A switch cabinet wrench contact fault diagnosis method, system, device and medium

Abstract

The present application belongs to the technical field of high-voltage circuit breaker component diagnosis, and specifically discloses a switch cabinet plum blossom contact fault diagnosis method, system, device and medium. The present application adopts digital twinning technology, establishes a three-dimensional physical model and a mathematical model, and constructs a digital twinning model according to the three-dimensional physical model and the mathematical model, simulates different working conditions, and obtains three-dimensional feature data. The three-dimensional feature data is processed, a fault model based on a convolutional neural network algorithm is constructed, and training is performed. Actual fault data is input into the trained fault model based on the convolutional neural network algorithm, and a fault diagnosis result is output, realizing fault diagnosis of the plum blossom contact. High-precision simulation data is obtained, real-time dynamic updating of the digital twinning body is realized, including updating of data and models, so that the digital twinning body is closer to a physical entity, error precision is smaller, and recognition accuracy is higher.

Description

Technical Field

[0001] This invention belongs to the field of high-voltage circuit breaker component diagnostic technology, specifically relating to a method, system, equipment, and medium for diagnosing faults in switchgear plum blossom contacts. Background Technology

[0002] High-voltage switchgear is numerous and widely used, playing a crucial role in the safe operation of the power grid. The pentagonal contact is one of the core components of high-voltage control devices, widely used to connect high-voltage control switches and the switchgear cabinet. It features multiple contacts, low resistance, and high conductivity. This device has become increasingly prevalent with the widespread use of high-voltage switchgear. However, faults and accidents caused by contact heat within the control circuit breaker cabinet are becoming more frequent. During operation, manufacturing and installation defects in high-voltage switchgear can lead to poor contact and overheating. If overheating is not detected and addressed promptly, it can result in failure or even an accident. Therefore, research on fault diagnosis based on the pentagonal contacts of switchgear is of great significance for improving the operational reliability of switchgear and the stability of power distribution networks.

[0003] Overheating of the plum blossom contact is a typical fault in high-voltage switchgear and a major hidden danger affecting the stable operation of the power system. Currently, research on plum blossom contacts is limited, mainly focusing on experimental measurement and simulation. Experimental measurement primarily employs improved temperature monitoring and pressure testing using measuring sensors, which suffers from high costs and expenses. Simulation studies mainly investigate the internal temperature distribution of high-current switchgear through electromagnetic, heat transfer, and convection coupling simulations. This method is highly convenient but has limitations in testing accuracy. Summary of the Invention

[0004] In order to overcome the shortcomings of the prior art, the present invention aims to provide a method, system, equipment and medium for diagnosing faults of switchgear plum blossom contacts, so as to solve the problems of high research cost and limited testing accuracy for plum blossom contact faults.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] In a first aspect, the present invention provides a method for diagnosing faults in the perforated contacts of a switchgear, comprising:

[0007] S1: Establish a physical three-dimensional model and a mathematical model. The physical three-dimensional model is specifically a three-dimensional structural model of the plum blossom contactor, and the mathematical model is specifically a temperature-magnetic field coupling mathematical model.

[0008] S2: Based on the three-dimensional structural model of the plum blossom contact and the temperature-magnetic field coupling mathematical model, a digital twin model is constructed to simulate different working conditions and obtain three-dimensional feature data;

[0009] S3: Transform and process the three-dimensional feature data, train the convolutional neural network algorithm, and obtain a trained fault model based on the convolutional neural network algorithm;

[0010] S4: Input the actual fault data into the trained fault model based on the convolutional neural network algorithm, and output the fault diagnosis results.

[0011] Furthermore, the method for establishing the three-dimensional structural model of the plum blossom contact is as follows:

[0012] Finite element modeling was performed using ANSYS software, and a three-dimensional structural model of the plum blossom contact was constructed based on the structural parameters of the plum blossom contact.

[0013] Furthermore, the method for establishing the temperature-magnetic field coupling mathematical model is as follows:

[0014] A temperature-magnetic field coupling mathematical model was established to calculate the heat generation rate of the plum blossom contactor;

[0015] The heat conduction equation for the plum blossom contact is:

[0016]

[0017] Where: ρ is the conductor density, with units of g / cm³. 3 C p λ is the specific heat capacity of the conductor, in J / (kg·K); T is the temperature, in °C; t is the simulation time, in s; λ is the thermal conductivity of the conductor, in W / (m·K); Q is the internal heat source.

[0018] The convective heat transfer coefficient α is:

[0019]

[0020] Where: a is the convective heat transfer coefficient, with units of W / (m²). 2 ·K); λ is the thermal conductivity of the conductor, in W / (m·K); L is the surface characteristic scale, in m; ΔT is the temperature difference between the finger touching the air, in °C.

[0021] Radiant surface heat transfer coefficient α r for:

[0022] α r =δ b T 4

[0023] Where: α r It is the heat transfer coefficient of the radiating surface, with units of W / (m²). 2 ·K); ε is the equivalent emissivity of the stylus contact on the conductor rod surface; δ b is Boltzmann's constant, in J / K; T is temperature, in °C.

[0024] The heat generation rate of the conductor is as follows:

[0025]

[0026] Where: Q is the heat production rate; the unit is W / m³. 3 I is current, measured in amperes (A); R is resistance, measured in ohms (Ω); and V is volume, measured in cubic meters (m³). 3 .

[0027] Furthermore, the method for constructing the digital twin model is as follows:

[0028] Electro-thermal structural coupling simulation was carried out using the electromagnetic simulation module, thermal simulation module, and structural field simulation module of ANSYS software. A digital twin model was constructed based on the three-dimensional structural model of the plum blossom contact and the temperature-magnetic field coupling mathematical model.

[0029] Furthermore, the simulation of different working conditions to obtain three-dimensional feature data specifically includes:

[0030] We selected common fault types such as loose contacts and increased contact resistance to perform electrothermal structure coupled finite element simulation of the temperature of the switch station contacts; based on the calculation results of multiphysics field coupling analysis, we obtained the temperature distribution of the plum blossom contacts under rated current and short-circuit current under different operating conditions, and obtained three-dimensional feature data.

[0031] Furthermore, the three-dimensional feature data transformation process specifically includes:

[0032] Data transformation: Convert 3D temperature data into 2D temperature data; convert 3D temperature data into a 2D heat map, project it onto the plane containing the axis, divide the 3D point cloud into the plane containing the axis, erase the height information of the point cloud based on the plane, and project it onto the plane to obtain a point matrix on a 2D plane.

[0033] Data completion: Construct a function with a custom resolution to classify the points in the matrix into each matrix cell based on the resolution as two criteria. Then, take the average temperature of all points in each cell as the temperature of the matrix element to obtain a pixelated heatmap matrix. The specific calculation formula is as follows:

[0034] Input n interpolation nodes: x1 <x2<…<x n The corresponding function values ​​are y1, y2, ..., y n The interpolation point to be determined is x0;

[0035] Calculate h j =x j+1 -x j (j = 1, 2, ..., n-1)

[0036] calculate

[0037] calculate

[0038] Solve the system of equations:

[0039] Substitute the values ​​into the interpolation function to solve for the interpolation:

[0040]

[0041] The spline interpolation function is calculated from the first effective value point to the last effective value point to complete the temperature of an entire row. Then, spline interpolation is performed on each column to complete the temperature, and finally, a complete heat map matrix image is obtained.

[0042] Furthermore, the training of the convolutional neural network algorithm specifically includes:

[0043] Convolutional neural networks (CNNs) are chosen for anomaly detection. The InceptionNet module is set up to encapsulate multiple convolutional kernels of different sizes to learn image features simultaneously. Four convolutional paths are constructed using 1x1, 3x3, 5x5, and pooled 1x1 kernels, and are encapsulated within the InceptionNet module for use by the CNN. The CNN then performs convolution and pooling operations to expand the image information into a final fully connected layer for prediction and classification.

[0044] The optimizer used is the Adam optimizer;

[0045] The simulated temperature three-dimensional feature data is processed into a heat map matrix and then fed into a convolutional neural network for learning. The convolutional neural network is then trained using normal operation data and fault data.

[0046] Finally, a well-trained fault model based on a convolutional neural network algorithm is obtained.

[0047] Secondly, the present invention provides a fault diagnosis system for switchgear Phillips-shaped contacts, comprising:

[0048] The physical 3D model and mathematical model building module is used to build physical 3D models and mathematical models. The physical 3D model is specifically a 3D structural model of the plum blossom contactor, and the mathematical model is specifically a temperature-magnetic field coupling mathematical model.

[0049] The digital twin model establishment and 3D feature data acquisition module is used to construct a digital twin model based on the 3D structural model of the plum blossom contact and the temperature-magnetic field coupling mathematical model, simulate different working conditions, and obtain 3D feature data.

[0050] The fault model building module based on the convolutional neural network algorithm is used to transform and process three-dimensional feature data, train the convolutional neural network algorithm, and obtain a trained fault model based on the convolutional neural network algorithm.

[0051] The fault diagnosis result output module is used to input actual fault data into a trained fault model based on a convolutional neural network algorithm and output fault diagnosis results.

[0052] Thirdly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a switch cabinet Phillips-shaped contact fault diagnosis method as described in any one of the above-mentioned methods.

[0053] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a method for diagnosing faults in a switchgear sprite contact as described above.

[0054] The present invention has at least the following beneficial effects:

[0055] 1. This invention employs digital twin technology to establish a three-dimensional physical model and a mathematical model, and constructs a digital twin model based on the three-dimensional physical model and mathematical model. Simulations are performed under different working conditions to obtain three-dimensional feature data. The three-dimensional feature data is then transformed and processed to construct a fault model based on a convolutional neural network algorithm, which is then trained. Actual fault data is input into the trained fault model based on the convolutional neural network algorithm, and fault diagnosis results are output, achieving fault diagnosis of the plum blossom contact point. High-precision simulation data is obtained, enabling real-time dynamic updates of the digital twin, including updates to the data and model, making it closer to the physical entity with smaller errors and higher recognition accuracy.

[0056] 2. Based on the obtained three-dimensional point cloud data, the present invention performs two-dimensional heat map data conversion and performs interpolation completion processing on the converted empty pixel data, which greatly improves the recognition accuracy of the fault diagnosis model.

[0057] 3. Based on a large amount of high-precision fault simulation data, this invention constructs a convolutional neural network algorithm using the Adam optimizer for fault diagnosis, which can eliminate the problem of overfitting due to excessively large pixels in the recognition image. Fault diagnosis is achieved by embedding intelligent modules and other means. Attached Figure Description

[0058] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0059] Figure 1 A schematic diagram of a fault diagnosis method for sprite contacts in a switchgear;

[0060] Figure 2 A three-dimensional structural model of the plum blossom contactor;

[0061] Figure 3 This is a schematic diagram of the 3D point cloud distribution;

[0062] Figure 4 This is a two-dimensional detail map of the point cloud.

[0063] Figure 5 Complete the image for the heatmap;

[0064] Figure 6 A schematic diagram of the InncepNet module;

[0065] Figure 7 A schematic diagram illustrating the construction of a CNN model;

[0066] Figure 8 Here is a pseudocode diagram of the Adam algorithm;

[0067] Figure 9 This is a schematic diagram of a switchgear plum blossom contact fault diagnosis system module. Detailed Implementation

[0068] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0069] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0070] Example 1

[0071] like Figure 1 As shown, a method for diagnosing faults in the perforated contacts of a switchgear includes:

[0072] S1: Establish a physical three-dimensional model and a mathematical model. The physical three-dimensional model is specifically a three-dimensional structural model of the plum blossom contactor, and the mathematical model is specifically a temperature-magnetic field coupling mathematical model.

[0073] S11: Finite element modeling is performed using ANSYS software. A three-dimensional structural model of the plum blossom contact is constructed based on the structural parameters of the plum blossom contact. Each plum blossom contact consists of 30 contact points. The Autodesk 3ds Max software platform is used to recreate the model one-to-one based on the CAD drawings or obj format model. Figure 2 This is a three-dimensional structural model of the plum blossom contactor.

[0074] S12: Establish a temperature-magnetic field coupling mathematical model to calculate the heat generation rate of the plum blossom contactor;

[0075] According to the first law of thermodynamics, the heat conduction equation for the plum blossom contact is:

[0076]

[0077] Where: ρ is the conductor density, with units of g / cm³. 3 C p λ is the specific heat capacity of the conductor, in J / (kg·K); T is the temperature, in °C; t is the simulation time, in s; λ is the thermal conductivity of the conductor, in W / (m·K); Q is the internal heat source.

[0078] Ignoring airflow within both moving and stationary contacts, this region considers only internal heat conduction, while external heat dissipation is assessed through convection and radiation. Therefore, the convective and radiative heat transfer coefficients need to be characterized based on a third heat transfer boundary condition. The convective heat transfer coefficient α is:

[0079]

[0080] Where: a is the convective heat transfer coefficient, with units of W / (m²). 2 ·K); λ is the thermal conductivity of the conductor, in W / (m·K); L is the surface characteristic scale, in m; ΔT is the temperature difference between the finger touching the air, in °C; A * A is a parameter value, without units. * The value depends only on the temperature and fluid type.

[0081] The heat transfer coefficient α of the radiating surface is calculated according to the Stefan-Boltzmann law. r for:

[0082] α r =δ b T 4

[0083] Where: α r It is the heat transfer coefficient of the radiating surface, with units of W / (m²). 2·K); ε is the equivalent emissivity of the stylus contact on the conductor rod surface, a ratio representing the ratio of the conductor's radiation intensity to that of a blackbody, and has no dimension; δ b is Boltzmann's constant, in J / K; T is temperature, in °C.

[0084] The heat generation rate of the conductor is as follows:

[0085]

[0086] Where: Q is the heat production rate; the unit is W / m³. 3 I is current, measured in amperes (A); R is resistance, measured in ohms (Ω); and V is volume, measured in cubic meters (m³). 3 .

[0087] S2: Based on the three-dimensional structural model of the plum blossom contact and the temperature-magnetic field coupling mathematical model, a digital twin model is constructed to simulate different working conditions and obtain three-dimensional feature data;

[0088] S21: Apply the electromagnetic simulation module, thermal simulation module and structural field simulation module of ANSYS software to carry out electro-thermal structural coupling simulation. Construct a digital twin model based on the three-dimensional structural model of the plum blossom contact and the temperature-magnetic field coupling mathematical model; the array twin model is the plum blossom contact simulation model.

[0089] S22: Common fault types such as loose contacts and increased contact resistance were selected to perform electrothermal structure coupled finite element simulation of the temperature of the switch station contacts; based on the calculation results of multiphysics field coupling analysis, the temperature distribution of the plum blossom contacts under rated current and short-circuit current under different operating conditions was obtained, and three-dimensional feature data were obtained.

[0090] S3: Transform and process the three-dimensional feature data, train the convolutional neural network algorithm, and obtain a trained fault model based on the convolutional neural network algorithm;

[0091] S31: Three-dimensional feature data transformation and processing

[0092] Data transformation: Converting 3D temperature data into 2D temperature data; since the original sample is 3D point cloud data of temperature, it is necessary to convert the 3D temperature data into a 2D thermogram. The 3D point cloud obtained from the through-wall sleeve temperature simulation is reproduced by Matlab, and the 3D point cloud distribution is as follows. Figure 3 As shown. To construct the captured two-dimensional temperature distribution, the three-dimensional point cloud needs to be projected onto a two-dimensional plane. Considering that the vertical axis provides the most temperature information, the projection is onto the plane containing the axis. The three-dimensional point cloud is divided along the plane containing the axis, and then the height information of the point cloud relative to the plane is erased. The projection onto the plane yields a point lattice on a two-dimensional plane, with the two-dimensional detail distribution as shown. Figure 4 As shown, the value of each point is the temperature information.

[0093] Data Completion: To address the issue of excessive sparseness when projecting onto a 2D point matrix, a function with customizable resolution is constructed. This function categorizes the point matrix into individual cells based on resolution, and then takes the average temperature of all points in each cell as the temperature of the matrix element, resulting in a pixelated heatmap matrix. While the heatmap can be generated with high flexibility through manual resolution selection, excessively high resolution leads to a sparse matrix with many matrix elements lacking temperature values ​​(empty pixels). These empty pixels need to be filled in by surrounding pixels. We use spline interpolation to fill in these empty pixels, obtaining the effective temperature points for each row of the matrix for interpolation. The specific calculation formula is as follows:

[0094] Input n interpolation nodes: x1 <x2<…<x n The corresponding function values ​​are y1, y2, ..., y n The interpolation point to be determined is x0;

[0095] Calculate h j =x j+1 -x j (j = 1, 2, ..., n-1)

[0096] calculate

[0097] calculate

[0098] Solve the system of equations:

[0099] Substitute the values ​​into the interpolation function to solve for the interpolation:

[0100]

[0101] Calculate the spline interpolation function from the first valid value point to the last valid value point, then directly fill in the temperature for an entire row. Repeat this process for each column, finally obtaining the complete heatmap matrix image. The completed image is shown below. Figure 5 As shown.

[0102] S32: Fault Model Training Based on Convolutional Neural Network Algorithm

[0103] Convolutional neural networks (CNNs) were chosen for anomaly detection. To automatically find the optimal combination of convolutions, the InceptionNet module was implemented. This module encapsulates multiple convolutional kernels of different sizes to simultaneously learn image features. Four convolutional paths are constructed using 1x1, 3x3, 5x5, and pooled 1x1 kernels, and these paths are encapsulated within the InceptionNet module for use by the CNN. The IncepNet module is shown below. Figure 6As shown. The convolutional neural network (CNN) unfolds the image information into a final fully connected layer through operations such as convolution and pooling, and then performs prediction and classification. The CNN model is constructed as follows... Figure 7 As shown.

[0104] Optimizer selection for training a convolutional neural network requires defining the loss function and the network's hyperparameters. Since this network's function is image classification, cross-entropy is used as the loss function for updating hyperparameters. Optimizers are based on stochastic gradient descent (SGD) and further optimization algorithms incorporating momentum. We chose the Adam optimizer, which combines the advantages of the adaptive gradient algorithm AdaGrad and the exponential moving average optimization algorithm RMSPro, comprehensively considering the first and second moment estimates of the gradient to calculate the update step size. The pseudocode for the Adam algorithm is as follows: Figure 8 As shown.

[0105] The simulated temperature 3D feature data was processed into a heat map matrix and then fed into a convolutional neural network for learning. The convolutional neural network was then trained using normal operation data and fault data. The initial learning rate of the Adam optimizer was set to 0.0001, and the number of training epochs (EPOCH) was 100. After testing, the final CNN model achieved a recognition accuracy of over 90%.

[0106] Finally, a well-trained fault model based on a convolutional neural network algorithm is obtained.

[0107] S4: Input the actual fault data into the trained fault model based on the convolutional neural network algorithm, and output the fault diagnosis results.

[0108] Example 2

[0109] A fault diagnosis system for spline contacts in switchgear includes:

[0110] The physical 3D model and mathematical model building module is used to build physical 3D models and mathematical models. The physical 3D model is specifically a 3D structural model of the plum blossom contactor, and the mathematical model is specifically a temperature-magnetic field coupling mathematical model.

[0111] The digital twin model establishment and 3D feature data acquisition module is used to construct a digital twin model based on the 3D structural model of the plum blossom contact and the temperature-magnetic field coupling mathematical model, simulate different working conditions, and obtain 3D feature data.

[0112] The fault model building module based on the convolutional neural network algorithm is used to transform and process three-dimensional feature data, train the convolutional neural network algorithm, and obtain a trained fault model based on the convolutional neural network algorithm.

[0113] The fault diagnosis result output module is used to input actual fault data into a trained fault model based on a convolutional neural network algorithm and output fault diagnosis results.

[0114] In the physical 3D model and mathematical model building module, a 3D structural model of the plum blossom contactor is created. The specific method for creating the model is as follows:

[0115] Finite element modeling was performed using ANSYS software, and a three-dimensional structural model of the plum blossom contact was constructed based on the structural parameters of the plum blossom contact.

[0116] A temperature-magnetic field coupling mathematical model is established based on the physical 3D model and the mathematical model. The specific method for establishing the model is as follows:

[0117] A temperature-magnetic field coupling mathematical model was established to calculate the heat generation rate of the plum blossom contactor;

[0118] The heat conduction equation for the plum blossom contact is:

[0119]

[0120] Where: ρ is the conductor density, with units of g / cm³. 3 C p λ is the specific heat capacity of the conductor, in J / (kg·K); T is the temperature, in °C; t is the simulation time, in s; λ is the thermal conductivity of the conductor, in W / (m·K); Q is the internal heat source.

[0121] The convective heat transfer coefficient α is:

[0122]

[0123] Where: a is the convective heat transfer coefficient, with units of W / (m²). 2 ·K); λ is the thermal conductivity of the conductor, in W / (m·K); L is the surface characteristic scale, in m; ΔT is the temperature difference between the finger touching the air, in °C;

[0124] Radiant surface heat transfer coefficient α r for:

[0125] α r =εδ b T 4

[0126] Where: α r It is the heat transfer coefficient of the radiating surface, with units of W / (m²). 2 ·K); ε is the equivalent emissivity of the stylus contact on the conductor rod surface; δ b is Boltzmann's constant, in J / K; T is temperature, in °C.

[0127] The heat generation rate of the conductor is as follows:

[0128]

[0129] Where: Q is the heat production rate; the unit is W / m³. 3 I is current, measured in amperes (A); R is resistance, measured in ohms (Ω); and V is volume, measured in cubic meters (m³). 3 .

[0130] The digital twin model is established in the module for establishing the digital twin model and acquiring 3D feature data. The specific method is as follows:

[0131] Electro-thermal structural coupling simulation was carried out using the electromagnetic simulation module, thermal simulation module, and structural field simulation module of ANSYS software. A digital twin model was constructed based on the three-dimensional structural model of the plum blossom contact and the temperature-magnetic field coupling mathematical model.

[0132] In the digital twin model establishment and 3D feature data acquisition module, simulations are performed under different working conditions to obtain 3D feature data, specifically including:

[0133] We selected common fault types such as loose contacts and increased contact resistance to perform electrothermal structure coupled finite element simulation of the temperature of the switch station contacts; based on the calculation results of multiphysics field coupling analysis, we obtained the temperature distribution of the plum blossom contacts under rated current and short-circuit current under different operating conditions, and obtained three-dimensional feature data.

[0134] The fault model building module based on convolutional neural network algorithm transforms and processes 3D feature data, specifically including:

[0135] Data transformation: Convert 3D temperature data into 2D temperature data; convert 3D temperature data into a 2D heat map, project it onto the plane containing the axis, divide the 3D point cloud into the plane containing the axis, erase the height information of the point cloud based on the plane, and project it onto the plane to obtain a point matrix on a 2D plane.

[0136] Data completion: Construct a function with a custom resolution to classify the points in the matrix into each matrix cell based on the resolution as two criteria. Then, take the average temperature of all points in each cell as the temperature of the matrix element to obtain a pixelated heatmap matrix. The specific calculation formula is as follows:

[0137] Input n interpolation nodes: x1 < x2 < ... < x n The corresponding function values ​​are y1, y2, ..., y n The interpolation point to be determined is x0;

[0138] Calculate h j =x j+1 -x j (j = 1, 2, ..., n-1)

[0139] calculate

[0140] calculate

[0141] Solve the system of equations:

[0142] Substitute the values ​​into the interpolation function to solve for the interpolation:

[0143]

[0144] The spline interpolation function is calculated from the first effective value point to the last effective value point to complete the temperature of an entire row. Then, spline interpolation is performed on each column to complete the temperature, and finally, a complete heat map matrix image is obtained.

[0145] The convolutional neural network algorithm is trained in the fault model building module, specifically including:

[0146] Convolutional neural networks (CNNs) are chosen for anomaly detection. The InceptionNet module is set up to encapsulate multiple convolutional kernels of different sizes to learn image features simultaneously. Four convolutional paths are constructed using 1x1, 3x3, 5x5, and pooled 1x1 kernels, and are encapsulated within the InceptionNet module for use by the CNN. The CNN then performs convolution and pooling operations to expand the image information into a final fully connected layer for prediction and classification.

[0147] The optimizer used is the Adam optimizer;

[0148] The simulated temperature three-dimensional feature data is processed into a heat map matrix and then fed into a convolutional neural network for learning. The convolutional neural network is then trained using normal operation data and fault data.

[0149] Finally, a well-trained fault model based on a convolutional neural network algorithm is obtained.

[0150] Example 3

[0151] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement a switch cabinet quincunx contact fault diagnosis method as described in Embodiment 1.

[0152] Example 4

[0153] The present invention provides a computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements a method for diagnosing faults in a switchgear sprite contact as described in Embodiment 1.

[0154] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0155] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0156] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0157] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0158] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A switchgear wobbler contact fault diagnosis method, characterized by, The method comprises the following steps: S1: establishing a physical three-dimensional model and a mathematical model, wherein the physical three-dimensional model is specifically a three-dimensional structure model of a quincunx contact, and the mathematical model is specifically a temperature-magnetic field coupling mathematical model; S2: constructing a digital twin model according to the three-dimensional structure model of the quincunx contact and the temperature-magnetic field coupling mathematical model, simulating different working conditions, and obtaining three-dimensional feature data; S3: performing transformation processing on the three-dimensional feature data, performing convolutional neural network algorithm training, and obtaining a trained fault model based on the convolutional neural network algorithm; The transformation processing of the three-dimensional feature data comprises the following steps: convert the three-dimensional temperature data into two-dimensional temperature data; convert the three-dimensional temperature data into a two-dimensional thermal map, project the three-dimensional point cloud onto the plane where the axis is located, divide the three-dimensional point cloud according to the plane where the axis is located, erase the height information of the point cloud based on the plane as a reference, project onto the plane, and obtain a dot matrix on a two-dimensional plane; construct a function with a custom resolution, classify the dot matrix into each matrix grid with the resolution as two references, take the average temperature of all points in each grid as the temperature of the matrix element, obtain a pixelated thermal map matrix, and the specific calculation formula is as follows: Input n interpolation nodes: x 1 x 2<… x n corresponding function values y 1, y 2,… y n The interpolation point to be solved is x 0; Computing Computing ， Computing , Solving the system of equations: plug in the interpolation function to solve the interpolation: calculate the spline interpolation function from the first valid value point to the last valid value point, fill in a whole row of temperature, then perform spline interpolation on each column to fill in, and finally obtain a complete thermal map matrix image; The convolutional neural network algorithm training comprises the following steps: select a convolutional neural network to perform abnormal state detection; set an InceptionNet module, encapsulate multiple sizes of convolution kernels to simultaneously learn image features, use 1*1 convolution kernels, 3*3 convolution kernels, 5*5 convolution kernels, 1*1 convolution kernels after pooling, four convolution kernel branches are formed to constitute four convolution routes and are encapsulated in the InceptionNet module for CNN calling; the convolutional neural network expands the information of the picture into the last full connection layer through convolution and pooling operations and performs prediction classification; the optimizer selects an Adam optimizer, comprehensively considers the first-order moment estimation and second-order moment estimation of the gradient, calculates the update step, and uses cross-entropy as the loss function of updating the hyperparameters; the simulation temperature three-dimensional feature data is processed into a thermal map matrix and then sent to the convolutional neural network for learning, and then the convolutional neural network is trained using normal operation data and fault data, the initial learning rate of the Adam optimizer is set to 0.0001, and the training rounds EPOCH are 100 rounds; finally, a trained fault model based on the convolutional neural network algorithm is obtained; S4: input actual fault data into the trained fault model based on the convolutional neural network algorithm, and output fault diagnosis results.

2. The switchgear wye contact fault diagnostic method of claim 1, wherein, The method for establishing the three-dimensional structure model of the quincunx contact comprises the following steps: finite element modeling is performed by using ANSYS software, and a three-dimensional structure model of the quincunx contact is constructed according to the structural parameters of the quincunx contact.

3. The switchgear wye contact fault diagnostic method of claim 1, wherein, The method for establishing the temperature-magnetic field coupling mathematical model comprises the following steps: establish a temperature-magnetic field coupling mathematical model, and calculate the heat generation rate of the quincunx contact; the heat conduction equation of the quincunx contact is as follows: wherein: is the density of the conductor, in g / cm 3 ; is the specific heat capacity of the conductor, in J / (kg·K); T is the temperature, in °C; t is the simulation time, in s; the heat generation rate of the conductor is as follows: is the thermal conductivity of the conductor, in W / (m·K); Q is the internal heat source; Convection heat transfer coefficient a is: wherein: a is the convective heat transfer coefficient with units of W / (m 2 ·K); the method for constructing the digital twin model comprises the following steps: is the thermal conductivity of the conductor with units of W / (m·K); L is the surface feature scale with units of m; T is the temperature difference between the touch finger and air with units of °C; Radiation surface heat transfer coefficient is: wherein: is the radiant surface heat transfer coefficient, in W / (m 2 ·K); is the equivalent emissivity of the conductor bar surface wye contact; is the Boltzmann constant, in J / K; T is the temperature, in °C; ​ wherein: Q is the heat production rate; the unit is , I is the current, the unit is A; R is the resistance, the unit is , V is the volume, the unit is .

4. The switchgear wrench contact fault diagnosis method according to claim 1, characterized in that, ​ The ANSYS software electromagnetic simulation module, thermal simulation module and structural field simulation module are used to carry out electric-thermal structure coupling simulation, and a digital twin model is constructed according to a three-dimensional structure model of the star contact and a temperature-magnetic field coupling mathematical model.

5. The switchgear wye contact fault diagnostic method of claim 1, wherein, The simulation of different working conditions and obtaining of three-dimensional feature data specifically include: The temperature of the switch station contact is simulated by selecting the common fault types of loose contact and increased contact resistance; according to the calculation results of the multi-physical field coupling analysis, the temperature distribution of the star contact under the rated current and short-circuit current under different working conditions is obtained, and three-dimensional feature data is obtained.

6. A switchgear wobbler contact fault diagnostic system characterized by, It includes: a physical three-dimensional model and mathematical model establishing module, which is used to establish a physical three-dimensional model and a mathematical model, the physical three-dimensional model is specifically a three-dimensional structure model of the star contact, and the mathematical model is specifically a temperature-magnetic field coupling mathematical model; a digital twin model establishing and three-dimensional feature data obtaining module, which is used to construct a digital twin model according to the three-dimensional structure model of the star contact and the temperature-magnetic field coupling mathematical model, simulate different working conditions, and obtain three-dimensional feature data; a fault model establishing module based on a convolutional neural network algorithm, which is used to process the three-dimensional feature data, train the convolutional neural network algorithm, and obtain a trained fault model based on the convolutional neural network algorithm; the three-dimensional feature data processing includes: converting the three-dimensional temperature data into two-dimensional temperature data; converting the three-dimensional temperature data into a two-dimensional thermal map, projecting it onto the plane where the axis is located, segmenting the three-dimensional point cloud with the plane where the axis is located, and erasing the height information of the point cloud with the plane as the reference, projecting it onto the plane to obtain a point array on a two-dimensional plane; constructing a self-defined resolution function, classifying the point array into each matrix grid with the resolution as the two reference matrices, taking the average temperature of all points in each grid as the temperature of the matrix element, and obtaining a pixelated thermal map matrix, the specific calculation formula is as follows: Input n interpolation nodes: x 1 x 2<… x n Corresponding function values y 1, y 2,… y n The interpolation point to be solved is x 0; Computing Computing ， Computing , Solving the system of equations: solving the interpolation by inserting the interpolation function: from the first effective value point to the last effective value point, the spline interpolation function is calculated, a whole row of temperature is filled, and each column is filled by spline interpolation, and finally a complete thermal map matrix image is obtained; the convolutional neural network algorithm training includes: selecting a convolutional neural network for abnormal state detection; setting an InceptionNet module, encapsulating multiple size convolution kernels to learn image features at the same time, using 1*1 convolution kernels, 3*3 convolution kernels, 5*5 convolution kernels, 1*1 convolution kernels after pooling, four convolution kernel branches to form four convolution routes and encapsulate them in the InceptionNet module for CNN calling; the convolutional neural network expands the information of the picture into the final full connection layer through convolution and pooling operations and performs prediction classification; the optimizer selects the Adam optimizer, which comprehensively considers the first moment estimation and second moment estimation of the gradient to calculate the update step; cross-entropy is used as the loss function of updating the hyperparameters. The simulation temperature three-dimensional feature data is processed into a heat map matrix and then sent to a convolutional neural network for learning, and the convolutional neural network is trained by using normal operation data and fault data, and the initial learning rate of the Adam optimizer is set to 0.0001, and the training round EPOCH is 100 rounds; The fault diagnosis result output module is used for inputting actual fault data into the trained fault model based on the convolutional neural network algorithm, and outputting a fault diagnosis result.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the switch cabinet wrench contact fault diagnosis method in any one of claims 1-5.

8. A computer-readable storage medium storing a computer program, the computer-readable storage medium comprising: The computer program is executed by the processor to realize the switch cabinet wrench contact fault diagnosis method in any one of claims 1-5.

Images