Intelligent fusion terminal topology identification method and system supporting multi-source data fusion

Abstract

The application discloses a method and system for identifying topology of intelligent fusion terminals supporting multi-source data fusion, and relates to the technical field of data processing. The method comprises the following steps: based on the line characteristics of the intelligent fusion terminal to be identified, making a dynamic signal injection decision to obtain an injection signal configuration strategy; injecting a detection signal into the power grid to identify local topology edges; taking the local topology edges as the first modal evidence, synchronously collecting steady-state electrical quantity measurement data to generate the second modal evidence, making a fusion decision, and outputting a preliminary identified circuit topology; when any node has a response anomaly, starting double-domain repair to generate a repaired topology structure; and according to the repaired topology structure, performing structure correlation and credible evolution to determine the final topology structure. The technical problems of line connection relationship confusion, terminal ownership relationship ambiguity and distorted topology identification results susceptible to abnormal signal interference in the prior art under complex power distribution scenarios are solved, and the technical effect of improving the topology identification accuracy under complex power distribution scenarios is achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically to a method and system for topology identification of intelligent fusion terminals that supports multi-source data fusion. Background Technology

[0002] With the continuous advancement of intelligent distribution networks and digital transformation of low-voltage distribution areas, a large number of intelligent integrated terminals, electricity meters, data acquisition devices, and edge sensing devices have been deployed in residential communities, industrial parks, and aging distribution areas to monitor line operation status, collect user electricity consumption data, and manage distribution. To support fault location, line loss analysis, load dispatching, and abnormal electricity consumption identification, the system typically needs to accurately obtain the line connection relationships and terminal affiliation relationships in the distribution network, i.e., construct the actual operating topology. However, in complex distribution scenarios, frequent historical modifications, missing line data, untimely changes in terminal wiring, and unauthorized on-site connections are common, leading to significant discrepancies between the actual line connection relationships and the records. Simultaneously, the distribution network also suffers from signal crosstalk, abnormal node communication, measurement fluctuations, and missing local data, making existing topology identification methods that rely on single measurement data or fixed rule matching prone to misjudgments, missed judgments, and unstable identification results, failing to meet the application requirements for accuracy and reliability in topology identification under complex distribution environments. Summary of the Invention

[0003] This application provides a method and system for intelligent fusion terminal topology identification that supports multi-source data fusion, which solves the technical problems in the prior art such as chaotic line connection relationships, unclear terminal affiliation relationships, and distorted topology identification results due to abnormal signal interference in complex power distribution scenarios.

[0004] The first aspect of this application provides a method for topology identification of intelligent fusion terminals supporting multi-source data fusion, the method comprising:

[0005] Based on the line characteristics of the intelligent fusion terminal to be identified, dynamic signal injection decision-making is performed to obtain the injection signal configuration strategy. Based on the injection signal configuration strategy, a detection signal carrying a unique identifier is injected into the power grid for signal detection feedback. Local topology edges are identified through multi-terminal dynamic interactive verification of the signal detection feedback information. These local topology edges are used as the first modal evidence. Steady-state electrical measurement data from each node are simultaneously collected, and second modal evidence is generated through time-series feature extraction and pattern matching. The first and second modal evidences are then used for fusion decision-making to output a preliminary identified circuit topology. When any node experiences an abnormal response, dual-domain repair is initiated. This involves time-domain repair using historical data records of the node, and spatial repair using synchronous detection data from other nodes through collaborative verification via spatial correlation, generating a repaired topology structure. Based on the repaired topology structure, a structural correlation and reliable evolution are performed to determine the final topology structure.

[0006] A second aspect of this application provides an intelligent fusion terminal topology identification system supporting multi-source data fusion, the system comprising: The system comprises the following modules: a dynamic signal injection module (using the line characteristics of the intelligent fusion terminal to be identified) and a local topology identification module (using the local topology edge as the local topology edge as the local topology edge as the local topology edge as the local topology edge as the local topology edge as the local topology edge as the local topology edge as the local topology edge as the local topology edge as the local topology edge as the first modal evidence, synchronously collecting steady-state electrical measurement data of each node, generating second modal evidence through time-series feature extraction and pattern matching, and using the first modal evidence and second modal evidence for fusion decision-making to output the preliminary identified circuit topology; a dual-domain repair module (when any node experiences an abnormal response, dual-domain repair is initiated, using the node's historical data records for time-domain repair, and using the synchronous detection data of other nodes for collaborative verification to perform spatial repair through spatial correlation, generating a repaired topology structure); and a topology determination module (based on the repaired topology structure, performing a structural correlation reliable evolution to determine the final topology structure).

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, based on the line characteristics of the intelligent fusion terminal to be identified, dynamic signal injection decision-making is performed to obtain the injection signal configuration strategy. Next, based on the injection signal configuration strategy, a detection signal carrying a unique identifier is injected into the power grid for signal detection feedback. Through multi-terminal dynamic interactive verification of the signal detection feedback information, local topology edges are identified. Then, using the local topology edges as the first modal evidence, steady-state electrical measurement data of each node are simultaneously collected. Second modal evidence is generated through time-series feature extraction and pattern matching. The first and second modal evidences are then fused for decision-making to output a preliminary identified circuit topology. When any node exhibits an abnormal response, dual-domain repair is initiated. This involves time-domain repair using historical data records of the node, and spatial repair using synchronous detection data from other nodes through collaborative verification via spatial correlation, generating a repaired topology structure. Finally, based on the repaired topology structure, a reliable evolution of structural association is performed to determine the final topology structure. This approach solves the technical problems of chaotic line connection relationships, unclear terminal affiliation, and susceptibility to abnormal signal interference in complex power distribution scenarios in existing technologies, achieving the technical effect of improving the accuracy of topology identification in complex power distribution scenarios. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a schematic flowchart of a method for topology identification of an intelligent fusion terminal that supports multi-source data fusion, provided in an embodiment of this application. Figure 2 This is a schematic diagram of the topology recognition system for an intelligent fusion terminal that supports multi-source data fusion, provided in an embodiment of this application.

[0010] Explanation of reference numerals in the attached diagram: 11 Dynamic signal injection module, 12 Local topology identification module, 13 Preliminary topology generation module, 14 Dual-domain repair module, 15 Topology determination module. Detailed Implementation

[0011] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0012] Example 1, as Figure 1 As shown, this application provides a topology identification method for intelligent fusion terminals that supports multi-source data fusion, wherein the method includes: Based on the line characteristics of the intelligent fusion terminal to be identified, dynamic signal injection decisions are made to obtain the injection signal configuration strategy.

[0013] In the low-voltage distribution network to be identified, the transformer outlet or concentrator node is used as the injection control starting point. The line operation characteristics corresponding to each monitoring node are periodically collected through an intelligent fusion terminal. These line operation characteristics include background noise power spectral density, line harmonic distortion rate, node voltage fluctuation value, line time-varying impedance, load change rate, and historical signal attenuation parameters. The collected line operation characteristics are time-aligned and subjected to sliding window statistical processing to obtain the current line state feature vector. Based on the current line state feature vector, the line interference level, signal attenuation level, and load disturbance level are calculated. The background noise power spectral density is used to identify high-interference and low-interference frequency bands in the current line, the time-varying impedance is used to assess the degree of signal attenuation in line branches, and the load change rate is used to assess the line state stability during the current period. Based on the line interference, etc. A dynamic signal injection decision matrix is ​​established based on the line interference level, signal attenuation level, and load disturbance level. Corresponding injection signal parameter combinations are generated according to preset mapping rules. Specifically, when the line interference level increases, the number of signal repetition codes and the frequency hopping range are increased; when the signal attenuation level increases, the transmit power and symbol width are increased; and when the load disturbance level increases, the duration of a single injection is reduced and the number of repetition injection rounds is increased. Subsequently, the generated candidate injection signal parameter combinations are constrained and verified. These constraints include power quality constraints, terminal communication capability constraints, and line safety operation constraints. Candidate combinations that may cause harmonic exceedances, voltage fluctuations exceeding limits, or excessive terminal processing load are eliminated. Finally, the target parameter combination with the highest communication coverage and lowest signal conflict probability is selected from the remaining candidate combinations as the injection signal configuration strategy for the current round, used for the generation and injection control of subsequent topology detection signals.

[0014] Furthermore, based on the characteristics of the power grid lines to be identified, dynamic signal injection decisions are made to obtain injection signal configuration strategies, including: At the intelligent fusion terminal side of the low-voltage distribution network to be identified, real-time operating characteristics of the lines are collected, including: background noise power spectral density, harmonic distortion rate, time-varying impedance, and load fluctuation mode of each monitoring node; based on the collected line characteristics, the injection signal parameter combination for the current round is dynamically generated, the injection signal parameter combination including signal carrier frequency, symbol width, transmit power, repetition count, and frequency hopping sequence; based on the injection signal parameter combination, the injection signal configuration strategy is obtained; wherein, the signal carrier frequency is adaptively configured in the valley frequency band of the background noise power spectrum to avoid strong interference, and the transmit power is dynamically adjusted to the lowest effective power that can penetrate all branch nodes without causing power quality exceedance limits.

[0015] Preferably, at the intelligent fusion terminal side of the low-voltage distribution network to be identified, the line status is monitored in real time through sampling circuits, carrier communication modules, and edge analysis modules. Specifically, a spectrum analysis unit performs a fast Fourier transform on the line background noise to obtain the background noise power spectral density distribution corresponding to each monitoring node; a harmonic analysis module performs harmonic decomposition on the collected voltage and current signals to obtain the line harmonic distortion rate; a periodic perturbation injection and response measurement method is used to calculate the equivalent time-varying impedance of each line branch under different time windows; simultaneously, a load fluctuation pattern sequence is established based on historical load curves and real-time power change trends; subsequently, the background noise power spectral density, harmonic distortion rate, time-varying impedance, and load fluctuation pattern corresponding to each monitoring node are time-aligned and normalized to form a current line status feature set; dynamic signal parameter decision-making is performed based on the current line status feature set, selecting multiple candidate carrier frequencies according to the low-noise interval in the background noise power spectral density, and eliminating frequency intervals overlapping with the main harmonic frequency bands based on the harmonic distribution; signal propagation attenuation in the line is estimated based on the line time-varying impedance and node level depth. The system dynamically determines the corresponding transmit power and symbol width. When the line impedance increases or the node branch level increases, the symbol width is increased and the transmit power is appropriately increased to enhance the detectability of the remote node signal. Based on the fluctuation frequency and duration in the load fluctuation mode, the repetition number and frequency hopping sequence of the detection signal are dynamically configured. When severe load fluctuations or periodic narrowband interference are detected, the repetition number is increased and the frequency hopping range is expanded to reduce the probability of the detection signal being submerged by random noise. Subsequently, the generated signal carrier frequency, symbol width, transmit power, repetition number, and frequency hopping sequence are combined to form multiple candidate injection signal parameter combinations. These combinations are then filtered based on communication coverage, signal collision probability, and power quality constraints to obtain the injection signal configuration strategy for the current round. The signal carrier frequency is adaptively configured in the valley band of the background noise power spectrum to avoid strong interference areas. The transmit power is dynamically adjusted through step-by-step probing and attenuation estimation to the minimum effective power that can cover all target branch nodes without causing voltage fluctuations to exceed limits or harmonic exceedances, thereby ensuring detection reliability while reducing the impact on the normal operation of the distribution network.

[0016] Furthermore, the strategy for obtaining the injection signal configuration also includes: The injection signal configuration strategy is structured as a multi-stage adaptive sequence consisting of multiple sub-strategies executed in chronological order. The first-stage sub-strategy uses a wide-bandwidth, low-transmit-power, long-symbol-width detection signal to perform a preliminary scan of the entire line, acquiring the basic response characteristics of each monitoring node. The second-stage sub-strategy, based on the response characteristics obtained from the first-stage sub-strategy, identifies local regions with a signal-to-noise ratio below a preset threshold or insufficient topology connectivity confidence, and dynamically generates a second injection signal parameter combination for these local regions. This second injection signal parameter combination includes one or more of narrowband, high-transmit-power, repetitive coding, or directional frequency hopping sequences, used to enhance the excitation of weak-response nodes. Subsequent stage sub-strategies, based on the fusion feedback results of the preceding stages, progressively focus injection on topology edges where arbitration conflicts still exist, until the topology identification confidence of all local regions reaches a preset convergence condition. The sub-strategies at each stage exhibit continuous, complementary, and progressive enhancement relationships, collectively constituting the injection signal configuration strategy.

[0017] Preferably, the injection signal configuration strategy is structured as a multi-stage adaptive sequence consisting of multiple sub-strategies executed in chronological order. Each stage sub-strategy is executed sequentially in a "global scan - local enhancement - conflict arbitration" manner. In the first stage sub-strategy, a wide-bandwidth, low-transmit-power, and long-symbol-width probe signal is used to perform a preliminary scan of the entire low-voltage distribution network. The wide bandwidth is used to improve frequency coverage under different line environments, the low transmit power is used to reduce the impact on normal carrier communication services and power quality, and the long symbol width is used to improve decoding stability under weak signal environments. After the first stage of probe, the corresponding signals of each monitoring node are obtained. The initial response characteristics include signal arrival delay, response strength, decoding success rate, frequency band interference level, and inter-node correlation response relationships. Subsequently, based on the initial response characteristics obtained from the first-stage sub-strategy, signal quality analysis and topology reliability assessment are performed on each local region. When the average signal-to-noise ratio of nodes within a local region is lower than a preset threshold, the node response missing rate is higher than a preset proportion, or the connection confidence of candidate topology edges is lower than a preset topology reliability threshold, the corresponding region is marked as a weak response region or a region awaiting arbitration. In the second-stage sub-strategy, a second injection signal parameter combination is dynamically generated for the weak response region or the region awaiting arbitration. The strategy employs several methods to enhance detection and excitation of weak-response nodes and complex branch regions. First, it improves anti-interference capabilities by narrowing carrier frequency bandwidth. Second, it enhances signal coverage of remote nodes by increasing transmission power. Third, it improves decoding reliability through repetitive coding mechanisms and avoids localized persistent interference frequency bands using directional frequency hopping sequences. Simultaneously, the second-stage sub-strategy only performs targeted injection on local target regions to reduce the communication load caused by repeated scanning across the entire network. In subsequent sub-strategies, based on the fusion feedback results from the previous stages, it performs step-by-step focused injection on topological edges that still have connection conflicts, ambiguities in attribution, or fluctuations in confidence. This is achieved by narrowing the detection range, improving time synchronization accuracy, and increasing... Increase the frequency of local node collaborative verification to repeatedly verify and dynamically arbitrate conflicting topological edges. After each stage, recalculate the topological identification confidence of the current local area. When the topological identification confidence of all local areas reaches the preset convergence condition, or the topological edge change rate after multiple iterations is lower than the preset change threshold, the subsequent stage injection is terminated. Among them, the sub-strategies of each stage share the response characteristics, confidence results and local topological state of the previous stage, and form a complementary and progressive enhancement relationship through a continuous feedback mechanism, which together constitute the injection signal configuration strategy to achieve progressively refined identification of topological structures in complex power distribution environments.

[0018] Based on the injection signal configuration strategy, a detection signal carrying a unique identifier is injected into the power grid for signal detection feedback. Through multi-terminal dynamic interactive verification of the signal detection feedback information, local topological edges are identified.

[0019] Specifically, a corresponding detection signal frame is generated according to the injection signal configuration strategy of the current round. This detection signal frame includes at least the injection source node identifier, the current injection round number, a time synchronization identifier, a frequency band parameter identifier, and verification coding information. The detection signal is coupled and injected into the target low-voltage distribution line according to a preset time window via the carrier communication module of the intelligent fusion terminal in the distribution area, allowing the detection signal to propagate along the distribution line and its branch nodes. Each monitoring node listens to the detection signal in the corresponding frequency band in real time through its local signal acquisition module and performs synchronous demodulation, timestamp recording, and signal quality analysis on the received signal to generate corresponding local response evidence. This local response evidence includes at least the signal arrival time, received signal strength, decoding success rate, bit error rate, and a unique identifier for the corresponding detection signal. Subsequently, a multi-terminal dynamic collaborative verification mechanism is established. Based on the location relationship, communication status, and historical connection reliability of the monitoring nodes participating in the detection in the current round, multiple monitoring nodes are dynamically divided into several collaborative verification groups. The monitoring nodes within the group broadcast and exchange their local response evidence via edge communication links or power line carrier communication links. Each node performs cross-consistency verification on the response evidence for the same detection event, comparing the signal arrival time difference, response strength attenuation trend, and decoding identifier consistency among the nodes. When multiple nodes exhibit continuous temporal correlation and stable strength attenuation along the same signal propagation path, a physical connection is determined to exist between the corresponding nodes. Furthermore, the topology edge connection confidence is calculated based on the response consistency between nodes. When the connection confidence exceeds a preset threshold, a corresponding local topology edge is generated and recorded as structured edge information containing the starting node, target node, propagation direction, and edge confidence. For node relationships with response conflicts or insufficient consistency, the corresponding topology edge is marked as an arbitration edge and fed back to the subsequent enhanced detection and repeated verification process for re-determination, thereby achieving dynamic identification and gradual confirmation of local topology structures in complex power distribution scenarios.

[0020] Furthermore, based on the injection signal configuration strategy, a probe signal carrying a unique identifier is injected into the power grid for signal detection feedback. Through multi-terminal dynamic interactive verification of the signal detection feedback information, local topological edges are identified, including: After injecting detection signals into the power grid according to the injection signal configuration strategy, the response data of each monitoring node to the detection signals is obtained as local response evidence of each node; a collaborative verification group is established, in which at least some monitoring nodes are dynamically divided into a group, and a two-way communication link is established between the nodes in the group; the local response evidence of each node is broadcast and exchanged within the collaborative verification group, a consistency determination is performed based on the exchanged response evidence of the nodes, and the local topology edge is generated based on the consistency determination result.

[0021] Preferably, after injecting probe signals into the power grid according to the injection signal configuration strategy, each monitoring node listens to the target frequency band in real time and performs synchronous demodulation, timing alignment, and bit error rate verification on the received probe signals. When a probe signal consistent with the current round's injection identifier is detected, the corresponding response data is recorded as local response evidence for each node. The local response evidence includes at least the unique identifier of the probe signal, the signal arrival timestamp, the received signal strength, the decoding success rate, the bit error rate, the frequency band occupancy status, and the node's own identification information. Subsequently, based on the current monitoring node's communication connectivity, geographical adjacency, historical topology association, and the similarity of the current round's probe response, a collaborative verification group is dynamically established. Multiple monitoring nodes with a response time window overlap exceeding a preset threshold and similar signal propagation characteristics are grouped into the same collaborative verification group. After the collaborative verification group is established, the nodes within the group establish bidirectional communication connections through edge wireless communication links, power line carrier communication links, or Ethernet links, and exchange their respective local responses using a broadcast method. Based on the evidence received, each node performs a consistency determination on the multi-node response information under the same detection event after receiving response evidence broadcast by other nodes in the group. This involves comparing the signal arrival time difference, signal strength attenuation relationship, consistency of detection signal decoding identifier, and error rate trend among the nodes. When multiple nodes meet the preset time-series propagation relationship and the signal strength change conforms to the line attenuation model, a valid connection relationship is determined to exist between the corresponding nodes. Furthermore, the connection confidence of the local topology edge is calculated based on the proportion of nodes that have reached a consensus determination, the stability of signal quality, and the continuity of the propagation path. When the connection confidence is higher than the preset determination threshold, the corresponding local topology edge is generated and represented as a structured connection record containing the starting node, target node, propagation direction, edge weight, and edge confidence. For node relationships that do not meet the consistency determination requirements, they are temporarily marked as connection relationships to be verified, and detection verification is re-executed in the subsequent enhancement injection stage to improve the accuracy and stability of local topology identification results in complex power distribution environments.

[0022] Furthermore, generating the local topological edge includes: Each node in the collaborative verification group performs intra-group broadcasting via a bidirectional communication link, exchanging its local response evidence with other nodes in the group and receiving response evidence broadcast by other nodes. Based on all received response evidence, each node in the group performs a consistency determination, comparing the signal arrival time, signal strength, and decoding identifier recorded by each node under the same injection event. When more than a preset proportion of nodes in the group reach a consensus on the existence of a certain topological edge, the corresponding topological edge is determined to be valid. Based on the valid result of the consistency determination, a local topological edge with a confidence mark is generated, wherein the confidence mark at least reflects the proportion of nodes that have reached a consensus or the signal quality.

[0023] Preferably, each node schedules its broadcast cycle according to a unified time synchronization benchmark and sends its own recorded local response evidence within a preset broadcast time window. This local response evidence includes at least a node identifier, a unique identifier for the probe signal, a signal arrival timestamp, received signal strength, bit error rate, and decoding result. Upon receiving the broadcast data, other nodes in the group write the corresponding response evidence into the group's shared response buffer and cluster the events according to the unique identifier of the probe signal, merging multiple node response data belonging to the same injection event into the same verification set. Subsequently, each node in the group performs a consistency determination based on all received response evidence, wherein, for the same injection event... The system compares the signal arrival time difference, signal strength attenuation trend, and decoding identifier consistency of the response data of each node. When the signal propagation delay between multiple nodes meets the preset delay continuity constraint, the signal strength change conforms to the line attenuation law, and the decoding identifier remains consistent, it is determined that there is a valid propagation association between the corresponding nodes. Furthermore, for each candidate topology edge, the number of nodes supporting its existence and the number of nodes opposing its existence are counted, and the consistency ratio is calculated. When more than a preset proportion of nodes in a group reach a consensus on the existence of a certain topology edge, the corresponding topology edge is determined to be valid. The consistency ratio is calculated according to the following formula: ,in, Indicates the consistency ratio. This indicates the number of nodes that determine the existence of the edge in this topology. This represents the total number of nodes participating in the current consensus determination. Subsequently, based on the valid results of the consensus determination, local topological edges with confidence labels are generated. The confidence labels comprehensively consider the consensus ratio, average signal strength, bit error rate, and historical stability indicators to calculate the comprehensive edge confidence of the corresponding topological edge. The comprehensive edge confidence is calculated according to the following formula: ,in, This represents the overall confidence level of the topological edges. Indicates the consistency ratio. This represents the average received signal strength of the nodes within the group. Indicates the preset maximum signal strength. This represents the average bit error rate. Indicators representing historical stability , , and Let be the weight coefficient, and satisfy... When the overall edge confidence is higher than the preset edge confirmation threshold, the corresponding node relationship is confirmed as a valid local topological edge, and the corresponding starting node, target node, propagation direction and edge confidence information are recorded in a structured form for subsequent multimodal fusion topology recognition process.

[0024] The local topology edge is used as the first modal evidence. Steady-state electrical measurement data of each node are collected synchronously. The second modal evidence is generated by time-series feature extraction and pattern matching. The first modal evidence and the second modal evidence are used to make a fusion decision and output the preliminary identification circuit topology.

[0025] Specifically, the local topological edges obtained through the propagation verification of the probe signal are used as the first modal evidence. This first modal evidence includes at least node connection relationships, edge propagation directions, edge confidence levels, and node hierarchical relationships. Simultaneously, steady-state electrical measurement data of each monitoring node are collected synchronously within the same time window. This steady-state electrical measurement data includes RMS voltage, RMS current, active power, reactive power, power factor angle, and phase offset. Subsequently, the collected steady-state electrical measurement data undergoes time synchronization correction, anomaly removal, normalization, and missing value compensation to form a unified time-series measurement dataset. Based on this unified time-series measurement dataset, time-series feature extraction is performed on each node. This involves extracting voltage fluctuation trend features, load change synchronization features, power transition direction features, and phase coupling features through a sliding time window, and calculating the correlation coefficient, load response synchronization rate, and phase offset difference between nodes to construct a time-series correlation feature matrix between nodes. Further, the time-series correlation feature matrix is ​​input into a preset topology pattern library for pattern matching. This preset topology pattern library contains node correlation patterns under different distribution topologies. The system employs two modalities: load coupling mode and voltage propagation mode. By calculating the matching similarity between current node characteristics and historical topology patterns, a node affiliation probability matrix and candidate connection relationships are generated as second modal evidence. Subsequently, a fusion decision is made between the first and second modal evidence. Local topological edges from the first modal evidence are used as active propagation verification results, while node association relationships from the second modal evidence are used as steady-state operation association results. Candidate connection relationships between the same node pairs are cross-validated. When both the first and second modal evidence support a node connection relationship, the overall credibility of the corresponding topological edge is increased. When there is a conflict between the two types of evidence, conflict arbitration is performed based on edge confidence, pattern matching similarity, and historical stability weights, and low-credibility topological edges are downweighted. Further, a node connection graph is constructed based on the fused candidate topological edges, and topology loop closure verification, node hierarchy continuity verification, and isolated node detection are performed on the node connection graph to eliminate abnormal connection relationships that do not meet grid connection constraints. Finally, the node connection relationships that have passed the fusion verification are output as a preliminary identification of the circuit topology for subsequent anomaly repair and reliable evolution processing.

[0026] Furthermore, the local topological edges are used as the first modal evidence. Steady-state electrical measurement data of each node are collected synchronously. Second modal evidence is generated through time-series feature extraction and pattern matching. The first and second modal evidences are then used for fusion decision-making to output a preliminary identification of the circuit topology, including: The system synchronously collects steady-state electrical measurement data from each node, including time-aligned voltage RMS value sequences, active power, reactive power, and power factor angle. It preprocesses the voltage RMS value sequences of each node to extract the correlation coefficient matrix of the voltage fluctuation curve and the power transition direction and amplitude at load abrupt changes. The extracted time-series features are matched with a preset topology pattern library to generate a node affiliation probability matrix, serving as the second modal evidence. Candidate edges from the first modal evidence and candidate edges from the second modal evidence are associated by node pairs. Edge fusion is performed based on the confidence level of the candidate edges to determine the retained topological relationship edges, generating a preliminary identification circuit topology.

[0027] Preferably, steady-state electrical measurement data from each node are collected synchronously. Each monitoring node, through its internal voltage sampling unit, current sampling unit, and synchronous clock module in an intelligent fusion terminal, periodically collects the corresponding node's effective voltage value sequence, active power, reactive power, and power factor angle under a unified time reference. Subsequently, the steady-state electrical measurement data collected from each node undergoes timestamp alignment, anomaly removal, normalization, and missing value compensation processing to form a unified time-series measurement dataset. The effective voltage value sequence of each node is preprocessed, including using moving average filtering and median filtering to remove high-frequency noise. The system suppresses and compensates for missing sampling points using linear interpolation or spline interpolation to generate a smooth voltage fluctuation curve. Based on the smooth voltage fluctuation curve, the temporal correlation between each node is calculated to obtain the correlation coefficient matrix of the voltage fluctuation curve. Simultaneously, load abrupt change detection is performed on the active power sequence of each node. The load abrupt change time is identified by calculating the power change rate between adjacent time points, and the corresponding power transition direction and amplitude are extracted. When multiple nodes exhibit power transitions in the same direction within a similar time window, a high load coupling correlation is determined between the nodes. Subsequently, the extracted... The obtained correlation coefficient matrix, power transition direction, and transition amplitude, among other time-series features, are input into a preset topology pattern library for pattern matching. This library pre-stores node correlation patterns, load propagation patterns, and hierarchical coupling patterns under different line topologies. By calculating the matching similarity between the current time-series features and each historical topology pattern, a node affiliation probability matrix is ​​generated as the second modality evidence. Further, candidate edges in the first modality evidence and candidate edges in the second modality evidence are associated according to node pairs. For candidate edges that exist in both types of modality evidence, the probability of corresponding edges is increased. For candidate edges supported by only a single modality, weighting is applied based on edge confidence, pattern matching similarity, and historical stability. Then, edge fusion is performed based on the overall edge confidence of each candidate edge. When the overall edge confidence is higher than a preset retention threshold, the corresponding candidate edge is retained as a valid topological relationship edge; when the overall edge confidence is lower than a preset removal threshold, the corresponding candidate edge is deleted. Candidate edges falling between these two thresholds are marked as edges awaiting arbitration and enter the subsequent enhancement verification stage. Finally, a node connection graph is constructed based on the retained topological relationship edges to generate a preliminary identification circuit topology.

[0028] Furthermore, the effective voltage value sequence of each node is preprocessed to extract the correlation coefficient matrix of the voltage fluctuation curve and the power transition direction and amplitude at the moment of load abrupt change, including: The voltage RMS sequence of each node is preprocessed, including denoising, normalization and missing value imputation, and the voltage fluctuation curve is extracted. Based on the voltage fluctuation curve, the time-series correlation coefficient between all node pairs is calculated to obtain the correlation coefficient matrix. Load change detection is performed on the active power sequence of each node to identify the load change time and determine the power transition direction and amplitude of each node at the corresponding change time.

[0029] Preferably, the voltage RMS value sequence of each node is preprocessed. Each monitoring node continuously collects voltage RMS data according to a unified sampling period, and the original voltage RMS value sequence undergoes denoising, normalization, and missing value imputation. The denoising process uses moving average filtering, median filtering, or low-pass filtering to remove random high-frequency noise and instantaneous abnormal spikes. The normalization process unifies the voltage sequences of different nodes according to a unified dimensional range to eliminate dimensional deviations between different nodes. The missing value imputation uses linear interpolation, spline interpolation, or neighborhood mean compensation to recover missing sampling points based on the duration of missing data. After preprocessing, the voltage fluctuation curve corresponding to each node is extracted, whereby the voltage fluctuation curve characterizes the dynamic trend of node voltage changes over time. Subsequently, based on the voltage fluctuation curve, the temporal correlation coefficient between all node pairs is calculated to obtain a correlation coefficient matrix. The correlation coefficient between any node i and node j is calculated according to the following formula: ,in, This represents the temporal correlation coefficient between node i and node j. This represents the effective voltage value of node i at time t. This represents the average voltage value of node i within the statistical time window, where T represents the length of the statistical window. This represents the effective voltage value of node j at time t. This represents the average voltage value of node j within the statistical time window. Based on the calculated correlation coefficient matrix, a set of nodes with high voltage fluctuation synchronization is identified to reflect the potential electrical connection relationships between nodes. Simultaneously, load mutation detection is performed on the active power sequence of each node. Specifically, the active power value of each node is continuously read according to a preset sampling period to construct the time series power curve for the corresponding node. The real-time power change is calculated based on the power difference between adjacent sampling times, and the power change trend is statistically analyzed using a sliding time window. When the power change at the current moment exceeds a preset mutation threshold and maintains the same direction of change over multiple consecutive sampling periods, a load mutation event is determined for the current time period, and the moment when the power change first exceeds the mutation threshold is identified as the load mutation moment. The power change is calculated according to the following formula: ,in, This represents the change in power at node i at time t. This represents the active power value of node i at time t; further, the power transition direction is determined according to the sign of the power change, wherein, when When, it is determined to be the direction of load increase, when When the load decreases, the direction is determined. At the same time, the absolute value of the power change is taken as the power transition amplitude of the corresponding node at the moment of the load change. Subsequently, the synchronization analysis of the power transition direction and power transition amplitude of multiple nodes within the same time window is performed. When multiple nodes show the same transition direction and similar trend of transition amplitude change within a close time, it is determined that there is a related load response relationship between the corresponding nodes, which is used for subsequent node affiliation analysis and topology relationship identification.

[0030] When any node experiences an abnormal response, a dual-domain repair is initiated. This involves temporal repair using historical data records from the node, and spatial repair using synchronous detection data from other nodes through collaborative verification, thereby generating a repaired topology.

[0031] During the signal injection and topology edge generation process, the response status of each monitoring node is continuously monitored. When any node is found to have missing response data, misaligned response timing, sudden signal strength changes, abnormally high bit error rate, or continuous decoding failures, the corresponding node is determined to be an abnormal response node, and a dual-domain repair process is triggered. First, time-domain repair is performed, in which historical response data records of the abnormal response node within a preset historical time window are read. The historical response data records include at least historical signal arrival time, historical signal strength, historical bit error rate, and historical topology edge association status. Subsequently, the historical response data records are processed according to time... A time-series change sequence is established sequentially, and the changing trend of historical response data is analyzed using a sliding time window. Based on the current anomaly type, a corresponding time-series compensation algorithm is used for repair. Specifically, when the anomaly type is a single missing point, linear interpolation or spline interpolation is used to recover the current missing data point; when the anomaly type is a continuous fluctuation anomaly, a weighted moving average or autoregressive prediction model is used to estimate the current response value, thereby generating a time-domain repaired response dataset. Subsequently, spatial repair is performed, which involves reading the synchronous detection data of other monitoring nodes within the collaborative verification group to which the response anomaly node belongs under the same injection event, and based on the nodes... A spatial association set is constructed based on historical topological relationships, physical adjacency relationships, and signal propagation paths. The response correlation between abnormal nodes and neighboring nodes is analyzed based on this spatial association set. Specifically, the expected response data of the abnormal node is estimated by comparing the signal propagation delay difference, signal strength attenuation trend, and load fluctuation synchronization among neighboring nodes. Further, the estimation results provided by each neighboring node are weighted and fused according to their spatial association weights to generate a spatially repaired response dataset. Subsequently, the temporally repaired response dataset and the spatially repaired response dataset are fused. Temporal and spatial weights are dynamically allocated based on the historical stability of the current node, the consistency of neighboring nodes, and the duration of the anomaly. When historical data stability is high, the temporal repair weight is increased; when neighboring node consistency is high, the spatial repair weight is increased. The original abnormal response data is replaced with the fused repair results, and the corresponding node's local topological edge generation, consistency determination, and edge confidence calculation processes are re-executed. Finally, the node connection graph is reconstructed based on the repaired effective topological edges to generate a repaired topology structure, thereby improving the continuity and stability of topology identification in anomaly scenarios under complex power distribution environments.

[0032] Furthermore, generating the repaired topology includes: When any monitoring node is detected to have missing response data, misaligned timing, or abnormal signal strength after the probe signal is injected, it is identified as a response anomaly node, and a dual-domain repair process is executed. This involves: acquiring the response data sequence recorded by the response anomaly node within a preset historical time window, using a temporal interpolation algorithm to complete the missing or abnormal data points, generating a temporally repaired response dataset; acquiring the response data synchronously detected by other monitoring nodes in the collaborative verification group to which the response anomaly node belongs, estimating the expected response data of the corresponding nodes based on spatial correlation, and generating a spatially repaired response dataset; merging the temporally repaired response dataset with the spatially repaired response dataset, replacing the original abnormal response data of the response anomaly node, and re-executing the topology edge generation process based on the repaired complete response data, outputting the repaired topology structure.

[0033] Preferably, after receiving the detection signal, each monitoring node performs real-time integrity verification, time synchronization verification, and signal quality analysis on the response data. If the target detection signal is not detected for multiple consecutive sampling periods, the signal arrival time deviates from the preset synchronization window, or the received signal strength deviates from the historical average by more than a preset anomaly threshold, the corresponding node is determined to have a response anomaly. Subsequently, the response data sequence recorded by the anomaly node within the preset historical time window is obtained. The response data sequence includes at least historical signal arrival time, historical signal strength, historical bit error rate, and historical side confidence. A historical time-series response curve is constructed from the response data sequence in chronological order, and a time-series interpolation algorithm is used to complete the currently missing or abnormal data points. When the number of abnormal data points is small, a linear interpolation algorithm is used for compensation. When the continuous anomaly time is long, spline interpolation or an autoregressive prediction algorithm based on a sliding window is used for estimation to generate a response dataset after time-domain repair. Subsequently, the response data synchronously detected by other monitoring nodes in the collaborative verification group to which the response anomaly node belongs is obtained, and the expected response data of the corresponding node is estimated based on spatial correlation. Specifically, a spatial correlation matrix is ​​established based on the historical topological connection relationship, physical adjacency relationship of lines, and historical signal propagation consistency between nodes. By comparing the signal propagation delay difference, signal strength attenuation trend, and load fluctuation synchronization between adjacent nodes, the spatial correlation weight of each neighboring node to the response anomaly node is calculated, and the expected response data of the current anomaly node is estimated using a weighted average method to generate a response dataset after spatial domain repair. The spatial correlation weight is calculated according to the following formula: ,in, This represents the spatial association weight of node j with the response anomaly node i. The coefficient of spatial correlation between node i and node j is represented by N, where N represents the number of neighboring nodes participating in spatial repair. The sum of spatial correlation coefficients between the response anomaly node i and all neighboring nodes participating in spatial repair is used to normalize the spatial correlation of each node. i represents the current response anomaly node number, j represents the target neighboring node number participating in spatial repair, and k represents the neighboring node index number participating in the summation operation. Subsequently, the corresponding response data are weighted and fused according to the spatial correlation weights of each neighboring node to estimate the spatial predicted response value of the response anomaly node. Further, the response dataset after temporal repair is fused with the response dataset after spatial repair, where temporal and spatial weights are dynamically allocated based on the stability of historical data and the consistency of neighboring nodes to obtain the final fused repair result. Then, the original abnormal response data of the response anomaly node is replaced with the final fused repair result, and the node consistency determination, local topology edge generation, and edge confidence calculation processes are re-executed based on the repaired complete response data. Finally, a node connection graph is constructed based on the newly generated valid topology edges, and the repaired topology structure is output to improve the continuity and reliability of topology identification in abnormal states under complex power distribution scenarios.

[0034] Based on the repaired topology, a reliable evolution of structural associations is performed to determine the final topology.

[0035] Furthermore, based on the repaired topology, a structural association reliable evolution is performed to determine the final topology, including: Based on the repaired topology, a confidence index is calculated for each topological edge and the overall topological relationship, including edge existence confidence, node-level consistency evaluation, and multimodal evidence consistency determination. Each topological relationship edge is determined based on the confidence index; edges that meet the preset confidence requirements are directly adopted as valid topological edges. For edges that do not meet the determination requirements, adaptive signal injection is re-executed to obtain new response evidence and update the corresponding local topological edges. If a local change in the power grid is detected, only the affected area is incrementally updated, while the existing topological relationships in the unchanged areas are retained. Based on the topological relationship edges determined after confidence determination and incremental update processing, the final topology is obtained.

[0036] The edge existence confidence is used to characterize the degree of credibility of a candidate topology edge in the process of multiple rounds of active detection and collaborative verification; the node hierarchy consistency evaluation is used to characterize whether the power supply hierarchy relationship between the nodes connected by the current candidate topology edge meets the hierarchical structure constraints of the actual power distribution network; the multimodal evidence consistency judgment is used to characterize the degree of consistency between the first modal evidence formed by active detection and the second modal evidence formed by steady-state measurement for the same candidate topology edge.

[0037] The edge existence confidence calculation includes: First, calculating the percentage of nodes whose existence is confirmed by the collaborative verification group in this round of detection to obtain the consistency ratio; then, calculating the continuous occurrence frequency of the candidate topology edge in multiple consecutive historical identification periods, and combining it with the average received signal strength and average bit error rate of the corresponding detection signal for comprehensive evaluation; the edge existence confidence is calculated according to the following formula: ,in, This indicates that the edge has a confidence level. This indicates the consensus rate in the current round. This indicates the number of times a candidate topological edge appears consecutively in a historical period. This represents the total number of historical periods in the statistics. This represents the average bit error rate of the signal corresponding to the current candidate edge. , and Let be the weight coefficient, and satisfy... .

[0038] The node-level consistency evaluation calculation includes: ,in, This represents the node-level consistency evaluation value. This indicates the power supply level number of node i. This indicates the power supply level number of node j. This represents the maximum level depth in the current topology; the smaller the node level difference and the more normal the hierarchical propagation relationship, the higher the node level consistency evaluation value.

[0039] The calculation of the consistency determination of multimodal evidence includes: ,in, This represents the consistency evaluation value of multimodal evidence. This represents the probability of the existence of the corresponding candidate edge in the first modality of evidence. This represents the pattern matching probability of the corresponding node connection relationship in the second modality evidence. The smaller the difference between the two probabilities, the more consistent the support of the two modal evidences for the current candidate edge, and the higher the corresponding multimodal evidence consistency evaluation value.

[0040] Preferably, a comprehensive confidence level is calculated for each candidate topological edge, and the comprehensive confidence level is calculated according to the following formula: ,in, This represents the overall credibility of candidate topological edges. This indicates that the edge has a confidence level. This represents the node-level consistency evaluation value. This represents the consistency evaluation value of multimodal evidence. , and Let be the weight coefficient, and satisfy... Subsequently, based on the aforementioned credibility index, each topological relationship edge is determined. When the overall credibility of a candidate topological edge is higher than a preset credibility confirmation threshold, the corresponding topological edge is directly adopted as a valid topological edge; when the overall credibility is lower than a preset rejection threshold, the corresponding topological edge is deleted. For candidate topological edges between the confirmation threshold and the rejection threshold, they are determined as low-credibility edges to be arbitrated, and adaptive signal injection is re-executed. Specifically, based on the interference level, node response stability, and historical misjudgment frequency of the local region where the edge to be arbitrated is located, the re-injected signal carrier frequency, transmit power, repetition count, and frequency hopping sequence are dynamically adjusted to obtain new response evidence and update the corresponding local topological edge. Subsequently, the edge existence confidence calculation and multimodal consistency determination are re-executed using the updated local topological edge to improve the complexity... The reliability of edge connections in the scenario is assessed. Furthermore, if a local change in the power grid is detected, incremental updates are performed only on the affected area. This involves identifying locally changed areas by monitoring node additions, node exits, load migrations, switch status changes, and line reconfiguration events. For areas without structural changes, existing stable topology relationships are retained, and the entire network topology identification is not re-executed. For affected areas, the detection signal injection, local topology edge generation, and reliability assessment processes are re-executed to update the corresponding local topology structure. Finally, based on the topology relationship edges determined after reliability assessment and incremental update processing, the entire network node connection graph is reconstructed to obtain the final topology structure. The corresponding topology edge set, node hierarchy relationships, and edge reliability results are output for subsequent distribution network operation analysis and topology management.

[0041] In summary, the embodiments of this application have at least the following technical effects: First, based on the line characteristics of the intelligent fusion terminal to be identified, dynamic signal injection decision-making is performed to obtain the injection signal configuration strategy. Next, based on the injection signal configuration strategy, a detection signal carrying a unique identifier is injected into the power grid for signal detection feedback. Through multi-terminal dynamic interactive verification of the signal detection feedback information, local topology edges are identified. Then, using the local topology edges as the first modal evidence, steady-state electrical measurement data of each node are simultaneously collected. Second modal evidence is generated through time-series feature extraction and pattern matching. The first and second modal evidences are then fused for decision-making to output a preliminary identified circuit topology. When any node exhibits an abnormal response, dual-domain repair is initiated. This involves time-domain repair using historical data records of the node, and spatial repair using synchronous detection data from other nodes through collaborative verification via spatial correlation, generating a repaired topology structure. Finally, based on the repaired topology structure, a reliable evolution of structural association is performed to determine the final topology structure. This approach solves the technical problems of chaotic line connection relationships, unclear terminal affiliation, and susceptibility to abnormal signal interference in complex power distribution scenarios in existing technologies, achieving the technical effect of improving the accuracy of topology identification in complex power distribution scenarios.

[0042] Example 2, based on the same inventive concept as the intelligent fusion terminal topology identification method supporting multi-source data fusion in the foregoing examples, such as... Figure 2 As shown, this application provides an intelligent fusion terminal topology identification system that supports multi-source data fusion, wherein the system includes: Dynamic signal injection module 11: Based on the line characteristics of the intelligent fusion terminal to be identified, it makes dynamic signal injection decisions and obtains the injection signal configuration strategy; Local topology identification module 12: Based on the injection signal configuration strategy, it injects a detection signal carrying a unique identifier into the power grid, performs signal detection feedback, and identifies local topology edges through multi-terminal dynamic interactive verification of the signal detection feedback information; Preliminary topology generation module 13: Using the local topology edges as the first modal evidence, it synchronously collects steady-state electrical measurement data of each node, generates second modal evidence through time-series feature extraction and pattern matching, and uses the first modal evidence and the second modal evidence to make fusion decisions and outputs the preliminary identified circuit topology; Dual-domain repair module 14: When any node has an abnormal response, dual-domain repair is initiated, in which time-domain repair is performed using the data records of the node's historical moments, and spatial repair is performed through spatial correlation using the synchronous detection data of other nodes to generate a repaired topology structure; Topology determination module 15: Based on the repaired topology structure, it performs structural association reliable evolution to determine the final topology structure.

[0043] Furthermore, the dynamic signal injection module 11 is used to perform the following method: At the intelligent fusion terminal side of the low-voltage distribution network to be identified, real-time operating characteristics of the lines are collected, including: background noise power spectral density, harmonic distortion rate, time-varying impedance, and load fluctuation mode of each monitoring node; based on the collected line characteristics, the injection signal parameter combination for the current round is dynamically generated, the injection signal parameter combination including signal carrier frequency, symbol width, transmit power, repetition count, and frequency hopping sequence; based on the injection signal parameter combination, the injection signal configuration strategy is obtained; wherein, the signal carrier frequency is adaptively configured in the valley frequency band of the background noise power spectrum to avoid strong interference, and the transmit power is dynamically adjusted to the lowest effective power that can penetrate all branch nodes without causing power quality exceedance limits.

[0044] Furthermore, the dynamic signal injection module 11 is used to perform the following method: The injection signal configuration strategy is structured as a multi-stage adaptive sequence consisting of multiple sub-strategies executed in chronological order. The first-stage sub-strategy uses a wide-bandwidth, low-transmit-power, long-symbol-width detection signal to perform a preliminary scan of the entire line, acquiring the basic response characteristics of each monitoring node. The second-stage sub-strategy, based on the response characteristics obtained from the first-stage sub-strategy, identifies local regions with a signal-to-noise ratio below a preset threshold or insufficient topology connectivity confidence, and dynamically generates a second injection signal parameter combination for these local regions. This second injection signal parameter combination includes one or more of narrowband, high-transmit-power, repetitive coding, or directional frequency hopping sequences, used to enhance the excitation of weak-response nodes. Subsequent stage sub-strategies, based on the fusion feedback results of the preceding stages, progressively focus injection on topology edges where arbitration conflicts still exist, until the topology identification confidence of all local regions reaches a preset convergence condition. The sub-strategies at each stage exhibit continuous, complementary, and progressive enhancement relationships, collectively constituting the injection signal configuration strategy.

[0045] Furthermore, the local topology identification module 12 is used to perform the following method: After injecting detection signals into the power grid according to the injection signal configuration strategy, the response data of each monitoring node to the detection signals is obtained as local response evidence of each node; a collaborative verification group is established, in which at least some monitoring nodes are dynamically divided into a group, and a two-way communication link is established between the nodes in the group; the local response evidence of each node is broadcast and exchanged within the collaborative verification group, a consistency determination is performed based on the exchanged response evidence of the nodes, and the local topology edge is generated based on the consistency determination result.

[0046] Furthermore, the local topology identification module 12 is used to perform the following method: Each node in the collaborative verification group performs intra-group broadcasting via a bidirectional communication link, exchanging its local response evidence with other nodes in the group and receiving response evidence broadcast by other nodes. Based on all received response evidence, each node in the group performs a consistency determination, comparing the signal arrival time, signal strength, and decoding identifier recorded by each node under the same injection event. When more than a preset proportion of nodes in the group reach a consensus on the existence of a certain topological edge, the corresponding topological edge is determined to be valid. Based on the valid result of the consistency determination, a local topological edge with a confidence mark is generated, wherein the confidence mark at least reflects the proportion of nodes that have reached a consensus or the signal quality.

[0047] Furthermore, the preliminary topology generation module 13 is used to perform the following method: The system synchronously collects steady-state electrical measurement data from each node, including time-aligned voltage RMS value sequences, active power, reactive power, and power factor angle. It preprocesses the voltage RMS value sequences of each node to extract the correlation coefficient matrix of the voltage fluctuation curve and the power transition direction and amplitude at load abrupt changes. The extracted time-series features are matched with a preset topology pattern library to generate a node affiliation probability matrix, serving as the second modal evidence. Candidate edges from the first modal evidence and candidate edges from the second modal evidence are associated by node pairs. Edge fusion is performed based on the confidence level of the candidate edges to determine the retained topological relationship edges, generating a preliminary identification circuit topology.

[0048] Furthermore, the preliminary topology generation module 13 is used to perform the following method: The voltage RMS sequence of each node is preprocessed, including denoising, normalization and missing value imputation, and the voltage fluctuation curve is extracted. Based on the voltage fluctuation curve, the time-series correlation coefficient between all node pairs is calculated to obtain the correlation coefficient matrix. Load change detection is performed on the active power sequence of each node to identify the load change time and determine the power transition direction and amplitude of each node at the corresponding change time.

[0049] Furthermore, the dual-domain repair module 14 is used to perform the following method: When any monitoring node is detected to have missing response data, misaligned timing, or abnormal signal strength after the probe signal is injected, it is identified as a response anomaly node, and a dual-domain repair process is executed. This involves: acquiring the response data sequence recorded by the response anomaly node within a preset historical time window, using a temporal interpolation algorithm to complete the missing or abnormal data points, generating a temporally repaired response dataset; acquiring the response data synchronously detected by other monitoring nodes in the collaborative verification group to which the response anomaly node belongs, estimating the expected response data of the corresponding nodes based on spatial correlation, and generating a spatially repaired response dataset; merging the temporally repaired response dataset with the spatially repaired response dataset, replacing the original abnormal response data of the response anomaly node, and re-executing the topology edge generation process based on the repaired complete response data, outputting the repaired topology structure.

[0050] Furthermore, the topology determination module 15 is used to perform the following method: Based on the repaired topology, a confidence index is calculated for each topological edge and the overall topological relationship, including edge existence confidence, node-level consistency evaluation, and multimodal evidence consistency determination. Each topological relationship edge is determined based on the confidence index; edges that meet the preset confidence requirements are directly adopted as valid topological edges. For edges that do not meet the determination requirements, adaptive signal injection is re-executed to obtain new response evidence and update the corresponding local topological edges. If a local change in the power grid is detected, only the affected area is incrementally updated, while the existing topological relationships in the unchanged areas are retained. Based on the topological relationship edges determined after confidence determination and incremental update processing, the final topology is obtained.

[0051] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for topology identification of intelligent fusion terminals supporting multi-source data fusion, characterized in that, The method includes: Based on the line characteristics of the intelligent fusion terminal to be identified, dynamic signal injection decision is made to obtain the injection signal configuration strategy; Based on the injection signal configuration strategy, a detection signal carrying a unique identifier is injected into the power grid for signal detection feedback. Local topological edges are identified through multi-terminal dynamic interactive verification of the signal detection feedback information. The local topological edge is used as the first modal evidence. Steady-state electrical measurement data of each node are collected synchronously. The second modal evidence is generated by time-series feature extraction and pattern matching. The first modal evidence and the second modal evidence are used to make a fusion decision and output the preliminary identification circuit topology. When any node experiences an abnormal response, a dual-domain repair is initiated. This involves temporal repair using historical data records from the node, and spatial repair using synchronous detection data from other nodes through collaborative verification, thereby generating a repaired topology. Based on the repaired topology, a reliable evolution of structural associations is performed to determine the final topology.

2. The intelligent fusion terminal topology identification method supporting multi-source data fusion according to claim 1, characterized in that, Based on the characteristics of the power grid lines to be identified, dynamic signal injection decisions are made to obtain injection signal configuration strategies, including: On the intelligent fusion terminal side of the low-voltage distribution network to be identified, the real-time operating characteristics of the line are collected, including: background noise power spectral density, harmonic distortion rate, time-varying impedance and load fluctuation mode of each monitoring node. Based on the collected line characteristics, the injection signal parameter combination for the current round is dynamically generated. The injection signal parameter combination includes signal carrier frequency, symbol width, transmit power, repetition count, and frequency hopping sequence. Based on the combination of injected signal parameters, an injection signal configuration strategy is obtained; The signal carrier frequency is adaptively configured in the valley band of the background noise power spectrum to avoid strong interference, and the transmission power is dynamically adjusted to the lowest effective power that can penetrate all branch nodes without causing power quality exceedances.

3. The intelligent fusion terminal topology identification method supporting multi-source data fusion according to claim 2, characterized in that, The strategy for obtaining the injection signal configuration also includes: The injection signal configuration strategy is configured as a multi-stage adaptive sequence consisting of multiple sub-strategies executed in chronological order; The first-stage sub-strategy uses a wide-bandwidth, low-power, long-symbol-width detection signal to perform a preliminary scan of the entire line and obtain the basic response characteristics of each monitoring node. The second-stage sub-strategy identifies local regions with signal-to-noise ratios below a preset threshold or insufficient topological connectivity confidence based on the response characteristics obtained from the first-stage sub-strategy. It then dynamically generates a second injection signal parameter combination for the local region. The second injection signal parameter combination includes one or more of narrowband, high transmit power, repetitive coding, or directional frequency hopping sequences to enhance the excitation of weak response nodes. The subsequent sub-strategy, based on the fusion feedback results of the previous stage, focuses on injecting topological edges that still have arbitration conflicts step by step until the topological identification confidence of all local regions reaches the preset convergence condition. Among them, there are continuous, complementary and progressively reinforcing relationships between the sub-strategies of each stage, which together constitute the injection signal configuration strategy.

4. The intelligent fusion terminal topology identification method supporting multi-source data fusion according to claim 1, characterized in that, Based on the injection signal configuration strategy, a probe signal carrying a unique identifier is injected into the power grid for signal detection and feedback. Through multi-terminal dynamic interactive verification of the signal detection feedback information, local topological edges are identified, including: After injecting the detection signal into the power grid according to the injection signal configuration strategy, the response data of each monitoring node to the detection signal is obtained as evidence of the local response of each node. Establish a collaborative verification group, wherein at least some monitoring nodes are dynamically divided into a group, and a two-way communication link is established between the nodes in the group. Within the collaborative verification group, each node broadcasts and exchanges its local response evidence, performs a consistency determination based on the exchanged response evidence, and generates the local topology edge based on the consistency determination result.

5. The intelligent fusion terminal topology identification method supporting multi-source data fusion according to claim 4, characterized in that, Generating the local topological edge includes: Each node in the collaborative verification group performs intra-group broadcasting through a two-way communication link, exchanging its local response evidence with other nodes in the group, and receiving response evidence broadcast by other nodes. Each node in the group performs a consistency determination based on all received response evidence, comparing the signal arrival time, signal strength, and decoding identifier recorded by each node under the same injection event. When more than a preset proportion of nodes in the group reach a consensus on the existence of a certain topological edge, the corresponding topological edge is determined to be valid. Based on the valid results of the consensus determination, local topological edges with confidence labels are generated, wherein the confidence labels at least reflect the proportion of nodes that have reached consensus or the signal quality.

6. The intelligent fusion terminal topology identification method supporting multi-source data fusion according to claim 5, characterized in that, Using the local topological edges as the first modal evidence, steady-state electrical measurement data of each node are collected synchronously. Second modal evidence is generated through time-series feature extraction and pattern matching. The first and second modal evidences are then used for fusion decision-making to output a preliminary identification of the circuit topology, including: The steady-state electrical measurement data of each node are collected synchronously, including time-aligned voltage RMS value sequence, active power, reactive power, and power factor angle. The voltage RMS value sequence of each node is preprocessed to extract the correlation coefficient matrix of the voltage fluctuation curve and the power transition direction and amplitude at the moment of load change. The extracted temporal features are matched with a preset topological pattern library to generate a node affiliation probability matrix, which serves as the second modality evidence. The candidate edges in the first modal evidence and the candidate edges in the second modal evidence are associated by node pairs. The edge fusion judgment is performed based on the confidence of the candidate edges to determine the retained topological relationship edges and generate a preliminary identification circuit topology.

7. The intelligent fusion terminal topology identification method supporting multi-source data fusion according to claim 6, characterized in that, The voltage RMS sequence of each node is preprocessed to extract the correlation coefficient matrix of the voltage fluctuation curve and the power transition direction and amplitude at the moment of load abrupt change, including: The voltage RMS sequence of each node is preprocessed, including denoising, normalization and missing value imputation, and the voltage fluctuation curve is extracted. Based on the voltage fluctuation curve, calculate the time correlation coefficient between all node pairs to obtain the correlation coefficient matrix; Load change detection is performed on the active power sequence of each node to identify the moment of load change and determine the power transition direction and magnitude of each node at the corresponding moment of change.

8. The intelligent fusion terminal topology identification method supporting multi-source data fusion according to claim 4, characterized in that, Generate a repaired topology, including: When any monitoring node is detected to have missing response data, misaligned timing, or abnormal signal strength after the detection signal is injected, it is identified as an abnormal response node and a dual-domain repair process is executed. Among them, the response data sequence recorded by the abnormal response node within a preset historical time window is obtained, and the missing or abnormal data points are filled in using a time-series interpolation algorithm to generate a time-domain repaired response dataset. Obtain the response data synchronously detected by other monitoring nodes within the collaborative verification group to which the response anomaly node belongs, estimate the expected response data of the corresponding node based on spatial correlation, and generate the response dataset after spatial domain repair. The response dataset after temporal domain repair is merged with the response dataset after spatial domain repair, the original abnormal response data of the abnormal response nodes is replaced, and the topology edge generation process is re-executed based on the repaired complete response data to output the repaired topology structure.

9. The intelligent fusion terminal topology identification method supporting multi-source data fusion according to claim 1, characterized in that, Based on the repaired topology, a reliable evolution of structural associations is performed to determine the final topology, including: Based on the repaired topology, a confidence index is calculated for each topological edge and the overall topological relationship, including edge existence confidence, node hierarchy consistency evaluation, and multimodal evidence consistency determination. Based on the credibility index, each topological relationship edge is determined. Edges that meet the preset credibility requirements are directly adopted as valid topological edges. For edges that do not meet the determination requirements, adaptive signal injection is triggered again to obtain new response evidence and update the corresponding local topological edges. If a local change in the power grid is detected, only the affected area will be incrementally updated, while the existing topology of the unchanged area will be preserved. The final topological structure is obtained based on the topological relationship edges determined after credibility determination and incremental update processing.

10. A topology recognition system for intelligent fusion terminals supporting multi-source data fusion, characterized in that, The system is used to implement the intelligent fusion terminal topology identification method supporting multi-source data fusion as described in any one of claims 1-9, the system comprising: Dynamic signal injection module: Based on the line characteristics of the intelligent fusion terminal to be identified, it makes dynamic signal injection decisions and obtains the injection signal configuration strategy; Local topology identification module: Based on the injection signal configuration strategy, a detection signal carrying a unique identifier is injected into the power grid, and signal detection feedback is performed. Through multi-terminal dynamic interactive verification of the signal detection feedback information, local topology edges are identified. Preliminary topology generation module: The local topology edge is used as the first modal evidence. Steady-state electrical measurement data of each node are collected synchronously. The second modal evidence is generated by time-series feature extraction and pattern matching. The first modal evidence and the second modal evidence are used to make a fusion decision and output the preliminary identification circuit topology. Dual-domain repair module: When any node experiences an abnormal response, dual-domain repair is initiated. This involves performing temporal repair using historical data records of the node, and performing spatial repair by leveraging the synchronous detection data of other nodes through collaborative verification to achieve spatial correlation, thereby generating a repaired topology. Topology determination module: Based on the repaired topology, perform a reliable evolution of structural associations to determine the final topology.

Images