Multi-dimensional heterogeneous technology atlas construction method, equipment and medium

Abstract

The invention relates to a multi-dimensional heterogeneous technology atlas construction method and device and a medium, the method comprises data acquisition and processing, heterogeneous graph model construction, mapping and similarity processing and multi-dimensional heterogeneous technology atlas construction, and the device and the medium are used for realizing the method. According to the method, the heterogeneous graph model and the similarity of the nodes in the model are combined to construct the technical graph, and compared with the prior art, the technical graph generated by the method is more accurate and reliable.

Description

Technical Field

[0001] The present invention relates to the field of power systems, and in particular to a method for constructing a multi-dimensional heterogeneous technology map. Background Art

[0002] In recent years, with the rapid development of my country's power industry, technological progress and market structure in key areas have undergone profound changes. The current technological innovation work of the power system is bound to become more scientific and precise, which means that the scientific and technological intelligence services required for innovation will shift from knowledge services to intelligent services. Under this general trend, scientific and technological innovation has put forward higher requirements for the identification of cutting-edge directions and trend judgment.

[0003] A technology map is a graphical or visual representation of the relationships and hierarchies of technology architecture, components, tools, frameworks, and skills in a specific field or industry. It is a graphical representation method used to describe the overall architecture of a technology and the relationships between its various parts, so as to help people better understand and navigate the knowledge in a specific technical field, locate the core technology in the technical field, understand the evolution of the technical field, and speculate on the development direction of the technical field, thereby facilitating the planning and execution of technical projects, training personnel, or evaluating the development direction of technology. Constructing a technology map generally requires eight steps: (1) determining the goal and scope; (2) collecting data and information, and performing data preprocessing; (3) entity recognition and relationship extraction; (4) data modeling and representation: modeling and representing the extracted entities and relationships; (5) map construction and fusion; (6) evaluation and verification; (7) visualization and interaction; and (8) updating and maintenance. Due to the continuous development of artificial intelligence technology, new methods have been developed in every link of constructing a technology map. The Chinese patent application "CN118095416A" discloses a method for constructing and maintaining a technology map, which includes data collection and cleaning, data modeling and association, knowledge map construction, visual display design, dynamic update and maintenance, and user interaction and application. Although the construction of the technology map is realized, the method directly constructs the technology map based on the results of data modeling. The process is complicated, the data integrity and accuracy cannot be guaranteed, and it cannot accurately describe the situation when facing complex relationships between entities.

[0004] Therefore, providing a more accurate and simple method for constructing a technology map is a problem that needs to be solved. Summary of the invention

[0005] The purpose of the present invention is to provide a method, device and medium for constructing a multi-dimensional heterogeneous technology map in order to overcome the defects of the above-mentioned prior art.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] According to a first aspect of the present invention, a method for constructing a multi-dimensional heterogeneous technology map is provided, the method comprising:

[0008] Data acquisition and processing: Acquire multi-source heterogeneous data and perform preprocessing to generate a database of key power fields. The multi-source heterogeneous data includes the title, abstract, text, citation relationship, publication time information or application time of papers and patents related to key power fields;

[0009] Heterogeneous graph model construction: construct a heterogeneous graph model based on the power key field database;

[0010] Mapping and similarity processing: mapping the nodes and the relationships between the nodes in the heterogeneous graph model to a low-dimensional vector space, and generating a similarity fusion network;

[0011] Construction of multi-dimensional heterogeneous technology map: Construct a multi-dimensional heterogeneous technology map based on the similarity fusion network and heterogeneous graph model.

[0012] As a preferred technical solution, the preprocessing includes: normalizing and structuring data acquired from different sources; and classifying, merging and cleaning the normalized and structured data.

[0013] As a preferred technical solution, the method for constructing the heterogeneous graph model is:

[0014] Analyze the aforementioned power key field database to extract keywords and research areas;

[0015] Determine entities and relationships between entities based on the keywords, research areas, and the power key area database;

[0016] Generate a corresponding relationship network based on the relationship between entities, use the related entities as nodes of the relationship network, add edges of related nodes based on the relationship between related entities, and gather all the relationship networks to obtain the initial construction of the heterogeneous graph model;

[0017] The initial heterogeneous graph model is verified and adjusted to obtain a heterogeneous graph model.

[0018] As a preferred technical solution, the method for extracting keywords and research fields is:

[0019] Preliminary extraction: using text mining technology to extract the keyword field and research field field of each paper and patent in the database of key power fields; the text mining technology includes word segmentation, part-of-speech tagging, stem extraction, part-of-speech restoration, word frequency statistics, named entity recognition and keyword extraction;

[0020] Verification or optimization: obtain the real keywords and research field data of relevant papers and patents, compare the keyword fields and research field fields with the real keywords and research field data, and perform verification or optimization based on the comparison results.

[0021] As a preferred technical solution, the method for determining entities and the relationship between entities is:

[0022] Extracting key entities from the power key domain database, assigning a unique identifier to each entity, and entering attribute information for each entity based on detailed power key domain database information;

[0023] The relationships between different entities are identified based on keywords, research fields and power key field database information, and the categories of the relationships include citation, collaboration and co-occurrence.

[0024] As a preferred technical solution, the mapping and similarity processing method is:

[0025] Using a graph neural network to encode and extract features from nodes of different types of relationship networks in the heterogeneous graph model, embed the nodes and corresponding relationships in different relationship networks into the same feature space according to the encoding and feature extraction results, and obtain fused node representation;

[0026] Calculating a similarity matrix between nodes in a relationship network of the same type based on the fused node representation, and constructing a corresponding similarity network according to the similarity matrix;

[0027] According to the prior knowledge and task objectives, weights are assigned to each similar network, and all similar networks are fused using the linear weighted fusion algorithm to obtain a similarity fusion network, which is expressed as:

[0028]

[0029] S fused represents the similarity fusion network, k represents the kth similarity network, w represents the similarity network weight, Sim k Represents the similarity matrix of the k-th similar network.

[0030] As a preferred technical solution, the method for calculating the similarity matrix between different nodes in the same type of network is:

[0031] For each type of network, the cosine similarity is used to calculate the similarity matrix between nodes, and its expression is:

[0032]

[0033] Among them, k represents the network type, i and j both represent nodes, and v iand v j Represents the edge of a node.

[0034] As a preferred technical solution, the method for constructing the multidimensional heterogeneous technology map is:

[0035] Based on the heterogeneous graph model and the results of mapping and similarity processing, classification is performed according to node types and relationship categories; the node types include authors, institutions, documents and keywords; the relationship categories include citations, collaborations and co-occurrences;

[0036] Generate a visual network using visualization technology based on the classification results;

[0037] A network layout algorithm is used to optimize the layout of the visualization network, and the optimized visualization network is verified; if the verification result is qualified, the visualization network is used as a multi-dimensional heterogeneous technology map, and the multi-dimensional heterogeneous technology map includes a panoramic technology map and a local technology map.

[0038] According to a second aspect of the present invention, there is provided an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and the method described above is implemented when the processor executes the program.

[0039] According to a third aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein the program implements the method described above when executed by a processor.

[0040] Compared with the prior art, the present invention provides a method for constructing a multidimensional technology map in key power areas. Different from the prior art in which a technology map is constructed directly through a constructed data model, the present invention utilizes a heterogeneous network structure to deeply analyze the complex relationships between power system data, generates a heterogeneous graph model, and then performs feature mapping and similarity processing based on the heterogeneous graph model to reveal the direct and indirect relationships between heterogeneous features. Based on the heterogeneous graph model, the results of mapping and similarity processing are used as a basis for constructing a technology map, so that the constructed multidimensional technology map is more reliable and accurate. In addition, when faced with complex relationships between entities, the present invention can accurately describe and ensure data integrity and accuracy based on the results of mapping and similarity processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of the method of the present invention;

[0042] Figure 2 A flowchart of constructing a heterogeneous graph of the present invention;

[0043] Figure 3 It is a multi-dimensional heterogeneous technology map of the relay protection of the present invention;

[0044] Figure 4 The citation network of the paper on the relay protection of the present invention;

[0045] Figure 5 is the keyword co-occurrence network of the relay protection of the present invention;

[0046] Figure 6 A collaborative network of authors for the relay protection of this invention;

[0047] Figure 7 A cooperative network of institutions for relay protection of the present invention;

[0048] Figure 8 It is a field crossover network of the relay protection of the present invention. DETAILED DESCRIPTION

[0049] 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 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 should fall within the scope of protection of the present invention.

[0050] With the increasing development of technologies in key power fields, single-dimensional analysis is difficult to fully reveal the laws of technological development. Multidimensional heterogeneous technology maps provide multi-level perspectives by integrating multi-dimensional data such as author cooperation, keyword co-occurrence, institutional collaboration, and field intersections. They play an important role in coping with technological complexity, supporting scientific decision-making, revealing disciplinary intersections, and discovering potential opportunities for technological integration. They help to fully understand the path of technological evolution and empower scientific researchers and enterprises. Building a multidimensional technology map can help enterprises effectively organize and manage technology-related knowledge assets, gain an in-depth understanding of the dynamics and trends in the technology field, and keep abreast of cutting-edge technologies and competitors. Therefore, building a multidimensional technology map is important for enterprises to cope with market competition.

[0051] Example 1

[0052] Based on the above reasons, this embodiment provides a method for constructing a multi-dimensional heterogeneous technology map. The process of the method is as follows: Figure 1 As shown, the detailed steps are:

[0053] S1. Data acquisition and processing.

[0054] S11. Obtaining multi-source heterogeneous data:

[0055] Multi-source heterogeneous data include papers and patent data in key fields.

[0056] S111. Obtain search terms for relevant highly cited review papers and patents in key areas of power generation, and adjust the search terms according to the research objectives.

[0057] S112. Crawl papers and patent data in key areas from the Web of Science paper database and the Derwent Innovations Index database, and obtain fields including "PT", "AU", "BA", "BE", "GP", "AF", "TI", "SO", "CT", "CY", "CL", "SP", "HO", "DE", "ID", "AB", "C1", "C3", "RP", "EM", "RI", "OI", "FU", "FP", "FX", "CR", "NR", "TC", "Z9", "U1", "U2", "PU", "PI", "PA", "SN", "EI", "BN", "J9", "JI", "PD", "VL", "IS", "BP", "EP", "AR", "DI", "DL", "D2", "EA", "PG", "WC", "WE", "SC", "GA", "PM", "OA", "BP", "EP", "AR", "DI", "DL", "D2", "EA", "PG", "WC", "WE", "SC", "GA", "PM", "OA".

[0058] Among them, paper and patent data at least include the title, abstract, text, citation relationship, publication time information or application time of the relevant papers and patents.

[0059] S12. Preprocess multi-source heterogeneous data to generate a database for key power areas.

[0060] S121. Normalize and structure data obtained from different sources.

[0061] S122. Classify, merge and clean the normalized and structured data, and eliminate data irrelevant to the target research field.

[0062] S2. Heterogeneous graph model construction.

[0063] The process of this step is as follows Figure 2 The detailed steps are as follows:

[0064] S21. Analyze the power key field database and extract keywords and research fields.

[0065] S211. Preliminary extraction: Use text mining techniques such as word segmentation, part-of-speech tagging, stem extraction, part-of-speech restoration, word frequency statistics, named entity recognition and keyword extraction to extract the keyword field and research field field of each paper and patent in the database of key power fields.

[0066] S212, verification or optimization: obtain the real keywords, WoS categories, International Patent Classification (IPC) and other data of relevant papers and patents, compare the keyword fields and research field fields with the real keywords, WoS categories, International Patent Classification (IPC) and other data, and perform verification or optimization based on the comparison results.

[0067] S22. Determine entities and relationships between entities based on keywords, research fields, and the power key field database.

[0068] S221, extract entities, such as papers, patents, authors, keywords and scientific terms, and determine entity categories, such as authors, institutions, documents, keywords, organizations, concepts and documents, and assign a unique identifier to each entity, and enter attribute information for each entity based on detailed power key field database information;

[0069] S222. Identify the relationships between different entities based on keywords, research fields, and power key field database information, where the categories of relationships include citation, collaboration, and co-occurrence.

[0070] S23, generating a corresponding relationship network according to the relationship between the entities, and taking the related entities as nodes of the relationship network; adding edges of the related nodes based on the relationship between the related entities, and setting the attributes of the edges, such as type, direction and weight;

[0071] S24. Collect all the relationship networks to obtain the initial heterogeneous graph model.

[0072] S25. Verify and adjust the initial heterogeneous graph model to obtain a heterogeneous graph model.

[0073] S251. Use actual case data to perform tests, perform queries and path analysis, test the responsiveness and accuracy of the heterogeneous graph model, and verify the practicality of the model.

[0074] S252. Collect feedback information, identify and further resolve possible problems in the heterogeneous graph model, such as failure to correctly identify certain node types and node relationships, omission of key nodes or relationships, etc., so as to improve the practicality and reliability of the graph model.

[0075] S3, mapping and similarity processing.

[0076] S31. Use graph neural networks to encode and extract features from nodes of different types of relationship networks in heterogeneous graph models. According to the encoding and feature extraction results, embed the nodes and corresponding relationships in different relationship networks into the same feature space to obtain fused node representation.

[0077] S32, based on the fusion node representation, the similarity matrix between the nodes in the same type of relationship network is calculated using cosine similarity, and the corresponding similarity network is constructed according to the similarity matrix, and the calculation formula is:

[0078]

[0079] Among them, k represents the network type, i and j both represent nodes, and v i and v j Represents the edge of a node.

[0080] S33. Assign weights to each similar network according to prior knowledge and task objectives, and use a linear weighted fusion algorithm to fuse all similar networks to obtain a similarity fusion network, which is expressed as:

[0081]

[0082] S fused represents the similarity fusion network, k represents the kth similarity network, w represents the similarity network weight, Sim k Represents the similarity matrix of the k-th similar network.

[0083] For the graph neural network used in step S3, the method provided in this embodiment designs a target optimization function of the graph neural network according to different task objectives, such as classification and clustering, and optimizes the weight parameters of the graph neural network through performance training based on the target optimization function to make its results more accurate.

[0084] S4. Construction of multi-dimensional heterogeneous technology map.

[0085] S41. Based on the heterogeneous graph model and the results of mapping and similarity processing, classification is performed according to node types and relationship categories, where node types include authors, institutions, documents and keywords, and relationship categories include citations, collaborations and co-occurrences.

[0086] S42. Generate a visualization network using the classification results using visualization technology.

[0087] S43. Use a network layout algorithm to optimize the layout of the visualization network, where the network layout algorithm includes a force-guided layout algorithm and a hierarchical layout algorithm.

[0088] S44. Verify the optimized visualization network. If the verification result is qualified, use the visualization network as a multi-dimensional heterogeneous technology map, wherein the multi-dimensional heterogeneous technology map includes a panoramic technology map and a local technology map.

[0089] The present embodiment also provides an electronic device, which includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0090] Multiple components in the device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks.

[0091] The processing unit performs the various methods and processes described above, such as methods S1 to S4. For example, in some embodiments, methods S1 to S4 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of methods S1 to S4 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S4 in any other appropriate manner (e.g., by means of firmware).

[0092] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0093] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0094] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0095] Example 2

[0096] In order to verify that the method provided in the above embodiment is feasible, in this embodiment, the field of relay protection is taken as an example, and the method provided in Example 1 is used to construct a technology map. Therefore, the construction process of the technology map will not be repeated in this embodiment. The analysis is carried out based on the relay protection technology map, and the steps include:

[0097] A1. Using the method provided in Example 1 to construct a technology map, the following can be obtained: Figure 3 The multi-dimensional heterogeneous technology map in the field of relay protection shown in FIG. includes: Figure 4 The paper citation network shown, Figure 5 The keyword co-occurrence network shown in Figure 6 The author collaboration network shown in Figure 7 The institutional collaboration network shown and Figure 8 The domain cross-network shown.

[0098] In the figure, red nodes and red lines represent papers, and the edges between red nodes represent citation relationships. Nodes that have edges with red nodes can be regarded as attributes of red nodes; green nodes represent keywords, and the lines between green nodes represent the co-occurrence relationship of keywords; blue nodes represent authors, and the edges between blue nodes represent cooperative relationships; purple nodes represent institutions, and the edges between purple nodes represent cooperation between institutions; orange nodes represent fields, and the lines between orange nodes represent intersections between fields.

[0099] A2. Analyze the constructed multi-dimensional heterogeneous technology map.

[0100] A21. Conduct multi-dimensional data feature analysis on data in key power areas, analyze node type distribution, relationship type distribution and basic attributes, outline the global structure and key features of the data, establish a preliminary understanding of the power outage protection field, and provide support for understanding the general situation in the power outage protection field.

[0101] A22, based on Figure 4 A multi-dimensional heterogeneous technology map is used to analyze the development of key power areas.

[0102] A23. Conduct network structure analysis on the paper citation network, author cooperation network, keyword co-occurrence network, institutional cooperation network and field cross-network in the field of power outage protection.

[0103] The specific analysis is as follows:

[0104] exist Figure 4 In the paper citation network shown, the nodes are the titles of the papers and the edges are the citation relationships of the papers. From the perspective of citation distribution, the journal IEEE T POWER DELIVER is the most widely cited in relay protection. Among them, the representative papers of IEEE Transactions on Power Transmission include "Allocation of Centralized Switch Fault Limiters Enabled by 5G in Transmission Systems" and "The Impact of Inverter-Based Resources on Negative Sequence-Based Protection Elements", and the main research contents of both are focused on renewable energy and relay misoperation problems; representative papers of Power System Research include "A Review and Future Trends of Protection Research on Renewable Energy Integrated AC Microgrids" and "Optimal Coordination of Directional Overcurrent Relays in Complex Distribution Networks Based on Sine-Cosine Algorithms", and the main research contents of both are focused on renewable energy and microgrids and overcurrent relay optimization problems.

[0105] exist Figure 5In the keyword co-occurrence network shown, the nodes are the keywords of the paper, and the edges are the co-occurrence relationships of the keywords. It can be seen from the figure that the high-frequency keywords in the field of relay protection include power system protection, fault detection and overcurrent protection. According to the representative papers containing high-frequency keywords, such as "An improved simulated annealing-linear programming hybrid algorithm applied to the optimization and coordination of directional overcurrent relays", the linear programming algorithm is used to solve the relay optimization problem, and "Microgrid intelligent protection strategy based on convolutional neural network" uses convolutional neural network to optimize the intelligent fault protection strategy. It can be seen that the overall trend in the field of relay protection is to use new methods such as linear programming and deep learning to optimize relay protection problems.

[0106] exist Figure 6 In the author cooperation network shown, the nodes are the authors of the papers, and the edges are the cooperation relationships between the authors. From the perspective of cooperation distribution, cooperation presents a block attribute, each cooperation forms a closed loop to form a small group, and the cooperation between groups is almost zero. According to the figure, the representative authors in the author cooperation network include: 1) Brahma Sukumar, who works at New Mexico State University, and his representative papers include "Enhancing distribution network protection strategy using communication protocols" and "Distributed generation microgrid protection based on multi-agent system". The main research content includes solving complex problems of microgrids based on collaborative simulation platform and computer-aided design; 2) Yin Xianggen / Yin Xianggen, who works at Huazhong University of Science and Technology, and his representative papers include "DFIG short-circuit model considering coordinated control strategy of grid-side and rotor-side converters". The main research content includes relay protection problems of renewable energy power grids.

[0107] exist Figure 7 In the institutional cooperation network shown, the nodes are the institutions to which the papers belong, and the edges are the cooperation relationships between institutions. Institutional cooperation presents a "core-periphery" feature. According to the figure, it can be concluded that the core is mainly cooperation with institutions in European and American countries, such as the University of California, and the periphery includes Indian Institute of Technology, State Grid, and Huazhong University of Science and Technology. Among them, representative institutions include: 1) State Grid, representative research includes "Grid Fault Diagnosis Based on Variational Mode Decomposition and Convolutional Neural Network" and "A New Distance Protection Scheme Based on Improved Virtual Voltage Measurement", and the main research content includes the use of advanced methods such as deep learning and reinforcement learning to solve grid fault diagnosis and relay protection problems; 2) Indian Institute of Technology, representative research includes "New Algorithm for Fault Detection and Location of Low-voltage DC Microgrids", and the main research content includes intelligent microgrids and overcurrent relay optimization problems.

[0108] exist Figure 8In the domain intersection network shown, the nodes are the fields to which the papers belong, and the edges are the intersection relations of the fields. From the perspective of domain intersection distribution, the fields of electrical engineering and energy fuels have the most intersections, and the related intersecting fields also include information systems, materials science and communications. The representative fields include: 1) Information system field, representative papers include "Efficient Communication Federated Learning and Permissioned Blockchain for Digital Dual Edge Networks" and "Cascading Failures in the Internet of Things: A Review and Outlook on Reliability and Resilience", and the main research content includes intelligent fault detection combined with blockchain and the Internet of Things; 2) Artificial intelligence field, representative papers include "Microgrid Adaptive Protection Scheme Based on SOM Clustering Technology" and "Dynamic FDB Selection Method and Its Application: Coordinated Modeling and Optimization of Directional Overcurrent Relays", and the main research content includes modeling and optimization combined with artificial intelligence clustering and classification algorithms.

[0109] In summary, the construction of a technology map using the method provided in Example 1 can cover multiple aspects, and the various relationships between nodes in the map can be used to explore key research directions, technical themes, and development trends in the field of power outage protection. At the same time, by identifying high-impact nodes in the field, exploring the synergy and influence paths between nodes, revealing the technical iteration rules and innovation patterns in the field, and providing a reference for cutting-edge research and application in the field of power outage protection. Then, by analyzing the network indicators such as the centrality and density of each relationship network in the technology map, the importance and connection strength of the nodes in their respective networks are evaluated, high-impact researchers, institutions, and concepts are identified, and potential research hotspots and innovation trends are explored.

[0110] It can be seen that the method provided in Example 1 can effectively extract key trends and evolution laws when dealing with large and complex key power areas, reveal the direct and indirect interactions in the key power areas, and provide a more comprehensive system protection perspective. Based on the technical map generated by this method, technicians can better understand the development status and trends in the key power areas, thereby maintaining a competitive advantage in research and development and innovation.

[0111] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for constructing a multi-dimensional heterogeneous technology map, characterized in that: The method includes: Data acquisition and processing: Acquire multi-source heterogeneous data and perform preprocessing to generate a database of key power fields. The multi-source heterogeneous data includes the title, abstract, text, citation relationship, publication time information or application time of papers and patents related to key power fields; Heterogeneous graph model construction: construct a heterogeneous graph model based on the power key field database; Mapping and similarity processing: mapping the nodes and the relationships between the nodes in the heterogeneous graph model to a low-dimensional vector space, and generating a similarity fusion network; Construction of multi-dimensional heterogeneous technology map: Construct a multi-dimensional heterogeneous technology map based on the similarity fusion network and heterogeneous graph model.

2. A method for constructing a multidimensional heterogeneous technology map according to claim 1, characterized in that: The preprocessing includes: normalizing and structuring the data obtained from different sources; and classifying, merging and cleaning the normalized and structured data.

3. A method for constructing a multidimensional heterogeneous technology map according to claim 1, characterized in that: The method for constructing the heterogeneous graph model is: Analyze the aforementioned power key field database to extract keywords and research areas; Determine entities and relationships between entities based on the keywords, research areas, and the power key area database; Generate a corresponding relationship network based on the relationship between entities, use the related entities as nodes of the relationship network, add edges of related nodes based on the relationship between related entities, and collect all the relationship networks to obtain the initial construction of the heterogeneous graph model; The initial heterogeneous graph model is verified and adjusted to obtain a heterogeneous graph model.

4. A method for constructing a multi-dimensional heterogeneous technology map according to claim 3, characterized in that: The method for extracting keywords and research fields is: Preliminary extraction: using text mining technology to extract the keyword field and research field field of each paper and patent in the database of key power fields; the text mining technology includes word segmentation, part-of-speech tagging, stem extraction, part-of-speech restoration, word frequency statistics, named entity recognition and keyword extraction; Verification or optimization: obtain the real keywords and research field data of relevant papers and patents, compare the keyword fields and research field fields with the real keywords and research field data, and perform verification or optimization based on the comparison results.

5. A method for constructing a multi-dimensional heterogeneous technology map according to claim 3, characterized in that: The method for determining entities and the relationship between entities is: Extracting key entities from the power key domain database, assigning a unique identifier to each entity, and entering attribute information for each entity based on detailed power key domain database information; The relationships between different entities are identified based on keywords, research fields and power key field database information, and the categories of the relationships include citation, collaboration and co-occurrence.

6. A method for constructing a multi-dimensional heterogeneous technology map according to claim 1, characterized in that: The mapping and similarity processing methods are: Using a graph neural network to encode and extract features from nodes of different types of relationship networks in the heterogeneous graph model, embed the nodes and corresponding relationships in different relationship networks into the same feature space according to the encoding and feature extraction results, and obtain fused node representation; Calculating a similarity matrix between nodes in a relationship network of the same type based on the fused node representation, and constructing a corresponding similarity network according to the similarity matrix; According to the prior knowledge and task objectives, weights are assigned to each similar network, and all similar networks are fused using the linear weighted fusion algorithm to obtain a similarity fusion network, which is expressed as: S fused represents the similarity fusion network, k represents the kth similarity network, w represents the similarity network weight, Sim k Represents the similarity matrix of the k-th similar network.

7. A method for constructing a multi-dimensional heterogeneous technology map according to claim 6, characterized in that: The method for calculating the similarity matrix between different nodes in the same type of network is: For each type of network, the cosine similarity is used to calculate the similarity matrix between nodes, and its expression is: Among them, k represents the network type, i and j both represent nodes, and v i and v j Represents the edge of a node.

8. The method for constructing a multi-dimensional heterogeneous technology map according to claim 1, characterized in that: The method for constructing the multidimensional heterogeneous technology map is: Based on the heterogeneous graph model and the results of mapping and similarity processing, classification is performed according to node types and relationship categories; the node types include authors, institutions, documents and keywords; the relationship categories include citations, collaborations and co-occurrences; Generate a visual network using visualization technology based on the classification results; A network layout algorithm is used to optimize the layout of the visualization network, and the optimized visualization network is verified; if the verification result is qualified, the visualization network is used as a multi-dimensional heterogeneous technology map, and the multi-dimensional heterogeneous technology map includes a panoramic technology map and a local technology map.

9. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images