Line loss segmented monitoring method and system based on power distribution automation

Abstract

The invention relates to the field of line loss monitoring, in particular to a line loss segmented monitoring method and system based on power distribution automation, and the method comprises the steps: constructing a calibration library containing multi-modal features and topological equivalent parameters through a calibration experiment; determining a key monitoring sensitive area and generating a broadband excitation signal adaptive to the sensitive area; an excitation signal is directionally injected into the head end of the line, effective signal extraction of original multi-modal signal data is completed through an introduced separation mechanism by a modal parameter identification algorithm, and a pure feature matrix is obtained; inverting topological equivalent parameters after directional injection of the wide-screen excitation signals through the pure characteristic matrix, and rehearsing topological parameter changes caused by line loss abnormity after directional injection of the excitation signals. Through the calibration library and real-time topology equivalent parameter extraction, impedance abrupt change point positioning and line physical structure marking sensitive areas are combined, monitoring blind areas caused by segmented solidification are avoided, redundant monitoring of impedance stable sections can be reduced, and monitoring efficiency and accuracy are improved.

Description

A method and system for segmented monitoring of line losses based on distribution automation Technical Field

[0001] This invention relates to the field of line loss monitoring, and specifically to a method and system for segmented line loss monitoring based on distribution automation. Background Technology

[0002] Monitoring line losses in power distribution lines is a core component of ensuring the energy efficiency and safe operation of the power grid. Its core requirements are to accurately locate abnormal line loss intervals, identify loss types, and provide timely warnings.

[0003] However, existing line loss monitoring technologies have significant limitations: the segmentation method is fixed, relying heavily on pre-defined fixed topology nodes to divide monitoring intervals, which cannot be dynamically adjusted according to real-time impedance distribution and load changes, easily leading to monitoring blind spots or redundancy; the signal anti-interference capability is weak, with multiple interference sources such as electromagnetic interference, environmental noise, and equipment interference present on-site, and the collected multi-mode signals are often mixed with interference signals, lacking an effective signal separation mechanism, making it difficult to extract effective signals, which in turn affects the accuracy of topology equivalent parameter calculation; and the adaptability to different scenarios and the ability to provide early warning of anomalies are insufficient.

[0004] Existing technologies are mostly designed for single or few loss scenarios, lacking calibration benchmarks that cover multiple scenarios such as normal line loss, leakage current, and insulation aging. Furthermore, the identification of anomaly types largely relies on measured data after a fault occurs, making it impossible to predict potential line loss anomalies through real-time parameter changes. This results in delayed fault response and increases the risk to power grid operation. Summary of the Invention

[0005] This invention addresses the technical problems existing in the prior art by providing a method and system for segmented monitoring of line losses based on power distribution automation.

[0006] The technical solution of this invention to solve the above-mentioned technical problems is as follows: A method for segmented monitoring of line losses based on distribution automation, comprising the following steps: The method includes:

[0007] S101. Conduct full-scenario surveys based on the target power distribution line, and build a calibration library containing multimodal features and topological equivalent parameters through calibration experiments;

[0008] S102. By calling the constructed calibration library through the modal parameter identification algorithm, the initial topology equivalent parameters of the line are extracted, the impedance distribution characteristics of the line in each section are obtained, and then the impedance change point is located by the fusion formation mechanism embedded in the modal parameter identification algorithm, the sensitive area to be monitored is determined, and a broadband excitation signal adapted to the sensitive area is generated.

[0009] S103. The excitation signal is injected into the beginning of the line. The miniature sensing nodes distributed in each section of the line collect the original multimodal signal data generated after the excitation signal is injected. The modal parameter identification algorithm completes the effective signal extraction of the original multimodal signal data with the introduced separation mechanism to obtain a pure feature matrix.

[0010] S104 The modal parameter identification algorithm inverts the topological equivalent parameters after the directional injection of the widescreen excitation signal through the pure feature matrix, and pre-simulates the changes in topological parameters caused by the abnormal line loss after the directional injection of the excitation signal, and completes dynamic segmentation based on the results of the changes in topological parameters.

[0011] In a preferred embodiment, step S101, the full-scene survey based on the target power distribution line, includes:

[0012] Conduct a full-scene survey of the target power distribution line to determine its inherent topology.

[0013] Miniature sensing nodes are deployed at equal intervals, and all miniature sensing nodes are connected to each other to form a distributed sensing network.

[0014] Then, calibration experiments are conducted based on the target power distribution line, including:

[0015] Weak broadband excitation signals corresponding to loss scenarios such as normal line loss, leakage loss, and insulation aging are injected into random lines. Under each loss scenario, a distributed sensing network formed by the communication connection of all micro sensing nodes collects multi-mode signal features including reflection coefficient, phase shift, energy attenuation rate, line surface micro-temperature, and electric field distortion, as well as measured topological equivalent parameters including node equivalent impedance, line distributed capacitance, and load coupling coefficient. The collected multi-mode signal features and measured topological equivalent parameters are used to construct a calibration library, which serves as the training dataset for the modal parameter identification algorithm. Since this line segment is located on the target distribution line, the data collected from this line segment is still representative.

[0016] It also includes the acquisition of interference signals in pure interference scenarios, including electromagnetic interference, environmental noise, and equipment interference, and the establishment of an interference sample set.

[0017] Furthermore, the normal line loss scenario includes four sub-scenarios: no-load, light load, medium load, and heavy load. In each sub-scenarios, the line is kept free of abnormal losses and stable for 30 minutes before data collection begins. The minor loss scenario includes minor leakage current, which can be simulated by an adjustable resistor to achieve a leakage current of 1A to 3A; minor insulation aging, which can be achieved by replacing the original wire with a 10-meter-long section of aged wire; and poor contact at the joint, which can be achieved by loosening the bolts to increase the contact resistance to 5Ω to 10Ω. There are a total of three sub-scenarios. The severe loss scenario includes severe leakage current, which can be simulated by a leakage current of 5A to 10A; severe insulation aging, which can be achieved by replacing the original wire with a 20-meter-long section of aged wire; and simultaneous loss at multiple nodes, which can be achieved by having poor contact at two to three joints simultaneously. There are a total of three sub-scenarios.

[0018] After constructing each loss scenario, in some other specific implementations, a weak wideband excitation signal with a peak value ≤ 5V can be used, with a frequency range of 1kHz to 1MHz, a frequency jump step size of 10kHz, and a signal duration of 0.1 seconds at each frequency point to ensure that the signal can cover the sensitive frequency range corresponding to different impedances of the line. The injection method is as follows: injection is carried out from the beginning of the line through an electromagnetic induction coupler, and auxiliary injection points are added at three key branch points to achieve multi-directional signal coverage. During the injection process, the changes in line voltage and current are monitored in real time to ensure that the excitation signal does not affect the normal power supply of the line.

[0019] In a preferred embodiment, after constructing the calibration library, the modal parameter identification algorithm transmits data to the distributed sensing network via a communication connection to synchronously obtain the current basic operating status of the line, avoiding the subsequent initial parameter extraction from being disconnected from the actual operating conditions. Then, the constructed calibration library is called, and the current basic operating status is used as the target. Several historical calibration experimental data that are closest to the current operating status are matched by cosine similarity, that is, the similarity values ​​are arranged in descending order. After weighting the similarity values ​​by weighting the weighted average method, the initial topology equivalent parameters of the line, including the initial node equivalent impedance, initial line distributed capacitance, and initial load coupling coefficient of each interval, are extracted to obtain an interval-parameter two-dimensional matrix representing the impedance distribution characteristics of each line interval. The rows correspond to the line intervals, and the columns correspond to the initial topology equivalent parameters of the line.

[0020] The fusion formation mechanism embedded in the modal parameter identification calculation uses the current basic operating status of the line synchronously acquired through a distributed sensing network to identify sensitive areas of abnormal line loss. It then uses electrical signal focusing control logic to achieve directional coupling of excitation signal energy, generating a broadband excitation signal. Specifically, this includes:

[0021] Using a two-dimensional interval-parameter matrix as input, the line interval is subdivided into multiple sub-units according to a fixed distance, such as subdividing the line interval by a fixed distance of 5 meters. The fixed distance is determined according to the overall length of the interval. Therefore, in actual power distribution lines, because there are too many line intervals and the fixed distance of each interval is different, the average value of the initial node equivalent impedance of each sub-unit is calculated according to the definition of the actual scenario. The impedance difference between adjacent sub-units is analyzed, and the difference exceeding the threshold is judged as the abrupt change point to complete the location of the impedance abrupt change point. Sensitive areas are marked in combination with the physical structure of the power distribution line obtained from the full-scene survey.

[0022] In a preferred embodiment, the interval is subdivided into multiple subdivision units according to a fixed interval, including:

[0023] Using the initial node equivalent impedance column in the interval-parameter two-dimensional matrix, associate the physical coordinates of the line interval corresponding to each row in the interval-parameter two-dimensional matrix, and use the beginning of the line interval as the origin to uniformly mark the start and end coordinates of the interval in a clockwise direction, with no overlap among all line intervals;

[0024] The line interval is subdivided by the initial node equivalent impedance of each line interval, which is distributed to all subdivided units within the interval according to the length of the subdivided unit. After the calculation is completed, the corresponding mean value is matched for each subdivided unit.

[0025] The calculation of the mean of its subdivided units is as follows: the product of the initial node equivalent impedance of the line section and the fixed distance length is divided by the total length of the line section.

[0026] After the equivalent impedance mean values ​​of the line section and its initial nodes are allocated, starting from the first sub-unit, the mean values ​​are extracted in pairs with its adjacent sub-units to form several pairs of adjacent unit mean values. Then, the relative difference method is used to calculate the impedance relative difference between adjacent units. The line impedance fluctuation range of the measured topology equivalent parameters in the calibration library is used as the threshold for sudden change judgment. If the impedance relative difference between adjacent units is greater than the set threshold for sudden change judgment, that is, the boundary coordinates of the two sub-units in the adjacent unit are the location of the impedance sudden change point. The trend of the impedance relative difference between adjacent units is used as the distribution of the impedance distribution curve.

[0027] During routine monitoring, the calibration library obtained before monitoring serves as a historical experience database. The impedance distribution curves of highly sensitive areas are subdivided and analyzed, with each subdivision unit consisting of several meters. The initial nodal equivalent impedance value of each subdivision unit is calculated. Then, through impedance difference analysis between adjacent subdivision units, three types of areas are classified and corresponding transition step sizes are matched. Specifically, this includes:

[0028] Impedance uniformity zone: If the impedance difference between adjacent subdivision units of several meters is ≤5%, it is determined to be an impedance stable zone. The frequency jump step size is set to 50kHz to improve the overall monitoring efficiency while ensuring monitoring coverage and avoiding unnecessary waste of accuracy.

[0029] Impedance change zone: The impedance difference between adjacent sub-units of several meters is between 5% and 15%, which is determined to be a region of slow impedance change, such as the edge section of a densely loaded area. The jump step size is set to 30kHz to balance monitoring accuracy and efficiency and ensure that the changes in line loss caused by the gradual change in impedance can be captured.

[0030] Impedance abrupt change zone: If the impedance difference between adjacent sub-units is ≥15%, it is determined to be an impedance abrupt change zone, such as line joints, branch access points, and insulation aging critical points. The jump step size is reduced to 10kHz, and three sub-units before and after the abrupt change point are set as fine monitoring sub-regions. The signal excitation of the abrupt change point is enhanced by dense frequency coverage to ensure that the multi-mode signal can clearly reflect the line loss characteristics of the area.

[0031] Harmonic signal data at the beginning of the line section are collected. The specific frequencies corresponding to the third, fifth and seventh harmonics with the highest current proportion are identified by Fourier transform. Among them, odd harmonics dominate in the power distribution system, and the third, fifth and seventh harmonics are the main sources of interference. The power load of the power distribution line is mainly nonlinear load. The harmonics generated by this type of load are mainly odd harmonics. The amplitude of even harmonics is usually less than 1% of the fundamental frequency, and the interference to the excitation signal can be ignored. The average frequency is taken as the adapted broadband excitation signal.

[0032] In a preferred embodiment, after acquiring the appropriate broadband excitation signal, it is directionally injected into the beginning of the line through an electromagnetic induction coupler. The micro-sensing nodes distributed along the line collect the original multimodal signal data generated after the excitation signal injection, including the original reflection coefficient, original phase shift, original energy attenuation rate, original line surface micro-temperature, and original electric field distortion. The five data mentioned above also come from the collection of the micro-sensing nodes. The difference is that after the broadband excitation signal is introduced, these five data change accordingly. Therefore, the original data is introduced as a distinction, and an initial five-dimensional original signal matrix is ​​formed. The separation mechanism takes the initial five-dimensional original signal matrix as input, establishes a hybrid matrix containing the original signal and the interference signal based on the feature vector of the interference signal, that is, splits the original signal into the effective signal and the interference signal, and establishes the phase stability objective function with the target condition that the length of the continuous phase fluctuation interval is greater than or equal to ten sampling periods. The matrix parameters are iteratively adjusted by the gradient descent method, and the hybrid matrix is ​​separated to obtain a separation matrix that maximizes the component weights that satisfy the phase stability characteristics.

[0033] Multiplying the original signal matrix by the separation matrix yields a line loss correlated effective signal submatrix containing signal components that satisfy phase stability characteristics and a mixed interference signal submatrix containing interference components that are phase unstable.

[0034] In a preferred embodiment, the matrix parameters are iteratively adjusted using the gradient descent method to separate the mixing matrix and obtain a separation matrix that maximizes the component weights while satisfying the phase stability characteristic, including:

[0035] Based on the preliminary inverse matrix calculation results of the hybrid matrix, the objective function value is the quotient of the length of the continuous stable phase interval of the effective signal and the total sampling interval length. The sampling period threshold is determined based on the historical phase characteristics of the effective signal in the calibration library. In each iteration, the original signal matrix is ​​decoupled based on the current hybrid matrix, and the objective function value containing the original signal and interference signal matrix is ​​calculated. The iteration is repeated until the objective function value exceeds the objective function threshold, at which point the iteration stops and the separation matrix is ​​output.

[0036] After obtaining the effective signal sub-matrix associated with line loss, feature extraction is performed, including:

[0037] Signals with amplitudes less than 0.1% of the injected signal are discarded;

[0038] Filter signals whose frequency exceeds the frequency band of the excitation signal;

[0039] Smooth the remaining signal;

[0040] The final output is a pure feature matrix containing five-dimensional features: pure reflection coefficient, pure phase shift, pure energy attenuation rate, pure line surface micro-temperature, and pure electric field distortion.

[0041] In a preferred embodiment, after obtaining the pure feature matrix, a standardization process is performed. The modal parameter recognition algorithm extracts the overall feature distribution in the pure feature matrix. The overall feature distribution includes the mean, variance, and peak value of the original multimodal features. It is then matched with the multimodal feature distribution under each loss scenario in the calibration library to determine the basic scenario type of the current line section.

[0042] For the five-dimensional features of the pure feature matrix, the cosine similarity between each feature and the corresponding feature in the calibration library is calculated. The average of all similarities is then taken as the overall similarity between the calibration library and the five-dimensional features of the current line interval, which is used as the multi-dimensional cosine similarity. Data with similarity exceeding the similarity threshold are selected as a similarity subset.

[0043] The measured values ​​of the topological equivalent parameters of each group of data within the similarity subset are extracted, weighted, and sorted. The average of the median of the measured values ​​of the topological equivalent parameters of each group of data is used as the topological equivalent parameter matrix corresponding to the pure feature matrix.

[0044] The inversion units are divided according to the inherent topology nodes and the deployment locations of micro-sensing nodes. Specifically, the entire line is divided into several inversion units based on the main line segmentation points, branch line access points, and load access points as the basic boundaries, combined with the deployment spacing of micro-sensing nodes. Finally, the three-dimensional topological equivalent parameter prediction matrix of each line section is output, including the predicted node equivalent impedance, predicted line distributed capacitance, and predicted load coupling coefficient of each line section.

[0045] In a preferred embodiment, the segmented deviation rate, which includes the deviation rate of the equivalent impedance of the nodes, the deviation rate of the distributed capacitance of the lines, and the deviation rate of the load coupling coefficient, is calculated using the predicted equivalent impedance of the nodes, the predicted distributed capacitance of the lines, and the predicted load coupling coefficient of each line section. The sliding window method is then used to divide the changing trend according to the segmented deviation rate, including the steady trend, the rising trend, and the sudden change trend, to complete the dynamic segmentation of each line section.

[0046] There are several line sections where the equivalent impedance deviation rate and load coupling coefficient deviation rate at the nodes show an upward trend simultaneously, which is a prediction of increased line loss due to overload in several line sections.

[0047] In a preferred embodiment, it further includes:

[0048] There are several line sections where the node equivalent impedance deviation rate and load coupling coefficient deviation rate exceed the preset threshold and show an upward trend, which is a prediction of increased line loss caused by overload in several line sections.

[0049] There are several line sections where the line distributed capacitance deviation rate exceeds the preset threshold, and the trend is stable and / or increasing, which is a prediction of leakage loss caused by the aging of the line insulation layer.

[0050] There are several line sections where the equivalent impedance deviation rate of the nodes exceeds the preset threshold and shows an abrupt change trend, which is a prediction of loose line joints and / or instantaneous leakage.

[0051] The node equivalent impedance deviation rate is defined as the quotient of the absolute difference between the predicted node equivalent impedance and the initial node equivalent impedance and the initial node equivalent impedance. The line distributed capacitance deviation rate is defined as the quotient of the absolute difference between the predicted line distributed capacitance and the initial line distributed capacitance and the initial line distributed capacitance. The load coupling coefficient deviation rate is defined as the quotient of the absolute difference between the predicted load coupling coefficient and the initial load coupling coefficient and the initial load coupling coefficient. During the calibration of the preset threshold, weak broadband excitation signals corresponding to loss scenarios such as normal line loss, leakage loss, and insulation aging are injected into random lines. Under each loss scenario, the lower limit of the interval of the deviation rate distribution data of the topology equivalent parameters under each scenario is extracted as the critical threshold, i.e., the preset threshold.

[0052] The present invention also provides a line loss segmentation monitoring system based on distribution automation, the system comprising:

[0053] Calibration library construction module: Collect multimodal signal characteristics, measured topological equivalent parameters, and interference signals under different loss scenarios through calibration experiments to construct a calibration library and interference sample set;

[0054] The initial topology parameter extraction and excitation signal generation module is used to extract the initial topology equivalent parameters by calling the calibration library and combining the current basic operating status of the line with the fusion, locate impedance change points and mark sensitive areas through impedance difference analysis, and generate broadband excitation signals.

[0055] Multimodal signal acquisition and purification module: It acquires raw multimodal signal data through miniature sensing nodes, separates effective signals from interference signals using a separation mechanism, and outputs a pure feature matrix through a feature purification process;

[0056] The topology parameter inversion and dynamic segmentation module is used to standardize and match the pure feature matrix through modal parameter recognition algorithms, invert the topology equivalent parameters, calculate the segmentation deviation rate and divide the trend of change, and complete the dynamic segmentation of the line and the pre-simulation of line loss anomaly types.

[0057] The beneficial effects of this invention are: by extracting equivalent parameters from the calibration library and real-time topology, combined with the location of impedance mutation points and the marking of sensitive areas by the physical structure of the line, the monitoring segment no longer depends on fixed nodes. It can be dynamically adjusted according to the line impedance distribution, load changes and other operating conditions, avoiding monitoring blind spots caused by segment fixation, and reducing redundant monitoring of impedance stable sections, thereby improving monitoring efficiency and accuracy.

[0058] By pre-constructing a set of interference samples for pure interference scenarios and combining it with a signal separation mechanism that incorporates a modal parameter recognition algorithm, the effective components and interference components in the acquired signal can be effectively separated. Then, through a purification process of removing invalid signals, filtering noise frequencies, and smoothing, the impact of noise on the data is greatly reduced, providing high-purity feature data for topological equivalent parameter inversion and significantly improving the accuracy of parameter calculation. Attached Figure Description

[0059] Figure 1 is a flowchart of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0061] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0062] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0063] This embodiment provides a method for segmented monitoring of line losses based on distribution automation, comprising the following steps:

[0064] S101. Conduct full-scenario surveys based on the target power distribution line, and build a calibration library containing multimodal features and topological equivalent parameters through calibration experiments;

[0065] Conduct a full-scene survey of the target power distribution line to determine its inherent topology.

[0066] Miniature sensing nodes are deployed at equal intervals, and all miniature sensing nodes are connected to each other to form a distributed sensing network.

[0067] Then, calibration experiments are conducted based on the target power distribution line, including:

[0068] Weak broadband excitation signals corresponding to loss scenarios such as normal line loss, leakage loss, and insulation aging are injected into random lines. Under each loss scenario, a distributed sensing network formed by the communication connection of all micro sensing nodes collects multi-mode signal features including reflection coefficient, phase shift, energy attenuation rate, line surface micro-temperature, and electric field distortion, as well as measured topological equivalent parameters including node equivalent impedance, line distributed capacitance, and load coupling coefficient. The collected multi-mode signal features and measured topological equivalent parameters are used to construct a calibration library, which serves as the training dataset for the modal parameter identification algorithm. Since this line segment is located on the target distribution line, the data collected from this line segment is still representative.

[0069] It also includes the acquisition of interference signals in pure interference scenarios, including electromagnetic interference, environmental noise, and equipment interference, and the establishment of an interference sample set.

[0070] Furthermore, the normal line loss scenario includes four sub-scenarios: no-load, light load, medium load, and heavy load. In each sub-scenarios, the line is kept free of abnormal losses and stable for 30 minutes before data collection begins. The minor loss scenario includes minor leakage current, which can be simulated by an adjustable resistor to achieve a leakage current of 1A to 3A; minor insulation aging, which can be achieved by replacing the original wire with a 10-meter-long section of aged wire; and poor contact at the joint, which can be achieved by loosening the bolts to increase the contact resistance to 5Ω to 10Ω. There are a total of three sub-scenarios. The severe loss scenario includes severe leakage current, which can be simulated by a leakage current of 5A to 10A; severe insulation aging, which can be achieved by replacing the original wire with a 20-meter-long section of aged wire; and simultaneous loss at multiple nodes, which can be achieved by having poor contact at two to three joints simultaneously. There are a total of three sub-scenarios.

[0071] After constructing each loss scenario, in some other specific implementations, a weak wideband excitation signal with a peak value ≤ 5V can be used, with a frequency range of 1kHz to 1MHz, a frequency jump step size of 10kHz, and a signal duration of 0.1 seconds at each frequency point to ensure that the signal can cover the sensitive frequency range corresponding to different impedances of the line. The injection method is as follows: injection is carried out from the beginning of the line through an electromagnetic induction coupler, and auxiliary injection points are added at three key branch points to achieve multi-directional signal coverage. During the injection process, the changes in line voltage and current are monitored in real time to ensure that the excitation signal does not affect the normal power supply of the line.

[0072] S102. By calling the constructed calibration library through the modal parameter identification algorithm, the initial topology equivalent parameters of the line are extracted, the impedance distribution characteristics of the line in each section are obtained, and then the impedance change point is located by the fusion formation mechanism embedded in the modal parameter identification algorithm, the sensitive area to be monitored is determined, and a broadband excitation signal adapted to the sensitive area is generated.

[0073] After constructing the calibration library, the modal parameter recognition algorithm transmits data through a communication connection and a distributed sensing network to synchronously obtain the current basic operating status of the line, avoiding the subsequent initial parameter extraction from being out of sync with the actual operating conditions. Then, it calls the constructed calibration library, takes the current basic operating status as the target, and uses cosine similarity matching to find several historical calibration experimental data that are closest to the current operating status. That is, the similarity values ​​are arranged in descending order. After weighting the similarity values ​​by weighting the weighted average method, the initial topological equivalent parameters of the line, including the initial node equivalent impedance, initial line distributed capacitance, and initial load coupling coefficient of each interval, are extracted to obtain an interval-parameter two-dimensional matrix representing the impedance distribution characteristics of each line interval. The rows correspond to the line intervals, and the columns correspond to the initial topological equivalent parameters of the line.

[0074] The fusion formation mechanism embedded in the modal parameter identification calculation uses the current basic operating status of the line synchronously acquired through a distributed sensing network to identify sensitive areas of abnormal line loss. It then uses electrical signal focusing control logic to achieve directional coupling of excitation signal energy, generating a broadband excitation signal. Specifically, this includes:

[0075] Using a two-dimensional interval-parameter matrix as input, the line interval is subdivided into multiple sub-units according to a fixed distance, such as subdividing the line interval by a fixed distance of 5 meters. The fixed distance is determined according to the overall length of the interval. Therefore, in actual power distribution lines, because there are too many line intervals and the fixed distance of each interval is different, the average value of the initial node equivalent impedance of each sub-unit is calculated according to the definition of the actual scenario. The impedance difference between adjacent sub-units is analyzed, and the difference exceeding the threshold is judged as the abrupt change point to complete the location of the impedance abrupt change point. Sensitive areas are marked in combination with the physical structure of the power distribution line obtained from the full-scene survey.

[0076] Using the initial node equivalent impedance column in the interval-parameter two-dimensional matrix, associate the physical coordinates of the line interval corresponding to each row in the interval-parameter two-dimensional matrix, and use the beginning of the line interval as the origin to uniformly mark the start and end coordinates of the interval in a clockwise direction, with no overlap among all line intervals;

[0077] The line interval is subdivided by the initial node equivalent impedance of each line interval, which is distributed to all subdivided units within the interval according to the length of the subdivided unit. After the calculation is completed, the corresponding mean value is matched for each subdivided unit.

[0078] The calculation of the mean of its subdivided units is as follows: the product of the initial node equivalent impedance of the line section and the fixed distance length is divided by the total length of the line section.

[0079] After the equivalent impedance mean values ​​of the line section and its initial nodes are allocated, starting from the first sub-unit, the mean values ​​are extracted in pairs with its adjacent sub-units to form several pairs of adjacent unit mean values. Then, the relative difference method is used to calculate the impedance relative difference between adjacent units. The line impedance fluctuation range of the measured topology equivalent parameters in the calibration library is used as the threshold for sudden change judgment. If the impedance relative difference between adjacent units is greater than the set threshold for sudden change judgment, that is, the boundary coordinates of the two sub-units in the adjacent unit are the location of the impedance sudden change point. The trend of the impedance relative difference between adjacent units is used as the distribution of the impedance distribution curve.

[0080] During routine monitoring, the calibration library obtained before monitoring serves as a historical experience database. The impedance distribution curves of highly sensitive areas are subdivided and analyzed, with each subdivision unit consisting of several meters. The initial nodal equivalent impedance value of each subdivision unit is calculated. Then, through impedance difference analysis between adjacent subdivision units, three types of areas are classified and corresponding transition step sizes are matched. Specifically, this includes:

[0081] Impedance uniformity zone: If the impedance difference between adjacent subdivision units of several meters is ≤5%, it is determined to be an impedance stable zone. The frequency jump step size is set to 50kHz to improve the overall monitoring efficiency while ensuring monitoring coverage and avoiding unnecessary waste of accuracy.

[0082] Impedance change zone: The impedance difference between adjacent sub-units of several meters is between 5% and 15%, which is determined to be a region of slow impedance change, such as the edge section of a densely loaded area. The jump step size is set to 30kHz to balance monitoring accuracy and efficiency and ensure that the changes in line loss caused by the gradual change in impedance can be captured.

[0083] Impedance abrupt change zone: If the impedance difference between adjacent sub-units is ≥15%, it is determined to be an impedance abrupt change zone, such as line joints, branch access points, and insulation aging critical points. The jump step size is reduced to 10kHz, and three sub-units before and after the abrupt change point are set as fine monitoring sub-regions. The signal excitation of the abrupt change point is enhanced by dense frequency coverage to ensure that the multi-mode signal can clearly reflect the line loss characteristics of the area.

[0084] Harmonic signal data at the beginning of the line section are collected. The specific frequencies corresponding to the third, fifth and seventh harmonics with the highest current proportion are identified by Fourier transform. Among them, odd harmonics dominate in the power distribution system, and the third, fifth and seventh harmonics are the main sources of interference. The power load of the power distribution line is mainly nonlinear load. The harmonics generated by this type of load are mainly odd harmonics. The amplitude of even harmonics is usually less than 1% of the fundamental frequency, and the interference to the excitation signal can be ignored. The average frequency is taken as the adapted broadband excitation signal.

[0085] S103. The excitation signal is injected into the beginning of the line. The miniature sensing nodes distributed in each section of the line collect the original multimodal signal data generated after the excitation signal is injected. The modal parameter identification algorithm completes the effective signal extraction of the original multimodal signal data with the introduced separation mechanism to obtain a pure feature matrix.

[0086] After acquiring the appropriate broadband excitation signal, it is directionally injected into the beginning of the line through an electromagnetic induction coupler. The micro-sensing nodes distributed along the line collect the original multi-mode signal data generated after the excitation signal injection, including the original reflection coefficient, original phase shift, original energy attenuation rate, original line surface micro-temperature, and original electric field distortion. The five data mentioned above also come from the collection of the micro-sensing nodes. The difference is that these five data change accordingly after the broadband excitation signal is introduced. Therefore, the original data is introduced as a distinction, and an initial five-dimensional original signal matrix is ​​formed. The separation mechanism takes the initial five-dimensional original signal matrix as input and establishes a hybrid matrix containing the original signal and the interference signal based on the feature vector of the interference signal. That is, the original signal is split into the effective signal and the interference signal. The phase stability objective function is established with the target condition that the length of the continuous phase fluctuation interval is greater than or equal to ten sampling periods. The matrix parameters are iteratively adjusted by the gradient descent method, and the hybrid matrix is ​​separated to obtain a separation matrix that maximizes the component weights that satisfy the phase stability characteristics.

[0087] Multiplying the original signal matrix by the separation matrix yields a line loss correlated effective signal submatrix containing signal components that satisfy phase stability characteristics and a mixed interference signal submatrix containing interference components that are phase unstable.

[0088] Furthermore, the separation mechanism uses the initial five-dimensional original signal matrix as the core input to decompose the effective signal and the interference signal. Based on the feature vectors of common interference signals in power distribution lines, a hybrid matrix containing the original signal and the interference signal is constructed. This matrix is ​​a linear combination of the effective component and the interference component in the original signal. The algorithm establishes a phase stability objective function with the phase continuous fluctuation interval length being greater than or equal to ten sampling periods as the objective condition. The effective signal associated with line loss is directly related to the changes in the equivalent parameters of the line topology, and its phase characteristics show continuous and regular fluctuations. In contrast, the phase fluctuations of the interference signal are irregular and have a shorter interval. The objective function provides a basis for signal separation based on these differences.

[0089] By iteratively adjusting the matrix parameters using gradient descent, a separation matrix is ​​obtained that maximizes the component weights while satisfying the phase stability characteristic. This includes:

[0090] Based on the preliminary inverse matrix calculation results of the hybrid matrix, the objective function value is the quotient of the length of the continuous stable phase interval of the effective signal and the total sampling interval length. The sampling period threshold is determined based on the historical phase characteristics of the effective signal in the calibration library. In each iteration, the original signal matrix is ​​decoupled based on the current hybrid matrix, and the objective function value containing the original signal and interference signal matrix is ​​calculated. The iteration is repeated until the objective function value exceeds the objective function threshold, at which point the iteration stops and the separation matrix is ​​output.

[0091] After obtaining the effective signal sub-matrix associated with line loss, feature extraction is performed, including:

[0092] Signals with amplitudes less than 0.1% of the injected signal are discarded;

[0093] Filter signals whose frequency exceeds the frequency band of the excitation signal;

[0094] Smooth the remaining signal;

[0095] The final output is a pure feature matrix containing five-dimensional features: pure reflection coefficient, pure phase shift, pure energy attenuation rate, pure line surface micro-temperature, and pure electric field distortion.

[0096] In some other specific implementations, the separation matrix is ​​obtained starting from the initial inverse matrix of the mixing matrix. The parameters are optimized by relying on the gradient descent method. During the iteration process, the algorithm decouples the initial five-dimensional original signal matrix based on the current mixing matrix, splits the mixed signal into potential effective signal components and interference signal components, and then calculates the objective function value corresponding to the decoupled signal.

[0097] The iterative process is repeated, and the parameters of the mixing matrix are adjusted according to the change of the objective function value in each iteration until the objective function value exceeds the preset threshold, indicating that the phase stability of the effective signal after decoupling has met the requirements. At this time, the iteration stops and the final separation matrix is ​​output.

[0098] S104 The modal parameter identification algorithm inverts the topological equivalent parameters after the directional injection of the widescreen excitation signal through the pure feature matrix, and pre-simulates the changes in topological parameters caused by the abnormal line loss after the directional injection of the excitation signal, and completes dynamic segmentation based on the results of the changes in topological parameters.

[0099] After obtaining the pure feature matrix, it is standardized. The modal parameter recognition algorithm extracts the overall feature distribution in the pure feature matrix. The overall feature distribution includes the mean, variance, and peak value of the original multimodal features. It is then matched with the multimodal feature distribution under each loss scenario in the calibration library to determine the basic scenario type of the current line section.

[0100] For the five-dimensional features of the pure feature matrix, the cosine similarity between each feature and the corresponding feature in the calibration library is calculated. The average of all similarities is then taken as the overall similarity between the calibration library and the five-dimensional features of the current line interval, which is used as the multi-dimensional cosine similarity. Data with similarity exceeding the similarity threshold are selected as a similarity subset.

[0101] The measured values ​​of the topological equivalent parameters of each group of data within the similarity subset are extracted, weighted, and sorted. The average of the median of the measured values ​​of the topological equivalent parameters of each group of data is used as the topological equivalent parameter matrix corresponding to the pure feature matrix.

[0102] The inversion units are divided according to the inherent topology nodes and the deployment locations of micro-sensing nodes. Specifically, the entire line is divided into several inversion units based on the main line segmentation points, branch line access points, and load access points as the basic boundaries, combined with the deployment spacing of micro-sensing nodes. Finally, the three-dimensional topological equivalent parameter prediction matrix of each line section is output, including the predicted node equivalent impedance, predicted line distributed capacitance, and predicted load coupling coefficient of each line section.

[0103] After clarifying the basic scenario types, for the five-dimensional features of the pure feature matrix, cosine similarity is calculated with the five-dimensional features of the corresponding scenario in the calibration library to quantify the matching degree of single-dimensional features. The average value of the cosine similarity of the five-dimensional features is taken as the multi-dimensional cosine similarity between the current line interval and the calibration library data to avoid misjudgment caused by single-dimensional feature bias. Then, based on the calibration library data, the similarity threshold is determined and the calibration library data with multi-dimensional cosine similarity exceeding the threshold are selected to form a similarity subset. For this subset, the measured values ​​of the topological equivalent parameters of each group of data are extracted and sorted by similarity level. This is used as the topological equivalent parameter matrix corresponding to the current pure feature matrix to ensure that the parameter values ​​not only conform to the historical measured patterns but also adapt to the current actual state of the line.

[0104] The segmented deviation rate, which includes the deviation rate of the equivalent impedance of the nodes, the deviation rate of the distributed capacitance of the lines, and the deviation rate of the load coupling coefficient, is calculated based on the predicted equivalent impedance of the nodes, the deviation rate of the distributed capacitance of the lines, and the deviation rate of the load coupling coefficient in each line section. The sliding window method is used to divide the changing trend according to the segmented deviation rate, including the steady trend, the rising trend and the sudden change trend, to complete the dynamic segmentation of each line section.

[0105] There are several line sections where the equivalent impedance deviation rate and load coupling coefficient deviation rate at the nodes show an upward trend simultaneously, which is a prediction of increased line loss due to overload in several line sections.

[0106] Also includes:

[0107] There are several line sections where the node equivalent impedance deviation rate and load coupling coefficient deviation rate exceed the preset threshold and show an upward trend, which is a prediction of increased line loss caused by overload in several line sections.

[0108] There are several line sections where the line distributed capacitance deviation rate exceeds the preset threshold, and the trend is stable and / or increasing, which is a prediction of leakage loss caused by the aging of the line insulation layer.

[0109] There are several line sections where the equivalent impedance deviation rate of the nodes exceeds the preset threshold and shows an abrupt change trend, which is a prediction of loose line joints and / or instantaneous leakage.

[0110] The node equivalent impedance deviation rate is defined as the quotient of the absolute difference between the predicted node equivalent impedance and the initial node equivalent impedance and the initial node equivalent impedance. The line distributed capacitance deviation rate is defined as the quotient of the absolute difference between the predicted line distributed capacitance and the initial line distributed capacitance and the initial line distributed capacitance. The load coupling coefficient deviation rate is defined as the quotient of the absolute difference between the predicted load coupling coefficient and the initial load coupling coefficient and the initial load coupling coefficient. During the calibration of the preset threshold, weak broadband excitation signals corresponding to loss scenarios such as normal line loss, leakage loss, and insulation aging are injected into random lines. Under each loss scenario, the lower limit of the interval of the deviation rate distribution data of the topology equivalent parameters under each scenario is extracted as the critical threshold, i.e., the preset threshold.

[0111] This invention also provides a line loss segmentation monitoring system based on distribution automation, comprising:

[0112] Calibration library construction module: Collect multimodal signal characteristics, measured topological equivalent parameters, and interference signals under different loss scenarios through calibration experiments to construct a calibration library and interference sample set;

[0113] The initial topology parameter extraction and excitation signal generation module is used to extract the initial topology equivalent parameters by calling the calibration library and combining the current basic operating status of the line with the fusion, locate impedance change points and mark sensitive areas through impedance difference analysis, and generate broadband excitation signals.

[0114] Multimodal signal acquisition and purification module: It acquires raw multimodal signal data through miniature sensing nodes, separates effective signals from interference signals using a separation mechanism, and outputs a pure feature matrix through a feature purification process;

[0115] The topology parameter inversion and dynamic segmentation module is used to standardize and match the pure feature matrix through modal parameter recognition algorithms, invert the topology equivalent parameters, calculate the segmentation deviation rate and divide the trend of change, and complete the dynamic segmentation of the line and the pre-simulation of line loss anomaly types.

[0116] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0117] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0118] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0119] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0120] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0121] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0122] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for segmented monitoring of line losses based on distribution automation, characterized in that, The method includes: S101, conducting a full-scene survey based on the target power distribution line, and constructing a calibration library containing multimodal features and topological equivalent parameters through calibration experiments; S102, using a modal parameter identification algorithm to call the constructed calibration library, extracting the initial topological equivalent parameters of the line, obtaining the impedance distribution characteristics of each section of the line, and then using a fusion formation mechanism embedded in the modal parameter identification algorithm to locate impedance mutation points, determine the key monitoring sensitive areas, and generate a broadband excitation signal adapted to the sensitive areas; S103, injecting the excitation signal directionally into the head end of the line, and collecting the original multimodal signal data generated after the excitation signal injection by micro-sensing nodes distributed in each section of the line, and using the introduced separation mechanism to complete the effective signal extraction of the original multimodal signal data to obtain a pure feature matrix; S104, using the pure feature matrix to invert the topological equivalent parameters after the directional injection of the broadband excitation signal, and pre-simulating the topological parameter changes caused by the abnormal line loss after the directional injection of the excitation signal, and completing dynamic segmentation based on the topological parameter change results.

2. The method for segmented monitoring of line losses based on distribution automation according to claim 1, characterized in that, In step S101, the full-scenario survey based on the target power distribution line includes: conducting a full-scenario survey of the target power distribution line to determine its inherent topology; deploying micro-sensing nodes at equal intervals, with all micro-sensing nodes communicating to form a distributed sensing network; and conducting calibration experiments based on the target power distribution line, including: injecting weak broadband excitation signals corresponding to loss scenarios such as normal line loss, leakage loss, and insulation aging into the line; under each loss scenario, collecting multi-mode signal characteristics including reflection coefficient, phase shift, energy attenuation rate, line surface micro-temperature, and electric field distortion degree through the distributed sensing network formed by the communication connection of all micro-sensing nodes, as well as measured topological equivalent parameters including node equivalent impedance, line distributed capacitance, and load coupling coefficient; accumulating the collected multi-mode signal characteristics and measured topological equivalent parameter data to construct a calibration library; and acquiring interference signals under pure interference scenarios, including electromagnetic interference, environmental noise, and equipment interference, and establishing an interference sample set.

3. The method for segmented monitoring of line losses based on distribution automation according to claim 1, characterized in that, After constructing the calibration library, the modal parameter identification algorithm synchronously acquires the current basic operating status of the line through communication connection and distributed sensing network. Then, it calls the constructed calibration library, using the current basic operating status as the target, and matches several historical calibration experimental data points closest to the current operating status using cosine similarity. After weighting the data according to similarity values ​​using a weighted average method, it extracts the initial topology equivalent parameters of the line, including the initial node equivalent impedance, initial line distributed capacitance, and initial load coupling coefficient for each interval. This yields an interval-parameter two-dimensional matrix representing the impedance distribution characteristics of each line interval, with rows corresponding to line intervals and columns corresponding to the initial topology equivalent parameters. The data is embedded in the fusion formation mechanism of modal parameter identification calculation. The current basic operating status of the line is synchronously acquired through the distributed sensing network to find sensitive areas with abnormal line loss. The excitation signal energy is directionally coupled through the electrical signal focusing control logic to generate a broadband excitation signal. Specifically, it includes: taking the interval-parameter two-dimensional matrix as input, subdividing the line interval into multiple sub-units at fixed intervals, calculating the average value of the initial node equivalent impedance of each sub-unit, and analyzing the impedance difference between adjacent sub-units. The difference exceeding the threshold is used as the abrupt change point to complete the location of the impedance abrupt change point. Sensitive areas are marked in combination with the physical structure of the distribution line obtained based on the full-scene survey.

4. The method for segmented monitoring of line losses based on distribution automation according to claim 3, characterized in that, The line interval is subdivided into multiple sub-units based on a fixed distance. This includes: associating the initial node equivalent impedance column in the interval-parameter two-dimensional matrix with the physical coordinates of the line interval corresponding to each row of the interval-parameter two-dimensional matrix, and uniformly labeling the start and end coordinates of the interval in a clockwise direction with the beginning of the line interval as the origin, ensuring that all line intervals do not overlap; the fixed-distance subdivision of the line interval involves distributing the total value of the initial node equivalent impedance of each line interval to all sub-units within its interval according to the proportion of the sub-unit length, and matching the corresponding mean value for each sub-unit after the calculation is completed; after the line interval and its initial node equivalent impedance mean value are distributed, starting from the first sub-unit, it is paired with its adjacent sub-units in sequence. The mean value is extracted to form several pairs of mean values ​​of adjacent units. The relative difference method is then used to calculate the relative impedance difference of adjacent units. The line impedance fluctuation range of the measured topology equivalent parameters in the calibration library is used as the threshold for abrupt change judgment. If the relative impedance difference of adjacent units is greater than the set threshold for abrupt change judgment, the boundary coordinates of the two sub-units in the adjacent unit are the location of the impedance abrupt change point. The trend of the relative impedance difference of adjacent units is used as the distribution of the impedance distribution curve. Harmonic signal data at the beginning of the line section are collected. The specific frequencies corresponding to the third, fifth and seventh harmonics with the highest current proportion are identified by Fourier transform. The average frequency is taken as the adapted broadband excitation signal.

5. The method for segmented monitoring of line losses based on distribution automation according to claim 1, characterized in that, After acquiring the appropriate broadband excitation signal, it is directionally injected into the beginning of the line via an electromagnetic induction coupler. Miniature sensing nodes distributed along the line collect the original multimodal signal data generated after the excitation signal injection, including the original reflection coefficient, original phase shift, original energy attenuation rate, original line surface micro-temperature, and original electric field distortion, forming an initial five-dimensional original signal matrix. The separation mechanism takes the initial five-dimensional original signal matrix as input, establishes a hybrid matrix containing the original signal and the interference signal based on the feature vector of the interference signal, and establishes a phase stability objective function with the target condition that the length of the continuous phase fluctuation interval is greater than or equal to ten sampling periods. The matrix parameters are iteratively adjusted using the gradient descent method to separate the hybrid matrix and obtain a separation matrix that maximizes the weight of the components that satisfy the phase stability characteristics. Multiplying the original signal matrix with the separation matrix yields a line loss correlated effective signal submatrix containing signal components that satisfy the phase stability characteristics and a hybrid interference signal submatrix containing interference components with unstable phases.

6. The method for segmented monitoring of line losses based on distribution automation according to claim 5, characterized in that, The matrix parameters are iteratively adjusted using the gradient descent method to separate the mixing matrix and obtain a separation matrix that maximizes the component weights that satisfy the phase stability characteristics. This includes: calculating the initial inverse matrix based on the mixing matrix, using the quotient of the length of the continuous stable phase interval of the effective signal to the total sampling interval length as the objective function value, decoupling the original signal matrix based on the current mixing matrix in each iteration, calculating the objective function value of the decoupled matrix containing the original signal and interference signal, repeating the iteration until the objective function value exceeds the objective function threshold, stopping the iteration, and outputting the separation matrix; after obtaining the effective signal sub-matrix associated with line loss, feature purification is performed, including: removing signals with amplitudes lower than 0.1% of the injected signal; filtering signals with frequencies exceeding the excitation signal frequency band; smoothing the remaining signals; finally, a pure feature matrix containing five-dimensional features including pure reflection coefficient, pure phase shift, pure energy attenuation rate, pure line surface micro-temperature, and pure electric field distortion.

7. The method for segmented monitoring of line losses based on distribution automation according to claim 1, characterized in that, After obtaining the pure feature matrix, standardization is performed. The modal parameter recognition algorithm extracts the overall feature distribution from the pure feature matrix, which includes the mean, variance, and peak value of the original multimodal features. This distribution is then matched with the multimodal feature distribution under each loss scenario in the calibration library to determine the basic scenario type of the current line section. For each of the five-dimensional features of the pure feature matrix, the cosine similarity between each feature and its corresponding feature in the calibration library is calculated. The average of all similarities is then taken as the overall similarity between the calibration library and the five-dimensional features of the current line section, serving as the multi-dimensional cosine similarity. Data with similarities exceeding a similarity threshold are selected as a similarity subset. Phase... The measured values ​​of the topological equivalent parameters of each data set within the similarity subset are weighted and sorted. The average of the median of the measured values ​​of the topological equivalent parameters of each data set is used as the topological equivalent parameter matrix corresponding to the pure feature matrix. The inversion units are divided according to the deployment locations of the inherent topological nodes and micro-sensing nodes of the line. Specifically, the entire line is divided into several inversion units based on the main line segment points, branch line access points, and load access points as the basic boundaries, combined with the deployment spacing of micro-sensing nodes. Finally, the three-dimensional topological equivalent parameter prediction matrix of each line section is output, including the predicted node equivalent impedance, predicted line distributed capacitance, and predicted load coupling coefficient of each line section.

8. The method for segmented monitoring of line losses based on distribution automation according to claim 7, characterized in that, The segmented deviation rate, which includes the deviation rate of the equivalent impedance of the nodes, the deviation rate of the distributed capacitance of the lines, and the deviation rate of the load coupling coefficient, is calculated based on the predicted equivalent impedance of the nodes, the deviation rate of the distributed capacitance of the lines, and the deviation rate of the load coupling coefficient in each line section. The sliding window method is used to divide the changing trend according to the segmented deviation rate, including the steady trend, the rising trend and the sudden change trend, to complete the dynamic segmentation of each line section. There are cases where the deviation rate of the equivalent impedance of the nodes and the deviation rate of the load coupling coefficient of several line sections show an upward trend at the same time, which is a prediction of the increase in line loss caused by overload in several line sections.

9. A method for segmented monitoring of line losses based on distribution automation according to claim 8, characterized in that, This also includes: the existence of several line sections where the equivalent impedance deviation rate and load coupling coefficient deviation rate at the nodes exceed the preset threshold and simultaneously show an upward trend, which is a prediction of increased line loss due to overload in several line sections; the existence of several line sections where the distributed capacitance deviation rate exceeds the preset threshold and shows a stable trend and / or an upward trend, which is a prediction of leakage loss due to aging of the line insulation layer; and the existence of several line sections where the equivalent impedance deviation rate at the nodes exceeds the preset threshold and shows an abrupt trend, which is a prediction of loose line joints and / or instantaneous leakage.

10. A line loss segmentation monitoring system based on distribution automation, applied to the line loss segmentation monitoring method based on distribution automation as described in any one of claims 1-9, the system comprising: Calibration library construction module; The calibration experiment collects multimodal signal characteristics, measured topological equivalent parameters, and interference signals under different loss scenarios, and constructs a calibration library and interference sample set; initial topological parameter extraction and excitation signal generation modules are also included. This module is used to extract initial topology equivalent parameters by calling the calibration library and combining the current basic operating status of the line, locate impedance change points and mark sensitive areas through impedance difference analysis, and generate broadband excitation signals; it also includes a multi-mode signal acquisition and purification module. Raw multimodal signal data is collected through miniature sensing nodes, and effective signals and interference signals are separated using a separation mechanism. A pure feature matrix is ​​then output through a feature purification process. The topology parameter inversion and dynamic segmentation module is used to standardize and match the pure feature matrix through modal parameter recognition algorithms, invert the topology equivalent parameters, calculate the segmentation deviation rate and divide the trend of change, and complete the dynamic segmentation of the line and the pre-simulation of line loss anomaly types.

Images