Method and system for identifying hidden fault of power distribution network

Abstract

The invention discloses a method and a system for identifying a hidden fault of a power distribution network, relates to the technical field of identification of the hidden fault, and aims to determine an abnormal identification system of a line based on a plurality of line early warning items, a section topological relation and the state of the power distribution network, so that the accuracy of the abnormal identification system of the line is improved. Determining a plurality of corresponding fault features according to the detection of the abnormity identification system of the line, determining a corresponding fault area based on the feature positions of the plurality of fault features, the corresponding feature forms and the corresponding line management and control nodes, and determining corresponding fault recording data based on the identification of the fault area; a plurality of fault nodes are determined based on the fault recording data, the data combination of the fault area and the working process of the power distribution network, the hidden fault of the power distribution network is determined according to the node position of each fault node, the corresponding hidden content and the working content of the fault area, and the accuracy of the hidden fault of the power distribution network is improved.

Description

Technical Field

[0001] This invention relates to the technical field of identifying hidden faults, and more particularly to a method and system for identifying hidden faults in a power distribution network. Background Technology

[0002] With the development of technology, power distribution networks are gradually being applied to people's lives and serve as components of transmission circuits. Power distribution networks have multiple lines, each extending along a preset path to transmit electricity in a directional manner. In existing technologies, data from each line is collected and its status is marked. Faults in the power distribution network are determined based on the status and abnormal signals of each line. However, the location of each fault node and its corresponding hidden nature are ignored, resulting in low accuracy in identifying hidden faults in the power distribution network. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and system for identifying hidden faults in power distribution networks.

[0004] This invention provides a method for identifying hidden faults in a power distribution network, comprising: During the operation of the distribution network, the corresponding line data combination is determined based on the detection of the distribution network lines, and the segment topology relationship is determined according to each line data combination and the line path of the distribution network. Based on the identification of the topology of the section, multiple line control nodes are identified, and the status of each line control node is marked. Based on the warning status of the line control node, the corresponding line warning event is determined. Based on the detection of the line early warning event, multiple corresponding line early warning items are determined, and based on the multiple line early warning items, the section topology relationship and the status of the distribution network, the line anomaly identification system is determined; Based on the detection of the anomaly identification system of the line, multiple fault features are determined. Based on the feature location, corresponding feature shape and corresponding line control node of the multiple fault features, the corresponding fault area is determined. Based on the identification of the fault area, the corresponding fault recording data is determined. Based on the fault recording data, the data combination of the fault area, and the working history of the distribution network, multiple fault nodes are identified. Based on the node location of each fault node, the corresponding hidden content, and the working content of the fault area, the hidden faults of the distribution network are determined.

[0005] This invention provides a system for identifying hidden faults in a power distribution network, which is applied to the aforementioned method for identifying hidden faults in a power distribution network.

[0006] Compared with the prior art, the beneficial effects of the present invention are: (1) During the operation of the distribution network, the corresponding line data combination is determined based on the detection of the distribution network lines, and the segment topology relationship is determined according to each line data combination and the line path of the distribution network; multiple line control nodes are determined based on the identification of the segment topology relationship, and the status of each line control node is marked. The corresponding line warning event is determined according to the warning status of the line control node; multiple line warning items are determined based on the detection of the line warning event, and the line anomaly identification system is determined based on multiple line warning items, segment topology relationship and distribution network status. The status of each line control node is introduced, which is compatible with the consideration of multiple line warning items, segment topology relationship and distribution network status, and improves the accuracy of the line anomaly identification system.

[0007] (2) Based on the detection of the abnormal identification system of the line, multiple fault features are determined. Based on the feature location, corresponding feature shape and corresponding line control node of the multiple fault features, the corresponding fault area is determined. Based on the identification of the fault area, the corresponding fault waveform data is determined. Based on the fault waveform data, the data combination of the fault area and the working history of the distribution network, multiple fault nodes are determined. Based on the node location of each fault node, the corresponding hidden content and the working content of the fault area, the hidden faults of the distribution network are determined. The fault waveform data is further controlled, and the overall consideration of the node location of each fault node, the corresponding hidden content and the working content of the fault area is realized, which improves the accuracy of the hidden faults of the distribution network. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating the method for identifying hidden faults in a power distribution network according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating step S11 in the method for identifying hidden faults in a power distribution network according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating step S12 in the method for identifying hidden faults in a power distribution network according to an embodiment of the present invention. Figure 4 This is a flowchart illustrating step S13 in the method for identifying hidden faults in a power distribution network according to an embodiment of the present invention. Figure 5 This is a flowchart illustrating step S14 in the method for identifying hidden faults in a power distribution network according to an embodiment of the present invention. Figure 6 This is a flowchart illustrating step S15 in the method for identifying hidden faults in a power distribution network according to an embodiment of the present invention. Figure 7 This is a schematic diagram of the structural composition of the system for identifying hidden faults in a power distribution network according to an embodiment of the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0010] Please see Figures 1 to 7 A method for identifying hidden faults in a power distribution network, applied to scenarios involving the identification of hidden faults; the method for identifying hidden faults in a power distribution network includes: Step S11: During the operation of the distribution network, the corresponding line data combination is determined based on the detection of the distribution network lines, and the segment topology relationship is determined according to each line data combination and the line path of the distribution network. Step S12: Based on the identification of the topology of the section, determine multiple line control nodes, mark the status of each line control node, and determine the corresponding line warning event based on the warning status of the line control node; Step S13: Based on the detection of the line warning event, determine the corresponding multiple line warning items, and based on the multiple line warning items, the section topology relationship and the status of the distribution network, determine the line anomaly identification system; Step S14: Based on the detection of the anomaly identification system of the line, determine the corresponding multiple fault features, determine the corresponding fault area based on the feature location, corresponding feature shape and corresponding line control node of the multiple fault features, and determine the corresponding fault recording data based on the identification of the fault area. Step S15: Based on the fault waveform data, the data combination of the fault area and the working history of the distribution network, determine multiple fault nodes, and determine the hidden faults of the distribution network according to the node location of each fault node, the corresponding hidden content and the working content of the fault area.

[0011] refer to Figure 2 In step S11, the specific steps are as follows: S111: Real-time monitoring of the operation of the distribution network, identification of multiple lines based on the distribution network distribution map, determination of multiple corresponding working data based on the detection of each line, and determination of corresponding line data combinations based on the matching of each line and the corresponding multiple working data. S112: Based on the detection of the distribution network distribution map, multiple line nodes are determined. Based on the node information and corresponding node locations of the multiple line nodes, the line path of the distribution network is determined. Based on the line path of the distribution network, the corresponding line, and the combination of corresponding line data, the segment topology relationship is determined.

[0012] In the embodiments of this application, the operation of the distribution network is monitored in real time. Multiple lines are identified based on the distribution network map, and multiple corresponding working data are determined based on the detection of each line. The corresponding line data combination is determined based on the matching of each line and the corresponding multiple working data. This approach takes into account the overall consideration of matching each line and the corresponding multiple working data, ensuring the accuracy of the corresponding line data combination.

[0013] At this time, intelligent monitoring devices deployed at key nodes of the distribution network, such as FTU (Feeder Terminal Unit), DTU (Distribution Terminal Unit), or dedicated fault indicators (FI), continuously collect real-time data from the power grid. These devices mainly collect electrical parameters such as three-phase current, three-phase voltage, active power, reactive power, power factor, and zero-sequence current / voltage. This data is recorded in the form of high-frequency sampling, forming raw fault waveform data, which provides the most original and richest information source for subsequent analysis.

[0014] The system needs to identify and determine multiple independent lines based on the distribution network map. The system maintains a digital distribution network geographic information system (GIS) or network topology model, which contains the physical attributes of all lines, such as line number, start and end nodes, line length, cable type, etc. By parsing this model, the system can automatically identify all independent line segments that need to be monitored. Here, a line usually refers to the physical connection between two adjacent monitoring devices, which is the smallest unit in topology analysis.

[0015] The system accurately associates the massive, continuous, real-time data it has collected with the specific lines it has identified. The system determines which line the data belongs to based on the data source (such as the installation location of the FTU). At the same time, the system performs preliminary processing on the raw data, such as calculating characteristic values ​​like RMS, peak value, abrupt changes, and harmonic content. This processed data can more directly reflect the operating status of the line and potential anomalies.

[0016] The system creates a data structure or object for each independent line. This structure cleverly integrates the line's static attributes (from the GIS model) and dynamic operational data (from real-time monitoring). This combination can be represented as a multi-dimensional feature vector, for example: Line Data Combination = [Line ID, Current RMS Value, Voltage RMS Value, Zero Sequence Current, Power Factor, ...]. This vector is the direct input for subsequent status assessment and anomaly identification.

[0017] Specifically, for the hidden fault identification scenario of the 10kV distribution network in area B, during the real-time monitoring phase, eight intelligent fault indicators (FIs) were installed on the main line and key branches of the distribution network, numbered F1 to F8. These FIs continuously monitor the line current at a sampling rate of 4kHz. When a sudden change in current, over-limit, or signal of a specific spectrum is detected, waveform recording is triggered and the data is uploaded to the master station system.

[0018] In the scenario of hidden fault identification in the 10kV distribution network of region B, during the real-time monitoring phase, we installed 8 intelligent fault indicators (FIs) on the main line and key branches of the distribution network, numbered F1 to F8. These FIs continuously monitor the line current at a sampling rate of 4kHz. When a sudden change in current, over-limit, or signal of a specific spectrum is detected, waveform recording is triggered and the data is uploaded to the main station system. FIF3 and F7 simultaneously triggered alarms and uploaded waveform data. The system extracted the following from the waveform data of F3: the effective value of the A-phase current is 152A (normal load is about 100A), there is a sudden change of 30A lasting about 20ms, and the zero-sequence current component is 5A. The system marked these data as working data belonging to lines L3 and L4. Similarly, the system also extracted the working data of L7 and L8 from the data of F7, completing the conversion of the raw data into specific line characteristics.

[0019] The system constructs data combinations for all eight line sections in real time. For sections experiencing anomalies, their data combinations will contain significant abnormal characteristics. For example, the data combination for line L3 is: [Line ID: L3, Current: 152A, Current Sudden Amount: 30A, Zero-Sequence Current: 5A, ...]; while the normal state data combination for line L1 is: [Line ID: L1, Current: 98A, Current Sudden Amount: 0A, Zero-Sequence Current: 0.1A, ...]. Through this process, we abstract the distribution network into a set composed of data combinations from eight lines. These structured data combinations will become a solid foundation for the next step of topology modeling and state analysis. For example, the high zero-sequence current appearing simultaneously in the data combinations of L3, L4, and L7 is a strong correlation characteristic pointing to a single-phase grounding fault, providing a key clue for subsequent accurate location.

[0020] Furthermore, multiple line nodes are identified based on the detection of the distribution network distribution map. The line paths of the distribution network are determined based on the node information and corresponding node locations of the multiple line nodes. The segment topology is determined based on the line paths of the distribution network, the corresponding lines, and the corresponding combinations of line data. This approach takes into account the overall consideration of the line paths of the distribution network, the corresponding lines, and the corresponding combinations of line data, ensuring the accuracy of the segment topology.

[0021] At this point, the system will parse the digital geographic information system (GIS) or network wiring diagram of the power distribution network. This diagram stores the physical structure of the power grid in the form of nodes and edges. Nodes represent the connection points of lines, such as substation busbars, towers, branch points, switching stations, transformer substations, etc. The system identifies all key nodes by traversing the network diagram and assigns a unique identifier (ID) to each node. At the same time, it records the node attributes, such as node type (power source point, T-junction, load point), geographical coordinates (latitude and longitude), and connected equipment information.

[0022] Line paths are defined based on the physical connections between nodes; a line path is a sequence of two or more nodes connected in an orderly manner; the system will clearly define the starting point and ending point of each path, as well as all the nodes passed through in between; for radial distribution networks, the path usually starts from the power source (substation) and extends to the end load; at the same time, the system will also record the physical parameters of the path, such as line length, conductor type, impedance, etc., to provide physical basis for subsequent fault analysis.

[0023] The system determines the segment topology based on the distribution network's line paths, corresponding lines, and the corresponding combinations of line data. The system uses the locations of monitoring devices such as fault indicators (FIs) or fault transfer units (FTUs) as logical dividing points to divide long line paths into multiple smallest, indivisible segments. Each segment is uniquely numbered, and a topology matrix is ​​established. This matrix describes the logical relationship of which segments are faulty if a monitoring point (FI) alarms. The matrix's rows typically correspond to monitoring points (FIs), and the columns correspond to line segments. The values ​​of matrix elements (such as 0 or 1) represent the logical association between the FI's alarm signal and the segment's fault state. The line data combinations generated in step S111 are bound to the corresponding segments, completing the deep integration of the data layer and the topology layer.

[0024] Specifically, the digital GIS model of the power distribution network is imported into the system; the system automatically identifies all key nodes, including: N0 (10kV outgoing bus of 110kV substation, power source), N1 (#1 tower, T-junction), N2 (#2 switch station), N3 (#3 tower, branch point), N4 (#4 distribution transformer area), and N5 (#5 tower, line end); the system records the precise geographical coordinates and equipment parameters of each node, completing the digital mapping of the physical skeleton of the power grid.

[0025] The system constructs the main line path based on the connection relationship between nodes; for example, the path of a main line can be described as: Path_Main=[N0->N1->N2->N3->N5]; at the same time, a branch path is branched from point N3: Path_Branch=[N3->N4]; the system records the length and conductor parameters of each path segment, for example, the section between N0 and N1 is a 2km long JKLYJ-185 type overhead line.

[0026] Fault indicators, named F1, F2, F3, and F4, were installed at four nodes N1, N2, N3, and N5 in the distribution network. The system uses these F1 and F2 nodes as logical dividing points to segment the line path: segment L1 is the line between N0 and F1 (N1); segment L2 is the line between F1 (N1) and F2 (N2); segment L3 is the line between F2 (N2) and F3 (N3); segment L4 is the line between F3 (N3) and F4 (N5); and segment L5 is the branch line between F3 (N3) and N4. Next, the system constructs a 4x5 topology matrix R, with rows corresponding to F1-F4 and columns corresponding to L1-L5. The rule for the value of matrix element r_ij is: if a fault occurs in section Lj, the fault current will flow through Fi, then r_ij=1, otherwise it is 0; for example, when L3 is faulty, the current will flow through F1, F2, and F3, but will not flow through F4; the system associates the line data combination generated in step S111 (such as the current mutation amount and zero-sequence current of section L3) with section L3 in the topology model; through this process, the distribution network is abstracted into a topology model with clear mathematical relationships consisting of 5 logical sections and 4 monitoring points. This model is the solid foundation for subsequent alarm vector construction and matching location, enabling the system to infer the fault section from the discrete alarm signals.

[0027] refer to Figure 3 In step S12, the specific steps are as follows: S121: Identify the topological relationship of this section, and determine multiple line control areas during the identification process. Based on the identification of each line control area, determine the corresponding line control node to identify multiple line control nodes. S122: Among multiple line control nodes, the status of the line control node is determined based on the identification of each line control node. If the status of the line control node is an early warning status, the corresponding line early warning identifier is determined based on the tracing of the line control node, and the corresponding line early warning event is determined based on the detection of the line early warning identifier.

[0028] In the embodiments of this application, the topological relationship of the segment is identified, and multiple line control areas are determined during the identification process. Based on the identification of each line control area, the corresponding line control node is determined, thereby determining multiple line control nodes. This approach takes into account the overall consideration of identifying each line control area and ensures the accuracy of the corresponding line control node.

[0029] At this point, the system performs an in-depth analysis of the segment topology relationships constructed in step S11. This analysis goes beyond simple connection relationships, combining multi-dimensional information such as the importance of the network structure, load level, line length, and historical fault data. The system identifies line control areas that require key attention through algorithms or preset rules. These areas are typically network hubs (such as T-junctions between trunk lines and important branch lines), long line segmentation points (midpoints of long lines segmented for fault isolation), critical load entry points (line entry points connecting to primary and secondary important loads), and equipment boundary points (connections between different types of equipment). A line control area logically represents a set of one or more adjacent segments, and its operating status has a significant impact on the reliability of the entire network.

[0030] The system maps these logical areas to specific locations in the physical world; the system queries equipment ledgers and GIS information to determine the physical location of the monitoring equipment (such as FTU, DTU, and fault indicator FI) actually installed in each control area or on its boundary. This physical location is the line control node; through this process, the system finally determines a set of multiple physical control nodes. These nodes are part of the entire fault identification system. All subsequent status assessments and early warnings are based on the data provided by these nodes, realizing the optimized allocation of monitoring resources.

[0031] Specifically, a topology model containing 5 sections (L1-L5) and 4 monitoring points (F1-F4) was constructed. In the stage of determining the line control area, the system analyzed the topology of the distribution network and, combined with the load importance information, identified the following three key line control areas: Control Area A (Main Hub Area), which is the T-junction of the main line and an important branch leading to the industrial area. A fault in this area would cause a large-scale power outage and is automatically marked as high priority by the system; Control Area B (Long Line Segment Area), where the line length from the substation to this area exceeds 10 kilometers. To facilitate fault isolation, the intermediate segment points are identified as critical; Control Area C (Important Load Entry Area), which is the entry point of the 10kV line leading to the only tertiary hospital in area B. Due to the extremely high load level, this entry point is identified as the highest priority control area.

[0032] Based on the identified control areas, the system queries equipment installation records to precisely match logical areas with physical devices. For control area A (main hub area), the system finds its corresponding physical node N3, and fault indicator F3 is installed at N3. Therefore, line control node CN-A is established, and its physical entity is F3. For control area B (long line segment area), the system finds its corresponding physical node N1, and fault indicator F1 is installed at N1. Therefore, line control node CN-B is established, and its physical entity is F1. For control area C (important load entry area), the system finds its corresponding physical node N4, and a dedicated FTU device is installed at N4. Therefore, line control node CN-C is established, and its physical entity is the FTU at N4. Through this step, the system no longer treats all monitoring points equally, but intelligently identifies the three line control nodes—F1, F3, and the FTU at N4—that are crucial to the safe operation of the distribution network. Subsequent monitoring and early warning resources will prioritize these nodes, laying a solid foundation for efficient and accurate identification of hidden faults.

[0033] Furthermore, among multiple line control nodes, the status of each line control node is determined based on its identification. If the status of a line control node is an early warning state, the corresponding line early warning identifier is determined based on the tracing of the line control node. The corresponding line early warning event is determined based on the detection of the line early warning identifier. This approach incorporates the overall consideration of tracing line control nodes and ensures the accuracy of the corresponding line early warning identifier.

[0034] At this time, the system continuously obtains the associated line data combinations (from S111) from each line control node (i.e., monitoring equipment such as FTU and FI). The system internally deploys a status evaluation algorithm for each node. This algorithm analyzes the data combinations in real time based on preset thresholds, logical rules, or more complex pattern recognition models. For example, it monitors whether the zero-sequence current exceeds the set threshold, whether the current harmonics exceed the standard, and whether specific transient waveforms appear. According to the analysis results, the status of the node is divided into discrete levels, such as normal, warning, and fault. When any one or more indicators trigger the preset abnormal conditions, the status of the node is determined to be a warning state.

[0035] If a line control node is in an alert state, the system will determine the corresponding line alert identifier based on the traceability of the line control node. Once the state of a node is determined to be in an alert state, the system will immediately initiate an event generation process. The first step of this process is to create a unique line alert identifier for the event. This identifier is usually a string containing a timestamp and a sequence number (such as EVT-YYYYMMDD-HHMMSS-NNN) to ensure its global uniqueness. At the same time, the system will record the strong association between this identifier and the control node ID that triggered it, realizing real-time traceability from event to node.

[0036] After generating the warning identifier, the system constructs a complete line warning event data structure. This event is the core object of all subsequent analyses. The event not only contains a unique warning identifier but also encapsulates rich contextual information, forming a standardized information package. The content typically includes the event ID (i.e., the line warning identifier), the triggering node (the physical ID and topological location of the line control node), the timestamp (the precise time when the warning occurred), the warning type (such as intermittent grounding, arc discharge, etc.), key characteristic values ​​(the specific data that triggered the warning), and the associated waveform file (a pointer or link to the complete fault waveform data).

[0037] Specifically, three key line control nodes of the distribution network were identified: CN-A (F3), CN-B (F1), and CN-C (FTU at N4). During the status determination phase of the line control nodes, it was assumed that on a wet, rainy night, an intermittent grounding fault caused by insulator flashover occurred in the distribution network. The system monitored all control nodes in real time. Data from control node CN-A (F3) showed that its zero-sequence current experienced multiple pulses exceeding 5A within a short period, with waveform characteristics consistent with the system's preset intermittent grounding pattern. Simultaneously, the data from its upstream CN-B (F1) and downstream CN-C (FTU at N4) both showed normal readings. Based on this, the system immediately updated the status of control node CN-A (F3) from normal to warning status.

[0038] The system detects that CN-A has entered an early warning state and immediately generates a unique line early warning identifier: EVT-20231027-221533-001. Internally, it establishes a mapping relationship between this identifier and node CN-A (F3), ensuring the uniqueness and traceability of the event. Based on the early warning identifier EVT-20231027-221533-001, the system constructs a detailed line early warning event and pushes it to the event processing queue. The event content is as follows: Event ID is EVT-20231027-221533-001. The triggering node is CN-A (physical device F3, located at trunk node N3); the timestamp is 2023-10-27 22:15:33.125; the warning type is intermittent grounding; the key characteristic is a zero-sequence current pulse with a peak value of 6.2A and a pulse interval of approximately 200ms; the associated waveform file is / data / waveforms / F3_20231027_221533.dat; through this step, the system successfully transformed a weak and complex electrical phenomenon into a clear, traceable, and information-rich line warning event.

[0039] refer to Figure 4 In step S13, the specific steps are as follows: S131: Monitor the early warning events of the line in real time, perform dynamic detection of the early warning events of the line, and determine multiple line early warning contents during the detection process. Based on the multiple line early warning contents, the corresponding early warning areas and the corresponding line control nodes, determine the corresponding multiple line early warning items. S132: Collect multiple working data of the distribution network, determine the status of the distribution network based on the multiple working data of the distribution network, determine the first level of anomaly identification content based on multiple line early warning projects and the status of the distribution network, determine the second level of anomaly identification content based on the section topology and the status of the distribution network, and determine the line anomaly identification system based on the first level of anomaly identification content and the second level of anomaly identification content.

[0040] In the embodiments of this application, the line warning event is monitored in real time, the line warning event is dynamically detected, and multiple line warning contents are determined during the detection process. Based on the multiple line warning contents, the corresponding warning areas and the corresponding line control nodes, multiple corresponding line warning items are determined. This approach takes into account the overall consideration of multiple line warning contents, the corresponding warning areas and the corresponding line control nodes, ensuring the accuracy of the corresponding multiple line warning items.

[0041] At this point, the system places the line warning event generated by S12 into an active monitoring queue. Dynamic detection is not a one-time reading, but a continuous and in-depth analysis of the fault waveform data associated with the event. The system calls various signal processing and pattern recognition algorithms to scan the waveform data stream, including time-domain feature extraction, frequency-domain analysis, time-frequency analysis, and waveform morphology matching. This dynamism is reflected in the fact that the system can adaptively load the most suitable set of detection algorithms according to the preliminary type of the warning event. For example, for grounding events, the system will focus on zero-sequence component and transient traveling wave analysis.

[0042] Through parallel or serial analysis of the aforementioned multiple algorithms, the system can identify multiple specific and quantifiable anomalies from complex waveform data. Each phenomenon is a line warning content. For example, a single hidden fault (such as insulator flashover) can trigger multiple warning content in the waveform data simultaneously, such as the instantaneous value of zero-sequence current exceeding the threshold, significant imbalance of three-phase current, and high-frequency noise in the voltage waveform. Each warning content is assigned a specific characteristic value, such as peak value of zero-sequence current: 8.5A, three-phase imbalance: 30%, etc., realizing a quantitative description of the anomaly.

[0043] The system determines multiple line warning items based on the warning content, corresponding warning areas, and corresponding line control nodes. The system creates an independent line warning item for each identified warning content. Each item is a structured data object that binds the warning content itself with contextual information. Its core elements include item ID, associated warning event ID, trigger node, warning area, warning content description, and quantitative feature value. A warning event is parsed into a set of multiple line warning items, realizing a multi-dimensional and multi-perspective description of the fault.

[0044] Specifically, the system has generated an early warning event EVT-20231027-221533-001, triggered by the control node CN-A (F3), and associated it with a fault recording file. During the dynamic detection of the early warning event, the system initiates dynamic detection of the recording file F3_20231027_221533.dat associated with the event. Since the initial early warning type is intermittent grounding, the system prioritizes loading the zero-sequence component analysis, transient pulse detection, and three-phase unbalance calculation modules. When scanning the recording data, the detection module simultaneously detects multiple abnormal features.

[0045] The system identified three independent anomalies: 1) The zero-sequence channel detected multiple pulse signals with amplitudes exceeding 5A; 2) The instantaneous value calculation of the three-phase current showed that at the same moment the zero-sequence pulse appeared, the vector sum of the three-phase currents was not zero, and the imbalance exceeded 25%; 3) The high-frequency harmonic analysis module found that during the pulse period, the content of the 3rd and 5th harmonics in the voltage waveform increased significantly. The system identified these three independent anomalies as three line warning items.

[0046] The system created structured line warning items for the three warning contents mentioned above. For example, warning item P-001 includes item ID, associated event ID, trigger node CN-A (F3), warning area (main line section monitored by F3), warning content description of zero-sequence current pulse exceeding the standard, and quantified characteristic value pulse peak value of 6.2A, with a repetition period of approximately 200ms. Similarly, the system also created item P-002 describing three-phase current imbalance and item P-003 describing voltage harmonic distortion. Through this step, the system refined a general intermittent grounding warning into three interrelated warning items, each with its own focus. This provides richer and more specific analytical materials for the subsequent construction of the anomaly identification system, enabling the system to cross-verify from multiple perspectives and thus more accurately determine the nature and location of the fault.

[0047] Furthermore, multiple operational data points from the distribution network are collected, and the distribution network status is determined based on these data points. The first level of anomaly identification content is determined based on multiple line early warning projects and the distribution network status. The second level of anomaly identification content is determined based on the segment topology and the distribution network status. Based on these first and second level anomaly identification content, a line anomaly identification system is established, incorporating both levels of anomaly identification to ensure accuracy. Simultaneously, the status of each line control node is introduced, taking into account multiple line early warning projects, segment topology, and the distribution network status, further improving the accuracy of the line anomaly identification system.

[0048] At this time, the system collects macroscopic working data that can reflect the overall operating status of the distribution network through SCADA system, meteorological interface, etc., including total active / reactive power of the whole network, average voltage level of the system, load factor, regional weather forecast (such as humidity, wind speed), etc. Based on these macroscopic data, the system conducts a comprehensive assessment of the current overall status of the distribution network and draws a qualitative conclusion, such as heavy load, normal load, light load, and the impact of severe weather. This status provides important background information for subsequent anomaly analysis and helps to distinguish whether it is an internal fault or an external environmental influence.

[0049] The system determines the first level of anomaly identification based on multiple line warning items and the state of the distribution network. The system cross-compares and logically infers the warning items generated by S131 with the current distribution network state. Its core is to determine whether there is a conflict or coordination relationship between the warning items and the power grid state. For example, if a current overload warning occurs under light load conditions, it is an internal fault. If an intermittent grounding warning occurs under severe weather conditions, it is caused by external factors. The first level of anomaly identification will give a context-based evaluation conclusion to verify the rationality of local anomalies.

[0050] The system determines the second level of anomaly identification based on the segment topology and the state of the distribution network. It uses distribution network state information to perform logical verification on non-alarm areas in the topology model to validate the rationality of alarm areas. It examines local anomalies from a global perspective; for example, if the overall load of the distribution network is low, but an abnormal current is detected at a downstream control node of a certain segment, this points to a fault in that segment itself, rather than downstream load fluctuations. The second level of anomaly identification provides a verification conclusion based on topological logic, further narrowing down the possible range of faults. The system integrates the first and second levels of identification to form a structured, multi-dimensional line anomaly identification system. This system not only includes detailed information on all early warning items, but more importantly, it integrates in-depth analysis conclusions based on the global state of the power grid and topological logic.

[0051] Specifically, the system has extracted three warning items (P-001, P-002, P-003) from the warning event EVT-20231027-221533-001. During the stage of determining the status of the distribution network, the system collects SCADA data of the distribution network, showing that the current network load rate is 40%, the system voltage is stable, and the meteorological interface shows that it is currently raining lightly with an air humidity of 85%. After comprehensive evaluation, the distribution network status is determined to be normal load and humid weather.

[0052] In the first stage of anomaly identification, the system analyzed warning items P-001 (zero-sequence current pulse) and P-002 (three-phase current imbalance). Under normal load conditions, the presence of such obvious zero-sequence current and imbalance basically ruled out the possibility of it being caused by normal operation such as system overload or large motor starting. At the same time, humid weather provided external conditions for insulator surface flashover, which is highly correlated with the warning phenomenon. Therefore, the first identification content is: {Warning items P-001 / P-002 are highly credible, suspected to be a single-phase grounding fault caused by insulation flashover due to humid weather, which is highly consistent with the overall state of the power grid and meteorological conditions}.

[0053] In the stage of determining the content of the second level of anomaly identification, the system analyzes the topology; there are no alarms from upstream node CN-B (F1) and downstream node CN-C (FTU at N4) of trigger node CN-A (F3); under normal load conditions, the downstream load fluctuation is insufficient to cause the zero-sequence current of F3; upstream F1 did not alarm, indicating that the fault point is not upstream of F1; therefore, the content of the second level of identification is: {based on topology logic, downstream load disturbance and upstream fault are ruled out, the fault source is highly localized and located within the section monitored by F3}.

[0054] In the stage of determining the line anomaly identification system, the system integrates the above two levels of identification content to construct the final line anomaly identification system. This system not only includes triggering events and early warning items, but more importantly, it integrates the first identification conclusion: under normal load and humid weather conditions, a high-confidence judgment of a single-phase ground fault caused by insulation flashover. It also integrates the second identification conclusion: based on topology logic, upstream and downstream disturbances are excluded, and the fault source is highly localized in the F3 monitoring section. The final comprehensive assessment is: a highly reliable, localized, concealed single-phase ground fault caused by insulation flashover has been identified, and the fault area is locked in the F3 monitoring section. Through this step, the system elevates multiple independent early warning items into an anomaly identification system that combines event context, global power grid status, and topology logic.

[0055] refer to Figure 5 In step S14, the specific steps are as follows: S141: In the anomaly identification system of this line, the anomaly identification system of this line is dynamically detected, and multiple fault events are identified during the detection process. In each fault event, the corresponding fault characteristics are determined based on the identification of the fault event, so as to collect multiple fault characteristics. S142: Among multiple fault features, the feature location of the fault feature is determined based on the identification of each fault feature. At the same time, the feature shape of the fault feature is determined based on the morphological detection of the fault feature. The first fault range is determined according to the feature location of multiple fault features and the corresponding line control node. S143: Determine the second fault range based on the characteristic morphology of multiple fault features and the corresponding line control node, and determine the corresponding fault area based on the matching of the first fault range and the second fault range; identify the fault area, and determine multiple fault data during the identification process, and determine the corresponding fault waveform data based on the detection of multiple fault data.

[0056] In the embodiments of this application, the anomaly identification system of the line is dynamically detected, and multiple fault events are identified during the detection process. In each fault event, the corresponding fault features are determined based on the identification of the fault event, so as to collect multiple fault features, which takes into account the overall consideration of fault event identification and ensures the accuracy of the corresponding fault features.

[0057] At this point, the system takes the anomaly identification system generated by S13 as an analysis object and initiates a multi-level, dynamic detection process. Based on the comprehensive evaluation conclusion of the anomaly identification system (such as a single-phase grounding suspected to be caused by insulation flashover), the system loads a set of in-depth analysis models specifically for this type of fault from its algorithm library. These models include arc models, traveling wave propagation models, high-frequency signal analysis models, etc., to perform targeted in-depth analysis.

[0058] During in-depth analysis, the system identifies independent fault events with clear physical meaning from complex waveform data. A fault event is a key stage or phenomenon in the fault development process. For example, a typical single-phase ground fault is deconstructed into a pre-breakdown event, an arc initiation event, an arc continuation and extinction event, and a transient traveling wave event. By matching specific waveform patterns and energy mutation points, the system identifies the occurrence time and duration of these events from the waveform data, decomposing the fault process into a series of analyzable sub-events.

[0059] For each identified fault event, the system uses corresponding signal processing techniques to extract a series of fault features that can accurately describe the physical characteristics of the event. These features are more diagnostically valuable than the warning content in S131 and are highly quantified. For example, the system extracts the fault initial phase angle, DC component decay time constant, and estimated fault transition resistance from arc initiation events; it extracts the arrival time of the traveling wave front and the wave front steepness from transient traveling wave events; and it extracts the spectral center frequency of arc discharge from high-frequency signal analysis. The system collects all these features with different physical meanings extracted from different events to form a rich set of fault features.

[0060] Specifically, the system has constructed an anomaly identification system for single-phase grounding caused by insulation flashover; in the dynamic detection anomaly identification system stage, the system loads a deep analysis model package for single-phase grounding faults, especially a sub-model for intermittent arc grounding, and begins to perform millisecond-level and microsecond-level scanning and analysis on the associated waveform file F3_20231027_221533.dat.

[0061] During the analysis, the system identified two key fault events in the waveform data: Fault event E-01, which detected a very steep voltage sag and current surge at time t0, consistent with the typical characteristics of an arc initiation event; Fault event E-02, immediately following E-01, detected another similar abrupt change at time t0+50ms, consistent with the characteristics of an arc reignition event, confirming the intermittent nature of the fault.

[0062] The system extracts high-precision quantitative features for the identified events. For event E-01 (arc initiation event), the system analyzes the voltage waveform at the time of the fault and extracts fault feature F-001: the initial phase angle of the fault is 63 degrees. By analyzing the current-voltage relationship in the fault circuit, the system estimates fault feature F-002: the initial fault transition resistance is approximately 150 ohms. For event E-02 (transient traveling wave event), the system uses high-speed sampling rate waveform recording data to accurately measure the time when the traveling wave front arrives at F3 and extracts fault feature F-003: the initial traveling wave front arrival time is t0+0.15ms. The system also calculates the rise rate of the wave front and extracts fault feature F-004: the current wave front steepness di / dt is approximately 5A / μs. In addition, the system performs spectral analysis on the high-frequency current throughout the fault process and extracts fault feature F-005: the center frequency of the arc discharge spectrum is 8kHz.

[0063] Furthermore, among multiple fault features, the feature location of the fault feature is determined based on the identification of each fault feature. At the same time, the feature shape of the fault feature is determined based on the morphological detection of the fault feature. The first fault range is determined according to the feature locations of multiple fault features and the corresponding line control nodes. This comprehensive consideration of the feature locations of multiple fault features and the corresponding line control nodes ensures the accuracy of the first fault range.

[0064] At this point, the system interprets each fault feature to determine its most indicative physical location. This process depends on the physical meaning of the feature itself and the power grid topology model. For example, for features (such as arc spectrum) directly monitored by a specific line control node (such as F3), the feature location usually points directly to the line segment monitored by that node. For some indirect features, the system will infer based on their properties. Each fault feature will be assigned one or more feature locations, usually represented in the form of line segment ID or node ID.

[0065] The system performs morphological analysis on the waveforms or values ​​of fault characteristics to obtain a qualitative description of their electrical behavior. The analysis includes waveform morphology (whether it is a steep pulse, a gradual decay, or an oscillating waveform), energy morphology (whether the energy is concentrated at low, medium, or high frequencies), and numerical morphology (whether the value is stable, gradually changing, or abruptly changing). Each fault characteristic is assigned a characteristic morphology label, such as high-frequency pulse type, low-frequency decay type, etc.

[0066] The system treats all fault features as independent evidence and overlays them on the topology model. The range determination logic includes: finding spatial intersections; if multiple features point to the same or adjacent segments, then this common area is the most suspicious range; considering evidence weights, as different types of features have different location weights; and performing node association, where the system focuses on analyzing which line control nodes provide the most consistent and directional feature evidence. The system ultimately outputs a first fault range, which is a single segment and a set of several adjacent segments, representing the fault location directly inferred from all current fault features.

[0067] Specifically, the system has collected five fault features (F-001 to F-005) from the fault waveform data. In the stage of determining the location and morphology of the fault features, the system analyzed each feature. Features F-001 (fault initial phase angle 63 degrees) and F-002 (transition resistance 150 ohms) were both monitored by F3, so their feature locations directly point to the segment L3 monitored by F3, and their morphology is numerical. The original data of features F-003 (traveling wave arrival time) and F-004 (wavefront steepness 5A / μs) came from F3, and were initially associated with segment L3, with morphologies of instantaneous pulse type and high-frequency pulse type, respectively. The arc of feature F-005 (arc spectrum center frequency 8kHz) is the fault source, and its spectral characteristics were captured by F3, strongly associated with segment L3, and its morphology is high-frequency oscillation type.

[0068] In the initial fault location determination phase, the system overlays all five features as evidence onto the distribution network topology model. The logical reasoning process is as follows: Features F-001, F-002, and F-005 directly or indirectly point to the area monitored by the line control node CN-A (F3); although features F-003 and F-004 are raw data used for precise location, their source node is also F3, further strengthening the suspicion of the area near F3. All evidence highly consistently points to the same area, with no contradictory evidence pointing to other sections. Based on this highly consistent evidence aggregation, the system determines the initial fault location to be the entire line segment L3 covered by the line control node CN-A (F3). Through this step, the system spatially correlates and aggregates five independent, scattered fault features to form a clear and highly confident preliminary fault location, laying a solid foundation for the next step of using electrical logic for more precise location.

[0069] Therefore, the second fault range is determined based on the characteristic morphology of multiple fault features and the corresponding line control nodes. The corresponding fault area is determined based on the matching of the first fault range and the second fault range. The fault area is identified, and multiple fault data are determined during the identification process. The corresponding fault waveform data is determined based on the detection of multiple fault data. This approach takes into account the overall consideration of matching the first fault range and the second fault range, ensuring the accuracy of the corresponding fault area.

[0070] At this point, the system will select fault features with timing or propagation characteristics (such as traveling waves and transient components) for calculation; the location logic includes: using the time difference of the transient traveling waves generated by the fault reaching different line control nodes, combined with the known line wave impedance, the distance from the fault point to each node can be accurately calculated; or the nature of the fault area can be inferred by analyzing the relationship between the frequency of certain fault modes and the fault loop parameters; through this calculation based on physical laws, the system can obtain an independent, usually more accurate, second fault range.

[0071] The system compares and matches the first fault range obtained in S142 (based on evidence aggregation) with the second fault range obtained in this step (based on electrical logic). The matching logic includes: overlap verification, if the two ranges have overlapping areas in the topology, the confidence of the overlapping area will be greatly improved and it will be identified as the final fault area; conflict resolution, if the two ranges do not overlap or there is a conflict, the system will arbitrate according to the feature weight and calculation accuracy; even if the first range is a long segment, the second range can greatly narrow it down, and finally lock the fault area into a very small range.

[0072] Once the fault area is finally located, the system will immediately execute a data aggregation task. The system will identify all monitoring devices within the fault area and its boundaries, and request all data generated by these devices within the fault occurrence time window. Among all fault data, the system will prioritize and select fault waveform data, which usually refers to waveform files with the highest sampling rate, the most complete information, and the best time synchronization. The system will generate a data package containing time-aligned waveform files from different observation points, providing the most original and comprehensive material for subsequent multi-view, in-depth diagnosis of hidden faults.

[0073] Specifically, the system has determined the first fault range to be the entire L3 section monitored by F3. In the stage of determining the second fault range, the system selects the fault feature F-003 (traveling wave arrival time) with time sequence characteristics extracted in S141. The system finds that the time for the fault traveling wave to arrive at the upstream node F1 is t1, and the time to arrive at F3 is t3. Through high-precision time synchronization, the time difference Δt = t3 - t1 = 0.025ms is calculated. Given that the line wave speed is approximately 300m / μs, according to the traveling wave ranging formula, the system calculates that the distance from the fault point to F1 is 3750m. According to the topology model, the total length from F1 to F3 is 5km. Therefore, the fault point is located between F1 and F3, approximately 3.75km from F1, or approximately 1.25km from F3. The system defines this precise point location as the second fault range.

[0074] The system matches the first fault range (the entire L3 section) with the second fault range (a point within L3 approximately 1.25 km from F3); the second range falls entirely within the first range, and the two are highly consistent with each other, with no conflict; the system ultimately pinpoints the fault area to be: within the L3 section, at a location 1.25 km from the F3 monitoring point.

[0075] The system identified the fault area as section L3 and determined that the monitoring devices directly related to section L3 were upstream device F1 and its own device F3. The system initiated data requests to the FTUs of F1 and F3, retrieving waveform data for five cycles before and after the fault. The system generated a data packet containing two files, aligned with timestamps, which is the final determined fault waveform data: F1_FaultRecord_20231027_221533.dat and F3_FaultRecord_20231027_221533.dat. Through this step, the system successfully transformed a vague abnormal signal into a fault area accurate to the meter level and prepared the most comprehensive multi-point observation data, providing irrefutable, high-quality input for finally determining the root cause of the hidden fault.

[0076] refer to Figure 6In step S15, the specific steps are as follows: S151: Based on real-time monitoring of the fault area, determine multiple data points for the fault area, determine the corresponding data combination based on the matching of multiple data points, determine the fault control framework based on the fault waveform data and the data combination of the fault area, and determine multiple fault nodes based on the matching of the fault control framework and the working history of the distribution network. S152: Based on the identification of each fault node, the node position of the fault node is determined, and the hidden content is determined according to the depth detection of the fault node. At this time, the hidden content is related to the corresponding hidden fault. S153: Determine the first fault coefficient based on the node location of each fault node and the work content of the fault area; determine the second fault coefficient based on the hidden content of each fault node and the work content of the fault area; and determine the hidden faults of the distribution network based on the mapping relationship between the first fault coefficient, the second fault coefficient and the hidden faults.

[0077] In the embodiments of this application, multiple data points of the fault area are determined based on real-time monitoring of the fault area, and corresponding data combinations are determined based on the matching of multiple data points. A fault management framework is determined based on the fault waveform data and the data combination of the fault area, and multiple fault nodes are determined based on the matching of the fault management framework and the working history of the distribution network. This approach is compatible with the overall consideration of the data combination of fault waveform data and the fault area, ensuring the accuracy of the fault management framework.

[0078] At this point, the system uses the fault area determined in S14 as the query scope and comprehensively collects all relevant data within that area through the data interface. The collected data is multi-dimensional, including real-time operating data (such as current, voltage, power, temperature, etc. of all monitoring points in the area), equipment ledger data (such as model, manufacturer, commissioning date, rated parameters, etc. of all equipment in the area), status monitoring data, and environmental data (such as temperature, humidity, wind speed, rainfall, etc. at the time of the fault). The system timestamps and spatially correlates these heterogeneous data to form a structured data combination that corresponds one-to-one with the fault area, providing complete contextual information for subsequent analysis.

[0079] The system takes the high-precision fault waveform data obtained in S14 as the core input and deeply integrates it with the data combination formed in the previous step. This fault management framework includes electrical dynamics (the precise change process of voltage and current in the area before and after the fault occurs), equipment status snapshot (the static attributes and known status of all equipment in the area at the moment the fault occurs), and environmental background (the specific environmental conditions at the time of the fault). This framework is equivalent to a digital twin of the fault moment, comprehensively recording all internal and external conditions of the fault occurrence.

[0080] The system cross-matches the fault management framework with the operational history of the distribution network, which includes historical load curves, historical fault records, inspection and maintenance records, and historical meteorological data. Through matching, the system can identify equipment that exhibits abnormal behavior within the framework or has a strong correlation with historical data. For example, a cable joint that exhibits abnormal temperature rise under low load conditions, or an insulator that is nearing or exceeding its design service life and located in a heavily polluted area. Each identified suspected device with potential defects is defined as a fault node.

[0081] Specifically, the system has pinpointed the fault area to within section L3, 1.25 kilometers from monitoring point F3. During the data combination phase, the system used this precise point as the center and queried the GIS and asset management system, obtaining the following data: the equipment ledger shows this location is pole #57, suspending a set of XP-70 suspension insulators, commissioned in 2008; environmental data indicates the real-time weather at the time of the fault was light rain, humidity 85%, and temperature 22℃. The system combines this data with the real-time load data of section L3 (load rate 40% at the time of the fault) to form a structured data package.

[0082] The system uses the fault recording data of F1 and F3 obtained by S14 as the core and merges it with the data packet from the previous step; the resulting fault management framework describes the following scenario: at 22:15:33, an arc grounding fault occurred on the line where the XP-70 insulator (which has been in operation for 15 years) at pole #57 is located, under an environment of 85% humidity and a load rate of 40%. The fault recording shows typical insulation flashover characteristics.

[0083] The system matched the framework with historical data; matching historical loads confirmed that a 40% load rate was within the normal range, ruling out overload factors; checking maintenance records revealed that the insulators had never been replaced or extensively cleaned since commissioning; matching geographical and meteorological information confirmed that the area belonged to a Class IV pollution zone, and historical data showed that unexplained transient grounding faults had occurred multiple times in similar high-humidity weather conditions; comprehensive comparison revealed that the XP-70 insulators at pole #57, with their 15-year history of not being cleaned, the Class IV pollution zone environment, and a high degree of consistency with the current insulation flashover fault, made them the most suspicious source of the fault; therefore, the system identified the XP-70 insulator string at pole #57 as the fault node identified in this study; through this step, the system successfully focused a physical fault area onto a specific, potentially defective fault node, thus pinpointing the target for the final root cause diagnosis.

[0084] Furthermore, the node position of each fault node is determined based on the identification of each fault node, and the hidden content is determined based on the depth detection of the fault node. At this time, the hidden content is related to the corresponding hidden fault, which is compatible with the overall consideration of the identification of each fault node and ensures the accuracy of the node position of the fault node.

[0085] At this point, the system extracts the precise geographical location (such as GIS coordinates) and topological location (its specific connection relationship and section affiliation in the single-line diagram of the distribution network) from the attribute information of the faulty node; clarifying the node location is to ensure that subsequent in-depth detection and analysis can be accurately associated with specific equipment in the physical world, and to provide a precise target for the final maintenance instructions.

[0086] The system will call a specialized analysis model for this type of fault node to further mine the fault waveform data acquired by S14. The deep detection methods include waveform morphology analysis (identifying fingerprints left by specific physical processes), spectral feature analysis (extracting its frequency domain features), and energy pattern recognition. Through the above deep detection, the system can deduce the physical essence from electrical phenomena. This deduced, hidden defect is the concealed content. For example, for an insulator node, the concealed content is that the composite insulator skirt has undergone electrolytic aging and has basically lost its hydrophobicity.

[0087] The system incorporates an expert knowledge base or a machine learning-based mapping model. This knowledge base defines strong correlations between various hidden contents and the final hidden fault types. For example, the knowledge base contains the rule: IF (hidden contents = loss of hydrophobicity of insulators) AND (environment = high humidity) THEN (hidden fault = pollution flashover). This correlation ensures that the system can not only discover defects but also accurately classify them into a known fault type with clear countermeasures, thus supporting the final decision.

[0088] Specifically, the system has identified the XP-70 insulator string at pole #57 as the fault node. During the node location determination phase, the system queried from the GIS system and determined the precise location of the fault node to be: 116.40°E, 39.90°N, located in the L3 section of the main distribution network, pole #57, phase A. This location information was uniquely determined, providing precise navigation for subsequent maintenance.

[0089] The system retrieves the fault waveform data recorded in F3 of S14 and initiates a deep analysis module for insulator faults. Through waveform morphology analysis, the system discovers tiny, repetitive pulses superimposed on the voltage waveform before the fault, a typical phenomenon of partial discharge on the insulator surface. Through spectral characteristic analysis, the system performs spectral analysis on the fault current, finding that the energy is mainly concentrated in the low-frequency band, but accompanied by obvious and wide-bandwidth harmonic components, which highly matches the spectral characteristics of a stable arc formed by surface flashover. Combining the background information of 15 years of uncleaned area and Class IV pollution zone in S151, the system makes a diagnosis: the insulator... Due to long-term accumulation of dirt on the surface of the insulator string, a conductive water film forms under humid weather, leading to a decrease in insulation level and ultimately causing surface discharge and developing into arc flashover. Therefore, the system identifies the hidden fault as follows: long-term accumulation of industrial pollutants on the surface of the porcelain insulator forms a highly conductive dirt layer under humid conditions, resulting in a sharp decrease in surface insulation strength. In the system's expert knowledge base, there is a clear mapping relationship: decreased surface insulation strength + humid environment is strongly correlated with the hidden fault type of pollution flashover. Therefore, when identifying the hidden fault, the system establishes a strong correlation between it and the final diagnostic result of pollution flashover.

[0090] Therefore, a first fault coefficient is determined based on the node location of each fault node and the working content of the fault area, and a second fault coefficient is determined based on the hidden content of each fault node and the working content of the fault area. The hidden faults of the distribution network are determined based on the mapping relationship between the first fault coefficient, the second fault coefficient, and the hidden faults. This approach takes into account the overall consideration of the mapping relationship between the first fault coefficient, the second fault coefficient, and the hidden faults, ensuring the accuracy of the hidden faults in the distribution network. At the same time, the fault recording data is further controlled, realizing the overall consideration of the node location of each fault node, the corresponding hidden content, and the working content of the fault area, thus improving the accuracy of the hidden faults in the distribution network.

[0091] At this point, the system calculates the spatial distance between the physical location of the fault node and the precise location of the fault area determined in S14; if the fault node is located at the center of the fault area, the first fault coefficient is 1.0 (maximum value); if the node is at a certain distance, the coefficient is reduced according to the distance and the fault impact range; if the node is far from the fault area in the topology, the coefficient is close to 0; the first fault coefficient presents a confidence assessment of the spatial dimension.

[0092] The system calls its built-in expert knowledge base or causal graph, which defines the probability or intensity of what kind of fault phenomenon different hidden content can cause; for example, the knowledge base defines that the probability of flashover caused by contamination on the insulator surface is 0.95; the system assigns a value to the second fault coefficient according to the matching degree between the current fault phenomenon and the hidden content; the second fault coefficient presents a confidence assessment of the causal dimension.

[0093] The system determines hidden faults in the distribution network based on the mapping relationship between the first fault coefficient, the second fault coefficient, and hidden faults. The system calculates a comprehensive confidence score for each fault node, which can be a weighted sum or product of the first and second fault coefficients. The system sorts all fault nodes by their comprehensive scores, and the node with the highest score is identified as the final fault source. Based on the hidden content of the node with the highest score and the mapping relationship in the knowledge base, the system generates a final, structured diagnostic report, i.e., the hidden fault of the distribution network. This report not only indicates the location and cause of the fault but also provides a confidence score, providing strong support for operation and maintenance decisions.

[0094] Specifically, the system has identified the XP-70 insulator string at pole #57 as the fault node and determined that its hidden cause is surface contamination leading to a decrease in insulation strength. In the stage of determining the first fault coefficient, the physical location of the fault node (insulator at pole #57) completely coincides with the precise location of the fault area determined in S14 (1.25 km from F3 in section L3). Since the spatial location is completely consistent and there is no deviation, the system evaluates the first fault coefficient (C1) to be 1.0.

[0095] In the stage of determining the second fault coefficient, the hidden content of the fault node is that surface contamination leads to a decrease in insulation strength, while the working content (electrical phenomenon) of the fault area is the arc grounding shown in the waveform data. The system consulted the expert knowledge base and found that there is a very strong causal relationship between insulator contamination and arc grounding (flashover), and its correlation strength was defined as 0.98. Therefore, the system rated the second fault coefficient (C2) as 0.98.

[0096] In the stage of identifying hidden faults in the distribution network, assuming the system uses a product model to calculate the comprehensive score S=C1XC2, the comprehensive score of this node is S=1.0X0.98=0.98, which is a very high confidence score. Since there is only one fault node in this example, this node wins with the highest score of 0.98. Based on the information of this node, the system generates the final diagnostic report, identifying the hidden fault in the distribution network as follows: Fault location: Phase A of pole #57 in section L3 of the distribution network; Faulty equipment: XP-70 insulator string; Root cause (hidden fault): Pollution flashover occurs on the surface of the insulator due to long-term accumulation of industrial pollution under humid weather conditions; Comprehensive confidence level: 98%.

[0097] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of a hidden fault identification system for a power distribution network according to an embodiment of the present invention; the hidden fault identification system for the power distribution network includes: The segment topology module 21 is used to determine the corresponding line data combination based on the detection of the lines in the distribution network during the operation of the distribution network, and to determine the segment topology relationship based on each line data combination and the line path of the distribution network. The line early warning event module 22 is used to identify multiple line control nodes based on the identification of the topology relationship of the section, mark the status of each line control node, and determine the corresponding line early warning event based on the early warning status of the line control node. The anomaly identification system module 23 is used to determine multiple corresponding line warning items based on the detection of the line warning event, and to determine the line anomaly identification system based on the multiple line warning items, the section topology relationship and the state of the distribution network; The quality level module 24 is used to determine multiple fault features based on the detection of the abnormality identification system of the line, determine the corresponding fault area based on the feature location, corresponding feature shape and corresponding line control node of the multiple fault features, and determine the corresponding fault recording data based on the identification of the fault area. The concealed fault module 25 is used to determine multiple fault nodes based on the fault waveform data, the data combination of the fault area and the working history of the distribution network, and to determine the concealed faults of the distribution network according to the node location of each fault node, the corresponding concealed content and the working content of the fault area.

[0098] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for identifying hidden faults in a power distribution network, characterized in that, include: During the operation of the distribution network, the corresponding line data combination is determined based on the detection of the distribution network lines, and the segment topology relationship is determined according to each line data combination and the line path of the distribution network. Based on the identification of the topology of the section, multiple line control nodes are identified, and the status of each line control node is marked. Based on the warning status of the line control node, the corresponding line warning event is determined. Based on the detection of the line early warning event, multiple corresponding line early warning items are determined, and based on the multiple line early warning items, the section topology relationship and the status of the distribution network, the line anomaly identification system is determined; Based on the detection of the anomaly identification system of the line, multiple fault features are determined. Based on the feature location, corresponding feature shape and corresponding line control node of the multiple fault features, the corresponding fault area is determined. Based on the identification of the fault area, the corresponding fault recording data is determined. Based on the fault recording data, the data combination of the fault area, and the working history of the distribution network, multiple fault nodes are identified. Based on the node location of each fault node, the corresponding hidden content, and the working content of the fault area, the hidden faults of the distribution network are determined.

2. The method for identifying hidden faults in a power distribution network according to claim 1, characterized in that, During the operation of the distribution network, the corresponding line data combinations are determined based on the detection of the distribution network lines, and the segment topology is determined according to each line data combination and the line path of the distribution network, including: The system monitors the operation of the distribution network in real time, identifies multiple lines based on the distribution network map, determines multiple corresponding working data based on the detection of each line, and determines the corresponding line data combination based on the matching of each line and the corresponding multiple working data. Multiple line nodes are identified based on the detection of the distribution network distribution map. The line paths of the distribution network are determined based on the node information and corresponding node locations of the multiple line nodes. The segment topology is determined based on the line paths of the distribution network, the corresponding lines, and the combination of data of each line.

3. The method for identifying hidden faults in a power distribution network according to claim 1, characterized in that, The process of identifying multiple line control nodes based on the topology of the section, marking the status of each line control node, and determining the corresponding line warning event based on the warning status of the line control node includes: The topological relationships of this section are identified, and multiple line control areas are determined during the identification process. Based on the identification of each line control area, the corresponding line control nodes are determined to identify multiple line control nodes. Among multiple line control nodes, the status of each line control node is determined based on the identification of each line control node. If the status of a line control node is an early warning status, the corresponding line early warning identifier is determined based on the tracing of the line control node, and the corresponding line early warning event is determined based on the detection of the line early warning identifier.

4. The method for identifying hidden faults in a power distribution network according to claim 1, characterized in that, The method for determining multiple line early warning items based on the detection of the line early warning event, and determining the line anomaly identification system based on multiple line early warning items, segment topology, and distribution network status, includes: The system monitors the warning events of the line in real time, performs dynamic detection of the warning events, and identifies multiple line warning contents during the detection process. Based on the multiple line warning contents, the corresponding warning areas, and the corresponding line control nodes, the system determines the corresponding multiple line warning items.

5. The method for identifying hidden faults in a power distribution network according to claim 4, characterized in that, The method for determining multiple corresponding line early warning items based on the detection of the line early warning event, and determining the line anomaly identification system based on multiple line early warning items, segment topology, and distribution network status, further includes: Collect multiple operational data from the distribution network, determine the status of the distribution network based on these data, determine the first level of anomaly identification content based on multiple line early warning projects and the status of the distribution network, determine the second level of anomaly identification content based on the segment topology and the status of the distribution network, and determine the line anomaly identification system based on the first and second level of anomaly identification content.

6. The method for identifying hidden faults in a power distribution network according to claim 1, characterized in that, The process involves determining multiple fault features based on the anomaly identification system of the line, identifying corresponding fault regions based on the feature locations, corresponding feature morphologies, and corresponding line control nodes of the multiple fault features, and determining corresponding fault recording data based on the identification of the fault regions, including: In the anomaly identification system of this line, dynamic detection is performed on the anomaly identification system of this line, and multiple fault events are identified during the detection process. In each fault event, the corresponding fault characteristics are determined based on the identification of the fault event, so as to collect multiple fault characteristics.

7. The method for identifying hidden faults in a power distribution network according to claim 6, characterized in that, The method of determining multiple fault features based on the detection of the anomaly identification system of the line, determining the corresponding fault area based on the feature location, corresponding feature shape and corresponding line control node of the multiple fault features, and determining the corresponding fault recording data based on the identification of the fault area, further includes: Among multiple fault features, the feature location of the fault feature is determined based on the identification of each fault feature. At the same time, the feature shape of the fault feature is determined based on the morphological detection of the fault feature. The first fault range is determined according to the feature location of multiple fault features and the corresponding line control node. The second fault range is determined based on the characteristic patterns of multiple fault features and the corresponding line control nodes. The corresponding fault area is determined based on the matching of the first fault range and the second fault range. The fault area is identified, and multiple fault data are determined during the identification process. The corresponding fault waveform data is determined based on the detection of multiple fault data.

8. The method for identifying hidden faults in a power distribution network according to claim 1, characterized in that, The method involves identifying multiple fault nodes based on the fault recording data, the data combination of the fault area, and the working history of the distribution network. Hidden faults in the distribution network are then identified based on the node location of each fault node, its corresponding concealed nature, and the working content of the fault area. These include: Based on real-time monitoring of the fault area, multiple data points for the fault area are determined. The corresponding data combination is determined by matching the multiple data points. The fault control framework is determined based on the fault waveform data and the data combination of the fault area. Multiple fault nodes are determined based on the matching of the fault control framework and the working history of the distribution network.

9. The method for identifying concealed faults in a power distribution network according to claim 8, characterized in that, The method of determining multiple fault nodes based on the fault recording data, the data combination of the fault area, and the working history of the distribution network, and determining the hidden faults of the distribution network based on the node location of each fault node, the corresponding hidden content, and the working content of the fault area, also includes: Based on the identification of each fault node, the node location of the fault node is determined, and the hidden content is determined based on the depth detection of the fault node. At this time, the hidden content is related to the corresponding hidden fault. The first fault coefficient is determined based on the node location of each fault node and the work content of the fault area. The second fault coefficient is determined based on the hidden content of each fault node and the work content of the fault area. The hidden faults of the distribution network are determined based on the mapping relationship between the first fault coefficient, the second fault coefficient and the hidden faults.

10. A system for identifying hidden faults in a power distribution network, characterized in that, The system for identifying hidden faults in the power distribution network is applied to the method for identifying hidden faults in the power distribution network as described in any one of claims 1-9.

Images