Construction method of power grid basic data knowledge graph

Abstract

The invention discloses a power grid basic data knowledge graph construction method, and the key points of the technical scheme are that data collection is used for obtaining various types of required power basic data, data cleaning is used for carrying out the preprocessing and knowledge extraction of the collected data, and the knowledge graph is obtained; and the data processing module is used for identifying concepts, entities, attributes and relationships in the power field from the preprocessed data and converting the concepts, entities, attributes and relationships into triads or knowledge representations in other standard formats, so that various power basic data can be cleaned, and then knowledge extraction and fusion can be performed on the preprocessed and cleaned data to obtain the data. According to the method, optimized data can be extracted to the greatest extent, the data capacity is reduced, the data redundancy is reduced, and the consistency and reliability of the data are improved, so that the formed knowledge graph has extremely high practicability and reliability, and various power basic data can be fully and effectively processed; and therefore, the value of various electric power basic data can be completely exerted.

Description

Technical Field

[0001] The present invention relates to the technical field of power data traceability and analysis, and in particular to a method for constructing a knowledge graph of basic power grid data. Background Art

[0002] Data is an important resource in today's Internet era. Through the analysis, interpretation, and induction of various data, sufficient and accurate information can be obtained, thereby providing strong support for social management, business operations, and future trend forecasts in specific fields.

[0003] In the process of providing power services, the power sector involves many fields such as power production, transmission and distribution, power sales, and energy management. Therefore, the basic data of the power grid is large in scale, diverse in types, and highly timely. This places high demands on the organization, management, and cognition of data. However, the data are scattered in different systems and platforms, and there is a phenomenon of "information islands", which leads to high data redundancy, poor consistency, and difficulty in exchange.

[0004] However, most of the existing methods for processing multi-source power data are unable to effectively structure and standardize the data, which makes it difficult to effectively mine and analyze multi-source data after data processing, and the value of the data cannot be fully utilized. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a method for constructing a knowledge graph of basic power grid data, which solves the problem that the existing methods for processing multi-source data are difficult to effectively mine and analyze multi-source data and cannot fully realize the value of the data.

[0006] The above technical objectives of the present invention are achieved through the following technical solutions: A method for constructing a knowledge graph of basic power grid data includes the following steps: S1, data collection: obtain the basic power data required for each type; S2, data cleaning: preprocessing the collected data; S3, knowledge extraction: identifying concepts, entities, attributes and relationships in the power field from the preprocessed data and converting them into triples or other standard formats of knowledge representation; S4, knowledge fusion: integrating knowledge from different sources and types to form a unified and consistent knowledge graph; S5, knowledge storage: store the formed knowledge graph.

[0007] The present invention cleans various types of basic power data and then extracts and fuses knowledge from the pre-processed and cleaned data, thereby maximally refining and optimizing data, reducing data capacity, reducing data redundancy, and improving data consistency and reliability. In this way, the formed knowledge graph has extremely high practicality and reliability, so that various types of basic power data can be fully and effectively processed, thereby fully bringing out the value of various types of basic power data.

[0008] Preferably, the basic power data includes unstructured data and structured data.

[0009] The data collection can obtain relevant data of power grid companies from various sources such as the power Internet of Things, enterprise information systems, public documents, etc., including structured data and unstructured data. The present invention ensures that various relevant basic power data can be obtained more fully, thereby obtaining a sufficiently large knowledge base, ensuring that the subsequently constructed power grid basic data knowledge graph has a wider applicability and improves the reliability of the knowledge graph.

[0010] Preferably, the data cleaning includes cleaning of natural language data and image data.

[0011] The present invention can remove useless or erroneous data, unify data formats and encodings, improve data quality, and also facilitate the subsequent extraction of target knowledge from various types of data, ensuring that subsequent construction work can be successfully completed.

[0012] Preferably, the method for cleaning the natural language data includes one or more of regular expression processing, word segmentation processing, part-of-speech tagging processing, or named entity recognition; the method for cleaning the image data includes one or more of image enhancement, image denoising, or image segmentation.

[0013] The present invention uses natural language processing technologies such as regular expressions, word segmentation, part-of-speech tagging, and named entity recognition to clean and normalize text data for the natural language data cleaning method; and uses image processing technologies such as image enhancement, denoising, and segmentation for the picture data cleaning method.

[0014] Preferably, the knowledge extraction includes knowledge extraction for text data and image data. The knowledge extraction method for text data is to extract concepts, entities, attributes and relationships from text data using rule-based or machine learning-based methods; the knowledge extraction method for image data is to extract entities and attributes using target detection, image classification and image description generation.

[0015] The present invention can extract effective information parts from various types of basic power data, so as to further clean the data and effectively improve the quality of the data.

[0016] Preferably, the knowledge fusion is to fuse knowledge from different sources and types using a vector representation-based or symbolic logic-based method.

[0017] The present invention can eliminate various conflicts, divergences and other problems between the extracted knowledge, and further improve the data quality.

[0018] Preferably, the knowledge storage method is to use a graph database or other storage system to store the knowledge graph in a structure suitable for query and analysis, and provide corresponding interfaces and services.

[0019] The present invention can facilitate quick and convenient querying of the knowledge graph when in use, thereby effectively improving the ease of use and practicality of the knowledge graph.

[0020] Preferably, it also includes: S6, knowledge reasoning: mining implicit knowledge and rules from the knowledge graph.

[0021] The present invention can further improve the mining and analysis of basic power data, thereby extracting deeper information and rules, and maximizing the value of basic power data.

[0022] Preferably, the specific method of knowledge reasoning is one or more of causality, anomaly detection or similarity calculation.

[0023] The present invention can reason about the extracted knowledge at multiple levels and dimensions, thereby obtaining the knowledge and rules implicit in the knowledge graph to the greatest extent.

[0024] Preferably, it also includes S7 knowledge update: timely updating of the concept layer and data layer in the knowledge graph, and the specific method of knowledge update is to update the concept layer and data layer in the knowledge graph using incremental update or full update method.

[0025] The present invention can maintain the freshness and effectiveness of the knowledge graph, thereby extending the service life of the knowledge graph as much as possible, further improving the practicality and economy of the knowledge graph.

[0026] In summary, the present invention mainly has the following beneficial effects: By cleaning various types of basic power data and then extracting and fusing the pre-processed and cleaned data, we can maximize the extraction and optimization of data, reduce data capacity, reduce data redundancy, and improve data consistency and reliability. This will make the formed knowledge graph extremely practical and reliable, so that we can fully and effectively process various types of basic power data, and then fully realize the value of various types of basic power data. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flow chart of the method for constructing the basic data knowledge graph of the power grid of the present invention. Figure 2 It is a flow chart of the method for reasoning and updating the knowledge graph of basic power grid data of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] Embodiment 1: This embodiment proposes a method for constructing a knowledge graph of basic power grid data. Figure 1 , including the following steps: S1, data collection: obtain the basic power data required for each type; The data acquisition can obtain relevant data of power grid companies from multiple sources such as the power Internet of Things, enterprise information systems, and public documents, including structured data and unstructured data. The present invention ensures that various types of relevant power basic data can be obtained more fully, thereby obtaining a sufficiently large knowledge base, ensuring that the knowledge graph of power grid basic data constructed subsequently has a wider applicability, and improving the reliability of the knowledge graph. Data can be obtained from the power business system using crawlers, APIs, databases, etc., or text information can be extracted from unstructured data such as pictures, videos, and voices using technologies such as OCR and ASR; by adopting the above technical solution, it is ensured that various types of relevant power basic data can be obtained more fully, thereby obtaining a sufficiently large knowledge base, ensuring that the knowledge graph of power grid basic data constructed subsequently has a wider applicability, and improving the reliability of the knowledge graph.

[0030] S2, data cleaning: preprocessing the collected data; the data cleaning includes cleaning of image data and natural language data. The present invention can remove useless or erroneous data, unify data formats and encodings, improve data quality, and facilitate the subsequent extraction of target knowledge from various types of data, ensuring that subsequent construction work can be completed smoothly. The present invention uses natural language data cleaning methods such as regular expressions, word segmentation, part-of-speech tagging, named entity recognition and other natural language processing technologies to clean and normalize text data; the image data cleaning method uses image enhancement, denoising, segmentation and other image processing technologies.

[0031] S3, knowledge extraction: Identify and extract concepts, entities, attributes and relationships in the power field from the preprocessed data, and convert them into triples or other standard formats of knowledge representation; S4, knowledge fusion: Integrate the extracted knowledge of different types and sources, solve problems such as entity disambiguation, relationship alignment and knowledge conflict, and form a unified and consistent knowledge graph; the knowledge fusion specifically uses a symbolic logic-based or vector-based representation method to fuse knowledge from different sources and types. Specifically, entity disambiguation and relationship alignment are performed through knowledge graph embedding models such as TransE, and ontology matching and mapping techniques are used for knowledge merging and conflict resolution. The present invention can eliminate various conflicts, disagreements and other problems between the extracted knowledge, and further improve data quality.

[0032] S5, knowledge storage: storing the formed knowledge graph. The knowledge storage method is: using a graph database or other storage system to store the knowledge graph in a structure suitable for analysis and query, while providing corresponding services and interfaces. Specifically, use graph databases such as Neo4j, OrientDB, or standard formats such as RDF, OWL, etc. to store knowledge graphs, and provide query languages ​​such as Cypher, SPARQL, and service interfaces such as RESTful API; the present invention can facilitate quick and convenient querying of the knowledge graph during use, thereby effectively improving the ease of use and practicality of the knowledge graph.

[0033] The present invention cleans various types of basic power data and then extracts and fuses knowledge from the pre-processed and cleaned data, thereby maximally refining and optimizing data, reducing data capacity, reducing data redundancy, and improving data consistency and reliability. In this way, the formed knowledge graph has extremely high practicality and reliability, so that various types of basic power data can be fully and effectively processed, thereby fully bringing out the value of various types of basic power data.

[0034] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0035] Embodiment 2: This embodiment optimizes S3 based on embodiment 1 and proposes a method for constructing a knowledge graph of basic power grid data. Figure 1 , including the following steps: S1, data collection: obtain the basic power data required for each type; The data acquisition can obtain relevant data of power grid companies from multiple sources such as the power Internet of Things, enterprise information systems, and public documents, including structured data and unstructured data. The present invention ensures that various types of relevant power basic data can be obtained more fully, thereby obtaining a sufficiently large knowledge base, ensuring that the knowledge graph of power grid basic data constructed subsequently has a wider applicability, and improving the reliability of the knowledge graph. Data can be obtained from the power business system using crawlers, APIs, databases, etc., or text information can be extracted from unstructured data such as pictures, videos, and voices using technologies such as OCR and ASR; by adopting the above technical solution, it is ensured that various types of relevant power basic data can be obtained more fully, thereby obtaining a sufficiently large knowledge base, ensuring that the knowledge graph of power grid basic data constructed subsequently has a wider applicability, and improving the reliability of the knowledge graph.

[0036] S2, data cleaning: preprocessing the collected data; the data cleaning includes cleaning of image data and natural language data. The present invention can remove useless or erroneous data, unify data formats and encodings, improve data quality, and facilitate the subsequent extraction of target knowledge from various types of data, ensuring that subsequent construction work can be completed smoothly. The present invention uses natural language data cleaning methods such as regular expressions, word segmentation, part-of-speech tagging, named entity recognition and other natural language processing technologies to clean and normalize text data; the image data cleaning method uses image enhancement, denoising, segmentation and other image processing technologies.

[0037] S3, knowledge extraction: Identify and extract concepts, entities, attributes and relationships in the power field from the preprocessed data, and convert them into triples or other standard formats of knowledge representation; The knowledge extraction includes knowledge extraction for image data and text data. The knowledge extraction method for text data is: using machine learning-based or rule-based methods to extract entities, concepts, attributes and relationships from text data; specifically, using dependency syntax analysis, semantic role labeling and other technologies to extract triples, or using pre-trained models such as BERT to perform tasks such as entity linking and relationship classification; the knowledge extraction method for image data is: using image classification, target detection, and image description generation methods to extract entities and attributes. The present invention can extract the effective information part of various types of power basic data, so that the data can be further cleaned, effectively improving the quality of the data.

[0038] S4, knowledge fusion: Integrate the extracted knowledge of different types and sources, solve problems such as entity disambiguation, relationship alignment and knowledge conflict, and form a unified and consistent knowledge graph; the knowledge fusion specifically uses a symbolic logic-based or vector-based representation method to fuse knowledge from different sources and types. Specifically, entity disambiguation and relationship alignment are performed through knowledge graph embedding models such as TransE, and ontology matching and mapping techniques are used for knowledge merging and conflict resolution. The present invention can eliminate various conflicts, disagreements and other problems between the extracted knowledge, and further improve data quality.

[0039] S5, knowledge storage: storing the formed knowledge graph. The knowledge storage method is: using a graph database or other storage system to store the knowledge graph in a structure suitable for analysis and query, while providing corresponding services and interfaces. Specifically, use graph databases such as Neo4j, OrientDB, or standard formats such as RDF, OWL, etc. to store knowledge graphs, and provide query languages ​​such as Cypher, SPARQL, and service interfaces such as RESTful API; the present invention can facilitate quick and convenient querying of the knowledge graph during use, thereby effectively improving the ease of use and practicality of the knowledge graph.

[0040] The present invention cleans various types of basic power data and then extracts and fuses knowledge from the pre-processed and cleaned data, thereby maximally refining and optimizing data, reducing data capacity, reducing data redundancy, and improving data consistency and reliability. In this way, the formed knowledge graph has extremely high practicality and reliability, so that various types of basic power data can be fully and effectively processed, thereby fully bringing out the value of various types of basic power data.

[0041] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0042] Embodiment 3: This embodiment adds a new step after S5 on the basis of Embodiment 2, and proposes a method for constructing a knowledge graph of basic data of power grids. Figure 1 , including the following steps: S1, data collection: obtain the basic power data required for each type; The data acquisition can obtain relevant data of power grid companies from multiple sources such as the power Internet of Things, enterprise information systems, and public documents, including structured data and unstructured data. The present invention ensures that various types of relevant power basic data can be obtained more fully, thereby obtaining a sufficiently large knowledge base, ensuring that the knowledge graph of power grid basic data constructed subsequently has a wider applicability, and improving the reliability of the knowledge graph. Data can be obtained from the power business system using crawlers, APIs, databases, etc., or text information can be extracted from unstructured data such as pictures, videos, and voices using technologies such as OCR and ASR; by adopting the above technical solution, it is ensured that various types of relevant power basic data can be obtained more fully, thereby obtaining a sufficiently large knowledge base, ensuring that the knowledge graph of power grid basic data constructed subsequently has a wider applicability, and improving the reliability of the knowledge graph.

[0043] S2, data cleaning: preprocessing the collected data; the data cleaning includes cleaning of image data and natural language data. The present invention can remove useless or erroneous data, unify data formats and encodings, improve data quality, and facilitate the subsequent extraction of target knowledge from various types of data, ensuring that subsequent construction work can be completed smoothly. The present invention uses natural language data cleaning methods such as regular expressions, word segmentation, part-of-speech tagging, named entity recognition and other natural language processing technologies to clean and normalize text data; the image data cleaning method uses image enhancement, denoising, segmentation and other image processing technologies.

[0044] S3, knowledge extraction: Identify and extract concepts, entities, attributes and relationships in the power field from the preprocessed data, and convert them into triples or other standard formats of knowledge representation; The knowledge extraction includes knowledge extraction for image data and text data. The knowledge extraction method for text data is: using machine learning-based or rule-based methods to extract entities, concepts, attributes and relationships from text data; specifically, using dependency syntax analysis, semantic role labeling and other technologies to extract triples, or using pre-trained models such as BERT to perform tasks such as entity linking and relationship classification; the knowledge extraction method for image data is: using image classification, target detection, and image description generation methods to extract entities and attributes. The present invention can extract the effective information part of various types of power basic data, so that the data can be further cleaned, effectively improving the quality of the data.

[0045] S4, knowledge fusion: Integrate the extracted knowledge of different types and sources, solve problems such as entity disambiguation, relationship alignment and knowledge conflict, and form a unified and consistent knowledge graph; the knowledge fusion specifically uses a symbolic logic-based or vector-based representation method to fuse knowledge from different sources and types. Specifically, entity disambiguation and relationship alignment are performed through knowledge graph embedding models such as TransE, and ontology matching and mapping techniques are used for knowledge merging and conflict resolution. The present invention can eliminate various conflicts, disagreements and other problems between the extracted knowledge, and further improve data quality.

[0046] S5, knowledge storage: storing the formed knowledge graph. The knowledge storage method is: using a graph database or other storage system to store the knowledge graph in a structure suitable for analysis and query, while providing corresponding services and interfaces. Specifically, use graph databases such as Neo4j, OrientDB, or standard formats such as RDF, OWL, etc. to store knowledge graphs, and provide query languages ​​such as Cypher, SPARQL, and service interfaces such as RESTful API; the present invention can facilitate quick and convenient querying of the knowledge graph during use, thereby effectively improving the ease of use and practicality of the knowledge graph.

[0047] refer to Figure 2 ,S6,Knowledge reasoning, is used to mine implicit rules and knowledge from the knowledge graph.

[0048] The present invention can further improve the mining and analysis of basic power data, thereby extracting deeper information and rules, and maximizing the value of basic power data.

[0049] The specific methods of knowledge reasoning include causality, anomaly detection, similarity calculation, etc. Specifically, such as using Prolog, SWRL and other rule languages ​​for logical reasoning, or using GCN, GAT and other graph neural network models for graph calculation; the present invention can reason on the extracted knowledge at multiple levels and dimensions, so as to obtain the knowledge and rules implicit in the knowledge graph to the greatest extent.

[0050] The present invention cleans various types of basic power data and then extracts and fuses knowledge from the pre-processed and cleaned data, thereby maximally refining and optimizing data, reducing data capacity, reducing data redundancy, and improving data consistency and reliability. In this way, the formed knowledge graph has extremely high practicality and reliability, so that various types of basic power data can be fully and effectively processed, thereby fully bringing out the value of various types of basic power data.

[0051] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0052] Embodiment 4: This embodiment adds a new step after S6 on the basis of embodiment 3, and proposes a method for constructing a knowledge graph of basic data of power grid. Figure 1 , including the following steps: S1, data collection: obtain the basic power data required for each type; The data acquisition can obtain relevant data of power grid companies from multiple sources such as the power Internet of Things, enterprise information systems, and public documents, including structured data and unstructured data. The present invention ensures that various types of relevant power basic data can be obtained more fully, thereby obtaining a sufficiently large knowledge base, ensuring that the knowledge graph of power grid basic data constructed subsequently has a wider applicability, and improving the reliability of the knowledge graph. Data can be obtained from the power business system using crawlers, APIs, databases, etc., or text information can be extracted from unstructured data such as pictures, videos, and voices using technologies such as OCR and ASR; by adopting the above technical solution, it is ensured that various types of relevant power basic data can be obtained more fully, thereby obtaining a sufficiently large knowledge base, ensuring that the knowledge graph of power grid basic data constructed subsequently has a wider applicability, and improving the reliability of the knowledge graph.

[0053] S2, data cleaning: preprocessing the collected data; the data cleaning includes cleaning of image data and natural language data. The present invention can remove useless or erroneous data, unify data formats and encodings, improve data quality, and facilitate the subsequent extraction of target knowledge from various types of data, ensuring that subsequent construction work can be completed smoothly. The present invention uses natural language data cleaning methods such as regular expressions, word segmentation, part-of-speech tagging, named entity recognition and other natural language processing technologies to clean and normalize text data; the image data cleaning method uses image enhancement, denoising, segmentation and other image processing technologies.

[0054] S3, knowledge extraction: Identify and extract concepts, entities, attributes and relationships in the power field from the preprocessed data, and convert them into triples or other standard formats of knowledge representation; The knowledge extraction includes knowledge extraction for image data and text data. The knowledge extraction method for text data is: using machine learning-based or rule-based methods to extract entities, concepts, attributes and relationships from text data; specifically, using dependency syntax analysis, semantic role labeling and other technologies to extract triples, or using pre-trained models such as BERT to perform tasks such as entity linking and relationship classification; the knowledge extraction method for image data is: using image classification, target detection, and image description generation methods to extract entities and attributes. The present invention can extract the effective information part of various types of power basic data, so that the data can be further cleaned, effectively improving the quality of the data.

[0055] S4, knowledge fusion: Integrate the extracted knowledge of different types and sources, solve problems such as entity disambiguation, relationship alignment and knowledge conflict, and form a unified and consistent knowledge graph; the knowledge fusion specifically uses a symbolic logic-based or vector-based representation method to fuse knowledge from different sources and types. Specifically, entity disambiguation and relationship alignment are performed through knowledge graph embedding models such as TransE, and ontology matching and mapping techniques are used for knowledge merging and conflict resolution. The present invention can eliminate various conflicts, disagreements and other problems between the extracted knowledge, and further improve data quality.

[0056] S5, knowledge storage: storing the formed knowledge graph. The knowledge storage method is: using a graph database or other storage system to store the knowledge graph in a structure suitable for analysis and query, while providing corresponding services and interfaces. Specifically, use graph databases such as Neo4j, OrientDB, or standard formats such as RDF, OWL, etc. to store knowledge graphs, and provide query languages ​​such as Cypher, SPARQL, and service interfaces such as RESTful API; the present invention can facilitate quick and convenient querying of the knowledge graph during use, thereby effectively improving the ease of use and practicality of the knowledge graph.

[0057] refer to Figure 2 ,S6,Knowledge reasoning, is used to mine implicit rules and knowledge from the knowledge graph.

[0058] The present invention can further improve the mining and analysis of basic power data, thereby extracting deeper information and rules, and maximizing the value of basic power data.

[0059] The specific methods of knowledge reasoning include causality, anomaly detection, similarity calculation, etc. Specifically, such as using Prolog, SWRL and other rule languages ​​for logical reasoning, or using GCN, GAT and other graph neural network models for graph calculation; the present invention can reason on the extracted knowledge at multiple levels and dimensions, so as to obtain the knowledge and rules implicit in the knowledge graph to the greatest extent.

[0060] S7, knowledge update, is used to timely update the data layer and concept layer in the knowledge graph. The specific method of knowledge update is: use the full update or incremental update method to update the data layer and concept layer in the knowledge graph. For example, use timestamps, version control and other technologies to record data changes, and choose to add, delete or modify corresponding knowledge according to the changes; the present invention can maintain the freshness and effectiveness of the knowledge graph, thereby extending the service life of the knowledge graph as much as possible, further improving the practicality and economy of the knowledge graph.

[0061] The present invention cleans various types of basic power data and then extracts and fuses knowledge from the pre-processed and cleaned data, thereby maximally refining and optimizing data, reducing data capacity, reducing data redundancy, and improving data consistency and reliability. In this way, the formed knowledge graph has extremely high practicality and reliability, so that various types of basic power data can be fully and effectively processed, thereby fully bringing out the value of various types of basic power data.

[0062] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a knowledge graph of basic power grid data, characterized in that: The steps include: S1, data collection: obtain the basic power data required for each type; S2, data cleaning: preprocessing the collected data; S3, knowledge extraction: identifying concepts, entities, attributes and relationships in the power field from the preprocessed data and converting them into triples or other standard formats of knowledge representation; S4, knowledge fusion: integrating knowledge from different sources and types to form a unified knowledge graph; S5, knowledge storage: store the formed knowledge graph.

2. The method for constructing a knowledge graph of basic power grid data according to claim 1, characterized in that: The basic power data includes unstructured data and structured data.

3. The method for constructing a knowledge graph of basic power grid data according to claim 1, characterized in that: The data cleaning includes cleaning of natural language data and image data.

4. The method for constructing a knowledge graph of basic power grid data according to claim 3, characterized in that: The method for cleaning the natural language data includes one or more of regular expression processing, word segmentation processing, part-of-speech tagging processing, or named entity recognition; the method for cleaning the image data includes one or more of image enhancement, image denoising, or image segmentation.

5. The method for constructing a knowledge graph of basic power grid data according to claim 4, characterized in that: The knowledge extraction includes knowledge extraction of text data and image data. The knowledge extraction method for text data is to use rule-based or machine learning-based methods to extract concepts, entities, attributes and relationships from text data; the knowledge extraction method for image data is to use target detection, image classification and image description generation to extract entities and attributes.

6. The method for constructing a knowledge graph of basic power grid data according to claim 1, characterized in that: The knowledge fusion specifically involves fusing knowledge from different sources and types using a vector representation-based or symbolic logic-based method.

7. The method for constructing a knowledge graph of basic power grid data according to claim 1, characterized in that: The knowledge storage method is to use a graph database or other storage system to store the knowledge graph in a structure suitable for query and analysis, and provide corresponding interfaces and services.

8. A method for constructing a knowledge graph of basic power grid data according to any one of claims 1 to 7, characterized in that: Also includes: S6, knowledge reasoning: mining implicit knowledge and rules from knowledge graphs.

9. The method for constructing a knowledge graph of basic power grid data according to claim 8, characterized in that: The specific method of the knowledge reasoning is one or more of causality, anomaly detection or similarity calculation.

10. The method for constructing a knowledge graph of basic power grid data according to claim 8, characterized in that: Also includes: S7, knowledge update: timely update the concept layer and data layer in the knowledge graph. The specific method of knowledge update is to update the concept layer and data layer in the knowledge graph using incremental update or full update method.

Images