Casing simulation frequency domain dielectric spectrum curve prediction method and device based on convolutional neural network, equipment and medium

Abstract

The invention provides a sleeve simulation frequency domain dielectric spectrum curve prediction method, device and equipment based on a convolutional neural network, and a medium, and relates to the field of sleeve testing, and the method comprises the steps: obtaining constant temperature frequency domain dielectric spectrum data of a to-be-tested sleeve at different temperatures; and inputting the constant-temperature frequency-domain dielectric spectrum data of the to-be-tested casing pipe at different temperatures into the trained frequency-domain dielectric spectrum curve prediction model to obtain a frequency-domain dielectric spectrum prediction curve of the to-be-tested casing pipe. Therefore, the model is used for analyzing the complex nonlinear transformation of the dielectric response under the temperature gradient of the sleeve, learning the correlation between different frequency points and the local action between adjacent temperature areas, and predicting the complete curve of the whole sleeve through different temperature areas, so that the prediction efficiency is improved, and the diagnosis period is shortened.

Description

Technical Field

[0001] This invention relates to the field of tubing testing technology, and more specifically, to a method, apparatus, equipment, and medium for predicting the frequency domain dielectric spectrum curve of tubing simulation based on a convolutional neural network. Background Technology

[0002] Energy and electricity are the foundation of economic and social development, and a safe, efficient, and clean power supply is a crucial guarantee for my country's modernization process. With the continuous growth in electricity load demand and the large-scale construction of ultra-high-voltage power grids, the insulation level of key equipment in power grid transmission and transformation faces higher requirements. High-voltage power transformer bushings, as key connecting components between external and internal conductors, mostly adopt an oil-impregnated paper capacitor core design. Research shows that over a quarter of transformer failures are caused by bushing failures. In recent years, epoxy resin, due to its superior performance, good heat and corrosion resistance, and ease of processing and molding, has been widely used in the field of power equipment insulation. Dry bushings, represented by epoxy resin, avoid insulation damage caused by insulation failure because there is no insulating oil as a cooling medium inside the core; however, their thermal conductivity is far lower than that of oil-impregnated paper bushings. During normal operation, the electric field strength of the inner layer of the impregnated paper bushing core and the end of the electrode plate is relatively high, the temperature of the central conductor is high and a temperature gradient is easily formed, and the insulation medium is difficult to dissipate heat after heating, which in turn leads to internal local overheating defects. Under the combined effect of electricity and heat, the insulation is prone to aging and shortens its lifespan. The electric field can also cause the field strength distortion of the deteriorated part of the core, induce partial discharge, and eventually form a discharge channel between the capacitor screens, causing short circuits between the capacitor screens, increased losses, and a significant reduction in dielectric strength, leading to bushing failure.

[0003] There are significant technical challenges in assessing the insulation status of paper-impregnated sleeves. As a composite insulation system, paper-impregnated insulation exhibits significant temperature dependence in its dielectric properties, with different temperatures at different locations. This makes it difficult to determine the overall insulation status based on local data, resulting in exceptionally difficult assessments based on frequency domain dielectric spectrum characteristics. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a method for predicting the frequency domain dielectric spectrum curve of a sleeve simulation based on a convolutional neural network, the method comprising:

[0005] Acquire isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures;

[0006] The isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures are input into the trained frequency domain dielectric spectrum curve prediction model to obtain the frequency domain dielectric spectrum prediction curve of the sleeve under test.

[0007] The frequency domain dielectric spectrum prediction model is trained through the following training operations:

[0008] Acquire isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures;

[0009] Based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures, the frequency domain dielectric spectrum data of the training sleeve under temperature gradient is calculated using simulation software.

[0010] Using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data, and the frequency domain dielectric spectrum data of the training sleeve under temperature gradient as label data, an initial convolutional neural network is trained to obtain a frequency domain dielectric spectrum curve prediction model.

[0011] As one possible implementation, the initial convolutional neural network includes an input layer, an output layer, and at least two convolutional layers, wherein a linear rectified function is added as an activation function to both the input layer and the convolutional layers.

[0012] As one possible implementation, the step of using isothermal frequency-domain dielectric spectrum data of the training sleeve at different temperatures as input data, and frequency-domain dielectric spectrum data of the training sleeve under temperature gradients as label data, to train an initial convolutional neural network to obtain a frequency-domain dielectric spectrum curve prediction model includes:

[0013] The isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures are used as input data and input into the initial convolutional neural network to obtain the output data.

[0014] The mean squared error loss function is used to calculate the loss function values ​​for the output data and the label data;

[0015] Based on the loss function value, the Adam optimization function is used to adjust the parameters of the initial convolutional neural network.

[0016] As one possible implementation, acquiring the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures includes:

[0017] Obtain the complex relative permittivity εr and conductivity σ of the training tube at multiple discrete frequencies f under multiple preset temperatures;

[0018] For each group of complex relative permittivity εᵣ and conductivity σ, calculate the dielectric loss factor tanδ.

[0019] As one possible implementation, the step of calculating the frequency domain dielectric spectrum data of the training sleeve under a temperature gradient using simulation software, based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures, includes:

[0020] Using simulation software, boundary conditions are applied to the training tube sleeve at multiple preset temperatures and multiple discrete frequencies f, and frequency domain analysis is performed. The admittance Y is calculated by current field simulation. The boundary conditions include adding an applied voltage V.

[0021] The dielectric loss factor tanδ is calculated based on each discrete frequency f and its corresponding admittance Y.

[0022] As one possible implementation, using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data includes:

[0023] The isothermal frequency domain dielectric spectrum data corresponding to b frequencies at a preset temperature are used to construct a b×a input matrix;

[0024] The input matrix is ​​increased in dimensionality to obtain three-dimensional tensor structure data related to sample index, frequency dimension, and temperature dimension;

[0025] The dimensionality of the three-dimensional tensor structure data is reduced to obtain the input data.

[0026] Secondly, embodiments of the present invention provide a sleeve simulation frequency domain dielectric spectrum curve prediction device based on a convolutional neural network, the device comprising:

[0027] The test data acquisition module is used to acquire the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures;

[0028] The prediction module is used to input the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures into the trained frequency domain dielectric spectrum curve prediction model to obtain the frequency domain dielectric spectrum prediction curve of the sleeve under test.

[0029] The frequency domain dielectric spectrum prediction model is trained through the following training operations:

[0030] Acquire isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures;

[0031] Based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures, the frequency domain dielectric spectrum data of the training sleeve under temperature gradient is calculated using simulation software.

[0032] Using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data, and the frequency domain dielectric spectrum data of the training sleeve under temperature gradient as label data, an initial convolutional neural network is trained to obtain a frequency domain dielectric spectrum curve prediction model.

[0033] As one possible implementation, the initial convolutional neural network includes an input layer, an output layer, and at least two convolutional layers, wherein a linear rectified function is added as an activation function to both the input layer and the convolutional layers.

[0034] Thirdly, embodiments of the present invention provide an electronic device, comprising:

[0035] One or more processors;

[0036] Storage device, on which one or more programs are stored,

[0037] When the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method as described in the first aspect.

[0038] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by one or more processors, implements the method described in the first aspect.

[0039] The embodiments of the present invention provide a method, apparatus, device, and medium for predicting the frequency domain dielectric spectrum curve of a sleeve based on a convolutional neural network. This involves first acquiring isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures; then inputting this data into a trained frequency domain dielectric spectrum curve prediction model to obtain the predicted frequency domain dielectric spectrum curve of the sleeve under test. Specifically, the frequency domain dielectric spectrum curve prediction model is trained through the following operations: acquiring isothermal frequency domain dielectric spectrum data of a training sleeve at different temperatures; calculating the frequency domain dielectric spectrum data of the training sleeve under a temperature gradient using simulation software based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures; and training an initial convolutional neural network using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data and the frequency domain dielectric spectrum data of the training sleeve under a temperature gradient as label data to obtain the frequency domain dielectric spectrum curve prediction model.

[0040] In this way, by using the model to analyze the complex nonlinear transformation of the dielectric response under the temperature gradient of the bushing, learning the correlation between different frequency points and the local interaction between adjacent temperature regions, and predicting the complete curve of the entire bushing through different temperature regions, the prediction efficiency is improved and the diagnosis cycle is shortened. Attached Figure Description

[0041] Figure 1 This is an exemplary system architecture diagram in which an embodiment of the present invention can be applied;

[0042] Figure 2 This is a flowchart of an embodiment of the sleeve simulation frequency domain dielectric spectrum curve prediction method based on convolutional neural network provided by the present invention;

[0043] Figure 3 This is an FDS test structure diagram of the sleeve simulation frequency domain dielectric spectrum curve prediction method based on convolutional neural network provided in the embodiments of the present invention;

[0044] Figure 4 These are the tanδ curves of the simulated bushing under three temperature gradients provided in the embodiments of the present invention;

[0045] Figure 5 This is a cross-sectional view of the 3D simulation model of bushing insulation provided in this embodiment of the invention;

[0046] Figure 6 This is a schematic diagram of the 3D simulation model of bushing insulation provided in an embodiment of the present invention;

[0047] Figure 7 This is an initial convolutional neural network structure diagram provided in an embodiment of the present invention;

[0048] Figure 8 This is a ReLU function image provided in an embodiment of the present invention;

[0049] Figure 9 This is a schematic diagram of the loss function values ​​provided in an embodiment of the present invention;

[0050] Figure 10 This is a structural diagram of the sleeve simulation frequency domain dielectric spectrum curve prediction device based on convolutional neural network provided in an embodiment of the present invention;

[0051] Figure 11 This is a schematic diagram of the structure of a computer system suitable for implementing embodiments of the present disclosure. Detailed Implementation

[0052] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0053] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0054] Figure 1 An exemplary system architecture 100 is shown, illustrating an embodiment of the sleeve simulation frequency domain dielectric spectrum prediction method, apparatus, electronic device, and storage medium based on the convolutional neural network of the present invention.

[0055] like Figure 1As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0056] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as prediction applications, voice recognition applications, short video social applications, audio and video conferencing applications, live video streaming applications, document editing applications, input method applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0057] Terminal devices 101, 102, and 103 can be either hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with displays, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), laptops, and desktop computers, etc. When terminal devices 101, 102, and 103 are software, they can be installed on the terminal devices listed above. They can be implemented as multiple software programs or software modules (e.g., used to provide prediction services) or as a single software program or software module. No specific limitations are imposed here.

[0058] In some cases, the sleeve simulation frequency domain dielectric spectrum curve prediction method based on convolutional neural networks provided by this invention can be executed by terminal devices 101, 102, and 103. Correspondingly, the sleeve simulation frequency domain dielectric spectrum curve prediction device based on convolutional neural networks can be set in terminal devices 101, 102, and 103. In this case, the system architecture 100 may not include server 105.

[0059] In some cases, the sleeve simulation frequency domain dielectric spectrum curve prediction method provided by the present invention can be jointly executed by terminal devices 101, 102, 103 and server 105. For example, the step of "acquiring isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures" can be executed by terminal devices 101, 102, 103, and the step of "inputting the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures into the trained frequency domain dielectric spectrum curve prediction model to obtain the frequency domain dielectric spectrum prediction curve of the sleeve under test" can be executed by server 105. The present invention does not limit this. Correspondingly, the sleeve simulation frequency domain dielectric spectrum curve prediction device based on convolutional neural network can also be respectively set in terminal devices 101, 102, 103 and server 105.

[0060] In some cases, the sleeve simulation frequency domain dielectric spectrum curve prediction method based on convolutional neural network provided by the present invention can be executed by server 105. Correspondingly, the sleeve simulation frequency domain dielectric spectrum curve prediction device based on convolutional neural network can also be set in server 105. In this case, the system architecture 100 may not include terminal devices 101, 102, and 103.

[0061] It should be noted that server 105 can be either hardware or software. When server 105 is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When server 105 is software, it can be implemented as multiple software programs or software modules (for example, used to provide distributed services), or as a single software program or software module. No specific limitations are made here.

[0062] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0063] Continue to refer to Figure 2 The document illustrates a flowchart 200 of an embodiment of a sleeve simulation frequency domain dielectric spectrum curve prediction method according to the present invention, which includes the following steps:

[0064] Step S101: Obtain the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures.

[0065] See Figure 3In the laboratory, insulating cardboard samples of the sleeve under test are prepared at different temperatures. The isothermal frequency domain spectroscopy (FDS) data corresponding to the complex relative permittivity of the insulating samples at different temperatures are measured and recorded. Specifically, the insulating cardboard samples are thoroughly dried and oil-impregnated. Then, the insulating samples are placed in cavities at different isothermal temperatures, and an FDS test is performed at preset temperature intervals. The tanδ values ​​(which can be calculated from the complex relative permittivity εᵣ and conductivity σ) of different frequency sampling points of the curves at different temperatures are obtained, serving as the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures for subsequent data structure processing. Optionally, the temperature setting range is 20℃-120℃, the preset temperature interval is 5℃, and the different frequencies can preferably be set to 21 frequencies.

[0066] The isothermal frequency domain dielectric spectrum data includes the dielectric loss factor tanδ, and specifically, the complex relative permittivity ε corresponding to multiple discrete frequencies f at multiple preset temperatures can be obtained. r And conductivity σ; for each group of corresponding complex relative permittivity εᵣ and conductivity σ, calculate the dielectric loss factor tanδ.

[0067] Step S102: Input the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures into the trained frequency domain dielectric spectrum curve prediction model to obtain the frequency domain dielectric spectrum prediction curve of the sleeve under test.

[0068] The trained frequency-domain dielectric spectrum prediction model can learn the dielectric response characteristics of the sleeve under different temperature gradients using a large amount of sample data, thereby accurately predicting the frequency-domain dielectric spectrum curve over the entire temperature range. Specifically, the model receives multi-temperature point FDS data obtained in step S101 as input. After extracting local frequency-domain features through convolutional layers, the data dimensionality is compressed through pooling layers, and then the temperature-frequency cross-features are integrated by fully connected layers. The final output is shown below. Figure 4 The diagram shows a continuous dielectric spectrum prediction curve covering the entire temperature range. This process eliminates the need for additional measurements at intermediate temperature points; only boundary temperature data is required to reconstruct the overall curve, significantly improving testing efficiency.

[0069] In this way, by using the model to analyze the complex nonlinear transformation of the dielectric response under the temperature gradient of the bushing, learning the correlation between different frequency points and the local interaction between adjacent temperature regions, and predicting the complete curve of the entire bushing through different temperature regions, the prediction efficiency is improved and the diagnosis cycle is shortened.

[0070] The training operation of this frequency domain dielectric spectrum curve prediction model is described in detail below.

[0071] Step S201: Obtain the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures.

[0072] Similar to the method of obtaining the test sleeve data in step S101, multiple sets of training sleeve samples are prepared in a laboratory environment. The ambient temperature of the samples is controlled by a precision temperature control system. A high-precision dielectric spectrum analyzer is used to measure the frequency domain dielectric response of each set of samples, which will not be elaborated here.

[0073] Step S202: Based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures, use simulation software to calculate the frequency domain dielectric spectrum data of the training sleeve under the temperature gradient.

[0074] Before proceeding, it is necessary to derive the control equations applicable to the frequency domain analysis of bushing insulation, starting from Maxwell's equations, to provide a theoretical basis for subsequent simulations.

[0075] The Maxwell quasi-static equations are as follows:

[0076]

[0077]

[0078] (1)

[0079] in, It is the gradient operator, where E is the electric field strength, H is the magnetic field strength, D is the electric displacement vector, J is the current density, and ω is the angular frequency.

[0080] (2)

[0081] (3)

[0082] Where E is the electric field strength, This is the gradient operator, where V is the electric potential, ε0 is the vacuum permittivity, and ε r Let be the relative permittivity of the insulating material, and D be the electric displacement vector. In order to assign values ​​to the material properties of the simulation model, it is necessary to obtain the relative permittivity and conductivity of each part of the model.

[0083] Combining equations (1) to (3) and the current conservation equation, the quasi-static field equation applicable to the dielectric frequency domain finite element analysis of bushing insulation is obtained as follows:

[0084] (4) (5)

[0085] in, This is the gradient operator, where J is the conduction current and D is the electric displacement vector. For displacement current, ε is the conductivity, j is the imaginary unit, ω is the angular frequency, ε₀ is the vacuum permittivity, and ε r is the relative permittivity of the insulating material, and V is the electric potential.

[0086] The isothermal frequency domain dielectric spectrum data of insulating paperboard samples measured at different temperatures were input into the material properties of the corresponding part of the simulation model. The computational region was meshed using a polyhedral mesh, and a 3D simulation model considering the temperature gradient of the sleeve was built. The current field was selected for solution using simulation software. Maxwell's quasi-static equation (Equation (5)) was used as the governing equation for finite element analysis. Boundary conditions were applied to the training sleeves corresponding to multiple discrete frequencies f at multiple preset temperatures. The established 3D simulation model was then used to solve the problem by applying boundary conditions to the training sleeves at multiple preset temperatures and multiple discrete frequencies f. Figure 6 The regions are divided and assigned corresponding material properties, including temperature-dependent conductivity σ and relative permittivity. The input is given to the simulation model to concretize formula (5). Frequency domain analysis is performed by applying voltage V and grounding boundary conditions to output admittance Y. Then, based on each discrete frequency f and its corresponding admittance Y, the dielectric loss factor tanδ is calculated. The simulation results are as follows: Figure 5 As shown, because the sleeve is highly symmetrical, the 2D model, when rotated, becomes a 3D image, as shown. Figure 6 Three-dimensional model.

[0087] The boundary conditions include adding an applied voltage V, for example, the conductive rod of the simulation model is used as the high voltage electrode, and an AC voltage of 200V is added; the outermost capacitor layer, i.e. the final screen, is used as the ground electrode, and grounding is added.

[0088] Multi-temperature FDS data obtained from experiments were imported into electromagnetic field simulation software. A three-dimensional electric-thermal field coupled model incorporating temperature gradient distribution was established to simulate the combined stress state of the bushing during actual operation. Reasonable boundary conditions needed to be set during the simulation, including but not limited to: applying rated voltage to the conductor, setting zero potential on the outer surface of the insulation, and considering the nonlinear effect of temperature on the dielectric parameters of the material. The electric field distribution under different temperature gradients was calculated through finite element analysis, thereby deriving the dielectric response characteristics over a continuous temperature range. Simulated dielectric spectrum data matching the dimensionality of the experimental data but covering a wider temperature range was generated and used as label data for model training.

[0089] Step S203: Using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data, and the frequency domain dielectric spectrum data of the training sleeve under temperature gradient as label data, train an initial convolutional neural network to obtain a frequency domain dielectric spectrum curve prediction model.

[0090] See Figure 7The initial convolutional neural network includes an input layer, an output layer, and at least two convolutional layers, wherein a linear rectified function is added as the activation function to both the input layer and the convolutional layers. Hyperparameters are set for the initial convolutional neural network, and two convolutional layers are selected and configured: convolutional layer 1 uses 16 3×3 convolutional kernels, and convolutional layer 2 uses 32 3×3 convolutional kernels. Each convolutional layer is followed by a ReLU activation function; the function graph is shown below. Figure 8 Following global average pooling, the predicted values ​​for 21 frequency points are output through a fully connected layer.

[0091] Using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data, the isothermal frequency domain dielectric spectrum data corresponding to b frequencies at a preset temperatures can be used to construct a b×a input matrix; the input matrix is ​​increased in dimension to obtain three-dimensional tensor structure data of sample index, frequency dimension and temperature dimension; the three-dimensional tensor structure data is reduced in dimension to obtain the input data.

[0092] For example, the values ​​of 21 frequency points in the simulated FDS curve are extracted into a 21×1 output matrix. Then, FDS data corresponding to 7 temperatures for each bushing simulation are selected from the FDS database of the insulation template to form a 21×7 input matrix. The input matrix is ​​transformed into a three-dimensional tensor structure of (sample index, 21, 7) by dimensionality increase. Its physical dimension is sample index × frequency dimension × temperature dimension. The output matrix is ​​transformed into a two-dimensional matrix of (sample index, 21, 1), so that the bushing simulation data and the data of the insulation template form a one-to-one correspondence as input data.

[0093] The specific training process may include: using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data, inputting it into an initial convolutional neural network to obtain output data; using the mean square error loss function to calculate the loss function value of the output data and the label data; and using the Adam optimization function to adjust the parameters of the initial convolutional neural network based on the loss function value.

[0094] As one possible implementation, a portion of the input data and its corresponding label data can be used as a training set for the aforementioned training operations, while another portion can be used as a validation set for real-time validation during model training. By comparing the prediction results of the validation set with the true labels, the generalization ability of the model can be dynamically evaluated. For details on calculating the loss function values ​​for the training and validation sets respectively, please refer to [link to relevant documentation]. Figure 9 .

[0095] In summary, the bushing simulation frequency domain dielectric spectrum curve prediction method provided in this embodiment generates extended training data by constructing a three-dimensional simulation model including temperature gradients. This data is then combined with measured multi-temperature FDS data to form input-label pairs. An improved convolutional neural network structure is used to achieve end-to-end dielectric spectrum prediction. This method overcomes the limitation of traditional testing requiring point-by-point measurements, reconstructing the dielectric response curve over a continuous temperature range using only boundary temperature data, significantly improving testing efficiency. In practical engineering applications, this technology can quickly assess the insulation status of bushings, providing crucial data support for condition-based maintenance of power equipment, and has high engineering practical value.

[0096] Further reference Figure 10 As an implementation of the methods shown in the above figures, the present invention provides an embodiment of a device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0097] like Figure 10 As shown, the sleeve simulation frequency domain dielectric spectrum curve prediction device 300 based on convolutional neural network in this embodiment includes: a test data acquisition module 301 and a prediction module 302.

[0098] The test data acquisition module 301 is used to acquire the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures;

[0099] Prediction module 302 is used to input the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures into the trained frequency domain dielectric spectrum curve prediction model to obtain the frequency domain dielectric spectrum prediction curve of the sleeve under test.

[0100] The frequency domain dielectric spectrum prediction model is trained through the following training operations:

[0101] Acquire isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures;

[0102] Based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures, the frequency domain dielectric spectrum data of the training sleeve under temperature gradient is calculated using simulation software.

[0103] Using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data, and the frequency domain dielectric spectrum data of the training sleeve under temperature gradient as label data, an initial convolutional neural network is trained to obtain a frequency domain dielectric spectrum curve prediction model.

[0104] In this embodiment, the specific processing of the sleeve simulation frequency domain dielectric spectrum curve prediction device 300 based on convolutional neural network and its resulting technical effects can be referred to separately. Figure 2The relevant descriptions of the steps in the corresponding embodiments will not be repeated here.

[0105] As one possible implementation, the initial convolutional neural network includes an input layer, an output layer, and at least two convolutional layers, wherein a linear rectified function is added as an activation function to both the input layer and the convolutional layers.

[0106] As one possible implementation, using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data, and the frequency domain dielectric spectrum data of the training sleeve under a temperature gradient as label data, an initial convolutional neural network is trained to obtain a frequency domain dielectric spectrum curve prediction model, including:

[0107] The isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures are used as input data and input into the initial convolutional neural network to obtain the output data.

[0108] The mean squared error loss function is used to calculate the loss function values ​​for the output data and the label data;

[0109] Based on the loss function value, the Adam optimization function is used to adjust the parameters of the initial convolutional neural network.

[0110] As one possible implementation, acquiring the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures includes:

[0111] Obtain the complex relative permittivity εr and conductivity σ of the training tube at multiple discrete frequencies f under multiple preset temperatures;

[0112] For each group of complex relative permittivity εᵣ and conductivity σ, calculate the dielectric loss factor tanδ.

[0113] As one possible implementation, the step of calculating the frequency domain dielectric spectrum data of the training sleeve under a temperature gradient using simulation software, based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures, includes:

[0114] Using simulation software, boundary conditions are applied to the training tube sleeve at multiple preset temperatures and multiple discrete frequencies f, and frequency domain analysis is performed. The admittance Y is calculated by current field simulation. The boundary conditions include adding an applied voltage V.

[0115] The dielectric loss factor tanδ is calculated based on each discrete frequency f and its corresponding admittance Y.

[0116] As one possible implementation, using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data includes:

[0117] The isothermal frequency domain dielectric spectrum data corresponding to b frequencies at a preset temperature are used to construct a b×a input matrix;

[0118] The input matrix is ​​increased in dimensionality to obtain three-dimensional tensor structure data related to sample index, frequency dimension, and temperature dimension;

[0119] The dimensionality of the three-dimensional tensor structure data is reduced to obtain the input data.

[0120] It should be noted that the implementation details and technical effects of each module and unit in the device provided in the embodiments of the present invention can be referred to the description of other embodiments of the present invention, and will not be repeated here.

[0121] The following is for reference. Figure 11 It shows a schematic diagram of the structure of a computer system 400 suitable for implementing the electronic device of the present invention. Figure 11 The computer system 400 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0122] like Figure 11 As shown, the computer system 400 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage device 408 into a random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the computer system 400. The processing device 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0123] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows computer system 400 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 A computer system 400 with various electronic devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0124] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a storage device 408, or installed from a ROM 402. When the computer program is executed by a processing device 401, it performs the functions defined in the methods of the embodiments of the present invention.

[0125] It should be noted that the computer-readable medium described above in this invention can be a computer-readable signal medium, a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0126] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0127] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the following functions: Figure 2 The methods illustrated in the embodiments and their alternative implementations are methods.

[0128] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0129] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0130] The units or modules described in the embodiments of the present invention can be implemented in software or hardware. The names of the units or modules do not necessarily limit the specific unit itself.

[0131] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

Claims

1. A method for predicting the frequency domain dielectric spectrum curve of a sleeve based on a convolutional neural network, characterized in that, The method includes: Acquire isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures; The isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures are input into the trained frequency domain dielectric spectrum curve prediction model to obtain the frequency domain dielectric spectrum prediction curve of the sleeve under test. The frequency domain dielectric spectrum prediction model is trained through the following training operations: Acquire isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures; Based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures, the frequency domain dielectric spectrum data of the training sleeve under temperature gradient is calculated using simulation software. Using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data, and the frequency domain dielectric spectrum data of the training sleeve under temperature gradient as label data, an initial convolutional neural network is trained to obtain a frequency domain dielectric spectrum curve prediction model.

2. The method according to claim 1, characterized in that, The initial convolutional neural network includes an input layer, an output layer, and at least two convolutional layers, wherein a linear rectified function is added as an activation function to both the input layer and the convolutional layers.

3. The method according to claim 1, characterized in that, The process of using isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data and frequency domain dielectric spectrum data of the training sleeve under temperature gradients as label data to train an initial convolutional neural network to obtain a frequency domain dielectric spectrum curve prediction model includes: The isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures are used as input data and input into the initial convolutional neural network to obtain the output data. The mean squared error loss function is used to calculate the loss function values ​​for the output data and the label data; Based on the loss function value, the Adam optimization function is used to adjust the parameters of the initial convolutional neural network.

4. The method according to claim 1, characterized in that, The acquisition of isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures includes: Obtain the complex relative permittivity ε corresponding to multiple discrete frequencies f at multiple preset temperatures for the training tube. r and conductivity σ; For each group of complex relative permittivity εᵣ and conductivity σ, calculate the dielectric loss factor tanδ.

5. The method according to claim 4, characterized in that, The step of calculating the frequency domain dielectric spectrum data of the training sleeve under a temperature gradient using simulation software, based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures, includes: Using simulation software, boundary conditions are applied to the training tube sleeve at multiple preset temperatures and multiple discrete frequencies f, and frequency domain analysis is performed. The admittance Y is calculated by current field simulation. The boundary conditions include adding an applied voltage V. The dielectric loss factor tanδ is calculated based on each discrete frequency f and its corresponding admittance Y.

6. The method according to claim 3, characterized in that, The step of using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data includes: The isothermal frequency domain dielectric spectrum data corresponding to b frequencies at a preset temperature are used to construct a b×a input matrix; The input matrix is ​​increased in dimensionality to obtain three-dimensional tensor structure data related to sample index, frequency dimension, and temperature dimension; The dimensionality of the three-dimensional tensor structure data is reduced to obtain the input data.

7. A device for predicting the frequency domain dielectric spectrum curve of a sleeve based on a convolutional neural network, characterized in that, The device includes: The test data acquisition module is used to acquire the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures; The prediction module is used to input the isothermal frequency domain dielectric spectrum data of the sleeve under test at different temperatures into the trained frequency domain dielectric spectrum curve prediction model to obtain the frequency domain dielectric spectrum prediction curve of the sleeve under test. The frequency domain dielectric spectrum prediction model is trained through the following training operations: Acquire isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures; Based on the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures, the frequency domain dielectric spectrum data of the training sleeve under temperature gradient is calculated using simulation software. Using the isothermal frequency domain dielectric spectrum data of the training sleeve at different temperatures as input data, and the frequency domain dielectric spectrum data of the training sleeve under temperature gradient as label data, an initial convolutional neural network is trained to obtain a frequency domain dielectric spectrum curve prediction model.

8. The apparatus according to claim 7, characterized in that, The initial convolutional neural network includes an input layer, an output layer, and at least two convolutional layers, wherein a linear rectified function is added as an activation function to both the input layer and the convolutional layers.

9. An electronic device, characterized in that, include: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, It stores a computer program thereon, wherein the computer program, when executed by one or more processors, implements the method as described in any one of claims 1-6.

Images