Work ticket information synchronization method and system based on semantic recognition

Abstract

The invention belongs to the field of information synchronization, and particularly relates to a work ticket information synchronization method and system based on semantic recognition. The method comprises the following steps: analyzing a work ticket text through a predefined equipment topology ontology, and constructing a task context knowledge graph; using the pre-training language model to identify a to-be-executed operation, and calculating a semantic ambiguity score and an operation risk coefficient; when the semantic ambiguity score exceeds a threshold value, adjusting a weight retrieval historical operation sequence according to an operation risk coefficient to eliminate ambiguity; determining a precondition checking level according to the operation risk coefficient, extracting a corresponding precondition logic condition from the task context knowledge graph, and comparing the precondition logic condition with a real-time state for verification; and encoding the disambiguated and verified operation instruction into a system compatible format and synchronously executing the disambiguated and verified operation instruction. According to the invention, deep semantic understanding and risk differentiation management and control of work ticket information are realized, and the operation accuracy, safety and synchronization efficiency are improved.

Description

Technical Field

[0001] This invention belongs to the field of information synchronization, and in particular relates to a method and system for synchronizing work ticket information based on semantic recognition. Background Technology

[0002] In complex operational scenarios such as power systems and industrial production, work permits are core technical documents that ensure on-site operational safety and standardize equipment operation procedures. The operational instructions they contain directly guide on-site personnel to perform critical tasks such as equipment start-up and shutdown, parameter adjustment, maintenance, and repair. Traditional work permit processing heavily relies on manual operation. Operators must first read and understand the natural language description in the work permit, and then manually convert it into specific instructions that the target control system can execute. The entire process is inefficient and susceptible to human factors.

[0003] Natural language itself is characterized by ambiguity and polysemy, often resulting in unclear wording and complex referential relationships in work orders. Different operators may interpret the same instruction differently, leading to operational deviations. Furthermore, the safety of on-site operations has long relied on the individual experience and sense of responsibility of operators, lacking a systematic risk assessment mechanism to identify potential safety hazards in advance. Once a misoperation occurs, it can not only cause equipment damage and production interruptions, but in severe cases, it can also lead to personal injury accidents, causing significant economic losses and adverse social impacts for the company.

[0004] To address the drawbacks of manual processing, methods for extracting and converting work order information based on natural language processing (NLP) have gradually emerged in the industry. Some solutions employ keyword matching or regular expressions to build rule engines, attempting to automatically parse operation instructions. However, these methods can only identify keyword associations on the surface of the text, failing to handle the flexible and varied language expressions within work orders and unable to delve into the deeper semantic logic behind the instructions. When operation instructions have multiple possible interpretations, existing methods lack reliable semantic disambiguation mechanisms, making it impossible to accurately pinpoint the true operational intent. More critically, existing technologies often neglect differentiated risk assessments of operations, failing to organically integrate operation instructions with key contextual information such as equipment topology, associated safety procedures, and historical fault records. Different devices exhibit significant differences in operational complexity, system correlation, and historical fault conditions, resulting in varying security risk levels. However, existing methods often employ uniform verification standards, failing to implement differentiated security verification based on risk levels. For high-risk operations, comprehensive pre-condition checks are lacking; for low-risk operations, redundant verification processes may exist, failing to guarantee the security of complex operations and impacting the execution efficiency of simple operations.

[0005] Therefore, how to achieve deep semantic understanding of work order text and combine it with risk assessment for intelligent disambiguation and precondition verification is a problem that current technology urgently needs to solve. Summary of the Invention

[0006] In view of this, the present invention proposes a work order information synchronization method and system based on semantic recognition, in order to solve the technical problems in the existing work order processing, such as the semantic ambiguity of natural language instructions making it difficult to accurately parse them, and the lack of a systematic risk assessment and pre-security verification mechanism that combines equipment context, which leads to the reliance on human experience and the risk of misoperation.

[0007] To address the above problems, the technical solution of the work ticket information synchronization method based on semantic recognition proposed in this invention is as follows: The work ticket information synchronization method based on semantic recognition includes the following steps: Based on a predefined device topology ontology, the operation entities, logical relationships and historical operation records in the work order text are parsed to construct a task context knowledge graph. Semantic analysis is performed on the work order text, a pre-trained language model is used to identify the operation to be executed, and a semantic ambiguity score is calculated based on the probability distribution entropy of the candidate interpretations of the operation to be executed; for the target entity of the operation to be executed, the operation risk coefficient is calculated by combining the node degree, the number of associated safety procedures and the historical failure rate in the task context knowledge graph. When the semantic ambiguity score is higher than a preset first threshold, the weight of the similarity retrieval algorithm is adjusted using the operation risk coefficient, and reference operation sequences with similar risk levels are retrieved from the historical operation records of the task context knowledge graph to eliminate semantic ambiguity. The precondition verification level is determined based on the operational risk coefficient, and preconditions of corresponding depth and breadth are extracted from the task context knowledge graph based on the precondition verification level. The extracted preconditions are compared and verified with the real-time status obtained from the target system. The operation instructions confirmed after disambiguation and security verification are encoded into a data format compatible with the target system and synchronized.

[0008] Furthermore, the process of constructing the task context knowledge graph includes: A named entity recognition model is used to extract equipment, switch, and instrument operation entities from the work order text as nodes of the task context knowledge graph; a relation extraction model is used to identify the connection, subordination, and control logic relationships between operation entities as edges of the task context knowledge graph; and operation steps and execution results in historical operation records are associated with relevant equipment nodes to complete the construction of the task context knowledge graph.

[0009] Further, the calculation of the semantic ambiguity score based on the probability distribution entropy of the candidate explanatory terms of the operation to be executed includes: The operation to be performed is input into a pre-trained language model to obtain the output probability distribution of a set of predefined candidate explanatory terms; the information entropy of the output probability distribution is calculated based on the Shannon entropy formula, and the calculated information entropy value is used as the semantic ambiguity score.

[0010] Furthermore, the calculation process for the operational risk coefficient includes: The node degree, the number of associated security procedures, and the historical failure rate of the target entity of the operation to be performed are obtained from the task context knowledge graph; the obtained node degree, the number of associated security procedures, and the historical failure rate are normalized; the normalized values ​​are weighted and summed according to preset weight coefficients to calculate the operation risk coefficient.

[0011] Furthermore, the method for obtaining the reference operation sequence includes: Based on the magnitude of the operational risk coefficient, the weights of features related to safety measures in the similarity retrieval algorithm are adjusted; the difference between the operational risk coefficient of each historical operational record in the retrieval results and the current operational risk coefficient is calculated, and historical operational records with differences within a preset range are determined as reference operational sequences.

[0012] Further, the step of determining the precondition verification level based on the operational risk coefficient, and extracting preconditions of corresponding depth and breadth from the task context knowledge graph based on the precondition verification level, includes: A tiered verification strategy is established. When the operational risk coefficient is lower than a preset second threshold, it is defined as a low-risk level, and the verification scope is the direct associated preceding nodes of the target entity in the task context knowledge graph. When the operational risk coefficient is between the preset second threshold and a preset third threshold, it is defined as a medium-risk level, and the verification scope is extended to the indirect associated preceding nodes of the target entity in the task context knowledge graph. When the operational risk coefficient is higher than the preset third threshold, it is defined as a high-risk level, and the verification scope is further extended to a wider range of system associated nodes.

[0013] Furthermore, the step of encoding the operation instructions, which have been confirmed after disambiguation and security verification, into a data format compatible with the target system and performing synchronization includes: The operation instructions are converted into structured data objects containing target entity identifiers, execution actions, and parameters; the structured data objects are then sent to the target control system via an industrial communication protocol for the system to parse and execute synchronous operations.

[0014] The technical solution of the work ticket information synchronization system based on semantic recognition proposed in this invention is as follows: The work order information synchronization system based on semantic recognition includes the following modules: The module is used to parse the operation entities, logical relationships and historical operation records in the work order text based on the predefined device topology ontology, and build a task context knowledge graph. The calculation module is used to perform semantic analysis on the work order text, identify the operation to be executed using a pre-trained language model, and calculate the semantic ambiguity score based on the probability distribution entropy of the candidate interpretations of the operation to be executed; for the target entity of the operation to be executed, the module calculates the operation risk coefficient by combining the node degree, the number of associated safety procedures, and the historical failure rate in the task context knowledge graph. The retrieval module is used to adjust the weight of the similarity retrieval algorithm using the operation risk coefficient when the semantic ambiguity score is higher than a preset first threshold, and to retrieve reference operation sequences with similar risk levels from the historical operation records of the task context knowledge graph to eliminate semantic ambiguity. The execution module is used to determine the precondition verification level based on the operation risk coefficient, and extract the corresponding depth and breadth of precondition logical conditions from the task context knowledge graph based on the precondition verification level. The extracted precondition logical conditions are compared and verified with the real-time status obtained from the target system. The operation instructions confirmed after disambiguation and security verification are encoded into a data format compatible with the target system and executed synchronously.

[0015] Furthermore, the process of constructing the task context knowledge graph includes: A named entity recognition model is used to extract equipment, switch, and instrument operation entities from the work order text as nodes of the task context knowledge graph; a relation extraction model is used to identify the connection, subordination, and control logic relationships between operation entities as edges of the task context knowledge graph; and operation steps and execution results in historical operation records are associated with relevant equipment nodes to complete the construction of the task context knowledge graph.

[0016] Further, the calculation of the semantic ambiguity score based on the probability distribution entropy of the candidate explanatory terms of the operation to be executed includes: The operation to be performed is input into a pre-trained language model to obtain the output probability distribution of a set of predefined candidate explanatory terms; the information entropy of the output probability distribution is calculated based on the Shannon entropy formula, and the calculated information entropy value is used as the semantic ambiguity score.

[0017] Furthermore, the calculation process for the operational risk coefficient includes: The node degree, the number of associated security procedures, and the historical failure rate of the target entity of the operation to be performed are obtained from the task context knowledge graph; the obtained node degree, the number of associated security procedures, and the historical failure rate are normalized; the normalized values ​​are weighted and summed according to preset weight coefficients to calculate the operation risk coefficient.

[0018] Furthermore, the method for obtaining the reference operation sequence includes: Based on the magnitude of the operational risk coefficient, the weights of features related to safety measures in the similarity retrieval algorithm are adjusted; the difference between the operational risk coefficient of each historical operational record in the retrieval results and the current operational risk coefficient is calculated, and historical operational records with differences within a preset range are determined as reference operational sequences.

[0019] Further, the step of determining the precondition verification level based on the operational risk coefficient, and extracting preconditions of corresponding depth and breadth from the task context knowledge graph based on the precondition verification level, includes: A tiered verification strategy is established. When the operational risk coefficient is lower than a preset second threshold, it is defined as a low-risk level, and the verification scope is the direct associated preceding nodes of the target entity in the task context knowledge graph. When the operational risk coefficient is between the preset second threshold and a preset third threshold, it is defined as a medium-risk level, and the verification scope is extended to the indirect associated preceding nodes of the target entity in the task context knowledge graph. When the operational risk coefficient is higher than the preset third threshold, it is defined as a high-risk level, and the verification scope is further extended to a wider range of system associated nodes.

[0020] Furthermore, the step of encoding the operation instructions, which have been confirmed after disambiguation and security verification, into a data format compatible with the target system and performing synchronization includes: The operation instructions are converted into structured data objects containing target entity identifiers, execution actions, and parameters; the structured data objects are then sent to the target control system via an industrial communication protocol for the system to parse and execute synchronous operations.

[0021] The beneficial effects of this invention are as follows: By constructing a task context knowledge graph, this invention achieves the structured integration of work order operation entities, logical relationships, and historical records, providing comprehensive contextual support for semantic understanding and risk assessment, and solving the pain point of traditional methods' difficulty in deeply associating information. This invention utilizes a pre-trained language model to identify operations to be executed and calculates semantic ambiguity scores through probability distribution entropy. Combined with operation risk coefficients to adjust weights and retrieve historical operation sequences, this effectively eliminates the ambiguity of natural language instructions and improves the accuracy of operation intent recognition. Furthermore, this invention classifies pre-condition verification levels based on operation risk coefficients, implementing differentiated verification scope and depth. This ensures comprehensive safety verification of high-risk operations while avoiding process redundancy in low-risk operations, balancing operation safety and execution efficiency. Finally, the disambiguated and verified operation instructions are encoded into a system-compatible format and executed synchronously, realizing the automated and intelligent conversion of work order information from natural language to system-executable instructions. This significantly reduces the risk of misoperation caused by human intervention and significantly improves the efficiency and reliability of work order processing in scenarios such as power systems and industrial production, providing systematic technical protection for on-site operation safety. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the steps of the work ticket information synchronization method based on semantic recognition of the present invention. Figure 2 A schematic diagram illustrating an example of calculating semantic ambiguity scores. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0024] Specific embodiments of the work ticket information synchronization method based on semantic recognition proposed in this invention: like Figure 1 As shown, the work ticket information synchronization method based on semantic recognition includes the following steps: S1, based on a predefined device topology ontology, parses the operation entities, logical relationships and historical operation records in the work order text to construct a task context knowledge graph.

[0025] Specifically, an ontology editing tool such as Protege is used to pre-construct the power equipment topology ontology, defining equipment entity categories such as transformers, circuit breakers, and disconnectors, as well as relationship types such as connection, dependency, and control. For the input work order text, such as "Operate the high-voltage side disconnector of main transformer No. 1," a BiLSTM-CRF-based named entity recognition model is used to extract the operation entity "High-voltage side disconnector of main transformer No. 1," and a relation extraction model is used to identify the logical relationship between the operation command and the operation entity. Past work orders, equipment defect records, and fault reports are imported from the historical database, and entity and relation extraction is performed in the same way. All parsed entities, relations, and historical records are stored as nodes and edges in graph databases such as Neo4j, forming a task context knowledge graph containing static equipment topology, dynamic operation, and historical information.

[0026] In an optional embodiment, the process of constructing a task context knowledge graph includes: A named entity recognition model is used to extract equipment, switch, and instrument operation entities from the work order text as nodes of the task context knowledge graph; a relation extraction model is used to identify the connection, subordination, and control logic relationships between operation entities as edges of the task context knowledge graph; and operation steps and execution results in historical operation records are associated with relevant equipment nodes to complete the construction of the task context knowledge graph.

[0027] Named entity recognition models can employ a structure combining bidirectional long short-term memory networks and conditional random fields, namely BiLSTM-CRF. For example, if the input work order text is "close switch 101 on the 10kV side of main transformer No. 1 and record the reading of ammeter A1", the named entity recognition model will identify main transformer No. 1 as a device entity, switch 101 as a switch entity, and ammeter A1 as an instrument entity. The identified entities will be added to the knowledge graph as initial nodes, and the logical relationships between these entities will be analyzed through a relation extraction model.

[0028] The relation extraction model is based on a graph neural network, which determines the relation type by learning the information transmission between nodes. In the example above, the model would identify a location relationship between switch 101 and main transformer 1, and a monitoring relationship between ammeter A1 and switch 101. These relationships are then used as directed edges connecting the corresponding entity nodes; for example, node 101 points to node 1 and is marked as "located," and node ammeter A1 points to node 101 and is marked as "monitored." Historical data is integrated into the knowledge graph. For instance, a historical record showing a successful closing operation on switch 101 at 9:00 AM on October 26, 2023, is processed as an event node and associated with the switch 101 node via a `has_history` edge. The event node's attributes include time, operation type, and result, thus giving the knowledge graph a dynamic temporal context.

[0029] S2, perform semantic analysis on the work order text, use a pre-trained language model to identify the operation to be executed, and calculate the semantic ambiguity score based on the probability distribution entropy of the candidate explanatory terms of the operation to be executed; for the target entity of the operation to be executed, calculate the operation risk coefficient by combining the node degree, the number of associated safety procedures and the historical failure rate in the task context knowledge graph.

[0030] Specifically, the instruction text in the work order, such as checking the switch status, is input into a BERT model fine-tuned using power industry corpus. This model outputs multiple possible specific operations and their confidence probabilities, such as checking the on / off indicator light (probability 0.7), checking the remote control status (probability 0.2), and checking the mechanical position (probability 0.1). Based on these probability values, the Shannon entropy formula is used to calculate a semantic ambiguity score; a higher entropy value indicates a more ambiguous instruction. Simultaneously, the target entity switch is located in the task context knowledge graph, the number of connected devices is counted as the node degree, the number of directly associated safety procedure nodes is queried, and its historical failure rate attribute value is read. The operational risk coefficient is then obtained through a weighted summation formula.

[0031] In an optional embodiment, calculating the semantic ambiguity score based on the probability distribution entropy of candidate explanatory terms for the operation to be performed includes: The operation to be performed is input into a pre-trained language model to obtain the output probability distribution of a set of predefined candidate explanatory terms; the information entropy of the output probability distribution is calculated based on the Shannon entropy formula, and the calculated information entropy value is used as the semantic ambiguity score.

[0032] An operation to be performed, such as operating a disconnect switch, is input into a finely tuned pre-trained language model, such as BERT. A set of candidate interpretations related to the disconnect switch is predefined, including closing the disconnect switch, opening the disconnect switch, checking the disconnect switch status, and cleaning the disconnect switch. Figure 2 As shown, the pre-trained language model calculates the similarity or generation probability between the input fuzzy command and each candidate interpretation, outputting a probability distribution vector. For example, the probability distribution output by the pre-trained language model might be: the probability of closing the disconnect switch. The probability of opening the disconnect switch is 0.35. The probability of checking the status of the disconnector switch is 0.40. The probability of cleaning the disconnect switch is 0.20. It is 0.05. Using the Shannon entropy formula... To calculate the information entropy of this probability distribution. Based on the example data, the calculated semantic ambiguity score is: A high entropy value indicates that the intent of the instruction is unclear and requires further disambiguation. A clear instruction, such as opening a disconnect switch, might have a probability distribution where the probability of opening the disconnect switch is 0.98, and the probabilities of other items are extremely low. The calculated entropy value would approach 0, indicating that the semantics are clear.

[0033] In an optional embodiment, the calculation process of the operational risk coefficient includes: The node degree, the number of associated security procedures, and the historical failure rate of the target entity of the operation to be performed are obtained from the task context knowledge graph; the obtained node degree, the number of associated security procedures, and the historical failure rate are normalized; the normalized values ​​are weighted and summed according to preset weight coefficients to calculate the operation risk coefficient.

[0034] Specifically, assuming the target entity of the operation to be performed is node 101 (switch) in the task context knowledge graph, the first step is to extract three key indicators of this node from the task context knowledge graph. For example, the query reveals that the degree of node 101 (switch) is 12, indicating that it is directly connected to 12 other devices or logical nodes; the number of safety procedure nodes associated with this switch node is 4; and through statistical analysis of the associated historical operation record nodes, the historical failure rate is found to be 2%, or 0.02. The second step is to normalize these three raw data to eliminate the influence of dimensions, specifically using the max-min normalization method formula: Assuming the nodal degree ranges from 5 to 25, the number of associated safety procedures ranges from 1 to 8, and the historical failure rate ranges from 0 to 0.05 across the entire power grid, then the normalized values ​​for the switch at node 101 are as follows: Nodal Degree =0.35, number of associated security procedures =0.43, historical failure rate =0.4. Based on preset weights such as node degree weights. =0.2, weight of the number of associated security procedures =0.5, historical failure rate weight =0.3, calculated using the weighted summation formula Calculate the operational risk coefficient. Substituting the above data into the weighted summation formula, we get R = 0.405. The operational risk coefficient represents the potential risk level of the current operation.

[0035] S3, when the semantic ambiguity score is higher than a preset first threshold, the weight of the similarity retrieval algorithm is adjusted using the operation risk coefficient, and reference operation sequences with similar risk levels are retrieved from the historical operation records of the task context knowledge graph to eliminate semantic ambiguity.

[0036] Before the system runs, a first threshold is set, for example, 0.9. If the semantic ambiguity score calculated above is greater than the set first threshold of 0.9, the disambiguation process is initiated. The calculated operation risk coefficient is used as a key parameter to adjust the ranking weight of text similarity retrieval algorithms such as BM25 or vector space models. For example, for a high-risk operation, the text similarity retrieval algorithm will increase the similarity score with operation sequences also marked as high-risk in the historical records; the task context knowledge graph is searched for historical operation sequences that are similar to the current fuzzy instruction text and have a similar risk coefficient. If a historical record is found that involves checking the device number, checking the open / close indicator light, and recording the status, this sequence is recommended to the user as the most likely reference operation, thus clarifying the specific execution steps for checking the switch status.

[0037] In an optional embodiment, the method for obtaining the reference operation sequence includes: Based on the magnitude of the operational risk coefficient, the weights of features related to safety measures in the similarity retrieval algorithm are adjusted; the difference between the operational risk coefficient of each historical operational record in the retrieval results and the current operational risk coefficient is calculated, and historical operational records with differences within a preset range are determined as reference operational sequences.

[0038] Specifically, a similarity retrieval algorithm is used to vectorize historical operation records. The resulting feature vector includes dimensions such as operation type, target entity, operation parameters, and associated security measures. Among these, the feature weights related to security measures are... Adjustments will be made based on the current operational risk coefficient R, using the adjustment formula. ,in, It is the basic weight. It is an amplification factor. Assuming the current operational risk coefficient R = 0.405, the base weight... , =1.5, then the adjusted weight of the safety measures is =0.32, indicating that for high-risk operations, when searching for similar cases, more attention will be paid to whether sufficient security measures were taken in the historical operations.

[0039] After initial similarity ranking and retrieval using adjusted security measure weights, a second screening step is performed to ensure that the reference cases match the risk level of the current operation. This step is based on the operation risk coefficient of each historical operation record pre-stored in the task context knowledge graph: assuming the current operation risk coefficient is 0.405 and the preset filtering difference range is 0.1, the system will traverse the initially retrieved historical operation list, retaining only records with operation risk coefficients between 0.305 and 0.505. For example, a historical operation with high similarity but an operation risk coefficient of 0.8 will be filtered out, while another record with slightly lower similarity but a risk coefficient of 0.45 will be retained, ultimately forming a sequence of reference operations with similar risk levels.

[0040] S4. Determine the precondition verification level based on the operational risk coefficient, and extract the corresponding depth and breadth of precondition logical conditions from the task context knowledge graph based on the precondition verification level. Compare and verify the extracted precondition logical conditions with the real-time status obtained from the target system. Encode the operation instructions confirmed after disambiguation and security verification into a data format compatible with the target system and execute synchronization.

[0041] This step establishes a tiered verification strategy based on operational risk coefficients. The coefficients are mapped to three verification levels: Level 1 for coefficients less than 0.4, Level 2 for coefficients between 0.4 and 0.7, and Level 3 for coefficients greater than 0.7. During Level 1 verification, only nodes at depth 1 in the task context knowledge graph are traversed to extract conditions directly related to the target entity, such as requiring the corresponding circuit breaker to be in the open position for operating a disconnector. During Level 3 verification, the traversal depth of the task context knowledge graph is expanded to 3, extending the verification scope upwards to the power supply side switch status and the activation status of related protection pressure plates. Finally, all extracted logical conditions, such as the open / closed status of circuit breakers and switches, are compared one by one with real-time remote signaling data from the SCADA system via the OPC interface. Execution continues only when all conditions are met.

[0042] Operation instructions that are confirmed to be unambiguous and have passed all precondition checks, such as setting switch #1 to the open position, are encoded into a standardized format compatible with the target system and encapsulated into a JSON or XML message containing fields such as the device's unique identifier, operation type, and timestamp. Subsequently, this message is sent via industrial Ethernet to the interface server of the DCS or remote control system. The server parses the message and issues control commands to the field PLC or smart terminal, thus completing the synchronous execution of the operation and finally receiving the execution result confirmation.

[0043] In an optional embodiment, the precondition verification level is determined based on the operational risk coefficient, and precondition logical conditions of corresponding depth and breadth are extracted from the task context knowledge graph based on the precondition verification level, including: A tiered verification strategy is established. When the operational risk coefficient is lower than a preset second threshold, it is defined as a low-risk level, and the verification scope is the direct associated preceding nodes of the target entity in the task context knowledge graph. When the operational risk coefficient is between the preset second threshold and a preset third threshold, it is defined as a medium-risk level, and the verification scope is extended to the indirect associated preceding nodes of the target entity in the task context knowledge graph. When the operational risk coefficient is higher than the preset third threshold, it is defined as a high-risk level, and the verification scope is further extended to a wider range of system associated nodes.

[0044] Specifically, the second threshold is set to 0.3, and the third threshold is set to 0.7. When the calculated operational risk coefficient R is less than 0.3, it is judged to be at a low risk level. Taking the target entity as switch 101 and its operational risk coefficient as 0.25 as an example, the system initiates a low-risk check, searches the task context knowledge graph for nodes directly connected to the switch 101 node through the is_precondition_of relationship, such as disconnector switch 1011 being in the open state and grounding switch 1012 being in the closed state, and only checks the direct preconditions.

[0045] If the calculated operational risk coefficient R is 0.405, falling between 0.3 and 0.7, it is classified as a medium-risk level. At this point, the scope of the verification will expand. In addition to verifying the aforementioned direct prerequisites, a second-order traversal will be performed along the edges of the task context knowledge graph to verify indirectly related nodes. For example, it will check whether the status of the upstream 35 kV 202 switch supplying power to the line containing switch 101 is normal, and simultaneously check whether the relevant protection devices have been deactivated as required. If the operational risk coefficient R is as high as 0.8, exceeding the third threshold of 0.7, a high-risk verification will be initiated. The verification scope will be further expanded to system-level related nodes. In addition to covering all the aforementioned electrical conditions, queries will be added for system-level information that is not directly electrically related but affects operational safety, such as whether the current overall power flow is stable and whether there are severe weather warnings.

[0046] In an optional embodiment, encoding the operation instructions, which have been confirmed after disambiguation and security verification, into a target system-compatible data format and performing synchronization includes: The operation instructions are converted into structured data objects containing target entity identifiers, execution actions, and parameters; the structured data objects are then sent to the target control system via an industrial communication protocol for the system to parse and execute synchronous operations.

[0047] First, the verified operation instructions are converted into standardized JSON objects. These objects contain a `target_id` field specifying the unique ID of the device, an `action` field specifying the action type, and a `parameter` field containing additional parameters. Then, the JSON object is sent to the field control unit via an industrial communication protocol (such as OPC UA). The programmable logic controller (PLC) or remote terminal unit (RTU) deployed in the field receives and parses the JSON object, locates the physical device based on the identifier, and executes the operation instructions. After the operation is completed, the control unit sends a confirmation message to the system containing the execution status, time, and feedback data, thus completing the closed-loop process of instruction execution and status synchronization.

[0048] Specific embodiments of the work ticket information synchronization system based on semantic recognition proposed in this invention are as follows: The work order information synchronization system based on semantic recognition includes the following modules: The module is used to parse the operation entities, logical relationships and historical operation records in the work order text based on the predefined device topology ontology, and build a task context knowledge graph. The calculation module is used to perform semantic analysis on the work order text, identify the operation to be executed using a pre-trained language model, and calculate the semantic ambiguity score based on the probability distribution entropy of the candidate interpretations of the operation to be executed; for the target entity of the operation to be executed, the module calculates the operation risk coefficient by combining the node degree, the number of associated safety procedures, and the historical failure rate in the task context knowledge graph. The retrieval module is used to adjust the weight of the similarity retrieval algorithm using the operation risk coefficient when the semantic ambiguity score is higher than a preset first threshold, and to retrieve reference operation sequences with similar risk levels from the historical operation records of the task context knowledge graph to eliminate semantic ambiguity. The execution module is used to determine the precondition verification level based on the operation risk coefficient, and extract the corresponding depth and breadth of precondition logical conditions from the task context knowledge graph based on the precondition verification level. The extracted precondition logical conditions are compared and verified with the real-time status obtained from the target system. The operation instructions confirmed after disambiguation and security verification are encoded into a data format compatible with the target system and executed synchronously.

[0049] It should be noted that the processing procedures of the construction module, calculation module, retrieval module and execution module are the same as the processing procedures of steps S1, S2, S3 and S4 in the various embodiments of the work ticket information synchronization method based on semantic recognition, and will not be described in detail here.

[0050] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A work order information synchronization method based on semantic recognition, characterized in that, Includes the following steps: Based on a predefined device topology ontology, the operation entities, logical relationships and historical operation records in the work order text are parsed to construct a task context knowledge graph. Semantic analysis is performed on the work order text, a pre-trained language model is used to identify the operation to be executed, and a semantic ambiguity score is calculated based on the probability distribution entropy of the candidate explanatory terms of the operation to be executed. For the target entity of the operation to be performed, the operation risk coefficient is calculated by combining the node degree, the number of associated security procedures, and the historical failure rate in the task context knowledge graph. When the semantic ambiguity score is higher than a preset first threshold, the weight of the similarity retrieval algorithm is adjusted using the operation risk coefficient, and reference operation sequences with similar risk levels are retrieved from the historical operation records of the task context knowledge graph to eliminate semantic ambiguity. The precondition verification level is determined based on the operational risk coefficient, and preconditions of corresponding depth and breadth are extracted from the task context knowledge graph based on the precondition verification level. The extracted preconditions are compared and verified with the real-time status obtained from the target system. The operation instructions confirmed after disambiguation and security verification are encoded into a data format compatible with the target system and synchronized.

2. The work order information synchronization method based on semantic recognition according to claim 1, characterized in that, The process of constructing the task context knowledge graph includes: A named entity recognition model is used to extract equipment, switch, and instrument operation entities from the work order text as nodes of the task context knowledge graph; a relation extraction model is used to identify the connection, subordination, and control logic relationships between operation entities as edges of the task context knowledge graph; and operation steps and execution results in historical operation records are associated with relevant equipment nodes to complete the construction of the task context knowledge graph.

3. The work order information synchronization method based on semantic recognition according to claim 1, characterized in that, The calculation of semantic ambiguity score based on the probability distribution entropy of candidate explanatory terms for the operation to be executed includes: The operation to be performed is input into a pre-trained language model to obtain the output probability distribution of a set of predefined candidate explanatory terms; the information entropy of the output probability distribution is calculated based on the Shannon entropy formula, and the calculated information entropy value is used as the semantic ambiguity score.

4. The work order information synchronization method based on semantic recognition according to claim 1, characterized in that, The calculation process for the operational risk coefficient includes: The node degree, the number of associated security procedures, and the historical failure rate of the target entity of the operation to be performed are obtained from the task context knowledge graph; the obtained node degree, the number of associated security procedures, and the historical failure rate are normalized; the normalized values ​​are weighted and summed according to preset weight coefficients to calculate the operation risk coefficient.

5. The work order information synchronization method based on semantic recognition according to claim 1, characterized in that, The method for obtaining the reference operation sequence includes: Based on the magnitude of the operational risk coefficient, the weights of features related to safety measures in the similarity retrieval algorithm are adjusted; the difference between the operational risk coefficient of each historical operational record in the retrieval results and the current operational risk coefficient is calculated, and historical operational records with differences within a preset range are determined as reference operational sequences.

6. The work order information synchronization method based on semantic recognition according to claim 1, characterized in that, The process of determining the precondition verification level based on the operational risk coefficient, and extracting preconditions of corresponding depth and breadth from the task context knowledge graph based on the precondition verification level, includes: A tiered verification strategy is established. When the operational risk coefficient is lower than a preset second threshold, it is defined as a low-risk level, and the verification scope is the direct associated preceding nodes of the target entity in the task context knowledge graph. When the operational risk coefficient is between the preset second threshold and a preset third threshold, it is defined as a medium-risk level, and the verification scope is extended to the indirect associated preceding nodes of the target entity in the task context knowledge graph. When the operational risk coefficient is higher than the preset third threshold, it is defined as a high-risk level, and the verification scope is further extended to a wider range of system associated nodes.

7. The work order information synchronization method based on semantic recognition according to claim 1, characterized in that, The process of encoding the operation instructions, which have been confirmed after disambiguation and security verification, into a data format compatible with the target system and performing synchronization includes: The operation instructions are converted into structured data objects containing target entity identifiers, execution actions, and parameters; the structured data objects are then sent to the target control system via an industrial communication protocol for the system to parse and execute synchronous operations.

8. A work order information synchronization system based on semantic recognition, characterized in that, Includes the following modules: The module is used to parse the operation entities, logical relationships and historical operation records in the work order text based on the predefined device topology ontology, and build a task context knowledge graph. The calculation module is used to perform semantic analysis on the work order text, identify the operation to be executed using a pre-trained language model, and calculate the semantic ambiguity score based on the probability distribution entropy of the candidate explanatory terms of the operation to be executed. For the target entity of the operation to be performed, the operation risk coefficient is calculated by combining the node degree, the number of associated security procedures, and the historical failure rate in the task context knowledge graph. The retrieval module is used to adjust the weight of the similarity retrieval algorithm using the operation risk coefficient when the semantic ambiguity score is higher than a preset first threshold, and to retrieve reference operation sequences with similar risk levels from the historical operation records of the task context knowledge graph to eliminate semantic ambiguity. The execution module is used to determine the precondition verification level based on the operation risk coefficient, and extract the corresponding depth and breadth of precondition logical conditions from the task context knowledge graph based on the precondition verification level. The extracted precondition logical conditions are compared and verified with the real-time status obtained from the target system. The operation instructions confirmed after disambiguation and security verification are encoded into a data format compatible with the target system and executed synchronously.

9. The work order information synchronization system based on semantic recognition according to claim 8, characterized in that, The process of constructing the task context knowledge graph includes: A named entity recognition model is used to extract equipment, switch, and instrument operation entities from the work order text as nodes of the task context knowledge graph; a relation extraction model is used to identify the connection, subordination, and control logic relationships between operation entities as edges of the task context knowledge graph; and operation steps and execution results in historical operation records are associated with relevant equipment nodes to complete the construction of the task context knowledge graph.

10. The work order information synchronization system based on semantic recognition according to claim 8, characterized in that, The calculation of semantic ambiguity score based on the probability distribution entropy of candidate explanatory terms for the operation to be executed includes: The operation to be performed is input into a pre-trained language model to obtain the output probability distribution of a set of predefined candidate explanatory terms; the information entropy of the output probability distribution is calculated based on the Shannon entropy formula, and the calculated information entropy value is used as the semantic ambiguity score.

Images