Power grid tidal current diagram element identification method and system based on deep learning

Abstract

The invention discloses a power grid tidal current diagram element identification method and system based on deep learning, and the method comprises the steps: segmenting an original tidal current diagram in a total coordinate system to obtain a first tidal current sub-diagram, and carrying out the editing and expansion of the first tidal current sub-diagram to obtain a second tidal current sub-diagram; inputting the second power flow sub-graph into a target detection model to obtain identification information of a plant station element, identifying a plant station candidate name through a pre-trained OCR character identification model, and performing position matching on the plant station candidate name and the plant station element to obtain a plant station name; mapping plant-station elements and plant-station names in the second power flow sub-graphs to a total coordinate system; identifying the name of a power transmission line in the second power flow sub-graph by using a Qwen-VL model, and adding a connection edge between two plant-station elements according to the name of the power transmission line to establish a plant-station topology relationship network; according to the method, high-precision identification and matching of plant station elements, routes and names in the power grid tidal current diagram are realized, so that a power grid topological structure is automatically reconstructed.

Description

Technical Field

[0001] The present invention belongs to the field of image recognition, and in particular relates to a method and system for recognizing elements in power grid flow diagrams based on deep learning. Background Art

[0002] In the operation and maintenance of modern power systems, power flow diagrams are essential tools for reflecting the operational status of power networks. These diagrams typically display various power plants (such as power stations, substations, and converter stations) and their connection routes, along with key naming information. However, in actual operation and maintenance, raw power grid structure data is often stored in bitmap formats (such as JPG and PNG) due to data conversion or system compatibility issues.

[0003] At present, power grid flow diagram recognition methods are mainly divided into two categories: traditional image processing methods and deep learning methods.

[0004] Traditional methods are implemented through a multi-step process, including image preprocessing (such as binarization and edge detection), symbol recognition (based on template matching or morphological operations), and text extraction (OCR technology); however, they have poor generalization capabilities for complex topological structures (such as intersecting routes), and OCR has low accuracy in the dense text background of power grid maps.

[0005] Deep learning methods (such as the Faster R-CNN model and the YOLO model) can directly detect components such as plants and stations in images end-to-end, and combined with improved OCR models (such as PaddleOCR) to extract text, significantly improving the level of automation. However, component detection is easily affected by scale changes and occlusion, and the relationship between text and symbols is difficult to model. Summary of the Invention

[0006] The present invention provides a method and system for identifying elements in power grid flow diagrams based on deep learning, which realizes high-precision identification and matching of plant and station elements, routes and their names in power grid flow diagrams, thereby automatically reconstructing the power grid topology.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is:

[0008] A first aspect of the present invention provides a method for identifying elements in a power flow diagram based on deep learning, comprising:

[0009] Obtain the original tidal current diagram and construct a global coordinate system, segment the original tidal current diagram within the global coordinate system to obtain a first tidal current sub-diagram, and edit and expand the first tidal current sub-diagram to obtain a second tidal current sub-diagram;

[0010] The second power flow subgraph is input into the target detection model, and the second power flow subgraph is identified to obtain identification information of the plant station element, wherein the identification information includes the location and type of the plant station element; a local identification image is extracted from the second power flow subgraph according to the location of the plant station element;

[0011] Input the local recognition image into the pre-trained OCR text recognition model to obtain several candidate plant station names, and perform position matching between the candidate plant station names and plant station elements to obtain the plant station name; map the plant station elements and plant station names in each second power flow sub-graph to the global coordinate system;

[0012] The Qwen-VL model is used to identify the names of transmission lines in the second power flow subgraph, and connecting edges are added between the two plant-station elements according to the names of the transmission lines to establish a plant-station topological relationship network.

[0013] Furthermore, the first power flow subgraph is edited and expanded to obtain a second power flow subgraph, specifically including:

[0014] Calculate the width and height of the first power flow subgraph using the following formula:

[0015]

[0016] In the formula, is the width of the first power flow subgraph, is the height of the first power flow subgraph; and are the width and height of the original tidal flow chart respectively; is the number of first power flow subgraphs included in the original power flow graph along the width direction; is the number of the first tidal subgraphs included in the original tidal graph along the height direction;

[0017] On the premise that the height of the first tidal subgraph is the same, the width of the first tidal subgraph is edited and expanded to obtain a part of the second tidal subgraph; on the premise that the width of the first tidal subgraph is the same, the height of the first tidal subgraph is edited and expanded to obtain another part of the second tidal subgraph.

[0018] Furthermore, the width of the first power flow subgraph is edited and expanded to obtain part of the second power flow subgraph, specifically including:

[0019] The column number of the first tidal subgraph in the original tidal graph is recorded as ;when =1, adjust the width coverage of the second power flow subgraph to ;

[0020] when When , adjust the width coverage of the second power flow subgraph to: ;

[0021] when When , adjust the width coverage of the second power flow subgraph to: ;

[0022] in, , is the starting horizontal coordinate of the first tidal subgraph in the i-th column of the original tidal graph, , is the adaptive expansion amount of the first power flow subgraph in the i-th column; is the adaptive expansion coefficient of the first power flow subgraph in the i-th column;

[0023] The adaptive expansion coefficient The calculation formula is:

[0024]

[0025] In the formula, is the adjustment coefficient, is the characteristic density of the plant element in the first power flow subgraph in the i-th column.

[0026] Furthermore, the height of the first power flow subgraph is edited and expanded to obtain part of the second power flow subgraph, specifically including:

[0027] The row number of the first tidal subgraph in the original tidal graph is recorded as ;when =1, adjust the height coverage of the second power flow subgraph to ;

[0028] when When , adjust the height coverage of the second power flow subgraph to: ;

[0029] when When , adjust the height coverage of the second power flow subgraph to: ;

[0030] in, , is the starting vertical coordinate of the first tidal subgraph in the jth row of the original tidal graph, , is the adaptive expansion amount of the first power flow subgraph in row j; is the adaptive expansion coefficient of the first power flow subgraph in the j-th row.

[0031] Furthermore, the second power flow subgraph is input into the target detection model, and the second power flow subgraph is identified to obtain identification information of the plant element. The identification information includes the location and type of the plant element, specifically including:

[0032] The second power flow subgraph is input into the target detection model, and the second power flow subgraph is identified to obtain the location of the plant element;

[0033] Extracting a sub-mask containing a single plant / station element and a sub-mask distribution sub-map from the second power flow sub-map according to the position of the plant / station element; splicing the sub-mask distribution sub-maps to obtain a sub-mask distribution map;

[0034] The sub-mask and the sub-mask distribution map are superimposed to obtain a single plant station image; the single plant station image is input into the target detection model, and the plant station elements are identified to obtain the type of the plant station elements.

[0035] Furthermore, the candidate plant station names are positionally matched with the plant station elements to obtain the plant station names, specifically including:

[0036] Mark the center coordinates of the candidate station names in the second power flow subgraph as , calculate the distance from the center coordinate of the candidate plant station name to the plant station element, and the expression formula is:

[0037]

[0038] In the formula, The coordinates of the plant element in the second power flow subgraph; is the distance from the center coordinate of the kth plant station candidate name to the rth plant station element;

[0039] The candidate plant station name closest to the plant station element is selected as the plant station name.

[0040] Furthermore, a connection edge is added between the two plant / station elements according to the transmission line name to establish a plant / station topology network, specifically including:

[0041] Extracting a plant sub-mask containing a single plant sub-element from the second power flow subgraph based on the position of the plant sub-element; inputting the plant sub-mask into a pre-trained instance segmentation model to obtain adjacent connection lines of the plant sub-element; screening the adjacent connection lines based on the positional relationship between the plant sub-element and the adjacent connection lines to obtain the plant sub-element connection lines; the plant sub-element connection lines include a plurality of transmission lines; and calculating the number of transmission lines connected to the plant sub-element based on the number and type of the plant sub-element connection lines;

[0042] Initialize the connection edges of the plant / station elements to 0, filter the transmission line names using regular expressions, and extract key indexes from the transmission line names. Search the plant / station names based on the key indexes to obtain a set of associated plant / station elements, and remove plant / station elements whose connection edges reach the set number of transmission lines from the associated plant / station set.

[0043] Generate a prompt sentence based on the associated plant-station set, input the prompt sentence and the original power flow diagram into the Qwen-VL model, determine the two plant-station elements with circuit connection relationship and add connecting edges;

[0044] Traverse all transmission line names and add the connection edges corresponding to the transmission line names to obtain the plant-station topology network.

[0045] Furthermore, the adjacent connection lines are screened according to the positional relationship between the plant-station elements and the adjacent connection lines to obtain the plant-station connection lines, specifically including:

[0046] Calculate the distance between the rth plant element and the sth adjacent connecting line , the expression formula is:

[0047]

[0048] In the formula, The coordinates of the plant and station elements in the plant and station mask; The coordinates of the sth adjacent connecting line in the plant-station mask;

[0049] Calculate the correlation threshold based on the mask size of the plant station , the expression formula is:

[0050]

[0051] In the formula, and The width and height of the plant station mask;

[0052] Filter out distance Less than the association threshold The adjacent connecting lines are used as plant-station connecting lines.

[0053] A second aspect of the present invention provides a power grid flow diagram element recognition system based on deep learning, comprising:

[0054] A segmentation module is used to obtain the original tidal current diagram and construct a global coordinate system, segment the original tidal current diagram in the global coordinate system to obtain a first tidal current sub-diagram, and edit and expand the first tidal current sub-diagram to obtain a second tidal current sub-diagram;

[0055] The plant element recognition module inputs the second power flow subgraph into the target detection model, identifies the second power flow subgraph to obtain identification information of the plant element, including the location and type of the plant element; and extracts a local recognition image from the second power flow subgraph based on the location of the plant element.

[0056] The plant name recognition module inputs the local recognition image into the pre-trained OCR text recognition model to obtain several candidate plant names, matches the candidate plant names with the plant elements to obtain the plant name; and maps the plant elements and plant names in each second power flow sub-graph to the global coordinate system.

[0057] Transmission line identification: The Qwen-VL model is used to identify the transmission line name in the second power flow subgraph. According to the transmission line name, a connection edge is added between the two plant and station elements to establish a plant and station topology relationship network.

[0058] The third aspect of the present invention provides an electronic terminal, comprising a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the power grid flow diagram element identification method described in the first aspect of the present invention.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The present invention segments the original power flow diagram and then performs recognition, reducing the complexity of single image processing, accelerating model calculations, quickly completing power flow diagram recognition and analysis, improving power grid operation and maintenance efficiency, and improving recognition accuracy. Editing and expanding the first power flow sub-graph can avoid damage to the integrity of some plant and station elements and character strings due to segmentation, ensuring that each plant and station element and character string can be accurately recognized and avoiding the loss of key content.

[0061] The present invention inputs the local recognition image into a pre-trained OCR text recognition model to obtain several candidate names of plant and station, performs position matching on the candidate names of plant and station with the plant and station elements to obtain the plant and station names, performs position matching on the candidate names of plant and station with the plant and station elements, and uses information such as spatial relationships and distances to screen out the most practical plant and station names, effectively avoiding errors caused by factors such as similar text, and ensuring that the plant and station names accurately correspond to the corresponding plant and station elements.

[0062] The present invention maps the plant elements and plant names in each second flow subgraph to the global coordinate system, realizes the conversion from local coordinates to global coordinates, and records the position information of the plant names and plant elements in the global coordinate system; uses the Qwen-VL model to identify the transmission line names in the second flow subgraph, and adds connection edges between the two plant elements according to the transmission line names to establish a plant-station topological relationship network, which more accurately reflects the actual connection relationship of the power grid and improves the reliability of power grid analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a flow chart of a method for identifying elements in a power grid flow diagram based on deep learning provided by the present invention;

[0064] Figure 2 It is the original tidal current segmentation diagram provided by the present invention;

[0065] Figure 3 The present invention provides a segmentation graph for editing and expanding the width of the first power flow subgraph;

[0066] Figure 4 The segmentation graph provided by the present invention is obtained by editing and expanding the height of the first power flow subgraph;

[0067] Figure 5 This is the general distribution diagram of plant and station masks provided by the present invention;

[0068] Figure 6 This is the actual plant and station distribution map provided by the present invention;

[0069] Figure 7 and Figure 8 The present invention provides a mask image. DETAILED DESCRIPTION

[0070] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0071] Example 1

[0072] like Figure 1 As shown, this embodiment provides a method for identifying elements in a power flow diagram based on deep learning, comprising the following steps:

[0073] like Figure 2 As shown, the original tidal current diagram is obtained and the overall coordinate system is constructed. The original tidal current diagram is segmented in the overall coordinate system to obtain A first power flow subgraph is obtained by editing and expanding the first power flow subgraph to obtain a second power flow subgraph; specifically, the following steps are performed:

[0074] Calculate the width and height of the first power flow subgraph using the following formula:

[0075]

[0076] In the formula, is the width of the first power flow subgraph, is the height of the first power flow subgraph; and are the width and height of the original tidal flow chart respectively; is the number of first power flow subgraphs included in the original power flow graph along the width direction; is the number of the first tidal subgraphs included in the original tidal graph along the height direction;

[0077] like Figure 3 As shown, under the premise that the height of the first power flow sub-graph is the same, the width of the first power flow sub-graph is edited and expanded to obtain part of the second power flow sub-graph, specifically including:

[0078] The column number of the first tidal subgraph in the original tidal graph is recorded as ;when =1, adjust the width coverage of the second power flow subgraph to ;

[0079] when When , adjust the width coverage of the second power flow subgraph to: ;

[0080] when When , adjust the width coverage of the second power flow subgraph to: ;

[0081] in, , is the starting horizontal coordinate of the first tidal subgraph in the i-th column of the original tidal graph, , is the adaptive expansion amount of the first power flow subgraph in the i-th column; is the adaptive expansion coefficient of the first power flow subgraph in the i-th column;

[0082] The adaptive expansion coefficient The calculation formula is:

[0083]

[0084] In the formula, is the adjustment coefficient, is the characteristic density of the plant element in the first power flow subgraph in the i-th column.

[0085] like Figure 4 As shown, under the premise that the width of the first power flow sub-graph is the same, the height of the first power flow sub-graph is edited and expanded to obtain another part of the second power flow sub-graph, specifically including:

[0086] The row number of the first tidal subgraph in the original tidal graph is recorded as ;when =1, adjust the height coverage of the second power flow subgraph to ;

[0087] when When , adjust the height coverage of the second power flow subgraph to: ;

[0088] when When , adjust the height coverage of the second power flow subgraph to: ;

[0089] in, , is the starting vertical coordinate of the first tidal subgraph in the jth row of the original tidal graph, , is the adaptive expansion amount of the first power flow subgraph in row j; is the adaptive expansion coefficient of the first power flow subgraph in the j-th row.

[0090] This embodiment segments the original power flow diagram and then performs recognition, reducing the complexity of single image processing, accelerating model calculations, quickly completing power flow diagram recognition and analysis, improving power grid operation and maintenance efficiency, and improving recognition accuracy. Editing and expanding the first power flow sub-diagram can avoid damaging the integrity of some plant and station elements and character strings due to segmentation, ensuring that each plant and station element and character string can be accurately recognized and avoiding the loss of key content.

[0091] The second power flow subgraph is input into the target detection model, and the second power flow subgraph is identified to obtain the identification information of the plant element. The identification information includes the location and type of the plant element, specifically including:

[0092] The second power flow subgraph is input into the target detection model, and the second power flow subgraph is identified to obtain the location of the plant element;

[0093] like Figure 5 and 6 As shown, according to the position of the power station element, each mask sub-graph containing a single power station element is extracted from the second power flow sub-graph, and then the sub-graphs are spliced ​​to obtain the overall power station mask distribution graph;

[0094] The masked distribution map of power stations and the original tidal flow map are superimposed to obtain the actual power station distribution map; then, each single power station image in the actual power station distribution map is input into the target detection model, and the power station elements are identified to obtain the type of the power station elements. Identifying the position and type in steps not only ensures the accuracy of position identification, but also ensures the reliability of type determination, thereby improving the overall system recognition performance.

[0095] In this embodiment, the target detection model is set to the YOLO model. A first training sample is obtained from the power flow diagram database. The first training sample is annotated and used to train the target detection model. The confidence calculation formula for the target detection model for plant type detection is:

[0096]

[0097] In the formula, is the confidence level of the plant type detection, c is the plant type, such as substation, power station, etc.; b is the detection coordinate, is the output score of the target detection model for type c, For multi-label classification function.

[0098] The confidence level of the target detection model output will serve as a key screening indicator. When the confidence level is lower than the preset threshold (such as 0.8), the system will automatically exclude the recognition result and initiate a compensation mechanism, including expanding or narrowing the recognition range to re-predict or adjusting hyperparameters to retrain. When large-scale recognition failures occur, the model parameter optimization process will be triggered to improve recognition performance by dynamically adjusting hyperparameters.

[0099] Based on the location of the plant / station elements, a local recognition image is extracted from the second power flow subgraph. This local recognition image is then fed into a pre-trained optical character recognition (OCR) model to obtain candidate plant / station names. The local recognition image is obtained by expanding the image data around the plant / station using the center of gravity of the mask rectangle of the single plant / station as the center. Because the plant / station names are horizontally oriented and clear, the OCR model can then easily identify the range and content of all text within the image.

[0100] The candidate plant station names are matched with the plant station elements to obtain the plant station names, including:

[0101] Mark the center coordinates of the candidate station names in the second power flow subgraph as , calculate the distance from the center coordinate of the candidate plant station name to the plant station element, and the expression formula is:

[0102]

[0103] In the formula, The coordinates of the plant element in the second power flow subgraph; is the distance from the center coordinate of the kth plant station candidate name to the rth plant station element;

[0104] The candidate plant name closest to the plant element is selected as the plant name. The distance between the center coordinates of the candidate plant name and the plant element is calculated, and the closest candidate name is selected as the plant name. This location-based matching method fully utilizes spatial relationships, avoiding errors caused by similar characters or OCR misrecognition, thereby ensuring accurate correspondence between plant names and plant elements.

[0105] The plant and station elements and plant and station names in each second power flow sub-map are mapped to the global coordinate system; the Qwen-VL model is used to identify the transmission line names in the second power flow sub-map. The transmission line names are composed of three parts: two Chinese characters, a line segment number, and the word "line". The two Chinese characters are the abbreviations of the plant and station names at both ends of the transmission line name.

[0106] Add connection edges between the two plant and station elements based on the transmission line name to establish a plant and station topology network, including:

[0107] Based on the location of the plant element, a plant sub-mask containing a single plant element is extracted from the second power flow subgraph. This plant sub-mask is input into a pre-trained instance segmentation model to obtain the adjacent connection lines of the plant element. The adjacent connection lines are categorized as one-line, two-line, three-line, or four-line.

[0108] The instance segmentation model is a neural convolutional model for image recognition. A power grid flow training graph is obtained from a power flow graph database. A training mask containing a single power station element is extracted from the power grid flow training graph. Annotations are added to the first training sample to obtain a second training sample. The instance segmentation model is trained using the second training sample. After repeated iterations reach a set number of iterations, the trained instance segmentation model is output.

[0109] Calculate the distance between the rth plant element and the sth adjacent connecting line , the expression formula is:

[0110]

[0111] In the formula, The coordinates of the plant and station elements in the plant and station mask; The coordinates of the sth adjacent connecting line in the plant-station mask;

[0112] Calculate the correlation threshold based on the mask size of the plant station , the expression formula is:

[0113]

[0114] In the formula, and The width and height of the plant station mask;

[0115] Filter out distance Less than the association threshold The adjacent connecting lines are used as plant-station connecting lines. Figure 7 As shown, the double line with a distance of 107.2 will be filtered out. Figure 8 As shown, the double line 106.1 will be filtered out.

[0116] The plant-station connection line includes several transmission lines. The number of transmission lines connected to the plant-station element is calculated according to the number and type of the plant-station connection lines. The expression formula is:

[0117]

[0118] In the formula, is the number of transmission lines connected to the plant-station element, t is the type number of the plant-station connection line, is the tth type of adjacent connection line connecting the plant station element; v is the total number of types of adjacent connection lines.

[0119] Initialize the connection edges of the plant / station elements to 0. After filtering the transmission line names using a regular expression, extract the key index from the transmission line names. Search the plant / station names based on the key index to obtain a set of associated plant / station elements. Remove plant / station elements from this set of associated plant / station elements whose connection edges reach the set number of transmission lines. Remove plant / station elements from this set of associated plant / station elements whose connection edges reach the set number of transmission lines. This step aims to avoid duplicate connections and ensure the accuracy of the connection relationship between each plant / station element.

[0120] After filtering transmission line names using regular expressions, the accuracy rate reached over 95.6%. All routes must be matched against the power plant. There are three types of matching between power plant elements: one-to-one, one-to-many, and many-to-many.

[0121] Table 1, prompt word template list

[0122]

[0123] As shown in Table 1, a prompt sentence is generated based on the associated plant-station set. The prompt sentence and the original power flow diagram are input into the Qwen-VL model to determine the two plant-station elements with circuit connection relationship and add connecting edges;

[0124] Traverse all transmission line names and add the connection edges corresponding to the transmission line names to obtain the plant-station topology network.

[0125] The Qwen-VL model and OCR text recognition model in this embodiment are both existing technologies, and the training and construction processes are not repeated here.

[0126] The plant and station elements and their names in each second power flow subgraph are mapped to the global coordinate system, achieving local-to-global coordinate conversion and accurately recording the location of the plant and station names and elements in the global coordinate system. The Qwen-VL model is used to identify transmission line names, and connecting edges are added between plant and station elements based on the transmission line names to establish a plant-station topology network. This more accurately reflects the actual connections in the power grid, thereby improving the reliability of power grid analysis.

[0127] Example 2

[0128] This embodiment further provides a power flow diagram element recognition system based on deep learning. The power flow diagram element recognition system can execute the power flow diagram element recognition method described in Example 1. The power flow diagram element recognition system includes:

[0129] A segmentation module is used to obtain the original tidal current diagram and construct a global coordinate system, segment the original tidal current diagram in the global coordinate system to obtain a first tidal current sub-diagram, and edit and expand the first tidal current sub-diagram to obtain a second tidal current sub-diagram;

[0130] The plant element recognition module inputs the second power flow subgraph into the target detection model, identifies the second power flow subgraph to obtain identification information of the plant element, including the location and type of the plant element; and extracts a local recognition image from the second power flow subgraph based on the location of the plant element.

[0131] The plant name recognition module inputs the local recognition image into the pre-trained OCR text recognition model to obtain several candidate plant names, matches the candidate plant names with the plant elements to obtain the plant name; and maps the plant elements and plant names in each second power flow sub-graph to the global coordinate system.

[0132] Transmission line identification: The Qwen-VL model is used to identify the transmission line name in the second power flow subgraph. According to the transmission line name, a connection edge is added between the two plant and station elements to establish a plant and station topology relationship network.

[0133] Example 3

[0134] This embodiment also provides an electronic terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the power grid flow diagram element identification method described in Example 1.

[0135] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0136] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0137] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0138] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0139] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for identifying elements in power flow diagrams based on deep learning, characterized in that: include: Obtain the original tidal current diagram and construct a global coordinate system, segment the original tidal current diagram within the global coordinate system to obtain a first tidal current sub-diagram, and edit and expand the first tidal current sub-diagram to obtain a second tidal current sub-diagram; The second power flow subgraph is input into the target detection model, and the second power flow subgraph is identified to obtain identification information of the plant station element, wherein the identification information includes the location and type of the plant station element; a local identification image is extracted from the second power flow subgraph according to the location of the plant station element; Input the local recognition image into the pre-trained OCR text recognition model to obtain several candidate plant and station names, and perform position matching between the candidate plant and station names and the plant and station elements to obtain the plant and station names; Map the plant and station elements and plant and station names in each second power flow sub-graph to the global coordinate system; The Qwen-VL model is used to identify the names of transmission lines in the second power flow subgraph, and connecting edges are added between the two plant-station elements according to the names of the transmission lines to establish a plant-station topological relationship network.

2. The method for identifying elements in a power grid flow diagram according to claim 1, wherein: Edit and expand the first power flow subgraph to obtain the second power flow subgraph, specifically including: Calculate the width and height of the first power flow subgraph using the following formula: ； In the formula, is the width of the first power flow subgraph, is the height of the first power flow subgraph; and are the width and height of the original tidal flow chart respectively; is the number of first power flow subgraphs included in the original power flow graph along the width direction; is the number of the first tidal subgraphs included in the original tidal graph along the height direction; On the premise that the height of the first tidal subgraph is the same, the width of the first tidal subgraph is edited and expanded to obtain a part of the second tidal subgraph; on the premise that the width of the first tidal subgraph is the same, the height of the first tidal subgraph is edited and expanded to obtain another part of the second tidal subgraph.

3. The method for identifying elements in a power grid flow diagram according to claim 2, wherein: Edit and expand the width of the first power flow subgraph to obtain part of the second power flow subgraph, specifically including: The column number of the first tidal subgraph in the original tidal graph is recorded as ;when =1, adjust the width coverage of the second power flow subgraph to ; when When , adjust the width coverage of the second power flow subgraph to: ; when When , adjust the width coverage of the second power flow subgraph to: ; in, , is the starting horizontal coordinate of the first tidal subgraph in the i-th column of the original tidal graph, , is the adaptive expansion amount of the first power flow subgraph in the i-th column; is the adaptive expansion coefficient of the first power flow subgraph in the i-th column; The adaptive expansion coefficient The calculation formula is: ； In the formula, is the adjustment coefficient, is the characteristic density of the plant element in the first power flow subgraph in the i-th column.

4. The method for identifying elements in a power grid flow diagram according to claim 2, wherein: Edit and expand the height of the first power flow subgraph to obtain part of the second power flow subgraph, specifically including: The row number of the first tidal subgraph in the original tidal graph is recorded as ;when =1, adjust the height coverage of the second power flow subgraph to ; when When , adjust the height coverage of the second power flow subgraph to: ; when When , adjust the height coverage of the second power flow subgraph to: ; in, , is the starting vertical coordinate of the first tidal subgraph in the jth row of the original tidal graph, , is the adaptive expansion amount of the first power flow subgraph in row j; is the adaptive expansion coefficient of the first power flow subgraph in the j-th row.

5. The method for identifying elements of a power grid flow diagram according to claim 1, wherein: The second power flow subgraph is input into the target detection model, and the second power flow subgraph is identified to obtain the identification information of the plant element. The identification information includes the location and type of the plant element, specifically including: The second power flow subgraph is input into the target detection model, and the second power flow subgraph is identified to obtain the location of the plant element; Extracting a sub-mask containing a single plant / station element and a sub-mask distribution sub-map from the second power flow sub-map according to the position of the plant / station element; splicing the sub-mask distribution sub-maps to obtain a sub-mask distribution map; The sub-mask and the sub-mask distribution map are superimposed to obtain a single plant station image; the single plant station image is input into the target detection model, and the plant station elements are identified to obtain the type of the plant station elements.

6. The method for identifying elements of a power grid flow diagram according to claim 1, characterized in that: The candidate plant station names are matched with the plant station elements to obtain the plant station names, including: Mark the center coordinates of the candidate station names in the second power flow subgraph as , calculate the distance from the center coordinate of the candidate plant station name to the plant station element, and the expression formula is: ； In the formula, The coordinates of the plant element in the second power flow subgraph; is the distance from the center coordinate of the kth plant station candidate name to the rth plant station element; The candidate plant station name closest to the plant station element is selected as the plant station name.

7. The method for identifying elements in a power grid flow diagram according to claim 1, characterized in that: Add connection edges between the two plant and station elements based on the transmission line name to establish a plant and station topology network, including: Extracting a plant sub-mask containing a single plant sub-element from the second power flow subgraph based on the position of the plant sub-element; inputting the plant sub-mask into a pre-trained instance segmentation model to obtain adjacent connection lines of the plant sub-element; screening the adjacent connection lines based on the positional relationship between the plant sub-element and the adjacent connection lines to obtain the plant sub-element connection lines; the plant sub-element connection lines include a plurality of transmission lines; and calculating the number of transmission lines connected to the plant sub-element based on the number and type of the plant sub-element connection lines; Initialize the connection edges of the plant / station elements to 0, filter the transmission line names using regular expressions, and extract key indexes from the transmission line names. Search the plant / station names based on the key indexes to obtain a set of associated plant / station elements, and remove plant / station elements whose connection edges reach the set number of transmission lines from the associated plant / station set. Generate a prompt sentence based on the associated plant-station set, input the prompt sentence and the original power flow diagram into the Qwen-VL model, determine the two plant-station elements with circuit connection relationship and add connecting edges; Traverse all transmission line names and add the connection edges corresponding to the transmission line names to obtain the plant-station topology network.

8. The method for identifying elements in a power grid flow diagram according to claim 7, characterized in that: According to the positional relationship between the plant and station elements and the adjacent connection lines, the adjacent connection lines are screened to obtain the plant and station connection lines, specifically including: Calculate the distance between the rth plant element and the sth adjacent connecting line , the expression formula is: ； In the formula, The coordinates of the plant and station elements in the plant and station mask; The coordinates of the sth adjacent connecting line in the plant-station mask; Calculate the correlation threshold based on the mask size of the plant station , filter out the distance Less than the association threshold The adjacent connecting lines are used as plant-station connecting lines.

9. A power grid flow diagram element recognition system based on deep learning, characterized in that: include: A segmentation module is used to obtain the original tidal current diagram and construct a global coordinate system, segment the original tidal current diagram in the global coordinate system to obtain a first tidal current sub-diagram, and edit and expand the first tidal current sub-diagram to obtain a second tidal current sub-diagram; The plant element recognition module inputs the second power flow subgraph into the target detection model, identifies the second power flow subgraph to obtain identification information of the plant element, including the location and type of the plant element; and extracts a local recognition image from the second power flow subgraph based on the location of the plant element. The plant and station name recognition module inputs the local recognition image into the pre-trained OCR text recognition model to obtain several candidate plant and station names, and then matches the candidate plant and station names with the plant and station elements to obtain the plant and station names; Map the plant and station elements and plant and station names in each second power flow sub-graph to the global coordinate system; Transmission line identification: The Qwen-VL model is used to identify the transmission line name in the second power flow subgraph. According to the transmission line name, a connection edge is added between the two plant and station elements to establish a plant and station topology relationship network.

10. An electronic terminal, characterized in that: It comprises a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the power grid flow diagram element identification method according to any one of claims 1 to 8.

Images