Fault Diagnosis Method, System, Device and Storage Medium Based on Fast Graph Calculation

Abstract

The present invention discloses a fault diagnosis method, system, device and storage medium based on fast graph calculation, including: clustering nodes in the knowledge graph of the power operation and maintenance field to generate a node group sequence of the graph; determining sparse representation matrices of each pooling layer in the multi-scale graph neural network according to the node group sequence of the graph, and constructing a multi-scale graph neural network; inputting the node group sequence of the graph into the multi-scale graph neural network to complete the calculation of the fast graph neural network, obtaining the output result of the multi-scale graph neural network, and completing the diagnosis of the fault type of the power equipment. The method, system, device and storage medium can accurately and efficiently diagnose the fault type of the power equipment in the knowledge graph of the power operation and maintenance field.

Description

Technical Field

[0001] The present invention belongs to the technical field of deep learning, and relates to a fault diagnosis method, system, device and storage medium based on fast graph calculation. Background Art

[0002] When diagnosing the fault types of power equipment in the knowledge graph of the power operation and maintenance field, the existing technology generally uses a graph neural network for diagnosis. When constructing a graph neural network model for graph classification and regression problems, graph pooling is a crucial step. Because for graph structure inputs whose size and topology are constantly changing, what is more needed is a unified representation at the layer level rather than at the node level. The most direct pooling method provided by the graph convolutional layer is to use the global average value and sum of node features as a simple layer-level representation. This pooling operation treats all nodes equally and uses the global geometric information of the graph. However, the above global pooling method does not use the hierarchical structure of the graph and omits the effective geometric information that may be carried in the graph structure data.

[0003] It is worth noting that if a differentiable, data-dependent pooling layer containing learnable operations or parameters can be constructed, it can bring a substantial improvement to the graph classification problem. In this regard, the spectral-based pooling method proposes another design pattern, which can perform graph pooling operations in the frequency domain such as the Fourier domain or the wavelet domain. In essence, the spectral-based pooling method can combine the graph structure and node information, but its potential defect lies in the bottleneck in computational efficiency and accuracy, resulting in the inability to diagnose the types of power equipment faults efficiently and accurately. Summary of the Invention

[0004] The purpose of the present invention is to overcome the above-mentioned shortcomings of the existing technology, and provides a fault diagnosis method, system, device and storage medium based on fast graph calculation, which can accurately and efficiently diagnose the fault types of power equipment in the knowledge graph of the power operation and maintenance field.

[0005] The calculation efficiency and calculation accuracy are relatively high.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] On the one hand, the present invention provides a fault diagnosis method based on fast graph calculation, including:

[0008] Clustering the nodes in the knowledge graph of the power operation and maintenance field to generate a sequence of node groups of the graph;

[0009] Determine the sparse representation matrices of each pooling layer in the multi-scale graph neural network according to the sequence of node groups of the graph, and construct a multi-scale graph neural network;

[0010] Input the node group sequence of the said graph into a multi-scale graph neural network to complete the calculation of the fast graph neural network, and obtain the fault types of the power equipment in the knowledge graph of the power operation and maintenance field.

[0011] A further improvement of the fault diagnosis method based on fast graph calculation according to the present invention lies in:

[0012] The specific process of clustering the nodes in the knowledge graph of the power operation and maintenance field to generate a node group sequence of the graph is as follows:

[0013] Cluster the nodes in the knowledge graph G of the power operation and maintenance field through a clustering algorithm to generate a node group sequence (G0, G1,..., G K ), where G0 = G, and the nodes of graph G j+1 correspond to the node clusters in graph G j , and j = 0,..., K - 1.

[0014] The specific process of determining the sparse representation matrix of each pooling layer in the multi-scale graph neural network according to the node group sequence of the said graph and constructing the multi-scale graph neural network is as follows:

[0015] Perform vector representation on each node group in the node group sequence of the said graph, and form the representation matrix of the node group through the vector representations of each node group;

[0016] Respectively use the representation matrices of each node group as the sparse representation matrices of each pooling layer in the multi-scale graph neural network to construct the multi-scale graph neural network.

[0017] Before respectively using the representation matrices of each node group as the sparse representation matrices of each pooling layer in the multi-scale graph neural network, it further includes:

[0018] Discard the high-frequency fine information in the representation matrix through the eigenvalues of the representation matrices of each node group.

[0019] The multi-scale graph neural network is expressed as:

[0020]

[0021] where, Φ j is the sparse representation matrix of size N j+1 ×N j of the j-th pooling layer, is the input feature matrix of size N j ×d, is the output feature matrix of size N j+1 ×d, and K is the number of pooling layers in the multi-scale graph neural network.

[0022] In the process of inputting the node group sequence of the graph into the multi-scale graph neural network to obtain the output result of the multi-scale graph neural network,

[0023] Pool the node group sequence of the graph through each pooling layer, aggregate the output results of each pooling layer, and transmit the aggregation result to the readout operation and the multi-layer perceptron.

[0024] In a second aspect of the present invention, there is provided a fault diagnosis system based on fast graph calculation, including:

[0025] A clustering module for clustering the nodes in the knowledge graph of the power operation and maintenance field to generate a node group sequence of the graph;

[0026] A determination module for determining the sparse representation matrix of each pooling layer in the multi-scale graph neural network according to the node group sequence of the graph, and constructing a multi-scale graph neural network;

[0027] A calculation module for inputting the node group sequence of the graph into the multi-scale graph neural network, completing the calculation of the fast graph neural network, and obtaining the fault type of the power equipment in the knowledge graph of the power operation and maintenance field.

[0028] A further improvement of the fault diagnosis system based on fast graph calculation according to the present invention is that:

[0029] The determination module includes:

[0030] A data processing module for vector-representing each node group in the node group sequence of the graph, and forming a representation matrix of the node group through the vector representations of each node group;

[0031] A construction module for respectively using the representation matrix of each node group as the sparse representation matrix of each pooling layer in the multi-scale graph neural network to construct a multi-scale graph neural network.

[0032] In a third aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the fault diagnosis method based on fast graph calculation are implemented.

[0033] In a fourth aspect of the present invention, there is provided a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the fault diagnosis method based on fast graph calculation are implemented.

[0034] The present invention has the following beneficial effects:

[0035] When the fault diagnosis method, system, device and storage medium based on fast graph calculation according to the present invention are specifically operated, the sparse representation matrices of each pooling layer in the multi-scale graph neural network are determined according to the node group sequence of the graph. Based on the sparse representation method, while filtering fine information, the structural information of the knowledge graph in the field of power operation and maintenance is retained to solve the problem of too low calculation efficiency of the traditional graph neural network. At the same time, the multi-scale graph neural network is constructed based on the sparse representation matrices of each pooling layer, hierarchically combining the topological information of different sizes of sets in the graph to solve the problem of low calculation accuracy of the traditional graph neural network, and achieving the purpose of accurately and efficiently diagnosing the fault types of power equipment in the knowledge graph of the power operation and maintenance field. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings forming a part of this invention are used to provide a further understanding of the invention. The illustrative embodiments and descriptions thereof of the invention are used to explain the invention and do not constitute an improper limitation of the invention. In the drawings:

[0037] Figure 1 is the flowchart of the method of the present invention;

[0038] Figure 2 is the processing flowchart of the pooling layer;

[0039] Figure 3 is the structural diagram of the multi-scale graph neural network;

[0040] Figure 4 is the system structural diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. 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.

[0042] It should be noted that the terms "first", "second", etc. in the specification, claims and above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0043] The present invention will be further described in detail below with reference to the accompanying drawings:

[0044] Embodiment 1

[0045] Reference Figure 1 , the fault diagnosis method based on fast graph calculation described in the present invention includes:

[0046] 1) Obtain the structural data of the knowledge graph in the field of power operation and maintenance, cluster the nodes in the knowledge graph in the field of power operation and maintenance, and generate a sequence of node groups of the graph;

[0047] Specifically, for the nodes in the knowledge graph G in the field of power operation and maintenance, a sequence of node groups of the graph (G0, G1,..., G K ) is generated through a clustering algorithm, where G0 = G, and for j = 0,..., K - 1, there are nodes in graph G j+1 corresponding to the node clusters in graph G j . Let N j = |V(G j )| be the number of nodes in graph G j . Then, for a graph G j containing N j+1 nodes and a graph G j and graph G j+1 , there is N j > N j+1 . The clustering algorithm is used to analyze the characteristics of the graph structure data, and a sequence of node groups of the graph (G0, G1,..., G K ) is generated.

[0048] It should be noted that, for example, in the field of power operation and maintenance inspection, the clustering algorithm can adopt candidate algorithms such as spectral clustering (Shi & Malik, 2000), k-means clustering (Pakhira, 2014), DBSCAN (Ester et al., 1996), OPTICS (Ankerst et al., 1999), and METIS (Karypis & Kumar, 1998), or can also adopt the clustering structure information inherent in the graph data of the knowledge graph in the field of power operation and maintenance inspection.

[0049] The knowledge graph in the field of power operation and maintenance inspection contains data on common faults and defects of power equipment. According to the current standards and definitions, there are subordinate relationships among the entities in the knowledge graph, and a hierarchical structure can be naturally formed without a clustering algorithm. For example, the insulator category includes porcelain insulators, glass insulators, and composite insulators, etc., and the porcelain insulator further includes porcelain insulator contamination, porcelain insulator self-explosion, and porcelain insulator zero value, etc., and can be further divided downward according to the different degrees of defects.

[0050] 2) Perform vector representation on each node group in the node group sequence of the graph, and form the representation matrix of the node group through the vector representations of each node group;

[0051] Specifically, for the node group sequence (G0, G1,..., G K ) generated after clustering, since the N j+1 nodes in graph G j+1 are selected and clustered from the N j nodes in graph G j , each node in graph G j can be represented as a vector of size N j+1 , and the N j vectors can form a sparse representation of size N j+1 × N j , that is, the sparse representation matrix Φ j of the j-th pooling layer.

[0052] It should be noted that in the present invention, the eigenvalues are calculated through spectral graph transformation, and the eigenvalues are used as the representation matrix Φ of the graph structure data. The representation matrix Φ has orthogonality and parameter learnability, and represents information of different frequencies through the eigenvalues. By discarding the high-frequency fine information in the eigenvalues of the representation matrix Φ, the sparse representation matrix Φ j ;

[0053] It should be noted that the spectral graph transformation includes Fourier transform, dyadic wavelet transform, and Daubechies wavelet transform.

[0054] The Fourier transform of the spectrogram is similar to the traditional Fourier transform, but the corresponding Fourier basis and Laplacian matrix need to be found on the graph structure data. Among them, the Laplacian matrix \(L = D - A\) is calculated from the degree matrix \(D\) and the adjacency matrix \(A\) of the graph; reference Figure 2 , for any signal \(x\), the spectral graph Fourier transform is:

[0055]

[0056] where \(\varPhi=[\varphi_1,...,\varphi N \) is the eigenvector of \(L\), and \(L = \varPhi\varLambda\varPhi T , is the Fourier coefficient of the signal \(x\) on the \(k\)-th Fourier basis. Its essence is the projection of the signal on the Fourier basis. Through the frequency definition of the graph signal, the Fourier coefficient can reflect the intensity of the graph signal on different frequency components, providing conditions for subsequent filtering.

[0057] 3) Use the representation matrices of each node group as the sparse representation matrices of each pooling layer in the multi-scale graph neural network to construct the multi-scale graph neural network;

[0058] Specifically, refer to Figure 3 , take the node group sequence of the graph \((G_0, G_1,..., G K ) as the input and input it into the multi-scale graph neural network. Use the pooling layer in the multi-scale graph neural network for pooling to obtain the corresponding coarsened graph after pooling A node of the coarsened graph is a clustering of a node group before coarsening. Therefore, the number of nodes in the coarsened graph is relatively reduced, and it can only reflect approximate feature information. The result of the final pooling layer consists of one node, and this node contains all the node information of the previous layer, that is

[0059] It should be noted that the present invention aggregates the output results of each pooling layer and then transmits them to the readout operation and the multi-layer perceptron, thereby retaining the feature information of the graph structure data at each level, making the final graph representation result contain the sum of the node information of different layers, and solving the problem of insufficient accuracy of the traditional graph neural network.

[0060] 4) Input the node group sequence of the graph into the multi-scale graph neural network to complete the calculation of the fast graph neural network, and obtain the fault types of the power equipment in the knowledge graph of the power operation and maintenance field.

[0061] Example Two

[0062] Refer to Figure 4 , the fault diagnosis system based on fast graph calculation described in the present invention includes:

[0063] The clustering module 1 is used to cluster the nodes in the knowledge graph of the power operation and maintenance field to generate a sequence of node groups of the graph;

[0064] The determination module 2 is used to determine the sparse representation matrices of each pooling layer in the multi-scale graph neural network according to the sequence of node groups of the graph, and construct the multi-scale graph neural network;

[0065] The calculation module 3 is used to input the sequence of node groups of the graph into the multi-scale graph neural network to obtain the fault types of power equipment in the knowledge graph of the power operation and maintenance field.

[0066] Preferably, the determination module 2 includes:

[0067] The data processing module 21 is used to perform vector representation on each node group in the sequence of node groups of the graph, and form the representation matrix of the node group through the vector representations of each node group;

[0068] The construction module 22 is used to respectively use the representation matrices of each node group as the sparse representation matrices of each pooling layer in the multi-scale graph neural network to construct the multi-scale graph neural network.

[0069] Embodiment III

[0070] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the fault diagnosis method based on fast graph calculation are implemented. Among them, the memory may include a memory, such as a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk memory, etc.; the processor, network interface, and memory are interconnected through an internal bus, and the internal bus may be an Industry Standard Architecture bus, a Peripheral Component Interconnect standard bus, an Extended Industry Standard Architecture bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory may include a memory and a non-volatile memory, and provide instructions and data to the processor.

[0071] Embodiment IV

[0072] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the fault diagnosis method based on fast graph calculation are implemented. Specifically, the computer-readable storage medium includes, but is not limited to, for example, a volatile memory and / or a non-volatile memory. The volatile memory may include a random access memory (RAM) and / or a cache memory, etc. The non-volatile memory may include a read-only memory (ROM), a hard disk, a flash memory, an optical disc, a magnetic disk, etc.

[0073] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an all-hardware embodiment, an all-software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0074] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0075] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that realizes the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0076] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, such that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable devices provide steps for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: still, the specific implementation manners of the present invention can be modified or equivalently replaced, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A fault diagnosis method based on fast graph calculation, characterized in that, Including: Clustering the nodes in the knowledge graph of the power operation and maintenance field to generate a sequence of node groups of the graph; Determining the sparse representation matrices of each pooling layer in the multi-scale graph neural network according to the sequence of node groups of the graph, and constructing the multi-scale graph neural network; Inputting the sequence of node groups of the graph into the multi-scale graph neural network to obtain the fault types of power equipment in the knowledge graph of the power operation and maintenance field; The specific process of clustering the nodes in the knowledge graph of the power operation and maintenance field to generate a sequence of node groups of the graph is as follows: Cluster the nodes in the knowledge graph G in the power operation and maintenance field through a clustering algorithm to generate a sequence of node groups of the graph (G0, G1,..., G K ), where G0 = G, and the nodes of the graph G j+1 correspond to the node clusters in the graph G j , j = 0,..., K - 1; The specific process of determining the sparse representation matrices of each pooling layer in the multi-scale graph neural network according to the sequence of node groups of the graph and constructing the multi-scale graph neural network is as follows: Performing vector representation on each node group in the sequence of node groups of the graph, and forming the representation matrix of the node group through the vector representations of each node group; Taking the representation matrices of each node group as the sparse representation matrices of each pooling layer in the multi-scale graph neural network respectively, and constructing the multi-scale graph neural network.

2. The fault diagnosis method based on fast graph calculation according to claim 1, wherein Before taking the representation matrices of each node group as the sparse representation matrices of each pooling layer in the multi-scale graph neural network respectively, it further includes: Discarding the high-frequency fine information in the representation matrix through the eigenvalues of the representation matrix of each node group.

3. The fault diagnosis method based on fast graph calculation according to claim 1, wherein, The multi-scale graph neural network is expressed as: Among them, Φ j is the sparse representation matrix of size N j+1 ×N j for the j-th pooling layer, is the input feature matrix of size N j ×d, is the output feature matrix of size N j+1 ×d, and K is the number of pooling layers in the multi-scale graph neural network.

4. The fault diagnosis method based on fast graph calculation according to claim 1, characterized in that In the process of inputting the sequence of node groups of the graph into the multi-scale graph neural network to obtain the output result of the multi-scale graph neural network, pooling the sequence of node groups of the graph through each pooling layer, aggregating the output results of each pooling layer, and transmitting the aggregation result to the readout operation and the multi-layer perceptron.

5. A fault diagnosis system based on fast graph calculation, characterized in that, Including: A clustering module (1) for clustering the nodes in the knowledge graph of the power operation and maintenance field to generate a sequence of node groups of the graph; A determination module (2) for determining the sparse representation matrices of each pooling layer in the multi-scale graph neural network according to the sequence of node groups of the graph and constructing the multi-scale graph neural network; A calculation module (3) for inputting the sequence of node groups of the graph into the multi-scale graph neural network to obtain the fault types of power equipment in the knowledge graph of the power operation and maintenance field; The determination module (2) includes: A data processing module (21) for performing vector representation on each node group in the sequence of node groups of the graph and forming the representation matrix of the node group through the vector representations of each node group; A construction module (22) for taking the representation matrices of each node group as the sparse representation matrices of each pooling layer in the multi-scale graph neural network respectively and constructing the multi-scale graph neural network; The specific process of clustering the nodes in the knowledge graph of the power operation and maintenance field to generate a sequence of node groups of the graph is as follows: Cluster the nodes in the knowledge graph graph G in the power operation and maintenance field through a clustering algorithm to generate a sequence of node groups of the graph (G0, G1,..., G K ), where G0 = G, and the nodes of graph G j+1 correspond to the node clusters in graph G j , and j = 0,..., K - 1.

6. A computer 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 fault diagnosis method based on fast graph calculation according to any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the fault diagnosis method based on fast graph calculation according to any one of claims 1 to 4.

Images