Fault detection method of gas insulated switchgear based on multi-map fusion
Through multi-map fusion and 18-layer residual neural network, the problem of low GIS fault detection efficiency is solved, efficient and accurate fault detection and prediction is achieved, and the maintenance efficiency and safety of the equipment are improved.
Patent Information
- Application Number
- CN202510615250.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-13
- Publication Date
- 2025-08-08
AI Technical Summary
In the prior art, the fault detection efficiency of gas insulated switchgear (GIS) is low and the process is complicated, making it difficult to effectively detect internal mechanical structure defects, resulting in the development of potential faults into serious consequences.
The multi-map fusion method is adopted to obtain the time-frequency vibration signals of the gas insulated switch equipment, perform short-time Fourier transform and Gram angle field transformation, extract color and texture information, generate fusion maps, and use 18-layer residual neural network for fault detection.
It realizes efficient and accurate fault detection, improves the accuracy and robustness of fault diagnosis, ensures real-time monitoring and fault prediction of equipment, and improves maintenance efficiency and operation safety.
Smart Images
Figure CN120448791A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of power transmission technology, and in particular to a fault detection method for gas insulated switchgear (GIS) based on multi-graph fusion. Background Art
[0002] GIS is widely used in the power transmission field due to its advantages such as strong power transmission capacity, small footprint and low maintenance. However, defects in the internal mechanical structure of GIS equipment are difficult to detect. During long-term energized operation, defects can easily develop into failures, resulting in serious consequences.
[0003] In existing research, numerical feature extraction or neural network methods are mainly used to solve the above-mentioned fault detection problem by using a one-dimensional feature vector composed of multiple features.
[0004] However, the existing methods have disadvantages such as large computational complexity of the network model, resulting in low detection efficiency and complicated detection process. Summary of the Invention
[0005] The embodiments of the present application provide a gas-insulated switchgear fault detection method based on multi-graph fusion to achieve efficient and convenient GIS fault detection.
[0006] In a first aspect, an embodiment of the present application provides a fault detection method for a gas-insulated switchgear based on multi-graph fusion, comprising:
[0007] Obtaining the target time-frequency vibration signal of the gas-insulated switchgear to be inspected;
[0008] Determining a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal;
[0009] Determining a target fusion spectrum according to the target time-frequency domain spectrum and the target coding spectrum;
[0010] The target fusion map is input into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected. The fault detection model is obtained by training an 18-layer residual neural network based on at least one time-frequency vibration signal of the gas-insulated switchgear with different mechanical defect types.
[0011] In one or more embodiments, determining a target fusion spectrum according to the target time-frequency domain spectrum and the target coding spectrum includes:
[0012] Extracting target color information from the target time-frequency domain spectrum;
[0013] Extracting target texture information from the target coding atlas;
[0014] The target texture information and the target color information are reconstructed and fused to obtain the target fusion atlas.
[0015] In one or more embodiments, extracting target color information from the target time-frequency domain spectrum includes:
[0016] Performing Gaussian smoothing and downsampling on the target time-frequency domain spectrum for a preset number of times to obtain target color information;
[0017] Accordingly, the extracting target texture information from the target coding atlas includes:
[0018] Performing Gaussian smoothing and downsampling processing on the target coding atlas for the preset number of times to obtain target texture information;
[0019] Accordingly, the target texture information and the target color information are reconstructed and fused to obtain the target fusion atlas, including:
[0020] The target texture information and the target color information are subjected to restoration processing corresponding to Gaussian smoothing processing and downsampling processing to obtain the target fusion atlas.
[0021] In one or more embodiments, determining a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal includes:
[0022] Performing a short-time Fourier transform on the target time-frequency vibration signal to obtain the target time-frequency domain spectrum;
[0023] Performing Gram angle field transform on the target time-frequency vibration signal to obtain the target coding spectrum.
[0024] In one or more embodiments, before inputting the target fusion graph into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected, the method further includes:
[0025] acquiring at least one time-frequency vibration signal of different mechanical defect types;
[0026] The 18-layer residual neural network is trained according to at least one time-frequency vibration signal of different mechanical defect types to obtain the fault detection model.
[0027] In one or more embodiments, the training of the 18-layer residual neural network based on at least one time-frequency vibration signal of different mechanical defect types to obtain the fault detection model includes:
[0028] For each time-frequency vibration signal, determining a time-frequency domain spectrum and a coding spectrum corresponding to the time-frequency vibration signal;
[0029] Determining a fusion spectrum according to the time-frequency domain spectrum and the coding spectrum;
[0030] The 18-layer residual neural network is trained according to a fusion graph corresponding to at least one time-frequency vibration signal to obtain the fault detection model.
[0031] In one or more embodiments, different mechanical defect types include at least one of the following: normal operating conditions, loose bellows bolts, and foreign matter in busbar air chamber bolts.
[0032] In a second aspect, an embodiment of the present application provides a fault detection device for a gas-insulated switchgear based on multi-graph fusion, comprising:
[0033] An acquisition module, used for acquiring a target time-frequency vibration signal of the gas-insulated switchgear to be tested;
[0034] A first determination module is used to determine a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal;
[0035] A second determining module is used to determine a target fusion spectrum according to the target time-frequency domain spectrum and the target coding spectrum;
[0036] A processing module is used to input the target fusion map into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected. The fault detection model is obtained by training an 18-layer residual neural network based on at least one time-frequency vibration signal of the gas-insulated switchgear with different mechanical defect types.
[0037] In one or more embodiments, the second determining module is specifically configured to:
[0038] Extracting target color information from the target time-frequency domain spectrum;
[0039] Extracting target texture information from the target coding atlas;
[0040] The target texture information and the target color information are reconstructed and fused to obtain the target fusion atlas.
[0041] In one or more embodiments, the second determination module extracts target color information from the target time-frequency domain spectrum, specifically for:
[0042] Performing Gaussian smoothing and downsampling on the target time-frequency domain spectrum for a preset number of times to obtain target color information;
[0043] Accordingly, the second determination module extracts target texture information from the target coding atlas, specifically for:
[0044] Performing Gaussian smoothing and downsampling processing on the target coding atlas for the preset number of times to obtain target texture information;
[0045] Correspondingly, the second determination module reconstructs and fuses the target texture information and the target color information to obtain the target fusion atlas, which is specifically used to:
[0046] The target texture information and the target color information are subjected to restoration processing corresponding to Gaussian smoothing processing and downsampling processing to obtain the target fusion atlas.
[0047] In one or more embodiments, the first determining module is specifically configured to:
[0048] Performing a short-time Fourier transform on the target time-frequency vibration signal to obtain the target time-frequency domain spectrum;
[0049] Performing Gram angle field transform on the target time-frequency vibration signal to obtain the target coding spectrum.
[0050] In one or more embodiments, before inputting the target fusion graph into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected, the processing module is further configured to:
[0051] acquiring at least one time-frequency vibration signal of different mechanical defect types;
[0052] The 18-layer residual neural network is trained according to at least one time-frequency vibration signal of different mechanical defect types to obtain the fault detection model.
[0053] In one or more embodiments, the processing module trains the 18-layer residual neural network based on at least one time-frequency vibration signal of different mechanical defect types to obtain the fault detection model, specifically for:
[0054] For each time-frequency vibration signal, determining a time-frequency domain spectrum and a coding spectrum corresponding to the time-frequency vibration signal;
[0055] Determining a fusion spectrum according to the time-frequency domain spectrum and the coding spectrum;
[0056] The 18-layer residual neural network is trained according to a fusion graph corresponding to at least one time-frequency vibration signal to obtain the fault detection model.
[0057] In one or more embodiments, different mechanical defect types include at least one of the following: normal operating conditions, loose bellows bolts, and foreign matter in busbar air chamber bolts.
[0058] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor;
[0059] The memory stores computer-executable instructions;
[0060] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementations of the first aspect.
[0061] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementation methods of the first aspect.
[0062] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.
[0063] The present invention provides a method for fault detection of gas-insulated switchgear based on multi-spectrum fusion. The method obtains a target time-frequency vibration signal of the gas-insulated switchgear to be detected; determines a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal; determines a target fusion spectrum based on the target time-frequency domain spectrum and the target coding spectrum; and inputs the target fusion spectrum into a pre-trained fault detection model to obtain a target fault type of the gas-insulated switchgear to be detected. The fault detection model is obtained by training an 18-layer residual neural network based on at least one time-frequency vibration signal of the gas-insulated switchgear with different mechanical defect types. In this technical solution, by obtaining the time-frequency vibration signal of the gas-insulated switchgear to be detected, characteristic information generated by the device during operation due to faults can be captured. After processing, these signals can generate corresponding time-frequency domain spectra and coding spectra, reflecting the frequency changes and fault modes of the device under different states. Then, based on the time-frequency domain spectrum and coding spectrum, a target fusion spectrum is determined. This spectrum integrates information from different frequency bands, thereby more comprehensively describing the operating status of the device. The target fusion graph is then fed into a pre-trained fault detection model, enabling accurate identification of the equipment's fault type. This method not only efficiently detects equipment faults but also improves the accuracy and robustness of fault diagnosis, ensuring real-time monitoring and fault prediction for equipment, ultimately enhancing maintenance efficiency and operational safety. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0065] Figure 1 Schematic diagram of the process of the fault detection method of gas insulated switchgear based on multi-graph fusion provided in the embodiment of the present application Figure 1 ;
[0066] Figure 2 Schematic diagram of the process of the fault detection method of gas insulated switchgear based on multi-graph fusion provided in the embodiment of the present application Figure 2 ;
[0067] Figure 3 A schematic diagram of determining the target fusion map provided in the embodiment of the present application;
[0068] Figure 4 Schematic diagram of the process of the fault detection method of gas insulated switchgear based on multi-graph fusion provided in the embodiment of the present application Figure 3 ;
[0069] Figure 5 A schematic diagram of the confusion matrix of the test set diagnosis results provided in an embodiment of the present application;
[0070] Figure 6 Schematic diagram of the process of the fault detection method of gas insulated switchgear based on multi-graph fusion provided in the embodiment of the present application Figure 4 ;
[0071] Figure 7 A schematic structural diagram of a fault detection device for a gas-insulated switchgear based on multi-graph fusion provided in an embodiment of the present application;
[0072] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.
[0073] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION
[0074] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.
[0075] GIS is widely used in the power transmission sector due to its advantages such as high transmission capacity, small footprint, and minimal maintenance. However, mechanical structural defects within GIS equipment are difficult to detect. During long-term energized operation, these defects can easily develop into failures, resulting in serious consequences. Because GIS mechanical structural defects are subtle, latent, and often lack obvious characteristics, research on mechanical defect diagnosis in GIS equipment has attracted considerable attention.
[0076] In existing technologies, machine learning can automatically adjust structural parameters and change their mapping relationships. It has strong generalization capabilities and has been widely used in the field of fault diagnosis. For example, it extracts feature vectors from GIS acoustic wave data and adopts a fault diagnosis algorithm based on an online sequential extreme learning machine to achieve accurate diagnosis of mechanical defects.
[0077] However, in the above implementation, numerical feature extraction or neural network methods are mainly used to solve the above fault detection problem using a one-dimensional feature vector composed of multiple features. However, there are disadvantages such as a large amount of computational complexity of the network model, which leads to low detection efficiency and a complicated detection process.
[0078] Based on the above technical problems, the inventors' technical conception is as follows: the image pyramid principle can be used to convert the time-frequency vibration signal into a short-time Fourier time-frequency spectrum and a Gram angular field spectrum, and perform adaptive weight fusion. Specifically, color information and texture information can be extracted from the two spectra respectively and adaptively upsampled and reconstructed to obtain a fused spectrum with complementarity and two-spectrum characteristics, which can more accurately represent the characteristics of the time-frequency vibration signal. After that, an 18-layer residual neural network is used to analyze the features, which can obtain more accurate detection results and avoid the problem that the neural network model requires a lot of calculations.
[0079] The following specific embodiments describe in detail the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.
[0080] Figure 1 Schematic diagram of the process of the fault detection method of gas insulated switchgear based on multi-graph fusion provided in the embodiment of the present application Figure 1 ,like Figure 1 As shown, the method includes:
[0081] Step 11: Acquire a target time-frequency vibration signal of the gas-insulated switchgear to be tested;
[0082] In this step, a piezoelectric acceleration sensor may be used to collect a vibration signal within a preset time period of the gas-insulated switchgear that currently needs to be detected, and the signal is recorded as a target time-frequency vibration signal.
[0083] After that, the collected target time-frequency vibration signal can be filtered out of interference signals based on the wavelet packet decomposition filtering method to ensure the accuracy of subsequent fault type detection.
[0084] Step 12: determining a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal;
[0085] In this step, the collected target time-frequency vibration signal may be processed differently to obtain a target time-frequency domain spectrum corresponding to the target time-frequency vibration signal and a target coding spectrum corresponding to the target time-frequency vibration signal.
[0086] The purpose of this implementation is to directly acquire different vibration spectra of the target time-frequency vibration signal. After obtaining the target time-frequency domain spectrum and the target coding spectrum, the above two spectra can be fused and feature mined, thereby avoiding the information extraction limitations of a single spectrum.
[0087] Optionally, step 12 may be implemented as follows:
[0088] Step 1: Perform short-time Fourier transform on the target time-frequency vibration signal to obtain the target time-frequency domain spectrum;
[0089] In this implementation, the collected target time-frequency vibration signal is mapped into a time-frequency spectrum, that is, a target time-frequency domain spectrum, through short-time Fourier transform.
[0090] Step 2: Perform Gram angle field transform on the target time-frequency vibration signal to obtain the target coding spectrum.
[0091] In this implementation, the collected target time-frequency vibration signal is mapped into a coding spectrum, namely, a target coding spectrum, through Gram angular field transform.
[0092] It should be understood that the execution order of step 1 and step 2 is not limited.
[0093] Step 13: Determine the target fusion spectrum according to the target time-frequency domain spectrum and the target coding spectrum;
[0094] In this step, based on the target time-frequency domain atlas and the target coding atlas determined above, a target fused atlas having the color information of the target time-frequency domain atlas and the texture information of the target coding atlas is obtained by fusion.
[0095] In one possible implementation, the scale-invariant feature transform (SIFT) is implemented through an image pyramid, and color information and texture information are extracted from the target time-frequency domain atlas and the target coding atlas respectively, and then adaptively upsampled and reconstructed to obtain a target fusion atlas with complementary and dual-spectrum characteristics.
[0096] Step 14: Input the target fusion graph into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected;
[0097] Among them, the fault detection model is obtained by training an 18-layer residual neural network based on at least one time-frequency vibration signal of the gas-insulated switchgear with different mechanical defect types.
[0098] In this step, after obtaining the target fusion map corresponding to the target time-frequency vibration signal, the target fusion map is input into the fault detection model, so that the fault detection model processes the target fusion map to obtain the fault type of the gas-insulated switchgear to be detected, which is recorded as the target fault type.
[0099] The present invention provides a method for fault detection of gas-insulated switchgear (GIS) based on multi-spectrum fusion. The method comprises obtaining a target time-frequency vibration signal of the GIS to be detected; determining a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal; and determining a target fusion spectrum based on the target time-frequency domain spectrum and the target coding spectrum. The target fusion spectrum is then input into a pre-trained fault detection model to obtain a target fault type of the GIS to be detected. The fault detection model is obtained by training an 18-layer residual neural network based on at least one time-frequency vibration signal of the GIS with different mechanical defect types. In this technical solution, by obtaining the time-frequency vibration signal of the GIS to be detected, characteristic information generated by the device during operation due to faults can be captured. After processing these signals, corresponding time-frequency domain spectra and coding spectra can be generated, reflecting the frequency changes and fault modes of the device under different states. Subsequently, a target fusion spectrum is determined based on the time-frequency domain spectrum and coding spectrum. This spectrum integrates information from different frequency bands, thereby more comprehensively describing the operating status of the device. The target fusion spectrum is then input into the pre-trained fault detection model to accurately identify the fault type of the device. This method can not only efficiently detect equipment failures, but also improve the accuracy and robustness of fault diagnosis, ensure real-time monitoring and fault prediction of equipment, and ultimately improve equipment maintenance efficiency and operational safety.
[0100] Based on the above embodiments, Figure 2Schematic diagram of the process of the fault detection method of gas insulated switchgear based on multi-graph fusion provided in the embodiment of the present application Figure 2 ,like Figure 2 As shown, step 13 may include:
[0101] Step 21: extracting target color information from the target time-frequency domain spectrum;
[0102] Optionally, a possible implementation of step 21 may be: performing Gaussian smoothing and downsampling processing on the target time-frequency domain spectrum for a preset number of times to obtain target color information.
[0103] In this implementation, the target time-frequency domain map is subjected to five Gaussian smoothing and downsampling processes to obtain five groups of images with progressively decreasing scales and progressively increasing smoothness. The top approximate image of the target time-frequency domain map is retained (the difference is less than a first preset threshold) as the color feature during reconstruction to form the Gaussian pyramid-color, that is, the target color information.
[0104] Step 22: extracting target texture information from the target coding atlas;
[0105] Optionally, a possible implementation of step 22 may be: performing Gaussian smoothing and downsampling processing on the target coding atlas for the preset number of times to obtain target texture information.
[0106] In this implementation, the target coding atlas is Gaussian smoothed and downsampled five times to collect the texture information lost in the five-time smoothing and downsampling process of the Gram angular field atlas to form a Laplacian pyramid-texture, which is recorded as the target texture information.
[0107] Step 23: reconstruct and fuse the target texture information and the target color information to obtain a target fusion atlas.
[0108] Optionally, a possible implementation of step 23 may be: performing Gaussian smoothing and downsampling corresponding restoration processing on the target texture information and the target color information to obtain a target fusion atlas.
[0109] In this implementation, the target texture information extracted from the target coding atlas and the target color information of the target time-frequency domain atlas are combined to restore the image to the original target time-frequency domain atlas or the original scale corresponding to the target coding atlas through upsampling. The restored image contains the texture information of the target coding atlas and the color information of the target time-frequency domain atlas.
[0110] Based on the embodiment of the present application, the Laplacian pyramid image calculation process is expressed as follows:
[0111]
[0112] Where Li represents the Laplacian pyramid image of the i-th layer, Gi represents the Gaussian image of the i-th layer, Up represents the upsampling position mapping process, represents convolution, k 5×5 For the convolution kernel used, the overall adaptive weight atlas fusion process is as follows Figure 3 shown.
[0113] Figure 3 A schematic diagram of determining the target fusion map provided in the embodiment of the present application is shown as follows: Figure 3 As shown, the Gram angular field spectrum (ie, target coding spectrum) and the short-time Fourier time spectrum (ie, target time-frequency domain spectrum), as well as the adaptive weight fusion spectrum (ie, target fusion spectrum).
[0114] The fault detection method for gas-insulated switchgear based on multi-atlas fusion provided in an embodiment of the present application extracts target color information from the target time-frequency domain atlas, extracts target texture information from the target coding atlas, and reconstructs and fuses the target texture information and the target color information to obtain the target fusion atlas. This technical solution can effectively combine color and texture information to enhance the expressive power of the atlas, making it more accurately reflect the operating status of the equipment and potential fault characteristics. In this way, the fault detection model can more comprehensively analyze the time-frequency signals of the equipment, thereby improving the accuracy and reliability of fault diagnosis and enhancing the accuracy and robustness of equipment monitoring.
[0115] Based on the above embodiments, Figure 4 Schematic diagram of the process of the fault detection method of gas insulated switchgear based on multi-graph fusion provided in the embodiment of the present application Figure 3 ,like Figure 4 As shown, before step 14, the following steps may also be included:
[0116] Step 41: Acquire at least one time-frequency vibration signal of different mechanical defect types;
[0117] In this step, for different mechanical defect types, at least one time-frequency vibration signal corresponding to each mechanical defect type is obtained.
[0118] In this implementation, the collected time-frequency vibration signal can also be filtered out of interference signals based on the wavelet packet decomposition filtering method to ensure the reliability of subsequent fault detection model training.
[0119] Optionally, the different types of mechanical defects include at least one of the following: normal operating conditions, loose bellows bolts, and foreign matter in the busbar air chamber bolts.
[0120] Step 42: Train an 18-layer residual neural network based on at least one time-frequency vibration signal of different mechanical defect types to obtain a fault detection model.
[0121] In this implementation, the ResNet18 model is mainly composed of a feature extraction layer and a defect classification layer. The feature extraction layer consists of 18 layers of convolution and maximum pooling layers. The specific structural parameters are shown in Table 1 (Table 1 is a parameter diagram of the ResNet model feature extraction layer provided in the embodiment of the present application).
[0122] Table 1
[0123]
[0124] Optionally, the processed fusion map is subjected to convolution operation to obtain a feature map, and the map is downsampled through a maximum pooling layer to reduce the scale of the feature map. The features extracted by convolution are sent to a defect classification layer composed of average pooling, a fully connected layer and Softmax to obtain the probability value of each fault category. The category with the highest probability is the recognition result, that is, the mechanical defect type corresponding to the fusion map.
[0125] Optionally, a possible implementation of step 42 is:
[0126] Step 1: for each time-frequency vibration signal, determine the time-frequency domain spectrum and coding spectrum corresponding to the time-frequency vibration signal;
[0127] In this implementation, reference may be made to the method for determining the time-frequency domain spectrum and the coding spectrum in the above-mentioned manner, and the implementation principle is similar.
[0128] Step 2: Determine the fusion spectrum based on the time-frequency domain spectrum and the coding spectrum;
[0129] In this implementation, the method of determining the fusion map in the above method can still be referred to, and the implementation principle is similar.
[0130] Step 3: Based on the fusion graph corresponding to at least one time-frequency vibration signal, the 18-layer residual neural network is trained to obtain a fault detection model.
[0131] In this implementation, an 18-layer residual neural network is trained using a fusion map corresponding to at least one time-frequency vibration signal until the classification accuracy and recall rate corresponding to the training results are both within a preset threshold, such as above 99%.
[0132] In one possible implementation, under the conditions of different mechanical defect types, the number of samples for each load current is set to 200, and the number of samples for the five currents is 1000. All samples are randomly divided into a training set (2250) and a test set (750) in a ratio of 3:1. An adaptive weight algorithm fusion map of the normal signal and the two defect vibration signals is established and fed into the GIS mechanical defect recognition and classification model as input. The confusion matrix of the classification result is as follows: Figure 3 As shown in the figure, the classification accuracy and recall rate of the test set are both above 99%, and the classification effect is relatively good.
[0133] Optional, Figure 5 A schematic diagram of the confusion matrix of the test set diagnosis results provided in the embodiment of this application is shown in FIG. Figure 5 As shown, the precision is 99.6% and the recall is 99.6%.
[0134] Accuracy and recall are used as evaluation indicators for the proposed mechanical defect classification model. Accuracy refers to the ratio of correctly classified samples, while recall refers to the proportion of correctly predicted faults among events identified as fault samples, measuring the model's ability to correctly identify faults. The expressions for accuracy and recall are as follows:
[0135]
[0136] Among them, TN and TP are the correctly identified normal and faulty samples, while FN and FP are the incorrectly identified normal and faulty samples.
[0137] It should be understood that in the embodiment of the present application, fault detection of gas-insulated switchgear under the same current is implemented. When detecting gas-insulated switchgear under different currents, the time-frequency vibration signals collected from the gas-insulated switchgear under the corresponding current can be used to retrain the fault detection model, and the implementation principle is similar.
[0138] The multi-graph fusion-based fault detection method for gas-insulated switchgear provided in an embodiment of the present application obtains at least one time-frequency vibration signal for different mechanical defect types and trains an 18-layer residual neural network based on the at least one time-frequency vibration signal for each mechanical defect type to obtain a fault detection model. In this technical solution, the fault detection model can learn complex time-frequency features from multiple mechanical defect types and optimize the model's learning capabilities through a deep residual network, thereby effectively improving the accuracy and robustness of fault diagnosis.
[0139] Based on the above embodiments, Figure 6 Schematic diagram of the process of the fault detection method of gas insulated switchgear based on multi-graph fusion provided in the embodiment of the present application Figure 4 ,like Figure 6As shown, the method includes:
[0140] Step 61: Collection of vibration signals of typical mechanical defects of GIS equipment and sample preprocessing;
[0141] Step 62: Establishing a fusion spectrum of multiple vibration signal spectra of GIS equipment using an adaptive weight algorithm;
[0142] Step 63: constructing a GIS equipment mechanical defect identification and classification model (i.e., a fault detection model);
[0143] Step 64: Construction and evaluation of performance evaluation indicators for the GIS equipment mechanical defect identification and classification model.
[0144] The fault detection method for gas-insulated switchgear based on multi-graph fusion provided in the embodiment of the present application has similar implementation principles and technical effects to those of the above-mentioned embodiment and will not be repeated here.
[0145] The following is a description of the device embodiments provided in the embodiments of the present application.
[0146] Figure 7 A schematic diagram of the structure of a fault detection device for a gas-insulated switchgear based on multi-graph fusion provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the fault detection device for gas-insulated switchgear based on multi-graph fusion provided in this embodiment includes:
[0147] An acquisition module 71 is used to acquire a target time-frequency vibration signal of the gas-insulated switchgear to be detected;
[0148] A first determination module 72 is configured to determine a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal;
[0149] The second determining module 73 is used to determine the target fusion spectrum according to the target time-frequency domain spectrum and the target coding spectrum;
[0150] Processing module 74 is used to input the target fusion map into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected. The fault detection model is obtained by training an 18-layer residual neural network based on at least one time-frequency vibration signal of the gas-insulated switchgear with different mechanical defect types.
[0151] In one or more embodiments, the second determining module 73 is specifically configured to:
[0152] Extract target color information from the target time-frequency domain spectrum;
[0153] Extracting target texture information from the target coding atlas;
[0154] The target texture information and target color information are reconstructed and fused to obtain a target fusion map.
[0155] In one or more embodiments, the second determination module 73 extracts target color information from the target time-frequency domain spectrum, specifically for:
[0156] Perform Gaussian smoothing and downsampling on the target time-frequency domain spectrum for a preset number of times to obtain the target color information;
[0157] Accordingly, the second determination module 73 extracts the target texture information in the target coding atlas, specifically for:
[0158] Performing Gaussian smoothing and downsampling on the target coding map for a preset number of times to obtain target texture information;
[0159] Accordingly, the second determination module 73 reconstructs and fuses the target texture information and the target color information to obtain a target fusion atlas, which is specifically used to:
[0160] The target texture information and target color information are subjected to Gaussian smoothing and downsampling corresponding restoration processing to obtain a target fusion atlas.
[0161] In one or more embodiments, the first determining module 72 is specifically configured to:
[0162] Perform short-time Fourier transform on the target time-frequency vibration signal to obtain the target time-frequency domain spectrum;
[0163] Perform Gram angle field transform on the target time-frequency vibration signal to obtain the target coding spectrum.
[0164] In one or more embodiments, before inputting the target fusion graph into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected, the processing module 74 is further configured to:
[0165] acquiring at least one time-frequency vibration signal of different mechanical defect types;
[0166] According to at least one time-frequency vibration signal of different mechanical defect types, an 18-layer residual neural network is trained to obtain a fault detection model.
[0167] In one or more embodiments, the processing module 74 trains an 18-layer residual neural network based on at least one time-frequency vibration signal of different mechanical defect types to obtain a fault detection model, specifically for:
[0168] For each time-frequency vibration signal, determining a time-frequency domain spectrum and a coding spectrum corresponding to the time-frequency vibration signal;
[0169] Determine a fusion spectrum according to the time-frequency domain spectrum and the coding spectrum;
[0170] According to the fusion graph corresponding to at least one time-frequency vibration signal, an 18-layer residual neural network is trained to obtain a fault detection model.
[0171] In one or more embodiments, different mechanical defect types include at least one of the following: normal operating conditions, loose bellows bolts, and foreign matter in busbar air chamber bolts.
[0172] The fault detection device for gas-insulated switchgear based on multi-graph fusion provided in this embodiment can execute the fault detection method for gas-insulated switchgear based on multi-graph fusion provided in the above method embodiment. Its implementation principle and technical effects are similar, and are not described in detail in this embodiment.
[0173] Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present application. Figure 8 As shown, the electronic device provided in this embodiment includes:
[0174] At least one processor 81 and memory 82 .
[0175] Optionally, the device further includes a communication component 83 , wherein the processor 81 , the memory 82 and the communication component 83 are connected via a bus 84 .
[0176] During the specific implementation process, at least one processor 81 executes the computer-executable instructions stored in the memory 82, so that the at least one processor 81 performs the above method.
[0177] The specific implementation process of the processor 81 can be found in the above-mentioned method embodiment. Its implementation principle and technical effects are similar, and will not be repeated here in this embodiment.
[0178] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the present invention may be directly executed by a hardware processor or by a combination of hardware and software modules in the processor.
[0179] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory.
[0180] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be classified as address buses, data buses, and control buses. For ease of illustration, the buses in the drawings of this application are not limited to just one bus or just one type of bus.
[0181] The present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.
[0182] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.
[0183] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile memory device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.
[0184] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be an integral part of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist in a device as discrete components.
[0185] The division of units is merely a logical functional division; actual implementations may employ alternative divisions, such as combining or integrating multiple units or components into another system, or omitting or disabling certain features. Furthermore, any direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units, either through an interface, electrical, mechanical, or other means.
[0186] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.
[0187] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0188] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.
[0189] Those skilled in the art will appreciate that all or part of the steps in the above-described method embodiments can be implemented using hardware associated with program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.
[0190] Finally, it should be noted that those skilled in the art will readily identify other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art not disclosed herein. The present invention is not limited to the precise structure described above and illustrated in the accompanying drawings, and various modifications and variations may be made without departing from the scope thereof. The scope of the present invention is limited solely by the appended claims.
Claims
1. A fault detection method for gas-insulated switchgear based on multi-atlas fusion, characterized in that: include: Obtaining the target time-frequency vibration signal of the gas-insulated switchgear to be inspected; Determining a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal; Determining a target fusion spectrum according to the target time-frequency domain spectrum and the target coding spectrum; The target fusion map is input into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected. The fault detection model is obtained by training an 18-layer residual neural network based on at least one time-frequency vibration signal of the gas-insulated switchgear with different mechanical defect types.
2. The method according to claim 1, characterized in that The determining of a target fusion spectrum according to the target time-frequency domain spectrum and the target coding spectrum includes: Extracting target color information from the target time-frequency domain spectrum; Extracting target texture information from the target coding atlas; The target texture information and the target color information are reconstructed and fused to obtain the target fusion atlas.
3. The method according to claim 2, characterized in that The extracting target color information from the target time-frequency domain spectrum includes: Performing Gaussian smoothing and downsampling on the target time-frequency domain spectrum for a preset number of times to obtain target color information; Accordingly, the extracting target texture information from the target coding atlas includes: Performing Gaussian smoothing and downsampling processing on the target coding atlas for the preset number of times to obtain target texture information; Accordingly, the target texture information and the target color information are reconstructed and fused to obtain the target fusion atlas, including: The target texture information and the target color information are subjected to restoration processing corresponding to Gaussian smoothing processing and downsampling processing to obtain the target fusion atlas.
4. The method according to claim 1, wherein The determining of a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal includes: Performing short-time Fourier transform on the target time-frequency vibration signal to obtain the target time-frequency domain spectrum; Performing Gram angle field transform on the target time-frequency vibration signal to obtain the target coding spectrum.
5. The method according to any one of claims 1 to 4, characterized in that Before inputting the target fusion graph into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected, the method further includes: acquiring at least one time-frequency vibration signal of different mechanical defect types; The 18-layer residual neural network is trained according to at least one time-frequency vibration signal of different mechanical defect types to obtain the fault detection model.
6. The method according to claim 5, characterized in that The 18-layer residual neural network is trained according to at least one time-frequency vibration signal of different mechanical defect types to obtain the fault detection model, including: For each time-frequency vibration signal, determining a time-frequency domain spectrum and a coding spectrum corresponding to the time-frequency vibration signal; Determining a fusion spectrum according to the time-frequency domain spectrum and the coding spectrum; The 18-layer residual neural network is trained according to a fusion graph corresponding to at least one time-frequency vibration signal to obtain the fault detection model.
7. The method according to any one of claims 1 to 3, characterized in that Different mechanical defect types include at least one of the following: normal operating conditions, loose bellows bolts, and foreign matter in busbar air chamber bolts.
8. A fault detection device for gas-insulated switchgear based on multi-atlas fusion, characterized in that: include: An acquisition module, used for acquiring a target time-frequency vibration signal of the gas-insulated switchgear to be tested; A first determination module is used to determine a target time-frequency domain spectrum and a target coding spectrum corresponding to the target time-frequency vibration signal; A second determining module is used to determine a target fusion spectrum according to the target time-frequency domain spectrum and the target coding spectrum; A processing module is used to input the target fusion map into a pre-trained fault detection model to obtain the target fault type of the gas-insulated switchgear to be detected. The fault detection model is obtained by training an 18-layer residual neural network based on at least one time-frequency vibration signal of the gas-insulated switchgear with different mechanical defect types.
9. An electronic device, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.