Intelligent substation equipment fault diagnosis method, device and equipment

Abstract

The application discloses a smart substation equipment fault diagnosis method, device and equipment, relates to the technical field of smart substation fault diagnosis, and comprises the following steps: based on the equipment connection relation, the spatial alignment is carried out to the electric power sampling signal, and the structured graph signal is obtained; the feature transformation is carried out to the structured graph signal, the statistical correlation matrix is obtained and is mapped into the real-time state characteristic point in the Riemann manifold space; according to the curvature characteristic of the Riemann manifold space, the geometric deviation of the real-time state characteristic point and the preset ideal state point is obtained by using a logarithmic mapping operator; the fault contribution degree of each equipment is calculated by using the geometric deviation, and the fault equipment is determined, and the equipment diagnosis result is output. The application is used to solve the problems that the prior art has insufficient weak fault sensing sensitivity under complex working conditions and it is difficult to accurately trace the source of the fault.

Description

Technical Field

[0001] This invention relates to the field of intelligent substation fault diagnosis technology, and more specifically, to intelligent substation equipment fault diagnosis methods, devices, and equipment. Background Technology

[0002] With the deepening construction of smart substations, online monitoring and fault diagnosis of primary equipment have become core aspects of ensuring power grid security.

[0003] However, existing diagnostic technologies still have the following shortcomings in practical applications: First, substation equipment operating data has high-dimensional nonlinear characteristics. Existing technologies mostly use linear space models and simple linear distance metrics, which are difficult to accurately reflect the true state evolution trajectory of the equipment, resulting in insufficient sensitivity to detect minor initial faults. Second, affected by fluctuations in operating conditions, equipment operating parameters exhibit normalized drift characteristics. Existing methods mostly calculate fault contribution based on linear residuals, lacking effective means to separate the true deviation under dynamic operating conditions, and cannot eliminate the mapping distortion caused by calculating fault contribution based on numerical differences. This makes it difficult for the diagnostic system to quickly pinpoint the source of the fault, reducing the efficiency of on-site fault investigation in substations.

[0004] Therefore, there is an urgent need for a method, device, and equipment for diagnosing faults in intelligent substation equipment that takes into account both nonlinear feature perception and accurate source tracing. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method, apparatus, and equipment for diagnosing faults in intelligent substation equipment. By constructing a structured graph signal and mapping it to a Riemannian manifold space to calculate geometric deviation and contribution, the present invention addresses the problems of insufficient sensitivity in detecting weak faults under complex operating conditions and difficulty in accurately tracing the source of faults.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for diagnosing faults in intelligent substation equipment includes the following steps:

[0008] Spatial alignment of power sampling signals based on device connection relationships yields structured graph signals;

[0009] The structured graph signal is subjected to feature transformation to obtain a statistical correlation matrix and mapped to real-time state feature points in the Riemannian manifold space;

[0010] Based on the curvature characteristics of the Riemannian manifold space, the geometric deviation between the real-time state feature points and the preset ideal state points is obtained using a logarithmic mapping operator.

[0011] The fault contribution of each device is calculated using the geometric deviation, the faulty device is identified, and the device diagnosis results are output.

[0012] In a preferred embodiment, obtaining the structured graph signal includes: constructing a topological correlation matrix based on the physical connection relationships of the substation equipment; constructing a node feature matrix corresponding to the node indexes of the correlation matrix based on the power sampling signals of the equipment; and combining the correlation matrix and the feature matrix to obtain the structured graph signal.

[0013] In a preferred embodiment, obtaining the statistical correlation matrix and mapping it to real-time state feature points in the Riemannian manifold space includes: performing a graph Fourier transform on the structured graph signal to obtain graph frequency domain features; performing sample covariance calculation on the graph frequency domain features to obtain a statistical correlation matrix; and mapping the statistical correlation matrix to a symmetric positive definite matrix Riemannian manifold space to obtain the real-time state feature points.

[0014] In a preferred embodiment, obtaining the geometric deviation between the real-time state feature point and the preset ideal state point using a logarithmic mapping operator includes: using a Riemann logarithmic mapping operator to map the real-time state feature point to a reference tangent space with the preset ideal state point as the tangent point, generating a Riemann logarithmic tangent vector; and calculating the norm of the Riemann logarithmic tangent vector as the geometric deviation of the real-time state feature point relative to the preset ideal state point.

[0015] In a preferred embodiment, the step of calculating the fault contribution of each device using the geometric deviation includes: calculating the partial derivative of the geometric deviation with respect to the node feature matrix in reverse based on the calculation path of the geometric deviation to obtain the node sensitivity matrix; and calculating the norm of each row vector in the node sensitivity matrix as the fault contribution of the corresponding device node.

[0016] In a preferred embodiment, the output device diagnostic results include: mapping the geometric deviation to a preset state range to obtain the substation's operating state level; determining an adaptive threshold based on the statistical characteristic parameters of the fault contribution, and comparing the fault contribution with the adaptive threshold to determine the faulty device; and outputting device diagnostic results that include the operating state level and the faulty device.

[0017] A fault diagnosis device for intelligent substation equipment includes the following units: a signal graphing unit, used to spatially align power sampling signals based on equipment connection relationships to obtain a structured graph signal; a manifold mapping unit, used to perform feature transformation on the structured graph signal to obtain a statistical correlation matrix and map it to real-time state feature points in a Riemannian manifold space; a geometric measurement unit, used to obtain the geometric deviation between the real-time state feature points and a preset ideal state point using a logarithmic mapping operator based on the curvature characteristics of the Riemannian manifold space; and a location diagnosis unit, used to calculate the fault contribution of each device using the geometric deviation and determine the faulty device, and output the device diagnosis result.

[0018] An electronic device includes: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the intelligent substation equipment fault diagnosis method.

[0019] A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the intelligent substation equipment fault diagnosis method.

[0020] The technical effects and advantages of the intelligent substation equipment fault diagnosis method of this invention are as follows:

[0021] 1. This invention maps the statistical correlation matrix of structured graph signals to the Riemannian manifold space and uses a logarithmic mapping operator to extract the geometric deviation of real-time state feature points relative to preset ideal state points. This solves the mapping distortion problem of traditional linear measures when describing the high-dimensional nonlinear state evolution of equipment, improves the sensitivity of sensing weak faults in the early stage of equipment operation, and realizes refined early warning of the health status of power equipment.

[0022] 2. This invention calculates the fault contribution of each device by utilizing geometric deviation and identifies the faulty device. It directly transforms the state geometric differences in the Riemannian manifold space into fault weight indicators for physical devices, thereby achieving precise location of the fault source and providing a direct quantitative basis for the refined maintenance of substation equipment. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the fault diagnosis method for intelligent substation equipment provided in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of three-phase power sampling signals in an embodiment of the present invention.

[0025] Figure 3 This is a schematic diagram of the changed state intensity scalar sequence in an embodiment of the present invention.

[0026] Figure 4This is a schematic diagram comparing the contribution of equipment failure to the adaptive threshold in an embodiment of the present invention.

[0027] Figure 5 This is a schematic diagram of a fault diagnosis device unit for intelligent substation equipment provided in an embodiment of the present invention.

[0028] Figure 6 This is a structural block diagram of an exemplary electronic device provided for implementing embodiments of the present disclosure.

[0029] Figure 7 This is a schematic diagram of an exemplary storage medium that can be used to implement embodiments of the present disclosure, as provided in the embodiments of the present invention. Detailed Implementation

[0030] 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. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0031] Example 1, Figure 1 This invention provides a method for diagnosing faults in intelligent substation equipment, comprising the following steps:

[0032] S1. Spatially align the power sampling signal based on the device connection relationship to obtain a structured graph signal, including:

[0033] S101. Construct a topological correlation matrix based on the physical connection relationships of the substation equipment, as follows:

[0034] The configuration description file of the intelligent substation is parsed to identify the connection and layout relationships of all equipment in the station. Then, the primary equipment such as the main transformer, circuit breaker, disconnector, current transformer and voltage transformer in the substation are instantiated into a set of nodes in the graph model, and the actual physical electrical connections between the equipment (such as bus connection and transmission line connection) are mapped into a set of edges in the graph model.

[0035] An N×N adjacency matrix is ​​constructed based on the node set and edge set, where N is the total number of nodes of the monitored equipment in the substation. The degree matrix is ​​calculated based on this adjacency matrix, and the Laplace matrix, which reflects the differential characteristics of the graph structure, is further derived as the final topological inclination matrix. The specific calculation process of the topological inclination matrix is ​​as follows:

[0036] ,

[0037] ,

[0038] ,

[0039] in, Adjacency matrix The elements in the table represent the connection status between nodes; Specifically, the degree matrix is ​​a diagonal matrix. The diagonal element represents the first element. The degree of a node, i.e., the number of edges connected to it; It is a Laplace matrix of dimension N×N, i.e., a topological correlation matrix.

[0040] S102. Based on the power sampling signal of the device, construct a node feature matrix corresponding to the node index of the correlation matrix, as follows:

[0041] High-frequency synchronous sampling is performed on each equipment node through the substation process layer network or merging unit to obtain power sampling signals containing the three-phase current sequence of each monitoring point;

[0042] The state intensity scalar of each device node at the current moment is calculated based on the three-phase current sequence; and all state intensity scalars of a single device within the entire sampling time window are arranged in chronological order to obtain the first... The timing feature vector of the device; the first device's timing feature vector; The temporal feature vector of the device is directly used as the first node feature matrix. Row vectors are used to ultimately obtain a node feature matrix reflecting the spatiotemporal status of all equipment at the station; the mathematical expression of the node feature matrix is ​​as follows:

[0043] ,

[0044] ,

[0045] in, For dimension The node feature matrix, The number of sampling points. Indicates the first Each device node during the sampling time window Temporal feature vectors within; For the first The device node at the ... The state intensity scalar of each sampling point , , The first The device in the The current amplitudes of phases A, B, and C were collected at each sampling point.

[0046] To intuitively illustrate the calculation and feature extraction effects of the state strength scalar, this embodiment uses a specific branch circuit breaker in a substation as an example; for example... Figure 2 and Figure 3 As shown, Figure 2 These are the original three-phase current waveforms (A, B, and C) acquired by the circuit breaker in this branch within the sampling window. Figure 3 This is a single-state intensity scalar sequence based on the original three-phase current waveform. By comparison, it can be seen that the transformed state intensity scalar effectively achieves dimensionality reduction and feature aggregation of the data while preserving the current fluctuation characteristics, providing standardized input data for constructing the node feature matrix.

[0047] S103. Combine the correlation matrix and the feature matrix to obtain the structured graph signal, as follows:

[0048] The node feature matrix is ​​defined as a multidimensional graph domain signal attached to the physical topology of the substation, and the topological correlation matrix is ​​used as the underlying support structure of the structured graph signal space, i.e., the graph displacement operator.

[0049] By using the correspondence between the row vector indices of the node feature matrix and the node indices of the topological association matrix, the association matrix and the feature matrix are combined into a binary tuple containing both "topological-attribute" information, i.e., a structured graph signal.

[0050] This step transforms isolated single-point sampling signals into a global object with global topological features by constructing a structured graph signal with spatial coupling characteristics. This enables subsequent processing to perform frequency domain analysis on the signal using the graph topology, thereby effectively extracting the spatial correlation features contained in the device connection relationships.

[0051] S2. Perform feature transformation on the structured graph signal to obtain a statistical correlation matrix and map it to real-time state feature points in the Riemannian manifold space, including:

[0052] S201. Perform a graph Fourier transform on the structured graph signal to obtain the graph frequency domain features, as follows:

[0053] The topological incidence matrix is ​​subjected to eigenvalue decomposition to obtain eigenvalues ​​and corresponding orthogonal eigenvectors; the orthogonal eigenvectors are then sorted in ascending order according to their corresponding eigenvalues, and an eigenvector matrix is ​​constructed using the sorted orthogonal eigenvectors as column vectors; the mathematical expressions for the eigenvalue decomposition and matrix construction are as follows:

[0054] ,

[0055] ,

[0056] ,

[0057] ,

[0058] in, For eigenvalues, For graph frequency index, For the corresponding to the first Eigenvectors with eigenvalues; It is an eigenvalue diagonal matrix. The eigenvector matrix;

[0059] The transpose of the eigenvector matrix is ​​used as the graph Fourier transform operator to perform a left multiplication operation on the node feature matrix, yielding the graph frequency domain features; the calculation formula for the graph frequency domain features is as follows:

[0060] ,

[0061] in, This is the frequency domain feature matrix of the graph.

[0062] S202. Perform sample covariance calculation on the frequency domain features of the graph to obtain the statistical correlation matrix, as follows:

[0063] The frequency domain feature matrix of the graph is compressed using the sample covariance algorithm, and the statistical correlation between different graph frequency components is calculated to obtain the statistical correlation matrix. The calculation formula is as follows:

[0064] ,

[0065] in, To statistically analyze the correlation matrix, The number of sampling points. This is the transpose of the frequency domain characteristic matrix of the graph; The preset regularization coefficient is usually a very small positive number, for example... ; It is an identity matrix with the same dimension as the statistical correlation matrix, used to ensure that the statistical correlation matrix is ​​full rank and invertible.

[0066] S203. Map the statistical correlation matrix to a symmetric positive definite matrix Riemannian manifold space to obtain the real-time state feature points, as follows:

[0067] Based on the symmetric positive definite property of the statistical correlation matrix, and using the affine invariant Riemannian metric as the basis, a symmetric positive definite matrix Riemannian manifold space with non-positive curvature properties is constructed. The affine invariant Riemannian metric is specifically embodied in the form of a Riemann inner product; that is, for any matrix point in the Riemannian manifold space and any two tangent vectors in its tangent space, the Riemann inner product satisfying the affine invariant property is defined as follows:

[0068] ,

[0069] in, For Riemann inner product, , Let be any matrix point and its inverse matrix in the Riemannian manifold space. , The tangent vector, Represents the matrix trace operation;

[0070] Based on the definition of the Riemann inner product, the statistical correlation matrix can be directly regarded as a Riemannian manifold space. A specific geometric point, namely a real-time state feature point. .

[0071] It should be noted that traditional substation equipment condition analysis is usually based on linear metrics in Euclidean space to measure feature differences, which can lead to geometric distortion of the nonlinear statistical characteristics of the equipment, making it difficult to accurately capture early weak fault features. This invention, by converting structured signals to Riemannian manifold space, restores the true topological structure of the data, overcomes the bias of linear space in representing nonlinear features, and improves the identification and robustness of weak fault signals in strong noise environments.

[0072] S3. Based on the curvature characteristics of the Riemannian manifold space, the geometric deviation between the real-time state feature points and the preset ideal state points is obtained using a logarithmic mapping operator, including:

[0073] S301. Using the Riemann logarithmic mapping operator, the real-time state feature points are mapped to a reference tangent space with a preset ideal state point as the tangent point, generating a Riemann logarithmic tangent vector, as follows:

[0074] Based on the statistical correlation matrix sample set of substation equipment under historical normal operating conditions, an objective function is constructed with the goal of minimizing the sum of squared Riemann distances from all sample points to the center point. The specific objective function is as follows:

[0075] ,

[0076]

[0077] in, To preset the ideal state point, This represents the total number of samples in the statistical correlation matrix under historical normal operating conditions. Let be any matrix variable in the Riemannian manifold space, used as the independent variable for the optimization search; For the first in the sample set A statistical correlation matrix sample; For the geodesic distance function based on the affine invariant Riemannian metric, Represents the Frobenius norm operation;

[0078] The objective function is solved iteratively using the Riemann gradient descent method: In each iteration, the gradient direction of the current ideal state estimation point on the Riemann manifold (i.e., the average direction of the tangent vector pointing to each sample point) is calculated, and the estimated point position is updated by moving along the opposite direction of the gradient by a preset step size, the preset range of which is 0.1 to 0.5; the update process is repeated until the convergence condition is met, which can be set as the Riemann distance between the estimated points obtained in two adjacent iterations being less than 1 / 2. The latest ideal state estimate point after convergence is the globally optimal Riemann geometric mean, and is defined as the preset ideal state point. In the specific programming implementation, the iterative solution process relies on the underlying matrix operation support provided by the numerical computing library, including eigenvalue decomposition and singular value decomposition. At the same time, a manifold optimization algorithm framework can be used to perform gradient calculation and scaling operations based on Riemann geometry to ensure that the iterative process is numerically stable and converges efficiently in the curved manifold space.

[0079] A reference tangent space is constructed using a preset ideal state point as the tangent point. This reference tangent space is an Euclidean linear space tangent to the curved manifold surface at the tangent point. The Riemann logarithmic mapping operator is used to project and expand the real-time state feature points located on the curved manifold surface into the reference tangent space. Geometrically, this projection process is equivalent to finding the shortest path (i.e., a geodesic) connecting the ideal state point and the real-time state feature points within the manifold space, and then straightening and projecting this geodesic onto the reference tangent plane. This yields a matrix vector that preserves the original manifold topological differences, i.e., the Riemann logarithmic tangent vector. The formula for the Riemann logarithmic mapping is as follows:

[0080] ,

[0081] in, Let the Riemann logarithmic tangent vector be... It is a Riemann logarithmic mapping operator with a preset ideal state point as the reference point; For matrix logarithm operations, , These are the square root matrix and its inverse matrix of the ideal state point, respectively.

[0082] S302. Calculate the norm of the Riemann logarithmic tangent vector as the geometric deviation of the real-time state feature point relative to the preset ideal state point, as follows:

[0083] Since the Euclidean distance in the tangent space is locally isomorphic to the geodesic distance on the manifold, the geometric distance, i.e., the geometric deviation, from the ideal state point in the real-time operating state can be directly quantified by calculating the Frobenius norm of the Riemann logarithmic tangent vector. The calculation formula is as follows:

[0084] ,

[0085] in, For geometric deviation, Represents the matrix trace operation. Let be the transpose of the Riemann logarithmic tangent vector.

[0086] This step transforms the complex nonlinear manifold measurement problem into a linear norm calculation problem in the tangent space, and uses geodesic distance instead of traditional Euclidean distance, so that the calculated geometric deviation can truly reflect the topological difference between the substation equipment status and the ideal normal state, providing a high-fidelity quantitative indicator for subsequent accurate fault identification.

[0087] S4. Calculate the fault contribution of each device using the geometric deviation and determine the faulty device, then output the device diagnosis results;

[0088] In this embodiment, step S4, which calculates the fault contribution of each device using the geometric deviation, includes:

[0089] S401. Based on the calculation path of the geometric deviation, the partial derivative of the geometric deviation with respect to the node feature matrix is ​​calculated in reverse to obtain the node sensitivity matrix, as follows:

[0090] Based on the geometric deviation calculation link specifically "node feature matrix → graph frequency domain features → statistical correlation matrix → scalar geometric deviation", starting from the final geometric deviation and deriving backward along the calculation link, the gradient of the geometric deviation relative to the node feature matrix is ​​calculated, resulting in an N×M dimension gradient matrix. The gradient quantifies the contribution of data fluctuations at each sampling time of each device node to the final global deviation. The calculation process can be implemented using a computational framework that supports automatic differentiation (such as a modern deep learning framework): the system automatically constructs a forward computation graph from "node feature matrix" to "scalar geometric deviation", and uses the framework's built-in backpropagation engine to automatically derive and calculate the gradient values ​​of each node, thereby avoiding the tediousness and errors of manually deriving complex matrix differentiation formulas.

[0091] The gradient matrix is ​​used as the node sensitivity matrix, and the calculation logic of the node sensitivity matrix is ​​expressed as follows:

[0092] ,

[0093] in, Here is the node sensitivity matrix. Represents matrix differentiation operations;

[0094] S402. Calculate the norm of each row vector in the node sensitivity matrix as the fault contribution of the corresponding device node, as follows:

[0095] The first node sensitivity matrix Row vector, recording the row vector of the first row. The temporal sensitivity distribution of data fluctuations of individual devices to the overall abnormal state within the sampling time window;

[0096] By calculating the first row vectors The norm transforms the time-series sensitivity distribution into a single scalar index, and serves as the first norm. The fault contribution of each device to the global geometric deviation; the larger the fault contribution value, the more likely the data distortion of the corresponding device is the main cause of the real-time state deviating from the ideal state point of the Riemannian manifold, and the higher its priority in fault tracing; the formula for calculating the fault contribution is as follows:

[0097] ,

[0098] in, Indicates the first Fault contribution of each device node express Norm operations; The first node in the node sensitivity matrix represents the... row vectors Indicates the first The row vector The values ​​of each sampling point.

[0099] In this embodiment, the output device diagnostic results in step S4 include:

[0100] S403. Map the geometric deviation to a preset state range to obtain the substation's operating state level, as follows:

[0101] Based on a sample set of geometric deviations of substation equipment under historical normal operating conditions, the statistical mean and standard deviation of historical geometric deviations are calculated, and threshold boundaries for three state intervals—normal, warning, and fault—are set. For example, setting Let the sum of the historical mean and twice the historical standard deviation be set. It is the sum of the historical mean and three times the historical standard deviation;

[0102] Compare the current geometric deviation with the threshold boundary: If If the operating status level is determined to be "normal", the diagnostic result of "normal status" will be output directly; if The operational status level is determined to be "early warning status"; if If so, the operating status level is determined to be "fault status".

[0103] S404. Based on the statistical characteristic parameters of the fault contribution, an adaptive threshold is determined, and the faulty equipment is determined by comparing the fault contribution with the adaptive threshold, as follows:

[0104] If the operating status level is "warning status" or "fault status", the mean and standard deviation of the fault contribution of all devices are calculated to determine an adaptive threshold, and devices with a fault contribution greater than the adaptive threshold are identified as faulty devices. The formula for the adaptive threshold is as follows:

[0105] ,

[0106] in, To adaptively determine the threshold, The mean, Standard deviation; This is the sensitivity coefficient, which can be 1 or 1.5.

[0107] like Figure 4 As shown, this embodiment calculates the fault contribution of each monitoring device in a substation. Figure 4 The horizontal axis represents the names of different equipment in the substation, and the vertical axis represents the fault contribution value. It can be seen that under the current sampling window, the fault contribution of equipment "circuit breaker QF2" is significantly higher than that of other equipment and exceeds the adaptive threshold by a large margin, while the fault contribution of other equipment is below the threshold, thus locking in the source of the fault.

[0108] S405. Generate equipment diagnostic results including operating status levels and faulty equipment, as follows:

[0109] The operating status level, faulty equipment name, faulty equipment contribution value, and timestamp of the anomaly are encapsulated into a structured equipment diagnostic result, and pop-up alarms are issued through a visual human-machine interface or sent to the remote control backend through the station's communication network.

[0110] This step introduces a gradient-based sensitivity analysis mechanism and uses the backpropagation algorithm to accurately quantify the contribution weight of each device to the overall anomaly. At the same time, combined with dynamic statistical threshold determination, it establishes the priority of fault handling and realizes adaptive screening of single-point or multi-point concurrent faults, thereby achieving accurate location and interpretable diagnosis of substation equipment faults.

[0111] like Figure 5As shown, a fault diagnosis device for intelligent substation equipment includes the following units:

[0112] The signal graphing unit is used to spatially align the power sampling signal based on the device connection relationship to obtain a structured graph signal.

[0113] The manifold mapping unit is used to perform feature transformation on the structured graph signal to obtain a statistical correlation matrix and map it to real-time state feature points in the Riemannian manifold space.

[0114] A geometric measurement unit is used to obtain the geometric deviation between the real-time state feature point and the preset ideal state point by using a logarithmic mapping operator based on the curvature characteristics of the Riemannian manifold space.

[0115] The positioning and diagnostic unit is used to calculate the fault contribution of each device using the geometric deviation and to identify the faulty device, and output the device diagnostic results.

[0116] like Figure 6 As shown, an electronic device includes:

[0117] Memory, used to store computer programs;

[0118] A processor is used to implement the steps of the intelligent substation equipment fault diagnosis method when executing the computer program.

[0119] like Figure 7 As shown, a computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of the intelligent substation equipment fault diagnosis method.

[0120] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0121] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0122] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0123] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0124] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0125] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for diagnosing faults in intelligent substation equipment, characterized in that, Includes the following steps: Spatial alignment of power sampling signals based on device connection relationships yields structured graph signals; Perform a graph Fourier transform on the structured graph signal to obtain the graph frequency domain features, and then perform sample covariance calculation to obtain the statistical correlation matrix; The statistical correlation matrix is ​​mapped to the symmetric positive definite matrix Riemannian manifold space to obtain real-time state feature points; The Riemann logarithmic mapping operator is used to map the real-time state feature points to a reference tangent space with a preset ideal state point as the tangent point, thereby generating a Riemann logarithmic tangent vector. The norm of the Riemann logarithmic tangent vector is calculated and used as the geometric deviation of the real-time state feature point relative to the preset ideal state point. The fault contribution of each device is calculated using the geometric deviation, the faulty device is identified, and the device diagnosis results are output.

2. The intelligent substation equipment fault diagnosis method according to claim 1, characterized in that, The obtained structured graph signal includes: Construct a topology association matrix based on the physical connection relationships of the substation equipment; Based on the power sampling signal of the device, a node feature matrix corresponding to the node index of the correlation matrix is ​​constructed; The correlation matrix and the feature matrix are combined to obtain the structured graph signal.

3. The intelligent substation equipment fault diagnosis method according to claim 2, characterized in that, The calculation of the fault contribution of each device using the geometric deviation includes: Based on the calculation path of the geometric deviation, the partial derivative of the geometric deviation with respect to the node feature matrix is ​​calculated in reverse to obtain the node sensitivity matrix; Calculate the norm of each row vector in the node sensitivity matrix, which is used as the fault contribution of the corresponding device node.

4. The intelligent substation equipment fault diagnosis method according to claim 1, characterized in that, The output device diagnostic results include: The geometric deviation is mapped to a preset state range to obtain the substation's operating state level; An adaptive threshold is determined based on the statistical characteristic parameters of the fault contribution, and the faulty equipment is determined by comparing the fault contribution with the adaptive threshold. The output includes the device diagnostic results for the operating status level and faulty devices.

5. A fault diagnosis device for intelligent substation equipment, characterized in that, Includes the following units: The signal graphing unit is used to spatially align the power sampling signal based on the device connection relationship to obtain a structured graph signal. The manifold mapping unit is used to perform a graph Fourier transform on the structured graph signal, obtain graph frequency domain features, perform sample covariance calculation to obtain a statistical correlation matrix, and map the statistical correlation matrix to a symmetric positive definite matrix Riemann manifold space to obtain real-time state feature points. A geometric metric unit is used to map the real-time state feature points to a reference tangent space with a preset ideal state point as the tangent point using the Riemann logarithmic mapping operator, thereby generating a Riemann logarithmic tangent vector. The norm of the Riemann logarithmic tangent vector is calculated and used as the geometric deviation of the real-time state feature point relative to the preset ideal state point. The positioning and diagnostic unit is used to calculate the fault contribution of each device using the geometric deviation and to identify the faulty device, and output the device diagnostic results.

6. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the intelligent substation equipment fault diagnosis method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of the intelligent substation equipment fault diagnosis method as described in any one of claims 1 to 4.

Images