Power transmission and transformation reduction verification method and system based on knowledge graph
By using a knowledge graph-based intelligent review method, the problems of low efficiency and poor accuracy of traditional manual review methods in power transmission and transformation projects have been solved. This has enabled efficient and systematic design document management and review, improving the accuracy and security of the review process.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-08
- Publication Date
- 2026-04-07
AI Technical Summary
Traditional manual review methods are inefficient, prone to omissions, and lack standardized criteria in power transmission and transformation projects. They are difficult to guarantee the accuracy and traceability of the review, lack the ability to process multi-source heterogeneous data in a unified manner, and lack effective risk prediction and anomaly judgment mechanisms.
A knowledge graph-based approach is adopted, which uses optical character recognition technology to process unstructured data, constructs a knowledge graph subgraph, generates a list of design review standard documents, and uses natural language processing technology for text segmentation and semantic role labeling to establish an intelligent review and verification model. The review report is generated by combining the intelligent prompt word library.
It has enabled efficient and systematic management of power transmission and transformation design documents, improved the coverage and accuracy of design review, reduced subjective bias, supported incremental learning and dynamic updates of knowledge graphs, improved review efficiency and report preparation efficiency, and reduced safety hazards.
Smart Images
Figure CN121809614A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of power transmission and transformation project review, and particularly relates to a power transmission and transformation reduction verification method and system based on a knowledge graph. BACKGROUND
[0002] With the continuous expansion of power transmission and transformation projects, the design files are large in quantity and complex in structure, involve multiple dimensions such as electrical systems, power transmission lines and environmental constraints, and the traditional manual design review method faces problems such as low efficiency, easy omission and non-uniform standards. The existing review methods mainly rely on manual checking of paper or electronic documents, lack of unified processing capability for multi-source heterogeneous data, and have difficulties in structured analysis of drawing, table and text information, which leads to that the review results are easily affected by subjective factors and it is difficult to guarantee the accuracy and traceability of the review. In addition, with the continuous updating of national power grid and industry standards, it is time-consuming and laborious for manual tracking of specification provisions and parameter matching, and there is a lack of effective risk prediction and abnormality judgment mechanism. Under this background, how to realize intelligent and systematic review of power transmission and transformation design files, quickly identify potential violations and risk points, and generate structured review reports has become a technical problem to be solved. In recent years, artificial intelligence technologies such as knowledge graph, natural language processing and graph neural network have achieved certain application in the field of engineering design and document analysis, but there is still a lack of overall solution in the power transmission and transformation design review scene, which organically combines multi-modal design data, standard specifications and review experience. Therefore, developing a technical method capable of automatically analyzing non-structured review files, modeling knowledge graphs, verifying review points and intelligently generating reports has become an important research direction to improve the efficiency of power transmission and transformation design review and guarantee design compliance and safety. SUMMARY
[0003] To solve the above problems in the prior art, the application provides a power transmission and transformation reduction verification method based on a knowledge graph, The object of the application can be achieved by the following technical solutions: S1: obtaining power transmission and transformation station review file data, identifying text content through optical character recognition technology for unstructured data, and constructing a knowledge graph subgraph according to the natural language processing technology to extract the text content and generate a design review standard file list; S2: performing text segmentation based on the design review standard file list, annotating according to semantic roles, constructing a knowledge graph review dimension, and obtaining a structured review template; S3: establishing an intelligent review verification model according to the structured review template, inputting the power transmission and transformation station review file data into the model, combining file integrity with the structured review template, and verifying the review points of the structured power transmission and transformation project through a loss function combined with cross entropy. S4: Constructing an intelligent review prompt library, taking the review points as prompt words, writing into an intelligent review verification model, and analyzing review suggestions according to semantic retrieval to drive the intelligent review prompt library to generate a power transmission review report.
[0004] Specifically, the method for obtaining the power transmission review file data is: calling the synchronous data acquisition online review data of the message queue, triggering the data retrieval instruction execution structure query by setting the access token for the local archived data, and attaching a timestamp and a version identifier to the obtained data record.
[0005] Specifically, the design review standard file list is created according to the review file data, a parameterized template is created, parameters are grouped according to the review scene, a review data-entity-review relationship table is established, and historical review archives, real-time review reports and specification standards are integrated according to the review data-entity-review relationship table.
[0006] Specifically, the method for reviewing the review content is: The semantic feature coding of the review parameters extracted from the design file is performed, the text specification content is mapped to a feature vector representation that can be used for comparison operation, and a structured correspondence relationship between the design parameters and the standard rules is constructed; By analyzing the association path between the design parameter nodes and the standard constraint nodes, the semantic correlation degree and the logical consistency deviation are calculated, and a potential violation mark list is formed according to the weight level of the deviation amount, and the design parameters that do not meet the specification requirements are automatically marked as risk nodes; According to the review matching result, a risk identification strategy model is established, the risk events triggered by the parameter deviation are evaluated, and the evaluation threshold is calibrated combined with the historical review data feedback to form the review judgment boundary.
[0007] Specifically, the method for generating the design review standard file list is: according to the historical review state data of the power transmission station, the data is classified according to the voltage level and the file type, an initial data set of the review file is established, the review file information is identified based on the initial data set, and the scanned documents are analyzed by optical character recognition technology, the entities are mapped to nodes, and the relationships are mapped to edges, and the design review standard file list is generated.
[0008] Specifically, the verification method of the review points of the structured power transmission project is: The review point set formed in the structured review template is taken as a verification reference, the design parameters, associated standard items and their logical relationships corresponding to each review point are taken as a check unit, and the input power transmission design file data is compared and searched for consistency piece by piece; The unmatched items identified by the retrieval are subjected to semantic consistency analysis and normative compliance detection using a validation model. A graded label is set for the degree of deviation, and the matching rules and deviation thresholds are dynamically adjusted in combination with historical review sample data.
[0009] Specifically, the intelligent review prompt word library performs semantic clustering according to the dimensions of electrical parameter verification, structural stability verification, and environmental compliance review. The description of review violation characteristics and corresponding standard words are structurally encapsulated in the form of prompt units to generate a set of prompt word entries that can be retrieved and called. Through semantic retrieval engine and knowledge graph context reasoning mechanism, a mapping relationship is established between the set of prompt word entries and the output results of the review verification model. The corresponding prompt word template is triggered according to the review deviation type and risk level.
[0010] Specifically, the power transmission and transformation review report groups and organizes the verification results according to the review dimension, and records the matching status and triggered prompt words of each review node as review items to form the basic data structure of the review report. Based on the template rules in the prompt word library, the review items are reorganized according to the standardized semantic expression method. The review report content is uniformly encapsulated and rendered by the message generation engine, and timestamps and review process hash signature information are attached.
[0011] Specifically, the method for constructing the intelligent review and verification model is as follows: A training sample set is constructed based on the set of review points extracted from the structured review template and the actual power transmission and transformation design document data. The design parameters, text description information and their corresponding standard item tags in the sample set are feature-encoded, and a training data index structure is established according to the review dimensions, risk categories and standard item mapping relationships. By establishing a validation model framework for graph-structured data, the node information and relationship paths in the knowledge graph are used as model inputs, and the key points of the review are automatically identified through node embedding calculation and semantic association analysis. The review validity model is calibrated by combining actual review feedback data, and the model judgment output is re-labeled and weighted based on the confirmation results in historical review cases. The model review weights and deviation thresholds are iteratively updated based on the incremental learning mechanism.
[0012] Specifically, the optical character recognition technology divides and slices text, tables, and annotation areas based on layout analysis algorithms, and performs character recognition and symbol extraction on each area; after the recognition results are generated, semantic error correction and field matching are performed by combining engineering domain lexicon and standardized encoding, and finally the effectively recognized content is output in a structured manner and automatically mapped to the corresponding entity nodes in the knowledge graph.
[0013] Specifically, the initial dataset of the review documents includes original design documents, parameter specifications, standard clause reference records, and field survey records extracted from the power transmission and transformation engineering design management platform, document management system, and historical review archive. Before access, the data is categorized and identified by timestamp, version number, and source system code, and is divided into three categories according to data type: text files, image files, and structured parameter files, generating a unified basic input data set.
[0014] Specifically, a knowledge graph-based power transmission and transformation reduction verification system is characterized by comprising: Data acquisition module: Acquires data from review documents of power transmission and substations, identifies text content for unstructured data using optical character recognition technology, extracts the text content using natural language processing technology to construct a knowledge graph subgraph, and generates a list of design review standard documents; Content parsing module: Based on the design review standard document list, the text is segmented, labeled according to semantic roles, and a knowledge graph review dimension is constructed to sort out the review content of the design documents in the pre-review and review stages, and a structured review template is obtained; Review and Verification Module: Based on the structured review template, an intelligent review and verification model is established. The review document data of the power transmission and substation is input into the model and combined with the document integrity and the structured review template. The review points of the structured power transmission and substation project are verified by combining the loss function with cross-entropy. Lexicon Building Module: Constructs an intelligent review prompt lexicon, writes the review points as prompt words into the intelligent review verification model, and drives the intelligent review prompt lexicon to generate a power transmission and transformation review report based on the review suggestions analyzed by semantic retrieval.
[0015] The beneficial effects of this invention are as follows: This invention presents a knowledge graph-based method for reviewing and verifying power transmission and transformation designs. Through automated parsing of multi-source heterogeneous review documents, knowledge graph modeling, construction of structured review templates, and intelligent review verification, it achieves efficient and systematic management and review of power transmission and transformation design documents. Compared to traditional manual review methods, this method significantly improves the coverage and accuracy of design reviews, reduces subjective biases caused by differences in human experience, and ensures comprehensive checking of key review points such as electrical parameters, substation lines, and environmental constraints. Simultaneously, through intelligent prompting terminology and semantic retrieval-driven review report generation, it automates and enables traceable management of the review process, improving report preparation efficiency and information visualization. Utilizing graph neural networks and Bayesian risk modeling, this method can probabilistically assess potential non-compliant parameters, enabling early risk warnings and anomaly detection, thereby effectively reducing safety hazards caused by design flaws. Furthermore, the system supports incremental learning and dynamic updates to the knowledge graph, adapting to industry standard updates and the expansion of new equipment parameters, enhancing the long-term adaptability and maintainability of the review model. Overall, this invention not only improves the efficiency of power transmission and transformation design review by more than 30%, but also achieves the standardization, structuring, and intelligentization of review results, providing strong technical support for the safety, compliance, and management efficiency of power transmission and transformation engineering design. Attached Figure Description
[0016] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.
[0017] Figure 1 This is a schematic diagram of the structure of a knowledge graph-based power transmission and transformation reduction verification method and system according to the present invention.
[0018] Figure 2 This is a schematic diagram illustrating the execution of the intelligent review and verification model described in the knowledge graph-based power transmission and transformation reduction verification method and system of the present invention. Detailed Implementation
[0019] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.
[0020] Please see Figure 1 A knowledge graph-based method for power transmission and transformation reduction verification: S1: Obtain the review document data of the power transmission and substation, identify the text content of the unstructured data through optical character recognition technology, and extract the text content according to natural language processing technology to construct a knowledge graph subgraph and generate a list of design review standard documents; S2: Based on the design review standard document list, perform text segmentation, label according to semantic roles, construct a knowledge graph review dimension to sort out the review content of design documents in the pre-review and review stages, and obtain a structured review template; S3: Establish an intelligent review verification model based on the structured review template, input the review document data of the power transmission and substation into the model and combine the document integrity with the structured review template, and verify the review points of the structured power transmission and substation project through the loss function and cross-entropy verification. S4: Construct an intelligent review prompt word library, write the review points as prompt words into the intelligent review verification model, and drive the intelligent review prompt word library to generate a power transmission and transformation review report based on the review suggestions according to semantic retrieval analysis.
[0021] Specifically, the method for obtaining the power transmission and transformation review document data is as follows: call the synchronous data of the message queue to obtain online review data, and for the local archived data, trigger the data retrieval instruction by setting an access token to execute a structured query, and attach a timestamp and version identifier to the obtained data records.
[0022] Specifically, the design review standard document list creates a parameterized template based on the review document data, groups the parameters according to the review scenario, establishes a review data-entity-review relationship table, and integrates historical review archives, real-time review reports, and standard specifications based on the review data-entity-review relationship table.
[0023] In this embodiment, the key points for reviewing structured power transmission and transformation projects and the feature vector set of design documents are analyzed to define an intelligent review dataset: , Among them, u i Let x be the design text input sample corresponding to the i-th review task. i Design parameters and equipment specification vector, y i Let z be the set of constraint factors for the review rules. i To score the review deviation, t i For timestamps.
[0024] Let the time series of design parameters in the i-th review be: , Among them, f ij For time point t j The corresponding parameter values or design indicators. Combining historical review records and multi-dimensional knowledge graph constraints, when the deviation between the design parameters and review rules exceeds a threshold, the recursive feedback module initiates an optimization algorithm to correct the weights and matching parameters in the intelligent review verification model. The specific calculation method is as follows: , Where D is the design parameter dependence weight coefficient, Δt is the review time interval, and R ref For reference to the evaluation criteria value, R obs λ represents the actual review deviation value, and λ is the dynamic correction factor.
[0025] Specifically, the method for organizing the review content is as follows: Semantic feature encoding is performed on the review parameters extracted from the design documents, and the text specification content is mapped into feature vector representations that can be used for comparison operations to construct a structured correspondence between design parameters and standard rules. By analyzing the association path between design parameter nodes and standard constraint nodes, calculating their semantic relevance and logical consistency deviation, and forming a potential violation marker list based on the weight level of the deviation, design parameters that do not meet the specification requirements are automatically marked as risk nodes. Based on the review matching results, a risk identification strategy model is established to assess the level of risk events triggered by parameter deviations. The assessment threshold is then calibrated by combining historical review data to form the review judgment boundary.
[0026] Specifically, the method for generating the design review standard document list is as follows: the data is classified according to voltage level and document type based on the historical review status data of the power transmission and transformation station, an initial dataset of review documents is established, the review document information is identified based on the initial dataset, and the scanned document is parsed using optical character recognition technology, entities are mapped as nodes and relationships are used as edges, and a design review standard document list is generated.
[0027] Specifically, the verification method for the review criteria of the structured power transmission and transformation project is as follows: The set of review points formed in the structured review template is used as the verification benchmark. The design parameters, related standard items and their logical relationships corresponding to each review point are used as the verification unit to perform feature comparison and consistency retrieval on the input power transmission and transformation design document data. The unmatched items identified by the retrieval are subjected to semantic consistency analysis and normative compliance detection using a validation model. A graded label is set for the degree of deviation, and the matching rules and deviation thresholds are dynamically adjusted in combination with historical review sample data.
[0028] In this embodiment, the intelligent review and verification model is deployed on the edge computing nodes and cloud microservice architecture of the power transmission and transformation design review system. It achieves real-time verification of key design review points and generation of review reports through design document data collection, feature parsing, and a rule engine. The system first continuously collects design document parameters, equipment specifications, and standard entries through multi-source design data interfaces (such as PDF document parsers and database query interfaces), and processes them in real time using a streaming computing framework, including document parsing, parameter normalization, time series alignment, and outlier removal. Subsequently, based on knowledge graph construction and embedding vectorization methods, the model maps design parameters, review rules, and dependencies into a structured feature set, extracting key review indicators and deviation-sensitive parameters to form an intelligent review and verification feature set.
[0029] The technology stack is as follows: Text processing: Jieba, HanLP, Python (pandas, re); Semantic annotation: AllenNLP, HanLP, BERT-based SRL; Knowledge graph construction: Neo4j, JanusGraph, RDFLib, Protégé; Intelligent review models: PyTorch / TensorFlow, BERT / ERNIE, scikit-learn; Hint database: FAISS, ElasticSearch, Redis, MongoDB; Report generation: Jinja2, python-docx, ReportLab; The specific technical flow is as follows: design parameter extraction → knowledge graph mapping → feature vectorization → rule matching → deviation judgment → risk marking → prompt word generation → structured review report closed loop.
[0030] Based on this, the model uses a rule engine (such as Drools) to match review standard thresholds, design parameter constraints, and priority weights, dynamically generating review anomaly markers and rectification suggestions. These are then written into knowledge graph nodes through field mapping methods, achieving coupled storage with structured review data. When the deviation between the actual design parameters and the model output exceeds a set threshold, the system automatically triggers recursive feedback logic. Based on incremental learning and rolling optimization methods, the model weights, embedded parameters, and review rules are dynamically adjusted, thereby achieving iterative optimization and continuous accuracy of the intelligent review verification model.
[0031] Specifically, the intelligent review prompt word library performs semantic clustering according to the dimensions of electrical parameter verification, structural stability verification, and environmental compliance review. The description of review violation characteristics and corresponding standard words are structurally encapsulated in the form of prompt units to generate a set of prompt word entries that can be retrieved and called. Through semantic retrieval engine and knowledge graph context reasoning mechanism, a mapping relationship is established between the set of prompt word entries and the output results of the review verification model. The corresponding prompt word template is triggered according to the review deviation type and risk level.
[0032] Specifically, the power transmission and transformation review report groups and organizes the verification results according to the review dimension, and records the matching status and triggered prompt words of each review node as review items to form the basic data structure of the review report. Based on the template rules in the prompt word library, the review items are reorganized according to the standardized semantic expression method. The review report content is uniformly encapsulated and rendered by the message generation engine, and timestamps and review process hash signature information are attached.
[0033] Specifically, the method for constructing the intelligent review and verification model is as follows: A training sample set is constructed based on the set of review points extracted from the structured review template and the actual power transmission and transformation design document data. The design parameters, text description information and their corresponding standard item tags in the sample set are feature-encoded, and a training data index structure is established according to the review dimensions, risk categories and standard item mapping relationships. By establishing a validation model framework for graph-structured data, the node information and relationship paths in the knowledge graph are used as model inputs, and the key points of the review are automatically identified through node embedding calculation and semantic association analysis. The review validity model is calibrated by combining actual review feedback data, and the model judgment output is re-labeled and weighted based on the confirmation results in historical review cases. The model review weights and deviation thresholds are iteratively updated based on the incremental learning mechanism.
[0034] Specifically, the optical character recognition technology divides and slices text, tables, and annotation areas based on layout analysis algorithms, and performs character recognition and symbol extraction on each area; after the recognition results are generated, semantic error correction and field matching are performed by combining engineering domain lexicon and standardized encoding, and finally the effectively recognized content is output in a structured manner and automatically mapped to the corresponding entity nodes in the knowledge graph.
[0035] Specifically, the initial dataset of the review documents includes original design documents, parameter specifications, standard clause reference records, and field survey records extracted from the power transmission and transformation engineering design management platform, document management system, and historical review archive. Before access, the data is categorized and identified by timestamp, version number, and source system code, and is divided into three categories according to data type: text files, image files, and structured parameter files, generating a unified basic input data set.
[0036] Specifically, a knowledge graph-based power transmission and transformation reduction verification system is characterized by comprising: Data acquisition module: Acquires data from review documents of power transmission and substations, identifies text content for unstructured data using optical character recognition technology, extracts the text content using natural language processing technology to construct a knowledge graph subgraph, and generates a list of design review standard documents; Content parsing module: Based on the design review standard document list, the text is segmented, labeled according to semantic roles, and a knowledge graph review dimension is constructed to sort out the review content of the design documents in the pre-review and review stages, and a structured review template is obtained; Review and Verification Module: Based on the structured review template, an intelligent review and verification model is established. The review document data of the power transmission and substation is input into the model and combined with the document integrity and the structured review template. The review points of the structured power transmission and substation project are verified by combining the loss function with cross-entropy. Lexicon Building Module: Constructs an intelligent review prompt lexicon, writes the review points as prompt words into the intelligent review verification model, and drives the intelligent review prompt lexicon to generate a power transmission and transformation review report based on the review suggestions analyzed by semantic retrieval.
[0037] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.
Claims
1. A knowledge graph-based method for power transmission and transformation reduction verification, characterized in that, include: S1: Obtain the review document data of the power transmission and substation, identify the text content of the unstructured data through optical character recognition technology, and extract the text content according to natural language processing technology to construct a knowledge graph subgraph and generate a list of design review standard documents; S2: Based on the design review standard document list, perform text segmentation, label according to semantic roles, construct a knowledge graph review dimension to sort out the review content of design documents in the pre-review and review stages, and obtain a structured review template; S3: Establish an intelligent review verification model based on the structured review template, input the review document data of the power transmission and substation into the model and combine the document integrity with the structured review template, and verify the review points of the structured power transmission and substation project through the loss function and cross-entropy verification. S4: Construct an intelligent review prompt word library, write the review points as prompt words into the intelligent review verification model, and drive the intelligent review prompt word library to generate a power transmission and transformation review report based on the review suggestions according to semantic retrieval analysis.
2. The method according to claim 1, characterized in that, The method for obtaining the power transmission and transformation review document data is as follows: call the synchronous data of the message queue to obtain online review data, and for the local archived data, trigger the data retrieval instruction by setting an access token to execute a structured query, and attach a timestamp and version identifier to the obtained data records.
3. The method according to claim 1, characterized in that, The design review standard document list creates a parameterized template based on the review document data, groups the parameters according to the review scenario, establishes a review data-entity-review relationship table, and integrates historical review archives, real-time review reports, and standard specifications based on the review data-entity-review relationship table.
4. The method according to claim 1, characterized in that, The method for organizing the review content is as follows: Semantic feature encoding is performed on the review parameters extracted from the design documents, and the text specification content is mapped into feature vector representations that can be used for comparison operations to construct a structured correspondence between design parameters and standard rules. By analyzing the association path between design parameter nodes and standard constraint nodes, calculating their semantic relevance and logical consistency deviation, and forming a potential violation marker list based on the weight level of the deviation, design parameters that do not meet the specification requirements are automatically marked as risk nodes. Based on the review matching results, a risk identification strategy model is established to assess the level of risk events triggered by parameter deviations. The assessment threshold is then calibrated by combining historical review data to form the review judgment boundary.
5. The method according to claim 2, characterized in that, The method for generating the design review standard document list is as follows: based on the historical review status data of the power transmission and transformation station, the data is classified according to voltage level and document type to establish an initial dataset of review documents. Based on the initial dataset, the review document information is identified, and the scanned document is parsed using optical character recognition technology. Entities are mapped as nodes and relationships are used as edges to generate the design review standard document list.
6. The method according to claim 5, characterized in that, The verification method for the key points of the review of the structured power transmission and transformation project is as follows: The set of review points formed in the structured review template is used as the verification benchmark. The design parameters, related standard items and their logical relationships corresponding to each review point are used as the verification unit to perform feature comparison and consistency retrieval on the input power transmission and transformation design document data. The unmatched items identified by the retrieval are subjected to semantic consistency analysis and normative compliance detection using a validation model. A graded label is set for the degree of deviation, and the matching rules and deviation thresholds are dynamically adjusted in combination with historical review sample data.
7. The method according to claim 4, characterized in that, The intelligent review prompt word library performs semantic clustering according to the dimensions of electrical parameter verification, structural stability verification, and environmental compliance review. It encapsulates the description of review violation characteristics and corresponding standard words in the form of prompt units, generating a set of prompt word entries that can be retrieved and called. Through semantic retrieval engine and knowledge graph context reasoning mechanism, it establishes a mapping relationship between the prompt word entry set and the output results of the review verification model, and triggers the corresponding prompt word template according to the review deviation type and risk level.
8. The method according to claim 2, characterized in that, The power transmission and transformation review report groups and organizes the verification results according to the review dimension, and records the matching status and triggered prompt words of each review node as review items, forming the basic data structure of the review report. Based on the template rules in the prompt word library, the review items are reorganized according to the standardized semantic expression method. The review report content is uniformly encapsulated and rendered by the message generation engine, and timestamps and review process hash signature information are attached.
9. The method according to claim 4, characterized in that, The method for constructing the intelligent review and verification model is as follows: A training sample set is constructed based on the set of review points extracted from the structured review template and the actual power transmission and transformation design document data. The design parameters, text description information and their corresponding standard item tags in the sample set are feature-encoded, and a training data index structure is established according to the review dimensions, risk categories and standard item mapping relationships. By establishing a validation model framework for graph-structured data, the node information and relationship paths in the knowledge graph are used as model inputs, and the key points of the review are automatically identified through node embedding calculation and semantic association analysis. The review validity model is calibrated by combining actual review feedback data, and the model judgment output is re-labeled and weighted based on the confirmation results in historical review cases. The model review weights and deviation thresholds are iteratively updated based on the incremental learning mechanism.
10. The method according to claim 4, characterized in that, The optical character recognition technology is based on a layout analysis algorithm to divide and slice text, tables and annotation areas, and performs character recognition and symbol extraction in each area. After the recognition results are generated, semantic error correction and field matching are performed by combining an engineering domain lexicon and standardized encoding. Finally, the effectively recognized content is output in a structured manner and automatically mapped to the corresponding entity nodes in the knowledge graph.
11. The method according to claim 7, characterized in that, The initial dataset of the review documents includes original design documents, parameter specifications, standard clause reference records, and field survey records extracted from the power transmission and transformation engineering design management platform, document management system, and historical review archive. Before access, the data is categorized and identified by timestamp, version number, and source system code, and is divided into three categories according to data type: text files, image files, and structured parameter files, generating a unified basic input data set.
12. A knowledge graph-based power transmission and transformation reduction verification system, characterized in that, include: Data acquisition module: Acquires data from review documents of power transmission and substations, identifies text content for unstructured data using optical character recognition technology, extracts the text content using natural language processing technology to construct a knowledge graph subgraph, and generates a list of design review standard documents; Content parsing module: Based on the design review standard document list, the text is segmented, labeled according to semantic roles, and a knowledge graph review dimension is constructed to sort out the review content of the design documents in the pre-review and review stages, and a structured review template is obtained; Review and Verification Module: Based on the structured review template, an intelligent review and verification model is established. The review document data of the power transmission and substation is input into the model and combined with the document integrity and the structured review template. The review points of the structured power transmission and substation project are verified by combining the loss function with cross-entropy. Lexicon Building Module: Constructs an intelligent review prompt lexicon, writes the review points as prompt words into the intelligent review verification model, and drives the intelligent review prompt lexicon to generate a power transmission and transformation review report based on the review suggestions analyzed by semantic retrieval.