Transformer bushing temperature rise simulation method and device

Abstract

The invention discloses a transformer bushing temperature rise simulation method and device, and relates to the technical field of power transmission and transformation and digital twinborn simulation, and the method comprises the steps: obtaining a diffusion model which is obtained through the training of an obtained training data set; obtaining a target sleeve heating power parameter; and temperature field simulation is carried out based on the heating power parameters of the target bushing and the diffusion model to obtain a two-dimensional temperature field simulation result of the target bushing, and the simulation result can accurately reflect the temperature rise condition of the transformer bushing in the actual operation process. Through the method, the accuracy and efficiency of temperature rise simulation of the transformer bushing can be effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power transmission and digital twin simulation technology, in particular to a transformer bushing temperature rise simulation method and device. BACKGROUND

[0002] Transformer bushing temperature rise simulation is a key technology to ensure the safe operation of the power grid. As a channel for energy exchange inside and outside the transformer, the bushing is prone to insulation aging, dielectric breakdown and other faults due to local overheating under long-term high load, and may even cause explosion or fire. Through simulation, the internal current density distribution, heat conduction path and temperature field changes of the bushing can be accurately predicted, and design defects or operation risks can be identified in advance. For example, optimizing the structure of the conductive rod, improving the proportion of the insulating material or adjusting the heat dissipation structure can be verified through simulation to avoid the cost waste caused by repeated trial production. In addition, the simulation results can provide a theoretical threshold for state monitoring and assist in developing differentiated operation and maintenance strategies to extend the service life of the equipment. With the growth of ultra-high voltage projects and the demand for new energy grid connection, bushing temperature rise simulation has become a core technical means to improve the reliability of transformers and ensure the stable operation of the power grid.

[0003] At present, the temperature rise simulation of the transformer bushing mainly relies on finite volume simulation software, finite element simulation software, etc. For specific working condition parameters, a complete simulation process is carried out. This will result in a large time consumption for each simulation, and the simulation efficiency is low when the design is frequently changed. SUMMARY

[0004] The main purpose of the present application is to provide a transformer bushing temperature rise simulation method and device, which aims to solve the technical problem of low efficiency of the current transformer bushing temperature rise simulation.

[0005] To achieve the above-mentioned purpose, the present application provides a transformer bushing temperature rise simulation method, which comprises:

[0006] obtaining a diffusion model, wherein the diffusion model is obtained by training a training data set;

[0007] obtaining a target bushing heat power parameter;

[0008] performing temperature field simulation based on the target bushing heat power parameter and the diffusion model to obtain a target bushing two-dimensional temperature field simulation result.

[0009] In an embodiment, the step of obtaining a diffusion model comprises:

[0010] obtaining a plurality of power working condition parameters of the transformer bushing, wherein each power working condition parameter is uniformly distributed;

[0011] performing bushing temperature field simulation and processing on each power working condition parameter to obtain a training data set;

[0012] constructing a diffusion network structure and training the diffusion network structure by using the training data set to obtain a diffusion model.

[0013] In an embodiment, the step of performing casing temperature field simulation and processing for each of the power operating condition parameters to obtain a training data set comprises:

[0014] performing casing temperature field simulation for each of the power operating condition parameters to obtain an initial simulation result;

[0015] obtaining a preset number of two-dimensional temperature distribution maps based on the initial simulation result on the axial symmetric plane of the casing;

[0016] performing data enhancement processing on the preset number of two-dimensional temperature distribution maps to obtain a training data set.

[0017] In an embodiment, the step of performing data enhancement processing on the preset number of two-dimensional temperature distribution maps to obtain a training data set comprises:

[0018] cutting the two-dimensional temperature distribution map into a distribution map of a first pixel size based on the center line of the casing in the preset number of two-dimensional temperature distribution maps;

[0019] scaling the distribution map of the first pixel size into a distribution map of a second pixel size;

[0020] merging and quantifying color channels in the distribution map of the second pixel size into grayscale channels and normalizing to a preset range to obtain a preset number of two-dimensional numerical tensors;

[0021] obtaining a training data set according to the preset number of two-dimensional numerical tensors.

[0022] In an embodiment, the step of constructing a diffusion network structure and training the diffusion network structure by using the training data set to obtain a diffusion model comprises:

[0023] constructing a diffusion network structure, wherein the diffusion network structure comprises a weight constant part and a weight iterative updating part;

[0024] obtaining a plurality of two-dimensional numerical tensors according to the training data set, and obtaining power operating condition parameters corresponding to the plurality of two-dimensional numerical tensors;

[0025] taking the plurality of two-dimensional numerical tensors as network input images, and taking the power operating condition parameters as network input parameters;

[0026] training the diffusion network structure by using the network input images and the network input parameters to obtain a diffusion model.

[0027] In an embodiment, the step of training the diffusion network structure by the network input image and the network input parameter to obtain a diffusion model comprises:

[0028] performing word embedding operation on the network input parameter in the diffusion network structure to map the network input parameter into a first vector of a preset dimension, and performing broadcast operation on the first vector to expand it into a tensor structure of a preset size;

[0029] performing un-stitching on the network input image to separate at least three channel value matrices;

[0030] performing binaryzation operation on the value matrix of the second channel to obtain a binaryzation tensor;

[0031] performing network weight update by the value matrix of the first channel, in which, a noise tensor is randomly generated, the noise tensor and the value matrix of the first channel are summed according to a weight ratio, and the summed result is stitched with the preset size tensor structure to obtain an input tensor, wherein the weight ratio linearly changes in the iteration round;

[0032] performing convolution transformation, residual block processing, self-attention weighted summation, inverse convolution transformation and tensor stitching on the input tensor in sequence to obtain an output tensor;

[0033] calculating the mean square error of the output tensor and a target tensor, determining the error gradient according to the mean square error, and updating the network weight and bias by an optimization algorithm based on the error gradient, and iteratively performing until a training termination condition is met to obtain a diffusion model.

[0034] In an embodiment, the step of performing temperature field simulation based on the target sleeve heat power parameter and the diffusion model to obtain a target sleeve two-dimensional temperature field simulation result comprises:

[0035] generating an initial noise tensor;

[0036] inputting the initial noise tensor and the target sleeve heat power parameter into the diffusion model to perform temperature field simulation to obtain a current binaryzation tensor and a power feature tensor;

[0037] stitching the initial noise tensor and the power feature tensor into a current input tensor;

[0038] performing forward propagation on the current input tensor to obtain a correction tensor;

[0039] correcting the initial noise tensor based on the correction tensor and the weight ratio until the iteration is performed until a simulation termination condition is met to obtain a corrected noise tensor;

[0040] obtaining a target bushing two-dimensional temperature field simulation result based on the current binarization tensor and the corrected noise tensor.

[0041] In an embodiment, the step of obtaining the target bushing two-dimensional temperature field simulation result based on the current binarization tensor and the corrected noise tensor comprises:

[0042] respectively broadcasting the current binarization tensor and the corrected noise tensor to three channels to obtain a first tensor and a second tensor;

[0043] performing logical operation and numerical scaling processing on the first tensor and the second tensor, replacing positions with a first numerical value in the three channels with a target numerical value to obtain the target bushing two-dimensional temperature field simulation result.

[0044] In an embodiment, the step of obtaining the plurality of power working condition parameters of the transformer bushing comprises:

[0045] determining an active power value range of the transformer bushing;

[0046] adjusting the active power value range to obtain a target power value range;

[0047] uniformly selecting a preset number of power sampling points in the target power value range to obtain the plurality of power working condition parameters of the transformer bushing.

[0048] In addition, to achieve the above-mentioned purpose, the present application also proposes a transformer bushing temperature rise simulation device, the transformer bushing temperature rise simulation device comprises:

[0049] an acquisition module configured to acquire a diffusion model, wherein the diffusion model is obtained by training a training data set;

[0050] The acquisition module is further configured to acquire a target bushing heat power parameter.

[0051] a simulation module configured to perform temperature field simulation based on the target bushing heat power parameter and the diffusion model to obtain a target bushing two-dimensional temperature field simulation result.

[0052] In addition, to achieve the above-mentioned purpose, the present application also proposes a transformer bushing temperature rise simulation device, the device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the transformer bushing temperature rise simulation method as described above.

[0053] In addition, to achieve the above object, the present application also provides a storage medium, which is a computer readable storage medium, and a computer program is stored on the storage medium, and the computer program realizes the steps of the transformer bushing temperature rise simulation method when executed by a processor.

[0054] In addition, to achieve the above object, the present application also provides a computer program product, which comprises a computer program, and the computer program realizes the steps of the transformer bushing temperature rise simulation method when executed by a processor.

[0055] The one or more technical solutions provided by the present application can obtain a diffusion model, wherein the diffusion model is obtained by training a training data set; obtain a target bushing heat power parameter; and perform temperature field simulation based on the target bushing heat power parameter and the diffusion model to obtain a target bushing two-dimensional temperature field simulation result, which can accurately reflect the temperature rise of the transformer bushing in the actual operation process. By combining the preset bushing heat power parameter with the diffusion model, the two-dimensional temperature field simulation result of the target bushing can be quickly obtained, and the cumbersome calculation process in the traditional numerical simulation method is avoided. The method can accurately reflect the temperature rise of the transformer bushing in the actual operation process, significantly reduces the simulation time, and improves the simulation efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0056] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and serve to explain the principles of the present application together with the specification.

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, for those skilled in the art, other drawings can also be obtained without creative labor.

[0058] Figure 1 A flowchart is provided for the transformer bushing temperature rise simulation method embodiment one of the present application;

[0059] Figure 2 A flowchart is provided for the transformer bushing temperature rise simulation method embodiment two of the present application;

[0060] Figure 3 A structure diagram of a diffusion network structure is provided for the transformer bushing temperature rise simulation method one embodiment of the present application;

[0061] Figure 4 A flowchart is provided for the transformer bushing temperature rise simulation method embodiment three of the present application;

[0062] Figure 5 Brief flowchart schematic diagram provided for an embodiment of the transformer bushing temperature rise simulation method of the present application;

[0063] Figure 6 Module structure schematic diagram of the transformer bushing temperature rise simulation device of the embodiment of the present application;

[0064] Figure 7 Device structure schematic diagram of the hardware running environment involved in the transformer bushing temperature rise simulation method in the embodiment of the present application.

[0065] The object implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0066] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application, and are not used to limit the present application.

[0067] In order to better understand the technical solutions of the present application, the specific embodiments will be described in detail below with reference to the drawings and the specific embodiments.

[0068] The main solution of the embodiment of the present application is: obtaining a plurality of power operating condition parameters of a transformer bushing, wherein each power operating condition parameter is uniformly distributed; performing bushing temperature field simulation and processing on each power operating condition parameter to obtain a training data set; constructing a diffusion network structure and training the diffusion network structure through the training data set to obtain a diffusion model; obtaining a target bushing heating power parameter; performing temperature field simulation based on the target bushing heating power parameter and the diffusion model to obtain a target bushing two-dimensional temperature field simulation result.

[0069] The existing technology mainly relies on finite volume simulation software, finite element simulation software, etc. for the temperature rise simulation of the transformer bushing. For specific operating condition parameters, a complete simulation process is performed. This will result in a large time consumption for each simulation, resulting in a very low efficiency when frequently changing the design.

[0070] The present application provides a solution. Through the data set composed of the two-dimensional bushing temperature field pre-calculated by the given uniformly distributed operating condition parameters, the diffusion model is used to splice the tensor of the operating condition parameters and the mixed noise data as the model input, and the mean square error of the model output and the known noise is used as the loss function for model training. Based on the above trained model, starting from random noise, based on the guidance of the given operating condition parameters, the noise components are repeatedly canceled out, and the two-dimensional temperature field distribution diagram of the transformer bushing under the given operating condition parameters can be obtained.

[0071] It should be noted that the execution subject of the embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, or an electronic device capable of realizing the above functions, a transformer bushing temperature rise simulation device, etc. The transformer bushing temperature rise simulation device is taken as an example to describe the embodiment and the following embodiments.

[0072] Based on this, the embodiment of the present application provides a transformer bushing temperature rise simulation method, which refers to Figure 1 , Figure 1 The flowchart of the first embodiment of the transformer bushing temperature rise simulation method of the present application is shown in the figure.

[0073] In the embodiment, the transformer bushing temperature rise simulation method includes steps S10-S30:

[0074] Step S10: Obtain a diffusion model, wherein the diffusion model is obtained by training a training data set.

[0075] It should be noted that the diffusion model is used to simulate the temperature rise of the transformer bushing, and the diffusion model is trained by the training data set collected in advance.

[0076] The training data set is a data set processed from the collected multiple power operating condition parameters of the transformer bushing.

[0077] In a possible implementation, step S10 can include steps A11-A13:

[0078] Step A11: Obtain multiple power operating condition parameters of the transformer bushing, wherein each power operating condition parameter is uniformly distributed.

[0079] It can be understood that the multiple power operating condition parameters of the transformer bushing, wherein each power operating condition parameter is uniformly distributed.

[0080] It should be noted that the number of power operating condition parameters can be set according to requirements, and the transformer is a uniformly shaped bushing. The power operating condition of the transformer bushing can be set in advance to obtain multiple power parameters of the transformer bushing. Specifically, when obtaining the power operating condition parameters, the parameters can be uniformly collected according to the power operating condition range to obtain uniformly distributed power operating condition parameters.

[0081] In a possible implementation, step A11 can include steps A111-A113:

[0082] Step A111: Determine the active power value range of the transformer bushing;

[0083] It can be understood that the active power value range of the transformer bushing can be determined first, and the active power is the part of the power consumed on the resistance element, which is not reversibly converted (such as converted into heat, light or mechanical energy), so the possible active power value range [P1, P2] of the transformer bushing can be determined first.

[0084] Step A112: adjusting the active power value range to obtain a target power value range;

[0085] In a specific implementation, the [P1, P2] can be adjusted to obtain a target power value range, and the target power value range is [max(0, 0.5*P1), 1.5*P2].

[0086] Step A113: uniformly selecting a preset number of power sampling points in the target power value range to obtain a plurality of power working condition parameters of the transformer bushing.

[0087] It can be understood that a preset number of power sampling points can be uniformly selected in the target power value range, and the preset number can be set according to requirements, for example, the preset number is 400, and 400 power sampling points are uniformly selected, so that the data corresponding to the power sampling points are used as the power working condition parameters of the transformer bushing.

[0088] Step A12: bushing temperature field simulation and processing of each power working condition parameter to obtain a training data set.

[0089] It can be understood that after obtaining 400 power working condition parameters, bushing temperature field simulation can be performed to obtain simulation calculation results, and the simulation calculation results are processed to obtain a training data set, which is a data set for subsequent model training.

[0090] In a feasible implementation, step A12 can include steps A121-A123:

[0091] Step A121: bushing temperature field simulation of each power working condition parameter to obtain an initial simulation result;

[0092] It should be noted that the bushing temperature field simulation can be performed on each power working condition parameter to obtain an initial simulation result, which reflects the temperature distribution of the transformer bushing under different power working conditions.

[0093] Step A122: based on the initial simulation result, a plane temperature distribution is intercepted on the axisymmetric plane of the bushing to obtain a preset number of two-dimensional temperature distribution maps;

[0094] In a specific implementation, based on the initial simulation results, a plane temperature distribution can be intercepted on the axisymmetric plane of the bushing, and the preset number can be set according to actual needs, and specifically corresponds to the number of power operating parameters described above, for example, 400 two-dimensional temperature distribution graphs are intercepted, which can more intuitively present the temperature conditions of a specific plane of the bushing under different power operating conditions.

[0095] Step A123: performing data enhancement processing on the preset number of two-dimensional temperature distribution graphs to obtain a training data set.

[0096] In a specific implementation, to improve the diversity of data and the stability of model training, the 400 two-dimensional temperature distribution graphs can be subjected to data enhancement processing such as rotation, flipping, scaling, etc., so as to expand the data set and increase the variability of the data, and finally obtain a training data set, which will be used for subsequent diffusion model training.

[0097] In a feasible implementation, step A123 can include: taking the bushing center line in the preset number of two-dimensional temperature distribution graphs as a reference, cropping the two-dimensional temperature distribution graphs into distribution graphs of a first pixel size; scaling the distribution graphs of the first pixel size into distribution graphs of a second pixel size; merging and quantizing color channels in the distribution graphs of the second pixel size into grayscale channels and standardizing them to a preset range to obtain a preset number of two-dimensional numerical tensors; and obtaining a training data set according to the preset number of two-dimensional numerical tensors.

[0098] It should be noted that for the above-mentioned 400 two-dimensional distribution graphs, the two-dimensional temperature distribution graphs can be cropped into distribution graphs of a first pixel size with the bushing center line in the graphs as a reference, and the first pixel size can be 474*2844 pixels. Then the distribution graphs of 474*2844 pixels can be scaled into distribution graphs of a second pixel size, and the second pixel size can be 128*768 pixels.

[0099] After obtaining the distribution graphs of the second pixel size, the color channels in the distribution graphs of the second pixel size can be merged and quantized into grayscale channels and standardized to a preset range, the color channels are red, green and blue three color channels, and the preset range is [0.0, 1.0], so that the red, green and blue three color channels are merged and quantized into one grayscale channel and standardized to the range [0.0, 1.0], and finally 400 two-dimensional numerical tensors of 128*768 are obtained, the values of the matrix range in [0.0, 1.0], so that the preset number of two-dimensional numerical tensors are used as a training data set.

[0100] Step A13: constructing a diffusion network structure and training the diffusion network structure through the training data set to obtain a diffusion model.

[0101] It should be noted that after obtaining the training data set, the diffusion network structure can be constructed, and the diffusion network structure is trained by the training data set to obtain the final diffusion model. The diffusion model is used for subsequent temperature field simulation of a given bushing heating power parameter to obtain a corresponding simulation result, thereby improving the efficiency and accuracy of the simulation. The diffusion network structure can receive input data and gradually remove noise to recover the original temperature field distribution.

[0102] It should be noted that the diffusion network structure is a specially designed and optimized diffusion network structure, which includes a constant weight part and a weight iterative updating part. By continuously updating the weight, the optimal diffusion model is obtained. Training the diffusion network structure includes word embedding operation, deconcatenation, binarization, convolution, residual, self-attention weighted summation, deconvolution, and tensor concatenation processes.

[0103] The diffusion network structure is trained by the training data set to adjust the network parameters, so that the model can accurately simulate the temperature rise of the transformer bushing under different power conditions. Finally, a trained diffusion model is obtained.

[0104] Step S20: Obtain a target bushing heating power parameter.

[0105] It can be understood that the target bushing heating power parameter is a given working condition parameter, which can be obtained according to user demand.

[0106] Step S30: Perform temperature field simulation based on the target bushing heating power parameter and the diffusion model to obtain a target bushing two-dimensional temperature field simulation result.

[0107] In specific implementation, the target bushing heating power parameter can be input into the diffusion model for temperature field simulation, thereby obtaining a target bushing two-dimensional temperature field simulation result corresponding to the target bushing heating power parameter.

[0108] The diffusion model will perform temperature field simulation according to the input power parameter and the internal learned temperature field distribution rule, and finally output a two-dimensional temperature field simulation result of the target bushing under a given power condition. This result can directly show the temperature rise distribution of the transformer bushing in the actual operation process, providing an important reference for engineering design and operation and maintenance.

[0109] The embodiment provides a transformer bushing temperature rise simulation method, obtains a diffusion model, wherein the diffusion model is obtained by training a training data set; obtains a target bushing heat power parameter; performs temperature field simulation based on the target bushing heat power parameter and the diffusion model to obtain a target bushing two-dimensional temperature field simulation result. Through the given bushing heat power parameter and the diffusion model, the two-dimensional temperature field simulation result of the target bushing can be quickly obtained, and the cumbersome calculation process in the traditional numerical simulation method is avoided. The method can accurately reflect the temperature rise of the transformer bushing in the actual operation process, significantly reduces the simulation time, and improves the simulation efficiency.

[0110] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as the above embodiment one can refer to the above introduction, and the following will not be repeated. On this basis, please refer to Figure 2 , step A13 includes steps A131-A134:

[0111] Step A131: constructing a diffusion network structure, wherein the diffusion network structure includes a weight constant part and a weight iterative updating part.

[0112] It should be noted that in the constructed diffusion network structure, the weight constant part is mainly responsible for processing the basic feature extraction of the input data. It performs preliminary transformation on the input data with fixed weight parameters to provide stable feature representation for subsequent processing. The weight iterative updating part adjusts its weight to adapt to the complex characteristics of the transformer bushing temperature field simulation under different power conditions. In the training process, the weight iterative updating part can gradually optimize the weight parameters according to the information of the training data set, so that the diffusion network structure can more accurately simulate the distribution law of the temperature field.

[0113] Step A132: obtaining a plurality of two-dimensional numerical tensors according to the training data set, and obtaining power condition parameters corresponding to the plurality of two-dimensional numerical tensors.

[0114] It should be noted that a plurality of two-dimensional numerical tensors can be obtained from the training data set. The two-dimensional numerical tensor is a bushing symmetric section temperature distribution diagram, which is scaled to a two-dimensional numerical tensor with a size of 128x768.

[0115] In specific implementation, each two-dimensional numerical tensor corresponds to a power condition parameter, so the power condition parameter corresponding to the two-dimensional numerical tensor can be obtained.

[0116] Step A133: taking the plurality of two-dimensional numerical tensors as network input images, and taking the power condition parameters as network input parameters.

[0117] It can be understood that multiple two-dimensional numerical tensors can be taken as network input images, and power working condition parameters can be taken as network input parameters.

[0118] Step A134: training the diffusion network structure by the network input images and the network input parameters to obtain a diffusion model.

[0119] It should be noted that by taking multiple two-dimensional numerical tensors as network input images and power working condition parameters as network input parameters, the diffusion network structure is input together. In the training process, the constant weight part of the diffusion network structure first extracts basic features from the input images to obtain stable feature representations. Then, the weight iterative updating part adjusts its own weight according to the input parameters and the basic features to adapt to the complex characteristics of the transformer bushing temperature field simulation under different power working conditions. With the training, the diffusion network structure gradually learns the distribution law of the temperature field and can accurately simulate the corresponding temperature field distribution according to the input power working condition parameters. Finally, after sufficient training, a diffusion model that can accurately simulate the temperature rise of the transformer bushing under different power working conditions is obtained.

[0120] In a possible implementation, step A134 can include steps C11-C16.

[0121] Step C11: performing word embedding operation on the network input parameters in the diffusion network structure to map the network input parameters into a first vector of a preset dimension, and performing broadcast operation on the first vector to expand it into a tensor structure of a preset size;

[0122] It should be noted that the network input parameters I2 can be subjected to word embedding operation in the diffusion network structure to be mapped into a first vector of a preset dimension, i.e., I2 is mapped into a first vector I3 of a size of 256 after word embedding operation, and then the first vector I3 is subjected to broadcast operation to be expanded into a tensor structure I4 of a preset size, and the preset size is 256x 768x 128, so that the first vector is expanded into a tensor structure I4 of 256x 768x 128.

[0123] Step C12: performing inverse stitching on the network input images to separate at least three channel numerical matrices;

[0124] It should be noted that the network input images I1 can be subjected to inverse stitching to separate a numerical matrix I5 of channel 1 (standardized to a range of [-1, 1]), a numerical matrix I6 of channel 2 (not standardized), and a numerical matrix of channel 3. The numerical matrix of channel 3 is directly discarded.

[0125] Step C13: performing binaryzation operation on the numerical matrix of the second channel to obtain a binaryzation tensor;

[0126] In practice, the numerical matrix I6 of the second channel can be binarized, updating all values ​​greater than 150 to 255 and all values ​​less than or equal to 150 to 0, thus obtaining the binarized tensor structure I7.

[0127] Step C14: Update the network weights using the numerical matrix of the first channel. During the weight update process, a noise tensor is randomly generated. The noise tensor is summed with the numerical matrix of the first channel according to the weight ratio and then concatenated with the tensor structure of the preset size to obtain the input tensor. The weight ratio changes linearly within the iteration rounds.

[0128] In practice, the numerical matrix I5 of the first channel participates in the network weight update. Each round of weight update involves two processes: forward propagation and backward propagation. For example, in the k-th round (out of K rounds) of forward propagation, a noise tensor N of size 768x128 is randomly generated according to a standard normal distribution. k , connect I5 and N k Summing according to weighted proportions (i.e., Then it is concatenated with a tensor structure I4 of a preset size to obtain an input tensor I8 with dimensions of 257x768x128, where the weight ratio, β, decreases from its minimum value β in K steps. min Linear change to the maximum value β max , (i.e., β) min +k / K·(β max -β min )).

[0129] Step C15: Perform convolution transformation, residual block processing, self-attention weighted summation, deconvolution transformation, and tensor concatenation on the input tensor in sequence to obtain the output tensor;

[0130] It should be noted that the input tensor can be sequentially subjected to convolution transformation, residual block processing, self-attention weighted summation, deconvolution transformation, and tensor concatenation to obtain the output tensor. The specific steps are as follows:

[0131] C151: After convolution transformation, I8 is transformed into tensor I9 with a size of 64 x 768 x 128;

[0132] C152: I9 is ​​transformed into a tensor I of size 64 x 768 x 128 after passing through the residual block. 10 (The residual block includes two convolutional layers. The first convolutional layer changes the number of channels if possible, and the second convolutional layer maintains the number of channels. Group normalization is performed before each convolutional layer. The residual blocks in this application all adopt this two-layer structure.)

[0133] C153: I10 Through residual block, transform into tensor I of 64 x 768 x 128 11 ;

[0134] C154: I 11 Through convolutional transform, transform into tensor I of 64 x 384 x 64 12 ;

[0135] C155: I 12 Through residual block, transform into tensor I of 128 x 384 x 64 13 ;

[0136] C156: I 13 Through residual block, transform into tensor I of 128 x 384 x 64 14 ;

[0137] C157: I 14 Through convolutional transform, transform into tensor I of 128 x 192 x 32 15 ;

[0138] C158: I 15 Through residual block, transform into tensor I of 256 x 192 x 32 16 ;

[0139] C159: I 16 Through residual block, transform into tensor I of 256 x 192 x 32 17 ;

[0140] C160: I 17 Through convolutional transform, transform into tensor I of 256 x 96 x 16 18 ;

[0141] C161: I 18 Through residual block, and weighted sum based on self-attention weight values, transform into tensor I of 512 x 96 x 16 19 ;

[0142] C162: I 19 Through residual block, and weighted sum based on self-attention weight values, transform into tensor I of 512 x 96 x 16 20 ;

[0143] C163: I 20 Through convolutional transform, transform into tensor I of 512 x 48 x 8 21 ;

[0144] C164: I 21Pass through residual block and weighted sum based on self-attention weight values, transforming to tensor I of 1024 x 48 x 8 22 ;

[0145] C165: I 22 Pass through residual block and weighted sum based on self-attention weight values, transforming to tensor I of 1024 x 48 x 8 23 ;

[0146] C166: I 23 Pass through residual block and weighted sum based on self-attention weight values, transforming to tensor I of 1024 x 48 x 8 24 ;

[0147] C167: I 24 Pass through convolutional transform, transforming to tensor I of 1024 x 48 x 8 25 ;

[0148] C168: I 25 Concatenate with I 23 Pass through residual block and weighted sum based on self-attention weight values, transforming to tensor I of 1024 x 48 x 8 26 ;

[0149] C169: I 26 Concatenate with I 22 Pass through residual block and weighted sum based on self-attention weight values, transforming to tensor I of 1024 x 48 x 8 27 ;

[0150] C170: I 27 Concatenate with I 21 Pass through residual block and weighted sum based on self-attention weight values, transforming to tensor I of 1024 x 48 x 8 28 ;

[0151] C171: I 28 Pass through deconvolutional transform, transforming to tensor I of 1024 x 96 x 16 29 ;

[0152] C172: I 29 Concatenate with I 20 Pass through residual block and weighted sum based on self-attention weight values, transforming to tensor I of 512 x 96 x 16 30 ;

[0153] C173: I 30 Concatenate with I 19The components are concatenated, then processed through residual blocks, and weighted and summed based on self-attention weights, transforming the result into a 512x96x16 tensor I. 31 ;

[0154] C174: I 31 with I 18 The components are concatenated, then processed through residual blocks, and weighted and summed based on self-attention weights, transforming the result into a 512x96x16 tensor I. 32 ;

[0155] C175: I 32 After deconvolution, it is transformed into a 512x192x32 tensor I. 33 ;

[0156] C176: I 33 with I 17 The components are concatenated and then processed through residual blocks, transforming the result into a tensor I of size 256 x 192 x 32. 34 ;

[0157] C177: I 34 with I 16 The components are concatenated and then processed through residual blocks, transforming the result into a tensor I of size 256 x 192 x 32. 35 ;

[0158] C178: I 35 with I 15 The components are concatenated and then processed through residual blocks, transforming the result into a tensor I of size 256 x 192 x 32. 36 ;

[0159] C179: I 36 After deconvolution, it is transformed into a 256x384x64 tensor I. 37 ;

[0160] C180: I 37 with I 14 The components are concatenated and then processed through residual blocks, transforming the result into a tensor I of size 128 x 384 x 64. 38 ;

[0161] C181: I 38 with I 13 The components are concatenated and then processed through residual blocks, transforming the result into a tensor I of size 128 x 384 x 64. 39 ;

[0162] C182: I 39 with I 12 The components are concatenated and then processed through residual blocks, transforming the result into a tensor I of size 128 x 384 x 64. 40 ;

[0163] C183: I 40 After deconvolution transformation, it is transformed into a 128x 768x 128 tensor I 41 ;

[0164] C184: I 41 Spliced with I 11 After the residual block, it is transformed into a 64x 768x 128 tensor I 42 ;

[0165] C185: I 42 Spliced with I 10 After the residual block, it is transformed into a 64x 768x 128 tensor I 43 ;

[0166] C186: I 43 Spliced with I9, and then after the residual block, it is transformed into a 64x 768x 128 tensor I 44 ;

[0167] C187: I 44 After convolution transformation, it is transformed into a 1x 768x 128 tensor I 45 .

[0168] Through the above steps of C151-C187, an output tensor I of 1x 768x 128 is obtained 45 .

[0169] Step C16: Calculate the mean square error of the output tensor and the target tensor, determine the error gradient according to the mean square error, and update the network weight and bias based on the error gradient through the optimization algorithm, and iterate until the training termination condition is met. Get the diffusion model.

[0170] In specific implementation, the target tensor is N k , the mean square error E k of the output tensor and the target tensor can be calculated, so as to determine the error gradient according to the mean square error, that is, according to E k , the gradient value in steps C14 to C15 is calculated.

[0171] In specific implementation, the optimization algorithm is an adaptive optimization algorithm (AdamW), which can update the weight and bias in steps C14 to C15 through the optimization algorithm. Through iterative updating, until the training termination condition is met. The training termination condition can be set to the number of iterations reaching a preset number of iterations, for example, the number of times of updating the network weight and bias reaches the preset number of iterations, then stop updating, and get the diffusion model.

[0172] As shown in Figure 3 Figure 3 is a structural diagram of a diffusion network structure, including convolutional layers, down-sampling layers, up-sampling layers, residual connection layers, self-attention layers, hidden layers, embedding layers, broadcast layers, and splicing / anti-splicing layers. Through convolution, down-sampling, up-sampling, residual connection, self-attention weighted summation, feature extraction, embedding, broadcast, splicing / anti-splicing, binarization, bitwise logical operation, and jump connection processing on network input parameters and network input images, weight updates are continuously performed, and finally a diffusion model is obtained.

[0173] The embodiment constructs a diffusion network structure, wherein the diffusion network structure includes a weight constant part and a weight iterative update part; a plurality of two-dimensional numerical tensors are obtained according to the training data set, and power working condition parameters corresponding to the plurality of two-dimensional numerical tensors are obtained; the plurality of two-dimensional numerical tensors are taken as network input images, and the power working condition parameters are taken as network input parameters; the diffusion network structure is trained through the network input images and the network input parameters to obtain a diffusion model. By continuously adjusting the weight parameters, the prediction result of the diffusion network structure gradually approaches the true label, thereby improving the accuracy and generalization ability of the model. After multiple iterative training, the diffusion network structure can learn the temperature field distribution rule of the transformer bushing under different power working conditions, and obtain a trained diffusion model, thereby improving the efficiency and accuracy of bushing temperature rise simulation.

[0174] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar contents as the above-mentioned first embodiment can be referred to the above introduction, and will not be described in detail. On this basis, please refer to Figure 4 , step S30 includes steps S301-S306:

[0175] Step S301: generating an initial noise tensor.

[0176] It should be noted that before inputting the target bushing heat power parameter into the diffusion model, an initial noise tensor needs to be generated, which serves as the starting point of the diffusion process and provides a basis for subsequent step-by-step denoising and reconstruction. This initial noise tensor usually has similar dimensions and structure as the target output (i.e., transformer bushing temperature rise distribution), but contains random noise to simulate the initial state of the unknown temperature field. By introducing initial noise, the diffusion model can learn how to gradually restore the true temperature field distribution from the disordered noise state during the training process.

[0177] Specifically, an initial noise tensor N' of 1x 768x 128 can be randomly generated based on a standard normal distribution g When the iteration number k = 1, it is a randomly generated N' g ​The noise tensor is inherited from subsequent updates in subsequent loops.

[0178] Step S302: Input the initial noise tensor and the target sleeve heating power parameters into the diffusion model to perform temperature field simulation, and obtain the current binarized tensor and power feature tensor.

[0179] It should be noted that the initial noise tensor and the target sleeve heating power parameters can be input into the diffusion model for temperature field simulation. Specifically, the target sleeve heating power parameters can be gradually denoised and feature extracted through forward and backward propagation in the diffusion model, which can accurately map the relationship between the input target sleeve heating power parameters and the output temperature field, and obtain the current binarized tensor I′7.

[0180] By inputting the initial noise tensor and the target sleeve heating power parameters into the diffusion model for temperature field simulation, the same embedding weights and broadcasting operations as in step C11 above are performed on the preset heating power parameters, embedding them into a 256-dimensional vector, and then, after broadcasting operations, expanding them into a power feature tensor structure I′4 with a size of 256x768x128.

[0181] Step S303: Concatenate the initial noise tensor and the power feature tensor to form the current input tensor.

[0182] In practical implementation, the initial noise tensor and the power feature tensor can be concatenated, and N′ can be used to... g The current input tensor I′8 is concatenated with I′4 to form a tensor of 257x768x128.

[0183] Step S304: Perform forward propagation on the current input tensor to obtain the corrected tensor.

[0184] Understandably, after obtaining the current input tensor I′8, forward propagation can be performed by referring to steps C151-C187 above, ultimately obtaining the 1x768x128 corrected tensor I′. 45 .

[0185] Step S305: Correct the initial noise tensor based on the corrected tensor and weight ratio until the simulation termination condition is met, and obtain the corrected noise tensor.

[0186] In practical implementation, the modified tensor I′ can be used. 45 and the weight ratio β with respect to the initial noise tensor N′ g Make corrections, that is Wherein, β changes linearly from its maximum value (e.g., 1e-2) to its minimum value (e.g., 1e-4) within K steps. K is a preset threshold for the number of iterations.

[0187] If the iteration number k < K, k = k + 1, and return to step S501 until k = K, and then obtain the revised noise tensor.

[0188] Step S306: obtaining a target sleeve two-dimensional temperature field simulation result based on the current binarization tensor and the revised noise tensor.

[0189] It should be noted that the current binarization tensor I'7 and the revised noise tensor N'7 can be broadcasted (copied) to three channels respectively. g The processing includes channel expansion, logical operation and numerical scaling processing, so as to obtain the target sleeve two-dimensional temperature field simulation result.

[0190] After multiple iterations and post-processing, the model generates a temperature field distribution matching the target sleeve heat power parameter. This result can be used to evaluate the temperature rise characteristics of the sleeve under different working conditions, and provide key data support for transformer design and operation. By comparing the simulation result with the actual measurement data, the accuracy and reliability of the model can be further verified.

[0191] In a feasible implementation, step S306 can include steps D11-D12:

[0192] Step D11: respectively broadcasting the current binarization tensor and the revised noise tensor to three channels to obtain a first tensor and a second tensor;

[0193] It should be noted that the current binarization tensor I'7 and the revised noise tensor N'7 can be broadcasted (copied) to three channels respectively. g The current binarization tensor I'7 is broadcasted to three channels to become a first tensor I 50 , and the revised noise tensor N'7 is broadcasted to three channels (1 channel is copied, 2 channel is 1-N'7, and 3 channel is 0*N'7) to become a second tensor I g . g g 51

[0194] Step D12: performing logical operation and numerical scaling processing on the first tensor and the second tensor, replacing the positions where all three channels are the first value with a target value, to obtain the target sleeve two-dimensional temperature field simulation result.

[0195] In specific implementation, the first tensor I 50 and the second tensor I 51 can be subjected to logical operation and numerical scaling processing, scaled to the range of [0, 255], and the positions where all three channels are the first value are replaced with a target value, the first value being 0 and the target value being 255, so as to obtain I 52 , I 52 ​​​The target sleeve two-dimensional temperature field simulation result is a simulated sleeve two-dimensional temperature field.

[0196] The embodiment can more accurately obtain the current binary tensor and the power feature tensor by generating the initial noise tensor and combining the target sleeve heat power parameter to input the diffusion model for temperature field simulation, thereby providing reliable data basis for subsequent temperature field analysis. The initial noise tensor and the power feature tensor are spliced into the current input tensor and forward propagation is performed to obtain the correction tensor. Then, the initial noise tensor is corrected based on the correction tensor and the weight proportion until the simulation termination condition is met, and the corrected noise tensor is obtained. This process can continuously optimize the simulation result and improve the accuracy and reliability of the simulation. Finally, the target sleeve two-dimensional temperature field simulation result obtained based on the current binary tensor and the corrected noise tensor can more accurately reflect the temperature distribution of the sleeve, thereby providing strong support for the design, optimization and operation of the sleeve.

[0197] For example, in order to help understand the implementation process of the transformer sleeve temperature rise simulation method obtained after the above embodiment one, please refer to Figure 5 , Figure 5 A brief flowchart of a transformer sleeve temperature rise simulation method is provided, specifically: the input is the basic heat power range of the sleeve, the power is discretized, numerical simulation is performed based on the conventional finite volume method, thereby obtaining a training data set, word embedding operation is performed on the potential representation of the working condition parameter (heat power) set, an extended tensor structure is obtained, data set A and data set B are obtained by data augmentation of the training data set, data set A is used for model training, data set A is input into a specially designed network for structure, and random noise and the extended tensor structure are input for model training. The diffusion model is obtained by continuously updating the weight after reaching the predetermined number of iterations. If the predetermined number of iterations is not reached, the output tensor is obtained, and the random noise and the given heat power are input again to obtain the two-dimensional tensor intermediate data representing the temperature field. The weight is continuously optimized until the number of iterations reaches the predetermined number of iterations. Data set B is the current given target sleeve heat power parameter. The initial noise tensor and the target sleeve heat power parameter are processed by the diffusion model, thereby obtaining the final target sleeve two-dimensional temperature field simulation result, that is, the output sleeve temperature field simulation two-dimensional image.

[0198] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the transformer sleeve temperature rise simulation method of the present application. More forms of simple transformation based on this technical concept are within the protection scope of the present application.

[0199] The present application also provides a transformer sleeve temperature rise simulation device, please refer to Figure 6The transformer bushing temperature rise simulation device comprises:

[0200] The acquisition module 10 is configured to acquire a diffusion model.

[0201] The acquisition module 10 is further configured to acquire a target bushing heat power parameter.

[0202] The simulation module 20 is configured to perform temperature field simulation based on the target bushing heat power parameter and the diffusion model to obtain a target bushing two-dimensional temperature field simulation result.

[0203] The transformer bushing temperature rise simulation device provided in the present application adopts the transformer bushing temperature rise simulation method in the above embodiments, and can solve the technical problem of low simulation efficiency of the transformer bushing temperature rise at present. Compared with the prior art, the transformer bushing temperature rise simulation device provided in the present application has the same beneficial effects as the transformer bushing temperature rise simulation method provided in the above embodiments, and other technical features in the transformer bushing temperature rise simulation device are the same as the features disclosed in the above embodiments, which will not be described herein.

[0204] In an embodiment, the acquisition module 10 is further configured to acquire a plurality of power working condition parameters of the transformer bushing, wherein each power working condition parameter is uniformly distributed; perform bushing temperature field simulation and processing on each power working condition parameter to obtain a training data set; construct a diffusion network structure, and train the diffusion network structure through the training data set to obtain a diffusion model.

[0205] In an embodiment, the acquisition module 10 is further configured to perform bushing temperature field simulation on each power working condition parameter to obtain an initial simulation result; obtain a preset number of two-dimensional temperature distribution maps by intercepting a plane temperature distribution on an axisymmetric plane of the bushing based on the initial simulation result; and perform data enhancement processing on the preset number of two-dimensional temperature distribution maps to obtain a training data set.

[0206] In an embodiment, the acquisition module 10 is further configured to take the bushing center line in the preset number of two-dimensional temperature distribution maps as a reference to crop the two-dimensional temperature distribution maps into a distribution map of a first pixel size; scale the distribution map of the first pixel size into a distribution map of a second pixel size; merge and quantize color channels in the distribution map of the second pixel size into grayscale channels and standardize them to a preset range to obtain a preset number of two-dimensional numerical tensors; and obtain a training data set according to the preset number of two-dimensional numerical tensors.

[0207] In an embodiment, the obtaining module 10 is further configured to construct a diffusion network structure, wherein the diffusion network structure comprises a constant weight part and an iterative weight updating part; obtain a plurality of two-dimensional numerical tensors from the training data set, and obtain power working condition parameters corresponding to the plurality of two-dimensional numerical tensors; take the plurality of two-dimensional numerical tensors as network input images, and take the power working condition parameters as network input parameters; train the diffusion network structure by using the network input images and the network input parameters, and obtain a diffusion model.

[0208] In an embodiment, the obtaining module 10 is further configured to perform word embedding operation on the network input parameters in the diffusion network structure, map the network input parameters to a first vector of a preset dimension, and perform broadcast operation on the first vector to expand it to a tensor structure of a preset size; perform inverse splicing on the network input images to separate at least three channel numerical matrices; perform binaryzation operation on a second channel numerical matrix to obtain a binaryzation tensor; perform network weight updating by using a first channel numerical matrix, and in the weight updating process, randomly generate a noise tensor, sum the noise tensor and the first channel numerical matrix according to a weight ratio, and splice the sum with the tensor structure of the preset size to obtain an input tensor, wherein the weight ratio linearly changes in an iteration round; sequentially perform convolution transformation, residual block processing, self-attention weighted summation, inverse convolution transformation and tensor splicing on the input tensor to obtain an output tensor; calculate a mean square error of the output tensor and a target tensor, determine an error gradient according to the mean square error, and update network weights and biases by using an optimization algorithm based on the error gradient, and iteratively execute until a training termination condition is met to obtain the diffusion model.

[0209] In an embodiment, the simulation module 20 is further configured to generate an initial noise tensor; input the initial noise tensor and the target sleeve heat power parameter to the diffusion model to perform temperature field simulation, and obtain a current binaryzation tensor and a power feature tensor; splice the initial noise tensor and the power feature tensor to obtain a current input tensor; perform forward propagation on the current input tensor to obtain a modified tensor; modify the initial noise tensor based on the modified tensor and a weight ratio until an iteration is performed until a simulation termination condition is met to obtain a modified noise tensor; and obtain a target sleeve two-dimensional temperature field simulation result based on the current binaryzation tensor and the modified noise tensor.

[0210] In an embodiment, the simulation module 20 is further configured to broadcast the current binaryzation tensor and the modified noise tensor to three channels respectively to obtain a first tensor and a second tensor; perform logical operation and numerical scaling processing on the first tensor and the second tensor, replace positions with a first numerical value in the three channels with a target numerical value to obtain the target sleeve two-dimensional temperature field simulation result.

[0211] In an embodiment, the obtaining module 10 is further configured to determine an active power value range of the transformer bushing; adjust the active power value range to obtain a target power value range; and uniformly select a preset number of power sampling points in the target power value range to obtain a plurality of power operating condition parameters of the transformer bushing.

[0212] The present application provides a transformer bushing temperature rise simulation device, which comprises at least one processor and a memory in communication connection with the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the transformer bushing temperature rise simulation method in Embodiment I.

[0213] Reference will be made to the following description of the embodiments of the present application. Figure 7 which shows a structural diagram of a transformer bushing temperature rise simulation device suitable for implementing the embodiments of the present application. The transformer bushing temperature rise simulation device in the embodiments of the present application can include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), and vehicle terminal (e.g., vehicle navigation terminal) and fixed terminals such as digital TVs and desktop computers. Figure 7 The transformer bushing temperature rise simulation device shown is merely an example and should not impose any limitation on the functions and use range of the embodiments of the present application.

[0214] As Figure 7As shown, the transformer bushing temperature rise simulation device can include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.) that can perform various appropriate actions and processes according to programs stored in a ROM (Read Only Memory) 1002 or programs loaded from a storage device 1003 into a RAM (Random Access Memory) 1004. Various programs and data required for operation of the transformer bushing temperature rise simulation device are also stored in the RAM 1004. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, an LCD (Liquid Crystal Display), a speaker, a vibrator, etc.; the storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the transformer bushing temperature rise simulation device to communicate with other devices wirelessly or by wire to exchange data. Although the transformer bushing temperature rise simulation device with various systems is shown in the figure, it should be understood that all the shown systems are not required to be implemented or possessed. More or fewer systems can be alternatively implemented or possessed.

[0215] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing program codes for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network through the communication device, or installed from the storage device 1003, or installed from the ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the methods of the embodiments disclosed in the present application are performed.

[0216] The transformer bushing temperature rise simulation device provided by the present application adopts the transformer bushing temperature rise simulation method in the above-mentioned embodiments, and can solve the technical problem of low simulation efficiency of the current transformer bushing temperature rise simulation. Compared with the prior art, the transformer bushing temperature rise simulation device provided by the present application has the same beneficial effects as the transformer bushing temperature rise simulation method provided by the above-mentioned embodiments, and other technical features in the transformer bushing temperature rise simulation device are the same as the features disclosed in the previous embodiment method, which will not be described here.

[0217] It should be understood that various aspects of the disclosure can be implemented in hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any appropriate manner in any one or more embodiments or examples.

[0218] The above description is merely illustrative of the application and is not intended to limit the scope of the application. Any variations and modifications that can be made by any person skilled in the art within the spirit and scope of the application are intended to be encompassed by the application. The scope of the application is defined by the appended claims.

[0219] The application provides a computer readable storage medium having computer readable program instructions (i.e. computer programs) stored thereon, the computer readable program instructions being used to execute the transformer bushing temperature rise simulation method in the above embodiments.

[0220] The computer readable storage medium provided by the application may, for example, be a U disk, but is not limited to an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system, system or device, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electric connection with one or more conductive wires, a portable computer disk, a hard disk, a RAM (Random Access Memory), a ROM (Read Only Memory), an EPROM (Erasable Programmable Read Only Memory or flash memory), an optical fiber, a CD-ROM (CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present embodiment, the 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, system or device. The program code contained on the computer readable storage medium can be transmitted by any appropriate medium, including but not limited to an electric wire, an optical cable, an RF (Radio Frequency), etc., or any suitable combination of the above.

[0221] The above computer readable storage medium can be contained in the transformer bushing temperature rise simulation device; or can exist separately and not be assembled into the transformer bushing temperature rise simulation device.

[0222] The computer readable storage medium described above carries one or more programs, when the one or more programs are executed by the transformer bushing temperature rise simulation device, the transformer bushing temperature rise simulation device: obtains a diffusion model, wherein the diffusion model is obtained by training a training data set; obtains a target bushing heat power parameter; performs temperature field simulation based on the target bushing heat power parameter and the diffusion model to obtain a target bushing two-dimensional temperature field simulation result.

[0223] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0224] The flow diagrams and the block diagrams in the drawings are illustrations of possible architectures, functions, and operations for systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams or block diagrams can represent a module, a segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks may

[0225] The modules involved in the embodiments of the present application can be implemented in a software manner or in a hardware manner. In some cases, the name of the module does not constitute a limitation on the module itself.

[0226] The readable storage medium provided by the application is a computer readable storage medium, and the computer readable storage medium stores computer readable program instructions (namely, a computer program) for executing the transformer bushing temperature rise simulation method, and can solve the technical problem of low efficiency of the current transformer bushing temperature rise simulation. Compared with the prior art, the computer readable storage medium provided by the application has the same beneficial effects as the transformer bushing temperature rise simulation method provided by the above-mentioned embodiments, and will not be repeated here.

[0227] The application further provides a computer program product comprising a computer program, which, when executed by a processor, implements the steps of the transformer bushing temperature rise simulation method as described above.

[0228] The computer program product provided by the application can solve the technical problem of low efficiency of the current transformer bushing temperature rise simulation. Compared with the prior art, the computer program product provided by the application has the same beneficial effects as the transformer bushing temperature rise simulation method provided by the above-mentioned embodiments, and will not be repeated here.

[0229] The above-mentioned is only part of the embodiments of the application, and does not limit the patent scope of the application, and any equivalent structural transformation, direct / indirect application in other related technical fields under the technical concept of the application, or the content of the specification and the drawings of the application are included in the patent protection scope of the application.

Claims

1. A method of transformer bushing temperature rise simulation, characterized by, The transformer bushing temperature rise simulation method comprises: obtaining a diffusion model, wherein the diffusion model is obtained by training a training data set; obtaining a target bushing heat power parameter; based on the target bushing heat power parameter and the diffusion model, temperature field simulation is carried out to obtain a target bushing two-dimensional temperature field simulation result.

2. The method of claim 1, wherein, The step of obtaining the diffusion model comprises: obtaining a plurality of power operating condition parameters of the transformer bushing, wherein each power operating condition parameter is uniformly distributed; carrying out bushing temperature field simulation and processing on each power operating condition parameter to obtain a training data set; constructing a diffusion network structure and training the diffusion network structure by the training data set to obtain a diffusion model.

3. The method of claim 2, wherein, The step of carrying out bushing temperature field simulation and processing on each power operating condition parameter to obtain a training data set comprises: carrying out bushing temperature field simulation on each power operating condition parameter to obtain an initial simulation result; based on the initial simulation result, a plane temperature distribution is obtained on the axial symmetric plane of the bushing, and a preset number of two-dimensional temperature distribution maps are obtained; data enhancement processing is carried out on the preset number of two-dimensional temperature distribution maps to obtain a training data set.

4. The method of claim 2, wherein, The step of carrying out data enhancement processing on the preset number of two-dimensional temperature distribution maps to obtain a training data set comprises: taking the bushing center line in the preset number of two-dimensional temperature distribution maps as a reference, the two-dimensional temperature distribution maps are cropped into distribution maps of a first pixel size; the distribution maps of the first pixel size are scaled into distribution maps of a second pixel size; color channels in the distribution maps of the second pixel size are merged and quantized into grayscale channels and standardized to a preset range to obtain a preset number of two-dimensional numerical tensors; a training data set is obtained according to the preset number of two-dimensional numerical tensors.

5. The method of claim 1, wherein, The step of constructing a diffusion network structure and training the diffusion network structure by the training data set to obtain a diffusion model comprises: constructing a diffusion network structure, wherein the diffusion network structure comprises a weight constant part and a weight iterative updating part; a plurality of two-dimensional numerical tensors are obtained according to the training data set, and a plurality of power operating condition parameters corresponding to the two-dimensional numerical tensors are obtained; the plurality of two-dimensional numerical tensors are taken as network input images, and the power operating condition parameters are taken as network input parameters; the diffusion network structure is trained by the network input images and the network input parameters to obtain a diffusion model.

6. The method of claim 5, wherein, The step of training the diffusion network structure by the network input images and the network input parameters to obtain a diffusion model comprises: word embedding operation is performed on the network input parameters in the diffusion network structure, the network input parameters are mapped into a first vector of a preset dimension, and the first vector is expanded into a tensor structure of a preset size by broadcast operation; the network input images are de-stitched to separate at least three channel numerical matrices; a binaryzation operation is performed on the second channel numerical matrix to obtain a binaryzation tensor; The network weight value is updated through the numerical matrix of the first channel. In the weight value updating process, a noise tensor is randomly generated, the noise tensor is summed with the numerical matrix of the first channel according to a weight ratio, and the input tensor is obtained after splicing the summed tensor with the preset size tensor structure, wherein the weight ratio linearly changes in the iteration round. The input tensor is sequentially subjected to convolution transformation, residual block processing, self-attention weighted summation, inverse convolution transformation and tensor splicing to obtain an output tensor. The mean square error of the output tensor and a target tensor is calculated, the error gradient is determined according to the mean square error, and the network weight value and bias are updated by an optimization algorithm based on the error gradient. The iteration is performed until the training termination condition is met to obtain a diffusion model.

7. The method of claim 1, wherein, The step of simulating the temperature field based on the target sleeve heat power parameter and the diffusion model to obtain the target sleeve two-dimensional temperature field simulation result comprises: generating an initial noise tensor; inputting the initial noise tensor and the target sleeve heat power parameter into the diffusion model to simulate the temperature field, to obtain a current binary tensor and a power feature tensor; splicing the initial noise tensor and the power feature tensor into a current input tensor; forward propagating the current input tensor to obtain a modified tensor; modifying the initial noise tensor based on the modified tensor and a weight ratio until the iteration is performed until the simulation termination condition is met to obtain a modified noise tensor; obtaining the target sleeve two-dimensional temperature field simulation result based on the current binary tensor and the modified noise tensor.

8. The method of claim 7, wherein, The step of obtaining the target sleeve two-dimensional temperature field simulation result based on the current binary tensor and the modified noise tensor comprises: broadcasting the current binary tensor and the modified noise tensor to three channels respectively to obtain a first tensor and a second tensor; performing logical operation and numerical scaling processing on the first tensor and the second tensor, replacing positions with a first numerical value in the three channels with a target numerical value to obtain the target sleeve two-dimensional temperature field simulation result.

9. The method of any one of claims 2 to 6, wherein, The step of obtaining a plurality of power working condition parameters of the transformer sleeve comprises: determining an active power value range of the transformer sleeve; adjusting the active power value range to obtain a target power value range; uniformly selecting a preset number of power sampling points in the target power value range to obtain a plurality of power working condition parameters of the transformer sleeve.

10. A transformer bushing temperature rise simulation device, characterized by, The device comprises: an acquisition module configured to acquire a diffusion model, wherein the diffusion model is obtained by training a training data set; the acquisition module is further configured to acquire a target sleeve heat power parameter; a simulation module configured to simulate a temperature field based on the target sleeve heat power parameter and the diffusion model to obtain a target sleeve two-dimensional temperature field simulation result.

Images