A high-voltage circuit breaker mechanical fault diagnosis method and system

Abstract

The application discloses a high-voltage circuit breaker mechanical fault diagnosis method and system, the method first acquires the vibration signal of the high-voltage circuit breaker; then the vibration signal of the high-voltage circuit breaker is input into a pre-constructed adaptive mapping graph convolutional neural network and is processed to acquire the fault type of the high-voltage circuit breaker. The model proposed in the application can fully utilize the numerical characteristics and node structure characteristics of the high-voltage circuit breaker mechanical fault signal, adopts the GCN network to identify the fault, and thus realizes high-precision robust diagnosis of the high-voltage circuit breaker mechanical fault.

Description

Technical Field

[0001] This invention belongs to the field of high-voltage circuit breaker mechanical fault diagnosis technology, and relates to a method and system for diagnosing high-voltage circuit breaker mechanical faults. Background Technology

[0002] High-voltage circuit breakers (HVCBs), as crucial protection and control devices in power systems, have a significant impact on the safety and stability of transmission and distribution systems. Abnormal HVCB operation can not only cause major power grid accidents but also seriously threaten the safety of operators. Therefore, strengthening HVCB fault diagnosis is of great importance for ensuring the safe and stable operation of the entire power grid and improving its reliability. During operation, HVCBs generate signals such as sound waves, vibrations, and currents, which contain a wealth of equipment status information. By analyzing these signals, the HVCB's condition can be assessed, and potential or early-stage faults can be detected. Among these, vibration signals have strong anti-interference capabilities and high sensitivity, making them widely applicable. Therefore, a series of circuit breaker mechanical fault diagnosis methods based on vibration signal analysis have been extensively researched and applied. A commonly used method is the diagnostic method based on convolutional neural networks (CNNs).

[0003] CNN-based diagnostic methods only utilize numerical information from the vibration signals of high-voltage circuit breakers (HVCBs), neglecting the structural and topological information between fault features. This leads to insufficient information utilization, resulting in reliance on a large number of training samples and a significant performance degradation under imbalanced sample conditions. Obtaining HVCB fault samples in the field is difficult, with a limited number of samples that are often imbalanced compared to balanced samples, leading to poor performance in the field. To address these shortcomings, Graph Convolutional Networks (GCNs) can be used for fault identification. First, the HVCB fault signal is acquired, then converted into a graph signal, and finally, the GCN algorithm is used for data analysis and state identification. HVCNs can extract not only the data information of the samples but also their structural information, making more comprehensive use of the sample data. This reduces reliance on a large number of samples and maintains good performance even under imbalanced sample conditions.

[0004] The flowchart for fault identification using GCN shows that graph construction is a crucial step in the entire process. Existing graph construction methods are all based on the same graph construction idea: connecting each node to its K most similar nodes. This method has only one hyperparameter K, which makes the choice of k value directly determine the performance of the algorithm. Currently, the selection of k value mainly relies on empirical judgment, which has certain limitations. In addition, this single k value is not suitable for all nodes, especially in imbalanced datasets, where it often performs poorly. This is because when the samples are imbalanced, this algorithm will favor the class with the larger number of samples, and the accuracy of the classification of the class with fewer samples is very low. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a method and system for diagnosing mechanical faults in high-voltage circuit breakers, thereby solving the technical problems in the prior art where the selection of the k value mainly relies on empirical judgment when using GCN for fault identification, which has limitations. In addition, in imbalanced datasets, a single k value is not suitable for all nodes, resulting in a low accuracy rate of judgment.

[0006] This invention is achieved through the following technical solution:

[0007] A method for diagnosing mechanical faults in high-voltage circuit breakers includes the following steps:

[0008] S1: Acquire the vibration signal of the high-voltage circuit breaker;

[0009] S2: Input the vibration signal of the high-voltage circuit breaker into a pre-constructed adaptive graph convolutional neural network and process it to obtain the fault type of the high-voltage circuit breaker.

[0010] Preferably, the propagation expression of the adaptive graph convolutional neural network is:

[0011]

[0012]

[0013]

[0014]

[0015] In the formula, A is the adjacency matrix, h k w represents the feature matrix of a node in the k-th layer of the graph neural network. k This is the trainable weight matrix within the network. This is the normalized matrix of the adjacency matrix. To add node self-joins to the adjacency matrix, for The degree matrix, I N Let be the identity matrix, and ReLU be the activation function.

[0016] Preferably, the construction process of the adaptive graph convolutional neural network is as follows:

[0017] S21: A graph structure constructed based on an adaptive method, with 70% of the graph structure used as the training set and the remainder as the test set;

[0018] S22: Input the graph structure of the training set into the GCN network to train the GCN network, and test the GCN network using the graph structure of the test set. When the loss function of the GCN network converges, stop training and complete the construction of the adaptive graph convolutional neural network.

[0019] Preferably, the expression for the construction process of the graph structure is:

[0020]

[0021]

[0022]

[0023]

[0024]

[0025]

[0026] Among them W ij d represents the weight of the edge between node i and node j. ij The node represents the Euclidean distance between node i and node j. i Let p represent the set of the 10 neighboring nodes with the highest weights among the neighbors of node i, where sort represents sorting from largest to smallest. i (t) represents the accuracy p corresponding to node i having t connecting edges, and is a set containing 10 values, Y. i Indicates the type to which node i belongs. This represents the type of node with the m-th largest weight among the neighbors of node i. p i The t,ki corresponding to the maximum value in (t) is the optimal number of connected edges for node i, θ is a calculation coefficient positively correlated with the node weight, and γ is the weight coefficient. This represents the average value of the node weights.

[0027] Preferably, the graph structure is constructed based on an adaptive method, specifically as follows:

[0028] S211: Sets the range of values ​​for the number of edges connected to any node;

[0029] S212: Determine the weight between any two nodes;

[0030] S213: The process of constructing the graph structure using the adaptive method is completed by taking into account the range of the number of edges connected to any node and the weight between any two nodes.

[0031] Preferably, step S211 specifically involves: the number of edges connected to any node is in the range of 1-z, where z is 2log2(N) and N is the number of nodes;

[0032] Preferably, the construction of the adaptive graph convolutional neural network further includes forward propagation; the forward propagation expression is:

[0033]

[0034] in, N represents the feature of node v in the k-th layer of the neural network. v Let w represent all neighboring nodes of node v, where N is the number of neighbors of the central node v, and w k-1 and B k-1 These are the learnable weight parameters in a neural network. This is the activation function.

[0035] A high-voltage circuit breaker mechanical fault diagnosis system, comprising:

[0036] The data acquisition unit is used to acquire the vibration signal of the high-voltage circuit breaker;

[0037] Data processing unit: The data processing unit is used to input the vibration signal of the high-voltage circuit breaker into a pre-constructed adaptive graph convolutional neural network and process it to obtain the fault type of the high-voltage circuit breaker.

[0038] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described above.

[0039] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0040] Compared with the prior art, the present invention has the following beneficial technical effects:

[0041] This invention discloses a method for diagnosing mechanical faults in high-voltage circuit breakers. First, the vibration signal of the high-voltage circuit breaker is acquired. Then, the vibration signal is input into a pre-constructed adaptive graph convolutional neural network (GCN) for processing. Finally, the fault type of the high-voltage circuit breaker is obtained. This method can adaptively select the k-value (i.e., the number of edges connected to a node) based on the neighbor information of each node, fully utilizing the numerical and label information of the data to uncover deeper hidden information. The model proposed in this invention can fully utilize the numerical features and node structure features of the high-voltage circuit breaker mechanical fault signal, thereby achieving high-precision and robust diagnosis of high-voltage circuit breaker mechanical faults. First, the high-voltage circuit breaker mechanical vibration signal is converted into a graph signal using a dynamic adaptive k-nearest neighbor (kNN) graph construction method. The dynamic adaptive kNN method can assign different k-values ​​to each node based on its condition, fully utilizing the numerical and label information of the samples to construct a more information-rich structural graph, avoiding information loss during the graph construction process, and improving the overall performance of the model. To fully exploit the features of the graph structure and achieve adaptive high-precision mechanical fault diagnosis of high-voltage circuit breakers, this invention employs a GCN network for fault identification, achieving high-precision and robust diagnosis of high-voltage circuit breaker mechanical faults. Attached Figure Description

[0042] 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.

[0043] Figure 1 This is a flowchart illustrating a mechanical fault diagnosis method for high-voltage circuit breakers according to the present invention.

[0044] Figure 2 This is a schematic diagram of the structure of a high-voltage circuit breaker mechanical fault diagnosis system according to the present invention. Detailed Implementation

[0045] 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.

[0046] 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.

[0047] 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.

[0048] 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.

[0049] 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.

[0050] 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.

[0051] The present invention will now be described in further detail with reference to the accompanying drawings:

[0052] like Figure 1 As shown, this invention discloses a method for diagnosing mechanical faults in high-voltage circuit breakers, comprising the following steps:

[0053] S1: Acquire the vibration signal of the high-voltage circuit breaker;

[0054] S2: Input the vibration signal of the high-voltage circuit breaker into a pre-constructed adaptive graph convolutional neural network and process it to obtain the fault type of the high-voltage circuit breaker.

[0055] The construction process of the adaptive graph convolutional neural network is as follows:

[0056] S21: A graph structure constructed based on an adaptive method, with 70% of the graph structure used as the training set and the remainder as the test set;

[0057] S22: Input the graph structure of the training set into the GCN network to train the GCN network, and test the GCN network using the graph structure of the test set. When the loss function of the GCN network converges, stop training and complete the construction of the adaptive graph convolutional neural network.

[0058] Specifically, the graph structure is constructed using an adaptive method, as follows:

[0059] S211: Sets the range of values ​​for the number of edges connected to any node;

[0060] S212: Determine the weight between any two nodes;

[0061] S213: The process of constructing the graph structure using the adaptive method is completed by taking into account the range of the number of edges connected to any node and the weight between any two nodes.

[0062] Specifically, in the adaptive graph construction method of this application, the value of k, i.e., the number of edges connected to the node, is adaptively selected based on the neighbor information of each node. This fully utilizes the numerical and label information of the data to uncover deeper hidden information. The algorithm introduces two new variables: the weight W of the neighbors and the accuracy p. The graph construction process is as follows:

[0063] First, define the range of values ​​for k, i.e., set the number of edges connected to each node to be 1-z, where the value of z depends on the number of nodes. Preferably, z is 2log2(N), where N is the number of nodes.

[0064] Then, the weights of the node's neighbors and the accuracy p corresponding to different k values ​​for each node are calculated, thereby determining the number of connecting edges for each node. The graph construction process is expressed as follows:

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071] Among them W ij d represents the weight of the edge between node i and node j. ij The node represents the Euclidean distance between node i and node j. i Let p represent the set of the 10 neighboring nodes with the highest weights among the neighbors of node i, where sort represents sorting from largest to smallest. i (t) represents the accuracy p corresponding to node i having t connecting edges, and is a set containing 10 values, Y. i Indicates the type to which node i belongs. This represents the type of node with the m-th largest weight among the neighbors of node i. p i The t,ki corresponding to the maximum value in (t) is the optimal number of connected edges for node i, θ is a calculation coefficient positively correlated with the node weight, and γ is the weight coefficient. This represents the average value of the node weights.

[0072] Furthermore, specifically in the graph construction process, firstly, the range of the number of edges connected to each node is set to 1-z, where z is 2log2(N), and N is the number of nodes; then... To calculate the weight between any two nodes; for each node, extract the z nodes with the largest weights to form a set; assuming a node has t connected edges, extract the t nodes with the largest weights. If the labels of these two nodes are the same, add their weights; otherwise, ignore the weights. Use the final sum of weights divided by the power of the coefficient of t as the value of p. Finally, the value of t corresponding to the largest p value is the final number of edges for that node.

[0073] The constructed graph structure is then input into the GCN network. GCN is a feature extraction method for graph structures, which has many advantages over CNN: traditional CNN networks rely on a large amount of labeled data, and traditional CNNs can only extract the data information of samples and cannot effectively utilize the structural information of samples, resulting in insufficient utilization of sample information and unsatisfactory diagnostic accuracy. Graph convolution can make full use of the structural and numerical information of data, reduce dependence on samples, reduce the impact of imbalanced datasets on the model, and improve the accuracy of the model.

[0074] Graph construction and forward propagation are crucial components of GCN. The forward propagation process in GCN is similar to a convolution operation; that is, for a central node, information from its neighbors and itself is considered as the feature information of that central node. This can be expressed as:

[0075]

[0076] in N represents the feature of node v in the k-th layer of the neural network. v Let w represent all neighboring nodes of node v, where N is the number of neighbors of the central node v, and w k-1 and B k-1 These are the learnable weight parameters in a neural network. This is the activation function.

[0077] The mathematical principle of graph convolutional neural networks mainly involves exploring the properties of graphs using the eigenvalues ​​and eigenvectors of the Laplacian matrix. After a series of optimizations, including Chebyshev polynomial approximation, the propagation expression of the graph convolutional network can be expressed as:

[0078]

[0079]

[0080]

[0081]

[0082] In the formula, A is the adjacency matrix, h k w represents the feature matrix of a node in the k-th layer of the graph neural network. k This is the trainable weight matrix within the network. This is the normalized matrix of the adjacency matrix. To add node self-joins to the adjacency matrix, for The degree matrix, I N Let be the identity matrix and ReLU be the activation function. This enables the forward propagation of node features in a graph convolutional neural network.

[0083] like Figure 2 As shown, the present invention also discloses a high-voltage circuit breaker mechanical fault diagnosis system, comprising:

[0084] The data acquisition unit is used to acquire the vibration signal of the high-voltage circuit breaker;

[0085] Data processing unit: The data processing unit is used to input the vibration signal of the high-voltage circuit breaker into a pre-constructed adaptive graph convolutional neural network and process it to obtain the fault type of the high-voltage circuit breaker.

[0086] 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.

[0087] 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.

[0088] 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.

[0089] 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.

[0090] 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.

[0091] 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.

[0092] 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 method for diagnosing mechanical faults in high-voltage circuit breakers, characterized in that, Includes the following steps: S1: Acquire the vibration signal of the high-voltage circuit breaker; S2: Input the vibration signal of the high-voltage circuit breaker into a pre-constructed adaptive graph convolutional neural network and process it to obtain the fault type of the high-voltage circuit breaker; The propagation expression for the adaptive graph convolutional neural network is: In the formula, A is the adjacency matrix. This represents the feature matrix of a node in the k-th layer of the graph neural network. This is the trainable weight matrix within the network. This is the normalized matrix of the adjacency matrix. To add node self-joins to the adjacency matrix, for The degree matrix, It is the identity matrix. For activation functions; The construction process of the adaptive graph convolutional neural network is as follows: S21: A graph structure constructed based on an adaptive method, with 70% of the graph structure used as the training set and the remainder as the test set; S22: Input the graph structure of the training set into the GCN network, train the GCN network, and test the GCN network using the graph structure of the test set. When the loss function of the GCN network converges, stop training and complete the construction of the adaptive graph convolutional neural network. The graph structure is constructed based on an adaptive method, specifically as follows: S211: Sets the range of values ​​for the number of edges connected to any node; S212: Determine the weight between any two nodes; S213: The process of constructing the graph structure based on the adaptive method is completed by using the range of the number of edges connected to any node and the weight between any two nodes.

2. The method for diagnosing mechanical faults in a high-voltage circuit breaker according to claim 1, characterized in that, The expression for the construction process of the graph structure is as follows: ) in This represents the weight of the edge between node i and node j. This represents the Euclidean distance between node i and node j. Let i represent the set of the 10 neighboring nodes with the highest weights among the neighbors of node i, where sort represents sorting from largest to smallest. Let p represent the accuracy of node i when it has t connecting edges. It is a set of 10 values. Indicates the type to which node i belongs. This represents the type of node with the m-th largest weight among the neighbors of node i. express The t corresponding to the maximum value in the middle, That is, the optimal number of connecting edges for node i. It is a calculated coefficient that is positively correlated with the node weight. These are the weighting coefficients. This represents the average value of the node weights.

3. The method for diagnosing mechanical faults in a high-voltage circuit breaker according to claim 1, characterized in that, The specific steps of step S211 are as follows: the number of edges connected to any node is in the range of 1-z, where z is 2log2(N) and N is the number of nodes.

4. The method for diagnosing mechanical faults in a high-voltage circuit breaker according to claim 1, characterized in that, The construction of the adaptive graph convolutional neural network also includes forward propagation; the forward propagation expression is: in, This represents the feature of node v in the k-th layer of the neural network. This represents all neighboring nodes of node v. Let v be the number of neighbors of the central node v. and These are the learnable weight parameters in a neural network. This is the activation function.

5. A mechanical fault diagnosis system for high-voltage circuit breakers, characterized in that, A method for diagnosing mechanical faults in a high-voltage circuit breaker as described in any one of claims 1 to 4 includes: The data acquisition unit is used to acquire the vibration signal of the high-voltage circuit breaker; Data processing unit: The data processing unit is used to input the vibration signal of the high-voltage circuit breaker into a pre-constructed adaptive graph convolutional neural network and process it to obtain the fault type of the high-voltage circuit breaker.

6. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-4.

Images