Intelligent review method and system for feasibility research of power grid infrastructure project

Abstract

The invention relates to an intelligent review method and system for feasibility research of a power grid infrastructure project. The method comprises the following steps: training a Qwen-7B model through an engineering feasibility research report and review opinions of the power grid infrastructure project to obtain a large language model; performing historical project key index extraction on the project feasibility research report and review suggestions of the historical power grid infrastructure project by using the large language model to obtain a historical project key index library; performing engineering key index extraction on the engineering feasibility research report and the review opinion of the new power grid infrastructure project by using the large language model to obtain a new engineering key index; performing systematic comparative analysis on the new project key indexes in a historical project key index library to judge whether the new project key indexes are reasonable or not; according to the method, the rationality of related index data can be quickly judged, data reference is provided for review personnel, and the feasibility review efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of engineering evaluation, and in particular to a feasibility study intelligent review method and system for a power grid infrastructure project. Background Art

[0002] The feasibility study of power grid infrastructure projects is an important content and achievement of preliminary work, and is an important basis for project establishment, approval and construction. The main task of the pre-evaluation is to review whether the project meets the evaluation conditions and whether there are subversive factors that affect the project progress before the feasibility study review work is officially carried out. In the past power grid infrastructure projects that were approved, there were some cases that were returned because the economic, technical and other indicators did not meet the project approval conditions, resulting in a waste of project review resources. With the continuous development of the power industry, more professional technical means and information system models are needed for the feasibility study review of power grid projects to meet the lean and standardized development of power grid analysis and decision-making, feasibility study review and other work. It is urgent to establish a set of efficient and accurate intelligent feasibility study review solutions for power grid infrastructure projects to improve the efficiency of feasibility study review. Summary of the invention

[0003] The present invention provides a feasibility study intelligent review method and system for power grid infrastructure projects to solve at least one of the above technical problems.

[0004] The technical solution of the present invention to solve the above technical problems is as follows: A feasibility study intelligent review method for a power grid infrastructure project, comprising:

[0005] S1, train the Qwen-7B model through the engineering feasibility study reports and review opinions of power grid infrastructure projects in the first preset historical time period to obtain a large language model suitable for the feasibility study review field;

[0006] S2, using the large language model to extract historical engineering key indicators from engineering feasibility study reports and review opinions of historical power grid infrastructure projects within a second preset historical time period, to obtain a historical engineering key indicator library;

[0007] S3, using the large language model to extract key engineering indicators from the engineering feasibility study report and review opinions of the new power grid infrastructure project to obtain new engineering key indicators;

[0008] S4, performing a systematic comparative analysis on the new engineering key indicators in the historical engineering key indicator library to determine whether the new engineering key indicators are reasonable;

[0009] S5, generating an internal audit report based on the results of whether the key indicators of the new project are reasonable.

[0010] Based on the above technical solution, the present invention can also be improved as follows.

[0011] Further, the S1 is specifically:

[0012] S11, collecting engineering feasibility study reports and review opinions of power grid infrastructure projects within a first preset historical time period to obtain a sample data set;

[0013] S12, performing data cleaning on the sample data set;

[0014] S13, performing engineering key indicator labeling processing on the cleaned sample data set to obtain the sample data set with labels;

[0015] S14, dividing the sample data set with labels into a training set, a validation set and a test set;

[0016] S15, training the pre-trained model using the training set to obtain an initial large language model;

[0017] S16, using the verification set and the test set to verify and test the initial large language model, so as to optimize the parameters of the initial large language model and obtain a large language model suitable for the feasibility study review field.

[0018] Further, the S2 is specifically:

[0019] S21, using the large language model to extract key indicators of historical projects from engineering feasibility reports and review opinions of historical power grid infrastructure projects within a second preset historical time period to obtain key indicators of historical projects;

[0020] S22, creating a database and determining a data table structure storing the key indicators of the historical projects in the database;

[0021] S23, based on the data table structure, keyword matching and filtering are performed on the historical engineering key indicators to store them in the database, thereby obtaining the historical engineering key indicator library.

[0022] Furthermore, the key indicators of the project include basic indicators, technical indicators and economic indicators; the technical indicators include numerical technical indicators and character technical indicators; S4 is specifically:

[0023] S41, performing consistency evaluation on the basic indicators in the new project key indicators in the historical project key indicator library to find out the historical power grid infrastructure projects similar to the new power grid infrastructure project; defining the historical power grid infrastructure projects similar to the new power grid infrastructure project as similar power grid infrastructure projects;

[0024] S42, extracting technical indicator statistical features from the numerical technical indicators of the similar power grid infrastructure projects, and performing Z-score analysis using the technical indicator statistical features and the numerical technical indicators of the new power grid infrastructure project to determine whether the numerical technical indicators of the new power grid infrastructure project are reasonable;

[0025] S43, completely matching and comparing the character-based technical indicators of the new power grid infrastructure project with the character-based technical indicators of the similar power grid infrastructure project to determine whether the character-based technical indicators of the new power grid infrastructure project are reasonable;

[0026] S44, extracting statistical characteristics of economic indicators from the economic indicators of the similar power grid infrastructure projects, and performing Z-score analysis using the statistical characteristics of economic indicators and the economic indicators of the new power grid infrastructure project to determine whether the economic indicators of the new power grid infrastructure project are reasonable.

[0027] Further, the S41 is specifically as follows:

[0028] S411, using a basic indicator consistency evaluation formula, calculating a basic indicator similarity score between the new project key indicator and the historical project key indicator in the historical project key indicator library;

[0029] S412, determining whether the basic indicator similarity score reaches a preset similarity score threshold; if not, determining the corresponding historical power grid infrastructure project as being dissimilar to the new power grid infrastructure project; if so, determining the corresponding historical power grid infrastructure project as being similar to the new power grid infrastructure project, and defining the historical power grid infrastructure project that is similar to the new power grid infrastructure project as a similar power grid infrastructure project.

[0030] Further, in S411, the basic indicator consistency evaluation formula is:

[0031]

[0032] Wherein, x represents the new grid infrastructure project, y represents the historical grid infrastructure project; S(x,y) represents the similarity score between the new grid infrastructure project and the historical grid infrastructure project; i represents the i-th basic indicator, x i represents the index value of the i-th basic index in the new power grid infrastructure project, y i represents the index value of the i-th basic index in the historical power grid infrastructure project; f i (x i ,y i ) represents the matching function between the new power grid infrastructure project and the historical power grid infrastructure project based on the i-th basic indicator; w irepresents the weight of the ith basic indicator; n represents the total number of the basic indicators.

[0033] Furthermore, the basic indicators include numerical basic indicators and character basic indicators;

[0034] When the basic indicator is the numerical basic indicator:

[0035]

[0036] When the basic index is the character-type basic index:

[0037]

[0038] Among them, x max represents the maximum value of the i-th basic indicator in the new power grid infrastructure project, x min It represents the minimum index value of the ith basic index in the new power grid infrastructure project.

[0039] Further, in said S42, said technical indicator statistical characteristics include mean and standard deviation;

[0040] The formula for Z-score analysis using the statistical characteristics of the technical indicators and the numerical technical indicators of the new power grid infrastructure project is:

[0041]

[0042] Wherein, Z represents the Z score, μ represents the mean, σ represents the standard deviation; x T Numerical technical indicators representing the new power grid infrastructure project;

[0043] Whether the numerical technical indicators of the new power grid infrastructure project are reasonable is determined specifically as follows: when the Z score is within the preset numerical range, the numerical technical indicators of the new power grid infrastructure project are determined to be reasonable; when the Z score is not within the preset numerical range, the numerical technical indicators of the new power grid infrastructure project are determined to be unreasonable.

[0044] Further, there is at least one similar power grid infrastructure project; S43 specifically comprises:

[0045] Comparing the character-based technical indicators of the new power grid infrastructure project with the character-based technical indicators of the similar power grid infrastructure project;

[0046] If there is at least one character-based technical indicator of the similar power grid infrastructure project that is completely consistent with the character-based technical indicator of the new power grid infrastructure project, then it is determined that the character-based technical indicator of the new power grid infrastructure project is reasonable;

[0047] If the character-type technical indicators of all the similar power grid infrastructure projects are inconsistent with the character-type technical indicators of the new power grid infrastructure project, it is determined that the character-type technical indicators of the new power grid infrastructure project are unreasonable.

[0048] Based on the above-mentioned intelligent feasibility study review method for power grid infrastructure projects, the present invention also provides an intelligent feasibility study review system for power grid infrastructure projects.

[0049] A feasibility study intelligent review system for power grid infrastructure projects, comprising:

[0050] A model building module, which is used to train the Qwen-7B model through the engineering feasibility study reports and review opinions of power grid infrastructure projects within a first preset historical time period to obtain a large language model suitable for the feasibility study review field;

[0051] An indicator library construction module, which is used to extract historical engineering key indicators from engineering feasibility study reports and review opinions of historical power grid infrastructure projects within a second preset historical time period using the large language model to obtain a historical engineering key indicator library;

[0052] An indicator extraction module is used to extract key engineering indicators from the engineering feasibility study report and review opinions of the new power grid infrastructure project using the large language model to obtain new engineering key indicators;

[0053] An analysis module, which is used to systematically compare and analyze the new engineering key indicators in the historical engineering key indicator library to determine whether the new engineering key indicators are reasonable;

[0054] A report generation module is used to generate an internal audit report based on whether the key indicators of the new project are reasonable.

[0055] The beneficial effects of the present invention are as follows: the intelligent feasibility study review method and system of the present invention for power grid infrastructure projects is proposed based on the current status analysis of feasibility study plans for power grid infrastructure projects in reality. According to the principles of establishing a key indicator system and comparative analysis with historical similar projects, a comprehensive analysis of the rationality evaluation of key indicators is adopted, which can quickly judge the rationality of relevant indicator data, provide data reference for reviewers, and improve the efficiency of feasibility study review. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A flowchart of a feasibility study intelligent review method for a power grid infrastructure project of the present invention;

[0057] Figure 2 The present invention is a structural block diagram of a feasibility study intelligent review system for power grid infrastructure projects. DETAILED DESCRIPTION

[0058] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0059] like Figure 1 As shown, a feasibility study intelligent review method for a power grid infrastructure project includes:

[0060] S1, train the Qwen-7B model through the engineering feasibility study reports and review opinions of power grid infrastructure projects in the first preset historical time period to obtain a large language model suitable for the feasibility study review field;

[0061] S2, using the large language model to extract historical engineering key indicators from engineering feasibility study reports and review opinions of historical power grid infrastructure projects within a second preset historical time period, to obtain a historical engineering key indicator library;

[0062] S3, using the large language model to extract key engineering indicators from the engineering feasibility study report and review opinions of the new power grid infrastructure project to obtain new engineering key indicators;

[0063] S4, performing a systematic comparative analysis on the new engineering key indicators in the historical engineering key indicator library to determine whether the new engineering key indicators are reasonable;

[0064] S5, generating an internal audit report based on the results of whether the key indicators of the new project are reasonable.

[0065] The invention provides an intelligent feasibility study review method for power grid infrastructure projects. The method is proposed based on the current status analysis of feasibility study plans for power grid infrastructure projects in reality. According to the principles of establishing a key indicator system and comparative analysis with historical similar projects, a comprehensive analysis of the rationality evaluation of key indicators is adopted. The method can quickly judge the rationality of relevant indicator data, provide data reference for reviewers, and improve the efficiency of feasibility study review.

[0066] In some embodiments, the S1 is specifically:

[0067] S11, collecting engineering feasibility study reports and review opinions of power grid infrastructure projects within a first preset historical time period to obtain a sample data set;

[0068] S12, performing data cleaning on the sample data set;

[0069] S13, performing engineering key indicator labeling processing on the cleaned sample data set to obtain the sample data set with labels;

[0070] S14, dividing the sample data set with labels into a training set, a validation set and a test set;

[0071] S15, training the pre-trained model using the training set to obtain an initial large language model;

[0072] S16, using the verification set and the test set to verify and test the initial large language model, so as to optimize the parameters of the initial large language model and obtain a large language model suitable for the feasibility study review field.

[0073] In the S11, in the process of collecting engineering feasibility study reports and review opinions of power grid infrastructure projects within the first preset historical time period, it is necessary to ensure that the data covers projects of different types and sizes. The engineering feasibility study reports and review opinions of power grid infrastructure projects within the first preset historical time period are collected, and information is extracted to form a sample data set. Among them, the core process of information extraction relies on entity recognition (Named Entity Recognition, NER) and relationship extraction (Relation Extraction, RE) technology in natural language processing (NLP). Entity recognition (NER) is to identify and annotate key information from unstructured text. The key information in the feasibility study report mainly includes basic indicators, technical indicators (such as substation capacity, transmission line length), economic indicators (such as total investment, static investment), etc. For example, from the sentence in the report "The transmission line length of this project is 1,200 kilometers, and the total investment is 800 million yuan", the model can identify "transmission line length = 1,200 kilometers" and "total investment = 800 million yuan" as two independent entities. Relation extraction (RE) is the logical or causal relationship between entities. By analyzing the grammatical structure and contextual information in the sentence, the association between entities is extracted to form more in-depth information. For example, in the sentence "Due to the increase in substation capacity, the power supply area is expected to expand to 1,500 square kilometers", the causal relationship between "substation capacity increase" and "power supply area expansion" is identified.

[0074] In S12, data cleaning is to remove noise, sensitive information or irrelevant content in the data, such as personal information, project numbers, etc.

[0075] In the S13, the key indicators of the project are labeled, specifically, the sample data in the sample data set are divided into different categories and labels, such as project type, basic indicators, technical indicators, economic indicators, suggestions, etc. In order to further improve the knowledge coverage of the model and cope with the data richness in complex project scenarios, this paper introduces the Retrieval-Augmented Generation (RAG) technology. RAG retrieves the document content related to the current task from the external knowledge base before generating information, and provides more complete context support in the generation stage. The RAG model first obtains documents related to the current task from the knowledge base through the retrieval stage, such as power supply capacity and investment return data in historical power grid infrastructure projects. The retrieved content provides context for model generation, so that the model can generate more reliable analysis results in combination with the information in the current feasibility study report. For example, when the model processes a project involving "substation expansion", the RAG model can retrieve substation expansion instances similar to power grid infrastructure projects from the knowledge base, and generate more targeted expansion plan recommendations in combination with the environmental parameters of the current power grid infrastructure project.

[0076] In S14, the division ratio of the training set, the validation set and the test set is 8:1:1 to ensure the generalization and accuracy of the model.

[0077] In the S15, before model training, the training set, validation set and test set need to be converted into a format that can be processed by Qwen-7B, usually organized in JSON or CSV format, including "input text" and "output label" fields. During the model training process, fine-tuning parameters such as learning rate, batch size, training rounds, etc. are set according to project requirements, and fine-tuning is performed on the training data through appropriate optimizers (such as AdamW) and loss functions (such as cross entropy loss). In the analysis of the feasibility study report of the power grid infrastructure project, the data set on which the model fine-tuning depends consists of industry technical documents, standard specifications and historical power grid feasibility study reports. In order to ensure the performance of the model in terms of technical indicators, economic benefits, environmental impact, etc., the present invention has refined the annotation of the training data. For example, key information such as substation capacity and transmission line parameters in technical indicators, and core data such as return on investment and dynamic investment in economic indicators are manually annotated. These annotated data serve as supervisory signals for model training, helping the model to better learn entity and relationship patterns in the field during the fine-tuning process. Through fine-tuning, the model can more accurately identify specific key indicators in the power grid field, especially when it comes to sentences with professional terms and complex grammatical structures. For example, "the internal rate of return of the project is 10%, and the power supply capacity is expected to be 2500MW." The fine-tuning model can accurately identify "10%" as the internal rate of return of the investment and associate it with the economic benefit evaluation of the specific project.

[0078] In some embodiments, S2 is specifically:

[0079] S21, using the large language model to extract key indicators of historical projects from engineering feasibility reports and review opinions of historical power grid infrastructure projects within a second preset historical time period to obtain key indicators of historical projects;

[0080] S22, creating a database and determining a data table structure storing the key indicators of the historical projects in the database;

[0081] S23, based on the data table structure, keyword matching and filtering are performed on the historical engineering key indicators to store them in the database, thereby obtaining the historical engineering key indicator library.

[0082] In S21, the second preset historical time period may completely cover the first preset historical time period, or may partially cover the first preset historical time period. When using the large language model to extract key indicators of historical projects from the engineering feasibility study reports and review opinions of historical power grid infrastructure projects within the second preset historical time period, NER and RE technologies are used to parse the engineering feasibility study reports and review opinions to identify all technical, economic, environmental and other key indicators of the line / substation project. For complex or uncertain indicators, relevant information is retrieved from an external knowledge base through RAG technology for enhanced generation.

[0083] In S22, the database may be a SQL, NoSQL or other database. The data table structure includes field names, types and constraints. For example, the fields include project number, project name, key indicator name (such as investment amount, land area, etc.), indicator value, data source (feasibility study report or review opinion), date and timestamp.

[0084] In said S23, a corresponding table is created according to the designed data table structure to store the extracted key indicators of historical projects, and keyword matching and filtering are performed to ensure that the extracted content is the required key indicators (such as investment amount, project scale, construction period, etc.).

[0085] In some embodiments, after completing the intelligent feasibility review of the new power grid infrastructure project, the new engineering key indicators extracted from S3 will also be called historical engineering key indicators. Therefore, after completing the intelligent feasibility review of the new power grid infrastructure project, the new engineering key indicators extracted from S3 can be stored in the database as historical engineering key indicators.

[0086] In some embodiments, the key engineering indicators include basic indicators, technical indicators and economic indicators; the technical indicators include numerical technical indicators and character technical indicators; S4 is specifically:

[0087] S41, performing consistency evaluation on the basic indicators in the new project key indicators in the historical project key indicator library to find out the historical power grid infrastructure projects similar to the new power grid infrastructure project; defining the historical power grid infrastructure projects similar to the new power grid infrastructure project as similar power grid infrastructure projects;

[0088] S42, extracting technical indicator statistical features from the numerical technical indicators of the similar power grid infrastructure projects, and performing Z-score analysis using the technical indicator statistical features and the numerical technical indicators of the new power grid infrastructure project to determine whether the numerical technical indicators of the new power grid infrastructure project are reasonable;

[0089] S43, completely matching and comparing the character-based technical indicators of the new power grid infrastructure project with the character-based technical indicators of the similar power grid infrastructure project to determine whether the character-based technical indicators of the new power grid infrastructure project are reasonable;

[0090] S44, extracting statistical characteristics of economic indicators from the economic indicators of the similar power grid infrastructure projects, and performing Z-score analysis using the statistical characteristics of economic indicators and the economic indicators of the new power grid infrastructure project to determine whether the economic indicators of the new power grid infrastructure project are reasonable.

[0091] The consistency of basic indicators is used to verify the similarity determining factors between new power grid infrastructure projects and historical power grid infrastructure projects, and to find similar power grid infrastructure projects. When evaluating the consistency of basic indicators, the basic indicators of new power grid infrastructure projects should be basically consistent with those of historical power grid infrastructure projects. If the difference is large, it may be caused by changes in the background conditions of the project within the time period. The greater the difference between the basic indicators and the key indicator library, the lower the score. When the score of the basic indicator library is greater than 80%, it is determined that the historical power grid infrastructure project is basically consistent with the new power grid infrastructure project, otherwise it is determined to be dissimilar. Power grid infrastructure projects include historical projects of substation engineering and historical projects of line engineering; among them, the basic indicator set of historical projects of substation engineering includes: project type, voltage level, project scale, province and city, etc.; the basic indicator set of historical projects of line engineering includes: project type, voltage level, project scale, line length, province and city, terrain (flat land, hills, river network swamps, mountains, high mountains and ridges, deserts, etc.), wind speed and other indicators.

[0092] In some embodiments, the S41 is specifically:

[0093] S411, using a basic indicator consistency evaluation formula, calculating a basic indicator similarity score between the new project key indicator and the historical project key indicator in the historical project key indicator library;

[0094] S412, determining whether the basic indicator similarity score reaches a preset similarity score threshold; if not, determining the corresponding historical power grid infrastructure project as being dissimilar to the new power grid infrastructure project; if so, determining the corresponding historical power grid infrastructure project as being similar to the new power grid infrastructure project, and defining the historical power grid infrastructure project that is similar to the new power grid infrastructure project as a similar power grid infrastructure project.

[0095] In S411, the basic indicator consistency evaluation formula is:

[0096]

[0097] Wherein, x represents the new grid infrastructure project, y represents the historical grid infrastructure project; S(x,y) represents the similarity score between the new grid infrastructure project and the historical grid infrastructure project; i represents the i-th basic indicator, x i represents the index value of the i-th basic index in the new power grid infrastructure project, y i represents the index value of the i-th basic index in the historical power grid infrastructure project; f i (x i ,y i ) represents the matching function between the new power grid infrastructure project and the historical power grid infrastructure project based on the i-th basic indicator; w i Represents the weight of the i-th basic indicator, and the sum of all weights is 1, that is, n represents the total number of the basic indicators.

[0098] The basic indicators include numerical basic indicators (such as investment amount, project scale, etc.) and character basic indicators (such as project type, voltage level, etc.); for numerical basic indicators (such as investment amount, project scale), the similarity is calculated using the matching function of the normalized difference method; for character basic indicators (such as project type, voltage level), the similarity is calculated using the matching function of the complete match or level difference method; specifically,

[0099] When the basic indicator is the numerical basic indicator:

[0100]

[0101] When the basic index is the character-type basic index:

[0102]

[0103] Among them, x max represents the maximum value of the i-th basic indicator in the new power grid infrastructure project, x min It represents the minimum index value of the ith basic index in the new power grid infrastructure project.

[0104] Therefore, in the end, if the value of the similarity score S(x,y) is closer to 1, it means that the new power grid infrastructure project x is more similar to the historical power grid infrastructure project y. According to experience, the preset similarity score threshold is set to 0.8. If the value of S(x,y) is greater than the threshold, the new power grid infrastructure project x and the historical power grid infrastructure project y are considered to be similar projects.

[0105] Technical indicator consistency analysis, technical indicators are divided into character-type technical indicators and numerical-type technical indicators. They are calculated and analyzed separately through the preset model algorithm to comprehensively determine whether the various technical indicators of the new power grid infrastructure project are within a reasonable range. Numerical indicators can be extracted from similar power grid infrastructure projects by extracting their statistical characteristics, including mean (μ), standard deviation (σ), etc., and calculating the Z score to determine whether the technical indicators of the new power grid infrastructure project are within a reasonable range. The Z score is used to determine whether the technical indicators of the new power grid infrastructure project are within a reasonable range. Usually, a Z score between [-2,2] indicates that the technical indicators of the new power grid infrastructure project are within a reasonable range. Character-type technical indicators (such as steel model, concrete model, protective cap model, conductor model, etc.) can be judged to be reasonable by complete matching. If the character-type indicators of the technical indicators of the new power grid infrastructure project are completely consistent with the character-type technical indicators of similar power grid infrastructure projects, they can be considered reasonable.

[0106] In some embodiments, in S42, the technical indicator statistical characteristics include a mean and a standard deviation; wherein the mean is the mean of the numerical technical indicators of the similar power grid infrastructure projects, and the calculation formula is:

[0107]

[0108] The standard deviation is a measure of the dispersion of data values ​​from the mean, and its calculation formula is:

[0109]

[0110] Among them, μ represents the mean, y j represents the index value of the jth numerical technical index in similar power grid infrastructure projects, M represents the total number of numerical technical indicators, and σ represents the standard deviation;

[0111] The formula for Z-score analysis using the statistical characteristics of the technical indicators and the numerical technical indicators of the new power grid infrastructure project is:

[0112]

[0113] Wherein, Z represents the Z score, μ represents the mean, σ represents the standard deviation; x T Numerical technical indicators representing the new power grid infrastructure project;

[0114] Whether the numerical technical indicators of the new power grid infrastructure project are reasonable is determined specifically as follows: when the Z score is within the preset numerical range, the numerical technical indicators of the new power grid infrastructure project are determined to be reasonable; when the Z score is not within the preset numerical range (the preset numerical range is generally set to [-2,2]), the numerical technical indicators of the new power grid infrastructure project are determined to be unreasonable.

[0115] In some embodiments, there is at least one similar power grid infrastructure project; S43 specifically comprises:

[0116] Comparing the character-based technical indicators of the new power grid infrastructure project with the character-based technical indicators of the similar power grid infrastructure project;

[0117] If there is at least one character-based technical indicator of the similar power grid infrastructure project that is completely consistent with the character-based technical indicator of the new power grid infrastructure project, then it is determined that the character-based technical indicator of the new power grid infrastructure project is reasonable;

[0118] If the character-type technical indicators of all the similar power grid infrastructure projects are inconsistent with the character-type technical indicators of the new power grid infrastructure project, it is determined that the character-type technical indicators of the new power grid infrastructure project are unreasonable.

[0119] For example, after S41, historical power grid infrastructure projects 1, 2 and 3 similar to the new power grid infrastructure project are obtained.

[0120] If the steel model (a character-type technical indicator) of the new power grid infrastructure project is Q345, the historical power grid infrastructure project 1 and the historical power grid infrastructure project 3 are also Q345, and the historical power grid infrastructure project 2 is Q235; then the steel model of the new power grid infrastructure project completely matches the steel model of the historical power grid infrastructure project 1 and the historical power grid infrastructure project 3, then the indicator of the new power grid infrastructure project is considered to be reasonable.

[0121] If the steel model of the new power grid infrastructure project is Q235, the historical power grid infrastructure project 1 and the historical power grid infrastructure project 3 are also Q345, and the historical power grid infrastructure project 2 is Q235; then the steel model of the new power grid infrastructure project completely matches the steel model of the historical power grid infrastructure project 2, then the indicator of the new power grid infrastructure project is considered to be reasonable.

[0122] If the steel model of the new power grid infrastructure project is Q135, the historical power grid infrastructure projects 1 and 3 are Q345, and the historical power grid infrastructure project 2 is Q235; then the steel model of the new power grid infrastructure project does not completely match the steel models of the historical power grid infrastructure projects 1, 2, and 3, then this indicator of the new power grid infrastructure project is considered unreasonable.

[0123] In some embodiments, S44 is an economic indicator analysis, which compares economic indicators such as the return on investment and financial internal rate of return of new power grid infrastructure projects. Economic indicators are numerical indicators, so to evaluate the rationality of the economic indicators of new power grid infrastructure projects, the numerical indicator similarity analysis method in the technical indicator similarity analysis can be used. If the economic indicators of the new power grid infrastructure project are similar to those of historical successful projects, the economic feasibility is considered to be high; otherwise, an economic risk warning will be generated. For the feasibility study stage, the most critical economic indicators are the static investment amount and the dynamic investment amount. For line projects, it is necessary to calculate the amount per kilometer.

[0124] In some embodiments, S5 is intelligent review report generation, which specifically includes:

[0125] S51, based on the comparison analysis results, use the model to generate targeted review opinions, including similarity analysis, risk warnings and improvement suggestions.

[0126] S52: Report template design: Determine the content structure of the review report, including similarity analysis between new grid infrastructure projects and historical grid infrastructure projects, key risk points and optimization suggestions.

[0127] S53: Content filling: Based on the analysis results of the review module, the report content is automatically filled in. The content can be dynamically generated using a template engine (such as Jinja2).

[0128] Based on the above-mentioned intelligent feasibility study review method for power grid infrastructure projects, the present invention also provides an intelligent feasibility study review system for power grid infrastructure projects.

[0129] like Figure 2 As shown, a feasibility study intelligent review system for power grid infrastructure projects includes:

[0130] A model building module, which is used to train the Qwen-7B model through the engineering feasibility study reports and review opinions of power grid infrastructure projects within a first preset historical time period to obtain a large language model suitable for the feasibility study review field;

[0131] An indicator library construction module, which is used to extract historical engineering key indicators from engineering feasibility study reports and review opinions of historical power grid infrastructure projects within a second preset historical time period using the large language model to obtain a historical engineering key indicator library;

[0132] An indicator extraction module is used to extract key engineering indicators from the engineering feasibility study report and review opinions of the new power grid infrastructure project using the large language model to obtain new engineering key indicators;

[0133] An analysis module, which is used to systematically compare and analyze the new engineering key indicators in the historical engineering key indicator library to determine whether the new engineering key indicators are reasonable;

[0134] A report generation module is used to generate an internal audit report based on whether the key indicators of the new project are reasonable.

[0135] The specific functions of each module in the feasibility study intelligent review system for a power grid infrastructure project of the present invention refer to the steps of the feasibility study intelligent review method for a power grid infrastructure project of the present invention, which will not be repeated here.

[0136] The present invention conducts intelligent feasibility study review for power grid infrastructure projects based on a large language model and multi-dimensional data analysis, establishes a key indicator library for historical projects through collection, information extraction, and data analysis, studies intelligent evaluation algorithms for power engineering construction projects, establishes corresponding mathematical models, determines the reference range of key indicators, assists experts in conducting feasibility study reviews, and improves the efficiency of feasibility study reviews.

[0137] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A feasibility study intelligent review method for power grid infrastructure projects, characterized in that: include: S1, train the Qwen-7B model through the engineering feasibility study reports and review opinions of power grid infrastructure projects in the first preset historical time period to obtain a large language model suitable for the feasibility study review field; S2, using the large language model to extract historical engineering key indicators from engineering feasibility study reports and review opinions of historical power grid infrastructure projects within a second preset historical time period, to obtain a historical engineering key indicator library; S3, using the large language model to extract key engineering indicators from the engineering feasibility study report and review opinions of the new power grid infrastructure project to obtain new engineering key indicators; S4, performing a systematic comparative analysis on the new engineering key indicators in the historical engineering key indicator library to determine whether the new engineering key indicators are reasonable; S5, generating an internal audit report based on the results of whether the key indicators of the new project are reasonable.

2. The feasibility study intelligent review method for power grid infrastructure projects according to claim 1 is characterized in that: The S1 is specifically: S11, collecting engineering feasibility study reports and review opinions of power grid infrastructure projects within a first preset historical time period to obtain a sample data set; S12, performing data cleaning on the sample data set; S13, performing engineering key indicator labeling processing on the cleaned sample data set to obtain the sample data set with labels; S14, dividing the sample data set with labels into a training set, a validation set and a test set; S15, training the pre-trained model using the training set to obtain an initial large language model; S16, using the verification set and the test set to verify and test the initial large language model, so as to optimize the parameters of the initial large language model and obtain a large language model suitable for the feasibility study review field.

3. The feasibility study intelligent review method for power grid infrastructure projects according to claim 1 is characterized in that: The S2 is specifically: S21, using the large language model to extract key indicators of historical projects from engineering feasibility reports and review opinions of historical power grid infrastructure projects within a second preset historical time period to obtain key indicators of historical projects; S22, creating a database and determining a data table structure storing the key indicators of the historical projects in the database; S23, based on the data table structure, keyword matching and filtering are performed on the historical engineering key indicators to store them in the database, thereby obtaining the historical engineering key indicator library.

4. The feasibility study intelligent review method for power grid infrastructure projects according to claim 1 is characterized in that: The key indicators of the project include basic indicators, technical indicators and economic indicators; the technical indicators include numerical technical indicators and character technical indicators; S4 is specifically: S41, performing consistency evaluation on the basic indicators in the new project key indicators in the historical project key indicator library to find out the historical power grid infrastructure projects similar to the new power grid infrastructure project; defining the historical power grid infrastructure projects similar to the new power grid infrastructure project as similar power grid infrastructure projects; S42, extracting technical indicator statistical features from the numerical technical indicators of the similar power grid infrastructure projects, and performing Z-score analysis using the technical indicator statistical features and the numerical technical indicators of the new power grid infrastructure project to determine whether the numerical technical indicators of the new power grid infrastructure project are reasonable; S43, completely matching and comparing the character-based technical indicators of the new power grid infrastructure project with the character-based technical indicators of the similar power grid infrastructure project to determine whether the character-based technical indicators of the new power grid infrastructure project are reasonable; S44, extracting statistical characteristics of economic indicators from the economic indicators of the similar power grid infrastructure projects, and performing Z-score analysis using the statistical characteristics of economic indicators and the economic indicators of the new power grid infrastructure project to determine whether the economic indicators of the new power grid infrastructure project are reasonable.

5. The feasibility study intelligent review method for power grid infrastructure projects according to claim 4 is characterized in that: The S41 is specifically: S411, using a basic indicator consistency evaluation formula, calculating a basic indicator similarity score between the new project key indicator and the historical project key indicator in the historical project key indicator library; S412, determining whether the basic indicator similarity score reaches a preset similarity score threshold; if not, determining the corresponding historical power grid infrastructure project as being dissimilar to the new power grid infrastructure project; If so, the corresponding historical power grid infrastructure project is determined to be similar to the new power grid infrastructure project, and the historical power grid infrastructure project similar to the new power grid infrastructure project is defined as a similar power grid infrastructure project.

6. The feasibility study intelligent review method for power grid infrastructure projects according to claim 5 is characterized in that: In S411, the basic indicator consistency evaluation formula is: Wherein, x represents the new grid infrastructure project, y represents the historical grid infrastructure project; S(x,y) represents the similarity score between the new grid infrastructure project and the historical grid infrastructure project; i represents the i-th basic indicator, x i represents the index value of the i-th basic index in the new power grid infrastructure project, y i represents the index value of the i-th basic index in the historical power grid infrastructure project; f i (x i ,y i ) represents the matching function between the new power grid infrastructure project and the historical power grid infrastructure project based on the i-th basic indicator; w i represents the weight of the ith basic indicator; n represents the total number of the basic indicators.

7. The feasibility study intelligent review method for power grid infrastructure projects according to claim 6 is characterized in that: The basic indicators include numerical basic indicators and character basic indicators; When the basic indicator is the numerical basic indicator: When the basic index is the character-type basic index: Among them, x max represents the maximum value of the i-th basic indicator in the new power grid infrastructure project, x min It represents the minimum index value of the ith basic index in the new power grid infrastructure project.

8. The feasibility study intelligent review method for power grid infrastructure projects according to claim 4 is characterized in that: In said S42, said technical indicator statistical characteristics include mean and standard deviation; The formula for Z-score analysis using the statistical characteristics of the technical indicators and the numerical technical indicators of the new power grid infrastructure project is: Wherein, Z represents the Z score, μ represents the mean, σ represents the standard deviation; x T Numerical technical indicators representing the new power grid infrastructure project; Whether the numerical technical indicators of the new power grid infrastructure project are reasonable is determined specifically as follows: when the Z score is within the preset numerical range, the numerical technical indicators of the new power grid infrastructure project are determined to be reasonable; when the Z score is not within the preset numerical range, the numerical technical indicators of the new power grid infrastructure project are determined to be unreasonable.

9. The feasibility study intelligent review method for power grid infrastructure projects according to claim 4 is characterized in that: There is at least one similar power grid infrastructure project; S43 specifically includes: Comparing the character-based technical indicators of the new power grid infrastructure project with the character-based technical indicators of the similar power grid infrastructure project; If there is at least one character-based technical indicator of the similar power grid infrastructure project that is completely consistent with the character-based technical indicator of the new power grid infrastructure project, then it is determined that the character-based technical indicator of the new power grid infrastructure project is reasonable; If the character-type technical indicators of all the similar power grid infrastructure projects are inconsistent with the character-type technical indicators of the new power grid infrastructure project, it is determined that the character-type technical indicators of the new power grid infrastructure project are unreasonable.

10. An intelligent feasibility study review system for power grid infrastructure projects, characterized in that: include: A model building module, which is used to train the Qwen-7B model through the engineering feasibility study reports and review opinions of power grid infrastructure projects within a first preset historical time period to obtain a large language model suitable for the feasibility study review field; An indicator library construction module, which is used to extract historical engineering key indicators from engineering feasibility study reports and review opinions of historical power grid infrastructure projects within a second preset historical time period using the large language model to obtain a historical engineering key indicator library; An indicator extraction module is used to extract key engineering indicators from the engineering feasibility study report and review opinions of the new power grid infrastructure project using the large language model to obtain new engineering key indicators; An analysis module, which is used to systematically compare and analyze the new engineering key indicators in the historical engineering key indicator library to determine whether the new engineering key indicators are reasonable; A report generation module is used to generate an internal audit report based on whether the key indicators of the new project are reasonable.

Images