Low-voltage power distribution area topology dimension reduction and variable house identification method and system

Abstract

The application provides a low-voltage power distribution area topology dimension reduction and transformer identification method and system, relates to the technical field of low-voltage power distribution network topology identification, and comprises the following steps: acquiring voltage time series data, node identification information and phase information of each node in a low-voltage power distribution area; identifying a healthy phase set of each node based on the voltage time series data; based on the common healthy phase set of any two nodes, constructing a general electrical distance between the two nodes; based on the general electrical distance, clustering the nodes by using a two-stage agglomerative hierarchical clustering to obtain a subdivided cluster; for each subdivided cluster, constructing a cluster-level voltage centroid based on the voltage time series data of the nodes in the subdivided cluster; based on the general electrical distance between the cluster-level voltage centroid and a transformer measurement object, determining the transformer to which the subdivided cluster belongs, and completing transformer identification. The application can achieve good topology dimension reduction effect, the transformer identification result has strong interpretability, and is suitable for engineering deployment.

Description

Technical Field

[0001] This invention relates to the field of low-voltage distribution network topology identification technology, specifically to a method and system for low-voltage distribution substation topology dimensionality reduction and substation identification based on steady-state healthy phase and two-stage hierarchical clustering, and more particularly to a method and system for low-voltage distribution substation topology dimensionality reduction and substation identification. Background Technology

[0002] With the continuous improvement of the digitalization and intelligence level of distribution networks, accurate topology identification of low-voltage distribution substations has become an important foundation for line loss analysis, power flow calculation, fault location, distributed generation access analysis, and lean management of distribution assets. Especially in low-voltage substations, frequent changes in load cutover, equipment maintenance, branch adjustments, and user access lead to discrepancies between the actual operating topology and the archived topology, making it imperative to use data-driven methods to automatically identify topology relationships.

[0003] Existing low-voltage distribution transformer area topology identification technologies can be broadly divided into two categories: one is based on physical mechanisms or network models, and the other is based on the statistical characteristics of measurement data.

[0004] Physical mechanism or model-driven methods typically rely on line parameters, node impedances, power flow equations, or Kirchhoff's laws for topology inversion. While these methods have a clear theoretical foundation, they often face the following problems in practical low-voltage distribution areas: First, they are highly dependent on prior parameters. Line parameters, wiring methods, phase information, and equipment status in low-voltage distribution areas are often missing or inaccurate, making it difficult to reliably implement identification methods based on accurate physical models. Second, the electrical distance between nodes in low-voltage distribution networks is too small. The physical and electrical distances between nodes in low-voltage distribution areas are usually short, and the node voltage amplitudes are very similar, making traditional methods that use voltage drop characteristics to determine parent-child relationships or path relationships insufficient in their discriminative power. Third, instantaneous measurements are difficult to synchronize. Actual smart meters often suffer from inconsistent timestamps, asynchronous sampling, missing measurements, and noise in their current and power measurements, resulting in poor engineering applicability of methods that directly rely on instantaneous quantities to satisfy Kirchhoff's laws for verification.

[0005] Data-driven methods typically build identification models based on the correlation, similarity, or statistical structure between voltage, current, power, or energy sequences. While these methods reduce reliance on precise network parameters, existing methods still have the following shortcomings: First, they lack a unified mechanism for handling phase differences and abnormal phases. In actual low-voltage distribution areas, different nodes may only collect partial phase data, or some phases may be in an abnormal state for a long time. Existing methods often assume that the phases between nodes are consistent, making it difficult to stably calculate the similarity between nodes under conditions of missing phases, out-of-phase phases, and abnormal measurements. Second, the clustering process struggles to balance noise resistance and purity. Single clustering algorithms are often either easily affected by isolated outliers, leading to cluster expansion distortion, or they become overly subdivided, resulting in too many clusters and losing their dimensionality reduction value. Existing solutions rarely balance topology dimensionality reduction efficiency with the intra-cluster homogeneity required for subsequent identification. Third, the clustering results are not sufficiently integrated with subsequent refined identification. Existing solutions rarely use the feature of close electrical distance between nodes within a cluster for refined identification, but only for transformer identification, phase identification, etc., failing to provide the prerequisite for cluster-level energy matching in the subsequent line-to-line identification process.

[0006] Therefore, the existing technology needs a topology reduction and variable household identification method that can adapt to engineering characteristics such as similar voltage levels, inconsistent phases, and unstable measurement quality in low-voltage distribution area nodes. Under the premise of relying only on the voltage time series data of smart meters, a robust, interpretable, and applicable topology reduction and variable household identification scheme can be realized. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method and system for low-voltage distribution transformer area topology reduction and variable user identification.

[0008] According to the present invention, a method and system for topology reduction and variable subscriber identification of low-voltage distribution substations are provided, the scheme of which is as follows: Firstly, a method for topology reduction and variable subscriber identification in low-voltage distribution substations is provided, the method comprising: Step S1: Obtain the voltage timing data, node identification information, and phase information of each node in the low-voltage distribution substation area; Step S2: Based on the voltage timing data, node identification information, and phase information, identify the healthy phase set of each node; Step S3: Based on the common health phase set of any two nodes, construct the general electrical distance between the two nodes; Step S4: Based on the general electrical distance, the nodes are clustered using two-stage agglomerative hierarchical clustering to obtain sub-clusters. The two stages include a first-stage clustering and a second-stage clustering performed sequentially. Step S5: For each sub-cluster, construct the cluster-level voltage centroid based on the voltage timing data of the nodes within that sub-cluster; Step S6: Based on the general electrical distance between the cluster-level voltage centroid and the transformer measurement object, determine the transformer to which the sub-cluster belongs, and complete the transformer user identification.

[0009] Preferably, in step S2, identifying the healthy phase set of each node includes: For each node, calculate the average value of the voltage time series of that node under each phase; When the average value of the voltage time series under a certain phase is within the preset healthy phase judgment interval, that phase is determined to be the healthy phase of that node; The set of healthy phases of a node consists of all healthy phases of that node.

[0010] Preferably, in step S3, constructing the universal electrical distance between the two nodes includes: When the two nodes share a common healthy phase, for each phase in the set of common healthy phases, calculate the voltage timing Pearson correlation coefficient for the two nodes in that phase; based on the mean of the Pearson correlation coefficient over the set of common healthy phases, determine the general electrical distance: When users With users When a common health phase exists, that is When, its general electrical distance is defined as

[0011] in, Indicates user With users At parting The time-series voltage Pearson correlation coefficient, The number of phases representing the set of common healthy phases; When the two nodes do not have a common healthy phase, the general electrical distance is set to infinity.

[0012] Preferably, in step S4, the first-stage clustering uses uniform linkage as the inter-cluster distance metric, where uniform linkage is defined as:

[0013] in, and There are two candidate clusters. and These represent the number of nodes within the corresponding cluster; Set coarse segmentation cutoff threshold When the average link distance between two candidate clusters to be merged is greater than the preset coarse segmentation truncation threshold, merging is stopped, and coarse clusters are obtained.

[0014] Preferably, in step S4, the second-stage clustering is performed within each coarse cluster obtained from the first-stage clustering, and full linkage is used as the inter-cluster distance metric:

[0015] Set the subdivision cutoff threshold When the full link distance between two candidate clusters to be merged is greater than a preset subdivision truncation threshold, merging is stopped, and the subdivided clusters are obtained.

[0016] Preferably, in step S5, constructing the cluster-level voltage centroid includes: For any sub-cluster In a certain phase At a certain sampling time The voltage centroid under the given conditions is defined as:

[0017] in, Indicates subdivision cluster Zhong Xiang Bie The set of nodes that are determined to be in a healthy state. Represents a node At parting ,time The voltage measurement value.

[0018] Preferably, in step S6, determining the transformer to which the subdivision cluster belongs includes: Calculate the general electrical distance between the cluster-level voltage centroid of the subdivided cluster and each transformer measurement object; For each sub-cluster Take the minimum value among the general electrical distances between it and each transformer measurement object. Set threshold They believe that as long as The distance to a transformer is within the range Within, determine the sub-cluster. It is assigned to the corresponding transformer measurement object, and all nodes within the sub-cluster are assigned to that transformer.

[0019] Preferably, after step S1 and before step S2, a step of preprocessing the voltage timing data is further included, the preprocessing including: The voltage timing data is organized according to a unified time axis; The outliers, missing values, and invalid measurements in the voltage time series data are cleaned to form standardized voltage time series data.

[0020] Secondly, a low-voltage distribution substation topology reduction and variable user identification system is provided, including: The data reading module is used to acquire voltage timing data, node identification information, and phase information of each node in the low-voltage distribution substation area; A healthy phase identification module is used to identify the set of healthy phases for each node based on the voltage timing data; The electrical distance calculation module is used to construct the general electrical distance between two nodes based on the common healthy phase set of any two nodes; A two-stage clustering module is used to cluster nodes based on the general electrical distance using a two-stage agglomerative hierarchical clustering method to obtain subdivided clusters. The two stages include a first-stage clustering and a second-stage clustering executed sequentially. The cluster centroid construction module is used to construct a cluster-level voltage centroid for each sub-cluster based on the voltage timing data of the nodes within that sub-cluster. The transformer identification module is used to determine the transformer to which the sub-cluster belongs based on the general electrical distance between the cluster-level voltage centroid and the transformer measurement object, thereby completing the transformer identification.

[0021] Preferably, the system further includes: The data preprocessing module is used to perform time axis alignment, outlier removal, and missing value processing on the voltage time series data before the health phase identification module performs identification, so as to form standardized voltage time series data. The results output module is used to output the topology dimensionality reduction results, node cluster results, and variable household identification results.

[0022] Compared with the prior art, the advantages of the present invention are as follows: 1. Adapting to the engineering characteristics of similar voltage levels in low-voltage distribution areas; This invention does not rely on traditional voltage drop amplitude differences, but only uses voltage data for electrical distance calculation and topology dimensionality reduction, avoiding the problem of insufficient distinguishability caused by similar voltage amplitudes of nodes in low-voltage distribution areas, and can stably complete topology identification under conditions of similar voltage levels; 2. It can uniformly handle the problems of different phases and abnormal phases; by introducing a steady-state healthy phase and common healthy phase set mechanism, this invention can stably complete distance calculation and cluster analysis when the phases at different nodes are inconsistent, some phases are abnormal or missing, avoiding distance distortion caused by phase inconsistency, and uniformly handling the actual scenarios where single phase, two phase, three phase and phase abnormality coexist. 3. Balancing noise resistance and cluster purity: A two-stage agglomerative hierarchical clustering strategy is adopted, consisting of coarse division with uniform linkage and fine division with full linkage. In the first stage, uniform linkage is used to reduce the interference of individual outliers or local noise on the clustering process by utilizing the average distance of all node pairs between clusters, thus separating regions with significant differences in electrical characteristics. In the second stage, full linkage is used to constrain the distance between the farthest members within a cluster, resulting in higher consistency within the generated fine clusters. This not only suppresses the erroneous expansion caused by outliers but also ensures strong correlation within the fine clusters, making it suitable for subsequent hierarchical topology identification. 4. It has a good dimensionality reduction effect; the present invention replaces the original nodes with subdivided clusters to participate in subsequent analysis, which can significantly reduce the computational scale of subsequent line identification. For example, in a specific embodiment, 267 effective nodes are compressed into 22 subdivided clusters, realizing effective topological dimensionality reduction of the original node set. 5. The identification results are highly interpretable; the present invention matches the cluster voltage centroid with the transformer measurement object, the identification chain is clear, and it is easy to explain from an engineering perspective why a node belongs to a certain transformer. The identification results have good engineering interpretability. 6. The system is simple to implement and suitable for engineering deployment. This invention mainly relies on the existing voltage timing data of smart meters and feeder cabinets. It does not require additional high-cost synchronous sampling devices or precise line parameters, and has strong engineering practical value. It is easy to deploy and implement on the existing digital infrastructure of low-voltage distribution substations.

[0023] Other beneficial effects of the present invention will be explained in detail through the introduction of specific technical features and technical solutions in specific embodiments. Those skilled in the art should be able to understand the beneficial technical effects brought about by these technical features and technical solutions through the introduction of these technical features and technical solutions. Attached Figure Description

[0024] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart of the variable household identification algorithm according to an embodiment of the present invention; Figure 2 This is a low-voltage distribution network topology diagram according to an embodiment of the present invention. Detailed Implementation

[0025] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the scope of protection of the present invention.

[0026] This invention provides a method for topology reduction and transformer identification in low-voltage distribution transformer areas. It addresses existing problems in low-voltage distribution transformer area topology identification methods, such as strong reliance on prior parameters, insufficient discrimination due to small electrical distances between nodes, difficulty in synchronizing instantaneous measurements, lack of a unified processing mechanism for phase differences and abnormal phases, difficulty in balancing noise resistance and purity during clustering, and insufficient connection between clustering results and subsequent refined identification. The aim is to achieve robust, interpretable topology reduction and transformer identification that can be used for subsequent transformer identification, relying solely on smart meter voltage time-series data. (Refer to...) Figure 1 As shown, this invention addresses the tree-like structure composed of transformers, primary nodes, and secondary nodes in low-voltage distribution substations. It constructs a universal electrical distance for different phase scenarios by uniformly preprocessing the node voltage time series. Then, based on a two-stage agglomerative hierarchical clustering strategy of coarse division for uniform links and fine division for full links, it achieves node cluster partitioning, completing topology dimensionality reduction. Subsequently, it constructs cluster-level voltage centroids and matches them with transformer voltage time series data, thereby achieving the corresponding identification of transformers and node clusters, i.e., completing transformer customer identification. The method specifically includes steps S1 to S6, as follows: Step S1: Obtain the voltage timing data, node identification information, and phase information of each node in the low-voltage distribution substation area; Among them, node voltage time sequence data is used to represent the voltage measurement values ​​of each node in the low-voltage distribution area at each sampling time and each phase. It is usually collected by smart meters deployed at the low-voltage side of the transformer, feeder cabinet, branch box and user meter box at fixed sampling intervals. Node identification information is used to uniquely identify each node and its corresponding measurement object, including node number, node type (transformer, primary node, secondary node, etc.), and the identification of the measuring device to which the node belongs. Phase information is used to indicate the measurement phase types that each node has, including phase A, phase B, phase C and their combinations. Different nodes may only have the measurement capability for some phases.

[0027] In a preferred embodiment, after step S1 and before step S2, a preprocessing step for the voltage time series data is included. The preprocessing includes: organizing the voltage time series data along a unified time axis to ensure that voltage measurements at different nodes and in different phases are aligned in the time dimension; cleaning outliers, missing values, and invalid measurements from the voltage time series data, where outliers include measurements that significantly exceed physically reasonable ranges, missing values ​​include measurements missing due to communication interruptions or equipment failures, and invalid measurements include zero values, constant values, or obvious noise sequences, etc., to form standardized voltage time series data that can be used for subsequent correlation analysis. Preprocessing effectively improves the stability and accuracy of subsequent healthy phase identification, electrical distance calculation, and cluster analysis.

[0028] Step S2: Based on voltage timing data, node identification information, and phase information, identify the set of healthy phases for each node; Since different nodes in the low-voltage distribution area may have incomplete phases, some phase distortions, or abnormal offsets, this invention introduces the concept of "steady-state healthy phase".

[0029] For each user The average value of each of the three-phase voltage time series is calculated to identify the set of healthy phases. Assume the user... At parting The average time-series voltage is When it satisfies At that time, this phase was considered a healthy phase. This led to the construction of a user... Health Collection .

[0030] For any two users and Define their common health set as .

[0031] By calculating similarity only within the common healthy phase range of two nodes, this invention can uniformly handle real-world scenarios where single-phase, two-phase, three-phase, and phase-specific anomalies coexist, thereby avoiding distance distortion caused by phase inconsistency.

[0032] Step S3: Based on the common health phase set of any two nodes, construct the general electrical distance between the two nodes; Constructing a general electrical distance between two nodes includes: When the two nodes have a common healthy phase, for each phase in the set of common healthy phases, calculate the voltage timing Pearson correlation coefficient of the two nodes in that phase; based on the mean of the Pearson correlation coefficient over the set of common healthy phases, determine the general electrical distance.

[0033] Specifically, in order to characterize the degree of voltage correlation between users on the healthy phase, users are defined. With users General Electric distance.

[0034] When users With users When a common health phase exists, that is When, its general electrical distance is defined as

[0035] in, Indicates user With users At parting The time-series voltage Pearson correlation coefficient, This represents the number of phases in the common healthy phase set; since the Pearson correlation coefficient ranges from... Therefore, when a common healthy phase exists, the aforementioned electrical distance The range of values ​​is When two nodes do not have a common healthy phase, the general electrical distance is set to infinity to indicate that the two nodes are completely incomparable in terms of comparable phases and to maintain isolation during the clustering process.

[0036] It should be noted that the Pearson correlation coefficient described in this invention is only a preferred embodiment of node similarity measurement. In other embodiments, in addition to the Pearson correlation coefficient, other correlation measurement indicators such as Spearman rank correlation coefficient, Kendall correlation coefficient, and cosine similarity can also be used, all of which are within the scope of protection of this invention.

[0037] By defining the general electrical distance as described above, this invention can uniformly characterize the degree of electrical correlation between nodes in different phase scenarios, providing a unified distance metric basis for subsequent cluster analysis.

[0038] Step S4: Based on the general electrical distance, two-stage agglomerative hierarchical clustering is used to cluster the nodes to obtain sub-clusters. The two stages include the first stage clustering and the second stage clustering executed sequentially. This invention uses two-stage agglomerative hierarchical clustering to achieve topological dimensionality reduction. Its core idea is to first use the uniform linkage method with strong noise resistance to complete coarse-grained region isolation, and then use the fully linkage method with stricter constraints to complete fine-grained purification.

[0039] Phase 1: Coarse Linkage Analysis In the first phase, the system treats all nodes as clusters, performs agglomerative hierarchical clustering based on a general electrical distance matrix, and uses uniform linkage as the definition of inter-cluster distance:

[0040] in, and There are two candidate clusters. and These represent the number of nodes within the corresponding cluster; Set coarse segmentation cutoff threshold When the average link distance between two candidate clusters to be merged is greater than the preset coarse segmentation truncation threshold, merging is stopped, and coarse clusters are obtained.

[0041] The average distance between all node pairs in a cluster can be used to reduce the interference of individual outliers or local noise on the clustering process and separate regions with significant differences in electrical characteristics.

[0042] Phase Two: Full Linkage Segmentation After obtaining the coarse clustering results, a second-stage clustering is performed within each coarse cluster, using full connectivity as the inter-cluster distance:

[0043] Set the subdivision cutoff threshold When the full link distance between two candidate clusters to be merged is greater than the preset subdivision truncation threshold, merging is stopped, and subdivided clusters are obtained.

[0044] Fully linked clusters are constrained by the distance to the furthest member within the cluster, resulting in higher consistency within the sub-clusters and making it easier to ensure that nodes within the cluster have a higher voltage correlation, thereby improving the accuracy of subsequent variable household identification.

[0045] In this embodiment, the coarse threshold is determined using the controlled variable method. With segmentation threshold Sensitivity analysis was conducted to verify the rationality of its selection. The analysis showed that... When the clustering quality remains constant within the effective range, it proves that it only plays a preliminary role in isolating regions of extreme differences and is not sensitive. As a core sensitive parameter, its value needs to be balanced between homogeneity (intra-cluster purity) and the number of node clusters (dimensionality reduction efficiency).

[0046] Step S5: For each sub-cluster, construct the cluster-level voltage centroid based on the voltage timing data of the nodes within that sub-cluster; After completing the two-stage clustering, each sub-cluster represents a group of nodes with highly homogeneous voltage timing. To further map this node cluster to a transformer, this invention constructs a cluster-level voltage centroid for each sub-cluster.

[0047] For any sub-cluster In a certain phase At a certain sampling time The voltage centroid under the given conditions is defined as:

[0048] in, Indicates subdivision cluster Zhong Xiang Bie The set of nodes that are determined to be in a healthy state. Represents a node At parting ,time The voltage measurement value.

[0049] Step S6: Based on the general electrical distance between the cluster-level voltage centroid and the transformer measurement object, determine the transformer to which the sub-cluster belongs, and complete the transformer user identification.

[0050] In step S6, determining the transformer to which the subdivision cluster belongs includes: Calculate the general electrical distance between the cluster-level voltage centroid of the subdivided cluster and each transformer measurement object; For each sub-cluster Take the minimum value among the general electrical distances between it and each transformer measurement object. Set threshold They believe that as long as The distance to a transformer is within the range Within, determine the sub-cluster. It is assigned to the corresponding transformer measurement object, and all nodes within the sub-cluster are assigned to that transformer.

[0051] Since the sub-cluster belongs to a certain transformer, all nodes within that sub-cluster also belong to that transformer, thus completing the transformer identification.

[0052] This invention also provides a low-voltage distribution transformer area topology reduction and change-of-service identification system. This system can be implemented by executing the steps of the aforementioned low-voltage distribution transformer area topology reduction and change-of-service identification method. Therefore, those skilled in the art can understand the aforementioned low-voltage distribution transformer area topology reduction and change-of-service identification method as a preferred embodiment of the system. The system specifically includes: The data reading module is used to acquire voltage timing data, node identification information, and phase information of each node in the low-voltage distribution substation area; The data preprocessing module is used to perform time axis alignment, outlier removal, and missing value processing on the voltage time series data before the health phase identification module performs identification, so as to form standardized voltage time series data. A healthy phase identification module is used to identify the set of healthy phases for each node based on the voltage timing data; The electrical distance calculation module is used to construct the general electrical distance between two nodes based on the common healthy phase set of any two nodes; A two-stage clustering module is used to cluster nodes based on the general electrical distance using a two-stage agglomerative hierarchical clustering method to obtain subdivided clusters. The two stages include a first-stage clustering and a second-stage clustering executed sequentially. The cluster centroid construction module is used to construct a cluster-level voltage centroid for each sub-cluster based on the voltage timing data of the nodes within that sub-cluster. The transformer identification module is used to determine the transformer to which the sub-cluster belongs based on the general electrical distance between the cluster-level voltage centroid and the transformer measurement object, thus completing the transformer identification.

[0053] The results output module is used to output the topology dimensionality reduction results, node cluster results, and variable household identification results.

[0054] Next, this specific implementation method will be described in more detail using the scenario of low-voltage distribution transformer area topology identification.

[0055] Reference Figure 2 As shown, in one specific embodiment, a low-voltage distribution substation in a Shanghai industrial park is selected as the research object. This substation contains 272 nodes, including one substation, four transformers, four incoming line cabinets / feeder cabinets, 82 primary nodes, and 181 secondary nodes. The system collects 60 consecutive days of smart meter measurement data from this substation, with a sampling interval of 5 minutes. The measurement content includes at least the phase voltage time series data of each node at each sampling time. The collected raw data is first preprocessed, including time axis unification, outlier removal, missing value handling, and invalid node screening, to form standardized voltage time series data that can be used for subsequent analysis.

[0056] In this embodiment, the substation and four incoming line cabinets do not participate in the node clustering process because they lack complete user-side meter data. Therefore, the final number of effective nodes used for topology dimensionality reduction and transformer-user identification is 267. The system calculates the time mean of the phase voltage time series for each effective node and identifies a set of healthy phases based on a preset voltage range. In this embodiment, if the mean of the phase voltage time series of a certain node is between 210V and 250V, then that phase is determined to be a healthy phase, and a general electrical distance matrix between nodes is constructed.

[0057] After obtaining the general electrical distance matrix, the system performs a two-stage agglomerative hierarchical clustering. The first stage uses a uniformly linked clustering method for coarse segmentation, with a coarse segmentation truncation threshold set to 0.05, to separate node regions with significant differences in electrical characteristics. The second stage further performs fully linked clustering within each coarse cluster, with a subdivision truncation threshold set to 0.01, to form subdivided clusters with highly consistent voltage fluctuations among nodes within each cluster. After these two stages of clustering, this embodiment compresses 267 effective nodes into 22 subdivided clusters, achieving topological dimensionality reduction of the original node set. Nodes in each subdivided cluster exhibit strong correlation in voltage timing characteristics and can serve as high-purity candidate objects for subsequent transformer identification.

[0058] After completing the sub-cluster division, the system constructs a cluster-level voltage centroid for each sub-cluster. Specifically, for any sub-cluster, the cluster voltage centroid at a certain phase and sampling time is taken as the average voltage of all nodes within that sub-cluster that are determined to be healthy phases in that phase. Then, the system compares the electrical distance between the voltage centroids of each sub-cluster and the transformer-side measurement objects. In this embodiment, the transformer-side measurement objects can be characterized by the voltage measurement sequence of the corresponding feeder cabinet. The system calculates the general electrical distance between the centroid of each sub-cluster and each transformer-side measurement object, and obtains the set of transformers to which each sub-cluster may belong based on preset matching rules, thereby completing transformer customer identification.

[0059] In the actual operation results of this embodiment, after determining the transformer user affiliation of the 22 sub-clusters formed by 267 valid nodes, the set of transformers to which the sub-clusters may belong completely matches the correct topology, and the transformer user identification accuracy reaches 100%. In addition, the clustering results show that transformer-side measurement objects with strong electrical coupling can form an independent top-level region in the clustering structure, while each transformer downstream primary node and its related user nodes are divided into one or a few independent sub-clusters. This indicates that the general electrical distance and two-stage clustering strategy constructed in this invention can effectively reflect the electrical topology relationship in the actual transformer area.

[0060] This embodiment demonstrates that, even when the voltage amplitudes of nodes in a low-voltage distribution substation are highly similar, the phases differ, and the measurement quality is not completely stable, the present invention can still stably complete topology dimensionality reduction and variable user identification solely based on voltage time series data, providing reliable clustering information for subsequent line user identification.

[0061] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, enabling the system and its various devices, modules, and units provided in this specific embodiment to be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided in this specific embodiment can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0062] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for topology reduction and variable subscriber identification in low-voltage distribution substations, characterized in that, include: Step S1: Obtain the voltage timing data, node identification information, and phase information of each node in the low-voltage distribution substation area; Step S2: Based on the voltage timing data, node identification information, and phase information, identify the healthy phase set of each node; Step S3: Based on the common health phase set of any two nodes, construct the general electrical distance between the two nodes; Step S4: Based on the general electrical distance, the nodes are clustered using two-stage agglomerative hierarchical clustering to obtain sub-clusters. The two stages include a first-stage clustering and a second-stage clustering performed sequentially. Step S5: For each sub-cluster, construct the cluster-level voltage centroid based on the voltage timing data of the nodes within that sub-cluster; Step S6: Based on the general electrical distance between the cluster-level voltage centroid and the transformer measurement object, determine the transformer to which the sub-cluster belongs, and complete the transformer user identification.

2. The method for topology reduction and variable customer identification of low-voltage distribution transformer areas according to claim 1, characterized in that, In step S2, identifying the healthy phase set of each node includes: For each node, calculate the average value of the voltage time series of that node under each phase; When the average value of the voltage time series under a certain phase is within the preset healthy phase judgment interval, that phase is determined to be the healthy phase of that node; The set of healthy phases of a node consists of all healthy phases of that node.

3. The method for low-voltage distribution transformer area topology reduction and variable user identification according to claim 1, characterized in that, In step S3, constructing the universal electrical distance between the two nodes includes: When the two nodes share a common healthy phase, for each phase in the set of common healthy phases, calculate the voltage timing Pearson correlation coefficient for the two nodes in that phase; based on the mean of the Pearson correlation coefficient over the set of common healthy phases, determine the general electrical distance: When users With users When a common health phase exists, that is When, its general electrical distance is defined as in, Indicates user With users At parting The time-series voltage Pearson correlation coefficient, The number of phases representing the set of common healthy phases; When the two nodes do not have a common healthy phase, the general electrical distance is set to infinity.

4. The method for low-voltage distribution transformer area topology reduction and variable user identification according to claim 3, characterized in that, In step S4, the first-stage clustering uses uniform linkage as the inter-cluster distance metric, where uniform linkage is defined as: in, and There are two candidate clusters. and These represent the number of nodes within the corresponding cluster; Set coarse segmentation cutoff threshold When the average link distance between two candidate clusters to be merged is greater than the preset coarse segmentation truncation threshold, merging is stopped, and coarse clusters are obtained.

5. The method for low-voltage distribution transformer area topology reduction and variable user identification according to claim 4, characterized in that, In step S4, the second-stage clustering is performed within each coarse cluster obtained from the first-stage clustering, and full linkage is used as the inter-cluster distance metric. Set the subdivision cutoff threshold When the full link distance between two candidate clusters to be merged is greater than a preset subdivision truncation threshold, merging is stopped, and the subdivided clusters are obtained.

6. The method for topology reduction and variable customer identification of low-voltage distribution transformer areas according to claim 1, characterized in that, In step S5, constructing the cluster-level voltage centroid includes: For any sub-cluster In a certain phase At a certain sampling time The voltage centroid under the given conditions is defined as: in, Indicates subdivision cluster Zhong Xiang Bie The set of nodes that are determined to be in a healthy state. Represents a node At parting ,time The voltage measurement value.

7. The method for topology reduction and variable customer identification of low-voltage distribution transformer areas according to claim 1, characterized in that, In step S6, determining the transformer to which the subdivision cluster belongs includes: Calculate the general electrical distance between the cluster-level voltage centroid of the subdivided cluster and each transformer measurement object; For each sub-cluster Take the minimum value among the general electrical distances between it and each transformer measurement object. Set threshold They believe that as long as The distance to a transformer is within the range Within, determine the sub-cluster. It is assigned to the corresponding transformer measurement object, and all nodes within the sub-cluster are assigned to that transformer.

8. The method for topology reduction and variable customer identification of low-voltage distribution transformer areas according to claim 1, characterized in that, After step S1 and before step S2, a preprocessing step for the voltage timing data is further included, the preprocessing including: The voltage timing data is organized according to a unified time axis; The outliers, missing values, and invalid measurements in the voltage time series data are cleaned to form standardized voltage time series data.

9. A low-voltage distribution substation topology reduction and variable user identification system, characterized in that, include: The data reading module is used to acquire voltage timing data, node identification information, and phase information of each node in the low-voltage distribution substation area; A healthy phase identification module is used to identify the set of healthy phases for each node based on the voltage timing data; The electrical distance calculation module is used to construct the general electrical distance between two nodes based on the common healthy phase set of any two nodes; A two-stage clustering module is used to cluster nodes based on the general electrical distance using a two-stage agglomerative hierarchical clustering method to obtain subdivided clusters. The two stages include a first-stage clustering and a second-stage clustering executed sequentially. The cluster centroid construction module is used to construct a cluster-level voltage centroid for each sub-cluster based on the voltage timing data of the nodes within that sub-cluster. The transformer identification module is used to determine the transformer to which the sub-cluster belongs based on the general electrical distance between the cluster-level voltage centroid and the transformer measurement object, thereby completing the transformer identification.

10. The low-voltage distribution transformer area topology reduction and variable user identification system according to claim 9, characterized in that, The system also includes: The data preprocessing module is used to perform time axis alignment, outlier removal, and missing value processing on the voltage time series data before the health phase identification module performs identification, so as to form standardized voltage time series data. The results output module is used to output the topology dimensionality reduction results, node cluster results, and variable household identification results.

Images