A power intelligent orchestration ticketing method fusing semantic intermediate representation

Abstract

The application discloses a power intelligent arrangement ticketing method fusing semantic intermediate representation and belongs to the technical field of power dispatching. The method obtains a single mode ticketing request, completes intention recognition, extracts business semantics, constructs semantic intermediate representation and splits operation groups; candidate devices are matched in multiple dimensions, device information is supplemented in combination with a database, and topological path checking is carried out through a power grid weighted graph; operation task sequences are generated according to the order of dispatching regulations, device state memory is maintained, and cross-operation group state chain checking is performed; task quality scores are calculated by comprehensively considering device supplement, topology, state chain and semantic analysis, ticketing support levels are divided, and in-storage management and control are realized. The application solves the problems of device misplacement caused by various single mode expressions and multi-group state connection confusion, establishes a traceable analysis link, reduces the risk of wrong and missed tickets through multiple checking, and improves the accuracy and reliability of automatic generation of power operation tickets.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching, and in particular to a method for intelligent power orchestration and ticketing that integrates semantic intermediate representations. Background Technology

[0002] Power dispatch operation tickets are crucial for implementing power outages, busbar switching, power transfer, and decommissioning in power grid dispatch operations. Their content must accurately express the equipment, operational tasks, pre- and post-operation states, and relevant safety constraints. With the continuous accumulation of mode sheets, maintenance requests, and equipment model data in the dispatch operation management system, dispatchers must simultaneously understand natural language mode sheets, verify equipment records, query topology connections, and determine the state transitions between different operation groups when creating operation tickets. Because mode sheet texts come from multiple sources and have significant variations in expression, and the same equipment may have multiple forms of expression such as abbreviated names, dispatch numbers, and system equipment identifiers, the traditional method of relying on manual reading and applying typical tickets is easily affected by factors such as differences in experience, inconsistent textual expressions, equipment aliases, and changes in operating modes.

[0003] Existing automated operation ticket generation technologies typically rely on fixed templates, typical ticket databases, manual rules, or matching with single equipment ledgers. These technologies can only provide auxiliary ticket generation capabilities for some standardized scenarios and still have significant shortcomings in complex scenarios: the semantic understanding of the operation ticket depends on manual reading, the understanding results vary from person to person, and the correspondence between the generated result and the original operation ticket is difficult to trace, making verification and location difficult; when equipment names, substations, dispatch numbers, and equipment identifiers coexist, relying on text keyword matching is prone to equipment mismatch, and detailed equipment needs to be manually checked item by item, which is prone to omissions; when the operation ticket contains multiple time periods or multiple power outage and restoration groups, the connection of equipment status needs to be manually remembered and calculated, which is labor-intensive and prone to inconsistencies between previous and subsequent statuses; the reliability of the ticket generation results is low, the ticket generation results are inaccurate, and there are scheduling safety risks. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a more accurate method for intelligent power scheduling and ticketing.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] A method for intelligent power ticket orchestration that integrates semantic intermediate representations, the key of which includes the following steps:

[0007] Step S1: Obtain the method order ticketing request, identify its intent, and then apply the corresponding ticketing skill;

[0008] Step S2: Extract business semantics from the structured fields in the method sheet, split the business semantics into operation groups, and construct a semantic intermediate representation based on the business semantics; the semantic intermediate representation records the device object, semantic source, target status, whether it is explicitly named, whether it belongs to the secondary device boundary, and whether there is any incompletely parsed content in the structured fields, and records the operation group to which each business semantic belongs;

[0009] Step S3: Match the semantic fragments in the semantic intermediate representation with each device candidate in the device candidate set to obtain the matching device with the highest matching degree with the semantic fragment;

[0010] Step S4: Query the database based on the matched devices to complete the device information and obtain the device completion result;

[0011] When a candidate device can be uniquely mapped to a device ID after verification of the plant, device type, voltage level, dispatch number, and alias information, and the operation task does not require association with device details, the device is used as the matching result; otherwise, the system further queries the device's power status, adjacent devices, connection paths, and topology range, determines the effective topology path with the smallest cumulative weight among the candidate topology paths that meet the business constraints, and cross-validates the database completion results with the topology discovery results.

[0012] Step S5: Organize the operation task sequence according to the semantic intermediate representation and device completion result. The operation task sequence is generated in a preset order, wherein the pre-operation state of the next operation task inherits the post-operation state of the previous operation task. Further, the state chain verification of the generated operation tasks is performed across operation groups to obtain the state chain verification result.

[0013] Step S6: Calculate the task quality score based on the generation status of the operation task, the completeness of the equipment completion result, the pass rate of the topology verification result, the continuity of the state chain verification result, and the alarm items. Determine the support level of the ticketing result based on the task quality score and the alarm status, and perform warehousing control according to the support level.

[0014] Preferably, in step S1, the method order ticketing request includes at least one parameter among the method order number, method order JSON, whether to formally save, and whether to allow alarm entry into the database; and based on the parameters in the request, the business scenario to which the method order ticketing request belongs is identified, the business scenario includes ticket generation, task pre-inspection, formal entry into the database, rollback verification, and result review, and then the corresponding ticketing skill is selected according to the business scenario.

[0015] Preferably, in step S2, the structured fields include the moderation order number, work content, power outage scope, approval start time, approval end time, power outage mode arrangement, power supply mode arrangement, coordination with power outage matters, and relay protection adjustment;

[0016] The split operation grouping refers to the system splitting a mode order with multiple time periods or multiple power outage ranges into multiple operation groups according to the time overlap relationship and power outage and restoration logic.

[0017] Preferably, in step S3, the semantic fragment in the semantic intermediate representation is matched with each device candidate in the device candidate set to obtain the matching device with the highest matching degree with the semantic fragment. The matching degree is determined by text similarity, device type consistency, plant consistency and contextual business consistency. If the matching degree of all candidate devices is lower than the preset requirement or the matching degree of multiple candidate devices is close, an alarm is output.

[0018] Preferably, in step S4, the equipment data completion specifically involves querying the plant / station table, equipment mapping table, line endpoint table, and equipment basic information table based on the equipment name, plant / station name, voltage level, equipment type, dispatch number, and alias information to complete the main equipment ID, equipment type, affiliated plant / station, and dispatch name required for the operation task.

[0019] The specific steps for determining the effective topology path with the minimum cumulative weight among candidate topology paths that satisfy business constraints are as follows:

[0020] The power grid equipment topology is abstracted as a weighted graph G = (V, E, W), where V is the set of equipment nodes, E is the set of equipment connection edges, and W is the set of weights between nodes in the weighted graph. For any two equipment nodes u and v, in the candidate topology path set PathsΩ(u, v) that satisfies the business constraint Ω, the path with the smallest cumulative value of the connection edge weight w(a, b) of each pair of adjacent equipment nodes (a, b) is selected as the effective topology path P*(u, v).

[0021] When there is no valid topology path, or when the device type, operating status, and affiliated plant of the device in the valid topology path are inconsistent with the semantics of the mode sheet, a topology verification alarm will be output.

[0022] Preferably, in step S5, the operation task sequence is organized based on the semantic intermediate representation and the device completion result;

[0023] The operation tasks are formed in the following order: power outage mode task, line or switch busbar switching task, busbar switching task, power outage range task, power restoration range task, busbar restoration task, and power restoration mode restoration task. For each task generated, the system will record the current status of the relevant equipment so that the pre-operation status of the next task can inherit the post-operation status of the previous task.

[0024] Then, the generated operation task is checked for state chain verification. The state before and after the operation of the same operation task is checked against the state remembered by the device. If the verification fails, a state verification alarm is issued.

[0025] The beneficial effects of adopting the above technical solution are as follows:

[0026] This invention integrates the semantics of the moderation form, equipment ledger, power grid topology, and equipment status rules into a unified calculation and verification process. First, the business semantics of the moderation form are mapped into a structured semantic intermediate representation. Equipment objects, target states, and the original text of the moderation form are matched item by item, quantifying the equipment matching degree and filtering out suspicious matches. This solves the problem of equipment mismatch caused by multiple source aliases and expression differences, making equipment identification more accurate. In the generated task sequence, the state transition of each device is based on state memory and verified through cross-group state chains, making the state connection before and after the operation more accurate and avoiding errors, omissions, and state inconsistencies. As a result, the generated operation tasks have multiple verification guarantees at the three levels of equipment objects, task details, and operation status, significantly reducing the error and omission rate, and resulting in high reliability and accuracy of the ticketing results. Attached Figure Description

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0028] Figure 1 This is a flowchart of a power intelligent orchestration ticketing method that integrates semantic intermediate representation, as proposed in this invention. Detailed Implementation

[0029] To make the above-mentioned objectives, features, and advantages of the present invention more apparent and understandable, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] This invention proposes a power intelligent orchestration ticketing method that integrates semantic intermediate representation. It comprehensively utilizes the natural language understanding, skill orchestration, and tool invocation capabilities of a large language model agent, and combines semantic intermediate representation for power dispatch mode sheets with a state chain verification mechanism based on power grid topology. This achieves closed-loop processing from mode sheet input, semantic parsing, equipment completion, task orchestration, state verification, task quality evaluation to rule pre-checking and database entry. The overall system flowchart is as follows. Figure 1 As shown, it includes:

[0031] Step S1: Obtain the method order ticketing request, identify its intent, and then apply the corresponding ticketing skill;

[0032] The method order ticketing request includes at least one parameter among the following: method order number, method order JSON, whether to save formally, and whether to allow alarm entry into the database. Based on the parameters in the request, the business scenario to which the method order ticketing request belongs is identified. The business scenario includes ticket generation, task pre-inspection, formal entry into the database, rollback verification, and result review. Then, the corresponding ticketing skill is selected according to the business scenario.

[0033] In this embodiment, the dispatcher can submit a request in natural language through a dialogue interface, or the upstream system can submit a request message carrying parameters through an API interface. The message must contain at least one of the method slip number or method slip JSON. The system first performs permission verification on the request to confirm whether the submitting account has the permission to complete the ticketing operation; then it performs a field integrity check. Subsequently, the agent inputs the request text and parameters into the intent recognition stage: parameters such as method slip number, method slip JSON, whether to save formally, and whether to allow alarm entry into the database are normalized into executable skill invocation tasks. The large language model, combined with the documentation of each ticketing skill, determines whether the current request belongs to a ticketing generation, task pre-inspection, formal entry into the database, rollback verification, or result review scenario. For example, if the request parameters "whether to save formally" are no and "whether to allow alarm entry into the database" are no, it can be identified as a task pre-inspection scenario, and the corresponding ticketing skill is invoked. The input parameters, tool access records, and generated results of each skill invocation are written to the audit log, forming a closed-loop link of "user intent - skill selection - controlled execution - result auditing". The process does not rely on fixed examples, nor is the final operation ticket directly written by the model.

[0034] Step S2: Extract business semantics from the structured fields in the method sheet, split the business semantics into operation groups, and construct a semantic intermediate representation based on the business semantics; the semantic intermediate representation records the device object, semantic source, target status, whether it is explicitly named, whether it belongs to the secondary device boundary, and whether there is any incompletely parsed content in the structured fields, and records the operation group to which each business semantic belongs;

[0035] The structured fields include the moderation order number, work content, power outage scope, approved start time, approved end time, power outage mode arrangement, power supply mode arrangement, coordination matters for power outage, and relay protection adjustments;

[0036] Operation grouping refers to the system splitting operation orders with multiple time periods or multiple power outage areas into multiple operation groups according to time overlap and power outage / restoration logic. Operation groups are a set of maintenance tasks formed by merging power outage areas with direct or cascading overlap in the approved time intervals. Each operation group has a group number, start time, end time, power outage area, power outage method arrangement, and power restoration arrangement.

[0037] "Time overlap relationship" refers to the judgment of whether two time periods overlap in time; "Power outage and power supply logic" refers to the business correspondence between power outage operations and power supply operations, such as the pairing of power outage groups and power supply groups for the same equipment.

[0038] In this embodiment, after the ticket processing skill reads the content of the method slip, it first performs standardization processing, including but not limited to date normalization format and correction of obvious typos according to the preset power equipment naming standard. Then, business semantics are extracted from the structured fields, including: for the method slip JSON, it is directly read by key name; for natural language text, the large language model extracts and backfills each item according to the field definition, and the content that cannot be extracted is marked as empty.

[0039] The rules for grouping and splitting are as follows: extract the approval start time and approval end time of each power outage item to form a time interval. If the two intervals have an intersection and the corresponding power outage ranges are logically related in terms of power outage and restoration, they are merged into the same operation group; otherwise, they are split into different groups.

[0040] Construct a semantic intermediate representation, which retains structured fields such as power outage range, power outage method change, power supply method change, and relay protection method. These fields are used to record the device object, semantic source, target status, whether it is explicitly named, whether it belongs to the secondary equipment boundary, whether there is any incompletely parsed content, and to mark the operation group number to which it belongs.

[0041] Based on the time interval intersection judgment and the grouping and splitting of power outage and restoration logic, the complex task with multiple time periods and multiple power outage ranges can be decomposed into operation groups that can generate tasks independently and are interconnected; the semantic intermediate representation records the semantic source and parsing status one by one, providing a structured traceability basis for subsequent state chain verification and manual review.

[0042] Step S3: Match the semantic fragment in the semantic intermediate representation with each device candidate in the device candidate set to obtain the matching device with the highest matching degree of the semantic fragment; match the semantic fragment in the semantic intermediate representation with each device candidate in the device candidate set to obtain the matching device with the highest matching degree of the semantic fragment, wherein the matching degree is determined by text similarity, device type consistency, plant consistency and contextual business consistency. If the matching degree of all candidate devices is lower than the preset requirement or the matching degree of multiple candidate devices is close, an alarm is output.

[0043] Text similarity refers to the degree of literal or semantic closeness between the text of a semantic fragment and the candidate name of an equipment; equipment type consistency refers to whether the equipment category referred to by the semantic fragment matches the type of the candidate equipment in the ledger; plant consistency refers to whether the plant to which the semantic fragment belongs is the same as the plant to which the candidate equipment belongs; contextual business consistency refers to the degree of business matching between the candidate equipment and the context in which the semantic fragment is located (such as the line mentioned in the same sentence, the power outage range of the group to which it belongs).

[0044] In the embodiment, let the set of semantic fragments be... The candidate set of devices is The matching degree is represented by the overall matching score, as shown below.

[0045] (1)

[0046] Among all candidate devices, the system selects the device with the highest score that meets the threshold requirement as the matching result, as shown below.

[0047] (2)

[0048] in, For the first A semantic fragment, The first device candidate; , , and These represent text similarity, device type consistency, plant / site consistency, and contextual business consistency, respectively. The overall matching of the content between the devices in the candidate set and the current semantic segment is evaluated based on the similarity of multiple dimensions. , , , The corresponding weights are obtained from historical sample data; Set a threshold for the device. The optimal matching device is selected. If the candidate score is insufficient or multiple candidate scores are close, the system will record the corresponding content as an uncertain item and generate an alarm.

[0049] Among them, text similarity is a weighted sum of semantic vector cosine similarity and edit distance normalization value to take into account cases such as "5021 switch" and "5021 circuit breaker" which have the same meaning but different words; equipment type consistency and plant consistency are taken as two values, with consistency being 1 and inconsistency being 0; contextual business consistency is based on whether the candidate device has a direct topological connection relationship with the confirmed device in the same sentence, with existence being 1, non-existence being 0.5, and conflict being 0.

[0050] The system iterates through all candidate devices and selects the candidate with the highest score as the matching result, but it must meet two conditions simultaneously: its score is not lower than the matching threshold, and the difference between its score and the second-highest score candidate is not less than 0.05. If all candidate scores are lower than the threshold, or the difference between the highest and second-highest scores is less than 0.05 (i.e., multiple candidate scores are close), the system records the corresponding semantic fragment as an uncertain item and outputs an alarm, retaining the original text information for manual specification, rather than forcibly matching.

[0051] Equipment matching is determined by four dimensions: text, type, plant / site, and context. This effectively solves the problem of multi-source aliases caused by the coexistence of abbreviated equipment names, scheduling numbers, and system identifiers, and significantly reduces the mismatch rate caused by relying solely on keyword matching. The weights and thresholds are all derived from historical data calibration, and the matching strategy can be continuously optimized as samples accumulate. Two abnormal situations, namely insufficient scores and close candidates, are explicitly identified and transformed into uncertain item alarms, preventing suspicious matches from silently entering the subsequent task generation stage and providing a pre-emptive guarantee for the accuracy of equipment completion.

[0052] Step S4: Query the database based on the matched devices to complete the device information and obtain the device completion result;

[0053] Relying solely on the device name is insufficient to confirm the task details. The system further queries the device's power status, adjacent devices, connection paths, and topology range. Among the candidate topology paths that meet the business constraints, it determines the effective topology path with the smallest cumulative weight and cross-validates the database completion results with the topology discovery results.

[0054] The equipment data completion process involves querying the plant / station table, equipment mapping table, line endpoint table, and equipment basic information table based on the equipment name, plant / station name, voltage level, equipment type, dispatch number, and alias information to complete the main equipment ID, equipment type, affiliated plant / station, and dispatch name required for the operation task. All of the above tables are stored in a relational database, and the main equipment ID refers to the unique identifier of the equipment in the system.

[0055] In this embodiment, based on the equipment name, substation name, voltage level, equipment type, dispatch number, and alias information in the semantic fragment, the equipment mapping table is first consulted to map the alias and dispatch number to candidate equipment identifiers. Then, the substation table, line endpoint table, and equipment basic information table are combined to verify and complete the main equipment ID, equipment type, affiliated substation, and dispatch name required for the operation task. For cases where only equipment abbreviations appear in the text or multiple objects are listed consecutively in the same sentence, the context inheritance rule is adopted. For example, if the substation is omitted in the second equipment in the text "500kV Jia Substation 5011, 5012 switches", the substation and voltage level of the preceding text are inherited to narrow down the candidate range. For scenarios such as line busbar switching, switch transfer, busbar switching, and main transformer power outage, relying solely on equipment names is insufficient to confirm task details. The system further calls topology services to query the equipment's energized status, adjacent equipment, connection paths, and topology range, automatically completing the associated detailed equipment such as switches, trolleys, disconnectors, busbars, and line endpoints. The introduction of topology services enables the automatic completion of detailed equipment such as switches and disconnectors in complex scenarios such as busbar switching and power transfer, reducing the burden of manual verification.

[0056] During topology path verification, the power grid equipment topology is abstracted into a weighted graph. ,in Represents a set of device nodes. Represents the set of device connection edges. Represent the set of edge weights;

[0057] For any two device nodes u and v, in the candidate topology path set PathsΩ(u,v) that satisfies the business constraint Ω, the shortest path algorithm selects the path with the smallest cumulative value of the connection edge weight w(a,b) of each adjacent node pair (a,b) as the valid topology path P*(u,v). If no valid path exists, or if the device type, operating status, or affiliated plant / station of the device in the path is inconsistent with the semantics of the mode sheet, the system outputs a topology verification alarm. Simultaneously, the system compares the device list supplemented in the database with the device list discovered by the topology discovery one by one. Devices lacking unique identifiers or whose results are inconsistent are retained as alarms and not directly entered into the formal database. The business constraint Ω here is a path selection condition generated based on the semantic intermediate representation, basic device information, real-time operating status, and scheduling rules, including device type constraints, voltage level constraints, plant / station constraints, device number constraints, operating status constraints, and topology uniqueness constraints.

[0058] Weighted graph shortest path verification transforms business constraints into a computable graph search problem. Cross-verification from two independent sources—database and topology—forms a mutual verification mechanism, exposing and alerting risks such as missing equipment, topology breaks, and semantic inconsistencies before task generation, effectively preventing erroneous details from entering formal operation tasks.

[0059] Step S5: Organize the operation task sequence based on the semantic intermediate representation and device completion results. The operation task sequence is generated in a preset order, where the pre-operation state of the next operation task inherits the post-operation state of the previous operation task. Further, the generated operation tasks are checked across operation groups to obtain the state chain check result. The device memory state refers to the current state recorded in the state memory object maintained by the system for each device.

[0060] In this embodiment, the system determines the target status of each device according to the target status classification rules: for specific primary equipment clearly included in the power outage scope, the power outage target status is generated as maintenance; for equipment such as switches, trolleys, and handcarts not explicitly named but associated with main transformer, line, or busbar tasks, the target status is generated as cold standby, to avoid mistakenly extending the power outage of the main equipment to the maintenance of associated equipment; the operation of disconnectors and grounding switches is generated as opening or closing; PT type equipment is only generated as running to maintenance tasks when it is clearly included in the power outage scope in the mode sheet. The operation tasks are formed in the following order: power outage mode task, line or switch busbar switching task, busbar busbar switching task, power outage scope task, power restoration scope task, busbar restoration task, and power restoration mode restoration task. This order is arranged according to the safety procedures for dispatching switching operations, ensuring that mode adjustment precedes power outage operation and restoration operation follows power restoration operation. The system establishes a state memory object for each device, recording the device's initial state, current state, target state, task source, whether it is explicitly named, whether it is completed by topology association, and its corresponding operation group. Each time a task is generated, the current state of the relevant devices is updated, ensuring that the pre-operation state of the subsequent task inherits the post-operation state of the previous task. For example, when a line transitions from operation to maintenance, tasks such as "switch transitioning from operation to hot standby," "switch transitioning from hot standby to cold standby," "line-side disconnector opened," and "line transitioning to maintenance" are generated sequentially. The pre-operation state of the subsequent task is precisely the post-operation state of the previous task.

[0061] When the mode order declares "mode not restored", the system suppresses mirror restoration tasks that are the opposite of the mode arrangement; the power supply range tasks are organized in reverse order of the power outage tasks, so that the equipment can be restored from the state after the power outage to the initial state before the power outage; for multi-time period groups, the initial state of the later group inherits the end state of the previous group.

[0062] During state chain verification, the system compares the pre-operation state, post-operation state of each task on the same device with the device's memory state one by one. For example, each task is measured by a state consistency factor, which is determined by the product of two indicator functions: the pre-operation state and the device's memory state, and the post-operation state and the state transition result. The state chain is continuous when the product of all task factors is 1. Once a chain break occurs, for example, if the pre-operation state of a task is running but the device's memory state is already cold standby, the system will generate a state verification alarm and block the saving.

[0063] The beneficial effects of the above technical solution are as follows: the target state classification rule distinguishes between explicitly named equipment and associated supplementary equipment, avoiding excessive operations where a power outage of the main equipment is mistakenly extended to maintenance of auxiliary equipment; the fixed task generation order conforms to the scheduling safety procedure, ensuring the sequential relationship of each stage of mode adjustment, power outage, power restoration, and restoration; the equipment state memory object runs through multiple operation groups, enabling subsequent tasks to automatically inherit the state of the previous one, and in conjunction with the mode non-restoration suppression and power restoration reverse order mechanism, effectively eliminating state chain breaks, duplicate restorations, and missed restorations in cross-group scenarios; the state chain consistency check intercepts state connection errors before entering the database, significantly reducing the risk of incorrect or missed tickets.

[0064] Step S6: Calculate the task quality score based on the generation status of the operation task, the completeness of the equipment completion result, the pass rate of the topology verification result, the continuity of the state chain verification result, and the alarm items. Determine the support level of the ticketing result based on the task quality score and the alarm status, and perform warehousing control according to the support level.

[0065] Based on task generation, equipment completion completeness, topology verification pass rate, state chain continuity, semantic parsing completeness, and alarm status, the result quality is categorized into fully supported, partially supported, and unsupported. Fully supported indicates a valid task has been generated without alarms; partially supported indicates a primary equipment task has been generated, but equipment completion, topology verification, or semantic parsing alarms still exist; unsupported indicates that a valid task cannot be formed. Secondary equipment content such as relay protection, automatic transfer switch, setting settings, and pressure plates can be parsed as notes or debugging fields for manual interpretation, but it is not written into the primary equipment operation task and does not directly lead to the result being judged as partially supported.

[0066] If the equipment completeness is The pass rate for topology verification is The continuity of the state chain is Semantic parsing completeness is The warning and penalty items are: Task quality score It can be expressed by formula (6).

[0067] (6)

[0068] After obtaining the task quality score, the system gives the final support level according to formula (7).

[0069] (7)

[0070] in, , , and These represent device completion completeness, topology verification pass rate, state chain continuity, and semantic parsing completeness, respectively. For warning and penalty items, to These are the weighting coefficients, which can be calibrated using historical regression samples and the validation set. Considering power operation safety, and Set the weight to be no less than that of device name matching and semantic parsing, but the specific value should be based on test results. Score the quality of the task; and These are the thresholds for full support and partial support, and the thresholds can be determined based on historical data. A set of operation tasks; To support the level determination function, the determination rule is as follows: , and When it is determined to be supported; when , And if the support conditions are not met, it is judged as partial support; when or If an error occurs, it is deemed unsupported. Rule validation checks fields such as device identifier, pre- and post-operation status, task name, and task type. If an alarm is detected, the official save is blocked by default, and rollback validation does not retain official data.

[0071] First, define the quality indicators. :

[0072]

[0073] in, To account for the number of items to be inspected, To check the number of projects that passed.

[0074] Equipment completeness :

[0075]

[0076] Among them, the necessary fields such as the main equipment ID, detailed equipment ID, equipment type, and affiliated plant / station are complete and unique, and are included in the calculation. .

[0077] Topology check pass rate :

[0078]

[0079] If the topology path exists, the equipment type and voltage level meet the requirements, the plant-station relationship is consistent, and the database completion result is consistent with the topology discovery result, it is considered as passing.

[0080] State chain continuity :

[0081]

[0082] Among them, the real-time status verification is passed when the device first appears, and subsequent tasks can inherit the status after the operation of the previous task, which is recorded. .

[0083] Semantic parsing completeness :

[0084]

[0085] Among them, the necessary semantic fields such as device object, operation type, target status, and operation group were successfully parsed and included in the calculation. .

[0086] Alarm penalty items can be defined as:

[0087]

[0088] in, This is a non-blocking alarm set. To configure penalty weights according to alarm levels, This represents the total number of items checked in the task. Hard errors such as missing device IDs, failed real-time status queries, status chain conflicts, and failed data entry pre-verification should not be handled by reducing the score, but should be set as a separate blocking flag.

[0089] In this invention, the parsing results of the form do not directly enter the database. Instead, they are first mapped to a semantic intermediate representation, creating a traceable mapping relationship between the original form, the device object, and the operation task. This facilitates manual review and problem localization. At the same time, the task sequence is generated and maintained in a preset order, ensuring the continuity and consistency of the state before and after the operation in scenarios with multiple time periods and multiple power outages / restors. Finally, the quantitative evaluation of quality scores and support levels, combined with the database entry control, prevents incomplete or risky content from being formally entered into the database, thus balancing the efficiency of automated ticketing with the security of scheduling operations.

[0090] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent power ticket arrangement that integrates semantic intermediate representation, characterized in that, Includes the following steps: Step S1: Obtain the method order ticketing request, identify its intent, and then apply the corresponding ticketing skill; Step S2: Extract business semantics from the structured fields in the method sheet, split the business semantics into operation groups, and construct a semantic intermediate representation based on the business semantics; the semantic intermediate representation records the device object, semantic source, target status, whether it is explicitly named, whether it belongs to the secondary device boundary, and whether there is any incompletely parsed content in the structured fields, and records the operation group to which each business semantic belongs; Step S3: Match the semantic fragments in the semantic intermediate representation with each device candidate in the device candidate set to obtain the matching device with the highest matching degree with the semantic fragment; Step S4: Query the database based on the matched devices to complete the device information and obtain the device completion result; When a candidate device can be uniquely mapped to a device ID after verification of the plant, device type, voltage level, dispatch number, and alias information, and the operation task does not require association with device details, the device is used as the matching result. Otherwise, the system further queries the power status of the device, adjacent devices, connection paths and topology range, determines the effective topology path with the smallest cumulative weight among the candidate topology paths that meet the business constraints, and cross-validates the database completion results with the topology discovery results. Step S5: Organize the operation task sequence according to the semantic intermediate representation and device completion result. The operation task sequence is generated in a preset order, wherein the pre-operation state of the next operation task inherits the post-operation state of the previous operation task. Further, the state chain verification of the generated operation tasks across operation groups is performed to obtain the state chain verification result. Step S6: Calculate the task quality score based on the generation status of the operation task, the completeness of the equipment completion result, the pass rate of the topology verification result, the continuity of the state chain verification result, and the alarm items. Determine the support level of the ticketing result based on the task quality score and the alarm status, and perform warehousing control according to the support level.

2. The method for intelligent power ticket arrangement based on fused semantic intermediate representation according to claim 1, characterized in that, In step S1, the method order ticketing request includes at least one parameter among the method order number, method order JSON, whether to save formally, and whether to allow alarm entry into the database; and based on the parameters in the request, the business scenario to which the method order ticketing request belongs is identified. The business scenario includes ticket generation, task pre-inspection, formal entry into the database, rollback verification, and result review, and then the corresponding ticketing skill is selected according to the business scenario.

3. The method for intelligent power ticket arrangement based on fused semantic intermediate representation according to claim 1, characterized in that, In step S2, the structured fields include the moderation order number, work content, power outage scope, approval start time, approval end time, power outage mode arrangement, power supply mode arrangement, coordination with power outage matters, and relay protection adjustment; The split operation grouping refers to the system splitting a mode order with multiple time periods or multiple power outage ranges into multiple operation groups according to the time overlap relationship and power outage and restoration logic.

4. The method for intelligent power ticket arrangement based on fused semantic intermediate representation according to claim 1, characterized in that, In step S3, the semantic fragment in the semantic intermediate representation is matched with each device candidate in the device candidate set to obtain the matching device with the highest matching degree with the semantic fragment. The matching degree is determined by text similarity, device type consistency, plant consistency and contextual business consistency. If the matching degree of all candidate devices is lower than the preset requirement or the matching degree of multiple candidate devices is close, an alarm is output.

5. The method for intelligent power ticket arrangement based on fused semantic intermediate representation according to claim 1, characterized in that, In step S4, the equipment data completion specifically involves querying the plant / station table, equipment mapping table, line endpoint table, and equipment basic information table based on the equipment name, plant / station name, voltage level, equipment type, dispatch number, and alias information to complete the main equipment ID, equipment type, affiliated plant / station, and dispatch name required for the operation task. The specific steps for determining the effective topology path with the minimum cumulative weight among candidate topology paths that satisfy business constraints are as follows: The power grid equipment topology is abstracted as a weighted graph G = (V, E, W), where V is the set of equipment nodes, E is the set of equipment connection edges, and W is the set of weights between nodes in the weighted graph. For any two equipment nodes u and v, in the candidate topology path set PathsΩ(u, v) that satisfies the business constraint Ω, the path with the smallest cumulative value of the connection edge weight w(a, b) of each pair of adjacent equipment nodes (a, b) is selected as the effective topology path P*(u, v). When there is no valid topology path, or when the device type, operating status, and affiliated plant of the device in the valid topology path are inconsistent with the semantics of the mode sheet, a topology verification alarm will be output.

6. The method for intelligent power ticket arrangement based on fused semantic intermediate representation according to claim 1, characterized in that, In step S5, the operation task sequence is organized based on the semantic intermediate representation and the device completion result. The operation tasks are formed in the following order: power outage mode task, line or switch busbar switching task, busbar switching task, power outage range task, power restoration range task, busbar restoration task, and power restoration mode restoration task. For each task generated, the system records the current status of the relevant equipment so that the pre-operation status of the next task can inherit the post-operation status of the previous task. Then, the generated operation task is checked for state chain verification. The state before and after the operation of the same operation task is checked against the state remembered by the device. If the verification fails, a state verification alarm is issued.

Images