A Fault Diagnosis Method and Device for Switchgear Based on Deep Learning

Abstract

The present invention discloses a switch cabinet fault diagnosis method and device based on deep learning, belonging to the technical field of power equipment. The method includes: S1, obtaining PD data; S2, preprocessing the PD data; the preprocessing content includes: extracting global statistical features for the PD data to obtain global statistical features; and performing data type conversion on the PD data to obtain a grayscale image; S3, using a ResNet model combined with a CBAM attention mechanism to process the grayscale image, and extracting key potential features as key local discriminant features; S4, splicing the global statistical features and the key local discriminant features to obtain multi-scale features; S5, using an MLP classifier for classification processing to obtain a switch cabinet fault diagnosis result. The present invention can analyze signal features from different levels and scales through multi-scale feature fusion, improving the accuracy and adaptability of fault diagnosis.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment, and particularly relates to a switch cabinet fault diagnosis method and device based on deep learning. Background Technique

[0002] In the process of power generation, power transmission, power distribution and power conversion in the power system, the main function of the switch cabinet is to switch on and off, control and protect electrical equipment, which directly affects the safety and stability of the high-voltage transmission process. However, during the process from production to transportation, assembly, operation to aging of the switch cabinet, it is inevitable to generate some minor physical defects, which are likely to cause distortion of the electric field intensity inside the switch cabinet equipment during power transmission, resulting in partial discharge (PD) phenomena. As the discharge phenomenon gradually deteriorates, it will seriously affect the stability of power transmission and even lead to large-scale power supply interruption.

[0003] Li Zhaoshu et al. designed an on-line partial discharge monitoring system for high-voltage switch cabinets based on the pulse current method to monitor the operating state of high-voltage switch cabinets and judge whether they are in a fault state. This system collects the ultra-high frequency signals of high-voltage switch cabinets through an ultra-high frequency sensor module and uses the second-generation wavelet denoising technology for signal denoising. At the same time, the optimal binary tree complex wavelet decomposition method is used to extract the characteristics of the discharge signal, and finally a local discharge feature fusion recognition model based on neural network and evidence theory is constructed.

[0004] Jia Kunqi et al. determined the partial discharge feature vector by collecting the ultrasonic signals of high-voltage switch cabinets, extracting waveform feature vectors, the first and second correlation coefficients, and the peak values of each signal segment. Subsequently, multi-information fusion is carried out by combining the ultra-high frequency detection method and these feature vectors, so as to realize the diagnosis of the partial discharge fault type of high-voltage switch cabinets.

[0005] These algorithms involve complex feature extraction processes and knowledge in related fields, and usually require expert knowledge guidance to achieve accurate pattern recognition, with a high dependence on experience and technology. In addition, only extracting features manually does not enable the model to fully understand and utilize the features of the original data, and it is difficult to achieve good fault diagnosis effects when the partial discharge patterns are similar. Summary of the Invention

[0006] The present invention aims to solve at least one of the technical problems in the above related technologies to a certain extent.

[0007] For this reason, the purpose of the present invention is to provide a switch cabinet fault diagnosis method and device based on deep learning, which can analyze signal features from different levels and scales through multi-scale feature fusion, and improve the fault diagnosis accuracy and adaptability.

[0008] To solve the above technical problems, the present invention is implemented as follows:

[0009] An embodiment of the present invention provides a switchgear fault diagnosis method based on deep learning, and the method includes:

[0010] S1. Obtain PD data;

[0011] S2. Preprocess the obtained PD data;

[0012] The preprocessing content includes: extracting global statistical features for the PD data to obtain global statistical features; and converting the data type of the PD data to obtain a grayscale image;

[0013] S3. Process the grayscale image using a ResNet model combined with a CBAM attention mechanism, and extract key potential features as key local discriminant features;

[0014] S4. Concatenate the global statistical features and the key local discriminant features to obtain multi-scale features;

[0015] S5. Use an MLP classifier to classify the multi-scale features to obtain a switchgear fault diagnosis result.

[0016] In addition, according to the switchgear fault diagnosis method based on deep learning of the present invention, the following additional technical features may also be included:

[0017] In some of the embodiments, the global statistical features include the number of pulses, the average absolute amplitude of the pulses, the standard deviation of the pulses, and the maximum and minimum values of the pulses.

[0018] In some of the embodiments, the PD data is a pulse sequence resolved by partial discharge phase or phase-resolved partial discharge data.

[0019] In some of the embodiments, the CBAM attention mechanism includes a channel attention mechanism and a spatial attention mechanism, and by combining the channel attention mechanism and the spatial attention mechanism, the key potential features related to the PD pattern are effectively extracted.

[0020] In some of the embodiments, the MLP classifier is a multi-layer perceptron designed based on multi-scale features, and the switchgear fault diagnosis result judged by it includes whether there is a fault and the type of the fault.

[0021] In some embodiments, the data type conversion in step S2 includes: organizing the PD data into a two-dimensional data matrix, normalizing the data matrix, multiplying the normalized data by 255 and converting it into an integer format to adapt to the grayscale image pixel value range of 0 to 255, thereby obtaining a grayscale image that can clearly display the spatial distribution and intensity of partial discharge activity.

[0022] In some embodiments, the CBAM attention mechanism is set in the basic block of the ResNet model.

[0023] In some embodiments, the channel attention mechanism includes an average pooling layer and a maximum pooling layer, a fully connected layer, a ReLU module and a fully connected layer arranged in parallel;

[0024] The formula of the channel attention mechanism is:

[0025]

[0026] Where F represents the input feature map, σ represents the sigmoid activation function, W0 and W1 represent the weights of the two fully connected layers respectively, and M c Represents the output channel attention feature map.

[0027] In some embodiments, the spatial attention mechanism includes a maximum pooling layer, an average pooling layer, and a convolutional layer arranged sequentially and adjacently;

[0028] The formula of the spatial attention mechanism is:

[0029]

[0030] In the formula, F represents the input feature map, σ represents the sigmoid activation function, and f 7×7 Represents a 7x7 convolution operation, M s Represents the output spatial attention feature map.

[0031] An embodiment of the present invention also provides a switch cabinet fault diagnosis device based on deep learning, including a processor and a memory, wherein a program is stored on the memory, and when the processor runs the program on the memory, it can implement the content of the switch cabinet fault diagnosis method based on deep learning as described in any one of the above items.

[0032] Compared with the prior art, the present invention has at least the following beneficial effects:

[0033] In an embodiment of the present invention, a switchgear fault diagnosis method based on deep learning is provided, which can effectively mine the key features of partial discharge patterns by making full use of the information in the original data, thereby enhancing the recognition ability of the model;

[0034] In an embodiment of the present invention, the provided switchgear fault diagnosis method based on deep learning enables the model to analyze signal features from different levels and scales through multi-scale feature fusion, improving the adaptability to complex patterns;

[0035] In an embodiment of the present invention, the provided switchgear fault diagnosis method based on deep learning adopts an automated feature extraction process of deep learning, reducing the dependence on domain knowledge and enhancing the universality and application potential of the system;

[0036] In an embodiment of the present invention, the provided switchgear fault diagnosis method based on deep learning combines global statistical features with local discriminative features extracted by deep learning, reflecting different dimensional states of partial discharge signals, effectively reducing the influence of noise and abnormal data on the recognition results, and ensuring reliable fault recognition in complex environments;

[0037] In an embodiment of the present invention, the provided switchgear fault diagnosis method based on deep learning combines multi-level feature extraction, attention mechanism and deep learning model, showing high accuracy in fault detection and type recognition. Through precise feature expression and information integration, the model can quickly identify potential problems and provide accurate fault analysis, thus significantly improving the diagnosis efficiency;

[0038] In an embodiment of the present invention, the provided switchgear fault diagnosis method based on deep learning, due to the universality of the structure design and feature extraction method, enables the model to adapt to the monitoring requirements of different environments, can be extended to fault diagnosis in different application scenarios and environments, and has good application potential.

[0039] The switchgear fault diagnosis device based on deep learning of the present invention can implement the content of the above-mentioned switchgear fault diagnosis method based on deep learning, and thus at least has all the features and advantages of the above-mentioned switchgear fault diagnosis method based on deep learning, which will not be elaborated herein. The additional aspects and advantages of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flowchart of the switchgear fault diagnosis method based on deep learning disclosed in an embodiment of the present invention;

[0041] Figure 2 is a schematic diagram of data preprocessing disclosed in an embodiment of the present invention;

[0042] Figure 3 is a schematic diagram of the ResNet18 network structure disclosed in an embodiment of the present invention;

[0043] Figure 4 Schematic diagram of the CBAM attention mechanism structure disclosed in an embodiment of the present invention;

[0044] Figure 5 Schematic diagram of the channel attention mechanism structure disclosed in an embodiment of the present invention;

[0045] Figure 6 Schematic diagram of the spatial attention mechanism structure disclosed in an embodiment of the present invention;

[0046] Figure 7 Confusion matrix of PRPS fault diagnosis disclosed in an embodiment of the present invention;

[0047] Figure 8 Confusion matrix of PRPS fault mode classification disclosed in an embodiment of the present invention;

[0048] Figure 9 Confusion matrix of PRPD fault diagnosis disclosed in an embodiment of the present invention;

[0049] Figure 10 Confusion matrix of PRPD fault mode classification disclosed in an embodiment of the present invention. Specific embodiments

[0050] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, rather than all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0051] Next, the embodiments of the present invention will be described in detail through specific embodiments and their application scenarios in conjunction with the accompanying drawings.

[0052] Please refer to Figure 1 As shown, in some embodiments of the present invention, a switchgear fault diagnosis method based on deep learning is provided. Feature learning is performed using a deep neural network, channel attention and spatial attention are combined, and discriminative latent features related to partial discharge patterns are focused on. In addition, the signal statistical features extracted manually and the deep latent features are combined to determine whether a fault exists, and further determine the fault type when a fault exists.

[0053] Based on the locally acquired partial discharge phase-resolved pulse sequence (Phase Resolved Pluse Sequence, PRPS) or phase-resolved partial discharge (Phase Resolved Partial Discharge, PRPD) data, the following steps are processed.

[0054] (1) PD data preprocessing mainly includes statistical feature extraction and local data transformation, extracting the global statistical information of PD data and transforming the original data into a single-channel three-dimensional image matrix, i.e., grayscale image data;

[0055] (2) For the transformed grayscale image, feature learning is carried out using the ResNet network combined with the attention mechanism; to solve the problem of insufficient feature extraction ability of the traditional ResNet network, the present invention introduces the CBAM (Convolutional Block Attention Module) attention mechanism, focusing on the useful information related to the PD pattern, and obtaining the key local discriminant features expressing the PD category pattern through deep feature extraction; the statistical feature variables expressing the global information of the PD signal and the deep extraction information expressing the local information are spliced and fused to obtain multi-scale features;

[0056] (3) Design a multi-layer perceptron (MLP) based on the multi-scale features, which can simultaneously realize the diagnosis of the presence or absence of faults and the fault types.

[0057] In some embodiments of the present invention, the content of data preprocessing includes:

[0058] 1) Extraction of global statistical information

[0059] The number of pulses as a global feature for partial discharge recognition has significant advantages. It directly reflects the activity degree of partial discharge and provides an important indication of the equipment health status. By monitoring the change in the number of pulses per unit time, potential faults can be recognized in real time, the trend can be analyzed, and it can be judged whether the discharge phenomenon is gradually intensifying. In addition, the calculation of the number of pulses is relatively simple, easy to understand, convenient for quick decision-making, and not easily affected by single abnormal data, thus improving the reliability of recognition. The comparison of the number of pulses between different devices can also help to quickly identify abnormal situations and improve the monitoring efficiency.

[0060] The average absolute amplitude of the pulses refers to the average value of the absolute amplitudes of all partial discharge pulses within a certain time window, which can effectively reflect the overall intensity of the partial discharge signal. By calculating the absolute amplitude of each pulse and averaging it, this feature not only provides an important basis for fault diagnosis but also has strong anti-interference ability because using the absolute value can reduce the influence of negative amplitudes on the result. In addition, the calculation of the average absolute amplitude is simple and easy to understand, which helps to quickly evaluate the operating state of the equipment, so it has important application value in partial discharge feature extraction.

[0061] The standard deviation of the pulses can effectively reflect the volatility of the partial discharge signal amplitude and provide a strong indication of the equipment's health status. A higher standard deviation usually means that there is greater instability or a higher risk of failure in the equipment. In addition, this feature is easy to calculate and understand in feature analysis, which helps to quickly evaluate the operating status of the equipment. Generally speaking, the standard deviation of the pulses is an important indicator in partial discharge identification and fault diagnosis.

[0062] The maximum and minimum values of the pulses can help quickly identify the abnormal behavior and fault status of the equipment. The level of the maximum value can indicate whether there is a serious fault, while the minimum value can provide information on the signal stability. These two features are simple and easy to calculate, which helps in real-time monitoring and quick decision-making, so they have important application values in partial discharge identification and fault diagnosis.

[0063] 2) Transformation of data image form

[0064] During the monitoring process, PD signals are usually collected by sensors or detection devices. These signals can be organized into a two-dimensional data matrix. In order to effectively convert the data into an image, the data matrix needs to be normalized. This operation scales the amplitude of the signal to the range of [0, 1], which is convenient for subsequent image generation. The steps of normalization include finding the minimum and maximum values of the data matrix, and then applying the following formula:

[0065]

[0066] Through normalization, the contrast of the image can be enhanced, making the signal features more obvious in the image. The normalized data matrix can be converted into a grayscale image through an image processing library. In this step, the normalized data is multiplied by 255 and converted to an integer format (uint8) to adapt to the pixel value range of the grayscale image (0 to 255). The finally generated image can clearly show the spatial distribution and intensity of the partial discharge activity, and the process is as Figure 2 shown.

[0067] In some embodiments of the present invention, the content of the ResNet model combined with the CBAM attention mechanism includes:

[0068] The deep neural network for PD signal pattern feature extraction adopts as Figure 3The shown ResNet18 structure inputs the preprocessed grayscale image data into the ResNet model. The image passes through multiple residual blocks, and each residual block consists of multiple convolutional layers, which are responsible for extracting features at different levels. By introducing skip connections, the network can retain the input information during training, ensuring that each layer can learn useful features. During the feature extraction process, batch normalization is applied after the convolutional layer to accelerate training and stabilize the performance of the model. It standardizes the output of each layer, making the distribution of the input to each layer more consistent, thereby reducing the fluctuations during training. In addition, the ReLU activation function introduces non-linear features, enabling the network to learn more complex patterns. As the image is passed layer by layer in the network, the model gradually extracts multi-level information from low-level features (such as edges and textures) to high-level features (such as partial discharge patterns).

[0069] To enhance the key feature extraction ability of the model in large-scale data processing, the CBAM attention module is introduced into ResNet as Figure 4 shown. This module mainly includes a channel attention mechanism and a spatial attention mechanism. By combining and complementing the two attention mechanisms, it effectively extracts key information related to the PD pattern.

[0070] The operating mechanism of channel attention is as Figure 5 shown. The input feature map passes through parallel average pooling layer (AvgPool) and max pooling layer (MaxPool). Without changing the channels, the spatial size is compressed to obtain two feature maps. Then, through a shared MLP module, the channels of the feature maps are first compressed and then expanded to the original number of channels, enhancing the channel feature extraction ability through dimensional transformation. After passing through the ReLU activation function, they are added element by element. In order to convert the features into attention scores, a sigmoid activation function is calculated. The attention formula for this process is shown as follows:

[0071]

[0072] In the formula, F represents the input feature map, σ represents the sigmoid activation function, W0 and W1 respectively represent the weights of two fully connected layers, and M c represents the output channel attention feature map.

[0073] Global average pooling and max pooling reflect the importance of different channels in the feature extraction process. The model uses these descriptors to calculate the weights of each channel, enabling the network to focus on the channels that are most crucial for differentiating partial discharge features. This attention not only enhances the response to effective signals but also suppresses the influence of redundant or noisy information.

[0074] In addition to the above compression and dilation at the channel level, in order to extract the key features of PD signals at the spatial level, the output features of the channel attention mechanism are transmitted into Figure 6 the spatial attention module shown in the figure. First, the feature channels are compressed through max pooling and average pooling to reflect the importance of each spatial position in the feature map. Then, the two features are concatenated through a Concat operation and transformed into a feature map with a channel of 1 through a convolutional layer. After passing through the sigmoid function, the feature variables are converted into attention scores, enabling the network to emphasize the area where the partial discharge activity is most concentrated, thereby more effectively extracting spatial features. The formula is shown as follows:

[0075]

[0076] In the formula, F represents the input feature map, σ represents the sigmoid activation function, and f 7×7 represents the 7x7 convolutional operation, and M s represents the output spatial attention feature map.

[0077] Finally, the CBAM module is placed in the basic block (Basic Block) of the ResNet network, weakening the model's attention to irrelevant information. The model can not only more accurately identify the partial discharge pattern but also maintain robustness in complex backgrounds and noise environments, reducing the risk of false alarms and missed detections in the actual monitoring data processing, thereby improving the overall diagnostic accuracy.

[0078] In some embodiments of the present invention, the content of the MLP classifier based on multi-scale features includes:

[0079] Based on the global statistical features of the partial discharge (PD) signal and the key local discriminant features extracted by the deep neural network, a diagnostic classifier using MLP is designed. The classifier mainly includes a shared fully connected layer and two independent fully connected classifiers, which are used to determine whether there is a fault in the PD signal and to determine the specific type of the fault, respectively.

[0080] 1) Shared fully connected layer: This layer receives the input from the feature extraction part, performs preliminary feature fusion, and extracts a more abstract feature representation. The shared layer can make full use of the complementary information of the global statistical features and local discriminant features to enhance the learning ability of the model.

[0081] 2) Fully connected classifier: The fault detection classifier is used to determine whether there is a fault in the PD signal. By outputting a binary classification result (faulty / non-faulty), it helps to quickly identify potential problems; the fault type classifier is used to further subdivide the fault type, such as internal discharge, external discharge, and corona discharge, etc. This layer provides multi-level fault classification, which helps to more accurately locate and diagnose the specific fault cause.

[0082] Through this multi-level classification design, the model can not only improve the accuracy of fault detection but also conduct in-depth analysis in fault type recognition. The shared fully connected layer effectively fuses different features, enhancing the model's recognition ability for complex patterns and ultimately improving the overall diagnostic accuracy. The flexibility and hierarchy of this structure enable the model to quickly and accurately make judgments when facing diverse PD signals, thus providing more reliable support in power equipment maintenance.

[0083] The present invention combines global statistical features with local discriminant features extracted by deep learning. Through multi-level information fusion, it captures the multi-dimensional information of the signal, helps the model to more comprehensively understand the partial discharge pattern, and enhances the model's feature expression ability. The adopted CBAM attention module combines channel attention and spatial attention, effectively strengthening the extraction of key information. The channel attention mechanism weights features according to the importance of different channels, and the spatial attention mechanism emphasizes the important spatial positions in the feature map. This mechanism ensures that the model can focus on the area where the partial discharge activity is most concentrated. The present invention adopts the design of a shared fully connected layer and an independent classifier, making the model more flexible in fault detection and type recognition. The shared layer can fuse different features and enhance the learning ability, while the independent classifier separately processes the presence of faults and type recognition, improving the model's performance ability in complex patterns.

[0084] The following are two experiments of the actual application of the present invention and their results.

[0085] Example 1:

[0086] Based on the PRPS data collected by ultrasonic sensors, deep feature extraction and pattern recognition tasks are carried out, and the model parameters are set as shown in Table 1.

[0087] Table 1 Model parameter settings for PRPS data experiments

[0088]

[0089] The MLP classifier designed through global features and deep key features can effectively conduct effective fault diagnosis and pattern classification on the equipment data of actual substation switchgear. Specifically, PRPS data monitoring is carried out on three different switchgears on-site. It is known that the partial discharge patterns of the three switchgears are different, divided into no partial discharge, low-intensity partial discharge, and high-intensity partial discharge. Feature extraction and fault diagnosis are carried out on the PRPS data of the three different switchgears. The input data is the grayscale image obtained by normalizing and image-transforming the PRPS data. After deep feature extraction by the deep neural network, the discriminative latent features and the global statistical features of the PRPS data are fused and input into the MLP classifier, and the classifier can effectively conduct fault diagnosis and pattern recognition. The recognition results are shown in Table 2.

[0090] Table 2 PRPS data experimental pattern recognition accuracy

[0091]

[0092] Among them, the fault diagnosis accuracy rate reaches 99.71%, the fault classification accuracy rate reaches 98.84%, and the fault diagnosis confusion matrix is as shown in Figure 7 shown, and the fault mode classification confusion matrix is as shown in Figure 8 shown. 0 represents no fault and no partial discharge occurs. 1 indicates that partial discharge occurs in the switch cabinet with a low partial discharge intensity. 2 indicates that partial discharge occurs in the switch cabinet with a high partial discharge intensity. The partial discharge pattern recognition algorithm proposed by the present invention can effectively classify the monitored PRPS data for partial discharge, and the classification accuracy is relatively high.

[0093] Example 2:

[0094] Based on the PRPD data collected by ultrasonic sensors, deep feature extraction and pattern recognition tasks are carried out, and the model parameters are set as shown in Table 3.

[0095] Table 3 PRPD data experimental model parameter settings

[0096]

[0097] The MLP classifier designed through global features and deep key features can effectively perform effective fault diagnosis and pattern classification on the equipment data of actual substation switch cabinets. Specifically, the PRPD data of three different switch cabinets on site are monitored. It is known that the partial discharge patterns of the three switch cabinets are different, and are divided into no partial discharge, low-intensity partial discharge, and high-intensity partial discharge. Feature extraction and fault diagnosis are carried out on the PRPD data of the three different switch cabinets. The input data is the grayscale image obtained by normalizing and image-transforming the PRPD data. Through the deep feature extraction of the deep neural network, the discriminative latent features and the global statistical features of the PRPD data are fused and input into the MLP classifier, and the classifier can effectively perform fault diagnosis and pattern recognition. The recognition results are shown in Table 4.

[0098] Among them, the fault diagnosis accuracy rate reaches 99.71%, the fault classification accuracy rate reaches 98.55%, and the fault diagnosis confusion matrix is as shown in Figure 9 shown, and the fault mode classification confusion matrix is as shown in Figure 10 shown. 0 represents no fault and no partial discharge occurs. 1 indicates that partial discharge occurs in the switch cabinet with a low partial discharge intensity. 2 indicates that partial discharge occurs in the switch cabinet with a high partial discharge intensity. The partial discharge pattern recognition algorithm proposed by the present invention can effectively classify the monitored PRPS data for partial discharge, and the classification accuracy is relatively high.

[0099] Table 4 PRPD data experimental pattern recognition accuracy

[0100]

[0101] For those parts not described in detail in the present invention, reference may be made to the prior art or to the well-known techniques to those skilled in the art.

[0102] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the spirit of the present invention and the scope protected by the claims, and all of them fall within the protection scope of the present invention.

Claims

1. A switch cabinet fault diagnosis method based on deep learning, characterized in that: The method comprises: S1. Obtain PD data; S2. Preprocessing the acquired PD data; The preprocessing includes: extracting global statistical features of the PD data to obtain global statistical features; and converting the data type of the PD data to obtain a grayscale image; S3. Processing the grayscale image using a ResNet model combined with a CBAM attention mechanism to extract key latent features as key local discriminant features; S4, performing feature splicing on the global statistical features and the key local discriminant features to obtain multi-scale features; S5. Using an MLP classifier to classify the multi-scale features to obtain a switchgear fault diagnosis result; The global statistical features include the number of pulses, the average absolute amplitude of the pulses, the standard deviation of the pulses, and the maximum and minimum values of the pulses; The PD data is a phase-resolved pulse sequence of partial discharge or phase-resolved partial discharge data; The MLP classifier is a multi-layer perceptron designed based on multi-scale features, and the switch cabinet fault diagnosis results determined by the MLP classifier include the presence or absence of a fault and the type of fault.

2. The switch cabinet fault diagnosis method based on deep learning according to claim 1 is characterized in that: The CBAM attention mechanism includes a channel attention mechanism and a spatial attention mechanism. By combining the channel attention mechanism and the spatial attention mechanism, the key potential features related to the PD pattern can be effectively extracted.

3. The switch cabinet fault diagnosis method based on deep learning according to claim 1 is characterized in that: The data type conversion in step S2 includes: organizing the PD data into a two-dimensional data matrix, normalizing the data matrix, multiplying the normalized data by 255 and converting it into an integer format to adapt to the grayscale image pixel value range of 0 to 255, thereby obtaining a grayscale image that can clearly show the spatial distribution and intensity of partial discharge activity.

4. The switch cabinet fault diagnosis method based on deep learning according to claim 1, characterized in that: The CBAM attention mechanism is set in the basic block of the ResNet model.

5. The switch cabinet fault diagnosis method based on deep learning according to claim 2, characterized in that: The channel attention mechanism includes an average pooling layer and a maximum pooling layer set in parallel, a fully connected layer, a ReLU module and a fully connected layer; The formula of the channel attention mechanism is: Where F represents the input feature map, σ represents the sigmoid activation function, W0 and W1 represent the weights of the two fully connected layers respectively, and M c Represents the output channel attention feature map.

6. The switch cabinet fault diagnosis method based on deep learning according to claim 2, characterized in that: The spatial attention mechanism includes a maximum pooling layer, an average pooling layer, and a convolutional layer that are arranged sequentially and adjacently; The formula of the spatial attention mechanism is: In the formula, F represents the input feature map, σ represents the sigmoid activation function, and f 7×7 Represents a 7x7 convolution operation, M s Represents the output spatial attention feature map.

7. A switch cabinet fault diagnosis device based on deep learning, comprising a processor and a memory, wherein a program is stored in the memory, characterized in that: When the processor runs the program on the memory, it can implement the content of the switch cabinet fault diagnosis method based on deep learning as described in any one of claims 1 to 6.

Images