Low-voltage transformer area topology level identification method and system based on master station big data

Abstract

The invention discloses a main station big data-based low-voltage transformer area topology hierarchy identification method and system, and the method comprises the steps: constructing a multi-dimensional voltage feature vector through the voltage time sequence data, collected by a main station side, of a transformer area general meter and a user ammeter, and forming an electrical feature matrix through feature extraction; and on the basis, judging the station area affiliation and the phase line affiliation of the user electric meter through similarity calculation. Then, generating a stable consensus cluster based on multi-time window clustering and the copolymerization matrix, and mapping the cluster into a virtual electric meter box node; constructing a minimum spanning tree and carrying out directionalization to obtain an initial radial topological structure; further introducing an electrical consistency criterion based on correlation of voltage drop and subtree current, performing multi-candidate structure optimization on the preliminary topology, and finally obtaining an optimal transformer area topology hierarchical structure consistent with a physical rule; high-precision automatic identification of the low-voltage transformer area topology can be realized without manual participation, and effective support is provided for digital operation and maintenance of a power distribution network.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to a method and system for identifying the topology hierarchy of low-voltage distribution areas based on big data from a main station. Background Technology

[0002] With the continuous advancement of intelligent and information-based construction of distribution networks, low-voltage distribution substations have become the core units for power system perception and management. However, due to long historical construction cycles, frequent upgrades, and flexible user access, there are often inconsistencies or even serious gaps between the actual wiring structure and archival data of low-voltage substations. This makes it impossible for the master station system to accurately grasp the true topological connection relationships between each user's meter and the substation's main meter and branch lines. Currently, traditional substation topology identification methods mainly rely on manual on-site verification or modeling based on local pilot projects. These methods are not only time-consuming and costly to implement, but also difficult to achieve dynamic updates and large-scale promotion, and can no longer meet the practical needs of the rapid development of the distribution Internet of Things and refined operation and maintenance management.

[0003] On the other hand, with the widespread deployment of smart meters and electricity information collection systems, the master station has accumulated massive amounts of high-frequency time-series power data, covering multi-dimensional information such as voltage, current, active power, reactive power, and electrical energy from user meters and the main meter for the distribution area. This data contains rich electrical correlation characteristics: for example, voltage fluctuations at monitoring nodes within the same distribution area are highly correlated, voltage changes between different phase lines exhibit specific phase differences, and power transfer relationships at the branch level conform to energy conservation constraints. If, based on the big data resources of the master station, these physical correlation characteristics can be automatically extracted and corresponding mathematical models constructed, it is expected that user meter distribution area identification, phase line allocation judgment, and electrical hierarchy relationship restoration can be achieved without manual intervention, thereby completing the self-learning and dynamic updating of the distribution area topology.

[0004] Existing research and practice have made relevant attempts in voltage similarity analysis, energy balance calculation and graph structure inference, but several key problems remain unsolved: First, existing algorithms mostly rely on single features (such as only voltage correlation or power correlation) for analysis, lacking a fusion judgment mechanism of multi-dimensional electrical features, and are susceptible to measurement noise and load fluctuation interference; Second, the reliability of the identification results is difficult to quantify and verify, and there is a lack of physical constraints and verification mechanisms based on the principle of electrical consistency; Third, a systematic implementation path from data acquisition and processing, topology algorithm calculation to result visualization output has not yet been formed, making it difficult to form a complete closed loop of technical application.

[0005] Therefore, there is an urgent need for a low-voltage transformer area topology hierarchical identification system that can be based on big data from the main station, integrate multi-source electrical characteristics, take into account physical constraints and algorithm optimization, and achieve fully automated processing and visualized output. Summary of the Invention

[0006] The purpose of this invention is to address the problems of unclear topology, missing data, and low efficiency of manual verification in existing low-voltage distribution transformer areas. It proposes a method and system for identifying the topology hierarchy of low-voltage transformer areas based on big data from the master station, realizing an intelligent process of automatically completing transformer area topology identification and hierarchical reconstruction starting from master station data.

[0007] The technical solution of this invention is: In a first aspect, the present invention provides a method for identifying the topology hierarchy of low-voltage transformer areas based on big data from a master station, comprising the following steps: S1. Steps for acquiring electricity consumption data: Obtain the voltage time sequence data of the main meter and each user's meter in the target area on a unified time axis from the main station's electricity consumption information acquisition system. S2. Voltage feature construction steps: Normalize, perform first-order difference operation, determine monotonicity, and calculate local variance based on sliding window for the voltage time series data of each meter in sequence to generate multi-dimensional feature vectors; combine the multi-dimensional feature vectors of all meters to form a voltage feature matrix. S3. Transformer Area Phase Line Identification Step: Based on the voltage feature matrix, calculate the similarity between each user's meter and each candidate transformer area master meter on each phase line; determine the transformer area affiliation of each user's meter according to the similarity, and determine the phase line affiliation of each user's meter within the assigned transformer area. S4. Virtual Node Construction Steps: Within the same transformer area whose affiliation has been determined, perform cluster analysis on the voltage feature matrix of user meters at multiple time windows; construct a co-cluster matrix based on the clustering results of all time windows and extract multiple consensus clusters; map consensus clusters with a size not less than the preset minimum cluster size to a virtual meter box node, and use the mean of the voltage feature matrices of all user meters in the cluster as the representative feature vector of the virtual meter box node; S5. Preliminary Topology Construction Steps: Take the main meter of the transformer area as the root node, and form a node set together with all virtual meter box nodes; calculate the similarity between nodes based on the representative feature vector of each node and convert it into distance; use the distance as the edge weight, perform the minimum spanning tree algorithm on the complete graph formed by the node set to obtain an undirected spanning tree; orient the edges in the undirected spanning tree according to the magnitude relationship of the average voltage of each node, and complete the directional processing starting from the root node to obtain a preliminary radially directed topology structure. S6. Electrical topology optimization steps: For each directed edge in the initial radially directed topology, calculate the voltage drop sequence between parent and child nodes and the subtree current sequence rooted at the child node. On the valid sample set where the voltage drop is greater than a preset threshold, the correlation coefficient between the voltage drop sequence and the subtree current sequence is calculated, and the overall electrical consistency score of the topology is defined based on the correlation coefficient of all directed edges. By adjusting the connection relationships, multiple candidate topologies are generated. The overall electrical consistency score of each candidate topology is calculated, and the candidate topology with the highest score is selected as the final low-voltage distribution area topology hierarchy identification result.

[0008] Furthermore, the generation of multidimensional feature vectors in S2 specifically includes: For the first The meter is at a certain time voltage The first feature component is obtained by performing Min–Max normalization. , ;in, Indicates the meter number, Indicates the time series number. and Electricity meters Minimum and maximum voltage values ​​during the entire observation period; The second feature component is obtained by performing a first-order difference operation on the normalized sequence. , ; Based on the relationship between the normalized voltages of two adjacent points, a monotonicity characteristic is defined, resulting in the third characteristic component. Its value selection rule is: if ,but ;like ,but If the two are equal, then ; Using a preset number of consecutive time points as a sliding window, the second feature component is... Calculate the local variance to obtain the fourth eigencomponent. To characterize the intensity of fluctuations within a short timescale; The first, second, third, and fourth characteristic components are concatenated in chronological order to form the comprehensive voltage characteristic matrix of the meter. .

[0009] Furthermore, S3 specifically includes: S31, For user electricity meters eigenvectors With the phase line of the Gp in the substation eigenvectors on Calculate cosine similarity , Indicates phase lines A, B, and C; S32, Install the user's electricity meter The cosine similarity of the transformer area Gp under phases A, B, and C is weighted and summed according to preset weights to obtain the comprehensive similarity. The preset weights of phases A, B, and C are non-negative real numbers and their sum is 1. S33. Based on the principle of maximizing comprehensive similarity, the user's electricity meter... It is assigned to the area with the highest overall similarity; Within the assigned transformer area, the phase line with the highest cosine similarity is selected as the user's electricity meter. The phase line is assigned to.

[0010] Furthermore, S33 also includes: If the user's electricity meter If the overall similarity with all candidate transformer substations is lower than the preset threshold, the user's meter will be marked as an unidentified object. The phase line identification step for the transformer area is repeated over multiple subsequent time windows.

[0011] When the distribution area affiliation results of the same user's electricity meter are consistent across multiple time window periods, and the corresponding comprehensive similarity is consistently higher than the preset threshold, the distribution area affiliation relationship is confirmed as the final result.

[0012] Furthermore, in S4, constructing the co-cluster matrix and extracting multiple consensus clusters specifically includes: Cluster analysis was performed on the voltage feature matrix of user meters in the same area at multiple time windows; Statistics on any two user electricity meters Frequency of being assigned to the same cluster in all clustering runs and the number of times they jointly participated in clustering. ; Construct a co-convergence matrix A, whose elements Defined as: when hour, ;when hour, And the diagonal elements of the matrix ; Set a similarity threshold An undirected graph is constructed based on the co-convergence matrix A, where the vertices are user meters and the edges satisfy the following conditions: The elements constitute the undirected graph, and all connected components are extracted. Each connected component corresponds to a consensus cluster.

[0013] Furthermore, S5 includes: S51, Representative feature vectors based on any two nodes Calculate similarity and define the distance matrix. ,in, Indicates the node number of the virtual meter box; S52. On the complete graph containing the main meter node of the transformer area and each virtual meter box node, execute the minimum spanning tree algorithm with the distance matrix as the edge weight to obtain the minimum spanning tree algorithm. The smallest undirected spanning tree; S53. For any connected node in an undirected spanning tree Calculate the nodes of the edges. average voltage ;like Then the edge will be oriented to the node. Pointing to node ; S54. Starting from the root node corresponding to the overall table of the transformer area, perform orientation processing on all edges in the entire undirected spanning tree to form a one-way downward path starting from the root node.

[0014] Furthermore, obtaining the overall electrical conformance score in S6 specifically includes: S61. For each directed edge in the initial radially directed topology At each moment Calculate voltage drop ;in, and Parent nodes and child nodes At any moment The voltage value; S62, Calculate child nodes Subtree currents of all its descendant nodes ; S63. Define the valid sample set ;in, The preset voltage drop threshold is used; within the valid sample set. Above, calculate the voltage drop sequence. With subtree current sequence Pearson correlation coefficient ; S64. Define the overall electrical consistency score of the current topology T by weighting the sum of the correlation coefficients of all edges. , .

[0015] Furthermore, the process of adjusting connectivity to generate multiple candidate topologies in S6 specifically includes: S65. The generation range of candidate topologies is limited to the first-level child nodes under the root node and their downstream branches in the initial radially directed topology. S66. Enumerate the non-empty proper subsets of the child node set of any first-level child node. For each enumerated non-empty proper subset, perform a structure adjustment operation: The nodes contained in the non-empty proper subset are retained under the original first-level child nodes, while the remaining nodes in the child node set are reattached to the root node. Each of the above structural adjustment operations generates a new candidate topology.

[0016] Secondly, the present invention provides a low-voltage transformer area topology hierarchy identification system based on main station big data, used to implement the method, including: The electricity data acquisition module is used to collect voltage time sequence data from the main meter of the transformer area and user meters and upload the data. The voltage feature construction module is used to normalize, perform first-order difference, monotonicity and local variance calculations on voltage time series data to form a multi-dimensional voltage feature matrix. The transformer area phase line identification module is used to calculate the similarity between the user's electricity meter and the main meter of each transformer area and each phase line based on the multi-dimensional voltage feature matrix, and to perform transformer area affiliation and phase line affiliation determination. The virtual node construction module is used to perform multi-window clustering analysis, construct a co-clustering matrix, generate consensus clusters, and map the consensus clusters to virtual meter box nodes. The topology reconstruction module is used to construct a minimum spanning tree based on the similarity between nodes and to perform orientation, thereby obtaining a preliminary directed topology structure. The topology optimization module is used to generate and evaluate multiple candidate topologies based on the electrical consistency criterion of voltage drop and subtree current correlation, and select the topology with the highest electrical consistency score as the final topology. The main station display module is used to receive the final topology structure and display the topology hierarchy of the transformer area as the recognition result in a graphical manner.

[0017] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor performs the aforementioned method for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station.

[0018] The beneficial effects of this invention are: This invention deeply integrates data-driven feature learning with the physical laws of circuits, achieving automatic identification, hierarchical inference, and physical rationality correction of low-voltage distribution area topology without on-site manual intervention. The method effectively captures electrical coupling relationships using multi-dimensional voltage features, improves identification stability through multi-window clustering and co-clustering matrices, constructs a globally connected skeleton using minimum spanning trees, and introduces an electrical consistency criterion based on Ohm's law to optimize key hierarchical connections. This overcomes the systemic defects of existing technologies, such as misconnection of trunk line first nodes, hierarchical ambiguity, and topology distortion caused by reliance on single features, neglect of directional constraints, and lack of physical verification mechanisms. The system architecture is complete, the process is closed-loop, and the logic is rigorous, providing high-precision, robust, and highly interpretable topology data support for advanced applications on the main station side.

[0019] The method of this invention collects voltage sampling data from the main meter and user meters in the transformer substation area, constructs a time-series matrix, and preprocesses it to form a voltage feature matrix. Then, it utilizes voltage curve similarity, combined with multi-window clustering and co-cluster matrix fusion, to achieve accurate identification of transformer substation affiliation and phase lines. Based on this, it introduces power transfer direction and energy conservation constraints, and uses a minimum spanning tree algorithm to construct a preliminary tree-like topology rooted at the main meter. Finally, through an electrical consistency optimization mechanism, it calculates the correlation between voltage difference and current, selects the structure with the highest total correlation from candidate topologies as the final topology, and transmits it back to the main station for dynamic visualization. This invention requires no additional hardware investment and can automatically identify complex electrical connection relationships in transformer substations, improving topology accuracy and maintenance efficiency, and achieving real-time perception and intelligent management of transformer substation structures.

[0020] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0021] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the invention.

[0022] Figure 1 The flowchart illustrating the data interaction between the master station system and the control center in this invention is shown.

[0023] Figure 2 The flowchart of the transformer area topology phase line hierarchy identification based on meter timing data is shown.

[0024] Figure 3 This diagram illustrates the clustering analysis results of voltage time-series data from different transformer substations in this invention. Detailed Implementation

[0025] Preferred embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0026] The following specific embodiments further illustrate a method and system for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station, but the present invention is not limited to the following embodiments.

[0027] The method of the present invention generally includes multiple steps such as multi-layer feature extraction based on voltage signals, identification of transformer area and phase line affiliation, construction of virtual nodes based on multi-window clustering, topology reconstruction based on similarity minimum spanning tree, and topology optimization based on electrical consistency. It can be implemented by a computing module deployed in the control center in software or software plus hardware.

[0028] S1. Electricity data acquisition steps: Obtain the voltage time series data of the main meter and each user's meter in the target area on a unified time axis from the main station's electricity information acquisition system; specifically, based on the voltage time series data, form a set of original voltage observation values ​​of each meter under the same time series, and perform preprocessing, including missing value interpolation, outlier removal and timestamp alignment.

[0029] S2. Voltage feature construction steps: Normalize, perform first-order difference operation, determine monotonicity, and calculate local variance based on sliding window for the voltage time series data of each meter in sequence to generate multi-dimensional feature vectors; combine the multi-dimensional feature vectors of all meters to form a voltage feature matrix. Specifically, generating multidimensional feature vectors includes: (1) Characteristics of normalized voltage curve: For the first The meter is at a certain time voltage The first feature component is obtained by performing Min–Max normalization. , ;in, Indicates the meter number, Indicates the time series number. and Electricity meters The minimum and maximum voltage values ​​throughout the entire observation period; this feature primarily reflects the relative change pattern of the voltage curve.

[0030] (2) First-order difference feature: The voltage change rate and response direction are obtained by performing a first-order difference operation on the normalized sequence, which are used as the second feature component. , ; (3) Monotonicity characteristics: The monotonicity characteristics are defined based on the relationship between the normalized voltages of two adjacent points, and the third characteristic component is obtained. Its value selection rule is: if ,but ;like ,but If the two are equal, then ; (4) Local variance characteristics: Using three consecutive time points as a sliding window, the second characteristic component is subjected to local variance characteristics. Calculate the local variance to obtain the fourth eigencomponent. To characterize the intensity of fluctuations within a short timescale; ; The first, second, third, and fourth characteristic components are concatenated in chronological order to form the comprehensive voltage characteristic matrix of the meter. .

[0031] All electricity meters form a voltage characteristic matrix:

[0032] in, This represents the total number of electricity meters.

[0033] S3. Transformer Area Phase Line Identification Step: Based on the voltage feature matrix, calculate the similarity between each user's meter and each candidate transformer area master meter on each phase line; determine the transformer area affiliation of each user's meter according to the similarity, and determine the phase line affiliation of each user's meter within the assigned transformer area. Specifically, including: S31, For user electricity meters eigenvectors With the phase line of the Gp in the substation eigenvectors on Calculate cosine similarity , Indicates phase lines A, B, and C; S32, Install the user's electricity meter The cosine similarity of the transformer area Gp under phases A, B, and C is weighted and summed according to preset weights to obtain the comprehensive similarity. ; ,

[0034] in, The preset weights for phases A, B, and C can be configured based on the three-phase load ratio or data completeness. S33. Based on the principle of maximizing comprehensive similarity, assign the user's electricity meter to the transformer area with the highest comprehensive similarity; within the assigned transformer area, select the phase line with the highest cosine similarity as the phase line assignment for the user's electricity meter. If the overall similarity between the user's meter and all candidate transformer substations is lower than a preset threshold, the user's meter is marked as an unidentified object. The transformer substation phase line identification step is repeated in multiple subsequent time window periods. When the transformer substation affiliation results of the same user's meter are consistent in multiple time window periods, and the corresponding overall similarity is consistently higher than the preset threshold, the transformer substation affiliation relationship is confirmed as the final result.

[0035] In one embodiment, S3 includes substation area attribution identification and phase line attribution determination. First, the district affiliation is identified by setting the master table set as follows: Each master table has a characteristic vector under the three phases A, B, and C. .

[0036] For any user table Calculate the cosine similarity between it and the three phases of the overall table for each transformer substation. , To describe the similarity between the user table and the overall area, the three-phase similarity is weighted and summed, where... Configuration can be made based on the three-phase load ratio or data completeness: , The user table determines the area affiliation of the following:

[0037] In a preferred embodiment, a similarity threshold is introduced to avoid forced attribution. If for all All If so, the user table will be marked as unrecognized, as an exception or subject to manual review.

[0038] Secondly, phase line attribution determination; after identifying the user table Subordinate area Then, the phase line with the highest similarity among the three phases in this transformer area is selected:

[0039] Thus obtain The dual-layer mapping results enable integrated identification of transformer substation affiliation and phase line affiliation.

[0040] In another preferred embodiment, the system repeatedly executes the above-mentioned attribution process in multiple time windows. When a user table is stably assigned to the same transformer area and phase line in most windows and the similarity is consistently higher than the threshold, the attribution relationship is confirmed as the final result to improve the robustness of recognition.

[0041] S4. Virtual Node Construction Steps: Within the same transformer area whose affiliation has been determined, perform cluster analysis on the voltage feature matrix of user meters at multiple time windows; construct a co-cluster matrix based on the clustering results of all time windows and extract multiple consensus clusters; map consensus clusters with a size not less than the preset minimum cluster size to a virtual meter box node, and use the mean of the voltage feature matrices of all user meters in the cluster as the representative feature vector of the virtual meter box node; Specifically, cluster analysis is performed on the voltage feature matrix of user meters within the same distribution area at multiple time windows; Statistics on any two user electricity meters Frequency of being assigned to the same cluster in all clustering runs and the number of times they jointly participated in clustering. ; Construct a co-convergence matrix A, whose elements Defined as: when hour, ;when hour, And the diagonal elements of the matrix ; Set a similarity threshold An undirected graph is constructed based on the co-convergence matrix A, where the vertices are user meters and the edges satisfy the following conditions: The elements constitute the undirected graph, and all connected components are extracted. Each connected component corresponds to a consensus cluster.

[0042] In one embodiment, to further reflect the physical proximity within the transformer substation, the present invention aggregates user meters with similar voltage characteristics to form virtual meter box nodes, which are used as the node basis for topology reconstruction.

[0043] First, multi-window clustering and co-clustering matrix construction: clustering is performed on the feature vectors of all nodes in the transformer area at multiple time windows.

[0044] Based on the results of multiple runs, a co-aggregation matrix is ​​constructed. ,in Number of nodes: , , ; in, This is an indicator function.

[0045] The co-occurrence matrix reflects the consistency of node co-occurrence in multiple clusters, which is equivalent to smoothing the voltage similarity relationship over time by voting.

[0046] Secondly, consensus cluster generation and virtual node construction; setting a similarity threshold. ,when The node is considered to be and There are stable associations between them, thus constructing an undirected graph. :

[0047] In the figure The connected components are extracted, and each connected component is considered as a consensus cluster; a minimum cluster size is also set. Only retain those that meet the requirements. The consensus clusters are treated as virtual meter box nodes, with the average voltage characteristics of users within the cluster serving as the representative feature of that virtual node.

[0048] S5. Preliminary Topology Construction Steps: Take the main meter of the transformer area as the root node, and form a node set together with all virtual meter box nodes; calculate the similarity between nodes based on the representative feature vector of each node and convert it into distance; use the distance as the edge weight, perform the minimum spanning tree algorithm on the complete graph formed by the node set to obtain an undirected spanning tree; orient the edges in the undirected spanning tree according to the magnitude relationship of the average voltage of each node, and complete the directional processing starting from the root node to obtain a preliminary radially directed topology structure. Specifically, this includes S51 and representative feature vectors based on any two nodes. Calculate similarity and define the distance matrix. ,in, S52. On the complete graph containing the main meter node of the transformer area and each virtual meter box node, perform the minimum spanning tree algorithm with the distance matrix as the edge weight to obtain the minimum spanning tree algorithm. Minimum undirected spanning tree; S53, for any connected node in an undirected spanning tree Calculate the nodes of the edges. average voltage ;like Then the edge will be oriented to the node. Pointing to node S54. Starting from the root node corresponding to the overall table of the transformer area, perform orientation processing on all edges in the entire undirected spanning tree to form a one-way downward path starting from the root node.

[0049] In one embodiment, topological reconstruction and directionalization based on the minimum spanning tree of similarity includes the following steps: First, the node set and similarity matrix are constructed; the overall table of transformer substations is used as the root node. With all virtual meter box nodes Together they form a set of nodes:

[0050] For any two nodes Based on its feature vector Calculate similarity The corresponding distance matrix is ​​defined as follows:

[0051] Secondly, minimum spanning tree topology reconstruction; in the node set Construct a complete graph above, and with Apply the Minimum Spanning Tree (MST) algorithm to the edge weights to obtain an undirected tree. :

[0052] That is, select the connected tree with the smallest overall difference between nodes, so as to form a topological skeleton that satisfies the greatest overall similarity.

[0053] Finally, topology directionality and hierarchy determination are performed; to ensure that the topology conforms to the physical law of monotonically decreasing voltage along the feeder, the average voltage of each node is defined: , in For nodes At any moment The voltage.

[0054] For any undirected edge ,like Then orient the edge as Otherwise, the orientation is Using the master table node Using the root as the root, the entire tree is oriented to obtain a directed radial topology. This clarifies the hierarchical position of each virtual meter box within the transformer substation.

[0055] S6. Electrical Topology Optimization Steps: For each directed edge in the initial radially directed topology, calculate the voltage drop sequence between parent and child nodes and the subtree current sequence rooted at the child node; on the valid sample set where the voltage drop is greater than a preset threshold, calculate the correlation coefficient between the voltage drop sequence and the subtree current sequence, and define the overall electrical consistency score of the topology based on the correlation coefficient of all directed edges; by adjusting the connection relationship, generate multiple candidate topologies, calculate the overall electrical consistency score of each candidate topology, and select the candidate topology with the highest score as the final low-voltage distribution area topology hierarchy identification result.

[0056] Specifically, obtaining the overall electrical conformance score includes: S61. For each directed edge in the initial radially directed topology At each moment Calculate voltage drop ;in, and Parent nodes and child nodes At any moment The voltage value; S62, Calculate child nodes Subtree currents of all its descendant nodes ; S63. Define the valid sample set ;in, The preset voltage drop threshold is used; within the valid sample set. Above, calculate the voltage drop sequence. With subtree current sequence Pearson correlation coefficient ; S64. Define the overall electrical consistency score of the current topology T by weighting the sum of the correlation coefficients of all edges. , .

[0057] According to claim 7, the method for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station is characterized in that, in step S6, adjusting the connection relationships to generate multiple candidate topologies specifically includes: S65. The generation range of candidate topologies is limited to the first-level child nodes under the root node and their downstream branches in the initial radially directed topology. S66. Enumerate the non-empty proper subsets of the child node set of any first-level child node. For each enumerated non-empty proper subset, perform a structure adjustment operation: The nodes contained in the non-empty proper subset are retained under the original first-level child nodes, while the remaining nodes in the child node set are reattached to the root node. Each of the above structural adjustment operations generates a new candidate topology.

[0058] In one embodiment, to address the problem of mis-attached trunk first nodes in multi-trunk transformer areas where MSTs constructed solely based on similarity are prone to such issues, this invention further introduces a consistency criterion based on Ohm's law to optimize the initial topology under physical constraints.

[0059] First, the electrical consistency criterion; for any candidate directed tree edge At any moment Definition: Node voltage: , Voltage drop: Subtree current: ,in For nodes and all its descendants; instantaneous equivalent resistance: .

[0060] Under ideal electrical conditions, a rise in load leads to An increase should be accompanied by a pressure drop in that section. As the length increases, the two should be positively correlated in the time dimension. Based on this, the consistency score of an edge is defined as:

[0061] Among them, only adopt Greater than the threshold The time of the sample is used as the effective sample to suppress noise.

[0062] Topology The overall electrical conformance score is defined as: .

[0063] This involves candidate topology generation and optimal topology selection. Based on engineering experience, erroneous connections are mainly concentrated in the first-level trunk nodes below the root node. Therefore, in the initial directed tree... Based on the hierarchical structure, local adjustment operations are limited to the first-level child nodes and their sub-branches of the root node to generate a finite set of candidate topologies. This includes the original topology and several deformed topologies obtained by redistributing the first-level child nodes.

[0064] For each candidate topology Calculate all edges according to their parent-child relationships. and ; Calculate the edge values ​​on the valid samples. And calculate the overall score. The final selection made Largest candidate topology: .

[0065] This yields the most consistent final topology in an electrical sense, effectively correcting systematic deviations such as trunk head node bridging caused by simple similarity-driven approaches.

[0066] Finally, the main station performs data interaction and topology display; the interaction process between the main station and the control center backend is shown in the attached figure. Figure 1As shown, the master station packages and pushes meter data within the distribution area to the control center API at the agreed scheduling time. The pushed content is granular at the distribution area level, including 96 voltage curves and their identifiers (such as meter ID, distribution area ID, and phase information) from the master meter and user meters; the data is aligned by natural day, meeting the basic requirements of monotonous timestamps, complete data points, and consistent data caliber. After receiving the data, the control center performs local data solidification and time alignment, and then performs data quality control: cleaning and repairing missing, abnormal, out-of-bounds, and inconsistent records, and deduplicating duplicate or out-of-order batches by batch number (or business time) to ensure that the data entering the calculation stage is calculable and comparable. When the data reaches the minimum coverage conditions required for identification (such as 96 complete data points in a single day, or the cumulative arrival of data within a multi-day window), the control center enters the distribution area topology identification process.

[0067] During the identification phase, the control center constructs a feature space for curves within the same transformer area and phase. Consensus clusters (virtual meter boxes) are formed based on multi-window clustering. A similarity matrix is ​​jointly constructed using cluster-level curves and the main meter curve, and a preliminary radial skeleton is obtained through minimum spanning tree analysis. Subsequently, an electrical consistency criterion is introduced to enumerate and identify local candidate topologies for locations prone to misconnection, such as the main line's first node. The structure with the highest correlation between "voltage drop and subtree current" is selected as the optimal topology. The identification results are returned as a structured load, including transformer area ID, phase, phase line node location (node ​​name / path identifier / level), and a list of directed edges (parent-child relationships). The control center immediately transmits the results back upon completion of the calculation, enabling the master station to obtain the latest topology status at the same business time.

[0068] After receiving the results, the master station renders the radial topology of the current transformer area in real time based on the node path identifiers and parent-child relationships: it expands hierarchically with the master table as the root, visually presenting the branch structure, branch depth, and node adjacency relationships. The rendered topology serves as the structural foundation for various applications within the platform, directly supporting subsequent analysis modules such as line loss calculation, fault location, abnormal waveform identification, and electricity theft assessment, achieving a closed-loop linkage of "data—model—results—applications." The entire process operates on a cycle of "timed push from the master station, on-site judgment and real-time feedback from the control center, and instant presentation from the master station," providing stable, interpretable, and iterative topology recognition capabilities while ensuring physical consistency.

[0069] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A method for identifying the topology hierarchy of low-voltage distribution areas based on big data from a main station, characterized in that, Includes the following steps: S1. Steps for acquiring electricity consumption data: Obtain the voltage time sequence data of the main meter and each user's meter in the target area on a unified time axis from the main station's electricity consumption information acquisition system. S2. Voltage feature construction steps: Normalize, perform first-order difference operation, determine monotonicity, and calculate local variance based on sliding window for the voltage time series data of each meter in sequence to generate multi-dimensional feature vectors; combine the multi-dimensional feature vectors of all meters to form a voltage feature matrix. S3. Transformer Area Phase Line Identification Step: Based on the voltage feature matrix, calculate the similarity between each user's meter and each candidate transformer area's master meter on each phase line; determine the transformer area affiliation of each user's meter according to the similarity, and determine the phase line affiliation of each user's meter within the assigned transformer area. S4. Virtual Node Construction Steps: Within the same transformer area whose affiliation has been determined, perform cluster analysis on the voltage feature matrix of user meters at multiple time windows; construct a co-cluster matrix based on the clustering results of all time windows and extract multiple consensus clusters; map consensus clusters with a size not less than the preset minimum cluster size to a virtual meter box node, and use the mean of the voltage feature matrices of all user meters in the cluster as the representative feature vector of the virtual meter box node; S5. Preliminary Topology Construction Steps: Take the main meter of the transformer area as the root node, and form a node set together with all virtual meter box nodes; calculate the similarity between nodes based on the representative feature vector of each node and convert it into distance; use the distance as the edge weight, perform the minimum spanning tree algorithm on the complete graph formed by the node set to obtain an undirected spanning tree; orient the edges in the undirected spanning tree according to the magnitude relationship of the average voltage of each node, and complete the directional processing starting from the root node to obtain a preliminary radially directed topology structure. S6. Electrical topology optimization steps: For each directed edge in the initial radially directed topology, calculate the voltage drop sequence between parent and child nodes and the subtree current sequence rooted at the child node. On the valid sample set where the voltage drop is greater than a preset threshold, the correlation coefficient between the voltage drop sequence and the subtree current sequence is calculated, and the overall electrical consistency score of the topology is defined based on the correlation coefficient of all directed edges. By adjusting the connection relationships, multiple candidate topologies are generated. The overall electrical consistency score of each candidate topology is calculated, and the candidate topology with the highest score is selected as the final low-voltage distribution area topology hierarchy identification result.

2. The method for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station according to claim 1, characterized in that... In S2, generating multidimensional feature vectors specifically includes: For the The meter is at a certain time voltage The first feature component is obtained by performing Min–Max normalization. , ;in, Indicates the meter number, Indicates the time series number. and Electricity meters Minimum and maximum voltage values ​​during the entire observation period; The second feature component is obtained by performing a first-order difference operation on the normalized sequence. , ; Based on the relationship between the normalized voltages of two adjacent points, a monotonicity characteristic is defined, resulting in the third characteristic component. Its value selection rule is: if ,but ;like ,but If the two are equal, then ; Using a preset number of consecutive time points as a sliding window, the second feature component is... Calculate the local variance to obtain the fourth eigencomponent. To characterize the intensity of fluctuations within a short timescale; The first, second, third, and fourth characteristic components are concatenated in chronological order to form the comprehensive voltage characteristic matrix of the meter. .

3. The method for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station according to claim 1, characterized in that... S3 specifically includes: S31, For user electricity meters eigenvectors With the phase line of the Gp area eigenvectors on Calculate cosine similarity , Indicates phase lines A, B, and C; S32, Install the user's electricity meter The cosine similarity of the transformer area Gp under phases A, B, and C is weighted and summed according to preset weights to obtain the comprehensive similarity. The preset weights of phases A, B, and C are non-negative real numbers and their sum is 1. S33. Based on the principle of maximizing comprehensive similarity, the user's electricity meter... It is assigned to the area with the highest overall similarity; Within the assigned transformer area, the phase line with the highest cosine similarity is selected as the user's electricity meter. The phase line is assigned to.

4. The method for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station according to claim 3, characterized in that... S33 also includes: If the user's electricity meter If the overall similarity with all candidate transformer substations is lower than the preset threshold, the user's meter will be marked as an unidentified object. The phase line identification step for the transformer area is repeated over multiple subsequent time windows. When the distribution area affiliation results of the same user's electricity meter are consistent across multiple time window periods, and the corresponding comprehensive similarity is consistently higher than the preset threshold, the distribution area affiliation relationship is confirmed as the final result.

5. The method for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station according to claim 1, characterized in that... In S4, constructing the co-aggregation matrix and extracting multiple consensus clusters specifically includes: Cluster analysis was performed on the voltage feature matrix of user meters in the same area at multiple time windows; Statistics on any two user electricity meters Frequency of being assigned to the same cluster in all clustering runs and the number of times they jointly participated in clustering. ; Construct a co-convergence matrix A, whose elements Defined as: when hour, ;when hour, And the diagonal elements of the matrix ; Set a similarity threshold An undirected graph is constructed based on the co-convergence matrix A, where the vertices are user meters and the edges satisfy the following conditions: The elements constitute the undirected graph, and all connected components are extracted. Each connected component corresponds to a consensus cluster.

6. The method for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station according to claim 1, characterized in that S5 include: S51, Representative feature vectors based on any two nodes Calculate similarity and define the distance matrix. ,in, Indicates the node number of the virtual meter box; S52. On the complete graph containing the main meter node of the transformer area and each virtual meter box node, execute the minimum spanning tree algorithm with the distance matrix as the edge weight to obtain the minimum spanning tree algorithm. The smallest undirected spanning tree; S53. For any connected node in an undirected spanning tree Calculate the nodes of the edges. average voltage ;like Then the edge will be oriented to the node. Pointing to node ; S54. Starting from the root node corresponding to the overall table of the transformer area, perform orientation processing on all edges in the entire undirected spanning tree to form a one-way downward path starting from the root node.

7. The method for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station according to claim 1, characterized in that... In S6, obtaining the overall electrical conformance score specifically includes: S61. For each directed edge in the initial radially directed topology At each moment Calculate voltage drop ;in, and Parent nodes and child nodes At any moment The voltage value; S62, Calculate child nodes Subtree currents of all its descendant nodes ; S63. Define the valid sample set ;in, The preset voltage drop threshold is used; within the valid sample set. Above, calculate the voltage drop sequence. With subtree current sequence Pearson correlation coefficient ; S64. Define the overall electrical consistency score of the current topology T by weighting the sum of the correlation coefficients of all edges. , .

8. The method for identifying the topology hierarchy of low-voltage distribution areas based on big data from the main station according to claim 7, characterized in that... In S6, adjusting the connection relationships to generate multiple candidate topologies specifically includes: S65. The generation range of candidate topologies is limited to the first-level child nodes under the root node and their downstream branches in the initial radially directed topology. S66. Enumerate the non-empty proper subsets of the child node set of any first-level child node. For each enumerated non-empty proper subset, perform a structure adjustment operation: The nodes contained in the non-empty proper subset are retained under the original first-level child nodes, while the remaining nodes in the child node set are reattached to the root node. Each of the above structural adjustment operations generates a new candidate topology.

9. A low-voltage distribution area topology hierarchy identification system based on main station big data, used to implement the method according to any one of claims 1 to 8, characterized in that, include: The electricity data acquisition module is used to collect voltage time sequence data from the main meter of the transformer area and user meters and upload the data. The voltage feature construction module is used to normalize, perform first-order difference, monotonicity and local variance calculations on voltage time series data to form a multi-dimensional voltage feature matrix. The transformer area phase line identification module is used to calculate the similarity between the user's electricity meter and the main meter of each transformer area and each phase line based on the multi-dimensional voltage feature matrix, and to perform transformer area affiliation and phase line affiliation determination. The virtual node construction module is used to perform multi-window clustering analysis, construct a co-clustering matrix, generate consensus clusters, and map the consensus clusters to virtual meter box nodes. The topology reconstruction module is used to construct a minimum spanning tree based on the similarity between nodes and to perform orientation, thereby obtaining a preliminary directed topology structure. The topology optimization module is used to generate and evaluate multiple candidate topologies based on the electrical consistency criterion of voltage drop and subtree current correlation, and select the topology with the highest electrical consistency score as the final topology. The main station display module is used to receive the final topology structure and display the topology hierarchy of the transformer area as the recognition result in a graphical manner.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it causes the processor to perform the low-voltage distribution area topology hierarchy identification method based on the main station big data as described in any one of claims 1 to 8.

Images