Power question and answer pair generation method and device of power system, computer equipment and readable storage medium

Abstract

This application relates to a method, apparatus, computer equipment, and readable storage medium for generating power question-and-answer pairs in a power system. It includes: retrieving a subgraph from a power knowledge graph based on a target power business scenario; extracting power entity relationship groups from the subgraph; determining a rule verification template based on the power entity relationship groups and a power operation procedure rule base; filling the rule verification template with the attribute values ​​of device nodes from the power entity relationship groups to obtain a logical deduction chain; mapping the logical deduction chain to a standard power terminology dictionary to obtain a semantic description sequence; filtering the semantic description sequence using a pre-set question-and-answer sentence template library to obtain a target logical framework; and obtaining candidate question-and-answer content blocks based on the target logical framework and the semantic description sequence, verifying the candidate question-and-answer blocks to obtain verified target power question-and-answer pairs. This method can improve the accuracy of question-and-answer pair generation.

Description

Technical Field

[0001] This application relates to the field of power system technology, and in particular to a method, apparatus, computer equipment, and readable storage medium for generating power question-and-answer pairs in a power system. Background Technology

[0002] With the development of intelligent and digital technologies in power systems, intelligent power question-and-answer technology has emerged. This technology features accurate response and efficient interaction, and can provide professional information answers and operational guidance for core business links such as power operation and maintenance, dispatching and repair. It is an important application technology in the digital transformation of power systems, which leads to the current power question-and-answer pair generation method based on preset template library and keyword matching.

[0003] Traditional technologies primarily rely on manually pre-setting power-related question-and-answer template libraries to pre-define the question-and-answer content for various power business scenarios, thereby meeting the basic question-and-answer needs in simple power business scenarios. However, existing methods are limited by the coverage of fixed templates and the mechanical nature of keyword matching, resulting in question-and-answer pairs that are prone to logical inconsistencies and low accuracy. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, and readable storage medium for generating power question-answer pairs in a power system, which can improve the accuracy and rationality of question-answer pair generation, in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a method for generating power question-and-answer pairs in a power system, including:

[0006] Based on the target power business scenario, a search is performed in the pre-built power knowledge graph to obtain a subgraph containing the target power equipment nodes and associated attribute edges, and the power entity relationship group to be processed is extracted from the subgraph.

[0007] Based on the type of attribute edge in the power entity relationship group, and combined with the preset power operation procedure rule library, a rule verification template for the power entity relationship group is determined, and the attribute values ​​of the equipment nodes in the power entity relationship group are filled into the rule verification template to obtain the logical deduction chain.

[0008] The semantic description sequence is obtained by mapping the causal logic nodes in the logical deduction chain with the preset standard dictionary of power terminology. Then, based on the logical transition relationships and causal connectors in the semantic description sequence, and combined with the preset question and answer template library, the target logical framework is obtained.

[0009] Based on the target logic framework, the semantic description sequence is reorganized into a logical closed loop of candidate question-answer content blocks, and the core device nodes in the candidate question-answer blocks are backtracked to the power knowledge graph to verify node connectivity, thus obtaining the verified target power question-answer pairs.

[0010] In one embodiment, based on the attribute edge types in the power entity relationship group and in conjunction with a preset power operation procedure rule base, a rule verification template for the power entity relationship group is determined, including:

[0011] Based on the type identifier of the associated attribute edges in the power entity relationship group, a search and matching is performed in the power operation procedure rule base to obtain a preliminary matching candidate rule pool. The applicable equipment level tags in the candidate rule pool are compared and filtered one by one with the equipment level attributes of the target power equipment nodes in the power entity relationship group to obtain the effective rule subgroups that conform to the current equipment level. The logical dependency relationship description field of each effective rule in the effective rule subgroup is identified to obtain the pre-dependent nodes and post-dependent nodes between rules. Based on the pre-dependent nodes and post-dependent nodes, a rule dependency directed graph reflecting the order of rule execution is constructed. The topological sorting traversal is performed with the starting node in the rule dependency directed graph to obtain an ordered rule arrangement list. Based on the rule arrangement list and combined with the equipment connection topology path of the target power equipment nodes in the power entity relationship group, a template is constructed and adapted to obtain a rule verification template for the power entity relationship group.

[0012] In one embodiment, based on the rule arrangement list, and combined with the device connection topology path of the target power equipment node in the power entity relationship group, a template is constructed and adapted to obtain a rule verification template for the power entity relationship group, including:

[0013] For each rule entry in the rule list, a standardized logical structure template is extracted to obtain an initial template component. This standardized logical structure template includes condition judgment slots, action execution slots, and result verification slots. Based on the device connection topology path of the target power equipment node in the power entity relationship group, the condition judgment slots of each standardized logical structure template in the initial template component are sequentially concatenated to obtain a coherent logical condition chain. Operation instruction codes and verification standard codes are extracted from the rule list and filled into the corresponding action execution slots and result verification slots in the logical condition chain to obtain an instantiated rule verification prototype. Based on the variable placeholder identifiers in the rule verification prototype, and combined with the device node attribute values ​​of the power entity relationship group, a mapping and binding relationship is constructed and transformed to obtain the rule verification template.

[0014] In one embodiment, the attribute values ​​of device nodes in the power entity relationship group are filled into the rule validation template to obtain a logical deduction chain, including:

[0015] Based on the variable placeholder identifiers in the rule validation template and the type identifiers of attribute edges in the power entity relationship group, a unique mapping relationship between the variable placeholder identifiers and the attribute key names of equipment nodes is constructed, resulting in an attribute filling index table. Based on the unique mapping relationship recorded in the attribute filling index table, the original equipment node attribute values ​​are extracted from the power entity relationship group to obtain the attribute value group to be filled. Based on the data format characteristics of the original equipment node attribute values ​​in the attribute value group to be filled, and combined with the data type constraints limited by the variable placeholder identifiers in the rule validation template, the original equipment node attribute values ​​are converted to a format consistency, resulting in a standardized attribute value sequence. Based on the pre-set logical operator connection structure in the rule validation template and the filling order specified in the attribute filling index table, the standardized attribute values ​​in the standardized attribute value sequence are substituted into the rule validation template to obtain an instantiated rule expression group. Based on the logical operator connection relationship within each expression in the instantiated rule expression group, the Boolean logic judgment result is propagated step by step to obtain a logical deduction chain from the premise to the final conclusion.

[0016] In one embodiment, a semantic description sequence is obtained by mapping the causal logic nodes in the logical deduction chain to a preset power terminology standard dictionary, including:

[0017] Based on the device object identifiers and operation action identifiers carried by the causal logic nodes in the logical deduction chain, and by searching in the standard dictionary of power terminology, a preliminary standardized term group is obtained. Based on the voltage level attribute tags and safety procedure attribute tags preset in the standard dictionary of power terminology for each term in the preliminary standardized term group, standardized qualifying modifiers are extracted to obtain complete term units with attribute qualifications. Based on the pre-set logical connection relationship type identifiers between adjacent causal logic nodes in the logical deduction chain, and by searching in the standard dictionary of power terminology, causal connectors are extracted to obtain standardized causal connectors. Based on the node arrangement order of the logical deduction chain, standardized causal connectors are inserted between two adjacent complete term units to obtain a preliminary semantic description fragment group. Based on the part-of-speech attribute annotations of the complete term units in the preliminary semantic description fragment group, and by combining the rigid word order modification rules in the standard dictionary of power terminology, the word arrangement of the preliminary semantic description fragment group is adjusted to obtain a semantic description sequence.

[0018] In one embodiment, the core device nodes in the candidate question-answer block are traced back to the power knowledge graph to verify node connectivity, resulting in verified target power question-answer pairs, including:

[0019] Based on the standardized equipment name entries contained in the semantic description sequence of the candidate question-and-answer blocks, a reverse retrieval is performed using a standard dictionary of power terminology to obtain the core node identifier group to be verified. For each node identifier in the core node identifier group, pre-stored attributes of the node identifier are extracted from the power knowledge graph to obtain a multi-dimensional physical attribute group. The pre-stored attributes include voltage level attributes, substation attributes, and equipment type classification attributes. Based on the equipment type classification attributes in the multi-dimensional physical attribute group, and combined with the preset cross-voltage level connection rules and cross-regional connection rules in the power knowledge graph, the edge relationship storage area of ​​the power knowledge graph is filtered to obtain a physically feasible association edge group. Based on the physically feasible association edge group, path search and question-and-answer pair generation are performed using the candidate question-and-answer blocks to obtain the verified target power question-and-answer pair.

[0020] In one embodiment, based on physically feasible association edge groups, path search and question-answer pair generation are performed in conjunction with candidate question-answer blocks to obtain verified target power question-answer pairs, including:

[0021] Based on the causal logical flow of the semantic description sequence in the candidate question-and-answer block, combined with the core node identifier group, the logical order of each node is determined, and the logical order is mapped to the starting node identifier and target node identifier pair of graph traversal, resulting in a directed traversal task queue. Based on each starting node identifier and target node identifier pair in the directed traversal task queue, path search is performed in combination with the physical feasible associated edge group, resulting in a path existence Boolean flag group and the corresponding original connected path record group. Based on the original connected path record group marked as true in the path existence Boolean flag group, the intermediate node identifier, associated edge type identifier, and the list of allowed operation action attributes of associated edges on the continuous path are extracted, resulting in a physical topology evidence chain. Based on the associated edge type identifier and the list of allowed operation action attributes in the physical topology evidence chain, a double comparison is performed in combination with the standardized causal connector words and standardized operation behavior words of the semantic description sequence in the candidate question-and-answer block, resulting in a double consistency verification result group. The double comparison includes logical consistency comparison and action compatibility comparison. Based on the records that pass the verification in the verification result group, the semantic description sequence of the corresponding candidate question-and-answer block is bound and encapsulated with the physical topology evidence chain, resulting in a verified target power question-and-answer pair.

[0022] Secondly, this application also provides an apparatus for generating power question-and-answer pairs in a power system, which implements the steps of the method as described in the first aspect.

[0023] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method as described in the first aspect.

[0024] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method as described in the first aspect.

[0025] The aforementioned power system question-and-answer pair generation method, device, computer equipment, and readable storage medium, based on the target power business scenario, retrieves data from a pre-constructed power knowledge graph to obtain a subgraph containing target power equipment nodes and associated attribute edges, and extracts power entity relationship groups to be processed from the subgraph; based on the type of attribute edges in the power entity relationship groups, combined with a pre-set power operation procedure rule base, a rule verification template for the power entity relationship groups is determined, and the attribute values ​​of the equipment nodes in the power entity relationship groups are filled into the rule verification template to obtain a logical deduction chain; based on the causal logic nodes in the logical deduction chain, terminology mapping is performed with a pre-set power terminology standard dictionary to obtain a semantic description sequence, and based on the logical transition relationships and causal connectors in the semantic description sequence, combined with a pre-set question-and-answer sentence template library, a target logical framework is obtained; based on the target logical framework, the semantic description sequence is reorganized into candidate question-and-answer content blocks with logical closed loops, and the core equipment nodes in the candidate question-and-answer blocks are backtracked to the power knowledge graph to verify node connectivity, resulting in verified target power question-and-answer pairs. This application utilizes a comprehensive design encompassing knowledge graph retrieval and extraction rule base template construction, terminology standardization mapping, and graph backtracking connectivity verification to generate power-related question-and-answer pairs that are logically sound, rigorous, and validated. This significantly improves the accuracy and reliability of the question-and-answer pairs, reduces manual maintenance costs, and meets the power system's demand for high-quality question-and-answer pairs. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart illustrating a method for generating power question-and-answer pairs in a power system, as shown in one embodiment.

[0028] Figure 2 This is a flowchart illustrating the process of determining a rule verification template for a group of power entity relationships in one embodiment.

[0029] Figure 3 This is a block diagram of a power system question-and-answer pair generation device in one embodiment;

[0030] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0032] In one embodiment, such as Figure 1 As shown, a method for generating power-related question-and-answer pairs in a power system is provided. This embodiment illustrates the method by applying it to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0033] Step 102: Based on the target power business scenario, search the pre-built power knowledge graph to obtain a subgraph containing the target power equipment nodes and associated attribute edges, and extract the power entity relationship group to be processed from the subgraph.

[0034] Optionally, target power business scenarios can be obtained through direct input from power business operators or through automatic identification of real-time data within the power system. These target power business scenarios refer to specific operational scenarios in power operation, dispatching, and maintenance, such as transformer maintenance at a 10kV substation, circuit breaker operation at a 35kV line, and topology verification of a 110kV busbar. Using these target power business scenarios as the retrieval basis, a targeted search is performed in a pre-constructed power knowledge graph. Based on the clearly defined core features of the target power business scenario, such as the type of power equipment and the type of business operation, corresponding nodes and edges in the power knowledge graph are matched. A subgraph containing the target power equipment node and its associated attribute edges is then selected. Nodes represent various types of power equipment, and edges represent the attributes of the equipment and the relationships between equipment. After determining the subgraph, it undergoes structured parsing to extract all power equipment nodes, their corresponding physical attributes, grid topology relationships between nodes, and operational rule associations. The extracted combinations of equipment, attributes, and relationships are then defined as power entity relationship groups to be processed. This ensures that subsequent generation of power question-and-answer pairs relies solely on these power entity relationship groups, thus aligning with the target power business scenario. These power entity relationship groups are a collection of structured power entities and various relationships between them within the subgraph. The pre-built power knowledge graph is a directed attribute graph formed by digitally modeling primary equipment, secondary equipment, grid topology relationships, physical attributes, and operational rules within the power system, and it is updated periodically.

[0035] Step 104: Based on the type of attribute edge in the power entity relationship group and combined with the preset power operation procedure rule library, determine the rule verification template for the power entity relationship group, and fill the attribute values ​​of the device nodes in the power entity relationship group into the rule verification template to obtain the logical deduction chain.

[0036] Optionally, based on the types of all attribute edges identified in the power entity relationship group, and in conjunction with the mandatory logical constraint rules in the pre-set power operation procedure rule base, a unique corresponding rule verification template is matched for the power entity relationship group, as described in steps 201 to 204. The power operation procedure rule base is a pre-built structured database storing mandatory logical constraint rules and industry standard rules for the operation and execution of various equipment and services in the power industry. Mandatory logical constraint rules refer to the core and inviolable subset of rules in the power operation procedure rule base, representing the logical principles and constraints that must be strictly followed in the operation and execution of equipment and services in the power system.

[0037] Step 106: Based on the causal logic nodes in the logical deduction chain, perform term mapping with the preset standard dictionary of power terminology to obtain a semantic description sequence. Then, based on the logical transition relationships and causal connectors in the semantic description sequence, and combined with the preset question-and-answer sentence template library, perform filtering to obtain the target logical framework.

[0038] The formation of a standardized semantic description sequence is specifically illustrated in steps 301 to 305. The power terminology standard dictionary refers to a pre-constructed, structured professional dictionary that stores standardized power industry terms and their corresponding definitions, application scenarios, and ambiguity resolution rules. Logical transition relationships refer to the opposite, relative, or concessive logical connections between two semantic units in a semantic description sequence; that is, the content of the latter semantic unit differs from the content of the former semantic unit in terms of semantic transition or contrast, or is supplemented by a concession based on the preceding semantic unit. Causal connectors refer to words used in a semantic description sequence to connect two semantic units with a cause-and-effect logical relationship. They are used to explicitly identify causal logic such as "condition-result," "operation-consequence," and "fault-phenomenon" in power business and to clearly define the causal derivation relationship between semantic units. Furthermore, the question-and-answer sentence template library is a pre-built, structured database that stores standardized question-and-answer sentence structures for all business scenarios in the power industry. It is used to match and adapt sentence frameworks for semantic description sequences, ensuring that the generated power question-and-answer pairs are standardized and consistent in sentence structure, clear and coherent in logical expression, and in line with the professional expression habits of the power industry.

[0039] For example, suppose the obtained logical derivation chain contains causal logic nodes: 10kV Chengdong Substation No. 1 Transformer (maintenance, premise), bus-side circuit breaker (opening, process), differential protection device (engagement, process), 10kV Section I busbar (disconnection, result). After mapping the above nodes to the standard dictionary of power terminology for deambiguity, the resulting semantic description sequence is "When 10kV Chengdong Substation No. 1 Transformer is undergoing maintenance, the 10kV Section I busbar-side circuit breaker connected to it needs to be opened, the transformer differential protection device remains engaged, and the 10kV Section I busbar is disconnected from the transformer." The causal connectives in this sequence are "when" and "need," with no logical transition relationship. Based on this condition, the question-and-answer sentence template library is filtered, and the sentence template "In the XX scenario, the XX device needs to perform the XX operation, the XX device is in the XX state, and finally achieves the XX result" is matched. Therefore, the target logical framework formed by combining with the semantic description sequence is "In the 10 kV Chengdong Substation No. 1 transformer maintenance scenario, the 10 kV I section bus side circuit breaker connected to the transformer needs to perform a tripping operation, the transformer differential protection device remains in the engaged state, and finally achieves the disconnection of the 10 kV I section bus from the transformer."

[0040] Step 108: Based on the target logical framework, the semantic description sequence is reorganized into a candidate question-and-answer content block with a logical closed loop, and the core device nodes in the candidate question-and-answer block are backtracked to the power knowledge graph to verify the node connectivity, thus obtaining the verified target power question-and-answer pair.

[0041] Based on the target logical framework, the semantic units in the semantic description sequence are rearranged, combined and expressed according to the preset logical order and sentence structure of the target logical framework to obtain candidate question-and-answer content blocks with a logical closed loop that have a question-and-answer correspondence. The candidate question-and-answer content blocks refer to content combinations that have the initial form of questions and corresponding answers, and are logically complete and semantically coherent. They include two parts: specific question expressions for the target power business scenario and answer expressions derived based on power knowledge graphs and operating procedure rules.

[0042] Continuing with the above embodiments, based on the target logic framework, the semantic description sequence is reorganized into candidate question-and-answer content blocks: {Question: When the No. 1 transformer of the 10kV Chengdong Substation is under maintenance, what operations need to be performed and what is the final effect? ​​Answer: When the No. 1 transformer of the 10kV Chengdong Substation is under maintenance, the circuit breaker on the 10kV I-section busbar side connected to it needs to be tripped, and the transformer differential protection device remains in the activated state, ultimately achieving the disconnection of the 10kV I-section busbar from the transformer}. The core equipment nodes in this candidate question-and-answer content block are extracted: the No. 1 transformer of the 10kV Chengdong Substation, the circuit breaker on the 10kV I-section busbar side, the 10kV I-section busbar, and the differential protection device. These core equipment nodes are traced back to the power knowledge graph to verify node connectivity. The verification result shows that the topological connectivity of each node conforms to the modeling relationship in the power knowledge graph. The verification passes, and this candidate question-and-answer content block is the target power question-and-answer pair.

[0043] The aforementioned power system question-and-answer pair generation method, device, computer equipment, and readable storage medium, based on the target power business scenario, retrieves data from a pre-constructed power knowledge graph to obtain a subgraph containing target power equipment nodes and associated attribute edges, and extracts power entity relationship groups to be processed from the subgraph; based on the type of attribute edges in the power entity relationship groups, combined with a pre-set power operation procedure rule base, a rule verification template for the power entity relationship groups is determined, and the attribute values ​​of the equipment nodes in the power entity relationship groups are filled into the rule verification template to obtain a logical deduction chain; based on the causal logic nodes in the logical deduction chain, terminology mapping is performed with a pre-set power terminology standard dictionary to obtain a semantic description sequence, and based on the logical transition relationships and causal connectors in the semantic description sequence, combined with a pre-set question-and-answer sentence template library, a target logical framework is obtained; based on the target logical framework, the semantic description sequence is reorganized into candidate question-and-answer content blocks with logical closed loops, and the core equipment nodes in the candidate question-and-answer blocks are backtracked to the power knowledge graph to verify node connectivity, resulting in verified target power question-and-answer pairs. This application utilizes a comprehensive design encompassing knowledge graph retrieval and extraction rule base template construction, terminology standardization mapping, and graph backtracking connectivity verification to generate power-related question-and-answer pairs that are logically sound, rigorous, and validated. This significantly improves the accuracy and reliability of the question-and-answer pairs, reduces manual maintenance costs, and meets the power system's demand for high-quality question-and-answer pairs.

[0044] In an exemplary embodiment, based on the attribute edge types in the power entity relationship group and in conjunction with a preset power operation procedure rule base, a rule verification template for the power entity relationship group is determined, including:

[0045] Step 202: Based on the type identifier of the associated attribute edge in the power entity relationship group, perform a search and matching in the power operation procedure rule base to obtain a preliminary matching candidate rule pool.

[0046] Among them, the type identifier is a pre-defined unique feature identifier for different categories of associated attribute edges in the power knowledge graph, used to distinguish the relationship type corresponding to the attribute edge, such as equipment connection relationship identifier, equipment affiliation relationship identifier, etc. The candidate rule pool refers to the collection of all power operation procedure rules that have a matching relationship with the attribute edge type of the power entity relationship group.

[0047] In one embodiment, assuming the target power service scenario is "overhaul of Transformer No. 1 at the 10kV Chengdong Substation", the associated attribute edge types and type identifiers in its power entity relationship group include: equipment connection relationship between the transformer and the busbar (001), equipment affiliation relationship between the transformer and the protection device (002), operation specification rule relationship corresponding to the transformer overhaul (003), and physical attribute relationship corresponding to the equipment voltage level (004). Based on this, the above four type identifiers are used as search keywords to search in the power operation procedure rule base. The matching rules include: "When the transformer is overhauled, the circuit breaker on the busbar side connected to it needs to be opened," "The protection device needs to maintain linkage with its main equipment," "The auxiliary equipment's on / off status needs to be confirmed before the 10kV equipment is overhauled," and "The equipment connection topology within the substation needs to match the operation rules."

[0048] Step 204: Compare and filter the applicable device level tags in the candidate rule pool with the device level attributes of the target power equipment nodes in the power entity relationship group one by one to obtain the valid rule subgroups that meet the current device level.

[0049] The applicable equipment level label refers to the identifier of the voltage level or functional level of the power equipment to which the rule applies, which is used to label each rule in the power operation procedure rule base. Examples include 10 kV level, 35 kV level, 110 kV level, transformer main equipment level, and protection device auxiliary equipment level. The equipment level attribute refers to the inherent attribute of each power equipment node in the power knowledge graph, reflecting the equipment's voltage level or functional level, and has a one-to-one matching relationship with the applicable equipment level label.

[0050] Continuing with the above embodiments, the applicable equipment level labels for each rule in the candidate rule pool are as follows: "When a transformer is under maintenance, the circuit breaker on its connected bus side needs to be tripped" (10kV level, transformer main equipment level), "The protection device needs to maintain linkage with its main equipment" (10kV level, protection device auxiliary equipment level), "The auxiliary equipment's on / off status needs to be confirmed before 10kV equipment maintenance" (10kV level), and "The equipment connection topology within the substation needs to match the operation rules" (35kV and above level). The equipment level attribute of the target power equipment nodes in the power entity relationship group is all 10kV level, with transformers at the 10kV main equipment level and protection devices at the 10kV auxiliary equipment level. After comparing the tags and attributes one by one, the rule “The equipment connection topology in the substation must match the operation rules” with the applicable equipment level tag “35 kV and above” was removed. The remaining three rules were retained, and the effective rule subgroups were: {“When the transformer is under maintenance, the circuit breaker on the bus side connected to it must be opened”, “The protection device must be linked with the main equipment”, and “The auxiliary equipment must be confirmed to be in operation or not before the maintenance of 10 kV equipment”}.

[0051] Step 206: Identify the logical dependency description field of each valid rule in the valid rule subgroup to obtain the preceding dependent nodes and subsequent dependent nodes between rules, and construct a directed rule dependency graph that reflects the order of rule execution based on the preceding dependent nodes and subsequent dependent nodes.

[0052] Among them, the logical dependency description field refers to the structured field preset in each rule of the power operation procedure rule base, which is used to describe the execution order dependency between the rule and other rules.

[0053] Continuing with the above embodiment, the three rules within the effective rule subgroup are assigned unique node identifiers as follows: Rule 1 (When the transformer is under maintenance, the circuit breaker on the bus side connected to it needs to be tripped) - Node 1; Rule 2 (The protection device needs to maintain linkage with its main equipment) - Node 2; Rule 3 (Before the maintenance of 10 kV equipment, the status of auxiliary equipment needs to be confirmed) - Node 3. The logical dependency description fields of each rule are extracted and identified: the dependent node of Rule 3 is Node 1, meaning "confirming the status of auxiliary equipment" is a prerequisite for "tripping the bus side circuit breaker"; the dependent node of Rule 1 is Node 2, meaning "tripping the bus side circuit breaker" is a prerequisite for "linking the protection device with the main equipment"; Rule 2 has no dependent node, and Rule 3 has no prerequisite node. Using Nodes 1, 2, and 3 as the core, and "Node 3 → Node 1" and "Node 1 → Node 2" as directed edges, a directed rule dependency graph is constructed. This directed graph visually reflects the rule execution order as: Rule 3 → Rule 1 → Rule 2.

[0054] Step 208: Perform topological sorting and traversal of the starting node in the rule dependency directed graph to obtain an ordered list of rules. Based on the list of rules, combine the device connection topology path of the target power equipment node in the power entity relationship group to construct and adapt a template to obtain a rule verification template for the power entity relationship group.

[0055] Continuing with the above embodiment, following the rule-dependent directed graph of step 206, the starting node is node 3 (rule 3). A topological sorting traversal is performed starting from node 3, with the traversal order being node 3 → node 1 → node 2. The ordered rule list formed after the traversal is as follows: 1. Before the maintenance of 10 kV equipment, the status of auxiliary equipment must be confirmed; 2. When the transformer is under maintenance, the circuit breaker on the bus side connected to it must be opened; 3. The protection device must maintain a linkage state with the main equipment to which it belongs.

[0056] The rule verification template generated by the embodiments of the present invention not only conforms to the mandatory logical constraints of the power operation procedures, but also fits the actual physical relationship of the power equipment, thus avoiding problems such as the rule template being disconnected from the actual power business operation and the rule execution order being chaotic.

[0057] In an exemplary embodiment, based on the rule arrangement list, and combined with the device connection topology path of the target power equipment node in the power entity relationship group, a template is constructed and adapted to obtain a rule verification template for the power entity relationship group, including:

[0058] Step 302: Extract the standardized logical structure template for each rule entry in the rule arrangement list to obtain the initial template component; the standardized logical structure template includes condition judgment slots, action execution slots and result verification slots.

[0059] The standardized logical structure template refers to a general structure template with fixed logical modules that is preset for power operation rules. It includes condition judgment slots, action execution slots, and result verification slots. The condition judgment slots are structured fields used to fill in the preconditions required for the execution of the rule. The action execution slots are structured fields used to fill in the core operation actions of the rule. The result verification slots are structured fields used to fill in the result standards that need to be verified after the execution of the rule. The initial template components refer to a set of standardized logical structure templates that correspond one-to-one with the rule entries in the rule arrangement list and have not been spliced ​​or filled with content.

[0060] Continuing with the above embodiments, the rule list is as follows: 1. Before overhauling 10kV equipment, the auxiliary equipment's on / off status must be confirmed; 2. During transformer overhaul, the circuit breaker on the bus side connected to it must be tripped; 3. The protection device must maintain linkage with its main equipment. Each rule entry is disassembled and a standardized logic structure template is extracted. The resulting initial template components correspond one-to-one with the rule entries, and the arrangement order is consistent. Specifically, it includes: Template 1: [Condition Judgment Slot] → [Action Execution Slot: Confirm Auxiliary Equipment On / Off Status] → [Result Verification Slot: Auxiliary Equipment On / Off Status Meets Overhaul Requirements] Template 2: [Condition Judgment Slot] → [Action Execution Slot: Perform Tripping Operation on Bus Side Circuit Breaker] → [Result Verification Slot: Bus Side Circuit Breaker Successfully Tripped]; Template 3: [Condition Judgment Slot] → [Action Execution Slot: Maintain Linkage Between Protection Device and Main Equipment] → [Result Verification Slot: Linkage Between Protection Device and Main Equipment is Normal].

[0061] Step 304: Based on the device connection topology path of the target power equipment node in the power entity relationship group, the condition judgment slots of each standardized logical structure template in the initial template component are sequentially connected to obtain a coherent logical condition chain.

[0062] Optionally, the device connection topology path refers to the topological connection order between target power device nodes in the power knowledge graph. During the splicing process, the association logic of the device connection topology path is followed, and the result verification slots of the previous template are logically connected with the condition judgment slots of the next template.

[0063] Continuing with the above embodiment, the device connection topology path of the target power equipment node extracted from the power entity relationship group is: differential protection device → 10kV Chengdong Substation No. 1 Transformer → 10kV Section I Busbar → Busbar-side Circuit Breaker. The initial template assembly includes Template 1, Template 2, and Template 3 arranged in this order. Based on this device connection topology path, the condition judgment slots of each template are connected in series. The result verification slot of the previous template is logically connected with the condition judgment slot of the next template. The resulting logical condition chain is: "When carrying out 10kV transformer maintenance work, first confirm that the operation status of the differential protection device (auxiliary equipment) meets the maintenance requirements, then perform the operation on the 10kV Section I busbar-side circuit breaker connected to the 10kV transformer, and ensure that the circuit breaker's tripping status meets the requirements. Finally, maintain the association status between the differential protection device and its associated 10kV transformer."

[0064] Step 306: Extract the operation instruction code and verification standard code from the rule arrangement list, and fill the corresponding action execution slot and result verification slot in the logical condition chain to obtain the instantiated rule verification prototype.

[0065] Optionally, the operation instruction code refers to a unique numerical identifier preset for the specific operation action corresponding to each power operation rule, corresponding one-to-one with the standardized operation action description in the power operation procedure rule base; the verification standard code refers to a unique numerical identifier preset for the result judgment standard corresponding to each power operation rule, corresponding one-to-one with the standardized verification standard description in the power operation procedure rule base. The operation instruction code and verification standard code are respectively filled one-to-one with the action execution slots and result verification slots of each standardized logical structure template in the initial template component after the logical condition chain is assembled. After all slots are filled, the assembled logical condition chain and the filled action execution slots and result verification slots are integrated into a rule verification framework to obtain an instantiated rule verification prototype. The slots of this rule verification prototype still contain variable placeholder identifiers that are not bound to specific device attributes.

[0066] Continuing with the above embodiment, the operation instruction codes and verification standard codes corresponding to the three rule entries in the rule arrangement list are as follows: Rule 1 (operation instruction code: C001 - confirm activation / deactivation; verification standard code: Y001 - activation / deactivation compliant), Rule 2 (operation instruction code: C002 - execute tripping; verification standard code: Y002 - tripping successful), Rule 3 (operation instruction code: C003 - maintain linkage; verification standard code: Y003 - linkage normal). Fill the above code into the corresponding slots of each template in the initial template component, and then integrate it with the logical condition chain obtained in step 304. The resulting instantiation rule verification prototype is: "When carrying out 10 kV transformer maintenance work, first [condition judgment], execute the operation of [C001-confirm commissioning / discharging] auxiliary equipment commissioning / discharging status, and verify [Y001-commendation / discharging compliance]; then [condition judgment], execute the operation of [C002-execute tripping] bus side circuit breaker, and verify [Y002-tripping successful]; finally [condition judgment], execute the operation of [C003-maintain linkage] protection device and main equipment status, and verify [Y003-linkage normal]."

[0067] Step 308: Based on the variable placeholder identifier in the rule validation prototype, the mapping and binding relationship is constructed and transformed by combining the device node attribute values ​​of the power entity relationship group to obtain the rule validation template.

[0068] Optionally, the variable placeholder identifier refers to the placeholder identifier in the rule validation prototype that is used to refer to a specific power equipment node and has not been filled with actual attribute values; the equipment node attribute value refers to the inherent specific attribute of the target power equipment node in the power knowledge graph.

[0069] Continuing with the above embodiments, the variable placeholders in the rule validation prototype are identified as: "10 kV transformer", "auxiliary equipment", "busbar side circuit breaker", and "main equipment". The corresponding equipment node attribute values ​​in the power entity relationship group are: "10 kV Chengdong Substation No. 1 transformer", "10 kV Chengdong Substation No. 1 transformer differential protection device", "10 kV Chengdong Substation 10 kV Section I busbar side circuit breaker", and "10 kV Chengdong Substation No. 1 transformer". After establishing the mapping and binding relationship, the placeholder identifier is converted into a specific attribute value, and the final rule verification template for this power entity relationship group is as follows: "When carrying out maintenance work on the No. 1 transformer of the 10 kV Chengdong Substation, first confirm the on / off status of the differential protection device of the No. 1 transformer of the 10 kV Chengdong Substation, and verify that the on / off status meets the maintenance requirements; then perform a tripping operation on the circuit breaker on the 10 kV I section busbar side of the 10 kV Chengdong Substation connected to the No. 1 transformer of the 10 kV Chengdong Substation, and verify that the circuit breaker has been successfully tripped; finally, maintain the linkage status between the differential protection device of the No. 1 transformer of the 10 kV Chengdong Substation and the No. 1 transformer of the 10 kV Chengdong Substation, and verify that the linkage status is normal."

[0070] This invention enables the transformation from a general standardized template to a customized rule verification template that fits the actual business scenario, equipment topology, and equipment attributes. This results in a rule verification template that has both a standardized power operation logic framework and fits the equipment association characteristics and actual operation requirements of specific power businesses.

[0071] In an exemplary embodiment, the attribute values ​​of device nodes in the power entity relationship group are filled into the rule validation template to obtain a logical deduction chain, including:

[0072] Step 402: Based on the variable placeholder identifier in the rule validation template and combined with the type identifier of the attribute edge in the power entity relationship group, construct a unique mapping relationship between the variable placeholder identifier and the attribute key name of the device node to obtain the attribute filling index table;

[0073] Among them, the device node attribute key name refers to the unique naming identifier used to characterize the specific attributes of the device node in the power entity relationship group.

[0074] Continuing with the above embodiment, the variable placeholder identifiers in the rule validation template are: {#device1#, #device2#, #device3#, #voltage level#, #belonging site#, #connected object#}; the type identifiers of the associated attribute edges in the power entity relationship group include: connection relationship 001, membership relationship 002, maintenance operation rule 003, and voltage level attribute 004. A unique device node attribute key name is matched to each variable placeholder identifier to construct a unique mapping relationship and generate an attribute-filled index table, specifically:

[0075]

[0076] Step 404: Based on the unique mapping relationship recorded in the attribute filling index table, extract the original equipment node attribute values ​​from the power entity relationship group to obtain the attribute value group to be filled.

[0077] The original device node attribute values ​​refer to the specific attribute information of device nodes directly stored in the power entity relationship group without formatting, which are the inherent attribute data of device nodes in the power knowledge graph. All extracted original device node attribute values ​​are arranged according to the order of variable placeholders in the attribute filling index table to form a set of attribute values ​​that correspond one-to-one with the positions to be filled in the template, and finally obtain the attribute value group to be filled.

[0078] Continuing with the above embodiment, following step 402 of filling the attribute index table, using the device node attribute key name in the table as the keyword, the original device node attribute values ​​are extracted from the power entity relationship group corresponding to "10 kV Chengdong Substation No. 1 Transformer Overhaul", and arranged according to the index table order. The resulting attribute value group to be filled is: {10 kV Chengdong Substation No. 1 Transformer, 10 kV Chengdong Substation No. 1 Transformer Differential Protection Device, 10 kV Chengdong Substation 10 kV Section I Busbar Side Circuit Breaker, 10 kV, Chengdong Substation, 10 kV Chengdong Substation 10 kV Section I Busbar}. Each attribute value corresponds to #Equipment 1#, #Equipment 2#, #Equipment 3#, #Voltage Level#, #Substation#, and #Connected Object# in the index table.

[0079] Step 406: Based on the data format characteristics of the original device node attribute values ​​in the attribute value group to be filled, and combined with the data type constraints limited by the variable placeholder identifier in the rule validation template, the original device node attribute values ​​are converted to a format consistency to obtain a standardized attribute value sequence.

[0080] Data format characteristics refer to the format attributes of the original attribute values, such as the representation, character length, unit labeling, and naming conventions. For example, whether it includes units, whether it is a full name, and whether it conforms to the naming conventions for power equipment. Data type constraints refer to the format specifications that the rule validation template pre-sets for each variable placeholder identifier, which must be followed when filling attribute values. For example, equipment names must be complete and standardized names, voltage levels must include numerical values ​​and units, and connected objects must be labeled with specific equipment types. Using data type constraints as the conversion standard, the original equipment node attribute values ​​are converted to be format-consistent, that is, the representation and storage format of the original equipment node attribute values ​​are adjusted to a form that completely matches the data type constraints defined by the variable placeholder identifiers.

[0081] Continuing with the above embodiments, following the group of attribute values ​​to be filled in step 404, after analyzing and extracting the data format characteristics of the original attribute values, the format is converted by combining the data type constraints of the variable placeholders in the rule validation template (the equipment name is a complete standardized name with no redundant descriptions). Since the original attribute values ​​already conform to the power standardization specifications, no correction is needed, and a standardized attribute value sequence is directly formed: {10kV Chengdong Substation No. 1 Transformer, 10kV Chengdong Substation No. 1 Transformer Differential Protection Device, 10kV Chengdong Substation 10kV Section I Busbar Side Circuit Breaker, 10kV, Chengdong Substation, 10kV Chengdong Substation 10kV Section I Busbar}. If the original attribute values ​​contain abbreviations such as "No. 1 Transformer" or "Chengdong Transformer", the device will be converted to the complete standardized name "10kV Chengdong Substation No. 1 Transformer" or "Chengdong Substation".

[0082] Step 408: Based on the pre-defined logical operator connection structure and the filling order specified in the attribute filling index table in the rule validation template, the standardized attribute values ​​in the standardized attribute value sequence are substituted into the rule validation template to obtain the instantiated rule expression group.

[0083] The logical operator connection structure refers to the combination and structural form of operators used in the rule validation template to connect the variable placeholder identifiers and represent the logical relationships between the attribute values, such as "when..., perform... operation", "if..., then validate...", "first..., then..., finally...", etc. Simultaneously, the filling order specified in the attribute filling index table is retrieved. The filling order refers to the arrangement order of the variable placeholder identifiers in the attribute filling index table, that is, the order in which the standardized attribute values ​​are filled. Using the logical operator connection structure as a framework and the filling order as a basis, each standardized attribute value in the standardized attribute value sequence is precisely substituted into the corresponding variable placeholder identifier position in the rule validation template to complete the attribute value filling of the template.

[0084] Continuing with the above embodiment, assume that the logical operator connection structure of the rule verification template is as follows: "When carrying out maintenance work on equipment 1# of #voltage level## site##, first confirm the commissioning / discharging status of #equipment 2# and verify that the commissioning / discharging status meets the maintenance requirements; then perform a tripping operation on #equipment 3# on the #connection object# side connected to #equipment 1# and verify that #equipment 3# has successfully tripped; finally, maintain the linkage status between #equipment 2# and its parent #equipment 1# and verify that the linkage status is normal"; the filling order specified in the attribute filling index table is #equipment 1#, #equipment 2#, #equipment 3#, #voltage level#, #site#, #connection object#. The device substitutes the standardized attribute value sequence into the corresponding position of the template, and the resulting set of instantiated rule expressions is as follows: Expression 1: When carrying out maintenance work on the No. 1 transformer of the 10 kV Chengdong Substation, confirm the on / off status of the differential protection device of the No. 1 transformer of the 10 kV Chengdong Substation, and verify that the on / off status meets the maintenance requirements; Expression 2: Perform a tripping operation on the circuit breaker on the 10 kV I section busbar side of the 10 kV Chengdong Substation connected to the No. 1 transformer of the 10 kV Chengdong Substation, and verify that the circuit breaker has been successfully tripped; Expression 3: Maintain the linkage status between the differential protection device of the No. 1 transformer of the 10 kV Chengdong Substation and the No. 1 transformer of the 10 kV Chengdong Substation, and verify that the linkage status is normal.

[0085] Step 410: Based on the logical operator connection relationship within each expression in the instantiation rule expression group, the Boolean logic judgment result is deduced step by step to obtain the logical deduction chain from the premise to the final conclusion.

[0086] The logical operator connection relationship refers to the logical association between conditions, actions, and results within a single expression, while also extracting the logical operator connection relationships between expressions. The Boolean logic judgment result is either "true / compliant / satisfied" or "false / incompatible / not satisfied". Following the logical order of the instantiated rule expression group, the Boolean logic judgment results of each expression are propagated and deduced level by level. Only when the judgment result of the preceding expression is "true / compliant / satisfied" will the subsequent expression be triggered for execution; otherwise, the deduction terminates and outputs a deduction failure result.

[0087] Continuing with the above embodiments, following the instantiation of the rule expression group in step 408, the logical operator connection relationships within and between each expression are extracted: Expression 1 and Expression 2 have a "sequential progression" relationship, Expression 2 and Expression 3 have a "sequential progression" relationship, and each expression has a logical relationship of "condition → action → verification result". Boolean logic judgment and step-by-step deduction are performed as follows: For Expression 1, the condition that the maintenance work on Transformer No. 1 of the 10kV Chengdong Substation is carried out is considered a prerequisite (valid). Confirming the differential protection device's on / off state is considered an action (compliant with the rules). The compliance of the on / off state with the maintenance requirements is considered a verification result (satisfied). The judgment result is "valid". Based on the judgment result of Expression 1, for Expression 2, the condition that the circuit breaker on the designated bus side is opened is considered an action (compliant with the rules). The successful opening of the circuit breaker is considered a verification result (satisfied). The judgment result is "valid". Based on the judgment result of Expression 2, for Expression 3, the condition that the protection device and main equipment are linked is considered an action (compliant with the rules). The normal linkage state is considered a verification result (satisfied). The judgment result is "valid". Integrating the deduction process and results, the logical deduction chain is as follows: "Prerequisite: Carry out maintenance work on transformer No. 1 of Chengdong Substation → Execution step 1: Confirm the differential protection device of transformer No. 1 of Chengdong Substation is in operation or not, and the verification result is that it meets the maintenance requirements → Execution step 2: Perform a tripping operation on the 10kV I-section busbar side circuit breaker connected to transformer No. 1 of Chengdong Substation, and the verification result is that the tripping was successful → Execution step 3: Maintain the linkage status between the differential protection device of transformer No. 1 of Chengdong Substation and its main transformer, and the verification result is that the linkage is normal → Final conclusion: The preliminary operation and status verification of the maintenance of transformer No. 1 of Chengdong Substation both meet the requirements of the power operation procedures."

[0088] The embodiments of the present invention ensure that the generation of the logical deduction chain relies on accurate equipment attribute data and follows strict power operation logic and Boolean deduction rules, thus ensuring that the generated logical deduction chain has rigorous business logic, standardized attribute descriptions, and continuous causal progression.

[0089] In an exemplary embodiment, a semantic description sequence is obtained by mapping the causal logic nodes in the logical deduction chain to a preset standard dictionary of power terminology, including:

[0090] Step 502: Based on the device object identifier and operation action identifier carried by the causal logic node in the logical deduction chain, and combined with the power terminology standard dictionary, a preliminary standardized term group is obtained.

[0091] Here, causal logic nodes refer to the nodes in the logical deduction chain that represent the conditions, actions, and results of power business operations; equipment object identifiers refer to the identifiers in causal logic nodes that uniquely identify specific power equipment objects; and operation action identifiers refer to the feature identifiers in causal logic nodes that uniquely identify specific power operation actions. All equipment object identifiers and operation action identifiers are used as search keywords to search in a pre-defined standard dictionary of power terminology, filtering out matching standardized power professional terms.

[0092] In one embodiment, the causal logic nodes in the logical deduction chain are assumed to be: {carry out maintenance work on transformer No. 1 of 10 kV Chengdong Substation, confirm the differential protection device of transformer No. 1 of 10 kV Chengdong Substation in operation status, verify that the operation status meets the maintenance requirements, perform a tripping operation on the 10 kV I-section bus side circuit breaker connected to transformer No. 1 of 10 kV Chengdong Substation, verify that the circuit breaker has been successfully tripped, maintain the linkage status between the differential protection device of transformer No. 1 of 10 kV Chengdong Substation and its main transformer, and verify that the linkage status is normal}. The equipment object identifiers carried by each node are {10 kV transformer, differential protection device, bus side circuit breaker}, and the operation action identifiers are {carry out maintenance, confirm the operation status, verify compliance, perform tripping, maintain linkage status, verify normal}. Using the above identifiers as search keywords, a preliminary standardized term group was obtained by searching the standard dictionary of power terminology: {Carry out maintenance work on a 10 kV transformer, confirm the activation / deactivation status of the differential protection device, verify that the activation / deactivation status meets the maintenance requirements, perform a tripping operation on the bus-side circuit breaker connected to the 10 kV transformer, verify that the circuit breaker tripping was successful, maintain the linkage status between the differential protection device and the main transformer, and verify that the linkage status is normal}.

[0093] Step 504: Based on the voltage level attribute tags and safety regulation attribute tags preset in the standard dictionary of power terminology for each term in the preliminary standardized terminology group, extract standardized qualifying modifiers to obtain complete terminology units with attribute qualifying.

[0094] Optionally, voltage level attribute labels refer to standardized attribute labels in the standard dictionary of power terminology that characterize the voltage level of each power equipment term, such as 10 kV, 35 kV, etc.; safety procedure attribute labels refer to standardized attribute labels in the standard dictionary of power terminology that characterize the power safety procedures followed by each power operation term, such as maintenance procedures, operating procedures, calibration procedures, etc. Standardized qualifying modifiers refer to standardized terms that correspond one-to-one with the attribute labels and are used to modify the terminology entries.

[0095] Continuing with the above embodiments, following the preliminary standardized terminology group in step 301, the voltage level attribute tag and safety procedure attribute tag are retrieved from the standard dictionary of power terminology for each entry, and the corresponding standardized limiting modifiers are extracted. For example, the tag for the entry "carry out 10 kV transformer maintenance work" is {10 kV, maintenance procedure}, and the modifier is {10 kV, under the power equipment maintenance procedure}; the tag for the entry "confirm the differential protection device's on / off status" is {10 kV, operating procedure}, and the modifier is {10 kV, under the power equipment operating procedure}. Then, the modifiers are combined with the corresponding terms to obtain a complete term unit with attribute limitations: {Carry out transformer maintenance work under the 10 kV power equipment maintenance procedure, confirm the differential protection device's on / off status under the 10 kV power equipment operation procedure, verify that the on / off status under the 10 kV power equipment verification procedure meets the maintenance requirements, perform tripping operation on the bus-side circuit breaker connected to the transformer under the 10 kV power equipment operation procedure, verify that the circuit breaker tripping is successful under the 10 kV power equipment verification procedure, maintain the linkage status between the differential protection device and the main transformer under the 10 kV power equipment operation procedure, and verify that the linkage status under the 10 kV power equipment verification procedure is normal}.

[0096] Step 506: Based on the pre-set logical connection relationship type identifier between adjacent causal logical nodes in the logical deduction chain, and combined with the standard dictionary of power terminology, extract causal connection words to obtain standardized causal connection words.

[0097] Optionally, the logical connection type identifier refers to a unique identifier pre-defined to characterize the logical relationship between causal logical nodes. Connecting words matching the identifier are retrieved from the standardized causal connection terminology database of the power terminology standard dictionary. Then, the connecting words corresponding one-to-one with the logical connection type identifier are organized into standardized causal connecting words. These standardized causal connecting words conform to power industry language standards and are used to connect adjacent power operation logics. The number of standardized causal connecting words is consistent with the number of groups of adjacent causal logical nodes in the logical deduction chain.

[0098] Continuing with the above embodiment, the logical connection relationship type identifier between adjacent causal logical nodes is: {sequential execution relationship, result verification relationship, sequential execution relationship, result verification relationship, sequential execution relationship, result verification relationship}. Based on this, the above identifier is searched in the standard dictionary of power terminology, and the standardized causal connection words obtained are: {first, verification result is, second, verification result is, subsequently, verification result is}, and the order of the connection words is consistent with the logical connection relationship type identifier.

[0099] Step 508: Based on the node arrangement order of the logical deduction chain, standardized causal connectors are inserted between two adjacent complete term units to obtain a preliminary semantic description fragment group.

[0100] Continuing with the above embodiments, based on the complete terminology unit of step 504 and the standardized causal connectives of step 506, the connectives are inserted between adjacent complete terminology units according to the node arrangement order of the logical deduction chain. The resulting preliminary semantic description fragment group is as follows: 1. To carry out transformer maintenance work under the 10 kV power equipment maintenance procedure, firstly, confirm the differential protection device's on / off status under the 10 kV power equipment operation procedure; 2. Confirm the differential protection device's on / off status under the 10 kV power equipment operation procedure. The verification result is that the on / off status under the 10 kV power equipment verification procedure meets the maintenance requirements; 3. The on / off status under the 10 kV power equipment verification procedure meets the maintenance requirements. The requirements are as follows: 1. Perform a tripping operation on the busbar-side circuit breaker connected to the transformer under the 10 kV power equipment operation procedure; 2. Perform a tripping operation on the busbar-side circuit breaker connected to the transformer under the 10 kV power equipment operation procedure, and verify that the circuit breaker tripping under the 10 kV power equipment operation procedure is successful; 3. Verify that the circuit breaker tripping under the 10 kV power equipment operation procedure is successful, and then maintain the linkage state between the differential protection device and the main transformer under the 10 kV power equipment operation procedure; 4. Maintain the linkage state between the differential protection device and the main transformer under the 10 kV power equipment operation procedure, and verify that the linkage state under the 10 kV power equipment operation procedure is normal.

[0101] Step 510: Based on the part-of-speech attribute annotation of complete term units in the preliminary semantic description fragment group, and combined with the rigid word order modification rules in the standard dictionary of power terminology, the word arrangement of the preliminary semantic description fragment group is adjusted to obtain the semantic description sequence.

[0102] Among them, part-of-speech tagging refers to the grammatical part-of-speech attributes marked for each standardized power term in the standard dictionary of power terminology, such as noun, verb, attributive, adverbial, etc. Rigid word order modification rules refer to the pre-set grammatical word order rules in the standard dictionary of power terminology that conform to the expression norms of power professional language. The rules are formulated in combination with the professional expression habits of the power industry and clarify the arrangement order of words with different parts of speech such as nouns, verbs, and attributives.

[0103] Continuing with the above embodiments, after identifying the part-of-speech attribute annotations of each complete term unit, the word order is adjusted according to the rigid word order modification rules of the standard dictionary of power terminology (in power operation, voltage level modifiers must be adjacent to equipment nouns, and procedure adverbs must be placed before the operation action verbs). Redundant modifier expressions in the fragment are corrected, and the adjusted semantic fragments are integrated to obtain the semantic description sequence as follows: "Carry out maintenance work on a 10 kV transformer under the power equipment maintenance procedure. First, confirm the 10 kV differential protection device's on / off status under the power equipment operation procedure. The verification result is that the on / off status meets the maintenance requirements under the 10 kV power equipment verification procedure. Second, perform the tripping operation under the power equipment operation procedure on the bus-side circuit breaker connected to the 10 kV transformer. The verification result is that the 10 kV circuit breaker has successfully achieved tripping under the power equipment verification procedure. Subsequently, maintain the linkage status between the 10 kV differential protection device and the 10 kV main transformer under the power equipment operation procedure. The verification result is that the linkage status meets the normal requirements under the 10 kV power equipment verification procedure."

[0104] The semantic description sequence generated by the embodiments of the present invention not only possesses complete power operation logic that strictly follows the logical deduction chain, but also adopts standardized terminology, attribute constraints and language order of the power industry, thus achieving a deep integration of power operation logic and professional language norms.

[0105] In an exemplary embodiment, the core device nodes in the candidate question-answer block are traced back to the power knowledge graph to verify node connectivity, resulting in a verified target power question-answer pair, including:

[0106] Step 602: Based on the standardized equipment name entries contained in the semantic description sequence in the candidate question-answer block, and combined with the standard dictionary of power terminology, a reverse retrieval is performed to obtain the core node identifier group to be verified.

[0107] Standardized equipment name entries refer to equipment names in the candidate question-and-answer content blocks that conform to power industry standards after being mapped from the standard dictionary of power terminology. Reverse retrieval refers to a retrieval method that matches the corresponding unique identifiers of equipment in the dictionary using standardized terms as search terms. Equipment node identifiers refer to the feature identifiers preset for each power equipment node in the power knowledge graph, used to uniquely identify the equipment. Core node identifier groups refer to the set of unique identifiers corresponding to all core power equipment in the candidate question-and-answer content blocks that can be identified in the power knowledge graph. Here, core power equipment refers to the power equipment necessary to realize the core functions of this business.

[0108] In one embodiment, the standardized equipment name entries included in the semantic description sequence of the candidate question-and-answer content block are: {10 kV transformer, 10 kV differential protection device, 10 kV busbar circuit breaker, 10 kV main transformer, 10 kV bus}. The above entries are then retrieved in reverse from the standard dictionary of power terminology to extract matching equipment node identifiers, resulting in the core node identifier group to be verified: {B001, B002, B003, B001, B004}, where B001 is the identifier for a 10 kV transformer / main transformer, B002 is the identifier for a 10 kV differential protection device, B003 is the identifier for a 10 kV busbar circuit breaker, and B004 is the identifier for a 10 kV bus, and the identifiers are arranged in the same order as the standardized equipment name entries.

[0109] Step 604: For each node identifier in the core node identifier group, extract the pre-stored attributes of the node identifier in the power knowledge graph to obtain a multi-dimensional physical attribute group; the pre-stored attributes include voltage level attributes, substation attributes, and equipment type classification attributes.

[0110] Specifically, the process involves retrieving pre-stored attributes of equipment node identifiers in the power knowledge graph, namely, extracting the voltage level attribute, substation attribute, and equipment type classification attribute corresponding to each equipment node identifier. The voltage level attribute refers to the inherent voltage operation level attribute of the power equipment; the substation attribute refers to the name or number attribute of the substation to which the power equipment belongs, representing the physical deployment location of the equipment; and the equipment type classification attribute refers to the attribute classified according to the function and type of the power equipment.

[0111] Continuing with the above embodiments, the core node identifier group is {B001, B002, B003, B001, B004}. Pre-stored attributes corresponding to each identifier are extracted from the power knowledge graph. The resulting multi-dimensional physical attribute group, ordered by identifier, is as follows: 1. B001: Voltage level attribute is 10 kV, substation attribute is Chengdong Substation, equipment type classification attribute is primary equipment - transformer; 2. B002: Voltage level attribute is 10 kV, substation attribute is Chengdong Substation, equipment type classification attribute is secondary equipment - protection device; 3. B003: Voltage level attribute is 10 kV, substation attribute is Chengdong Substation, equipment type classification attribute is primary equipment - circuit breaker; 4. B001: Voltage level attribute is 10 kV, substation attribute is Chengdong Substation, equipment type classification attribute is primary equipment - transformer; 5. B004: Voltage level attribute is 10 kV, substation attribute is Chengdong Substation, equipment type classification attribute is primary equipment - busbar.

[0112] Step 606: Based on the equipment type classification attribute in the multidimensional physical attribute group, and combined with the preset cross-voltage level connection rules and cross-regional connection rules in the power knowledge graph, filter in the edge relationship storage area of ​​the power knowledge graph to obtain physically feasible association edge groups.

[0113] Among them, the cross-voltage level connection rule refers to the pre-set rules in the power knowledge graph that regulate whether physical connections are allowed between devices of different voltage levels. For example, 10 kV devices can only be connected to 10 kV devices or specific voltage regulating devices, and direct connection to 35 kV and above devices without voltage regulation is prohibited. The cross-regional connection rule refers to the pre-set rules in the power knowledge graph that regulate whether physical connections are allowed between devices deployed in different substations. For example, devices within the same substation can be directly connected, while devices in different substations can only be connected through transmission lines. Next, based on the equipment type classification attribute as the basic screening condition, and combined with cross-voltage level connection rules and cross-regional connection rules, the edge relationship storage area of ​​the power knowledge graph is used to screen the connection edges between equipment nodes that simultaneously satisfy the equipment type matching, cross-voltage level connection rules, and cross-regional connection rules. All the selected valid connection edges are organized into a set containing the start node, end node, and connection type of the connection edge. This set is the physically feasible association edge group. The edge relationship storage area refers to the area in the power knowledge graph that stores the topological connection relationships between all equipment nodes, including the start node, end node, connection type, and constraint conditions of the connection edge.

[0114] Continuing with the above embodiment, all device nodes have a voltage level of 10 kV, a substation attribute of Chengdong Substation, and are classified as primary equipment or secondary equipment. Based on the preset rules of the power knowledge graph (equipment of the same voltage level and within the same substation can directly establish physical connections; secondary protection devices can establish subordinate connections with primary transformers; primary transformers can establish connections with primary busbars; primary busbars can establish connections with primary circuit breakers), the feasible physical association edge groups obtained by filtering in the graph edge relationship storage area are: {B001-B002 (subordinate connection), B001-B004 (physical connection), B004-B003 (physical connection), B001-B002 (subordinate connection)}, corresponding to feasible association edges between differential protection devices and transformers, transformers and busbars, busbars and circuit breakers, and differential protection devices and transformers, respectively.

[0115] Step 608: Based on the physically feasible associated edge group, path search and question-answer pair generation are performed in combination with candidate question-answer blocks to obtain the verified target power question-answer pair.

[0116] This invention verifies and optimizes question-and-answer pairs from the perspective of power physical topology and connection rules, solving the problems that may exist in candidate question-and-answer blocks, such as missing device topology connectivity and inconsistencies between the description of connection relationships and actual business operations.

[0117] In an exemplary embodiment, based on physically feasible association edge groups, path search and question-answer pair generation are performed in conjunction with candidate question-answer blocks to obtain verified target power question-answer pairs, including:

[0118] Step 702: Based on the causal logical flow of the semantic description sequence in the candidate question-answer block, combined with the core node identifier group, determine the logical order of each node, and map the logical order to the starting node identifier and target node identifier pair of graph traversal to obtain the directed traversal task queue.

[0119] Here, causal logical flow refers to the overall logical direction representing the sequence and causal relationships of power operation processes within the semantic description sequence. Logical sequence refers to the order in which the operations corresponding to each core equipment node appear in the power business process. The starting node identifier refers to the identifier of the core equipment node that serves as the starting point in each traversal path, and the target node identifier refers to the identifier of the core equipment node that serves as the ending point in each traversal path. One starting node identifier can correspond to one or more target node identifiers, and the order of the node pairs is consistent with the logical sequence.

[0120] In one embodiment, the causal logic flow extracted from the semantic description sequence of the candidate question-and-answer block is: confirm the status of the differential protection device → operate the transformer → operate the bus → operate the circuit breaker, and the core node identifier group is {B001 (transformer), B002 (differential protection device), B003 (circuit breaker), B004 (bus)}. The logical order of each node is determined as: B002→B001→B004→B003, which is mapped to start-target node identifier pairs, resulting in a directed traversal task queue of: {(B002, B001), (B001, B004), (B004, B003)}.

[0121] Step 704: Based on each pair of starting node identifiers and target node identifiers in the directed traversal task queue, and combined with the physically feasible associated edge group, a path search is performed to obtain the path existence Boolean flag group and the corresponding original connected path record group.

[0122] The topology path search refers to the process of searching for a connected path from the starting node identifier to the target node identifier based on the physical connection relationships between device nodes recorded in the physically feasible associated edge group. If a valid connected path exists, it is marked as "true"; otherwise, it is marked as "false," resulting in a Boolean flag group for path existence. Simultaneously, for each node pair marked as "true," its complete connected path information from the starting node to the target node is recorded, including all node identifiers on the path and the associated edge identifiers between nodes. All path information records are arranged in the node pair order of the directed traversal task queue to obtain the original connected path record group.

[0123] Continuing with the above embodiment, the directed traversal task queue is {(B002, B001), (B001, B004), (B004, B003)}, and the physically feasible edge pair is {B001-B002 (membership connection), B001-B004 (physical connection), B004-B003 (physical connection)}. A path search is performed on each node pair, resulting in a Boolean flag group for path existence: {true, true, true}. Simultaneously, the original connected path information is recorded, resulting in the original connected path record group: {B002→B001, B001→B004, B004→B003}. Each path information corresponds one-to-one with a node pair.

[0124] Step 706: Based on the original connected path record group marked as true in the path existence Boolean flag group, extract the intermediate node identifier, associated edge type identifier and the list of allowed operation action attributes of the associated edge on the continuous path to obtain the physical topology evidence chain.

[0125] Optionally, the intermediate node identifier refers to the identifier of other device nodes in the connected path besides the starting node and the target node; if there are no intermediate nodes, this field is blank; the associated edge type identifier refers to the type feature identifier corresponding to the associated edge between nodes in the connected path; the allowed operation action attribute list refers to the set of power operation actions preset for each associated edge in the power knowledge graph, which are allowed to be executed between the device nodes corresponding to the edge, representing the device operation permissions corresponding to the associated edge.

[0126] Continuing with the above embodiment, all Boolean flag groups for the paths are "true", and the original connected path record groups are {B002→B001, B001→B004, B004→B003}, with no intermediate nodes in any of the paths. Parsing and extracting the information from each path yields the following physical topology evidence chain: 1. Path B002→B001: Intermediate node identifier (none), associated edge type identifier (membership connection), list of allowed operation actions (confirmation status, maintaining linkage); 2. Path B001→B004: Intermediate node identifier (none), associated edge type identifier (physical connection), list of allowed operation actions (disconnection, connection); 3. Path B004→B003: Intermediate node identifier (none), associated edge type identifier (physical connection), list of allowed operation actions (opening, closing, performing operation).

[0127] Step 708: Based on the associated edge type identifier and the list of allowed operation action attributes in the physical topology evidence chain, a double comparison is performed on the standardized causal connector words and standardized operation behavior words of the semantic description sequence in the candidate question and answer block to obtain a double consistency verification result group; the double comparison includes logical consistency comparison and action compatibility comparison.

[0128] The standardized operational behavior terms refer to the standardized vocabulary representing power operation actions in the semantic description sequence. A dual comparison operation is performed on the above four types of information: logical consistency comparison and action compatibility comparison. Logical consistency comparison compares the inherent logical relationship corresponding to the associated edge type identifier in the physical topology evidence chain with the logical relationship represented by the standardized causal connection words between corresponding nodes in the semantic description sequence to determine if they are consistent. Action compatibility comparison compares the standardized operational behavior terms in the semantic description sequence with the list of allowed operational action attributes for the corresponding associated edges in the physical topology evidence chain to determine if the operational behavior is within the allowed operational action attribute list, i.e., whether the operational behavior conforms to the device operation permissions of the associated edge. If both the logical consistency comparison and the action compatibility comparison pass, the verification is considered "passed"; if either comparison fails, the verification is considered "failed".

[0129] Continuing with the above embodiment, the physical topology evidence chain includes the associated edge type identifiers of three paths and a list of allowed operation action attributes. The standardized causal connection words in the candidate question-and-answer block semantic description sequence are {belonging to, connecting, connecting}, and the standardized operation behavior words are {confirmation status, delisting, circuit breaking}. A dual comparison is performed: 1. Path B002→B001: Logical consistency comparison (belonging connection matches "belonging to"), action compatibility comparison ("confirmation status" is in the allowed list) → verification passed; 2. Path B001→B004: Logical consistency comparison (physical connection matches "connecting"), action compatibility comparison ("delisting" is in the allowed list) → verification passed; 3. Path B004→B003: Logical consistency comparison (physical connection matches "connecting"), action compatibility comparison ("circuit breaking" is in the allowed list) → verification passed. The final dual-consistency verification result group is {verification passed, verification passed, verification passed}.

[0130] Step 710: Based on the records that have passed verification in the verification result group, bind and encapsulate the semantic description sequence of the corresponding candidate question and answer block with the physical topology evidence chain to obtain the target power question and answer pair that has passed verification.

[0131] Among them, binding and encapsulation refers to associating and integrating the business logic information of the semantic description sequence with the topological feature information of the physical topological evidence chain.

[0132] Continuing with the above embodiments, all verification result groups are "verification passed," and the corresponding physical topology evidence chain consists of complete topology information for three paths. The semantic description sequence of the candidate question-and-answer block is the operational logic description of the overhaul of transformer No. 1 at the 10kV Chengdong Substation. The device binds and encapsulates this semantic description sequence with the physical topology evidence chain, and simultaneously performs format standardization correction on the question-and-answer content. The final verified target power question-and-answer pair is: {Question: When overhauling transformer No. 1 at the 10kV Chengdong Substation, what operations need to be performed and what is the final effect?} Answer: When overhauling the No. 1 transformer at the 10kV Chengdong Substation, first confirm that the 10kV differential protection device belonging to the transformer meets the overhaul requirements (topology evidence: B002-B001 subordinate connection, allowing confirmation operation). Then, perform a disconnection operation on the 10kV I-section busbar physically connected to the transformer (topology evidence: B001-B004 physical connection, allowing disconnection operation). By performing a tripping operation on the 10kV I-section busbar side circuit breaker physically connected to the busbar, the two are disconnected (topology evidence: B004-B003 physical connection, allowing tripping operation). Finally, maintain the linkage state between the differential protection device and the main transformer, and ensure that the linkage is normal. Ultimately, the physical disconnection of the No. 1 transformer and the 10kV I-section busbar at the 10kV Chengdong Substation is achieved, meeting the safety operation requirements for overhaul.

[0133] The embodiments of the present invention strengthen the physical topology verification and logical action verification of power question-answer pairs, so that the rationality and accuracy of the target power question-answer pairs are directly supported by physical topology evidence, and solve the problems of logical and physical mismatch and operation actions not matching device connection permissions that may exist in candidate question-answer blocks.

[0134] Based on the same inventive concept, such as Figure 3 As shown in the illustration, this application also provides a power database generation device for implementing the power database generation method for the power system described above, comprising: a subgraph retrieval and entity extraction module 801, a logic deduction chain generation module 802, a target logic framework filtering module 803, and a backtracking verification module 804. The solution provided by this device is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more embodiments of the power database generation device for the power system provided below can be found in the limitations of the power database generation method for the power system described above, and will not be repeated here.

[0135] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores a power knowledge graph. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for generating power question-and-answer pairs in a power system.

[0136] 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.

[0137] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0138] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

Claims

1. A method for generating power question-and-answer pairs in a power system, characterized in that, The method includes: Based on the target power business scenario, a search is performed in the pre-constructed power knowledge graph to obtain a subgraph containing the target power equipment nodes and associated attribute edges, and the power entity relationship group to be processed is extracted from the subgraph; Based on the type of attribute edge in the power entity relationship group, and combined with the preset power operation procedure rule library, a rule verification template for the power entity relationship group is determined, and the attribute values ​​of the device nodes in the power entity relationship group are filled into the rule verification template to obtain the logical deduction chain. The semantic description sequence is obtained by mapping the causal logic nodes in the logical deduction chain with the preset standard dictionary of power terminology. Then, based on the logical transition relationships and causal connectors in the semantic description sequence, and combined with the preset question and answer template library, the target logical framework is obtained. Based on the target logical framework, the semantic description sequence is reorganized into a logically closed-loop candidate question-and-answer content block, and the core device nodes in the candidate question-and-answer block are backtracked to the power knowledge graph to verify node connectivity, thus obtaining the verified target power question-and-answer pair.

2. The method according to claim 1, characterized in that, The rule verification template for the power entity relationship group is determined based on the attribute edge type in the power entity relationship group and in conjunction with a preset power operation procedure rule base, including: Based on the type identifier of the associated attribute edge in the power entity relationship group, a search and matching is performed in the power operation procedure rule base to obtain a preliminary matching candidate rule pool; The applicable equipment level labels in the candidate rule pool are compared and filtered one by one with the equipment level attributes of the target power equipment nodes in the power entity relationship group to obtain the effective rule subgroups that meet the current equipment level. The logical dependency description field of each valid rule in the valid rule subgroup is identified to obtain the preceding dependent nodes and the subsequent dependent nodes between rules, and a rule dependency directed graph reflecting the order of rule execution is constructed based on the preceding dependent nodes and the subsequent dependent nodes. A topological sorting traversal is performed on the starting node of the rule-dependent directed graph to obtain an ordered list of rules. Based on the list of rules, a template is constructed and adapted by combining the device connection topology path of the target power equipment node in the power entity relationship group to obtain a rule verification template for the power entity relationship group.

3. The method according to claim 2, characterized in that, The process involves constructing and adapting a template based on the rule arrangement list and the device connection topology path of the target power equipment node in the power entity relationship group, resulting in a rule verification template for the power entity relationship group, including: Each rule entry in the rule arrangement list is extracted using a standardized logical structure template to obtain an initial template component; the standardized logical structure template includes condition judgment slots, action execution slots, and result verification slots. Based on the device connection topology path of the target power equipment node in the power entity relationship group, the condition judgment slots of each standardized logical structure template in the initial template component are sequentially connected to obtain a coherent logical condition chain. Extract the operation instruction code and verification standard code from the rule arrangement list, and fill the corresponding action execution slot and result verification slot in the logical condition chain to obtain the instantiated rule verification prototype. Based on the variable placeholder identifier in the rule validation prototype, the mapping and binding relationship is constructed and transformed by combining the device node attribute values ​​of the power entity relationship group to obtain the rule validation template.

4. The method according to claim 3, characterized in that, The step of filling the device node attribute values ​​in the power entity relationship group into the rule verification template to obtain the logical deduction chain includes: Based on the variable placeholder identifier in the rule validation template and the type identifier of the attribute edge in the power entity relationship group, a unique mapping relationship between the variable placeholder identifier and the attribute key name of the device node is constructed to obtain the attribute filling index table. Based on the unique mapping relationship recorded in the attribute filling index table, the original equipment node attribute values ​​are extracted from the power entity relationship group to obtain the attribute value group to be filled. Based on the data format characteristics of the original device node attribute values ​​in the attribute value group to be filled, and combined with the data type constraints limited by the variable placeholder identifier in the rule validation template, the original device node attribute values ​​are converted to a format consistency to obtain a standardized attribute value sequence. Based on the pre-defined logical operator connection structure and the filling order specified in the attribute filling index table in the rule validation template, the standardized attribute values ​​in the standardized attribute value sequence are substituted into the rule validation template to obtain an instantiated rule expression group. Based on the logical operator connection relationship within each expression in the instantiation rule expression group, the Boolean logic judgment result is deduced step by step to obtain the logical deduction chain from the premise to the final conclusion.

5. The method according to claim 1, characterized in that, The semantic description sequence is obtained by mapping the causal logic nodes in the logical deduction chain to a pre-set standard dictionary of power terminology, including: Based on the device object identifier and operation action identifier carried by the causal logic node in the logical deduction chain, and combined with the power terminology standard dictionary, a preliminary standardized term group is obtained by retrieval. Based on the voltage level attribute tags and safety procedure attribute tags preset in the standard dictionary of power terminology for each term in the preliminary standardized terminology group, standardized qualifying modifiers are extracted to obtain complete terminology units with attribute qualifying; Based on the pre-set logical connection relationship type identifier between adjacent causal logical nodes in the logical deduction chain, and combined with the power terminology standard dictionary, causal connection words are extracted to obtain standardized causal connection words; Based on the node arrangement order of the logical deduction chain, the standardized causal connector is inserted between two adjacent complete term units to obtain a preliminary semantic description fragment group; Based on the part-of-speech tagging of complete term units in the preliminary semantic description fragment group, and combined with the rigid word order modification rules in the standard dictionary of power terminology, the word arrangement of the preliminary semantic description fragment group is adjusted to obtain the semantic description sequence.

6. The method according to claim 1, characterized in that, The step of backtracking the core device nodes in the candidate question-answer block to the power knowledge graph to verify node connectivity, and obtaining the verified target power question-answer pairs, includes: Based on the standardized equipment name terms contained in the semantic description sequence of the candidate question and answer block, a reverse retrieval is performed in conjunction with the standard dictionary of power terminology to obtain the core node identifier group to be verified. For each node identifier in the core node identifier group, the pre-stored attributes of the node identifier are extracted from the power knowledge graph to obtain a multi-dimensional physical attribute group; the pre-stored attributes include voltage level attributes, substation attributes, and equipment type classification attributes. Based on the equipment type classification attributes in the multidimensional physical attribute group, and combined with the preset cross-voltage level connection rules and cross-regional connection rules in the power knowledge graph, the edge relationship storage area of ​​the power knowledge graph is filtered to obtain physically feasible association edge groups. Based on the physically feasible associated edge group, path search and question-answer pair generation are performed in conjunction with the candidate question-answer block to obtain the verified target power question-answer pair.

7. The method according to claim 6, characterized in that, The process of performing path search and question-answer pair generation based on the physically feasible association edge group and the candidate question-answer blocks to obtain verified target power question-answer pairs includes: Based on the causal logical flow of the semantic description sequence in the candidate question-answer block, combined with the core node identifier group, the logical order of each node is determined, and the logical order is mapped to the starting node identifier and target node identifier pair of graph traversal to obtain the directed traversal task queue. Based on each pair of start node identifiers and target node identifiers in the directed traversal task queue, and combined with the physical feasible associated edge group, path search is performed to obtain the path existence Boolean flag group and the corresponding original connected path record group. Based on the original connected path record group marked as true in the Boolean flag group, extract the intermediate node identifier, associated edge type identifier and the list of allowed operation actions of associated edges on the continuous path to obtain the physical topology evidence chain; Based on the associated edge type identifier and the list of allowed operation action attributes in the physical topology evidence chain, a double comparison is performed on the standardized causal connect words and standardized operation behavior words of the semantic description sequence in the candidate question and answer block to obtain a double consistent verification result group; the double comparison includes logical consistency comparison and action compatibility comparison. Based on the records that have passed verification in the verification result group, the semantic description sequence of the corresponding candidate question and answer block is bound and encapsulated with the physical topology evidence chain to obtain the target power question and answer pair that has passed verification.

8. A power system question-and-answer pair generation device, characterized in that, The device includes: The subgraph retrieval and entity extraction module is used to retrieve data from a pre-built power knowledge graph based on the target power business scenario, obtain a subgraph containing the target power equipment nodes and associated attribute edges, and extract the power entity relationship group to be processed from the subgraph. The logic deduction chain generation module is used to determine the rule verification template for the power entity relationship group based on the type of attribute edge in the power entity relationship group and in combination with the preset power operation procedure rule library, and fill the device node attribute values ​​in the power entity relationship group into the rule verification template to obtain the logic deduction chain; The target logic framework filtering module is used to map the causal logic nodes in the logical deduction chain to a preset standard dictionary of power terminology to obtain a semantic description sequence. Based on the logical transition relationships and causal connectors in the semantic description sequence, and combined with a preset question-and-answer template library, the module filters to obtain the target logic framework. The backtracking verification module is used to reorganize the semantic description sequence into a logical closed loop of candidate question-answer content blocks based on the target logical framework, and backtrack the core device nodes in the candidate question-answer blocks to the power knowledge graph to verify the node connectivity, thereby obtaining the verified target power question-answer pairs.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images