Power grid operation ticket generation method and device, storage medium and computer equipment

Abstract

The power grid operation ticket generation method and device, storage medium and computer equipment provided by the application accurately identify complex and changeable operation intentions by performing semantic analysis and semantic disambiguation on the obtained task information through the preset power grid knowledge graph. On this basis, the task intention information obtained by analysis is used for multi-hop knowledge retrieval in the power grid knowledge graph to obtain structured knowledge closely related to the current operation task. Then, according to the structured knowledge, the primary device operation is reasoned first, and then the corresponding secondary device operation sequence is reasoned based on the primary device operation sequence linkage, thereby realizing automatic modeling and deduction of the deep logical relationship between the primary and secondary devices. Finally, a complete, executable and safety specification-compliant power grid operation ticket is automatically generated. Therefore, the scheme improves the intelligent level of the operation ticket generation process, reduces the dependence on manual experience and template matching, and can adapt to diversified switching operation scenarios.

Description

Technical Field

[0001] This application relates to the field of power system technology, and in particular to a method, apparatus, storage medium and computer equipment for generating power grid operation tickets. Background Technology

[0002] Against the backdrop of new power system construction, power grid topologies are becoming increasingly complex and operational modes are becoming more flexible, placing higher demands on the safety and intelligence of switching operations. For a long time, operation ticket preparation has relied primarily on manual experience, which is time-consuming and susceptible to human error. To address this issue, the industry has gradually introduced automated operation ticket generation systems. These systems establish a fixed rule base and a typical ticket template base, and generate initial draft operation tickets by matching key information to user tasks.

[0003] However, existing solutions have limited ability to understand operational tasks, making it difficult to accurately interpret complex and ever-changing operational intentions. Furthermore, they lack the ability to reason about the deep logical relationships between power grid equipment, resulting in a relatively low level of intelligence in the power grid operation ticket generation process. Summary of the Invention

[0004] The purpose of this application is to at least address one of the aforementioned technical deficiencies, particularly the limitations of existing technologies in understanding operational tasks, making it difficult to accurately interpret complex and ever-changing operational intentions, and lacking the ability to reason about the deep logical relationships between power grid equipment, resulting in a relatively low level of intelligence in the power grid operation ticket generation process.

[0005] In a first aspect, this application provides a method for generating power grid operation tickets, the method comprising:

[0006] Obtain task information for the operation ticket generation task, as well as the preset power grid knowledge graph;

[0007] Based on the power grid knowledge graph, semantic parsing and semantic disambiguation are performed on the task information to obtain task intent information;

[0008] The task intent information is used to perform multi-hop knowledge retrieval in the power grid knowledge graph to obtain structured knowledge related to the operation ticket generation task;

[0009] Based on the structured knowledge, a primary equipment operation reasoning is performed on the operation ticket generation task to obtain a primary equipment operation sequence. Based on the primary equipment operation sequence, a secondary equipment operation reasoning is performed on the operation ticket generation task to obtain a secondary equipment operation sequence.

[0010] The primary equipment operation sequence and the secondary equipment operation sequence are structurally integrated to form a power grid operation ticket.

[0011] In one embodiment, the step of performing semantic parsing and disambiguation on the task information based on the power grid knowledge graph to obtain task intent information includes:

[0012] Obtain a preset large-scale power model, which includes a semantic parsing module and a graph attention module;

[0013] The task information is input into the semantic parsing module, so that the semantic parsing module calls the power grid knowledge graph to perform semantic parsing on the task information to obtain task semantic information;

[0014] The task information and the task semantic information are input into the graph attention module, so that the graph attention module calls the power grid knowledge graph to perform relational analysis on the task information, and completes or corrects the task semantic information based on the relational analysis results to obtain task intent information.

[0015] In one embodiment, the step of using the task intent information to perform multi-hop knowledge retrieval in the power grid knowledge graph to obtain structured knowledge related to the operation ticket generation task includes:

[0016] Extract the target object from the task intent information;

[0017] Starting from the target object, the device association, state mapping relationship, normative constraint relationship and historical operation experience of the target object are sequentially retrieved in the power grid knowledge graph.

[0018] Based on the device association, the state mapping, the specification constraint, and the historical operation experience, structured knowledge related to the operation ticket generation task is generated.

[0019] In one embodiment, the step of performing a device operation reasoning on the operation ticket generation task based on the structured knowledge to obtain a device operation sequence includes:

[0020] Based on the structured knowledge, the primary equipment involved in the operation ticket generation task is determined;

[0021] Based on the structured knowledge, the pre-defined primary equipment operation agent is used to reason about the state transitions and operation sequences of each primary equipment to obtain the primary equipment operation sequence.

[0022] In one embodiment, the step of performing secondary device operation reasoning on the operation ticket generation task based on the primary device operation sequence to obtain the secondary device operation sequence includes:

[0023] Obtain the protection coordination relationships and operating requirements of secondary equipment;

[0024] Combining the protection coordination relationship of the secondary equipment and the operation requirements of the secondary equipment, the secondary equipment operation sequence is obtained by using a preset secondary equipment operation intelligent agent to reason about the secondary safety measures in the operation sequence of the primary equipment.

[0025] In one embodiment, before structurally integrating the primary device operation sequence and the secondary device operation sequence, the method further includes:

[0026] The primary device operation sequence and the secondary device operation sequence are validated using a preset safety constraint intelligent agent to obtain the rule validation result.

[0027] Furthermore, by using a preset risk prediction intelligent agent to simulate the changes in the power grid state after executing the primary equipment operation sequence and the secondary equipment operation sequence, risk prediction results are obtained;

[0028] If both the rule verification result and the risk prediction result meet expectations, then the primary equipment operation sequence and the secondary equipment operation sequence are structurally integrated; otherwise, the primary equipment operation reasoning and the secondary equipment operation reasoning are re-performed.

[0029] In one embodiment, the method further includes:

[0030] When the operation ticket generation task involves cross-regional collaborative reasoning, the operation ticket generation task is divided into multiple sub-tasks according to different regions;

[0031] Each subtask is distributed to the corresponding region for local primary equipment operation reasoning and secondary equipment operation reasoning, and the reasoning results returned by each region are aggregated to form a power grid operation ticket.

[0032] Secondly, this application provides a power grid operation ticket generation device, the device comprising:

[0033] The information acquisition module is used to acquire task information for the operation ticket generation task, as well as a preset power grid knowledge graph;

[0034] The intent recognition module is used to perform semantic parsing and semantic disambiguation on the task information based on the power grid knowledge graph to obtain task intent information;

[0035] The knowledge retrieval module is used to perform multi-hop knowledge retrieval in the power grid knowledge graph using the task intent information to obtain structured knowledge related to the operation ticket generation task;

[0036] The multi-agent reasoning module is used to perform primary equipment operation reasoning on the operation ticket generation task based on the structured knowledge to obtain a primary equipment operation sequence, and to perform secondary equipment operation reasoning on the operation ticket generation task based on the primary equipment operation sequence to obtain a secondary equipment operation sequence.

[0037] The operation ticket generation module is used to structurally integrate the primary equipment operation sequence and the secondary equipment operation sequence to form a power grid operation ticket.

[0038] Thirdly, this application provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the power grid operation ticket generation method as described in any of the above embodiments.

[0039] Fourthly, this application provides a computer device, including: one or more processors, and a memory;

[0040] The memory stores computer-readable instructions, and when the one or more processors execute the computer-readable instructions, they perform the steps of the power grid operation ticket generation method as described in any of the above embodiments.

[0041] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0042] The power grid operation ticket generation method, device, storage medium, and computer equipment provided in this application utilize a pre-defined power grid knowledge graph to perform semantic parsing and disambiguation on the acquired task information, accurately identifying complex and ever-changing operational intentions. Based on this, the parsed task intention information is used to perform multi-hop knowledge retrieval within the power grid knowledge graph to obtain structured knowledge closely related to the current operation task, including deep information such as power grid topology, equipment status, and operating rules. Subsequently, based on this structured knowledge, the operation of primary equipment is first inferred, and then the corresponding operation sequence of secondary equipment is inferred based on the sequence of primary equipment operations, thereby achieving automatic modeling and deduction of the deep logical relationship between primary and secondary equipment. Finally, the operation sequences of primary and secondary equipment are structurally integrated to automatically generate a complete, executable, and safety-compliant power grid operation ticket. Therefore, this solution improves the intelligence level of the operation ticket generation process, not only reducing reliance on manual experience and template matching, thus lowering the risk of human error, but also adapting to diverse switching operation scenarios, effectively improving the efficiency and scenario coverage of operation ticket preparation. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 A flowchart illustrating a method for generating power grid operation tickets provided in this application embodiment;

[0045] Figure 2 An example diagram illustrating a method for generating power grid operation tickets provided in this application embodiment;

[0046] Figure 3 This is a schematic diagram of the structure of a power grid operation ticket generation device provided in an embodiment of this application;

[0047] Figure 4 This is an internal structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0048] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0049] In one embodiment, this application provides a method for generating power grid operation tickets. The following embodiments illustrate the application of this method to an operation ticket generation system, hereinafter referred to as the system. It is understood that the system can be deployed in a computer device to execute the power grid operation ticket generation method provided in this application.

[0050] like Figure 1 As shown, this application provides a method for generating power grid operation tickets, the method comprising:

[0051] S101: Obtain task information for the operation ticket generation task, as well as the preset power grid knowledge graph.

[0052] Task information refers to the structured elements extracted from the task generation process of the operation ticket, including but not limited to the operation object, initial state, target state, and operation scenario. The power grid knowledge graph is a semantic knowledge base organized in a graph structure, comprising an entity layer, a relation layer, and an attribute layer. The entity layer includes core entities in the power grid switching operation domain, specifically categorized into four main types: equipment entities, operation entities, state entities, and specification entities. The relation layer defines various relationships between entities, such as equipment relationships, operation logic relationships, state mapping relationships, and specification constraint relationships. The attribute layer supplements the attribute information of each entity, including inherent attributes and dynamic attributes.

[0053] In this step, when a user needs to generate a power grid operation ticket, they can input the ticket information into the client through text input, voice input, or interface operation. The client then creates an operation ticket generation task based on the ticket information and transmits the task to the system. When the system receives the operation ticket generation task, it retrieves relevant information such as the operation object, initial state, and target state—that is, the task information—from the task. Simultaneously, it retrieves the pre-built and stored power grid knowledge graph.

[0054] Specifically, the power grid knowledge graph can be created using a combination of bottom-up and top-down construction strategies. It uses equipment ledger databases, power grid topology databases, real-time operation databases, operation specification databases, historical operation ticket databases, and protection configuration databases as raw data sources, and maps and generates them according to a three-layer architecture: entity layer, relationship layer, and attribute layer. The construction of each layer is as follows:

[0055] (1) Entity layer construction

[0056] The construction of equipment entities primarily originates from the equipment ledger database and the protection configuration database. The system extracts records of primary equipment (such as circuit breakers, disconnectors, transformers, and busbars) and secondary equipment (such as protection devices, monitoring and control devices, and pressure plates) from the equipment ledger database, generating a unique graph node for each equipment record. Simultaneously, records related to protection devices are extracted from the protection configuration database and also incorporated into the graph as equipment entities. These nodes carry inherent attributes of the equipment, such as model, specifications, and installation location.

[0057] The construction of operational entities originates from the operational specification database and the historical operational ticket database. The system extracts standard operational verbs from the operational specification database, such as "close," "disconnect," "engage," and "exit," treating each verb as an independent node. Simultaneously, it extracts operational actions from typical operational steps from the historical operational ticket database to supplement and validate the operational entities. Each operational entity node is associated with terminology definitions from the operational specification database as attributes.

[0058] The construction of state entities originates from the equipment ledger database and the real-time operation database. The equipment ledger database records the initial state of the equipment (such as running, hot standby, cold standby, and maintenance), while the real-time operation database provides the current actual state information of the equipment. The system generates these states as independent state nodes, serving as intermediary nodes between equipment entities and operation entities.

[0059] The construction of normative entities primarily originates from the operational specification database, particularly the "five-prevention" interlocking rules, operational procedure requirements, and operational terminology standards. The system constructs each rule or specification clause into a normative entity node, such as "Five-prevention rule_Preventing the opening and closing of disconnect switches under load," for subsequent constraint reasoning.

[0060] (2) Relationship layer construction

[0061] The construction of equipment relationships originates from the relationship fields in the equipment ledger database and the power grid topology database. The equipment ledger database records the hierarchical relationships between equipment (e.g., a circuit breaker belongs to a certain bay), while the power grid topology database records electrical connection relationships (e.g., series and parallel connections) and topology partitioning information. By parsing this data, relationship edges such as "Circuit Breaker_5011 — Series — Disconnect Switch_5012" are generated.

[0062] The construction of operational logic relationships originates from the historical operation ticket database and the operation specification database. The historical operation ticket database stores a large number of executed switching operation tickets, which include the sequence of operation steps. By analyzing typical operation tickets, the sequential order, cooperative operation relationships, and mutually exclusive operation relationships between operations are extracted, generating relational edges such as "next operation" or "mutually exclusive".

[0063] The state mapping relationship is constructed from the operation specification database and the real-time operation database. The operation specification database defines the state transitions caused by each legal operation; for example, "closing the circuit breaker" changes the equipment from the "maintenance" state to the "operation" state. The real-time operation database provides a snapshot of the current state, used to verify and supplement the state mapping relationship.

[0064] The construction of normative constraint relationships originates from the operation specification database and the protection configuration database. The system associates operational entities with the normative entities that must be followed, generating "must-follow" relationship edges. Simultaneously, it extracts the coordination relationships between protection devices and primary equipment from the protection configuration database, generating "protection range" or "coordination constraint" relationship edges.

[0065] (3) Attribute layer construction

[0066] The inherent attributes are primarily populated from the equipment ledger database and the operation specification database. The equipment ledger database provides information that does not change frequently, such as equipment model, specifications, installation location, and commissioning time, which is directly assigned to the corresponding equipment entity nodes. The operation specification database provides definitions and specification numbers for operation terms, which are assigned to operation entity nodes.

[0067] The dynamic attributes are populated from a real-time operational database. This database interfaces with an EMS (Energy Management System) / DMS (Distribution Management System) to collect equipment operating status (switch opening / closing, disconnector position, switch engagement / disengagement) and power grid flow data (bus voltage, line power, current, frequency) within seconds. The system synchronizes these dynamic attributes to the corresponding equipment entity nodes in real time and sets dynamic refresh markers to ensure that the dynamic attributes in the graph are consistent with the operational data.

[0068] Furthermore, the power grid knowledge graph can be updated periodically in batches to ensure its reliability.

[0069] S102: Based on the power grid knowledge graph, perform semantic parsing and semantic disambiguation on the task information to obtain the task intent information.

[0070] Among them, task intent information refers to the structured and unambiguous description of the operational target extracted after semantic parsing and ambiguity elimination of task information.

[0071] In this step, the power grid knowledge graph can be used as prior knowledge to perform word segmentation, entity recognition, relation extraction, and semantic understanding on the text content of the task information, resulting in semantic parsing results. Based on this, to address issues such as ambiguity, vague expressions, and omitted abbreviations in the task information, information such as entity attributes and relationships in the power grid knowledge graph is used to eliminate ambiguities and complete the content of the semantic parsing results, thereby obtaining the task intent information.

[0072] For example, when a dispatcher inputs task information such as "disconnect 5011 switch" which has an abbreviation or vague expression, the system will rely on the knowledge of the equipment type, bay, associated equipment and operating procedures corresponding to 5011 in the power grid knowledge graph to automatically parse and eliminate ambiguity, clarify that the operation object is 5011 circuit breaker, the operation action is disconnection, and the target state is tripping, thereby forming accurate task intent information.

[0073] In one example, complete task intent information can be represented as:

[0074] Operation target: DL-01 circuit breaker of 220kV XX line, and automatically associate its topology-related devices from the knowledge graph, including disconnecting switches GL-02 and GL-03, grounding switch JD-01, and protection device B-01;

[0075] Initial state: Running;

[0076] Target status: Maintenance;

[0077] Operational scenario: Equipment maintenance;

[0078] Special requirements: Before maintenance, close the grounding switch and ensure proper grounding protection.

[0079] S103: Utilize task intent information to perform multi-hop knowledge retrieval in the power grid knowledge graph to obtain structured knowledge related to the operation ticket generation task.

[0080] Structured knowledge refers to a set of knowledge with clear logical relationships and organizational forms obtained by traversing nodes and edges in the power grid knowledge graph, starting from the task intent information. It includes, but is not limited to, the topological relationships between devices, the order and mutual exclusion of operation steps, the legal paths of state transitions, the safety specifications that must be followed in operation, and historical success cases.

[0081] In this step, the system starts with the explicitly defined operation target entity from the task intent information as the initial node. It then navigates to directly connected device nodes via association edges to obtain their topological adjacency information. Next, using these device nodes and the current state as conditions, it navigates along state mapping edges to the valid target state node and records the operation sequence along the path. Simultaneously, the system also starts from the operation entity and navigates along normative constraint edges to the relevant safety specification nodes to obtain the conditions that each operation must meet. This series of navigations forms a retrieval subgraph starting from the initial node and covering multiple types of related knowledge, thereby obtaining structured knowledge related to the operation ticket generation task.

[0082] For example, suppose the task intent is "to switch circuit breaker DL-01 from operating state to maintenance state". The system first starts with node "DL-01", tracing the associated disconnectors GL-02 and GL-03 via a series connection, then the grounding switch JD-01 via a grounding connection, and finally the protection device B-01 via a protection coordination connection. This is the first hop. Next, the system uses the "operating" and "maintaining" state nodes as endpoints to search the power grid knowledge graph for valid state transition paths. By reverse-searching via the "post-operation state" connection, it finds that the transition from "operating" to "maintaining" must pass through two intermediate states: "hot standby" and "cold standby". Each state transition corresponds to a standard operation. This is the second hop. Finally, for each retrieved operation, such as "disconnecting DL-01", the system jumps to the "five-prevention rules" node along the "must follow" connection to obtain the constraint that "the disconnector can only be opened after the circuit breaker is disconnected". This is the third hop. Finally, the knowledge obtained from multi-hop knowledge retrieval is summarized to obtain structured knowledge.

[0083] S104: Based on structured knowledge, perform primary equipment operation reasoning on the operation ticket generation task to obtain a primary equipment operation sequence, and based on the primary equipment operation sequence, perform secondary equipment operation reasoning on the operation ticket generation task to obtain a secondary equipment operation sequence.

[0084] In this context, a primary equipment operation sequence refers to the ordered set of steps for primary equipment, derived through structured knowledge reasoning, to complete grid operations. A secondary equipment sequence refers to the ordered set of steps for secondary equipment, derived through reasoning based on the generated primary equipment operation sequence, to complete grid operations.

[0085] In this step, primary equipment operation reasoning is a planning process based on state transition paths. The system uses the initial and target states from the task intent information as endpoints and the legal state transition sequences from the retrieved structured knowledge as a basis to progressively generate the primary equipment operation steps necessary to reach the target state from the initial state. Secondary equipment operation reasoning, on the other hand, is a mapping process based on protection coordination relationships. The system uses key operation nodes in the primary equipment operation sequence as triggering conditions, and based on the coordination relationships between protection devices and primary equipment in the structured knowledge, as well as the requirements of the operating specifications for secondary equipment, inserts corresponding secondary equipment operation steps at appropriate locations, forming a secondary equipment operation sequence that closely coordinates with the primary equipment operation.

[0086] Specifically, when performing reasoning for primary equipment operations, a corresponding primary equipment operation agent can be used to reason about the actions of the primary equipment. Similarly, when performing reasoning for secondary equipment operations, a corresponding secondary equipment operation agent can be used to reason about the actions of the secondary equipment. It is understood that when constructing primary and secondary equipment operation agents, entity relationships and operation rule bases for primary and secondary equipment can be constructed based on the power grid knowledge graph, respectively. Then, the primary and secondary equipment operation agents are initialized according to equipment type, operation logic, and constraints, completing the configuration of capabilities such as reasoning strategies, timing matching, and security verification, forming dedicated agents capable of independently executing reasoning and collaborative interaction. A more detailed construction process can be found in the training process of general agents; this application does not impose specific limitations on this.

[0087] For example, suppose the task requires switching a 220 kV line from "operation" to "maintenance". In the primary equipment operation reasoning, the system obtains the state transition path "operation → hot standby → cold standby → maintenance" from structured knowledge. Based on this, the system sequentially deduces the first step "disconnect the circuit breaker", the second step "open the line-side disconnect switch", the third step "open the bus-side disconnect switch", and the fourth step "close the grounding switch", thus forming the primary equipment operation sequence. In the secondary equipment operation reasoning, based on the above primary equipment operation sequence, the system identifies that the critical operation "disconnecting the circuit breaker" will disconnect the line load and may trigger protection action. Therefore, according to the protection coordination relationship, before the "disconnecting the circuit breaker" step, the system infers and inserts the step "deactivate the line protection device trip output pressure plate". At the same time, according to the maintenance operation safety specifications, after "closing the grounding switch", the system infers and inserts the step "activate the maintenance protection device", thus forming a secondary equipment operation sequence that matches the timing and logic of the primary equipment operation sequence.

[0088] S105: Structure and integrate the primary equipment operation sequence and the secondary equipment operation sequence to form a power grid operation ticket.

[0089] Among them, the power grid operation ticket refers to a standardized instruction document that can serve as the legal basis for on-site switching operations, generated by structurally integrating the operation sequences of primary equipment and secondary equipment.

[0090] In this step, after determining the primary and secondary equipment operation sequences, the system performs time-sequence alignment, step merging, content verification, and format standardization of the primary and secondary equipment operation steps. It can also supplement necessary elements such as operation tasks, operation numbers, confirmation requirements, and safety precautions. Finally, the above data is filled into a preset operation ticket template to form a standardized, complete, and directly executable power grid operation ticket.

[0091] In the above embodiments, a pre-defined power grid knowledge graph is used to perform semantic parsing and disambiguation on the acquired task information, accurately identifying complex and ever-changing operational intentions. Based on this, the parsed task intention information is used to perform multi-hop knowledge retrieval in the power grid knowledge graph to obtain structured knowledge closely related to the current operational task, including deep information such as power grid topology, equipment status, and operating rules. Subsequently, based on this structured knowledge, the operation of primary equipment is first inferred, and then the corresponding operation sequence of secondary equipment is inferred based on the sequence of primary equipment operations, thereby realizing the automatic modeling and deduction of the deep logical relationship between primary and secondary equipment. Finally, the operation sequences of primary and secondary equipment are structurally integrated to automatically generate a complete, executable, and safety-compliant power grid operation ticket. Therefore, this solution improves the intelligence level of the operation ticket generation process, not only reducing reliance on manual experience and template matching, thus lowering the risk of human error, but also adapting to diverse switching operation scenarios, effectively improving the efficiency and scenario coverage of operation ticket preparation.

[0092] In one embodiment, based on a power grid knowledge graph, semantic parsing and semantic disambiguation are performed on the task information to obtain task intent information, including:

[0093] S1: Obtain the preset large power model.

[0094] S2: Input the task information into the semantic parsing module so that the semantic parsing module can call the power grid knowledge graph to perform semantic parsing on the task information and obtain the task semantic information.

[0095] S3: Input the task information and task semantic information into the graph attention module so that the graph attention module can call the power grid knowledge graph to parse the relationship between the task information and complete or correct the task semantic information based on the relationship parsing results to obtain the task intent information.

[0096] The power grid big data model includes a semantic parsing module and a graph attention module. The semantic parsing module is the text understanding unit in the power grid big data model, oriented towards power grid operation tasks. It is used for word segmentation, entity recognition, and basic semantic extraction of task information. The graph attention module is the association reasoning unit in the power grid big data model, deeply integrated with the power grid knowledge graph. It is used to perform attention-weighted calculations based on relationships such as equipment associations, operational constraints, and state mappings in the power grid knowledge graph, achieving fuzzy information completion and ambiguous semantic correction.

[0097] In this embodiment, a large-scale power model that has undergone fine-tuning and functional configuration in the power sector is first loaded. This model has a built-in semantic parsing module and a graph attention module. Task information is first input into the semantic parsing module, which calls upon entity information such as equipment, operations, and states from the power grid knowledge graph to perform basic semantic decomposition and structuring processing on the task information, obtaining preliminary task semantic information. Based on this, the task information and task semantic information are input together into the graph attention module. The graph attention module, based on the topological relationships, operational logic, and normative constraints in the power grid knowledge graph, performs relationship mining and weight calculation on elements such as equipment, actions, and objects in the task information. Based on the parsed relationships, it performs missing content completion, ambiguity correction, and error correction on the task semantic information, ultimately forming accurate, complete, and unambiguous task intent information.

[0098] Specifically, the graph attention module includes a lightweight graph attention mechanism whose attention calculations directly rely on the relational edge information of the power grid knowledge graph. Specifically, when parsing task information, the graph attention module extracts from the power grid knowledge graph the relationships (such as series, parallel, and subordinate relationships), operational logic relationships (such as sequence and mutually exclusive operations), and state mapping relationships (such as the correspondence between operations and legal state transitions) related to the devices mentioned in the task information. This relational information serves as structured prior knowledge for the graph attention mechanism, helping the module select the interpretation that best matches the actual power grid topology and operational specifications from multiple possible semantic interpretations. For example, when the task information contains the ambiguous expression "disconnect 5011," the graph attention mechanism queries the power grid knowledge graph for the type of device 5011 (circuit breaker or disconnector), its current state, and the associated five-prevention constraint rules, thereby accurately determining the specific semantic meaning of the verb "disconnect." Meanwhile, leveraging the standardized entities and operational logic relationships within the knowledge graph, the module can effectively handle out-of-vocabulary words and fuzzy input scenarios, such as non-standard equipment names appearing in dialect voice commands and operational parameters omitted in simplified, colloquial task descriptions. These are completed and corrected through entity aliases and synonym mappings in the power grid knowledge graph. Ultimately, the task intent information is obtained.

[0099] In one example, the power big model can use the State Grid's self-developed Guangming Power Big Model as the initial model, then fine-tune the instructions on the power dispatch corpus, and integrate the power grid knowledge graph so that the model can obtain structured prior knowledge in the power grid domain when reasoning.

[0100] It is understandable that the semantic parsing module of the power grid big data model completes the basic semantic extraction of task information, and then the graph attention module, combined with the power grid knowledge graph, performs relational analysis and semantic completion correction, ultimately outputting accurate and reliable task intent information. In this process, natural language understanding and graph-structured knowledge are deeply integrated, retaining the flexibility of text understanding while possessing the rigor of power grid professional logic. This effectively solves problems such as brief descriptions of power operation tasks, ambiguous terminology, and missing information, thus improving the accuracy of task intent recognition.

[0101] In one embodiment, multi-hop knowledge retrieval is performed in the power grid knowledge graph using task intent information to obtain structured knowledge related to the operation ticket generation task, including:

[0102] S1: Extract the target object from the task intent information.

[0103] S2: Starting from the target object, retrieve the target object's equipment associations, state mapping relationships, normative constraint relationships, and historical operation experience in the power grid knowledge graph in sequence.

[0104] S3: Generate structured knowledge related to the operation ticket generation task based on equipment association, state mapping, specification constraints, and historical operation experience.

[0105] Here, the target object refers to the operational entity extracted from the task intent information. For example, a power grid device, bay, or line unit to which state transitions are to be performed. Device relationships refer to the topological and hierarchical relationships in the power grid knowledge graph describing the electrical connections, dependencies, series connections, parallel connections, protection correspondences, etc., between the target object and other devices. State mapping relationships refer to the legal state transition rules for devices before and after operation. Normative constraint relationships refer to the rule constraints that power grid operations must adhere to. Historical operational experience refers to historical case knowledge stored in the power grid knowledge graph, including the successful execution steps, sequences, and verification results of similar tasks and devices in similar scenarios.

[0106] In this embodiment, the system can accurately extract the target object of the operation from the clearly defined task intent information, using it as the starting node for knowledge graph retrieval. Then, centered on this target object, the system sequentially performs multi-dimensional searches in the power grid knowledge graph according to a preset path. First, it obtains the device associations such as electrical connections and topological affiliation between the target object and adjacent devices. Next, it retrieves the legal transition paths from the target object's current state to the target state, i.e., state mapping relationships. Then, it retrieves the applicable five-prevention rules, operating specifications, protection requirements, and other normative constraints for the target object and the operation task. Simultaneously, it matches and retrieves historical successful operation cases of similar devices and tasks to form historical operation experience. Finally, it integrates and organizes the retrieved device associations, state mapping relationships, normative constraints, and historical operation experience to form structured knowledge that can be directly used for reasoning.

[0107] For example, when performing multi-hop knowledge retrieval, for the task intent information of "changing bay 501 of substation A from operation to maintenance", the retrieval starts from bay 501. It then sequentially explores its topologically related devices, the current operating status of each device, the legal state transition path from "operation" to "maintenance", the five-prevention constraint rules corresponding to each operation step in the path, and successful experiences from similar historical operation cases within the power grid knowledge graph. Ultimately, it obtains structured knowledge corresponding to this operation ticket generation task.

[0108] Specifically, by using a point-to-surface, layer-by-layer retrieval method, it can accurately extract and integrate knowledge highly relevant to the operation task, comprehensively covering equipment topology, state rules, safety constraints, and historical experience. This avoids reasoning bias caused by knowledge gaps or insufficient connections, and provides complete, accurate, and reliable knowledge support for subsequent operation reasoning of primary and secondary equipment.

[0109] In one embodiment, based on structured knowledge, a device operation reasoning is performed on the operation ticket generation task to obtain a device operation sequence, including:

[0110] S1: Based on structured knowledge, determine the primary equipment involved in the operation ticket generation task.

[0111] S2: Based on structured knowledge, the system uses a pre-defined primary equipment operation agent to reason about the state transitions and operation sequence of each primary equipment, thus obtaining the primary equipment operation sequence.

[0112] Among them, the primary equipment operation intelligent agent refers to the autonomous decision-making unit for reasoning about the operation of primary equipment in the power grid. It uses the equipment association and state mapping relationship in the knowledge graph as the reasoning basis and is responsible for the state transition and operation sequence reasoning of primary equipment such as circuit breakers, disconnect switches, and busbars.

[0113] In this embodiment, based on the device association and state mapping relationships contained in the structured knowledge, all primary devices involved in this operation ticket generation task are located and determined, including circuit breakers, disconnect switches, grounding switches, busbars, transformers, and other devices that directly participate in power transmission. Then, the structured knowledge is used as the basis for reasoning to call the pre-set primary device operation intelligent agent. The intelligent agent deduces and judges the initial state, target state, legal transition path, and execution order of each primary device according to the power grid operation procedures, the five-prevention interlocking requirements, and the state transition logic, and finally forms a primary device operation sequence that is timely, compliant with constraints, and complete in steps.

[0114] Specifically, by relying on professional intelligent agents, the system automates and standardizes decision-making for primary equipment operations, strictly adheres to power grid safety constraints and state transition logic, accurately locates the operating equipment, and generates a reasonable operating sequence. This effectively reduces human judgment bias and rule omissions, and significantly improves the efficiency and reliability of operation ticket preparation.

[0115] In one embodiment, based on the primary equipment operation sequence, a secondary equipment operation reasoning is performed on the operation ticket generation task to obtain a secondary equipment operation sequence, including:

[0116] S1: Obtain the protection coordination relationship of secondary equipment and the operation requirements of secondary equipment.

[0117] S2: Combining the protection coordination relationship of secondary equipment and the operation requirements of secondary equipment, the secondary equipment operation agent is used to reason about the secondary safety measures in the operation sequence of primary equipment to obtain the operation sequence of secondary equipment.

[0118] Among them, the secondary equipment protection coordination relationship refers to the fixed constraints and linkage relationships formed between primary and secondary equipment to meet the requirements of power grid safe operation and relay protection logic. The secondary equipment operation intelligence agent refers to the autonomous decision-making unit for power grid secondary equipment operation reasoning. It is responsible for the collaborative operation reasoning of secondary equipment such as protection devices, pressure plates, and measurement and control devices, based on the protection coordination relationship mapped by the protection configuration database and the secondary equipment operation requirements in the operation specification library.

[0119] In this embodiment, the secondary equipment protection coordination relationship and secondary equipment operation requirements corresponding to this task can first be obtained from the power grid knowledge graph or real-time configuration library. This includes the correspondence between protection devices and primary equipment, the activation and deactivation logic of protection plates, the measurement and control interlocking conditions, and safety measures specifications. Then, this information, along with the generated primary equipment operation sequence, is input into a preset secondary equipment operation intelligent agent. The intelligent agent uses the primary equipment operation sequence as a benchmark, combined with the protection coordination relationship and operation requirements, to perform timing alignment and logical reasoning on the secondary safety measures required for each primary operation step. This determines the timing and content of secondary operations such as protection plate activation and deactivation, protection function activation and deactivation, setting value switching, and measurement and control interlocking, ultimately forming a secondary equipment operation sequence that coordinates with the primary equipment operation sequence.

[0120] It is understandable that the protection coordination relationship and operation requirements of secondary equipment are first obtained, and then the secondary equipment operation agent is combined with the operation sequence of primary equipment to complete the reasoning of secondary safety measures. Finally, a time-coordinated and logically compliant secondary equipment operation sequence is generated to achieve accurate matching between primary and secondary operations and to follow the protection coordination logic and safety measure specifications, thereby ensuring that secondary safety measures are timely, correct and complete.

[0121] In one embodiment, before the primary equipment operation sequence and the secondary equipment operation sequence are structurally integrated, the power grid operation ticket generation method further includes:

[0122] S1: Use a pre-set safety constraint agent to perform rule verification on the primary device operation sequence and the secondary device operation sequence, and obtain the rule verification result.

[0123] S2: Use a preset risk prediction agent to simulate the changes in the power grid state after executing primary and secondary equipment operation sequences to obtain risk prediction results.

[0124] S3: If the rule verification results and risk prediction results both meet expectations, then the primary equipment operation sequence and the secondary equipment operation sequence are structurally integrated; otherwise, the primary equipment operation reasoning and the secondary equipment operation reasoning are re-performed.

[0125] Among them, the safety constraint agent refers to an autonomous decision-making unit specifically responsible for verifying the rule compliance of operation sequences. It uses the normative entities and constraint relationships in the operation specification database, as well as the dynamic attributes in the real-time operation database, as verification criteria, and is responsible for the pre-verification of the five-prevention rules and power flow constraints. The risk prediction agent refers to an autonomous decision-making unit specifically responsible for simulating the changes in the power grid state after the operation. It uses the historical state evolution patterns in the power grid knowledge graph and real-time operation data as the basis for prediction, and is responsible for inferring the changes in the power grid state after the operation.

[0126] In this embodiment, a preset safety constraint agent is first invoked. Taking the primary and secondary equipment operation sequences as input, the agent performs rule verification according to the five-prevention interlocking rules, operational logic, regulatory clauses, and interlocking constraints, determining whether the operation sequence is legal, the order is reasonable, and the coordination is correct, and outputs a pass / fail rule verification result. Next, a preset risk prediction agent is invoked. Using the real-time operating state of the power grid as initial conditions, the agent simulates the execution of the primary and secondary equipment operation sequences, deduces the changes in power grid topology, power flow distribution, and equipment operating states, identifies potential risks such as exceeding limits, instability, and malfunction, and generates risk prediction results. When both the rule verification results and the risk prediction results meet the expected requirements, the two types of operation sequences are structurally integrated to form the final operation ticket. If any item fails to meet the expectations, the process returns to the primary and secondary equipment operation reasoning stage for re-correction.

[0127] It is understood that the common constraint boundary for all intelligent agents used in this application is the task intent information. All operational steps generated by all intelligent agents must be aligned with the task intent information. Each intelligent agent independently deduces based on the shared power grid knowledge graph context, while maintaining consistency through a cooperative communication mechanism.

[0128] Specifically, a safety constraint agent performs logical compliance verification of the operation sequence, while a risk prediction agent simulates changes in the power grid's operating state and identifies risks. Only when both results meet expectations are they integrated to form an operation ticket; otherwise, the process is revised and re-reasoned. This approach not only eliminates violations of the five defenses and logical errors at the rule level but also anticipates and eliminates potential risks such as power flow exceeding limits and improper protection coordination at the system level, thereby improving the safety and reliability of operation tickets.

[0129] In one embodiment, the power grid operation ticket generation method further includes:

[0130] S1: When the operation ticket generation task involves cross-regional collaborative reasoning, the operation ticket generation task is divided into multiple sub-tasks according to different regions.

[0131] S2: Distribute each subtask to the corresponding region for local primary equipment operation reasoning and secondary equipment operation reasoning, and summarize the reasoning results returned by each region to form a power grid operation ticket.

[0132] In this embodiment, when the operation ticket generation task involves cross-regional collaborative reasoning, a federated privacy reasoning mechanism can be adopted to avoid data leakage. The knowledge carrier of this mechanism is not the original grid data, but rather the encrypted parameters and gradient information of the local power grid knowledge graph of each region. This allows each region to participate in the joint training and optimization of global reasoning without exposing sensitive information such as local power grid topology, equipment parameters, and operating status.

[0133] For example, the inference agents of each substation and power supply bureau complete training and updates locally, uploading only encrypted model parameters. Simultaneously, controllable noise is added to the model parameters using differential privacy technology to prevent attackers from inferring the original data through parameter inversion. Furthermore, homomorphic encryption technology can be used to allow necessary cross-regional knowledge aggregation operations to be performed in an encrypted state, ensuring that the original data never leaves the domain throughout the entire inference process.

[0134] In one embodiment, such as Figure 2 As shown, Figure 2 This is an example diagram illustrating a method for generating power grid operation tickets provided in an embodiment of this application. Figure 2 In this application, the power grid knowledge graph, which is deeply integrated and structurally associated with power grid switching operations, is the power grid knowledge graph. The intelligent reasoning layer includes a multimodal task parsing module for the large power model and a multi-agent hybrid reasoning engine module. The former includes a large power model semantic understanding engine, a lightweight graph attention mechanism, and multimodal task input. The large power model semantic understanding engine performs semantic parsing of task information, the lightweight graph attention mechanism parses the task information, and corrects or supplements the semantic parsing results, while the multimodal task input can parse input data from different modalities. The latter includes a knowledge retrieval and reasoning mechanism, multi-agent collaborative reasoning, and a federated privacy reasoning mechanism. Finally, the output operation ticket can be re-verified through a multi-dimensional verification layer.

[0135] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0136] The power grid operation ticket generation device provided in the embodiments of this application is described below. The power grid operation ticket generation device described below can be referred to in correspondence with the power grid operation ticket generation method described above.

[0137] like Figure 3 As shown, this application provides a power grid operation ticket generation device 200, the device comprising:

[0138] The information acquisition module 201 is used to acquire task information of the operation ticket generation task, as well as the preset power grid knowledge graph;

[0139] The intent recognition module 202 is used to perform semantic parsing and semantic disambiguation on task information based on the power grid knowledge graph to obtain task intent information;

[0140] The knowledge retrieval module 203 is used to perform multi-hop knowledge retrieval in the power grid knowledge graph using task intent information to obtain structured knowledge related to the operation ticket generation task;

[0141] The multi-agent reasoning module 204 is used to perform primary equipment operation reasoning on the operation ticket generation task based on structured knowledge to obtain a primary equipment operation sequence, and to perform secondary equipment operation reasoning on the operation ticket generation task based on the primary equipment operation sequence to obtain a secondary equipment operation sequence.

[0142] The operation ticket generation module 205 is used to structurally integrate the operation sequences of primary equipment and secondary equipment to form a power grid operation ticket.

[0143] In the above embodiments, a pre-defined power grid knowledge graph is used to perform semantic parsing and disambiguation on the acquired task information, accurately identifying complex and ever-changing operational intentions. Based on this, the parsed task intention information is used to perform multi-hop knowledge retrieval in the power grid knowledge graph to obtain structured knowledge closely related to the current operational task, including deep information such as power grid topology, equipment status, and operating rules. Subsequently, based on this structured knowledge, the operation of primary equipment is first inferred, and then the corresponding operation sequence of secondary equipment is inferred based on the sequence of primary equipment operations, thereby realizing the automatic modeling and deduction of the deep logical relationship between primary and secondary equipment. Finally, the operation sequences of primary and secondary equipment are structurally integrated to automatically generate a complete, executable, and safety-compliant power grid operation ticket. Therefore, this solution improves the intelligence level of the operation ticket generation process, not only reducing reliance on manual experience and template matching, thus lowering the risk of human error, but also adapting to diverse switching operation scenarios, effectively improving the efficiency and scenario coverage of operation ticket preparation.

[0144] In one embodiment, the intent recognition module includes:

[0145] The model acquisition submodule is used to acquire a preset large-scale power model, which includes a semantic parsing module and a graph attention module.

[0146] The semantic parsing submodule is used to input task information into the semantic parsing module, so that the semantic parsing module can call the power grid knowledge graph to perform semantic parsing on the task information and obtain the task semantic information;

[0147] The semantic disambiguation module is used to input task information and task semantic information into the graph attention module, so that the graph attention module can call the power grid knowledge graph to parse the relationship between the task information and complete or correct the task semantic information based on the relationship parsing results to obtain the task intent information.

[0148] In one embodiment, the knowledge retrieval module includes:

[0149] The object extraction submodule is used to extract the target object from the task intent information;

[0150] The knowledge retrieval submodule is used to retrieve, starting from the target object, the equipment association relationship, state mapping relationship, normative constraint relationship and historical operation experience of the target object in the power grid knowledge graph.

[0151] The knowledge generation submodule is used to generate structured knowledge related to the operation ticket generation task based on device associations, state mapping relationships, specification constraints, and historical operation experience.

[0152] In one embodiment, the multi-agent inference module includes:

[0153] The equipment determination submodule is used to determine the primary equipment involved in the operation ticket generation task based on structured knowledge;

[0154] The first reasoning submodule is used to reason about the state transitions and operation sequences of each primary device using structured knowledge as the basis and a preset primary device operation agent, so as to obtain the primary device operation sequence.

[0155] In one embodiment, the multi-agent inference module includes:

[0156] The data acquisition submodule is used to acquire the protection coordination relationship of secondary equipment and the operation requirements of secondary equipment;

[0157] The second reasoning submodule is used to combine the secondary equipment protection coordination relationship and the secondary equipment operation requirements, and use a preset secondary equipment operation intelligent agent to reason about the secondary safety measures in the primary equipment operation sequence to obtain the secondary equipment operation sequence.

[0158] In one embodiment, before the primary equipment operation sequence and the secondary equipment operation sequence are structurally integrated, the power grid operation ticket generation device further includes:

[0159] The rule verification module is used to perform rule verification on the primary device operation sequence and the secondary device operation sequence using a preset safety constraint intelligent agent, and obtain the rule verification result.

[0160] The risk prediction module is used to simulate the changes in the power grid state after executing primary and secondary equipment operation sequences using a preset risk prediction intelligent agent, and to obtain risk prediction results.

[0161] The condition judgment module is used to structurally integrate the primary equipment operation sequence and the secondary equipment operation sequence if both the rule verification result and the risk prediction result meet the expectations; otherwise, it will re-perform the primary equipment operation reasoning and the secondary equipment operation reasoning.

[0162] In one embodiment, the power grid operation ticket generation device further includes:

[0163] The task partitioning module is used to divide the operation ticket generation task into multiple sub-tasks according to different regions when the operation ticket generation task involves cross-regional collaborative reasoning.

[0164] The federated inference module is used to distribute each subtask to the corresponding region for local primary equipment operation inference and secondary equipment operation inference, and to aggregate the inference results returned by each region to form a power grid operation ticket.

[0165] The division of modules in the above-described power grid operation ticket generation device is merely illustrative. In other embodiments, the power grid operation ticket generation device can be divided into different modules as needed to complete all or part of its functions. Each module in the above-described power grid operation ticket generation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0166] In one embodiment, this application also provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the power grid operation ticket generation method as described in any of the above embodiments.

[0167] In one embodiment, this application also provides a computer device storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the power grid operation ticket generation method as described in any of the above embodiments.

[0168] Indicatively, such as Figure 4 As shown, Figure 4 This is a schematic diagram of the internal structure of a computer device 300 provided in an embodiment of this application. The computer device 300 can be provided as a server. (Refer to...) Figure 4The computer device 300 includes a processing component 302, which further includes one or more processors, and memory resources represented by memory 301 for storing instructions, such as application programs, that can be executed by the processing component 302. The application programs stored in memory 301 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 302 is configured to execute instructions to perform the power grid operation ticket generation method of any of the above embodiments.

[0169] The computer device 300 may also include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate on an operating system stored in memory 301, such as Windows Server™, Mac OS X™, Unix™, Linux™, Free BSD™, or similar.

[0170] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0171] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a…" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. In this document, the singular forms "a," "an," and "the" may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “comprising / including” or “having” specify the presence of the stated features, wholes, steps, operations, components, parts or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0172] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0173] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. 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 this application. Therefore, this application 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 generating power grid operation tickets, characterized in that, The method includes: Obtain task information for the operation ticket generation task, as well as the preset power grid knowledge graph; Based on the power grid knowledge graph, semantic parsing and semantic disambiguation are performed on the task information to obtain task intent information; The task intent information is used to perform multi-hop knowledge retrieval in the power grid knowledge graph to obtain structured knowledge related to the operation ticket generation task; Based on the structured knowledge, a primary equipment operation reasoning is performed on the operation ticket generation task to obtain a primary equipment operation sequence. Based on the primary equipment operation sequence, a secondary equipment operation reasoning is performed on the operation ticket generation task to obtain a secondary equipment operation sequence. The primary equipment operation sequence and the secondary equipment operation sequence are structurally integrated to form a power grid operation ticket.

2. The method for generating power grid operation tickets according to claim 1, characterized in that, The step of performing semantic parsing and disambiguation on the task information based on the power grid knowledge graph to obtain task intent information includes: Obtain a preset large-scale power model, which includes a semantic parsing module and a graph attention module; The task information is input into the semantic parsing module, so that the semantic parsing module calls the power grid knowledge graph to perform semantic parsing on the task information to obtain task semantic information; The task information and the task semantic information are input into the graph attention module, so that the graph attention module calls the power grid knowledge graph to perform relational analysis on the task information, and completes or corrects the task semantic information based on the relational analysis results to obtain task intent information.

3. The method for generating power grid operation tickets according to claim 1, characterized in that, The step of using the task intent information to perform multi-hop knowledge retrieval in the power grid knowledge graph to obtain structured knowledge related to the operation ticket generation task includes: Extract the target object from the task intent information; Starting from the target object, the device association, state mapping relationship, normative constraint relationship and historical operation experience of the target object are sequentially retrieved in the power grid knowledge graph. Based on the device association, the state mapping, the specification constraint, and the historical operation experience, structured knowledge related to the operation ticket generation task is generated.

4. The method for generating power grid operation tickets according to claim 1, characterized in that, The step of performing a single-stage equipment operation reasoning on the operation ticket generation task based on the structured knowledge to obtain a single-stage equipment operation sequence includes: Based on the structured knowledge, the primary equipment involved in the operation ticket generation task is determined; Based on the structured knowledge, the pre-defined primary equipment operation agent is used to reason about the state transitions and operation sequences of each primary equipment to obtain the primary equipment operation sequence.

5. The method for generating power grid operation tickets according to claim 1, characterized in that, The step of performing secondary equipment operation reasoning on the operation ticket generation task based on the primary equipment operation sequence to obtain the secondary equipment operation sequence includes: Obtain the protection coordination relationships and operating requirements of secondary equipment; Combining the protection coordination relationship of the secondary equipment and the operation requirements of the secondary equipment, the secondary equipment operation sequence is obtained by using a preset secondary equipment operation intelligent agent to reason about the secondary safety measures in the operation sequence of the primary equipment.

6. The method for generating power grid operation tickets according to claim 1, characterized in that, Before structurally integrating the primary equipment operation sequence and the secondary equipment operation sequence, the method further includes: The primary device operation sequence and the secondary device operation sequence are validated using a preset safety constraint intelligent agent to obtain the rule validation result. Furthermore, by using a preset risk prediction intelligent agent to simulate the changes in the power grid state after executing the primary equipment operation sequence and the secondary equipment operation sequence, risk prediction results are obtained; If both the rule verification result and the risk prediction result meet expectations, then the primary equipment operation sequence and the secondary equipment operation sequence are structurally integrated; otherwise, the primary equipment operation reasoning and the secondary equipment operation reasoning are re-performed.

7. The method for generating power grid operation tickets according to any one of claims 1 to 6, characterized in that, The method further includes: When the operation ticket generation task involves cross-regional collaborative reasoning, the operation ticket generation task is divided into multiple sub-tasks according to different regions; Each subtask is distributed to the corresponding region for local primary equipment operation reasoning and secondary equipment operation reasoning, and the reasoning results returned by each region are aggregated to form a power grid operation ticket.

8. A power grid operation ticket generation device, characterized in that, The device includes: The information acquisition module is used to acquire task information for the operation ticket generation task, as well as a preset power grid knowledge graph; The intent recognition module is used to perform semantic parsing and semantic disambiguation on the task information based on the power grid knowledge graph to obtain task intent information; The knowledge retrieval module is used to perform multi-hop knowledge retrieval in the power grid knowledge graph using the task intent information to obtain structured knowledge related to the operation ticket generation task; The multi-agent reasoning module is used to perform primary equipment operation reasoning on the operation ticket generation task based on the structured knowledge to obtain a primary equipment operation sequence, and to perform secondary equipment operation reasoning on the operation ticket generation task based on the primary equipment operation sequence to obtain a secondary equipment operation sequence. The operation ticket generation module is used to structurally integrate the primary equipment operation sequence and the secondary equipment operation sequence to form a power grid operation ticket.

9. A storage medium, characterized in that: The storage medium stores computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the power grid operation ticket generation method as described in any one of claims 1 to 7.

10. A computer device, characterized in that, include: One or more processors, and memory; The memory stores computer-readable instructions, which, when executed by the one or more processors, perform the steps of the power grid operation ticket generation method as described in any one of claims 1 to 7.

Images