Overhead transmission line special construction scheme key technology cue word generation method based on work decomposition structure

Abstract

The invention discloses an overhead transmission line special construction scheme key technology cue word generation method based on a work decomposition structure. The method comprises the steps that text analysis based on knowledge enhancement is carried out, deep semantic analysis is carried out on an overhead transmission line special construction project summary, and text analysis input cue words are generated; based on the construction key technology in the work decomposition structure, using a semantic extraction algorithm to generate construction key technology input cues; and evaluating the similarity of the two groups of input cue words and the fixed template by adopting a similarity analysis method, and completing the combination of the two groups of input cue words by means of semantic matching and alignment processing, thereby obtaining the final input cue word of the special construction scheme of the overhead transmission line. According to the method, the comprehensiveness, accuracy and pertinence of the input cue word can be improved, so that the finally generated special construction scheme of the overhead transmission line is more comprehensive and reliable.

Description

Technical Field

[0001] This invention relates to a method for generating prompt words, specifically a method for generating key technical prompt words for overhead transmission line construction schemes based on work breakdown structures, belonging to the field of engineering design and construction technology. Background Technology

[0002] In recent years, the construction scale of overhead transmission lines has been expanding year by year, and the construction of ultra-high voltage lines has significantly improved the power grid coverage and transmission capacity. However, with the expansion of scale, for specialized construction projects crossing rivers, mountains, railways, and national highways, it is often necessary to take into account factors such as natural conditions (e.g., hydrology, geology, slope), human conditions (e.g., traffic control, ecological protection), and policy and legal restrictions (e.g., approval processes), which leads to an increase in the complexity, special nature, and workload of the construction.

[0003] In the generation of overhead transmission line construction plans based on generative AI (Artificial Intelligence), specific plans are often generated directly from only the project overview. However, the project overview is often too simplistic and lacks comprehensive and accurate input prompts. These limitations are mainly reflected in the following three aspects: 1) The project overview mainly describes the basic requirements and general information of the project, but lacks key construction information and technologies that need attention; 2) The prompts are not targeted enough, and there is a lack of description of abnormal environments and technical difficulties included in the construction scenario of overhead transmission line construction; 3) The information contained in the project overview is vague and easily overlooks some key construction links and information. Due to these limitations, construction plans generated directly from the project overview often require a lot of subsequent supplementation and improvement, increasing the complexity and time cost of construction preparation, and may also affect the efficiency and quality of construction. Therefore, for the generation of overhead transmission line construction plans, more comprehensive and detailed input prompts are needed to ensure the accuracy and completeness of the plan, thereby better guiding actual construction operations.

[0004] The design of input prompts generated by existing overhead transmission line construction schemes still has the following defects and shortcomings in today's complex application environments:

[0005] First, there are problems in generating overhead transmission line construction plans based on project overviews: Given the characteristics of plan generation, the key to the final plan lies in the design of input prompts. Traditional methods often directly use the project overview as input, which suffers from inaccurate and incomplete input information. This results in an incomplete structure, unclear content, and ambiguous construction objectives in the final overhead transmission line construction plan. Furthermore, the design of prompts fails to adequately consider the dynamic and uncertain nature of the project. For example, in existing technologies, such as the intelligent transmission line construction method based on precise line length deployment disclosed in CN115085089A, the focus is on precise deployment and sag control using intelligent line length measurement equipment. While this improves the accuracy and efficiency of the stringing process and reduces high-altitude work in certain scenarios, the source of input information and the generation of prompts still rely on the project's basic overview and the measurement of specific physical parameters. This fails to address the fundamental problem of how to automatically and comprehensively extract and integrate key technical elements from complex project texts and multi-dimensional construction environments in the early stages of plan generation to form high-quality input prompts. When faced with specialized construction projects that require comprehensive consideration of multiple constraints such as geology, hydrology, and traffic control, the input information for this method still appears to be limited and insufficient.

[0006] Secondly, there is a problem with the weak correlation between the design of input prompts for generating special construction schemes for overhead transmission lines and specific key construction technologies: When designing input prompts for generating special construction schemes for overhead transmission lines, there is insufficient consideration of key technologies within the work breakdown structure of the construction task, resulting in a lack of interrelationships. Traditional input prompt design methods do not fully consider the key technical elements within the work breakdown structure. Specifically, key information such as construction methods and steps is not effectively integrated, leading to an inaccurate and specific generated construction scheme regarding the specific content of construction methods, lacking relevance. Furthermore, some existing technologies dedicated to intelligent construction processes or scheme generation, such as the mechanized construction method for overhead transmission lines disclosed in publication number CN120724714A, are also problematic. Figure 1The intelligent data generation method integrates multi-source data (such as digital orthophoto models and elevation data) and uses multi-objective optimization algorithms to plan construction paths and generate road network overview maps, feature overview maps, and equipment overview maps. Although this method has made progress in visualizing and planning specific construction elements using 3D scenes and optimization algorithms, its technical approach focuses on the macro-configuration of geospatial data and equipment resources. It does not delve into how to deeply analyze and extract key technical features such as "construction methods, technical points, and key steps" from the work breakdown structure (WBS) of the construction project and transform them into core prompts for generating guidance schemes. Therefore, the generated schemes may lack in-depth correlation and precise guidance at the level of construction process and key technological steps. Summary of the Invention

[0007] The purpose of this invention is to provide a method for generating key technology prompts for overhead transmission line construction schemes based on a work breakdown structure (WBS) to address at least one of the aforementioned technical problems. This method comprehensively utilizes text analysis and key technologies from the WBS. By integrating these two aspects of information, the resulting input prompts are more applicable and targeted. This significantly improves the comprehensiveness and accuracy of generating overhead transmission line construction schemes, providing strong support for optimizing construction plans.

[0008] This invention achieves the above objective through the following technical solution: a method for generating key technical keywords for overhead transmission line construction schemes based on work breakdown structures, wherein the generation of key technical keywords includes the following steps:

[0009] S1. Based on knowledge-enhanced text analysis, generate text analysis input prompts for overhead transmission line construction projects;

[0010] S2. Based on the key construction technologies in the work breakdown structure, use the Text-Rank model to generate input prompts for key construction technologies.

[0011] S3. The similarity analysis method is used to evaluate the similarity between the two sets of input prompts and the fixed template. Then, the semantic matching and alignment processing are used to combine the two sets of input prompts to obtain the final input prompts for the special construction plan of overhead transmission lines.

[0012] As a further aspect of the present invention: In S1, when generating text analysis input prompts, a knowledge enhancement model is constructed that includes a Transformer-based text analysis model and a knowledge graph based on the overhead power transmission line construction field.

[0013] As a further aspect of the present invention: In S1, the generation of text analysis input prompts specifically includes the following steps:

[0014] S11. Data Preparation and Preprocessing: Collect relevant project overview data on overhead transmission line construction schemes and professional knowledge in the field of overhead transmission line construction. After preprocessing the collected data, transform it into word vectors that are readable by the text analysis model for subsequent text analysis and processing. At the same time, integrate the collected professional knowledge in the field of overhead transmission line construction to construct a knowledge graph for subsequent data augmentation.

[0015] S12. Data-enhanced model fine-tuning based on knowledge graph of overhead transmission line construction: The text analysis model is fine-tuned and trained using the data obtained from the previous data preparation and preprocessing. At the same time, the knowledge graph of overhead transmission line construction obtained earlier is integrated into the training of the text analysis model, so that the trained text analysis model can perform better in the task of extracting text features and key information in the special construction field of overhead transmission lines, and improve the text analysis model's understanding of the professional terms and construction logic contained therein.

[0016] S13. Input Prompt Generation Stage: After training the text analysis model, the input overview of the overhead transmission line construction project is analyzed. Based on the key information extracted by the model, combined with domain knowledge and construction specifications, specific prompts are generated to guide the generation of construction plans.

[0017] As a further aspect of the present invention: In S11, the collected data includes, but is not limited to, design specifications, typical construction methods, risk databases and historical cases for special construction of overhead transmission lines, and data preprocessing includes, but is not limited to, data cleaning, text analysis and labeling of professional terms.

[0018] As a further aspect of the present invention: In S2, the generation of input prompts for key construction technologies specifically includes the following steps:

[0019] S21. Preprocessing of work breakdown structure data for special construction of overhead transmission lines: dividing the text paragraphs in the work breakdown structure and identifying the key technologies contained therein.

[0020] S22. The Text-Rank model is fine-tuned and trained using the preprocessed data to generate prompt words for the special construction plan of overhead transmission lines. The prompt words generated by this model are more accurate and targeted.

[0021] S23, Output of input prompts for key construction technologies, including requirements for construction methods and steps in key construction technologies.

[0022] As a further aspect of the present invention: S21, the key technologies include construction methods, construction steps, construction resources, and technical parameters.

[0023] As a further aspect of the present invention: In S3, a similarity analysis method is used to evaluate the similarity between the two sets of input prompts and the fixed template based on a cosine similarity calculation method; the combination of the two sets of input prompts is completed by semantic matching and alignment processing using a Transformer-based semantic extraction algorithm.

[0024] As a further aspect of the present invention: In S3, the method for calculating cosine similarity is as follows:

[0025] Transform the two sets of input prompt words into two sets of word vectors, denoted as input prompt word vector A and input prompt word vector B, respectively. The cosine similarity between the two sets of word vectors can then be represented by the ratio of the inner product of input prompt word vectors A and B to the product of their magnitudes. The calculation expression is as follows:

[0026] ;

[0027] Received The numerical range of is [-1, 1]. The specific meaning of the value is that the closer it is to 1, the more similar it is, and vice versa.

[0028] As a further aspect of the present invention: In S3, the semantic extraction algorithm based on Transformer specifically includes:

[0029] By leveraging the cross-attention mechanism introduced by the Transformer decoder, the two sets of sentences are encoded separately to obtain the semantic vector of each word;

[0030] Cross-attention is used to calculate the attention weights between one set of sentences and another set of sentences, thereby finding the most semantically relevant words or phrases and achieving fine-grained semantic alignment.

[0031] The semantically aligned input prompts are used to input the prompts generated for the overhead transmission line construction plan according to the fixed input prompt template.

[0032] The beneficial effects of this invention are:

[0033] 1) This invention addresses the problem of inaccurate and incomplete information in traditional methods that directly use project overviews as input prompts. It proposes a knowledge-enhanced text analysis design. By constructing a professional knowledge graph of overhead transmission line construction and fine-tuning the Transformer model on this basis, the invention performs deep semantic analysis on the project overview text. This method can accurately extract text content related to the transmission line construction project, paying particular attention to the characteristics of the construction scenario, and transforming it into highly targeted and practical input prompts. This effectively guides the formulation of construction plans and improves construction efficiency and quality.

[0034] 2) To address the problem of insufficient correlation between traditional input prompts and key construction technologies, this invention designs an input prompt framework based on key technologies in the Work Breakdown Structure (WBS). By using a Text-Rank model to extract semantics from the work breakdown structure of overhead transmission line construction, it accurately captures key information such as construction methods, steps, and resource requirements, and thereby generates prompts that are highly correlated with construction technologies. This input prompt design ensures that the generation of methods, steps, and other parts in the subsequent construction plan is more accurate and comprehensive, effectively improving the quality of the construction plan.

[0035] 3) This invention addresses the potential data redundancy between two sets of input prompts by proposing a fixed template-based input prompt design method. It utilizes cosine similarity analysis to evaluate the semantic similarity of the two sets of prompts, identifying duplicate or similar content. Based on this, semantic matching and alignment operations are used to remove duplicates, and the unique content of each set of prompts is matched and integrated to ensure a more comprehensive final prompt. Finally, the fixed template standardizes the prompt format, making it easier to manage and use, thus improving the efficiency and quality of construction plan generation.

[0036] 4) From a performance perspective, this invention comprehensively utilizes Transformer technology from the field of deep learning, knowledge-enhanced text analysis technology, Text-Rank model, and semantic similarity analysis and semantic matching methods. It can efficiently and accurately design prompt words for overhead transmission line construction schemes. It can quickly analyze the key technical requirements in the project overview and work breakdown structure, providing comprehensive reference for prompt word generation. During the prompt word generation process, similarity analysis and semantic matching methods can effectively distinguish the repetition of different prompt words and merge and remove repetitive parts, resulting in more accurate and comprehensive final prompt words, significantly improving the efficiency and accuracy of input prompt word generation.

[0037] 5) From a cost perspective, traditional methods of generating input prompts often directly use project overviews or rely on manual labor and expert experience to extract them. This approach not only requires a large number of professionals to invest time and effort but may also lead to increased labor costs and significant resource consumption. In contrast, the input prompt framework designed in this invention can efficiently generate input prompts after model training, greatly reducing reliance on manual labor. The final generated prompts only require review and confirmation by a small number of professionals, eliminating the need for a large number of people to spend a lot of time and effort during the generation process, thus significantly reducing cost input and achieving efficient and low-cost prompt generation.

[0038] 6) From a safety and reliability perspective, the input prompt word generation framework designed in this invention has significant advantages. By combining deep learning, semantic analysis, and knowledge enhancement technologies, it accurately captures key safety information in construction texts, integrates domain knowledge and experience, and generates comprehensive and accurate prompt words. Compared with traditional manual summarization, it reduces human error. Through semantic matching and fusion optimization, it removes duplicate prompts, making the expression of safety requirements more accurate and easy to understand. The final input prompt words can be put into use after being reviewed by a small number of professionals, which not only ensures construction safety but also reduces labor costs. Attached Figure Description

[0039] Figure 1 A schematic diagram of the design framework for generating input prompts for the special construction scheme of overhead transmission lines in this invention;

[0040] Figure 2 A schematic diagram illustrating the input of prompt words for the knowledge-enhanced text analysis-based power transmission line construction scheme in this invention;

[0041] Figure 3 This is a schematic diagram of the input prompts for key construction technologies in the work breakdown structure based on power transmission line construction in this invention;

[0042] Figure 4 A schematic diagram for inputting prompt words for the special construction scheme of overhead transmission lines based on fixed templates in this invention;

[0043] Figure 5 This is a flowchart of the process for generating input prompts for the overhead transmission line construction scheme of this invention. Detailed Implementation

[0044] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Example 1: Generally speaking, the construction of overhead power transmission lines is divided into specialized construction and conventional construction. The distinction between the two is mainly based on the complexity and special nature of the construction environment and the amount of construction work. Specifically, conventional construction usually refers to situations where the terrain is flat, transportation is convenient, construction conditions are standardized, and no special treatment is required. Specialized construction, on the other hand, includes complex terrains or special environments such as crossing rivers, mountains, railways, and national highways. These scenarios often require additional consideration of natural conditions (such as a height difference of more than 20m or a slope of more than 15 degrees in the construction area, or construction located in an earthquake-prone area), human conditions (such as crossing railways, national highways, or other traffic-controlled areas), and policy and legal restrictions (such as approval requirements related to environmental protection, ecological protection, and legal restrictions in the construction area).

[0046] When undertaking specialized construction tasks for overhead transmission lines, the aforementioned factors must be fully considered, leading to a significant increase in the workload of generating corresponding construction plans, which traditional methods struggle to meet. Therefore, the design of input prompts for generating specialized construction plans for overhead transmission lines must fully consider the scenario characteristics involved in the specialized construction. Furthermore, to address the technical issues in the design of prompts for generating specialized construction plans for overhead transmission lines, a method for generating key technology prompts for specialized construction plans based on a work breakdown structure (WBS) is proposed to improve the detail and practicality of the construction plans. This key technology prompt generation method comprises three parts: 1) Text analysis employs a knowledge-enhanced approach to conduct an in-depth analysis of the project overview, extracting information crucial to the generation of specialized construction plans for overhead transmission lines, providing a foundation for subsequent prompt generation; 2) Key technologies in the work breakdown structure focus on extracting feature information such as construction methods, technical points, and key steps from the decomposition structure of the construction task, thereby determining the prompts for specialized construction of overhead transmission lines; 3) Through a fixed prompt template, the key information obtained from text analysis is combined with the key technical features extracted based on the work breakdown structure to generate comprehensive prompts. These prompts are designed to provide more targeted and instructive input for the generation of specialized construction plans for overhead power transmission lines, thereby improving the comprehensiveness and accuracy of the plans.

[0047] like Figure 1As shown, this embodiment provides a method for generating key technology prompts for overhead transmission line construction schemes based on work breakdown structures. This method first uses a Transformer-based text analysis model with knowledge enhancement to generate specific input prompts. Secondly, based on the key construction technologies in the work breakdown structure of overhead transmission line construction, a Text-Rank model is used to generate targeted prompts for specific construction methods and steps. Finally, the input prompts obtained from the two methods are combined according to a fixed template to obtain the final input prompts. The method includes the following steps:

[0048] First, such as Figure 2 As shown, based on knowledge-enhanced text analysis, text analysis input prompts are generated for overhead power transmission line construction projects.

[0049] The knowledge-enhanced text analysis prompts for overhead transmission line construction plans consist of two parts: a Transformer-based text analysis model and knowledge enhancement based on a knowledge graph in the field of overhead transmission line construction. Generating the text analysis input prompts is achieved through the following steps:

[0050] 1) Data Preparation and Preprocessing: Collect relevant project overview data of overhead transmission line construction schemes in various scenarios, as well as professional knowledge in the field of overhead transmission line construction. Perform appropriate preprocessing on the collected data (relevant project overview data of overhead transmission line construction schemes), such as data cleaning, text analysis, and professional term labeling. Then, transform it into word vectors readable by the text analysis model (Transformer-based text analysis model) for subsequent text analysis processing. At the same time, integrate the collected professional knowledge in the field of overhead transmission line construction to build a knowledge graph for subsequent data augmentation. Through data augmentation, input prompt words applicable to specific overhead transmission line construction schemes can be obtained, enabling the text analysis model to perform better in the task of extracting text features and key information in the field of specific overhead transmission line construction, and improving the text analysis model's understanding of professional terms and construction logic.

[0051] The preprocessed data is transformed into word vectors readable by the text analysis model through encoding, completing the text vector encoding. The encoded word vectors are represented as follows:

[0052] ;

[0053] Where Token represents the encoded word vector, and E represents the encoding weight matrix. This represents a 0-1 vector group where the t-th bit is 1;

[0054] After encoding the text vectors, positional encoding is performed in the text analysis model, which involves standardizing the data length of the word vectors in the text.

[0055]

[0056] Where L represents the final length of the text vector. This represents the actual length of the text vector. This indicates the preset maximum length of the text vector; when When the text is long enough, it needs to be padded with zeros to keep the final length consistent; otherwise, it needs to be truncated.

[0057] 2) Data-enhanced model fine-tuning based on knowledge graph of overhead transmission line construction: The text analysis model is fine-tuned and trained using the data obtained from the previous data preparation and preprocessing. At the same time, the knowledge graph of overhead transmission line construction obtained earlier is integrated into the training of the text analysis model, so that the trained text analysis model can perform better in the task of extracting text features and key information in the special construction field of overhead transmission lines, and improve the text analysis model's understanding of the professional terms and construction logic contained therein.

[0058] 3) Input prompt word generation stage: After completing the training of the text analysis model, the text analysis is performed on the input overview of the overhead transmission line construction project. Based on the key information extracted by the text analysis model, combined with domain knowledge and construction specifications, specific prompt words are generated to guide the generation of construction plans.

[0059] A knowledge-enhanced text analysis framework can effectively improve the quality of prompts for overhead transmission line construction plans. Data preparation and preprocessing ensure the basic quality of the text analysis model training; knowledge graphs enhance the model's understanding capabilities, enabling it to accurately grasp the professional details of the construction field; finally, the fine-tuned text analysis model can generate highly targeted and practical prompts based on the project overview, effectively guiding the formulation of construction plans, thereby improving construction efficiency and quality, and providing solid technical support for transmission line engineering construction.

[0060] Second, such as Figure 3 As shown, based on the key construction technologies in the work breakdown structure, the Text-Rank model is used to generate input prompts for key construction technologies.

[0061] To address the issue of insufficient correlation between the input prompts generated for overhead transmission line construction plans and the key technologies such as construction methods and steps in the work breakdown structure (WBS), this step provides a design framework for input prompts based on the key construction technologies in the WBS of overhead transmission line construction. It reverse-engineers the corresponding input prompts by analyzing the key construction technologies, main steps, and specific details of construction resources within the WBS hierarchical structure. The main implementation method is to use a Text-Rank model to extract the core information of the key construction technologies in the WBS of the overhead transmission line construction task. The specific implementation steps are as follows:

[0062] 1) Preprocessing of work breakdown structure data for special construction of overhead transmission lines: dividing the text paragraphs in the work breakdown structure and distinguishing logical paragraphs containing key construction technologies, such as specific construction tasks, construction methods and construction resources in some special scenarios.

[0063] 2) The Text-Rank model is fine-tuned and trained using the preprocessed data to generate prompts for power transmission line construction plans. The prompts generated by this Text-Rank model are more accurate and targeted.

[0064] The semantics of text word vectors containing key construction technologies are extracted using the Text-Rank model. The extracted results are then used to back-map project implementation conditions (such as machinery type, personnel configuration, and environmental constraints). Based on this, concise and targeted prompts are generated. The semantic extraction includes the following steps:

[0065] 21) Text content mapping: Divide different types of text content in key construction technologies, connect keywords and phrases in the same paragraph, level, and line together, and thus transform the text content into a network of word vectors;

[0066] 22) Text-Rank Model Iteration: The obtained "word network" is segmented and dwell words are removed. A sliding window with a fixed window length is used to truncate the text segments, and it is determined whether each pair of words is in the same window. Then, the weight coefficient is obtained by dividing by the number of windows. The specific calculation formula is as follows:

[0067]

[0068] in, ;

[0069] 23) By iterating through multiple loops to score important word vectors, we can obtain important words and phrases (high-scoring words and phrases) that contain key construction technologies. The specific iterative formula is as follows:

[0070]

[0071] 3) After obtaining high-scoring words and phrases, the final key construction technology input prompts are obtained through reverse mapping. These key construction technology input prompts mainly include the requirements for construction methods and steps in the key construction technologies.

[0072] The key construction technology input prompts include three types: construction-specific task prompts (descriptions of key construction task arrangements and requirements), construction-recommended construction method prompts (descriptions of recommended construction methods and key process requirements), and construction resource allocation prompts (descriptions of construction resource arrangements and allocation requirements).

[0073] In summary, this step effectively addresses the issue of insufficient correlation between input prompts and key technologies such as construction methods and procedures. Through a three-step process—data preprocessing, Text-Rank model fine-tuning, and final prompt output—core information from key construction technologies is accurately extracted, generating highly targeted and practical input prompts. This provides strong support for the development of power transmission line construction plans, enhancing the scientific rigor and operability of the plans.

[0074] Third, such as Figure 4 As shown, similarity analysis is used to evaluate the similarity between the two sets of input prompts and the fixed template. Semantic matching and alignment are then used to combine the two sets of input prompts to obtain the final input prompts for the overhead transmission line construction plan.

[0075] Two sets of input prompts are obtained through the "first" and "second" steps respectively. These two sets of input prompts may have overlaps or complements. In order to fully reflect the characteristics of each set of prompts, this step designs a design framework for input prompts for overhead transmission line construction schemes based on a fixed template. The similarity analysis method is used to evaluate the similarity between the two sets of input prompts and the fixed template. Then, semantic matching and alignment processing are used to combine the two sets of input prompts, thereby obtaining a more accurate and comprehensive set of input prompts for overhead transmission line construction schemes.

[0076] This step mainly consists of two parts: one part is to evaluate the similarity between the two sets of input prompts and the fixed template using similarity analysis, which is based on cosine similarity calculation; the other part is to combine the two sets of input prompts using semantic matching and alignment, which is based on the Transformer semantic extraction algorithm.

[0077] Among them, methods based on cosine similarity calculation include:

[0078] 1) Transform the two sets of input prompt words into two sets of word vectors, denoted as input prompt word vector A and input prompt word vector B respectively. The cosine similarity between the two sets of word vectors can be represented by the ratio between the inner product of input prompt word vector A and input prompt word vector B and the product of the magnitudes of the two vectors. The calculation expression is as follows:

[0079]

[0080] Received The numerical range of the similarity index is [-1, 1]. Specifically, a value closer to 1 indicates greater similarity, and a value further away indicates less similarity. Through similarity analysis, duplicate or similar prompts can be efficiently filtered out, reducing unnecessary processing steps. This enables more accurate content fusion and optimization when dealing with complex semantic relationships, thereby improving the efficiency and accuracy of the entire semantic extraction process.

[0081] Transformer-based semantic extraction algorithms include:

[0082] 1) By using the cross-attention mechanism introduced by the Transformer decoder, the two sets of sentences are encoded separately to obtain the semantic vector of each word;

[0083] 2) Utilize cross-attention to calculate the attention weights between one set of sentences and another set of sentences, thereby finding the most semantically relevant words or phrases and achieving fine-grained semantic alignment;

[0084] 3) Input the semantically aligned input prompts according to the fixed input prompt template to generate prompts for the overhead transmission line construction plan.

[0085] Working process and principle: such as Figure 5 As shown, to achieve accurate and efficient design of input prompts for overhead transmission line construction schemes, this solution combines deep learning, data similarity analysis, semantic extraction, and matching alignment techniques. It proposes a method for generating input prompts based on knowledge-enhanced text analysis and key technologies in the work breakdown structure. The solution mainly consists of three parts: a design framework for input prompts for overhead transmission line construction schemes based on knowledge-enhanced text analysis; a design framework for input prompts for overhead transmission line construction schemes based on key technologies in the work breakdown structure; and a design framework for input prompts for overhead transmission line construction schemes based on a fixed template. The specific implementation process of this solution is as follows:

[0086] 1. Collect specialized knowledge (design specifications, typical construction methods, risk database) and historical cases related to typical overhead power transmission line construction projects, including railway crossings, river crossings, and mountain crossings. Construct a knowledge graph encompassing project overview, key construction technologies, and resource allocation. Use this knowledge graph to augment the original training corpus, expanding the training dataset. Based on the expanded dataset, fine-tune the pre-trained model for overhead power transmission line construction, enabling it to understand semantics within this domain. Simultaneously, construct a database containing the work breakdown structure of overhead power transmission line construction and its key construction technologies for fine-tuning the Text-Rank model to ensure accurate extraction of construction tasks, methods, and resource requirements, laying a data foundation for subsequent prompt word generation.

[0087] 2. Parallel generation of input prompts based on text analysis of the overhead transmission line construction project overview and semantic extraction of the work breakdown structure (WBS). The project overview text is preprocessed using a knowledge-enhanced text analysis model, transforming it into word vectors readable by the model. The model then identifies key textual information representing transmission line construction, completes and refines semantics using a knowledge graph, and outputs input prompts in a four-tag format: "Construction Scene - Construction Task - Construction Method - Precautions". The input WBS is first extracted, containing content on construction methods, steps, resources, and technical parameters. After preprocessing, a "word network" is constructed. A Text-Rank model is used for iterative calculation to select high-scoring keyword vectors. Finally, the extracted results are used to back-map project implementation conditions (such as machinery type, personnel configuration, and environmental constraints), outputting input prompts in a three-tag format: "Construction-Specific Task Requirements - Construction-Specific Work Requirements - Construction Resource Configuration Requirements". The input prompts obtained through this approach ensure that the final generated overhead transmission line construction plan is more accurate and comprehensive.

[0088] 3. The two sets of input prompts obtained earlier are preprocessed and converted into two sets of word vectors. Then, word vector similarity analysis is performed on the two sets of input prompts. The cosine similarity calculation formula is used to calculate the corresponding data. The similarity is judged by the calculation results. A certain threshold is set as the standard for removing duplicate data. Then, the semantics of the two sets of input prompts are extracted through the corresponding model. The extracted semantic information is matched and aligned. The final design of the input prompts for the special construction plan of overhead transmission lines is completed according to the fixed input prompt template.

[0089] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0090] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for generating key technical prompts for overhead transmission line construction schemes based on work breakdown structure, characterized in that, The method for generating key technology prompts includes the following steps: S1. Based on knowledge-enhanced text analysis, generate text analysis input prompts for overhead transmission line construction projects; S2. Based on the key construction technologies in the work breakdown structure, use the Text-Rank model to generate input prompts for key construction technologies. S3. The similarity analysis method is used to evaluate the similarity between the two sets of input prompts and the fixed template. Then, the semantic matching and alignment processing are used to combine the two sets of input prompts to obtain the final input prompts for the special construction plan of overhead transmission lines.

2. The key technology prompt word generation method according to claim 1, characterized in that: In step S1, when generating text analysis input prompts, a knowledge enhancement model is constructed that includes a Transformer-based text analysis model and a knowledge graph based on the overhead power transmission line construction domain.

3. The method for generating key technical prompts according to claim 2, characterized in that: In step S1, the generation of text analysis input prompts specifically includes the following steps: S11. Data preparation and preprocessing: Collect data on overhead transmission line construction schemes and professional knowledge in the field of overhead transmission line construction. After preprocessing the collected data, transform it into word vectors that are readable by the text analysis model for subsequent text analysis and processing. At the same time, integrate the collected professional knowledge in the field of overhead transmission line construction to construct a knowledge graph for subsequent data augmentation. S12. Data-enhanced model fine-tuning based on knowledge graph in the field of overhead transmission line construction: The text analysis model is fine-tuned and trained using the data obtained from the data preparation and preprocessing, and the knowledge graph is integrated into the training of the text analysis model. S13. Input Prompt Generation Stage: After completing the training of the text analysis model, the input overview of the overhead transmission line construction project is analyzed. Based on the key information extracted by the text analysis model, combined with domain knowledge and construction specifications, specific prompts are generated to guide the generation of construction plans.

4. The key technology prompt word generation method according to claim 3, characterized in that: In S11, the collected data includes, but is not limited to, design specifications, typical construction methods, risk databases and historical cases for special construction of overhead transmission lines. Preprocessing includes, but is not limited to, data cleaning, text analysis and labeling of professional terms.

5. The key technology prompt word generation method according to claim 1, characterized in that, In step S2, the generation of input prompts for key construction technologies specifically includes the following steps: S21. Preprocessing of work breakdown structure data for special construction of overhead transmission lines: dividing the text paragraphs in the work breakdown structure and distinguishing key construction technologies. S22. The Text-Rank model is fine-tuned and trained using the preprocessed data, thereby enabling the Text-Rank model to generate prompt words for special construction plans of overhead transmission lines; S23. Output of key construction technology input prompts, wherein the key construction technology input prompts include the requirements for construction methods and steps in the key construction technologies.

6. The method for generating key technical prompts according to claim 1, characterized in that: The key technologies mentioned in S21 include construction methods, construction steps, construction resources, and technical parameters.

7. The method for generating key technical prompts according to claim 1, characterized in that: In step S3, a similarity analysis method is used to evaluate the similarity between the two sets of input prompts and the fixed template, based on a cosine similarity calculation method; the combination of the two sets of input prompts is completed by semantic matching and alignment processing using a Transformer-based semantic extraction algorithm.

8. The method for generating key technical prompts according to claim 7, characterized in that: The method based on cosine similarity calculation includes: The two sets of input prompt words are transformed into two sets of word vectors, denoted as input prompt word vector A and input prompt word vector B, respectively. The cosine similarity between the two sets of word vectors is then expressed as the ratio of the inner product of input prompt word vectors A and B to the product of their magnitudes. The calculation expression is as follows: ； Received The numerical range of is [-1, 1]. The closer the value is to 1, the more similar the two are, and vice versa.

9. The method for generating key technical prompts according to claim 7, characterized in that: In S3, the semantic extraction algorithm based on Transformer specifically includes: By leveraging the cross-attention mechanism introduced by the Transformer decoder, the two sets of sentences are encoded separately to obtain the semantic vector of each word; Cross-attention is used to calculate the attention weights between one set of sentences and another set of sentences, thereby finding the most semantically relevant words or phrases and achieving fine-grained semantic alignment. The semantically aligned input prompts are used to input the prompts generated for the overhead transmission line construction plan according to the fixed input prompt template.

Images