Data recommendation method and system based on electric power engineering project knowledge graph

Abstract

The invention discloses a data recommendation method and system based on an electric power engineering project knowledge graph, and the method comprises the steps: carrying out the structural processing of isomerized data in an electric power engineering project, obtaining the structural data and entity information, and constructing a structural database; the method comprises the following steps: capturing key elements of structured data, and combining mass text data of an electric power engineering project to construct an electric power engineering project knowledge graph with the structured data and associated information thereof; compared with single data storage and query, the electric power engineering project knowledge graph constructed by the method can reveal the deep semantic relationship and hidden association between the data. The understandability of the data is enhanced, the integrity and accuracy of the knowledge base are further improved through the reasoning algorithm and the large language model, and dynamic and intelligent knowledge management and expansion are achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of natural language processing, and in particular relates to a data recommendation method and system based on a knowledge graph of an electric power engineering project. Background Art

[0002] With the deepening of the digital transformation of the power industry, power engineering project management is facing unprecedented challenges. The traditional project management model mainly relies on manual collation and analysis of data, which is inefficient and difficult to effectively mine the deep laws behind the data. The data of power engineering projects usually covers multiple levels and dimensions, such as equipment operation status, personnel allocation, material supply, etc. Faced with these multi-source heterogeneous data, hidden correlations and potential patterns are often hidden in the complex patterns and trends of the data. It is difficult to fully and accurately identify and utilize this information by manual analysis alone. At present, the demand for power grid engineering data management in terms of effectiveness, intelligent analysis and real-time retrieval is becoming increasingly urgent. Most of the existing intelligent data analysis systems are based on traditional database architecture and analyze data from a mathematical perspective. For example, the surface relationship between data is revealed by calculating statistical indicators such as mean, median, standard deviation, and visualized through front-end display tools. Although this type of statistical analysis method can provide certain information, its limitations are also very obvious. Especially in a complex environment such as the power system, traditional data analysis methods are difficult to effectively mine the deep connections behind the data, resulting in a lack of depth and breadth in the analysis results.

[0003] Power grid data is highly complex and multi-dimensionally correlated. Traditional statistical analysis methods often focus too much on the surface patterns of data and have difficulty in deeply understanding the potential connections and structural patterns between data. For example, in power engineering projects, the connections between projects are not only reflected in quantitative indicators such as investment and progress, but also involve more complex technical relationships, equipment dependencies, geographical distribution, and historical maintenance records. These multi-dimensional information are difficult to effectively integrate and analyze in traditional database analysis methods, resulting in one-sided data and failure to fully reveal the deep connections between projects.

[0004] In addition, with the rapid expansion of power grid data and the increasing diversity of data sources (such as IoT devices, sensor data, historical operation and maintenance records, etc.), how to quickly extract valuable information from massive heterogeneous data has become an important challenge for intelligent power grid data analysis. Summary of the invention

[0005] In order to solve the problems existing in the prior art, the present invention proposes a data recommendation method and system based on a knowledge graph of electric power engineering projects.

[0006] The technical solution of the present invention is as follows:

[0007] A data recommendation method based on a knowledge graph of electric power engineering projects, comprising:

[0008] Structural processing of heterogeneous data in power engineering projects, acquisition of structured data and its entity information, and construction of a structured database;

[0009] By capturing key elements of structured data and combining it with massive text data of power engineering projects, a knowledge graph of power engineering projects with structured data and its related information is constructed;

[0010] Identify the indicator intent of the user input data, query the structured data and its entity information in the structured database with the identified indicator intent, and query the associated information in the power engineering project knowledge graph;

[0011] The structured data and entity information obtained by querying the structured database and the associated information obtained by querying the knowledge graph of the power engineering project are input into the large language model, and the output value of the large language model is used as the data recommendation result.

[0012] Furthermore, the specific method of structuring the heterogeneous data in the power engineering project, obtaining the structured data and its entity information, and building a structured database includes:

[0013] Divide the heterogeneous data in the power engineering project into text data and image data;

[0014] Extracting JSON data from the text data using information extraction, dependency analysis, and relationship extraction methods in turn;

[0015] Using a multimodal large language model to understand the image data, obtaining text data corresponding to the image data, and extracting JSON data from the text data corresponding to the image data using information extraction, dependency analysis, and relationship extraction methods in sequence;

[0016] Integrate the JSON data obtained from the text data and the JSON data obtained from the image data, and use the integrated JSON data in the form of key-value pairs as structured data and its entity information;

[0017] Eliminate noise from structured data and its entity information;

[0018] A structured database is constructed using the noise-eliminated structured data and its entity information.

[0019] Further, the noise elimination includes normalizing the structured data and its entity information by comparing semantic similarity using a large language model, calculating similarity using cosine similarity, and merging the structured data and its entity information whose similarity is greater than or equal to a set threshold;

[0020] The normalized structured data and its entity information are processed based on pre-set rules in relation to time data format, outliers, detection range and duplicate records.

[0021] Furthermore, the specific method of capturing key elements of structured data and combining massive text data of power engineering projects to construct a knowledge graph of power engineering projects with structured data and its associated information includes:

[0022] Obtain massive text data of power engineering projects;

[0023] Based on the structured data and its entity information, a large language model is used to capture the corresponding key elements in the massive text data of the power engineering project, and a plurality of triples associated with the structured data and its entity information are formed;

[0024] Based on the several triples, a knowledge graph of electric power engineering projects with structured data and related information is constructed.

[0025] Furthermore, the massive text data of the power engineering project includes project specifications, budget estimates and technical reports.

[0026] Furthermore, the specific method of identifying the indicator intent of the user input data, querying the structured data and its entity information in the structured database with the identified indicator intent, and querying the associated information in the power engineering project knowledge graph includes:

[0027] The indicator intent is identified for the user input data, and the structured data and its entity information are queried in the structured database based on the indicator intent, and the related information is queried in the power engineering project knowledge graph using CQL statements with the queried structured data and its entity information.

[0028] Furthermore, the specific method of inputting the structured data and entity information thereof queried in the structured database and the associated information queried in the power engineering project knowledge graph into the large language model and using the output value of the large language model as the data recommendation result includes:

[0029] Classify the indicator intent into numerical data and non-numerical data;

[0030] For numerical data, the associated information obtained by querying in the power engineering project knowledge graph is subjected to data statistical analysis, and the data statistical analysis results are combined as associated information with the structured data and entity information obtained by querying in the structured database and input into the large language model, and the output value of the large language model is used as the data recommendation result; the data statistical analysis includes calculations using five mathematical statistical methods: mean, median, coefficient of dispersion, standard deviation, and discrete interval.

[0031] For non-numeric data, the structured data and its entity information obtained by querying the structured database and the associated information obtained by querying the power engineering project knowledge graph are input into the large language model, and the output value of the large language model is used as the data recommendation result.

[0032] A data recommendation system based on a knowledge graph of electric power engineering projects, comprising a structured database construction module, a knowledge graph construction module, an association query module and a recommendation module;

[0033] The structured database construction module is used to perform structured processing on heterogeneous data in the power engineering project, obtain structured data and its entity information, and construct a structured database;

[0034] The knowledge graph construction module is used to capture key elements of structured data and combine massive text data of power engineering projects to construct a knowledge graph of power engineering projects with structured data and its associated information;

[0035] The associated query module is used to identify the indicator intent of the user input data, query the structured data and its entity information in the structured database with the identified indicator intent, and query the associated information in the power engineering project knowledge graph;

[0036] The recommendation module is used to input the structured data and its entity information obtained by querying the structured database and the associated information obtained by querying the knowledge graph of the power engineering project into the large language model, and use the output value of the large language model as the data recommendation result.

[0037] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute any of the methods described above.

[0038] A computer-readable storage medium stores a computer program, wherein the computer program implements the steps of any of the above methods when executed by a processor.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention proposes a data recommendation method and system based on a knowledge graph of electric power engineering projects, which structures the heterogeneous data in the electric power engineering projects, obtains structured data and its entity information, and constructs a structured database; by capturing key elements of the structured data and combining the massive text data of the electric power engineering projects, a knowledge graph of the electric power engineering projects with structured data and its associated information is constructed. Compared with single data storage and query, the knowledge graph of the electric power engineering projects constructed by the present invention can reveal the deep semantic relationships and hidden associations between data. This not only enhances the comprehensibility of the data, but also further improves the integrity and accuracy of the knowledge base through reasoning algorithms and large language models, and realizes dynamic and intelligent knowledge management and expansion.

[0041] The method of the present invention identifies the indicator intent of the user input data, queries the structured data and its entity information in the structured database with the identified indicator intent, and queries the associated information in the knowledge graph of the electric power engineering project; the structured data and its entity information obtained by querying in the structured database and the associated information obtained by querying in the knowledge graph of the electric power engineering project are input into the large language model, and the output value of the large language model is used as the data recommendation result. Through natural language processing technology, the method of the present invention can understand complex natural language queries, quickly locate entities and relationships in the text, and provide semantic query results. At the same time, the system combines the deep semantic relationships in the knowledge graph to realize the intelligent recommendation of query indicators, greatly improving the accuracy and efficiency of the query. Compared with the existing retrieval system based on keyword matching, the present invention can perform semantic reasoning and associated queries, and provide more insightful retrieval results.

[0042] In the process of structuring heterogeneous data, the method of the present invention faces diversified and heterogeneous data sources (such as text, drawings, project reports, etc.) in the field of power engineering, and structures them into standardized data formats (such as JSON). Compared with traditional methods, the present invention can automatically extract key information from multimodal data by integrating information extraction, relationship extraction, and topic modeling framework technology, improve the efficiency and accuracy of data processing, reduce manual intervention, and significantly enhance the availability of data. At the same time, large language models (such as qwen2.5-72b, GPT-4o, etc.) are used to handle inconsistencies, missing or errors in data and improve data quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a data recommendation method based on a knowledge graph of electric power engineering projects in an embodiment;

[0044] Figure 2 A schematic diagram of constructing a structured database in an embodiment;

[0045] Figure 3 This is a schematic diagram of the knowledge representation architecture in the power field in the embodiment;

[0046] Figure 4 This is a schematic diagram of the fusion recommendation algorithm framework in the embodiment;

[0047] Figure 5 This is an example diagram of the fusion recommendation algorithm process in the embodiment. DETAILED DESCRIPTION

[0048] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent modifications of the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0049] Embodiment 1:

[0050] The present invention provides a data recommendation method based on the knowledge graph of power engineering projects, such as Figure 1 As shown, including:

[0051] S1. Structural processing of heterogeneous data in power engineering projects, acquisition of structured data and its entity information, and construction of a structured database;

[0052] S2. By capturing key elements of structured data and combining it with massive text data of power engineering projects, a knowledge graph of power engineering projects with structured data and its related information is constructed;

[0053] S3. Identify the indicator intent of the user input data, query the structured data and its entity information in the structured database with the identified indicator intent, and query the related information in the knowledge graph of the power engineering project;

[0054] S4. Input the structured data and its entity information obtained from the structured database and the associated information obtained from the knowledge graph of the power engineering project into the large language model, and use the output value of the large language model as the data recommendation result.

[0055] Embodiment 2:

[0056] This example is further designed on the basis of the first embodiment in that in this example, the heterogeneous data in the power engineering project is structured, the structured data and its entity information are obtained, and the specific method of constructing a structured database includes:

[0057] In the field of power engineering, intelligent recommendation of diversified data is the key to improving data utilization efficiency and decision-making level. The entire data processing process can be divided into two main parts. The specific steps and methods are as follows: Figure 2 The specific steps are as follows:

[0058] S101, dividing the heterogeneous data in the power engineering project into text data and image data;

[0059] S102, extract JSON data from the text data using information extraction (BERT+GRU), dependency analysis (PCFG), and relationship extraction (large language model qwen2.5-72b);

[0060] S103, using a multimodal large language model to understand the image data, obtaining the text data corresponding to the image data, and extracting JSON data from the text data corresponding to the image data using information extraction (BERT+GRU), dependency analysis (PCFG), and relationship extraction (large language model qwen2.5-72b) methods in sequence;

[0061] S104, integrating the JSON data obtained from the text data and the JSON data obtained from the image data, and using the integrated JSON data in the form of key-value pairs as structured data and its entity information;

[0062] S105, performing noise elimination on the structured data and its entity information;

[0063] S106: construct a structured database using the structured data after noise elimination and its entity information.

[0064] Embodiment three:

[0065] This example is further designed on the basis of the second embodiment in that the noise elimination in this example includes using a large language model (qwen2.5-72b) to compare the semantic similarity of the structured data and its entity information (iterative cleaning of existing indicators and inserted indicators) for normalization, using cosine similarity to calculate the similarity, merging the structured data and its entity information whose similarity is greater than or equal to a set threshold, and the threshold is generally set to 0.9;

[0066] The normalized structured data and its entity information are processed based on pre-set rules in relation to time data format, outliers, detection range and duplicate records.

[0067] The following is an example to illustrate the noise elimination:

[0068] The JSON data for this project is:

[0069]

[0070]

[0071]

[0072] Polling is performed to compare the similarity of each key-value pair in JSON. For example, when inserting "building structure type": "reinforced concrete", the large language model is used to calculate the text encoding of "building structure type": "reinforced concrete" and "building structure form": "reinforced concrete frame structure" to obtain vectors x1 and x2 and perform similarity comparison. The formula is as follows. The calculated cosine similarity is 0.93, which exceeds the threshold and is not inserted.

[0073]

[0074] At the same time, the following rules are used for noise elimination:

[0075] 1. Ensure that the time data format is correct. If the time format is xxxx-xx-xx, use regular expressions to standardize it if it is not in the standard format.

[0076] 2. Range detection and outlier processing: manually define a reasonable range for each field and remove unreasonable ones;

[0077] 3. Delete redundant data information and spaces;

[0078] 4. Fixed fields have fixed types. For example, a single project number must be a number. For such key fields, if they are not standardized, this field will be set as incorrect and await manual screening.

[0079] 5.…

[0080] Get the final JSON

[0081]

[0082] Embodiment 4:

[0083] This example is further designed on the basis of the third embodiment. In this example, by capturing key elements of structured data and combining massive text data of power engineering projects, a specific method for constructing a knowledge graph of power engineering projects with structured data and its associated information includes:

[0084] First, we integrate various data sources of power engineering projects to obtain massive text data of power engineering projects; by comprehensively collecting these heterogeneous data, we ensure the breadth and depth of the data;

[0085] Next, the collected unstructured massive text data is intelligently processed, especially the technical documents and reports related to the project. The large language model qwen2.5-72b is used to build a framework structure with the project as the core. Based on the structured data and its entity information, the large language model is used to capture the corresponding key elements (such as equipment, technical standards, budget, progress, etc.) in the massive text data of the power engineering project, forming several triples associated with the structured data and its entity information;

[0086] Based on several triples, a knowledge graph of power engineering projects with structured data and associated information is constructed. The associated information is triples. The constructed knowledge graph will cover the core concepts in power engineering (such as equipment, materials, faults, etc.) and their attributes and relationships. The knowledge representation architecture designed in this example is as follows: Figure 3 As shown in the figure, the architecture is divided into the open domain and the power domain. The open domain is divided into the concept layer and the instance layer. The general concepts in the concept layer are concretized into specific instances. These instances are specific data points, representing the specific occurrence of entities in the concept layer. In the instance layer, the general concepts in the concept layer are concretized into specific instances. These instances are specific data points, representing the specific occurrence of entities in the concept layer. The power domain is divided into the concept layer, the business layer, and the instance layer. The business layer contains business concepts or entities unique to the power domain, such as power stations, transmission lines, substations, etc. It organically connects modular knowledge units with operational businesses through the business layer to ensure the applicability of the model. With the deepening and expansion of the power domain, the knowledge graph structure will facilitate the migration and update of the ontology level and timely reflect the development and changes in the field.

[0087] Embodiment five:

[0088] This example is further designed on the basis of the fourth embodiment in that the massive text data of the power engineering project in this example includes project specifications, budget estimates and technical reports.

[0089] Embodiment six:

[0090] This example is further designed on the basis of the fifth embodiment in that, Figure 4 As shown, in this example, the user input data is used to identify the indicator intent using natural language processing technology, and the identified indicator intent is used to query the structured data and its entity information in the structured database using information retrieval technology, and the specific method for querying related information in the power engineering project knowledge graph includes:

[0091] The indicator intent is identified for the user input data, and the structured data and its entity information are queried in the structured database based on the indicator intent. The CQL statement is used to query the related information in the power engineering project knowledge graph with the queried structured data and its entity information.

[0092] Embodiment seven:

[0093] This example is further designed on the basis of the sixth embodiment in that in this example, the structured data and its entity information obtained by querying in the structured database and the associated information obtained by querying in the knowledge graph of the power engineering project are input into the large language model, and the specific method of using the output value of the large language model as the data recommendation result includes:

[0094] Classify the indicator intent into numerical data and non-numerical data;

[0095] For numerical data, the related information obtained from the search in the knowledge graph of the power engineering project is subjected to data statistical analysis, and the results of the data statistical analysis are used as related information combined with the structured data and its entity information obtained from the search in the structured database and input into the large language model, and the output value of the large language model is used as the data recommendation result; the data statistical analysis includes calculations using five mathematical statistical methods: mean, median, coefficient of dispersion, standard deviation, and discrete interval.

[0096] For non-numeric data, the structured data and its entity information obtained from the structured database and the associated information obtained from the knowledge graph of the power engineering project are input into the large language model, and the output value of the large language model is used as the data recommendation result.

[0097] The following is an example to further illustrate this design. Figure 5 As shown, in this example, the indicator is intended to be numerical, and the user input data is "the system's design elevation for the site of the new substation project from 2021 to 2024"

[0098] (1) First, query the site design elevation from 2021 to 2024 and under the conditions of new substation construction in the Mysql database. The SQL statement is as follows:

[0099] Select value from tb_project where plan_review_time between'2021-01-01'and'2024-12-31'and index like "% Site design elevation %" and project_type = "New substation project";

[0100] The specific numerical data obtained are [35.6, 3.8, 2.78, 17.67, 4.87~5.30, 2.25, 7.73~8.61, 13.10, 23.11, 8.28…].

[0101] (2) The technical indicator field of the natural elevation of the site is queried in the knowledge graph to obtain other indicators or fields that may affect the value of the natural elevation of the site. If the content of the field is linked in the knowledge graph, it is extracted from the knowledge graph. If not, it is searched from the MySQL database.

[0102] (3) The values ​​of the above two steps are submitted to the large language model for integration and the recommended value of the natural elevation of the site is obtained. Here, the qwen2.5 open source large language model is used to balance performance and consumption.

[0103] Embodiment eight:

[0104] A data recommendation system based on a knowledge graph of a power engineering project of the present invention comprises a structured database construction module, a knowledge graph construction module, an association query module and a recommendation module;

[0105] The structured database construction module is used to structure the heterogeneous data in the power engineering project, obtain the structured data and its entity information, and build a structured database;

[0106] The knowledge graph construction module is used to capture key elements of structured data and combine it with massive text data of power engineering projects to build a knowledge graph of power engineering projects with structured data and its related information;

[0107] The associated query module is used to identify the indicator intent of the user input data, query the structured data and its entity information in the structured database based on the identified indicator intent, and query the associated information in the knowledge graph of the power engineering project;

[0108] The recommendation module is used to input the structured data and its entity information obtained by querying in the structured database and the associated information obtained by querying in the knowledge graph of the power engineering project into the large language model, and use the output value of the large language model as the data recommendation result.

[0109] This example builds an intelligent data analysis upper-level application based on the data recommendation system of the power engineering project knowledge graph of the present invention. The application covers five core functions: self-correlation of indicators, project recommendation of queries, discovery and identification of hidden relationships, visualization platform, and review experience library based on the power grid engineering knowledge graph. By integrating the structured data relationships built in the database and the deep semantic structure in the knowledge graph, more in-depth data display, recommendation and intelligent analysis are achieved. Through this fusion of structured data and deep semantics, the depth and breadth of data analysis are improved, providing strong technical support for efficient management, decision support and intelligent recommendation of power engineering projects.

[0110] The function is implemented as follows:

[0111] 1. Build a self-correlation model for data indicators: By building a self-correlation model based on knowledge graphs, using natural language processing (NLP) technology to semantically annotate data indicators, accurately model the relationship between different indicators, and solve the problem of inaccurate analysis caused by data ambiguity in traditional data association analysis. And apply intelligent algorithms and visualization tools to fully understand and display the relationship between data. Timely updates and dynamic adjustments ensure that the system continues to adapt to actual needs.

[0112] 2. Query indicator recommendation: Using knowledge graph technology, when performing data analysis, the present invention first queries the specific value of a certain indicator in the database and identifies other indicators associated with the indicator. These related information are presented through the knowledge graph, providing rich context. Subsequently, the large language model is used to jointly analyze and reason these data, thereby generating accurate data indicator recommendations in specific situations.

[0113] 3. Discover and identify hidden relationships: Use cross-level, multi-dimensional data analysis technology, combined with time series and cross-domain data, to identify hidden relationships between data elements. Through deep learning models and association analysis technology, the system can cross multiple data dimensions to discover potential relationships in patterns and trends, especially in the recognition of abnormal patterns in power engineering data.

[0114] 4. Build a visualization platform: Build a visualization platform based on knowledge graphs, use graphics processing algorithms and data visualization technology to transform complex power data into intuitive charts and graphic displays, and help users understand data structures and relationships more clearly.

[0115] 5. Establish an evaluation experience library based on the knowledge graph of power grid engineering: Through the application of the knowledge graph technology of power grid engineering, it is possible to conduct comparative analysis with the relevant knowledge in the knowledge graph, quickly identify abnormal situations in the use of funds, and use the graph knowledge to analyze the causes of abnormalities. At the same time, the historical engineering case knowledge in the knowledge graph also provides a useful reference for the rationality review of funds.

[0116] Embodiment nine:

[0117] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute any of the methods described above.

[0118] A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of any of the above methods.

[0119] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a technician familiar with the technical field within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A data recommendation method based on a knowledge graph of electric power engineering projects, characterized in that: include: Structural processing of heterogeneous data in power engineering projects, acquisition of structured data and its entity information, and construction of a structured database; By capturing key elements of structured data and combining it with massive text data of power engineering projects, a knowledge graph of power engineering projects with structured data and its related information is constructed; Identify the indicator intent of the user input data, query the structured data and its entity information in the structured database with the identified indicator intent, and query the associated information in the power engineering project knowledge graph; The structured data and entity information obtained by querying the structured database and the associated information obtained by querying the knowledge graph of the power engineering project are input into the large language model, and the output value of the large language model is used as the data recommendation result.

2. The data recommendation method based on the knowledge graph of electric power engineering projects according to claim 1 is characterized in that: The specific method of structuring the heterogeneous data in the power engineering project, obtaining the structured data and its entity information, and building a structured database includes: Divide the heterogeneous data in the power engineering project into text data and image data; Extracting JSON data from the text data using information extraction, dependency analysis, and relationship extraction methods in turn; Using a multimodal large language model to understand the image data, obtaining text data corresponding to the image data, and extracting JSON data from the text data corresponding to the image data using information extraction, dependency analysis, and relationship extraction methods in sequence; Integrate the JSON data obtained from the text data and the JSON data obtained from the image data, and use the integrated JSON data in the form of key-value pairs as structured data and its entity information; Eliminate noise from structured data and its entity information; A structured database is constructed using the noise-eliminated structured data and its entity information.

3. The data recommendation method based on the knowledge graph of electric power engineering projects according to claim 2 is characterized in that: The noise elimination includes using a large language model to compare the semantic similarity of the structured data and its entity information for normalization, using cosine similarity to calculate the similarity, and merging the structured data and its entity information whose similarity is greater than or equal to a set threshold; The normalized structured data and its entity information are processed based on pre-set rules in relation to time data format, outliers, detection range and duplicate records.

4. The data recommendation method based on the knowledge graph of electric power engineering projects according to claim 3 is characterized in that: The specific method of capturing key elements of structured data and combining massive text data of power engineering projects to construct a knowledge graph of power engineering projects with structured data and related information includes: Obtain massive text data of power engineering projects; Based on the structured data and its entity information, a large language model is used to capture the corresponding key elements in the massive text data of the power engineering project, and a plurality of triples associated with the structured data and its entity information are formed; Based on the several triples, a knowledge graph of electric power engineering projects with structured data and related information is constructed.

5. The data recommendation method based on the knowledge graph of electric power engineering projects according to claim 4 is characterized in that: The massive text data of the power engineering project includes project specifications, budget estimates and technical reports.

6. The data recommendation method based on the knowledge graph of electric power engineering projects according to claim 5 is characterized in that: The specific method of identifying the indicator intent of the user input data, querying the structured data and its entity information in the structured database based on the identified indicator intent, and querying the associated information in the power engineering project knowledge graph includes: The indicator intent is identified for the user input data, and the structured data and its entity information are queried in the structured database based on the indicator intent, and the related information is queried in the power engineering project knowledge graph using CQL statements with the queried structured data and its entity information.

7. The data recommendation method based on the knowledge graph of electric power engineering projects according to claim 6 is characterized in that: The specific method of inputting the structured data and entity information obtained by querying the structured database and the associated information obtained by querying the knowledge graph of the power engineering project into the large language model and using the output value of the large language model as the data recommendation result includes: Classify the indicator intent into numerical data and non-numerical data; For numerical data, the associated information obtained by querying in the power engineering project knowledge graph is subjected to data statistical analysis, and the data statistical analysis results are combined as associated information with the structured data and entity information obtained by querying in the structured database and input into the large language model, and the output value of the large language model is used as the data recommendation result; the data statistical analysis includes calculation using five mathematical statistical methods: mean value, median, coefficient of dispersion, standard deviation, and discrete interval; For non-numeric data, the structured data and its entity information obtained by querying the structured database and the associated information obtained by querying the power engineering project knowledge graph are input into the large language model, and the output value of the large language model is used as the data recommendation result.

8. A data recommendation system based on a knowledge graph of electric power engineering projects, characterized in that: It includes structured database building module, knowledge graph building module, association query module and recommendation module; The structured database construction module is used to perform structured processing on heterogeneous data in the power engineering project, obtain structured data and its entity information, and construct a structured database; The knowledge graph construction module is used to capture key elements of structured data and combine massive text data of power engineering projects to construct a knowledge graph of power engineering projects with structured data and its associated information; The associated query module is used to identify the indicator intent of the user input data, query the structured data and its entity information in the structured database with the identified indicator intent, and query the associated information in the power engineering project knowledge graph; The recommendation module is used to input the structured data and its entity information obtained by querying the structured database and the associated information obtained by querying the knowledge graph of the power engineering project into the large language model, and use the output value of the large language model as the data recommendation result.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to call and run the computer program stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method as claimed in any one of claims 1 to 7 are implemented.

Images