An electromagnetic simulation modeling method for the connection between GIS and power cables

Abstract

The present invention belongs to the field of simulation modeling technology, and specifically relates to an electromagnetic simulation modeling method for the connection between a GIS and a power cable. The method comprises the following steps: constructing a three-dimensional geometric model in proportion; importing the three-dimensional geometric model into COMSOL to obtain a simulation model; adding magnetic and electric field physical fields to the simulation model and performing meshing; constructing a simulation model containing different knife mark defects; experimentally obtaining the internal electric field strength and magnetic field strength of the GIS terminal when different knife mark defects occur, and comparing them with the simulation results; constructing a data set based on the simulation model; utilizing feature transformation to convert the one-dimensional spatiotemporal signal in the data set into a two-dimensional topological feature image; utilizing a residual network to train and identify and classify the two-dimensional topological feature image, and establishing a network model for determining different main insulation knife mark defects. The present invention achieves efficient and accurate identification of main insulation knife mark defects on GIS cable terminals, thereby improving the safety and stability of the power system.

Description

Technical Field

[0001] The invention belongs to the technical field of simulation modeling, and in particular relates to an electromagnetic simulation modeling method for connecting a GIS with a power cable. Background Art

[0002] Driven by the modern urbanization process and industrial development, power cables are widely used in urban power grids and power supply systems of large and medium-sized enterprises due to their efficient and safe transmission characteristics, becoming a key component of the power transmission and distribution network. However, cable accessories, especially cable terminals, have gradually become the weak link in the cable system due to their complex installation process, material dispersion, electrical stress concentration, and multi-composite interface structure. According to statistics, cable terminal failures account for a considerable proportion of cable failures. These problems not only affect the stable operation of transmission lines, but may also cause serious power accidents, resulting in huge economic losses. At present, there are relatively few studies on the electric field distribution of cable terminals. Existing research mainly focuses on achieving electric field homogenization by optimizing the internal structure of cable terminals, but there are still many problems that need to be solved. For example, phenomena such as stress cone breakdown are particularly prominent in GIS cable terminals.

[0003] During the installation of cable terminals, the metal sheath of the cable body needs to be cut open, which can lead to electric field concentration at the end of the insulation shield. Uneven electric field distribution can cause aging, breakdown, or discharge of internal terminal materials, seriously affecting the insulation performance and service life of the cable terminal. Furthermore, the existing design and manufacturing process of cable terminals still has many deficiencies in terms of material selection and processing, structural optimization, and other aspects, making it difficult for cable terminals to meet the requirements of use under complex operating conditions such as high voltage and high current in actual operation. Therefore, in-depth research on the electric field distribution characteristics of cable terminals and optimization of their internal structure and material properties are of great significance for improving the reliability and safety of cable systems. Summary of the Invention

[0004] In view of the above shortcomings in the existing technology, the purpose of the present invention is to provide an electromagnetic simulation modeling method for the connection between GIS and power cables, which integrates simulation modeling and deep learning, realizes efficient and accurate identification of knife mark defects in the main insulation of GIS cable terminals, and can improve the safety and stability of the power system.

[0005] To achieve the above objectives, the present invention provides an electromagnetic simulation modeling method for connecting GIS and power cables, comprising the following steps:

[0006] S1. Use SolidWorks to build a 3D geometric model based on the actual GIS terminal and cable body;

[0007] S2. Import the 3D geometric model into COMSOL to obtain a simulation model, and add material properties to each component of the simulation model;

[0008] S3. Add magnetic field and electric field physics fields to the simulation model. The range of the magnetic field and electric field physics fields corresponds to all domains of the simulation model, and set the boundary conditions for electromagnetic field analysis.

[0009] S4, meshing the simulation model;

[0010] S5. Add different knife mark defects to the main insulation of the simulation model, build a simulation model containing different knife mark defects, and simulate to obtain the internal electric field strength and magnetic field strength of the GIS terminal;

[0011] S6. Build an experimental platform and set knife mark defects that match the simulation model on GIS terminal samples and cable body samples. Through experiments, obtain the internal electric field strength and magnetic field strength of the GIS terminal when different knife mark defects are present. Compare the results with the simulation results to determine whether the simulation model needs to be adjusted.

[0012] S7. Based on the simulation model verified by the experiment, the simulation results are analyzed to establish a data set of the internal electric field strength and magnetic field strength of the GIS terminal with different main insulation knife mark defects;

[0013] S8, using feature transformation to convert the one-dimensional spatiotemporal signal in the data set into a two-dimensional topological feature image;

[0014] S9. Use the residual network ResNet101 model to train and identify and classify two-dimensional topological feature images, and establish a network model for judging different main insulation knife mark defects.

[0015] As a preferred embodiment of the present invention, in the S1, in the three-dimensional geometric model, the GIS terminal includes a Havers block, a shielding cover, an epoxy sleeve, a retaining ring, a silicone rubber stress cone, reinforced insulation, a cone support, a transition device and a tail pipe, wherein the epoxy sleeve includes an epoxy sleeve high-voltage insert, an epoxy sleeve low-voltage insert and epoxy resin; the cable body includes a conductor core, an insulation layer and a semi-conductive shielding layer, wherein the insulation layer is XLPE, and the semi-conductive shielding layer includes an inner semi-conductive shielding layer and an outer semi-conductive shielding layer, the outer semi-conductive shielding layer is outside the insulation layer, and the inner semi-conductive shielding layer is between the conductor core and the insulation layer.

[0016] As a preferred embodiment of the present invention, in S2, the material of the Haversian block is set to copper, the materials of the shielding cover, epoxy sleeve high-pressure insert, epoxy sleeve low-pressure insert, cone support, transition device and tail pipe are set to aluminum alloy, and the material of the retaining ring is set to polytetrafluoroethylene, and the setting method is selected from the COMSOL material library;

[0017] Experimental tests were conducted on silicone rubber stress cones and reinforced insulation, and their electrical conductivity, relative magnetic permeability, and relative dielectric constant were measured and input into the material properties of COMSOL.

[0018] As a preferred embodiment of the present invention, in the above S3, all domains of the simulation model include the conductor domain, the insulation domain, the shielding domain, the support structure domain, and the boundary domain. The boundary conditions for the electromagnetic field analysis are specifically set as follows: setting the external boundary magnetic insulation and electrical insulation, applying a voltage of 64 kV to the conductor core, and setting the outer semi-conductive shielding layer to be grounded;

[0019] In S4, when performing mesh division, a regular tetrahedron mesh is used for the semi-conductive shielding layer, and the mesh size is selected to be finer; a regular tetrahedron mesh is used for the conductor core, insulation layer, silicone rubber stress cone, and reinforced insulation, and the mesh size is selected to be finer; for the remaining components of the simulation model, a regular tetrahedron mesh is used, and the mesh size is selected to be regular.

[0020] As a preferred embodiment of the present invention, in said S5, the main insulation is the insulation layer, and the added knife scratch defects are micro scratches, medium scratches and severe scratches of different knife scratch lengths and depths;

[0021] For knife scratch defects with a length of less than 1 mm and a depth of less than 0.05 mm, it is considered that there is no scratch; otherwise, it is considered that there is a scratch;

[0022] When scratches exist, scratches with a length greater than or equal to 1 mm and less than 5 mm and a depth greater than or equal to 0.05 mm and less than 0.25 mm are considered minor scratches; scratches with a length greater than or equal to 15 mm and a depth greater than or equal to 0.75 mm are considered severe scratches; and the rest are considered moderate scratches.

[0023] As a preferred embodiment of the present invention, in S6, the experimental platform includes an electromagnetic field measurement device, a data acquisition system, a central processing unit, and an environmental control unit. The GIS terminal sample and the cable body sample to be tested are fixed on the experimental platform. The electromagnetic field measurement device includes a magnetic field intensity meter and an electric field intensity meter for measuring the electromagnetic field distribution of the GIS terminal sample. The data acquisition system includes a wireless signal device, a wired signal device, and current transformers arranged at multiple points inside the GIS terminal sample for collecting current transformer data. The central processing unit is used to receive and process data transmitted from the data acquisition system for analysis and comparison. The environmental control unit is used to control the environmental conditions of the experimental platform.

[0024] The test process is divided into the following steps: sample preparation, environment setting, electromagnetic field testing, data acquisition and data analysis; sample preparation, installing the GIS terminal sample and cable body sample on the experimental platform; environment setting, setting the required environmental conditions through the environmental control unit; electromagnetic field testing, using electromagnetic field measurement equipment to test the GIS terminal sample; data acquisition, collecting current transformer data and electromagnetic field data through the data acquisition system; data analysis, analyzing the current transformer data and electromagnetic field data through the central processing unit, comparing them with the simulation model, and determining whether the simulation model needs to be adjusted.

[0025] As a preferred embodiment of the present invention, the electromagnetic field measurement equipment has a total of 30 measurement points, 10 of which are set at the interface between the conductor core and the inner semi-conductive shielding layer, 10 of which are set at the interface between the inner semi-conductive shielding layer and the insulating layer, and 10 of which are set at the interface between the insulating layer and the silicone rubber stress cone. 30 corresponding measurement points are configured in the simulation model.

[0026] Based on the experimental platform, the electric field strength and magnetic field strength at the interface between the conductor core and the inner semi-conductive shielding layer were tested for different knife mark defects. The experimental results were compared with the corresponding simulation results. When the error was within 5%, the simulation model did not need to be adjusted.

[0027] Based on the experimental platform, the electric field strength and magnetic field strength at the measurement point on the interface between the inner semi-conductive shielding layer and the insulating layer were tested for different knife mark defects. The experimental results were compared with the corresponding simulation results. When the error was within 5%, the simulation model did not need to be adjusted.

[0028] Based on the experimental platform, the electric field strength and magnetic field strength at the interface between the insulating layer and the silicone rubber stress cone were tested for different knife mark defects. The experimental results were compared with the corresponding simulation results. When the error was within 5%, the simulation model did not need to be adjusted.

[0029] When adjustments are needed, the simulation model can be adjusted by correcting the dimensional parameters of components, adjusting the electromagnetic parameters of components, optimizing meshing, and adjusting boundary conditions.

[0030] As a preferred embodiment of the present invention, in the above S7, for the simulation model under different tool mark defects after experimental verification, the corresponding data set is obtained based on the simulation of 30 measurement points, the horizontal axis of the data set is time, and the vertical axis is the electric field strength or magnetic field strength.

[0031] As a preferred embodiment of the present invention, in said S8, the conversion process is to pre-process the data set using the Gram Angular Field (GAF):

[0032] S8.1. Normalize the electric field strength and magnetic field strength at each moment and express them as:

[0033] ;

[0034] Where, It represents the signal amplitude of node i at time t after normalization. The signal amplitude is the electric field intensity or magnetic field intensity, and the node is the measurement point; is the signal amplitude of node i at time t; 、 are the maximum and minimum values ​​of the signal amplitude of node i respectively;

[0035] S8.2. Use polar coordinates to represent the normalized node signal amplitude data:

[0036] ;

[0037] ;

[0038] Where, Represents the angle corresponding to node i at time t in the polar coordinate system; represents the radius corresponding to node i in the polar coordinate system; N is the number of nodes;

[0039] S8.3. Perform trigonometric transformation on the amplitude data of different nodes in the polar coordinate system:

[0040] ;

[0041] ;

[0042] Where, represents the inner product operation; represents the signal amplitude of node j at time t after normalization; G represents the final generated two-dimensional topological feature image; Represents the angle corresponding to node j at time t in the polar coordinate system.

[0043] As a preferred solution of the present invention, in S9, the residual network ResNet101 model includes an input layer, a feature extraction network and a classification output layer, wherein:

[0044] The input size of the input layer is 224×224×2, and the initial feature extraction is performed through a 7×7 convolution kernel, and downsampling is achieved with a 3×3 maximum pooling;

[0045] The feature extraction network consists of 33 Bottleneck blocks, each of which contains 3 layers of convolution and integrates the SE attention module and skip connection;

[0046] Classification output layer: After compressing the feature vector using global average pooling, the defect probability is output through Dropout and Softmax.

[0047] The simulation and algorithm involved in the present invention can be executed by an electronic device, which includes a memory, a processor, and a computer program stored in the memory and run on the processor. The above-mentioned simulation and algorithm calculation are realized by executing the software through the processor.

[0048] The beneficial effects of the present invention are:

[0049] The present invention can realize the rapid and accurate identification of the main insulation knife mark defects of the GIS cable terminal by combining simulation modeling with deep learning technology. By using SolidWorks software to construct a three-dimensional geometric model and importing it into COMSOL for simulation, the electromagnetic field distribution inside the GIS terminal can be accurately simulated. On this basis, a variety of simulation models are constructed by adding different main insulation knife mark defects, and the simulation results are verified and corrected using experimental data to ensure the reliability and accuracy of the model. Furthermore, the one-dimensional spatiotemporal signal is converted into a two-dimensional topological feature image through feature transformation, and then the residual network ResNet101 model is used for training and identification classification to establish a network model that can judge different main insulation knife mark defects. This method not only effectively improves the detection efficiency, but also significantly improves the accuracy of the detection results, providing strong technical support for fault diagnosis and status assessment of GIS terminals.

[0050] The present invention provides solid data support for the establishment of defect diagnosis standards through cross-validation of measured data and simulation results. This method can detect potential main insulation knife mark defects in GIS terminals early. By simulating the electromagnetic field distribution under different main insulation knife mark defects, it provides a theoretical basis for GIS terminal fault location, thereby improving the reliability and safety of GIS terminals. Predicting the trend of abnormal changes in the electromagnetic field makes it possible to take preventive measures in a timely manner, thereby ensuring the safe and stable operation of the power system. In addition, this method can be extended to the diagnosis of insulation defects in other high-voltage equipment, which has important practical significance for power system status inspection and fault warning, and helps to reduce the maintenance cost of the power system and improve the overall operating efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the process of the present invention;

[0052] Figure 2 is a schematic diagram of a simulation model in an embodiment of the present invention;

[0053] Figure 3 Schematic diagram of mesh division of a simulation model in an embodiment of the present invention.

[0054] In the figure: 1. Haver block; 2. Shielding cover; 3. Epoxy sleeve; 4. Retaining ring; 5. Silicone rubber stress cone; 6. Reinforced insulation; 7. Cone support; 8. Transition device; 9. Tail pipe; 10. Conductor core; 11. Insulation layer; 12. Inner semi-conductive shielding layer; 13. Outer semi-conductive shielding layer. DETAILED DESCRIPTION

[0055] The embodiments of the present invention are further described below with reference to the accompanying drawings:

[0056] like Figure 1 As shown, an electromagnetic simulation modeling method for connecting GIS and power cables includes the following steps:

[0057] S1. Use SolidWorks to build a 3D geometric model based on the actual GIS terminal and cable body;

[0058] S2. Import the 3D geometric model into COMSOL to obtain a simulation model, and add material properties to each component of the simulation model;

[0059] S3. Add magnetic field and electric field physics fields to the simulation model. The range of the magnetic field and electric field physics fields corresponds to all domains of the simulation model, and set the boundary conditions for electromagnetic field analysis.

[0060] S4, meshing the simulation model;

[0061] S5. Add different knife mark defects to the main insulation of the simulation model, build a simulation model containing different knife mark defects, and simulate to obtain the internal electric field strength and magnetic field strength of the GIS terminal;

[0062] S6. Build an experimental platform and set knife mark defects that match the simulation model on GIS terminal samples and cable body samples. Through experiments, obtain the internal electric field strength and magnetic field strength of the GIS terminal when different knife mark defects are present. Compare the results with the simulation results to determine whether the simulation model needs to be adjusted.

[0063] S7. Based on the simulation model verified by the experiment, the simulation results are analyzed to establish a data set of the internal electric field strength and magnetic field strength of the GIS terminal with different main insulation knife mark defects;

[0064] S8, using feature transformation to convert the one-dimensional spatiotemporal signal in the data set into a two-dimensional topological feature image;

[0065] S9. Use the residual network ResNet101 model to train and identify and classify two-dimensional topological feature images, and establish a network model for judging different main insulation knife mark defects.

[0066] To obtain simulation results, add a frequency-domain-transient study and solve the problem using the transient solver.

[0067] In S1, in the three-dimensional geometric model, the GIS terminal includes a Haver block 1 (connecting the GIS terminal and the cable body to ensure the sealing and conductivity of the connection), a shielding cover 2 (shielding the electromagnetic field, preventing external electromagnetic interference, and uniformly distributing the electric field), an epoxy sleeve 3 (insulation, mechanical support and sealing), a retaining ring 4 (fixing and supporting internal components to prevent component displacement), a silicone rubber stress cone 5 (controlling the electric field stress distribution and reducing the electric field strength on the surface of the insulation layer 11), reinforced insulation 6 (improving the insulation performance of the cable terminal and enhancing the insulation strength), a cone support 7 (supporting the silicone rubber stress cone 5 to ensure the stability and correct position of the stress cone), and a transition device. The cable body comprises a conductor core 10, an insulating layer 11 and a semi-conductive shielding layer, wherein the insulating layer 11 is XLPE (cross-linked polyethylene), and the semi-conductive shielding layer comprises an inner semi-conductive shielding layer 12 and an outer semi-conductive shielding layer 13, wherein the outer semi-conductive shielding layer 13 is outside the insulating layer 11, and the inner semi-conductive shielding layer 12 is between the conductor core 10 and the insulating layer 11.

[0068] In S2, the material of the Harvard block 1 is set to copper, the materials of the shielding cover 2, the high-voltage insert of the epoxy sleeve 3, the low-voltage insert of the epoxy sleeve 3, the cone support 7, the transition device 8, and the tail pipe 9 are set to aluminum alloy, and the material of the retaining ring 4 is set to polytetrafluoroethylene. The setting method is to select from the COMSOL material library;

[0069] Experimental tests were conducted on the silicone rubber stress cone 5 and the reinforced insulation 6, and their electrical conductivity, relative magnetic permeability, and relative dielectric constant were measured and input into the material properties of COMSOL.

[0070] In S3, all domains of the simulation model include the conductor domain (including the conductive areas such as the conductor core 10 and the conductive parts of the transition device 8), the insulation domain (including the insulating parts such as the insulation layer 11, the silicone rubber stress cone 5, the reinforced insulation 6, and the epoxy sleeve 3), the shielding domain (including the inner semi-conductive shielding layer 12 and the outer semi-conductive shielding layer 13), the support structure domain (the area where the support structures such as the retaining ring 4 and the cone support 7 are located), and the boundary domain (the outer boundary of the model and the interface with the external environment). The boundary conditions for the electromagnetic field analysis are specifically set as follows: setting magnetic insulation and electrical insulation for the outer boundary, applying a voltage of 64 kV to the conductor core 10 (simulating the voltage conditions under actual conditions), and setting the outer semi-conductive shielding layer 13 to ground (in line with the actual grounding method);

[0071] In S4, when performing mesh division, a regular tetrahedron mesh is used for the semi-conductive shielding layer, and the mesh size is selected to be finer; a regular tetrahedron mesh is used for the conductor core 10, the insulating layer 11, the silicone rubber stress cone 5, and the reinforced insulation 6, and the mesh size is selected to be finer; for the remaining components of the simulation model, a regular tetrahedron mesh is used, and the mesh size is selected to be regular.

[0072] The knife marks on the insulating layer 11 refer to scratches or damages caused on the surface of the insulating layer 11 of the cable due to improper operation or other reasons during the production, installation, operation and maintenance of the cable.

[0073] In S5, the main insulation is the insulation layer 11, and the added knife scratch defects are small scratches, medium scratches, and severe scratches of different knife scratch lengths and depths;

[0074] If the length of the knife mark is less than 1mm and the depth of the knife mark is less than 0.05mm, it is considered that there is no scratch (normal); otherwise, it is considered that there is a scratch;

[0075] When scratches exist, scratches with a length greater than or equal to 1 mm and less than 5 mm and a depth greater than or equal to 0.05 mm and less than 0.25 mm are considered minor scratches; scratches with a length greater than or equal to 15 mm and a depth greater than or equal to 0.75 mm are considered severe scratches; and the rest are considered moderate scratches.

[0076] In S6, the experimental platform includes electromagnetic field measurement equipment, a data acquisition system, a central processing unit, and an environmental control unit. The GIS terminal sample and cable body sample to be tested are fixed on the experimental platform (to ensure stability during electromagnetic field testing). The electromagnetic field measurement equipment includes a magnetic field intensity meter and an electric field intensity meter, which are used to measure the electromagnetic field distribution of the GIS terminal sample. The data acquisition system includes a wireless signal device, a wired signal device, and current transformers arranged at multiple points inside the GIS terminal sample to collect current transformer data. The central processing unit is used to receive and process data transmitted from the data acquisition system for analysis and comparison. The environmental control unit is used to control the environmental conditions of the experimental platform.

[0077] The test process is divided into the following steps: sample preparation, environment setting, electromagnetic field testing, data acquisition and data analysis; sample preparation, installing the GIS terminal sample and cable body sample on the experimental platform; environment setting, setting the required environmental conditions through the environmental control unit; electromagnetic field testing, using electromagnetic field measurement equipment to test the GIS terminal sample; data acquisition, collecting current transformer data and electromagnetic field data through the data acquisition system; data analysis, analyzing the current transformer data and electromagnetic field data through the central processing unit, comparing them with the simulation model, and determining whether the simulation model needs to be adjusted.

[0078] The electromagnetic field measurement equipment has a total of 30 measurement points, 10 of which are set at the interface between the conductor core 10 and the inner semi-conductive shielding layer 12, 10 of which are set at the interface between the inner semi-conductive shielding layer 12 and the insulating layer 11, and 10 of which are set at the interface between the insulating layer 11 and the silicone rubber stress cone 5. 30 corresponding measurement points are configured in the simulation model.

[0079] Based on the experimental platform, the electric field strength and magnetic field strength at the measurement point on the interface between the conductor core 10 and the inner semi-conductive shielding layer 12 were tested for different knife mark defects. The experimental results were compared with the corresponding simulation results. When the error was within 5%, the simulation model did not need to be adjusted.

[0080] Based on the experimental platform, the electric field strength and magnetic field strength at the interface between the inner semi-conductive shielding layer 12 and the insulating layer 11 were tested for different knife mark defects. The experimental results were compared with the corresponding simulation results. When the error was within 5%, the simulation model did not need to be adjusted.

[0081] Based on the experimental platform, the electric field strength and magnetic field strength at the interface between the insulating layer 11 and the silicone rubber stress cone 5 were tested for different knife mark defects. The experimental results were compared with the corresponding simulation results. When the error was within 5%, the simulation model did not need to be adjusted.

[0082] When adjustments are needed, the simulation model can be adjusted by correcting the dimensional parameters of components, adjusting the electromagnetic parameters of components, optimizing meshing, and adjusting boundary conditions.

[0083] In S7, for the simulation models under different tool mark defects after experimental verification, the corresponding data sets are obtained based on the simulation of 30 measurement points. The horizontal axis of the data set is time, and the vertical axis is the electric field intensity or magnetic field intensity.

[0084] In S8, the conversion process is to preprocess the data set using Gram Angular Field (GAF) to enhance the data classification features:

[0085] S8.1. Normalize the electric field strength and magnetic field strength at each moment and express them as:

[0086] ;

[0087] Where, It represents the signal amplitude of node i at time t after normalization. The signal amplitude is the electric field intensity or magnetic field intensity, and the node is the measurement point; is the signal amplitude of node i at time t; 、 are the maximum and minimum values ​​of the signal amplitude of node i respectively;

[0088] S8.2. Use polar coordinates to represent the normalized node signal amplitude data:

[0089] ;

[0090] ;

[0091] Where, Represents the angle corresponding to node i at time t in the polar coordinate system; represents the radius corresponding to node i in the polar coordinate system; N is the number of nodes, that is, N=30;

[0092] S8.3. Perform trigonometric transformation on the amplitude data of different nodes in the polar coordinate system:

[0093] ;

[0094] ;

[0095] Where, represents the inner product operation; represents the signal amplitude of node j at time t after normalization; G represents the final generated two-dimensional topological feature image; Represents the angle corresponding to node j at time t in the polar coordinate system.

[0096] In S9, the residual network ResNet101 model includes an input layer, a feature extraction network, and a classification output layer, where:

[0097] The input layer uses a dual-channel Gramian Angular Field (GAF) feature. Channel 1 is the GASF matrix of the electric field signal (corresponding to the electric field intensity G), and channel 2 is the GADF matrix of the magnetic field signal (corresponding to the magnetic field intensity G). The input size is 224 × 224 × 2. Initial feature extraction is performed using a 7 × 7 convolution kernel (stride = 2), and downsampling is achieved with a 3 × 3 maximum pooling (stride = 2).

[0098] The feature extraction network consists of 33 Bottleneck blocks (divided into 4 levels: 3-4-23-3). Each Bottleneck block contains 3 layers of convolution (1×1 dimension reduction → 3×3 convolution → 1×1 dimension increase), and integrates the SE attention module (compression ratio 16) and skip connections.

[0099] Classification output layer: After compressing the feature vector using global average pooling, the defect probability (normal, minor scratches, moderate scratches, severe scratches) is output through Dropout and Softmax.

[0100] The training phase uses mixed precision training (FP16+FP32) and a phased transfer learning strategy (freeze the shallow layer for 10 rounds first), and the loss function uses Focal Loss ( ) Solve the problem of class imbalance.

[0101] After completing the network model for judging different main insulation knife mark defects, the field inspection data of the GIS cable terminal to be inspected (GIS terminal and cable body) can be input. This data needs to undergo the same preprocessing steps as in the simulation modeling process, including operations such as feature transformation, and is converted into a format that the model can recognize. The final output is whether the cable terminal to be inspected has a main insulation knife mark defect, as well as the type of defect. In addition, combined with simulation analysis and experimental data, the specific location of the defect in the GIS cable terminal can be pointed out, providing accurate guidance for on-site maintenance.

Claims

1. An electromagnetic simulation modeling method for connecting GIS and power cables, characterized in that The following steps are involved: S1. Use SolidWorks to build a 3D geometric model based on the actual GIS terminal and cable body; S2. Import the 3D geometric model into COMSOL to obtain a simulation model, and add material properties to each component of the simulation model; S3. Add magnetic field and electric field physics fields to the simulation model. The range of the magnetic field and electric field physics fields corresponds to all domains of the simulation model, and set the boundary conditions for electromagnetic field analysis. S4, meshing the simulation model; S5. Add different knife mark defects to the main insulation of the simulation model, build a simulation model containing different knife mark defects, and simulate to obtain the internal electric field strength and magnetic field strength of the GIS terminal; S6. Build an experimental platform and set knife mark defects that match the simulation model on GIS terminal samples and cable body samples. Through experiments, obtain the internal electric field strength and magnetic field strength of the GIS terminal when different knife mark defects are present. Compare the results with the simulation results to determine whether the simulation model needs to be adjusted. The experimental platform includes electromagnetic field measurement equipment, a data acquisition system, a central processing unit, and an environmental control unit. The GIS terminal sample and cable body sample to be tested are fixed on the experimental platform. The electromagnetic field measurement equipment includes a magnetic field intensity meter and an electric field intensity meter, which are used to measure the electromagnetic field distribution of the GIS terminal sample. The data acquisition system includes a wireless signal device, a wired signal device, and current transformers arranged at multiple points inside the GIS terminal sample to collect current transformer data. The central processing unit is used to receive and process data transmitted from the data acquisition system for analysis and comparison. The environmental control unit is used to control the environmental conditions of the experimental platform. The test process is divided into the following steps: sample preparation, environment setting, electromagnetic field testing, data acquisition and data analysis; Sample preparation: GIS terminal samples and cable body samples are installed on the experimental platform; Environmental setting: setting the required environmental conditions through the environmental control unit; Electromagnetic field testing: Using electromagnetic field measurement equipment to test GIS terminal samples; data acquisition: Using a data acquisition system to collect current transformer data and electromagnetic field data; data analysis: Using a central processing unit to analyze current transformer data and electromagnetic field data, and compare them with the simulation model to determine whether the simulation model needs to be adjusted; S7. Based on the simulation model verified by the experiment, the simulation results are analyzed to establish a data set of the internal electric field strength and magnetic field strength of the GIS terminal with different main insulation knife mark defects; S8, using feature transformation to convert the one-dimensional spatiotemporal signal in the data set into a two-dimensional topological feature image; S9. Use the residual network ResNet101 model to train and identify and classify two-dimensional topological feature images, and establish a network model for judging different main insulation knife mark defects.

2. The electromagnetic simulation modeling method for connecting GIS and power cables according to claim 1, characterized in that: In the S1, in the three-dimensional geometric model, the GIS terminal includes a Haver block, a shielding cover, an epoxy sleeve, a retaining ring, a silicone rubber stress cone, reinforced insulation, a cone support, a transition device and a tail pipe, wherein the epoxy sleeve includes an epoxy sleeve high-voltage insert, an epoxy sleeve low-voltage insert and epoxy resin; the cable body includes a conductor core, an insulation layer and a semi-conductive shielding layer, wherein the insulation layer is XLPE, and the semi-conductive shielding layer includes an inner semi-conductive shielding layer and an outer semi-conductive shielding layer, the outer semi-conductive shielding layer is outside the insulation layer, and the inner semi-conductive shielding layer is between the conductor core and the insulation layer.

3. The electromagnetic simulation modeling method for connecting GIS and power cables according to claim 2, characterized in that: In the aforementioned S2, the material of the Haver block is set to copper, the materials of the shielding cover, epoxy sleeve high-pressure insert, epoxy sleeve low-pressure insert, cone support, transition device, and tail pipe are set to aluminum alloy, and the material of the retaining ring is set to polytetrafluoroethylene, and the setting method is selected in the COMSOL material library; Experimental tests were conducted on silicone rubber stress cones and reinforced insulation, and their electrical conductivity, relative magnetic permeability, and relative dielectric constant were measured and input into the material properties of COMSOL.

4. The electromagnetic simulation modeling method for connecting GIS and power cables according to claim 2, characterized in that: In the above S3, all domains of the simulation model include the conductor domain, the insulation domain, the shielding domain, the support structure domain, and the boundary domain. The boundary conditions for the electromagnetic field analysis are specifically set as follows: setting the external boundary magnetic insulation and electrical insulation, applying a voltage of 64 kV to the conductor core, and setting the outer semi-conductive shielding layer to be grounded; In S4, when performing mesh division, a regular tetrahedron mesh is used for the semi-conductive shielding layer, and the mesh size is selected to be finer; a regular tetrahedron mesh is used for the conductor core, insulation layer, silicone rubber stress cone, and reinforced insulation, and the mesh size is selected to be finer; for the remaining components of the simulation model, a regular tetrahedron mesh is used, and the mesh size is selected to be regular.

5. The electromagnetic simulation modeling method for connecting GIS and power cables according to claim 2, characterized in that: In the S5, the main insulation is the insulation layer, and the added knife scratch defects are small scratches, medium scratches and severe scratches of different knife scratch lengths and depths; For knife scratch defects with a length of less than 1 mm and a depth of less than 0.05 mm, it is considered that there is no scratch; otherwise, it is considered that there is a scratch; When scratches exist, scratches with a length greater than or equal to 1 mm and less than 5 mm and a depth greater than or equal to 0.05 mm and less than 0.25 mm are considered minor scratches; scratches with a length greater than or equal to 15 mm and a depth greater than or equal to 0.75 mm are considered severe scratches; and the rest are considered moderate scratches.

6. The electromagnetic simulation modeling method for connecting GIS and power cables according to claim 1, characterized in that: The electromagnetic field measurement equipment has a total of 30 measurement points, 10 of which are set at the interface between the conductor core and the inner semi-conductive shielding layer, 10 at the interface between the inner semi-conductive shielding layer and the insulating layer, and 10 at the interface between the insulating layer and the silicone rubber stress cone. These 30 measurement points are configured in the simulation model accordingly. Based on the experimental platform, the electric field strength and magnetic field strength at the interface between the conductor core and the inner semi-conductive shielding layer were tested for different knife mark defects. The experimental results were compared with the corresponding simulation results. When the error was within 5%, the simulation model did not need to be adjusted. Based on the experimental platform, the electric field strength and magnetic field strength at the measurement point on the interface between the inner semi-conductive shielding layer and the insulating layer were tested for different knife mark defects. The experimental results were compared with the corresponding simulation results. When the error was within 5%, the simulation model did not need to be adjusted. Based on the experimental platform, the electric field strength and magnetic field strength at the interface between the insulating layer and the silicone rubber stress cone were tested for different knife mark defects. The experimental results were compared with the corresponding simulation results. When the error was within 5%, the simulation model did not need to be adjusted. When adjustments are needed, the simulation model can be adjusted by correcting the dimensional parameters of components, adjusting the electromagnetic parameters of components, optimizing meshing, and adjusting boundary conditions.

7. The electromagnetic simulation modeling method for connecting GIS and power cables according to claim 6, characterized in that: In the above-mentioned S7, for the simulation model under different tool mark defects after experimental verification, the corresponding data set is obtained based on the simulation of 30 measurement points, where the horizontal axis of the data set is time and the vertical axis is the electric field strength or magnetic field strength.

8. The electromagnetic simulation modeling method for connecting GIS and power cables according to claim 1, characterized in that: In the above S8, the conversion process is to pre-process the data set using the Gram Angular Field (GAF): S8.

1. Normalize the electric field strength and magnetic field strength at each moment and express them as: ； Where, It represents the signal amplitude of node i at time t after normalization. The signal amplitude is the electric field intensity or magnetic field intensity, and the node is the measurement point; is the signal amplitude of node i at time t; 、 are the maximum and minimum values ​​of the signal amplitude of node i respectively; S8.

2. Use polar coordinates to represent the normalized node signal amplitude data: ； ； Where, Represents the angle corresponding to node i at time t in the polar coordinate system; represents the radius corresponding to node i in the polar coordinate system; N is the number of nodes; S8.

3. Perform trigonometric transformation on the amplitude data of different nodes in the polar coordinate system: ； ； Where, represents the inner product operation; represents the signal amplitude of node j at time t after normalization; G represents the final generated two-dimensional topological feature image; Represents the angle corresponding to node j at time t in the polar coordinate system.

9. The electromagnetic simulation modeling method for connecting GIS and power cables according to claim 1, characterized in that: In the above S9, the residual network ResNet101 model includes an input layer, a feature extraction network, and a classification output layer, wherein: The input size of the input layer is 224×224×2, and the initial feature extraction is performed through a 7×7 convolution kernel, and downsampling is achieved with a 3×3 maximum pooling; The feature extraction network consists of 33 Bottleneck blocks, each of which contains 3 layers of convolution and integrates the SE attention module and skip connection; Classification output layer: After compressing the feature vector using global average pooling, the defect probability is output through Dropout and Softmax.

Images