Power grid key technology breakthrough point prediction method, equipment and medium

Abstract

The invention relates to a power grid key technology breakthrough point prediction method and device and a medium, and the method comprises the steps: collecting and preprocessing power grid technology related patents or literature texts, dividing data according to a preset time window, and forming a time sequence text set; based on the time sequence text set, obtaining technical topic distribution of the corresponding time window and keywords of all technical topics, and constructing topic evolution characteristics of all the technical topics; based on the topic probability features, considering strategy influence and technology collaborative correction, fitting an S-shaped curve, and obtaining substitution parameters; based on a pre-established topic evolution feature-substitution parameter rule, matching a technical breakthrough point direction, clustering keywords of a technical topic, obtaining breakthrough point sub-directions, quantitatively evaluating the priority of the breakthrough point sub-directions according to technical feasibility, economic feasibility and strategy suitability, and outputting power grid key technical breakthrough point prediction. Compared with the prior art, the method achieves the precise prediction of the key technology breakthrough point of the power grid.

Description

Technical Field

[0001] This invention relates to the fields of text mining and document analysis technology, and in particular to a method, equipment and medium for predicting key technological breakthroughs in power grids. Background Technology

[0002] Against the backdrop of intensifying global technological competition, the power industry, as a core sector of energy transition, urgently needs accurate prediction of key technological breakthroughs to optimize R&D resource allocation, mitigate the risks of technological iteration, and seize the commanding heights of the industry. Determining the direction of breakthroughs in key power grid technologies has become a core requirement for power grid companies in formulating strategic plans and conducting core technology research. However, existing technology prediction methods have significant limitations in adapting to the characteristics of the power grid sector, integrating multi-dimensional information, and outputting actionable breakthroughs, making it difficult to meet actual needs.

[0003] Power grid technology possesses unique attributes such as strong interdisciplinary collaboration, significant policy-driven technological iteration, and long equipment lifecycles. This interdisciplinary nature leads to blurred boundaries between technological themes. Traditional LDA models, lacking an optimized dictionary for the power grid domain, are prone to misclassifying interdisciplinary keywords. Policy interventions can cause short-term fluctuations in technology penetration rates exceeding 30%, and existing models, failing to incorporate policy influence factors, exhibit fitting errors exceeding 15%. Furthermore, existing methods often remain at the level of identifying directions. For example, Chinese patent CN120068882B discloses a method and system for analyzing and predicting thematic trends in scientific literature, which aims to understand the deep semantics of literature, effectively identify key areas of focus, and predict future evolution trends. However, it fails to refine the direction and priority of breakthrough ideas, and only a small portion of the predicted results can be translated into actual R&D projects, resulting in poor feasibility.

[0004] Traditional power grid technology forecasting often employs qualitative methods such as the Delphi method and expert judgment, relying heavily on the experience and knowledge of domain experts to determine the technology stage and breakthrough direction. For example, expert seminars are organized to determine whether a certain energy storage technology has entered the mature stage, or subjective assessments by experts suggest vague breakthrough directions such as material innovation. These methods have two major drawbacks: First, expert knowledge is limited by individual research fields and experience, making it prone to bias and error. For instance, the stage judgment deviation for interdisciplinary collaborative technologies (such as virtual power plants + photovoltaic grid connection) can reach 20%-30%. Second, they cannot quantify the technology's growth potential and the pace of replacement, making it difficult to accurately identify the inflection point of technology replacement (such as the specific time window for the replacement of traditional DC transmission by flexible DC transmission). This leads to a misallocation of R&D investment by power grid companies, either prematurely abandoning still valuable mature technologies or delaying the deployment of high-potential replacement technologies, resulting in an annual waste of billions of dollars in resources.

[0005] In summary, power grid companies urgently need a predictive method that can accurately extract features from massive amounts of data, quantitatively assess potential, and output actionable breakthroughs to support the forward-looking deployment of key technologies and fill the technological gap in predicting breakthroughs in key power grid technologies. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the existing technology by providing a method, device and medium for predicting breakthroughs in key technologies of power grids, thereby achieving accurate prediction of breakthroughs.

[0007] The objective of this invention can be achieved through the following technical solutions: A method for predicting breakthroughs in key technologies for power grids, the method comprising: Collect and preprocess patent or literature texts related to power grid technology, divide the data according to a preset time window, and form a time series text set; Based on the time series text set, the distribution of technical topics and keywords of each technical topic in the corresponding time window are obtained, and the topic evolution features of each technical topic are constructed; the topic evolution features include topic probability features, technology optimization features and technology substitution features; Based on the aforementioned topic probability characteristics, considering the impact of strategies and technical collaborative corrections, an S-shaped curve is fitted to obtain alternative parameters. Based on pre-established theme evolution characteristics-alternative parameter rules, the direction of technological breakthroughs is matched, and the keywords of the technological themes are clustered to obtain breakthrough sub-directions. According to technical feasibility, economic feasibility and strategy adaptability, the priority of breakthrough sub-directions is quantitatively evaluated, and the prediction of key technology breakthroughs in the power grid is output.

[0008] Furthermore, the preset time window is divided by year, and the time series text set is organized by year. The text set in each time window contains all power grid technology-related patents or literature texts for the corresponding year. If the number of texts in a certain time window is less than a preset threshold, it is merged with the adjacent time window with fewer texts.

[0009] Furthermore, the process of obtaining the distribution of the technical topics and the keywords of each technical topic within the corresponding time window includes: Based on the time-series text set, the core text of each patent or document within each time window is encoded to obtain word vectors; By performing average pooling on all word vectors, a fixed-dimensional document semantic vector is generated. For all document vectors within a certain time window, an unsupervised clustering algorithm is used to cluster them by technical topics, obtaining technical topic clusters; each cluster represents a semantically closely related technical topic. All documents in each technical topic cluster are treated as a sub-corpus. The TF-IDF value of each word in the sub-corpus is calculated, and the N words with the highest values ​​are selected as the keywords of the technical topic. The calculation of the TF-IDF value incorporates a technical field weighting correction factor and a time decay factor, and its specific calculation expression is as follows: in, Let be the word frequency of word t in document d. Inverse document frequency, This is a domain weight correction factor. This is the time decay factor; in, The number of times word t appears in document d. This represents the sum of the occurrences of all words in document d. It is the total number of documents in the sub-corpus. It is the number of documents containing the word t. For domain weight coefficients, The weight level of word t in the power grid field professional dictionary s. The time decay coefficient, For the current point in time in the analysis, The time point when word t first appears.

[0010] Furthermore, the topic probability feature is the annual average probability of the technology topic within each time window; the technology optimization feature is calculated by statistically analyzing the proportion of application optimization, efficiency improvement, cost reduction, and reliability enhancement in the keywords of each technology topic; the technology substitution feature is calculated by statistically analyzing the proportion of descriptions of technology substitution, new substitution, and substitution materials in the keywords of each technology topic, as well as the growth rate of the frequency of occurrence of substitution technology topics.

[0011] Furthermore, the S-curve is fitted using the Levenberg-Marquardt nonlinear least squares method. Before fitting the S-curve, the topic probability features are Min-Max standardized and mapped to the interval [0.01, 0.99]. After fitting, the curve fit is verified by residual analysis. The substitution parameters include the substitution inflection point time and substitution progress indicators.

[0012] Furthermore, the fitting formula for the S-curve is: in, Theme In the time window penetration rate Theme The upper limit of penetration saturation, Theme The rate of technological growth Theme The replacement inflection point time, As a strategy influence factor, To revise the theme Technological synergy correction factor with upstream and downstream technologies; in, To be related to the theme A collection of related technical topics, For related topics penetration rate For related topics The upper limit of saturation; The strategy influencing factors The acquisition process includes: collecting relevant power grid technology strategies, generating strategy influence factors based on text sentiment analysis algorithms and preset strategy strength scores.

[0013] Furthermore, in the process of quantitatively evaluating the priority of breakthrough ideas, the Analytic Hierarchy Process (AHP) is used to determine the weights of technical feasibility, economic feasibility, and strategy adaptability, and the priority of breakthrough ideas is evaluated by calculating a comprehensive score. The technical feasibility includes technology maturity, technical barriers, and R&D cycle; The economic feasibility includes the potential for unit cost reduction and the size of market demand. The policy adaptability includes the level of policy support and environmental compliance.

[0014] Furthermore, the process of obtaining the direction of the breakthrough point includes: Using a preset number of high-occurrence keywords of the aforementioned technical topic as input, calculate the Euclidean distance between the keywords; use the method of minimizing intra-class variance to perform hierarchical clustering of the high-occurrence keywords, and determine the number of clusters through a dendrogram; calculate the co-occurrence frequency of keywords within each cluster, and if the co-occurrence frequency is greater than a preset number, the clustering is deemed reasonable; otherwise, adjust the number of clusters and re-cluster to obtain the direction of breakthrough ideas.

[0015] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the power grid key technology breakthrough prediction method as described above.

[0016] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the power grid key technology breakthrough prediction method as described above.

[0017] Compared with the prior art, the beneficial effects of the present invention include: 1. This invention addresses the problems of large subjective bias, poor model adaptability, and insufficient practicality in existing power grid technology breakthrough prediction methods. Through a prediction process involving multi-dimensional feature construction, composite curve fitting, and quantitative priority evaluation, it achieves accurate breakthrough prediction. Power grid technology possesses unique attributes such as strong interdisciplinary collaboration, significant policy intervention, and long equipment lifecycles. This invention incorporates policy influence factors and technology synergy correction factors into the S-curve fitting process, adapting to policy-driven and technology-synergistic characteristics, making breakthrough prediction more realistic. Furthermore, this invention refines breakthrough directions into executable sub-directions, outputting complete breakthrough solutions that can be directly applied to technology breakthrough prediction, demonstrating strong applicability.

[0018] 2. This invention employs a process of text encoding, document vector generation, and unsupervised clustering. First, the core text of the patent or document is encoded to obtain word vectors. Then, fixed-dimensional document semantic vectors are generated through average pooling, avoiding the shortcomings of traditional bag-of-words models that ignore semantic relationships. The invention uses unsupervised clustering to automatically generate technical topics, eliminating the need for manual preset quantities, adapting to the dynamic evolution characteristics of power grid technology, and making semantic quantification more accurate.

[0019] 3. This invention introduces a technical field weight correction factor and a time decay factor into the TF-IDF value calculation to optimize the weight calculation; the technical field weight correction factor distinguishes the level of terms through a professional dictionary in the power grid field to strengthen the weight of core terms and avoid interference from common words; the time decay factor reduces the weight of old terms through exponential decay to ensure that keywords reflect the latest technology trends and solves the problem of traditional TF-IDF emphasizing frequency and neglecting timeliness.

[0020] 4. When fitting the S-shaped curve, this invention introduces a strategy influence factor and a technology synergy correction factor. The strategy influence factor quantifies external strategy intervention through text sentiment analysis and strategy strength scoring. It generates strategy influence factors according to level, timeliness, and relevance, which can reduce fitting error. The technology synergy correction factor corrects the impact of technology synergy on the penetration rate by calculating the penetration rate ratio of the technology theme and its upstream and downstream themes, avoiding the bias of isolated prediction of a single technology and solving the defect of traditional models that ignore technology correlation.

[0021] 5. In the priority evaluation of the breakthrough sub-directions of this invention, the analytic hierarchy process is used to determine the feasibility weight. Multiple sub-indicators such as technology maturity, potential for reducing unit cost, and strength of strategic support are quantitatively scored to calculate the comprehensive score and classify priorities, providing a clear basis for the allocation of R&D resources and avoiding resource misallocation.

[0022] 6. This invention is not limited to a single power grid technology, but can be widely applied to predicting breakthroughs in various key technologies such as virtual power plants, flexible DC transmission, electrochemical energy storage, and photovoltaic grid connection. Attached Figure Description

[0023] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0025] Example 1 This embodiment discloses a method for predicting breakthroughs in key technologies for power grids. The method is as follows: Figure 1 As shown, it includes steps S1-S4, and the specific steps are as follows: Step S1: Collect and preprocess patent or literature texts related to power grid technology, divide the data according to a preset time window, and form a time series text set.

[0026] The preprocessing process includes: To remove duplicate records: The single comparison logic of "title + abstract" has limitations. Therefore, it has been upgraded to a multi-core field combination comparison, using cross-validation of multi-dimensional information to reduce errors in duplicate determination. Patent number, title, abstract, IPC classification number, and applicant are prioritized as core comparison dimensions. Among these, the patent number is the primary criterion; if two records have the same patent number, they are directly determined to be duplicates; if the patent numbers are different, further verification is performed using other fields. For the title and abstract, text similarity analysis is used, combined with technical field relevance determination, and standardized matching of applicant information is also referenced to form a multi-dimensional cross-validation system to ensure the accuracy of duplicate record determination. For records determined to be duplicates, priorities are set based on information completeness and technical value—records published later and with more complete technical information are retained first, while other duplicate records are marked and uniformly removed, with the selection criteria retained.

[0027] Deleting Invalid Data: Clearly define the criteria for core key fields to avoid inconsistent cleaning standards due to ambiguous judgments and ensure the standardization of invalid data removal. The definition of core key fields, based on the technical topic extraction requirements, includes patent number, abstract, IPC classification number, publication date, and claims. The invalid data determination logic is as follows: If a record has missing core fields, the impact of the missing fields on subsequent topic extraction must be assessed. If the missing field is an irreplaceable key information source, or if the absence of multiple core fields renders the record unsupportive for analysis, it is determined to be invalid data. If only non-critical core fields are missing, attempts can be made to supplement them using authoritative patent data sources. If supplementation fails and does not affect the core analysis, it can be marked and retained; otherwise, it should be removed.

[0028] Noise Removal: This requires moving beyond the single definition of noise as "missing abstract + unclear classification." A multi-type noise judgment logic needs to be constructed, taking into account the characteristics of patent data, to prevent noise from interfering with the accuracy of topic extraction. Invalid abstracts include not only missing abstracts but also abstracts lacking actual technical information or completely unrelated to the patent topic, judged through technical information relevance verification. Disorganized classification includes cases where the classification number does not reach the basic level or the classification number does not match the patent's technical topic, verified using the IPC classification system's technical field division rules. Non-technical records refer to non-patent documents or non-technical patents without technical solution descriptions, judged through a dual assessment of patent type attributes and technical topic relevance. Secondary noise verification involves a multi-person independent review mechanism for initially judged noisy records to ensure consistent judgment results and prevent misjudgments that could lead to the loss of valid data.

[0029] To ensure the effectiveness of the data cleaning process, a quality verification system needs to be established from three dimensions: uniqueness, completeness, and accuracy, forming a closed-loop control system for the cleaning process. Uniqueness verification checks for duplicate records after cleaning, ensuring each record is unique and free of redundant information. Completeness verification checks for missing core fields, ensuring no critical information is missing and that records have the information foundation to support subsequent analysis. Accuracy verification uses random sampling to confirm the consistency between classification numbers and technical topics, and the validity of abstract technical information, ensuring the data has no significant bias.

[0030] The preset time window is divided by year, and the time series text set is organized by year. The text set in each time window contains all power grid technology-related patents or literature texts for the corresponding year. If the number of texts in a certain time window is less than the preset threshold, it is merged with the adjacent time window with fewer texts.

[0031] Step S2: Based on the time series text set, obtain the distribution of technical topics and keywords of each technical topic in the corresponding time window, and construct the topic evolution features of each technical topic; the topic evolution features include topic probability features, technology optimization features and technology substitution features.

[0032] The process of obtaining the distribution of technical topics and the keywords of each technical topic within a corresponding time window includes: Based on the time-series text set, the core text of each patent or document within each time window is encoded to obtain word vectors; in this embodiment, a pre-trained Chinese BERT model is used to encode the core text such as the abstract and claims of each patent or document within each time window. By performing average pooling on all word vectors, a fixed-dimensional document semantic vector is generated. For all document vectors within a certain time window, unsupervised clustering algorithms, such as Gaussian Mixture Model (GMM) or HDBSCAN density clustering algorithm, are used to cluster technical topics to obtain technical topic clusters; each cluster represents a semantically closely related technical topic. All documents in each technical topic cluster are treated as a sub-corpus. The TF-IDF value of each word in the sub-corpus is calculated, and the N words with the highest values ​​are selected as the keywords of the technical topic.

[0033] In another embodiment, the keyword extraction process involves calculating the center vector of the cluster; calculating the cosine similarity between the word vector of each word in the vocabulary and this center vector; and selecting the N words with the highest similarity as keywords. This method can extract words that are semantically closest to the core of the topic, even if they do not have a high literal frequency in the document of the cluster.

[0034] The calculation of TF-IDF values ​​incorporates a technical field weighting correction factor and a time decay factor. The specific calculation expression is as follows: in, Let be the word frequency of word t in document d. Inverse document frequency, This is a domain weight correction factor. This is the time decay factor; in, The number of times word t appears in document d. This represents the sum of the occurrences of all words in document d. It is the total number of documents in the sub-corpus. It is the number of documents containing the word t. For domain weight coefficients, The weight level of word t in the power grid field professional dictionary s. The time decay coefficient, For the current point in time in the analysis, The time point when word t first appears.

[0035] In another embodiment, the theme evolution features are constructed by tracking the movement of cluster center vectors and changes in keyword sets across different time windows. For example, if the cluster center vectors of two adjacent years are very close and the keywords have a high degree of overlap, they are considered to be a continuation of the same theme; if the center vectors shift significantly or the keywords change significantly, it may indicate a shift or breakthrough in technology.

[0036] Step S3: Based on the topic probability characteristics, considering the impact of strategies and technical collaborative corrections, fit an S-shaped curve to obtain alternative parameters.

[0037] The probability feature of a theme is the annual average probability of a technology theme within each time window; the technology optimization feature is calculated by statistically analyzing the proportion of application optimization, efficiency improvement, cost reduction, and reliability enhancement in the keywords of each technology theme; the technology substitution feature is calculated by statistically analyzing the proportion of descriptions of technology substitution, new substitution, and substitution materials in the keywords of each technology theme, as well as the growth rate of the frequency of occurrence of substitution technology themes.

[0038] The S-shaped curve was fitted using the Levenberg-Marquardt nonlinear least squares method. Before fitting the S-shaped curve, the topic probability characteristics were standardized using Min-Max and mapped to the interval [0.01, 0.99]. After fitting, the curve fit was verified by residual analysis. Alternative parameters included the substitution inflection point time and substitution progress indicators.

[0039] The fitting formula for the S-curve is: in, Theme In the time window penetration rate Theme The upper limit of penetration saturation, Theme The rate of technological growth Theme The replacement inflection point time, As a strategy influence factor, To revise the theme Technological synergy correction factor with upstream and downstream technologies; in, To be related to the theme A collection of related technical topics, For related topics penetration rate For related topics The upper limit of saturation; Strategy Influence Factor The acquisition process includes: collecting relevant power grid technology strategies, generating strategy influence factors based on text sentiment analysis algorithms and preset strategy strength scores.

[0040] Step S4: Based on the pre-established theme evolution feature-alternative parameter rules, match the direction of technological breakthroughs, cluster the keywords of the technological theme to obtain breakthrough sub-directions, and quantitatively evaluate the priority of breakthrough sub-directions according to technical feasibility, economic feasibility and strategy adaptability, and output the prediction of key technology breakthroughs in the power grid.

[0041] The process of breaking through the direction of ideas includes: Using a preset number of high-occurrence keywords of technical topics as input, calculate the Euclidean distance between keywords; use the method of minimizing intra-class variance to perform hierarchical clustering of high-occurrence keywords, and determine the number of clusters through a dendrogram; calculate the co-occurrence frequency of keywords within each cluster, if the co-occurrence frequency is greater than the preset number, the clustering is considered reasonable, otherwise adjust the number of clusters and re-cluster to obtain the direction of breakthrough ideas.

[0042] In practical applications, the number of clusters is set to 3-5 based on the complexity of the breakthrough point direction. For example, under the material / process innovation direction, lithium-ion batteries, capacity decay and thermal management are clustered into breakthrough point directions of cathode material optimization, electrolyte innovation and thermal management material upgrading. During the clustering process, the clustering effect is verified by the silhouette coefficient. A silhouette coefficient ≥ 0.6 is considered to indicate that the clustering is effective.

[0043] In the process of quantitatively evaluating the priority of breakthrough ideas, the Analytic Hierarchy Process (AHP) is used to determine the weights of technical feasibility, economic feasibility, and strategy adaptability, and the priority of breakthrough ideas is evaluated by calculating a comprehensive score. Technical feasibility includes technology maturity, technical barriers, and R&D cycle; Economic feasibility includes the potential for reducing unit costs and the size of market demand; Strategy adaptability includes the level of policy support and environmental compliance.

[0044] A prediction report is generated when predicting key technological breakthroughs in the output power grid. The prediction report includes: Technical breakthrough points and corresponding sub-breakthrough points (with priority); Technical goals for each sub-direction: For example, the optimization goal for cathode materials is to increase capacity by 30%; Key metrics: such as capacity increasing from 200mAh / g to 260mAh / g; Feasibility assessment details: scores for each dimension and overall score; Short-term (1-2 years), medium-term (3-5 years), and long-term (5-10 years) R&D recommendations for each high-priority sub-direction, clarifying the core tasks and expected results for each stage.

[0045] Example 2 This embodiment is based on the power grid key technology breakthrough prediction method disclosed in Embodiment 1 above, and takes electrochemical energy storage technology as a specific application scenario to disclose the actual implementation process of each step.

[0046] ① Data Acquisition and Preprocessing For the field of power grid technology, the following types of data are collected: Patent text: By searching the State Intellectual Property Office's patent search system and the USPTO database, keywords such as electrochemical energy storage, lithium-ion batteries, and energy storage materials were retrieved, with the time range set from 2010 to 2024. A total of 12,680 relevant patents were obtained, including core fields such as patent name, abstract, and claims.

[0047] Document text: By searching the CNKI and WebofScience databases, journal articles and conference papers corresponding to the above keywords were retrieved, totaling 8350 documents, including fields such as title, abstract, and keywords.

[0048] Data cleaning: Duplicate text (such as different format versions of the same patent) was removed, and unique text was retained in 8,920 patents and 5,630 documents; Filter out non-technical text (such as news reports and product advertisements) and retain technical text through keyword matching (including terms such as technical research and development methods).

[0049] Time window division: The preset time window is the year, forming 15 time series text sets from 2010 to 2024; Check the text volume of each window: The text volume of each year from 2010 to 2012 was less than 500 (320, 410, and 480 respectively), and was merged into a merged window for 2010-2012 according to the rules; the text volume of each window from 2013 to 2024 was greater than 600, and each window was kept as an independent window, ultimately forming 13 effective time windows.

[0050] Text structuring: Extract the core text of the patent / document, i.e., extract the abstract and claims of the patent, and extract the abstract and keywords of the document; use open-source word segmentation tools for Chinese text and NLTK tool for English text to remove stop words and low-frequency words.

[0051] ② Extraction of technical themes and construction of evolutionary features Word vector and document vector generation: Word vectors were trained using the Word2Vec model (Skip-gram algorithm). The corpus consisted of all preprocessed text. The vector dimension was set to 200, the window size to 5, and the number of iterations to 10. Average pooling is performed on all word vectors of a single document to generate a 200-dimensional document semantic vector. For example, the document vector of a silicon-based anode patent is [0.12, -0.08, ..., 0.23].

[0052] Technical topic clustering: For the document vectors of each time window, the K-means unsupervised clustering algorithm is used, where the optimal K value is determined by the silhouette coefficient: K=3 for the 2013 window, K=5 for the 2020 window, and K=6 for the 2024 window. Each cluster represents a technical theme. For example, the 2024 window clustering yielded six themes, including silicon-based anode materials, electrolyte safety, and energy storage system integration.

[0053] Keyword extraction, i.e. TF-IDF calculation: For each topic cluster, calculate the TF-IDF value according to the formula, and select the top 10 words with the highest TF-IDF value for each topic as keywords. For example, the keywords for the topic of silicon-based anode materials are: silicon-based anode, capacity decay, cycle life, nano-coating, volume expansion, etc.

[0054] ③ Construction of theme evolution characteristics Topic probability features: Calculate the annual average probability of a technical topic within each time window: Probability of a topic = Number of documents contained in that topic / Total number of documents in the window; Example: The topic of silicon-based anode materials in 2024 contains 180 documents, with a total of 600 documents in the window, and a probability of 0.3.

[0055] Technical optimization features: A pre-defined optimization terminology library is provided, which includes 30 terms such as efficiency improvement, cost reduction, and reliability enhancement. The percentage of optimization-related terms in the statistical topic keywords: In the topic of silicon-based anode materials, 2 out of 10 keywords are "cycle life improvement" and "cost reduction", accounting for 20%.

[0056] Features of technological substitution: A pre-defined terminology database for alternatives includes 25 terms such as alternative materials, new alternatives, and technological iterations. The percentage of substitution-related terms in the keyword list for graphite anodes is 10%, with one substitution material mentioned. The frequency growth rate of alternative themes is calculated as follows: silicon-based anodes had 80 occurrences in 2023 and 180 occurrences in 2024, with a growth rate of (180-80) / 80 = 125%.

[0057] ④ S-curve fitting and calculation of substitution parameters Data preprocessing: The theme probability features of each window from 2010 to 2024 are standardized by Min-Max and mapped to the interval [0.01, 0.99]. For example, the original value of the theme probability of silicon-based anode materials is in the range of 0.02-0.3, and the original value of 0.3 in 2024 corresponds to the standardized value of 0.99.

[0058] S-curve fitting: The Levenberg-Marquardt algorithm was used to fit the curve. Industry experts assessed that the saturation upper limit K for silicon-based anode materials is 0.9, and they are expected to account for 90% of the energy storage anode material market. The strategy impact factor was calculated based on the data collection strategy, with an average score of 0.42. Therefore, the strategy impact factor... The value is 0.42; for related topics such as energy storage battery systems and electrode manufacturing processes, the penetration rates in 2024 are 0.6 and 0.5, respectively, with upper saturation limits of 0.9 and 0.8, respectively. The technology synergistic correction factor is calculated and obtained. =1.129, obtained by fitting historical data, the growth rate r=0.15 / year, and the replacement inflection point time t0=2028.3.

[0059] The sum of squared residuals = 0.032 < 0.05, indicating a valid fit.

[0060] Replacement inflection point time: 2028.3; Replacement progress indicator: 2024 progress = current penetration rate / saturation limit = 0.3 / 0.9 ≈ 33.3%.

[0061] ⑤ Priority Quantitative Assessment Weight settings: Technical feasibility (40%): Technology maturity (40%), technical barriers (30%), R&D cycle (30%); Economic feasibility (35%): Potential for reducing unit costs (50%), Market demand size (50%); Strategy adaptability (25%): policy support (60%), environmental compliance (40%).

[0062] Perform sub-direction priority determination: High priority (4-5 points): Optimization of silicon-based anode nano-coating; Medium priority (2.5-4 points): Improved cycle life, suppression of volume expansion; Output priority order: Silicon-based anode nanocoating optimization > Cycle lifetime improvement > Volume expansion suppression.

[0063] ⑥ Output of prediction results The final output report predicts key technological breakthroughs in the power grid, including: Technology theme evolution trend: The penetration rate of silicon-based anode materials is expected to reach 45% in 2028 and 85% in 2035; Key breakthrough points and directions: Optimization of silicon-based anode nano-coatings is the highest priority; Implementation Recommendation: Optimize the allocation of 60% of R&D resources for silicon-based anode nano-coatings, complete small-batch production verification within one year, with a target cost reduction of 18% and cycle life increased to 3000 cycles.

[0064] Example 3 Based on Embodiment 1, this embodiment provides an electronic device, including: one or more processors and a memory, wherein the memory stores one or more programs, the one or more programs including instructions for executing the aforementioned method for predicting key breakthroughs in power grid technologies.

[0065] At the hardware level, the electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the aforementioned method for predicting key technology breakthroughs in the power grid. Of course, in addition to software implementation, this invention does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0066] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0067] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0068] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting breakthroughs in key technologies for power grids, characterized in that, The method includes: Collect and preprocess patent or literature texts related to power grid technology, divide the data according to a preset time window, and form a time series text set; Based on the time series text set, the distribution of technical topics and keywords of each technical topic in the corresponding time window are obtained, and the topic evolution features of each technical topic are constructed. The topic evolution features include topic probability features, technology optimization features and technology substitution features. Based on the aforementioned topic probability characteristics, considering the impact of strategies and technical collaborative corrections, an S-shaped curve is fitted to obtain alternative parameters. Based on pre-established theme evolution features-alternative parameter rules, the direction of technological breakthroughs is matched, and the keywords of the technological themes are clustered to obtain breakthrough sub-directions. According to technical feasibility, economic feasibility and strategy adaptability, the priority of breakthrough sub-directions is quantitatively evaluated, and the prediction of key technology breakthroughs in the power grid is output.

2. The method for predicting breakthroughs in key power grid technologies according to claim 1, characterized in that, The preset time window is divided by year, and the time series text set is organized by year. The text set in each time window contains all power grid technology-related patents or literature texts for the corresponding year. If the number of texts in a certain time window is less than a preset threshold, it is merged with the adjacent time window with fewer texts.

3. The method for predicting breakthroughs in key power grid technologies according to claim 1, characterized in that, The process of obtaining the distribution of the technical topics and the keywords of each technical topic within the corresponding time window includes: Based on the time-series text set, the core text of each patent or document within each time window is encoded to obtain word vectors; By performing average pooling on all word vectors, a fixed-dimensional document semantic vector is generated. For all document vectors within a certain time window, an unsupervised clustering algorithm is used to cluster them by technical topics, obtaining technical topic clusters; each cluster represents a semantically closely related technical topic. All documents in each technical topic cluster are treated as a sub-corpus. The TF-IDF value of each word in the sub-corpus is calculated, and the N words with the highest values ​​are selected as the keywords of the technical topic. The calculation of the TF-IDF value incorporates a technical field weighting correction factor and a time decay factor, and its specific calculation expression is as follows: in, Let be the word frequency of word t in document d. Inverse document frequency, This is a domain weight correction factor. This is the time decay factor; in, The number of times word t appears in document d. This represents the sum of the occurrences of all words in document d. It is the total number of documents in the subcorpus. It is the number of documents containing the word t. For domain weight coefficients, The weight level of word t in the power grid field professional dictionary s. The time decay coefficient, For the current point in time in the analysis, The time point when word t first appears.

4. The method for predicting breakthroughs in key power grid technologies according to claim 1, characterized in that, The topic probability feature is the annual average probability of the technology topic within each time window; the technology optimization feature is calculated by statistically analyzing the proportion of application optimization, efficiency improvement, cost reduction, and reliability enhancement in the keywords of each technology topic; the technology substitution feature is calculated by statistically analyzing the proportion of descriptions of technology substitution, new substitution, and substitution materials in the keywords of each technology topic, as well as the growth rate of the frequency of occurrence of substitution technology topics.

5. The method for predicting breakthroughs in key power grid technologies according to claim 1, characterized in that, The S-curve was fitted using the Levenberg-Marquardt nonlinear least squares method. Before fitting the S-curve, the probability features of the topic were standardized using Min-Max and mapped to the interval [0.01, 0.99]. After fitting, the curve fit was verified by residual analysis. The substitution parameters included the substitution inflection point time and substitution progress indicators.

6. The method for predicting breakthroughs in key power grid technologies according to claim 5, characterized in that, The fitting formula for the S-curve is: in, Theme In the time window penetration rate Theme The upper limit of penetration saturation, Theme The rate of technological growth Theme The replacement inflection point time, As a strategy influence factor, To revise the theme Technological synergy correction factor with upstream and downstream technologies; in, To be related to the theme A collection of related technical topics, For related topics penetration rate For related topics The upper limit of saturation; The strategy influencing factors The acquisition process includes: collecting relevant power grid technology strategies, generating strategy influence factors based on text sentiment analysis algorithms and preset strategy strength scores.

7. The method for predicting breakthroughs in key power grid technologies according to claim 1, characterized in that, In the process of quantitatively evaluating the priority of breakthrough ideas, the Analytic Hierarchy Process (AHP) is used to determine the weights of technical feasibility, economic feasibility, and strategy adaptability, and the priority of breakthrough ideas is evaluated by calculating a comprehensive score. The technical feasibility includes technology maturity, technical barriers, and R&D cycle; The economic feasibility includes the potential for unit cost reduction and the size of market demand. The policy adaptability includes the level of policy support and environmental compliance.

8. The method for predicting breakthroughs in key power grid technologies according to claim 1, characterized in that, The process of obtaining the direction of the breakthrough point includes: Using a preset number of high-occurrence keywords of the aforementioned technical topic as input, calculate the Euclidean distance between the keywords; use the method of minimizing intra-class variance to perform hierarchical clustering of the high-occurrence keywords, and determine the number of clusters through a dendrogram; calculate the co-occurrence frequency of keywords within each cluster, and if the co-occurrence frequency is greater than a preset number, the clustering is deemed reasonable; otherwise, adjust the number of clusters and re-cluster to obtain the direction of breakthrough ideas.

9. An electronic 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 computer program, it implements the steps of the power grid key technology breakthrough prediction method as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for predicting breakthroughs in key technologies of power grids as described in any one of claims 1-8.

Images