Innovative demand extraction method and system for power grid enterprise

Abstract

The invention relates to the technical field of power technology and innovation management, and provides an innovation demand extraction method and system for a power grid enterprise. The method comprises the following steps: extracting keywords of innovation demand multi-source data, and converting the keywords into keyword embedding vectors; performing semantic matching on the keyword embedded vector and innovation demand multi-source data in a database, and calculating association similarity; clustering the innovation demand multi-source data corresponding to the association similarity greater than a first set threshold to obtain a first demand list; defining demand screening indexes, performing weighted comprehensive scoring on all demand screening indexes of the first demand list by using a fuzzy logic method, and screening a second demand list of which the score is greater than a second set threshold value; and based on the second demand list, constructing a demand association graph, determining initial weights of edges in the demand association graph, optimizing the weights of the edges in the demand association graph by adopting a GNN network, sorting the priorities of the demand nodes, and outputting a demand sorting result.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology and innovation management technology, and in particular to a method and system for extracting innovation demands of a power grid enterprise. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The innovation demand extraction work of power grid companies faces complex challenges in multiple fields, levels and dimensions. Traditional demand identification methods have exposed obvious shortcomings in practice and cannot meet the urgent requirements of current technological innovation and energy transformation.

[0004] Innovation and efficiency demand refers to the innovative matters that power grid enterprises can simultaneously meet the strategic needs of enterprise development and have significant economic and social benefits during the process of technological innovation. With the rapid development of energy transformation and digital technology, the methods of identifying innovative needs of power grid enterprises are constantly being optimized. At present, demand identification technology based on big data, artificial intelligence and collaborative innovation platforms has gradually become mainstream. Typical methods include the following: data-driven analysis, machine learning-assisted decision-making, collaborative innovation platform, and expert consultation and scenario verification combined. However, in the application of innovative demand identification in power grid enterprises, these methods have the following problems: (1) Low efficiency in demand acquisition: Traditional methods often rely on expert experience and limited survey data, with a single collection method and limited coverage. Due to the wide range of demand sources and complex distribution, traditional manual analysis methods require a lot of time and manpower, and it is difficult to efficiently capture the potential needs of power grid companies in the innovation process. In addition, some hidden needs are difficult to accurately explore through traditional survey methods, resulting in the efficiency and comprehensiveness of demand acquisition far from expectations.

[0005] (2) Insufficient accuracy of demand analysis: The existing technology generally adopts static analysis methods, which rely on fixed models or indicator systems. This method is difficult to adapt to the rapidly changing external environment. In addition, the iteration speed of market demand and industry technology is constantly accelerating, and static analysis methods are difficult to meet the high standards for demand accuracy.

[0006] (3) Low level of intelligence: At present, the application of advanced technologies such as big data and artificial intelligence in the field of demand identification is still in its early stages, and the overall level of intelligence is low. Traditional methods are difficult to use massive data resources for in-depth analysis, and usually only stay at the level of data collection and simple statistics. They lack the ability to intelligently mine multidimensional data and analyze complex associations, and cannot achieve automation, intelligence, and dynamic optimization of demand identification. Summary of the invention

[0007] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a method and system for extracting innovation needs of power grid enterprises, which aims to comprehensively and accurately obtain the innovation needs in the field of power grid business. Through multi-dimensional data fusion, intelligent algorithm screening and dynamic modeling, a demand framework for innovation and efficiency of power grid enterprises is constructed to provide a scientific basis for power grid technology research and development and resource allocation.

[0008] In order to achieve the above object, the present invention adopts the following technical solution: A first aspect of the present invention provides a method for extracting innovative demands of a power grid enterprise.

[0009] A method for extracting innovative requirements of a power grid enterprise, comprising: Obtain multi-source data on innovation needs and build a database; extract keywords from multi-source data on innovation needs, build a keyword library, and convert them into keyword embedding vectors; Semantically matching the keyword embedding vector with the multi-source data of innovation requirements in the database to calculate the correlation similarity; clustering the multi-source data of innovation requirements corresponding to the correlation similarity greater than a first set threshold to obtain a first requirement list; Demand screening indicators are defined, and a fuzzy logic method is used to perform a weighted comprehensive score on all demand screening indicators of the first demand list, and a second demand list having a score greater than a second set threshold is screened; Based on the second demand list, a demand association graph is constructed, the initial weights of the edges in the demand association graph are determined, the GNN network is used to optimize the weights of the edges in the demand association graph, the priorities of the demand nodes in the optimized demand association graph are sorted, and the demand sorting results are output.

[0010] Further, the initial weights of the edges in the demand association graph are determined, the GNN network is used to optimize the weights of the edges in the demand association graph, the priorities of the demand nodes in the optimized demand association graph are sorted, and the demand sorting results are output; the method includes: Calculate the semantic similarity between the embedding vectors of demand nodes, quantify the upstream and downstream dependency weights between demand nodes, and calculate the initial weight of the edge between two demand nodes ; In the message passing layer of the GNN network, according to the changes in the external environment, a weight correction mechanism based on external environmental factors is designed to calculate the correction factor between two demand nodes , expressed by the following formula:

[0011] in, , and represent the policy relevance, market trend and technology popularity of demand nodes i and j respectively; γ, δ and They represent the weight coefficients of policy relevance, market trend and technology enthusiasm in the revision process respectively; According to the initial weight of the edge between two demand nodes Correction factor between two demand nodes The sum of , calculates the weight of the edge in the updated demand association graph , the optimized demand association graph has been obtained; The GAT attention mechanism is introduced in the GNN network to distinguish the importance of different neighbor nodes in the optimized demand association graph; according to the importance of the neighbor nodes, the weight information of the corrected edges and the feature information of the neighbor nodes are weighted and aggregated to the corresponding nodes, the feature information of the nodes is updated, the priority of the demand nodes is calculated according to the feature information, the priority of the demand nodes is re-sorted, and the demand sorting results are output.

[0012] Further, the demand association graph is optimized, and the method also includes: when the external environment changes and introduces new demands, calculating the association weights of new nodes and adding nodes or edges to the demand association graph; deleting low-priority nodes or edges; and re-ordering the priorities of demand nodes according to the updated edge weights.

[0013] Further, it is converted into a keyword embedding vector; the method includes: semantically classifying and supplementing the initially extracted keywords in combination with the opinions of field experts, adding supervision information to the keywords, screening and optimizing long-tail keywords in the keyword library; using a neural network model to map the keywords to a low-dimensional vector space and converting them into a keyword embedding vector.

[0014] Furthermore, based on the second requirement list, a requirement association graph is constructed; the method includes: using a natural language processing tool to parse the second requirement list and extract the associations between the requirements; and constructing a requirement association graph with the requirements as nodes and the associations between the requirements as edges.

[0015] Furthermore, keywords of the multi-source data of innovation needs are extracted, including: based on the multi-source data of innovation needs, a deep learning model is used to extract keywords of the multi-source data of innovation needs.

[0016] Furthermore, the obtained multi-source data of innovation needs are pre-processed, including data cleaning and standardization; Furthermore, the multi-source data on innovation needs include policy documents, industry trends, patent texts, market research reports, and technical documents of historical enterprise projects.

[0017] Furthermore, the first requirement list includes requirement content, data source and correlation similarity.

[0018] Further, demand screening indicators are defined, and the method includes: based on the Drools rule engine, demand screening indicators are defined, and the demand screening indicators include: technology maturity, market value and policy adaptability.

[0019] A second aspect of the present invention provides an innovative demand extraction system for a power grid enterprise.

[0020] An innovative demand extraction system for power grid enterprises, comprising: The data acquisition and preprocessing module is configured to: acquire multi-source data of innovation needs and build a database; extract keywords from the multi-source data of innovation needs, build a keyword library, and convert them into keyword embedding vectors; A clustering module is configured to: semantically match the keyword embedding vector with the multi-source data of innovation requirements in the database, and calculate the correlation similarity; cluster the multi-source data of innovation requirements corresponding to the correlation similarity greater than a first set threshold, and obtain a first requirement list; The screening module is configured to: define demand screening indicators, use fuzzy logic method to perform weighted comprehensive scoring on all demand screening indicators of the first demand list, and screen the second demand list whose score is greater than a second set threshold; The output module is configured to: construct a demand association graph based on the second demand list, determine the initial weights of the edges in the demand association graph, use the GNN network to optimize the weights of the edges in the demand association graph, sort the priorities of the demand nodes in the optimized demand association graph, and output the demand sorting results.

[0021] A third aspect of the present invention provides a computer-readable storage medium.

[0022] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the innovative demand extraction method for a power grid enterprise as described in the first aspect above.

[0023] A fourth aspect of the present invention provides a computer device.

[0024] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the innovative demand extraction method for a power grid enterprise as described in the first aspect above are implemented.

[0025] A fifth aspect of the present invention provides a computer program product or a computer program.

[0026] The present invention provides a computer program product or a computer program, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the innovative demand extraction method for a power grid enterprise as described in the first aspect above.

[0027] Compared with the prior art, the present invention has the following beneficial effects: (1) Improve the comprehensiveness and dynamism of innovation demand identification: Build an innovation demand keyword library based on multi-source data, use deep learning technology and text segmentation models to extract technical innovation keywords from policies, information, literature and other scientific and technological intelligence, achieve comprehensive scanning and extraction, and ensure the comprehensiveness of innovation demand. At the same time, the use of dynamic optimization algorithms and real-time monitoring mechanisms can quickly respond to changes in the external environment, and update the analysis model and indicator system in real time, thereby improving dynamism.

[0028] (2) Improving the accuracy and scientificity of innovation demand identification: By screening demand through intelligent algorithms, the accuracy and scientificity of identification are significantly improved, thereby effectively avoiding the subjectivity of manual judgment.

[0029] (3) Achieving intelligence and improving efficiency: By introducing artificial intelligence technology, combined with natural language processing (NLP) and graph neural networks (GNN), potential needs can be automatically extracted from the data, and the needs can be scientifically prioritized through intelligent optimization models to complete in-depth analysis and pattern recognition of massive data. Ultimately, the entire process of demand identification can be intelligentized, reducing the cost of manual intervention and greatly improving the efficiency of identifying innovative needs.

[0030] (4) Supporting power grid companies in building a new power system for the future: The present invention focuses on the innovative needs in the power grid business field, and provides a new perspective and direction for the technical layout, discipline construction, platform facilities and project deployment of power grid companies. By accurately identifying projects, it helps power grid companies seize the initiative in the construction of new power systems and promote the energy revolution. The present invention also takes into account the technical needs of the digital and intelligent transformation of power grids, and provides strong technical support and decision-making assistance for power grid companies to build a new power system for the future. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0032] Figure 1 It is a flow chart of the innovative demand extraction method of the power grid enterprise shown in the present invention; Figure 2 It is a flow chart of constructing the innovation demand keyword library shown in the present invention. DETAILED DESCRIPTION

[0033] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0034] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0035] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0036] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and systems according to various embodiments of the present disclosure. It should be noted that each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of a code may include one or more executable instructions for implementing the logical functions specified in each embodiment. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and / or block diagram, and the combination of boxes in the flowchart and / or block diagram can be implemented using a dedicated hardware-based system that performs a specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0037] Embodiment 1 The present embodiment provides a method for extracting innovative demands of a power grid enterprise. The present embodiment takes the application of the method to a server as an example. It is understandable that the method can also be applied to a terminal, and can also be applied to a system including a terminal, a server, and a server, and is implemented through the interaction between the terminal and the server. The server can be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communications, middleware services, domain name services, security services CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, which is not limited in this application. In the present embodiment, the method includes the following steps: Obtain multi-source data on innovation needs and build a database; extract keywords from multi-source data on innovation needs, build a keyword library, and convert them into keyword embedding vectors; Semantically matching the keyword embedding vector with the multi-source data of innovation requirements in the database to calculate the correlation similarity; clustering the multi-source data of innovation requirements corresponding to the correlation similarity greater than a first set threshold to obtain a first requirement list; Demand screening indicators are defined, and a fuzzy logic method is used to perform a weighted comprehensive score on all demand screening indicators of the first demand list, and a second demand list having a score greater than a second set threshold is screened; Based on the second demand list, a demand association graph is constructed, the initial weights of the edges in the demand association graph are determined, the GNN network is used to optimize the weights of the edges in the demand association graph, the priorities of the demand nodes in the optimized demand association graph are sorted, and the demand sorting results are output.

[0038] The present invention constructs a keyword library of grid innovation demand, scans and analyzes the full amount of innovation demand of grid enterprises, studies the future technical form of core grid professions, constructs demand filtering rules, and constructs a demand integration model, and proposes a demand framework for innovation and efficiency in the grid business field. This method improves the comprehensiveness and dynamism of innovation demand identification, improves the accuracy and scientificity of innovation demand identification, and realizes intelligence and efficiency improvement, supporting grid enterprises to build a new power system for the future.

[0039] The present embodiment is described in detail below with reference to the accompanying drawings. Figure 1 As shown: Step 1: Data collection and integration Multi-source data acquisition: Use crawler technology (such as Python's Scrapy framework) to collect data from policy databases, industry report websites, patent databases, corporate intranets and other related data source websites. The data includes policy documents, industry trends, patent texts, market research reports, and technical documents of corporate historical projects.

[0040] Data cleaning and standardization: The collected data is cleaned using Pandas and NLTK tools in the Python library, including removing irrelevant fields, unifying the format, and processing noisy data.

[0041] Data storage and management: Apply database technology to store the cleaned data in a database (such as MongoDB or Elasticsearch) for quick retrieval and analysis.

[0042] Step 2: Build a keyword library for innovation needs (see the attached process framework) Figure 2 ) Keyword extraction: Use technology based on deep learning pre-trained language models, such as the BERT pre-trained model or the GPT model, to perform semantic analysis on the data collected in the database, parse the semantic relationship of keywords, extract high-frequency technical vocabulary based on semantics, and build a keyword library.

[0043] Keyword optimization: Combined with the opinions of experts in the field, the initially extracted keywords are semantically classified and supplemented, and supervision information is added to the keywords to ensure that they cover the core areas of power grid technology, such as power transmission and distribution, energy storage, and dispatching. At the same time, long-tail keywords are screened and optimized in the keyword library to improve the quality of the keyword library.

[0044] Keyword vectorization: After the keyword library is built, the Word2Vec model is used to map the keywords into a low-dimensional vector space through a neural network model, and the keywords in the vocabulary are converted into embedded vectors to facilitate subsequent semantic matching and clustering analysis.

[0045] Step 3: Scan and analyze all innovation needs Semantic matching and association analysis: Use the cosine similarity algorithm to semantically match the keyword embedding vector with the multi-source data in the database, associate the keywords with the data source, obtain the association similarity score, and extract the innovation requirements with high similarity.

[0046] Demand clustering: The K-Means algorithm is used to cluster the extracted innovation demands and classify similar demands into the same topic.

[0047] Generate a requirements list: Output the clustered innovation requirements to form a preliminary requirements list, including requirement content, data source, and associated scores.

[0048] Step 4: Build demand filtering rules Rule engine design: Based on the Drools rule engine, we define demand screening indicators, including technology readiness level (TRL) (evaluating the development stage of technology based on technology application cases described in the literature, see Table 1 for details), market value (evaluating the commercial potential of demand through external market analysis data) and policy adaptability (matching the priority support direction of national policies).

[0049] Table 1 Technology Readiness Level (TRL) and Application Stage and Commercialization Potential

[0050] Fuzzy logic scoring (see Table 2 for details): Use fuzzy logic method to give a weighted comprehensive score to the three indicators of technology maturity, market value and policy adaptability, and perform normalization to eliminate low-priority requirements based on the score.

[0051] Table 2 Fuzzy logic scoring criteria

[0052] Step 5: Build a demand integration model Constructing a demand association graph: In the construction of the demand association graph, the goal of the model is to extract and display the semantic associations and technical dependencies between grid innovation needs. Based on the initial demand data, natural language processing tools are first used to deeply analyze the demand text and extract potential associations. For example, by semantically analyzing the demand description text, it can be determined that there is a direct technical dependency between the demand for "high proportion of new energy access" and the demand for "distributed energy storage optimization". This dependency relationship not only comes from the demand text, but can also be further verified by matching the technical constraints mentioned in policy documents or technical literature. These relationships are represented in the form of a graph, where nodes represent specific needs and edges represent the associations between needs.

[0053] Optimize the demand graph structure: Dynamically adjust the weights of edges during graph construction. The initial weights are calculated based on semantic similarity and upstream and downstream dependencies. The relevant calculation formula is as follows: Semantic similarity is used to measure the semantic relevance between requirement texts. Common calculation methods include similarity metrics based on embedding vectors, such as cosine similarity: (1) in, Indicates the semantic similarity between demand nodes i and j; and Represent the semantic embedding vectors of demand nodes i and j respectively.

[0054] The upstream and downstream dependencies mainly reflect the technical or business logic relationship between requirements, which can be quantified in the following ways: ① Technical document analysis: Determine whether there is a direct dependency through the technical terms or process descriptions in the requirement text. For example, if the description of requirement A mentions the output of requirement B as input, it can be determined that A depends on B. ② Policy document reference: Determine whether there is a logical connection between the two requirements by analyzing the technical routes or goals mentioned in the policy documents. The calculation formula for the upstream and downstream dependency weights is as follows: (2) in, represents the upstream and downstream dependency weights of demand nodes i and j; It indicates the dependence intensity of requirement i on requirement j (obtained through statistics such as the number of citations, technical descriptions, etc., such as ).

[0055] Combining the combined influence of semantic similarity and upstream and downstream dependencies, the association weight of two demand nodes i and j can be expressed as: (3) Among them, α and β are the importance weight coefficients of semantic similarity and upstream and downstream dependencies respectively (which can be adjusted through expert experience or historical data).

[0056] In the process of building the demand graph, it will be revised in combination with changes in the external environment (such as policy adjustments, changes in market demand, the emergence of new technologies, etc.). External environmental factors can be expressed through quantitative indicators, such as: Policy relevance (P): how well the need matches new policy or guidance documents; Market trend (M): the potential commercial value of demand in market analysis; Technology heat (T): the research heat or number of patent applications for the technology corresponding to the demand.

[0057] During the optimization process, an automated edge weight adjustment algorithm is used to continuously update the edge weights of the demand graph by comparing the correlation between historical data and new requirements, so that it can reflect the latest business needs. In order to dynamically reflect these changes, it is necessary to design a weight correction mechanism based on external environmental factors. The specific formula is as follows: (4) in, represents the correction factor of demand nodes i and j; , and represent the policy relevance, market trend and technology popularity of demand nodes i and j respectively; γ, δ and They respectively represent the weight coefficients of external environmental factors in the correction process.

[0058] The association weight after correction in combination with the external environment can be expressed as: (5) The above dynamic update process can be summarized as follows: (1) Obtain external environment data in real time: Extract the latest data from policy databases, market analysis reports, and technical literature to quantify policy, market, and technology heat indicators.

[0059] (2) Weight correction module: calculates the correction factor based on the extracted data , and update the edge weights of the demand association graph.

[0060] (3) Iterative optimization: The association graph is continuously optimized iteratively through the following steps: Adding new nodes or edges: When changes in the external environment introduce new requirements, use the above formula to calculate the association weight of the new node and add it to the association graph; Delete low-priority nodes or edges: remove requirements or associated edges with lower weights after correction to keep the graph structure simple; Weight reordering: Apply the improved graph neural network to weight and aggregate the corrected edge weight information and the feature information of the neighboring nodes into the corresponding nodes, update the feature information of the nodes, calculate the priority of the demand nodes based on the feature information, and reorder the priority of the demand nodes. After building the demand association graph, the improved graph neural network is used to optimize it. The improved GNN model contains the following key components: (1) Input layer: Node feature vector: The feature of each demand node is composed of the embedding vector of its description text (such as the high-dimensional vector generated by BERT or Word2Vec), denoted as .

[0061] Edge weight vector: The weight of the edge in the demand association graph (combining semantic similarity and upstream and downstream dependencies), which is used to represent the strength of association between demand nodes, denoted as .

[0062] (2) Message passing layer: Each node updates its own state through the information of adjacent nodes, and the message transmission rules are improved as follows: Enhanced edge weight sensitivity: Introducing edge weights As a weighting factor for message passing, edges with higher associated weights have a greater impact on node updates; Dynamic context-aware mechanism: Correction factors through external environment Dynamically adjust edge weights to reflect the impact of the external environment on demand associations in real time.

[0063] The update formula for node i is: (6) in, represents the set of neighbor nodes represented by node b; σ represents the activation function (such as ReLU).

[0064] (3) Attention Mechanism: The improved GNN introduces the attention mechanism of Graph Attention Network (GAT) to distinguish the importance of different adjacent nodes. The attention weight calculation formula is: (7) in, represents the attention weights of nodes i and j; a represents the learnable attention vector; W represents the transformation matrix of node features; || represents the vector concatenation operation.

[0065] (4) Output layer: Node priority score: By performing nonlinear transformation on each node feature embedding, the priority score of the required node is output, denoted as ; Optimized edge weights: Outputs the optimized edge weights, reflecting the latest association strength between requirements. The core task of the network is to re-prioritize the demand nodes according to the iteratively updated node feature vector (Formula 6) based on the semantic features of the demand (such as cosine similarity, Formula 1) and the association relationship (Formula 2). Unlike the traditional graph model, the improved GNN model introduces a dynamic context-aware mechanism (Formula 6), which enables the model to instantly update the properties of existing nodes when processing newly added demands. For example, the newly added "new energy scheduling optimization" demand may affect the technical path of the existing "intelligent transmission and distribution" demand, so the network will readjust the strength of the association between the two. In network training, supervised learning is introduced, and the importance of nodes is annotated using historical data to guide the model to calibrate the accuracy of priority sorting. The improved graph neural network optimization process can be summarized as follows: (1) Initialize the association graph: Node initialization: Encode the text features of the required nodes into embedded vectors through pre-trained models (such as BERT or Word2Vec) as input features ; Edge weight initialization: Calculate the initial weight of each edge based on semantic similarity and upstream and downstream dependencies .

[0066] (2) Dynamic messaging: In the association graph, node features are updated through the message passing layer, the specific process is as follows: Message aggregation: Each node i receives messages from its adjacent node set When receiving information, the edge weight and external correction factors is used as a dynamic adjustment factor.

[0067] Node update: Use update formula (6) to aggregate the adjacent node information to node i, and update it based on its own characteristics.

[0068] (3) Dynamic weight optimization: In each round of iteration, the edge weights are adjusted according to changes in the external environment (such as policies and market demand). The specific process is described in the dynamic update process above.

[0069] (4) Supervised learning optimization: Using historical data and expert-annotated demand priorities, supervised training of the GNN model is performed to optimize the prediction accuracy of node priorities. Specific process: Loss function design: Predicting the priority of nodes and the true priority label Calculation error: (8) Optimize the associated weights of the edges, using the edge weight prediction error: (9) Joint optimization loss function: (10) Among them, λ is the balance coefficient, which is used to adjust the weights of the two losses.

[0070] By adjusting and optimizing the loss function, it not only focuses on the prediction accuracy of the nodes, but also enhances the sensitivity to the edge weights, ensuring that the model can better reflect the real business relationship when adjusting the graph structure. Optimizing the loss function not only improves the convergence speed of the graph model, but also significantly enhances its adaptability to new data.

[0071] (5) Output optimization results: Optimized demand association graph: Outputs a demand association graph containing optimized node features and edge weights; Demand priority sorting: based on the final node priority Output demand sorting results; Demand hierarchy: Combined with the optimized association graph, extract the first-level core demand and its related second-level and third-level demands. For example, under the demand of "high proportion of new energy access", it can be further refined into "new energy prediction technology", "distributed energy storage optimization" and "high-precision load regulation technology". These demands are output through the optimized association graph and classified according to different technical directions to provide data support for the construction of the demand framework.

[0072] Framework visualization and result output: Use tools such as Gephi to generate visualization charts of the demand framework, intuitively displaying the structured results and priorities of the demand. Output the final demand structured framework to provide a basis for scientific planning and efficient management for the innovative development of power grid companies.

[0073] The present invention discloses a method and system for extracting innovation needs of power grid enterprises, which are applicable to the full life cycle management of power grid enterprises in technology innovation and resource planning. The present invention constructs a keyword library of power grid innovation needs, scans and analyzes the full amount of innovation needs of power grid enterprises, studies the future technical form of core power grid professions, constructs demand filtering rules, and constructs a demand integration model, and proposes an innovation and efficiency demand framework in the power grid business field. The present invention improves the comprehensiveness and dynamism of innovation demand identification, improves the accuracy and scientificity of innovation demand identification, realizes intelligence, improves efficiency, and supports power grid enterprises in building a new power system for the future.

[0074] Embodiment 2 This embodiment provides an innovation demand extraction system for a power grid enterprise.

[0075] An innovative demand extraction system for power grid enterprises, comprising: The data acquisition and preprocessing module is configured to: acquire multi-source data of innovation needs and build a database; extract keywords from the multi-source data of innovation needs, build a keyword library, and convert them into keyword embedding vectors; A clustering module is configured to: semantically match the keyword embedding vector with the multi-source data of innovation requirements in the database, and calculate the correlation similarity; cluster the multi-source data of innovation requirements corresponding to the correlation similarity greater than a first set threshold, and obtain a first requirement list; The screening module is configured to: define demand screening indicators, use fuzzy logic method to perform weighted comprehensive scoring on all demand screening indicators of the first demand list, and screen the second demand list whose score is greater than a second set threshold; The output module is configured to: construct a demand association graph based on the second demand list, determine the initial weights of the edges in the demand association graph, use the GNN network to optimize the weights of the edges in the demand association graph, sort the priorities of the demand nodes in the optimized demand association graph, and output the demand sorting results.

[0076] In some embodiments, the output module is further configured to: Calculate the semantic similarity between the embedding vectors of demand nodes, quantify the upstream and downstream dependency weights between demand nodes, and calculate the initial weight of the edge between two demand nodes ; In the message passing layer of the GNN network, according to the changes in the external environment, a weight correction mechanism based on external environmental factors is designed to calculate the correction factor between two demand nodes , expressed by the following formula:

[0077] in, , and represent the policy relevance, market trend and technology popularity of demand nodes i and j respectively; γ, δ and They represent the weight coefficients of policy relevance, market trend and technology enthusiasm in the revision process respectively; According to the initial weight of the edge between two demand nodes Correction factor between two demand nodes The sum of , calculates the weight of the edge in the updated demand association graph , the optimized demand association graph has been obtained; The GAT attention mechanism is introduced in the GNN network to distinguish the importance of different neighbor nodes in the optimized demand association graph; according to the importance of the neighbor nodes, the weight information of the corrected edges and the feature information of the neighbor nodes are weighted and aggregated to the corresponding nodes, the feature information of the nodes is updated, the priority of the demand nodes is calculated according to the feature information, the priority of the demand nodes is re-sorted, and the demand sorting results are output.

[0078] In some embodiments, the output module is further configured to: when external environmental changes introduce new demands, calculate the associated weights of new nodes and add nodes or edges to the demand association graph; delete low-priority nodes or edges; and re-order the priorities of demand nodes according to the updated edge weights.

[0079] In some embodiments, the data acquisition and preprocessing module is further configured to: perform semantic classification and supplement on the initially extracted keywords in combination with the opinions of domain experts, add supervision information to the keywords, screen and optimize long-tail keywords in the keyword library; use a neural network model to map the keywords to a low-dimensional vector space and convert them into keyword embedding vectors.

[0080] In some embodiments, the output module is further configured to: parse the second requirement list using a natural language processing tool to extract the associations between the requirements; and construct a requirement association graph with the requirements as nodes and the associations between the requirements as edges.

[0081] In some embodiments, the data acquisition and preprocessing module is further configured to: extract keywords from the multi-source data of innovation needs using a deep learning model based on the multi-source data of innovation needs.

[0082] In some embodiments, the data acquisition and preprocessing module is further configured to: preprocess the acquired multi-source data of innovation requirements, including data cleaning and standardization; In some embodiments, the multi-source data on innovation needs include policy documents, industry trends, patent texts, market research reports, and technical documents of historical corporate projects.

[0083] In some embodiments, the first requirement list includes requirement content, data source, and associated similarity.

[0084] In some embodiments, the screening module is further configured to: define demand screening indicators based on the Drools rule engine, and the demand screening indicators include: technology maturity, market value and policy adaptability.

[0085] Embodiment 3 This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the method for extracting innovative demands of a power grid enterprise as described in the first embodiment above are implemented.

[0086] Embodiment 4 This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for extracting innovative demands of a power grid enterprise as described in the first embodiment above are implemented.

[0087] Embodiment 5 This embodiment provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the method for extracting innovative requirements of a power grid enterprise described in the first embodiment.

[0088] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0089] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0090] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0091] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0092] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0093] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for extracting innovation requirements of a power grid enterprise, characterized in that: include: Obtain multi-source data on innovation needs and build a database; Extract keywords from multi-source data on innovation needs, build a keyword library, and convert it into keyword embedding vectors; Semantically matching the keyword embedding vector with the multi-source data of innovation requirements in the database to calculate the correlation similarity; clustering the multi-source data of innovation requirements corresponding to the correlation similarity greater than a first set threshold to obtain a first requirement list; Demand screening indicators are defined, and a fuzzy logic method is used to perform a weighted comprehensive score on all demand screening indicators of the first demand list, and a second demand list having a score greater than a second set threshold is screened; Based on the second demand list, a demand association graph is constructed, the initial weights of the edges in the demand association graph are determined, the GNN network is used to optimize the weights of the edges in the demand association graph, the priorities of the demand nodes in the optimized demand association graph are sorted, and the demand sorting results are output.

2. The method for extracting innovation needs of power grid enterprises according to claim 1, characterized in that: Determine the initial weight of the edge in the demand association graph, use the GNN network to optimize the weight of the edge in the demand association graph, sort the priority of the demand nodes in the optimized demand association graph, and output the demand sorting result; the method includes: Calculate the semantic similarity between the embedding vectors of demand nodes, quantify the upstream and downstream dependency weights between demand nodes, and calculate the initial weight of the edge between two demand nodes ; In the message passing layer of the GNN network, according to the changes in the external environment, a weight correction mechanism based on external environmental factors is designed to calculate the correction factor between two demand nodes , expressed by the following formula: in, , and represent the policy relevance, market trend and technology popularity of demand nodes i and j respectively; γ, δ and They represent the weight coefficients of policy relevance, market trend and technology enthusiasm in the revision process respectively; According to the initial weight of the edge between two demand nodes Correction factor between two demand nodes The sum of , calculates the weight of the edge in the updated demand association graph , the optimized demand association graph has been obtained; The GAT attention mechanism is introduced in the GNN network to distinguish the importance of different neighbor nodes in the optimized demand association graph; according to the importance of the neighbor nodes, the weight information of the corrected edges and the feature information of the neighbor nodes are weighted and aggregated to the corresponding nodes, the feature information of the nodes is updated, the priority of the demand nodes is calculated according to the feature information, the priority of the demand nodes is re-sorted, and the demand sorting results are output.

3. The method for extracting innovation needs of power grid enterprises according to claim 1, characterized in that: The method for optimizing the demand association graph also includes: when the external environment changes and introduces new demands, calculating the association weights of new nodes and adding nodes or edges to the demand association graph; deleting low-priority nodes or edges; and re-ordering the priorities of demand nodes according to the updated edge weights.

4. The method for extracting innovation needs of power grid enterprises according to claim 1, characterized in that: Convert to keyword embedding vector; the method includes: semantically classifying and supplementing the initially extracted keywords in combination with the opinions of domain experts, adding supervision information to the keywords, screening and optimizing long-tail keywords in the keyword library; using a neural network model to map the keywords to a low-dimensional vector space and convert them into keyword embedding vectors.

5. The method for extracting innovation needs of power grid enterprises according to claim 1, characterized in that: The method of constructing a requirement association graph based on the second requirement list includes: using a natural language processing tool to parse the second requirement list and extract the associations between the requirements; and constructing a requirement association graph with the requirements as nodes and the associations between the requirements as edges.

6. The method for extracting innovation requirements of a power grid enterprise according to claim 1, characterized in that: Extracting keywords from the multi-source data of innovation needs, including: extracting keywords from the multi-source data of innovation needs by using a deep learning model based on the multi-source data of innovation needs; or, Pre-process the acquired multi-source data on innovation needs, including data cleaning and standardization; or, The multi-source data of innovation demand include policy documents, industry dynamics, patent texts, market research reports and technical documents of historical projects of enterprises; or, The first requirement list includes requirement content, data source and correlation similarity; or, Demand screening indicators are defined, and the method includes: based on the Drools rule engine, the demand screening indicators are defined, and the demand screening indicators include: technology maturity, market value and policy adaptability.

7. An innovative demand extraction system for power grid enterprises, characterized in that: include: A data acquisition and preprocessing module is configured to: acquire multi-source data on innovation needs and build a database; Extract keywords from multi-source data on innovation needs, build a keyword library, and convert it into keyword embedding vectors; A clustering module is configured to: semantically match the keyword embedding vector with the multi-source data of innovation requirements in the database, and calculate the correlation similarity; cluster the multi-source data of innovation requirements corresponding to the correlation similarity greater than a first set threshold, and obtain a first requirement list; The screening module is configured to: define demand screening indicators, use fuzzy logic method to perform weighted comprehensive scoring on all demand screening indicators of the first demand list, and screen the second demand list whose score is greater than a second set threshold; The output module is configured to: construct a demand association graph based on the second demand list, determine the initial weights of the edges in the demand association graph, use the GNN network to optimize the weights of the edges in the demand association graph, sort the priorities of the demand nodes in the optimized demand association graph, and output the demand sorting results.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the innovative demand extraction method for a power grid enterprise as described in any one of claims 1 to 6 are implemented.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the innovative demand extraction method for a power grid enterprise as described in any one of claims 1-6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps in the innovative demand extraction method for a power grid enterprise as claimed in any one of claims 1 to 6 are implemented.

Images