Graph and parameter capturing method and system for power transformation drawing

Abstract

The invention discloses a graph and parameter capturing method for a power transformation drawing. The method comprises the steps that existing power transformation drawing data information is acquired and preprocessed to construct a training data set; based on feature analysis, semantic analysis and a multi-layer perceptron, constructing a graph and parameter capturing initial model of the power transformation drawing, and training to obtain a graph and parameter capturing model of the power transformation drawing; and the obtained graph and parameter grabbing model of the power transformation drawing is adopted to grab the graph and parameters of the actual power transformation drawing. The invention further discloses a system for achieving the graph and parameter grabbing method of the power transformation drawing. According to the method, the data set is marked and sorted, and the graph and parameter capturing model is built and trained based on feature analysis, semantic analysis and a multi-layer perceptron, so that automatic capturing of the graph and the parameter of the power transformation drawing is realized, and the method is high in reliability, good in accuracy, more convenient and quicker.

Description

Technical Field

[0001] The present invention belongs to the field of electrical automation, and particularly relates to a method and system for capturing graphics and parameters of substation drawings. Background Art

[0002] With the development of economic technology and the improvement of people's living standards, electric energy has become an essential secondary energy source in people's production and life, bringing endless convenience to people's production and life. Therefore, ensuring the stable and reliable supply of electric energy has become one of the most important tasks of the power system.

[0003] In the field of substation engineering design of the power system, the design of substation graphics and their key parameters has always been an important link in the overall design process. During the substation design process, a large number of equipment graphics (such as circuit breakers, transformers, etc.) and complex key parameter designs are involved.

[0004] However, the current substation engineering design scheme has always had the following problems: the existing substation drawings are presented in a paper form, and they do not directly contain the key design parameters of the substation graphics, which makes the reading of substation drawings extremely dependent on manual experience and the efficiency is low; at the same time, even if the drawings and key parameters are digitized, the geometric information of electrical symbols and the key parameter information in the drawings are often stored in different databases and lack a corresponding association mechanism, which also brings great inconvenience. Summary of the Invention

[0005] One object of the present invention is to provide a method for capturing graphics and parameters of substation drawings with high reliability, good accuracy and convenience.

[0006] Another object of the present invention is to provide a system for implementing the method for capturing graphics and parameters of substation drawings.

[0007] The method for capturing graphics and parameters of substation drawings provided by the present invention includes the following steps:

[0008] S1. Obtain the existing substation drawing data information;

[0009] S2. Preprocess the data information obtained in step S1 to construct a training data set;

[0010] S3. Based on feature parsing, semantic analysis and a multi-layer perceptron, construct an initial model for capturing graphics and parameters of substation drawings;

[0011] S4. Use the training data set constructed in step S2 to train the initial model for capturing graphics and parameters of substation drawings constructed in step S3 to obtain a model for capturing graphics and parameters of substation drawings;

[0012] S5. Use the graphic and parameter extraction model of the substation drawing obtained in step S4 to perform actual graphic and parameter extraction of the substation drawing.

[0013] The preprocessing of the data information obtained in step S1 described in step S2 specifically includes the following steps:

[0014] Preprocess the data information obtained in step S1; the preprocessing includes data filtering, data interpolation, and data marking.

[0015] The initial model for graphic and parameter extraction of substation drawings constructed based on feature parsing, semantic analysis, and multi-layer perceptron in step S3 specifically includes the following steps:

[0016] Based on geometric feature parsing and semantic analysis, construct a primitive feature extraction module for synchronously extracting the dual attributes of primitives in the substation drawing;

[0017] Based on the multi-layer perceptron, construct a dynamic coupling module for associating and dynamically coupling graphics and parameters;

[0018] Based on the gating function, construct a coordinate conversion module for converting the drawing coordinate system to the engineering coordinate system to complete the graphic and parameter extraction of the substation drawing.

[0019] The construction of the primitive feature extraction module based on geometric feature parsing and semantic analysis specifically includes the following steps:

[0020] The processing process of the constructed primitive feature extraction module includes the following steps:

[0021] Set the geometric feature matrix as G and the semantic feature matrix as S;

[0022] Calculate the feature tensor T using the following formula:

[0023]

[0024] In the formula, Concat() is the concatenation operation; W g is the geometric feature weight matrix; W s is the semantic feature weight matrix; is the outer product operation;

[0025] Calculate the cross-modal attention weight using the following formula:

[0026]

[0027] In the formula, A ij is the attention weight between the geometric modality i and the semantic modality j; σ() is the softmax function; T i is the joint feature tensor of the i-th primitive; Tj is the combined feature tensor of the j-th primitive; d g is the dimension of the geometric feature vector; d s is the dimension of the semantic feature vector;

[0028] The following formula is used for feature adaptive fusion to complete primitive feature extraction:

[0029]

[0030] In the formula is the reconstructed geometric feature of the i-th primitive; n is the number of neighboring primitives participating in feature fusion; G j is the original geometric feature vector of the j-th primitive; S j is the original semantic feature vector of the j-th primitive; is the feature concatenation operation; ⊙ is the Hadamard product operation.

[0031] Based on a multi-layer perceptron, a dynamic coupling module is constructed, which specifically includes the following steps:

[0032] The processing process of the constructed dynamic coupling module includes the following steps:

[0033] The following formula is used to calculate the polar radius:

[0034]

[0035] In the formula, ρ is the polar radius; (x, y) is the device coordinate; (x0, y0) is the coordinate of the drawing origin;

[0036] The following formula is used to calculate the polar angle:

[0037]

[0038] In the formula, θ is the polar angle;

[0039] The following formula is used for non-linear fusion:

[0040] Ρ i = MLP φ (ρ i ||θ i )⊙ Sigmoid(W a AA i )

[0041] In the formula, Ρ i is the dynamic attribute coupling strength of the i-th device in the polar coordinate space; MLP φ () is the multi-layer perceptron operation with parameter φ; ρ i is the straight-line distance of device i relative to the reference point; θ iis the azimuth angle of device i relative to the reference point; Sigmoid() is the sigmoid function; W a is the device attribute weight matrix; AA i is the device attribute vector;

[0042] The polarization sensitivity is calculated using the following formula:

[0043]

[0044] where Γ(P i , P j ) is the polarization sensitivity between P i and P j ; ||P i - P j ||2 is the Euclidean distance between P i and P j , which is used to dynamically adjust the parameter correlation strength between devices; ρ ij is the spatial Euclidean distance between device i and device j; ∈ is the set smoothing factor;

[0045] The following rules are used for the association and dynamic coupling of graphics and parameters:

[0046] Associate and dynamically couple the graphics and parameters of adjacent devices where Γ(P i , P j ) ≤ τ, and perform a matching check on the electrical parameters; τ is the set threshold.

[0047] The following formula is used to calculate the set threshold τ:

[0048] τ = μ + 3σ

[0049] where μ is the expected value of the device parameters; σ is the tolerance factor of the device parameters.

[0050] Based on the gating function, a coordinate transformation module is constructed, which specifically includes the following steps:

[0051] Set the data feature tensor as F and the storage structure feature tensor as S;

[0052] The following formula is used to calculate the joint tensor:

[0053]

[0054] where T is the joint tensor; W f is the feature space importance weight matrix; W S is the spatial topology correlation weight matrix;

[0055] The following formula is used for dynamic transformation:

[0056]

[0057] where M i is the intermediate feature tensor that fuses the feature stream and the structure stream; Softmax() is the softmax function; F i is the feature vector of the device parameters and the spatial topology; W t is the dynamic weight matrix; d t is the dynamic decay factor;

[0058] Adopt the following formula as the gating function:

[0059]

[0060] where G i is the dynamic gating vector; ΔS i is the structure stream increment tensor;

[0061] Adopt the following formula for selective fusion:

[0062]

[0063] where is the gated enhanced feature; S i is the structure stream input feature;

[0064] Map the drawing coordinate system to the engineering coordinate system to achieve alignment; convert unstructured data into structured fields through a rule engine to support dynamic expansion;

[0065] Finally, realize the capture of the graphics and parameters of the substation drawings.

[0066] The present invention also provides a system for implementing the method for capturing graphics and parameters of the substation drawing, including a data acquisition module, a data processing module, a model construction module, a model training module, and a capture module; the data acquisition module, the data processing module, the model construction module, the model training module, and the capture module are connected in series in sequence; the data acquisition module is used to acquire the existing substation drawing data information and upload the data information to the data processing module; the data processing module is used to preprocess the acquired data information according to the received data information to construct a training data set and upload the data information to the model construction module; the model construction module is used to construct an initial model for capturing graphics and parameters of the substation drawing based on feature parsing, semantic analysis, and a multi-layer perceptron according to the received data information and upload the data information to the model training module; the model training module is used to train the constructed initial model for capturing graphics and parameters of the substation drawing by using the constructed training data set according to the received data information to obtain a model for capturing graphics and parameters of the substation drawing and upload the data information to the capture module; the capture module is used to actually capture the graphics and parameters of the substation drawing by using the obtained model for capturing graphics and parameters of the substation drawing according to the received data information.

[0067] The method and system for capturing graphics and parameters of the substation drawing provided by the present invention, through the marking and sorting of the data set, and constructing and training the graphics and parameter capture model based on feature parsing, semantic analysis, and a multi-layer perceptron, not only realizes the automatic capture of the graphics and parameters of the substation drawing, but also has high reliability, good accuracy, and is more convenient and fast. Brief Description of the Drawings

[0068] Figure 1 It is a schematic flowchart of the method of the present invention.

[0069] Figure 2 It is a schematic diagram of the functional modules of the system of the present invention. Detailed Embodiment

[0070] As Figure 1 shown is a schematic flowchart of the method of the present invention: The method for capturing graphics and parameters of the substation drawing disclosed by the present invention includes the following steps:

[0071] S1. Acquire the existing substation drawing data information;

[0072] S2. Preprocess the data information acquired in step S1 to construct a training data set; specifically, it includes the following steps:

[0073] Preprocess the data information acquired in step S1; the preprocessing includes data filtering, data interpolation, and data marking;

[0074] S3. Based on feature parsing, semantic analysis, and multi-layer perceptron, construct an initial model for capturing the graphics and parameters of substation drawings; specifically including the following steps:

[0075] Based on geometric feature parsing and semantic analysis, construct a primitive feature extraction module for synchronously extracting the dual attributes of primitives in substation drawings;

[0076] Based on multi-layer perceptron, construct a dynamic coupling module for associating and dynamically coupling graphics with parameters;

[0077] Based on the gating function, construct a coordinate transformation module for converting the drawing coordinate system into the engineering coordinate system to complete the capture of the graphics and parameters of substation drawings;

[0078] Specifically in implementation, the processing process of the constructed primitive feature extraction module includes the following steps:

[0079] Set the geometric feature matrix as G and the semantic feature matrix as S; among them, for the geometric feature matrix G, geometric feature vectors are extracted through all device primitives in the substation drawing, and the affine transformation matrix is used to convert the substation drawing coordinate system into the engineering coordinate system; for the semantic feature matrix S, device parameter labels such as type coding, voltage level, and device number are parsed through the extended data of device primitives in the substation drawing to form the semantic feature matrix;

[0080] Calculate the feature tensor T using the following formula:

[0081]

[0082] In the formula, Concat() is the concatenation operation; W g is the geometric feature weight matrix; W s is the semantic feature weight matrix; is the outer product operation; the feature dimension is extended through the feature tensor T, and the high-resolution information of the original features is retained; at the same time, different from traditional dimensionality reduction and compression, the high dimensionality of each stream feature is maintained during the tensor construction stage to reduce information loss and enhance the fine-grained expression ability;

[0083] Calculate the cross-modal attention weight using the following formula:

[0084]

[0085] In the formula, A ij is the attention weight between the geometric modality i and the semantic modality j; σ() is the softmax function; T i is the joint feature tensor of the i-th primitive; T j is the joint feature tensor of the j-th primitive; d g is the dimension of the geometric feature vector; d sis the dimension of the semantic feature vector; this mechanism breaks through the limitations of unimodal associations, captures the implicit associations between geometric parameters and semantic labels; at the same time, combined with Softmax, it enhances the contrast of key features and improves the sensitivity of the model to subtle differences;

[0086] The following formula is used for feature adaptive fusion to complete primitive feature extraction:

[0087]

[0088] In the formula is the geometric feature after reconstruction of the i-th primitive; n is the number of neighboring primitives participating in feature fusion; G j is the original geometric feature vector of the j-th primitive; S j is the original semantic feature vector of the j-th primitive; is the feature concatenation operation; ⊙ is the Hadamard product operation; the contribution degree of each feature stream is dynamically adjusted through attention weights to achieve the collaborative optimization of local features and global semantics;

[0089] The constructed dynamic coupling module has a processing process including the following steps:

[0090] The following formula is used for the calculation of the polar radius:

[0091]

[0092] In the formula, ρ is the polar radius; (x, y) are the device coordinates; (x0, y0) are the coordinates of the drawing origin; this step is used to quantify the spatial distance and enhance the representation ability of the device distribution density;

[0093] The following formula is used for the calculation of the polar angle:

[0094]

[0095] In the formula, θ is the polar angle; this step is used to describe the azimuth distribution characteristics of the device in the two-dimensional plane and solve the problem of insufficient direction sensitivity of the traditional Cartesian coordinates;

[0096] The following formula is used for non-linear fusion:

[0097] Ρ i = MLP φ (ρ i ||θ i )⊙Sigmoid(W a AA i )

[0098] In the formula, Ρ i is the dynamic attribute coupling strength of the i-th device in the polar coordinate space; MLP φ() is the operation of a multi-layer perceptron with parameter φ; ρ i is the straight-line distance of device i relative to the reference point; θ i is the azimuth angle of device i relative to the reference point; Sigmoid() is the sigmoid function; W a is the device attribute weight matrix; AA i is the device attribute vector; where the multi-layer perceptron maps the polar coordinate parameters to the feature space and retains the spatial locality features; the sigmoid function is used to compress the device attribute vector and generate dynamic gating weights to suppress the interference of redundant parameters; the adaptive fusion of spatial parameters and device attributes is achieved through the Hadamard product operation;

[0099] The polarization sensitivity is calculated using the following formula:

[0100]

[0101] In the formula, Γ(P i , P j ) is the polarization sensitivity between P i and P j ; ||P i - P j ||2 is the Euclidean distance between P i and P j , which is used to dynamically adjust the parameter correlation strength between devices; ρ ij is the spatial Euclidean distance between device i and device j; ∈ is the set smoothing factor; this step is used to quantify the spatial-attribute coupling strength between devices;

[0102] The following rules are used for the association and dynamic coupling of graphics and parameters:

[0103] Associate and dynamically couple the graphics and parameters of adjacent devices where Γ(P i , P j ) ≤ τ, and perform the matching check of electrical parameters; τ is the set threshold.

[0104] The following formula is used to calculate the set threshold τ:

[0105] τ = μ + 3σ

[0106] In the formula, μ is the expected value of the device parameters; σ is the tolerance factor of the device parameters.

[0107] For the constructed coordinate conversion module, the processing process includes the following steps:

[0108] Set the data feature tensor as F and the storage structure feature tensor as S. Among them, for the data feature tensor F, geometric parameters and extended attributes are extracted through the attribute parameters and monitoring data of substation equipment graphic elements in the substation drawing, and the discrete encoding is converted into a continuous vector by using a pre-trained embedding layer. Among them, for the storage structure feature tensor S, based on the spatial topological relationship and connection status of the equipment in the substation drawing, a spatial adjacency graph is constructed based on the equipment coordinates, and the engineering coordinates of each equipment are discretized to output a unique hash value, and a structure tensor is generated in combination with the spatial adjacency relationship.

[0109] The combined tensor is calculated using the following formula:

[0110]

[0111] In the formula, T is the combined tensor; W f is the importance weight matrix of the feature space; W S is the spatial topology correlation weight matrix; this step is used to retain the fine-grained information of heterogeneous data.

[0112] The following formula is used for dynamic conversion:

[0113]

[0114] In the formula, M i is the intermediate feature tensor that fuses the feature stream and the structure stream; Softmax() is the softmax function; F i is the feature vector of the equipment parameters and the spatial topology; W t is the dynamic weight matrix; d t is the dynamic decay factor;

[0115] The following formula is used as the gating function:

[0116]

[0117] In the formula, G i is the dynamic gating vector; ΔS i is the structure stream increment tensor; this step is used to quantify the correlation strength between the data features and the storage structure.

[0118] The following formula is used for selective fusion:

[0119]

[0120] In the formula is the gated enhanced feature; S i is the structure stream input feature; this step is used to automatically suppress redundant data and enhance key features.

[0121] Map the drawing coordinate system to the engineering coordinate system to achieve alignment; convert unstructured data into structured fields through a rule engine to support dynamic expansion;

[0122] Finally, capture the graphics and parameters of the substation drawings;

[0123] S4. Use the training dataset constructed in step S2 to train the initial model for capturing the graphics and parameters of the substation drawings constructed in step S3 to obtain a model for capturing the graphics and parameters of the substation drawings;

[0124] S5. Use the model for capturing the graphics and parameters of the substation drawings obtained in step S4 to perform actual capture of the graphics and parameters of the substation drawings.

[0125] As Figure 2 shown in the schematic diagram of the functional modules of the system of the present invention: The system for implementing the method for capturing the graphics and parameters of the substation drawings disclosed in the present invention includes a data acquisition module, a data processing module, a model construction module, a model training module, and a capture module; the data acquisition module, the data processing module, the model construction module, the model training module, and the capture module are connected in series in sequence; the data acquisition module is used to acquire existing substation drawing data information and upload the data information to the data processing module; the data processing module is used to preprocess the acquired data information according to the received data information to construct a training dataset and upload the data information to the model construction module; the model construction module is used to construct an initial model for capturing the graphics and parameters of the substation drawings based on feature parsing, semantic analysis, and a multi-layer perceptron according to the received data information and upload the data information to the model training module; the model training module is used to train the initial model for capturing the graphics and parameters of the substation drawings constructed by using the constructed training dataset according to the received data information to obtain a model for capturing the graphics and parameters of the substation drawings and upload the data information to the capture module; the capture module is used to perform actual capture of the graphics and parameters of the substation drawings by using the obtained model for capturing the graphics and parameters of the substation drawings according to the received data information.

Claims

1. A method for capturing graphics and parameters of substation drawings, comprising the following steps: S1. Obtain the existing substation drawing data information; S2. Preprocess the data information obtained in step S1 to construct a training data set; S3. Based on feature parsing, semantic analysis, and a multi-layer perceptron, construct an initial model for capturing graphics and parameters of substation drawings; S4. Use the training data set constructed in step S2 to train the initial model for capturing graphics and parameters of substation drawings constructed in step S3 to obtain a model for capturing graphics and parameters of substation drawings; S5. Use the model for capturing graphics and parameters of substation drawings obtained in step S4 to perform actual capture of graphics and parameters of substation drawings.

2. The method for capturing the graphics and parameters of the substation drawing according to claim 1, characterized in that The preprocessing of the data information obtained in step S1 described in step S2 specifically includes the following steps: Preprocess the data information obtained in step S1; the preprocessing includes data filtering, data interpolation, and data marking.

3. The method for capturing the graphics and parameters of the substation drawing according to claim 1 or 2, characterized in that The construction of the initial model for capturing graphics and parameters of substation drawings based on feature parsing, semantic analysis, and a multi-layer perceptron described in step S3 specifically includes the following steps: Based on geometric feature parsing and semantic analysis, construct a primitive feature extraction module for synchronously extracting the dual attributes of primitives in substation drawings; Based on a multi-layer perceptron, construct a dynamic coupling module for associating and dynamically coupling graphics and parameters; Based on a gating function, construct a coordinate transformation module for converting the drawing coordinate system to an engineering coordinate system to complete the capture of graphics and parameters of substation drawings.

4. The method for capturing the graphics and parameters of the substation drawing according to claim 3, characterized in that The construction of the primitive feature extraction module based on geometric feature parsing and semantic analysis specifically includes the following steps: The processing process of the constructed primitive feature extraction module includes the following steps: Set the geometric feature matrix as G and the semantic feature matrix as S; Calculate the feature tensor T using the following formula: where Concat() is the concatenation operation; W g is the geometric feature weight matrix; W s is the semantic feature weight matrix; is the outer product operation; Calculate the cross-modal attention weight using the following formula: where A ij is the attention weight between the geometric modality i and the semantic modality j; σ() is the softmax function; T i is the joint feature tensor of the i-th primitive; T j is the joint feature tensor of the j-th primitive; d g is the dimension of the geometric feature vector; d s is the dimension of the semantic feature vector; Perform feature adaptive fusion using the following formula to complete primitive feature extraction: In the formula is the geometric feature after reconstruction of the i-th primitive; n is the number of neighboring graph elements participating in feature fusion; G j is the original geometric feature vector of the j-th graph element; S j is the original semantic feature vector of the j-th primitive; is the feature concatenation operation; ⊙ is the Hadamard product operation.

5. The method for capturing the graphics and parameters of the substation drawing according to claim 4, characterized in that The construction of the dynamic coupling module based on a multi-layer perceptron specifically includes the following steps: The processing process of the constructed dynamic coupling module includes the following steps: Calculate the polar radius using the following formula: Where ρ is the polar radius; (x, y) is the device coordinate; (x0, y0) is the coordinate of the drawing origin; Calculate the polar angle using the following formula: Where θ is the polar angle; Perform non-linear fusion using the following formula: Ρ i = MLP φ (ρ i ||θ i )⊙Sigmoid(W a AA i ) where Ρ i is the dynamic property coupling strength of the i-th device in the polar coordinate space; MLP φ () is the multi-layer perceptron operation with parameter φ; ρ i is the straight-line distance of device i relative to the reference point; θ i is the azimuth angle of device i relative to the reference point; Sigmoid() is the sigmoid function; W a is the device property weight matrix; AA i is the device property vector; Calculate the polarization sensitivity using the following formula: where Γ(P i , P j ) is the polarization sensitivity between P i and P j ; ||P i - P j ||2 is the Euclidean distance between P i and P j , which is used to dynamically adjust the parameter correlation strength between devices; ρ ij is the spatial Euclidean distance between device i and device j; ∈ is the set smoothing factor; Associate and dynamically couple graphics and parameters using the following rules: Associate and dynamically couple the graphics and parameters of adjacent devices with Γ(P i , P j ) ≤ τ, and perform a matching check on the electrical parameters; τ is a set threshold value.

6. The method for capturing the graphics and parameters of the substation drawing according to claim 5, characterized in that Calculate the set threshold τ using the following formula: τ = μ + 3σ Where μ is the expected value of the device parameter; σ is the tolerance factor of the device parameter.

7. The method for capturing the graphics and parameters of the substation drawing according to claim 6, characterized in that The construction of the coordinate transformation module based on a gating function specifically includes the following steps: Set the data feature tensor as F and the storage structure feature tensor as S; Calculate the joint tensor using the following formula: where T is the joint tensor; W f is the importance weight matrix of the feature space; W S is the spatial topological correlation weight matrix; Perform dynamic transformation using the following formula: where M i is the intermediate feature tensor that fuses the feature stream and the structure stream; Softmax() is the softmax function; F i is the feature vector of the device parameters and the spatial topology; W t is the dynamic weight matrix; d t is the dynamic decay factor; Use the following formula as the gating function: where G i is the dynamic gating vector; ΔS i is the structural flow increment tensor; Perform selective fusion using the following formula: In the formula is the gated enhancement feature; S i is a structural flow input feature; Map the drawing coordinate system to the engineering coordinate system for alignment; convert unstructured data to structured fields through a rule engine to support dynamic expansion; Finally, the graphics and parameters of the substation drawings are captured.

8. A system for implementing the method for capturing graphics and parameters of the substation drawing according to any one of claims 1 to 7, characterized in that It includes a data acquisition module, a data processing module, a model construction module, a model training module, and a capture module; the data acquisition module, the data processing module, the model construction module, the model training module, and the capture module are connected in series in sequence; the data acquisition module is used to acquire the existing data information of the substation drawings and upload the data information to the data processing module; The data processing module is used to preprocess the acquired data information according to the received data information to construct a training data set and upload the data information to the model construction module; The model construction module is used to construct an initial model for capturing the graphics and parameters of the substation drawings based on feature parsing, semantic analysis, and a multi-layer perceptron according to the received data information, and upload the data information to the model training module; The model training module is used to train the initial model for capturing the graphics and parameters of the substation drawings constructed by using the constructed training data set according to the received data information to obtain a model for capturing the graphics and parameters of the substation drawings, and upload the data information to the capture module; The capture module is used to actually capture the graphics and parameters of the substation drawings by using the obtained model for capturing the graphics and parameters of the substation drawings according to the received data information.

Images