A 10KV power distribution network fault automatic diagnosis positioning method

Abstract

The application provides a 10KV power distribution network fault automatic diagnosis positioning method, constructs a confidence evaluation model by collecting current direction and voltage polarity data, and carries out cross verification combined with a historical fault feature database to obtain a preliminary and adjusted confidence sequence; in view of the interference monitoring point, the weight of the interference monitoring point is reduced, a weighted polarity sequence is generated, and mode matching is carried out to determine a preliminary fault section; a multi-dimensional verification matrix is constructed by fusing the power distribution automation switch action and the user repair geographic position, the spatial deviation is quantified, and when the threshold is exceeded, deep analysis is started, the deviation sequence is corrected by means of the consistency filtering of the power distribution network topological connection relationship and the business standby path, finally, the information fusion technology is adopted to integrate the multi-source results to output the accurate fault position, and the positioning result is fed back to update the fault feature database and the confidence model parameters, adaptive optimization diagnosis is realized, and the diagnosis precision is improved.

Description

Technical Field

[0001] This invention relates to an automatic fault diagnosis and location method for 10kV distribution networks. Background Technology

[0002] In power grid transmission, the 10kV distribution network serves as a crucial link connecting high-voltage transmission lines and low-voltage users. Its operational stability and reliability directly impact the electricity safety and economic benefits of countless households. Especially in urban and industrial areas, if distribution network faults are not detected and addressed promptly, they can trigger widespread power outages and even cause significant economic losses. Current fault location methods for distribution networks often struggle to adapt to complex environments and changing operating conditions. They are particularly susceptible to electromagnetic interference, equipment aging, or changes in the external environment, leading to biased information and inaccurate location results. Furthermore, existing technologies lack effective integration mechanisms when processing information from multiple sources, making it difficult to comprehensively consider the reliability and interrelationships of different information sources, thus affecting the overall effectiveness of fault diagnosis.

[0003] A deeper challenge lies in accurately assessing and utilizing the key information characteristics collected from monitoring points, such as the features of current direction and voltage changes. These characteristics are often unstable in practical applications due to interference, leading to errors in the initial judgment of the fault area. This instability further affects the subsequent accurate confirmation of the fault location, especially when the power grid structure is complex and monitoring points are widely distributed. Relying solely on one type of information characteristic is unlikely to fully reflect the true location of the fault.

[0004] Therefore, how to accurately assess the reliability of key information characteristics in complex environments, and on this basis integrate information from multiple sources to achieve accurate fault location confirmation, has become a key issue in improving the efficiency of power distribution network fault diagnosis. Summary of the Invention

[0005] This invention proposes an automatic fault diagnosis and location method for 10KV distribution networks, which effectively overcomes the problems of polarity signal distortion and location deviation in complex electromagnetic environments, and significantly improves the location accuracy and robustness of faults such as single-phase grounding in distribution networks.

[0006] The technical solution of this invention is implemented as follows: An automatic fault diagnosis and location method for 10kV distribution networks includes: S1. Collect current direction and voltage polarity data to reflect line load changes and short-circuit signal characteristics, establish a confidence assessment model, determine the confidence weight of each polarity feature, and obtain a preliminary weight distribution. S2, combined with the pre-established fault feature database of polarity patterns of historical faults in the distribution network, data cross-validation is performed, the polarity discrimination confidence level of each node is adjusted, and the adjusted confidence sequence is obtained. S3. If there are monitoring points affected by electromagnetic interference in the adjusted confidence sequence, the weight of the monitoring point data is reduced, a weighted polarity sequence is generated by comprehensively considering the interference intensity factor, and the sequence pattern is matched and compared with the historical polarity pattern in the fault feature database to determine the preliminary fault section location. S4. Extract switch action information and user repair location data from the power distribution automation system. By constructing a multi-dimensional verification matrix, compare the correlation between time and space dimensions to obtain spatial deviation index, calculate the location offset, and determine the deviation quantification score. S5. If the deviation quantification score exceeds the preset threshold, the deep analysis unit is activated to re-evaluate the effectiveness of the polarity characteristics of each monitoring point, and the distribution network topology connection relationship is integrated with the data filtering mechanism. The consistency attribute is filtered through the business backup path to correct the redundant deviation of the deviation sequence and obtain the deviation correction sequence. S6 uses information fusion technology to integrate multi-source verification results, process the reliability of different data sources, output the verified precise location of the fault, and determine the final location result; S7. Update the typical cases in the fault feature database with the final location results, and optimize the evaluation framework for the parameter generation of the confidence assessment model. Preferably, in step S1, the current direction and voltage polarity data are initially classified by line load changes and short-circuit signal characteristics, the signal characteristic analysis results of each data point are determined, and a confidence evaluation model is established. By weighted summation of polarity feature vectors, the confidence weight of each polarity feature is obtained from the confidence evaluation model, and a preliminary weight distribution is obtained.

[0007] Outliers in the initial weight distribution are obtained. If the outliers exceed a preset threshold as deviations from the average, the weights are adjusted through power distribution network monitoring to determine the corrected weight distribution. Fault location auxiliary information is extracted from the corrected weight distribution, and the matching degree between the information and short-circuit signal features is determined by similarity comparison to obtain the final weight distribution.

[0008] Preferably, in step S3, by obtaining the adjusted confidence sequence, the monitoring points affected by electromagnetic interference are identified, the weight of the monitoring point data is reduced, and the interference intensity factor is comprehensively considered. The interference intensity factor is obtained by weighted summation of the monitoring point signal attenuation amplitude and the interference duration to generate a weighted polarity sequence and determine the sequence polarity characteristics. By combining historical polarity patterns in the fault feature database, sequence pattern matching and comparison are performed to determine the preliminary fault section location. Interference source localization verification is then performed, which involves fusing signal source data from multiple monitoring points using triangulation and fusing section boundary data to obtain the accurate section boundary.

[0009] Preferably, in step S4, a multi-dimensional verification matrix is ​​constructed using the preliminary fault section results, switch action information, and user repair geographical location data. The time dimension is then compared and correlated to obtain a time consistency index. For the time consistency index, the spatial deviation index is calculated by comparing the geographic location data and the preliminary fault section results in the spatial dimension. The spatial deviation index is obtained by comparing the difference in geographic coordinates. Historical fault record data is obtained and the location offset is calculated. The location offset is obtained by the vector difference between historical data and the current deviation index. The deviation quantification score is obtained by weighted averaging the offset and historical records.

[0010] Preferably, in step S6, the initial fault location is obtained from the multi-source verification results through the deviation correction sequence, and the initial fault location is integrated using information fusion technology to obtain a preliminary fused location; for the preliminary fused location, the reliability of the data source is obtained, and the reliability weight allocation is determined. Different data sources are processed using reliability weight allocation to obtain a weighted fused location. Location deviation correction is applied, and verification data is integrated. If the deviation exceeds a preset threshold, correction is performed to obtain the corrected location. From the corrected location, a source credibility assessment is obtained. The source credibility assessment and reliability weight allocation are processed to determine the positioning accuracy verification, obtain the verified location, output the verified fault location, and determine the final positioning result.

[0011] Preferably, in step S7, Typical case data are obtained from the fault feature database to determine the updated set of typical cases. The parameters of the confidence assessment model are adjusted. The confidence assessment model adopts the logistic regression model. The input is the feature vector in the set of typical cases, and the output is the adjusted parameter value to obtain the optimized parameter value. Based on the optimized parameter values, the output of the confidence assessment algorithm is obtained. The confidence assessment algorithm calculates the probability score through the logistic regression model to determine the confidence of the anomaly pattern recognition. The anomaly pattern recognition extracts fault feature patterns from a set of typical cases. If the confidence level exceeds a preset threshold, the verification criteria are extracted from the case matching rules, wherein the case matching rules are pre-established as a correspondence table based on similarity comparison to obtain a draft framework structure. The proposed framework is adapted by incorporating an optimization iteration loop and an application extension of the results. The optimization iteration loop adjusts the structure by repeatedly comparing the draft with a set of typical cases, while the application extension of the results adds a verification path for anomaly pattern recognition, thereby generating an optimized evaluation framework.

[0012] In another aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the method described thereon.

[0013] In another aspect, the present invention also provides a computer program product, comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method described herein.

[0014] The beneficial effects of this invention are as follows: This invention addresses the problem of misjudgment and location deviation of fault sections caused by electromagnetic interference, data noise, and spatiotemporal inconsistencies in multi-source information in polarity detection devices in power distribution lines. It constructs a confidence assessment model by collecting current direction and voltage polarity data and cross-validates it using a historical fault feature database to obtain preliminary and adjusted confidence sequences. For interference-affected monitoring points, its weight is reduced and a weighted polarity sequence is generated. Pattern matching is then performed to determine the preliminary fault section. A multi-dimensional verification matrix is ​​constructed by integrating the actions of distribution automation switches and the geographical location of user repair requests. Spatial deviations are quantified, and in-depth analysis is initiated when they exceed a threshold. The deviation sequence is filtered and corrected using the consistency of distribution network topology connections and backup business paths. Finally, information fusion technology is used to integrate multi-source results to output the precise fault location. The location results are fed back to update the fault feature database and confidence model parameters, achieving adaptive optimization diagnosis and improving diagnostic accuracy. Detailed Implementation

[0015] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0016] An automatic fault diagnosis and location method for 10kV distribution networks includes: S1. Collect current direction and voltage polarity data to reflect line load changes and short-circuit signal characteristics, establish a confidence assessment model, determine the confidence weight of each polarity feature, and obtain a preliminary weight distribution.

[0017] The current direction and voltage polarity data are initially classified based on line load changes and short-circuit signal characteristics. The signal characteristic analysis results for each data point are determined, and a confidence assessment model is established. By weighted summation of polarity feature vectors, the confidence weight of each polarity feature is obtained from the confidence assessment model, resulting in a preliminary weight distribution. Outliers in the preliminary weight distribution are identified. If the outliers exceed a preset threshold as deviations from the average, the weights are adjusted through power distribution network monitoring to determine the corrected weight distribution. Fault location auxiliary information is extracted from the corrected weight distribution, and the matching degree between the information and short-circuit signal characteristics is determined by similarity comparison to obtain the final weight distribution.

[0018] In one specific embodiment, in a 10kV distribution line, data is acquired from a polarity detection device installed at the substation outlet and user-side branch node to monitor the current direction in real time. For example, positive current flow indicates normal load, while reverse current flow may indicate reverse power or fault injection. At the same time, voltage polarity data captures positive and negative waveform changes to identify phase shifts. Preliminary classification is performed based on line load changes. For example, assuming a line load fluctuates from a normal 50A to a sudden 150A, combined with the signal of voltage polarity changing from positive to negative, the system extracts frequency components through Fourier transform to determine the feature vector of each data point, including amplitude, phase, and duration, thereby obtaining signal feature analysis results, such as "short-circuit characteristics dominated by high-frequency harmonics" or "normal characteristics of low-frequency load changes".

[0019] When establishing a confidence assessment model based on signal feature analysis results, a machine learning-based framework can be used, such as a hybrid model combining support vector machines and Bayesian networks. The polarity feature vector includes current direction vector, voltage polarity vector, and time series vector, with weights obtained through weighted summation. For example, current direction can be weighted at 0.4, voltage polarity at 0.3, and time series at 0.3. Then, the inner product of the vectors is calculated to obtain the confidence score for each feature. For instance, the confidence score for current direction is 0.85, indicating its reliability in the overall polarity judgment, thus forming a preliminary weight distribution.

[0020] Assuming the analysis shows a significant deviation in the voltage polarity vector during short-circuit events, the model is trained using historical data. After calculating a weighted sum, a preliminary distribution is derived, where the current characteristic weight dominates at 0.6, while voltage is only 0.2. This reflects the more reliable nature of the current signal under high load conditions. The model's development involves data preprocessing, feature extraction, and parameter optimization. For example, vector values ​​are first normalized, and then weights are adjusted using gradient descent to minimize confidence error and ensure the model's adaptability to different network topologies. When identifying outliers in the preliminary weight distribution, in the above case, it is assumed that the voltage polarity weight is abnormally high at 0.5, while the average weight is 0.25. If the deviation exceeds a preset threshold, such as 0.2, it is considered an anomaly. Weights are adjusted through distribution network monitoring, such as by accessing real-time monitoring data like transformer temperature and line impedance changes. The weights are recalculated, reducing the outlier to 0.3, thus determining the corrected distribution.

[0021] When extracting fault location auxiliary information from the corrected weight distribution, the subset dominated by the current direction weight is extracted and compared with the short-circuit signal features such as peak current and duration. The matching degree is calculated using the cosine similarity formula. If it exceeds 0.8, a high match is confirmed, and the final weight distribution is obtained.

[0022] S2. Combine the pre-established fault feature database of polarity modes of historical faults in the distribution network, perform data cross-validation, adjust the polarity discrimination confidence level of each node, obtain the adjusted confidence sequence, and determine the fault mode matching degree by comparing the confidence sequence with the polarity mode.

[0023] Pre-established fault characteristic databases typically store a large amount of historical data, including current and voltage polarity change patterns under various short-circuit events. The database is built based on long-term monitoring, collecting data through sensors installed at substations and line nodes, and labeling patterns as positive polarity reversal or negative polarity stability, etc. When a current fault needs to be addressed, historical polarity patterns related to similar load conditions are first extracted from the database. For example, when a ground fault occurs on a high-voltage distribution line, historical records show that the current direction changes from positive to negative, while the voltage polarity momentarily reverses.

[0024] In one specific embodiment, a vector comparison method can be used when performing data cross-validation based on the extracted polarity pattern and the preliminary weight distribution. Specifically, the preliminary weight distribution is the confidence value of each polarity feature obtained from previous steps, for example, a current direction weight of 0.6 and a voltage polarity weight of 0.4. Cross-validation then compares these weights with historical patterns. Assuming that the historical pattern shows an 80% probability of current reversal at the upstream node of the short circuit point, while the preliminary weight distribution estimates it to be 70%, consistency is verified by calculating a deviation value, such as an absolute difference of 0.1. If the deviation is too large, the system will mark it as a potential anomaly and further iterate to improve accuracy.

[0025] The process of adjusting the polarity confidence level of each node based on cross-validation results can be understood as a dynamic optimization step. Specifically, in a 10kV distribution line, the initial confidence level of node A is 0.75. After validation, it is found to deviate from the historical pattern by 0.2, so it is adjusted to 0.65. A weighted average formula is used, such as the new level equals the old level minus the deviation multiplied by an adjustment factor of 0.5. This adjustment considers the correlation between multiple nodes in the network; for example, the downstream influence of node B will trigger an adjustment of node A, ensuring that the confidence level of the entire sequence is more consistent with reality.

[0026] When determining the fault mode matching degree by comparing the adjusted confidence sequence with the polarity pattern, similarity calculation can be used. Specifically, a confidence sequence such as [0.65, 0.70, 0.55] corresponds to a network node, and the historical polarity pattern is [0.80, 0.75, 0.60]. During comparison, Euclidean distance or cosine similarity is calculated. For example, a similarity of 0.85 or higher is considered a high match, indicating that the current fault mode is similar to the historical ground fault. This comparison not only verifies the validity of the sequence but also provides precise guidance for fault location. In practical applications, it can effectively distinguish between single-phase short circuits and three-phase short circuits, improving diagnostic efficiency and reducing economic losses.

[0027] S3. If there are monitoring points affected by electromagnetic interference in the adjusted confidence sequence, the weight of the monitoring point data is reduced, a weighted polarity sequence is generated by comprehensively considering the interference intensity factor, and the sequence pattern is matched and compared with the historical polarity pattern in the fault feature database to determine the preliminary fault section location. By acquiring the adjusted confidence sequence, monitoring points affected by electromagnetic interference are identified, the weight of the monitoring point data is reduced, and the interference intensity factor is comprehensively considered. The interference intensity factor is obtained by weighted summation of the monitoring point signal attenuation amplitude and the interference duration to generate a weighted polarity sequence and determine the sequence polarity characteristics. Combined with historical polarity patterns in the fault feature database, sequence pattern matching and comparison are performed to determine the preliminary fault section location. Interference source localization verification is carried out, where interference source localization verification is performed by fusion of signal source data from multiple monitoring points using triangulation and fusion of section boundary data to obtain the accurate section boundary.

[0028] In one specific embodiment, identifying monitoring points affected by electromagnetic interference by acquiring an adjusted confidence sequence first requires analyzing the fluctuations in the confidence values ​​of each node in the sequence. If the confidence value of a monitoring point is abnormally low or unstable, it can be determined that it is affected by electromagnetic interference. Assuming multiple monitoring points are distributed along a high-voltage line, when the adjusted confidence sequence shows that the confidence value of monitoring point A drops sharply from 0.85 to 0.4, this may be due to electromagnetic radiation interference from a nearby substation. In this case, the system will automatically mark the point and reduce its data weight to 60% of its original value, thereby generating an interference adjustment sequence to ensure the reliability of the overall sequence and avoid interference data misleading fault diagnosis.

[0029] The interference intensity factor is obtained by weighted summation of the signal attenuation amplitude and interference duration at the monitoring point. Here, the signal attenuation amplitude refers to the percentage decrease in signal strength from the normal level, while the interference duration records the length of time the interference occurs. If the signal attenuation amplitude is 30% and the interference duration is 5 minutes, a weighted formula can be used, where the attenuation amplitude has a weight of 0.7 and the duration has a weight of 0.3. The calculated interference intensity factor is 30%*0.7 + 5*0.3 = 22.5, thus generating a weighted polarity sequence from the overall interference adjustment sequence. The specific process involves multiplying each element in the interference adjustment sequence by 1 and subtracting the proportion of the interference intensity factor. For example, if the original sequence value is positive polarity 1, the weighted result is 1*(1-0.225) = 0.775. Finally, the sequence polarity characteristics are determined. If the overall sequence shows an alternating positive and negative pattern, it indicates a potential grounding fault.

[0030] After obtaining the weighted polarity sequence, sequence pattern matching and comparison are performed in conjunction with historical polarity patterns in the fault feature database. First, the weighted sequence is compared one by one with various historical patterns stored in the database. For example, if the database contains continuous positive polarity patterns for short-circuit faults and alternating polarity patterns for overload faults, and the weighted polarity sequence is [+1,-1,+1,-1], while the historical patterns for overload faults correspond to similar alternating patterns, then the matching degree is calculated to be 80% similarity, thus determining the initial fault location to be between line segments B and C. This comparison helps to quickly locate potential problem areas and improve diagnostic efficiency.

[0031] Interference source localization verification was conducted to locate the initial fault section. This verification was performed using triangulation, which integrates signal data from multiple monitoring points. The principle of triangulation is to calculate the geometric location of the interference source using signal data from at least three monitoring points. For example, monitoring points D, E, and F are selected, and the distance or angle to the interference source is measured at each point. The data is then integrated by solving for the relationships between the sides and angles of the triangles. The specific process involves collecting the signal strength and direction at each point to form a coordinate system. For instance, if point D has coordinates (0,0) with a signal direction of east, point E has coordinates (10,0) with a signal direction of west, and point F has coordinates (5,5) with a signal direction of south, then the interference source is calculated to be located near position (5,2). Subsequently, the section boundary data, such as the coordinates of the line's starting and ending points, is integrated, and the boundary is adjusted from the initial BC section to a point 500 meters from the midpoint between points B and C. This ensures accurate section boundaries and targeted subsequent maintenance.

[0032] S4 extracts switch action information and user repair location data from the power distribution automation system, constructs a multi-dimensional verification matrix to compare the time and space dimensions, obtains the spatial deviation index, calculates the location offset, and determines the deviation quantification score.

[0033] A multi-dimensional verification matrix is ​​constructed by combining preliminary fault section results, switch action information, and user repair location data. The time dimension is correlated and compared to obtain a time consistency index. For the time consistency index, the location data and preliminary fault section results are compared in the spatial dimension to calculate a spatial deviation index. The spatial deviation index is obtained by comparing the differences in geographic coordinates. Historical fault record data is obtained to calculate the location offset. The location offset is obtained by the vector difference between historical data and the current deviation index. The deviation quantification score is obtained by weighted averaging the offset and historical records.

[0034] In one specific embodiment, when obtaining preliminary fault section results from the distribution automation system, data can be extracted through the system's built-in fault detection module. This module typically uses current and voltage anomaly monitoring to initially locate the fault section; for example, a section of cable from substation A to user group B might be identified as a potential fault area. Simultaneously, switch action information includes circuit breaker trip records and reclosing operation logs, reflecting the sequence of protection actions at the time of the fault. Furthermore, user-reported repair location data can be collected through a mobile app or customer service system. Addresses reported by users, such as "power outage near XX," are converted into GPS coordinates for subsequent processing. This acquisition of data ensures the initial integration of multi-source information, providing a foundation for subsequent verification.

[0035] When constructing a multidimensional verification matrix using preliminary fault section results, switch action information, and user repair location data, this data can be organized into a matrix structure. Rows represent timestamps, and columns correspond to section location, switch status, and user location, respectively. A single row in the matrix might show 1:00 AM on January 6, 2026, with the preliminary fault section being interval AB, the switch action recording showing the tripping of circuit breaker C, and user repair locations concentrated near point B. Then, the time dimension is correlated and compared, calculating the timestamp differences between data sources. If the difference is less than a preset threshold, such as 5 minutes, the time consistency index is high. This index can be obtained through simple averaging or weighted methods, such as using the switch action time as a reference benchmark for comparison with other data, thereby assessing the overall consistency of the event.

[0036] Regarding the time consistency index, when comparing geographic location data and preliminary fault segment results in the spatial dimension, GIS tools can be used to map coordinates and calculate spatial deviation indices. For example, the difference between the user's reported repair coordinates and the segment boundary can be compared using the Euclidean distance formula. If the user's location deviates from the segment center by more than 500 meters, the deviation index increases. This comparison helps identify spatial mismatches between data. The cause may be coordinate errors due to vague descriptions of the user's reported repair location, which may affect the accuracy of fault location. However, this index allows for timely adjustments to the verification strategy.

[0037] When acquiring historical fault records based on the spatial deviation index, fault logs for similar sections within the past year can be retrieved from the database, such as cable aging fault records that occurred in section AB. Then, the location offset is calculated as the vector difference between the current deviation index and historical data. For example, if the current deviation is 200 meters east and the historical average is 150 meters east, the offset is 50 meters. This calculation process emphasizes incorporating historical experience and avoids potential biases caused by relying solely on current data.

[0038] When determining the deviation quantification score using location offset and historical fault record data, a weighted average method can be used, where the offset has a weight of 0.6 and the historical record has a weight of 0.4. For example, the weighted result of an offset of 50 meters and a historical average deviation of 30 meters is 42 points. This score quantifies the degree of deviation, leading to more accurate fault location, reducing maintenance delays caused by misjudgments, and thus improving the reliability of the power distribution system.

[0039] S5. If the deviation quantification score exceeds the preset threshold, the deep analysis unit is activated to re-evaluate the effectiveness of the polarity characteristics of each monitoring point, and the distribution network topology connection relationship is integrated with the data filtering mechanism. The consistency attribute is filtered through the business backup path to correct the redundant deviation of the deviation sequence and obtain the deviation correction sequence.

[0040] In one specific embodiment, the deviation quantification score is compared with a preset threshold to determine whether the threshold has been exceeded. The core of this step is to quantify the severity of the deviation to determine subsequent processing. Assuming the deviation quantification score is a weighted average of the previously calculated position offset and historical fault records, if the score is 0.75 and the preset threshold is 0.5, then it is determined to exceed the threshold. This means that the initial fault segment result may have a large error and requires further verification. Through this comparison, potential problem areas can be effectively screened, thereby improving the accuracy of fault location.

[0041] When the deviation quantification score exceeds the threshold, the system automatically activates the deep analysis unit to collect polarity characteristic validity data from monitoring points distributed throughout the distribution network. These monitoring points typically include sensors installed in substations and feeders, which record real-time changes in current and voltage polarity. If a monitoring point detects that the polarity of an upstream switch is positive while that of a downstream switch is negative, it indicates that a fault may be located in a specific section.

[0042] For the collected polarity feature validity data, a data filtering mechanism is employed to fuse the distribution network topology connections, thereby obtaining a topology verification sequence. The data filtering mechanism is an algorithmic framework that first removes invalid data points, such as abnormal fluctuation values, and then matches the remaining data with the topology structure. For example, in a 10kV feeder, the topology connections describe the nodes and branches from the substation to the user end. By fusing these relationships, a sequence can be generated to verify whether the polarity data conforms to the network's physical layout, which helps identify inconsistencies in the topology. The consistency attributes of the topology verification sequence are filtered through service backup paths to correct redundant deviations in the deviation sequence and determine the dynamic adjustment sequence for backup paths. Service backup paths refer to pre-set backup power supply lines in the distribution network, such as switching to backup paths when the main path fails. Here, the consistency attribute check checks whether all polarity features in the sequence match the expected backup path. If redundant deviations, such as repeated offset points, are found, they are corrected through a filtering algorithm, such as calculating the average deviation and removing outliers, ultimately forming a dynamic adjustment sequence. This sequence records how paths are adjusted based on real-time data to minimize interruptions. The process of dynamically adjusting the sequence according to the backup path and generating the deviation correction sequence involves applying the adjusted sequence to the preliminary results. In actual operation, if the dynamic sequence indicates an offset of 200 meters, the correction sequence will update the boundary of the fault section and generate more accurate positioning data, thereby achieving the technical effect of rapid power restoration in the power distribution automation system.

[0043] S6 uses information fusion technology to integrate multi-source verification results, process the reliability of different data sources, output the verified precise location of the fault, and determine the final location result.

[0044] The initial fault location is obtained from the multi-source verification results using a deviation correction sequence. Information fusion technology is then used to integrate the initial fault location to obtain a preliminary fused location. For this preliminary fused location, the reliability of the data sources is assessed, and a reliability weight allocation is determined. Different data sources are processed using this reliability weight allocation to obtain a weighted fused location. Location deviation correction is applied, and the verification data is integrated. If the deviation exceeds a preset threshold, correction is performed to obtain the corrected location. From the corrected location, a source credibility assessment is obtained. This source credibility assessment and reliability weight allocation are processed to determine the positioning accuracy verification, resulting in the verified location. The verified accurate fault location is then output, determining the final positioning result.

[0045] In one specific embodiment, the process of obtaining the initial fault location from multi-source verification results through a deviation correction sequence can be understood as first collecting preliminary data from different monitoring devices, such as fault signals provided by sensor networks and historical data databases. These signals may include the location coordinates of abnormal voltage fluctuations or current interruptions. When integrating these initial fault locations using information fusion technology, data from multiple sources are cross-validated. For example, location information from smart meters and drone inspections is fused, and isolated errors are eliminated through averaging or voting mechanisms to obtain a preliminary fused location. This helps to initially pinpoint the fault area in a complex power grid environment, thus providing a reliable foundation for subsequent steps.

[0046] For the initial fusion location acquisition data source reliability step, it is necessary to evaluate the stability and historical accuracy of each data source. Sensor data reliability is scored; if a source's equipment is frequently affected by environmental interference, its reliability score may be low. After determining the reliability weight allocation, these weights are used to process different data sources, assigning greater weight to high-reliability sources. Weighted fusion location is calculated through weighted averaging, significantly improving the robustness of positioning, especially when handling power line faults under variable weather conditions, avoiding low-quality data dominating the results.

[0047] When applying weighted fusion location for location deviation correction, it integrates verification data such as real-time GPS coordinates and historical map information to determine whether the deviation exceeds a preset threshold. If it does, it corrects the deviation through an offset adjustment algorithm to obtain the corrected location. This can correct positioning deviations caused by building obstruction and ensure that the location information is closer to the actual fault point.

[0048] The process of obtaining a source credibility assessment from the corrected location involves further credibility review of the data source. When processing these assessments and assigning reliability weights, a comprehensive model such as a Bayesian inference framework is used to determine the location accuracy verification, resulting in a verified location. This step effectively filters noisy data and improves the overall system's reliability. When outputting the verified fault location, all the aforementioned steps are combined to generate the final location result. The final result may be accurate down to the specific tower number. This not only facilitates rapid response by the maintenance team but also plays a role in preventative maintenance, reducing power outage time and improving grid reliability.

[0049] S7. Update the typical cases in the fault feature database with the final location results, and optimize the evaluation framework for the parameter generation of the confidence assessment model.

[0050] Typical case data is obtained from the fault feature database to determine the updated typical case set. Parameters of the confidence assessment model are adjusted, where the confidence assessment model uses a logistic regression model. The input is the feature vector from the typical case set, and the output is the adjusted parameter values, resulting in optimized parameter values. Based on the optimized parameter values, the output of the confidence assessment algorithm is obtained. This algorithm calculates probability scores using the logistic regression model to determine the confidence level of anomaly pattern recognition, where anomaly pattern recognition extracts fault feature patterns from the typical case set. If the confidence level exceeds a preset threshold, verification criteria are extracted from the case matching rules. The case matching rules are pre-established as a correspondence table based on similarity comparison, resulting in a draft framework structure. For the draft framework structure, an optimization iteration loop and result application extension are incorporated. The optimization iteration loop adjusts the structure by repeatedly comparing the draft with the typical case set, and the result application extension adds verification paths for anomaly pattern recognition, generating an optimized assessment framework.

[0051] In one specific embodiment, based on the determined fault location, relevant typical case data is retrieved from the fault feature database. This data includes historical records of similar fracture events and environmental factors such as wind speed and temperature. The system compares this data with the current fault result, filters out cases with high matching degrees, and forms an updated set of typical cases, thus providing a richer reference basis for subsequent analysis. This updated set of typical cases is then used to adjust the parameters of the confidence assessment model. Here, the confidence assessment model is based on a logistic regression model, which maps the input feature vector to probability values. Logistic regression uses the sigmoid function to convert linear combinations into probabilities between 0 and 1. Model parameters such as weights and biases are iteratively optimized using gradient descent. The input feature vector may include fault duration, voltage fluctuation amplitude, etc. After extracting these vectors from the set of typical cases, the model adjusts the parameters according to the principle of minimizing the loss function. For example, for a conductor overload case, the initial parameters may lead to a lower probability. Through multiple iterations, the parameter values ​​are optimized to a more accurate weight combination, and the adjusted parameter values ​​are output, thereby improving the model's sensitivity to similar faults.

[0052] Based on these optimized parameter values, the confidence assessment algorithm calculates probability scores. Specifically, it first extracts fault feature patterns from a set of typical cases, such as abnormal voltage peak patterns, and then inputs them into a logistic regression model. The model uses the optimized parameters to calculate the output of the sigmoid function, i.e., the probability score, which is used to determine the confidence level of abnormal pattern recognition. For a short-circuit fault in a substation, the algorithm extracts feature patterns such as current surge curves. If the calculated probability score is 0.85, it indicates high confidence, which helps confirm the reliability of the fault pattern and avoid maintenance delays caused by misjudgments. If this confidence level exceeds a preset threshold, verification criteria are extracted from the case matching rules. The case matching rules are a pre-established correspondence table based on similarity comparison. The principle is to calculate the matching degree between different cases using cosine similarity or Euclidean distance. The correspondence table lists the mapping between various fault types and verification criteria.

[0053] The draft framework incorporates optimization iteration loops and result application extensions. The optimization iteration loop adjusts the structure by repeatedly comparing the draft with a set of typical cases, comparing the deviation between the output and the actual results, and modifying it if the deviation is large. After several rounds of iteration, it becomes more perfect.

[0054] In another aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0055] In another aspect, the present invention also proposes a computer program product, comprising a computer program, characterized in that the computer program implements the above-described method when executed by a processor.

[0056] In particular, according to some embodiments of this disclosure, the processes described above can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing methods. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processing device, it performs the functions defined in the methods of some embodiments of this disclosure.

[0057] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a task data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated task data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0058] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol, such as HTTP (Hypertext Transfer Protocol), and can interconnect with digital task data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or future-developed networks.

[0059] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: determine the network connection status of the switch production line management application in response to detecting a query operation on a production collaboration document in the switch production line management application; replace the webpage entry information corresponding to the production collaboration document with target entry file information and load target webpage resource information in response to determining that the network connection status of the switch production line management application indicates an offline state, so as to display the webpage of the production collaboration document offline in the switch production line management application, wherein the target entry file information is the file information of the entry file corresponding to the webpage of the production collaboration document downloaded in advance, and the target webpage resource information is the resource information corresponding to the webpage stored locally; in response to determining that the network connection status of the switch production line management application indicates an online state and that the webpage resource information corresponding to the production collaboration document is not stored locally, download the webpage resource information of the webpage from the production line document server, wherein the webpage resource information includes an entry file and resource information; display the webpage of the production collaboration document in the switch production line management application according to the webpage resource information, and store the webpage resource information in a local database.

[0060] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including product-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0061] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automatic fault diagnosis and location in a 10kV distribution network, characterized in that, include: S1. Collect current direction and voltage polarity data to reflect line load changes and short-circuit signal characteristics, establish a confidence assessment model, determine the confidence weight of each polarity feature, and obtain a preliminary weight distribution. S2, combined with the pre-established fault feature database of polarity patterns of historical faults in the distribution network, data cross-validation is performed, the polarity discrimination confidence level of each node is adjusted, and the adjusted confidence sequence is obtained. S3. If there are monitoring points affected by electromagnetic interference in the adjusted confidence sequence, the weight of the monitoring point data is reduced, a weighted polarity sequence is generated by comprehensively considering the interference intensity factor, and the sequence pattern is matched and compared with the historical polarity pattern in the fault feature database to determine the preliminary fault section location. S4. Extract switch action information and user repair location data from the power distribution automation system. By constructing a multi-dimensional verification matrix, compare the correlation between time and space dimensions to obtain spatial deviation index, calculate the location offset, and determine the deviation quantification score. S5. If the deviation quantification score exceeds the preset threshold, the deep analysis unit is activated to re-evaluate the effectiveness of the polarity characteristics of each monitoring point, and the distribution network topology connection relationship is integrated with the data filtering mechanism. The consistency attribute is filtered through the business backup path to correct the redundant deviation of the deviation sequence and obtain the deviation correction sequence. S6 uses information fusion technology to integrate multi-source verification results, process the reliability of different data sources, output the verified precise location of the fault, and determine the final location result; S7. Update the typical cases in the fault feature database with the final location results, and optimize the evaluation framework for the parameter generation of the confidence assessment model.

2. The automatic fault diagnosis and location method for a 10kV distribution network as described in claim 1, characterized in that, In step S1, the current direction and voltage polarity data are acquired and preliminarily classified by line load changes and short-circuit signal characteristics. The signal characteristic analysis results of each data point are determined, and a confidence evaluation model is established. By weighted summation of polarity feature vectors, the confidence weight of each polarity feature is obtained from the confidence evaluation model, and a preliminary weight distribution is obtained. Outliers in the initial weight distribution are obtained. If the outliers exceed a preset threshold as deviations from the average, the weights are adjusted through power distribution network monitoring to determine the corrected weight distribution. Fault location auxiliary information is extracted from the corrected weight distribution, and the matching degree between the information and short-circuit signal features is determined by similarity comparison to obtain the final weight distribution.

3. The automatic fault diagnosis and location method for a 10kV distribution network as described in claim 1, characterized in that, In step S3, by obtaining the adjusted confidence sequence, the monitoring points affected by electromagnetic interference are identified, the weight of the monitoring point data is reduced, and the interference intensity factor is comprehensively considered. The interference intensity factor is obtained by weighted summation of the monitoring point signal attenuation amplitude and the interference duration to generate a weighted polarity sequence and determine the sequence polarity characteristics. By combining historical polarity patterns in the fault feature database, sequence pattern matching and comparison are performed to determine the preliminary fault section location. Interference source localization verification is then performed, which involves fusing signal source data from multiple monitoring points using triangulation and fusing section boundary data to obtain the accurate section boundary.

4. The automatic fault diagnosis and location method for a 10kV distribution network as described in claim 1, characterized in that, In step S4, a multi-dimensional verification matrix is ​​constructed using the preliminary fault section results, switch action information, and user repair geographical location data. The time dimension is then compared and correlated to obtain a time consistency index. For the time consistency index, the spatial deviation index is calculated by comparing the geographic location data and the preliminary fault section results in the spatial dimension. The spatial deviation index is obtained by comparing the difference in geographic coordinates. Historical fault record data is obtained and the location offset is calculated. The location offset is obtained by the vector difference between historical data and the current deviation index. The deviation quantification score is obtained by weighted averaging the offset and historical records.

5. The automatic fault diagnosis and location method for a 10kV distribution network as described in claim 1, characterized in that, In step S6, the initial fault location is obtained from the multi-source verification results using the deviation correction sequence. Information fusion technology is then used to integrate the initial fault locations to obtain a preliminary fused location. For the preliminary fused location, the reliability of the data sources is assessed, and reliability weight allocation is determined. Different data sources are processed using reliability weight allocation to obtain a weighted fused location. Location deviation correction is applied, and verification data is integrated. If the deviation exceeds a preset threshold, correction is performed to obtain the corrected location. From the corrected location, a source credibility assessment is obtained. The source credibility assessment and reliability weight allocation are processed to determine the positioning accuracy verification, obtain the verified location, output the verified fault location, and determine the final positioning result.

6. The automatic fault diagnosis and location method for a 10kV distribution network as described in claim 1, characterized in that, In step S7, Typical case data are obtained from the fault feature database to determine the updated set of typical cases. The parameters of the confidence assessment model are adjusted. The confidence assessment model adopts the logistic regression model. The input is the feature vector in the set of typical cases, and the output is the adjusted parameter value to obtain the optimized parameter value. Based on the optimized parameter values, the output of the confidence assessment algorithm is obtained. The confidence assessment algorithm calculates the probability score through the logistic regression model to determine the confidence of the anomaly pattern recognition. The anomaly pattern recognition extracts fault feature patterns from a set of typical cases. If the confidence level exceeds a preset threshold, the verification criteria are extracted from the case matching rules, wherein the case matching rules are pre-established as a correspondence table based on similarity comparison to obtain a draft framework structure. The proposed framework is adapted by incorporating an optimization iteration loop and an application extension of the results. The optimization iteration loop adjusts the structure by repeatedly comparing the draft with a set of typical cases, while the application extension of the results adds a verification path for anomaly pattern recognition, thereby generating an optimized evaluation framework.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.