A method and system for operation and maintenance management of power distribution systems based on data fusion analysis technology

Abstract

This application discloses a power distribution system operation and maintenance management method based on data fusion analysis technology, relating to the field of operation and maintenance management. The method includes: acquiring a campus information distribution map and historical electricity consumption information; extracting campus electrical components; constructing a campus line topology map; selecting several key campus electrical components; marking key feeder terminal devices; acquiring key terminal information of key feeder terminal devices; locating faults in the campus line topology map based on a line fault judgment matrix to obtain several first fault sections; extracting several fault topology sub-graphs; locating faults in all fault topology sub-graphs based on a fault section location model to obtain several second fault sections; and using all second fault sections to correct all first fault sections to obtain the line fault sections. This application can effectively achieve accurate fault location of power distribution lines.

Description

Technical Field

[0001] This application relates to the field of operation and maintenance management, and in particular to a method and system for operation and maintenance management of power distribution systems based on data fusion analysis technology. Background Technology

[0002] With the development of the social economy and the deepening of the electricity market, the number of electricity users has exploded, placing increasingly higher demands on the stability of the power distribution system. Therefore, if a fault occurs in the power distribution system and cannot be located and repaired in a timely manner, it will severely impact various aspects of industrial production and residential life. However, as electricity demand continues to grow, the network scale of the power distribution system is gradually expanding and its structure is becoming increasingly complex, making fault location more difficult and time-consuming.

[0003] Traditional power distribution system operation and maintenance management methods primarily rely on manual inspection and troubleshooting of electrical equipment. However, due to the complex structure and large number of electrical devices in power distribution systems, manual inspection not only wastes manpower and resources but is also prone to oversights. Failure to promptly identify, repair, or replace faulty equipment may lead to errors in the power status data uploaded by the power system equipment, making it difficult to locate faults in the power system based on the power status data. Furthermore, although traditional power distribution system operation and maintenance management methods can collect and monitor power status data in real time through intelligent equipment, the analysis of this data still relies on manual analysis. Therefore, when a fault occurs in the power system, it is difficult to quickly and accurately locate the fault, which may cause serious social impacts. Summary of the Invention

[0004] This application provides a power distribution system operation and maintenance management method and system based on data fusion analysis technology, which is used to solve the problem that existing technologies are unable to quickly and accurately locate faults in power distribution systems.

[0005] To achieve the above objectives, the embodiments of this application adopt the following technical solutions:

[0006] Firstly, a method for operation and maintenance management of power distribution systems based on data fusion analysis technology is provided, the method comprising:

[0007] Obtain a campus information distribution map and historical electricity consumption information for the target campus, wherein the campus information distribution map includes a campus power line distribution map;

[0008] Extract all campus electrical components from the campus power distribution lines of the target campus from the campus power distribution map;

[0009] A line topology analysis is performed on the campus line distribution map to obtain the device connection relationship between all the campus electrical devices. All the campus electrical devices are taken as line nodes, and a campus line topology map is constructed based on the device connection relationship. The campus electrical devices include feeder terminal devices.

[0010] By combining the historical electricity consumption information and the campus line topology map, the critical coefficients of all the campus electrical devices are analyzed, and based on the critical coefficients of all the campus electrical devices, a number of campus electrical devices are selected as key campus electrical devices.

[0011] If any of the feeder terminal devices has a device connection relationship with any of the key campus electrical devices, then the redundant sensor pre-installed at the feeder terminal device is activated, and the feeder terminal device with the redundant sensor activated is designated as the key feeder terminal device.

[0012] Obtain key terminal information of all the key feeder terminal devices, as well as edge terminal information of all other devices in the feeder terminal devices except the key feeder terminal devices.

[0013] Based on the campus route topology map, a route fault judgment matrix is ​​constructed by integrating the key terminal information and edge terminal information. Based on the route fault judgment matrix, fault location is performed in the campus route topology map to obtain several first fault sections in the campus route topology map.

[0014] Based on all the first fault sections, several fault topology sub-graphs are extracted from the campus line topology map;

[0015] A fault segment location model is constructed based on a graph neural network model. The fault segment location model is used to locate faults in all the fault topology subgraphs to obtain several second fault segments in the campus route topology graph.

[0016] By correcting all the first fault sections using all the second fault sections, several line fault sections of the campus power distribution line are obtained.

[0017] Optionally, the historical electricity consumption information includes historical electricity consumption and historical electricity consumption duration. The campus information distribution map also includes a campus building distribution map. The step of analyzing the critical coefficients of all campus electrical devices by combining the historical electricity consumption information and the campus line topology map, and selecting a number of campus electrical devices as key campus electrical devices based on all the critical coefficients, includes the following steps:

[0018] Based on the campus building distribution map, the target campus is divided into multiple campus electricity consumption areas;

[0019] Based on the historical electricity consumption and the historical electricity consumption duration, the campus electricity consumption areas are divided into regional electricity consumption levels.

[0020] The campus wiring distribution map is matched with the campus building distribution map to obtain the device area locations of all the campus electrical devices;

[0021] For any of the campus electrical devices, an initial device key coefficient is assigned to the campus electrical device according to the power consumption level of the area corresponding to the device's location.

[0022] Calculate the node centrality of the campus electrical devices based on the device connection relationships;

[0023] The initial device critical coefficients are corrected to device critical coefficients using the node centrality.

[0024] Several campus electrical devices whose critical coefficients are greater than or equal to a preset coefficient threshold are marked as critical campus electrical devices.

[0025] Optionally, the step of constructing a line fault judgment matrix based on the campus line topology map, integrating the key terminal information and edge terminal information, and locating faults in the campus line topology map based on the line fault judgment matrix to obtain the first fault segment of the campus line topology map includes the following steps:

[0026] For any of the critical feeder terminal devices, calculate the information difference between the critical terminal information collected by the redundant sensor in the critical feeder terminal device and the critical terminal information collected by other sensors in the critical feeder terminal device other than the redundant sensor, and obtain multiple critical terminal information differences.

[0027] If the absolute value of any of the key terminal information differences is greater than or equal to a preset difference threshold, then it is determined that there is abnormal terminal information in the key terminal information.

[0028] If the absolute value of all the differences in the key terminal information is less than the difference threshold, then it is determined that there is no abnormal terminal information in the key terminal information.

[0029] If abnormal terminal information exists in the key terminal information, then the key terminal information collected by the redundant sensor is selected as the target terminal information.

[0030] The target terminal information and the edge terminal information are fused together to obtain the target feeder terminal information;

[0031] A line fault judgment matrix is ​​constructed by combining the campus line topology map and all the target feeder terminal information. Based on the line fault judgment matrix, several primary fault sections are located in the campus line topology map.

[0032] For any of the primary fault segments, the target terminal information is verified based on the primary fault segment and using an information verification function.

[0033] If the information verification result shows that there are no information errors in the target terminal information, then the primary fault section is determined to be the first fault section of the campus line topology map;

[0034] If the information verification result shows that there is an error in the target terminal information, then the target terminal information is updated according to all other information in the key terminal information except for the target terminal information to obtain the baseline terminal information;

[0035] By combining the baseline terminal information and the edge terminal information, fault location is performed in the campus route topology map to obtain several first fault sections in the campus route topology map.

[0036] Optionally, the step of constructing a line fault judgment matrix by combining the campus line topology map and all the target feeder terminal information, and locating several primary fault sections in the campus line topology map based on the line fault judgment matrix, includes the following steps:

[0037] Based on the target feeder terminal information, obtain the electrical parameters of all the campus electrical devices, and analyze the power direction of all the campus electrical devices by combining the electrical parameters and the topology type of the campus line topology diagram.

[0038] Based on the campus line topology map, a campus line description matrix of the campus power distribution line is constructed by combining the line power direction and the device connection relationship;

[0039] Based on the electrical parameters and using the threshold method to obtain device fault information in the target feeder terminal information, a campus fault information matrix of the campus power distribution line is constructed by combining the target feeder terminal information and the campus line topology map.

[0040] A line fault judgment matrix is ​​constructed by combining the campus line description matrix and the campus fault information matrix;

[0041] Based on the line fault judgment matrix and using preset fault location rules, fault location is performed in the campus line topology map to obtain several primary fault sections of the campus line topology map.

[0042] Optionally, the campus electrical equipment further includes main power distribution equipment, secondary power distribution equipment, and campus electrical equipment. The step of verifying the target terminal information based on the primary fault section and using an information verification function includes the following steps:

[0043] The primary fault section is transformed into a primary fault section matrix based on the line fault judgment matrix.

[0044] The campus power line topology is divided into an upstream section and a downstream section, with the primary fault section as the center point. The upstream section connects to the main power distribution equipment, and the downstream section connects to the campus power equipment.

[0045] Based on the campus line topology map, the power distribution status values ​​of all the secondary power distribution equipment are obtained according to the campus fault information matrix. The power distribution status values ​​include the first power distribution status value of the upstream section and the second power distribution status value of the downstream section.

[0046] Based on the analysis of all the primary fault sections, the status values ​​of all line sections in the campus line topology map are obtained, resulting in multiple section status values, including upstream section status values ​​and downstream section status values.

[0047] The upstream state value and the downstream state value are integrated to obtain the upstream total state value and the downstream total state value, respectively.

[0048] The primary fault segment matrix, the power distribution status value, the upstream segment status value, the downstream segment status value, the upstream total status value, and the downstream total status value are input into the information verification function to obtain the primary fault verification matrix. The information verification function is as follows:

[0049]

[0050] Where i is a matrix element in the primary fault section matrix, x represents the upstream line, y represents the downstream line, and D x D represents the first power distribution state value. y This indicates the second power distribution status value. This represents the upstream total state value. N represents the total downstream state value. x and N y M represents the total number of secondary power distribution equipment in the upstream line x and the downstream line y, respectively. x and M y Q represents the total number of all line segments within the upstream line x and the downstream line y, respectively. i，x Q represents the upstream segment status value of all the upstream segments within the upstream line x.i，y This represents the downstream segment status value of all the line segments within the downstream line y, and Π represents the logical OR operation.

[0051] If the primary fault verification matrix and the primary fault segment matrix are equal, then it is determined that the target terminal information does not contain any errors.

[0052] If the primary fault verification matrix and the primary fault segment matrix are not equal, then the target terminal information is determined to have an information error.

[0053] Optionally, the step of extracting several fault topology sub-graphs from the campus line topology map based on the first fault section includes the following steps:

[0054] Based on the segment location of all the first fault segments in the campus route topology map, the route nodes are marked with faults to obtain multiple first fault route nodes.

[0055] The first faulty line node is aggregated to obtain several faulty line node sets.

[0056] Based on the complete set of faulty line nodes, several faulty topology subgraphs are extracted from the campus line topology map.

[0057] Optionally, the step of constructing a fault segment location model based on a graph neural network model, and using the fault segment location model to locate faults in all the fault topology subgraphs to obtain several second fault segments in the campus route topology map, includes the following steps:

[0058] A fault segment localization model is constructed based on a graph neural network model. The fault segment localization model includes a fault feature extraction module, a fault feature weighting module, a feature dimension transformation module, and a fault classification learning module. The fault feature extraction module is constructed based on a spatiotemporal graph neural network model, the fault feature weighting module is constructed based on a graph attention network model, the feature dimension transformation module is constructed based on a one-dimensional convolutional neural network model, and the fault classification learning module is constructed based on a fully convolutional neural network model.

[0059] The fault section location model is trained using a pre-built fault location training set, and the fault topology subgraph is sequentially input into the fault section location model for fault location, thereby obtaining several second fault sections in the campus power distribution line.

[0060] Optionally, inputting the fault topology subgraph sequentially into the fault section location model for fault location to obtain several second fault sections in the campus power distribution line includes the following steps:

[0061] The fault topology subgraph is input into the fault feature extraction module of the fault segment localization model to perform a convolution operation on the fault topology subgraph on a spatiotemporal scale, thereby obtaining the fault spatiotemporal features of the fault topology subgraph.

[0062] The spatiotemporal features of the fault are sequentially input into the fault feature weighting module to calculate the attention coefficient, and weights are assigned to the spatiotemporal features of the fault based on the attention coefficient calculation results.

[0063] The fault spatiotemporal features that have been weighted are weighted and fused to obtain fault fusion features;

[0064] The fault fusion features are input into the feature dimension transformation module for feature mapping, and the fault fusion features after feature mapping are input into the fault classification learning module for classification learning to obtain several second fault sections in the campus line topology map.

[0065] Optionally, the step of correcting all the first fault sections using all the second fault sections to obtain several line fault sections of the campus power distribution line includes the following steps:

[0066] Based on the location of the second faulty section in the campus route topology map, the route node is marked as faulty to obtain the second faulty route node;

[0067] Based on the campus route topology diagram, analyze whether the second faulty route node overlaps with the first faulty route node;

[0068] If the second faulty line node overlaps with the first faulty line node, then the second faulty section corresponding to the overlapping second faulty line node is marked as a topology line faulty section.

[0069] All the faulty sections of the aforementioned topology lines are matched with the campus line distribution map in terms of location, and several faulty sections in the campus line distribution map are located based on the location matching structure.

[0070] Secondly, this application provides a power distribution system operation and maintenance management system based on data fusion analysis technology, characterized in that it includes:

[0071] The memory is configured to store instructions; and

[0072] A processor is configured to retrieve the instructions from the memory and, when executing the instructions, to implement the method for operation and maintenance management of a power distribution system based on data fusion analysis technology according to any one of the first aspects.

[0073] Through the above technical solution, this application utilizes historical electricity consumption information and campus building distribution maps to screen key campus electrical components. Feeder terminal devices connected to these key components are marked as key feeder terminal devices, and pre-installed redundant sensors in these devices are activated. These redundant sensors perform self-tests on the key feeder terminal devices, effectively detecting the presence of faulty equipment. This ensures the reliability of key terminal information while also enabling the detection of the key feeder terminal devices. A matrix method is used to first locate faults in the power distribution system lines, obtaining a first fault section. Based on this first fault section, a graph neural network model is used to refine it, resulting in a more accurate final fault section. In summary, this invention not only solves the problem of rapid and accurate fault location in power distribution systems using existing technologies, but also reduces the waste of power resources and improves the utilization rate of power resources to a certain extent.

[0074] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0075] Figure 1 A flowchart illustrating a power distribution system operation and maintenance management method based on data fusion analysis technology, provided as an embodiment of this application;

[0076] Figure 2 This is an example diagram of a campus route topology provided in an embodiment of this application. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustration and explanation of the embodiments of this application and are not intended to limit the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0078] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0079] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0080] Figure 1 This illustration schematically depicts a flowchart of a method for operation and maintenance management of a power distribution system based on data fusion analysis technology, according to an embodiment of this application. Figure 1 As shown in the figure, this application provides a method for operation and maintenance management of a power distribution system based on data fusion analysis technology. This method may include the following steps:

[0081] S101. Obtain the campus information distribution map and historical electricity consumption information of the target campus. The campus information distribution map includes the campus power line distribution map.

[0082] In this embodiment, the campus information distribution map includes a campus wiring distribution map and a campus building distribution map. The campus wiring distribution map refers to the power distribution network plan of the target campus, which includes multiple feeder terminal units, main power distribution equipment, secondary power distribution equipment, and campus electrical equipment, as well as the connecting lines between these electrical devices. The feeder terminal unit (FTU) has remote control, telemetry, remote signaling, and fault detection functions, and can provide information on the operation status of the power distribution system and various parameters, as well as information required for monitoring and control, including switch status, power parameters, phase-to-phase faults, grounding faults, and parameters during faults. The main power distribution equipment refers to the main distribution box, which is used to connect the branch circuits to the main line to realize power distribution and control. The secondary power distribution equipment includes primary distribution boxes, secondary distribution boxes, and tertiary distribution boxes, which are responsible for distributing the power output from the main distribution box to various levels of campus electrical equipment, such as computers, air conditioners, and experimental equipment (e.g., electric thermostats, mass spectrometers, nuclear magnetic resonance spectrometers). A campus building distribution map is a floor plan of the target campus, including the location of all buildings, such as teaching buildings, laboratories, and canteens. Historical electricity consumption information includes the historical electricity consumption and duration of each teaching building and its internal classrooms, offices, laboratories, etc. Historical electricity consumption is the daily average, and historical duration is the daily average duration. For example, classroom 306 in the first teaching building has an average daily electricity consumption of 7.52 kWh and an average daily duration of 12 hours.

[0083] S102. Extract all campus electrical components from the campus power distribution lines of the target campus from the campus power distribution map.

[0084] In this embodiment, campus electrical components such as circuit breakers, sectionalizing switches, tie switches, and power distribution unit (FTU) monitoring terminals are extracted from the campus wiring diagram. The extraction of electrical components can be performed using drawing analysis software, such as EasyEDA (JLCIC EDA) or Eplan (electrical design software).

[0085] S103. Perform a line topology analysis on the campus line distribution map to obtain the device connection relationship between all campus electrical devices. Take all campus electrical devices as line nodes and construct a campus line topology map based on the device connection relationship. Campus electrical devices include feeder terminal devices.

[0086] Algorithms such as tree search, shortest path algorithm, and node elimination are used to find the path between any two campus electrical components. Specifically, taking Dijkstra's algorithm in the shortest path algorithm as an example, a campus electrical component is randomly selected from the campus wiring map as the starting node. Two sets are initialized: a first set containing the shortest path node and a second set containing undetermined shortest path nodes. Initially, the first set only contains the starting node. All campus electrical components except the starting node are added to the second set as undetermined shortest path nodes. A distance array is initialized to record the estimated shortest path from the starting node to each node. Initially, the distances of all nodes except the starting node are set to infinity. The undetermined shortest path node A closest to the starting node is selected from the second set and added to the first set. The estimated distances from node A to all nodes in the second set are then updated. If the distance to a node via node A (with an undetermined shortest path) is shorter than the known shortest distance, then update the distance estimate for that node and record the path information. Repeat the steps of selecting nodes with undetermined shortest paths and updating distance estimates until the second set is empty, meaning all nodes have found their shortest paths. By recording the nodes directly connected to each node on the shortest path, a shortest path tree from the starting node to all other nodes can be constructed, thereby identifying the device connection relationships between campus electrical components in the campus route distribution map. By treating all campus electrical components in the campus route distribution map as route nodes and the device connection relationships as edges, a campus route topology map based on the campus route distribution map can be obtained.

[0087] Simultaneously, each campus electrical device can be numbered based on its distance from the main power source in the distribution network. Examples include circuit breaker #1, circuit breaker #2, sectionalizing switch #3, sectionalizing switch #4, sectionalizing switch #5, etc. Methods for calculating the distance between each campus electrical device and the main power source include Hadoop (distributed system infrastructure), breadth-first search, and shortest path algorithms. Taking breadth-first search as an example, the main power source is used as the source node. Starting from the source node, the search expands outwards layer by layer, simultaneously calculating the number of edges between the source node and other nodes in the campus network topology. This number is used to measure the distance between each campus electrical device and the main power source. If multiple campus electrical devices have the same number of edges (i.e., distance) to the main power source, they are randomly numbered. The purpose of assigning numbers is to distinguish different campus electrical devices, facilitating rapid fault location during subsequent fault diagnosis.

[0088] By constructing a campus circuit topology map, the complex relationships between campus electrical components in the campus circuit distribution map can be simplified into a topology map, which facilitates the subsequent analysis of the importance of each campus electrical component. At the same time, by identifying the connection relationships between campus electrical components, the propagation path of faults can be determined. When a campus electrical component fails, the area affected by the fault can be identified by tracing the path of current flow, which helps to quickly locate the source of the fault.

[0089] S104. Based on historical electricity consumption information and campus line topology analysis, identify the critical coefficients of all campus electrical devices, and select several campus electrical devices as key campus electrical devices based on the critical coefficients.

[0090] In this embodiment, the campus building distribution map refers to a floor plan of the target campus, which includes the location of all buildings on the target campus, such as teaching buildings, laboratories, canteens, etc. Based on the campus building distribution map, the target campus is divided into multiple power consumption areas. Specifically, each building is divided into one power consumption area; for example, the first teaching building is power consumption area 1, and the canteen is power consumption area 2.

[0091] Based on the historical electricity consumption and duration of each power consumption area, the area is classified into regional power consumption levels. After the level classification of each power consumption area is completed, the campus power distribution map is matched with the campus building distribution map to obtain the corresponding location of each campus electrical device in the campus building distribution map. For example, secondary distribution equipment No. 16 is located in power consumption area No. 6. Specifically, the coordinate system and size of the campus power distribution map and the campus building distribution map are unified. The coordinates of each campus electrical device are matched with the coordinate range of the campus power consumption area in the campus building distribution map. If the coordinates of the campus electrical device are within the coordinate range of the power consumption area, the corresponding campus power consumption area of ​​the campus electrical device in the campus building distribution map is obtained, thus obtaining the device area location of the campus electrical device. Since it is only necessary to use the regional power consumption level of the campus power consumption area as the initial device key coefficient for the allocation of campus electrical devices in the campus power consumption area, it is not necessary to obtain the precise location of the campus electrical devices in the campus building distribution map. It is only necessary to determine the campus power consumption area where each campus electrical device is located in the campus building distribution map.

[0092] After obtaining the power consumption levels of all campus electrical devices, initial criticality coefficients are assigned to each device based on its corresponding regional power consumption level. The node centrality of each device is calculated based on its connection relationships with other nodes in the campus network topology. Node centrality includes degree centrality, eigenvector centrality, intermediateness centrality, and compactness centrality. Degree centrality depends on the node's degree. A weighted average of degree centrality, eigenvector centrality, intermediateness centrality, and compactness centrality is calculated to obtain average centrality. The initial criticality coefficients are then adjusted based on the average centrality of each device to obtain the criticality coefficient for each device. After obtaining the criticality coefficients for all campus electrical devices, those with criticality coefficients greater than or equal to a preset threshold are marked as critical campus electrical devices.

[0093] S105. If there is a device connection relationship between any feeder terminal device and any critical campus electrical device, then the redundant sensor pre-installed at the feeder terminal device shall be activated, and the feeder terminal device with the redundant sensor activated shall be designated as the critical feeder terminal device.

[0094] In this embodiment, if there is a device connection between the feeder terminal device and critical campus electrical components, it indicates that the information monitored by the feeder terminal device originates from the critical campus electrical components. Therefore, if a sensor inside the feeder terminal device malfunctions, it may cause errors in the critical terminal information monitored by the feeder terminal device. Subsequent fault location results obtained based on this erroneous critical terminal information may also be incorrect. Therefore, it is necessary to monitor whether the feeder terminal devices connected to the critical campus electrical components are faulty. Specifically, a set of redundant sensors is pre-installed in each feeder terminal device. These redundant sensors are almost identical to the original basic sensors inside the feeder terminal device in terms of sensor type, sensor model, and sensor performance. If a connection is detected between the feeder terminal and critical campus electrical components, the redundant sensors inside the feeder terminal are immediately activated, and the feeder terminal is marked as a critical feeder terminal. During the subsequent monitoring of the power distribution system by the critical feeder terminal, two sets of critical terminal information of the same type but with potentially different data will be obtained. For example, both sets of critical terminal information may include information such as phase current, phase voltage, active power, and reactive power, but the specific values ​​of these information may be different, or even significantly different, but the information type is the same. Based on the information difference between the above-mentioned information of the same type, it is possible to detect whether there is a device fault in the critical feeder terminal.

[0095] S106. Obtain key terminal information of all key feeder terminal devices, as well as edge terminal information of all other devices in the feeder terminal devices except for the key feeder terminal devices.

[0096] In this embodiment, key terminal information of all key feeder terminal devices is obtained. Key terminal information includes information such as phase current, phase voltage, active power and reactive power, as well as edge terminal information of other feeder terminal devices that are not connected to key campus electrical devices. Edge terminal information also includes information such as phase current, phase voltage, active power and reactive power.

[0097] S107. Based on the campus line topology map, a line fault judgment matrix is ​​constructed by integrating key terminal information and edge terminal information. Based on the line fault judgment matrix, fault location is performed in the campus line topology map to obtain several first fault sections in the campus line topology map.

[0098] In this embodiment, for any critical feeder terminal device, the information difference between the critical terminal information collected by the redundant sensors inside the critical feeder terminal device and the critical terminal information collected by other sensors in the critical feeder terminal device (excluding the redundant sensors) is calculated, resulting in multiple critical terminal information differences. Based on these differences, it is determined whether there is abnormal terminal information in the critical terminal information. If abnormal terminal information is found, the critical terminal information collected by the redundant sensors is selected as the target terminal information. This is because redundant sensors are used less frequently and have a lower probability of failure; therefore, the critical terminal information collected by the redundant sensors is used as the target terminal information first. Of course, a faulty sensor may also be a redundant sensor; therefore, an information verification function will be used subsequently to verify the target terminal information collected by the redundant sensors.

[0099] The target terminal information and edge terminal information are merged to obtain complete feeder terminal information of the power distribution system, namely the target feeder terminal information. Then, based on the campus line topology map, a campus line description matrix and a campus fault information matrix are constructed according to the target feeder terminal information. The matrix product of the campus line description matrix and the campus fault information matrix is ​​normalized to obtain the line fault judgment matrix. The line fault judgment matrix is ​​used to locate faults in the campus line topology map to obtain several primary fault sections located in the campus line topology map. The primary fault sections are verified using an information verification function. If the verification result shows that there are no information errors in the target terminal information, the primary fault section can be directly determined as the first fault section in the campus line topology map.

[0100] S108. Extract several fault topology sub-graphs from the campus line topology map based on all first fault sections.

[0101] In this embodiment, based on the first fault section located in the campus route topology map, the route nodes located in the first fault section are marked as faults to obtain multiple first fault route nodes. Adjacent fault nodes are aggregated to obtain several fault route node sets. Commonly used node aggregation methods include DBSCAN (density-based clustering algorithm) and spectral clustering algorithm. Based on all the several fault route node sets, several fault topology subgraphs are extracted from the campus route topology map.

[0102] S109. Construct a fault section location model based on a graph neural network model, and use the fault section location model to locate faults in all fault topology subgraphs to obtain several second fault sections in the campus line topology map.

[0103] In this embodiment, a fault segment localization model is constructed based on a graph neural network model. The fault segment localization model includes a fault feature extraction module, a fault feature weighting module, a feature dimension transformation module, and a fault classification learning module. The fault feature extraction module is based on a spatiotemporal graph neural network model; the fault feature weighting module is based on a graph attention network model; the feature dimension transformation module is based on a one-dimensional convolutional neural network model; and the fault classification learning module is based on a fully convolutional neural network model. The fault feature extraction module captures the spatial correlation between nodes and the correlation between fault topology subgraphs at different times. The fault feature weighting module assigns weights to fault features of different dimensions based on the concept of learning attention weights and performs feature fusion on the weighted fault features. The feature dimension transformation module extracts and transforms the dimensions of the fault features. The fault classification learning module maps the extracted high-dimensional fault features to a low-dimensional space containing fault segment information for classification learning. Finally, the maximum activation value (argmax) obtained by feature classification using the Softmax function is the second fault segment.

[0104] S110. By using all the second fault sections to correct all the first fault sections, several line fault sections of the campus power distribution line are obtained.

[0105] In this embodiment, the line nodes are marked as faults according to the location of the second fault section in the campus line topology map to obtain the second fault line node. The second fault line node is analyzed according to the campus line topology map to see if it overlaps with the first fault line node. If the second fault line node overlaps with the first fault line node, the second fault section corresponding to the overlapping second fault line node is marked as the topology line fault section. All topology line fault sections are matched with the campus line distribution map to locate several line fault sections in the campus line distribution map according to the location matching results.

[0106] In one embodiment, historical electricity consumption information includes historical electricity consumption and historical electricity consumption duration. The campus information distribution map also includes a campus building distribution map. Combining historical electricity consumption information and the campus line topology map, the critical coefficients of all campus electrical devices are analyzed. Based on the critical coefficients of all devices, a number of campus electrical devices are selected as key campus electrical devices, including the following steps:

[0107] Based on the campus building distribution map, the target campus is divided into multiple campus electricity consumption areas;

[0108] Based on historical electricity consumption and historical electricity usage duration, the entire campus electricity consumption area is divided into regional electricity consumption levels;

[0109] By matching the campus wiring map with the campus building map, the location of all electrical components on campus can be obtained.

[0110] For any campus electrical device, an initial device criticality coefficient is assigned to the campus electrical device according to the power consumption level of the area corresponding to the device's location.

[0111] Calculate the node centrality of campus electrical components based on their connection relationships;

[0112] The initial device criticality coefficients are corrected to device criticality coefficients using node centrality.

[0113] Several campus electrical devices whose critical coefficients are greater than or equal to a preset threshold are marked as critical campus electrical devices.

[0114] In this embodiment, the campus building distribution map refers to a floor plan of the target campus, which includes the location of all buildings on the target campus, such as teaching buildings, laboratories, canteens, etc. Based on the campus building distribution map, the target campus is divided into multiple power consumption areas. Specifically, each building is divided into one power consumption area; for example, the first teaching building is power consumption area 1, and the canteen is power consumption area 2.

[0115] Based on the historical electricity consumption and historical electricity usage duration of each electricity consumption area, each area is classified into different electricity consumption levels. Specifically, areas with historical electricity consumption exceeding a preset first electricity consumption threshold and historical electricity usage duration exceeding a preset first electricity usage duration threshold are classified as Level 1 electricity consumption areas; areas with historical electricity consumption exceeding the first electricity consumption threshold or historical electricity usage duration exceeding the first electricity usage duration threshold are classified as Level 2 electricity consumption areas; and areas with historical electricity consumption less than or equal to the first electricity consumption threshold but greater than a preset second electricity consumption threshold, and historical electricity usage duration less than the first electricity usage duration threshold, are classified as Level 3 electricity consumption areas. Electricity consumption areas with a historical power consumption threshold greater than a preset second power consumption duration threshold are classified into three-level electricity consumption areas. Electricity consumption areas with historical power consumption less than or equal to a first power consumption threshold but greater than a second power consumption threshold, or historical power consumption duration less than a first power consumption duration threshold but greater than a second power consumption duration threshold, are classified into four-level electricity consumption areas. Electricity consumption areas with historical power consumption less than a second power consumption threshold and historical power consumption duration less than a second power consumption duration threshold are classified into five-level electricity consumption areas. Among these, the first power consumption threshold is greater than the second power consumption threshold, and the first power consumption duration threshold is greater than the second power consumption duration threshold.

[0116] After classifying each power consumption area, the campus wiring map and the campus building map are matched to determine the location of each electrical device in the campus building map. For example, secondary distribution equipment number 16 is located in power consumption area number 6. Specifically, the coordinate system and dimensions of the campus wiring map and the campus building map are unified. The coordinates of each electrical device are matched with the coordinate range of the campus power consumption area in the campus building map. If the coordinates of the electrical device are within the coordinate range of the power consumption area, the corresponding campus power consumption area in the campus building map is obtained, thus determining the device's location. Since only the regional power consumption level of the campus power consumption area is needed as the initial key coefficient for allocating the electrical devices within that area, it is not necessary to obtain the precise location of the electrical devices in the campus building map; only the campus power consumption area in the campus building map needs to be determined.

[0117] After obtaining the corresponding levels of all campus electrical devices, an initial device critical coefficient is assigned to each campus electrical device according to the regional power consumption level corresponding to each campus electrical device. Among them, the initial device critical coefficient assigned to campus electrical devices located in the first-level power consumption area is the largest, followed by the second-level, third-level, and fourth-level power consumption areas, and the initial device critical coefficient assigned to the fifth-level power consumption area is the smallest. For example, the initial device critical coefficient assigned to campus electrical devices located in the first-level power consumption area is 1, that of campus electrical devices in the second-level power consumption area is 0.8, that of campus electrical devices in the third-level power consumption area is 0.6, that of campus electrical devices in the fourth-level power consumption area is 0.4, and that of campus electrical devices in the fifth-level power consumption area is 0.2.

[0118] The node centrality of each campus electrical device is calculated based on its connectivity with other network nodes in the campus network topology. Node centrality includes degree centrality, eigenvector centrality, betweenness centrality, and compact centrality. Degree centrality depends on the node's degree, which is the number of edges connecting it to other network nodes. Eigenvector centrality depends on the degree of its neighbors; higher neighbor degrees indicate higher eigenvector centrality and greater importance. Betweenness centrality measures the extent to which a network node lies on the shortest path between other nodes; nodes with high betweenness centrality play a crucial role in information transmission and resource flow, thus indicating greater importance. Compact centrality is based on the sum of the shortest paths from the current network node to all other network nodes; a smaller sum of shortest paths indicates higher compact centrality and greater importance. The degree centrality, eigenvector centrality, betweenness centrality, and compact centrality of all line nodes are calculated. A weighted average of these values ​​is then taken to obtain the average centrality. The initial criticality coefficients of each campus electrical device are adjusted based on the average centrality, resulting in the criticality coefficient for each device. After obtaining the criticality coefficients for all campus electrical devices, those with criticality coefficients greater than or equal to a preset threshold are marked as critical campus electrical devices.

[0119] In one embodiment, a line fault judgment matrix is ​​constructed based on a campus route topology map, integrating key terminal information and edge terminal information. Fault location is then performed on the campus route topology map based on this matrix to obtain the first fault segment of the campus route topology map. This process includes the following steps:

[0120] For any critical feeder terminal device, calculate the information difference between the critical terminal information collected by the redundant sensors in the critical feeder terminal device and the critical terminal information collected by other sensors in the critical feeder terminal device other than the redundant sensors, and obtain multiple critical terminal information differences.

[0121] If the absolute value of any difference in key terminal information is greater than or equal to a preset difference threshold, then it is determined that there is abnormal terminal information in the key terminal information.

[0122] If the absolute value of the difference between all key terminal information is less than the difference threshold, it is determined that there is no abnormal terminal information in the key terminal information.

[0123] If abnormal terminal information exists in the key terminal information, the key terminal information collected by the redundant sensor is selected as the target terminal information.

[0124] The target terminal information and the edge terminal information are fused to obtain the target feeder terminal information;

[0125] A line fault judgment matrix is ​​constructed by combining the campus line topology map and information of all target feeder terminals. Based on the line fault judgment matrix, several primary fault sections are located in the campus line topology map.

[0126] For any primary fault segment, the target terminal information is verified based on the primary fault segment and using the information verification function.

[0127] If the information verification result shows that there are no information errors in the target terminal information, then the primary fault section is determined to be the first fault section in the campus line topology map;

[0128] If the information verification result shows that there is an error in the target terminal information, the target terminal information is updated based on all other information in the key terminal information except for the target terminal information to obtain the baseline terminal information.

[0129] By combining baseline terminal information and edge terminal information, fault location is performed in the campus line topology map to obtain the first fault section of the campus line topology map.

[0130] In this embodiment, for any critical feeder terminal device, the information difference between the critical terminal information collected by the redundant sensors inside the critical feeder terminal device and the critical terminal information collected by other sensors in the critical feeder terminal device (excluding the redundant sensors) is calculated to obtain multiple critical terminal information differences. For example, the critical terminal information collected by the redundant sensors is phase current a1, phase voltage b1, etc., while the critical terminal information collected by other sensors (the basic sensors in the critical feeder terminal device) is phase current a2, phase voltage b2, etc. Then, phase current a1 is subtracted from phase current a2, and phase voltage b1 is subtracted from phase voltage b2. This is because the redundant sensors and the basic sensors are identical in model, type, and performance, and are located in... In the same critical feeder terminal device, theoretically, the critical terminal information collected by the redundant sensor and the basic sensor should be equal or have negligible differences. That is, the absolute value of the difference between all critical terminal information should be less than the difference threshold. However, if the difference between the two collected information is large, that is, the absolute value of the difference between the critical terminal information is greater than or equal to the preset difference threshold, it indicates that there is a sensor fault in the redundant sensor or the basic sensor. Therefore, it is determined that there is abnormal terminal information in the critical terminal information, and the critical feeder terminal device is marked as an abnormal feeder terminal device. The number of the abnormal feeder terminal device is uploaded to the power distribution system management platform to remind the staff of the power distribution system management platform to repair the abnormal feeder terminal device in a timely manner.

[0131] If abnormal terminal information is found in the critical terminal information, the critical terminal information collected by the redundant sensor is selected as the target terminal information first. This is because redundant sensors are used less frequently and have a lower probability of failure. However, the faulty sensor may also be a redundant sensor, so an information verification function will be used to verify the target terminal information collected by the redundant sensor.

[0132] The target terminal information and edge terminal information are merged to obtain complete feeder terminal information of the power distribution system, namely the target feeder terminal information. Then, based on the campus line topology map, a campus line description matrix and a campus fault information matrix are constructed according to the target feeder terminal information. The matrix product of the campus line description matrix and the campus fault information matrix is ​​normalized to obtain the line fault judgment matrix. The line fault judgment matrix is ​​used to locate faults in the campus line topology map to obtain several primary fault sections located in the campus line topology map. The primary fault sections are verified using an information verification function. If the verification result shows that there are no information errors in the target terminal information, the primary fault section can be directly determined as the first fault section of the campus line topology map. If the information verification result shows that there are information errors in the target terminal information, it indicates that the redundant sensor is faulty. Therefore, the target terminal information needs to be replaced with the key terminal information collected by the basic sensor. The target terminal information after the replacement is named the reference terminal information. In other words, the reference terminal information is the key terminal information collected by the sensors other than the redundant sensors in the key feeder terminal device. Next, the same fault segment location steps are performed. The baseline terminal information and edge terminal information are fused to obtain fused terminal information. A line fault judgment matrix is ​​constructed by combining the campus line topology map and all fused terminal information. Based on the line fault judgment matrix, several second primary fault segments are located in the campus line topology map. The baseline terminal information is then verified using the obtained primary fault segments and an information verification function. Similarly, if the verification result shows that there are no errors in the baseline terminal information, the second primary fault segment is output as the first fault segment. If the verification result shows that there are errors in the baseline terminal information, it indicates that the basic sensor in the critical feeder terminal device has also failed. At this time, the number of the critical feeder terminal device needs to be uploaded to the power distribution system management platform again, and an alarm message needs to be output to remind the staff of the power distribution system management platform to go to the location of the abnormal feeder terminal device immediately and repair it immediately.

[0133] In one embodiment, constructing a line fault judgment matrix by combining the campus line topology map and all target feeder terminal information, and locating several primary fault sections in the campus line topology map based on the line fault judgment matrix includes the following steps:

[0134] Based on the target feeder terminal information, obtain the electrical parameters of all campus electrical devices, and analyze the power direction of all campus electrical devices by combining the electrical parameters and the topology type of the campus line topology diagram.

[0135] Based on the campus power line topology map, a campus power distribution line description matrix is ​​constructed by combining the power direction of the lines and the connection relationship of the devices.

[0136] Based on electrical parameters and using the threshold method to obtain device fault information in the target feeder terminal information, a campus fault information matrix of the campus power distribution line is constructed by combining the target feeder terminal information and the campus line topology map.

[0137] A line fault judgment matrix is ​​constructed by combining the campus line description matrix and the campus fault information matrix;

[0138] Based on the line fault judgment matrix and using the preset fault location rules, fault location is performed in the campus line topology map to obtain several primary fault sections in the campus line topology map.

[0139] In this embodiment, the feeder terminal unit (FTU) is an intelligent terminal device specifically designed for power distribution systems. It collects key data such as voltage and current in the power distribution system in real time through built-in sensors and data processing modules. This data undergoes preliminary processing, such as filtering and calibration, to improve data reliability, and is then used to calculate information such as phase difference. Therefore, the electrical parameters of all campus electrical devices can be directly extracted from the target feeder terminal information. These electrical parameters include current, voltage, and phase difference. Based on the topology of the campus circuit topology, the flow direction of current and voltage is analyzed. Based on the flow direction of current and voltage, current, voltage, phase difference, and using the power calculation formula, the power direction can be analyzed. The power calculation formula is P = UIcosθ, where U is voltage, I is current, and θ is the phase difference between voltage and current. If the calculated power is positive, the power is in the positive direction; if the calculated power is negative, the power is in the negative direction. After calculating the power direction of all line nodes in the campus line topology diagram, if the power of a line node is in the positive direction and there is a device connection relationship between it and other line nodes, then the line node Dij = 1 is defined; otherwise, the line node Dij = 0 is defined. After defining all line nodes, the campus line description matrix is ​​obtained.

[0140] When a line node experiences a current exceeding a preset current threshold, it indicates a fault in the power distribution system. After a fault occurs, the power flow direction changes. The FTUs installed at each node will upload three logic signals: "1," "0," and "-1," depending on whether a fault current is detected and whether its direction is the same as the network's forward direction. 1, 0, and -1 represent a line node experiencing a forward fault overcurrent, a line node not experiencing a fault overcurrent, and a line node experiencing a reverse fault overcurrent, respectively. After obtaining the fault status of all line nodes, a campus fault information matrix is ​​constructed based on the fault status. By combining the campus fault information matrix and the campus line description matrix, a line fault judgment matrix is ​​obtained. The line fault judgment matrix is ​​as follows:

[0141]

[0142] Among them, P ij Let D be the line fault judgment matrix between line node i and line node j. ij Let G be the campus route description matrix between route node i and route node j. i1 This is the campus fault information matrix between line node i and line node j.

[0143] Based on the line fault judgment matrix and using the preset fault location rules, the fault location is performed in the campus line topology map to obtain several primary fault sections in the campus line topology map. The preset fault location rules include: (1) For j with Pii=1 and Pij=1, Pjj=0 or -1 (j≠i); (2) For j with Pii=-1 and Pji=1, Pjj=0 (j≠i); (3) For j with Pii=0 and Pij=1, Pjj=-1 (j≠i); If any one of the above three conditions is met, and the section between line node i and line node j is not a T-connection section, then the feeder section formed by line node i and line node j can be determined as a primary fault section. If any one of the above three conditions is met, and the section between line node i and line node j is a T-connection section (for example, line node i is connected to line node j and also to line node f), then it is further determined whether the section between line node i and line node f meets any one of the above three conditions. If it is determined that the section between line node i and line node f also meets any one of the above three conditions, then the T-connection section between line node i and line node j is determined to be a primary fault section.

[0144] Reference Figure 2 In the figure, the arrow direction is the specified positive direction, Z is the main power distribution equipment, S is the campus power equipment; 1-7 represent node numbers, (1)-(7) represent section numbers, and the campus line description matrix is ​​as follows:

[0145]

[0146] Assuming segment (3) is a primary fault segment, the campus fault information matrix is ​​as follows:

[0147] G = [1 1 1 -1 0 0 0] T

[0148] Where T represents the transpose of the matrix.

[0149] Therefore, the line fault judgment matrix can be obtained as follows:

[0150]

[0151] Using the preset fault location rules, the fault judgment matrix of the line is judged. P33=1, P34=1, P44=-1. The section between line node 3 and line node 4 meets the above condition (1). However, since the section between line node 3 and line node 4 is a T-connection section, it is necessary to further judge whether there is a fault between line node 3 and line node 7. Since P33=1, P37=1, P77=0, the above condition (1) is met. Therefore, there is also a fault between line node 3 and line node 7. According to the fault location rules, it can be judged that section (3) is a primary fault section, which is consistent with the assumption. Therefore, the matrix method can be used to locate faults in the campus line topology map.

[0152] In one embodiment, the campus electrical equipment further includes main power distribution equipment, secondary power distribution equipment, and campus electrical equipment. The information verification of the target terminal information based on the primary fault section and utilizing an information verification function includes the following steps:

[0153] The primary fault sections are transformed into a primary fault section matrix based on the line fault judgment matrix.

[0154] The campus power line topology is divided into upstream and downstream sections, with the primary fault section as the center point. The upstream section connects to the main power distribution equipment, and the downstream section connects to the campus power equipment.

[0155] Based on the campus line topology map, the power distribution status values ​​of all secondary power distribution equipment are obtained according to the campus fault information matrix. The power distribution status values ​​include the first power distribution status value of the upstream section and the second power distribution status value of the downstream section.

[0156] Based on the analysis of all primary fault sections, the status values ​​of all line sections within the campus line topology map are obtained, resulting in multiple section status values, including upstream section status values ​​and downstream section status values.

[0157] The total upstream state value and the total downstream state value are obtained by integrating the state values ​​of all upstream segments and all downstream segments respectively.

[0158] The primary fault section matrix, distribution status values, upstream section status values, downstream section status values, upstream total status value, and downstream total status value are input into the information verification function to obtain the primary fault verification matrix. The information verification function is as follows:

[0159]

[0160] Where i is a matrix element in the primary fault section matrix, x represents the upstream line, y represents the downstream line, and D x D represents the first power distribution state value. y Indicates the second power distribution state value. This represents the overall state value upstream. N represents the downstream total state value. x and N y M represents the total number of secondary power distribution equipment in upstream line x and downstream line y, respectively. x and M y Q represents the total number of all line segments within upstream line x and downstream line y, respectively. i，x Q represents the upstream segment status value of all line segments within upstream line x. i，y This represents the downstream segment status value of all line segments within the downstream line y, and Π represents the logical OR operation.

[0161] If the primary fault check matrix and the primary fault segment matrix are equal, then it is determined that there are no information errors in the target terminal information.

[0162] If the primary fault check matrix and the primary fault segment matrix are not equal, then the target terminal information is determined to have an error.

[0163] In this embodiment, a partial matrix corresponding to the primary fault section in the line fault matrix is ​​extracted, namely the primary fault section matrix. For example, section (3) is the primary fault section, and its corresponding primary fault section matrix is ​​as follows:

[0164] G = [1 1 1 -1 0 0 0] T

[0165] Using the fault section as the dividing point, the campus power line topology is divided. Figure 1 Divide the topology into two parts: the section connected to the main power distribution equipment is named the upstream segment, and the section connected to the campus power equipment is named the downstream segment, referring to... Figure 2 Assuming the faulty section is section (3), the upstream section includes section (1), section (2), section (5), and section (6), and the downstream section is section (4), section (5), and section (7).

[0166] The power distribution status values ​​of all secondary power distribution equipment are obtained based on the campus fault information matrix. The power distribution status value indicates whether the secondary power distribution equipment is connected to the primary power distribution equipment. If it is disconnected from the primary power distribution equipment, the secondary power distribution equipment cannot distribute power to other electrical devices, and its power distribution status value is 0; otherwise, the power distribution status value is 1. The first power distribution status value indicates whether there is a secondary power distribution device in the upstream section that is disconnected from the primary power distribution equipment. If it exists, the power distribution status value is 0; otherwise, it is 1. The second power distribution status value indicates whether there is a secondary power distribution device in the downstream section that is disconnected from the primary power distribution equipment. If it exists, the power distribution status value is 0; otherwise, it is 1. Section status values ​​refer to three logic signals: "1, 0, -1". 1 indicates that a line node experiences a forward fault overcurrent, and the corresponding section status value is 1. 0 indicates that a line node does not experience a fault overcurrent, and the corresponding section status value is 0. -1 indicates that a line node experiences a reverse fault overcurrent, and the corresponding section status value is -1. For example, if line node 1 experiences a forward fault overcurrent, the section status value of section 1 is 1. For ease of distinction, the section status values ​​of upstream sections are named "upstream section status values," and the section status values ​​of downstream sections are named "downstream section status values." All upstream section status values ​​are integrated to obtain the upstream total status value, and all downstream section status values ​​are integrated to obtain the downstream total status value. In other words, the upstream total status value contains all upstream section status values, and the downstream total status value contains all downstream section status values.

[0167] The primary fault section matrix, distribution status value, upstream section status value, downstream section status value, upstream total status value, and downstream total status value are input into the information verification function. If the output of the information verification function, i.e., the primary fault verification matrix, is equal to the primary fault section matrix, then the target terminal information is determined to have no information error. If the primary fault verification matrix is ​​not equal to the primary fault section matrix, then the target terminal information is determined to have an information error. An equal matrix means that the rows and columns of the matrix are the same, and the corresponding elements are all equal. Assume that the fault information of the above section (3) is distorted and becomes the following distorted matrix:

[0168] g = [1 0 1 -1 0 1 0] T

[0169] The distortion fault judgment matrix is ​​constructed based on the distortion matrix, and the distortion fault judgment matrix is ​​as follows:

[0170]

[0171] The distortion fault judgment matrix is ​​analyzed according to the preset fault location rules, and the fault segments are obtained as segments (1), (3), and (6). Segments (1), (3), and (6) are converted into a primary fault segment matrix and input into the primary fault verification matrix to obtain the primary fault verification matrix as follows:

[0172] T = [1 0 0 -1 0 0 0] T

[0173] Since matrix g and matrix T are not equal, it indicates that the fault information in segment (3) is distorted, which is consistent with the hypothesis.

[0174] In one embodiment, extracting several fault topology sub-graphs from the campus line topology map based on the first fault section includes the following steps:

[0175] Based on the location of all first-fault sections in the campus line topology map, the line nodes are marked with faults to obtain multiple first-fault line nodes.

[0176] Node aggregation is performed on the first faulty line node to obtain several faulty line node sets;

[0177] Based on the set of all faulty line nodes, several faulty topology subgraphs are extracted from the campus line topology map.

[0178] In this embodiment, based on the first fault segment located in the campus route topology map, the route nodes located within the first fault segment are marked as faulty, resulting in multiple first faulty route nodes. Adjacent faulty nodes are then aggregated to obtain several sets of faulty route nodes. Common node aggregation methods include DBSCAN (density-based clustering), spectral clustering, and topological clustering. Taking topological clustering as an example, the node values ​​are updated by traversing the adjacency list to determine the relationships between nodes, and clustering is performed based on these relationships. In topological clustering, nodes that are close together can be considered to belong to the same cluster or community, thereby achieving node aggregation and obtaining several sets of faulty route nodes. For each set of faulty route nodes, all route nodes and edges in the campus route topology map are traversed. If a route node belongs to the currently processed set of faulty route nodes, it is added to the fault topology subgraph. Simultaneously, if two route nodes connected by an edge both belong to the current set of faulty route nodes, the edge is also added to the fault topology subgraph, thus obtaining a fault topology subgraph for all sets of faulty route nodes.

[0179] In one embodiment, a fault segment localization model is constructed based on a graph neural network model. The fault segment localization model is then used to locate faults in all fault topology subgraphs to obtain several second fault segments in the campus route topology map. This includes the following steps:

[0180] A fault segment localization model is constructed based on a graph neural network model. The fault segment localization model includes a fault feature extraction module, a fault feature weighting module, a feature dimension transformation module, and a fault classification learning module. The fault feature extraction module is constructed based on a spatiotemporal graph neural network model, the fault feature weighting module is constructed based on a graph attention network model, the feature dimension transformation module is constructed based on a one-dimensional convolutional neural network model, and the fault classification learning module is constructed based on a fully convolutional neural network model.

[0181] The fault section location model is trained using a pre-built fault location training set, and the fault topology subgraph is sequentially input into the fault section location model for fault location, thereby obtaining several second fault sections in the campus power distribution line.

[0182] In this embodiment, the Spatiotemporal Graph Neural Network (ST-GNN) is a graph neural network model specifically designed for processing spatiotemporal data. It combines spatial and temporal dimensions, effectively capturing spatiotemporal dependencies in the data. The ST-GNN structure mainly comprises a graph neural network (GNN) and a sequence model (such as RNN, LSTM, etc.). The GNN updates the features of line nodes through a message-passing mechanism, aggregates information from the neighboring nodes of line nodes through graph convolution operations, and learns the spatial dependencies between line nodes to obtain fault spatial features. Then, the temporal features of the fault topology subgraph can be extracted using a recurrent neural network model (RNN) or a long short-term memory network model (LSTM). Taking RNN as an example, the recurrent structure of RNN allows the network to remember previous information and adapt to the temporal dependencies of sequential data, thereby enabling the extraction of the temporal features of the fault topology subgraph to obtain fault temporal features. The graph attention network model comprises an input layer, an attention layer, an aggregation layer, and an output layer. The input layer receives the spatiotemporal features of the fault. The attention layer learns the relationships between nodes by calculating attention coefficients, thus obtaining attention coefficients. The aggregation layer uses these attention coefficients to weighted aggregate the features of neighboring nodes, resulting in fault fusion features. The output layer outputs these fault fusion features. The one-dimensional convolutional neural network model mainly consists of convolutional layers and pooling layers. After the fault fusion features are input into the convolutional layers of the one-dimensional convolutional neural network for convolution, pooling layers downsample the output of the convolutional layers to reduce data dimensionality and redundant information. After passing through multiple convolutional and pooling layers, multiple feature maps are generated. These feature maps are then concatenated and input into a fully convolutional neural network model for feature classification. Fully Convolutional Neural Networks (FCNs) are a special type of convolutional neural network architecture where all fully connected layers are replaced with convolutional layers. The convolutional layers in a FCN perform convolution operations on the high-dimensional fault features of the input, i.e., the fault fusion features that have completed feature mapping, to extract local features. By stacking multiple convolutional layers, higher-level feature representations are gradually extracted. Then, pooling layers are used for downsampling to further reduce the feature dimensionality while retaining key information. After the features are mapped to a low-dimensional space, the maximum activation value obtained after passing through the Softmax function is the second fault segment.

[0183] The pre-built fault location training set contains historical campus route topology maps with pre-marked fault sections. The fault location training set is used to train the fault section location model. At the same time, an appropriate loss function (such as cross-entropy) is selected to evaluate the model performance. Optimizers such as Adam or SGD are selected for model training. The model parameters are updated through the backpropagation algorithm. The loss and accuracy are monitored during the training process. During the model training process, the model parameters of the fault section location model are continuously adjusted, such as the number of convolutional layers and the selection of activation functions, until the preset maximum number of training iterations is reached, and the model training is completed.

[0184] The fault topology subgraphs are sequentially input into the fault segment localization model for fault localization, resulting in several second fault segments in the campus power distribution lines. Specifically, the fault feature extraction module captures the spatial correlation between each line node in the fault topology subgraph and the correlation between fault topology subgraphs at different times. Then, the fault feature weighting module assigns weights to fault features of different dimensions based on the idea of ​​learning attention weights, and performs feature fusion on the fault features after weight assignment. The feature dimension transformation module performs feature extraction and dimension transformation on the fault features. The fault classification learning module maps the extracted high-dimensional fault features to a low-dimensional space containing fault segment information for classification learning. Finally, the maximum activation value (argmax) obtained after passing through the Softmax function is the second fault segment.

[0185] In one embodiment, the process of sequentially inputting the fault topology subgraph into the fault section location model for fault location to obtain several second fault sections in the campus power distribution line includes the following steps:

[0186] The fault topology subgraphs are sequentially input into the fault feature extraction module of the fault section localization model to perform convolution operations on the fault topology subgraphs on the spatiotemporal scale, thereby obtaining the fault spatiotemporal features of the fault topology subgraphs.

[0187] The spatiotemporal features of the fault are input into the fault feature weighting module to calculate the attention coefficient, and weights are assigned to the spatiotemporal features of the fault based on the calculation results of the attention coefficient.

[0188] The spatiotemporal features of the faults after weight allocation are weighted and fused to obtain fault fusion features;

[0189] The fault fusion features are input into the feature dimension transformation module for feature mapping, and the fault fusion features after feature mapping are input into the fault classification learning module for classification learning, resulting in several second fault sections in the campus line topology map.

[0190] In this embodiment, each fault topology subgraph is input into the fault feature extraction module. A graph convolution method using Chebyshev multinomials as the kernel is employed to capture the spatial correlation features of the fault topology subgraph, i.e., fault spatial features. Simultaneously, a two-dimensional convolution operation is introduced along the temporal axis to merge information from adjacent time steps, obtaining sequence features reflecting dynamic spatiotemporal characteristics, i.e., fault temporal features. The fault spatial and fault temporal features are then integrated to obtain the fault spatiotemporal features. These spatiotemporal features are then input into the fault feature weighting module for weight allocation. Specifically, the Graph Attention Network (GAT) model performs a linear transformation on the fault spatiotemporal features of each line node, mapping them to a higher dimension to obtain feature vectors for all line nodes. These feature vectors are then concatenated and inner-producted with a learnable vector. An activation function (such as LeakyReLU) is then used to obtain attention coefficients. Finally, GAT performs a weighted summation of the line node features based on the calculated attention coefficients to obtain the fault fusion features. The fault fusion features are input into the feature dimension transformation module for feature mapping. Specifically, the convolution kernels of the one-dimensional convolutional neural network (1D-CNN) slide upwards along the sequence axis of the fault fusion features. Each translation performs a weighted summation of the elements within the corresponding sequence window. Then, nonlinear mapping is performed on the data after convolution to extract features. Finally, the sub-features extracted by each convolution kernel are concatenated to obtain the fault fusion features with complete feature mapping. The fault fusion features with complete feature mapping are then input into the fault classification learning module. As the last link of the fault segment localization model, the fault classification learning module is used to map the extracted high-dimensional fault features to a low-dimensional space containing fault segment information and perform classification learning through the Softmax function. That is, it performs feature classification on the fault fusion features with complete feature mapping to obtain the fault localization result, which is the second fault segment.

[0191] In one embodiment, correcting all first fault sections using all second fault sections to obtain several line fault sections of the campus power distribution line includes the following steps:

[0192] Based on the location of the second faulty section in the campus line topology map, the line nodes are marked with faults to obtain the second faulty line node.

[0193] Analyze the campus network topology diagram to determine whether the second faulty line node overlaps with the first faulty line node.

[0194] If the second faulty line node overlaps with the first faulty line node, then the second faulty section corresponding to the overlapping second faulty line node is marked as the topology line faulty section.

[0195] All faulty sections of the topology lines are matched with the campus line distribution map to pinpoint their locations. Based on the matching results, several faulty sections in the campus line distribution map are located.

[0196] In this embodiment, based on the second fault section located in the campus line topology map, the line nodes located in the second fault section are marked as faults, resulting in multiple second fault line nodes. Based on the position information of the line nodes corresponding to the first and second fault line nodes in the campus line topology map, it is determined whether there is a second fault line node that overlaps with the first fault line node. If the two overlap, it indicates that the campus electrical device corresponding to the line node is likely to be faulty. Therefore, the second fault section corresponding to the overlapping second fault line node is marked as a topology line fault section. The reason for marking the topology line fault section based on the second fault section is that the second fault section located based on the graph neural network model is more accurate than the first fault section located based on the matrix method. The final obtained faulty sections of all topological lines are matched with the campus line distribution map to determine their actual locations within the target campus, i.e., several faulty sections on the campus line distribution map. These faulty sections are then marked on the campus line distribution map. The marked campus line distribution map, along with the corresponding campus electrical device numbers for the line nodes within each faulty section, is uploaded to the power distribution system management platform. Subsequently, staff on the power distribution system management platform can determine the faulty area based on the marked campus line distribution map and identify the faulty campus electrical device based on the uploaded device numbers, significantly saving manpower. The reduced time for staff to troubleshoot power distribution system faults, coupled with their ability to quickly locate and repair faulty electrical components on campus, significantly saves electricity. This is because faulty electrical components, even when supplied with a large amount of power, cannot continue to function, resulting in wasted energy. Furthermore, if the faulty component is a critical piece of equipment in the power distribution system, such as a distribution box, failure to repair it promptly could lead to short circuits, overloads, and other safety issues. In severe cases, this could damage electrical equipment, cause fires, or even disrupt the entire power system for extended periods, impacting daily life and production activities.

[0197] In addition, if the second faulty line node does not overlap with the first faulty line node, both the second faulty line node and the first faulty line node will be marked as suspected faulty nodes. If, during the subsequent fault location process of the campus power distribution lines, the suspected faulty node is marked as the second faulty line node or the first faulty line node again, it is determined that the campus electrical device corresponding to the suspected faulty node has a fault. Its number will also be uploaded to the power distribution system management platform, and the staff of the power distribution system management platform will inspect and repair it according to the number.

[0198] This application also discloses a power distribution system operation and maintenance management system based on data fusion analysis technology, characterized in that it includes:

[0199] The memory is configured to store instructions; and

[0200] The processor is configured to retrieve instructions from memory and, when executing the instructions, to implement a method for operation and maintenance management of a power distribution system based on data fusion analysis technology according to any one of the above.

[0201] The processor can be a central processing unit (CPU). Of course, depending on the actual use, it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc., and this application does not limit it in this regard.

[0202] The memory can be an internal storage unit of a computer device, such as a hard disk or RAM, or an external storage device, such as a plug-in hard disk, smart memory card (SMC), secure digital card (SD), or flash memory card (FC) provided on the computer device. Furthermore, the memory can be a combination of internal storage units and external storage devices of a computer device. The memory is used to store computer programs and other programs and data required by the computer device. The memory can also be used to temporarily store data that has been output or will be output. This application does not limit this.

[0203] This application also provides a machine-readable storage medium storing instructions that cause a machine to execute the above-described method for operation and maintenance management of a power distribution system based on data fusion analysis technology.

[0204] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0205] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0206] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0207] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0208] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0209] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0210] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0211] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0212] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A power distribution system operation and maintenance management method based on data fusion analysis technology, characterized in that, The method comprises the following steps: Obtain the campus information distribution map and historical power consumption information of the target campus, and the campus information distribution map comprises a campus line distribution map; Extract all campus electrical devices in the campus power distribution line of the target campus from the campus line distribution map; Perform line topology analysis on the campus line distribution map to obtain the device connection relationship between all the campus electrical devices, take all the campus electrical devices as line nodes, and construct a campus line topology map according to the device connection relationship, wherein the campus electrical devices include feeder terminal devices; Analyze the device key coefficients of all the campus electrical devices in combination with the historical power consumption information and the campus line topology map, and screen a plurality of the campus electrical devices from all the campus electrical devices as key campus electrical devices according to all the device key coefficients; If there is a device connection relationship between any feeder terminal device and any key campus electrical device, enable the redundant sensor pre-installed at the feeder terminal device, and take the feeder terminal device with the enabled redundant sensor as a key feeder terminal device; Obtain the key terminal information of all the key feeder terminal devices and the edge terminal information of all the feeder terminal devices except the key feeder terminal devices; On the basis of the campus line topology map, construct a line fault judgment matrix by fusing the key terminal information and the edge terminal information, and perform fault positioning in the campus line topology map based on the line fault judgment matrix to obtain a plurality of first fault sections in the campus line topology map; Extract a plurality of fault topology subgraphs from the campus line topology map according to all the first fault sections; Construct a fault section positioning model based on a graph neural network model, perform fault positioning in all the fault topology subgraphs by using the fault section positioning model, and obtain a plurality of second fault sections in the campus line topology map; Correct all the first fault sections by using all the second fault sections to obtain a plurality of line fault sections of the campus power distribution line.

2. The method of claim 1, wherein, The historical power consumption information comprises historical power consumption and historical power consumption duration, the campus information distribution map further comprises a campus building distribution map, and the step of analyzing the device key coefficients of all the campus electrical devices in combination with the historical power consumption information and the campus line topology map, and screening a plurality of the campus electrical devices from all the campus electrical devices as key campus electrical devices comprises the following steps: Divide the target campus into a plurality of campus power consumption areas according to the campus building distribution map; Divide area power consumption levels for all the campus power consumption areas in combination with the historical power consumption and the historical power consumption duration; Match the campus line distribution map with the campus building distribution map in position to obtain the device area position of all the campus electrical devices; For any campus electrical device, assign an initial device key coefficient to the campus electrical device according to the area power consumption level corresponding to the device area position; According to the device connection relationship, the node centrality of the campus electrical device is calculated; The node centrality is used to correct the initial device key coefficient to a device key coefficient; If the device key coefficient is greater than or equal to a preset coefficient threshold, the campus electrical device is marked as a key campus electrical device.

3. The method of claim 1, wherein, The campus line topology graph is taken as a basis, the key terminal information and the edge terminal information are fused to construct a line fault judgment matrix, and fault positioning is performed in the campus line topology graph based on the line fault judgment matrix, so that a first fault section of the campus line topology graph is obtained. For any key feeder terminal device, the information difference between the key terminal information collected by the redundant sensor in the key feeder terminal device and the key terminal information collected by other sensors in the key feeder terminal device is calculated, so as to obtain a plurality of key terminal information differences. If the absolute value of any key terminal information difference is greater than or equal to a preset difference threshold, it is determined that there is abnormal terminal information in the key terminal information. If the absolute value of all the key terminal information differences is less than the difference threshold, it is determined that there is no abnormal terminal information in the key terminal information. If there is abnormal terminal information in the key terminal information, the key terminal information collected by the redundant sensor is selected as target terminal information. The target terminal information and the edge terminal information are fused to obtain target feeder terminal information. The campus line topology graph and all the target feeder terminal information are combined to construct a line fault judgment matrix, and a plurality of primary fault sections are located in the campus line topology graph based on the line fault judgment matrix. For any primary fault section, the target terminal information is information-verified based on the primary fault section and by using an information verification function. If the information verification result shows that the target terminal information has no information error, the primary fault section is determined to be a first fault section of the campus line topology graph. If the information verification result shows that the target terminal information has information error, the target terminal information is updated according to all the information in the key terminal information except the target terminal information, so as to obtain reference terminal information. The reference terminal information and the edge terminal information are combined to perform fault positioning in the campus line topology graph, so that a first fault section of the campus line topology graph is obtained.

4. The method of claim 3, wherein, The target feeder terminal information is used to obtain electrical parameters of all the campus electrical devices, and the power direction of all the campus electrical devices is analyzed based on the electrical parameters and the topology structure type of the campus line topology graph. ​ Based on the campus line topology map, a campus line description matrix of the campus power distribution line is constructed in combination with the line power direction and the device connection relationship; According to the electrical parameters and by using a threshold method, device fault information in the target feeder terminal information is obtained, and a campus fault information matrix of the campus power distribution line is constructed in combination with the target feeder terminal information and the campus line topology map; A line fault judgment matrix is constructed in combination with the campus line description matrix and the campus fault information matrix; According to the line fault judgment matrix and by using a preset fault positioning rule, fault positioning is performed in the campus line topology map to obtain a plurality of primary fault sections in the campus line topology map.

5. The method of claim 3, wherein, The campus electrical device further includes a main power distribution device, a secondary power distribution device, and a campus electrical device, and information checking of the target terminal information based on the primary fault section and by using an information checking function includes the following steps: The primary fault section is converted into a primary fault section matrix according to the line fault judgment matrix; The campus line topology map is divided into an upstream section and a downstream section with the primary fault section as a center point, the upstream section is connected to the main power distribution device, and the downstream section is connected to the campus electrical device; According to the campus fault information matrix, power distribution state values of all the secondary power distribution devices are obtained based on the campus line topology map, the state power distribution values include a first power distribution state value of the upstream section and a second power distribution state value of the downstream section; According to all the primary fault sections, state values of all line sections in the campus line topology map are analyzed to obtain a plurality of section state values, the section state values include upstream section state values and downstream section state values; All the upstream section state values and all the downstream section state values are integrated to obtain an upstream total state value and a downstream total state value, respectively; The primary fault section matrix, the power distribution state values, the upstream section state values, the downstream section state values, the upstream total state value, and the downstream total state value are input into an information checking function to obtain a primary fault checking matrix, the information checking function is as follows: ， wherein represents a matrix element within the primary fault section matrix, represents an upstream line, represents a downstream line, represents the first power distribution state value, represents the second power distribution state value, represents the upstream total state value, represents the downstream total state value, and respectively represent the total number of secondary power distribution devices within the upstream line and the downstream line and respectively represent the total number of all line sections within the upstream line and the downstream line represents the upstream section state value of all line sections within the upstream line represents a downstream section state value of all line sections within the downstream line represents a logical operation or operation;​​​​ If the primary fault checking matrix and the primary fault section matrix are equal matrices, it is determined that the target terminal information does not have information errors; If the primary fault checking matrix and the primary fault section matrix are not equal matrices, it is determined that the target terminal information has information errors.

6. The method of claim 1, wherein, The extraction of a plurality of fault topology subgraphs from the campus line topology map according to the first fault section includes the following steps: According to the section positions of all the first fault sections in the campus line topology map, the line nodes are marked for faults to obtain a plurality of first fault line nodes; The first fault line nodes are aggregated to obtain a plurality of fault line node sets; A plurality of fault topology subgraphs are extracted from the campus line topology map according to all the fault line node sets.

7. The method of claim 6, wherein, The fault section positioning model is constructed based on the graph neural network model, and the fault section positioning model is used for fault positioning in all the fault topology subgraphs to obtain a plurality of second fault sections in the campus line topology graph. The fault section positioning model is constructed based on the graph neural network model, and the fault section positioning model includes a fault feature extraction module, a fault feature weighting module, a feature dimension transformation module and a fault classification learning module, the fault feature extraction module is constructed based on a space-time graph neural network model, the fault feature weighting module is constructed based on a graph attention network model, the feature dimension transformation module is constructed based on a one-dimensional convolutional neural network model, and the fault classification learning module is constructed based on a full convolutional neural network model. The fault section positioning model is trained by using a pre-constructed fault positioning training set, and the fault topology subgraph is sequentially input into the fault section positioning model for fault positioning to obtain a plurality of second fault sections in the campus distribution line.

8. The method of claim 7, wherein, The fault section positioning model is constructed based on the graph neural network model, and the fault section positioning model is used for fault positioning in all the fault topology subgraphs to obtain a plurality of second fault sections in the campus line topology graph. The fault topology subgraph is sequentially input into the fault feature extraction module of the fault section positioning model to perform convolution operation on the fault topology subgraph in the space-time scale to obtain fault space-time features of the fault topology subgraph. The fault space-time features are input into the fault feature weighting module for attention coefficient calculation, and the fault space-time features are assigned weights according to the attention coefficient calculation result. The fault space-time features with completed weight assignment are weighted and fused to obtain fault fusion features. The fault fusion features are input into the feature dimension transformation module for feature mapping, and the fault fusion features with completed feature mapping are input into the fault classification learning module for classification learning to obtain a plurality of second fault sections in the campus line topology graph.

9. The method of claim 8, wherein, The second fault sections are used to correct all the first fault sections to obtain a plurality of line fault sections of the campus distribution line, including the following steps: The line nodes are marked according to the positions of the second fault sections in the campus line topology graph to obtain second fault line nodes; Whether the second fault line nodes and the first fault line nodes coincide is analyzed according to the campus line topology graph; If the second fault line nodes and the first fault line nodes coincide, the second fault sections corresponding to the second fault line nodes with coincidence are marked as topology line fault sections; All the topology line fault sections are matched with the campus line distribution graph in position, and a plurality of line fault sections in the campus line distribution graph are positioned according to the position matching structure.

10. A power distribution system operation and maintenance management system based on data fusion analysis technology, characterized in that, It includes: a memory configured to store instructions; and a processor configured to call the instructions from the memory and capable of realizing the power distribution system operation and maintenance management method based on data fusion analysis technology according to any one of claims 1 to 9 when executing the instructions.

Images