A method for automatically generating a power station model

Abstract

The application discloses a kind of factory station model automatic generation method, based on electric power primary connection diagram, with the aid of computer and supporting software program is realized, by automatically identifying the engineering type in PDF / JPG format electrical primary main wiring drawing paper, primary and secondary equipment graphic element, graphic element identification, graphic element and identification matching relationship, graphic element topological connection relationship, and generate agreed XML file, then automatically parse XML data, generate station map, and correctly generate topological connection relationship etc., generate setting calculation power grid model.The application adopts the method of drawing automatic identification intelligent analysis automatic modeling, instead of setting calculation software traditional manual modeling, greatly improve the efficiency, accuracy of relay protection setting calculation software power grid modeling.

Description

Technical Field

[0001] This invention relates to the field of power grid relay protection and setting calculation technology, and to a method for automatically generating plant and substation models, particularly a method that can automatically and intelligently generate plant and substation model diagrams without relying solely on expert experience. Background Technology

[0002] The relay protection setting calculation first requires establishing the primary and secondary power grid models in the relay protection setting calculation software. Currently, the main methods for drawing the station diagram in the setting calculation software are to manually draw the equipment, station lines, and inter-station lines based on the electrical primary wiring diagram of the project to be put into operation, or to draw them using typical experience templates, and then manually set the names of the maintenance equipment, etc.

[0003] Current modeling methods rely on the experience of engineers and require accurate drawing interpretation. Furthermore, manual drawing is time-consuming and inefficient when dealing with complex wiring methods or a large number of equipment elements within the station.

[0004] Image recognition, a branch of artificial intelligence technology, refers to the technology of using computers to process, analyze, and understand images to identify targets and objects of various patterns. It is a practical application of deep learning algorithms. Currently, image recognition technology is generally divided into facial recognition and product recognition. Facial recognition is mainly used in security checks, identity verification, and mobile payments; product recognition is mainly used in the commodity circulation process, especially in unmanned retail fields such as unmanned shelves and smart retail cabinets, where it has achieved good results. However, the recognition of electrical diagrams still faces the following problems.

[0005] 1. The data in power engineering drawings is quite complex and large in scale. The circuit diagrams contained therein are intricate and inconsistent, with closely spaced text. Traditional artificial intelligence algorithms alone cannot solve the logically complex drawings, resulting in poor recognition accuracy.

[0006] 2. Traditional power drawing recognition only identifies primary equipment such as busbars, circuit breakers, disconnectors, transformers, generators, and outgoing lines in the primary electrical wiring drawings. It is mainly used for primary power grids, lacks some information, and has no effective data that can be used for relay protection setting calculation and modeling. Summary of the Invention

[0007] The purpose of this invention is to improve the efficiency of relay protection setting calculation software modeling and the accuracy of power line drawing recognition. It proposes an automatic modeling method for relay protection setting calculation software based on a combination of artificial intelligence technology and expert experience. This method automatically identifies the project type, primary and secondary equipment elements, element identifiers, element and identifier matching relationships, and element topology connections in PDF / JPG format electrical primary main wiring drawings, generates a predefined XML file, and then automatically parses the XML data, generates station diagrams, and correctly generates topology connections, thus creating a power grid model for setting calculations.

[0008] The technical solution adopted in this invention is an automatic power plant model generation method, based on a primary power wiring diagram and implemented using a computer and supporting software programs. The automatic power plant model generation method includes the following steps:

[0009] S1. Extract information from the primary wiring diagram of the power system, including extracting information from the table in the lower right corner of the diagram and the graphic element information of the primary and secondary power equipment, and obtain the connection topology of the graphic elements.

[0010] S2. Extract a large amount of historical power primary wiring diagram information multiple times using the above steps, and construct an information model for power grid modeling using power primary wiring diagram extraction and setting calculation software through training;

[0011] S3. Using the information model constructed in step S2, extract the key information used for setting calculation modeling from the power drawings and save it as an XML format file. The key information includes project type, primary and secondary equipment, text, equipment text matching relationship, and topology relationship.

[0012] S4. Based on the above key information, the XML format file is parsed using tuning calculation software to generate a plant model that can be directly used by the tuning calculation software.

[0013] In step S1, the connection topology of the elements is obtained through the following steps:

[0014] S11. In the table at the bottom right of the power drawing, determine the project type based on the suffix of the project name, including: new construction project, expansion project, and renovation project;

[0015] S12. Based on the convex optimization theory and the principle of data probability distribution, the original input drawing is standardized and multi-angle data is amplified. The data is centered by removing the mean and the training set samples are increased. The FCOS algorithm for object detection is used to remove the predefined anchor boxes. The amplified image is subjected to primitive detection. Based on the monitoring results, the location and category information of primary and secondary equipment primitives in the drawing are given.

[0016] S13. Use OCR technology to identify different text directions in drawings, including graphic codes of text, symbols, and numbers, as well as voltage levels and their positions in the drawings;

[0017] S14. Based on the identified graphic element information and text information, match the nearest text and graphic element according to Euclidean distance, and then match them with the bus graphic elements according to the naming rules to obtain the text label and graphic element matching set.

[0018] S15. Convert the RGB image of the mask and primitive topological connections into a binary image. By marking the white pixel targets in the binary image, each individual connected region forms an identified block. Further obtain the geometric parameters of these blocks, including contour, bounding rectangle, centroid, and invariant moments. Then, use the Two-Pass algorithm or Seed-Filling algorithm to identify the connected regions and finally obtain the connection topology of the primitives.

[0019] Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0020] The method of automatic drawing recognition, intelligent analysis and automatic modeling is adopted to replace the traditional manual modeling of setting calculation software, which greatly improves the efficiency and accuracy of power grid modeling of relay protection setting calculation software.

[0021] By combining artificial intelligence algorithms with expert experience, traditional power grid rules and professional knowledge are integrated with artificial intelligence technology. This approach effectively solves the problems of complex and large-scale data in power engineering drawings, intricate and inconsistent circuit diagrams, and poor recognition accuracy caused by closely spaced text. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0023] In its specific implementation, this invention first uses OCR technology to identify the text in the table at the bottom right of the primary power connection line drawing, and then determines the project type based on the project name suffix, such as new construction project, expansion project, or renovation project. Different project types will be processed differently in subsequent software processing.

[0024] Next, we perform primitive and text-based primitive recognition. For primitive recognition, the original input image undergoes standardization and multi-angle data augmentation. The FCOS algorithm for object detection is then used to detect primitives in the augmented image, providing the location and category information of the primitives in the image based on the detection results. For text-based primitive recognition, OCR technology is used to identify different text directions, including primitive codes for text, symbols, numbers, voltage levels, and their locations on the drawing.

[0025] Then, the matching of graphic elements and text is implemented. Based on the identified graphic element information and graphic element text information, the matching process is carried out by matching the nearest text and graphic element according to the Euclidean distance.

[0026] Finally, by synchronously generating corresponding masks from the results of primitive text matching, the masks and the drag links on the original image are converted into binary images. The connection relationship between all masks is determined based on the connected component algorithm, special intersections are processed, and the topological relationship is finally obtained.

[0027] By repeatedly performing the above process and continuously training with a large amount of historical drawing information, an information model is built.

[0028] Extract key information used for setting calculation modeling from power drawings and save it as an XML file;

[0029] Using tuning calculation software to parse XML format files, a plant model that can be directly used by the software is generated. If the project type is "New Project," and if there is already equipment in the plant, all database and cache data related to that equipment are deleted, including lines connecting to other plants. Then, a complete site diagram is generated according to the content of the XML file. If the project type is "Expansion Project," after parsing the XML file, it is necessary to determine whether there is equipment with the same name and type in the current site diagram, including lines. If so, the equipment is not redrawn, but the equipment parsed from the XML file must be correctly connected to the equipment topology in the original site diagram; only equipment existing in the original diagram is retained.

Claims

1. A method for automatically generating a power plant model, based on a primary power wiring diagram and implemented using a computer and supporting software programs, characterized in that: The method for automatically generating plant models includes the following steps: S1. Extract information from the primary wiring diagram of the power system, including extracting information from the table in the lower right corner of the diagram and the graphic element information of the primary and secondary power equipment, and obtain the connection topology of the graphic elements. S2. Extract a large amount of historical power primary wiring diagram information multiple times using the above steps, and construct an information model for power grid modeling using power primary wiring diagram extraction and setting calculation software through training; S3. Using the information model constructed in step S2, extract the key information used for setting calculation modeling from the power drawings and save it as an XML format file. The key information includes project type, primary and secondary equipment, text, equipment text matching relationship, and topology relationship. S4. Based on the above key information, use tuning calculation software to parse the XML format file and finally generate a plant model that can be directly used by the tuning calculation software. In step S1, the connection topology of the elements is obtained through the following steps: S11. In the table at the bottom right of the power drawing, determine the project type based on the suffix of the project name, including: new construction project, expansion project, and renovation project; S12. Based on the convex optimization theory and the principle of data probability distribution, the original input drawing is standardized and multi-angle data is amplified. The data is centered by removing the mean and the training set samples are increased. The FCOS algorithm for object detection is used to remove the predefined anchor boxes. The amplified image is subjected to primitive detection. Based on the monitoring results, the location and category information of primary and secondary equipment primitives in the drawing are given. S13. Use OCR technology to identify different text directions in drawings, including the graphic codes of text, symbols, and numbers, as well as voltage levels and their positions in the drawings; S14. Based on the identified graphic element information and text information, match the nearest text and graphic element according to Euclidean distance, and then match them with the bus graphic elements according to the naming rules to obtain the text label and graphic element matching set. S15. Convert the RGB image of the mask and primitive topological connections into a binary image. By marking the white pixel targets in the binary image, each individual connected region forms an identified block. Further obtain the geometric parameters of these blocks, including contour, bounding rectangle, centroid, and invariant moments. Then, use the Two-Pass algorithm or Seed-Filling algorithm to identify the connected regions and finally obtain the connection topology of the primitives.