High-voltage power equipment partial discharge fault diagnosis method and system

Abstract

The invention belongs to the technical field of electric power equipment discharge fault diagnosis, and discloses a high-voltage electric power equipment partial discharge fault diagnosis method and system.The method deeply fuses ultrahigh frequency and ultrasonic signal features in signal processing, the complementarity of the ultrahigh frequency and ultrasonic signal features under strong electromagnetic interference is utilized, the system still keeps high accuracy in a complex interference scene, and the fault diagnosis accuracy is improved. The environment adaptability limitation is broken through. The convolutional neural network architecture adopts a double-branch 1D-CNN structure to respectively process two signals, enhance the feature extraction capability, accurately capture key features, and avoid feature loss caused by inter-signal interference. A two-way cross attention mechanism is introduced in the signal fusion link, deep semantic association is mined, the advantages of multiple sensors are exerted, and the analysis accuracy is improved. A residual classifier is introduced in the aspect of mode recognition, the gradient problem is relieved, network degradation is dealt with, feature transfer multiplexing is enhanced, system robustness is improved, multi-classification tasks are adapted, and diagnosis accuracy is remarkably improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of partial discharge fault diagnosis of high-voltage power equipment, and relates to a partial discharge fault diagnosis method and system for high-voltage power equipment. BACKGROUND

[0002] High-voltage power equipment is the core of power system to realize power transmission, distribution, control and protection, ensures the safe and efficient transmission of power from power plants to user end, and is the key infrastructure to maintain the stable operation of power grid. However, omissions may occur in many aspects such as manufacturing, installation, operation and maintenance, structural design and material selection, such as insufficient conductor surface machining precision, long-term operation leading to aging of the insulation material of high-voltage power equipment, which will cause uneven distribution of electric field inside the high-voltage power equipment, and then cause partial discharge, and even lead to internal insulation breakdown of high-voltage power equipment, causing serious electrical faults and endangering the safety of power grid, so it is necessary to develop an effective partial discharge fault diagnosis method for high-voltage power equipment.

[0003] At present, there are many partial discharge detection methods for research and practical engineering application, such as pulse current method, ultra-high frequency method, ultrasonic method, optical detection method and chemical detection method. The ultra-high frequency method is widely used due to its strong anti-noise and anti-interference performance, and good online monitoring ability. Field application experience shows that the electromagnetic environment of high-voltage power equipment and its vicinity is complex, especially when the background electromagnetic noise interference is serious or the tested product capacity is large, the sensitivity of the ultra-high frequency sensor to some defects is reduced, so only using a single ultra-high frequency signal cannot accurately identify the partial discharge fault of high-voltage power equipment, and then the deep learning algorithm with high recognition accuracy is invalid. The partial discharge identification method for high-voltage power equipment based on ultrasonic wave has strong anti-electromagnetic interference ability, and can accurately locate the partial discharge, and has been studied and applied more. However, the ultrasonic waveforms generated by different types of partial discharge are less different, and it is difficult to effectively distinguish the defect types only by amplitude and frequency. The method of combining ultra-high frequency-ultrasonic detection can compensate for the shortcomings of ultra-high frequency sensor in the presence of electromagnetic interference, and on the other hand, the ultra-high frequency signal can help the ultrasonic signal to better distinguish different partial discharge types, which is beneficial to improve the reliability of partial discharge identification and the accuracy of defect type judgment.

[0004] In order to fully tap the diagnostic information contained in the ultrahigh frequency signal and the ultrasonic signal, it is necessary to fuse the features extracted from the two kinds of signals to improve the accuracy and robustness of the diagnostic model. Feature fusion can be divided into three categories: data layer fusion, feature layer fusion and result layer fusion. The data layer fusion method directly splices or combines the original signal or the low-level features of the signal, retaining the most abundant and original information. Common methods include signal-level splicing, combined input after uniform normalization, and dimension reduction methods such as principal component analysis. This method has good information integrity, but when dealing with high-dimensional and nonlinear signals such as ultrahigh frequency and ultrasonic, the computational complexity is high, and the feature extraction ability of the subsequent model is also high. Feature layer fusion is to extract a certain level of signal features and then fuse them using deep learning algorithms. Typical methods such as Convolutional Neural Network (CNN) extract spatial features from the time-frequency graph of ultrahigh frequency and ultrasonic signals, and then fuse them through a fully connected layer. This method can effectively deal with complex nonlinear feature distribution and improve the feature expression ability of the model. However, only extracting features from the signal cannot fully capture the deep-level correlation between the two types of signals in the time and space dimensions, and there may be problems such as feature redundancy or weakening of key features when fusing two different signals. The introduction of attention mechanism can dynamically allocate the importance weight of each feature during the fusion process, making the model focus more on the areas or dimensions of ultrahigh frequency and ultrasonic signals that are more discriminative for diagnosis. Combining attention mechanism with CNN can help enhance the model's ability to recognize and select key features, achieving higher precision in joint diagnosis. Result layer fusion is to extract features and train models for ultrahigh frequency and ultrasonic signals respectively, and then fuse the final judgment results using ensemble learning, voting mechanism or weighted average method. This method is simple to implement and easy to expand, but it requires experience and knowledge, and its adaptability in complex fault scenarios is relatively weak. SUMMARY

[0005] In view of the problems in the prior art, the present application provides a high-voltage power equipment partial discharge fault diagnosis method and system, thereby solving the technical problem of low accuracy in the prior art when using ultrahigh frequency and ultrasonic combined method for high-voltage power equipment partial discharge fault diagnosis.

[0006] The present application is realized by the following technical solutions: A high-voltage power equipment partial discharge fault diagnosis method, comprising the following steps: S1: collecting ultrahigh frequency signals and ultrasonic signals of high-voltage power equipment; S2: completing fault diagnosis of high-voltage power equipment partial discharge based on the ultrahigh frequency signals, ultrasonic signals and a pre-constructed signal fusion model; The signal fusion model first uses a double-branch 1D-CNN network to extract features of the ultra-high frequency signal and the ultrasonic signal respectively, and outputs an ultra-high frequency signal feature vector and an ultrasonic signal feature vector; secondly, the ultra-high frequency signal feature vector and the ultrasonic signal feature vector are fused based on a cross-attention mechanism, and a fusion feature vector is output; finally, the fusion feature vector is linearly transformed based on a residual classifier to realize high-voltage power equipment partial discharge fault diagnosis.

[0007] Preferably, in step S1, an ultra-high frequency sensor is used to collect an ultra-high frequency signal, and an ultrasonic sensor is used to collect an ultrasonic signal.

[0008] Preferably, in step S2, the collected ultra-high frequency signal and ultrasonic signal are put into a signal fusion model pre-constructed in one-dimensional time domain signal.

[0009] Preferably, the double-branch 1D-CNN network includes a first branch and a second branch, the first branch includes three blocks, each block includes a convolution layer, a batch normalization layer and a pooling layer, and is used for processing the ultra-high frequency signal; the second branch also includes three blocks, each block includes a convolution layer, a batch normalization layer and a pooling layer, and is used for processing the ultrasonic signal.

[0010] Preferably, in the first branch and the second branch, the number of convolution kernels in each block is consistent, and is [16, 32, 64] respectively; for the pooling layer, maximum pooling is adopted in the first two blocks, and global average pooling is adopted in the last block.

[0011] Preferably, the ultra-high frequency signal feature vector and the ultrasonic signal feature vector are fused based on the cross-attention mechanism, and a fusion feature vector is output, specifically:

[0012]

[0013] wherein, is the attention of the ultrasonic channel to the ultra-high frequency channel, is the query matrix of the ultrasonic, is the key matrix of the ultra-high frequency, is the value matrix of the ultra-high frequency; is the dimension of the key vector; When constructing the attention of the ultrasonic-ultra-high frequency, Q comes from the ultra-high frequency, K and V come from the ultrasonic; that is, is the attention of the ultra-high frequency channel to the ultrasonic channel, is the query matrix of the ultra-high frequency, is the key matrix of the ultrasonic, is the value matrix of the ultrasonic.

[0014] A partial discharge fault diagnosis system for high-voltage power equipment, comprising: Data acquisition module: used to acquire ultra-high frequency signals and ultrasonic signals from high-voltage power equipment; Data processing module: used to complete the fault diagnosis of partial discharge in high-voltage power equipment based on the UHF signal, ultrasonic signal, and a pre-built signal fusion model; the signal fusion model first uses a dual-branch 1D-CNN network to extract features from the UHF signal and ultrasonic signal respectively, and outputs UHF signal feature vector and ultrasonic signal feature vector; secondly, it fuses the UHF signal feature vector and ultrasonic signal feature vector based on the cross-attention mechanism, and outputs a fused feature vector; finally, it performs a linear transformation on the fused feature vector based on a residual classifier to realize the fault diagnosis of partial discharge in high-voltage power equipment.

[0015] A computer device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described above.

[0016] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method.

[0017] A computer program product includes a computer program / instructions that, when executed by a processor, implement the steps of the above-described method.

[0018] Compared with the prior art, the present invention has the following beneficial technical effects: This invention discloses a method for diagnosing partial discharge faults in high-voltage power equipment. It effectively addresses the low accuracy problem often encountered in existing methods that combine UHF and ultrasound for partial discharge fault diagnosis in high-voltage power equipment. Firstly, at the signal processing level, this invention deeply integrates the features of UHF and ultrasound signals. Leveraging the complementary characteristics of these two signals in strong electromagnetic interference environments, the system maintains high diagnostic accuracy even in complex interference scenarios, overcoming the limitations of traditional methods in adaptability to different environments. Secondly, regarding the convolutional neural network architecture, this invention constructs a dual-branch 1D-CNN structure, designing processing paths separately for UHF and ultrasound signals. This invention specifically enhances feature extraction capabilities, accurately capturing key signal features and avoiding feature loss due to signal interference. Furthermore, in the signal fusion stage, it introduces a bidirectional cross-attention mechanism to deeply explore the semantic relationships between the two types of signals, fully leveraging the advantages of multiple sensors to make signal analysis more comprehensive and accurate, thus improving the accuracy of fault diagnosis. Finally, in pattern recognition, this invention introduces a residual classifier, using residual connections to alleviate gradient problems, address network degradation, enhance feature transfer and reuse, improve system robustness and training convenience, and perfectly adapt to multi-classification tasks, thereby significantly improving the accuracy of partial discharge fault diagnosis in high-voltage power equipment. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating a method for diagnosing partial discharge faults in high-voltage power equipment according to one embodiment of the present invention. Figure 2 This is a schematic diagram of the network structure of the signal fusion model in this invention; Figure 3 This is a flowchart illustrating a method for diagnosing partial discharge faults in high-voltage power equipment according to a specific embodiment of the present invention. Figure 4 This is a schematic diagram of the structure of a partial discharge fault diagnosis system for high-voltage power equipment according to the present invention; Figure 5 This is a schematic diagram of the structure of a partial discharge fault diagnosis system for high-voltage power equipment in a specific embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of a partial discharge fault diagnosis system for high-voltage power equipment in another specific embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0023] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0024] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper," "lower," "horizontal," or "inner" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of the invention is in use, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, terms such as "first" and "second" are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] Furthermore, the use of the term "horizontal" does not imply that the component must be absolutely horizontal, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0026] In the description of the embodiments of the present invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances.

[0027] The present invention will now be described in further detail with reference to the accompanying drawings: Example 1 like Figure 1 As shown, this invention discloses a method for diagnosing partial discharge faults in high-voltage power equipment, comprising the following steps: S1: Acquires ultra-high frequency signals and ultrasonic signals from high-voltage power equipment; S2: Based on the UHF signal, ultrasonic signal and the pre-built signal fusion model, complete the fault diagnosis of partial discharge in high-voltage power equipment; The signal fusion model first uses a dual-branch 1D-CNN network to extract features from UHF and ultrasonic signals respectively, outputting UHF signal feature vectors and ultrasonic signal feature vectors; secondly, it fuses the UHF signal feature vectors and ultrasonic signal feature vectors based on a cross-attention mechanism, outputting a fused feature vector; finally, it performs a linear transformation on the fused feature vector based on a residual classifier to achieve partial discharge fault diagnosis of high-voltage power equipment.

[0028] like Figure 2 As shown, the above process includes signal acquisition, feature extraction, feature fusion, and pattern recognition. Specifically, the above process is as follows: (1) In step S1, the signal acquisition process involves using an ultra-high frequency sensor to acquire ultra-high frequency signals and using an ultrasonic sensor to acquire ultrasonic signals. The acquired ultra-high frequency signals and ultrasonic signals are then transmitted as one-dimensional time-domain signals into the partial discharge fault diagnosis network, i.e., the signal fusion model.

[0029] (2) In the feature extraction process, this invention uses a two-branch one-dimensional convolutional neural network (1D-CNN) to extract features from ultra-high frequency (UHF) and acoustic emission (AE) signals respectively. Different sized convolutional kernels are used to capture multi-scale information of the signals. After processing by the feature extraction module, the two-branch one-dimensional convolutional neural network outputs feature vectors for the ultra-high frequency signals respectively. and the eigenvectors of ultrasound signals ; More specifically, a dual-branch 1D-CNN network module is used to process UHF signals and ultrasound signals respectively, extracting features from each channel of the two signals. The first branch is used to process UHF signals and includes three blocks. Each block is a combination of three layers: a convolutional layer, a batch normalization (BN) layer, and a pooling layer. The kernel sizes of the three convolutional layers are [7, 5, 3]. For high-frequency signals, smaller kernels are used to capture their rapidly changing signal features. The second branch is used to process ultrasound signals. Its structure is similar to the first branch, but its parameters are independent. The kernel sizes of the three convolutional layers are [11, 7, 5]. For low-frequency signals, a larger kernel is used to capture a wider range of features.

[0030] For both types of signals, the number of convolution kernels in the three block convolutional layers is distributed in the same way, namely [16, 32, 64]. As the number increases, the number of extracted feature maps also increases, which is beneficial for extracting deeper features.

[0031] Batch Normalization (BN) layer: Located after each convolutional layer and before the activation function. Default parameters are used (momentum = 0.99, epsilon = 0.001). BN layers help accelerate training convergence and improve model stability.

[0032] Pooling layer: Max pooling is used in the first two blocks and global average pooling is used in the last block. The window size and stride of each layer are set to 2. The pooling layer helps to reduce the feature dimension and enhance the translation invariance of the features.

[0033] (3) Then, in the feature fusion process, based on the cross-attention mechanism, the dependency and complementarity between the UHF signal and the ultrasound signal are established, the deep semantic association between the two signals is explored, and the attention-enhanced partial discharge signal identification features are obtained, i.e., the fused feature vector. Furthermore, specifically: This invention employs a cross-attention mechanism to fuse features of UHF and ultrasonic signals. This mechanism allows features from one sensor signal to focus on features from another sensor signal, achieving bidirectional flow of feature information. In the scenario of partial discharge fault diagnosis in high-voltage power equipment, the method in this invention constructs attention in two directions: 1. UHF to Ultrasound: Enabling ultrasound features to focus on important characteristic information in UHF.

[0034] 2. Ultrasound to UHF: Enabling UHF features to focus on important characteristic information in ultrasound.

[0035] Where Q is the query matrix from the first sequence, K is the key matrix from the second sequence, and V is the value matrix from the second sequence. is the dimension of the key vector, used for scaling and normalization. Where, This represents a row-level normalization operation performed on the input matrix, used to measure the degree of association between the query and each key, thereby obtaining the attention weight.

[0036] When constructing attention from UHF to ultrasound, Q comes from UHF, and K and V come from ultrasound; that is... Focusing attention on the UHF channel for the ultrasound channel. This is the query matrix for ultrasound. It is a very high frequency bond matrix. This is a matrix of values ​​for extremely high frequencies; When constructing attention from ultrasound to ultra-high frequency (UHF), Q comes from ultrasound, and K and V come from UHF; that is... To focus attention on the ultrasound channel in the ultra-high frequency channel. This is a query matrix with extremely high frequency. The bond matrix of ultrasound. This is the value matrix for ultrasound.

[0037] Through the aforementioned bidirectional cross-attention mechanism, this method can effectively capture the deep complementary information between UHF and ultrasonic signals, enhance the characterization ability of partial discharge features, and significantly improve the classifier's ability to distinguish between typical defects of partial discharge in high-voltage power equipment.

[0038] (4) Finally, the present invention performs a linear transformation on the fused feature vector output by the feature fusion module based on the residual classifier, and maps the features to the dimension space of different types of partial discharge defects, thereby realizing the fault diagnosis of partial discharge of high-voltage power equipment.

[0039] Furthermore, specifically: After completing feature fusion, this invention constructs a classifier with a residual connection structure, mapping the fused features of UHF and ultrasound to specific fault types. The specific steps are as follows: A one-dimensional feature vector is compressed into a fixed-length feature vector through a global pooling layer (max pooling); After global pooling, three residual blocks are stacked, and the structure of each residual block is as follows: Fully connected layer 1: Keeps the input and output dimensions consistent to avoid dimension mismatch, and its weight matrix is ​​W1; The Rectified Linear Unit (ReLU) performs a nonlinear transformation on the output of fully connected layer 1. Dropout layer: Randomly discards some neuron connections to avoid the model relying too much on certain features during training. In this method, the dropout probability is set to 0.2. Fully connected layer 2: Keeps the input and output dimensions consistent to avoid dimension mismatch, and its weight matrix is ​​W2.

[0040] After three residual blocks, a skip connection is introduced: the output of the global pooling layer is added to the output of the fully connected layer 2 of the last residual block; then the ReLU activation function is used to activate the output of the previous step.

[0041] After the above preprocessing, the output of the previous layer is finally mapped to the fault type space through the classification layer: the feature vector dimension is transformed into the number of target fault types through the fully connected layer, and the output of the fully connected layer is normalized using the normalized exponential function (i.e., the softmax activation function) to output the probability distribution of each fault type for partial discharge fault detection of high-voltage power equipment.

[0042] The network structure of the signal fusion model in this invention is as follows: Figure 2 As shown. The network model designed in this method is implemented based on Python 3.9 and the PyTorch deep learning framework. The specific environment configuration is: Intel Core i5-13600KF CPU processor, 32GB of RAM, and an NVIDIA GEFORCE RTX 1650 GPU with 24GB of storage. The model training uses the Adam optimizer, with an initial learning rate of 0.0001, a batch size of 16, and 100 epochs. After every 10 epochs, the learning rate is decayed to 0.5 times its original value to ensure more stable convergence and better model performance.

[0043] The UHF dataset consists of 100 corona discharge signals, 100 levitation potential discharge signals, 100 air gap discharge signals, and 100 metal particle discharge signals. The ultrasound dataset is the same as the UHF dataset. 70% of the total dataset is used for training, and 30% is used for testing.

[0044] Compared with existing single UHF detection methods and UHF-ultrasound combined detection methods, the detection method in this invention has significant advantages: In terms of signal processing, this method deeply integrates the features of UHF and ultrasound signals, fully utilizing the complementarity of the two signals. Even in environments with strong electromagnetic interference, the system can still maintain a high diagnostic accuracy. In terms of convolutional neural network architecture, an innovative dual-branch structure is built to process UHF and ultrasound signals separately, and the feature extraction capability for different signals is specifically enhanced to ensure accurate capture of key signal features.

[0045] Meanwhile, a bidirectional cross-attention mechanism is introduced to delve into the deep semantic relationships between the two types of signals, effectively leveraging the unique advantages of multi-sensor detection and making signal analysis more comprehensive and accurate. In the pattern recognition stage, a residual classifier is introduced, which, through residual connections, alleviates the gradient problem, addresses network degradation, enhances feature transfer and reuse, improves the system's robustness and training convenience, and perfectly adapts to the multi-classification task of high-voltage power equipment fault diagnosis.

[0046] Furthermore, the high-voltage power equipment partial discharge fault detection system proposed in this invention also possesses excellent application scalability. It can be used as an algorithm layer, forming part of an embedded diagnostic device for high-voltage power equipment. This system features high integration, low power consumption, and good industrial feasibility, enabling convenient deployment in substations and other field locations, providing robust fault diagnosis support for the stable operation of high-voltage power equipment.

[0047] Example 2 To further explain the technical solution of the present invention, the following embodiments are provided for detailed illustration: This embodiment provides a method for diagnosing partial discharge faults in high-voltage power equipment by fusing ultra-high frequency (UHF) and ultrasonic signals. This method fully utilizes the complementarity of UHF and ultrasonic signal information to achieve partial discharge fault diagnosis in high-voltage power equipment. Figure 3 As shown, the specific steps are as follows: UHF sensors are used to collect UHF signals from high-voltage power equipment, and ultrasonic sensors are used to collect ultrasonic signals from high-voltage power equipment. The UHF signals and ultrasonic signals are then transmitted to the partial discharge fault diagnosis network as one-dimensional time-domain signals.

[0048] A two-branch 1D-CNN network is used to extract features from the acquired Ultra High Frequency (UHF) and Acoustic Emission (AE) signals, respectively, and convolutional kernels of different sizes are used to capture multi-scale information of the signals. After processing by the feature extraction module, the two-branch 1D-CNN network outputs feature vectors for the UHF signals. and the eigenvectors of ultrasound signals ; Based on the cross-attention mechanism, the dependency and complementarity between UHF signals and ultrasound signals are established, the deep semantic association between the two signals is explored, the recognition features of partial discharge signals after attention enhancement are obtained, and the fused feature vector is output. By performing a linear transformation on the output fused feature vector based on the residual classifier, the features are mapped to the dimension space of different types of partial discharge defects, thereby realizing the fault diagnosis of partial discharge in high-voltage power equipment.

[0049] Example 3 In addition, such as Figure 4 As shown, the present invention also discloses a partial discharge fault diagnosis system for high-voltage power equipment, comprising: The data acquisition module is used to acquire ultra-high frequency signals and ultrasonic signals from high-voltage power equipment; The data processing module is used to perform fault diagnosis of partial discharge in high-voltage power equipment based on the ultra-high frequency signal, ultrasonic signal, and a pre-built signal fusion model. The signal fusion model first uses a dual-branch 1D-CNN network to extract features from UHF and ultrasonic signals respectively, outputting UHF signal feature vectors and ultrasonic signal feature vectors; secondly, it fuses the UHF signal feature vectors and ultrasonic signal feature vectors based on a cross-attention mechanism, outputting a fused feature vector; finally, it performs a linear transformation on the fused feature vector based on a residual classifier to achieve partial discharge fault diagnosis of high-voltage power equipment.

[0050] More specifically, such as Figure 5 As shown, the above system can effectively realize the signal acquisition, feature extraction, feature fusion and pattern recognition processes in this invention by setting a signal acquisition module, a feature fusion module, a pattern recognition module and a feature extraction module.

[0051] Example 4 Furthermore, to further explain the solution of the present invention, the following embodiments are provided: like Figure 6 As shown, this embodiment discloses a partial discharge fault diagnosis system for high-voltage power equipment, including: (1) Data Acquisition Module: A UHF data acquisition card is used to acquire UHF signals, and an ultrasonic data acquisition card is used to acquire ultrasonic signals, thus collecting sensor data. The sensor acquisition module is deployed on the external UHF sensor on the basin-type insulator of the high-voltage power equipment and the external high-voltage power equipment partial discharge interferometric fiber optic ultrasonic sensor on the casing of the high-voltage power equipment, used to simultaneously acquire UHF and ultrasonic signals. The sensors are signal-conditioned by a preamplifier circuit and output to the acquisition card.

[0052] (2) Intelligent diagnosis module: The signal fusion model described in this invention is deployed on the host computer to perform online identification and classification of high-voltage power equipment signals, so as to realize the diagnosis of partial discharge faults in high-voltage power equipment.

[0053] Additionally, a schematic diagram of a terminal device according to an embodiment of the present invention is provided. This terminal device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the various method embodiments described above. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the various device embodiments described above.

[0054] The computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention.

[0055] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0056] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0057] The memory can be used to store the computer program and / or module. The processor implements various functions of the terminal device by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory.

[0058] If the modules / units integrated into the terminal device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0059] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A partial discharge fault diagnosis method for high voltage power equipment, characterized by, The method comprises the following steps: S1: collecting ultra-high frequency signals and ultrasonic signals of high-voltage power equipment; S2: completing fault diagnosis of partial discharge of high-voltage power equipment based on the ultra-high frequency signals, ultrasonic signals and a pre-constructed signal fusion model; In the signal fusion model, first, a double-branch 1D-CNN network is used to extract features of the ultra-high frequency signals and ultrasonic signals respectively, and output ultra-high frequency signal feature vectors and ultrasonic signal feature vectors; second, the ultra-high frequency signal feature vectors and ultrasonic signal feature vectors are fused based on a cross-attention mechanism, and a fusion feature vector is output; finally, the fusion feature vector is linearly transformed based on a residual classifier, and fault diagnosis of partial discharge of high-voltage power equipment is realized.

2. The partial discharge fault diagnosis method for high voltage power equipment according to claim 1, characterized in that, In step S1, ultra-high frequency signals are collected by using an ultra-high frequency sensor, and ultrasonic signals are collected by using an ultrasonic sensor.

3. The partial discharge fault diagnosis method for high voltage power equipment according to claim 1, characterized in that, In step S2, the collected ultra-high frequency signals and ultrasonic signals are input into the pre-constructed signal fusion model in the form of one-dimensional time domain signals.

4. The partial discharge fault diagnosis method for high voltage power equipment according to claim 1, characterized in that, The double-branch 1D-CNN network comprises a first branch and a second branch, the first branch comprises three blocks, each block comprises a convolution layer, a batch normalization layer and a pooling layer, and is used for processing ultra-high frequency signals; the second branch also comprises three blocks, each block comprises a convolution layer, a batch normalization layer and a pooling layer, and is used for processing ultrasonic signals.

5. The partial discharge fault diagnostic method for a high voltage power equipment according to claim 4, characterized in that, In the first branch and the second branch, the number of convolution kernels in each block is consistent, and is [16, 32, 64] respectively; for the pooling layer, maximum pooling is adopted in the first two blocks, and global average pooling is adopted in the last block.

6. The partial discharge fault diagnostic method for high voltage power equipment according to claim 1, characterized in that, The ultra-high frequency signal feature vectors and ultrasonic signal feature vectors are fused based on the cross-attention mechanism, and a fusion feature vector is output, specifically as follows: wherein, attention for the ultrasound channel to the attention of the very high frequency channel, is a query matrix for the ultrasound, is a key matrix for the very high frequency, is a value matrix for the very high frequency; is a dimension of the key vector. To construct the attention of ultrasound-to-very high frequency, Q comes from very high frequency, K and V come from ultrasound; that is pay attention to the ultrasound channel for the very high frequency channel, query matrix for very high frequency, key matrix for ultrasound, value matrix for ultrasound.

7. A partial discharge fault diagnostic system for high voltage power equipment, characterized by, The method comprises the following steps: A data acquisition module is configured to collect ultra-high frequency signals and ultrasonic signals of high-voltage power equipment; A data processing module is configured to complete fault diagnosis of partial discharge of high-voltage power equipment based on the ultra-high frequency signals, ultrasonic signals and a pre-constructed signal fusion model; in the signal fusion model, first, a double-branch 1D-CNN network is used to extract features of the ultra-high frequency signals and ultrasonic signals respectively, and output ultra-high frequency signal feature vectors and ultrasonic signal feature vectors; second, the ultra-high frequency signal feature vectors and ultrasonic signal feature vectors are fused based on a cross-attention mechanism, and a fusion feature vector is output; finally, the fusion feature vector is linearly transformed based on a residual classifier, and fault diagnosis of partial discharge of high-voltage power equipment is realized.

8. A computer apparatus comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program, when executed by the processor, causes the processor to perform the method of any one of claims 1 to 7. The processor executes the computer program to realize the steps of the method according to any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program / instructions are executed by the processor to realize the steps of the method according to any one of claims 1-6.

10. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instructions are executed by the processor to realize the steps of the method according to any one of claims 1-6.

Images