Power engineering drawing defect positioning method and device, electronic equipment and storage medium

Abstract

The application relates to the technical field of computer vision, and discloses a power engineering drawing defect positioning method and device, an electronic device and a storage medium. The method comprises the following steps: performing defect analysis on a power engineering drawing through a preset multi-level progressive analysis architecture, and sequentially generating a drawing analysis result, a functional block recognition result, an equipment component detection result and an element parameter verification result; adopting a cross-level feature fusion mechanism to perform feature fusion based on the drawing analysis result, the functional block recognition result, the equipment component detection result and the element parameter verification result, and generating a comprehensive defect positioning result; and based on the comprehensive defect positioning result, generating a corresponding defect positioning report and performing visual display. Through the multi-level progressive analysis architecture and the cross-level feature fusion mechanism, accurate positioning from the whole drawing to the element parameter is realized, the efficiency and accuracy of power engineering drawing defect positioning are effectively improved, and the automation level and reliability of drawing review are improved.

Description

Technical Field

[0001] This application relates to the field of computer vision technology, and in particular to a method, apparatus, electronic device and storage medium for locating defects in power engineering drawings. Background Technology

[0002] The review of power engineering design drawings is a crucial step in ensuring the safe and stable operation of power systems. Traditional review methods rely primarily on manual processes, requiring reviewers to meticulously check the equipment layout, connections, and parameter settings to ensure compliance with design specifications. However, manual review suffers from inefficiencies, inconsistent adherence to standards, and susceptibility to omissions or misjudgments due to visual fatigue or experience differences. This is especially true when dealing with complex power system drawings, which typically contain numerous components, multi-layered connections, and dense textual parameters. Manual review struggles to comprehensively cover all details, easily overlooking critical defects and creating potential safety hazards.

[0003] In recent years, with the development of computer vision and artificial intelligence technologies, some automated drawing review technologies have been introduced into the field of power engineering. Existing technologies mostly employ general object detection, image segmentation, or text recognition algorithms for preliminary analysis of drawings. However, these methods still have significant shortcomings in practical applications. For example, they cannot achieve multi-level progressive analysis from the overall drawing to functional blocks, equipment components, and even component parameters, resulting in insufficient defect location accuracy and difficulty in improving overall detection reliability through information complementarity. Furthermore, general computer vision algorithms have low accuracy in identifying key components in power engineering drawings, cannot resolve the logical relationships between parameters, struggle to integrate graphic structure and parameter semantic information, and have limited ability to identify complex defects. Therefore, existing technologies for reviewing power engineering drawings suffer from inefficiency and insufficient accuracy.

[0004] The aforementioned problems severely restrict the efficiency and accuracy of power engineering drawing review. Especially in scenarios such as high-voltage substations and complex power distribution networks, design flaws can lead to major safety accidents. Therefore, there is an urgent need for a defect location method specifically designed for power engineering drawings, capable of achieving multi-level, progressive, and precise detection from overall layout to component parameters, while effectively integrating multi-modal information to comprehensively improve the automation level and reliability of drawing review.

[0005] The preceding description is intended to provide general background information and does not necessarily constitute prior art. Summary of the Invention

[0006] To address the aforementioned technical problems, this application provides a method, device, electronic device, and storage medium for locating defects in power engineering drawings, thereby resolving the issues of low efficiency and insufficient accuracy in the review of power engineering drawings in the prior art.

[0007] To address the aforementioned technical problems, this application provides a method for locating defects in power engineering drawings, comprising the following steps:

[0008] The power engineering drawings are analyzed for defects through a pre-set multi-level progressive analysis architecture, and the results of the drawing analysis, functional block identification, equipment component detection and component parameter verification are generated in sequence.

[0009] A cross-level feature fusion mechanism is used to perform feature fusion based on the drawing analysis results, the functional block identification results, the equipment component detection results, and the component parameter verification results to generate a comprehensive defect location result;

[0010] Based on the comprehensive defect location results, a corresponding defect location report is generated and visualized.

[0011] Furthermore, in some embodiments of this application, the multi-level progressive analysis architecture includes a drawing overall analysis layer, a functional block identification layer, an equipment component detection layer, and a component parameter verification layer executed sequentially. The step of performing defect analysis on power engineering drawings through a preset multi-level progressive analysis architecture, and sequentially generating drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results, includes:

[0012] The power engineering drawings are analyzed using the overall drawing analysis layer to obtain the drawing analysis results;

[0013] The functional block identification result is obtained by performing functional block identification on the drawing analysis result through the functional block identification layer.

[0014] The device component detection layer performs device component detection on the functional block identification result to obtain the device component detection result;

[0015] The component parameter verification layer verifies the component parameters of the device component by verifying the detection results, and obtains the component parameter verification results.

[0016] Furthermore, in some embodiments of this application, the step of performing drawing analysis on the power engineering drawings through the overall drawing analysis layer to obtain the drawing analysis results includes:

[0017] The power engineering drawings are classified and their types are determined using a power drawing type identification model.

[0018] Extract the global layout features of the power engineering drawings, including regional distribution features, component density features, and visual statistical features;

[0019] An attention heatmap is generated based on the global layout features;

[0020] An anomaly detection algorithm is used to filter out abnormal regions in the attention heatmap, and the weight of the abnormal regions is dynamically adjusted according to the preset importance of the power components.

[0021] Furthermore, in some embodiments of this application, the step of performing functional block identification on the drawing analysis results through the functional block identification layer to obtain the functional block identification results includes:

[0022] The power engineering drawings are semantically segmented to identify several functional blocks in the drawings, as well as the boundaries and types of each functional block.

[0023] Based on the boundaries and types of the functional blocks, verify whether the spatial relationships and connection methods between the functional blocks meet the preset standards.

[0024] Block layouts and connections that do not conform to preset standards are marked, generating block-level defect information.

[0025] Furthermore, in some embodiments of this application, the step of performing device component detection on the functional block identification result through the device component detection layer to obtain the device component detection result includes:

[0026] The location and type of power equipment are detected within each functional block by combining the prior positional relationships of power components;

[0027] An electrical topology diagram is constructed based on the extracted electrical connection relationships between the power equipment, and it is verified whether the connection methods between the power equipment conform to the connection rules in the power system design specifications.

[0028] Power equipment that has been marked as having incorrect equipment selection or abnormal topology connection logic is marked, and component-level defect information is generated.

[0029] Furthermore, in some embodiments of this application, the step of verifying the component parameters of the device component detection results through the component parameter verification layer to obtain the component parameter verification results includes:

[0030] Text recognition is performed on the power engineering drawings to obtain the text parameter information associated with the power equipment;

[0031] The text parameter information is converted into numerical values ​​to obtain the corresponding parameter values.

[0032] Based on the type and contextual semantics of the power equipment, verify the matching of the parameter values ​​with the preset specifications;

[0033] Parameter values ​​that exceed limits, have incorrect units, or are logically contradictory are marked, generating parameter-level defect information.

[0034] Furthermore, in some embodiments of this application, the method of employing a cross-level feature fusion mechanism to perform feature fusion based on the drawing analysis results, the functional block identification results, the equipment component detection results, and the component parameter verification results to generate a comprehensive defect location result includes:

[0035] The feature information of the overall drawing analysis layer, the functional block identification layer, the equipment component detection layer, and the component parameter verification layer are mapped to a unified feature dimension to form a standardized feature representation;

[0036] Based on the preset importance weight table of power components, the attention weights corresponding to the standardized features at each level are calculated respectively.

[0037] A bottom-up recursive approach is adopted to fuse features step by step according to the attention weights to obtain a comprehensive feature map;

[0038] Based on the comprehensive feature mapping, the drawing analysis results, the functional block identification results, the equipment component detection results, and the component parameter verification results, a comprehensive defect location result is generated.

[0039] Accordingly, this application provides a device for locating defects in power engineering drawings, comprising:

[0040] The analysis module is used to perform defect analysis on power engineering drawings through a preset multi-level progressive analysis architecture, and sequentially generate drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results.

[0041] The fusion module is used to perform feature fusion based on the drawing analysis results, the functional block identification results, the equipment component detection results, and the component parameter verification results using a cross-level feature fusion mechanism to generate a comprehensive defect location result.

[0042] The reporting module is used to generate a corresponding defect location report based on the comprehensive defect location results and to display it visually.

[0043] This application also provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the power engineering drawing defect location method as described above.

[0044] This application also provides a storage medium storing a computer program that can be loaded by a processor and executed as described above for the method of locating defects in power engineering drawings.

[0045] Implementing the embodiments of this application has the following beneficial effects:

[0046] As described above, this application provides a method, device, electronic device, and storage medium for locating defects in power engineering drawings. The method includes: first, performing defect analysis on the power engineering drawings through a preset multi-level progressive analysis architecture, sequentially generating drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results; then, using a cross-level feature fusion mechanism to perform feature fusion based on the drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results, generating a comprehensive defect location result; finally, based on the comprehensive defect location result, generating a corresponding defect location report and displaying it visually. The power engineering drawing defect location solution provided in this application performs a progressively detailed analysis of power engineering drawings from overall to specific levels by sequentially executing the overall drawing analysis layer, functional block identification layer, equipment component detection layer, and component parameter verification layer. This enables a comprehensive and systematic identification of potential defects in the drawings, ensuring the comprehensiveness and accuracy of defect location and avoiding the omission of important information. The solution integrates the analysis results at each level to generate a comprehensive defect location result, compensating for the shortcomings of single-level analysis, improving the accuracy of defect location, reducing false alarms and missed detections, while retaining key feature information. Finally, a report is generated based on the comprehensive defect location result, and the defect location and severity are visually displayed, facilitating engineers' quick understanding of problems in the drawings, improving drawing review efficiency, reducing the burden of manual review, and ensuring the quality and safety of power engineering design. Therefore, the power engineering drawing defect location solution provided in this application can effectively improve the efficiency and accuracy of defect location in power engineering drawings, thereby enhancing the automation level and reliability of drawing review. Attached Figure Description

[0047] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0048] Figure 1 This is a schematic diagram illustrating an application scenario of the power engineering drawing defect location method provided in the embodiments of this application;

[0049] Figure 2 This is a flowchart illustrating the method for locating defects in power engineering drawings provided in an embodiment of this application;

[0050] Figure 3This is a schematic diagram of the process for generating an attention heatmap provided in an embodiment of this application;

[0051] Figure 4 This is a schematic diagram of the parameter verification layer processing provided in the embodiments of this application;

[0052] Figure 5 This is a schematic diagram of the cross-level feature fusion mechanism provided in the embodiments of this application;

[0053] Figure 6 This is a schematic diagram of the multi-level progressive analysis architecture provided in the embodiments of this application;

[0054] Figure 7 This is an example diagram of the multi-level defect location results provided in the embodiments of this application;

[0055] Figure 8 This is a schematic diagram of the structure of the power engineering drawing defect location device provided in the embodiments of this application;

[0056] Figure 9 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application.

[0057] The realization of the objectives, functional features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and textual descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0058] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0059] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, components, features, and elements with the same names in different embodiments of this application may have the same meaning or different meanings, the specific meaning of which must be determined by its interpretation in that specific embodiment or further in conjunction with the context of that specific embodiment.

[0060] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0061] In the following description, the use of suffixes such as "module," "part," or "unit" to denote elements is solely for the purpose of illustrative purposes and has no specific meaning in itself. Therefore, "module," "part," or "unit" may be used interchangeably.

[0062] This application provides a method, device, electronic device, and storage medium for locating defects in power engineering drawings.

[0063] Specifically, the defect location device for power engineering drawings can be integrated into an electronic device, which can be a smartphone, tablet, laptop, or desktop computer, but is not limited to these. The electronic device can be directly or indirectly connected to a server via wired or wireless communication. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. This application does not impose any restrictions on these aspects.

[0064] Please see Figure 1 , Figure 1 This is an application environment diagram of a defect location method for power engineering drawings in one embodiment. (Refer to...) Figure 1This method for locating defects in power engineering drawings can be applied to a power engineering drawing defect location system. The system includes a terminal 110 and a server 120. The terminal 110 and server 120 are connected via a network. The terminal 110 can be a desktop terminal or a mobile terminal, specifically a mobile phone, tablet, or laptop. The server 120 can be a standalone server or a server cluster consisting of multiple servers. The terminal 110 performs defect analysis on the power engineering drawings using a pre-defined multi-level progressive analysis architecture, sequentially generating drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results. A cross-level feature fusion mechanism is used to fuse features based on the drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results to generate a comprehensive defect location result. Based on the comprehensive defect location result, a corresponding defect location report is generated and visualized.

[0065] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the priority of the embodiments.

[0066] This application provides a method for locating defects in power engineering drawings, comprising: performing defect analysis on power engineering drawings through a preset multi-level progressive analysis architecture, and sequentially generating drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results; employing a cross-level feature fusion mechanism to perform feature fusion based on the drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results to generate a comprehensive defect location result; and generating a corresponding defect location report based on the comprehensive defect location result and displaying it visually.

[0067] Please see Figure 2 , Figure 2 This is a flowchart illustrating the method for locating defects in power engineering drawings provided in this embodiment. The method for locating defects in power engineering drawings provided in this embodiment may specifically include the following steps:

[0068] S1. Perform defect analysis on power engineering drawings through a preset multi-level progressive analysis architecture, and generate drawing analysis results, functional block identification results, equipment component detection results and component parameter verification results in sequence;

[0069] Specifically, for step S1, this embodiment provides a multi-level progressive analysis architecture to perform progressive analysis on power engineering drawings, thereby sequentially generating drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results corresponding to different levels. The multi-level progressive analysis architecture includes a drawing overall analysis layer, a functional block identification layer, an equipment component detection layer, and a component parameter verification layer.

[0070] S2. A cross-level feature fusion mechanism is adopted to perform feature fusion based on drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results to generate comprehensive defect location results;

[0071] Specifically, for step S2, a cross-level feature fusion mechanism is adopted to map the feature information of the overall drawing analysis layer, functional block identification layer, equipment component detection layer, and component parameter verification layer to a unified feature dimension, forming a standardized feature representation. Based on the preset importance weight table of power components, the attention weights corresponding to the standardized features at each level are calculated respectively. Then, a bottom-up recursive approach is used to fuse features level by level according to the attention weights to obtain a comprehensive feature mapping.

[0072] S3. Based on the comprehensive defect location results, generate the corresponding defect location report and display it visually;

[0073] Specifically, for step S3, based on the comprehensive defect location results, a corresponding defect location report is generated and displayed visually. Visualization can take the form of heatmaps, etc., to intuitively present the location and severity of defects on the drawings, facilitating quick understanding and handling by engineers.

[0074] This embodiment achieves precise positioning from the overall drawing to the component parameters through a multi-level progressive analysis architecture and cross-level feature fusion mechanism, significantly improving the efficiency and quality of drawing review, reducing omissions and errors in manual review, and is applicable to various types of power engineering drawings. It has a high level of intelligence and automation, providing an efficient and accurate solution for the review of power engineering design drawings.

[0075] Furthermore, in some embodiments, the multi-level progressive analysis architecture in this embodiment may specifically include a drawing overall analysis layer, a functional block identification layer, an equipment component detection layer, and a component parameter verification layer executed sequentially. Step S1, "performing defect analysis on power engineering drawings through a preset multi-level progressive analysis architecture, and sequentially generating drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results," may specifically include:

[0076] S11. Perform drawing analysis on the power engineering drawings through the overall drawing analysis layer to obtain the drawing analysis results;

[0077] Specifically, the power engineering drawings are scanned globally through the overall drawing analysis layer to identify the drawing type and extract global layout features, including regional distribution features, component density features, and visual statistical features. Based on these features, an attention heatmap is generated, and a statistical anomaly detection algorithm is used to filter out abnormal areas in the heatmap. The weight of abnormal areas is dynamically adjusted according to the preset importance of power components.

[0078] S12. The functional block identification results are obtained by performing functional block identification on the drawing analysis results through the functional block identification layer;

[0079] Specifically, the functional block identification layer performs semantic segmentation on the drawing analysis results to identify several functional blocks in the power engineering drawings, as well as their corresponding boundaries and types. It verifies whether the spatial relationship and connection method between blocks meet the preset standards, marks those that do not meet the standards, and generates block-level defect information.

[0080] S13. Perform device component detection on the functional block identification results through the device component detection layer to obtain the device component detection results;

[0081] Specifically, based on the functional block identification results, the device component detection layer detects the location and type of power equipment by combining the prior positional relationship of power components, constructs an electrical topology diagram, verifies whether the connection method between devices complies with the power system design specifications, marks devices with incorrect selection or abnormal topology connection logic, and generates component-level defect information.

[0082] S14. Verify the component parameters of the equipment component test results through the component parameter verification layer to obtain the component parameter verification results;

[0083] Specifically, the component parameter verification layer performs text recognition on the equipment component detection results, extracts the text parameter information associated with the power equipment, converts it into numerical values, verifies the matching of parameter values ​​with preset specifications based on equipment type and contextual semantics, marks parameter values ​​that exceed limits, have incorrect units, or are logically contradictory, and generates parameter-level defect information.

[0084] The multi-level progressive analysis architecture provided in this embodiment can comprehensively and systematically analyze defects in power engineering drawings, ensuring that potential problems at all levels from the overall to the local can be discovered. This effectively improves the comprehensiveness and accuracy of defect location, avoids omissions that may be caused by traditional single-level analysis, and provides a rich and accurate information foundation for subsequent feature fusion and comprehensive defect location result generation.

[0085] Furthermore, in some embodiments, step S11, "performing drawing analysis on the power engineering drawings through the overall drawing analysis layer to obtain drawing analysis results," may specifically include:

[0086] S111. Classify power engineering drawings and determine the drawing type using a power drawing type identification model;

[0087] Specifically, a power engineering drawing type recognition model is used to classify and determine the drawing type. This model is specifically designed for power engineering drawings and can identify various types of drawings, such as substation primary system diagrams and distribution system diagrams. Its core is a lightweight convolutional neural network, designed to balance model complexity and performance, ensuring efficient processing of large numbers of power engineering drawings without sacrificing classification accuracy. During model training, a large dataset of labeled power engineering drawings is used, enabling the model to learn feature representations of different drawing types, thus accurately classifying input power engineering drawings in practical applications.

[0088] S112. Extract the global layout features of the power engineering drawings. The global layout features include regional distribution features, component density features, and visual statistical features.

[0089] Specifically, the global layout features of power engineering drawings are extracted, including regional distribution features, component density features, and visual statistical features. Regional distribution features reflect the spatial distribution of different functional areas in the drawings, such as the location and proportion of the main transformer area and switchgear area. Component density features describe the density of electrical components in each area, helping to identify anomalies such as excessively dense or sparse component placement. Visual statistical features encompass the overall visual characteristics of the drawings, such as color distribution and texture features, which reflect the drawing style and quality. By comprehensively extracting these global layout features, a preliminary understanding and analysis of power engineering drawings can be achieved at a macro level, providing a comprehensive feature foundation for subsequent anomaly area screening.

[0090] S113. Generate attention heatmaps based on global layout features;

[0091] Specifically, firstly, using the extracted global layout features, the importance and anomaly level of each area in the drawing are calculated through specific algorithms (such as improved Gradient Weighted Class Activation Mapping (Grad-CAM) technology) or models. Then, based on these calculation results, a heatmap of the same size as the original drawing is generated, where the value of each pixel represents the importance and anomaly level of that location within the global layout. Heatmaps are typically presented as grayscale or color images; areas with higher values ​​appear brighter or more vibrant on the map, indicating a greater need for attention. Through attention heatmaps, potential anomaly areas in power engineering drawings can be visually identified, providing a visual basis for subsequent anomaly area screening.

[0092] S114. A statistical anomaly detection algorithm is used to screen out abnormal regions in the attention heatmap, and the weight of the abnormal regions is dynamically adjusted according to the preset importance of the power components;

[0093] Specifically, the pixel values ​​in the attention heatmap are first statistically analyzed to calculate the overall distribution characteristics, such as mean and variance. Then, based on set thresholds or anomaly detection rules, regions whose pixel values ​​significantly deviate from the normal distribution are identified as anomalous areas. During the screening process, the weight of anomalous areas is dynamically adjusted according to the preset importance of the power components. For example, for areas containing critical power components, even if their anomalousness is slightly lower, they will be given higher weight due to the importance of the components, thus ensuring that these areas are not overlooked. In this way, truly noteworthy anomalous areas can be screened more accurately.

[0094] In this embodiment, the overall drawing analysis layer achieves a comprehensive understanding and preliminary defect location of power engineering drawings through steps such as power drawing type identification model, global layout feature extraction, attention heatmap generation, and abnormal area screening. This level of analysis provides an important foundation and guidance for subsequent functional block identification, equipment component inspection, and component parameter verification, ensuring the systematic nature and efficiency of the entire defect location process. Through progressively in-depth analysis from global to local perspectives, potential defects in power engineering drawings can be comprehensively and accurately discovered, improving the quality and safety of power engineering design.

[0095] Among them, such as Figure 3 As shown, the specific process for generating the attention heatmap is as follows:

[0096] Input includes:

[0097] Power Engineering Drawing Image I: As the power engineering drawing to be analyzed, it serves as the raw input data for the entire process in the form of an image.

[0098] Convolutional Network Model F: A pre-trained convolutional neural network model used for tasks such as feature extraction and classification of power grid drawings.

[0099] Classification score Sc: The score obtained by model F after classifying the power drawing image I, reflecting the confidence level that the image belongs to a certain category.

[0100] The output includes:

[0101] Attention Heatmap L: A heatmap of the same size as the input image, where the value of each pixel represents the importance and degree of anomaly of that location in the global layout, visually highlighting potential anomalies in the power drawing.

[0102] (1) Calculate the gradient of the classification score Sc with respect to the feature map Ak.

[0103] The goal is to determine which parts of the feature map Ak contribute most to the classification score Sc, i.e., to find the feature regions that have the greatest influence on the classification decision. The gradient of the classification score Sc with respect to the feature map Ak is then calculated using the backpropagation algorithm. This step quantifies the influence of each location in the feature map on the classification result through mathematical gradient calculation.

[0104] (2) Perform global average pooling on the feature dimensions to obtain the weights αk.

[0105] Gradient information is transformed into importance weights for each channel in the feature map, so that the feature map can be weighted subsequently based on these weights. (This refers to the process of applying gradients.) Global average pooling is performed, which calculates the average value across the spatial dimension of the feature map to obtain the weight αk. Global average pooling can compress the gradient information of each feature channel into a scalar value, representing the overall importance of that channel in the entire image.

[0106] (3) Employ a power sector-specific weighting adjustment function Optimize weights

[0107] Based on the domain knowledge of power engineering drawings, the initial weights αk are adjusted to better reflect the importance and risk levels of components in the drawings. A pre-designed power domain-specific weight adjustment function is applied. The initial weights αk are mapped to new weight values. This adjustment function takes into account factors such as the type, location, and function of the power components, assigning higher weights to feature channels related to key power components.

[0108] (4) Calculate the weighted feature map

[0109] The optimized weights are applied to the feature map to generate a weighted feature map that highlights important regions. The weights for each feature channel are then adjusted. Multiply by the corresponding feature map Ak, then sum the weighted feature maps of all channels, and finally perform non-linear processing using the ReLU activation function to obtain the weighted feature map. The ReLU activation function sets negative values ​​to zero and retains positive values, highlighting feature regions that contribute to classification.

[0110] (5) Normalized thermogram

[0111] The value range of the weighted feature map L is normalized to the interval [0,1] for easier subsequent visualization and processing. The weighted feature map L is normalized using the formula L = (L - min(L)) / (max(L) - min(L)). This step uses a linear transformation to map the minimum value in the feature map to 0, the maximum value to 1, and other values ​​to a proportional adjustment.

[0112] Compared to general Grad-CAM technology, this embodiment adds a specialized weight adjustment function tailored to the characteristics of power drawings. This function adjusts the weights based on the importance and risk level of the power components, giving safety-critical components (such as main transformers and circuit breakers) a higher attention weight.

[0113] Furthermore, in some embodiments, step S12, "performing functional block identification on the drawing analysis results through the functional block identification layer to obtain functional block identification results," may specifically include:

[0114] S121. Perform semantic segmentation on the power engineering drawings to identify several functional blocks in the power engineering drawings, as well as the boundaries and types of each functional block;

[0115] Specifically, an improved U-Net network is used to perform pixel-level classification of the drawings, assigning each pixel to its corresponding functional block category. For example, it identifies different functional areas such as the main transformer area, switchgear area, and cable trench area, and obtains the precise boundary and type information of each area. Combining knowledge from the power engineering field, the functional blocks obtained from semantic segmentation are further identified and classified to determine the specific functional type of each block. For example, it identifies whether a block is a main transformer area, switchgear area, or cable trench area. This process requires comprehensively considering features such as the shape, size, location, and types of components contained in the blocks, and matching them with the characteristics of standard functional blocks in power engineering.

[0116] S122. Based on the boundaries and types of the functional blocks, verify whether the spatial relationships and connection methods between the functional blocks meet the preset standards.

[0117] Specifically, based on the boundary and type information of the identified functional blocks, the spatial relationships and connection methods between the blocks are analyzed. This includes checking whether the relative positions, distances, routing of connecting lines, and connection points between different functional blocks comply with power design specifications and safety standards. For example, verifying whether the connection lines between the main transformer area and the switchgear area meet the specified path and distance requirements, and checking whether the positional relationship between the cable trench area and other blocks is reasonable.

[0118] S123. Mark the block layout and connection relationship that do not conform to the preset standard, and generate block-level defect information;

[0119] Specifically, when an anomaly is detected in the spatial relationship or connection method between blocks, it is marked at the corresponding location on the drawing, and the type and description of the defect are recorded. For example, the area where the connection path between the main transformer area and the switchgear area does not meet the specified requirements is marked, and the defect is noted as "the connection path does not meet the specifications".

[0120] This embodiment's functional block identification layer achieves precise division and layout verification of functional blocks in power engineering drawings through semantic segmentation, functional area identification, inter-block relationship detection, and block-level anomaly marking. This level of analysis provides clear area division and contextual information for subsequent equipment component inspection and component parameter verification, ensuring the systematic nature and efficiency of the entire defect localization process. By performing defect detection at the functional block level, layout and connection problems in the drawings can be identified in a timely manner, improving the quality and safety of power engineering design.

[0121] In addition, this embodiment also provides a method for power sector-specific improvements to the U-Net network, mainly including:

[0122] (1) Design an encoder structure specifically for power drawings.

[0123] Custom convolution kernels: Design convolution kernels with specific shapes and sizes to better match the shapes and dimensions of common elements in electrical drawings. For example, use elongated convolution kernels to capture the features of linear elements such as busbars.

[0124] Multi-scale feature extraction: Multi-scale features are extracted at different levels of the encoder to accommodate functional blocks and components of varying sizes in power drawings. This helps to accurately identify and locate key elements at different resolutions.

[0125] Electrical Element Enhancement Module: Specialized modules are added to the encoder to enhance the feature extraction of electrical-specific elements. These modules can leverage prior knowledge, such as the typical shape, color, and texture of the element, to improve recognition accuracy.

[0126] (2) Introduce attention gating mechanism

[0127] Attention Gating Module: An attention gating module is introduced into the decoder part of U-Net. This module can dynamically adjust the degree of attention given to different regions based on the features extracted by the encoder. Specifically, the attention gating module learns a weight matrix and assigns different weights to each pixel in the feature map. Pixels with higher weights represent regions that require more attention.

[0128] Incorporating prior knowledge: Prior knowledge from the field of power engineering, such as the typical characteristics of key connection points and intersections, is integrated into the attention gating module to guide the network to better focus on these important areas.

[0129] (3) Add a special boundary-sensitive loss function

[0130] Boundary detection module: A dedicated boundary detection module is added to the network to predict the pixels at block boundaries. This module can be implemented using edge detection operators or other specialized network structures.

[0131] Boundary-sensitive loss function: Design a loss function that gives greater weight to segmentation errors in boundary regions. For example, a boundary-weighted cross-entropy loss function can be used, where boundary pixels have higher weights than non-boundary pixels. This ensures the network prioritizes segmentation accuracy in boundary regions during optimization.

[0132] This embodiment achieves accurate identification and segmentation of functional blocks in power engineering drawings by making power-domain-specific improvements to the U-Net network. A dedicated encoder structure for power engineering drawings is designed to optimize the identification of power-specific elements; an attention gating mechanism is introduced to enable the network to focus on key connection points and intersections; and a special boundary-sensitive loss function is added to improve the accuracy of block boundary segmentation. These improvements allow the functional block identification layer to provide more accurate and reliable information for subsequent inter-block relationship detection and defect marking, thereby improving the performance and effectiveness of the overall power engineering drawing defect localization method.

[0133] Furthermore, in some embodiments, step S13, "performing device component detection on the functional block identification result through the device component detection layer to obtain the device component detection result," may specifically include:

[0134] S131. Based on the prior positional relationships of power components, detect the location and type of power equipment within each functional block;

[0135] Specifically, within functional blocks, the location and type of power equipment are detected using a Faster R-CNN network specialized for power equipment, incorporating prior positional relationships of power components. Prior positional relationships of power components refer to the common locations and layout patterns of various types of power equipment in drawings, predetermined based on power engineering design specifications and practical experience. For example, transformers are typically located in specific locations within the main transformer area, and circuit breakers are usually adjacent to switchgear. By integrating this prior knowledge into the detection model, the accuracy and efficiency of detection can be improved. Faster R-CNN is a deep learning model widely used in object detection, its core comprising a Region Proposal Network (RPN) and a Fast R-CNN detection network. In power equipment detection, the Faster R-CNN network has been specially modified to adapt to the characteristics of power drawings and the features of power components. For example, the size and proportion of anchor frames are adjusted to better match the size and shape of the power equipment; prior positional relationships of power components are incorporated into the RPN to guide the network in generating region proposals more accurately.

[0136] S132. Construct an electrical topology diagram based on the extracted electrical connection relationships between power equipment, and verify whether the connection methods between power equipment conform to the connection rules in the power system design specifications;

[0137] Specifically, based on the extracted electrical connection relationships between power equipment, an electrical topology diagram is constructed, and the connection methods between the equipment are verified to conform to the connection rules in the power system design specifications. Specifically, firstly, image processing and pattern recognition technologies are used to extract information such as connection lines and connection points between power equipment; then, a graph structure is constructed based on this information, where nodes represent power equipment and edges represent the electrical connection relationships between equipment; finally, according to the power system design specifications, the connection methods of the equipment in the graph are checked to ensure they meet the requirements, such as whether the connection sequence and connection method (series, parallel, etc.) are correct.

[0138] S133. Mark power equipment that has been marked as having incorrect equipment selection or abnormal topology connection logic, and generate component-level defect information;

[0139] Specifically, power equipment with incorrect selection or abnormal topology connection logic is marked, and component-level defect information is generated. In particular, when it is detected that the selection of a power device does not conform to design specifications (e.g., the device's capacity, model, or other parameters do not meet requirements), or its connection method in the electrical topology diagram has logical abnormalities (e.g., incorrect connection sequence, loop formation, etc.), it is marked at the corresponding location on the drawing, and the type and description of the defect are recorded.

[0140] The equipment component detection layer in this embodiment achieves accurate detection and comprehensive verification of the connection relationships of power equipment within a functional block by combining the prior positional relationships of power components, constructing an electrical topology diagram and verifying connection methods, and marking power equipment with incorrect equipment selection or abnormal topology connection logic. This level of analysis provides accurate equipment information and connection relationships for subsequent component parameter verification, ensuring the systematic nature and efficiency of the entire defect location process. By performing defect detection at the equipment component level, equipment selection and connection logic problems in drawings can be identified in a timely manner, improving the quality and safety of power engineering design.

[0141] Furthermore, this embodiment incorporates prior knowledge of power components into the detection network, such as the common proximity of transformers to switching equipment. This prior knowledge helps the detection model to be more targeted when searching for targets, improving detection accuracy and efficiency. Specifically, the common layout and relationships of power components are encoded into the detection model, enabling the model to make judgments based on this prior knowledge when predicting the location and type of power equipment. Traditional Region Proposal Networks (RPNs) primarily generate candidate regions based on low-level image features, while Relationship-Aware RPNs further consider the spatial and functional relationships between power equipment. Specifically, when generating region proposals, not only are the visual features of the image considered, but also the prior relationships between power components are combined to ensure that the generated candidate regions better match the actual layout and connection patterns of power equipment. For equipment with significant size differences in power drawings (such as small relays to large transformers), a multi-scale feature enhancement module is designed. This module extracts and fuses features at different scales, ensuring that the model can effectively detect power equipment of various sizes. Specifically, the multi-scale feature enhancement module can extract features from multiple convolutional layers and fuse these features together through operations such as upsampling and downsampling to form a feature representation rich in multi-scale information.

[0142] Furthermore, in some embodiments, step S14, "verifying the component parameters of the device component detection results through the component parameter verification layer to obtain the component parameter verification results," may specifically include:

[0143] S141. Perform text recognition on power engineering drawings to obtain text parameter information associated with power equipment;

[0144] Specifically, such as Figure 4As shown, this step primarily utilizes Optical Character Recognition (OCR) technology, optimized for the characteristics of power engineering drawings, to accurately recognize the text content within the drawings, including equipment models, parameter annotations, and technical specifications. While OCR technology is relatively mature in the field of text recognition, the text in power engineering drawings has unique characteristics, such as containing numerous technical terms, symbols, and special formats. Therefore, it is necessary to improve and optimize general OCR technologies. For example, specialized recognition models can be trained to achieve higher accuracy in recognizing power engineering terms and symbols; simultaneously, combining this with layout analysis of the drawings can improve the recognition performance of text in different areas.

[0145] S142. Perform parameter numerical conversion on the text parameter information to obtain the corresponding parameter values;

[0146] Specifically, such as Figure 4 As shown, this step transforms the identified text content into a structured data format, facilitating subsequent calculations and verification. For example, "Rated Current: 630A" is converted into key-value pairs, where the key is "Rated Current" and the value is "630A". Parameter numerical conversion involves Natural Language Processing (NLP) techniques and data processing methods. In parameter extraction from power engineering drawings, specialized parsing rules need to be designed to convert the text into numerical data based on its semantics and format. For example, identifying the combination patterns of parameter names and units, extracting the numerical components, and performing standardization.

[0147] S143. Based on the type of power equipment and contextual semantics, verify the matching of parameter values ​​with preset specifications;

[0148] Specifically, such as Figure 4 As shown, this step compares the converted parameter values ​​with the requirements in the power design code to check whether the parameters are within the allowable range, whether the units are correct, and whether the logical relationships are reasonable. Parameter matching verification requires establishing a complete set of verification rules and a knowledge base, combining power design codes and equipment characteristic requirements. In practical applications, rule engines or knowledge graph technology can be used to transform design codes into computable rules for automatic parameter verification. Simultaneously, the correlation between parameters should be considered for comprehensive verification.

[0149] S144. Mark parameter values ​​that exceed limits, have incorrect units, or are logically contradictory, and generate parameter-level defect information;

[0150] Specifically, parameter values ​​that exceed limits, have incorrect units, or exhibit logical inconsistencies are marked, generating parameter-level defect information. During the verification process, once a parameter is found to be non-compliant with specifications, it is immediately marked at the corresponding location on the drawing, and detailed defect information is recorded, including the parameter name, non-compliance item, and severity. A robust defect marking and reporting mechanism needs to be established for parameter-level anomaly marking. In the automated inspection system, the marking results can be displayed through a visual interface, allowing engineers to quickly locate and view defects. Simultaneously, a detailed defect report is generated, providing a basis for subsequent corrections and reviews.

[0151] The component parameter verification layer in this embodiment achieves fine-grained verification of component parameters through steps such as text recognition, parameter numerical conversion, specification-based parameter matching verification, and parameter-level anomaly marking on power engineering drawings. This level of analysis provides a final, meticulous check for the entire defect localization process, ensuring the accuracy and compliance of parameter settings in power engineering drawings. By detecting defects at the parameter level, potential design problems can be identified, preventing equipment malfunctions and safety hazards during operation, and further improving the quality and safety of power engineering design.

[0152] Furthermore, in some embodiments, step S2, "using a cross-level feature fusion mechanism to perform feature fusion based on drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results to generate a comprehensive defect location result," may specifically include:

[0153] S21. Map the feature information of the overall drawing analysis layer, functional block identification layer, equipment component detection layer and component parameter verification layer to a unified feature dimension to form a standardized feature representation;

[0154] Specifically, such as Figure 5 As shown, the feature information of the overall drawing analysis layer, functional block identification layer, equipment component detection layer, and component parameter verification layer is mapped to a unified feature dimension to form a standardized feature representation. Specifically, through linear interpolation or other suitable feature resampling methods, the feature maps of different levels are adjusted to the same size and number of channels so that subsequent feature fusion operations can proceed smoothly.

[0155] S22. Based on the preset importance weight table of power components, calculate the attention weights corresponding to the standardized features at each level;

[0156] Specifically, the importance weight table for power components is pre-defined based on the criticality of different components in the power system to the overall operational safety and functional realization. For example, key equipment such as main transformers and circuit breakers are assigned higher weights, while some auxiliary equipment has relatively lower weights. When calculating attention weights, feature information is combined with these pre-defined weights, giving key components a greater influence in the fusion process.

[0157] S23. A bottom-up recursive approach is adopted to fuse features step by step according to attention weights to obtain a comprehensive feature map;

[0158] Specifically, such as Figure 5 As shown, a bottom-up recursive approach is used to fuse features level by level according to attention weights to obtain a comprehensive feature map. Specifically, starting from the features at the lowest level (the component parameter verification layer), features are fused sequentially upwards with features from the previous level. In each fusion step, the features of the current level are combined with the features of the previous level after attention weight adjustment, gradually constructing a comprehensive feature map containing multi-level information. Recursive feature fusion is an effective multi-level feature integration strategy. In deep learning, architectures such as Feature Pyramid Networks (FPNs) enhance performance by combining high-resolution features from low levels with semantic features from high levels through bottom-up path enhancement, improving the performance of tasks such as object detection. Recursive residual connections are a variant of feature fusion. By introducing residual connections, the gradient vanishing problem is avoided, while retaining detailed information from different levels of features. The recursive nested structure retains feature information at different granularities, avoiding information loss caused by simple superposition.

[0159] The calculation formula for recursive residual join includes:

[0160] F 34 =F3 + α4·F4

[0161] F 234 =F2 + α3·F 34

[0162] F 1234 =F1+α2·F 234

[0163] Among them, F i Let α represent the i-th level feature map. i This represents the attention weight.

[0164] Feature mapping involves extracting feature information at each level. For example, F1 represents the features of the overall drawing analysis layer, F2 represents the features of the functional block identification layer, F3 represents the features of the equipment component detection layer, and F4 represents the features of the component parameter verification layer. Attention weights are weight values ​​calculated based on the importance of power components, used to adjust the influence of features at different levels during the fusion process. They are obtained through an attention weight calculation method specialized for the power industry, reflecting the importance of each level of feature in defect localization.

[0165] Specifically, F 34 =F3+α4·F4, which fuses the feature F3 of the third level with the feature F4 of the fourth level (after attention weight adjustment).

[0166] F 234 =F2 + α3·F 34 The second-level feature F2 is fused with the previously obtained F 34 (After attention weighting) the mixture is fused.

[0167] F 1234 =F1+α2·F 234 The features F1 of the first level are fused with the previously obtained F 234 (After attention weighting) the mixture is fused.

[0168] S24. Based on the comprehensive feature mapping, drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results, generate comprehensive defect location results;

[0169] Specifically, by combining comprehensive feature mapping with the analysis results at each level, and using machine learning methods such as classification and regression, the location and type of defects in the drawings are determined, and a final defect location report is generated.

[0170] The cross-level feature fusion mechanism in this embodiment effectively integrates multi-level analysis results through steps such as feature dimension unification, attention weight calculation based on the importance weight table of power components, bottom-up recursive feature fusion, and generation of comprehensive defect location results. This mechanism ensures that feature information from different levels can complement and verify each other, improving the accuracy, comprehensiveness, and reliability of defect location. Through cross-level feature fusion, defects in power engineering drawings can be located more accurately, reducing missed detections and false alarms, and further improving the quality and safety of power engineering design.

[0171] Furthermore, in some embodiments, step S3, "based on the comprehensive defect location results, generating a corresponding defect location report and visually displaying it," may specifically include:

[0172] S31. Generate a multi-layered heat map overlaid on the original drawing. The color gradient of the heat map indicates the severity of the defect.

[0173] Specifically, the location information and severity score of each defect are extracted from the comprehensive defect location results. Location information can be coordinate points or area ranges, and the severity score is used to determine the color depth of the heatmap. Appropriate heatmap generation algorithms are selected, such as Gaussian kernel density estimation and kernel density smoothing. These algorithms can calculate the heat value of each pixel on the entire drawing based on the defect's location and severity. Multi-layered heatmaps are generated according to different defect types or severity levels. Each layer of the heatmap highlights a specific type of defect or the severity of defects within a certain range. The generated multi-layered heatmaps are overlaid onto the original power engineering drawings. By adjusting the transparency and color mapping of the heatmaps, the layers of the heatmaps blend naturally with the original drawings, clearly showing the defect distribution without obscuring key details of the drawings. A color gradient is set for the heatmaps to intuitively reflect the severity of defects. Gradients from light to dark and from cool to warm colors are typically used to represent defect severity from mild to severe. Simultaneously, visualization parameters such as contrast and brightness are adjusted to enhance readability.

[0174] S32. Output a structured defect report, which includes defect location, defect type, associated parameters and specification basis, and supports interactive location query;

[0175] Specifically, a structured report template is designed based on the standards for reviewing power engineering drawings and user needs. The template should include a cover, table of contents, main text framework, chart insertion locations, and table styles. Data from the comprehensive defect location results is then populated according to the report template format. Each defect item should include detailed information such as defect location, defect type, related parameters, and relevant specifications. Simultaneously, descriptive text is automatically generated, providing a detailed description of each defect, including possible causes, impacts, and recommended corrective measures. An interactive location query function is embedded in the report. Users can quickly locate specific defect items within the report through hyperlinks, index tables, etc., and the results can be linked with heatmaps on the drawings. The generated report is formatted to ensure it is standardized, aesthetically pleasing, and easy to read and print. The layout of text, tables, and charts is optimized for clarity and readability.

[0176] This embodiment, through comprehensive, intuitive, and detailed recording and display of defects in power engineering drawings, not only improves the efficiency and quality of the review process but also provides engineers with rich information support and a convenient operating experience. This helps to promptly identify and correct design problems, thereby enhancing the overall safety and reliability of power engineering designs. Simultaneously, standardized reports and clear visualizations facilitate team collaboration and communication, promoting the smooth progress of power engineering projects.

[0177] This embodiment also provides a specific implementation method for locating defects in power engineering drawings, the specific process of which is as follows:

[0178] The system first receives power engineering design drawings as input, supporting various common formats (such as PNG, JPG, PDF, etc.).

[0179] Then, the input drawings are standardized, including:

[0180] Image enhancement: Adjust brightness and contrast, eliminate noise.

[0181] Size standardization: Adjust to a uniform size (e.g., 1024×1024 pixels).

[0182] Coordinate system unification: Establishing a unified coordinate reference system

[0183] Next, the system performs a four-level progressive analysis, with each level building upon the results of the previous level to perform more granular detection, such as... Figure 6 As shown, Figure 6 This is a schematic diagram of a multi-level progressive positioning architecture.

[0184] Next, cross-level feature fusion is performed, using a recursive residual connection method to fuse features from different levels, such as... Figure 5 As shown.

[0185] Finally, multi-level defect location results are output, including multi-granularity defect location results from overall to detailed levels, and are visually displayed in the form of a heatmap, such as... Figure 7 As shown.

[0186] Taking the review of a 220kV substation design drawing as an example, the application process of this application is explained. The system receives the substation primary system diagram input and performs preprocessing; the overall drawing analysis layer identifies the drawing type as "substation primary system diagram", generates an attention heatmap, and initially identifies potential problems in the connection area between the main transformer area and the GIS room; the functional block identification layer performs semantic segmentation on the drawing, identifies functional blocks such as the main transformer area and switchgear area, and detects that the equipment layout in the main transformer area does not meet the specifications; the equipment component detection layer identifies each power device in the functional block and detects that a certain circuit breaker model is improperly selected; the component parameter verification layer extracts equipment parameters and finds that the rated current parameter of the circuit breaker (630A) is lower than the line design load requirement (800A); the cross-level feature fusion mechanism integrates the analysis results of the four levels and generates a comprehensive defect location report; the system outputs defect location results with intuitive heatmap markings, including: unreasonable main transformer area layout (block-level defect), improper circuit breaker model selection (component-level defect), and insufficient rated current parameter of the circuit breaker (parameter-level defect).

[0187] In summary, the power engineering drawing defect location method provided in this embodiment performs a progressively detailed analysis of power engineering drawings from the overall to the local by sequentially executing the overall drawing analysis layer, functional block identification layer, equipment component detection layer, and component parameter verification layer. This enables a comprehensive and systematic identification of potential defects in the drawings, ensuring the comprehensiveness and accuracy of defect location and avoiding the omission of important information. The method integrates the analysis results at each level to generate a comprehensive defect location result, compensating for the shortcomings of single-level analysis, improving the accuracy of defect location, reducing false alarms and missed detections, while retaining key feature information. Finally, a report is generated based on the comprehensive defect location result, and the defect location and severity are visually displayed, facilitating engineers' quick understanding of problems in the drawings, improving drawing review efficiency, reducing the burden of manual review, and ensuring the quality and safety of power engineering design.

[0188] To facilitate better implementation of the power engineering drawing defect location method according to the embodiments of this application, this application also provides a power engineering drawing defect location device. The meanings of the terms used are the same as in the power engineering drawing defect location method described above, and specific implementation details can be found in the descriptions in the method embodiments.

[0189] Please see Figure 8 , Figure 8 The diagram below illustrates the structure of a power engineering drawing defect location device provided in this embodiment. Specifically, the power engineering drawing defect location device may include an analysis module 201, a fusion module 202, and a reporting module 203, as follows:

[0190] Analysis module 201 is used to perform defect analysis on power engineering drawings through a preset multi-level progressive analysis architecture, and generate drawing analysis results, functional block identification results, equipment component detection results and component parameter verification results in sequence.

[0191] The fusion module 202 is used to perform feature fusion based on drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results using a cross-level feature fusion mechanism to generate comprehensive defect location results;

[0192] Report module 203 is used to generate a corresponding defect location report based on the comprehensive defect location results and to display it visually.

[0193] In summary, the power engineering drawing defect location device provided in this embodiment, through the analysis module 201, sequentially executes the overall drawing analysis layer, functional block identification layer, equipment component detection layer, and component parameter verification layer, performing a progressively detailed analysis of the power engineering drawings from the whole to the parts. This enables a comprehensive and systematic identification of potential defects in the drawings, ensuring the comprehensiveness and accuracy of defect location and avoiding the omission of important information. The fusion module 202 integrates the analysis results of each level to generate a comprehensive defect location result, compensating for the shortcomings of single-level analysis, improving the accuracy of defect location, reducing false alarms and missed detections, while retaining key feature information. Finally, the report module 203 generates a report based on the comprehensive defect location results, and visually displays the defect location and severity, facilitating engineers to quickly understand the problems in the drawings, improving drawing review efficiency, reducing the burden of manual review, and ensuring the quality and safety of power engineering design.

[0194] Furthermore, embodiments of this application also provide an electronic device, such as... Figure 9 The diagram illustrates the structure of an electronic device according to an embodiment of this application. Specifically, the electronic device may include components such as a processor 301 with one or more processing cores, a memory 302 with one or more computer-readable storage media, a power supply 303, and an input unit 304. Those skilled in the art will understand that... Figure 9 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0195] The processor 301 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the memory 302, and by calling data stored in the memory 302, thereby providing overall monitoring of the electronic device. Optionally, the processor 301 may include one or more processing cores; preferably, the processor 301 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 301.

[0196] The memory 302 can be used to store software programs and modules. The processor 301 executes various functional applications and methods for locating defects in power engineering drawings by running the software programs and modules stored in the memory 302. The memory 302 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device. In addition, the memory 302 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 302 may also include a memory controller to provide the processor 301 with access to the memory 302.

[0197] The electronic device also includes a power supply 303 that supplies power to various components. Preferably, the power supply 303 can be logically connected to the processor 301 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 303 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0198] The electronic device may also include an input unit 304, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0199] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 301 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 302 according to the following instructions, and the processor 301 runs the applications stored in the memory 302 to realize various functions, as follows:

[0200] The power engineering drawings are analyzed for defects through a pre-set multi-level progressive analysis architecture, generating drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results in sequence. A cross-level feature fusion mechanism is used to fuse features based on the drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results to generate a comprehensive defect location result. Based on the comprehensive defect location result, a corresponding defect location report is generated and visualized.

[0201] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0202] This embodiment of the application performs a progressively detailed analysis of power engineering drawings from the overall to the local by sequentially executing the overall drawing analysis layer, functional block identification layer, equipment component detection layer, and component parameter verification layer. This enables a comprehensive and systematic identification of potential defects in the drawings, ensuring the comprehensiveness and accuracy of defect location and avoiding the omission of important information. The analysis results of each level are integrated to generate a comprehensive defect location result, which compensates for the shortcomings of single-level analysis, improves the accuracy of defect location, reduces false alarms and missed detections, and retains key feature information. Finally, a report is generated based on the comprehensive defect location result, and the defect location and severity are displayed intuitively through visualization, which helps engineers quickly understand the problems in the drawings, improves the efficiency of drawing review, reduces the burden of manual review, and ensures the quality and safety of power engineering design.

[0203] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0204] Therefore, embodiments of this application provide a storage medium storing multiple instructions that can be loaded by a processor to execute steps in any of the power engineering drawing defect location methods provided in this application. For example, the instructions can execute the following steps:

[0205] The power engineering drawings are analyzed for defects through a pre-set multi-level progressive analysis architecture, generating drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results in sequence. A cross-level feature fusion mechanism is used to fuse features based on the drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results to generate a comprehensive defect location result. Based on the comprehensive defect location result, a corresponding defect location report is generated and visualized.

[0206] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0207] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk, or optical disk, etc. Since the instructions stored in the storage medium can execute the steps in any of the power engineering drawing defect location methods provided in this application embodiment, the beneficial effects achievable by any of the power engineering drawing defect location methods provided in this application embodiment can be realized. See the preceding embodiments for details, which will not be repeated here.

[0208] The above provides a detailed description of a method, apparatus, electronic device, and storage medium for locating defects in power engineering drawings, as provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for locating defects in power engineering drawings, characterized in that, Includes the following steps: The power engineering drawings are subjected to defect analysis through a pre-set multi-level progressive analysis architecture, which sequentially generates drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results. The multi-level progressive analysis architecture includes a drawing overall analysis layer, a functional block identification layer, an equipment component detection layer, and a component parameter verification layer, which are executed sequentially. The feature information corresponding to the drawing analysis results, the functional block identification results, the equipment component detection results, and the component parameter verification results are mapped to a unified feature dimension to form a standardized feature representation; based on the preset importance weight table of power components, the attention weights corresponding to the standardized features at each level are calculated respectively. A comprehensive feature mapping is obtained by progressively fusing features according to the attention weights, starting from the component parameter verification layer, the equipment component detection layer, the functional block identification layer and the overall drawing analysis layer; Based on the comprehensive feature mapping, the drawing analysis results, the functional block identification results, the equipment component detection results, and the component parameter verification results, a comprehensive defect location result is generated; Based on the comprehensive defect location results, a corresponding defect location report is generated and visualized.

2. The method for locating defects in power engineering drawings according to claim 1, characterized in that, The method involves performing defect analysis on power engineering drawings using a pre-defined multi-level progressive analysis architecture, sequentially generating drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results, including: The power engineering drawings are analyzed using the overall drawing analysis layer to obtain the drawing analysis results; The functional block identification result is obtained by performing functional block identification on the drawing analysis result through the functional block identification layer. The device component detection layer performs device component detection on the functional block identification result to obtain the device component detection result; The component parameter verification layer verifies the component parameters of the device component by verifying the detection results, and obtains the component parameter verification results.

3. The method for locating defects in power engineering drawings according to claim 2, characterized in that, The step of performing drawing analysis on the power engineering drawings through the overall drawing analysis layer to obtain the drawing analysis results includes: The power engineering drawings are classified and their types are determined using a power drawing type identification model. Extract the global layout features of the power engineering drawings, including regional distribution features, component density features, and visual statistical features; An attention heatmap is generated based on the global layout features; An anomaly detection algorithm is used to filter out abnormal regions in the attention heatmap, and the weight of the abnormal regions is dynamically adjusted according to the preset importance of the power components.

4. The method for locating defects in power engineering drawings according to claim 3, characterized in that, The step of performing functional block identification on the drawing analysis results through the functional block identification layer to obtain the functional block identification results includes: The power engineering drawings are semantically segmented to identify several functional blocks in the drawings, as well as the boundaries and types of each functional block. Based on the boundaries and types of the functional blocks, verify whether the spatial relationships and connection methods between the functional blocks meet the preset standards. Block layouts and connections that do not conform to preset standards are marked, generating block-level defect information.

5. The method for locating defects in power engineering drawings according to claim 4, characterized in that, The step of performing device component detection on the functional block identification result through the device component detection layer to obtain the device component detection result includes: The location and type of power equipment are detected within each functional block by combining the prior positional relationships of power components; An electrical topology diagram is constructed based on the extracted electrical connection relationships between the power equipment, and it is verified whether the connection methods between the power equipment conform to the connection rules in the power system design specifications. Power equipment with incorrect equipment selection or abnormal topology connection logic is marked, and component-level defect information is generated.

6. The method for locating defects in power engineering drawings according to claim 2, characterized in that, The step of verifying the component parameters of the device component through the component parameter verification layer to obtain the component parameter verification result includes: Text recognition is performed on the power engineering drawings to obtain text parameter information associated with the power equipment; The text parameter information is converted into numerical values ​​to obtain the corresponding parameter values. Based on the type and contextual semantics of the power equipment, verify the matching of the parameter values ​​with the preset specifications; Parameter values ​​that exceed limits, have incorrect units, or are logically contradictory are marked, generating parameter-level defect information.

7. A device for locating defects in power engineering drawings, characterized in that, include: The analysis module is used to perform defect analysis on power engineering drawings through a preset multi-level progressive analysis architecture, and generate drawing analysis results, functional block identification results, equipment component detection results, and component parameter verification results in sequence; the multi-level progressive analysis architecture includes a drawing overall analysis layer, a functional block identification layer, an equipment component detection layer, and a component parameter verification layer executed in sequence. The fusion module is used to map the feature information corresponding to the drawing analysis results, the functional block identification results, the equipment component detection results, and the component parameter verification results to a unified feature dimension to form a standardized feature representation; based on the preset importance weight table of power components, the attention weight corresponding to the standardized features at each level is calculated respectively. A comprehensive feature mapping is obtained by progressively fusing features according to the attention weights, starting from the component parameter verification layer, the equipment component detection layer, the functional block identification layer and the overall drawing analysis layer; Based on the comprehensive feature mapping, the drawing analysis results, the functional block identification results, the equipment component detection results, and the component parameter verification results, a comprehensive defect location result is generated; The reporting module is used to generate a corresponding defect location report based on the comprehensive defect location results and to display it visually.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for locating defects in power engineering drawings as described in any one of claims 1-6.

9. A storage medium, characterized in that, The computer program is stored and can be loaded by a processor and executed as described in any one of claims 1-6 for the method of locating defects in power engineering drawings.

Images