Capacitive voltage transformer overreach monitoring method, device and equipment and storage medium

Abstract

The application relates to a method and device for monitoring over-error of a capacitive voltage transformer, electronic equipment and a computer readable storage medium. The method comprises: performing kernel principal component analysis based on historical sample data to determine an SPE statistical quantity threshold; calculating the SPE statistical quantity of current sample data collected in real time; if the SPE statistical quantity is less than or equal to the SPE statistical quantity threshold, determining that the operation state of the transformer group is normal; and if the SPE statistical quantity is greater than the SPE statistical quantity threshold, calculating the contribution rate of each element in the current sample data to the square prediction error statistical quantity, determining a fault element, and locating an over-error transformer in the transformer group. The application effectively processes the nonlinear problem of data by using KPCA, more accurately extracts data features, realizes fault element separation, accurately locates the over-error CVT, performs online real-time monitoring on the CVT error, and improves the stability and reliability of the power system.

Description

Technical Field

[0001] This application relates to the field of power system technology, and in particular to a method, apparatus, electronic device, and computer-readable storage medium for monitoring the out-of-tolerance of a capacitive voltage transformer. Background Technology

[0002] Capacitive voltage transformers (CVTs) are key equipment in power systems, playing an indispensable role in energy metering, relay protection, and power system condition monitoring. CVTs convert high voltage to low voltage suitable for measurement and protection equipment through capacitive voltage division, ensuring stable operation and precise control of all aspects of the power system.

[0003] In actual operation, the performance of CVTs is affected by various factors, such as capacitor aging and insulation degradation caused by long-term operation, drastic changes in ambient temperature and humidity, and external electromagnetic interference. These factors can cause errors in CVTs, and when the errors exceed the specified range (i.e., exceed the tolerance), a series of serious problems will occur. In terms of electricity metering, exceeding the tolerance will lead to inaccurate metering, causing economic disputes between power companies and users, and affecting fair transactions in the electricity market. In the field of relay protection, exceeding the tolerance may cause protection devices to malfunction or fail to operate, threatening the safe and stable operation of the power system, and even causing large-scale power outages.

[0004] Currently, CVT error detection technologies are mainly divided into two categories: offline detection and online detection. Offline detection, specifically the standard comparison method, requires comparing the CVT under test with a standard current transformer offline. This method not only fails to achieve long-term real-time monitoring of the CVT, but also suffers from significant limitations in the portability of standard current transformers, making it difficult to meet the testing needs of complex field environments. Furthermore, the operating conditions during offline detection differ from actual operating conditions, resulting in test results that do not accurately reflect the CVT's actual operating status. Online detection, particularly the live-line comparison scheme, while achieving real-time monitoring to some extent, involves high-voltage live-line operation, posing significant safety hazards. It demands extremely high levels of professional skills from operators and adequate protective equipment, and the operation process is complex and prone to accidents.

[0005] Furthermore, CVT output data exhibits complex distribution characteristics, making it difficult to meet the preconditions such as normal distribution relied upon by traditional monitoring methods. Traditional linear model-based monitoring methods often fail to accurately extract key information from CVT data, which possesses nonlinear characteristics, thus affecting the accuracy and reliability of monitoring results. With the continuous expansion of power system scale and increasing intelligence, higher demands are placed on the accuracy, real-time performance, and reliability of CVT out-of-tolerance monitoring. Traditional monitoring technologies can no longer meet these requirements, necessitating a new monitoring method to ensure the safe, stable, and efficient operation of power systems. Summary of the Invention

[0006] Therefore, it is necessary to provide a method, device, electronic equipment, and computer-readable storage medium for monitoring the out-of-tolerance of capacitive voltage transformers (CVTs) to address the aforementioned technical problems. This method can effectively handle data nonlinearity issues, accurately locate out-of-tolerance CVTs, and improve the stability and reliability of power systems.

[0007] In a first aspect, this application provides a method for monitoring the out-of-tolerance of a capacitive voltage transformer, including:

[0008] Kernel principal component analysis was performed based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data; the historical sample data included multiple sample data collected within a historical time period, and the sample data included three-phase voltage data within the transformer group.

[0009] For the current sample data collected in real time, calculate the squared prediction error statistic of the current sample data;

[0010] If the squared prediction error statistic is less than or equal to the squared prediction error statistic threshold, the current transformer group is considered to be operating normally.

[0011] If the squared prediction error statistic is greater than the squared prediction error statistic threshold, calculate the contribution rate of each element in the current sample data to the squared prediction error statistic to determine the faulty element.

[0012] The faulty element is used to locate the out-of-tolerance transformer in the transformer group.

[0013] In one embodiment, kernel principal component analysis is performed based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data, including:

[0014] Based on a preset kernel function, a kernel matrix is ​​constructed using historical sample data;

[0015] The kernel matrix is ​​centered, and the eigenvalues ​​and eigenvectors of the centered kernel matrix are decomposed.

[0016] Select k eigenvalues ​​from multiple eigenvalues ​​to form a variance contribution diagonal matrix, and select k eigenvectors from multiple eigenvectors to form a principal component direction matrix;

[0017] For the i-th sample data, the principal component projection of the i-th sample data is calculated based on the row vector of the i-th row of the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix.

[0018] Based on the preset kernel function, the value of the kernel function for the i-th sample data is determined, and the squared prediction error statistic for the i-th sample data is determined based on the kernel function value and the principal component projection.

[0019] Based on the squared prediction error statistics corresponding to multiple sample data, the threshold of the squared prediction error statistics corresponding to the preset confidence level is determined by the kernel density estimation strategy.

[0020] In one embodiment, the method further includes:

[0021] The mapping matrix is ​​calculated based on the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix.

[0022] The mapping features of the i-th sample data are determined based on the mapping matrix, and the residual vector of the i-th sample data is calculated based on the mapping features, the mapping matrix, and the principal component projection.

[0023] Calculate the squared prediction error statistic for the i-th sample data based on the residual vector of the i-th sample data.

[0024] In one embodiment, k eigenvalues ​​are selected from multiple eigenvalues ​​to form a variance contribution diagonal matrix, and k eigenvectors are selected from multiple eigenvectors to form a principal component direction matrix, including:

[0025] Select the k largest eigenvalues ​​from multiple eigenvalues ​​so that the cumulative contribution rate reaches a set threshold, and construct a variance contribution diagonal matrix based on the selected k eigenvalues;

[0026] Select the k largest eigenvectors from multiple eigenvectors, and construct the principal component direction matrix based on the selected k eigenvectors.

[0027] In one embodiment, calculating the squared prediction error statistic for the current sample data includes:

[0028] Based on the preset kernel function, construct the current kernel matrix of the current sample data and the historical sample data;

[0029] The current kernel matrix is ​​centered.

[0030] Based on the current kernel matrix after centering, the variance contribution diagonal matrix, and the principal component direction matrix, calculate the current principal component projection of the current sample data;

[0031] Based on the preset kernel function, the current self-kernel function value of the current sample data is determined, and the squared prediction error statistic of the current sample data is determined according to the current self-kernel function value and the current principal component projection.

[0032] In one embodiment, the contribution rate of each element in the current sample data to the squared prediction error statistic is calculated using the following formula:

[0033] ;

[0034] in, This represents the contribution rate of the i-th element to the squared prediction error statistic. This represents the value of the i-th element. This represents the j-th sample data in the historical sample data. This represents the current sample data. Represents the kernel function;

[0035] Accordingly, the faulty elements are identified, including:

[0036] The percentage contribution of each element is determined using the following formula:

[0037] ;

[0038] The element with the largest percentage contribution rate is identified as the faulty element.

[0039] In one embodiment, after locating the out-of-tolerance transformer in the transformer group based on the fault element, the method further includes:

[0040] An alarm message is issued based on the out-of-tolerance current transformer.

[0041] Secondly, this application also provides an over-tolerance monitoring device for a capacitive voltage transformer, comprising:

[0042] The historical data processing module is used to perform kernel principal component analysis based on historical sample data and determine the threshold of the squared prediction error statistic corresponding to the historical sample data. The historical sample data includes multiple sample data collected within a historical time period, and the sample data includes three-phase voltage data within the transformer group.

[0043] The real-time data processing module is used to calculate the squared prediction error statistic of the current sample data for the current sample data collected in real time.

[0044] The out-of-tolerance monitoring module is used to determine that the current transformer group is operating normally when the square prediction error statistic is less than or equal to the square prediction error statistic threshold; when the square prediction error statistic is greater than the square prediction error statistic threshold, it calculates the contribution rate of each element in the current sample data to the square prediction error statistic and identifies the faulty element; and locates the out-of-tolerance current transformer in the current transformer group based on the faulty element.

[0045] Thirdly, this application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in the first aspect above.

[0046] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect above.

[0047] The aforementioned method, device, electronic equipment, and computer-readable storage medium for monitoring the out-of-tolerance of capacitive voltage transformers (CVTs) utilizes kernel principal component analysis (KPCA) based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data. The historical sample data includes multiple sample data points collected within a historical time period, encompassing three-phase voltage data within the transformer group. For the currently collected sample data, the squared prediction error statistic is calculated. If the squared prediction error statistic is less than or equal to the threshold, the transformer group is determined to be operating normally. If the squared prediction error statistic exceeds the threshold, the contribution rate of each element in the current sample data to the squared prediction error statistic is calculated to identify the fault element. Based on the fault element, the out-of-tolerance transformer in the transformer group is located. Through this method, kernel principal component analysis (KPCA) is used to mine the three-phase voltage data characteristics of CVT operation, effectively handling the nonlinearity of the data, overcoming the limitations of traditional linear methods, extracting data features more accurately, improving the accuracy of out-of-tolerance monitoring, and providing a reliable basis for timely detection of CVT out-of-tolerance situations. By calculating the contribution rate of each element, faulty elements are separated, and out-of-tolerance transformers in the transformer group are located. This provides maintenance personnel with precise troubleshooting directions, improves maintenance efficiency, reduces power outage time and economic losses, and ensures the continuity and stability of power supply. It also enables online real-time monitoring of CVT errors, continuously tracking the CVT's operating status, and promptly capturing error changes, thereby enhancing the stability and reliability of the power system. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart illustrating an out-of-tolerance monitoring method for a capacitive voltage transformer in one embodiment.

[0050] Figure 2This is a flowchart illustrating the out-of-tolerance monitoring method for a capacitive voltage transformer in another embodiment;

[0051] Figure 3 This is a structural block diagram of an over-tolerance monitoring device for a capacitive voltage transformer in one embodiment;

[0052] Figure 4 This is a diagram of the internal structure of an electronic device in one embodiment. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] In one exemplary embodiment, such as Figure 1 As shown, a method for monitoring the out-of-tolerance of a capacitive voltage transformer is provided. This embodiment illustrates the application of this method to electronic equipment, and the method includes the following steps:

[0055] Step 102: Perform kernel principal component analysis based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data; the historical sample data includes multiple sample data collected within a historical time period, and the sample data includes three-phase voltage data within the transformer group.

[0056] Among them, the three-phase voltage data within the transformer group is used as the sample data for a single acquisition. The electronic equipment using the out-of-tolerance monitoring method for capacitive voltage transformers provided in this embodiment collects multiple sample data within a historical time period through voltage sensors installed at each capacitive voltage transformer (CVT) to form historical sample data.

[0057] A kernel matrix is ​​constructed and centered on historical sample data. Steady-state conditions are established through KPCA eigenvalue decomposition and component decomposition calculations, and the threshold of the squared prediction error (SPE) statistic for historical sample data is obtained. It is understood that traditional principal component analysis (PCA) can introduce biases when applied to nonlinear processes. In this embodiment, KPCA is used with a kernel trick to map the data from the original space to a higher-dimensional space, and then PCA is performed in the higher-dimensional space, thereby solving the dimensionality reduction problem for nonlinear data.

[0058] Optionally, the squared prediction error statistic threshold is configured as a parameter of the monitoring model. Electronic devices collect sample data in real time. If the power system is detected to be in a steady state, the monitoring model is continuously updated based on the real-time collected CVT operating data. Steady state refers to a state in which parameters such as voltage, current, and power remain relatively stable during system operation.

[0059] Step 104: Calculate the squared prediction error statistic for the current sample data collected in real time.

[0060] Step 106: If the squared prediction error statistic is less than or equal to the squared prediction error statistic threshold, determine that the current transformer group is operating normally.

[0061] The monitoring model compares the real-time SPE statistic with the SPE statistic threshold. If the SPE statistic does not exceed the SPE statistic threshold, the current transformer group is determined to be operating normally.

[0062] Step 108: If the squared prediction error statistic is greater than the squared prediction error statistic threshold, calculate the contribution rate of each element in the current sample data to the squared prediction error statistic and determine the faulty element.

[0063] Step 110: Locate the out-of-tolerance transformer in the transformer group based on the fault element.

[0064] If the SPE statistic exceeds the SPE statistic threshold, the operating status of the current transformer group is determined to be abnormal. Further source tracing is achieved through a fault separation strategy. Based on the contribution rate of each element in the current sample data to the squared prediction error statistic, the faulty element is identified, and the out-of-tolerance current transformer in the current transformer group is located.

[0065] Understandably, in calculating the SPE statistic threshold of historical sample data, accurate feature extraction makes the monitoring model's judgment of CVT out-of-tolerance conditions more reliable. In complex power environments, it can accurately identify minute error changes, significantly improving the accuracy of out-of-tolerance monitoring compared to traditional methods, reducing false positives and false negatives, and providing stronger protection for the stable operation of the power system.

[0066] The aforementioned method for monitoring the out-of-tolerance of capacitive voltage transformers utilizes kernel principal component analysis to mine the voltage data characteristics of three-phase voltage CVT operation within the transformer group. This effectively addresses the nonlinearity of the data, overcomes the limitations of traditional linear methods, and extracts data features more accurately, improving the accuracy of out-of-tolerance monitoring and providing a reliable basis for timely detection of CVT out-of-tolerance situations. By calculating the contribution rate of elements, fault elements are separated, and out-of-tolerance transformers in the transformer group are located, providing maintenance personnel with precise fault diagnosis directions, improving maintenance efficiency, reducing power outage time and economic losses, and ensuring the continuity and stability of power supply. Furthermore, it enables online real-time monitoring of CVT errors, continuously tracking the CVT's operating status, and promptly capturing error changes, thereby enhancing the stability and reliability of the power system.

[0067] In an exemplary embodiment, step 102 includes: constructing a kernel matrix based on a preset kernel function and historical sample data; centering the kernel matrix and performing eigenvalue decomposition on the centered kernel matrix to determine multiple eigenvalues ​​and multiple eigenvectors; selecting k eigenvalues ​​from the multiple eigenvalues ​​to form a variance contribution diagonal matrix and selecting k eigenvectors from the multiple eigenvectors to form a principal component direction matrix; for the i-th sample data, calculating the principal component projection of the i-th sample data based on the row vector of the i-th row of the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the self-kernel function value of the i-th sample data based on the preset kernel function and determining the squared prediction error statistic of the i-th sample data based on the self-kernel function value and the principal component projection; and determining the threshold of the squared prediction error statistic corresponding to a preset confidence level through a kernel density estimation strategy based on the squared prediction error statistics corresponding to multiple sample data.

[0068] Assuming the historical sample data includes n sample data points, the kernel matrix is ​​constructed using the following formula:

[0069] ;

[0070] In the formula, Indicates the preset kernel function. This represents the i-th sample data. This represents the j-th sample data.

[0071] The preset kernel function can be the Gaussian radial basis function (RBF) kernel function, the p-order polynomial kernel function, the sigmoid kernel function, etc., and a specific kernel function can be selected according to actual needs. Taking the Gaussian kernel function as the preset kernel function as an example, it can handle complex nonlinear relationships, and its expression is as follows:

[0072] ;

[0073] In the formula, is a key parameter of the Gaussian kernel function.

[0074] The kernel matrix is ​​transformed using a centering method to eliminate mean shift in the feature space. The specific transformation method is as follows:

[0075] ;

[0076] In the formula, It is an n×n identity matrix.

[0077] The eigenvalues ​​of the centered kernel matrix are decomposed based on the following formula:

[0078] ;

[0079] In the formula, Represents eigenvalues. This represents the eigenvector.

[0080] Select k eigenvalues ​​to construct a diagonal matrix of variance contribution. And construct the principal component direction matrix from the k eigenvectors corresponding to the k eigenvalues. .

[0081] For the i-th sample data, its principal component projection is calculated using the following formula:

[0082] ;

[0083] In the formula, express The i-th row vector, This is used to standardize the eigenvalues ​​in the variance contribution diagonal matrix, ensuring that the principal component directions have unit variance. This represents the principal component direction matrix.

[0084] The SPE statistic for the i-th sample data is determined using the following formula:

[0085] ;

[0086] In the formula, This represents the value of the kernel function.

[0087] Based on the method provided in this embodiment, the SPE statistic corresponding to each of the n sample data can be calculated. A kernel density estimation strategy is then used to determine the SPE statistic threshold corresponding to a pre-set confidence level. The pre-set confidence level refers to the confidence level set in advance according to actual needs. This embodiment does not limit the specific value; taking a pre-set confidence level of 95% as an example, the kernel density estimation method is used to fit the SPE statistic of the n sample data to obtain the probability density function f(x) of the SPE. Integrating the probability density function f(x) yields the cumulative distribution function F(x). Searching for the x value that makes F(x) = 0.95 yields the SPE statistic threshold corresponding to a 95% confidence level. This method is also effective for data where the original measurement points do not follow a normal distribution.

[0088] In an exemplary embodiment, the method further includes: calculating a mapping matrix based on the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the mapping features of the i-th sample data based on the mapping matrix, and calculating the residual vector of the i-th sample data based on the mapping features, the mapping matrix, and the principal component projection; and calculating the squared prediction error statistic of the i-th sample data based on the residual vector of the i-th sample data.

[0089] It is understandable that step 102 includes: constructing a kernel matrix based on a preset kernel function and historical sample data; centering the kernel matrix and performing eigenvalue decomposition on the centered kernel matrix to determine multiple eigenvalues ​​and multiple eigenvectors; selecting k eigenvalues ​​from the multiple eigenvalues ​​to form a variance contribution diagonal matrix and selecting k eigenvectors from the multiple eigenvectors to form a principal component direction matrix; calculating a mapping matrix based on the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; for the i-th sample data, calculating the principal component projection of the i-th sample data based on the i-th row vector of the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the mapping features of the i-th sample data based on the mapping matrix, and calculating the residual vector of the i-th sample data based on the mapping features, the mapping matrix, and the principal component projection; calculating the squared prediction error statistic of the i-th sample data based on the residual vector of the i-th sample data; and determining the threshold of the squared prediction error statistic corresponding to the preset confidence level through a kernel density estimation strategy based on the squared prediction error statistics corresponding to multiple sample data.

[0090] In the specific implementation, assuming that the historical sample data includes n sample data points, the kernel matrix is ​​constructed using the following formula:

[0091] ;

[0092] In the formula, Indicates the preset kernel function. This represents the i-th sample data. This represents the j-th sample data.

[0093] The preset kernel function can be the Gaussian radial basis function (RBF) kernel function, the p-order polynomial kernel function, the sigmoid kernel function, etc., and a specific kernel function can be selected according to actual needs. Taking the Gaussian kernel function as the preset kernel function as an example, it can handle complex nonlinear relationships, and its expression is as follows:

[0094] ;

[0095] In the formula, is a key parameter of the Gaussian kernel function.

[0096] The kernel matrix is ​​transformed using a centering method to eliminate mean shift in the feature space. The specific transformation method is as follows:

[0097] ;

[0098] In the formula, It is an n×n identity matrix.

[0099] The eigenvalues ​​of the centered kernel matrix are decomposed based on the following formula:

[0100] ;

[0101] In the formula, Represents eigenvalues. This represents the eigenvector.

[0102] Select k eigenvalues ​​to construct a diagonal matrix of variance contribution. And construct the principal component direction matrix from the k eigenvectors corresponding to the k eigenvalues. .

[0103] The mapping matrix is ​​calculated using the following formula:

[0104] ;

[0105] In the formula, This represents the mapping matrix of the first k principal components. This is used to standardize the eigenvalues ​​in the variance contribution diagonal matrix, ensuring that the principal component directions have unit variance. This represents the principal component direction matrix.

[0106] For the i-th sample data, its principal component projection is calculated using the following formula:

[0107] ;

[0108] In the formula, Indicates the projection of principal components. express The i-th row vector.

[0109] Reconstruct the original feature space and compute the residual vector of the i-th sample data:

[0110] ;

[0111] In the formula, This represents the residual vector.

[0112] The SPE statistic for the i-th sample data is calculated using the sum of squares of the residual vectors.

[0113] ;

[0114] Based on the method provided in this embodiment, the SPE statistic corresponding to each of the n sample data can be calculated, and the SPE statistic threshold corresponding to the preset confidence level can be determined through a kernel density estimation strategy. This method is also effective for data where the original measurement points do not follow a normal distribution.

[0115] In an exemplary embodiment, selecting k eigenvalues ​​from multiple eigenvalues ​​to form a variance contribution diagonal matrix, and selecting k eigenvectors from multiple eigenvectors to form a principal component direction matrix, includes: selecting the top k largest eigenvalues ​​from multiple eigenvalues ​​such that the cumulative contribution rate reaches a set threshold, and forming a variance contribution diagonal matrix based on the selected k eigenvalues; selecting the eigenvectors corresponding to the top k largest eigenvalues ​​from multiple eigenvectors, and forming a principal component direction matrix based on the selected k eigenvectors.

[0116] The cumulative contribution rate strategy involves selecting the k largest eigenvalues ​​to achieve a set threshold cumulative contribution rate, and then constructing a variance contribution diagonal matrix based on these k eigenvalues. And construct the principal component direction matrix from the corresponding k eigenvectors. The cumulative contribution rate refers to the proportion of the sum of the top k largest eigenvalues ​​to the total sum of all eigenvalues. The preset threshold is a parameter set in advance, and it is generally selected as any value between 85% and 95%.

[0117] In an exemplary embodiment, calculating the squared prediction error statistic of the current sample data includes: constructing a current kernel matrix of the current sample data and historical sample data based on a preset kernel function; centering the current kernel matrix; calculating the current principal component projection of the current sample data based on the centered current kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the current self-kernel function value of the current sample data based on the preset kernel function, and determining the squared prediction error statistic of the current sample data based on the current self-kernel function value and the current principal component projection.

[0118] Understandably, the statistics used in the PCA process include the T2 statistic and the Q statistic. The T2 statistic is more sensitive to errors in the principal component space and can reflect the systematic changes in the measurement process. The Q statistic is the total change in the residual space, also known as the squared prediction error (SPE), which reflects the changes related to measurement noise.

[0119] Wherein, it is assumed that the current sample data is The kernel matrix of the current sample data is calculated using the same kernel function as the historical sample data. The elements in the matrix are:

[0120] ;

[0121] The current kernel matrix is ​​transformed using a centering method to eliminate mean shift in the feature space. The specific transformation method is as follows:

[0122] ;

[0123] Map the current sample data onto the principal component directions in the high-dimensional feature space, and calculate the current principal component projection:

[0124] ;

[0125] Simplify the calculation of the SPE statistic for the current sample data using kernel functions:

[0126] ;

[0127] Optionally, in addition to simplifying the calculation of the SPE statistic based on the kernel function, the current mapping matrix can be calculated based on the current kernel matrix after centering, the variance contribution diagonal matrix, and the principal component direction matrix; the current mapping feature of the current sample data can be determined based on the current mapping matrix; and the current residual vector of the current sample data can be calculated based on the current mapping feature, the current mapping matrix, and the current principal component projection; and the SPE statistic of the current sample data can be calculated based on the current residual vector of the current sample data.

[0128] In an exemplary embodiment, the contribution rate of each element in the current sample data to the squared prediction error statistic is calculated using the following formula:

[0129] ;

[0130] in, This represents the contribution rate of the i-th element to the squared prediction error statistic. This represents the value of the i-th element. This represents the j-th sample data in the historical sample data. This represents the current sample data. Represents the kernel function;

[0131] Accordingly, the faulty elements are identified, including:

[0132] The percentage contribution of each element is determined using the following formula:

[0133] ;

[0134] The element with the largest percentage contribution rate is identified as the faulty element.

[0135] Understandably, since the input variables (i.e., the elements in the current sample data) and the kernel principal components cannot be explicitly represented, the contribution rate of the input variables to the SPE statistic can be solved by the partial derivative of the kernel function. Assume the current sample data is V=[v1,…,v…]. m ] T ∈R m×1 The single element is v. i(i=1,…,m), with the selected kernel function being the Gaussian kernel function, the contribution rate of each variable in the input space to the statistics T2 and SPE is derived (where i represents the i-th element in the current sample data and j represents the j-th measurement in the historical sample data):

[0136] (1);

[0137] (2);

[0138] In the formula, Let tr(*) represent the contribution rate of the i-th element in the current sample data to the T2 statistic, tr(*) denotes the trace of matrix *, and α is the standardized eigenvector. This is the current kernel matrix after centralization, where any element is... (p, q = 1, ..., n) It is a diagonal matrix; The contribution rate of the i-th element in the current sample data to the SPE statistic. The parameters of the Gaussian kernel function are... (j=1,…,n) represents historical sample data. For the current sample data, for The SPE value.

[0139] Specifically, for Finding the partial derivative can be decomposed into finding the partial derivative of each element, that is:

[0140] ;

[0141] Similarly, for Finding partial derivatives can also be decomposed into finding partial derivatives with respect to each element, that is:

[0142] ;

[0143] In the formula:

[0144] ;

[0145] ;

[0146] ;

[0147] The derivation results show that the calculation process for obtaining the contribution rate of each variable is extremely complex, with the computational burden mainly stemming from the matrix. Take the partial derivative with respect to the variable. Assume the matrix. for:

[0148] ;

[0149] Equation (1) represents finding the matrix. The trace, that is, the contribution rate of the i-th element in the input space to the change in the principal component space, is a matrix. The trace. Let's break down equation (2) and derive the second term as follows:

[0150] ;

[0151] because If it is a diagonal matrix, then express The linear transformation, and because Let represent the contribution rate of the i-th element in the input space to the changes in the principal component space. That is, the second term in equation (2) also represents the influence of the i-th element in the input space on the change of the principal component space. Therefore, the first term in equation (2) after removing this term can be expressed as the influence of the i-th element in the input space on the change of the residual space. Based on the above derivation, the contribution rate of each element to the residual space can be defined as:

[0152] ;

[0153] In practical applications, normalization is used to calculate the specific contribution percentage of each element in the current sample data:

[0154] ;

[0155] The element with the largest percentage contribution is identified as the fault element. In this way, the contribution rate of the input variables (i.e., each element in the current sample data) to the SPE statistic is calculated using the partial derivative of the kernel function. When multiple transformers fail simultaneously, the influence of each variable on the fault can be clearly identified, thus accurately separating the fault variable and quickly locating the faulty transformer. This feature provides maintenance personnel with precise fault diagnosis directions, significantly shortening fault diagnosis time, improving maintenance efficiency, reducing power outage time and economic losses caused by power system faults, and ensuring the continuity and stability of power supply.

[0156] In one exemplary embodiment, after step 110, the method further includes: issuing an alarm message based on the out-of-tolerance current transformer.

[0157] Understandably, in actual operation, timely detection of dynamic changes in CVT error is crucial. Once an error exceeds the tolerance limit, an alarm message should be promptly issued and feedback signal provided. This real-time monitoring function enables maintenance personnel to take immediate action, preventing the impact of CVT error exceeding tolerance on energy metering and the safe and stable operation of the power system. This is of great significance for ensuring the efficient operation of the power system, improving the accuracy of energy metering, and maintaining fair trading in the electricity market, and it meets the needs of modern power system development towards intelligence and efficiency.

[0158] In one alternative implementation, refer to Figure 2 The method for monitoring the out-of-tolerance of capacitive voltage transformers includes the following steps:

[0159] Using the three-phase voltage data within the same transformer group as a single acquisition sample, historical data is collected during the steady-state establishment of the system. KPCA feature decomposition is performed on the historical data to determine the threshold of the SPE statistic.

[0160] Collect real-time data and calculate the SPE statistic of the collected data through high-dimensional projection.

[0161] Compare the obtained SPE statistic with the SPE statistic threshold to determine whether it exceeds the tolerance. If it does not exceed the tolerance, it indicates that the current transformer group is operating normally.

[0162] If the three-phase voltage in the transformer group exceeds the tolerance, the contribution rate of the input variable to the SPE statistic is solved by the partial derivative of the kernel function to separate the fault variable, locate the transformer with the excess tolerance (i.e., the faulty transformer) in the transformer group and trigger an alarm.

[0163] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0164] Based on the same inventive concept, this application also provides an over-tolerance monitoring device for a capacitive voltage transformer used to implement the over-tolerance monitoring method for the capacitive voltage transformer described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the over-tolerance monitoring device for capacitive voltage transformers provided below can be found in the limitations of the over-tolerance monitoring method for capacitive voltage transformers described above, and will not be repeated here.

[0165] In one exemplary embodiment, such as Figure 3 As shown, an over-tolerance monitoring device for a capacitive voltage transformer is provided, comprising:

[0166] The historical data processing module 302 is used to perform kernel principal component analysis based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data. The historical sample data includes multiple sample data collected within a historical time period, and the sample data includes three-phase voltage data within the transformer group.

[0167] The real-time data processing module 304 is used to calculate the squared prediction error statistic of the current sample data for the current sample data collected in real time.

[0168] The out-of-tolerance monitoring module 306 is used to determine that the operating status of the current transformer group is normal when the square prediction error statistic is less than or equal to the square prediction error statistic threshold; when the square prediction error statistic is greater than the square prediction error statistic threshold, it calculates the contribution rate of each element in the current sample data to the square prediction error statistic and determines the fault element; and locates the out-of-tolerance current transformer in the current transformer group based on the fault element.

[0169] The aforementioned over-tolerance monitoring device for capacitive voltage transformers utilizes kernel principal component analysis (KPCA) to mine the three-phase voltage data characteristics of CVT operation. This effectively addresses the nonlinearity of the data, overcomes the limitations of traditional linear methods, and extracts data features more accurately, improving the accuracy of over-tolerance monitoring and providing a reliable basis for timely detection of CVT over-tolerance situations. By calculating the contribution rate of elements, fault elements are separated, locating the over-tolerance transformers in the transformer group. This provides maintenance personnel with precise fault diagnosis directions, improving maintenance efficiency, reducing power outage time and economic losses, and ensuring the continuity and stability of power supply. Furthermore, it enables online real-time monitoring of CVT errors, continuously tracking the CVT's operating status, and promptly capturing error changes, thereby enhancing the stability and reliability of the power system.

[0170] In an exemplary embodiment, the historical data processing module 302 is further configured to: construct a kernel matrix based on a preset kernel function and historical sample data; center the kernel matrix and perform eigenvalue decomposition on the centered kernel matrix to determine multiple eigenvalues ​​and multiple eigenvectors; select k eigenvalues ​​from the multiple eigenvalues ​​to form a variance contribution diagonal matrix and select k eigenvectors from the multiple eigenvectors to form a principal component direction matrix; for the i-th sample data, calculate the principal component projection of the i-th sample data based on the row vector of the i-th row of the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determine the self-kernel function value of the i-th sample data based on the preset kernel function and determine the squared prediction error statistic of the i-th sample data based on the self-kernel function value and the principal component projection; and determine the squared prediction error statistic threshold corresponding to a preset confidence level through a kernel density estimation strategy based on the squared prediction error statistics corresponding to multiple sample data.

[0171] In an exemplary embodiment, the historical data processing module 302 is further configured to calculate a mapping matrix based on the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determine the mapping features of the i-th sample data based on the mapping matrix; calculate the residual vector of the i-th sample data based on the mapping features, the mapping matrix, and the principal component projection; and calculate the squared prediction error statistic of the i-th sample data based on the residual vector of the i-th sample data.

[0172] In an exemplary embodiment, the historical data processing module 302 is further configured to select the top k largest feature values ​​from multiple feature values, such that the cumulative contribution rate reaches a set threshold, and construct a variance contribution diagonal matrix based on the selected k feature values; and select the feature vectors corresponding to the top k largest feature values ​​from multiple feature vectors, and construct a principal component direction matrix based on the selected k feature vectors.

[0173] In an exemplary embodiment, the real-time data processing module 304 is further configured to: construct a current kernel matrix of the current sample data and historical sample data based on a preset kernel function; perform centering processing on the current kernel matrix; calculate the current principal component projection of the current sample data based on the centered current kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determine the current self-kernel function value of the current sample data based on the preset kernel function; and determine the squared prediction error statistic of the current sample data based on the current self-kernel function value and the current principal component projection.

[0174] In an exemplary embodiment, the out-of-tolerance monitoring module 306 is further configured to calculate the contribution rate of each element in the current sample data to the squared prediction error statistic using the following formula:

[0175] ;

[0176] in, This represents the contribution rate of the i-th element to the squared prediction error statistic. This represents the value of the i-th element. This represents the j-th sample data in the historical sample data. This represents the current sample data. Represents the kernel function;

[0177] The percentage contribution of each element is determined using the following formula:

[0178] ;

[0179] The element with the largest percentage contribution rate is identified as the faulty element.

[0180] In one exemplary embodiment, the out-of-tolerance monitoring module 306 is also configured to issue an alarm message based on the out-of-tolerance current transformer.

[0181] The various modules in the aforementioned over-tolerance monitoring device for capacitive voltage transformers can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the electronic device in hardware form or independent of it, or stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of each module.

[0182] In one exemplary embodiment, an electronic device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, this electronic device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational 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 the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for monitoring the over-tolerance of a capacitive voltage transformer.

[0183] Those skilled in the art will understand that Figure 4The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0184] In an exemplary embodiment, an electronic device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps: performing kernel principal component analysis based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data; the historical sample data includes multiple sample data collected within a historical time period, and the sample data includes three-phase voltage data within the transformer group; for the currently collected sample data, calculating the squared prediction error statistic of the current sample data; if the squared prediction error statistic is less than or equal to the squared prediction error statistic threshold, determining that the operating status of the transformer group is normal; if the squared prediction error statistic is greater than the squared prediction error statistic threshold, calculating the contribution rate of each element in the current sample data to the squared prediction error statistic, and determining the faulty element; locating the out-of-tolerance transformer in the transformer group based on the faulty element.

[0185] In one embodiment, when the processor executes the computer program, it further performs the following steps: constructing a kernel matrix based on a preset kernel function and historical sample data; centering the kernel matrix and performing eigenvalue decomposition on the centered kernel matrix to determine multiple eigenvalues ​​and multiple eigenvectors; selecting k eigenvalues ​​from the multiple eigenvalues ​​to form a variance contribution diagonal matrix and selecting k eigenvectors from the multiple eigenvectors to form a principal component direction matrix; for the i-th sample data, calculating the principal component projection of the i-th sample data based on the row vector of the i-th row of the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the self-kernel function value of the i-th sample data based on the preset kernel function, and determining the squared prediction error statistic of the i-th sample data based on the self-kernel function value and the principal component projection; and determining the threshold of the squared prediction error statistic corresponding to a preset confidence level through a kernel density estimation strategy based on the squared prediction error statistics corresponding to multiple sample data.

[0186] In one embodiment, when the processor executes the computer program, it further performs the following steps: calculating a mapping matrix based on the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the mapping features of the i-th sample data based on the mapping matrix, and calculating the residual vector of the i-th sample data based on the mapping features, the mapping matrix, and the principal component projection; and calculating the squared prediction error statistic of the i-th sample data based on the residual vector of the i-th sample data.

[0187] In one embodiment, when the processor executes the computer program, it further performs the following steps: selecting the top k largest eigenvalues ​​from multiple eigenvalues ​​such that the cumulative contribution rate reaches a set threshold, and constructing a variance contribution diagonal matrix based on the selected k eigenvalues; selecting the eigenvectors corresponding to the top k largest eigenvalues ​​from multiple eigenvectors, and constructing a principal component direction matrix based on the selected k eigenvectors.

[0188] In one embodiment, when the processor executes the computer program, it further performs the following steps: constructing a current kernel matrix of the current sample data and historical sample data based on a preset kernel function; centering the current kernel matrix; calculating the current principal component projection of the current sample data based on the centered current kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the current self-kernel function value of the current sample data based on the preset kernel function, and determining the squared prediction error statistic of the current sample data based on the current self-kernel function value and the current principal component projection.

[0189] In one embodiment, when the processor executes the computer program, it further performs the following steps: calculating the contribution rate of each element in the current sample data to the squared prediction error statistic using the following formula:

[0190] ;

[0191] in, This represents the contribution rate of the i-th element to the squared prediction error statistic. This represents the value of the i-th element. This represents the j-th sample data in the historical sample data. This represents the current sample data. Represents the kernel function;

[0192] Accordingly, the faulty elements are identified, including determining the percentage contribution of each element using the following formula:

[0193] ;

[0194] The element with the largest percentage contribution rate is identified as the faulty element.

[0195] In one embodiment, the processor, when executing the computer program, also performs the following steps: issuing an alarm message based on the out-of-tolerance current transformer.

[0196] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it performs the following steps: performing kernel principal component analysis based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data; the historical sample data includes multiple sample data collected within a historical time period, and the sample data includes three-phase voltage data within the transformer group; for the currently collected sample data, calculating the squared prediction error statistic of the current sample data; if the squared prediction error statistic is less than or equal to the squared prediction error statistic threshold, determining that the operating status of the transformer group is normal; if the squared prediction error statistic is greater than the squared prediction error statistic threshold, calculating the contribution rate of each element in the current sample data to the squared prediction error statistic, and determining the fault element; locating the out-of-tolerance transformer in the transformer group based on the fault element.

[0197] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: constructing a kernel matrix based on a preset kernel function and historical sample data; centering the kernel matrix and performing eigenvalue decomposition on the centered kernel matrix to determine multiple eigenvalues ​​and multiple eigenvectors; selecting k eigenvalues ​​from the multiple eigenvalues ​​to form a variance contribution diagonal matrix and selecting k eigenvectors from the multiple eigenvectors to form a principal component direction matrix; for the i-th sample data, calculating the principal component projection of the i-th sample data based on the row vector of the i-th row of the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the self-kernel function value of the i-th sample data based on the preset kernel function, and determining the squared prediction error statistic of the i-th sample data based on the self-kernel function value and the principal component projection; and determining the threshold of the squared prediction error statistic corresponding to a preset confidence level through a kernel density estimation strategy based on the squared prediction error statistics corresponding to multiple sample data.

[0198] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: calculating a mapping matrix based on the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the mapping features of the i-th sample data based on the mapping matrix, and calculating the residual vector of the i-th sample data based on the mapping features, the mapping matrix, and the principal component projection; and calculating the squared prediction error statistic of the i-th sample data based on the residual vector of the i-th sample data.

[0199] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: selecting the top k largest eigenvalues ​​from multiple eigenvalues ​​such that the cumulative contribution rate reaches a set threshold, and constructing a variance contribution diagonal matrix based on the selected k eigenvalues; selecting the eigenvectors corresponding to the top k largest eigenvalues ​​from multiple eigenvectors, and constructing a principal component direction matrix based on the selected k eigenvectors.

[0200] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: constructing a current kernel matrix of the current sample data and historical sample data based on a preset kernel function; centering the current kernel matrix; calculating the current principal component projection of the current sample data based on the centered current kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determining the current self-kernel function value of the current sample data based on the preset kernel function, and determining the squared prediction error statistic of the current sample data based on the current self-kernel function value and the current principal component projection.

[0201] In one embodiment, when the computer program is executed by a processor, it further performs the following steps: calculating the contribution rate of each element in the current sample data to the squared prediction error statistic using the following formula:

[0202] ;

[0203] in, This represents the contribution rate of the i-th element to the squared prediction error statistic. This represents the value of the i-th element. This represents the j-th sample data in the historical sample data. This represents the current sample data. Represents the kernel function;

[0204] Accordingly, the faulty elements are identified, including determining the percentage contribution of each element using the following formula:

[0205] ;

[0206] The element with the largest percentage contribution rate is identified as the faulty element.

[0207] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: issuing an alarm message based on the out-of-tolerance current transformer.

[0208] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0209] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. The memory, database, or other media mentioned in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0210] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0211] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring the out-of-tolerance of a capacitive voltage transformer, characterized in that, The method includes: Kernel principal component analysis is performed based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data; the historical sample data includes multiple sample data collected during the historical time period when the system is in steady state, with the three-phase voltage data within the same transformer group as a single collection sample. For the current sample data collected in real time, calculate the squared prediction error statistic of the current sample data; If the squared prediction error statistic is less than or equal to the squared prediction error statistic threshold, the current transformer group is determined to be operating normally. If the squared prediction error statistic is greater than the squared prediction error statistic threshold, the contribution rate of each element in the current sample data to the squared prediction error statistic is calculated by the partial derivative of the kernel function to determine the faulty element. Based on the fault element, locate the out-of-tolerance current transformer in the current transformer group; The step of performing kernel principal component analysis based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data includes: Based on a preset kernel function, a kernel matrix is ​​constructed using historical sample data; The kernel matrix is ​​centered, and the centered kernel matrix is ​​decomposed to determine multiple eigenvalues ​​and multiple eigenvectors. Select k eigenvalues ​​from the plurality of eigenvalues ​​to form a variance contribution diagonal matrix, and select k eigenvectors from the plurality of eigenvectors to form a principal component direction matrix; For the i-th sample data, the principal component projection of the i-th sample data is calculated based on the row vector of the i-th row of the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix. Based on the preset kernel function, the self-kernel function value of the i-th sample data is determined, and the squared prediction error statistic of the i-th sample data is determined according to the self-kernel function value and the principal component projection; Based on the squared prediction error statistics corresponding to the multiple sample data, the threshold of the squared prediction error statistics corresponding to the preset confidence level is determined by the kernel density estimation strategy. Alternatively, the step of performing kernel principal component analysis based on historical sample data to determine the threshold of the squared prediction error statistic corresponding to the historical sample data includes: Based on a preset kernel function, a kernel matrix is ​​constructed using historical sample data; The kernel matrix is ​​centered, and the centered kernel matrix is ​​decomposed to determine multiple eigenvalues ​​and multiple eigenvectors. Select k eigenvalues ​​from the plurality of eigenvalues ​​to form a variance contribution diagonal matrix, and select k eigenvectors from the plurality of eigenvectors to form a principal component direction matrix; The mapping matrix is ​​calculated based on the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix. For the i-th sample data, the principal component projection of the i-th sample data is calculated based on the row vector of the i-th row of the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix. The mapping features of the i-th sample data are determined based on the mapping matrix, and the residual vector of the i-th sample data is calculated based on the mapping features, the mapping matrix, and the principal component projection. Calculate the squared prediction error statistic of the i-th sample data based on the residual vector of the i-th sample data; Based on the squared prediction error statistics corresponding to the multiple sample data, the threshold of the squared prediction error statistics corresponding to the preset confidence level is determined by the kernel density estimation strategy.

2. The method according to claim 1, characterized in that, The step of selecting k eigenvalues ​​from the plurality of eigenvalues ​​to form a variance contribution diagonal matrix, and selecting k eigenvectors from the plurality of eigenvectors to form a principal component direction matrix, includes: Select the k largest eigenvalues ​​from the plurality of eigenvalues, such that the cumulative contribution rate reaches a set threshold, and construct a variance contribution diagonal matrix based on the selected k eigenvalues; Select the k largest eigenvectors from the plurality of eigenvectors, and construct the principal component direction matrix based on the selected k eigenvectors.

3. The method according to claim 1, characterized in that, The calculation of the squared prediction error statistic for the current sample data includes: Based on the preset kernel function, construct the current kernel matrix of the current sample data and the historical sample data; The current kernel matrix is ​​centered. Based on the current kernel matrix after centering, the variance contribution diagonal matrix, and the principal component direction matrix, calculate the current principal component projection of the current sample data; Based on the preset kernel function, the current self-kernel function value of the current sample data is determined, and the squared prediction error statistic of the current sample data is determined according to the current self-kernel function value and the current principal component projection.

4. The method according to any one of claims 1 to 3, characterized in that, The contribution rate of each element in the current sample data to the squared prediction error statistic is calculated using the following formula: ； in, This represents the contribution rate of the i-th element to the squared prediction error statistic. This represents the value of the i-th element. This represents the j-th sample data in the historical sample data. This represents the current sample data. Represents the kernel function; Accordingly, the determination of the fault element includes: The percentage contribution of each element is determined using the following formula: ； The element with the largest percentage contribution rate is identified as the faulty element.

5. The method according to any one of claims 1 to 3, characterized in that, After locating the out-of-tolerance transformer in the transformer group based on the fault element, the method further includes: An alarm message is issued based on the out-of-tolerance current transformer.

6. A device for monitoring the out-of-tolerance of a capacitive voltage transformer, characterized in that, The device includes: The historical data processing module is used to perform kernel principal component analysis based on historical sample data, determine the squared prediction error statistic corresponding to the historical sample data, and determine the threshold of the squared prediction error statistic corresponding to the preset confidence level through a kernel density estimation strategy. The historical sample data includes multiple sample data collected within a historical time period when the system is in steady state, with the three-phase voltage data within the same transformer group as a single collection sample. The real-time data processing module is used to calculate the squared prediction error statistic of the current sample data collected in real time. The out-of-tolerance monitoring module is used to determine that the current transformer group is operating normally when the square prediction error statistic is less than or equal to the square prediction error statistic threshold; and to determine the faulty element by calculating the contribution rate of each element in the current sample data to the square prediction error statistic through the partial derivative of the kernel function when the square prediction error statistic is greater than the square prediction error statistic threshold; and to locate the out-of-tolerance current transformer in the current transformer group based on the faulty element. The historical data processing module is further configured to: construct a kernel matrix based on a preset kernel function and historical sample data; center the kernel matrix and perform eigenvalue decomposition on the centered kernel matrix to determine multiple eigenvalues ​​and multiple eigenvectors; select k eigenvalues ​​from the multiple eigenvalues ​​to form a variance contribution diagonal matrix and select k eigenvectors from the multiple eigenvectors to form a principal component direction matrix; for the i-th sample data, calculate the principal component projection of the i-th sample data based on the row vector of the i-th row of the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix; determine the self-kernel function value of the i-th sample data based on the preset kernel function, and determine the squared prediction error statistic of the i-th sample data based on the self-kernel function value and the principal component projection; determine the squared prediction error statistic threshold corresponding to a preset confidence level through a kernel density estimation strategy based on the squared prediction error statistics corresponding to the multiple sample data; or... Based on a preset kernel function, a kernel matrix is ​​constructed according to historical sample data. The kernel matrix is ​​then centered, and eigenvalues ​​and eigenvectors are decomposed to determine multiple eigenvalues ​​and eigenvectors. k eigenvalues ​​are selected from the eigenvalues ​​to form a variance contribution diagonal matrix, and k eigenvectors are selected from the eigenvectors to form a principal component direction matrix. A mapping matrix is ​​calculated based on the centered kernel matrix, the variance contribution diagonal matrix, and the principal component direction matrix. For the i-th sample data, the mapping matrix is ​​calculated based on the i-th row vector of the centered kernel matrix and the... Using the variance contribution diagonal matrix and the principal component direction matrix, calculate the principal component projection of the i-th sample data; determine the mapping feature of the i-th sample data based on the mapping matrix, and calculate the residual vector of the i-th sample data based on the mapping feature, the mapping matrix, and the principal component projection; calculate the squared prediction error statistic of the i-th sample data based on the residual vector of the i-th sample data; and determine the squared prediction error statistic threshold corresponding to the preset confidence level using a kernel density estimation strategy based on the squared prediction error statistics corresponding to the multiple sample data.

7. The apparatus according to claim 6, characterized in that, The historical data processing module is further configured to select the k largest feature values ​​from the plurality of feature values, such that the cumulative contribution rate reaches a set threshold, and construct a variance contribution diagonal matrix based on the selected k feature values; and select the feature vectors corresponding to the k largest feature values ​​from the plurality of feature vectors, and construct a principal component direction matrix based on the selected k feature vectors.

8. The apparatus according to claim 6, characterized in that, The historical data processing module is further configured to construct a current kernel matrix of the current sample data and the historical sample data based on the preset kernel function; and to perform centering processing on the current kernel matrix; Based on the current kernel matrix after centering, the variance contribution diagonal matrix, and the principal component direction matrix, calculate the current principal component projection of the current sample data; Based on the preset kernel function, the current self-kernel function value of the current sample data is determined, and the squared prediction error statistic of the current sample data is determined according to the current self-kernel function value and the current principal component projection.

9. An electronic 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 method according to any one of claims 1 to 5.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images