0.9997, demonstrating that filtering has a significant effect on the effective extraction of data features.

[0143] Before filtering While the result of 0.890 is good, it is actually underestimated.

[0144] The reason is that the random noise in power measurement and the random noise in line loss measurement are uncorrelated, and their existence together "dilutes" the strong correlation between the main trends of the two variables.

[0145] By using Bessel filtering, high-frequency noise was removed from both sets of data, retaining only their smooth, core trends. When comparing the trends of these two sets of clean data, a near-perfect linear relationship (r≈0.9997) can be observed, thus enabling a more accurate correlation between line loss and abnormal meters when locating line loss anomalies.

[0146] In summary, this comparison demonstrates that when processing noisy measured time series data (such as electricity meter data), performing filtering preprocessing followed by correlation analysis is a method that yields more accurate and reliable conclusions. This complete "data filtering-correlation analysis" workflow allows for a more accurate assessment of the intrinsic relationships between variables, providing reliable data support for refined line loss management, anomaly diagnosis, and electricity theft analysis.

[0147] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for real-time monitoring and location of abnormal power consumption based on LMU, characterized in that: include, Based on the historical voltage data of each meter in the distribution area, features are extracted to calculate the correlation between meters, anomaly probability ranking is generated, and differentiated collection frequencies are set. Based on the collection frequency, voltage, current, and power data are collected from the target meter to form a dataset for anomaly identification; Perform zero-live wire ratio calculation and correlation analysis on the dataset, as well as correlation analysis between electricity consumption and line loss data, to identify preliminary abnormal meters; For the initially abnormal electricity meters, adjust the data collection strategy and recalculate various indicators. Based on the changes in the results, classify the confidence level and determine whether the anomaly location has been completed. When the anomaly cannot be located, cluster analysis is performed on the meter data within the transformer area to determine the installation location of the line monitoring unit.

2. The method for real-time monitoring and location of abnormal power consumption based on LMU as described in claim 1, characterized in that: The process of extracting features from historical voltage data of each meter in the transformer substation to calculate the correlation between meters, generating an anomaly probability ranking, and setting differentiated data collection frequencies includes... For a single meter voltage data sequence ,definition The abnormal features of the electricity meter extracted based on voltage data are expressed as follows: ； in, For the convolution result, voltage sequence Using convolution kernels Convolution operation, For convolution kernel, , For the first Voltage value at each point For the first The convolution kernel weights at each point, The kernel length is [length]. The Pearson correlation coefficient is calculated based on the abnormal characteristic data of the electricity meter extracted from the voltage data. The expression is as follows: ； in, It is the Pearson correlation coefficient between meter A and meter B. and These are electricity meters A and electricity meter B After convolution, the i-th data point in the sequence, It is the number of data points in the voltage data sequence. and These are electricity meters A and electricity meter B The mean of the voltage data sequence after convolution; The convolution results of all electricity meters are used to calculate the Pearson correlation coefficient with other electricity meters. The average correlation between each electricity meter and the two electricity meters with the lowest correlation in the distribution area is determined. The electricity meters are sorted in ascending order according to the average correlation to generate a list of abnormal electricity consumption probability for each electricity meter. The electricity meters with the highest ranking are identified as high-priority objects, and the collection frequency of high-priority objects is increased.

3. The method for real-time monitoring and location of abnormal power consumption based on LMU as described in claim 2, characterized in that: The process of collecting voltage, current, and electricity data from the target meter based on the collection frequency to form an anomaly identification dataset includes, Under the set differentiated sampling frequency, periodic sampling operations are performed on the target electricity meter to obtain time series data of voltage, current, and electricity consumption; The original sampled values ​​of each meter within the sampling period are associated with the corresponding feature extraction results to construct a structured data set containing timestamps, data fields, and feature fields.

4. The method for real-time monitoring and location of abnormal power consumption based on LMU as described in claim 3, characterized in that: The process of performing zero-live-line ratio calculation and correlation analysis on the data set, as well as correlation analysis between electricity consumption and line loss data, to identify preliminary abnormal meters includes: Calculate the mean ratio of neutral current to live current of the target meter within a set time window, and the Pearson correlation coefficient between the neutral current and the live current. Determine whether the mean falls within the first set range and whether the current Pearson correlation coefficient is less than the first threshold. If either condition is met, the live and neutral wires are abnormal, and the meter with the initial abnormality is locked. The power consumption data of the target meter and the line loss data of the corresponding transformer area are subjected to third-order Bessel filtering, that is, the filtered data sequence is obtained by substituting into the difference equation; the Pearson correlation coefficient between the filtered power consumption data and the line loss data is calculated. If the Pearson correlation coefficient is greater than the first threshold, it is determined that the line loss is abnormal and the meter with the initial abnormality is identified.

5. The method for real-time monitoring and location of abnormal power consumption based on LMU as described in claim 4, characterized in that: The expression for the difference equation is: ； in, For a sequence of length L, the first... The result after filtering the point sequence For a sequence of length L, the first... Input data sequence at points, and Determine the coefficients of the filter characteristics respectively. ; 、 、 、 The first The point, the first The point, the first The point, the first One point.

6. The method for real-time monitoring and location of abnormal power consumption based on LMU as described in claim 4, characterized in that: The process of adjusting the data collection strategy for initially abnormal electricity meters, recalculating various indicators, classifying confidence levels based on changes in results, and determining whether anomaly localization has been completed includes... For meters that have been initially identified as abnormal, voltage, current and power data are collected again by increasing the corresponding collection frequency and optimizing the collection sequence. The mean value of the ratio of neutral and live wire current, the correlation between neutral and live wire current, and the correlation index between electricity consumption and line loss were recalculated for the re-collected data using the same anomaly judgment method as the initial judgment. The recalculated indicators are compared with the initial judgment results. If the average value of the ratio of live and neutral current is within the set range and the correlation index of live and neutral current is greater than or equal to the first threshold, or the correlation index of power consumption and line loss is greater than or equal to the first threshold, it is marked as a high-confidence anomaly, and the anomaly location result is directly output. If the mean of the ratio of live and neutral currents is not within the set range and the correlation index of live and neutral currents decreases and is less than or equal to the first threshold, or the correlation index of power consumption and line loss decreases and is less than the first threshold, then it is marked as a low confidence anomaly and no anomaly location result is output. If the mean of the neutral-live wire current ratio is not within the set range and the neutral-live wire current correlation index is less than the first threshold and the power consumption and line loss correlation index is greater than or equal to the first threshold, or if the mean of the neutral-live wire current ratio is not within the set range and the neutral-live wire current correlation index is greater than or equal to the first threshold and the power consumption and line loss correlation index is less than the first threshold, or if the mean of the neutral-live wire current ratio is not within the set range and the neutral-live wire current correlation index increases and is less than or equal to the first threshold, or if the power consumption and line loss correlation index increases and is less than the first threshold, then it is marked as a medium confidence anomaly. No anomaly location result is output. A new round of voltage, current and power consumption data is collected repeatedly. The mean of the neutral-live wire current ratio, the neutral-live wire current correlation, and the power consumption and line loss correlation index are calculated. The judgment is iteratively judged until the high confidence anomaly or low confidence anomaly is met.

7. The method for real-time monitoring and location of abnormal power consumption based on LMU as described in claim 4, characterized in that: When the anomaly cannot be located, cluster analysis is performed on the meter data within the transformer area to determine the installation location of the line monitoring unit, including... When a new round of voltage, current and power data is collected and the various indicators are recalculated, but the precise location of the abnormal meter has not yet been output, it is necessary to perform convolution smoothing on the voltage data of all meters in the transformer area and calculate the Pearson correlation coefficient between the meters. Based on the correlation results, the DBSCAN clustering algorithm is used to complete the meter clustering. According to the clustering results, a line monitoring unit is installed at the line bifurcation point corresponding to each meter cluster. The expression for convolution smoothing is: ； in, To smooth the convolution result, the convolution kernel is defined as follows: , The kernel length is 1. voltage sequence Using convolution kernels Convolution operation, For the first Voltage value at each point For convolution smoothing The convolution kernel weights at each point.

8. A real-time monitoring and location system for abnormal power consumption based on an LMU, employing the real-time monitoring and location method for abnormal power consumption based on an LMU as described in any one of claims 1 to 7, characterized in that, include: The system includes a feature extraction module, an anomaly detection module, an analysis module, an anomaly localization module, and an output module. The feature extraction module is based on the historical voltage data of each meter in the transformer area, extracts features to calculate the correlation between meters, generates an anomaly probability ranking, and sets differentiated collection frequencies. The anomaly identification module collects voltage, current, and power data from the target meter based on the collection frequency, forming a data set for anomaly identification. The analysis module performs zero-live-wire ratio calculation and correlation analysis on the data set, as well as correlation analysis between electricity consumption and line loss data, to identify preliminary abnormal meters; The anomaly location module adjusts the data collection strategy for the initial abnormal electricity meter and recalculates various indicators, classifies the confidence level based on the changes in the results, and determines whether the anomaly location has been completed. The output module performs cluster analysis on the meter data within the transformer area to determine the installation location of the line monitoring unit when the anomaly cannot be located.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the LMU-based real-time monitoring and location method for abnormal power consumption as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the LMU-based real-time monitoring and location method for abnormal power consumption as described in any one of claims 1 to 7.

Images