An ultra-high voltage GIS internal electric field probe intervention interference optimization method and system

Abstract

The application discloses an ultra-extra-high-voltage GIS internal electric field probe intervention interference optimization method and system, relates to the field of power equipment state monitoring and fault diagnosis, inputs ultra-extra-high-voltage GIS parameters into an electric field prediction model for prediction to obtain an electric field distribution before probe intervention; obtains an intervention electric field prediction model after migration learning and fine-tuning of the electric field prediction model, and further predicts the electric field distribution after probe intervention at each probe arrangement candidate point; based on the electric field distribution before probe intervention and the electric field distribution after probe intervention, the interference intensity of each probe arrangement candidate point is calculated; based on the minimization of electric field interference intensity and the minimization of the original electric field distribution uneven coefficient, a multi-objective optimization problem is constructed, a multi-objective genetic algorithm is used to solve the multi-objective optimization problem, and the parameters of the optimal probe arrangement point are obtained. The method effectively reduces the threat of probe intervention to the insulation safety of GIS, improves the accuracy and reliability of electric field measurement, and provides a scheme for state monitoring of GIS equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power equipment state monitoring and fault diagnosis, and particularly relates to a method and system for optimizing intervention interference of an internal electric field probe of an ultra and extra high voltage GIS. BACKGROUND

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute the prior art.

[0003] The ultra high voltage gas insulated switchgear (GIS) is a key device in the power system and is crucial to the safe and stable operation of the power system. The internal electric field distribution of the GIS is complex, and the strength thereof directly affects the insulation performance and operation reliability of the device. Therefore, accurate measurement of the internal electric field parameters of the GIS is of great significance for evaluating the insulation state of the device, achieving state monitoring and fault warning.

[0004] The internal environment of the ultra and extra high voltage GIS is a very strong closed electric field environment, and the intervention of the electric field probe will inevitably cause distortion of the original electric field, which may threaten the insulation safety of the GIS. Therefore, it is necessary to seek the optimal arrangement of the probe to minimize the intervention interference thereof. In addition, the sensitivity of the existing probe is usually calibrated under a uniform electric field, while in actual application, the distorted electric field formed after the intervention of the probe and the fixing member thereof is completely different from the calibration environment, and the measurement sensitivity of the probe will inevitably shift, and the measurement accuracy is difficult to guarantee.

[0005] In the prior art, static electric field simulation calculation is mostly relied on the finite element method, and there is a lack of optimization of the intervention interference of the internal electric field measurement sensor probe of the GIS. In terms of optimizing the arrangement of the probe to reduce the interference, the existing methods mostly rely on the experience of engineers or perform limited trial simulation, and lack a systematic and automated optimization framework. At the same time, the existing technology often only focuses on a single target, which easily leads to the fact that the final scheme cannot effectively guarantee the accuracy of the electric field measurement. SUMMARY

[0006] In order to overcome the deficiencies of the prior art, the present application provides a method and system for optimizing intervention interference of an internal electric field probe of an ultra and extra high voltage GIS, which replaces the traditional finite element method with a deep neural network to obtain the electric field distribution before and after the intervention of the probe, improves the calculation efficiency, and then determines the optimal arrangement scheme of the probe based on multi-objective optimization.

[0007] To achieve the above object, one or more embodiments of the present application provide the following technical solutions:

[0008] In a first aspect, the present application provides a method for optimizing intervention interference of an internal electric field probe of an ultra and extra high voltage GIS, comprising:

[0009] Obtaining the parameters of the GIS, inputting the parameters of the GIS into the electric field prediction model to obtain the electric field distribution before the probe intervention;

[0010] The electric field prediction model is obtained by transferring learning and fine-tuning the electric field prediction model based on the probe parameters, and the electric field distribution after the probe intervention is predicted based on the electric field prediction model.

[0011] Based on the electric field distribution before the probe intervention and the electric field distribution after the probe intervention, the interference intensity of each probe arrangement candidate point is calculated.

[0012] Based on the minimization of the electric field interference intensity and the minimization of the original electric field distribution non-uniformity coefficient, a multi-objective optimization problem is constructed, and a multi-objective genetic algorithm is used to solve the multi-objective optimization problem to obtain the parameters of the optimal probe arrangement point.

[0013] Further technical solutions, the parameters of the GIS include the cone angle of the basin type insulator, the umbrella skirt height, and the high voltage conductor end curvature radius.

[0014] Further technical solutions, the electric field prediction model adopts a generative neural network with an encoder-decoder structure, and a multi-objective loss function is designed, including a main reconstruction loss and a gradient consistency loss.

[0015] Further technical solutions, the probe parameters include the probe position, the probe size, and the relative dielectric constant.

[0016] Further technical solutions, the interference intensity is represented as:

[0017]

[0018]

[0019]

[0020] wherein, represents the probe intervention distortion coefficient, represents the spatial electric field non-uniformity coefficient after the probe intervention, represents the interference intensity, represents the weight coefficient matrix, represents the spatial electric field average value after the probe intervention, represents the spatial electric field maximum value after the probe intervention, represents the spatial electric field maximum value.

[0021] Further technical solutions, an arc-shaped fixing member is designed for the probe, the arc surface of the arc-shaped fixing member is consistent with the curvature of the inner wall of the GIS shell, the front surface is provided with a probe mounting groove, and the probe mounting groove is used to accommodate the sensing probe.

[0022] Further technical solutions, based on the optimal probe arrangement point parameter installation probe, the measurement sensitivity of the installed probe is calibrated, specifically:

[0023] In the three-dimensional model of the ultra-high voltage GIS, based on the parameter of the optimal probe arrangement point, the arc-shaped fixed part with the probe is introduced, and a sensitivity calibration simulation model is obtained;

[0024] In the sensitivity calibration simulation model, a high voltage electrode and a ground electrode are set, the electric field intensity at the probe intervention site and the signal output value of the probe are calculated;

[0025] Based on the electric field intensity and the signal output value, the sensitivity of the probe is calculated, and the sensitivity calibration of the probe is realized.

[0026] In a second aspect, the present application provides an ultra-high voltage GIS internal electric field probe intervention interference optimization system, comprising:

[0027] The original electric field prediction module is configured to: obtain the parameters of the ultra-high voltage GIS, input the GIS parameters into the electric field prediction model for prediction, and obtain the electric field distribution before the probe intervention;

[0028] The intervention electric field prediction module is configured to: based on the probe parameters, the electric field prediction model is migrated and fine-tuned to obtain an intervention electric field prediction model, and based on the intervention electric field prediction model, the electric field distribution after the probe intervention at each probe arrangement candidate point is predicted;

[0029] The interference intensity calculation module is configured to: based on the electric field distribution before the probe intervention and the electric field distribution after the probe intervention, the interference intensity of each probe arrangement candidate point is calculated;

[0030] The interference optimization module is configured to: based on the minimization of the electric field interference intensity and the minimization of the original electric field distribution non-uniformity coefficient, a multi-objective optimization problem is constructed, a multi-objective genetic algorithm is used to solve the multi-objective optimization problem, and the parameters of the optimal probe arrangement point are obtained.

[0031] In a third aspect, the present application provides a computer readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the ultra-high voltage GIS internal electric field probe intervention interference optimization method according to the first aspect.

[0032] In a fourth aspect, the present application provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the steps of the ultra-high voltage GIS internal electric field probe intervention interference optimization method according to the first aspect.

[0033] The one or more technical solutions have the following beneficial effects:

[0034] The application innovatively uses an electric field prediction model based on deep learning to replace the traditional finite element method, realizes fast and intelligent prediction of electric field distribution, and through migration learning and fine-tuning strategies, quickly adapts and constructs a high-precision intervention electric field prediction model based on the existing GIS ontology electric field prediction model, which can accurately predict the electric field distortion caused by the probe intervention, and provide basic data for intervention interference optimization; and the probe arrangement problem is constructed as a multi-objective optimization problem with the core of minimizing the electric field interference intensity and maximizing the field intensity sensitivity, and a multi-objective genetic algorithm is used for solving. This method can comprehensively weigh the two key requirements of insulation safety and measurement accuracy, and obtain a series of Pareto optimal solution sets from global search, thereby helping to select the most suitable point arrangement parameters for actual engineering requirements, and avoiding the one-sidedness of single-objective optimization. Through the above optimization, the threat of probe intervention to GIS insulation safety is effectively reduced, and the accuracy and reliability of electric field measurement are improved, providing a solution for the state monitoring of GIS equipment.

[0035] The application scientifically evaluates the advantages and disadvantages of different point arrangement schemes through the quantitative index of interference intensity, and considers the GIS spatial electric field average value, spatial electric field maximum value, spatial electric field distortion coefficient after probe intervention, and spatial electric field non-uniformity coefficient after probe intervention when calculating the intervention interference intensity, wherein the spatial electric field maximum value and the spatial electric field distortion coefficient after probe intervention are closely related to the GIS insulation safety, and the threat of probe intervention to the GIS insulation safety is maximally reduced while ensuring the measurement accuracy, thereby providing a clear basis for interference optimization.

[0036] The application also proposes a method for installing the electric field probe at the optimal point, designs a cambered surface fixing part, the cambered surface of the cambered surface fixing part is attached to the inner wall of the GIS, the electric field probe is placed in the groove on the other surface of the fixing part, and the inner wall of the GIS and the cambered surface of the fixing part and the electric field probe and the fixing part are fixed by epoxy glue, the geometric structure and material of the cambered surface fixing part are innovatively considered to affect the electric field at the optimal point, and the influence of the fixing part on the spatial electric field is maximally reduced on the basis of reliably installing the probe.

[0037] The application also proposes a sensitivity calibration method for the electric field probe under the intervention distortion electric field in the GIS, a simulation model including the inner wall of the GIS and the probe fixing part is constructed, the probe is calibrated under the intervention distortion electric field, the electric field distortion effects of the inner wall of the GIS shell and the probe fixing part are considered, and a sensitivity calibration coefficient of the probe under the distortion electric field is defined. The accuracy of the probe in the GIS internal electric field measurement is improved. BRIEF DESCRIPTION OF DRAWINGS

[0038] The accompanying drawings, which form a part of this specification, are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification. The embodiments of the application, and their

[0039] Figure 1 is a flow chart of an internal electric field probe intervention interference optimization method of an embodiment of the application;

[0040] Figure 2 is a 1100kV GIS geometric model of an embodiment of the application;

[0041] Figure 3 is a curve graph of changes of each evaluation index when the probe is intervened at different positions of an embodiment of the application;

[0042] Figure 4 is a structure diagram of an arc-shaped fixing member of a sensing probe of an embodiment of the application, wherein 1 is a probe mounting groove, 2 is epoxy glue, and 3 is a sensing probe. DETAILED DESCRIPTION

[0043] It should be noted that the following detailed description is merely exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0044] It is to be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments of the application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, devices, components and / or combinations thereof, but do not preclude the presence or addition of one or more other features, steps, operations, devices, components and / or combinations thereof.

[0045] The embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0046] Embodiment One

[0047] As shown in Figure 1 The embodiment discloses an internal electric field probe intervention interference optimization method of an ultra-high voltage GIS, and the method comprises the following steps:

[0048] S1: Obtain the parameters of the ultra-high voltage GIS, input the GIS parameters into an electric field prediction model for prediction, and obtain the electric field distribution before the probe intervention;

[0049] Ultra-high voltage GIS refers to gas insulated switchgear, which is a full enclosed high voltage device integrating circuit breakers, disconnectors, transformers and other components in a metal shell with sulfur hexafluoride gas as the insulating medium, and is widely used in ultra-high voltage power transmission systems.

[0050] In this embodiment, as shown in Figure 2 According to the precise size and material parameters of the ultra-high voltage GIS, a three-dimensional model of the GIS including the basin insulator, high voltage center conductor, metal flange and grounding shell is constructed. Specifically, the material parameters of each component in the GIS segment three-dimensional model are set, the high voltage electrode is set to aluminum, and the insulator is set to epoxy resin; the high voltage conductor potential and the shell grounding potential in the GIS segment three-dimensional model are set, the high voltage conductor potential is set to 1100kV, and the shell potential is set to 0V.

[0051] The existing conventional GIS electrostatic field calculation method is solved by finite element method, which is time-consuming and has low calculation efficiency, and the calculation process is fixed and cannot continue to learn and optimize. To solve the defects of the conventional method, the present application proposes a self-adaptive multi-physical field collaborative simulation framework based on deep learning, which uses deep neural network to replace finite element calculation, realizes fast prediction of electric field, and updates the simulation process from fixed calculation to intelligent process of self-learning and optimization.

[0052] The specific steps of calculating the electric field distribution based on deep learning are as follows:

[0053] (1) Define the parameters of ultra-high voltage GIS and sample to obtain sample points.

[0054] The key design variables of ultra-high voltage GIS are defined as parameters, including the cone angle of basin insulator, umbrella skirt height, high voltage conductor end curvature radius, etc. A large number of sample points are generated in the defined range of the above parameters using Latin hypercube sampling, forming sample data, and each sample point represents a possible GIS design configuration.

[0055] (2) Based on the sample points, the three-dimensional model of the ultra-high voltage GIS is simulated, and the electric field distribution data of each sample point is calculated.

[0056] Based on the parameter values of the sample points, the simulation software updates the three-dimensional model of the ultra-high voltage GIS, updates the material properties, calls the FEM solver for meshing and electrostatic field calculation, and extracts the electric field distribution data.

[0057] (3) Based on the electric field distribution data and the parameters of the ultra-high voltage GIS, a dataset is constructed, and is divided into a training set and a test set according to a certain proportion.

[0058] The parameter and electric field distribution data of all sample points and their corresponding relationship are constructed into a data set, which is stored in a large database for electric field prediction model training.

[0059] (4) The electric field prediction model is trained based on the data set to obtain a trained electric field prediction model.

[0060] Each parameter in the data set is normalized and preprocessed to obtain preprocessed parameter data, and the deep neural network is trained based on the preprocessed parameter data.

[0061] The electric field prediction model adopts a deep neural network, which is a generative neural network with an encoder-decoder structure, and the input data is a parameter vector of GIS design (ultra-high voltage GIS parameters), and the output data is a three-dimensional electric field distribution tensor (electric field distribution data), including an encoder part and a decoder part. The encoder part includes an input layer and a feature extraction layer in turn, and the decoder part includes a feature mapping conversion layer and an output layer in turn.

[0062] The input layer is implemented by a fully connected layer, which receives the normalized and preprocessed parameter vector and transmits the parameter vector to the feature extraction layer. The feature extraction layer is stacked by multiple fully connected layers, including a first hidden layer, a second hidden layer, a third hidden layer and a fully connected layer. Each hidden layer includes linear transformation (using a fully connected layer to perform linear transformation on the input through a weight matrix and a bias vector), batch normalization, ReLU activation function and Dropout layer. After receiving the parameter vector, the parameter vector is processed by multiple hidden layers, each layer learns and extracts more abstract and higher-level features. The last fully connected layer maps the features to a specified dimension of the latent space, and outputs a latent feature vector.

[0063] The feature mapping conversion layer is used to expand the latent feature vector to a high-dimensional latent vector, and reshape the high-dimensional latent vector to a 4D tensor. The feature mapping layer includes a first 3D upsampling block, a second 3D upsampling block and a third 3D upsampling block. Each upsampling block is composed of a normalization layer, a ReLU layer and a 3D convolution layer, which increases the number of channels respectively. The output layer is implemented by a 3D convolution, which outputs the electric field intensity distribution, i.e. a three-dimensional electric field distribution tensor.

[0064] A multi-objective loss function is designed during training, which is a weighted fusion of the main reconstruction loss and the gradient consistency loss , which is expressed as:

[0065]

[0066] wherein, the total multi-objective loss value is represented by is the weight coefficient of the main reconstruction loss, which is a hyperparameter, is a weight coefficient representing the gradient consistency loss, which is a hyperparameter.

[0067] The main reconstruction loss adopts mean squared error loss, which is used to measure the direct difference in numerical value between the predicted electric field distribution and the real electric field distribution, and is represented as:

[0068]

[0069] wherein, represents the number of samples, represents the size of the depth dimension (the number of voxels) of the electric field distribution tensor, represents the size of the height dimension of the electric field distribution tensor, represents the size of the width dimension of the electric field distribution tensor, represents the electric field intensity value predicted by the model at the spatial position of the th sample, represents the real electric field intensity value at the spatial position of the th sample. The main reconstruction loss makes the predicted electric field value of the model as close as possible to the real value in the whole, which is the basis for ensuring the prediction accuracy.

[0070] The gradient consistency loss is based on the spatial gradient difference between the predicted field and the real field, which is used to ensure that the predicted electric field distribution is consistent with the real field in the spatial variation trend (such as the edge, extreme value region), and is represented as:

[0071]

[0072] wherein, represents the index of the spatial direction, which traverses the three spatial dimensions, i.e. depth , height and width ; represents the spatial gradient of the predicted electric field distribution of the model at the spatial position of the th sample along the direction, represents the spatial gradient of the real electric field distribution of the th sample at the spatial position along the direction.

[0073] The gradient consistency loss focuses on the local change of the electric field distribution, which can effectively eliminate the blur phenomenon in the prediction result, make the predicted electric field distribution more clear and accurate in the areas where the electric field changes sharply such as the edge of the conductor and the interface of the insulator, and thus improve the physical rationality of the prediction result.

[0074] (5) Input the parameters of the target GIS into the trained electric field prediction model for prediction to obtain the predicted electric field distribution. This step obtains the distribution of the original spatial electric field inside the GIS before the probe intervention, and records the average value of the spatial electric field and the maximum value of the spatial electric field .

[0075] S2: Obtain an intervention electric field prediction model by performing transfer learning and fine-tuning on the electric field prediction model based on the probe parameters, and predict the electric field distribution at each probe arrangement candidate point after the probe intervention based on the intervention electric field prediction model;

[0076] The specific steps for obtaining the electric field distribution after the probe intervention are as follows:

[0077] (1) Define the probe parameters and sample to obtain sample points;

[0078] The probe parameters include probe size, position and relative dielectric constant, the size parameters include probe diameter, length and curvature radius, etc., and the position parameters include distance, angle and depth relative to the high-voltage conductor, etc.

[0079] Add the probe to the original GIS three-dimensional model, perform Latin hypercube sampling on the probe position, size and relative dielectric constant, and generate new FEM simulation data. Expand the input vector, add the probe parameters to the super-high voltage GIS parameters, and the input vector includes the probe position, size, relative dielectric constant code, etc.

[0080] (2) Model transfer learning and fine-tuning

[0081] Load the trained electric field prediction model in S1, freeze the weights of the encoder part to maintain the understanding of the GIS body, expand the input layer dimension of the model to receive the probe parameters, and add an attention mechanism in the decoder part to focus on the area near the probe.

[0082] After improving the architecture of the electric field prediction model, fine-tune it using a hierarchical fine-tuning strategy to obtain a fine-tuned intervention electric field prediction model. The hierarchical fine-tuning strategy divides the process into three stages: the first stage only trains the newly added probe-related layers, the second stage unfreezes part of the encoder, fine-tunes the body network, and the third stage fine-tunes the entire network end-to-end.

[0083] (3) Fine-tune the intervention electric field prediction model based on multi-scale training data.

[0084] Increase the grid density of the area near the probe in FEM simulation, and focus on labeling the electric field values of the probe surface and the surrounding gas in the training data, especially for the high-curvature areas such as the probe tip. During training, embed physical constraints for the probe, set the probe surface boundary conditions, handle the probe tip singularity, and set the far-field asymptotic behavior.

[0085] (4) Place the sensing probe into the cavity of the GIS three-dimensional model. Pre-set candidate points for probe placement in the cavity of the three-dimensional model. Move the probe model to each candidate point inside the shell in sequence to perform electric field simulation, obtain the probe parameters of each candidate point, perform normalization preprocessing, and input the finely tuned intervention electric field prediction model for prediction to obtain the electric field distribution after probe intervention at each candidate point, and then obtain the average spatial electric field after probe intervention. With the maximum value of the electric field in space .

[0086] like Figure 3 As shown, we can see the curves of the variation of the maximum value of the spatial electric field, the spatial electric field distortion coefficient, and the spatial electric field non-uniformity coefficient when the probe is inserted at different positions.

[0087] S3: Based on the electric field distribution before and after probe intervention, the interference intensity of each candidate probe placement point is calculated.

[0088] The interference intensity is calculated using the following formula, expressed as:

[0089]

[0090]

[0091]

[0092] in, This indicates the probe's distortion coefficient during intervention. This represents the coefficient of non-uniformity of the electric field in space after probe intervention. Indicates the intensity of interference. This represents the weighting coefficient matrix, which was determined through previous simulations, highlighting the critical impact of the maximum electric field and distortion coefficient on insulation safety.

[0093] S4: A multi-objective optimization problem is constructed based on minimizing the electric field interference intensity and minimizing the non-uniformity coefficient of the original electric field distribution. A multi-objective genetic algorithm is used to solve the multi-objective optimization problem to obtain the parameters of the optimal probe placement point.

[0094] In this embodiment, the multi-objective optimization problem is constructed as follows:

[0095] (1) Define decision variables

[0096] The decision variables are the parameters of the probe placement points, including the probe position coordinates, probe orientation angle, and probe size parameters such as radius and length.

[0097] (2) Construct the objective function

[0098] Objective function is to minimize the electric field interference intensity and minimize the coefficient of unevenness of the original electric field distribution .

[0099] The coefficient of unevenness of the original electric field distribution will affect the accuracy of the probe measurement, the smaller the coefficient of unevenness, the higher the measurement accuracy, so as to maximize the accuracy of the probe measurement, the coefficient of unevenness of the original electric field distribution before the probe intervention at each candidate point is calculated :

[0100]

[0101] Wherein, represents the average value of the space electric field.

[0102] Compare the values of all points in the region , finally select the position with the minimum value as the optimal layout position.

[0103] (3) Set the constraint condition

[0104] The constraint condition includes geometric constraint, electrical constraint, mechanical constraint and measurement constraint, the geometric constraint limits the probe from interfering with the internal components of GIS; the electrical constraint is the safety distance requirement, the field strength cannot exceed the threshold value; the mechanical constraint is the limitation of installation space; the measurement constraint is the signal measurability requirement.

[0105] The specific steps of using multi-objective genetic algorithm to solve the multi-objective optimization problem are as follows:

[0106] (1) Design multi-objective comprehensive fitness

[0107]

[0108] Wherein, is the comprehensive fitness of the i-th individual, , is the multi-objective weight, reflecting the relative importance of different objectives.

[0109] (2) Set the algorithm parameters

[0110] Set the parameters, including population size, crossover probability, mutation probability, maximum iteration number, and use tournament selection strategy, simulated binary crossover and polynomial mutation.

[0111] (3) Generate the initial population

[0112] Randomly generate the initial population in the candidate point region of the probe layout, ensure that each individual meets the basic geometric constraint, and use Latin hypercube sampling to ensure the diversity of the initial population; and encode each individual.​​

[0113] (4) Iterative optimization

[0114] For each individual in the population, the interference intensity value and the coefficient of non-uniformity of the original electric field distribution of each probe candidate point are used to calculate the overall fitness.

[0115] The solutions are hierarchically sorted according to the Pareto dominance relationship, and the crowding degree of individuals within the same non-dominated layer is calculated to ensure that the algorithm converges to the Pareto front, and then the genetic operation is performed.

[0116] (5) Extract the Pareto optimal solution set

[0117] In the final population, non-dominated solutions are identified, an approximate Pareto front is constructed, and the uniformity of the distribution of the front solutions is evaluated. The solution closest to the ideal solution is selected as the optimal solution, and the parameters of the optimal probe placement point are output to obtain the optimal probe placement scheme, that is, the optimal probe position, orientation and size are obtained.

[0118] In this embodiment, the present invention designs and installs an arc-shaped probe fixing component, and the specific steps are as follows:

[0119] Based on the measured radius of curvature of the GIS shell, the design is as follows: Figure 4 The arc-shaped fastener is shown. The arc surface of the fastener fits against the inner wall of the GIS housing, and the size of the probe mounting slot 1 on the front matches the body of the selected sensor probe 3. The probe mounting slot 1 is used to accommodate the probe.

[0120] In the GIS simulation model, a geometric model of an arc-shaped fastener is introduced. The relative permittivity of the fastener material is set as a variable 'a', and a parametric scan is performed within the range of 1 to 50. For each value of 'a', the interference intensity is calculated. Simulation results show that the interference intensity is minimized when the relative permittivity 'a' = 4.2 (close to epoxy resin). Therefore, an epoxy resin-based composite material with a relative permittivity of 4.2 is selected as the material for the fastener.

[0121] Use epoxy adhesive 2 to firmly bond the prepared arc-shaped fastener to the inner wall of the GIS housing. Then, carefully insert the sensor probe 3 into the probe mounting groove 1 of the fastener, and fix it with a small amount of the same epoxy adhesive 2. During installation, ensure that there are no visible gaps at the contact surfaces of the housing and the fastener, and that the adhesive layer is fully cured.

[0122] In this embodiment, the probe is installed based on the parameters of the optimal probe placement point, and the measurement sensitivity of the installed probe is calibrated. The specific steps are as follows:

[0123] (1) In the three-dimensional model of the ultra-high voltage GIS, the arc-shaped fixed part with the probe is introduced based on the parameters of the optimal probe arrangement point, to obtain a sensitivity calibration simulation model. The sensitivity calibration simulation model includes the basin-type insulator, the high-voltage central conductor, the metal flange, the grounding shell, and the arc-shaped fixed part which has been installed and whose material attribute is set as the relative dielectric constant of 4.2.

[0124] (2) In the above calibration model, the high-voltage electrode is set to 1100 kV, and the ground electrode is set to 0 kV. The electric field intensity at the probe installation site is calculated and the signal output value of the probe is calculated , and the probe sensitivity is calculated , wherein, is the electric field intensity, and the sensitivity calibration of the probe is realized accordingly.

[0125] Embodiment Two

[0126] The embodiment discloses an ultra-high voltage GIS internal electric field probe intervention interference optimization system, comprising:

[0127] An original electric field prediction module is configured to: obtain the parameters of the ultra-high voltage GIS, input the GIS parameters into an electric field prediction model for prediction, and obtain the electric field distribution before the intervention of the probe;

[0128] An intervention electric field prediction module is configured to: obtain an intervention electric field prediction model by performing transfer learning and fine-tuning on the electric field prediction model based on the probe parameters, and predict the electric field distribution after the intervention of the probe at each probe arrangement candidate point based on the intervention electric field prediction model;

[0129] An interference intensity calculation module is configured to: calculate the interference intensity of each probe arrangement candidate point based on the electric field distribution before the intervention of the probe and the electric field distribution after the intervention of the probe;

[0130] An interference optimization module is configured to: construct a multi-objective optimization problem based on the minimization of the electric field interference intensity and the minimization of the original electric field distribution non-uniformity coefficient, solve the multi-objective optimization problem by using a multi-objective genetic algorithm, and obtain the parameters of the optimal probe arrangement point.

[0131] Embodiment Three

[0132] The purpose of the embodiment is to provide a computing device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method of embodiment one when executing the program.

[0133] Embodiment Four

[0134] The embodiment aims to provide a computer readable storage medium, and a computer readable storage medium storing a computer program, which is executed by a processor to perform the steps of the method of the embodiment.

[0135] The steps involved in the devices of the above embodiments three and four correspond to the method of the embodiment one, and the specific embodiments can refer to the related description of the embodiment one. The term "computer readable storage medium" should be understood as including a single medium or multiple media of one or more instruction sets; it should also be understood as including any medium capable of storing, encoding or carrying instruction sets for execution by a processor and causing the processor to perform any method in the present application.

[0136] Those skilled in the art should understand that the above-mentioned modules or steps of the present application can be realized by a general computer device, alternatively, they can be realized by program codes executable by a computing device, so that they can be stored in a storage device for execution by a computing device, or they can be respectively made into individual integrated circuit modules, or a plurality of modules or steps among them can be made into a single integrated circuit module to realize. The present application is not limited to any specific combination of hardware and software.

[0137] The above only describes the preferred embodiments of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0138] The above describes the specific embodiments of the present application in combination with the drawings, but is not used to limit the protection scope of the present application. Those skilled in the art should understand that various modifications or changes made on the basis of the technical solutions of the present application without creative labor are still within the protection scope of the present application.

Claims

1. An ultra-extra-high voltage GIS internal electric field probe intervention interference optimization method, characterized in that, The method comprises the following steps: obtaining the parameters of the extra-high voltage GIS, inputting the GIS parameters into an electric field prediction model for prediction to obtain the electric field distribution before the probe intervention; obtaining the electric field prediction model after transfer learning and fine-tuning based on the probe parameters, and predicting the electric field distribution after the probe intervention at each probe arrangement candidate point based on the electric field prediction model after the intervention; the specific steps for obtaining the electric field distribution after the probe intervention are as follows: (1) defining the probe parameters and sampling to obtain sample points; the probe parameters include probe size, position and relative dielectric constant, the size parameters include probe diameter, length and curvature radius, and the position parameters are distance, angle and depth relative to the high-voltage conductor; adding the probe to the original GIS three-dimensional model, performing Latin hypercube sampling on the probe position, size and relative dielectric constant to generate new FEM simulation data; extending the input vector, adding the probe parameters to the extra-high voltage GIS parameters, and the input vector includes the probe position, size and relative dielectric constant code; (2) model transfer learning and fine-tuning; loading the trained electric field prediction model, freezing the weights of the encoder part to maintain the understanding of the GIS body, extending the input layer dimension of the model to receive the probe parameters, and adding an attention mechanism in the decoder part; after improving the architecture of the electric field prediction model, the model is fine-tuned using a hierarchical fine-tuning strategy to obtain a fine-tuned electric field prediction model after the intervention; the hierarchical fine-tuning strategy divides the process into three stages, the first stage only trains the newly added probe-related layers, the second stage unfreezes part of the encoder, fine-tunes the body network, and the third stage fine-tunes the whole network end to end; (3) fine-tuning the electric field prediction model after the intervention based on multi-scale training data; increase the grid density of the area near the probe in the FEM simulation, and mark the electric field values of the probe surface and the surrounding gas in the training data, especially for the high-curvature area; during training, embed the probe with physical constraints, set the probe surface boundary condition, handle the probe tip singularity and far-field asymptotic behavior; (4) The sensing probe is placed into the cavity of the GIS three-dimensional model, candidate points for probe arrangement are set in the three-dimensional model cavity in advance, the probe model is moved to each candidate point inside the shell in turn for electric field simulation, the probe parameters of each candidate point are obtained, after normalization pretreatment, the fine-tuned intervention electric field prediction model is inputted for prediction, the electric field distribution after intervention of the probe at each candidate point is obtained, and then the average value of the space electric field after intervention of the probe is obtained and the maximum value of the space electric field ; based on the electric field distribution before the probe intervention and the electric field distribution after the probe intervention, the interference intensity of each probe arrangement candidate point is calculated; based on the minimization of the electric field interference intensity and the minimization of the original electric field distribution non-uniformity coefficient, a multi-objective optimization problem is constructed, a multi-objective genetic algorithm is used to solve the multi-objective optimization problem, and the parameters of the optimal probe arrangement point are obtained.

2. The method of claim 1, wherein the method is characterized by, The extra-high voltage GIS parameters include the cone angle of the basin-type insulator, the umbrella skirt height, and the high-voltage conductor end curvature radius.

3. The method of claim 2, wherein the method is characterized by, The electric field prediction model uses a generative neural network with an encoder-decoder structure, and a multi-objective loss function is designed, including a main reconstruction loss and a gradient consistency loss.

4. The method of claim 1, wherein the method is characterized by, The probe parameters include probe position, probe size and relative dielectric constant.

5. The method of claim 1, wherein the method is characterized by: The interference intensity is represented as: wherein, represents a probe intervention distortion coefficient, represents a spatial electric field inhomogeneity coefficient after probe intervention, represents an interference intensity, represents a weight coefficient matrix, represents a spatial electric field average value after probe intervention, represents a spatial electric field maximum value after probe intervention, represents a spatial electric field maximum value.

6. The method of claim 1, wherein the method is characterized by: An arc-shaped fixing member is also designed for the probe, the arc surface of the arc-shaped fixing member is consistent with the curvature of the inner wall of the extra-high voltage GIS shell, the front surface is provided with a probe mounting groove, and the probe mounting groove is used to accommodate the sensing probe.

7. The method of claim 6, wherein the method further comprises: determining the position of the probe in the GIS; and determining the position of the probe in the GIS based on the voltage difference between the first and second electrodes. Based on the parameters of the optimal probe arrangement point, the probe is installed, and the measurement sensitivity of the installed probe is calibrated, specifically: In the three-dimensional model of ultra-high voltage GIS, an arc-shaped fixing component with fixed probes is introduced based on the parameters of the optimal probe layout points to obtain a sensitivity calibration simulation model. In the sensitivity calibration simulation model, a high-voltage electrode and a ground electrode are set, and the electric field strength at the probe insertion point and the signal output value of the probe are calculated. The probe sensitivity is calculated based on the electric field strength and signal output value, thereby enabling probe sensitivity calibration.

8. An ultra extra high voltage GIS internal electric field probe intervention interference optimization system, characterized in that, include: The original electric field prediction module is configured to: acquire ultra-high voltage GIS parameters, input the GIS parameters into the electric field prediction model for prediction, and obtain the electric field distribution before probe intervention; The interventional electric field prediction module is configured to: obtain the interventional electric field prediction model by performing transfer learning and fine-tuning on the electric field prediction model based on the probe parameters, and predict the electric field distribution after probe intervention at each candidate probe placement point based on the interventional electric field prediction model. The specific steps to obtain the electric field distribution after probe intervention are as follows: (1) Define probe parameters and obtain sample points; The probe parameters include probe size, position, and relative permittivity. Size parameters include probe diameter, length, and radius of curvature. Position parameters are distance, angle, and depth relative to the high-voltage conductor. A probe is added to the original GIS 3D model, and the probe position, size, and relative permittivity are sampled using Latin hypercube to generate new FEM simulation data; the input vector is expanded by adding probe parameters to the ultra-high voltage GIS parameters, and the input vector includes probe position, size, and relative permittivity encoding. (2) Model transfer learning and fine-tuning; Load the trained electric field prediction model, freeze the encoder part weights to maintain the understanding of GIS ontology, expand the model input layer dimension to receive probe parameters, and add an attention mechanism in the decoder part. After improving the electric field prediction model architecture, a hierarchical fine-tuning strategy was adopted to fine-tune it, resulting in a finely tuned intervention electric field prediction model. The layered fine-tuning strategy divides the process into three stages: the first stage trains only the newly added probe-related layers; the second stage unfreezes part of the encoder and fine-tunes the entire network; and the third stage performs end-to-end fine-tuning of the entire network. (3) Fine-tuning the intervention electric field prediction model based on multi-scale training data; In FEM simulation, the mesh density in the region near the probe is increased, and the electric field values ​​of the probe surface and the surrounding gas are highlighted in the training data, especially the high curvature region is oversampled; during training, physical constraints are embedded in the probe, and boundary conditions of the probe surface, singularity treatment of the probe tip and far-field asymptotic behavior are set. (4) The sensing probe is placed into the cavity of the GIS three-dimensional model, candidate points for probe arrangement are set in the three-dimensional model cavity in advance, the probe model is moved to each candidate point inside the shell in turn for electric field simulation, the probe parameters of each candidate point are obtained, after normalization pretreatment, the fine-tuned intervention electric field prediction model is inputted for prediction, the electric field distribution after intervention of the probe at each candidate point is obtained, and then the average value of the space electric field after intervention of the probe is obtained and the maximum value of the space electric field ; The interference intensity calculation module is configured to calculate the interference intensity of each candidate probe placement point based on the electric field distribution before and after probe insertion. The interference optimization module is configured to: construct a multi-objective optimization problem based on minimizing the electric field interference intensity and minimizing the non-uniformity coefficient of the original electric field distribution, and solve the multi-objective optimization problem using a multi-objective genetic algorithm to obtain the parameters of the optimal probe placement point.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, When the program is executed by the processor, it implements the steps in the interference optimization method for the internal electric field probe of ultra-high voltage GIS as described in any one of claims 1-7.

10. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the steps in the method for optimizing the interference of an internal electric field probe in an ultra-high voltage GIS according to any one of claims 1-7 when executing the program.

Images