A method for predicting the development path of electrical trees in solid dielectrics

Abstract

This invention relates to a method for predicting the development path of electrical trees in solid dielectrics, belonging to the field of reliability analysis and life prediction of high-voltage insulation materials. The method includes: numerically simulating the development path of electrical trees in solid dielectrics using the phase-field method under different load levels; constructing a dataset based on generated time-series images, and grouping and storing the dataset by load level; processing the classified datasets to obtain training and training sets; constructing and training a machine learning model for predicting the development path of electrical trees, learning the changing characteristics of electrical tree development paths under different load levels; acquiring images of the solid dielectric and load information in a real-world scenario and inputting them into the trained machine learning model to predict the development path of electrical trees in the solid dielectric under that load and generate a sequence of electrical tree development images. This invention improves the prediction accuracy of electrical tree propagation paths and expands the applicability of the model, enabling it to adapt to complex working conditions under different load levels.

Description

Technical Field

[0001] This invention belongs to the field of reliability analysis and lifetime prediction of high voltage insulation materials, and relates to the prediction of electrical tree propagation in solid insulation materials. In particular, it relates to a method for predicting the development path of electrical trees in solid dielectrics by integrating numerical simulation and machine learning. Background Technology

[0002] Polymer insulating materials are widely used in electrical equipment due to their excellent insulation properties, corrosion resistance, and processing flexibility. They are particularly favored in critical components such as high-voltage cables, transformers, and insulators due to their high dielectric strength and good mechanical properties. However, during long-term operation, polymer insulating materials can still be affected by various environmental factors and electric fields, leading to a gradual decrease in their insulation performance and consequently affecting the stable operation of the equipment. Therefore, studying the aging and damage mechanisms of polymer insulating materials, especially the development process of electrical treeing within the polymer, is crucial for improving the reliability of electrical equipment.

[0003] Under high electric fields and various types of mechanical loads, electrical trees may gradually form within polymer insulation materials. Electrical trees are branching damage structures, and their propagation process typically exhibits significant nonlinear characteristics as loads accumulate. Load factors such as temperature, electric field strength, and mechanical stress significantly influence the development path and rate of electrical trees. If electrical trees are not effectively controlled, their propagation may eventually lead to localized breakdown of the insulation material, and in severe cases, even cause the failure of the entire electrical equipment. Therefore, predicting and controlling the propagation of electrical trees under load conditions is crucial for improving the lifespan of insulation materials and the safety of equipment.

[0004] To ensure the long-term safe operation of electrical equipment, real-time monitoring of electrical tree expansion is crucial. Traditional methods for predicting electrical tree propagation paths are typically based on numerical simulations (such as the phase-field method), which can visualize and simulate electrical tree expansion under laboratory conditions. However, these methods are limited by computational resources and are difficult to apply in real-time to actual operating conditions. With the development of data-driven technologies, prediction methods combining numerical simulation and machine learning can provide faster and more accurate predictions of electrical tree propagation paths, potentially enabling real-time monitoring and prediction of the electrical tree expansion process. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method for predicting the development path of electrical trees in solid dielectrics. By integrating numerical simulation and machine learning to predict the development path of electrical trees, this method can efficiently predict the expansion path of electrical trees under complex load conditions, thereby providing important technical support for the safe maintenance of electrical equipment.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for predicting the propagation path of electrical trees in solid dielectrics, the method comprising:

[0008] S1. Under different load levels, the phase field method is used to numerically simulate the development path of electric trees in solid dielectrics, and a time series image containing the development path of electric trees is generated.

[0009] S2. Construct a dataset based on the generated image data, and group and store the dataset by load level as a label;

[0010] S3. Perform uniform frame sampling on the classification dataset, convert the grayscale image into a three-channel RGB image and normalize it, then group the image sequence according to the preceding and following order to construct the training set and the dataset.

[0011] S4. Construct a machine learning model for predicting the development path of electrical tree branches, and train the model using the training set so that the model learns the changing characteristics of the development path of electrical tree branches under different load levels.

[0012] S5. Obtain images and load information of the solid medium in the actual scene, input them into the trained machine learning model to predict the development path of electric tree in the solid medium under the load and generate a sequence of electric tree development images several time steps later.

[0013] Furthermore, in step S1, the phase-field method is used to numerically simulate the development path of electrical trees in solid dielectrics under different load levels, including:

[0014] 1) Determine the simulation parameters and phase-field parameters of the solid medium; calculate the field variables based on energy distribution and physical laws;

[0015] The simulation parameters include the relative permittivity of the air gap and the dielectric phase, the breakdown electric field strength, and the energy barrier for the development of electric trees. The phase field parameters of the solid dielectric include the simulation step size and the mesh size set according to the size of the target solid dielectric.

[0016] 2) Apply the phase field dynamics equation to update the sequence parameters and obtain the electric tree development state at each time step;

[0017] 3) Iterate the simulation until the electrical tree causes the solid medium to break down, forming a complete time series image.

[0018] Furthermore, in step S1, the different load levels are different load levels formed by the combined loads applied to the material under different loading voltages, applied mechanical stresses, and applied temperatures.

[0019] Furthermore, in step S4, the constructed machine learning model includes a spatial encoder, a temporal module, and a spatial decoder. The spatial encoder extracts spatial features from the input image sequence and encodes them into a latent spatial representation, which is then input into the temporal module. The temporal module converts the encoded data into temporal information usable by the spatial decoder, capturing the time-dependent characteristics of the electrical tree extension process and enhancing the perception of temporal information. The spatial decoder decodes the data output by the temporal module, integrates the feature information from each time step, and generates the ability to learn and predict the global features of electrical tree development.

[0020] The beneficial effects of this invention are as follows: By combining numerical simulation and machine learning, this invention not only significantly improves the prediction accuracy of electrical tree propagation paths but also expands the applicability of the model, enabling it to adapt to complex working conditions with different load levels. Furthermore, the introduction of simulation data makes model training more effective, while the time-series prediction capability of machine learning significantly enhances the ability to capture the nonlinear propagation paths of electrical trees, providing a new technical means for the reliability research of high-voltage insulation materials.

[0021] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0022] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0023] Figure 1 This is a flowchart illustrating a method for predicting the propagation path of electrical trees in a solid dielectric according to an embodiment of the present invention.

[0024] Figure 2 Example of a time-series image of the development path of an electric tree generated from a numerical simulation;

[0025] Figure 3 This refers to the preprocessing of image data.

[0026] Figure 4 An example of predicting the expansion path of electrical tree branches for a machine learning model. Detailed Implementation

[0027] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0028] like Figure 1 As shown, this invention provides a method for predicting the evolution path of electric tree branches by integrating numerical simulation and machine learning. The method is as follows:

[0029] 1. Under different load levels, the development path of electric trees in solid dielectrics is numerically simulated based on the phase field method, and time series images reflecting the development path of electric trees are generated.

[0030] Different load levels refer to the combined loads applied to materials under different loading voltages, applied mechanical stresses, and applied temperatures, thus forming different load levels.

[0031] Phase-field method simulations initialize the simulation by setting the material's load and physical properties (such as relative permittivity, breakdown electric field strength, and energy barrier). By calculating the system and local energies, the minimum energy path for electric tree propagation is obtained based on fracture variational theory. Phase-field method simulations include:

[0032] 1) Determine the simulation parameters and phase field parameters of the solid medium, including determining the relative permittivity of the air gap and the medium phase, the breakdown electric field strength, the energy barrier for the development of electric trees, and setting the simulation step size and mesh size according to the target medium size.

[0033] The relative permittivity is calculated using the following formula:

[0034]

[0035] In the formula, For order parameters, The dielectric constant at the current position. The dielectric constant of the material at complete breakdown. is the dielectric constant of the initial material.

[0036] 2) Calculate field variables based on energy distribution and physical laws:

[0037]

[0038] In the formula, Joule heat energy density, Energy density, such as mechanical stress, Electrostatic energy density, For gradient energy density, For phase separation energy density, For a unit volume of solid medium.

[0039] 3) Update the sequence parameters using the phase-field dynamics equation to obtain the electric tree development state at each time step; the phase-field dynamics equation is expressed as:

[0040]

[0041] In the formula, To break down the threshold energy density, This is the updatable coefficient for interface migration rate. Let be a step function, and its independent variable be... .

[0042] The energy density and related material parameters mentioned above can all be characterized by phase field order parameters, which reflects the uniformity brought by the phase field method and can also be well adapted to machine learning.

[0043] 4) Iterative simulation continues until electrical treeing causes breakdown of the polymer solid medium, forming a complete time-series image dataset, such as... Figure 2 As shown.

[0044] 2. After generating time-series images of electrical tree evolution paths under different load levels, a dataset is constructed and grouped and stored using load level as the label. This method of constructing a dataset can provide well-labeled training data for machine learning models.

[0045] 3. Further processing of the grouped image datasets, such as... Figure 3 As shown, it mainly includes:

[0046] (1) Sample frames uniformly from the time-series image to ensure data consistency;

[0047] (2) Convert the grayscale image into a three-channel RGB image and perform normalization processing to normalize the image value range to between 0 and 1;

[0048] (3) The image sequence is sampled in groups according to the preceding and following order to construct the training set and validation set of the machine learning model.

[0049] 4. Based on the processed dataset, design and train a machine learning model with time series modeling capabilities to predict the propagation of electrical tree paths, learn the changing characteristics of electrical tree development paths under different load levels, thereby improving prediction accuracy.

[0050] In this embodiment, since the development path of electrical tree branches exhibits obvious temporal characteristics, and its path changes gradually accumulate with time and load conditions, presenting a nonlinear process, the machine learning model chosen is a time-series prediction model, which achieves the prediction output for the target time point through the mapping function shown in the following formula:

[0051]

[0052] That is, by inputting an initial image sequence of several frames, the development path of electrical tree branches after the target time point is predicted.

[0053] like Figure 4 As shown, the time-series prediction model includes a spatial encoder, a time-series module, and a spatial decoder. The spatial encoder is represented as:

[0054]

[0055] The spatial decoder is represented as:

[0056]

[0057] The aforementioned timing module is a timing prediction module, such as an RNN network.

[0058] The training process of this time series prediction model includes:

[0059] 1) Input the image sequence after frame sampling into the spatial encoder, use operations such as convolution and normalization to extract spatial features, and encode the image information into a latent spatial representation to facilitate the learning of subsequent temporal features;

[0060] 2) The extracted features are input into the temporal module to enhance the perception of temporal information; at the same time, the temporal module acts as a translator to convert the encoded data into temporal information that the decoder can use, thereby capturing the time-dependent characteristics of the electrical tree expansion process.

[0061] 3) The decoder decodes the data output by the timing module, integrates the feature information of each time step, and thus generates the ability to learn and predict the global features of electrical tree development.

[0062] 5. After model training is complete, the model parameters with the minimum loss function are saved locally. In practical applications, input data containing load level information and initial crack images are fed into the trained model. The model outputs a sequence of electrical tree development images several time steps in the future to predict the development path of electrical trees under the load condition, effectively assisting in the condition prediction and maintenance of electrical equipment.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the development path of electrical trees in solid dielectrics, characterized in that, Under different load levels, the phase-field method was used to numerically simulate the propagation path of electric trees in solid dielectrics, generating time-series images containing the propagation path of electric trees; A dataset is constructed based on the generated image data, and the dataset is grouped and stored using the load level as a label. Uniform frame sampling is performed on the classification dataset to convert grayscale images into three-channel RGB images and normalize them. Then, the image sequences are grouped and sampled according to pre-order and post-order to construct the training set and the dataset. A machine learning model for predicting the development path of electrical tree branches is constructed, and the model is trained using the training set to learn the changing characteristics of the development path of electrical tree branches under different load levels. The image and load information of the solid medium in the real scene are obtained and input into the trained machine learning model to predict the development path of the electric tree under the load and generate an image sequence of the electric tree development several time steps in the future. Different load levels refer to the combined loads applied to materials under different loading voltages, applied mechanical stresses, and applied temperatures, thus forming different load levels. Phase-field simulation includes: First, determine the simulation parameters and the phase field parameters of the solid medium, including determining the relative permittivity of the air gap and the medium phase, the breakdown electric field strength, and the energy barrier for the development of electric trees. It also includes setting the simulation step size and mesh size according to the target medium size. The relative permittivity is calculated using the following formula: In the formula, For order parameters, The dielectric constant at the current position. The dielectric constant of the material at complete breakdown. The dielectric constant of the initial material; Then, the field variables are calculated based on the energy distribution and physical laws: In the formula, Joule heat energy density, Energy density, such as mechanical stress, For electrostatic energy density, For gradient energy density, For phase separation energy density, For a unit volume of solid medium; Finally, the sequence parameters are updated using the phase-field dynamics equation to obtain the electric tree development state at each time step; the phase-field dynamics equation is expressed as: In the formula, To break down the threshold energy density, This is the updatable coefficient for interface migration rate. Let be a step function, and its independent variable be... ; The machine learning model is a temporal prediction model, which includes a spatial encoder, a temporal module, and a spatial decoder; the spatial encoder is represented as follows: The spatial decoder is represented as: The timing module is an RNN network.

2. The method according to claim 1, characterized in that, The machine learning model includes a spatial encoder, a temporal module, and a spatial decoder; the spatial encoder is used to extract spatial features of the input image sequence and encode them into a latent spatial representation, which is then input into the temporal module; The timing module converts the encoded data into timing information used by the spatial decoder, captures the time-dependent characteristics of the electrical tree extension process, and enhances the perception of timing information. The spatial decoder decodes the data output by the temporal module, integrates the feature information of each time step, and generates the ability to learn and predict the global features of electrical tree development.

Images