A distribution network engineering design drawing specification review method and system based on a large language model

Abstract

The application discloses a kind of based on large language model's network engineering design drawing specification review method and system, it is related to electric power engineering design review technical field, including: drawing analysis module, for receiving the multiple design drawings of network engineering, through large language model to each drawing is semantically analyzed, extract structured design element set, the design element includes equipment object and its attribute parameter, line object and its attribute parameter and annotation information;The drawing analysis module is assigned global unique identifier for each identified design element, and establishes cross-drawing association index;The application is mapped by establishing the association index of cross-drawing graph element, the information of same equipment, same line in multiple drawings is associated, realizes the automatic consistency check of equipment parameter, material quantity, pole position coordinate etc. Key information between multiple drawings, compared with traditional method, reduce the omission rate of cross-drawing consistency review, improve review efficiency.

Description

Technical Field

[0001] This invention relates to the field of power engineering design review technology, specifically to a method and system for reviewing distribution network engineering design drawings based on a large language model. Background Technology

[0002] Distribution network engineering is a crucial link in the power system connecting the transmission network and users, encompassing various project types such as 10kV transformer substation renovation and 110kV line construction and upgrading. Distribution network engineering design drawings are the core technical documents for project implementation, and their quality directly affects the safety, economy, and compliance of the construction. Currently, the review of distribution network engineering design drawings is still mainly based on manual review, and traditional manual review methods have the following prominent problems:

[0003] First, reviewers need to check the equipment parameters, material quantities, line routes, and other information on each drawing and cross-compare multiple drawings. The manual review of a typical 10kV transformer substation renovation project document (including 6-8 drawings and design specifications) usually takes more than 120 minutes, which is inefficient.

[0004] Secondly, during manual review, reviewers need to rely on memory or repeatedly review and compare the same equipment or parameters in multiple drawings, which can easily lead to omissions due to information overload. For example, the transformer capacity marked on the primary wiring diagram of a distribution area is 400kVA, but the rated capacity corresponding to the transformer model in the material statistics table is 315kVA. Such inconsistencies across drawings are easily overlooked during manual review.

[0005] Third, reviewers can usually only qualitatively determine whether a design deviates from the standard design, lacking a quantitative assessment method for the degree of deviation. For designs with small deviations (such as conductor cross-sections that are one specification larger than the recommended value), whether they should be considered violations often varies from person to person, and the review standards lack consistency.

[0006] Fourth, the renovation project of the transformer substation involves a comparison of the plans before and after the renovation. Reviewers need to determine whether the renovation is necessary and whether the scale of the renovation is reasonable. Due to the lack of standardized economic rationality assessment methods, different reviewers may give completely different review conclusions for the same renovation plan, making the review results highly subjective.

[0007] Existing technologies for intelligent review of power engineering drawings primarily utilize computer vision to identify graphic elements in CAD drawings or large language models to check the compliance of design specifications. However, these solutions are mainly geared towards substation engineering or general drawing review scenarios, lacking deep adaptation to the characteristics of distribution network engineering. Specifically: existing solutions mostly review individual drawings independently, failing to achieve collaborative review and consistency verification across multiple drawings; existing solutions lack deep integration with distribution network design databases, making it impossible to quantitatively assess the degree of deviation between design schemes and design specifications; and existing solutions do not support automated comparative review of schemes before and after modifications, failing to assist reviewers in determining the economic rationality of modification schemes. Summary of the Invention

[0008] To address the aforementioned technical issues, this paper provides a method and system for reviewing the design drawings of power distribution network projects based on a large language model. This technical solution solves the problems of independently reviewing only a single drawing and being unable to quantitatively assess the degree of deviation between the design scheme and the standard design.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0010] A system for reviewing design drawings for power distribution network projects based on a large language model, comprising:

[0011] The drawing parsing module receives multiple design drawings of the power distribution network project, performs semantic parsing on each drawing using a large language model, and extracts a set of structured design elements. The design elements include equipment objects and their attribute parameters, line objects and their attribute parameters, and annotation information. The drawing parsing module assigns a globally unique identifier to each identified design element and establishes a cross-drawing association index.

[0012] The standard knowledge base stores standard clause knowledge, typical design parameter knowledge, and review rule knowledge. The standard clause knowledge consists of clause blocks that are converted into high-dimensional vectors and stored in a vector database after being divided into clause blocks according to current distribution network design specifications through a text vectorization model. The typical design parameter knowledge consists of a set of standard parameters obtained by structurally decomposing typical schemes in the typical distribution network design library. The review rule knowledge consists of executable rules encoded in the form of condition-action correspondence.

[0013] The rules engine module is used to receive the set of structured design elements output by the drawing parsing module, match the corresponding review rules from the review rule knowledge according to the type and attributes of the design elements, perform multi-drawing consistency verification and quantitative evaluation of design deviation in sequence, and output structured review conclusions.

[0014] The review results output module is used to summarize the review conclusions of the rule engine module and generate a structured review report containing a list of violations and a deviation analysis of the design.

[0015] Preferably, the cross-drawing association index includes three types of association relationships:

[0016] For devices with the same name, when the same device is identified in different drawings, the device name and device type are matched to establish an association between the globally unique identifiers of the same device in each drawing;

[0017] Electrical connection association: Based on the wiring connections in the primary wiring diagram, establish the electrical topology diagram between devices;

[0018] Physical location association: Based on the pole coordinate information in the plan layout and pole position list, establish a physical location mapping between equipment.

[0019] Preferably, in the standard knowledge base, each clause block of the standard clause knowledge includes a number, the original text of the clause, the applicable voltage level, the applicable power supply area type, and the relevant parameter constraints;

[0020] Each typical scheme of the above-mentioned design parameter knowledge corresponds to a specific voltage level and power supply area type, including a standard parameter set of recommended conductor cross-section, transformer capacity, switch selection and pole type combination. The standard parameter set is stored as both structured data and text description.

[0021] Each rule in the review rule knowledge includes a rule number, triggering conditions, inspection logic, judgment threshold, and violation level.

[0022] Preferably, the multi-drawing consistency check performed by the rule engine module includes:

[0023] The equipment parameter consistency check iterates through all pairs of equipment with the same name in the cross-drawing association index, compares the key parameters of the same equipment in different drawings one by one, and marks the equipment parameter inconsistency violation when the parameters do not match.

[0024] Quantity consistency check: count the number of times each type of equipment appears in each drawing, calculate the sum of the number of times each type of equipment appears in each drawing, compare the registered quantity of each type of equipment in the material statistics table with the sum of the number of times it appears, and mark it as a quantity inconsistency violation when the difference between the two is not equal to zero;

[0025] Topology consistency check: Extract the electrical topology adjacency table from the primary wiring diagram, and extract the physical topology adjacency table from the floor plan and pole location list. Compare whether the topology structures of the two are isomorphic. If inconsistency exists, mark it as a topology inconsistency violation.

[0026] A method for reviewing the design drawings of a power distribution network project based on a large language model, comprising the following steps: S1, receiving multiple design drawings of a power distribution network project, performing semantic analysis on each drawing using a large language model, extracting a set of structured design elements, assigning a globally unique identifier to each design element, and establishing a cross-drawing association index based on the globally unique identifier;

[0027] Step S2: Based on the cross-drawing association index, perform multi-drawing consistency verification, including equipment parameter consistency verification, quantity consistency verification, and topology consistency verification;

[0028] Step S3: Extract key design parameters from the set of design elements, compare them with the typical design parameters in the pre-built standard knowledge base, calculate the deviation index of each parameter, and output the graded review conclusion based on the deviation index.

[0029] Step S4: Summarize the review conclusions of steps S2 and S3 to generate a structured review report.

[0030] Preferably, in step S2:

[0031] The device parameter consistency verification is as follows: traverse all pairs of devices with the same name in the cross-drawing association index, compare the key parameters of the same device in different drawings one by one, and mark the device parameter inconsistency violation when the key parameters of the same device do not match in different drawings.

[0032] The quantity consistency verification is as follows: For the j-th type of equipment, count the number of times this type of equipment appears in each drawing, calculate the sum of the number of times this type of equipment appears in each drawing, compare the registered quantity of this type of equipment in the material statistics table with the sum of the number of times it appears, and mark it as a quantity inconsistency violation when the difference between the two is not equal to zero;

[0033] The topology consistency verification is as follows: extract the electrical topology adjacency table from the primary wiring diagram, extract the physical topology adjacency table from the plan layout diagram and the pole location list, compare whether the topology structures of the two are isomorphic, and mark the inconsistency as a topology inconsistency violation when there is an inconsistency.

[0034] Preferably, the calculation of the deviation index of each parameter in step S3 includes:

[0035] For discrete parameters, the relative deviation between the design value and the closest standard value in the standard sequence is calculated as the deviation index of the parameter. When the design value is in the standard sequence, the deviation index of the parameter is zero.

[0036] For continuously valued parameters, the deviation index of the parameter is zero when the design value falls within the allowable range. When the design value exceeds the allowable range, the relative deviation between the design value and the nearest allowable range boundary value is calculated as the deviation index of the parameter.

[0037] Preferably, step S3 further includes:

[0038] Based on the deviation index of each parameter and the preset weighting coefficient, a comprehensive deviation index is calculated, which is the weighted average of the deviation indices of each parameter.

[0039] Based on the comprehensive deviation index, the following graded review conclusions are output: When the comprehensive deviation index is zero, it is determined to be fully compliant with the design; when the comprehensive deviation index is greater than zero and less than or equal to the first threshold, it is determined to be a slight deviation, and a list of deviation parameters and suggested adjustment directions are output; when the comprehensive deviation index is greater than the first threshold and less than or equal to the second threshold, it is determined to be a significant deviation, and a deviation analysis and modification suggestions are output, requiring the designer to provide reasons for the deviation; when the comprehensive deviation index is greater than the second threshold, it is determined to be a serious deviation, and the review is marked as unsuccessful.

[0040] The first threshold is 5%, and the second threshold is 15%.

[0041] Preferably, between steps S3 and S4, for the transformer substation renovation project, the existing transformer rated capacity and current maximum load are extracted from the pre-renovation design drawings and current status survey data, and the post-renovation transformer rated capacity and predicted maximum load are extracted from the post-renovation design drawings.

[0042] Calculate the load rate before and after the renovation based on the maximum load before and after the renovation and the rated capacity of the transformer, and calculate the difference between the two as the load margin improvement amount.

[0043] The power supply reliability improvement index is calculated based on the average power outage time of the system before and after the renovation. The power supply reliability improvement index is the difference between the average power outage time of the system before the renovation and the average power outage time of the system after the renovation, divided by the average power outage time of the system before the renovation.

[0044] The load margin improvement and the power supply reliability improvement index are weighted and combined to obtain the comprehensive index of the necessity of the renovation.

[0045] Based on the comprehensive index of necessity for renovation, the following assessment conclusions are output: when the comprehensive index of necessity for renovation is greater than or equal to 0.2, the necessity for renovation is deemed sufficient; when the comprehensive index of necessity for renovation is greater than or equal to 0.1 and less than 0.2, the necessity for renovation is deemed somewhat necessary but requires further justification; when the comprehensive index of necessity for renovation is less than 0.1, the necessity for renovation is deemed insufficient.

[0046] Preferably, the method further includes step S5:

[0047] Regularly monitor information updates through the channels where the standards are published, and identify new, revised, and repealed clauses in the new version of the standards;

[0048] For new clauses, clause blocks are automatically generated and added to the vector database; for revised clauses, the old version is marked as obsolete and the new version is added to the database; for obsolete clauses, they are marked as obsolete and historical versions are retained for retrospective purposes.

[0049] After the specification is updated, the system automatically scans the existing review rule base to identify the affected rule entries. When the judgment threshold of a rule references the revised specification clause, the system prompts the administrator to confirm whether to update the threshold.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0051] This invention establishes a cross-drawing element association index to map information about the same equipment or line scattered across multiple drawings, enabling automated consistency verification of key information such as equipment parameters, material quantities, and pole coordinates across multiple drawings. Compared to the traditional method of reviewing each drawing independently, this invention can effectively detect cross-drawing errors such as inconsistencies between parameters marked in drawing A and those in drawing B, significantly reducing the omission rate in cross-drawing consistency reviews and improving review efficiency.

[0052] By deeply integrating with the typical distribution network design database, key parameters in the design scheme, including conductor cross-section, transformer capacity, and switch rated current, are compared with the recommended values ​​of the typical design. The deviation index is calculated, and review conclusions are output in a tiered manner. This transforms the original qualitative judgment of "whether it deviates from the typical design" into a quantitative assessment of deviation, thereby eliminating subjective differences and making the review standards more consistent.

[0053] By extracting information on differences in equipment parameters, grid structure, and load indicators before and after the transformer substation renovation, and combining this with load forecast data to calculate the economic rationality indicators of the renovation plan, this assists reviewers in determining whether the renovation is necessary and whether its scale is reasonable, reducing the subjectivity of review conclusions. Furthermore, it allows for the structured storage of review knowledge, such as distribution network design rules, cross-drawing consistency constraints, and renovation rationality criteria, in a knowledge base. This enables the digital accumulation and inheritance of review experts' tacit experience, preventing fluctuations in review capabilities due to personnel changes. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the overall system architecture of the present invention;

[0055] Figure 2 This is a schematic diagram of the cross-drawing paper association index of the present invention;

[0056] Figure 3 This is a flowchart of the multi-drawing consistency verification process of the present invention;

[0057] Figure 4This is a flowchart illustrating the deviation quantification evaluation process of the present invention;

[0058] Figure 5 This is a flowchart of the comparative review process for the transformer substation renovation scheme of the present invention;

[0059] Figure 6 This is a flowchart illustrating the overall process of reviewing and executing this invention. Detailed Implementation

[0060] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0061] Reference Figure 1 and Figure 2 As shown, the power distribution network engineering design drawing specification review system includes four core components: drawing parsing module, specification knowledge base, rule engine module, and review result output module.

[0062] The drawing parsing module receives multiple design drawings for the power distribution network project, including transformer substation primary wiring diagrams, transformer substation layout plans, pole location lists, material statistics tables, overhead line route diagrams, and cable laying diagrams. For each drawing, visual recognition and semantic parsing are first performed using a large visual language model to extract a set of structured design elements. For CAD format drawings, their vector image metadata is directly parsed to obtain element types, coordinate positions, attribute annotations, and layer information; for PDF or scanned drawings, image preprocessing is performed first, followed by content recognition using a large visual language model.

[0063] The design elements extracted from each drawing include: equipment objects (such as transformers, switches, poles, conductors, etc.) and their attribute parameters (such as model, rated capacity, rated voltage, conductor cross-section, etc.), line objects (such as overhead line sections, cable sections) and their attribute parameters (such as length, conductor type, laying method), and labeling information (such as pole number, line name, voltage level, etc.).

[0064] Cross-drawing association index is the core data structure of this embodiment. The system assigns a globally unique identifier (Element ID) to each identified design element and establishes the following three association relationships:

[0065] Firstly, the same-named equipment is associated. When the same equipment is identified in different drawings (e.g., transformer T1 in a primary wiring diagram and transformer T1 in a floor plan), the equipment name and type are matched to establish an association between their Element IDs.

[0066] Secondly, electrical connection association: based on the wiring connections in the primary wiring diagram, an electrical topology diagram between devices is established. For example, the overhead line section connected to the high-voltage side of the transformer and the distribution line section connected to the low-voltage side form an electrical path from the power source to the load.

[0067] Third, physical location association: Based on the pole coordinate information in the layout plan and pole position list, a physical location mapping between equipment is established. Equipment on the same pole and line segments between adjacent poles are all associated with each other through pole numbers.

[0068] The three relationships mentioned above together constitute a cross-drawing association index, which enables the system to automatically trace the corresponding information of the device in other drawings and perform consistency verification when an abnormality is found in the parameter of a certain device on any drawing.

[0069] The canonical knowledge base consists of three types of knowledge:

[0070] The first category is the knowledge of regulatory clauses. Current power distribution network design standards such as DL / T 5729, GB / T 45418, GB 50053, GB 50217, and GB50060 are divided into blocks according to clauses. Each block contains a clause number, the original text of the clause, applicable conditions, and constraint parameters. Each clause block is converted into a high-dimensional vector using a text vectorization model and stored in a vector database. For example, clause 6.3.2 of DL / T 5729-2023 states: "Cable trench covers should be made of reinforced concrete, with a thickness not less than 80mm, applicable to voltage levels of 10kV and above." This text is fed into a pre-trained Chinese vectorization model (e.g., BERT-based architecture, 768 hidden layer dimensions). The model outputs a 768-dimensional floating-point vector, where each number corresponds to a semantic feature dimension. Therefore, clauses with similar semantics are similar in the vector space.

[0071] The second category is the knowledge of typical design parameters, which involves the structured breakdown of typical schemes from the State Grid or China Southern Power Grid distribution network typical design databases. Each typical scheme corresponds to a specific voltage level, power supply area type, and load level, and includes a set of standard parameters such as recommended conductor cross-sections, transformer capacities, switch selection, and pole type combinations. The typical design parameters are stored simultaneously as structured data (for accurate comparison) and text descriptions (for vectorized retrieval).

[0072] The third category is review rule knowledge, which encodes the experience-based judgment logic in the review process into executable rules. Each rule includes a trigger condition, inspection logic, judgment threshold, and violation level. For example, the trigger condition for the rule "conductor cross-section verification" is "identification of an overhead line segment," the inspection logic is "whether the conductor cross-section meets the minimum cross-section requirements for this power supply area," the judgment threshold is based on the provisions of DL / T 5729, and the violation level is divided into two levels: "serious violation" and "general deviation."

[0073] The rules engine module receives the set of structured design elements output by the drawing parsing module, matches the corresponding review rules according to the type and attributes of the design elements, and then performs three types of review tasks in sequence: multi-drawing consistency verification, quantitative assessment of design deviation, and comparative review of modification schemes.

[0074] Example 2, as Figure 2 and 3 As shown in this embodiment, the implementation process of cross-drawing consistency verification is illustrated.

[0075] A typical 10kV transformer substation renovation project usually includes the following drawings: substation primary wiring diagram, substation layout plan, pole location list, material statistics table, and sometimes overhead line route diagram and low-voltage line plan. These drawings logically describe the same project plan, therefore the equipment parameters and quantities in each drawing must be consistent.

[0076] The system first identifies pairs of elements describing the same entity in different drawings using the cross-drawing association index in Example 1. Then, it automatically verifies the following three types of consistency constraints:

[0077] Category 1: Equipment parameter consistency. Key parameters of the same equipment must be consistent across different drawings. For example, in the transformer substation primary wiring diagram, the system identifies transformer T1 with a rated capacity of "400kVA." Simultaneously, the material statistics table identifies transformer S13-M-400 / 10, whose corresponding rated capacity should also be 400kVA. If these two parameters do not match, it is marked as a "device parameter inconsistency" violation, and the specific drawing location and parameter difference are indicated.

[0078] The consistency check covers the following equipment parameters: transformer rated capacity and model, switchgear rated current and type, conductor model and cross-section, cable model and cross-section, and pole type and length. The system traverses all pairs of equipment with the same name in the cross-drawing association index and compares their key parameters one by one.

[0079] Category Two: Quantity Consistency. The number of equipment items in the material statistics table should match the actual number of equipment items appearing in each drawing. The system counts the number of times each type of equipment appears in the wiring diagram and floor plan, comparing it with the quantity in the material statistics table. If a discrepancy is found, it is marked as a "Quantity Inconsistency" violation.

[0080] The quantity consistency check uses the following method: For the first Class of devices, assuming it is in the first The number of images identified in the drawing is The registered quantity of this equipment in the material statistics table is: Then the quantity deviation Defined as:

[0081]

[0082] in This represents the total number of drawings included in the statistics. As an indicator variable, when the first The drawing contains the first The value is 1 for this type of device, and 0 otherwise. When the discrepancy is recorded, it is marked as a violation due to quantity inconsistency; the absolute value of the deviation is considered. Record it in the review report.

[0083] The third category is topological consistency. The electrical connections in the primary wiring diagram should be consistent with the route in the layout plan. For example, if the primary wiring diagram shows that line L1 starts from the 10kV busbar of the substation, passes through pole 1 and pole 2, and reaches transformer T1 at pole 3, the layout plan should also show the same pole sequence and terminal equipment.

[0084] Topology consistency verification is achieved by comparing the electrical connection relationships and physical location relationships in the cross-drawing association index. The system extracts the electrical topology adjacency table from the primary wiring diagram and the physical topology adjacency table from the floor plan and pole location details table; the two should be isomorphic. If a topology inconsistency occurs (e.g., a section of line exists in the primary wiring diagram but is missing in the floor plan), it is marked as a "topology inconsistency" violation.

[0085] Example 3, as Figure 4 As shown in this embodiment, a quantitative comparison is performed between the design scheme and typical setting standards.

[0086] In the distribution network design database, each typical scheme provides a set of recommended design parameters for a specific power supply area type (e.g., Class A power supply area is a high-load-density urban area, and Class D power supply area is a rural area) and voltage level. These parameters are divided into two categories: discrete parameters and continuous parameters.

[0087] Discrete parameters are those whose values ​​can only be taken from a finite series of standard values. For example, the standard series for transformer rated capacity is 100, 200, 315, 400, 500, 630, 800, 1000 kVA; the standard series for conductor cross-section is 35, 50, 70, 95, 120, 150, 185, 240 mm². The values ​​of these parameters in the design scheme should fall within the standard series.

[0088] Continuously measured parameters refer to parameters that take values ​​within a continuous range. For example, the power supply radius (length of the main line) is usually required to be within a certain range, and the conductor sag is taken within a certain range based on meteorological conditions.

[0089] For the Design parameters Define its deviation index as follows:

[0090] For discrete-valued parameters:

[0091]

[0092] in, For the standard sequence and The closest standard value. If If it is already in the standard sequence, then .

[0093] For continuously variable parameters, let their allowable range be... :

[0094]

[0095] in, Distance The most recent allowed range boundary value (when) At that time ,when At that time ).

[0096] The physical meaning of the above deviation index: For discrete parameters, This reflects the relative magnitude of the design value deviating from the most recent standard value; for continuous parameters, This reflects the relative extent to which the design value exceeds the allowable range. The numerator and denominator have the same physical dimensions (as with the parameter). (same), therefore This is a dimensionless percentage value.

[0097] The deviation of a single parameter only reflects the local compliance status. To make a comprehensive evaluation of the overall design scheme, a comprehensive deviation index is defined. as follows:

[0098]

[0099] in, The total number of design parameters participating in the evaluation. For the first The weighting coefficients for each parameter are determined based on their impact on the safety and reliability of the distribution network project. For example, transformer capacity and conductor cross-section have relatively high weightings (recommended). The rated current of the switch is the next best (recommended). Other parameters, such as rod type selection, are taken with reference weights. ).

[0100] The physical meaning is the weighted average of the deviations of each parameter, and the dimension is a dimensionless percentage. This indicates that the design scheme fully meets the design requirements; The larger the value, the more serious the deviation of the design scheme from the original design.

[0101] The system according to The numerical range outputs the hierarchical review conclusion:

[0102] At that time, it was determined to be "fully compliant with the established rules"; If the deviation is deemed "slight", a list of deviation parameters and suggested adjustment direction will be output. If the deviation is deemed "significant," a detailed deviation analysis and modification suggestions will be output, requiring the designer to provide the reasons for the deviation. If the design is deemed "seriously off-target," it is marked as failing the review and requires a redesign.

[0103] The above thresholds can be configured according to the specific power supply area type and management requirements.

[0104] Example 4, as Figure 5 As shown in this embodiment, an automated comparative review method for the before-and-after schemes in a transformer substation renovation project is disclosed.

[0105] The core review objectives for transformer substation renovation projects are twofold: first, whether the renovation plan is necessary, i.e., whether the existing transformer substation has insufficient power supply capacity or safety hazards; and second, whether the scale of the renovation is reasonable, i.e., whether the renovation investment matches the expected results.

[0106] The system performs the following comparative review process for transformer substation renovation projects:

[0107] Step A: Extract parameters from the pre-modification plan. From the pre-modification design drawings and current status survey data, extract the following key parameters: number of existing transformers and rated capacity. (Unit: kVA), existing line conductor cross-section and length, current maximum load (Unit: kW), Current power supply radius (Unit: km), current line operating years and defect records.

[0108] Step B: Extract parameters of the modified scheme. From the modified design drawings, extract the corresponding parameters: the rated capacity of the modified transformer. (Unit: kVA), conductor cross-section and length after modification, predicted maximum load (Unit: kW) Power supply radius after renovation (Unit: km).

[0109] Step C: Calculate the load margin change, and define the load rate before the modification as follows:

[0110]

[0111] in, The power factor typically ranges from 100 to 1000. to Similarly, calculate the load rate after the modification. .

[0112] Define the load margin improvement amount for:

[0113]

[0114] The physical meaning of this indicator is the decrease in load rate after the renovation. This indicates that the transformer's load margin has increased and its power supply capacity has been improved after the upgrade. and Both are the ratio of active power to apparent power.

[0115] Step D: Calculate the improvement in power supply reliability. Based on the changes in the power grid structure before and after the upgrade, calculate the improvement in the System Average Outage Time Index (SAIDI). Let the SAIDI before the upgrade be... (Unit: hours / household / year), the predicted SAIDI after renovation is: (Unit: hours / household / year), defining the power supply reliability improvement index. for:

[0116]

[0117] in, The value represents the reduction in power outage time after the renovation compared to before the renovation. If... This indicates that the power supply reliability has improved after the upgrade.

[0118] Step E: Calculate the comprehensive index of necessity for renovation, weighting the improvement of load margin and the improvement of power supply reliability to define the comprehensive index of necessity for renovation. for:

[0119]

[0120] in, and Let be the weighting coefficient, satisfying Typical value When the primary goal of a certain type of modification is to solve overload, the size can be appropriately increased. When the primary goal is to solve frequent power outages, the power consumption can be appropriately increased. .

[0121] The physical meaning is the degree of comprehensive improvement of the renovation plan in terms of both load margin and power supply reliability. The higher the value, the stronger the necessity for transformation.

[0122] The system according to Value output review conclusion: At that time, the necessity of the renovation was deemed sufficient; At that time, it was determined that the renovation was necessary to some extent, but further evidence was needed. If the necessity for modification is deemed insufficient, it is recommended that the designers provide supplementary explanations or adjust the plan.

[0123] Example 5, as Figure 5 As shown in this embodiment, taking a 10kV transformer substation renovation project in Ezhou as an example, the project includes four drawings: a primary wiring diagram of the substation, a layout plan of the substation, a list of pole locations, and a material statistics table, as well as a design specification.

[0124] After receiving the above drawings and documents, the system will perform a review according to the following steps:

[0125] Step 1: Drawing parsing. For each drawing, a large language model is used for semantic parsing to extract equipment objects and their attribute parameters. In this embodiment, a total of 2 distribution transformers, 3 pole-mounted switches, 24 poles, 18 overhead line sections, and 4 cable sections were identified. Simultaneously, an Element ID was assigned to each identified object, establishing a cross-drawing association index.

[0126] Step 2, Multi-drawing Linkage Consistency Verification: Based on the cross-drawing association index, the three types of consistency verification in Example 2 are performed. In this example, the material statistics table shows that there are 2 transformers, which is consistent with the 2 transformers actually identified in the drawings; however, it is found that the capacity of transformer No. 1 in the primary wiring diagram is marked as 400kVA, while the rated capacity of the corresponding model in the material statistics table is 315kVA, triggering a "device parameter inconsistency" violation alarm.

[0127] Step 3: Evaluation of Design Deviation. Key parameters in the design scheme are extracted and compared with typical schemes for the corresponding power supply area type in the design library. In this embodiment, the transformer substation belongs to Class D power supply area, the recommended transformer capacity is 400kVA, and the design value is also 400kVA. The deviation is... The recommended cross-sectional area for the main conductor is 120mm², while the design value is 95mm², with a deviation of [missing information]. Overall Deviation Index The calculated weight of each parameter is 8.5%, which is considered a "significant deviation". The system outputs a modification suggestion that the conductor cross-section is too small.

[0128] Step 4: Compare and review the renovation plans, extract parameters from the plans before and after the renovation, and calculate the comprehensive index of the necessity of the renovation. In this embodiment, the transformer load rate before modification is... After renovation , Before renovation Hours / household / year, post-renovation forecast Hours / Household / Year .Pick ,but The necessity of the renovation was deemed sufficient.

[0129] Step 5: Generate a review report. The system summarizes the review conclusions from Steps 2 to 4 and generates a structured review report. The report includes: a list of violations (including specific locations and suggested modifications), a deviation analysis of design specifications (including details of deviations for each parameter), an assessment of the necessity of modifications, and overall review opinions.

[0130] In this embodiment, the entire review process took approximately 5 minutes and 30 seconds, a significant reduction compared to the more than 120 minutes required for manual review. The review identified one issue with inconsistent equipment parameters and one issue with conductor cross-sections deviating from the design specifications; both were confirmed and corrected by the designers before passing the second review.

[0131] Regarding the construction method and dynamic update mechanism of the above-mentioned standard knowledge base, in the initial construction phase: the system collects the current standard standards related to distribution network engineering design, including but not limited to DL / T 5729, GB / T 45418, GB 50053, GB50217, GB 50060, as well as typical distribution network design schemes and distribution network technical principles of State Grid or China Southern Power Grid.

[0132] Each specification document is segmented by clause. Each clause block contains the following structured information: clause number (e.g., "DL / T 5729-2023 Clause 6.3.2"), the original clause text, applicable voltage level, applicable power supply area type, and the names of the parameters involved and their constraints. The original clause text is converted into a high-dimensional vector using a text vectorization model and stored in a vector database along with the structured metadata.

[0133] For each typical scheme in the design library, its applicable voltage level, power supply area type, and all recommended parameters are extracted and stored in a structured format. Simultaneously, the textual description of the scheme is vectorized.

[0134] The review rules are coded in the form of "condition-action". Each rule includes: rule number, rule name, triggering condition (which types of design elements are matched), inspection logic (the specific verification algorithm), judgment threshold (extracted from the specification or set based on engineering experience), and violation level (serious violation / general deviation / suggestion).

[0135] Dynamic update phase: The system periodically monitors information updates from the specification release channels. When a new version of the specification is detected, the following update process is automatically executed:

[0136] Identify new, revised, and repealed clauses in the new specification. For new clauses, automatically generate clause blocks and add them to the vector database; for revised clauses, mark the old version as "repealed" and add the new version to the database; for repealed clauses, mark them as "repealed" but retain historical versions for traceability.

[0137] After the specification is updated, the system automatically scans the existing review rule base to identify the affected rule entries. If a rule's judgment threshold references a revised specification clause, the system prompts the administrator to confirm whether to update the threshold. Similarly, for updates to the design library, the system automatically identifies the affected typical scheme parameters and triggers a synchronous update of the design comparison benchmark.

[0138] Through the aforementioned dynamic update mechanism, the specification knowledge base always remains consistent with the latest standards and specifications, avoiding review errors caused by outdated specifications.

[0139] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A system for reviewing design drawings of power distribution network projects based on a large language model, characterized in that, include: The drawing parsing module receives multiple design drawings of the power distribution network project, performs semantic parsing on each drawing using a large language model, and extracts a set of structured design elements. The design elements include equipment objects and their attribute parameters, line objects and their attribute parameters, and annotation information. The drawing parsing module assigns a globally unique identifier to each identified design element and establishes a cross-drawing association index. The standard knowledge base stores standard clause knowledge, typical design parameter knowledge, and review rule knowledge. The standard clause knowledge consists of clause blocks that are converted into high-dimensional vectors and stored in a vector database after being divided into clause blocks according to current distribution network design specifications through a text vectorization model. The typical design parameter knowledge consists of a set of standard parameters obtained by structurally decomposing typical schemes in the typical distribution network design library. The review rule knowledge consists of executable rules encoded in the form of condition-action correspondence. The rules engine module is used to receive the set of structured design elements output by the drawing parsing module, match the corresponding review rules from the review rule knowledge according to the type and attributes of the design elements, perform multi-drawing consistency verification and quantitative evaluation of design deviation in sequence, and output structured review conclusions. The review results output module is used to summarize the review conclusions of the rule engine module and generate a structured review report containing a list of violations and a deviation analysis of the design.

2. The power distribution network engineering design drawing specification review system based on a large language model according to claim 1, characterized in that: The cross-drawing association index includes three types of association relationships: For devices with the same name, when the same device is identified in different drawings, the device name and device type are matched to establish an association between the globally unique identifiers of the same device in each drawing; Electrical connection association: Based on the wiring connections in the primary wiring diagram, establish the electrical topology diagram between devices; Physical location association: Based on the pole coordinate information in the plan layout and pole position list, establish a physical location mapping between equipment.

3. The power distribution network engineering design drawing specification review system based on a large language model according to claim 1, characterized in that: In the aforementioned standard knowledge base, each clause block of the standard clause knowledge includes a number, the original text of the clause, the applicable voltage level, the applicable power supply area type, and the relevant parameter constraints. Each typical scheme of the above-mentioned design parameter knowledge corresponds to a specific voltage level and power supply area type, including a standard parameter set of recommended conductor cross-section, transformer capacity, switch selection and pole type combination. The standard parameter set is stored as both structured data and text description. Each rule in the review rule knowledge includes a rule number, triggering conditions, inspection logic, judgment threshold, and violation level.

4. The power distribution network engineering design drawing specification review system based on a large language model according to claim 1, characterized in that: The multi-drawing consistency check performed by the rule engine module includes: The equipment parameter consistency check iterates through all pairs of equipment with the same name in the cross-drawing association index, compares the key parameters of the same equipment in different drawings one by one, and marks the equipment parameter inconsistency violation when the parameters do not match. Quantity consistency check: count the number of times each type of equipment appears in each drawing, calculate the sum of the number of times each type of equipment appears in each drawing, compare the registered quantity of each type of equipment in the material statistics table with the sum of the number of times it appears, and mark it as a quantity inconsistency violation when the difference between the two is not equal to zero; Topology consistency check: Extract the electrical topology adjacency table from the primary wiring diagram, and extract the physical topology adjacency table from the floor plan and pole location list. Compare whether the topology structures of the two are isomorphic. If inconsistency exists, mark it as a topology inconsistency violation.

5. A method for reviewing the specifications of power distribution network engineering design drawings based on a large language model, applied to the system described in any one of claims 1-4, characterized in that: Step S1: Receive multiple design drawings of the power distribution network project, perform semantic parsing on each drawing through a large language model, extract a set of structured design elements, assign a globally unique identifier to each design element, and establish a cross-drawing association index based on the globally unique identifier. Step S2: Based on the cross-drawing association index, perform multi-drawing consistency verification, including equipment parameter consistency verification, quantity consistency verification, and topology consistency verification; Step S3: Extract key design parameters from the set of design elements, compare them with the typical design parameters in the pre-built standard knowledge base, calculate the deviation index of each parameter, and output the graded review conclusion based on the deviation index. Step S4: Summarize the review conclusions of steps S2 and S3 to generate a structured review report.

6. The method for reviewing the design drawings of power distribution network projects based on a large language model according to claim 5, characterized in that: In step S2: The device parameter consistency verification is as follows: traverse all pairs of devices with the same name in the cross-drawing association index, compare the key parameters of the same device in different drawings one by one, and mark the device parameter inconsistency violation when the key parameters of the same device do not match in different drawings. The quantity consistency verification is as follows: For the j-th type of equipment, count the number of times this type of equipment appears in each drawing, calculate the sum of the number of times this type of equipment appears in each drawing, compare the registered quantity of this type of equipment in the material statistics table with the sum of the number of times it appears, and mark it as a quantity inconsistency violation when the difference between the two is not equal to zero; The topology consistency verification is as follows: extract the electrical topology adjacency table from the primary wiring diagram, extract the physical topology adjacency table from the plan layout diagram and the pole location list, compare whether the topology structures of the two are isomorphic, and mark the inconsistency as a topology inconsistency violation when there is an inconsistency.

7. The method for reviewing the design drawings of power distribution network projects based on a large language model according to claim 5, characterized in that: The calculation of the deviation index of each parameter in step S3 includes: For discrete parameters, the relative deviation between the design value and the closest standard value in the standard sequence is calculated as the deviation index of the parameter. When the design value is in the standard sequence, the deviation index of the parameter is zero. For continuously valued parameters, the deviation index of the parameter is zero when the design value falls within the allowable range. When the design value exceeds the allowable range, the relative deviation between the design value and the nearest allowable range boundary value is calculated as the deviation index of the parameter.

8. The method for reviewing the design drawings of power distribution network projects based on a large language model according to claim 7, characterized in that: Step S3 further includes: Based on the deviation index of each parameter and the preset weighting coefficient, a comprehensive deviation index is calculated, which is the weighted average of the deviation indices of each parameter. Based on the comprehensive deviation index, the following graded review conclusions are output: When the comprehensive deviation index is zero, it is determined to be fully compliant with the design; when the comprehensive deviation index is greater than zero and less than or equal to the first threshold, it is determined to be a slight deviation, and a list of deviation parameters and suggested adjustment directions are output; when the comprehensive deviation index is greater than the first threshold and less than or equal to the second threshold, it is determined to be a significant deviation, and a deviation analysis and modification suggestions are output, requiring the designer to provide reasons for the deviation; when the comprehensive deviation index is greater than the second threshold, it is determined to be a serious deviation, and the review is marked as unsuccessful. The first threshold is 5%, and the second threshold is 15%.

9. The method for reviewing the design drawings of power distribution network projects based on a large language model according to claim 5, characterized in that: Between steps S3 and S4, for the transformer substation renovation project, the existing transformer rated capacity and current maximum load are extracted from the pre-renovation design drawings and current status survey data, and the post-renovation transformer rated capacity and predicted maximum load are extracted from the post-renovation design drawings. Calculate the load rate before and after the renovation based on the maximum load before and after the renovation and the rated capacity of the transformer, and calculate the difference between the two as the load margin improvement amount. The power supply reliability improvement index is calculated based on the average power outage time of the system before and after the renovation. The power supply reliability improvement index is the difference between the average power outage time of the system before the renovation and the average power outage time of the system after the renovation, divided by the average power outage time of the system before the renovation. The load margin improvement and the power supply reliability improvement index are weighted and combined to obtain the comprehensive index of the necessity of the renovation. The necessity assessment conclusion for the renovation is output based on the comprehensive index of the necessity for renovation: when the comprehensive index of the necessity for renovation is greater than or equal to 0.2, the necessity for renovation is deemed sufficient; When the comprehensive index of necessity for renovation is greater than or equal to 0.1 and less than 0.2, it is determined that the renovation is necessary to a certain extent but requires further demonstration; when the comprehensive index of necessity for renovation is less than 0.1, it is determined that the necessity for renovation is insufficient.

10. The method for reviewing the design drawings of power distribution network projects based on a large language model according to claim 5, characterized in that: It also includes step S5: Regularly monitor information updates through the channels where the standards are published, and identify new, revised, and repealed clauses in the new version of the standards; For new clauses, clause blocks are automatically generated and added to the vector database; for revised clauses, the old version is marked as obsolete and the new version is added to the database; for obsolete clauses, they are marked as obsolete and historical versions are retained for retrospective purposes. After the specification is updated, the system automatically scans the existing review rule base to identify the affected rule entries. When the judgment threshold of a rule references the revised specification clause, the system prompts the administrator to confirm whether to update the threshold.

Images