A knowledge graph-based method and system for fault diagnosis during live-line work.

By constructing a knowledge graph for live-line work and monitoring topology change operations, calculating topology differences, determining boundary times, and combining on-site monitoring information to create a fault diagnosis subgraph, the problem of difficulty in distinguishing abnormal contact of bypass cable joints in traditional live-line work is solved, and accurate fault diagnosis is achieved.

CN123116111APending Publication Date: 2026-11-17JIANGSU FENGHAI NEW ENERGY ENG TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611639705.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-22
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional live-line maintenance (LNEM) fault diagnosis methods struggle to accurately distinguish whether abnormal contact at bypass cable joints exists before load transfer, occurs after load transfer, or persists across load transfer when performing LNEM bypass work on 10 kV distribution lines. This makes fault diagnosis difficult.

Method used

By constructing a knowledge graph for live-line work, monitoring topology change operations, calculating topology differences, determining boundary times, and combining the operating information of on-site monitoring terminals, a fault diagnosis subgraph is created to filter fault cause nodes associated with abnormal phenomenon nodes and determine the stage of fault occurrence.

Benefits of technology

It enables accurate differentiation of the occurrence time of abnormal contact at bypass cable joints during 10 kV bypass uninterrupted operation, reduces the involvement of live connections at different operation stages in fault diagnosis, and improves the accuracy and reliability of fault diagnosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116111A_ABST
    Figure CN123116111A_ABST
Patent Text Reader

Abstract

This invention relates to the field of fault diagnosis technology, specifically to a knowledge graph-based method and system for fault diagnosis of live-line work (LLO) operations. The invention establishes an LLO knowledge graph according to the LLO task and continuously updates the live connection status based on bypass connections, switch status changes, load transfers, equipment shutdowns, and equipment recovery. It then calculates topology differences and determines boundary times using graph changes, forming corresponding fault diagnosis subgraphs based on these boundary times. When an abnormality in operating current, contact temperature, or leakage current is detected by the field monitoring terminal, the corresponding fault diagnosis subgraph or a group of fault diagnosis subgraphs adjacent to the boundary time is selected using the temporal relationship between the abnormality occurrence time interval and the boundary time. Finally, the fault cause node and fault occurrence stage are determined by combining the fault correlation between the abnormal phenomenon node and the fault cause node, ensuring that the live connection status used for fault diagnosis corresponds to the operation stage at the time of the abnormality.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and in particular to a method and system for fault diagnosis of uninterrupted power supply operations based on knowledge graphs. Background Technology

[0002] The field of fault diagnosis technology mainly involves the identification and analysis of abnormal states of equipment, lines and operations. By collecting operating status information and combining fault phenomena and historical operation information, it can determine the possible fault types, fault locations and fault causes. It is mainly used for equipment operation monitoring, anomaly analysis, fault judgment and maintenance, and is widely used in power operation and maintenance, distribution line maintenance, substation equipment maintenance and live-line work and other related industries.

[0003] Traditional live-line maintenance (LLM) fault diagnosis typically begins with on-site monitoring terminals collecting equipment operating status data. Once abnormalities in current, temperature, or other parameters are detected, the system then combines this data with the existing equipment connections and work records to locate potential faults. While traditional LLM fault diagnosis can pinpoint faults based on equipment operational anomalies, it faces challenges when performing bypass LLM operations on 10kV distribution lines. Operations such as bypass cable connection, load transfer, removal of load switches under maintenance, and equipment restoration can rapidly alter the live connections between equipment. Anomalies may also occur between or near adjacent topology change operations. This makes it difficult to accurately correlate the equipment connection status used for fault diagnosis with the actual live connection status at the time of the anomaly. Consequently, it becomes challenging to distinguish whether the bypass cable joint contact anomaly existed before the load transfer, occurred after the load transfer, or persisted across load transfers. Summary of the Invention

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a method for fault diagnosis of live-line work based on knowledge graphs, comprising the following steps: The live-line work equipment and its connection relationships are obtained from the work tasks corresponding to the live-line work, and a live-line work knowledge graph is constructed. Then, the topology change operations in the live-line work are monitored to update the live-line work knowledge graph. Analyze the changes in the knowledge graph of live-line work to calculate the topological difference. Based on the topological differences and graph changes, determine the boundary time for changes in the graph state of the knowledge graph representing live-line work. Compare the corresponding information of the live-line work knowledge graph before and after the boundary time, then create a fault diagnosis subgraph based on the live-line work knowledge graph, and write the corresponding information that has changed in the comparison results into the fault diagnosis subgraph. The system acquires the operating information of the equipment used for live-line work collected by the on-site monitoring terminal to identify abnormal phenomenon nodes and abnormal occurrence time intervals. It then compares each boundary time with the abnormal occurrence time interval to determine the fault diagnosis sub-graph corresponding to the abnormal phenomenon node or the fault diagnosis sub-graph group adjacent to the boundary time. Based on the fault diagnosis subgraph or the fault diagnosis subgraph group with adjacent boundary times, filter the fault cause nodes associated with the abnormal phenomenon nodes, and determine the fault cause nodes and the fault occurrence stage in the power-off operation.

[0005] As a further aspect of the present invention, the process of constructing a knowledge graph for live-line work is as follows: Each of the aforementioned live-line working equipment is mapped to a corresponding device node in the live-line working knowledge graph, and the device nodes are interconnected according to the device connection relationship; wherein, the live-line working equipment includes the work object equipment, temporary connection equipment, and associated switch equipment; the device connection relationship is the physical connection relationship formed between the various live-line working equipment through lines, joints, or bypass cables.

[0006] As a further aspect of the present invention, the process of updating the knowledge graph for live-line work is as follows: Obtain the operation completion time corresponding to each topology change operation, as well as the switch status and energized status of each device node; determine the interconnected device node pairs by referring to the device connection relationship, and then determine whether there is an energized conduction path between each device node pair based on the switch status and the energized status; In the live-line work knowledge graph, a live connection relationship is established for each pair of equipment nodes. When a live-line path exists between the pairs of equipment nodes, the live connection relationship is determined to be active, and the corresponding operation completion time is recorded as the active time of the live connection relationship. When a live-line path does not exist between the pairs of equipment nodes, the live connection relationship is determined to be inactive, and the corresponding operation completion time is recorded as the inactive time of the live connection relationship. The switch status, the energized status, the energized connection relationship, the effective status, the ineffective status, the effective time and the ineffective time of the energized connection relationship are written into the live-line work knowledge graph to complete the update operation; wherein, the topology change operation includes bypass connection operation, switch status change operation, load transfer operation, equipment exit operation and equipment recovery operation.

[0007] As a further aspect of the present invention, the changes in the graph include the rate of change of newly added relationships, the rate of change of failed relationships, and the rate of change of node states, wherein the calculation process for each rate of change is as follows: The union of the pairs of device nodes whose live connections are active before and after the completion time of each operation is determined as the set of live connected device node pairs, and the total number of device node pairs in this set is counted; the ratio of the number of newly added live connected device node pairs to the total number of device node pairs is calculated to obtain the rate of change of newly added connections; then the ratio of the number of failed live connected device node pairs to the total number is calculated to obtain the rate of change of failed connections. The union of the device nodes that are in a charged state before and after the completion time of each operation is determined as the set of charged state comparison device nodes. The total number of device nodes in this set and the number of device nodes in the set whose charged state has changed are counted and the ratio of the two is calculated to obtain the node state change rate.

[0008] As a further aspect of the present invention, the topological difference is obtained by weighted summation of the rate of change of the newly added relationship, the rate of change of the failed relationship, and the rate of change of the node state.

[0009] As a further aspect of the present invention, the boundary time is determined according to the following conditions: Read the topology difference threshold. When the sum of the number of newly added live connection device node pairs and the number of failed live connection device node pairs is greater than zero, and the topology difference is greater than or equal to the topology difference threshold, the corresponding operation completion time is determined as the boundary time representing the change in the graph state of the live-line work knowledge graph.

[0010] As a further aspect of the present invention, the process of creating the fault diagnosis subgraph is as follows: Filter the device nodes whose energized state changes before and after each of the boundary times, as well as the device node pairs whose energized connection states change, and determine the effective time range of the corresponding energized connection based on the effective time and the failure time; A fault diagnosis subgraph is created based on the selected device nodes and device node pairs, and the corresponding energized state, the state of the energized connection relationship, and the effective time range are determined as the graph information of the fault diagnosis subgraph; then, two adjacent boundary times are determined as the start time and end time of the fault diagnosis subgraph, respectively.

[0011] As a further aspect of the present invention, the operating information includes the operating current, contact temperature, and leakage current of each of the device nodes; Read the operating current alarm threshold, contact temperature alarm threshold, and leakage current alarm threshold; when any of the operating information is greater than or equal to the corresponding alarm threshold, the corresponding operating information is determined as abnormal operating information, and the abnormal operating information is mapped to an abnormal phenomenon node in the live-line operation knowledge graph; The sampling time corresponding to the first occurrence of the abnormal phenomenon node is determined as the abnormal sampling time, the sampling time that is closest to the abnormal sampling time is determined as the preceding sampling time, and the period from the preceding sampling time to the abnormal sampling time is determined as the abnormal occurrence time interval. When the boundary time exists and the boundary time does not exist within the time interval of the anomaly occurrence, the fault diagnosis subgraph whose time range includes the anomaly sampling time is determined as the fault diagnosis subgraph corresponding to the anomaly node; when the boundary time exists within the time interval of the anomaly occurrence, the fault diagnosis subgraphs that end at and begin at the boundary time are determined as the fault diagnosis subgraph before the boundary time and the fault diagnosis subgraph after the boundary time, respectively, and the two are determined as the fault diagnosis subgraph group adjacent to the boundary time.

[0012] As a further aspect of the present invention, the process of determining the fault cause node and the fault occurrence stage in the power-off operation is specifically as follows: The number of abnormal phenomenon nodes corresponding to each candidate fault cause node is counted and the total number of all abnormal phenomenon nodes is calculated. The ratio between the two is used to obtain the fault correlation degree. The fault correlation degree threshold is read, and the candidate fault cause node with the largest fault correlation degree that is greater than or equal to the fault correlation degree threshold is determined as the fault cause node. When determining based on the fault diagnosis subgraph corresponding to the abnormal phenomenon node, the fault occurrence stage is determined according to the start time and end time of the fault diagnosis subgraph; When determining fault diagnosis subgraph groups adjacent to the boundary time, the fault correlation degree of the fault cause node before and after the boundary time is calculated respectively. When the fault correlation degree before the boundary time is less than the fault correlation degree threshold and the fault correlation degree after the boundary time increases to be greater than or equal to the fault correlation degree threshold, it is determined that the fault occurrence stage is located after the boundary time. When the fault correlation degree before and after the boundary time is greater than or equal to the fault correlation degree threshold, it is determined that the fault occurrence stage spans the boundary time.

[0013] A knowledge graph-based fault diagnosis system for live-line working, the system comprising: The knowledge graph construction module obtains the live-line work equipment and equipment connection relationships from the work tasks corresponding to live-line work, and constructs a live-line work knowledge graph; then it monitors the topology change operations in live-line work to update the live-line work knowledge graph. The difference analysis module analyzes the changes in the knowledge graph of live-line work in order to calculate the topological difference. The time determination module, referring to the topological differences and graph changes, determines the boundary time that represents the change in the graph state of the live-line work knowledge graph; The subgraph creation module compares the corresponding information of the live-line work knowledge graph before and after the boundary time, and then creates a fault diagnosis subgraph based on the live-line work knowledge graph, writing the corresponding information that has changed in the comparison results into the fault diagnosis subgraph. The anomaly location module acquires the operating information of the equipment working without power supply collected by the on-site monitoring terminal to determine the nodes of abnormal phenomena and the time intervals of abnormal occurrence. It compares each boundary time with the time intervals of abnormal occurrence to determine the fault diagnosis sub-graph corresponding to the node of abnormal phenomenon or the fault diagnosis sub-graph group adjacent to the boundary time. The fault diagnosis module filters fault cause nodes associated with abnormal phenomenon nodes based on fault diagnosis subgraphs or fault diagnosis subgraph groups with adjacent boundary times, and determines the fault cause nodes and fault occurrence stages in live-line work.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: 1. This invention establishes a knowledge graph for live-line work based on the live-line work tasks, and continuously updates the live connection status according to bypass connection, switch status changes, load transfer, equipment shutdown, and equipment recovery. Then, it uses graph changes to calculate topological differences and determine boundary times, forming corresponding fault diagnosis subgraphs based on the boundary times. After the field monitoring terminal detects abnormal operating current, contact temperature, or leakage current, it uses the time relationship between the abnormal occurrence time interval and the boundary time to select the corresponding fault diagnosis subgraph or a group of fault diagnosis subgraphs adjacent to the boundary time. Then, it combines the fault correlation between the abnormal phenomenon node and the fault cause node to determine the fault cause node and the fault occurrence stage, so that the live connection status used for fault diagnosis can correspond to the work stage when the abnormality occurs. This allows it to distinguish whether the abnormal contact of the bypass cable joint in 10 kV bypass live-line work occurs before the load transfer, after the load transfer, or persists across the load transfer.

[0015] 2. This invention records the active and inactive states of energized connections, along with their corresponding active and inactive times. It organizes only the equipment nodes and pairs that actually change before and after the boundary time into a fault diagnosis subgraph, enabling the differentiation of effective energized connections across different operational stages based on time ranges. Therefore, during continuous topology changes such as bypass cable connection, removal of the load switch under maintenance, and subsequent equipment restoration, it reduces the likelihood of energized connections from different operational stages simultaneously participating in the same fault assessment, ensuring that the selection range of fault cause nodes corresponds to the actual energized topology at that time.

[0016] 3. This invention maps abnormal operating information formed by operating current, contact temperature, and leakage current into abnormal phenomenon nodes, and uses the fault correlation between candidate fault cause nodes and abnormal phenomenon nodes for screening. This allows on-site operational anomalies to participate in fault judgment together with equipment connection changes in the fault diagnosis sub-graph. For abnormal contact temperature caused by abnormal contact of bypass cable joints, the fault cause node can be located by combining the time of anomaly occurrence, the corresponding operation stage, and the correlation between related abnormal phenomena. At the same time, the corresponding fault occurrence stage is given, so that the fault diagnosis results of live-line operation have both fault object and time stage information. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall data flow topology for fault diagnosis of live-line work based on knowledge graphs; Figure 2 A schematic diagram illustrating the mapping of live-line working equipment and their connection relationships to a live-line working knowledge graph; Figure 3 This is a schematic diagram illustrating the changes in the state of the graph and the formation of topological differences before and after the operation is completed. Figure 4 This is a schematic diagram showing the correspondence between boundary time, fault diagnosis sub-graph, and anomaly occurrence time intervals. Figure 5 A schematic diagram illustrating the relationship between abnormal phenomenon nodes, candidate fault cause nodes, and fault occurrence stages. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Please see Figure 1-5 This invention provides a technical solution: a method for fault diagnosis during live-line work based on knowledge graphs, comprising the following steps: S1: Obtain the live-line work equipment and equipment connection relationships from the work tasks corresponding to the live-line work, and construct a live-line work knowledge graph; then monitor the topology change operations in the live-line work to update the live-line work knowledge graph.

[0020] When retrieving live-line work equipment and its connection relationships from the work tasks corresponding to live-line work, the equipment to be worked on, temporary connection equipment, and associated switchgear are read item by item according to the equipment identifiers registered in the work tasks. The equipment to be worked on represents the equipment directly operated during the live-line work; temporary connection equipment represents equipment temporarily connected to the existing electrical connection during the live-line work; and associated switchgear represents equipment that has an electrical connection with the equipment to be worked on or the temporary connection equipment and participates in the connection on / off changes through its switch status. Subsequently, for each piece of live-line work equipment, the connection object and connection type are read, and the equipment identifiers corresponding to the two ends of the connection are combined into a connection record. The connection type is then registered as a line, connector, or bypass cable. Equipment connection relationships represent the physical connection formed between two pieces of live-line work equipment through a line, connector, or bypass cable. When multiple equipment connection relationships exist for the same piece of live-line work equipment, each connection is retained according to its different connection objects. In live-line work on 10 kV distribution lines, the equipment to be worked on is the load switch to be inspected, the temporary connection equipment is the bypass cable, and the associated switchgear is the bypass access side switch and the bypass exit side switch. Finally, the live-line work equipment data and equipment connection relationship data corresponding to the live-line work are obtained.

[0021] When constructing the knowledge graph for live-line work (LLDWork), equipment nodes are established one by one according to equipment identifiers. Each LLDWork equipment corresponds to one equipment node, and each equipment node represents the graph object in the LLDWork knowledge graph that corresponds to the equipment being performed on-site. After the equipment nodes are established, the equipment identifiers corresponding to both ends of each equipment connection relationship are read one by one, and the equipment nodes at both ends of the equipment connection relationship are located. Then, the connection type corresponding to the equipment connection relationship is read. When the connection type is a line, the physical connection relationship formed by the line is registered between the two equipment nodes; when the connection type is a joint, the physical connection relationship formed by the joint is registered between the two equipment nodes; when the connection type is a bypass cable, the physical connection relationship formed by the bypass cable is registered between the two equipment nodes. The equipment node of the load switch to be repaired corresponds to the equipment node of the work target equipment, the equipment node of the bypass cable corresponds to the equipment node of the temporary connection equipment, the equipment nodes of the bypass access side switch and the bypass exit side switch correspond to the equipment nodes of the associated switch equipment, and the actual connection between the bypass cable and the corresponding LLDWork equipment is registered as an equipment connection relationship. Finally, an LLDWork knowledge graph composed of equipment nodes and the equipment connection relationships between equipment nodes is formed.

[0022] When monitoring topology change operations during live-line work, operation records are read in the order of actual completion time for bypass connection operations, switch status change operations, load transfer operations, equipment exit operations, and equipment recovery operations. A bypass connection operation indicates a change in the physical connection corresponding to the bypass cable; a switch status change operation indicates a change in the switch status of the associated switchgear; a load transfer operation indicates a transfer of the energized connection corresponding to the energized load during live-line work; an equipment exit operation indicates that the equipment being worked on during live-line work has exited the current working connection; and an equipment recovery operation indicates that the equipment that has exited live-line work has re-entered the current working connection. After each topology change operation is completed, the operation completion time is read, and this completion time is used as the time position of the current topology status. Then, the switch status and energized status of the corresponding equipment node are read. Records are created corresponding to the operation completion time, equipment node, switch status, and energized status at the same time position, ultimately yielding the operation completion time for each topology change operation and the switch status and energized status of each equipment node.

[0023] When determining interconnected device node pairs based on device connection relationships, the device connection relationships in the live-line work knowledge graph are read one by one. For each device connection relationship read, the device nodes corresponding to both ends of the device connection relationship are obtained, and two device nodes physically connected by lines, connectors, or bypass cables are grouped into a device node pair. Subsequently, for each device node pair, the switch status and energized status corresponding to the current operation completion time are read. The switch status indicates the on / off state of the associated switch equipment, and the energized status indicates whether the device node is currently energized. During the judgment process, the physical connection is checked item by item from one device node to another along the device connection relationship to check whether the physical connection is continuous, then the switch status of the associated switch equipment involved in the connection on / off is checked, and then the energized status of the device node is checked. When the switch status and energized status together form a continuous energized connection, the corresponding device node pair is recorded as having an energized conductive path; when the switch status and energized status cannot together form a continuous energized connection, the corresponding device node pair is recorded as not having an energized conductive path. After the bypass cable is connected and the bypass access side switch and bypass exit side switch change their switch states, the corresponding switch states and energized states are checked item by item along the equipment connection relationship corresponding to the bypass cable. Finally, the energized conduction path judgment results of each equipment node at the corresponding operation completion time are obtained.

[0024] When establishing the live connection relationships between equipment node pairs in the live-line work knowledge graph, each live connection relationship is associated with the corresponding equipment node pair. A live connection relationship represents the live conduction state of two equipment nodes with a connection relationship at the corresponding operation completion time. When a live conduction path exists between equipment node pairs, the corresponding live connection relationship is registered as active, and the operation completion time that generated the current live conduction path state is read and written into the active time of the live connection relationship. When no live conduction path exists between equipment node pairs, the corresponding live connection relationship is registered as inactive, and the operation completion time that generated the current inactive state is read and written into the inactive time of the live connection relationship. After a live conduction path is formed between the equipment node pairs corresponding to bypass cables, the corresponding active time is registered. When subsequent load transfer, equipment removal, or equipment recovery causes a change in the corresponding live conduction path, the corresponding state and time are registered according to the changed live conduction path, ultimately obtaining the corresponding live connection relationship, active state, inactive state, active time, and inactive time for each equipment node pair.

[0025] When updating the live-line work knowledge graph, the operation completion time is used as the time position for writing the graph status. The device nodes corresponding to each operation completion time are located one by one, and their switch status and energized status are written to the corresponding device nodes. Then, according to the device connection relationship, the corresponding device node pairs are located, and their energized connection relationship, active status, inactive status, active time, and inactive time are written to the corresponding device node pairs. Subsequently, when new bypass connection operations, switch status change operations, load transfer operations, device exit operations, or device recovery operations occur, the new operation completion time and the corresponding switch status and energized status of the device nodes are read. The device node pair determination, energized conduction path judgment, and energized connection relationship status writing are then re-executed, ensuring that the device node status and energized connection relationship status corresponding to each operation completion time are saved in chronological order. Finally, the live-line work knowledge graph is updated with the topology change operations during live-line work.

[0026] S2: Analyze the changes in the knowledge graph of live-line work in order to calculate the topological difference.

[0027] When analyzing the changes in the knowledge graph of live-line working (LLM) operations, each operation completion time is used as the time position for comparing the graph state before and after the operation completion time. The energized connection relationships and equipment node energized states before the operation completion time are read, and then the energized connection relationships and equipment node energized states after the operation completion time are read. The energized connection relationships are aligned according to the equipment node pairs formed by the two ends of the equipment node, so that the energized connection relationship states of the same equipment node pair before and after the operation completion time form a set of comparison data. Similarly, the equipment node energized states are aligned according to the equipment node itself, so that the energized states of the same equipment node before and after the operation completion time form a set of comparison data. After completing the alignment of all equipment node pairs and equipment nodes, the data on changes in energized connection relationships and equipment node energized states are generated, ultimately yielding the comparison data of the graph state before and after the operation completion time for the rates of change in new relationships, changes in failed relationships, and changes in node states.

[0028] When calculating the rate of change of newly added relationships and the rate of change of failed relationships, first extract all device node pairs whose energized connections were in effect before the operation completion time, then extract all device node pairs whose energized connections were in effect after the operation completion time. Perform a union operation on the two sets of device node pairs before and after the operation completion time. If the same device node pair exists in both sets before and after the operation completion time, it is retained only once in the set of energized connected device node pairs. Then, count the number of different device node pairs in each set of energized connected device node pairs, and use the counted number as the total number of device node pairs.

[0029] Subsequently, the state of the energized connection relationships of each device node pair in the set of energized connection device node pairs is compared before and after the operation completion time. If a device node pair was not active before the operation completion time but became active after the operation completion time, it is counted in the number of newly added energized connection device node pairs. If a device node pair was active before the operation completion time but not active after the operation completion time, it is counted in the number of failed energized connection device node pairs. If a device node pair was active both before and after the operation completion time, it is not counted in the number of newly added or failed energized connection device node pairs. The ratio of the number of newly added energized connection device node pairs to the total number of device node pairs is calculated as the rate of change for newly added relationships. The ratio of the number of failed energized connection device node pairs to the total number of device node pairs is then calculated as the rate of change for failed relationships. Finally, the rates of change for newly added and failed relationships at the corresponding operation completion time are obtained.

[0030] When calculating the node state change rate, first extract all device nodes that were in the energized state before the operation completion time, then extract all device nodes that were in the energized state after the operation completion time. Perform a union operation on the two sets of device nodes before and after the operation completion time. If the same device node exists in both sets of device nodes before and after the operation completion time, it is only retained once in the set of device nodes for energized state comparison. Then, count the number of different device nodes in the set of device nodes for energized state comparison one by one, and use the counted number as the total number of device nodes in the set of device nodes for energized state comparison.

[0031] Subsequently, the energized states of each device node in the energized state comparison device node set before and after the operation completion time are compared one by one. When a device node changes from an energized state to a de-energized state or vice versa, the corresponding device node is counted in the number of device nodes whose energized state has changed; if the energized state of a device node does not change before and after the operation completion time, it is not counted in the number of device nodes whose energized state has changed. The ratio of the number of device nodes whose energized state has changed to the total number of device nodes in the energized state comparison device node set is calculated as the node state change rate, and finally, the node state change rate for the corresponding operation completion time is obtained.

[0032] When setting the weights for the rates of change in new relationships, changes in failed relationships, and changes in node states, set the weights for each of these rates separately, and ensure that the sum of these weights is 1. Each weight represents the relative proportion of the corresponding graph change in the total topological differences. With the sum of the three weights set to 1, the rates of change in new relationships, changes in failed relationships, and changes in node states collectively correspond to a complete weighted ratio.

[0033] The weights corresponding to the rate of change of newly added relationships are set according to the importance of the newly added live connection relationships to the determination of topology changes in live-line work. The more important the newly added live connection relationships generated by bypass connection operations or equipment restoration operations are as the main content for judging the current state change of the live-line work knowledge graph, the greater the weight of the rate of change of newly added relationships should be.

[0034] The weights corresponding to the failure relationship change rate are set according to the importance of the failed live connection relationship to the determination of topology changes in live-line work. The more important the failed live connection relationship caused by load transfer operation or equipment detachment operation is as a major judgment content of the current live-line work knowledge graph state change, the greater the weight of the failure relationship change rate is set.

[0035] The weights corresponding to the node state change rate are set according to the importance of the change in the energized state of the equipment node to the determination of state changes in the live-line work knowledge graph. The more important it is for an equipment node to change from an energized state to a de-energized state or vice versa as a primary judgment content of the current live-line work knowledge graph state change, the greater the weight of the node state change rate should be.

[0036] After setting one weight, two more weights are set based on the constraint that the sum of the three weights is 1. When one weight increases, another weight or two other weights are decreased while maintaining the sum of 1. The live connection of the bypass cable and the removal of the load switch under maintenance directly cause the addition and failure of live connection relationships, respectively. Therefore, the weights corresponding to the rate of change of the new relationship and the rate of change of the failure relationship can be set to the same or similar values, and the weight corresponding to the rate of change of the node state is set as the remaining weight. The three weights satisfy: ;in, This indicates the weight corresponding to the rate of change of the newly added relationship; The weights corresponding to the rate of change of the failure relationship; The weights corresponding to the node state change rate are represented, and the final weights corresponding to the new relationship change rate, the failed relationship change rate, and the node state change rate are obtained.

[0037] When calculating the topology difference, the rates of change for new relationships, failed relationships, and node states corresponding to the same operation completion time are read. Then, the weights corresponding to the rates of change for new relationships, failed relationships, and node states are read. These three rates are then weighted separately with their corresponding weights, and finally, the three weighted values ​​are summed. ;in, This represents the topology difference corresponding to the current operation completion time. This indicates the rate of change of the newly added relationship corresponding to the current operation completion time; This indicates the weight corresponding to the rate of change of the newly added relationship; This represents the rate of change of the failure relationship corresponding to the current operation completion time; The weights corresponding to the rate of change of the failure relationship; This represents the rate of change of the node state corresponding to the current operation completion time. This represents the weight corresponding to the node state change rate. When a bypass cable is connected to form a new energized connection, or a load switch under maintenance is disconnected to form a failed energized connection, accompanied by changes in the energized state of equipment nodes, the three types of graph changes are substituted into the corresponding weighting terms to finally obtain the topology difference corresponding to the completion time of each operation.

[0038] S3: Based on the topological differences and graph changes, determine the boundary time when the graph state of the knowledge graph representing live-line work changes.

[0039] When determining the boundary time, the topology difference threshold is first read. This threshold represents the comparison value used to determine whether the topology difference corresponding to the current operation completion time reaches the boundary condition for a change in the graph state. Then, for each operation completion time, the number of newly added energized connection pairs and the number of failed energized connection pairs are read. The sum of these is calculated, and then the topology difference corresponding to the same operation completion time is read. When the sum of the number of newly added and failed energized connection pairs is greater than 0, and the topology difference is greater than or equal to the topology difference threshold, the corresponding operation completion time is recorded as the time position where the live-line work knowledge graph state changes. After a change in energized connection is caused by the completion of energized connection of a bypass cable or the removal of a load switch under maintenance, a threshold comparison is performed on the topology difference for the corresponding operation completion time to finally obtain the boundary time representing the change in the live-line work knowledge graph state.

[0040] S4: Compare the corresponding information of the live-line work knowledge graph before and after the boundary time, then create a fault diagnosis subgraph based on the live-line work knowledge graph, and write the corresponding information that has changed in the comparison results into the fault diagnosis subgraph.

[0041] When comparing the corresponding information of the live-line work knowledge graph before and after the boundary time, for each boundary time, the energized state and energized connection relationship of the corresponding equipment node before the boundary time are read, and then the energized state and energized connection relationship of the corresponding equipment node after the boundary time are read. The energized state of each equipment node is aligned with the energized state before and after the boundary time. When the energized state of a device node changes before and after the boundary time, the corresponding device node is recorded in the set of device nodes whose energized state has changed. Similarly, the energized connection relationship of each equipment node is aligned with the energized state before and after the boundary time. When the corresponding energized connection relationship of a device node changes between an active and inactive state, the corresponding device node is recorded in the set of device node pairs whose energized connection relationship has changed. The states of the equipment nodes before and after the bypass cable is connected, and the equipment nodes and device node pairs before and after the load switch to be maintained is checked, ultimately yielding the equipment nodes whose energized state has changed and the device node pairs whose energized connection relationship has changed before and after the boundary time.

[0042] When determining the effective time range of a live connection, for each device node whose live connection status changes, the effective time and ineffective time of the corresponding live connection are read. The effective time represents the operation completion time corresponding to the live connection entering the effective state, and the ineffective time represents the operation completion time corresponding to the live connection entering the ineffective state. The effective and ineffective times of the same live connection are arranged chronologically. Starting from an effective time, the status records of the same live connection are read sequentially, and the ineffective time corresponding to the subsequent entry into the ineffective state is obtained. The effective time is used as the starting position of an effective time range, and the corresponding ineffective time is used as the ending position of an effective time range. When the same live connection subsequently enters the effective state again, a new effective time range is formed by combining the new effective time with the subsequent ineffective time. The live connection corresponding to a bypass cable becomes effective from the operation completion time formed by the bypass live connection and becomes ineffective when the bypass live connection exits the corresponding operation completion time, ultimately obtaining the effective time range for each live connection.

[0043] When creating a fault diagnosis subgraph, device nodes whose energized states change are designated as device nodes in the subgraph, and pairs of device nodes whose energized connection states change are designated as device node pairs. The energized states of the selected device nodes, the energized connection states of the selected device node pairs, and the determined effective time ranges are then read and associated with the corresponding device nodes or device node pairs. The fault diagnosis subgraph represents a local graph extracted from the live-line work knowledge graph based on changes before and after the boundary time. The graph information in the fault diagnosis subgraph consists of the selected device nodes, device node pairs, energized states, energized connection states, and effective time ranges. State change information directly corresponding to bypass cable access, load transfer, and the removal of load switches under maintenance is entered into the corresponding fault diagnosis subgraph, ultimately resulting in a fault diagnosis subgraph corresponding to the state changes at the boundary time.

[0044] When determining the start and end times of the fault diagnosis subgraph, all boundary times are arranged in chronological order, and then adjacent boundary times are read sequentially according to the arrangement order. For the same group of adjacent boundary times, the boundary time with the earlier time position is recorded as the start time of the fault diagnosis subgraph, and the boundary time with the later time position is recorded as the end time of the fault diagnosis subgraph. Then, the device nodes, device node pairs, energized states, energized connection relationships, and effective time ranges in the fault diagnosis subgraph are associated with the time range formed by the corresponding start time to end time, finally obtaining a fault diagnosis subgraph with start and end times.

[0045] S5: Obtain the operating information of the equipment for uninterrupted power supply from the on-site monitoring terminal to determine the abnormal phenomenon nodes and the time intervals of abnormal occurrence. Compare each boundary time with the time intervals of abnormal occurrence to determine the fault diagnosis sub-graph corresponding to the abnormal phenomenon node or the fault diagnosis sub-graph group adjacent to the boundary time.

[0046] When acquiring operational information of live-line working equipment (LLDAP) from the field monitoring terminal, the terminal receives operating current, contact temperature, and leakage current data for the corresponding LLDAP equipment. The operating current is collected by the current sensor of the corresponding LLDAP equipment, generating a value; the contact temperature is collected by a temperature sensor located at the contact point, generating a value; and the leakage current is collected by the leakage current sensor, generating a value. The field monitoring terminal simultaneously records the sampling time when receiving operating current, contact temperature, and leakage current. Based on the correspondence between LLDAP equipment and equipment nodes, the terminal associates the operating current, contact temperature, leakage current, and sampling time with the corresponding equipment node, and then arranges the operational information for the same equipment node in chronological order of sampling time. The terminal focuses on reading the operating current, contact temperature, and leakage current corresponding to bypass cables and bypass cable joints to ultimately obtain the operational information for each equipment node.

[0047] When identifying abnormal phenomenon nodes, the operating current alarm threshold, contact temperature alarm threshold, and leakage current alarm threshold are read separately. Then, the operating current, contact temperature, and leakage current are read again for the same equipment node and the same sampling time. The operating current is compared with the operating current alarm threshold, the contact temperature is compared with the contact temperature alarm threshold, and the leakage current is compared with the leakage current alarm threshold. When any of the operating current, contact temperature, and leakage current is greater than or equal to the corresponding alarm threshold, the operating information meeting the corresponding alarm condition is recorded as abnormal operating information. The corresponding equipment node and sampling time are then read, and the corresponding equipment node is located in the live-line working knowledge graph. The abnormal operating information is then mapped to the abnormal phenomenon node associated with the corresponding equipment node. After a contact abnormality occurs at a bypass cable joint, when the contact temperature corresponding to the bypass cable joint reaches the contact temperature alarm threshold, the corresponding contact temperature is recorded as abnormal operating information, ultimately obtaining the abnormal phenomenon node in the live-line working knowledge graph.

[0048] When determining the time interval for anomalies, the sampling records corresponding to the same device node are arranged from earliest to latest according to the sampling time. Then, the position of the first occurrence of the anomaly node is checked one by one along the sampling time sequence, and the sampling time corresponding to the first occurrence of the anomaly node is recorded as the anomaly sampling time. Subsequently, all sampling times before the anomaly sampling time are read, and the time difference between the anomaly sampling time and each previous sampling time is calculated. Then, each time difference is compared one by one, and the sampling time with the smallest time difference is closest to the anomaly sampling time in time sequence. This sampling time is recorded as the preceding sampling time. The preceding sampling time is used as the start time of the anomaly time interval, and the anomaly sampling time is used as the end time of the anomaly time interval. The sampling time corresponding to the first time the contact temperature of the bypass cable joint reaches the contact temperature alarm threshold is used as the anomaly sampling time. Then, the sampling time closest to the previous time is located, and finally, the anomaly time interval corresponding to the anomaly node is obtained.

[0049] When comparing boundary times with the anomaly occurrence time interval, each determined boundary time is read sequentially, and its position is compared with the preceding sampling time and the anomaly sampling time to check if a boundary time exists within the anomaly occurrence time interval. If a boundary time already exists but no boundary time exists within the anomaly occurrence time interval, the start and end times of each created fault diagnosis subgraph are read sequentially. The anomaly sampling time is compared with the time range corresponding to each fault diagnosis subgraph, and then fault diagnosis subgraphs whose time range covers the anomaly sampling time are selected. If the abnormal contact temperature of the bypass cable joint occurs between two boundary times, and there is no new boundary time between the preceding sampling time and the anomaly sampling time, the fault diagnosis subgraph covering the anomaly sampling time is selected, ultimately obtaining the fault diagnosis subgraph corresponding to the anomaly node.

[0050] When a boundary time exists within the anomaly occurrence time interval, the boundary time within the anomaly occurrence time interval is used as the time position for fault diagnosis subgraph retrieval. Among the already created fault diagnosis subgraphs, first, the subgraph with the same end time as the boundary time is searched, then the subgraph with the same start time as the boundary time is searched. The subgraph with the same end time as the boundary time is recorded as the fault diagnosis subgraph before the boundary time, and the subgraph with the same start time as the boundary time is recorded as the fault diagnosis subgraph after the boundary time. Then, a correspondence is established between the fault diagnosis subgraphs before and after the boundary time according to the same boundary time. When the anomaly occurrence time interval of the bypass cable joint crosses the boundary time corresponding to the load transfer, the fault diagnosis subgraphs before and after the load transfer are read separately, ultimately obtaining a group of fault diagnosis subgraphs with adjacent boundary times.

[0051] S6: Based on the fault diagnosis subgraph or the fault diagnosis subgraph group with adjacent boundary time, filter the fault cause nodes associated with the abnormal phenomenon nodes, and determine the fault cause nodes and the fault occurrence stage in the power-off operation.

[0052] When filtering fault cause nodes associated with abnormal phenomenon nodes, each abnormal phenomenon node is read one by one in the fault diagnosis subgraph corresponding to the abnormal phenomenon node. Then, the association relationships of the abnormal phenomenon nodes in the live-line work knowledge graph are read, and the corresponding fault cause node at the other end of the association relationship is searched. The association relationship represents the graph connection formed between abnormal phenomenon nodes and fault cause nodes, and the fault cause node represents the graph node that has a fault cause association with the abnormal phenomenon node. When the same fault cause node has association relationships with multiple abnormal phenomenon nodes, the multiple abnormal phenomenon nodes are grouped together to the same fault cause node, and the corresponding fault cause node is recorded as a candidate fault cause node.

[0053] When filtering based on fault diagnosis subgraph groups with adjacent boundary times, abnormal phenomenon nodes are read from the fault diagnosis subgraphs before and after the boundary time, respectively. Then, the corresponding fault cause nodes are searched according to the same association relationship, and the association records before and after the boundary time are saved. The abnormal phenomenon node of the contact temperature corresponding to the bypass cable joint is associated with the fault cause node corresponding to the abnormal contact of the bypass cable joint, and finally, the candidate fault cause nodes associated with the abnormal phenomenon nodes are obtained.

[0054] When calculating the fault correlation degree, for each candidate fault cause node, each associated abnormal node is read one by one. The number of abnormal nodes corresponding to each candidate fault cause node is accumulated, and then the total number of all abnormal nodes within the same comparison range is counted. An abnormal node associated with the same candidate fault cause node is counted only once. When calculating based on the fault diagnosis subgraph corresponding to the abnormal node, both the number of abnormal nodes corresponding to the candidate fault cause node and the total number of all abnormal nodes are counted from the same fault diagnosis subgraph. When calculating based on fault diagnosis subgraph groups with adjacent boundary times, the fault diagnosis subgraphs before and after the boundary time are counted separately. The ratio of the number of abnormal nodes corresponding to the candidate fault cause node to the total number of all abnormal nodes is calculated, and the resulting ratio is recorded as the fault correlation degree of the corresponding candidate fault cause node. Finally, the fault correlation degree corresponding to each candidate fault cause node is obtained.

[0055] When determining the fault cause node, the fault correlation threshold is read, and the fault correlation degree corresponding to each candidate fault cause node is compared with the fault correlation threshold. First, candidate fault cause nodes with a fault correlation degree greater than or equal to the fault correlation threshold are retained. Then, among the candidate fault cause nodes that meet the fault correlation threshold condition, the fault correlation degree values ​​are compared one by one, and the largest fault correlation degree is read. The candidate fault cause node corresponding to the largest fault correlation degree is then located. When the fault correlation degree formed between the candidate fault cause node corresponding to the abnormal contact of the bypass cable joint and the abnormal phenomenon node in the current fault diagnosis subgraph reaches the fault correlation threshold and is higher than the fault correlation degree corresponding to other candidate fault cause nodes, the fault cause node in the live-line operation is finally obtained.

[0056] When determining the fault occurrence stage based on the fault diagnosis subgraph corresponding to the abnormal phenomenon node, the start and end times recorded in the fault diagnosis subgraph corresponding to the abnormal phenomenon node are read. The start and end times are then verified to correspond to the same fault diagnosis subgraph. The start time is taken as the start position of the corresponding time range, and the end time is taken as the end position of the corresponding time range. A correspondence is established between the identified fault cause node and the time range formed by the start and end times. When the abnormal occurrence time interval corresponding to the bypass cable joint contact abnormality does not cross the boundary time, the start and end times corresponding to the fault diagnosis subgraph where the abnormal sampling time is located are used to finally obtain the fault occurrence stage.

[0057] When determining the fault occurrence stage based on fault diagnosis subgraph groups with adjacent boundary times, firstly, all abnormal phenomenon nodes in the fault diagnosis subgraph before the boundary time are read. The number of abnormal phenomenon nodes corresponding to the fault cause node is counted, along with the total number of all abnormal phenomenon nodes before the boundary time. The ratio of the number of abnormal phenomenon nodes corresponding to the fault cause node to the total number of all abnormal phenomenon nodes before the boundary time is calculated, and this ratio is recorded as the fault correlation degree before the boundary time. Subsequently, all abnormal phenomenon nodes in the fault diagnosis subgraph after the boundary time are read. The number of abnormal phenomenon nodes corresponding to the fault cause node is counted, along with the total number of all abnormal phenomenon nodes after the boundary time. The ratio of the number of abnormal phenomenon nodes corresponding to the fault cause node to the total number of all abnormal phenomenon nodes after the boundary time is calculated, and this ratio is recorded as the fault correlation degree after the boundary time.

[0058] After calculating the fault correlation degree before and after the boundary time, the fault correlation degree before and after the boundary time are compared with the fault correlation degree threshold, and the change in fault correlation degree before and after the boundary time is further compared. If the fault correlation degree before the boundary time is less than the fault correlation degree threshold, and the fault correlation degree after the boundary time increases to be greater than or equal to the fault correlation degree threshold relative to the fault correlation degree before the boundary time, the fault occurrence stage is recorded as after the boundary time. If both the fault correlation degree before and after the boundary time are greater than or equal to the fault correlation degree threshold, the fault occurrence stage is recorded as crossing the boundary time. If the abnormal contact of the bypass cable joint only causes the corresponding fault correlation degree to reach the fault correlation degree threshold after the load transfer is completed, the period after the boundary time corresponding to the load transfer is determined as the fault occurrence stage. If the abnormal contact of the bypass cable joint maintains a correlation state that reaches the fault correlation degree threshold before and after the boundary time corresponding to the load transfer, the fault occurrence stage is determined as crossing the boundary time.

[0059] A knowledge graph-based fault diagnosis system for live-line working, the system comprising: The knowledge graph construction module obtains the live-line work equipment and equipment connection relationships from the work tasks corresponding to live-line work, and constructs a live-line work knowledge graph; then it monitors the topology change operations in live-line work to update the live-line work knowledge graph. The difference analysis module analyzes the changes in the knowledge graph of live-line work in order to calculate the topological difference. The time determination module, referring to the topological differences and graph changes, determines the boundary time that represents the change in the graph state of the live-line work knowledge graph; The subgraph creation module compares the corresponding information of the live-line work knowledge graph before and after the boundary time, and then creates a fault diagnosis subgraph based on the live-line work knowledge graph, writing the corresponding information that has changed in the comparison results into the fault diagnosis subgraph. The anomaly location module acquires the operating information of the equipment working without power supply collected by the on-site monitoring terminal to determine the nodes of abnormal phenomena and the time intervals of abnormal occurrence. It compares each boundary time with the time intervals of abnormal occurrence to determine the fault diagnosis sub-graph corresponding to the node of abnormal phenomenon or the fault diagnosis sub-graph group adjacent to the boundary time. The fault diagnosis module filters fault cause nodes associated with abnormal phenomenon nodes based on fault diagnosis subgraphs or fault diagnosis subgraph groups with adjacent boundary times, and determines the fault cause nodes and fault occurrence stages in live-line work.

[0060] The knowledge graph-based fault diagnosis system for live-line work provided by this invention is used to execute the above-mentioned knowledge graph-based fault diagnosis method for live-line work, and has the same or corresponding technical features and effects. In this system, the graph construction module is used to execute step S1, the difference analysis module is used to execute step S2, the time determination module is used to execute step S3, the subgraph creation module is used to execute step S4, the anomaly location module is used to execute step S5, and the fault diagnosis module is used to execute step S6.

[0061] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for fault diagnosis of live-line working based on knowledge graph, characterized in that, Includes the following steps: The live-line work equipment and its connection relationships are obtained from the work tasks corresponding to the live-line work, and a live-line work knowledge graph is constructed. Then, the topology change operations in the live-line work are monitored to update the live-line work knowledge graph. Analyze the changes in the knowledge graph of live-line work to calculate the topological difference. Based on the topological differences and graph changes, determine the boundary time for changes in the graph state of the knowledge graph representing live-line work. Compare the corresponding information of the live-line work knowledge graph before and after the boundary time, then create a fault diagnosis subgraph based on the live-line work knowledge graph, and write the corresponding information that has changed in the comparison results into the fault diagnosis subgraph. The system acquires the operating information of the equipment used for live-line work collected by the on-site monitoring terminal to identify abnormal phenomenon nodes and abnormal occurrence time intervals. It then compares each boundary time with the abnormal occurrence time interval to determine the fault diagnosis sub-graph corresponding to the abnormal phenomenon node or the fault diagnosis sub-graph group adjacent to the boundary time. Based on the fault diagnosis subgraph or the fault diagnosis subgraph group with adjacent boundary times, filter the fault cause nodes associated with the abnormal phenomenon nodes, and determine the fault cause nodes and the fault occurrence stage in the power-off operation.

2. The knowledge graph-based fault diagnosis method for live-line work as described in claim 1, characterized in that, The specific process of constructing a knowledge graph for live-line working is as follows: Each of the aforementioned live-line working equipment is mapped to a corresponding device node in the live-line working knowledge graph, and the device nodes are interconnected according to the device connection relationship; wherein, the live-line working equipment includes the work object equipment, temporary connection equipment, and associated switch equipment; the device connection relationship is the physical connection relationship formed between the various live-line working equipment through lines, joints, or bypass cables.

3. The knowledge graph-based fault diagnosis method for live-line work as described in claim 2, characterized in that, The specific process of updating the knowledge graph for live-line work is as follows: Obtain the operation completion time corresponding to each topology change operation, as well as the switch status and energized status of each device node; determine the interconnected device node pairs by referring to the device connection relationship, and then determine whether there is an energized conduction path between each device node pair based on the switch status and the energized status; In the live-line work knowledge graph, a live connection relationship is established for each pair of equipment nodes. When a live-line path exists between the pairs of equipment nodes, the live connection relationship is determined to be active, and the corresponding operation completion time is recorded as the active time of the live connection relationship. When a live-line path does not exist between the pairs of equipment nodes, the live connection relationship is determined to be inactive, and the corresponding operation completion time is recorded as the inactive time of the live connection relationship. The switch status, the energized status, the energized connection relationship, the effective status, the ineffective status, the effective time and the ineffective time of the energized connection relationship are written into the live-line work knowledge graph to complete the update operation; wherein, the topology change operation includes bypass connection operation, switch status change operation, load transfer operation, equipment exit operation and equipment recovery operation.

4. The knowledge graph-based fault diagnosis method for live-line working as described in claim 3, characterized in that, The changes in the graph include the rate of change of newly added relationships, the rate of change of failed relationships, and the rate of change of node states. The calculation process for each rate of change is as follows: The union of the pairs of device nodes whose live connections are active before and after the completion time of each operation is determined as the set of live connected device node pairs, and the total number of device node pairs in this set is counted; the ratio of the number of newly added live connected device node pairs to the total number of device node pairs is calculated to obtain the rate of change of newly added connections; then the ratio of the number of failed live connected device node pairs to the total number is calculated to obtain the rate of change of failed connections. The union of the device nodes that are in a charged state before and after the completion time of each operation is determined as the set of charged state comparison device nodes. The total number of device nodes in this set and the number of device nodes in the set whose charged state has changed are counted and the ratio of the two is calculated to obtain the node state change rate.

5. The knowledge graph-based fault diagnosis method for live-line work as described in claim 4, characterized in that, The topological difference is obtained by weighted summation of the rate of change of the newly added relationship, the rate of change of the failed relationship, and the rate of change of the node state.

6. The knowledge graph-based fault diagnosis method for live-line work as described in claim 4, characterized in that, The boundary time is determined according to the following conditions: Read the topology difference threshold. When the sum of the number of newly added live connection device node pairs and the number of failed live connection device node pairs is greater than zero, and the topology difference is greater than or equal to the topology difference threshold, the corresponding operation completion time is determined as the boundary time representing the change in the graph state of the live-line work knowledge graph.

7. The knowledge graph-based fault diagnosis method for live-line work as described in claim 6, characterized in that, The process of creating the fault diagnosis subgraph is as follows: Filter the device nodes whose energized state changes before and after each of the boundary times, as well as the device node pairs whose energized connection states change, and determine the effective time range of the corresponding energized connection based on the effective time and the failure time; A fault diagnosis subgraph is created based on the selected device nodes and device node pairs, and the corresponding energized state, the state of the energized connection relationship, and the effective time range are determined as the graph information of the fault diagnosis subgraph; then, two adjacent boundary times are determined as the start time and end time of the fault diagnosis subgraph, respectively.

8. The knowledge graph-based fault diagnosis method for live-line work as described in claim 7, characterized in that, The operational information includes the operating current, contact temperature, and leakage current of each device node; Read the operating current alarm threshold, contact temperature alarm threshold, and leakage current alarm threshold; When any item in the operation information is greater than or equal to the corresponding alarm threshold, the corresponding operation information is determined as abnormal operation information, and the abnormal operation information is mapped to an abnormal phenomenon node in the live-line operation knowledge graph. The sampling time corresponding to the first occurrence of the abnormal phenomenon node is determined as the abnormal sampling time, the sampling time that is closest to the abnormal sampling time is determined as the preceding sampling time, and the period from the preceding sampling time to the abnormal sampling time is determined as the abnormal occurrence time interval. When the boundary time exists and the boundary time does not exist within the time interval of the anomaly occurrence, the fault diagnosis subgraph whose time range includes the anomaly sampling time is determined as the fault diagnosis subgraph corresponding to the anomaly node; when the boundary time exists within the time interval of the anomaly occurrence, the fault diagnosis subgraphs that end at and begin at the boundary time are determined as the fault diagnosis subgraph before the boundary time and the fault diagnosis subgraph after the boundary time, respectively, and the two are determined as the fault diagnosis subgraph group adjacent to the boundary time.

9. The knowledge graph-based fault diagnosis method for live-line work as described in claim 8, characterized in that, The specific process for determining the fault cause node and fault occurrence stage in the aforementioned live-line work is as follows: The number of abnormal phenomenon nodes corresponding to each candidate fault cause node is counted and the total number of all abnormal phenomenon nodes is calculated. The ratio between the two is used to obtain the fault correlation degree. The fault correlation degree threshold is read, and the candidate fault cause node with the largest fault correlation degree that is greater than or equal to the fault correlation degree threshold is determined as the fault cause node. When determining based on the fault diagnosis subgraph corresponding to the abnormal phenomenon node, the fault occurrence stage is determined according to the start time and end time of the fault diagnosis subgraph; When the fault diagnosis sub-graph group adjacent to the boundary time is determined, the fault correlation degree of the fault cause node before and after the boundary time is calculated respectively. When the fault correlation degree before the boundary time is less than the fault correlation degree threshold and the fault correlation degree after the boundary time increases to be greater than or equal to the fault correlation degree threshold, it is determined that the fault occurrence stage is after the boundary time. When the fault correlation degree before and after the boundary time is greater than or equal to the fault correlation degree threshold, it is determined that the fault occurrence stage spans the boundary time.

10. A knowledge graph-based fault diagnosis system for live-line working, characterized in that, The system includes: The knowledge graph construction module obtains the live-line work equipment and equipment connection relationships from the work tasks corresponding to live-line work, and constructs a live-line work knowledge graph; then it monitors the topology change operations in live-line work to update the live-line work knowledge graph. The difference analysis module analyzes the changes in the knowledge graph of live-line work in order to calculate the topological difference. The time determination module, referring to the topological differences and graph changes, determines the boundary time that represents the change in the graph state of the live-line work knowledge graph; The subgraph creation module compares the corresponding information of the live-line work knowledge graph before and after the boundary time, and then creates a fault diagnosis subgraph based on the live-line work knowledge graph, writing the corresponding information that has changed in the comparison results into the fault diagnosis subgraph. The anomaly location module acquires the operating information of the equipment working without power supply collected by the on-site monitoring terminal to determine the nodes of abnormal phenomena and the time intervals of abnormal occurrence. It compares each boundary time with the time intervals of abnormal occurrence to determine the fault diagnosis sub-graph corresponding to the node of abnormal phenomenon or the fault diagnosis sub-graph group adjacent to the boundary time. The fault diagnosis module filters fault cause nodes associated with abnormal phenomenon nodes based on fault diagnosis subgraphs or fault diagnosis subgraph groups with adjacent boundary times, and determines the fault cause nodes and fault occurrence stages in live-line work.