Intelligent verification method for preliminary design scheme of electric power engineering

Abstract

The invention provides an intelligent verification method for a preliminary design scheme of electric power engineering, and the method comprises the following steps: S1, obtaining the specification information of the electric power engineering design, and carrying out the preprocessing of the specification information of the electric power engineering design; s2, constructing a knowledge base based on the preprocessed standard information, determining the weight of the standard information according to the source of the standard information, and sorting the standard information in the knowledge base according to the weight of the standard information; s3, obtaining a to-be-calibrated preliminary design scheme of the electric power engineering, and preprocessing the preliminary design scheme of the electric power engineering; s4, based on the pre-processed preliminary design scheme, sequentially performing data retrieval matching according to the weight of the standard information in the knowledge base; s5, inputting the preliminary design scheme of the electric power engineering matched with the specification information in the knowledge base into the trained large language model for checking processing; and S6, performing logic verification and overdesign verification on the large language model approved as the compliant preliminary design scheme of the electric power engineering to obtain a final compliant preliminary design scheme of the electric power engineering.

Description

Technical Field

[0001] This invention relates to a review method, and more particularly to an intelligent review method for preliminary design schemes of power engineering projects. Background Technology

[0002] In the field of power engineering, preliminary design review is a core part of the early stages of a project, playing a crucial role in ensuring the quality of subsequent construction, effectively controlling project investment, and guaranteeing the safe and reliable operation of the power grid. The preliminary design documents for modern power transmission and transformation projects are extensive and complex, involving multiple technical fields such as primary electrical systems, secondary electrical systems, communications, civil engineering, transmission lines, environmental protection and soil and water conservation, and technical economics. Review experts need to conduct a comprehensive and in-depth evaluation of the design schemes within a short period of time, based on a vast amount of constantly updated national and industry mandatory standards, internal company standards, general design atlases, typical design cases, and rich historical review opinions and engineering data.

[0003] Traditional review models rely heavily on experts' personal experience and manual operations, facing significant challenges. Firstly, there is information overload and difficulty in knowledge retrieval; the knowledge sources needed for review are scattered across hundreds or even thousands of documents, requiring experts to spend considerable time manually reviewing, comparing, and verifying, resulting in extremely low efficiency. Secondly, the process is cumbersome, creating a significant efficiency bottleneck: from receiving design documents and checking their structure and the completeness of drawing labels, to verifying the compliance and rationality of technical parameters and economic indicators item by item, and then summarizing opinions from various disciplines to form a formal review report and meeting minutes, the entire process is highly dependent on manual labor, time-consuming, and labor-intensive, becoming a major bottleneck restricting the improvement of review efficiency.

[0004] Therefore, in order to solve the above-mentioned technical problems, it is urgent to propose a new technical approach. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide an intelligent review method for preliminary design schemes of power engineering projects. By pre-constructing a corresponding normative knowledge base, and then parsing the preliminary design scheme of the power engineering project, text matching is performed. The matched text information is then input into a trained large language model for normative prediction, thereby effectively improving the efficiency of reviewing preliminary design schemes of power engineering projects. Furthermore, after the review by the large language model, logical and over-design judgments are performed, thereby further improving the accuracy of the review results.

[0006] This invention provides an intelligent review method for preliminary design schemes in power engineering, comprising the following steps:

[0007] S1. Obtain the standard information for power engineering design and preprocess the standard information for power engineering design;

[0008] S2. Construct a knowledge base based on the preprocessed normative information, determine the weight of the normative information according to its source, and sort the normative information in the knowledge base according to its weight.

[0009] S3. Obtain the preliminary design scheme of the power engineering project to be reviewed and preprocess the preliminary design scheme of the power engineering project;

[0010] S4. Based on the preprocessed preliminary design scheme, data retrieval and matching are performed in the knowledge base according to the weight of the standardized information.

[0011] S5. Input the preliminary design scheme of the power engineering that matches the normative information in the knowledge base into the trained large language model for verification and review.

[0012] S6. The large language model is approved as a compliant preliminary power design scheme. Logical verification and over-design verification are then performed to obtain the final compliant preliminary power engineering design scheme.

[0013] Furthermore, the aforementioned specification information includes national standards and specifications, enterprise specifications, historical typical cases, and equipment technical manuals;

[0014] Preprocessing of standardized information includes removing duplicate information and segmenting the standardized information.

[0015] The reasonableness of segmentation is determined by the similarity threshold of the text structure segmentation:

[0016] ;in: Similarity between text segments and standard chapters The set of keywords for the text to be segmented. For a pre-defined set of standard chapter keywords; when This indicates that the current text segmentation is reasonable.

[0017] Furthermore, the constructed knowledge base includes a procedure and specification knowledge base, a typical case knowledge base, and a template knowledge base.

[0018] Furthermore, determining the weight of regulatory information based on its source specifically includes:

[0019] ;in: For the fusion weight of the data source, The importance coefficient of the data source. The reliability coefficient of this data source. This represents the total number of data sources.

[0020] Furthermore, the preprocessing of the preliminary design scheme for the power engineering project specifically includes:

[0021] A document parsing engine is used to parse the preliminary design scheme of the power engineering project to be reviewed and convert it into a structure corresponding to the text information in the knowledge base.

[0022] Furthermore, the data retrieval matching involves calculating the cosine similarity between the text vector of the preliminary design scheme of the power engineering project under review and the text vector in the knowledge base. The text vectors of the preliminary design scheme of the power engineering project with a cosine similarity greater than a set value and the text vectors in the knowledge base are used as matching vectors, and the matching vectors are input into the large language model.

[0023] Furthermore, the logical verification of the large language model as a compliant preliminary power design scheme specifically includes:

[0024] Calculate logical metrics:

[0025] ; Score for logical consistency; The weight of the k-th logical relation; Let M be the deviation value of the k-th logical relation; M is the weighted summation, when... When the value is less than 0.8, the preliminary design scheme of the power project is not compliant.

[0026] Furthermore, the over-design verification of large language models as compliant preliminary power design schemes specifically includes:

[0027] Calculate the overdesign index:

[0028] ;in: For capacity matching coefficient, Main transformer rated capacity, For maximum load, α is the load growth margin, when The preliminary design scheme for the power project must be compliant; otherwise, it is not compliant.

[0029] The beneficial effects of this invention are as follows: By constructing a corresponding standard knowledge base in advance, analyzing the preliminary design scheme of power engineering, performing text matching, and inputting the matched text information into a trained large language model for standard prediction, the efficiency of reviewing the preliminary design scheme of power engineering can be effectively improved. Moreover, after the review by the large language model, logical and over-design judgments are performed, thereby further improving the accuracy of the review results. Attached Figure Description

[0030] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0031] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0032] The present invention will be further described in detail below:

[0033] This invention provides an intelligent review method for preliminary design schemes in power engineering, comprising the following steps:

[0034] S1. Obtain the standard information for power engineering design and preprocess the standard information for power engineering design;

[0035] S2. Construct a knowledge base based on the preprocessed normative information, determine the weight of the normative information according to its source, and sort the normative information in the knowledge base according to its weight.

[0036] S3. Obtain the preliminary design scheme of the power engineering project to be reviewed and preprocess the preliminary design scheme of the power engineering project;

[0037] S4. Based on the preprocessed preliminary design scheme, data retrieval and matching are performed in the knowledge base according to the weight of the standardized information.

[0038] S5. Input the preliminary design scheme of the power engineering that matches the normative information in the knowledge base into the trained large language model for verification and review.

[0039] S6. The preliminary power engineering design scheme, approved by the large language model, undergoes logical verification and over-design verification to obtain the final compliant preliminary power engineering design scheme. This method involves pre-constructing a corresponding normative knowledge base, parsing the preliminary power engineering design scheme, performing text matching, and inputting the matched text information into the trained large language model for normative prediction. This effectively improves the efficiency of reviewing the preliminary power engineering design scheme. Furthermore, after the large language model's review, logical verification and over-design judgment are performed, further enhancing the accuracy of the review results. Large language models include BERT series (e.g., ROBERTa) and GPT series (e.g., GPT-4).

[0040] In this embodiment, the specification information includes national standards and specifications, enterprise specifications, historical typical cases, and equipment technical manuals;

[0041] Preprocessing of standardized information includes removing duplicate information and segmenting the standardized information.

[0042] The reasonableness of segmentation is determined by the similarity threshold of the text structure segmentation:

[0043] ;in: Similarity between text segments and standard chapters The set of keywords for the text to be segmented. For a pre-defined set of standard chapter keywords; when This indicates that the current text segmentation is reasonable.

[0044] The constructed knowledge base includes a procedure and specification knowledge base (which stores technical standards at the national, industry, and enterprise levels), a typical case knowledge base (which stores historical high-quality design cases and review opinions), a template knowledge base (which stores standardized output templates such as review reports and meeting minutes), and an equipment parameter knowledge base (which stores the technical indicators and selection requirements of various power equipment).

[0045] In this embodiment, determining the weight of the standard information based on its source specifically includes:

[0046] ;in: The fusion weights for the data sources; The importance coefficient of the data source is generally set to 1.0 for mandatory national regulations, 0.9 for recommended industry standards, 0.8 for internal enterprise guidelines, and 0.6 for historical cases. The reliability coefficient of the data source is 1.0 for officially released and currently valid data, 0.8 for unofficial data that has been verified by experts, and 0.5 for unverified experience summaries. This represents the total number of data sources.

[0047] In this embodiment, the preprocessing of the preliminary design scheme for the power engineering project specifically includes:

[0048] A document parsing engine is used to parse the preliminary design scheme of the power engineering project to be reviewed and convert it into a structure corresponding to the text information in the knowledge base. The text parsing engine adopts existing technologies, such as the EasyDoc intelligent document parsing engine, the Deep-Seek-OCR parsing engine, and the MinerU engine.

[0049] In this embodiment, data retrieval matching involves calculating the cosine similarity between the text vector of the preliminary design scheme of the power engineering project under review and the text vector in the knowledge base. The text vector of the preliminary design scheme of the power engineering project with a cosine similarity greater than a set value and the text vector in the knowledge base are used as matching vectors, and the matching vectors are input into the large language model.

[0050] In this example, the logical verification of the large language model as a compliant preliminary power design scheme specifically includes:

[0051] Calculate logical metrics:

[0052] ; Score for logical consistency; The weight of the k-th logical relation; Let M be the deviation value of the k-th logical relation; M is the weighted summation, when... When the value is less than 0.8, the preliminary design scheme of the power project is not compliant.

[0053] The over-design verification of large language models that have been approved as compliant preliminary power design schemes specifically includes:

[0054] Calculate the overdesign index:

[0055] ;in: For capacity matching coefficient, Main transformer rated capacity, For maximum load, α is the load growth margin, when The preliminary design scheme for the power project must be compliant; otherwise, it is not compliant.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent review and verification of preliminary design schemes for power engineering projects, characterized in that: Includes the following steps: S1. Obtain the standard information for power engineering design and preprocess the standard information for power engineering design; S2. Construct a knowledge base based on the preprocessed normative information, determine the weight of the normative information according to its source, and sort the normative information in the knowledge base according to its weight. S3. Obtain the preliminary design scheme of the power engineering project to be reviewed and preprocess the preliminary design scheme of the power engineering project; S4. Based on the preprocessed preliminary design scheme, data retrieval and matching are performed in the knowledge base according to the weight of the standardized information. S5. Input the preliminary design scheme of the power engineering that matches the normative information in the knowledge base into the trained large language model for verification and review. S6. The large language model is approved as a compliant preliminary power design scheme. Logical verification and over-design verification are then performed to obtain the final compliant preliminary power engineering design scheme.

2. The intelligent review method for preliminary design schemes of power engineering projects according to claim 1, characterized in that: The specified information includes national standards and specifications, enterprise specifications, historical typical cases, and equipment technical manuals; Preprocessing of standardized information includes removing duplicate information and segmenting the standardized information. The reasonableness of segmentation is determined by the similarity threshold of the text structure segmentation: ;in: Similarity between text segments and standard chapters. The set of keywords for the text to be segmented. For a pre-defined set of standard chapter keywords; when This indicates that the current text segmentation is reasonable.

3. The intelligent review method for preliminary design schemes of power engineering projects according to claim 1, characterized in that: The constructed knowledge base includes a procedure and specification knowledge base, a typical case knowledge base, and a template knowledge base.

4. The intelligent review method for preliminary design schemes of power engineering projects according to claim 1, characterized in that: Determining the weight of regulatory information based on its source specifically includes: ;in: For the fusion weight of the data source, The importance coefficient of the data source. The reliability coefficient of this data source. This represents the total number of data sources.

5. The intelligent review method for preliminary design schemes of power engineering projects according to claim 1, characterized in that: The preliminary design scheme for power engineering is preprocessed, specifically including: A document parsing engine is used to parse the preliminary design scheme of the power engineering project to be reviewed and convert it into a structure corresponding to the text information in the knowledge base.

6. The intelligent review method for preliminary design schemes of power engineering projects according to claim 1, characterized in that: Data retrieval matching involves calculating the cosine similarity between the text vector of the preliminary design scheme of the power engineering project under review and the text vector in the knowledge base. The text vectors of the preliminary design scheme of the power engineering project with a cosine similarity greater than a set value and the text vectors in the knowledge base are used as matching vectors, and the matching vectors are input into the large language model.

7. The intelligent review method for preliminary design schemes of power engineering projects according to claim 1, characterized in that: The logical verification of the large language model as a compliant preliminary power design scheme specifically includes: Calculate logical metrics: ; Score for logical consistency; The weight of the k-th logical relation; Let M be the deviation value of the k-th logical relation; M is the weighted summation, when... When the value is less than 0.8, the preliminary design scheme of the power project is not compliant.

8. The intelligent review method for preliminary design schemes of power engineering projects according to claim 1, characterized in that: The over-design verification of large language models that have been approved as compliant preliminary power design schemes specifically includes: Calculate the overdesign index: ;in: For capacity matching coefficient, Main transformer rated capacity, For maximum load, α is the load growth margin, when The preliminary design scheme for the power project must be compliant; otherwise, it is not compliant.

Images