A method and system for relay protection setting value analysis and verification

Abstract

The application discloses a kind of relay protection fixed value analysis verification method and system, it is related to relay protection technical field, including using exponential coupling formula method to calculate final coupling strength, using coupling matrix standardization method to construct coupling matrix;Using eigenvalue decomposition verification method to the updated manifold center is eigenvalue decomposition, organization is three order tensor, using Gaussian kernel similarity calculation method calculates the Gaussian kernel similarity between fault feature and historical fault feature, generates similarity score;Calculate fixed value deviation and update.The application introduces differential geometry, information geometry and quantum state coupling theory, constructs mode covariance tensor model based on symmetric positive definite manifold, proposes the exponential coupling mechanism of intermodal Riemann distance, tension index and perturbation coding, improves the structural integrity and coupling strength representation capability of mode information fusion, enhances the scientificity and operability of fixed value evaluation.

Description

Technical Field

[0001] This invention relates to the field of relay protection technology, and in particular to a method and system for analyzing and verifying relay protection settings. Background Technology

[0002] With the development of smart grids and automated power systems, relay protection technology, as one of the key technologies to ensure the safe and stable operation of the power grid, has a crucial impact on the operation of the power system due to the rationality and accuracy of its settings. Traditionally, relay protection settings are mainly calculated manually based on electrical parameters such as short-circuit current and load changes, and verified by simulation analysis or experience. With the rapid development of information and communication technology, a large amount of operation, maintenance and environmental data are collected in real time, providing a rich data foundation for the intelligent analysis and auxiliary verification of relay protection settings.

[0003] Existing methods for verifying relay protection settings still have shortcomings. The structural heterogeneity and temporal inconsistency of multimodal data make it impossible for traditional data fusion methods to accurately express the coupling mechanism between modes, resulting in error accumulation and uncertainty in the analysis results. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method and system for analyzing and verifying relay protection settings, which solves the problem that the structural heterogeneity and temporal inconsistency of multimodal data make it impossible for traditional data fusion methods to accurately express the coupling mechanism between modes, resulting in error accumulation and uncertainty in the analysis results.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for analyzing and verifying relay protection settings, comprising,

[0008] Three types of modal data were collected and preprocessed. The covariance matrix of each mode was calculated using the modal covariance matrix construction method. The covariance matrix of each mode was embedded into a symmetric positive definite manifold using the symmetric positive definite manifold embedding method, treating each mode as a point on the symmetric positive definite manifold. The Riemann distance between modes was calculated. The tension index was calculated using the matrix trace calculation method. Based on the Riemann distance and the tension index, the final coupling strength was calculated using the exponential coupling formula method. The coupling matrix was constructed using the coupling matrix normalization method.

[0009] The Karcher average iteration method is used to calculate the manifold center of all modal covariance matrices. The eigenvalue verification method is used to perform eigenvalue decomposition on the updated manifold center, which is organized into a third-order tensor. The modal slices of the third-order tensor are normalized and weighted using a coupling matrix to generate a fused feature matrix. The L1 regularized sparse coding method is used to construct an optimization problem. A support vector machine model is built to predict the fault detection results. The Gaussian kernel similarity calculation method is used to calculate the Gaussian kernel similarity between the fault features and historical fault features, and a similarity score is generated.

[0010] Linear projection is used to map fault features to a setpoint space, generating fault feature vectors, calculating setpoint deviations and updates, and constructing a visualization interface to display the three types of modal data.

[0011] As a preferred embodiment of the relay protection setting analysis and verification method of the present invention, the step of calculating the final coupling strength using the exponential coupling formula method and constructing the coupling matrix using the coupling matrix normalization method includes:

[0012] The multimodal feature matrix construction method is used to concatenate the three types of modal data into feature vectors according to time windows, and then stack the feature vectors of the modalities row by row to generate the modal feature matrix.

[0013] Based on the modality feature matrix, the phase time and mean vector of each feature column are calculated. The phase time is organized into a matrix according to the time window and feature dimension to obtain the priority matrix.

[0014] The covariance matrix of each mode is calculated using the modal covariance matrix construction method, and the covariance matrix is ​​eigenvalued using the singular value decomposition method to obtain the eigenvalue decomposition results.

[0015] The covariance matrix of each mode is embedded into a symmetric positive definite manifold using the symmetric positive definite manifold embedding method, treating it as a point on the symmetric positive definite manifold. Based on the eigenvalue decomposition results, the Riemann distance between modes is calculated, and the tension index is calculated using the matrix trace method.

[0016] The perturbation sequence matrix is ​​calculated using the perturbation sequence generation method. The perturbation coding matrix is ​​generated by adding the priority matrix and the perturbation sequence element by element. The modulation coupling strength between modes is calculated using the frequency domain inner product projection method. The final coupling strength is calculated using the exponential coupling formula method based on the Riemann distance and tension index. The coupling matrix is ​​constructed using the coupling matrix normalization method.

[0017] As a preferred embodiment of the relay protection setting analysis and verification method of the present invention, wherein: the updated manifold center is decomposed using the eigenvalue decomposition verification method, and organized into a third-order tensor, including:

[0018] The Karcher average iteration method is used to calculate the manifold center of all modal covariance matrices. The maximum number of iterations is set using the convergence threshold method. When the maximum number of iterations is reached, the iteration stops to obtain the updated manifold center. The eigenvalue verification method is used to perform eigenvalue decomposition on the updated manifold center. The principal component extraction method is used to extract the first r principal components to generate the fused feature vector.

[0019] The fused feature vectors are organized into third-order tensors using tensor organization.

[0020] As a preferred embodiment of the relay protection setting analysis and verification method of the present invention, the step of calculating the Gaussian kernel similarity between fault features and historical fault features using the Gaussian kernel similarity calculation method to generate a similarity score includes:

[0021] Modal slices are extracted from a third-order tensor using the tensor slice extraction method. The modal slices of the third-order tensor are normalized and weighted using a coupling matrix to generate a fusion feature matrix. A dictionary matrix is ​​randomly initialized. Based on the fusion feature matrix and the dictionary matrix, an optimization problem is constructed using the L1 regularized sparse coding method. The problem is solved using the proximal gradient method to obtain a sparse coefficient matrix. The sparse coefficient matrix is ​​transposed to generate a fixed-value verification feature.

[0022] Based on the dictionary matrix and the sparse coefficient matrix, matrix multiplication is used to calculate the reconstructed feature matrix and the reconstruction error is calculated. An empirical threshold setting method is used to set the error threshold. If the reconstruction error is less than the error threshold, the sparse coding is considered to be effective. Otherwise, the sparse regularization parameter is adjusted using the parameter iterative adjustment method until the reconstruction error is less than the error threshold and the adjustment stops.

[0023] A support vector machine model is constructed to predict fault detection results. Based on the fault detection results, fault features are selected from the fixed-value verification features and grouped according to fault type to form fault features.

[0024] Historical fault features are collected from a historical fault scenario database. The Gaussian kernel similarity between the fault features and the historical fault features is calculated using the Gaussian kernel similarity calculation method, and then normalized to generate a similarity score.

[0025] As a preferred embodiment of the relay protection setting analysis and verification method of the present invention, the calculation and updating of setting deviation includes:

[0026] The original setpoint parameters, including voltage setpoints, current setpoints, and power setpoints, are extracted from the relay protection system configuration of the substation using the setpoint initialization extraction method. The original setpoint feature vector is then generated using the direct splicing method.

[0027] Select the fault feature with the highest similarity score, use linear projection to map the fault feature to the fixed value space, generate the fault feature vector, calculate the fixed value deviation based on the original fixed value feature vector, use the fixed value deviation to update the original fixed value vector, and obtain the corrected fixed value vector.

[0028] As a preferred embodiment of the relay protection setting analysis and verification method of the present invention, the construction of a visual interface to display three types of modal data includes:

[0029] A visual interface is built to display the collected three types of modal data and the corrected constant vectors, allowing verified users to view them.

[0030] As a preferred embodiment of the relay protection setting analysis and verification method of the present invention, the step of collecting three types of modal data and performing preprocessing includes:

[0031] Three types of modal data were acquired from the relay protection system using a multimodal data acquisition method, and then denoised and normalized.

[0032] The three types of modal data include three-phase current, three-phase voltage, power factor, operating time, number of maintenance operations, number of failures, temperature, humidity, air pressure, and wind speed data.

[0033] Secondly, the present invention provides a relay protection setting analysis and verification system, comprising,

[0034] The data collection module is used to collect three types of modal data and perform preprocessing. It calculates the covariance matrix of each mode using the modal covariance matrix construction method, embeds the covariance matrix of each mode into a symmetric positive definite manifold using the symmetric positive definite manifold embedding method, treating each mode as a point on the symmetric positive definite manifold, calculates the Riemann distance between modes, calculates the tension index using the matrix trace calculation method, calculates the final coupling strength based on the Riemann distance and the tension index using the exponential coupling formula method, and constructs the coupling matrix using the coupling matrix normalization method.

[0035] The Tensor Score module is used to calculate the manifold center of all modal covariance matrices using the Karcher average iteration method, perform eigenvalue decomposition on the updated manifold center using the eigenvalue decomposition verification method, organize it into a third-order tensor, normalize and weight the modal slices of the third-order tensor using the coupling matrix to generate a fused feature matrix, construct an optimization problem using the L1 regularized sparse coding method, build a support vector machine model to predict the fault detection results, and calculate the Gaussian kernel similarity between fault features and historical fault features using the Gaussian kernel similarity calculation method to generate a similarity score.

[0036] The verification visualization module is used to map fault features to a setpoint space using linear projection, generate fault feature vectors, calculate and update setpoint deviations, and build a visualization interface to display three types of modal data.

[0037] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the relay protection setting analysis and verification method as described in the first aspect of the present invention.

[0038] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the relay protection setting analysis and verification method as described in the first aspect of the present invention.

[0039] The beneficial effects of this invention are as follows: By introducing differential geometry, information geometry and quantum state coupling theory, this invention constructs a modal covariance tensor model based on a symmetric positive definite manifold, proposes an exponential coupling mechanism combining intermodal Riemann distance, tension index and perturbation coding, improves the structural integrity and coupling strength characterization ability of modal information fusion, and enhances the scientificity and operability of the setpoint evaluation. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart of the relay protection setting analysis and verification method in Example 1.

[0042] Figure 2 This is a schematic diagram of the relay protection setting analysis and verification system in Example 1.

[0043] Figure 3 This is a schematic diagram of the embedded symmetrical positive definite manifold in the relay protection setting analysis and verification method in Example 1.

[0044] Figure 4 This is a schematic diagram of the visualization interface displaying three types of modal data in the relay protection setting analysis and verification method in Example 1. Detailed Implementation

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0046] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0047] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0048] Example 1, referring to Figures 1 to 4 This is the first embodiment of the present invention, which provides a method for analyzing and verifying relay protection settings, including the following steps:

[0049] S1. Collect three types of modal data and preprocess them. Use the modal covariance matrix construction method to calculate the covariance matrix of each mode. Use the symmetric positive definite manifold embedding method to embed the covariance matrix of each mode into a symmetric positive definite manifold, treating it as a point on the symmetric positive definite manifold. Calculate the Riemann distance between modes. Use the matrix trace calculation method to calculate the tension index. Based on the Riemann distance and the tension index, use the exponential coupling formula method to calculate the final coupling strength. Use the coupling matrix normalization method to construct the coupling matrix.

[0050] Specifically, three types of modal data are collected and preprocessed, including:

[0051] Three types of modal data were acquired from the relay protection system using a multimodal data acquisition method, and then denoised and normalized.

[0052] The three types of modal data include three-phase current, three-phase voltage, power factor, operating time, number of maintenance operations, number of failures, temperature, humidity, air pressure, and wind speed data.

[0053] Multimodal acquisition enhances the system's comprehensive state perception capability, the preprocessing process effectively improves the accuracy and robustness of subsequent feature analysis, and multimodal input enhances the system's redundancy and information diversity, providing a solid data foundation for complex state modeling.

[0054] Furthermore, the final coupling strength is calculated using the exponential coupling formula method, and the coupling matrix is ​​constructed using the coupling matrix normalization method, including:

[0055] The multimodal feature matrix construction method is used to concatenate the three types of modal data into feature vectors according to time windows (using the fixed time window method). The feature vectors of the modes are stacked row by row to generate the modal feature matrix (including electrical modes (three-phase current, three-phase voltage and power factor), operation and maintenance modes (running time, number of maintenance and number of failures) and environmental mode matrix (temperature, humidity, air pressure and wind speed)).

[0056] Based on the modality-based feature matrix, the phase time and mean vector of each feature column are calculated. The phase time is then organized into a matrix according to the time window and feature dimension to obtain the priority matrix, where rows represent samples and columns represent feature dimensions. The formula is as follows:

[0057]

[0058] Where P i [j,k] represents the phase time of mode i at the j-th time window and the k-th feature, X i [j,k] represents the k-th feature value of mode i in the j-th time window, rank(X) i [j,k]in X i [:,k]) is X i [j,k] is in the k-th column X i The ranking in [:,k] Let n be the feature mean vector of mode i, and n be the number of time windows. Let be the feature vector of the j-th time window of mode i;

[0059] The covariance matrix of each mode is calculated using the modal covariance matrix construction method, and the formula is as follows:

[0060]

[0061] Where Σ i Let be the covariance matrix of mode i, and T be the transpose;

[0062] The covariance matrix is ​​decomposed using singular value decomposition to obtain the eigenvalue decomposition results, including the eigenvector matrix and the eigenvalue diagonal matrix.

[0063] The covariance matrix of each mode is embedded into a symmetric positive definite manifold using the symmetric positive definite manifold embedding method, treating it as a point on the symmetric positive definite manifold. Based on the eigenvalue decomposition results, the Riemann distance between modes is calculated using the following formula:

[0064]

[0065] Where D R (Q i Q u Q is the Riemann distance between mode i and mode u.i and Q u Let λ be the covariance matrix of modes i and u, respectively, embedded at a point on a symmetric positive definite manifold. k Let be the k-th eigenvalue of the similarity transformation matrix, and d be the minimum common feature dimension of the similarity transformation matrix;

[0066] The inverse square root of the covariance matrix is ​​calculated using the matrix inverse square root method. The formula is as follows:

[0067]

[0068] in U is the inverse square root of the covariance matrix of mode i. i and Λ i These are the eigenvalue decomposition results for mode i;

[0069] The similarity transformation matrix is ​​calculated using the matrix similarity transformation method, and the formula is:

[0070]

[0071] Where A is the similarity transformation matrix, Σ u Let U be the covariance matrix of mode u;

[0072] The eigenvalues ​​of the similarity transformation matrix are obtained by using the eigenvalue extraction method to perform eigenvalue decomposition on the similarity transformation matrix.

[0073] The tension index is calculated using the matrix trace method, and the formula is as follows:

[0074]

[0075] Among them O iu Let tr(·) be the tension index between mode i and mode u, and let tr(·) be the matrix trace, the sum of the elements on the main diagonal.

[0076] The perturbation sequence matrix is ​​calculated using the perturbation sequence generation method, as shown in the formula:

[0077]

[0078] Where Δ i Let d be the perturbation sequence matrix of mode i, where rows represent samples and columns represent feature dimensions. i Let σ be the feature dimension of mode i. i,k For X i The standard deviation of [:,k], φ i The phase offset for mode i (1 represents electrical mode, 2 represents operation and maintenance mode, and 3 represents environmental mode, set using the fixed phase offset method);

[0079] The perturbation coding matrix is ​​generated by element-wise addition of the priority matrix and the perturbation sequence, as shown in the formula:

[0080]

[0081] Where M i Let P be the perturbation coding matrix for mode i. i Let be the priority matrix for mode i;

[0082] The modulation coupling strength between modes is calculated using the frequency domain inner product projection method, and the formula is as follows:

[0083]

[0084] in M represents the modulation coupling strength between modes i and u. i [j,k] and M u [j,k] represent the perturbation coding values ​​of modes i and u at the j-th time window and the k-th feature, respectively. and Let be the self-modulation coupling strength of modes i and u;

[0085] Based on the Riemann distance and tension index, the final coupling strength is calculated using the exponential coupling formula, and the coupling matrix is ​​constructed using the coupling matrix normalization method, as shown in the formula:

[0086]

[0087] Where C iu The final coupling strength between mode i and mode u is given by α, where α is the fusion weight, controlling the proportion of geometric and frequency domain contributions (set using a fixed weight assignment method), and C is the fusion weight. ii and C uu , respectively, represent the final coupling strengths of modes i and u.

[0088] By characterizing modes using covariance matrices, a statistical feature space of modes is constructed, enabling mathematical reduction of modes. Symmetric positive definite manifolds are used to improve the geometric accuracy of mode representation, ensuring that similarity calculations reflect statistical structural differences and avoiding errors caused by coordinate rotation or scale mismatch in Euclidean space. This enhances the robustness and reliability of similarity calculations between modes. Introducing matrix traces as a representation of modal coupling strength is an effective way to combine frequency domain response features; the sum of its main diagonal elements reflects the strength of the main components of modal coupling, avoiding direct comparison using high-dimensional frequency domain spectra and improving computational efficiency. By using traces, the principal component responses of modal coupling are focused on, ignoring non-primary perturbations, effectively improving the accuracy of modulation intensity estimation. Perturbation sequence encoding is introduced into modal coupling analysis, building upon traditional modal comparison... This invention introduces disturbance response sensitivity analysis to simulate the coupled response of modes to noise or external influences, enhancing the system's coupling analysis capability in dynamic environments and improving the practical adaptability of modeling. By simulating disturbances, it reflects the changing responses of different modes in real complex scenarios, thereby improving the dynamic adaptability of coupling strength assessment. Traditional modal coupling methods are mostly based on a single distance or correlation index. This invention integrates Riemann distance and tension index into the exponential coupling formula to construct a multi-level coupling strength model. The exponential function has nonlinear amplification characteristics, making strong coupling relationships more prominent on the exponential scale, while weak relationships approach zero, which is convenient for subsequent clustering or discrimination. It effectively solves the problem of excessively concentrated or excessively discrete numerical distribution of coupling degree between modes, giving the coupling matrix good distribution characteristics and facilitating structural analysis.

[0089] S2. Use the Karcher average iteration method to calculate the manifold center of all modal covariance matrices. Use the eigenvalue decomposition verification method to perform eigenvalue decomposition on the updated manifold center, organize it into a third-order tensor, use the coupling matrix to normalize and weight the modal slices of the third-order tensor to generate a fusion feature matrix, use the L1 regularized sparse coding method to construct an optimization problem, build a support vector machine model to predict the fault detection results, use the Gaussian kernel similarity calculation method to calculate the Gaussian kernel similarity between fault features and historical fault features, and generate a similarity score.

[0090] Specifically, the updated manifold centers are decomposed using the eigenvalue decomposition verification method, organized as third-order tensors, including:

[0091] The manifold centers are calculated using the Karcher average iterative method to obtain the modal covariance matrices. A convergence threshold is used to set the maximum number of iterations. Iteration stops when the maximum number of iterations is reached, yielding the updated manifold centers. The formula is as follows:

[0092]

[0093] Where Σ cLet be the manifold center covariance matrix, and let be the geometric center of the three modal covariance matrices on a symmetric positive definite manifold. Let Σ be the squared Riemann distance between the covariance matrix of mode i and the symmetric positive definite matrix Σ. (t) Let be the center of the manifold in the t-th iteration;

[0094] The updated manifold center is decomposed using the eigenvalue decomposition verification method, and the first r principal components are extracted using the principal component extraction method, where r is the number of principal components, to generate a fused feature vector.

[0095] The fused feature vectors are organized into third-order tensors using tensor organization.

[0096] The Karcher average iterative method utilizes the geometric properties of Riemannian manifolds to capture the statistical distribution characteristics of multimodal data by iteratively approximating the manifold center. This solves the mismatch problem of traditional mean methods on high-dimensional nonlinear data. The application of Riemann distance improves the representativeness of the mode center and enhances the accuracy of subsequent eigenvalue decomposition. The introduction of third-order tensors organizes multimodal data into a high-dimensional structure, preserves the interaction information between modes, and solves the problems of dimensional inconsistency and information redundancy in multimodal data.

[0097] Furthermore, the Gaussian kernel similarity calculation method is used to calculate the Gaussian kernel similarity between the fault features and historical fault features, generating a similarity score, including:

[0098] Modal slices are extracted from a third-order tensor using the tensor slicing method. A coupling matrix is ​​then used to normalize and weight the modal slices of the third-order tensor to generate a fusion feature matrix. The formula is as follows:

[0099]

[0100] in This is the weighted fusion feature matrix. For tensor The slice of the i-th mode;

[0101] The dictionary matrix is ​​randomly initialized. Based on the fused feature matrix and dictionary matrix, an optimization problem is constructed using L1 regularized sparse coding, with the following formula:

[0102]

[0103] Where A is the sparse coefficient matrix. The L1 regularization parameter controls sparsity, ||·|| F Given the Frobenius norm, calculate the square root of the sum of squares of the matrix;

[0104] The sparse coefficient matrix is ​​obtained by using the proximal gradient method. The sparse coefficient matrix is ​​then transposed to generate a constant value verification feature.

[0105] Based on the dictionary matrix and the sparse coefficient matrix, matrix multiplication is used to calculate the reconstructed feature matrix, and the reconstruction error is calculated. An empirical threshold method is used to set the error threshold. If the reconstruction error is less than the error threshold, the sparse coding is considered effective; otherwise, the sparse regularization parameters are adjusted using an iterative parameter adjustment method until the reconstruction error is less than the error threshold. The formula is as follows:

[0106]

[0107] Where E is the reconstruction error. To reconstruct the feature matrix, we have denoted as the features reconstructed through sparse coding.

[0108] Collect labeled historical three-modal data, calculate the fixed-value verification features, and generate a training set;

[0109] Construct a support vector machine model, train the support vector machine model using the training set, and iteratively optimize the support vector machine model parameters using a loss function and the Adam optimizer.

[0110] The fixed-value verification features are input into the trained support vector machine model to obtain the fault detection results and the corresponding fault types (such as normal, short circuit and ground).

[0111] Based on the fault detection results, fault features are selected from the set value verification features and grouped according to the fault type to form fault features;

[0112] Historical fault features are collected from a historical fault scenario database. The Gaussian kernel similarity between the fault features and the historical fault features is calculated using the Gaussian kernel similarity calculation method, and then normalized to generate a similarity score.

[0113] Normalized weighting ensures a balanced contribution of each mode (such as voltage and current) to the fused features. An optimization problem is constructed through L1 regularized sparse coding, and the sparse coefficient matrix is ​​solved by the near-end gradient method. The parameters are adaptively adjusted by reconstruction error and empirical threshold, avoiding model bias caused by manually setting regularization parameters in traditional sparse coding, thus enhancing the adaptive ability of feature representation. Gaussian kernels can capture complex relationships between features, improving the accuracy in ground fault identification and avoiding bias problems caused by information imbalance between modes. The coupling matrix enhances the structural consistency and semantic complementarity of modal information.

[0114] S3. Use linear projection to map fault features to the setpoint space, generate fault feature vectors, calculate and update setpoint deviations, and build a visualization interface to display three types of modal data.

[0115] Specifically, the calculation and update of the setpoint deviation includes:

[0116] The original setpoint parameters, including voltage setpoints, current setpoints, and power setpoints, are extracted from the relay protection system configuration of the substation using the setpoint initialization extraction method. The original setpoint feature vector is then generated using the direct splicing method.

[0117] The fault feature with the highest similarity score is selected, and linear projection is used to map the fault feature to the fixed value space to generate a fault feature vector. Based on the original fixed value feature vector, the fixed value deviation is calculated using the following formula:

[0118]

[0119] Where ΔV is the constant deviation vector. Set the learning rate (using the empirical learning rate setting method). V is the fault feature vector. set The original constant eigenvector;

[0120] The original constant vector is updated using the constant deviation to obtain the corrected constant vector, as shown in the formula:

[0121] V new =V set +ΔV,

[0122] Where V new This is the corrected constant vector.

[0123] The direct concatenation method integrates multi-dimensional setpoint parameters into a single vector, simplifying the complexity of subsequent calculations. Linear projection can reduce the dimensionality of high-dimensional and complex fault features to the same space as the setpoint parameters, preserving key information while reducing computational complexity. The adjustment direction and magnitude are clearly quantified through vector difference. The introduction of the learning rate avoids excessive adjustment leading to false triggering of protection logic. The incremental update strategy ensures that the setpoint adjustment is based on actual fault characteristics, which is more in line with real-time operating conditions compared to fixed rule adjustment. The update step size is controlled by the learning rate, avoiding excessive adjustment leading to false operation or failure to operate of the protection system.

[0124] Furthermore, a visual interface is constructed to display the three types of modal data, including:

[0125] A visual interface is built to display the collected three types of modal data and the corrected constant vectors, allowing verified users to view them.

[0126] The visualization interface can display modal data and setpoint change trends through charts (such as line charts and bar charts), helping operation and maintenance personnel to quickly identify anomalies. The real-name verification mechanism ensures that only authorized users can access sensitive data, which complies with power system information security standards.

[0127] This embodiment also provides a relay protection setting analysis and verification system, including:

[0128] The data collection module is used to collect three types of modal data and perform preprocessing. It calculates the covariance matrix of each mode using the modal covariance matrix construction method, embeds the covariance matrix of each mode into a symmetric positive definite manifold using the symmetric positive definite manifold embedding method, treating each mode as a point on the symmetric positive definite manifold, calculates the Riemann distance between modes, calculates the tension index using the matrix trace calculation method, calculates the final coupling strength based on the Riemann distance and the tension index using the exponential coupling formula method, and constructs the coupling matrix using the coupling matrix normalization method.

[0129] The Tensor Score module is used to calculate the manifold center of all modal covariance matrices using the Karcher average iteration method, perform eigenvalue decomposition on the updated manifold center using the eigenvalue decomposition verification method, organize it into a third-order tensor, normalize and weight the modal slices of the third-order tensor using the coupling matrix to generate a fused feature matrix, construct an optimization problem using the L1 regularized sparse coding method, build a support vector machine model to predict the fault detection results, and calculate the Gaussian kernel similarity between fault features and historical fault features using the Gaussian kernel similarity calculation method to generate a similarity score.

[0130] The verification visualization module is used to map fault features to a setpoint space using linear projection, generate fault feature vectors, calculate and update setpoint deviations, and build a visualization interface to display three types of modal data.

[0131] This embodiment also provides a computer device applicable to the relay protection setting analysis and verification method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the relay protection setting analysis and verification method proposed in the above embodiment.

[0132] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0133] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the relay protection setting analysis and verification method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0134] In summary, this invention introduces differential geometry, information geometry, and quantum state coupling theory to construct a modal covariance tensor model based on a symmetric positive definite manifold. It proposes an exponential coupling mechanism that combines intermodal Riemann distance, tension index, and perturbation coding, thereby improving the structural integrity and coupling strength characterization capability of modal information fusion and enhancing the scientificity and operability of the setpoint evaluation.

[0135] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for analyzing and verifying relay protection settings, characterized in that: include, Three types of modal data were collected and preprocessed. The covariance matrix of each mode was calculated using the modal covariance matrix construction method. The covariance matrix of each mode was embedded into a symmetric positive definite manifold using the symmetric positive definite manifold embedding method, treating each mode as a point on the symmetric positive definite manifold. The Riemann distance between modes was calculated. The tension index was calculated using the matrix trace calculation method. Based on the Riemann distance and the tension index, the final coupling strength was calculated using the exponential coupling formula method. The coupling matrix was constructed using the coupling matrix normalization method. The three types of modes include electrical mode, operation and maintenance mode, and environmental mode. The Karcher average iteration method is used to calculate the manifold center of all modal covariance matrices. The eigenvalue verification method is used to perform eigenvalue decomposition on the updated manifold center, which is organized into a third-order tensor. The modal slices of the third-order tensor are normalized and weighted using a coupling matrix to generate a fused feature matrix. The L1 regularized sparse coding method is used to construct an optimization problem. A support vector machine model is built to predict the fault detection results. The Gaussian kernel similarity calculation method is used to calculate the Gaussian kernel similarity between the fault features and historical fault features, and a similarity score is generated. Linear projection is used to map fault features to a setpoint space, generate fault feature vectors, calculate setpoint deviation and update, and build a visualization interface to display three types of modal data. The step of calculating the Gaussian kernel similarity between fault features and historical fault features using the Gaussian kernel similarity calculation method to generate a similarity score includes: Modal slices are extracted from a third-order tensor using the tensor slice extraction method. The modal slices of the third-order tensor are normalized and weighted using a coupling matrix to generate a fusion feature matrix. A dictionary matrix is ​​randomly initialized. Based on the fusion feature matrix and the dictionary matrix, an optimization problem is constructed using the L1 regularized sparse coding method. The problem is solved using the proximal gradient method to obtain a sparse coefficient matrix. The sparse coefficient matrix is ​​transposed to generate a fixed-value verification feature. Based on the dictionary matrix and the sparse coefficient matrix, matrix multiplication is used to calculate the reconstructed feature matrix and the reconstruction error is calculated. An empirical threshold setting method is used to set the error threshold. If the reconstruction error is less than the error threshold, the sparse coding is considered to be effective. Otherwise, the sparse regularization parameter is adjusted using the parameter iterative adjustment method until the reconstruction error is less than the error threshold and the adjustment stops. A support vector machine model is constructed to predict fault detection results. Based on the fault detection results, fault features are selected from the fixed-value verification features and grouped according to fault type to form fault features. Historical fault features are collected from a historical fault scenario database. The Gaussian kernel similarity between the fault features and the historical fault features is calculated using the Gaussian kernel similarity calculation method, and then normalized to generate a similarity score.

2. The relay protection setting analysis and verification method as described in claim 1, characterized in that: The calculation of the final coupling strength using the exponential coupling formula method and the construction of the coupling matrix using the coupling matrix normalization method include: The multimodal feature matrix construction method is used to concatenate the three types of modal data into feature vectors according to time windows, and then stack the feature vectors of the modalities row by row to generate the modal feature matrix. Based on the modality feature matrix, the phase time and mean vector of each feature column are calculated. The phase time is organized into a matrix according to the time window and feature dimension to obtain the priority matrix. The covariance matrix of each mode is calculated using the modal covariance matrix construction method, and the eigenvalue decomposition method is used to perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalue decomposition results. The covariance matrix of each mode is embedded into a symmetric positive definite manifold using the symmetric positive definite manifold embedding method, treating it as a point on the symmetric positive definite manifold. Based on the eigenvalue decomposition results, the Riemann distance between modes is calculated, and the tension index is calculated using the matrix trace method. The perturbation sequence matrix is ​​calculated using the perturbation sequence generation method. The perturbation coding matrix is ​​generated by adding the priority matrix and the perturbation sequence element by element. The modulation coupling strength between modes is calculated using the frequency domain inner product projection method. The final coupling strength is calculated using the exponential coupling formula method based on the Riemann distance and tension index. The coupling matrix is ​​constructed using the coupling matrix normalization method.

3. The relay protection setting analysis and verification method as described in claim 2, characterized in that: The updated manifold center is decomposed using the eigenvalue decomposition verification method, organized as a third-order tensor, including: The Karcher average iteration method is used to calculate the manifold center of all modal covariance matrices. The maximum number of iterations is set using the convergence threshold method. When the maximum number of iterations is reached, the iteration stops to obtain the updated manifold center. The eigenvalue verification method is used to perform eigenvalue decomposition on the updated manifold center. The principal component extraction method is used to extract the first r principal components to generate the fused feature vector. The fused feature vectors are organized into third-order tensors using tensor organization.

4. The relay protection setting analysis and verification method as described in claim 3, characterized in that: The calculation and updating of the setpoint deviation includes: The original setpoint parameters, including voltage setpoints, current setpoints, and power setpoints, are extracted from the relay protection system configuration of the substation using the setpoint initialization extraction method. The original setpoint feature vector is then generated using the direct splicing method. Select the fault feature with the highest similarity score, use linear projection to map the fault feature to the fixed value space, generate the fault feature vector, calculate the fixed value deviation based on the original fixed value feature vector, use the fixed value deviation to update the original fixed value vector, and obtain the corrected fixed value vector.

5. The relay protection setting analysis and verification method as described in claim 4, characterized in that: The constructed visualization interface displays three types of modal data, including: A visual interface is built to display the collected three types of modal data and the corrected constant vectors, allowing verified users to view them.

6. The relay protection setting analysis and verification method as described in claim 1, characterized in that: The collection and preprocessing of three types of modal data includes: Three types of modal data were acquired from the relay protection system using a multimodal data acquisition method, and then denoised and normalized. The three types of modal data include three-phase current, three-phase voltage, power factor, operating time, number of maintenance operations, number of failures, temperature, humidity, air pressure, and wind speed data.

7. A relay protection setting analysis and verification system, based on the relay protection setting analysis and verification method according to any one of claims 1 to 6, characterized in that: include, The data collection module is used to collect three types of modal data and perform preprocessing. It calculates the covariance matrix of each mode using the modal covariance matrix construction method, embeds the covariance matrix of each mode into a symmetric positive definite manifold using the symmetric positive definite manifold embedding method, treating each mode as a point on the symmetric positive definite manifold, calculates the Riemann distance between modes, calculates the tension index using the matrix trace calculation method, calculates the final coupling strength based on the Riemann distance and the tension index using the exponential coupling formula method, and constructs the coupling matrix using the coupling matrix normalization method. The Tensor Score module is used to calculate the manifold center of all modal covariance matrices using the Karcher average iteration method, perform eigenvalue decomposition on the updated manifold center using the eigenvalue decomposition verification method, organize it into a third-order tensor, normalize and weight the modal slices of the third-order tensor using the coupling matrix to generate a fused feature matrix, construct an optimization problem using the L1 regularized sparse coding method, build a support vector machine model to predict the fault detection results, and calculate the Gaussian kernel similarity between fault features and historical fault features using the Gaussian kernel similarity calculation method to generate a similarity score. The verification visualization module is used to map fault features to a setpoint space using linear projection, generate fault feature vectors, calculate and update setpoint deviations, and build a visualization interface to display three types of modal data.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the relay protection setting analysis and verification method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the relay protection setting analysis and verification method according to any one of claims 1 to 6.

Images