Low-voltage distribution network user-transformer relation checking method and related device

Abstract

The invention provides a low-voltage distribution network user-transformer relation checking method, which comprises the following steps of: sampling daily voltage data of users, forming a matrix by voltage sequence data of each user to be identified, performing dimensionality reduction by applying kernel principal component analysis, clustering feature samples of each user by utilizing a Gaussian hybrid clustering algorithm, and checking the user-transformer relation of the low-voltage distribution network. And a clustering result is compared with a system file, abnormal users are screened out, and verification of the transformer area user-transformer relation is realized. Through organic combination of kernel principal component analysis and Gaussian hybrid clustering, complex voltage data in the low-voltage distribution network can be better processed, verification of the transformer area user-transformer relation is effectively carried out, high practicability and advantages are achieved, the operation precision and efficiency of the distribution network can be improved, and the possibility of misrecognition is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems, and particularly relates to a method for verifying the household-transformer relationship in a low-voltage distribution network. Background Art

[0002] In a low-voltage distribution network, the correctness of the household-transformer relationship directly affects the statistics of the line loss in the substation area. Incorrect records of the household-transformer relationship will cause an increase in the line loss error, seriously affecting the calculation of the three-phase balance degree in the substation area. With the development of the power industry and the strong promotion of the urbanization process, the number of users in the low-voltage distribution network has increased rapidly. The household-transformer relationships of the users in the substation area change frequently, and the phase where they are located is also updated from time to time. However, due to possible errors in the investigation of the file information and the low efficiency of information update, it seriously affects the refined management of the distribution network. Therefore, the identification of the household-transformer relationship has become an urgent problem to be solved.

[0003] Traditionally, the method of manually inspecting the lines is generally used to verify the household-transformer relationship in the low-voltage substation area. The method of pulling the switch to observe and depicting on the spot is used to correct or supplement the household-transformer relationship in the substation area. However, with the continuous expansion of the scale of the low-voltage distribution network, the increase in the number of users, and the increasing complexity of the distribution network structure, the method of manually inspecting the lines not only consumes a large amount of resources, but also may have a large number of errors. Moreover, the manual method has problems such as untimely information update and difficulty in digitizing information. The topological information of the substation area recorded in this way brings difficulties to the intelligent construction and refined management of the low-voltage distribution network. To solve the deficiencies of the traditional method for identifying the topological information of the substation area, there are currently two types of methods for identifying the topology of the substation area in research and practical applications, namely, those based on communication devices and injected signals and those based on the electricity consumption information of intelligent monitoring terminals.

[0004] The topology identification method based on communication devices uses carrier communication technology or pulse current technology in the topology identification process. A small current signal or a small power signal is injected into the power grid using a portable device or an intelligent terminal, and this signal is received everywhere in the power grid, so as to obtain the topological information of the substation area. The inventor has found through research that currently, in applications, a topology identifier for the substation area developed based on carrier communication technology or pulse current technology is mainly used to identify the household-transformer relationship and the phase attribution relationship between the substation area and the users. However, this method has many drawbacks. For example, the signal is easily attenuated during transmission, resulting in difficulty in receiving the signal. Moreover, problems such as common-ground crosstalk and parallel transmission crosstalk in carrier communication may cause interference to the signal transmission, resulting in incorrect results in topology identification. In the pulse current method, it is difficult to set the injected current signal. If the signal is too small, it is easily submerged, and if the signal is too large, it is likely to affect the normal operation of the power grid. In addition, the topology identification method based on communication devices requires the purchase of a handheld topology identifier or the installation of equipment in the power grid, increasing the hardware cost and resource consumption. Summary of the Invention

[0005] The present invention provides a method for identifying the household-transformer relationship in a low-voltage distribution network by using kernel principal component analysis and Gaussian mixture clustering algorithm.

[0006] The technical solution adopted by the present invention is as follows: A method for verifying the household-transformer relationship in a low-voltage distribution network includes:

[0007] Obtain user voltage data to construct a voltage value matrix X;

[0008] Perform standardization processing on the voltage value matrix X to obtain a standardized voltage value matrix

[0009] Use sparse kernel principal component analysis on the standardized voltage value matrix to perform data dimensionality reduction to obtain a dimensionality reduction matrix Y;

[0010] Use the Gaussian mixture clustering method to cluster the dimensionality reduction matrix Y and compare with the archives for verification.

[0011] The specific process of obtaining user voltage data to construct the voltage value matrix X is as follows:

[0012] Obtain the daily voltage data of users to form a voltage time series column vector Xi of user i i ;

[0013] For n users to be identified, form a voltage value matrix X by combining the voltage time series of each user, X = [X 1 , X 2 ,..., X n T .

[0014] The specific process of performing standardization processing on the voltage value matrix X to obtain a standardized voltage value matrix is as follows:

[0015] The standardization method is where is the mean of the data in the j-th column of the voltage value matrix X, is the standard deviation of the data in the j-th column of the voltage value matrix X; the standardized voltage value matrix is

[0016] The specific process of using sparse kernel principal component analysis on the standardized voltage value matrix to perform data dimensionality reduction to obtain a dimensionality reduction matrix Y is as follows:

[0017] Use the Gaussian kernel function to obtain the corresponding kernel matrix W; specifically: And set the standard deviation σ of the Gaussian kernel function;

[0018] Perform centering processing on the kernel matrix W to obtain a new matrix K, K = W - NW - WN - NWN; where N is an n×n matrix with all elements being​ matrix;

[0019] Perform eigenvalue decomposition on matrix W to obtain the corresponding eigen-column vectors P 1 , P 2 ,..., P n and eigenvalues λ 1 , λ 2 ,..., λ n ; Sort the eigenvalues in descending order, retain the first m eigenvalues after sorting and their corresponding eigenvectors; The selected eigenvectors are concatenated to form the projection matrix P = [P 1 , P 2 ,..., P m ; where is the floor function;

[0020] Introduce sparsity and restrict the projection matrix P through L1 regularization to reduce the number of non-zero elements in the projection vector; The goal of the sparsity constraint is to optimize the sparsity by minimizing the objective function, and the objective function is: where α = 0.01, ‖·‖ 2 is the square of the matrix L2 norm, ‖· 1 is the L1 norm; Then the sparse projection matrix can be obtained

[0021] The dimensionality reduction matrix Y = M * X; The i-th column element in Y represents the sample data of user i after dimensionality reduction.

[0022] Using the Gaussian mixture clustering method to cluster the dimensionality reduction matrix Y to achieve the verification of the household change relationship, specifically:

[0023] Set the number of Gaussian distributions to the total number of substations to be identified, denoted as k;

[0024] The parameters of each Gaussian distribution are denoted as θ η ={π η , μ η , Σ η}, η represents the η-th distribution; where π η is the mixing coefficient of the η-th Gaussian distribution and satisfies Initialized as μ η is the center point of each Gaussian distribution, initialized as μ η = Y 1,η ; The covariance matrix Σ η represents the shape and distribution range of each Gaussian distribution, initialized as the m-dimensional identity matrix, i.e., Σ η = I m ;

[0025] Calculate the sample Y corresponding to user ω ω The posterior probability belonging to the η-th Gaussian distribution, that is where χ(Y ω |μ η , Σ η ) is the probability density function of the Gaussian distribution,

[0026] Iteratively update the parameters:

[0027]

[0028] At the end of each iteration, calculate the absolute value of the difference between the log-likelihood functions before and after the parameter iteration, that is, calculate ΔlogL(θ) = |logL(θ new ) - logL(θ old )|; where the log-likelihood function is:

[0029]

[0030] When ΔlogL(θ) < 0.001 or the number of iterations reaches 1000, the algorithm terminates the iteration;

[0031] After the algorithm terminates the iteration, according to the posterior probability γ ω,η Allocate the sample Y ω to the Gaussian distribution with the maximum probability, that is

[0032] Compare the household change identification results of all users with the system file. If the identified substation transformer number of a certain user is inconsistent with the substation transformer number recorded in the system file, then this user is identified as an abnormal user, and the household-substation relationship verification is realized.

[0033] The present invention also provides a low-voltage distribution network household-substation relationship verification device, including:

[0034] A voltage value matrix construction module, configured to obtain user voltage data and construct a voltage value matrix X;

[0035] A normalization module, configured to perform normalization processing on the voltage value matrix X to obtain a normalized voltage value matrix

[0036] A dimensionality reduction module, configured to perform data dimensionality reduction on the normalized voltage value matrix to obtain a dimensionality reduction matrix Y;

[0037] A clustering verification module, configured to perform clustering on the dimensionality reduction matrix Y by using Gaussian mixture clustering and compare with the file for verification.

[0038] The voltage value matrix construction module is specifically used for:

[0039] Obtain the daily voltage data of the user to form the voltage time series column vector X of user i i ;

[0040] For n users to be identified, form the voltage value matrix X by combining the voltage time series of each user, X = [X 1 , X 2 ,..., X n T .

[0041] The normalization module is specifically used for:

[0042] The normalization method is where is the mean value of the data in the j-th column of the voltage value matrix X, is the standard deviation of the data in the j-th column of the voltage value matrix X; the normalized voltage value matrix is X~.

[0043] The dimensionality reduction module is specifically used for:

[0044] Use the Gaussian kernel function to obtain the corresponding kernel matrix W; specifically: And set the standard deviation σ of the Gaussian kernel function;

[0045] Center the kernel matrix W to obtain the new matrix K, K = W - NW - WN - NWN; where N is an n×n matrix with all elements being ;

[0046] Perform eigenvalue decomposition on the matrix W to obtain the corresponding eigen-column vectors P 1 , P 2 ,..., P n and eigenvalues λ 1 , λ 2 ,..., λ n ; Sort the eigenvalues in descending order of magnitude, retain the first m eigenvalues after sorting and their corresponding eigenvectors; the selected eigenvectors are concatenated to form the projection matrix P = [P 1 , P 2 ,..., P m ; where [·] is the floor function;

[0047] Introduce sparsity and restrict the projection matrix P through L1 regularization to reduce the number of non-zero elements in the projection vector; the goal of the sparsity constraint is to optimize the sparsity by minimizing the objective function, and the objective function is: where α = 0.01, ||·|| 2 is the square of the matrix L2 norm, ||·||1 is the L1 norm; then a sparse projection matrix can be obtained

[0048] The dimensionality reduction matrix Y = M * X; The i-th column element in Y represents the sample data of user i after dimensionality reduction.

[0049] The clustering verification module is specifically used for:[[]]

[0050] Set the number of Gaussian distributions to the total number of substations to be identified, denoted as k;

[0051] The parameters of each Gaussian distribution are denoted as θ η ={π η , μ η , Σ η}, η represents the η-th distribution; where π η is the mixing coefficient of the η-th Gaussian distribution and satisfies Initialized as μ η is the center point of each Gaussian distribution, initialized as μ η = Y 1,η ; The covariance matrix Σ η represents the shape and distribution range of each Gaussian distribution, initialized as an m-dimensional identity matrix, i.e., Σ η = I m ;

[0052] Calculate the posterior probability that the sample Y ω corresponding to user ω belongs to the η-th Gaussian distribution, i.e., where χ(Y ω |μ η , Σ η ) is the probability density function of the Gaussian distribution,

[0053] Iteratively update the parameters:

[0054]

[0055] At the end of each iteration, calculate the absolute value of the difference between the log-likelihood functions before and after parameter iteration, i.e., calculate ΔlogL(θ) = |logL(θ new ) - logL(θ old )|; where the log-likelihood function is:

[0056]

[0057] When ΔlogL(θ) < 0.001 or the number of iterations reaches 1000, the algorithm terminates the iteration;

[0058] After the algorithm terminates the iteration, according to the posterior probability γω,η Assign the sample Y ω to the Gaussian distribution with the highest probability, that is

[0059] Compare the household change identification results of all users with the system file. If the identified substation transformer number of a certain user given by the identification result is inconsistent with the substation transformer number of this user recorded in the system file, then this user is identified as an abnormal user, and the verification of the household-substation relationship is realized.

[0060] On the other hand, the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor executes the computer program, the method for verifying the household-substation relationship in the low-voltage distribution network as described in the first aspect is realized.

[0061] On the other hand, the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for verifying the household-substation relationship in the low-voltage distribution network as described in the first aspect is realized.

[0062] The present invention samples the daily voltage data of users, forms a matrix of the voltage sequence data of each user to be identified, performs dimensionality reduction by using kernel principal component analysis, then clusters the characteristic samples of each user by using the Gaussian mixture clustering algorithm, and compares the clustering results with the system file to screen out abnormal users, so as to realize the verification of the household-substation relationship in the substation area. For users with highly similar voltage curves but not belonging to the same substation area, the linearity of the voltage curves between users is relatively high. Therefore, the method of identifying the household-substation relationship based on similarity often performs poorly and cannot accurately identify; while kernel principal component analysis introduces the kernel trick to map the data into a high-dimensional feature space, so as to better capture the non-linear relationship between data; kernel principal component analysis can realize dimensionality reduction. By calculating the principal components in the high-dimensional feature space, the original data can be projected into a low-dimensional subspace, thereby realizing the dimensionality reduction of the data. This can not only reduce the dimension of the data, improve the visualization and understandability of the data, but also accelerate the subsequent data analysis and processing process; Gaussian mixture clustering is a soft clustering method, that is, it assigns a probability of belonging to each cluster to each sample, rather than directly assigning the sample to a certain cluster. This soft clustering method enables it to better handle the uncertainty or overlap existing in the data set. Based on the soft clustering, relatively soft clustering boundaries can be generated, which means that the sample is not only assigned to the most likely cluster center, but the possibility of belonging to different clusters is described by probability, which has advantages in the case of fuzzy boundaries existing in some data sets.

[0063] The present invention combines the kernel principal component analysis (KPCA) and Gaussian mixture clustering (GMM) algorithms to conduct in-depth analysis of the voltage data in the low-voltage distribution network, and provides a new technical solution for the verification of the household-to-transformer relationship in the substation area, which has the following advantages:

[0064] 1) Overcoming the nonlinear problem of similar voltage curves: Users in the low-voltage distribution network may be mistakenly considered to belong to the same substation due to similar voltage fluctuation curves, and traditional similarity-based recognition methods often cannot accurately distinguish these users. By introducing kernel principal component analysis, the data is mapped to a high-dimensional feature space, which can capture the nonlinear relationship between voltage data and avoid the misidentification problem that cannot be handled by simple linear methods. Therefore, this method can provide more accurate household-to-substation relationship verification results when facing users with highly similar voltage curves but belonging to different substations.

[0065] 2) Improve the interpretability and efficiency of data dimensionality reduction: The dimensionality of user voltage data in low-voltage distribution networks is high, and traditional methods may face problems such as large computational workload and complex analysis. Kernel principal component analysis effectively compresses the dimensions of data through dimensionality reduction technology, making the analysis process more efficient and easier to understand. By extracting the principal components in high-dimensional space, this method can effectively retain key feature information while eliminating redundant data, thereby accelerating subsequent analysis and processing processes and improving the efficiency of the recognition process.

[0066] 3) Dealing with uncertainty and fuzzy boundary problems: In low-voltage distribution networks, there may be some overlap or uncertainty in the voltage data between users. As a soft clustering method, Gaussian mixture clustering can provide multiple possible attributes for each sample through probability allocation, so as to better deal with fuzzy cluster boundaries. For those users whose voltage data have high similarity but still belong to different substations, GMM can flexibly divide them into different groups according to their probability distribution, avoiding the errors caused by hard allocation.

[0067] The present invention can better process complex voltage data in low-voltage distribution networks and effectively verify the relationship between substations and households through the organic combination of kernel principal component analysis and Gaussian mixture clustering. It has strong practicality and advantages, can improve the accuracy and efficiency of distribution network operation, and reduce the possibility of misidentification. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] The first aspect of the present invention provides a method for verifying the relationship between users and transformers in a low-voltage distribution network, and the specific steps are as follows:

[0070] Step 1: Construct a voltage value matrix

[0071] Step 1.1: Starting from 0:00 every day, record the effective value of the user voltage every 10 minutes. Each user can record 144 voltage data per day. These 144 data can form the voltage time series column vector X of user i i .

[0072] Step 1.2: For n users to be identified, form a voltage value matrix X by combining the voltage time series of each user, X = [X 1 , X 2 ,..., X n T .

[0073] Step 2: Data standardization

[0074] Step 2.1: The standardization method is where is the mean of the data in the j-th column of the voltage value matrix X, is the standard deviation of the data in the j-th column of the voltage value matrix X. The standardized voltage value matrix is

[0075] Step 3: Use sparse kernel principal component analysis for data dimensionality reduction; introduce sparse kernel principal component analysis (SparseKPCA): Sparse KPCA makes the reduced feature vectors sparse by applying L1 regularization to the feature vectors, thereby improving the model's ability to capture key features. In voltage data, specific time points or features may be more important than other time points, and the sparsity constraint can help identify key voltage fluctuation patterns and reduce the influence of unimportant noise.

[0076] Step 3.1: Obtain the corresponding kernel matrix W using the Gaussian kernel function.

[0077] Specifically: And set the standard deviation σ of the Gaussian kernel function to 0.5.

[0078] Step 3.2: Centralize the kernel matrix W. Obtain a new matrix K, K = W - NW - WN - NWN. Where N is an n×n matrix with all elements being matrix.

[0079] Step 3.3: Perform eigen-decomposition on the matrix W to obtain the corresponding eigen-column vectors P 1 , P 2 ,..., P n and eigenvalues λ 1 , λ2 ,..., λ n . Sort the eigenvalues in descending order, and retain the first m eigenvalues and their corresponding eigenvectors after sorting. The selected eigenvectors are concatenated to form the projection matrix P = [P 1 , P 2 ,..., P m , where [·] is the floor function.

[0080] Step 3.4: Introduce sparsity in this step. Restrict the projection matrix P through L1 regularization to reduce the number of non-zero elements in the projection vector. The goal of the sparsity constraint is to optimize the sparsity by minimizing the objective function, and the objective function is: where α = 0.01, ||·|| 2 is the square of the matrix L2 norm, and ||·|| 1 is the L1 norm. Then the sparse projection matrix

[0081] Step 3.4: Obtain the dimensionality reduction matrix Y, Y = M * X. The i-th column element in Y represents the sample data of user i after dimensionality reduction.

[0082] Step 4: Use the Gaussian mixture clustering method for clustering to achieve the verification of the household change relationship

[0083] Step 4.1: Set the number of Gaussian distributions to the total number of substations to be identified, denoted as k.

[0084] Step 4.2: The parameters of each Gaussian distribution are denoted as θ η ={π η , μ η , ∑ η}, where η represents the η-th distribution. Among them, π η is the mixing coefficient of the η-th Gaussian distribution and satisfies Initialized as μ η is the center point of each Gaussian distribution, initialized as μ η = Y 1,η ; The covariance matrix ∑ η represents the shape and distribution range of each Gaussian distribution, initialized as the m-dimensional identity matrix, that is, ∑ η = I m .

[0085] Step 4.3: Calculate the posterior probability that the sample Y ω (the ω-th column data of the dimensionality reduction matrix Y) belongs to the η-th Gaussian distribution, that is where χ(Y ω |μ η, Σ η ) is the probability density function of the Gaussian distribution,

[0086] Step 4.4: Iteratively update the parameters:

[0087]

[0088] Step 4.5: At the end of each iteration, calculate the absolute value of the difference between the log-likelihood functions before and after the parameter iteration, that is, calculate ΔlogL(θ) = |logL(θ new ) - logL(θ old )|. Among them, the log-likelihood function is:

[0089]

[0090] Step 4.6: Repeat Steps 4.3 and 4.4. When ΔlogL(θ) < 0.001 or the number of iterations reaches 1000, the algorithm terminates the iteration.

[0091] Step 4.7: After the algorithm terminates the iteration, according to the posterior probability γ ω,η allocate the sample Y ω to the Gaussian distribution with the maximum probability, that is

[0092] Step 4.8: Compare the household transformation identification results of all users with the system file. If the identified substation transformer number of a certain user given by the identification result is inconsistent with the substation transformer number recorded in the system file for this user, then this user is identified as an abnormal user, and the household-substation relationship verification is realized.

[0093] Specific example: Three substations were selected in a certain community in Nanjing, and 20 users were selected from each substation within 3 days. A total of 60 users' voltage data were collected as experimental data to verify this method. And the clustering effect of this method and other methods for the experimental data was compared, and the standard mutual information was used as the evaluation index of the clustering result. The specific results are shown in the following table.

[0094]

[0095]

[0096] It can be seen from the table that this method can accurately identify the household-substation relationship, while the effects shown by other methods of reducing the dimension of data and performing clustering are relatively poor.

[0097] The embodiment of the present invention also provides a device for verifying the household-substation relationship of a low-voltage distribution network, including:

[0098] A voltage value matrix construction module, which is used to obtain user voltage data and construct a voltage value matrix X;

[0099] A normalization module, which is used to perform normalization processing on the voltage value matrix X to obtain a normalized voltage value matrix

[0100] A dimensionality reduction module, which is used to perform data dimensionality reduction on the normalized voltage value matrix by using sparse kernel principal component analysis to obtain a dimensionality reduction matrix Y;

[0101] A clustering verification module, which is used to perform clustering on the dimensionality reduction matrix Y by using the Gaussian mixture clustering method and verify by comparing with the archives.

[0102] The voltage value matrix construction module is specifically used for:

[0103] Obtain the daily voltage data of the user to form the voltage time series column vector Xi of user i i ;

[0104] For n users to be identified, form the voltage value matrix X by combining the voltage time series of each user, X = [X 1 , X 2 ,..., X n T 。

[0105] The normalization module is specifically used for:

[0106] The normalization method is where is the mean value of the data in the jth column of the voltage value matrix X, is the standard deviation of the data in the jth column of the voltage value matrix X; the normalized voltage value matrix is

[0107] The dimensionality reduction module is specifically used for:

[0108] Use the Gaussian kernel function to obtain the corresponding kernel matrix W; specifically: And set the standard deviation σ of the Gaussian kernel function;

[0109] Centralize the kernel matrix W to obtain a new matrix K, K = W - NW - WN - NWN; where N is an n×n matrix with all elements being ;

[0110] Perform eigenvalue decomposition on the matrix W to obtain the corresponding eigenvector columns P 1 , P 2 ,..., P n and eigenvalues λ 1 , λ 2 ,..., λ n ​; Sort the eigenvalues in descending order, and retain the first m eigenvalues and their corresponding eigenvectors after sorting; The selected eigenvectors are concatenated to form the projection matrix P = [P 1 , P 2 ,..., P m ; where [·] is the floor function;

[0111] Introduce sparsity, and restrict the projection matrix P through L1 regularization to reduce the number of non-zero elements in the projection vector; The goal of the sparsity constraint is to optimize the sparsity by minimizing the objective function, and the objective function is: where α = 0.01, ||·|| 2 is the square of the matrix L2 norm, and ||·|| 1 is the L1 norm; Then the sparse projection matrix

[0112] The dimensionality reduction matrix Y = M * X; The i-th column element in Y represents the sample data of user i after dimensionality reduction.

[0113] The clustering verification module is specifically used for:

[0114] Set the number of Gaussian distributions to the total number of substations to be identified, denoted as k;

[0115] The parameters of each Gaussian distribution are denoted as θ η = {π η , μ η , ∑ η}, where η represents the η-th distribution; among them, π η is the mixing coefficient of the η-th Gaussian distribution and satisfies Initialized as μ η is the center point of each Gaussian distribution, initialized as μ η = Y 1,η ; The covariance matrix ∑ η represents the shape and distribution range of each Gaussian distribution, and is initialized as the m-dimensional identity matrix, that is, ∑ η = I m ;

[0116] Calculate the posterior probability that the sample Y ω corresponding to user ω belongs to the η-th Gaussian distribution, that is where χ(Y ω |μ η , Σ η ) is the probability density function of the Gaussian distribution,

[0117] Iteratively update the parameters:

[0118]

[0119] At the end of each iteration, calculate the absolute value of the difference between the log-likelihood functions before and after the parameter iteration, that is, calculate ΔlogL(θ) = |logL(θ new ) - logL(θ old )); where the log-likelihood function is:

[0120]

[0121] When ΔlogL(θ) < 0.001 or the number of iterations reaches 1000, the algorithm terminates the iteration;

[0122] After the algorithm terminates the iteration, according to the posterior probability γ ω,η allocate the sample Y ω to the Gaussian distribution with the maximum probability, that is

[0123] Compare the household change identification results of all users with the system file. If the identified substation transformer number of a certain user given by the identification result is inconsistent with the substation transformer number recorded in the system file for this user, then this user is identified as an abnormal user, and the household-substation relationship verification is realized.

[0124] On the other hand, the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the low-voltage distribution network household-substation relationship verification method described in the first aspect is realized.

[0125] On the other hand, the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the low-voltage distribution network household-substation relationship verification method described in the first aspect is realized.

[0126] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can adopt the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0127] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of flows and / or blocks in the flowchart and / or block diagram can also be implemented. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in one or more flows in the flowchart and / or one or more blocks in the block diagram.

[0128] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the functions specified in one or more flows in the flowchart and / or one or more blocks in the block diagram.

[0129] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more flows in the flowchart and / or one or more blocks in the block diagram.

[0130] Finally, 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 them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent replacements. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A method for verifying the relationship between household transformers in a low-voltage distribution network, characterized in that: include: Obtain user voltage data to construct a voltage numerical matrix X; The voltage numerical matrix X is normalized to obtain the standardized voltage numerical matrix The standardized voltage numerical matrix is ​​analyzed using sparse kernel principal component analysis. Perform data dimensionality reduction to obtain the dimensionality reduction matrix Y; The Gaussian mixture clustering method is used to cluster the dimension reduction matrix Y and verify it by comparing it with the archives.

2. The method for verifying the relationship between household transformers in a low-voltage distribution network according to claim 1 is characterized in that: The method of obtaining user voltage data to construct a voltage numerical matrix X is specifically as follows: Get the user's daily voltage data to form the voltage time series column vector X of user i i ; For n users to be identified, the voltage time series of each user are combined into a voltage numerical matrix X, X = [X1, X2, ..., X n ] T .

3. The method for verifying the relationship between household transformers in a low-voltage distribution network according to claim 1 is characterized in that: The voltage numerical matrix X is standardized to obtain a standardized voltage numerical matrix Specifically: The standardized method is in is the mean value of the data in the jth column of the voltage numerical matrix X, is the standard deviation of the data in the jth column of the voltage numerical matrix X; the standardized voltage numerical matrix is 4. The method for verifying the relationship between household transformers in a low-voltage distribution network according to claim 1 is characterized in that: The standardized voltage numerical matrix is ​​analyzed by sparse kernel principal component analysis. The dimension reduction matrix Y obtained by data dimension reduction is as follows: The corresponding kernel matrix W is obtained using the Gaussian kernel function; specifically: And set the Gaussian kernel function standard deviation σ; Centralize the core matrix W to obtain a new matrix K, K = W-NW-WN-NWN; where N is an n×n matrix with all elements being Matrix of Perform eigendecomposition on the matrix W and obtain the corresponding eigenvalue column vectors P1, P2, ..., P n and eigenvalues ​​λ1,λ2,...,λ n ; Sort the eigenvalues ​​from large to small, retain the first m eigenvalues ​​and their corresponding eigenvectors after sorting; concatenate the selected eigenvalues ​​to form a projection matrix P = [P1, P2, ..., P m ];in is the floor rounding function; Introducing sparsity, restricting the projection matrix P through L1 regularization, so that the number of non-zero elements in the projection vector is reduced; The goal of the sparsity constraint is to optimize sparsity by minimizing the objective function, which is: where α = 0.01, ||·|| 2 is the square of the matrix L2 norm, ||·||1 is the L1 norm; then we can get the sparse projection matrix Dimensionality reduction matrix Y = M * The elements in the i-th column of X;Y represent the sample data of user i after dimensionality reduction.

5. The method for verifying the relationship between household transformers in a low-voltage distribution network according to claim 1 is characterized in that: The Gaussian mixture clustering method is used to cluster the dimension reduction matrix Y to realize the household change relationship verification, specifically: The number of Gaussian distributions is set to be the total number of areas to be identified, denoted as k; The parameter of each Gaussian distribution is denoted by θ η ={π η ,μ η ,∑ η }, η represents the ηth distribution; where π η is the mixing coefficient of the ηth Gaussian distribution and satisfies Initialize to μ η is the center point of each Gaussian distribution, initialized to μ η =Y 1,η ; Covariance matrix∑ η Represents the shape and distribution range of each Gaussian distribution, initialized as an m-dimensional identity matrix, that is, ∑ η =I m ; Calculate the sample Y corresponding to user ω ω The posterior probability of belonging to the ηth Gaussian distribution is Where χ(Y ω |μ η ,∑ η ) is the probability density function of the Gaussian distribution, Iterative update parameters: At the end of each iteration, the absolute value of the difference between the log-likelihood function before and after the parameter iteration is calculated, that is, ΔlogL(θ)=|logL(θ new )-logL(θ old )|; the log-likelihood function is: When ΔlogL(θ)<0.001 or the number of iterations reaches 1000, the algorithm terminates the iteration; After the algorithm terminates the iteration, according to the posterior probability γ ω,η The sample Y ω Assign to the Gaussian distribution with maximum probability, that is The household transformer identification results of all users are compared with the system files. If the transformer number of a user given by the identification result is inconsistent with the transformer number of the user recorded in the system file, the user is identified as an abnormal user and the household transformer relationship verification is realized.

6. A low voltage distribution network household transformer relationship verification device, characterized in that: include: A voltage numerical matrix construction module is used to obtain user voltage data to construct a voltage numerical matrix X; The standardization module is used to standardize the voltage numerical matrix X to obtain a standardized voltage numerical matrix Dimensionality reduction module for normalizing the voltage numerical matrix using sparse kernel principal component analysis Perform data dimensionality reduction to obtain the dimensionality reduction matrix Y; The clustering verification module is used to cluster the dimension reduction matrix Y using the Gaussian mixture clustering method and verify it by comparing the archives.

7. The low-voltage distribution network household transformer relationship verification device according to claim 6 is characterized in that: The voltage numerical matrix construction module is specifically used for: Get the user's daily voltage data to form the voltage time series column vector X of user i i ; For n users to be identified, the voltage time series of each user are combined into a voltage numerical matrix X, X = [X1, X2, ..., X n ] T .

8. The low-voltage distribution network household transformer relationship verification device according to claim 6 is characterized in that: The standardization module is specifically used for: The standardized method is in is the mean value of the data in the jth column of the voltage numerical matrix X, is the standard deviation of the data in the jth column of the voltage numerical matrix X; the standardized voltage numerical matrix is 9. The low-voltage distribution network household transformer relationship verification device according to claim 6, characterized in that: The dimension reduction module is specifically used for: The corresponding kernel matrix W is obtained using the Gaussian kernel function; specifically: And set the Gaussian kernel function standard deviation σ; Centralize the core matrix W to obtain a new matrix K, K = W-NW-WN-NWN; where N is an n×n matrix with all elements being Matrix of Perform eigendecomposition on the matrix W and obtain the corresponding eigenvalue column vectors P1, P2, ..., P n and eigenvalues ​​λ1,λ2,...,λ n ; Sort the eigenvalues ​​from large to small, retain the first m eigenvalues ​​and their corresponding eigenvectors after sorting; concatenate the selected eigenvalues ​​to form a projection matrix P = [P1, P2, ..., P m ];in [·] is the floor rounding function; Sparsity is introduced and the projection matrix P is restricted by L1 regularization so that the number of non-zero elements in the projection vector is reduced. The goal of the sparsity constraint is to optimize sparsity by minimizing the objective function, which is: where α = 0.01, ||·|| 2 is the square of the matrix L2 norm, ||·||1 is the L1 norm; then we can get the sparse projection matrix Dimensionality reduction matrix Y = M * The elements in the i-th column of X;Y represent the sample data of user i after dimensionality reduction.

10. The low-voltage distribution network household transformer relationship verification device according to claim 6, characterized in that: The cluster verification module is specifically used for: The number of Gaussian distributions is set to be the total number of areas to be identified, denoted as k; The parameter of each Gaussian distribution is denoted by θ η ={π η ,μ η ,∑ η }, η represents the ηth distribution; where π η is the mixing coefficient of the ηth Gaussian distribution and satisfies Initialize to μ η is the center point of each Gaussian distribution, initialized to μ η =Y 1,η ; Covariance matrix∑ η Represents the shape and distribution range of each Gaussian distribution, initialized as an m-dimensional identity matrix, that is, ∑ η =I m ; Calculate the sample Y corresponding to user ω ω The posterior probability of belonging to the ηth Gaussian distribution is Where χ(Y ω |μ η ,Σ η ) is the probability density function of the Gaussian distribution, Iterative update parameters: At the end of each iteration, the absolute value of the difference between the log-likelihood function before and after the parameter iteration is calculated, that is, ΔlogL(θ)=|logL(θ new )-logL(θ old )|; the log-likelihood function is: When ΔlogL(θ)<0.001 or the number of iterations reaches 1000, the algorithm terminates the iteration; After the algorithm terminates the iteration, according to the posterior probability γ ω,η The sample Y ω Assign to the Gaussian distribution with maximum probability, that is The household transformer identification results of all users are compared with the system files. If the transformer number of a user given by the identification result is inconsistent with the transformer number of the user recorded in the system file, the user is identified as an abnormal user and the household transformer relationship verification is realized.

11. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for verifying the relationship between users and transformers in a low-voltage distribution network as described in any one of claims 1 to 5 is implemented.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for verifying the relationship between users and transformers in a low-voltage distribution network as claimed in any one of claims 1 to 5 is implemented.

Images