Station area automatic division and affiliation judgment method based on communication node topology identification

Abstract

The invention relates to the technical field of power system communication, and discloses a transformer area automatic division and affiliation judgment method based on communication node topology identification, which specifically comprises the following steps: step 1, collecting communication topology data; step 2, constructing a multi-time-sequence topological snapshot; 3, identifying stable communication links, analyzing the topological snapshot sequence, calculating the occurrence frequency and duration of the connection relationship between each pair of communication nodes, and identifying the stable communication links with the frequency and duration exceeding preset thresholds; 4, constructing a weighted communication topological graph; 5, topological graph clustering segmentation is executed; and 6, determining the affiliation of the transformer area. According to the method, the multi-time-sequence topological snapshots are constructed, and the topological snapshot sequence is analyzed to identify the stable communication link with the frequency and the duration exceeding the preset threshold values, so that transient interference connection caused by noise interference such as load fluctuation and electric appliance starting and stopping is effectively filtered out.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system communication, in particular to a method for automatic division and attribution determination of transformer area based on communication node topology identification. BACKGROUND

[0002] With the deepening of smart grid and power internet of things technology, low-voltage distribution transformer area as the basic unit of power grid operation and management, its fine management and control capability is directly related to the line loss calculation accuracy, load prediction accuracy and fault response efficiency. The transformer area is powered by a single distribution transformer, and the electrical attribution relationship between the power users under the jurisdiction of the transformer and the transformer, i.e. the "transformer-user relationship", is the core basic data to realize the above functions. However, in the actual distribution network, due to the lack of historical archives, frequent line reconstruction and complex factors such as user private connection, the transformer-user relationship recorded in the transformer account is seriously out of touch with the real electrical connection on site, forming the widespread "relationship confusion" problem, which seriously restricts the intelligent operation and maintenance level of the distribution network.

[0003] Among them, the transformer area automatic division method based on communication node topology identification has become the mainstream technical path to solve the problem of unclear transformer-user relationship. This method relies on the intelligent meters and concentrators widely deployed in the advanced measurement system (AMI), uses the logical network structure constructed by communication methods such as power line carrier (PLC), analyzes the communication topology information such as the routing table and the neighbor node list obtained by the concentrator, infers the logical attribution relationship between each meter terminal and the concentrator, and then realizes the automatic identification of the transformer area boundary and the user attribution determination. The core assumption is that the communication connection relationship can accurately reflect the electrical connection relationship, i.e. the communication subnet and the electrical transformer area are one-to-one corresponding.

[0004] However, the existing technology has exposed significant defects in actual application. First, there is an essential inconsistency between the communication topology and the electrical topology: due to the adjacent laying of lines between different transformer areas in physical space, PLC signals are easy to produce cross-area crosstalk through electromagnetic coupling or common mode conduction, leading to the error access of meter terminals to the concentrators of adjacent transformer areas at the communication level, thus causing attribution misjudgment. Second, the communication topology is highly dynamic, affected by noise interference such as load fluctuation and appliance start-stop, the relay path and connection relationship of the meter frequently change, making the determination result based on a single time point topology snapshot extremely unstable, it is difficult to form a reliable and lasting transformer area file, the above problems together cause the existing method to have low accuracy and poor robustness in complex field environment, it is difficult to meet the urgent needs of the power grid for high-precision and high-stability determination of transformer-user relationship. SUMMARY

[0005] (I) Technical problems solved

[0006] In view of the deficiencies of the prior art, the present application provides a communication node topology identification-based automatic substation division and attribution determination method, which solves the problem of low accuracy and poor robustness of the prior art in complex field environments.

[0007] (II) Technical solutions

[0008] To achieve the above object, the present application is implemented by the following technical solutions: a communication node topology identification-based automatic substation division and attribution determination method, specifically comprising the following steps:

[0009] Step 1: Collect communication topology data, periodically collect communication topology data of all smart meters within the jurisdiction of the concentrator, the communication topology data including routing table, neighbor node list, signal strength and communication delay;

[0010] Step 2: Construct a multi-time sequence topology snapshot, continuously collect multiple sets of communication topology data at fixed time intervals within a preset time window to form a sequence of topology snapshots arranged in time sequence;

[0011] Step 3: Identify stable communication links, analyze the topology snapshot sequence, calculate the frequency and duration of the connection relationship between each pair of communication nodes, and identify stable communication links with frequency and duration exceeding a preset threshold;

[0012] Step 4: Construct a weighted communication topology graph, taking smart meters and concentrators as nodes, taking stable communication links as edges, and assigning weights to each edge, the weights being calculated based on signal strength, communication delay and link stability;

[0013] Step 5: Perform topology graph clustering and segmentation, based on the weighted communication topology graph, use an improved community discovery algorithm to cluster the nodes, and divide the entire topology graph into multiple communication subnets with tight internal connections and sparse external connections;

[0014] Step 6: Determine the substation attribution, map each identified communication subnet to an independent electrical substation, and determine the smart meters within the subnet as belonging to the corresponding distribution transformer of the substation.

[0015] Preferably, in step 1, the concentrator interacts with the smart meter through power line carrier communication mode, the collection period is adjustable from 5 minutes to 30 minutes, the routing table records the multi-hop path information of each meter, the neighbor node list records the adjacent node identifier of each meter for direct communication, the signal strength is represented by received signal strength indication, and the communication delay is recorded in milliseconds.

[0016] Preferably, in the step two, the preset time window length is 24 hours, and the fixed time interval is set to 15 minutes, so as to form a sequence containing 96 topology snapshots, each of which contains the complete connection relationship state of all nodes at the moment.

[0017] Preferably, in the step three, the identification threshold of the stable communication link is set to: the appearance frequency of the connection relationship in the topology snapshot sequence needs to be greater than or equal to 90%, and the duration of a single connection needs to be greater than or equal to 4 hours, and the connection below the threshold is regarded as a transient interference connection and is filtered out.

[0018] Preferably, in the step four, the weight calculation formula of the edge in the weighted communication topology graph is: the weight is equal to the standardized signal strength multiplied by the weight coefficient plus the standardized link stability multiplied by the weight coefficient minus the standardized communication delay multiplied by the weight coefficient , wherein is a preset weighting coefficient, and = 1, and the normalization process maps each parameter value to the interval of 0 to 1.

[0019] Preferably, in the step five, the improved community discovery algorithm is based on modularity optimization, and the edge weight is introduced into the objective function as a measure of the connection tightness within the community. In the algorithm iteration process, the modularity gain brought by the movement of the node to the adjacent community is calculated, and the community division is continuously optimized until the modularity no longer significantly improves.

[0020] Preferably, the improved community discovery algorithm also combines the node attribute similarity, the node attributes include the phase information, the voltage level and the geographic position coordinates of the electric meter, and the attribute similarity is taken as an auxiliary constraint condition in the community division process to improve the consistency of the division result with the electrical reality.

[0021] Preferably, the step six further comprises a result verification and correction step:

[0022] The preliminary determined substation attribution result is cross-verified with the limited account information provided by the distribution automation system;

[0023] For the contradictory judgment result, an auxiliary verification process based on voltage transient event correlation analysis is started, the correlation of the voltage waveform responses of each electric meter is analyzed by injecting a voltage disturbance signal with specific characteristics, and the attribution relationship is further confirmed or corrected.

[0024] Preferably, the auxiliary verification process based on voltage transient event correlation analysis specifically comprises:

[0025] A voltage dip event with a duration of 100 milliseconds and an amplitude of 5% of the rated voltage is generated by the smart switch at the output of the control area via the concentrator.

[0026] Simultaneously record the voltage waveforms monitored by all smart meters within the distribution area;

[0027] Calculate the cross-correlation coefficient between the voltage waveforms of every two meters. Meter pairs with a cross-correlation coefficient exceeding 0.95 are considered to be in the same transformer area.

[0028] The preliminary judgment results of step six are calibrated based on the cross-correlation matrix.

[0029] Preferably, the method also establishes a knowledge base for transformer area relationships to store historical judgment results, verification data, and on-site operation and maintenance feedback. Using the data in the knowledge base, a long short-term memory neural network model is trained. This model can predict the potential change trend of transformer area relationships based on real-time communication topology characteristics, thereby achieving predictive maintenance.

[0030] Preferably, the method is deployed on the power distribution network cloud master station platform, and communicates with the field concentrators through the standard 104 protocol or MQTT protocol. The cloud master station platform cleans, integrates and analyzes the received data, and finally sends the determined distribution area affiliation to the power consumption information collection system and related power distribution management system to complete the file update.

[0031] (III) Beneficial Effects

[0032] This invention provides a method for automatic substation division and affiliation determination based on communication node topology identification. It has the following beneficial effects:

[0033] (1) By constructing multiple time-series topology snapshots and analyzing the topology snapshot sequence to identify stable communication links whose frequency and duration both exceed preset thresholds, transient interference connections caused by noise interference such as load fluctuations and appliance start-up and shutdown are effectively filtered out, solving the problem that the judgment results based on a single point-in-time topology snapshot in the existing technology are extremely unstable.

[0034] (2) By combining the similarity of node attributes such as the phase information of the meter, voltage level and geographical coordinates in the topology graph clustering segmentation as auxiliary constraints, and starting the auxiliary verification process based on voltage transient event correlation analysis after the initial judgment, the problem of the essential inconsistency between the communication topology and the electrical topology caused by cross-regional interference of PLC signals was solved, and misjudgment of attribution was avoided.

[0035] (3) By constructing a weighted communication topology graph, using an improved community discovery algorithm for clustering and segmentation, and finally combining the result verification and correction steps, the shortcomings of existing technologies in low accuracy and poor robustness in complex field environments are effectively solved, and the urgent needs of the power grid for high-precision and high-stability determination of the relationship between the grid and the user are met. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the overall process of the present invention from data acquisition to final determination of station area affiliation;

[0037] Figure 2 This is a schematic diagram of the communication topology data acquisition process of the present invention;

[0038] Figure 3 This is a schematic diagram illustrating the process of constructing a multi-temporal topology snapshot for this invention;

[0039] Figure 4 This is a schematic diagram illustrating the identification of stable communication links in this invention;

[0040] Figure 5 A schematic diagram of the weighted communication topology constructed for this invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] Please see Figure 1 - Figure 5 This invention provides a method for automatic division and attribution determination of transformer substations based on communication node topology identification, specifically including the following steps:

[0043] Step 1: Collect communication topology data. The concentrator periodically collects communication topology data from all smart meters within its jurisdiction. This data includes routing tables, neighbor node lists, signal strength, and communication latency. Specifically:

[0044] The concentrator interacts with smart meters via power line carrier communication, with an adjustable data acquisition cycle of 5 to 30 minutes. The routing table records multiple paths to each meter, and the neighbor node list records the identifiers of adjacent nodes directly communicating with each meter. Signal strength is indicated by the received signal strength indicator, and communication latency is recorded in milliseconds. The concentrator is deployed at the low-voltage outgoing end of the distribution transformer, acting as the aggregation node of the communication network. It periodically sends topology detection commands to all smart meters within its logical coverage area via the power line carrier communication module. Upon receiving the command, each smart meter displays its currently maintained neighbor node list, its optimal route to the concentrator, the received signal strength indicators of each neighbor node, and its communication history with the concentrator. The return delay data is packaged and transmitted back to the concentrator via the power line carrier channel. The concentrator timestamps and performs preliminary verification on the raw data received from all meters to form a complete communication topology dataset. The structured storage format of the communication topology dataset includes the meter's unique identifier, a list of neighbor node identifiers, a sequence of routing path nodes, signal strength indicators, communication delay values, and data acquisition timestamps. The acquisition period needs to be set to balance data freshness and system communication load. A 5-minute period is suitable for highly dynamic distribution areas, while a 30-minute period is suitable for areas with relatively stable loads. The quantization accuracy of the signal strength indicator is 1 dB, and the measurement resolution of the communication delay is 1 millisecond, ensuring the data foundation for subsequent link stability analysis.

[0045] Step 2: Construct multiple time-series topology snapshots. Within a preset time window, continuously collect multiple sets of communication topology data at fixed time intervals to form a sequence of topology snapshots arranged in chronological order. Specifically:

[0046] The preset time window length is 24 hours, and the fixed time interval is set to 15 minutes, thus forming a sequence containing 96 topology snapshots. Each topology snapshot contains the complete connection relationship status of all nodes at that moment. The cloud master station platform sends an instruction to the concentrator to start a continuous data acquisition task for 24 hours. The timer inside the concentrator triggers a complete topology data acquisition process every 15 minutes, that is, the operation in step one is executed. After each acquisition is completed, the concentrator temporarily stores the complete communication topology data acquired this time as an independent topology snapshot in the local memory and records its sequence number (from 1 to 96). Each topology snapshot not only contains the connection relationship between nodes (i.e., which nodes are neighbors), but also contains dynamic attributes such as the signal strength and communication latency of each connection at that moment.

[0047] Step 3: Identify stable communication links, analyze the topology snapshot sequence, calculate the frequency and duration of the connection relationship between each pair of communication nodes, and identify stable communication links whose frequency and duration both exceed a preset threshold. Specifically:

[0048] The threshold for identifying stable communication links is set as follows: the frequency of a connection appearing in the topology snapshot sequence must be greater than or equal to 90%, and the duration of a single connection must be greater than or equal to 4 hours. Connections below this threshold are considered transient interference connections and are filtered out. After the cloud main station platform obtains the complete 96 topology snapshot sequences from the concentrator, it starts the link stability analysis module. The link stability analysis module first traverses all snapshots and establishes a global node pair connection matrix. For any pair of nodes (i,j), the link stability analysis module counts the number of times they appear simultaneously in each other's neighbor lists in the 96 snapshots, and calculates the frequency of occurrence as that number. Dividing by 96, the link stability analysis module analyzes the continuity of the connection relationship on the time axis, identifies all consecutively occurring intervals, and records the duration of each interval (calculated by multiplying the number of snapshots by 15 minutes). Only when the connection relationship between a pair of nodes satisfies the following conditions: occurrence frequency ≥ 90% (i.e., appearing in at least 87 snapshots) and the existence of at least one interval with a duration ≥ 4 hours (i.e., at least 16 consecutive snapshots), the connection is determined to be a stable communication link. All connections that do not meet this dual threshold, regardless of their signal strength, are considered as false connections caused by electromagnetic crosstalk or transient noise and are completely filtered out from subsequent analysis.

[0049] Step 4: Construct a weighted communication topology graph, using smart meters and concentrators as nodes and stable communication links as edges. Assign a weight to each edge, calculated based on signal strength, communication latency, and link stability. Specifically, in Step 4, the weight calculation formula for the edges in the weighted communication topology graph is: weight equals standardized signal strength multiplied by a weight coefficient. Add standardized link stability multiplied by a weighting factor Subtract the standardized communication delay and multiply by the weighting factor ,in The weighting coefficients are preset, and =1, the standardization process maps each parameter value to the interval between 0 and 1. For each identified stable communication link (i,j), its comprehensive weight needs to be calculated. Specifically:

[0050] First, the original parameters are standardized: signal strength Transform the value into the interval [0,1] using a linear mapping. Communication delay Converted to a value in the range [0,1] through reverse mapping. Link stability Then, its frequency of occurrence is used directly, and then a preset weighting coefficient is applied. =0.5、 =0.3、 =0.2, substitute into the following formula to calculate the weight:

[0051]

[0052] The higher the signal strength and link stability, the greater the weight; the longer the communication delay, the smaller the weight. The final weighted communication topology graph G=(V,E,W) is constructed, where V is the set of nodes consisting of all smart meters and concentrators, E is the set of edges consisting of all stable communication links, and W is the set of edge weights.

[0053] Step 5: Perform topology graph clustering and segmentation. Based on the weighted communication topology graph, an improved community detection algorithm is used to cluster the nodes, dividing the entire topology graph into multiple communication subnets with tight internal connections and sparse external connections. Specifically:

[0054] The improved community detection algorithm is based on modularity optimization. Its objective function incorporates edge weights as a measure of the tightness of connections within a community. During algorithm iteration, the modularity gain resulting from moving a node to an adjacent community is calculated, continuously optimizing community partitioning until the modularity no longer significantly increases. Furthermore, the improved algorithm combines node attribute similarity, including meter phase information, voltage level, and geographical coordinates. Attribute similarity is integrated as an auxiliary constraint into the community partitioning process to improve the consistency between the partitioning results and actual electrical conditions. The clustering and segmentation algorithm first initializes, treating each node as an independent community. The core of the algorithm is calculating the modularity Q, whose definition has been modified to consider edge weights.

[0055]

[0056] in, It is the sum of the weights of all edges in the graph. It is a node The weighted degree is the sum of the weights of all edges connected to it. It is a node The community to which it belongs The function is in = The value is 1 if the condition is positive and 0 otherwise. In each iteration, the algorithm traverses all nodes, attempts to move them to adjacent communities, and calculates the modularity gain ΔQ brought by the move. If ΔQ is positive, the move is executed. This process is repeated until no move can bring a positive ΔQ, that is, the modularity reaches a local maximum. Based on this, the algorithm further introduces node attribute constraints. For any two nodes... and Calculate their attribute similarity This value takes into account phase consistency, voltage level differences, and geographical distance. When calculating modularity gain, it is only applicable when... Nodes are only allowed when the value exceeds a preset threshold (e.g., 0.7). and They were assigned to the same community.

[0057] Step Six: Determine the distribution area affiliation. Map each identified communication subnet to an independent electrical distribution area, and determine the smart meters within the subnet to belong to the corresponding distribution transformer of that area. After the clustering and segmentation in Step Five, the weighted communication topology is divided into K disjoint communities. , ... Each community This is considered an independent communication subnet. Since each concentrator physically corresponds to a distribution transformer, and ideally, one concentrator should only manage one distribution area, the community containing concentrator nodes... That is, it is directly mapped to the electrical distribution area corresponding to the concentrator, the community. All smart meter nodes within the system are identified as belonging to the distribution transformer. For isolated communities that do not contain a concentrator, the system will mark them as abnormal and trigger an alarm.

[0058] Furthermore, following step six, there are also result verification and correction steps, specifically:

[0059] The preliminary determination of transformer substation affiliation is cross-validated with the limited ledger information provided by the distribution automation system. For contradictory determinations, an auxiliary verification process based on voltage transient event correlation analysis is initiated. This process involves injecting voltage disturbance signals with specific characteristics to analyze the correlation of voltage waveform responses of each meter, further confirming or correcting the affiliation. Specifically, the auxiliary verification process based on voltage transient event correlation analysis includes:

[0060] A voltage dip event with a duration of 100 milliseconds and an amplitude of 5% of the rated voltage is generated by the smart switch at the output of the control area via the concentrator.

[0061] Simultaneously record the voltage waveforms monitored by all smart meters within the distribution area;

[0062] Calculate the cross-correlation coefficient between the voltage waveforms of every two meters. Meter pairs with a cross-correlation coefficient exceeding 0.95 are considered to be in the same transformer area.

[0063] The preliminary judgment results of step 6 are calibrated based on the cross-correlation matrix.

[0064] Furthermore, the cloud master station platform first compares the preliminary attribution results obtained in step six with the potentially incomplete but relatively reliable ledger data obtained from the distribution automation system. If a conflict is found between the attribution determination of a meter and the ledger information, and the confidence level of the ledger information is high, the system marks the meter and its community as pending verification. Then, the system issues an instruction to the relevant concentrator, which controls the main switch of the control area to perform a standardized voltage sag operation: at a specified moment, the output voltage is instantaneously reduced to 95% of the rated value and maintained for 100 milliseconds before recovering. During this period, all smart meters synchronously record their voltage waveforms at a sampling rate of 1000 Hz. After the operation is completed, each meter uploads the recorded waveform data to the cloud master station, and the master station calculates the cross-correlation coefficients of the voltage waveforms between all pairs of meters to be verified. Because meters located in the same electrical distribution area will experience almost identical voltage dip events, their waveforms are highly similar. It will be very close to 1. In this embodiment, the threshold is set to 0.95. ≥0.95, then confirmed. and If the meters are in the same transformer substation, the meter is in the correct substation. Otherwise, the meter is incorrectly assigned. The main station will use this cross-correlation matrix to correct the initial assignment, for example, by reassigning meters that were incorrectly assigned to neighboring transformer substations to the correct community.

[0065] Furthermore, this invention establishes a transformer substation relationship knowledge base to store historical judgment results, verification data, and on-site operation and maintenance feedback. Using the data in this knowledge base, a long short-term memory (LSTM) neural network model is trained. This LTM model can predict potential changes in transformer substation relationships based on real-time communication topology characteristics, enabling predictive maintenance. The cloud master station platform maintains a structured transformer substation relationship knowledge base, which stores detailed results of each automatic partitioning, raw data for auxiliary verification, and manual correction records from on-site operation and maintenance personnel, forming a historical dataset with timestamps and labels. Using this dataset, an LTM neural network model can be trained offline. The input features of the LTM neural network model include real-time collected communication topology features, and the output is the probability of changes in transformer substation relationships. After sufficient training, the LTM neural network model can be deployed on the cloud master station to perform online inference on real-time data streams. When the stability of a transformer substation's relationship is predicted to be below a safety threshold, the system can issue an early warning, prompting operation and maintenance personnel to conduct preventative checks, thereby transforming passive response into proactive maintenance and further improving the intelligence level of the distribution network.

[0066] Furthermore, this method is deployed on a distribution network cloud master station platform. It communicates with field concentrators via the standard IEC 60870-5-104 protocol or MQTT protocol. The cloud master station platform cleans, integrates, and analyzes the received data, and finally distributes the determined transformer area affiliation to the electricity consumption information collection system and related distribution management systems to complete the file update. The software implementation of the entire method is deployed on a provincial or municipal distribution network cloud master station platform. The distribution network cloud master station platform communicates bidirectionally with concentrators located in various locations via a dedicated power communication network using the IEC 60870-5-104 protocol or MQTT protocol. After receiving the raw data, the platform sequentially performs data cleaning, data integration, and all the aforementioned analysis steps. The final generated, verified, and corrected transformer area affiliation results are automatically synchronized to the electricity consumption information collection system in the form of a standard data interface to update user files, and synchronized to the distribution management system to support advanced applications such as line loss calculation and load forecasting, realizing closed-loop management and value loop of transformer area relationship data.

[0067] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for automatic division and attribution determination of transformer substations based on communication node topology identification, characterized in that, Includes the following steps: Step 1: Collect communication topology data. The concentrator periodically collects communication topology data of all smart meters within its jurisdiction. The communication topology data includes routing table, neighbor node list, signal strength and communication delay. Step 2: Construct multiple time-series topology snapshots. Within a preset time window, continuously collect multiple sets of communication topology data at fixed time intervals to form a sequence of topology snapshots arranged in chronological order. Step 3: Identify stable communication links, analyze the topology snapshot sequence, calculate the frequency and duration of the connection relationship between each pair of communication nodes, and identify stable communication links whose frequency and duration both exceed a preset threshold. Step 4: Construct a weighted communication topology graph, with smart meters and concentrators as nodes and the stable communication links as edges, and assign weights to each edge. The weights are calculated based on a combination of signal strength, communication delay, and link stability. Step 5: Perform topology graph clustering and segmentation. Based on the weighted communication topology graph, an improved community detection algorithm is used to cluster the nodes, dividing the entire topology graph into multiple communication subnets with tight internal connections and sparse external connections. Step 6: Determine the affiliation of the distribution area. Map each identified communication subnet to an independent electrical distribution area, and determine the smart meters within the subnet as belonging to the distribution transformer corresponding to that distribution area.

2. The method for automatic division and attribution determination of transformer substations based on communication node topology identification according to claim 1, characterized in that: In step one, the concentrator interacts with the smart meter via power line carrier communication. The data collection period is adjustable from 5 to 30 minutes. The routing table records the multi-hop path information to each meter, the neighbor node list records the identifiers of the adjacent nodes that each meter communicates with directly, the signal strength is indicated by the received signal strength, and the communication delay is recorded in milliseconds.

3. The method for automatic division and attribution determination of transformer substations based on communication node topology identification according to claim 1, characterized in that: In step two, the preset time window length is 24 hours and the fixed time interval is set to 15 minutes, thereby forming a sequence containing 96 topological snapshots. Each topological snapshot contains the complete connection relationship status of all nodes at that moment.

4. The method for automatic division and attribution determination of transformer substations based on communication node topology identification according to claim 1, characterized in that: In step three, the threshold for identifying stable communication links is set as follows: the frequency of the connection relationship in the topology snapshot sequence must be greater than or equal to 90%, and the duration of a single connection must be greater than or equal to 4 hours.

5. The method for automatic division and attribution determination of transformer substations based on communication node topology identification according to claim 1, characterized in that: In step four, the weight calculation formula for the edges in the weighted communication topology graph is: weight equals the standardized signal strength multiplied by the weight coefficient. Add standardized link stability multiplied by a weighting factor Subtract the standardized communication delay and multiply by the weighting factor ,in The weighting coefficients are preset, and =1, the standardization process maps each parameter value to the range of 0 to 1.

6. The method for automatic division and attribution determination of transformer substations based on communication node topology identification according to any one of claims 1, characterized in that: In step five, the improved community detection algorithm is based on modularity optimization. Its objective function introduces edge weights as a measure of the tightness of connections within a community. During the algorithm iteration process, the modularity gain brought about by moving a node to an adjacent community is calculated, and the community division is continuously optimized until the modularity no longer increases significantly.

7. The method for automatic division and attribution determination of transformer substations based on communication node topology identification according to claim 1, characterized in that: The improved community detection algorithm also incorporates node attribute similarity, which includes the phase information of the electricity meter, voltage level, and geographical coordinates. Attribute similarity is integrated as an auxiliary constraint into the community partitioning process.

8. The method for automatic division and attribution determination of transformer substations based on communication node topology identification according to claim 1, characterized in that: Step six is ​​followed by a result verification and correction step: The preliminary determination of transformer substation affiliation is cross-validated with the limited ledger information provided by the distribution automation system. For contradictory judgment results, an auxiliary verification process based on voltage transient event correlation analysis is initiated. By injecting voltage disturbance signals with specific characteristics, the correlation of voltage waveform responses of each meter is analyzed to further confirm or correct the attribution relationship.

9. The method for automatic division and attribution determination of transformer substations based on communication node topology identification according to claim 1, characterized in that: The auxiliary verification process based on voltage transient event correlation analysis specifically includes: A voltage dip event with a duration of 100 milliseconds and an amplitude of 5% of the rated voltage is generated by the smart switch at the output of the control area via the concentrator. Simultaneously record the voltage waveforms monitored by all smart meters within the distribution area; Calculate the cross-correlation coefficient between the voltage waveforms of every two meters. Meter pairs with a cross-correlation coefficient exceeding 0.95 are considered to be in the same transformer area. The preliminary judgment results of step 6 are calibrated based on the cross-correlation matrix.

10. The method for automatic division and attribution determination of transformer substations based on communication node topology identification according to claim 1, characterized in that: The method also establishes a knowledge base for transformer area relationships to store historical judgment results, verification data, and on-site operation and maintenance feedback. Using the data in the knowledge base, a long short-term memory neural network model is trained. The long short-term memory neural network model can predict the potential change trend of transformer area relationships based on real-time communication topology characteristics, thereby achieving predictive maintenance.

Images