Topology identification method, system and equipment for low-voltage power distribution network court and medium

Abstract

The invention discloses a low-voltage power distribution network area topology identification method, system and device and a medium, and belongs to the field of electric power systems.The method comprises the steps that an incidence matrix is determined according to distribution transformer power data and power response intensity, a multi-dimensional similarity matrix is determined in combination with voltage characteristic data, and through multiple times of consensus spectral clustering and confidence coefficient analysis, a multi-dimensional similarity matrix is determined; obtaining a branch line division result; extracting a multi-dimensional electrical feature vector, determining a phase similarity matrix, constructing a user phase association graph, and performing iterative optimization on phase probability distribution through a graph smoothing algorithm to obtain an access phase attribution result; and the branch line division result and the access phase affiliation result are fused, and the topological structure of the low-voltage power distribution network area is determined based on the association relationship between the user and the target distribution transformer, so that accurate topology identification of the low-voltage power distribution network can be realized.

Description

Technical Field

[0001] This invention relates to the field of power systems, and in particular to a method, system, device and medium for identifying the topology of a low-voltage distribution network. Background Technology

[0002] Low-voltage distribution networks are the terminal devices connecting distribution transformers and end users in a power system. The accuracy of their distribution network topology is crucial for refined operation and maintenance work such as line loss calculation, three-phase imbalance management, and rapid fault location. The topology includes the hierarchical relationship between users and distribution transformers, the connection logic of branch lines, and the phase assignment of user access. With the development of smart grids, low-voltage distribution areas are large in scale and have complex user types. Traditional manual inspection methods are inefficient and prone to errors, and can no longer meet actual needs.

[0003] Currently, mainstream topology identification technologies include manual inspection, signal injection, and data-driven methods. Among them, a hierarchical clustering method based on voltage-power features can perform branching based on voltage residual similarity and use two-dimensional features of power factor and voltage deviation for phase identification. However, this method has significant limitations: it has a single feature dimension, and its identification accuracy drops significantly under load fluctuations or mixed load scenarios; moreover, it lacks an automatic correction mechanism for low-confidence results, resulting in insufficient robustness. Summary of the Invention

[0004] This invention provides a method, system, device, and medium for topology identification of low-voltage distribution network areas, which can solve the problem of how to improve the accuracy of topology identification.

[0005] This invention provides a method for identifying the topology of a low-voltage distribution network, comprising:

[0006] The correlation matrix is ​​determined based on the power data of the target transformer and the power response intensity of each user. Multiple power similarity matrices are determined based on the correlation matrix and the voltage characteristic data of each user. The power similarity matrices are weighted and fused to obtain a multi-dimensional similarity matrix. The branch line division results are obtained by performing multiple consensus spectrum clustering analysis and confidence analysis on the multi-dimensional similarity matrix. The correlation matrix is ​​used to characterize the correspondence between users and events.

[0007] For each user under the same branch in the branch line division result, a multi-dimensional electrical feature vector is extracted, a phase similarity matrix is ​​determined based on the multi-dimensional electrical feature vector, a user phase association map is constructed based on the phase similarity matrix, and the phase probability distribution of each user in the user phase association map is iteratively optimized by a graph smoothing algorithm to determine the access phase affiliation of users in each branch line, and the access phase affiliation result is obtained.

[0008] By integrating the branch line division results and the access phase assignment results, and based on the association relationship between each user and the target transformer, the topology of the low-voltage distribution network area corresponding to the target transformer is determined.

[0009] This invention generates an association matrix using "transformer power data + user power response intensity," transforming high-noise raw measurements into low-dimensional, event-driven response intensity. This amplifies the electrical coupling signal between users and transformers, reducing noise in subsequent clustering inputs. Multiple power similarity matrices are jointly generated using the association matrix and voltage feature data, simultaneously quantifying the two independent physical channels of power and voltage response. This compensates for the decreased discriminability of single voltage or power features during load fluctuations. A weighted fusion of these multiple similarity matrices yields a "multi-dimensional similarity matrix." Through matrix-level fusion, similarity information with different geometric / statistical characteristics is compressed into a unified metric, expanding the numerical differences between different user groups and providing richer discriminative dimensions for clustering. The multi-dimensional similarity matrix undergoes "multiple consensus spectrum clustering + confidence analysis." After multiple random initial value clusterings, the co-occurrence probability is statistically analyzed to quantify the stability of each user's affiliation. Low-confidence users are automatically identified and corrected secondaryly, reducing result drift caused by random errors in single clustering. Within the same branch, "multi-dimensional electrical feature vectors" are extracted to construct a phase correlation graph. A graph smoothing algorithm is then used to iteratively optimize the phase probability distribution, mapping high-dimensional electrical features to graph node signals. Through neighborhood smoothing iteration, the probabilities of adjacent nodes (real in-phase users) converge, while the probabilities of out-of-phase nodes diverge, improving phase identification accuracy. Finally, the branch division results and access phase attribution results are fused to form the transformer substation topology. The three-level connectivity relationship of "feeder-branch-phase" is output in one go, achieving accurate topology identification of the low-voltage distribution network.

[0010] Furthermore, the determination of the correlation matrix based on the distribution power data of the target transformer and the power response intensity corresponding to each user specifically involves:

[0011] Event detection is performed based on the active power time-series data in the power data of the distribution transformer, and a preset number of events are determined according to the detection results to obtain a global event set of the distribution transformer.

[0012] For each event in the global event set, based on a preset event window, the mean of the absolute value of the power difference for each user within the preset event window is calculated as the power response intensity of each user to each event;

[0013] The response intensity matrix is ​​determined based on the power response intensity of all users. The response intensity matrix is ​​then subjected to moving average processing and normalization processing to obtain the correlation matrix.

[0014] This process involves event detection on active power time series data, calculating the mean absolute value of user power differences within the event window as the response intensity, forming a response intensity matrix, and then obtaining the correlation matrix through moving average and normalization. The event triggering mechanism retains only "high signal-to-noise ratio" periods, filtering out daily random fluctuations and improving the robustness of response intensity calculation. Moving average smooths out random abrupt changes, and normalization eliminates dimensional differences, ensuring a uniform numerical range for the correlation matrix and providing a consistent benchmark for subsequent similarity comparisons, reducing clustering bias caused by dimensional inconsistencies.

[0015] Further, the step of determining multiple similarity matrices based on the correlation matrix and the voltage feature data corresponding to each user, and then weighting and fusing the similarity matrices to obtain a multi-dimensional similarity matrix, wherein the similarity matrix includes a basis matrix cosine similarity matrix, a voltage residual correlation matrix, a radial basis similarity matrix, and a Jaccard similarity matrix, specifically:

[0016] The correlation matrix is ​​decomposed using a nonnegative matrix factorization method to obtain a basis matrix and a coefficient matrix. Each column of the basis matrix represents a branch feature vector, and each row represents the weight vector of each user's association with each branch.

[0017] Based on the coefficient matrix, calculate the cosine similarity of the weight vectors to which any two users belong in the base matrix to obtain the cosine similarity matrix of the base matrix.

[0018] Based on the user voltage time series data of each user in the voltage characteristic data and the distribution transformer voltage time series data of the target distribution transformer, calculate the voltage residual time series sequence of each user, and then calculate the Pearson correlation coefficient of the voltage residual time series sequences of any two users to obtain the voltage residual correlation matrix.

[0019] Based on the radial basis function, the nonlinear similarity between any two users is calculated for the power response intensity of each user, and the radial basis similarity matrix is ​​obtained.

[0020] By comparing the power response intensity of each user with a preset threshold, it is determined whether each user responds, thus obtaining a set of user response events. Based on the set of user response events, the Jaccard similarity coefficient of any two sets of user response events is calculated, thus obtaining the Jaccard similarity matrix.

[0021] The basis matrix cosine similarity matrix, the voltage residual correlation matrix, the radial basis similarity matrix, and the Jaccard similarity matrix are weighted and fused according to a first preset weight ratio to obtain the multi-dimensional similarity matrix.

[0022] This process involves: obtaining the basis matrix and coefficient matrix through nonnegative matrix decomposition of the correlation matrix; calculating cosine similarity based on the coefficient vector → basis matrix cosine similarity matrix; obtaining the residual sequence from voltage time series reduction of transformer voltage → Pearson correlation coefficient matrix; mapping power response intensity using radial basis functions → nonlinear similarity matrix; calculating Jaccard coefficients after binarizing the response intensity → Jaccard similarity matrix; and fusing the four matrices with preset weights into a multidimensional similarity matrix. These four matrices respectively capture four different geometric or set features: "attribution weight directionality, voltage coupling, response intensity nonlinearity, and event co-occurrence rate." Weighted fusion provides a more comprehensive similarity profile, avoiding blind spots caused by single indicators. Nonnegative matrix decomposition compresses high-dimensional user-event data into low-dimensional "user-hidden branch" vectors, preserving physical interpretability while reducing the computational complexity of subsequent spectral clustering.

[0023] Furthermore, the branch line division result is obtained by performing multiple consensus spectrum clustering analyses and confidence analyses on the multi-dimensional similarity matrix, specifically as follows:

[0024] The target number of branches is determined based on the base matrix and the coefficient matrix. The target number of branches is used as the number of clusters. The spectral clustering algorithm is used to perform clustering analysis with different random seeds for a preset number of clusters. The branch affiliation label of each user in each cluster is recorded.

[0025] For each user, the number of times each user shares the same branch affiliation label with other users in multiple clusterings is counted. The ratio of the number of times any two users share the same branch affiliation label to the preset number of clusterings is used as the element value in the consensus matrix to obtain the consensus matrix.

[0026] Based on the consensus matrix, the cluster confidence of each user is calculated. For users to be classified whose cluster confidence is less than a preset confidence threshold, the feature vector of each user is calculated based on the voltage residual time series and the power response intensity. The branch feature center corresponding to each branch is calculated based on the mean of the feature vectors of all users in each branch. The similarity between the feature vector of the user to be classified and the feature center of each branch is calculated. The user to be classified is then reassigned to the branch with the highest similarity.

[0027] The number of users in each branch is counted, and users in branches with a number of users less than a preset threshold are merged into the neighboring branches with the most similar features to obtain the branch line division result.

[0028] This process involves determining the optimal number of branches by jointly using the proportion of zero elements in the base matrix and the silhouette coefficient → multiple random seed spectrum clustering → calculating the consensus matrix by counting co-occurrence times → calculating the user cluster confidence; low-confidence users are re-matched with branch centers based on the "voltage residual + power response" feature vector → reassigned; tiny branches with fewer than the threshold users are removed and merged into the nearest neighbor branch. The consensus matrix solidifies the results of multiple random clustering into a probabilistic form, significantly reducing the sensitivity to initial values; the confidence screening + re-matching mechanism performs secondary correction on "marginal users," reducing the misclassification rate; the joint optimization of the number of branches by the proportion of zero elements and the silhouette coefficient balances matrix sparsity (physical interpretability) and intra-cluster compactness (numerical quality), ensuring that the final number of branches is highly consistent with the actual number of feeders, thus improving the accuracy of the topology structure.

[0029] Furthermore, determining the number of target branches based on the basis matrix and the coefficient matrix specifically involves:

[0030] Obtain the user number limit corresponding to each branch, and determine the branch number range based on the user number limit and the total number of users of the target transformer;

[0031] For each branch number within the specified branch number range, the silhouette coefficient is calculated based on each of the specified attribution weight vectors, and the sparsity is obtained by calculating the proportion of zero elements in the basis matrix.

[0032] Based on the second preset weight ratio, the contour coefficient and sparsity of each branch number are weighted and summed to obtain the comprehensive score corresponding to each branch number. The branch number corresponding to the highest comprehensive score is taken as the target branch number.

[0033] This approach involves first calculating the upper and lower bounds of the number of branches based on the "user quantity limit" and the total number of users; then enumerating the number of branches within this range; calculating the corresponding contour coefficients and the sparsity of the basis matrix; finally, weighting the results according to a second preset weight to obtain a comprehensive score; and finally, using the number of branches corresponding to the highest score as the target number of branches. This transforms the "feeder capacity limit" into a hard upper and lower bound, avoiding overly fine or coarse segmentation; the contour coefficients ensure intra-class compactness, and the sparsity ensures physical interpretability. The weighted sum of these two factors automatically selects the optimal number of branches, eliminating the need for manual trial and error, improving recognition efficiency, and reducing the risk of over-segmentation.

[0034] Furthermore, for each user under the same branch in the branch line division result, the extraction of multi-dimensional electrical feature vectors is specifically as follows:

[0035] Based on the active power time-series data corresponding to each user, the power shape vector of each user is determined, the first Pearson correlation coefficient between the power shape vector and the three-phase power shape template is calculated, and the maximum value of the first Pearson correlation coefficient is taken as the power shape correlation feature value.

[0036] Acquire reactive power time-series data, calculate the median power factor for each user based on the active power time-series data and the reactive power time-series data, calculate the deviation between the median power factor and the center of the three-phase power factor, and take the minimum value of the deviation as the power factor matching degree feature value.

[0037] Calculate the median of the ratio of the active power time-series data to the reactive power time-series data, calculate the deviation of the median from the center value of the three ratios, and take the minimum value of the deviation as the characteristic value of the ratio.

[0038] The active power time-series data is subjected to a fast Fourier transform to obtain a frequency domain amplitude vector. The cosine similarity between the frequency domain amplitude vector and the three-phase frequency domain template is calculated, and the maximum value of the cosine similarity is taken as the frequency mode feature value.

[0039] Calculate the difference between the active power time-series data and the power mean, determine the power residual vector based on the difference, and calculate the second Pearson correlation coefficient between the power residual vector and the three-phase residual template. Take the maximum value of the second Pearson correlation coefficient as the residual shape feature value.

[0040] Based on the branch line division results, the mean feature matching degree between the user and all other users in the same branch is calculated as the clustering consistency feature value. The mean feature matching degree is the mean of the power shape correlation feature value, the power factor matching degree feature value, the ratio feature value, the frequency mode feature value, and the residual shape feature value.

[0041] The power shape correlation feature value, the power factor matching feature value, the ratio feature value, the frequency mode feature value, the residual shape feature value, and the cluster consistency feature value are normalized to obtain the multi-dimensional electrical feature vector.

[0042] This approach extracts five single-phase features: power shape correlation, power factor matching degree, active / reactive ratio deviation, frequency mode cosine similarity, and power residual shape correlation. These are then combined with the "mean of feature matching degree within the same branch" as a clustering consistency feature. After normalization, these six features form a multi-dimensional electrical feature vector. This six-dimensional feature vector covers five electrical subspaces: amplitude, phase, waveform, spectrum, and consistency, forming a "phase fingerprint." This maximizes the difference in features between users in different phases and minimizes the difference between users in the same phase. After normalization to eliminate dimensions, it can be directly used in graph smoothing algorithms, improving the convergence speed and accuracy of phase assignment.

[0043] Furthermore, determining the phase similarity matrix based on the multi-dimensional electrical feature vector specifically involves:

[0044] Based on the third preset weight ratio, the normalized feature values ​​in the multi-dimensional electrical feature vector are weighted and fused to obtain the initial assignment probability of each user to each phase, and the initial assignment probability is encoded to obtain the initial phase vector.

[0045] Based on the normalized power shape correlation feature value, power factor matching degree feature value and ratio feature value in the multi-dimensional electrical feature vector, the feature similarity between any two users in the same branch is calculated using the cosine similarity function to obtain the phase similarity matrix.

[0046] This process involves weighting and fusing six-dimensional features using a third preset weight to generate the user's initial probability of belonging to phases A, B, and C, which is then encoded into an initial phase vector. Cosine similarity is then calculated using three-dimensional features: power shape, power factor, and ratio, constructing a phase similarity matrix. The initial probability vector compresses high-dimensional features into a 3-dimensional "phase space," providing initial values ​​for the graph smoothing algorithm and reducing oscillations caused by random initialization. The subsequent similarity matrix selects only the most discriminative three-dimensional features, reducing the number of graph edges and improving the computational efficiency of graph smoothing. This combination ensures both accurate starting points for phase probabilities and rapid convergence during iteration, ultimately significantly reducing the phase recognition error rate.

[0047] Another embodiment of the present invention provides a low-voltage distribution network area topology identification system, including: a branch clustering identification module, a phase attribution identification module, and a topology construction module;

[0048] The branch clustering identification module is used to determine the correlation matrix based on the power data of the target transformer and the power response intensity corresponding to each user, and to determine multiple power similarity matrices based on the correlation matrix and the voltage feature data corresponding to each user. The power similarity matrices are weighted and fused to obtain a multi-dimensional similarity matrix. The branch line division result is obtained by performing multiple consensus spectrum clustering analysis and confidence analysis on the multi-dimensional similarity matrix. The correlation matrix is ​​used to characterize the correspondence between users and events.

[0049] The phase attribution identification module is used to extract multi-dimensional electrical feature vectors for each user under the same branch in the branch line division result, determine the phase similarity matrix based on the multi-dimensional electrical feature vectors, construct a user phase association map based on the phase similarity matrix, and iteratively optimize the phase probability distribution of each user in the user phase association map through a graph smoothing algorithm to determine the access phase attribution of users in each branch line and obtain the access phase attribution result.

[0050] The topology construction module is used to integrate the branch line division results and the access phase assignment results, and determine the topology of the low-voltage distribution network area corresponding to the target transformer based on the association relationship between each user and the target transformer.

[0051] This invention generates an association matrix using "transformer power data + user power response intensity," transforming high-noise raw measurements into low-dimensional, event-driven response intensity. This amplifies the electrical coupling signal between users and transformers, reducing noise in subsequent clustering inputs. Multiple power similarity matrices are jointly generated using the association matrix and voltage feature data, simultaneously quantifying the two independent physical channels of power and voltage response. This compensates for the decreased discriminability of single voltage or power features during load fluctuations. A weighted fusion of these multiple similarity matrices yields a "multi-dimensional similarity matrix." Through matrix-level fusion, similarity information with different geometric / statistical characteristics is compressed into a unified metric, expanding the numerical differences between different user groups and providing richer discriminative dimensions for clustering. The multi-dimensional similarity matrix undergoes "multiple consensus spectrum clustering + confidence analysis." After multiple random initial value clusterings, the co-occurrence probability is statistically analyzed to quantify the stability of each user's affiliation. Low-confidence users are automatically identified and corrected secondaryly, reducing result drift caused by random errors in single clustering. Within the same branch, "multi-dimensional electrical feature vectors" are extracted to construct a phase correlation graph. A graph smoothing algorithm is then used to iteratively optimize the phase probability distribution, mapping high-dimensional electrical features to graph node signals. Through neighborhood smoothing iteration, the probabilities of adjacent nodes (real in-phase users) converge, while the probabilities of out-of-phase nodes diverge, improving phase identification accuracy. Finally, the branch division results and access phase attribution results are fused to form the transformer substation topology. The three-level connectivity relationship of "feeder-branch-phase" is output in one go, achieving accurate topology identification of the low-voltage distribution network.

[0052] Another embodiment of the present invention provides a terminal device, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps of the low-voltage distribution network area topology identification method of the present invention.

[0053] Another embodiment of the present invention provides a computer-readable storage medium item, including: a stored computer program, which, when the computer program is running, controls the device where the computer-readable storage medium is located to perform the steps of the low-voltage distribution network area topology identification method of the present invention. Attached Figure Description

[0054] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0055] Figure 1 This is a flowchart illustrating a method for identifying the topology of a low-voltage distribution network provided in an embodiment of the present invention.

[0056] Figure 2 This is a schematic diagram of a low-voltage distribution network topology identification system provided in an embodiment of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0059] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0060] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0061] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0062] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple groups" refers to two or more (including two groups), and "multiple pieces" refers to two or more (including two pieces).

[0063] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0064] See Figure 1 To address the problem of improving the accuracy of topology identification in existing technologies, an embodiment of the present invention provides a method for topology identification of low-voltage distribution network areas, including steps S101-S103, specifically including:

[0065] Step S101: Determine the correlation matrix based on the power data of the target transformer and the power response intensity corresponding to each user, and determine multiple power similarity matrices based on the correlation matrix and the voltage characteristic data corresponding to each user. Perform weighted fusion on each power similarity matrix to obtain a multi-dimensional similarity matrix. Obtain the branch line division result by performing multiple consensus spectrum clustering analysis and confidence analysis on the multi-dimensional similarity matrix. The correlation matrix is ​​used to characterize the correspondence between users and events.

[0066] It should be noted that this application is based on the premise of knowing the subordinate relationship between users and distribution transformers (user-transformer relationship), and aims to obtain the correlation between "user-branch line" and "user-access phase". It carries out hierarchical and progressive topology identification. The overall process can be divided into two major stages: branch cluster identification and phase attribution identification.

[0067] As an example of an embodiment of the present invention, the step of determining the correlation matrix based on the power data of the target transformer and the power response intensity of each user specifically involves: performing event detection based on the active power time-series data in the transformer power data, and determining a preset number of events based on the detection results to obtain a global event set for the transformer; for each event in the global event set, calculating the mean of the absolute value of the power difference of each user within the preset event window based on a preset event window, which is used as the power response intensity of each user to each event; determining the response intensity matrix based on the power response intensity of all users, and performing moving average processing and normalization processing on the response intensity matrix to obtain the correlation matrix.

[0068] In this embodiment, the total active power time-series data of the target transformer (i.e., the core component of transformer power data) is first acquired. An event detection algorithm is used to filter global events that significantly impact the overall transformer power (the filtering criteria can be that the power fluctuation exceeds the transformer's rated power by 5% and conforms to the 3σ normal fluctuation range), forming a global event set for the transformer. For each event in this set, a preset event window of [e-2, e+2] (e being the event time) is set. The mean of the absolute value of the power difference for each user within the window is calculated as the power response intensity. The initial matrix formed by the power response intensities of all users is then smoothed using a 5-point moving average (to filter measurement noise) and normalized according to the 95th quantile (to eliminate data magnitude differences), resulting in the correlation matrix A∈R representing the correspondence between users and events. U×E (The dimension is U×E, where U is the total number of users in the target transformer and E is the number of global events).

[0069] As an example of an embodiment of the present invention, the step of determining multiple similarity matrices based on the correlation matrix and the voltage feature data corresponding to each user, and weighting and fusing the similarity matrices to obtain a multi-dimensional similarity matrix, wherein the similarity matrix includes a basis matrix cosine similarity matrix, a voltage residual correlation matrix, a radial basis similarity matrix, and a Jaccard similarity matrix, specifically: the correlation matrix is ​​decomposed based on a non-negative matrix decomposition method to obtain a basis matrix and a coefficient matrix, wherein each column in the basis matrix represents a branch feature vector, and each row represents the weight vector of each user to each branch; based on the coefficient matrix, the cosine similarity of the weight vectors of any two users in the basis matrix is ​​calculated to obtain the basis matrix cosine similarity matrix; based on the user voltage time series data of each user in the voltage feature data and the target transformer... The voltage residual time series data of the distribution transformer is used to calculate the voltage residual time series sequence for each user, and then the Pearson correlation coefficient of the voltage residual time series sequences of any two users is calculated to obtain the voltage residual correlation matrix. Based on the radial basis function, the nonlinear similarity between any two users is calculated for the power response intensity of each user to obtain the radial basis similarity matrix. By comparing the power response intensity of each user with a preset threshold, it is determined whether each user has responded to obtain a user response event set. Based on the user response event set, the Jaccard similarity coefficient between any two user response event sets is calculated to obtain the Jaccard similarity matrix. The basis matrix cosine similarity matrix, the voltage residual correlation matrix, the radial basis similarity matrix, and the Jaccard similarity matrix are weighted and fused according to a first preset weight ratio to obtain the multi-dimensional similarity matrix.

[0070] In this embodiment, based on the correlation matrix and the voltage characteristic data of each user (the difference between the voltage time-series data collected by the smart meter on the user side and the voltage time-series data of the low-voltage bus of the distribution transformer, i.e., the voltage residual sequence), four power similarity matrices are constructed: one is the basis matrix cosine similarity matrix, which obtains the basis matrix (U×k, k is the number of candidate branches) and coefficient matrix by non-negative matrix decomposition (NMF) of the correlation matrix, satisfying the matrix decomposition relation A≈WH (basis matrix W∈R). U×k The coefficient matrix H∈R k×EThe four power similarity matrices are: a base matrix (each column representing a branch feature vector and each row representing a user's weight for each branch), calculated by taking the cosine similarity of the weight vectors of any two users in the base matrix; a voltage residual correlation matrix, calculated by taking the Pearson correlation coefficient of the voltage residual sequences of any two users; an RBF similarity matrix, calculated by taking the nonlinear similarity of the power response intensity vectors of users in the correlation matrix using radial basis functions; and a Jaccard similarity matrix, calculated by taking the Jaccard similarity coefficient of any two sets of user response events determined by the correlation matrix. Then, the elements at corresponding positions of the four power similarity matrices are weighted and summed according to a weight ratio of 40% for the base matrix cosine similarity matrix, 20% for the voltage residual correlation matrix, 20% for the RBF similarity matrix, and 20% for the Jaccard similarity matrix, to obtain a multi-dimensional similarity matrix.

[0071] As an example of an embodiment of the present invention, the step of obtaining the branch line division result by performing multiple consensus spectral clustering analyses and confidence analyses on the multi-dimensional similarity matrix specifically involves: determining the target number of branches based on the base matrix and the coefficient matrix; using the target number of branches as the number of clusters; performing clustering analysis with different random seeds using a preset number of clustering operations using a spectral clustering algorithm; recording the branch affiliation label of each user in each clustering operation; for each user, counting the number of times each user has the same branch affiliation label as other users in multiple clustering operations; and using the ratio of the number of times corresponding to any two users to the preset number of clustering operations as the element value in the consensus matrix to obtain the consensus matrix. Based on the consensus matrix, the cluster confidence score of each user is calculated. For users to be classified whose cluster confidence score is less than a preset confidence threshold, the feature vector of each user is calculated based on the voltage residual time series and the power response intensity. The branch feature center corresponding to each branch is calculated based on the mean of the feature vectors of all users in each branch. The similarity between the feature vector of the user to be classified and the feature center of each branch is calculated. The user to be classified is reassigned to the branch with the highest similarity. The number of users in each branch is counted. Users in branches with a number of users less than a preset threshold are merged into the neighboring branches with the most similar features to obtain the branch line division result.

[0072] In this embodiment, a consensus spectral clustering algorithm is used to perform six consensus spectral clustering operations with different random seeds, using a multi-dimensional similarity matrix as input (the number of clusters can be determined based on the number of candidate branches estimated by NMF). The consensus matrix is ​​constructed by counting the number of times users belong to the same branch in the six clustering operations. (l r(i) is the label of user i in the r-th cluster, and I(·) is the indicator function (1 if equal, 0 otherwise). Calculate the cluster confidence of each user (mean of the consensus matrix elements within the same branch). For users with confidence < 0.25 (preset threshold), calculate the similarity between their feature vector based on the correlation matrix and voltage residual sequence and the feature center of each branch and reassign them. At the same time, merge small branches with a user size < 1.5% of the total number of users to obtain the final branch line division result.

[0073] Step S102: For each user under the same branch in the branch line division result, extract multi-dimensional electrical feature vectors, determine the phase similarity matrix based on the multi-dimensional electrical feature vectors, construct a user phase association map based on the phase similarity matrix, iteratively optimize the phase probability distribution of each user in the user phase association map using a graph smoothing algorithm, determine the access phase attribution of users in each branch line, and obtain the access phase attribution result.

[0074] As an example of an embodiment of the present invention, the step of extracting multi-dimensional electrical feature vectors for each user under the same branch in the branch line division result specifically includes: determining the power shape vector of each user based on the active power time-series data corresponding to each user; calculating the first Pearson correlation coefficient between the power shape vector and the three-phase power shape template; and taking the maximum value of the first Pearson correlation coefficient as the power shape correlation feature value; obtaining reactive power time-series data; calculating the median of the power factor corresponding to each user based on the active power time-series data and the reactive power time-series data; calculating the deviation value between the median of the power factor and the center of the three-phase power factor; and taking the minimum value of the deviation value as the power factor matching degree feature value; calculating the median of the ratio of the active power time-series data to the reactive power time-series data; calculating the deviation value between the median and the center value of the three-phase ratio; and taking the minimum value of the deviation value as the ratio feature value; and performing a fast Fourier transform on the active power time-series data to obtain the frequency domain amplitude. The system calculates the cosine similarity between the frequency domain amplitude vector and the three-phase frequency domain template, and uses the maximum value of the cosine similarity as the frequency mode feature value. It then calculates the difference between the active power time-series data and the power mean, determines the power residual vector based on the difference, and calculates the second Pearson correlation coefficient between the power residual vector and the three-phase residual template, using the maximum value of the second Pearson correlation coefficient as the residual shape feature value. Based on the branch line division results, it calculates the mean feature matching degree between the user and all other users within the same branch, using it as the clustering consistency feature value. The mean feature matching degree is the mean of the power shape correlation feature value, the power factor matching degree feature value, the ratio feature value, the frequency mode feature value, and the residual shape feature value. Finally, it normalizes the power shape correlation feature value, the power factor matching degree feature value, the ratio feature value, the frequency mode feature value, the residual shape feature value, and the clustering consistency feature value to obtain the multi-dimensional electrical feature vector.

[0075] In this embodiment, the electricity consumption time-series data of each user within the same branch is first acquired (active power, reactive power, and voltage time-series data collected by the user-side smart meter). Based on this data, six-dimensional multi-dimensional electrical features are extracted and a feature vector is constructed: the first is the power shape correlation feature value. The Pearson correlation coefficient between the user's active power shape vector and the A / B / C three-phase power shape template is calculated, and the maximum value is taken to obtain the power shape correlation feature value (weight 35%). The calculation formula for the power shape correlation feature value is as follows:

[0076] F shape (i,ph)=max(0,corr(P bin,i ,T bin,ph ));

[0077] Among them, P bin,i T is the power shape vector. bin,ph The power shape template for each phase is (ph∈{A,B,C}).

[0078] Second, the power factor matching degree characteristic value is calculated by determining the minimum deviation (weighted at 18%) between the median power factor of the user and the power factor centers of phase A (0.945), phase B (0.91), and phase C (0.9625). The formula for calculating the power factor matching degree is as follows:

[0079]

[0080] Among them, med(PF) i ) represents the median power factor for users, μ pf,ph It is the center of the phase power factor.

[0081] Thirdly, the QP ratio characteristic value is calculated by taking the minimum deviation (weighted at 15%) between the median of the user's reactive power to active power ratio and the center of the three-phase QP ratio. The formula for calculating the QP ratio characteristic value is as follows:

[0082]

[0083] Among them, med(Q) i / P i ) represents the median QP ratio for users, tan(arccos(μ) pf,ph ()) is the phase QP ratio center

[0084] Fourthly, frequency mode feature values ​​are used. The power timing of different phases differs in the frequency domain (e.g., the low-frequency component of the power of user A phase is more prominent). Frequency features are extracted by FFT: the user power shape vector is transformed by FFT to obtain the frequency domain amplitude vector; the first 8 frequency components (accounting for more than 90% of the total energy) are retained and normalized to obtain the frequency mode vector; the cosine similarity with the frequency template of each phase is calculated as the frequency mode feature (weight 12%).

[0085] Fifth, the residual shape eigenvalue is calculated by taking the maximum value (weight 12%) of the Pearson correlation coefficient between the user power residual (the difference between the power time series data and the mean) vector and the three-phase residual template. The formula for calculating the residual shape eigenvalue is as follows:

[0086] F resid (i,ph)=max(0,corr((P i -smooth(P i )) bin ,T resid,ph ));

[0087] Among them, Tresid,ph Phase residual template

[0088] Sixth is the cluster consistency feature value, which calculates the average matching degree (weight 8%) of the top five features between a user and other users in the same branch. The formula for calculating the cluster consistency feature value is as follows:

[0089]

[0090] Among them, C i The cluster to which user i belongs.

[0091] The six-dimensional features are normalized and mapped to the [0,1] interval using the min-max function to form a multi-dimensional electrical feature vector for each user.

[0092] As an example of an embodiment of the present invention, the step of determining the phase similarity matrix based on the multi-dimensional electrical feature vector specifically involves: weighting and fusing multiple normalized feature values ​​in the multi-dimensional electrical feature vector based on a third preset weight ratio to obtain the initial assignment probability of each user to each phase; encoding each initial assignment probability to obtain an initial phase vector; and using the normalized power shape correlation feature value, power factor matching feature value, and ratio feature value in the multi-dimensional electrical feature vector, calculating the feature similarity between any two users within the same branch using a cosine similarity function to obtain the phase similarity matrix.

[0093] In this embodiment, three core features are selected from the normalized feature vector: power shape correlation, power factor matching degree, and QP ratio. The cosine similarity function is used to calculate the feature similarity between any two users within the same branch, and a phase similarity matrix S∈R with dimension U×U is constructed. U×U (U represents the number of users within the same branch, and the diagonal elements of the matrix are 1, indicating that the similarity between users is 1). The formula for calculating the similarity matrix is ​​as follows:

[0094]

[0095] Where cosine(·) is the cosine similarity function, and the similarity matrix S∈R U×U (U is the number of users in the branch), and the diagonal element S(i,i) = 1.

[0096] The similarity matrix is ​​normalized to obtain the probability transition matrix, which is calculated using the following formula:

[0097]

[0098] Among them, P transThis represents the probability that user i propagates a feature to user j, used for information transmission on the graph.

[0099] A user phase association graph is constructed based on the probability transition matrix: nodes in the graph represent each user within the same branch, and the weight of the edge between nodes is the similarity value between the corresponding two users in the phase similarity matrix. The higher the similarity, the larger the edge weight, indicating a stronger association between the phase affiliations of the two users. The user phase probability distribution is iteratively optimized using a graph smoothing algorithm. The formula for calculating the phase probability distribution is as follows:

[0100]

[0101] Where w k The initial phase vector OneHot0∈R is obtained by encoding the initial probability distribution into six-dimensional feature weights. U×3 .

[0102] Perform multiple rounds of graph smoothing iterations, updating the phase vector through the transition matrix. The formula for calculating the transition matrix is ​​as follows:

[0103] OneHot k+1 =αOneHot k +(1-α)·P trans OneHot k ;

[0104] Where α = 0.6 is the smoothing coefficient (balancing the characteristics of the product itself and the characteristics of the propagation from its neighbors).

[0105] During the iteration process, the user's phase probability continuously "narrows" towards similar neighboring users, eliminating isolated outliers. Finally, the phase position confidence (i.e., the maximum probability value of a user belonging to a certain phase) is calculated for each user on the smoothed phase vector. For users with a confidence score <0.15, their initial multi-dimensional electrical feature scores are combined with the smoothed phase probability to recalculate the assignment probability. Ultimately, each user is assigned to the phase with the highest probability (phase A, phase B, or phase C), obtaining the access phase assignment results for users within each branch line.

[0106] Step S103: Integrate the branch line division results and the access phase assignment results, and determine the topology of the low-voltage distribution network area corresponding to the target transformer based on the association relationship between each user and the target transformer.

[0107] In this embodiment, the "user-branch" relationship obtained by branch clustering and the "user-phase" relationship obtained by phase identification are integrated, and combined with the known "user-transformer" relationship, to form a complete low-voltage distribution network area topology structure containing three levels of association.

[0108] As an example of an embodiment of the present invention, determining the target number of branches based on the base matrix and the coefficient matrix specifically involves: obtaining the user number limit corresponding to each branch; determining the range of the number of branches based on the user number limit and the total number of users of the target transformer; calculating the profile coefficient for each branch within the range of the number of branches based on each of the assigned weight vectors, and calculating the sparsity by calculating the proportion of zero elements in the base matrix; performing a weighted summation of the profile coefficient and sparsity for each branch based on a second preset weight ratio to obtain the comprehensive score corresponding to each branch, and taking the branch with the highest comprehensive score as the target number of branches.

[0109] In this embodiment, the user-event matrix is ​​decomposed into a basis matrix and a coefficient matrix based on NMF, satisfying the matrix decomposition relation A≈WH (basis matrix W∈R). U×k The coefficient matrix H∈R k×E The first step involves determining the optimal number of candidate branches, where each column of the base matrix represents the branch feature vector, and each row represents the user's affiliation weight to each branch. The second step involves determining the range of candidate branches based on the actual branch design specifications for each transformer station (5-20 users per branch). The third step involves calculating the silhouette coefficient (measuring cluster compactness and separation) and sparsity (measuring user affiliation concentration) for each candidate branch number k. The optimal branch number is selected by using the comprehensive evaluation index Score(k) = Silhouette(k) + 0.1 × Sparsity(k) (SilhouetteScore: measures the compactness and separation of the cluster, with a value range of [-1,1], the closer the value is to 1, the better the clustering effect. Sparsity: measures the sparsity of the base matrix, with a value range of [0,1], the larger the value, the more concentrated the user affiliation).

[0110] like Figure 2 As shown, based on the above method embodiments, corresponding system embodiments are provided;

[0111] One embodiment of the present invention provides a low-voltage distribution network transformer area topology identification system 200, including: a branch clustering identification module 201, a phase attribution identification module 202, and a topology construction module 203;

[0112] The branch clustering identification module 201 is used to determine the correlation matrix based on the power data of the target transformer and the power response intensity corresponding to each user, and to determine multiple power similarity matrices based on the correlation matrix and the voltage feature data corresponding to each user. The power similarity matrices are weighted and fused to obtain a multi-dimensional similarity matrix. The branch line division result is obtained by performing multiple consensus spectrum clustering analysis and confidence analysis on the multi-dimensional similarity matrix. The correlation matrix is ​​used to characterize the correspondence between users and events.

[0113] The phase attribution identification module 202 is used to extract multi-dimensional electrical feature vectors for each user under the same branch in the branch line division result, determine the phase similarity matrix based on the multi-dimensional electrical feature vectors, construct a user phase association map based on the phase similarity matrix, and iteratively optimize the phase probability distribution of each user in the user phase association map through a graph smoothing algorithm to determine the access phase attribution of users in each branch line and obtain the access phase attribution result.

[0114] The topology construction module 203 is used to integrate the branch line division results and the access phase assignment results, and determine the topology of the low-voltage distribution network area corresponding to the target distribution transformer based on the association relationship between each user and the target distribution transformer.

[0115] It is understood that the above system embodiments correspond to the method embodiments of the present invention, and can implement the low-voltage distribution network topology identification method provided by any of the above method embodiments of the present invention.

[0116] It should be noted that the system embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0117] For ease of description and brevity, the system embodiments of the present invention include all the implementation methods described in the above embodiments of the low-voltage distribution network area topology identification method, and will not be repeated here.

[0118] Based on the above embodiments of the low-voltage distribution network area topology identification method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the low-voltage distribution network area topology identification method of any embodiment of the present invention.

[0119] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0120] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0121] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting various parts of the terminal device via various interfaces and lines.

[0122] Based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the low-voltage distribution network area topology identification method described in any of the above-described method embodiments of the present invention.

[0123] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0124] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for identifying the topology of a low-voltage distribution network, characterized in that, include: The correlation matrix is ​​determined based on the power data of the target transformer and the power response intensity of each user. Multiple power similarity matrices are determined based on the correlation matrix and the voltage characteristic data of each user. The power similarity matrices are weighted and fused to obtain a multi-dimensional similarity matrix. The branch line division results are obtained by performing multiple consensus spectrum clustering analysis and confidence analysis on the multi-dimensional similarity matrix. The correlation matrix is ​​used to characterize the correspondence between users and events. For each user under the same branch in the branch line division result, a multi-dimensional electrical feature vector is extracted, a phase similarity matrix is ​​determined based on the multi-dimensional electrical feature vector, a user phase association map is constructed based on the phase similarity matrix, and the phase probability distribution of each user in the user phase association map is iteratively optimized by a graph smoothing algorithm to determine the access phase affiliation of users in each branch line, and the access phase affiliation result is obtained. By integrating the branch line division results and the access phase assignment results, and based on the association relationship between each user and the target transformer, the topology of the low-voltage distribution network area corresponding to the target transformer is determined.

2. The low-voltage distribution network topology identification method as described in claim 1, characterized in that, The determination of the correlation matrix based on the distribution power data of the target transformer and the power response intensity of each user is specifically as follows: Event detection is performed based on the active power time-series data in the power data of the distribution transformer, and a preset number of events are determined according to the detection results to obtain a global event set of the distribution transformer. For each event in the global event set, based on a preset event window, the mean of the absolute value of the power difference for each user within the preset event window is calculated as the power response intensity of each user to each event; The response intensity matrix is ​​determined based on the power response intensity of all users. The response intensity matrix is ​​then subjected to moving average processing and normalization processing to obtain the correlation matrix.

3. The low-voltage distribution network topology identification method as described in claim 1, characterized in that, The process involves determining multiple similarity matrices based on the correlation matrix and the voltage feature data corresponding to each user, and then weighting and fusing these similarity matrices to obtain a multi-dimensional similarity matrix. Specifically, the similarity matrix includes a basis matrix cosine similarity matrix, a voltage residual correlation matrix, a radial basis similarity matrix, and a Jaccard similarity matrix. The correlation matrix is ​​decomposed using a nonnegative matrix factorization method to obtain a basis matrix and a coefficient matrix. Each column of the basis matrix represents a branch feature vector, and each row represents the weight vector of each user's association with each branch. Based on the coefficient matrix, calculate the cosine similarity of the weight vectors to which any two users belong in the base matrix to obtain the cosine similarity matrix of the base matrix. Based on the user voltage time series data of each user in the voltage characteristic data and the distribution transformer voltage time series data of the target distribution transformer, calculate the voltage residual time series sequence of each user, and then calculate the Pearson correlation coefficient of the voltage residual time series sequences of any two users to obtain the voltage residual correlation matrix. Based on the radial basis function, the nonlinear similarity between any two users is calculated for the power response intensity of each user, and the radial basis similarity matrix is ​​obtained. By comparing the power response intensity of each user with a preset threshold, it is determined whether each user responds, thus obtaining a set of user response events. Based on the set of user response events, the Jaccard similarity coefficient of any two sets of user response events is calculated, thus obtaining the Jaccard similarity matrix. The basis matrix cosine similarity matrix, the voltage residual correlation matrix, the radial basis similarity matrix, and the Jaccard similarity matrix are weighted and fused according to a first preset weight ratio to obtain the multi-dimensional similarity matrix.

4. The low-voltage distribution network topology identification method as described in claim 3, characterized in that, The branch line division results are obtained by performing multiple consensus spectrum clustering analyses and confidence analyses on the multi-dimensional similarity matrix, specifically as follows: The target number of branches is determined based on the base matrix and the coefficient matrix. The target number of branches is used as the number of clusters. The spectral clustering algorithm is used to perform clustering analysis with different random seeds for a preset number of clusters. The branch affiliation label of each user in each cluster is recorded. For each user, the number of times each user shares the same branch affiliation label with other users in multiple clusterings is counted. The ratio of the number of times any two users share the same branch affiliation label to the preset number of clusterings is used as the element value in the consensus matrix to obtain the consensus matrix. Based on the consensus matrix, the cluster confidence of each user is calculated. For users to be classified whose cluster confidence is less than a preset confidence threshold, the feature vector of each user is calculated based on the voltage residual time series and the power response intensity. The branch feature center corresponding to each branch is calculated based on the mean of the feature vectors of all users in each branch. The similarity between the feature vector of the user to be classified and the feature center of each branch is calculated. The user to be classified is then reassigned to the branch with the highest similarity. The number of users in each branch is counted, and users in branches with a number of users less than a preset threshold are merged into the neighboring branches with the most similar features to obtain the branch line division result.

5. The low-voltage distribution network topology identification method as described in claim 3 or 4, characterized in that, The determination of the target branch number based on the base matrix and the coefficient matrix specifically involves: Obtain the user number limit corresponding to each branch, and determine the branch number range based on the user number limit and the total number of users of the target transformer; For each branch number within the specified branch number range, the silhouette coefficient is calculated based on each of the specified attribution weight vectors, and the sparsity is obtained by calculating the proportion of zero elements in the basis matrix. Based on the second preset weight ratio, the contour coefficient and sparsity of each branch number are weighted and summed to obtain the comprehensive score corresponding to each branch number. The branch number corresponding to the highest comprehensive score is taken as the target branch number.

6. The low-voltage distribution network topology identification method as described in claim 2, characterized in that, For each user under the same branch in the branch line division result, a multi-dimensional electrical feature vector is extracted, specifically as follows: Based on the active power time-series data corresponding to each user, the power shape vector of each user is determined, the first Pearson correlation coefficient between the power shape vector and the three-phase power shape template is calculated, and the maximum value of the first Pearson correlation coefficient is taken as the power shape correlation feature value. Acquire reactive power time-series data, calculate the median power factor for each user based on the active power time-series data and the reactive power time-series data, calculate the deviation between the median power factor and the center of the three-phase power factor, and take the minimum value of the deviation as the power factor matching degree feature value. Calculate the median of the ratio of the active power time-series data to the reactive power time-series data, calculate the deviation of the median from the center value of the three ratios, and take the minimum value of the deviation as the characteristic value of the ratio. The active power time-series data is subjected to a fast Fourier transform to obtain a frequency domain amplitude vector. The cosine similarity between the frequency domain amplitude vector and the three-phase frequency domain template is calculated, and the maximum value of the cosine similarity is taken as the frequency mode feature value. Calculate the difference between the active power time-series data and the power mean, determine the power residual vector based on the difference, and calculate the second Pearson correlation coefficient between the power residual vector and the three-phase residual template. Take the maximum value of the second Pearson correlation coefficient as the residual shape feature value. Based on the branch line division results, the mean feature matching degree between the user and all other users in the same branch is calculated as the clustering consistency feature value. The mean feature matching degree is the mean of the power shape correlation feature value, the power factor matching degree feature value, the ratio feature value, the frequency mode feature value, and the residual shape feature value. The power shape correlation feature value, the power factor matching feature value, the ratio feature value, the frequency mode feature value, the residual shape feature value, and the cluster consistency feature value are normalized to obtain the multi-dimensional electrical feature vector.

7. The low-voltage distribution network topology identification method as described in claim 1, characterized in that, The step of determining the phase similarity matrix based on the multi-dimensional electrical feature vector is specifically as follows: Based on the third preset weight ratio, the normalized feature values ​​in the multi-dimensional electrical feature vector are weighted and fused to obtain the initial assignment probability of each user to each phase, and the initial assignment probability is encoded to obtain the initial phase vector. Based on the normalized power shape correlation feature value, power factor matching degree feature value and ratio feature value in the multi-dimensional electrical feature vector, the feature similarity between any two users in the same branch is calculated using the cosine similarity function to obtain the phase similarity matrix.

8. A low-voltage distribution network area topology identification system, characterized in that, include: Branch clustering identification module, phase attribution identification module, and topology construction module; The branch clustering identification module is used to determine the correlation matrix based on the power data of the target transformer and the power response intensity corresponding to each user, and to determine multiple power similarity matrices based on the correlation matrix and the voltage feature data corresponding to each user. The power similarity matrices are weighted and fused to obtain a multi-dimensional similarity matrix. The branch line division result is obtained by performing multiple consensus spectrum clustering analysis and confidence analysis on the multi-dimensional similarity matrix. The correlation matrix is ​​used to characterize the correspondence between users and events. The phase attribution identification module is used to extract multi-dimensional electrical feature vectors for each user under the same branch in the branch line division result, determine the phase similarity matrix based on the multi-dimensional electrical feature vectors, construct a user phase association map based on the phase similarity matrix, and iteratively optimize the phase probability distribution of each user in the user phase association map through a graph smoothing algorithm to determine the access phase attribution of users in each branch line and obtain the access phase attribution result. The topology construction module is used to integrate the branch line division results and the access phase assignment results, and determine the topology of the low-voltage distribution network area corresponding to the target transformer based on the association relationship between each user and the target transformer.

9. A terminal device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the low-voltage distribution network area topology identification method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the low-voltage distribution network area topology identification method as described in any one of claims 1-7.

Images