AI-based low-voltage power distribution cabinet intelligent operation and maintenance decision support method and system

Abstract

The present application relates to the technical field of knowledge graph, specifically to a low-voltage power distribution cabinet intelligent operation and maintenance decision support method and system based on AI, comprising the following steps: obtaining low-voltage power distribution cabinet power-off nodes, electricity testing nodes, ground wire connection nodes and operation dependent edges, establishing an operation and maintenance standard graph, extracting field clock-in nodes and timestamp parameters, comparing node parameters in the operation and maintenance standard graph, and generating actual operation subgraphs. In the present application, the team reliability score is sorted in descending order, truncated and screened, a renewal list is generated, and the low-voltage power distribution cabinet node number is matched. The team credit evaluation result can be directly mapped to the high-risk operation permission identification and resource scheduling decision instruction, forming a closed-loop processing chain from operation process verification, abnormal collection, responsibility transmission to scheduling landing, which helps to improve the pertinence of violation identification and the discrimination degree of team screening.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph technology, and in particular to an AI-based intelligent operation and maintenance decision support method and system for low-voltage distribution cabinets. Background Technology

[0002] The field of knowledge graph technology is a data organization and reasoning technology with structured semantic expression as its core. By modeling information such as entities, relationships, and attributes in the form of graph structures, it realizes the unified expression and association fusion of multi-source heterogeneous data. In the power operation and maintenance scenario, knowledge graphs can semantically associate multi-dimensional information such as equipment nodes, operation processes, work behaviors, and personnel organization to build a queryable, reasonable, and scalable knowledge network structure.

[0003] While existing technologies can organize entities, relationships, and attributes into queryable, reasonable, and scalable knowledge network structures, in practice, they often remain at the level of relational expression and semantic association. They lack the ability to continuously depict specific work trajectories, such as whether on-site operations are performed in a standard sequence, whether there are inversions between different operation nodes, and whether the time consumption changes within the same work chain deviate from expectations. This results in a situation where the equipment and personnel relationships in the knowledge network are relatively clear, but the actual work process is difficult to fully map. For example, in the scenario of low-voltage distribution cabinet maintenance, on-site personnel have completed their check-in, and the knowledge network can identify the relationship between personnel and equipment, but it is difficult to further determine whether the voltage testing action occurred after the power-off action, or whether there was an abnormal delay in the grounding wire connection action. Therefore, it is easy to mistake risky work records for ordinary operational traces. Thus, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an AI-based intelligent operation and maintenance decision support method and system for low-voltage distribution cabinets.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an AI-based intelligent operation and maintenance decision support method for low-voltage distribution cabinets, comprising the following steps: Acquire the power outage nodes, voltage testing nodes, grounding wire connection nodes, and operation dependency edges of the low-voltage distribution cabinet, establish an operation and maintenance standard map, extract the on-site check-in nodes and timestamp parameters, compare the node parameters in the operation and maintenance standard map, and generate an actual operation sub-map. Extract the difference between the inverted amount and the time consumption inside the actual operation subgraph, calculate the process timing deviation, compare the process timing deviation with the preset alarm boundary value, and generate a violation and abnormal work order record. Extract the outsourced work group node and the violation and abnormal work order record, assign the violation and abnormal work order record to the negative node, associate and splice the outsourced work group node and the negative node, establish a supply chain knowledge graph, calculate the violation and abnormal work order record along the splicing edge in the supply chain knowledge graph to obtain the work group reliability score. Sort the reliability scores of each outsourced work group node in descending order, set a sorting cutoff value based on the descending sorting result, generate a core renewal list, and traverse and match the corresponding low-voltage distribution cabinet node numbers according to the core renewal list to establish resource scheduling decision instructions.

[0006] Preferably, the steps for obtaining the actual operation subgraph are as follows: Acquire the power outage nodes, voltage testing nodes, grounding wire connection nodes, and operation-dependent edges of the low-voltage distribution cabinet. Extract the node number, node type, and node sequence identifier of the power outage node, the node number, node type, and node sequence identifier of the voltage testing node, the node number, node type, and node sequence identifier of the grounding wire connection node, and the start-point number, end-point number, and edge direction identifier of the operation-dependent edges. Write the nodes into the node record according to the node number and write the edge connection relationship according to the start-point number and end-point number to form the operation and maintenance standard map. Extract the check-in node number, check-in node type, check-in device location identifier, and timestamp parameter of the check-in node. Locate candidate nodes with the same check-in node number in the operation and maintenance standard map according to the check-in node number. Filter out candidate nodes with inconsistent node types according to the check-in node type and candidate nodes with inconsistent device location identifiers according to the check-in device location identifier. Calculate the parameter comparison difference of the timestamp parameter corresponding to the node sequence identifier for each item. Keep the node records with a parameter comparison difference of zero to form a mapping node set. The mapping node set is arranged sequentially according to the order of the timestamp parameters. The corresponding splicing edges of the mapping node set in the operation and maintenance standard graph are extracted. It is verified whether the start number and end number of each corresponding splicing edge exist in the mapping node set at the same time. Edge records whose start number or end number does not fall into the mapping node set are deleted. The connected edge relationships within the mapping node set are retained. The mapping node set and the corresponding splicing edges are merged to form the actual operation subgraph.

[0007] Preferably, the step of obtaining the process timing deviation is as follows: Based on the actual operation subgraph, extract the preceding node number, following node number, preceding node timestamp parameter, and following node timestamp parameter corresponding to each operation dependency edge according to the connection direction of the operation dependency edge. Determine whether the preceding node timestamp parameter is later than the following node timestamp parameter. Record each operation dependency edge that has a later relationship. Count the number of operation dependency edges that have a later relationship. Extract the preceding inversion amount. Then calculate the node consumption time of each operation dependency edge according to the order of operation dependency edges. Subtract the node consumption time of the preceding operation dependency edge from the node consumption time of the following operation dependency edge to obtain the consumption time difference corresponding to each edge, forming a timing deviation calculation table. According to the timing deviation calculation table, the distribution record of the pre-inversion amount at each operation dependency edge position is read, the time difference value corresponding to each operation dependency edge is read, and the items are multiplied and accumulated according to the same operation dependency edge position. All multiply and accumulated results are extracted to obtain the pre-inversion penalty value. Then, the maximum time difference value and the minimum time difference value are extracted from the time difference value corresponding to each operation dependency edge. The maximum time difference value is calculated by subtracting the minimum time difference value from the minimum time difference value to obtain the time range value. The pre-inversion penalty value and the time range value are then written into the same deviation record to form the process timing deviation amount.

[0008] Preferably, the steps for obtaining the violation / abnormal work order record are as follows: Read the lower and upper bounds of the preset alarm boundary values, determine whether the process timing deviation is less than the lower bound, and determine whether the process timing deviation is greater than the upper bound. Filter out all deviation records that exceed the preset alarm boundary value range. Backtrack the node number, operation dependency edge number, job sequence identifier, timestamp parameter, pre-inversion amount, and time difference corresponding to each deviation record. Write the violation identifier, abnormality source identifier, deviation value identifier, and job parameter record according to the work order fields to form a violation abnormal work order record.

[0009] Preferably, the steps for acquiring the supply chain knowledge graph are as follows: Extract the team identifier, responsibility scope identifier, and associated low-voltage distribution cabinet node number from the outsourced team node; read the work order identifier, work order responsible entity identifier, violation identifier, abnormality source identifier, and operation parameter record from the violation and abnormal work order record; match each work order record with the team identifier and work order responsible entity identifier; filter out the matching team records and work order records to form a team work order correspondence table. Each of the aforementioned violation and abnormal work order records is assigned a negative node, and a negative node identifier, a negative level identifier, and a negative source identifier are written in. Outsourced work group nodes and negative nodes are spliced ​​together according to the matching relationship in the work group work order correspondence table. The starting point identifier, ending point identifier, and association sequence identifier of each splicing edge are recorded to form a supply chain knowledge graph.

[0010] Preferably, the steps for obtaining the core renewal list are as follows: Read the reliability scores of each outsourced work group node one by one, sort them in descending order of reliability scores, record the work group identifier, the reliability score of each group, and the score difference between adjacent groups for each group, retain the complete sorting results, and form a work group sorting sequence. Calculate the sorting cutoff value based on the sorting sequence of the work groups; Extract outsourced work group nodes from the work group sorting sequence whose descending positions are not greater than the sorting cutoff value, and form a core renewal list.

[0011] Preferably, the step of obtaining the resource scheduling decision instruction is as follows: According to the team identifier in the core renewal list, the associated low-voltage distribution cabinet node number is matched and the high-risk operation permission identifier is written into the associated low-voltage distribution cabinet node number. Then, the high-risk operation permission identifier and management control code are combined and written into the instruction field to form a resource scheduling decision instruction.

[0012] The present invention also provides a system comprising: The operation and maintenance map construction module is used to obtain the power outage nodes, voltage testing nodes, grounding wire connection nodes and operation dependency edges of the low-voltage distribution cabinet, establish the operation and maintenance standard map, extract the on-site check-in nodes and timestamp parameters, compare the node parameters in the operation and maintenance standard map, and generate the actual operation sub-map. The anomaly identification and analysis module is used to extract the difference between the pre-inversion amount and the time consumption within the actual operation sub-graph, calculate the process timing deviation, compare the process timing deviation with the preset alarm boundary value, and generate a violation and anomaly work order record. The team reliability assessment module is used to extract the outsourced team node and the violation and abnormal work order record, assign the violation and abnormal work order record to the negative node, associate and splice the outsourced team node and the negative node, establish a supply chain knowledge graph, and calculate the violation and abnormal work order record along the splicing edge in the supply chain knowledge graph to obtain the team reliability score. The resource scheduling decision module is used to sort the reliability scores of each outsourced work group node in descending order, set a sorting cutoff value based on the descending sorting result, generate a core renewal list, and traverse and match the corresponding low-voltage distribution cabinet node numbers according to the core renewal list to establish resource scheduling decision instructions.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by combining the low-voltage distribution cabinet's power-off node, voltage testing node, grounding wire connection node, and operation dependency edge, a standard operation and maintenance map is first formed. Then, the on-site check-in nodes and timestamp parameters are embedded into the map comparison process. This transforms abstract operation and maintenance relationships into a verifiable work chain, synchronizing the on-site execution sequence, node correspondence, and time sequence into the same judgment framework, thereby generating an actual operation sub-map and enhancing the accuracy of on-site operation reconstruction. Furthermore, by extracting the preceding inversion quantity and the time difference, and directly associating the time sequence deviation results with alarm boundary values, previously scattered violation signs can be compressed into targeted violation and abnormal work order records. This shifts anomaly identification from static record comparison to a process-oriented approach. Deviation identification; then, the records of violations and abnormal work orders are assigned as negative nodes and spliced ​​to the outsourced team nodes, forming a supply chain knowledge graph that can be accumulated and propagated along the splicing edge. This allows team risks to no longer be limited to a single work order evaluation, but to be deposited layer by layer along the responsibility association as team reliability scores, facilitating the identification of persistent non-compliant behaviors. Further, the team reliability scores are sorted in descending order, truncated and filtered, renewal list generated, and low-voltage distribution cabinet node number matched. This allows the team credit evaluation results to be directly mapped to high-risk operation permit authority identifiers and resource scheduling decision instructions, forming a closed-loop processing chain from operation process verification, abnormal collection, responsibility transmission to scheduling implementation. This helps to improve the targeting of violation identification and the differentiation of team screening. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 A scatter plot showing the distribution of alarms for process timing deviations; Figure 3 The time decay characteristic curve of the team reliability score; Figure 4 This is a schematic diagram of the actual operation subgraph mapping and spatiotemporal verification principle. Detailed Implementation

[0015] 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.

[0016] Please see Figure 1-4 This invention provides a technical solution: an AI-based intelligent operation and maintenance decision support method for low-voltage distribution cabinets, comprising the following steps: Obtain the power outage nodes, voltage testing nodes, grounding wire connection nodes, and operation dependency edges of the low-voltage distribution cabinet, establish an operation and maintenance standard map, extract the on-site check-in nodes and timestamp parameters, compare the node parameters in the operation and maintenance standard map, and generate an actual operation sub-map. Extract the difference between the inverted amount and the time consumption within the actual operation sub-graph, calculate the process timing deviation, compare the process timing deviation with the preset alarm boundary value, and generate a violation and abnormal work order record. Extract the outsourced work team nodes and violation / abnormal work order records, assign the violation / abnormal work order records to negative nodes, associate and splice the outsourced work team nodes and negative nodes, establish a supply chain knowledge graph, calculate the violation / abnormal work order records along the splicing edges in the supply chain knowledge graph, and obtain the work team reliability score. Sort the reliability scores of each outsourced work group node in descending order, set the sorting cutoff value according to the descending sorting result, generate the core renewal list, and traverse and match the corresponding low-voltage distribution cabinet node number according to the core renewal list to establish resource scheduling decision instructions.

[0017] The actual steps for obtaining the subgraph are as follows: Acquire the power outage nodes, voltage testing nodes, grounding wire connection nodes, and operation-dependent edges of the low-voltage distribution cabinet. Extract the node number, node type, and node sequence identifier of the power outage node, the node number, node type, and node sequence identifier of the voltage testing node, the node number, node type, and node sequence identifier of the grounding wire connection node, and the start-point number, end-point number, and edge direction identifier of the operation-dependent edges. Write the nodes into the node record according to the node number and write the edge connection relationship according to the start-point number and end-point number to form the operation and maintenance standard map. Extract the check-in node number, check-in node type, check-in device location identifier, and timestamp parameter of the check-in node. Locate candidate nodes with the same node number in the operation and maintenance standard map according to the check-in node number. Filter out candidate nodes with inconsistent node types according to the check-in node type and candidate nodes with inconsistent device location identifiers according to the check-in device location identifier. Calculate the parameter comparison difference of the timestamp parameter corresponding to the node sequence identifier for each item. Keep the node records with a parameter comparison difference of zero to form a mapping node set. The mapping node set is arranged sequentially according to the order of the timestamp parameters. The corresponding splicing edges of the mapping node set in the operation and maintenance standard graph are extracted. It is verified whether the start number and end number of each corresponding splicing edge exist in the mapping node set at the same time. The edge records whose start number or end number does not fall into the mapping node set are deleted. The connected edge relationships within the mapping node set are retained. The mapping node set and the corresponding splicing edges are merged to form the actual operation subgraph.

[0018] Specifically, the system retrieves the low-voltage distribution cabinet's power-off node, voltage testing node, grounding wire connection node, and operation dependency edges. It sequentially reads the basic attribute fields of each node from a pre-defined maintenance configuration library. For the power-off node, it reads the 6-digit node number, the node type field (set to 1), and the node sequence identifier representing the prerequisite step number in the standard operating procedure. For the voltage testing node, it reads its unique node number, the node type field (set to 2), and the node sequence identifier indicating that it must be executed after the power-off node. For the grounding wire connection node, it reads its node number, the node type field (set to 3), and the node sequence identifier indicating that it must be executed after the voltage testing node. The system parses the data structure of the operation dependency edges, extracting the start point number representing the beginning node of the dependency relationship, the end point number representing the completion node of the dependency relationship, and the edge direction identifier (using a Boolean value of 1 to represent unidirectional progression). It initializes a two-dimensional relational data table in memory, iterates through all extracted node data, and writes the node number as the primary key into the first column of the relational data table as a node. For each node record, the node type and sequence identifier are appended to the corresponding primary key attribute column. For example, the record for a power outage node is written as number 000111, type 1, and sequence 1; the record for a voltage testing node is written as number 000112, type 2, and sequence 2; and the record for a grounding wire connection node is written as number 000113, type 3, and sequence 3. A directed graph data structure is initialized, and each node record written to the relational data table is instantiated as a vertex in the directed graph. All operation dependency edges are traversed, and the start and end numbers of each edge are read. A directed connection is constructed between the corresponding start and end vertices in the directed graph. For example, based on the start number 000111 and the end number 000112, a directed connection is established between the power outage vertex and the voltage testing vertex. The corresponding adjacency matrix is ​​generated based on the edge connection relationship of the records. The row index and column index of the adjacency matrix correspond to the node number. When the element at the intersection of a row and a column in the matrix is ​​1, it represents the existence of an operation dependency relationship, and when it is 0, it represents the absence of a dependency relationship, thus forming a standard operation and maintenance graph.

[0019] Extract the check-in node number, check-in node type, check-in device location identifier, and timestamp parameters from the on-site check-in nodes. Iterate through all collected on-site check-in records, reading the 6-digit check-in node number associated with each on-site check-in node. Perform a global search within the previously established maintenance standard map, comparing the check-in node number with the node numbers of each vertex in the map. Extract vertices with identical numbers as candidate nodes. Read the check-in node type parameter from the on-site check-in nodes and compare it one by one with the preset node types of the candidate nodes. For example, if the check-in node type is 2, check if the candidate node type is also 2. If the values ​​are inconsistent, delete the candidate node from the candidate list. Read the check-in device location identifier, which is represented by the device's latitude and longitude coordinates. Read the standard device location coordinates corresponding to the candidate node and calculate the Euclidean distance between the check-in device location coordinates and the standard device location coordinates. Set the coordinate distance offset threshold to 5 meters, which is based on the standard power distribution room space dimensions and personnel operation activities. The radius is set at an empirical value of 5 meters. The system checks if the Euclidean distance is greater than 5 meters. If the distance is greater than 5 meters, the equipment location identifier is considered inconsistent, and the corresponding candidate nodes are eliminated. For the candidate nodes remaining after multiple eliminations, the timestamp parameter attached to each on-site check-in node is extracted. The timestamp parameter format is the precise time in year, month, day, hour, minute, and second. All on-site check-in nodes are sorted in ascending order from earliest to latest according to the timestamp parameter. Each sorted on-site check-in node is assigned an actual execution sequence number starting from 1. The node sequence identifier in the corresponding candidate node is read, and the actual execution sequence number is subtracted from the node sequence identifier. For example, if the actual execution sequence number of a check-in node is 2, the node sequence identifier of its candidate node is also 2. The difference between the two results in a parameter comparison difference of 0. The calculation results of all candidate nodes are checked, and node records with parameter comparison differences that are not 0 are deleted, while node records with parameter comparison differences of 0 are retained. These perfectly matching candidate nodes are packaged and stored in a one-dimensional array according to their original attributes and relationships, forming a mapping node set.

[0020] The mapping node set is sequentially arranged according to the order of timestamp parameters. The timestamp parameter bound to each node in the set is read, and a bubble sort algorithm is used to compare the timestamps of all nodes pairwise, placing nodes with earlier timestamps first and nodes with later timestamps last, resulting in a node sequence distributed according to the actual execution time. A traversal query operation is initiated in the edge relationship records of the previously generated operation and maintenance standard graph, comparing the node numbers contained in each operation dependency edge in the graph. All spliced ​​edges whose start and end numbers both appear in the mapping node set are extracted and temporarily stored in a transition buffer list. The start and end numbers of each corresponding spliced ​​edge are read sequentially from the transition buffer list. The read start number is cyclically matched with all node numbers contained in the mapping node set to determine if there are any records with equal values. The same cyclic matching operation is used to determine if the end number also exists in the mapping node set. When a corresponding spliced ​​edge's start number is found in the mapping node set... If no match is found, or if the endpoint number of the spliced ​​edge cannot be matched in the mapping node set, a deletion operation is triggered. For example, if the mapping node set only contains numbers 000111 and 000112, and a temporarily stored spliced ​​edge has a starting point of 000112 and an endpoint of 000113, since 000113 is not in the set, the spliced ​​edge record is removed from the transition buffer list. After verification, the transition buffer list only retains spliced ​​edge records whose starting and ending points are both contained within the mapping node set. These filtered spliced ​​edge records are read, and the edge direction identifier and connectivity attribute of the corresponding spliced ​​edge in the operation and maintenance standard graph are extracted. These connectivity relationships are converted into an edge set of the subgraph structure. The sorted mapping node set is read as the vertex set of the subgraph structure. A new directed graph object is initialized in memory, and the aforementioned vertex set and edge set are injected into the directed graph object. Each vertex node is assigned a value according to the attribute of the vertex set, and directed connections between vertices are constructed according to the edge set. The attribute superposition and binding are completed, forming the actual operation subgraph.

[0021] The steps for obtaining the process timing deviation are as follows: Based on the actual operation subgraph, extract the preceding node number, following node number, preceding node timestamp parameter, and following node timestamp parameter corresponding to each operation dependency edge according to the connection direction of the operation dependency edge. Determine whether the preceding node timestamp parameter is later than the following node timestamp parameter. Record each operation dependency edge that has a later relationship. Count the number of operation dependency edges that have a later relationship. Extract the preceding inversion amount. Then calculate the node consumption time of each operation dependency edge according to the order of operation dependency edges. Subtract the node consumption time of the preceding operation dependency edge from the node consumption time of the following operation dependency edge to obtain the consumption time difference corresponding to each edge, forming a timing deviation calculation table. Based on the timing deviation calculation table, the distribution record of the pre-inversion amount at each operation dependency edge position is read, the time difference value corresponding to each operation dependency edge is read, and the items are multiplied and accumulated according to the same operation dependency edge position. All multiply and accumulated results are extracted to obtain the pre-inversion penalty value. Then, the maximum and minimum time difference values ​​are extracted from the time difference values ​​corresponding to each operation dependency edge. The maximum time difference value is calculated by subtracting the minimum time difference value from the minimum time difference value to obtain the time range value. Finally, the pre-inversion penalty value and the time range value are written into the same deviation record to form the process timing deviation amount.

[0022] Specifically, based on the actual operation subgraph, the edge set and node set are extracted from the previously constructed actual operation subgraph. Each operation-dependent edge in the edge set is traversed. The execution order of nodes is confirmed according to the direction of the directed edges in the graph structure. The vertex attributes of the starting point of the directed edge are extracted as the predecessor node number, and the vertex attributes of the ending point of the directed edge are extracted as the successor node number. Simultaneously, the precise timestamp parameters of the operations performed on the predecessor node and the successor node are extracted from the attribute records of the corresponding vertices. The timestamp format is year, month, day, hour, minute, second. The timestamp parameters of the predecessor and successor nodes associated with the same operation-dependent edge are compared to determine whether the predecessor node timestamp parameter is later than the successor node timestamp parameter on the timeline, i.e., the predecessor operation is later than the successor operation. If a later situation exists, the starting and ending point numbers of the operation-dependent edges where the later relationship occurs are recorded, and a separate list is created to store these later-arriving operations. For the operation-dependent edges of a relationship, iterate through the list to count the total number of elements and obtain the number of operation-dependent edges that occurred later than the relationship. Set this number as the preceding inversion value. Then, sort all operation-dependent edges in the actual operation subgraph according to the timestamp order and calculate the node timeout of each operation-dependent edge. That is, subtract the preceding node's timestamp parameter from the subsequent node's timestamp parameter for each edge to obtain the node timeout in seconds. Calculate the adjacent difference according to the sorted operation-dependent edge sequence. Read the node timeout of the second operation-dependent edge and subtract the node timeout of the first operation-dependent edge, and so on. Subtract the node timeout of the preceding operation-dependent edge from the node timeout of the subsequent operation-dependent edge to obtain the timeout difference for each operation-dependent edge. The timeout difference for the first operation-dependent edge is recorded as 0. Combine and summarize the extracted edge number, preceding inversion value marking, and timeout difference to form a time series deviation calculation table.

[0023] Based on the timing deviation calculation table, data is read row by row from the previously established table. The distribution records of the pre-inversion amount at each operation-dependent edge are extracted, i.e., a Boolean flag variable indicating whether pre-inversion has occurred for each edge (1 for inversion, 0 for no inversion). Simultaneously, the time difference corresponding to each operation-dependent edge is extracted. The pre-inversion flag variable for each operation-dependent edge is multiplied by its time difference. When inversion occurs, the product is the time difference itself; when no inversion occurs, the product is 0. All operation-dependent edges are traversed, and all product results are summed together to obtain the total fluctuation of time difference caused by all inverted edges. The total sum of all multiplication results is extracted to obtain the pre-inversion penalty value. Then, the time difference calculation table is traversed again. The time difference sequence corresponding to each operation dependency edge is recorded. Using bubble sort or traversal comparison, the time difference with the largest value is selected as the maximum time difference, and the time difference with the smallest value is selected as the minimum time difference. The fluctuation range of this set of time difference is calculated by subtracting the minimum time difference from the maximum time difference, and the time difference range is obtained. The previously calculated pre-inversion penalty value is extracted, and the range value is also extracted. A new data structure instance is created in memory as a deviation record. The pre-inversion penalty value is assigned as one field to the deviation record, and the time difference value is assigned as another field to the deviation record. The calculation results of the two are stored in the same structure to form the process timing deviation.

[0024] The steps to obtain violation and abnormal work order records are as follows: Read the lower and upper bounds of the preset alarm boundary values, determine whether the process timing deviation is less than the lower bound, and determine whether the process timing deviation is greater than the upper bound. Filter out all deviation records that exceed the preset alarm boundary value range, and backtrack the node number, operation dependency edge number, job sequence identifier, timestamp parameter, pre-inversion amount, and time difference corresponding to each deviation record. Write the violation identifier, abnormality source identifier, deviation value identifier, and job parameter record according to the work order fields to form a violation abnormal work order record.

[0025] Specifically, the lower and upper bounds of the preset alarm boundary values ​​are read. These alarm boundary values ​​are calculated based on historical time-series deviation data extracted from 500 previous normal power distribution cabinet maintenance tasks. The lower bound is set by subtracting twice the standard deviation from the historical data average, and the upper bound is set by adding twice the standard deviation to the historical data average. For example, if the historical average deviation is 10 and the standard deviation is 2, then the lower bound is set to 6 and the upper bound to 14. The previously constructed process time-series deviation is extracted, and the simple sum of the pre-inversion penalty value and the time-consuming range value in the process time-series deviation is used as the overall deviation index. This is compared with the preset alarm boundary values ​​to determine whether the overall deviation index of the process time-series deviation is less than the set lower bound and whether it is greater than the set upper bound. If the process time-series deviation falls within the range of 6 to 14, it is considered normal; if it is less than... A value of 6 or greater than 14 is considered to exceed the allowable fluctuation range. All deviation records exceeding the preset alarm boundary value range are filtered out. For each filtered deviation record, a data tracing operation is triggered. Based on the primary key in the deviation record, a reverse lookup is performed in the various data tables generated earlier. The on-site check-in node number corresponding to each deviation record is extracted, the corresponding operation dependency edge number is extracted, the preset work sequence identifier is extracted, and the on-site collected timestamp parameter accurate to the second is extracted. Simultaneously, the preceding inversion quantity and the specific time difference data are extracted. Multiple record fields are defined in the newly created work order data table. The judgment result of exceeding the boundary is converted into the number 1 and written into the violation identifier field. The corresponding operation and maintenance business code is written as the abnormal source identifier. The specific difference value is written into the deviation value identifier. The node number and timestamp information extracted by backtracking are packaged and written into the work parameter record field to form a violation abnormal work order record.

[0026] The steps for acquiring a supply chain knowledge graph are as follows: Extract the team identifier, scope of responsibility identifier, and associated low-voltage distribution cabinet node number from the outsourced team nodes. Read the work order identifier, work order responsible entity identifier, violation identifier, abnormality source identifier, and operation parameter record from the violation and abnormal work order record. Match each work order record with the team identifier and work order responsible entity identifier, filter out the matching team records and work order records, and form a team work order correspondence table. Each violation / abnormal work order record is assigned a negative node, and a negative node identifier, a negative level identifier, and a negative source identifier are written in. Outsourced work group nodes and negative nodes are spliced ​​together according to the matching relationship in the work group work order correspondence table. The start point identifier, end point identifier, and association sequence identifier of each splicing edge are recorded to form a supply chain knowledge graph.

[0027] Specifically, the process involves extracting the team identifier, responsibility scope identifier, and associated low-voltage distribution cabinet node number from the outsourced team nodes. It then iterates through the database tables storing information for each outsourced team, parsing the fields of each record. This includes reading the team identifier (a unique string), the responsibility scope identifier (representing the coordinates of the polygonal vertices it maintains), and the associated low-voltage distribution cabinet node numbers corresponding to the multiple low-voltage distribution cabinets maintained by the team. These team attribute data are temporarily stored in a hash table in memory, using the team identifier as the key. Next, the process reads the work order identifier, work order responsible party identifier, violation identifier, abnormality source identifier, and work parameter records from the previously generated violation and abnormal work order records. It iterates through the list containing all violation and abnormal work order records, extracting the uniquely assigned work order identifier from each record, the work order responsible party identifier pointing to the party responsible for the violation, and the violation identifier representing the severity of the violation. The system retrieves the anomaly source identifier from the violation discovery channel, extracts various operation parameter records including timestamps and time difference, initializes a two-dimensional relational table in memory as a matching container, and compares each work order with the work group identifier in the hash map table and the work order responsible entity identifier in the work order record. The work order responsible entity identifier is extracted and checked for string equality with the key in the hash map table. When the two strings are found to be completely identical, the work group node information is merged with the work order record information. For example, if the work group identifier is BZ001 and the work order responsible entity identifier of a certain work order is also BZ001, then a match is determined to be valid. Matching work group records and work order records are filtered out, while detached work order records whose work order responsible entity identifiers cannot be found in the hash map table are discarded. The filtered matching records are written row by row into the established two-dimensional relational table, with the first column filled with work group-related attributes and the second column filled with work order-related attributes, merging to form a work group-work order correspondence table.

[0028] Each violation / abnormal work order record is assigned a negative node. The previously generated work order correspondence table is read, and for each merged record in the table, the work order portion is abstracted as an independent vertex in the graph structure. The attribute direction of this vertex is marked as negative. The work order identifier is extracted from the merged record and used as the unique primary key of this negative vertex, written into the negative node identifier field. Violation identifiers are extracted from the merged record, and the deviation level corresponding to the violation identifier is determined. If the deviation value is between 6 and 10, it is marked as a general level; if it is between 10 and 14, it is marked as a severe level. This level classification information is written into the negative level identifier. The abnormality source identifier is extracted, converted to its original form as the source attribute of the graph structure vertex, and written into the negative source identifier. A data structure supporting attribute graph storage is initialized in memory. All previously stored outsourced work group nodes are instantiated as positive vertices in this graph structure, and all assigned negative nodes are instantiated as positive vertices in the graph structure. For negative vertices, the vertex connection operation is initiated according to the matching relationship in the work order correspondence table. Starting from the first row of the correspondence table, the work group identifier and negative node identifier bound in the same row are identified. A directed edge is established between the corresponding positive and negative vertices in the graph structure to complete the splicing operation of outsourced work group nodes and negative nodes. Since one work group corresponds to multiple negative nodes, it is necessary to attach an order attribute to it when establishing the edge. The starting endpoint of the directed edge is the work group identifier, and the ending endpoint of the directed edge is the negative node identifier. The starting and ending identifiers of each spliced ​​edge are recorded. The edges are sorted according to the time stamp order of multiple work orders under the same work group in the correspondence table. Each edge is assigned an integer order value starting from 1 and incremented, which is written into the attribute list of the edge as the association order identifier of the edge. After traversing and connecting all the merged records, a network data graph containing the topological relationship between the work group and all its violation records is obtained, forming a supply chain knowledge graph.

[0029] The steps to obtain the team reliability score are as follows: Based on the supply chain knowledge graph, the reliability score of the work team is calculated using the following formula: ; in, For the first The reliability score of each outsourced work team node. For the first The preset base score value for each outsourced work team node. This is the counting index of the negative node. For the first The total number of negative nodes associated with each outsourced work group node. For the first The basic negative weights corresponding to each negative node. The preset time decay parameter, For the first Each negative node corresponds to the time difference between the occurrence time of the violation / abnormal work order record and the current evaluation time. For the first The penalty value for illegal aggregation corresponding to each negative node. In the first Within the preset time window preceding the occurrence of the violation / abnormal work order record corresponding to the first negative node, the... The number of other violation work orders corresponding to each outsourced work group node.

[0030] Specifically, the formula for calculating the reliability score of a work team introduces a time decay mechanism and a violation accumulation penalty mechanism to comprehensively evaluate the reliability level of the work team in the historical work order execution process. The time decay parameter reduces the impact of long-standing violation records on the current evaluation, which is in line with the laws of objective forgetting and management improvement. The violation accumulation penalty value applies an exponential negative weight to violations that occur continuously in a short period of time, effectively combating sudden violations in the work team and avoiding the masking of the true situation of short-term management failures within the work team by a single linear deduction. The parameter represents the first The preset base score for each outsourced work team node is obtained by collecting the total number of maintenance work orders undertaken by the target outsourced work team in the past calendar year, filtering out the number of normal work orders that were completed on time and in compliance with regulations, calculating the proportion of normal work orders to the total number of work orders undertaken, and then multiplying this proportion by 100 to convert it into a percentage base score. The calculation formula is as follows: ,in, This represents the preset base score. This represents the number of normal work orders that the target outsourced team completed on time and in compliance with regulations throughout the past calendar year. This represents the total number of maintenance work orders undertaken by the target outsourced team during the entire past calendar year. For example, if a team undertaken 100 work orders last year... The total is 100, of which 95 were completed on schedule and in compliance with regulations. If the value is 95, then the calculated basic full score is 95. The parameter represents the first The total number of negative nodes associated with a given outsourced work group node. For example, if a work group node is found to have 2 outgoing edges pointing to negative nodes, then... If the sum is 2, then by iterating through the list and adding the results, the total number is 2. The parameter represents the first The basic negative weight corresponding to each negative node, for example, the deviation value recorded by a negative node is 15. The value is 15, and the historical maximum deviation from the benchmark is 50. If the value is 50, then the basic negative weight is 15 divided by 50 and multiplied by 10 to get 3; The parameter represents the preset time decay parameter. The steps to obtain this parameter are as follows: read the half-life of the impact of the violation as set in the safety management regulations, calculate the decay rate using the natural logarithm function, ensuring that the penalty's impact decreases exponentially over time. The calculation formula is: ,in, This represents the preset time decay parameter. The natural logarithm constant is approximately 0.693. This represents the half-life of the impact of a violation as defined in safety management regulations. For example, if the half-life of the impact of a violation is specified as 180 days... If the value is 180, then the time decay parameter is 0.693 divided by 180, which gives approximately 0.00385. The parameter represents the first Each negative node corresponds to the time difference between the occurrence time of the violation / abnormal work order record and the current assessment time. For example, the current time converted to Julian days is 2460219. The work order number is 2460219, and the Julian date when the work order was generated is 2460189. If the value is 2460189, then the time difference is 30. The parameter represents the first time. Within the preset time window preceding the occurrence of the violation / abnormal work order record corresponding to the first negative node, the... The number of other violation work orders corresponding to each outsourced work group node is determined by the following steps: setting a time window of 30 days, and searching the database containing all work orders for items belonging to that work group and whose occurrence time is the [number missing]. The number of violation work order records within 30 days prior to the occurrence of each negative node is used as the aggregated work order count. The calculation formula is as follows: ,in, This represents the number of other violation work orders within the preset time window. The traversal index variable represents the set of negative nodes. This represents the total number of negative nodes belonging to this work group in the database containing all work orders. The representative function takes the value 1 if the internal condition is met, and takes the value 0 otherwise. Representing the The number of days since the occurrence of each negative node. Representing other The number of days since the occurrence of each negative node, where 30 represents the preset time window number of days. For example, if two violations are found in the work group within 30 days before the work order is issued, which meets the condition that the indicator function takes 1 twice, then the sum of the counts is 2. Calculations based on parameters: Substituting the aforementioned parameters into the formula, we set the first... The basic full score for each class group The total number of associated negative nodes is 95. The preset time decay parameter is 2. It is 0.00385; For the first negative node, that is Equal to 1, substitute the deviation value of 15 with the maximum baseline of 50 to obtain the basic negative weight. The value is 3, taking into account the time difference of occurrence. The value is 60, which is the number of other violations within the time window. =1; Calculate the penalty value for illegal aggregation of the first negative node: ; Calculate the decay weight product of the first negative node: ; Calculate the overall deduction for the first negative node: ; For the second negative node, that is Equal to 2, substituting the deviation value of 25 and the maximum baseline of 50, we obtain the basic negative weight. The value is 5, taking into account the time difference of occurrence. The value is 30, which is the number of other violations within the time window. It is 2; Calculate the penalty value for illegal aggregation of the second negative node: ; Calculate the decay weight product of the second negative node: ; Calculate the overall deduction for the second negative node: ; Sum the deductions for all negative nodes: ; The team reliability score is obtained by subtracting the summed result from the base full score: .

[0031] The results indicate that the calculated team reliability score of 84.98 comprehensively reflects the outsourced team's performance quality and safety management capabilities at the current assessment time. If the score is greater than 90, it means that the team's records are from a long time ago and are isolated events, indicating excellent renewal potential. If the score is less than 85, it means that the team has recently had continuous violations and a cluster of violations in a short period of time, proving that its management system has loopholes. This score is used as a quantitative indicator for subsequent ranking and generation of the core renewal list.

[0032] The steps to obtain the core renewal list are as follows: Read the reliability scores of each outsourced work group node one by one, sort them in order of reliability scores from largest to smallest, record the work group identifier, work group reliability score, and score difference between adjacent positions for each position, retain the complete ranking results, and form a work group ranking sequence. Based on the work group ranking sequence, calculate the ranking cutoff value using the following formula: ; in, To truncate the values ​​for sorting, To be in the first descending position to the th Select the position corresponding to the largest target value among the descending positions. The first in the sorting sequence of work groups A descending position index. This represents the total number of outsourced work group nodes in the work group sorting sequence. For the first The reliability score of the work group corresponding to each descending position. For the first The reliability score of the work group corresponding to each descending position. This represents the average reliability score of all work groups in the work group ranking sequence. This is the preset base score value; Extract outsourced work group nodes from the work group sorting sequence whose descending positions are not greater than the sorting cutoff value, and form a core renewal list.

[0033] Specifically, the reliability score of each outsourced work group node is read one by one. All outsourced work group assessment data records stored in the database are traversed, and the unique work group identifier bound to each outsourced work group node and the work group reliability score obtained in the previous calculation cycle are extracted. A one-dimensional array containing key-value pairs of work group identifiers and reliability scores is constructed in memory. The quicksort algorithm is called to sort all elements within this one-dimensional array in descending order of work group reliability scores. During the sorting process, the values ​​of every two adjacent elements are compared, and the element with the larger score is swapped to the front of the array, while the element with the smaller score is shifted to the back, until all elements have been traversed and no more records need to be swapped. This results in a node list sorted in descending order of scores. For this sorted node list, starting from the first position (1), an incrementing integer sequence number is assigned, and the current corresponding index for each position is extracted. The team identifier and its reliability score are calculated by extracting the reliability score of the current team and subtracting the reliability score of the next adjacent team from the first position in the list. For example, if the team identifier in the first position is A01 with a score of 92, and the team identifier in the second position is A02 with a score of 88, then subtracting 88 from 92 gives a score difference of 4. If it is the last position in the list, since there are no subsequent elements, its score difference is set to 0. Multiple data columns are created in the relational data table. The calculated and extracted position number, team identifier, team reliability score, and score difference of adjacent positions are recorded as a row and filled into the corresponding columns. The data is written into the table row by row in ascending order of position number to ensure that no outsourced team node data is missed and to retain complete arrangement and calculation information, forming a team sorting sequence.

[0034] In the formula for calculating the truncation value, the truncation threshold is dynamically determined by finding the largest gap between the rankings. This abandons the traditional method of eliminating teams with a fixed score line. It not only considers the absolute score difference between two adjacent teams, but also multiplies the difference by a relative advantage incentive term compared with the average score. This allows top teams with scores far exceeding the average to receive some compensation even if the gap is small, while the gap effect of teams lagging behind the average score will be weakened. This more scientifically identifies the true strength echelon dividing point between teams and avoids the problem of being accidentally injured when strong teams are clustered together. The parameter represents the position from the first descending position to the second position. Select the position corresponding to the largest target value among the descending positions. For example, after 5 iterations, the target value obtained in the 3rd iteration is the largest, which is 15.6. Then, the position corresponding to the largest target value is recorded as 3. The parameter represents the total number of outsourced work group nodes in the work group sorting sequence. For example, if there are 10 rows of work group data in the work group sorting sequence, then the total number of outsourced work group nodes obtained is 10. The parameter represents the first The reliability score of the work group corresponding to each descending position, for example, when the loop variable When the value is 2, look up the second row of data in the table, and extract the record with a reliability score of 88. Then the current... The value of the parameter is 88; The parameter represents the first The reliability score of the work group corresponding to each descending position, for example, the current position. If the value is 2, then adding 1 results in row number 3. Searching for the 3rd row in the table, the reliability score of that row is 82. Therefore, the current... The value of the parameter is 82; The parameter represents the average reliability score of all work groups in the work group ranking sequence. For example, if there are 5 work groups in the sequence... The score is 5, with scores of 96, 88, 82, 78, and 76 respectively. Adding these five numbers together gives 420, and dividing by the total number 5 gives an average of 84. The parameter represents the preset base score value, with a maximum score of 100 points. Calculations based on parameters: Set the total number of nodes for outsourced work teams The preset maximum score is 5. It is 100; The reliability scores for each work group corresponding to each priority level are as follows: 1st priority level It is 96, the second in line. It's 88, the 3rd in line. It is 82, the 4th in line. It is 78, the 5th in line. It is 76; Substitute all class group values ​​to calculate the average score : ; Calculate the constant term in the denominator : ; when At that time, substitute the corresponding score to calculate the target value: Calculate the score difference item: ; Calculate the advantage incentive: ; Calculate the product of the two: ; when At that time, substitute the corresponding score to calculate the target value: Calculate the score difference item: ; Calculate the advantage incentive: ; Calculate the product of the two: ; when At that time, substitute the corresponding score to calculate the target value: Calculate the score difference item: ; Calculate the advantage incentive: ; Calculate the product of the two: ; when At that time, substitute the corresponding score to calculate the target value: Calculate the score difference item: ; Calculate the advantage incentive: ; Calculate the product of the two: ; Compare all the obtained target product results (14, 7.5, 3.5, 1.25), with the maximum value being 14. The descending position number corresponding to this maximum value is... The value is 1, therefore the sorting cutoff value is obtained. The value is 1.

[0035] The results indicate that the calculated cutoff value of 1 means that there is the largest gap in overall strength between the first-ranked team and the second-ranked team. Teams ranked ahead of this position have a reliability performance far exceeding that of the group. Using this as the cutoff position can maximize the retention of high-quality resources and eliminate mediocre or poor-performing teams. This cutoff value provides a quantitative boundary basis for the subsequent division of the contract renewal list.

[0036] The steps for obtaining resource scheduling decision instructions are as follows: The system iterates through and matches the associated low-voltage distribution cabinet node numbers according to the team identifier in the core renewal list. The high-risk operation permit authorization identifier is written into the associated low-voltage distribution cabinet node number. Then, the high-risk operation permit authorization identifier and management control code are combined and written into the instruction field to form a resource scheduling decision instruction.

[0037] Specifically, the process involves extracting outsourced work group nodes whose descending positions in the work group ranking sequence are not greater than the truncation value. The previously generated work group ranking sequence data table is read, and the calculated truncation value is extracted. A blank data set is initialized in memory to temporarily store work group records that meet the criteria. Starting from the first row of the work group ranking sequence data table, a row-by-row traversal operation is initiated. The descending position index of the current row is extracted, and this index is compared with the truncation value. For example, if the truncation value is 1, when the record with index 1 is encountered, it is determined that 1 is less than or equal to 1, satisfying the condition. All attribute data of the outsourced work group nodes in this row are copied and added to the initialized data set. When the record with index 2 is encountered, it is determined that 2 is greater than 1, disqualifying the condition, so this row is skipped and copying stops. After the complete traversal and comparison, a set containing only the top-ranked high-quality work groups is obtained, and this set is saved. To create a core renewal list, the remaining work team identifiers are extracted from each item on the core renewal list. The underlying facility operation and maintenance relationship database is opened, and a traversal matching operation is initiated in the database according to the extracted work team identifiers. All associated low-voltage distribution cabinet node numbers maintained by the work team are searched, and these equipment number records are extracted. For each successfully matched associated low-voltage distribution cabinet node number, a high-risk operation permission identifier with the value of string 1 is added to the equipment's permission configuration attribute table. This represents granting the work team the right to perform high-risk operations such as live work or disconnecting and grounding the distribution cabinet. Next, the pre-set management control code is read. This code consists of 4 digits representing the operation type. The high-risk operation permission identifier and the management control code are concatenated together. For example, the permission identifier 1 and the control code 2048 are concatenated into the form of 1 bar 2048. This combined string is written into the instruction field sent to the mobile terminal to form a resource scheduling decision instruction.

[0038] 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. An AI-based intelligent operation and maintenance decision support method for low-voltage distribution cabinets, characterized in that, Includes the following steps: Acquire the power outage nodes, voltage testing nodes, grounding wire connection nodes, and operation dependency edges of the low-voltage distribution cabinet, establish an operation and maintenance standard map, extract the on-site check-in nodes and timestamp parameters, compare the node parameters in the operation and maintenance standard map, and generate an actual operation sub-map. Extract the difference between the inverted amount and the time consumption inside the actual operation subgraph, calculate the process timing deviation, compare the process timing deviation with the preset alarm boundary value, and generate a violation and abnormal work order record. Extract the outsourced work group node and the violation and abnormal work order record, assign the violation and abnormal work order record to the negative node, associate and splice the outsourced work group node and the negative node, establish a supply chain knowledge graph, calculate the violation and abnormal work order record along the splicing edge in the supply chain knowledge graph to obtain the work group reliability score. Sort the reliability scores of each outsourced work group node in descending order, set the sorting cutoff value according to the descending sorting result, generate a core renewal list, and traverse and match the corresponding low-voltage distribution cabinet node number according to the core renewal list to establish resource scheduling decision instructions. The steps for obtaining the process timing deviation are as follows: Based on the actual operation subgraph, extract the preceding node number, following node number, preceding node timestamp parameter, and following node timestamp parameter corresponding to each operation dependency edge according to the connection direction of the operation dependency edge. Determine whether the preceding node timestamp parameter is later than the following node timestamp parameter. Record each operation dependency edge that has a later relationship. Count the number of operation dependency edges that have a later relationship. Extract the preceding inversion amount. Then calculate the node consumption time of each operation dependency edge according to the order of operation dependency edges. Subtract the node consumption time of the preceding operation dependency edge from the node consumption time of the following operation dependency edge to obtain the consumption time difference corresponding to each edge, forming a timing deviation calculation table. According to the timing deviation calculation table, the distribution record of the pre-inversion amount at each operation dependency edge position is read, the time difference value corresponding to each operation dependency edge is read, and the items are multiplied and accumulated according to the same operation dependency edge position. All multiply and accumulated results are extracted to obtain the pre-inversion penalty value. Then, the maximum time difference value and the minimum time difference value are extracted from the time difference value corresponding to each operation dependency edge. The maximum time difference value is calculated by subtracting the minimum time difference value from the minimum time difference value to obtain the time range value. The pre-inversion penalty value and the time range value are written into the same deviation record to form the process timing deviation amount. The steps for obtaining the reliability score of the work group are as follows: Based on the aforementioned supply chain knowledge graph, the reliability score for the work team is calculated using the following formula: ； in, For the first The reliability score of each outsourced work team node. For the first The preset base score value for each outsourced work team node. This is the counting sequence number of the negative node. For the first The total number of negative nodes associated with each outsourced work group node. For the first The basic negative weights corresponding to each negative node. The preset time decay parameter, For the first Each negative node corresponds to the time difference between the occurrence time of the violation / abnormal work order record and the current evaluation time. For the first The penalty value for illegal aggregation corresponding to each negative node. In the first Within the preset time window preceding the occurrence of the violation / abnormal work order record corresponding to the first negative node, the... The number of other violation work orders corresponding to each outsourced work group node.

2. The AI-based intelligent operation and maintenance decision support method for low-voltage distribution cabinets according to claim 1, characterized in that, The steps for obtaining the actual operation subgraph are as follows: Acquire the power outage nodes, voltage testing nodes, grounding wire connection nodes, and operation-dependent edges of the low-voltage distribution cabinet. Extract the node number, node type, and node sequence identifier of the power outage node, the node number, node type, and node sequence identifier of the voltage testing node, the node number, node type, and node sequence identifier of the grounding wire connection node, and the start-point number, end-point number, and edge direction identifier of the operation-dependent edges. Write the nodes into the node record according to the node number and write the edge connection relationship according to the start-point number and end-point number to form the operation and maintenance standard map. Extract the check-in node number, check-in node type, check-in device location identifier, and timestamp parameter of the check-in node. Locate candidate nodes with the same check-in node number in the operation and maintenance standard map according to the check-in node number. Filter out candidate nodes with inconsistent node types according to the check-in node type and candidate nodes with inconsistent device location identifiers according to the check-in device location identifier. Calculate the parameter comparison difference of the timestamp parameter corresponding to the node sequence identifier for each item. Keep the node records with a parameter comparison difference of zero to form a mapping node set. The mapping node set is arranged sequentially according to the order of the timestamp parameters. The corresponding splicing edges of the mapping node set in the operation and maintenance standard graph are extracted. It is verified whether the start number and end number of each corresponding splicing edge exist in the mapping node set at the same time. Edge records whose start number or end number does not fall into the mapping node set are deleted. The connected edge relationships within the mapping node set are retained. The mapping node set and the corresponding splicing edges are merged to form the actual operation subgraph.

3. The AI-based intelligent operation and maintenance decision support method for low-voltage distribution cabinets according to claim 1, characterized in that, The steps for obtaining the violation and abnormal work order record are as follows: Read the lower and upper bounds of the preset alarm boundary values, determine whether the process timing deviation is less than the lower bound, and determine whether the process timing deviation is greater than the upper bound. Filter out all deviation records that exceed the preset alarm boundary value range. Backtrack the node number, operation dependency edge number, job sequence identifier, timestamp parameter, pre-inversion amount, and time difference corresponding to each deviation record. Write the violation identifier, abnormality source identifier, deviation value identifier, and job parameter record according to the work order fields to form a violation abnormal work order record.

4. The AI-based intelligent operation and maintenance decision support method for low-voltage distribution cabinets according to claim 1, characterized in that, The steps for obtaining the supply chain knowledge graph are as follows: Extract the team identifier, responsibility scope identifier, and associated low-voltage distribution cabinet node number from the outsourced team node; read the work order identifier, work order responsible entity identifier, violation identifier, abnormality source identifier, and operation parameter record from the violation and abnormal work order record; match each work order record with the team identifier and work order responsible entity identifier; filter out the matching team records and work order records to form a team work order correspondence table. Each of the aforementioned violation and abnormal work order records is assigned a negative node, and a negative node identifier, a negative level identifier, and a negative source identifier are written in. Outsourced work group nodes and negative nodes are spliced ​​together according to the matching relationship in the work group work order correspondence table. The starting point identifier, ending point identifier, and association sequence identifier of each splicing edge are recorded to form a supply chain knowledge graph.

5. The AI-based intelligent operation and maintenance decision support method for low-voltage distribution cabinets according to claim 1, characterized in that, The steps for obtaining the core renewal list are as follows: Read the reliability scores of each outsourced work group node one by one, sort them in descending order of reliability scores, record the work group identifier, the reliability score of each group, and the score difference between adjacent groups for each group, retain the complete sorting results, and form a work group sorting sequence. Calculate the sorting cutoff value based on the class group sorting sequence; Extract outsourced work group nodes from the work group sorting sequence whose descending positions are not greater than the sorting cutoff value, and form a core renewal list.

6. The AI-based intelligent operation and maintenance decision support method for low-voltage distribution cabinets according to claim 1, characterized in that, The steps for obtaining the resource scheduling decision instruction are as follows: According to the team identifier in the core renewal list, the associated low-voltage distribution cabinet node number is matched and the high-risk operation permission identifier is written into the associated low-voltage distribution cabinet node number. Then, the high-risk operation permission identifier and management control code are combined and written into the instruction field to form a resource scheduling decision instruction.

7. The system of the AI-based intelligent operation and maintenance decision support method for low-voltage distribution cabinets according to any one of claims 1-6, characterized in that, include: The operation and maintenance map construction module is used to obtain the power outage nodes, voltage testing nodes, grounding wire connection nodes and operation dependency edges of the low-voltage distribution cabinet, establish the operation and maintenance standard map, extract the on-site check-in nodes and timestamp parameters, compare the node parameters in the operation and maintenance standard map, and generate the actual operation sub-map. The anomaly identification and analysis module is used to extract the difference between the pre-inversion amount and the time consumption within the actual operation sub-graph, calculate the process timing deviation, compare the process timing deviation with the preset alarm boundary value, and generate a violation and anomaly work order record. The team reliability assessment module is used to extract the outsourced team node and the violation and abnormal work order record, assign the violation and abnormal work order record to the negative node, associate and splice the outsourced team node and the negative node, establish a supply chain knowledge graph, and calculate the violation and abnormal work order record along the splicing edge in the supply chain knowledge graph to obtain the team reliability score. The resource scheduling decision module is used to sort the reliability scores of each outsourced work group node in descending order, set a sorting cutoff value based on the descending sorting result, generate a core renewal list, and traverse and match the corresponding low-voltage distribution cabinet node numbers according to the core renewal list to establish resource scheduling decision instructions.

Images