A power distribution network fault monitoring system and method

Abstract

The application relates to the technical field of power distribution network fault monitoring, in particular to a power distribution network fault monitoring system and a monitoring method, which realizes real-time collection of multi-source data; edge calculation is used to preliminarily process and extract features from the multi-source data; distributed calculation nodes are used to further fuse and analyze the data by using multi-source data fusion technology, so as to generate a fault feature vector; the analysis results of the distributed calculation nodes are summarized, the fault feature vector is comprehensively analyzed, and the fault position is determined; the fault position result is compared and verified with historical data, if there is uncertainty, a secondary diagnosis process is started; according to the fault position result, fault early warning information is automatically sent to operation and maintenance personnel. By collecting multi-source data such as traveling wave signals, currents and voltages, and combining the data fusion technology of edge calculation and distributed calculation nodes, the accuracy of fault positioning is improved.

Description

Technical Field

[0001] This invention relates to the field of power distribution network fault monitoring technology, and in particular to a power distribution network fault monitoring system and monitoring method. Background Technology

[0002] Distribution networks play a crucial role in the power system. Faults in distribution networks, if not repaired promptly, can cause significant losses. In multi-terminal distribution networks, the time difference of traveling waves arriving at different measurement terminals is commonly used for precise location, but this method is inefficient.

[0003] The existing publication number CN114113948A discloses a method for monitoring faults in a distribution network, including: monitoring traveling wave signals at the beginning, end, and end of each branch of the distribution network as monitoring points; constructing a monitoring tree for each monitoring point; obtaining the first distance from each monitoring point to the fault point in response to a distribution network fault; obtaining the first set of fault branches on the monitoring tree for each monitoring point; filtering and reorganizing the first set of fault branches to obtain a second set of fault branches; adding the distances from the fault point to the first node among the elements with the same suspected branch in the second set of fault branches to obtain multiple judgment distances; selecting the maximum value among the judgment distances as the primary judgment distance, and the remaining judgment distances as secondary judgment distances; obtaining the fault distance based on the primary and secondary judgment distances; and thus obtaining the specific location of the fault point in the distribution network. This method improves the efficiency and accuracy of fault monitoring and location.

[0004] However, the aforementioned methods for monitoring power distribution network faults are highly dependent on the monitoring and analysis of traveling wave signals. If the traveling wave signal is interfered with during transmission, such as by electromagnetic interference or noise, its characteristics, such as waveform and propagation time, may change, thereby affecting the accuracy of fault location. Summary of the Invention

[0005] The purpose of this invention is to provide a power distribution network fault monitoring system and method, which solves the problem that existing power distribution network fault monitoring methods heavily rely on the monitoring and analysis of traveling wave signals. If the traveling wave signal is interfered with during transmission, such as by electromagnetic interference or noise, its characteristics, such as waveform and propagation time, may change, thereby affecting the accuracy of fault location.

[0006] To achieve the above objectives, the present invention provides a method for monitoring faults in a power distribution network, comprising the following steps:

[0007] Multi-source data monitoring devices are installed at the beginning, end, and end of each branch of the power distribution network to collect multi-source data in real time.

[0008] Edge computing is used to preprocess and extract features from the collected multi-source data, and the extracted feature data is uploaded to the nearest distributed computing node.

[0009] Distributed computing nodes receive feature data and use multi-source data fusion technology to further fuse and analyze the data to generate fault feature vectors;

[0010] By summarizing the analysis results of each distributed computing node, the fault feature vector is comprehensively analyzed to achieve accurate location and diagnosis of faults in the distribution network and determine the fault location.

[0011] The fault location results are compared and verified with historical data. If there is any uncertainty, a secondary diagnostic process is initiated.

[0012] Based on the fault location results, fault warning information is automatically sent to maintenance personnel.

[0013] The process includes installing multi-source data monitoring devices at the beginning, end, and branch ends of the power distribution network to collect multi-source data in real time. The steps also include:

[0014] The monitoring device includes a traveling wave signal sensor, a current transformer, and a voltage transformer, and the multi-source data includes traveling wave signal, current, and voltage.

[0015] The step of preprocessing and extracting features from the collected multi-source data using edge computing, and then uploading the extracted feature data to the nearest distributed computing node, further includes:

[0016] The collected multi-source data undergoes preliminary processing, including filtering, denoising, and normalization.

[0017] Extract the characteristic parameters of the multi-source data after preliminary processing, including the amplitude, frequency, phase, and abrupt change points of the traveling wave signal, and the effective value, peak value, harmonic content, and power factor of the voltage and current;

[0018] Principal component analysis or autoencoder can be used to reduce the dimensionality of the extracted feature parameters;

[0019] The dimensionality-reduced feature data is packaged into a standard format and timestamps and monitoring point identifiers are added.

[0020] Upload the feature data packet to the nearest distributed computing node.

[0021] The distributed computing nodes receive feature data, further fuse and analyze the data using multi-source data fusion technology, and generate fault feature vectors. The steps also include:

[0022] Distributed computing nodes unpack and verify feature data packets, extract feature data, and verify its integrity and consistency;

[0023] The feature data is time-aligned based on the timestamp and then normalized.

[0024] Weights are assigned to feature data from different sources, and the weights are dynamically adjusted based on the importance and reliability of the data;

[0025] The normalized feature data are fused to generate a comprehensive feature vector;

[0026] The fused feature data are concatenated into a high-dimensional fault feature vector;

[0027] The high-dimensional fault feature vector is optimized and stored after removing redundant information.

[0028] The process includes summarizing the analysis results from each distributed computing node, comprehensively analyzing the fault feature vectors to achieve accurate location and diagnosis of faults in the distribution network, and determining the fault location. The steps also include:

[0029] Based on the feature parameters in the high-dimensional fault feature vector, a matching analysis is performed using a predefined fault mode library to determine the specific type and severity of the fault.

[0030] By combining the topology of the distribution network and the traveling wave propagation model, the fault location can be determined. The location methods include the traveling wave method, the impedance method, or a machine learning-based location algorithm.

[0031] For fault location estimates provided by multiple distributed computing nodes, a weighted average method, voting method, or Bayesian fusion method is used to make a comprehensive judgment to determine the final fault location.

[0032] The process involves comparing the fault location results with historical data for verification. If uncertainty exists, a secondary diagnostic process is initiated. This process also includes:

[0033] Retrieve historical fault cases from the historical fault database that have similar characteristics to the current fault;

[0034] Compare the current fault location with the fault locations in historical fault cases, and calculate the location deviation;

[0035] If the positional deviation exceeds the preset threshold or there are significant differences in the fault characteristics, it is determined that there is uncertainty. In this case, multi-source data at the time of the fault occurrence is re-acquired and preliminarily processed and feature extracted.

[0036] The updated fault feature vectors are used to re-perform fault location and diagnostic analysis.

[0037] A power distribution network fault monitoring system includes a multi-source data monitoring module, an edge computing module, a distributed computing node, a fault diagnosis module, a historical data comparison module, and an early warning module. The edge computing module is connected to the multi-source data monitoring module, the distributed computing node is connected to the edge computing module, the fault diagnosis module is connected to the distributed computing node, the historical data comparison module is connected to both the fault diagnosis module and the multi-source data monitoring module, and the early warning module is connected to the historical data comparison module.

[0038] The multi-source data monitoring module is used to install multi-source data monitoring devices at the beginning, end and branch ends of the power distribution network to collect multi-source data in real time.

[0039] The edge computing module is used to preprocess and extract features from the collected multi-source data through edge computing, and upload the extracted feature data to the nearest distributed computing node;

[0040] The distributed computing node is used to receive feature data, further fuse and analyze the data using multi-source data fusion technology, and generate fault feature vectors.

[0041] The fault diagnosis module is used to summarize the analysis results of each distributed computing node, perform comprehensive analysis on the fault feature vector, realize the accurate location and diagnosis of faults in the distribution network, and determine the fault location.

[0042] The historical data comparison module is used to compare and verify the fault location results with historical data. If there is uncertainty, a secondary diagnostic process is initiated.

[0043] The early warning module is used to automatically send fault warning information to maintenance personnel based on the fault location results.

[0044] This invention discloses a power distribution network fault monitoring system and method. Multi-source data monitoring devices, including traveling wave signal sensors, current transformers, and voltage transformers, are installed at the beginning, end, and branch ends of the power distribution network to collect multi-source data such as traveling wave signals, current, and voltage in real time. Edge computing is used to filter, denoise, and normalize the collected multi-source data, extracting feature parameters. Principal component analysis or an autoencoder is used for dimensionality reduction, and the data is packaged and uploaded to the nearest distributed computing node. The distributed computing node receives the feature data, unpacks it, verifies it, aligns it in time, and normalizes it. After dynamically allocating weights, the data is fused to generate a high-dimensional fault feature vector, and redundant information is removed. The analysis results from each distributed computing node are summarized, and combined with a fault mode library, power distribution network topology, and traveling wave propagation model, the traveling wave method, impedance method, or machine learning algorithm is used to accurately locate and diagnose the fault. The fault location results are compared with historical data for verification. If uncertainty exists, data is re-collected and a secondary diagnostic process is initiated. Based on the finally determined fault location, a fault warning message is automatically sent to maintenance personnel. By collecting multi-source data such as traveling wave signals, current, and voltage, and combining edge computing and distributed computing node data fusion technologies, the system effectively reduces errors caused by single data sources, such as traveling wave signals affected by electromagnetic interference or noise, thus improving the accuracy of fault location. Edge computing is used for preliminary data processing and feature extraction, reducing data transmission volume. Simultaneously, efficient data fusion and analysis through distributed computing nodes significantly improve the efficiency of fault monitoring and location. Comparison with historical data and initiation of secondary diagnostic processes under uncertain conditions further enhance the reliability and accuracy of fault diagnosis. Employing traveling wave methods, impedance methods, and machine learning-based location algorithms, the system can adapt to fault location needs in different scenarios, enhancing its applicability and flexibility. Automatically sending fault warning information to maintenance personnel shortens fault response time and reduces losses caused by faults. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0046] Figure 1 This is a flowchart of the steps of the power distribution network fault monitoring method according to the first embodiment of the present invention.

[0047] Figure 2 This is a schematic diagram of the power distribution network fault monitoring system according to the second embodiment of the present invention.

[0048] In the diagram: 201-Multi-source data monitoring module, 202-Edge computing module, 203-Distributed computing node, 204-Fault diagnosis module, 205-Historical data comparison module, 206-Early warning module. Detailed Implementation

[0049] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0050] The first embodiment of this application is as follows:

[0051] Please see Figure 1 ,in, Figure 1 This is a flowchart of the steps of the power distribution network fault monitoring method according to the first embodiment of the present invention.

[0052] This invention provides a method for monitoring faults in a power distribution network, comprising the following steps:

[0053] S101: Install multi-source data monitoring devices at the beginning, end and branch ends of the distribution network to collect multi-source data in real time;

[0054] Specifically, the monitoring device includes a traveling wave signal sensor, a current transformer, a voltage transformer, and a smart terminal device. The traveling wave signal sensor captures transient voltage and current traveling wave signals generated during a fault. It can record the waveform characteristics of the fault traveling wave, including amplitude, frequency, phase, and abrupt changes, at a high sampling rate, such as 100kHz or higher. The current transformer monitors current changes in the distribution network in real time. It measures the effective value, peak value, harmonic content, and power factor of the current. The voltage transformer monitors voltage changes in the distribution network in real time. It measures the effective value, peak value, harmonic content, and power factor of the voltage. The smart terminal device integrates data acquisition, preprocessing, storage, and communication functions. It supports edge computing, performs preliminary processing and feature extraction on the acquired data, and uploads the processed data to the distributed computing node 203. Installing the multi-source data monitoring device at the beginning of the distribution network enables monitoring of changes in electrical parameters on the power supply side. Installed at the end of the distribution network, it can capture changes in electrical parameters on the load side. Installed at the end of each branch of the distribution network, it can monitor changes in electrical parameters of the branch lines and promptly detect faults in the branch lines. During monitoring, the traveling wave signal sensor uses a high sampling rate, such as 100kHz, to monitor transient voltage and current changes in real time to ensure the capture of high-frequency transient signals. The traveling wave signal is timestamped to record the precise time information of the signal. The current transformer monitors the current changes in the distribution network in real time, including RMS value, peak value, harmonic content, and power factor. The sampling frequency is set to 1kHz to 10kHz to capture dynamic changes in current. The current data is timestamped to record the precise time information of the data. The voltage transformer monitors the voltage changes in the distribution network in real time, including RMS value, peak value, harmonic content, and power factor. The sampling frequency is set to 1kHz to 10kHz to capture dynamic changes in voltage. The voltage data is timestamped to record the precise time information of the data.

[0055] S102: Preprocess and extract features from the collected multi-source data through edge computing, and upload the extracted feature data to the nearest distributed computing node 203;

[0056] Specifically, a low-pass filter is used to remove high-frequency noise and smooth the signal, and wavelet transform or adaptive filtering algorithms are employed to further remove noise while retaining the useful components of the signal. The processed data is then normalized to a uniform range, such as 0 to 1, for easier subsequent processing. The normalization formula is: Where X is the original data, and These represent the minimum and maximum values ​​of the data, respectively. Fault-related feature parameters are extracted from the preprocessed data, including: Traveling wave signal characteristic amplitude: the maximum value of the traveling wave signal; Frequency: the main frequency component of the traveling wave signal; Phase: the phase information of the traveling wave signal; Abrupt point: the abrupt change location of the traveling wave signal, used to capture the moment of fault occurrence; RMS current characteristic: the effective value of the current; Peak value: the maximum value of the current; Total harmonic distortion (THD): the total harmonic distortion of the current; Power factor: the power factor of the current. RMS voltage characteristic: the effective value of the voltage; Peak value: the maximum value of the voltage; Total harmonic distortion (THD): the total harmonic distortion of the voltage; Power factor: the power factor of the voltage. Principal component analysis (PCA) or an autoencoder is used to reduce the dimensionality of the extracted feature parameters, reducing the data volume while retaining key information. The dimensionality-reduced feature data can be transmitted and processed more efficiently. The dimensionality-reduced feature data is packaged into a standard format, with timestamps and monitoring point identifiers added. The packaged feature data is uploaded to the nearest distributed computing node 203 via wireless communication, such as 4G / 5G, or fiber optic communication. This ensures the reliability and real-time performance of data transmission, avoiding data loss or delay. Edge computing processes data near the data source, reducing data transmission latency and improving system real-time performance. Through feature extraction and dimensionality reduction, the amount of data uploaded to the distributed computing node 203 is reduced, lowering communication costs. Even under unstable communication link conditions, the edge computing module 202 can still complete preliminary processing, ensuring data integrity. Preliminary data processing locally reduces the risk of data exposure during transmission, enhancing system security.

[0057] S103: Distributed computing node 203 receives feature data, uses multi-source data fusion technology to further fuse and analyze the data, and generates fault feature vectors;

[0058] Specifically, distributed computing node 203 receives feature data packets from the edge computing system. The packet format should include a timestamp, monitoring point identifier, and feature data. The received data packets undergo integrity verification to ensure no data is lost or corrupted during transmission. The timestamp and monitoring point identifier are verified to ensure data source and time consistency. Feature data from different monitoring points is time-aligned based on the timestamp to ensure consistency across the time dimension. Alignment methods employ interpolation or sliding window techniques to handle time discrepancies. Feature data is normalized to unify data from different sources to the same range, such as 0 to 1. Weights are assigned to feature data from different sources, dynamically adjusted based on data importance and reliability. For example, traveling wave signals may have a higher weight than current or voltage data because traveling wave signals are more sensitive to fault responses. The normalized feature data is then weighted and averaged to generate a comprehensive feature vector. The calculation formula is as follows: ,in, Let i be the feature vector of the i-th monitoring point. The weights are used to concatenate the fused feature data into a high-dimensional fault feature vector. The fault feature vector should include feature parameters from multiple sources such as traveling wave signals, current, and voltage. Principal component analysis (PCA) or an autoencoder is used to reduce the dimensionality of the high-dimensional feature vector, removing redundant information. The optimized fault feature vector should retain key fault features to improve the efficiency of subsequent fault diagnosis. The generated fault feature vector is stored in the local memory of the distributed computing node 203 for subsequent analysis. The storage format should include a timestamp, monitoring point identifier, and fault feature vector. The fault feature vector is uploaded to the next step for global fault diagnosis and location. Through the distributed computing architecture, data processing tasks are distributed across multiple nodes, improving the overall efficiency of the system. Even if some nodes fail, the system can still operate normally, improving its fault tolerance. The distributed computing node 203 can quickly process and fuse data, ensuring real-time generation of the fault feature vector. The system can add or remove distributed computing nodes 203 as needed, flexibly adapting to different scales of distribution networks.

[0059] S104: Summarize the analysis results of each distributed computing node 203, perform comprehensive analysis on the fault feature vector, realize the accurate location and diagnosis of faults in the distribution network, and determine the fault location;

[0060] Specifically, a predefined fault mode library is constructed, containing feature vector templates for different fault types, such as short circuits, open circuits, and grounding faults. Each template contains typical values ​​and ranges of the fault feature vectors. The aggregated fault feature vectors are matched and analyzed against the templates in the fault mode library. Similarity metrics, such as Euclidean distance and cosine similarity, are used to calculate the similarity between the feature vectors and the templates. The template with the highest similarity is selected as the basis for fault type determination. Based on the fault feature vectors and the distribution network topology, the following method is used to accurately locate the fault: Traveling wave method: Utilizing the propagation characteristics of traveling wave signals, the distance from the fault point to each monitoring point is calculated. Based on the traveling wave time difference of arrival (TDOA) formula: ,

[0061] in, Let d be the time difference of arrival of the traveling wave, d be the distance from the fault point to the monitoring point, and v be the propagation speed of the traveling wave.

[0062] By combining time difference data from multiple monitoring points, the fault location is determined using geometric positioning algorithms (such as trilateration). Impedance method: The impedance of the fault point is calculated using current and voltage data. Based on the impedance formula: Where Z is the fault impedance, V is the fault voltage, and I is the fault current. The fault location is determined through impedance calculation based on the line parameters of the distribution network. A machine learning-based location algorithm is used: machine learning models, such as support vector machines and neural networks, are used to train and classify fault feature vectors. The input feature vectors are used, and the model outputs an estimate of the fault location. Model parameters are optimized using the training dataset to improve location accuracy. For fault location estimates provided by multiple distributed computing nodes 203, the following method is used for comprehensive judgment: Weighted average method: different weights are assigned to the fault location estimate of each node, and the weights are dynamically adjusted according to the reliability of the nodes. The weighted average is calculated as the final fault location. ,in, For the fault location estimation of the i-th node, Weights are used for each node. Voting method: Each node votes on the fault location, and the fault location with the most votes is selected as the final result. Bayesian fusion method: The fault location estimates from multiple nodes are fused using Bayes' theorem, and the posterior probability of each location is calculated. The location with the highest posterior probability is selected as the final fault location. Based on the fault location estimates generated by the above methods, combined with the distribution network topology and operating status, the final fault location is determined. The fault location should include specific line number, branch location, and distance information. The fault location information is output in a standardized format, including fault type, location coordinates, and estimation error range.

[0063] S105: Compare and verify the fault location results with historical data. If there is uncertainty, start the secondary diagnostic process.

[0064] Specifically, historical fault cases with similar characteristics to the current fault are retrieved from the historical fault database. Retrieval criteria include fault type, similarity of feature vectors (e.g., Euclidean distance, cosine similarity), and fault location information. The current fault location is compared with the fault locations in historical fault cases to calculate the location deviation. The location deviation formula is: Δd = |d| current -d history |, where d current d represents the current fault location. historyThe fault location is determined by comparing the current fault feature vector with the historical fault feature vector. Similarity metrics, such as Euclidean distance and cosine similarity, are used to assess the similarity of the feature vectors. Preset thresholds for location deviation and feature difference (e.g., similarity metric thresholds) are established. If the location deviation or feature difference exceeds the preset threshold, uncertainty is considered. If the verification results show uncertainty, a secondary diagnostic process is initiated to further confirm the fault location and type. This involves re-collecting data, preprocessing and extracting features, fusing and analyzing data, and comparing and confirming the results. The results of the secondary diagnosis are compared with the initial diagnosis to assess the differences. If the secondary diagnosis results are consistent with the initial diagnosis results, the fault location and type are confirmed. If the secondary diagnosis results still show uncertainty, a final judgment is made based on the experience of on-site maintenance personnel. After confirming the fault location and type, a final fault diagnosis report is generated. The report includes the fault type, location, diagnostic method, verification process, and reliability assessment. A reliability assessment is performed on the final fault diagnosis results, and the confidence level is calculated. The confidence level calculation formula is: Here, similarity is a measure of the similarity between the final fault feature vector and historical data. The fault diagnosis results are output in a standardized format, including information such as fault type, location, and confidence level.

[0065] By comparing historical data and re-collecting data, the possibility of misjudgments and omissions is reduced. Secondary diagnostics are initiated under uncertain conditions to ensure the reliability of fault diagnosis results. Detailed verification processes and confidence assessments are provided to help operations and maintenance personnel respond quickly to faults.

[0066] S106: Automatically send fault warning information to maintenance personnel based on the fault location results.

[0067] Specifically, the system pre-configures contact information for operations and maintenance personnel, including mobile phone numbers, email addresses, or instant messaging accounts. Based on the severity and location of the fault, alerts are sent to relevant operations and maintenance teams or individuals. Fault alerts are stored in the system database, recording the sending time, recipients, and fault details. The system generates detailed log files, recording the generation, sending, and receiving of fault alerts. It provides a fault alert traceability function, supporting historical record queries by time, fault type, location, and other criteria. Statistical analysis of fault alerts is performed to generate fault reports, helping operations and maintenance personnel understand fault distribution patterns and processing efficiency.

[0068] By collecting multi-source data such as traveling wave signals, current, and voltage, and combining edge computing and distributed computing node 203 data fusion technology, the errors caused by single data sources, such as traveling wave signals affected by electromagnetic interference or noise, are effectively reduced, improving the accuracy of fault location. Edge computing is used for preliminary data processing and feature extraction, reducing data transmission volume. Simultaneously, efficient data fusion and analysis through distributed computing node 203 significantly improves the efficiency of fault monitoring and location. Comparison with historical data and initiation of a secondary diagnostic process under uncertainty further improves the reliability and accuracy of fault diagnosis. Employing traveling wave method, impedance method, and machine learning-based location algorithms, the system can adapt to fault location needs in different scenarios, enhancing its applicability and flexibility. Automatically sending fault warning information to maintenance personnel shortens fault response time and reduces losses caused by faults.

[0069] The second embodiment of this application is as follows:

[0070] Based on the first embodiment, please refer to Figure 2 ,in, Figure 2 This is a schematic diagram of the power distribution network fault monitoring system according to the second embodiment of the present invention.

[0071] This embodiment of a power distribution network fault monitoring system includes a multi-source data monitoring module 201, an edge computing module 202, a distributed computing node 203, a fault diagnosis module 204, a historical data comparison module 205, and an early warning module 206.

[0072] In this specific embodiment, the edge computing module 202 is connected to the multi-source data monitoring module 201, the distributed computing node 203 is connected to the edge computing module 202, the fault diagnosis module is connected to the distributed computing node 203, the historical data comparison module 205 is connected to the fault diagnosis module 204 and the multi-source data monitoring module 201 respectively, and the early warning module 206 is connected to the historical data comparison module 205.

[0073] The multi-source data monitoring module 201 is used to install multi-source data monitoring devices at the beginning, end and branch ends of the power distribution network to collect multi-source data in real time.

[0074] The edge computing module 202 is used to preprocess and extract features from the collected multi-source data through edge computing, and upload the extracted feature data to the nearest distributed computing node 203.

[0075] The distributed computing node 203 is used to receive feature data, further fuse and analyze the data using multi-source data fusion technology, and generate fault feature vectors.

[0076] The fault diagnosis module 204 is used to summarize the analysis results of each distributed computing node 203, perform comprehensive analysis on the fault feature vector, realize the accurate location and diagnosis of the fault in the distribution network, and determine the fault location.

[0077] The historical data comparison module 205 is used to compare and verify the fault location results with historical data. If there is uncertainty, a secondary diagnostic process is initiated.

[0078] The early warning module 206 is used to automatically send fault warning information to maintenance personnel based on the fault location result.

[0079] Using a power distribution network fault monitoring system according to this embodiment, the multi-source data monitoring module 201 collects multi-source data in real time and uploads it to the edge computing module 202. The edge computing module 202 preprocesses and extracts features from the collected multi-source data and uploads the extracted feature data to the nearest distributed computing node 203. The distributed computing node 203 uses multi-source data fusion technology to further fuse and analyze the data, generate fault feature vectors, and uploads them to the fault diagnosis module 204. The fault diagnosis module 204 summarizes the analysis results of each distributed computing node 203, performs comprehensive analysis on the fault feature vectors, realizes accurate location and diagnosis of power distribution network faults, determines the fault location, and uploads it to the historical data comparison module 205. The historical data comparison module 205 compares and verifies the fault location result with historical data. If there is uncertainty, it uploads a signal to the multi-source data monitoring module 201 to start a secondary diagnosis process. Finally, the early warning module 206 receives the fault location result and automatically sends fault early warning information to the operation and maintenance personnel.

[0080] By combining multi-source data (traveling wave signals, current, voltage) and multi-source data fusion technology, errors from single data sources are reduced, improving the accuracy of fault location and diagnosis. Edge computing and distributed computing architectures are used to distribute data processing tasks across multiple nodes, reducing the computational burden on the central system and improving overall system efficiency and real-time performance. Historical data comparison and verification, along with secondary diagnostic processes, ensure the reliability of fault diagnosis results, reducing the possibility of misdiagnosis and missed diagnosis. Fault warning information is automatically sent to maintenance personnel, shortening fault response time and reducing power outage time and economic losses.

[0081] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.

Claims

1. A power distribution network fault monitoring method characterized by, The method comprises the following steps: installing multi-source data monitoring devices at the head end, tail end and end of each branch of the power distribution network to collect multi-source data in real time, wherein the monitoring device comprises a traveling wave signal sensor, a current transformer and a voltage transformer, and the multi-source data comprises a traveling wave signal, current and voltage; The collected multi-source data is preprocessed and feature extracted by edge computing, and the extracted feature data is uploaded to the nearest distributed computing node. The collected multi-source data is preliminarily processed, including filtering, denoising and normalization operation, wherein the normalization formula is: Wherein, X is the original data, And The minimum and maximum values of the data, respectively; the feature parameters of the multi-source data after preliminary processing, including the amplitude, frequency, phase and mutation point of the traveling wave signal, the effective value, peak value, harmonic content and power factor of the voltage and current; using principal component analysis or autoencoder for dimension reduction processing of the extracted feature parameters; packing the feature data after dimension reduction into a standard format, and attaching a timestamp and a monitoring point identifier; uploading the feature data packet to the nearest distributed computing node; The distributed computing node receives feature data, further fuses and analyzes the data by using a multi-source data fusion technology, and generates a fault feature vector, wherein the distributed computing node unpacks and verifies the feature data packet, extracts the feature data and verifies the integrity and consistency thereof; time aligns the feature data according to a time stamp, and performs normalization processing; assigns weights to feature data of different sources, and dynamically adjusts the weights according to the importance and reliability of the data; fuses the normalized feature data to generate a comprehensive feature vector, and the calculation formula is: wherein, is a feature vector of an i th monitoring point, is a weight; splices the fused feature data into a high-dimensional fault feature vector; optimizes the high-dimensional fault feature vector, and stores the high-dimensional fault feature vector after removing redundant information; The analysis results of each distributed computing node are summarized, the fault feature vector is comprehensively analyzed, the accurate positioning and diagnosis of the power distribution network fault are realized, and the fault position is determined, wherein according to the characteristic parameters in the high-dimensional fault feature vector, the pre-defined fault mode library is used for matching analysis, the specific type and severity of the fault are determined, the fault position is positioned in combination with the topological structure of the power distribution network and the traveling wave propagation model, the positioning method includes the traveling wave method, the impedance method or the positioning algorithm based on machine learning, for the fault position estimation provided by multiple distributed computing nodes, the weighted average method, the voting method or the Bayesian fusion method is used for comprehensive judgment to determine the final fault position, wherein the weighted average value is calculated as the final fault position: wherein, is the fault position estimation of the i th node, is the weight; comparing the fault location result with historical data to verify whether there is uncertainty, and if so, starting a secondary diagnosis process; automatically sending fault warning information to the operation and maintenance personnel according to the fault location result.

2. The power distribution network fault monitoring method of claim 1, wherein, comparing the fault location result with historical data to verify whether there is uncertainty, and if so, starting a secondary diagnosis process, the step further comprising: retrieving a historical fault case similar to the current fault characteristics from a historical fault database; comparing the current fault location with the fault location in the historical fault case to calculate a location deviation; if the location deviation exceeds a preset threshold or the fault characteristics have significant differences, it is determined that there is uncertainty, at which time the multi-source data at the time of fault occurrence is re-collected, and preliminary processing and feature extraction are performed; re-performing fault positioning and diagnosis analysis using the updated fault feature vector.

3. A power distribution network fault monitoring system suitable for the power distribution network fault monitoring method of claim 1, characterized in that, comprising a multi-source data monitoring module, an edge computing module, a distributed computing node, a fault diagnosis module, a historical data comparison module and a warning module, the edge computing module is connected with the multi-source data monitoring module, the distributed computing node is connected with the edge computing module, the fault diagnosis module is connected with the distributed computing node, the historical data comparison module is connected with the fault diagnosis module and the multi-source data monitoring module respectively, and the warning module is connected with the historical data comparison module; the multi-source data monitoring module is used for installing multi-source data monitoring devices at the head end, tail end and end of each branch of the power distribution network to collect multi-source data in real time; the edge computing module is used for pre-processing and feature extraction of the collected multi-source data through edge computing, and uploading the extracted feature data to the nearest distributed computing node; the distributed computing node is used for receiving feature data, further fusing and analyzing the data using multi-source data fusion technology to generate a fault feature vector; the fault diagnosis module is used for summarizing the analysis results of each distributed computing node, comprehensively analyzing the fault feature vector to realize accurate positioning and diagnosis of the power distribution network fault, and determining the fault location; the historical data comparison module is used for comparing the fault location result with historical data to verify whether there is uncertainty, and if so, starting a secondary diagnosis process; the warning module is used for automatically sending fault warning information to the operation and maintenance personnel according to the fault location result.

Images