A low-voltage transformer area topology intelligent identification method and system

Abstract

The application provides a low-voltage transformer area topology intelligent identification method and system, and relates to the technical field of electrical monitoring. The method comprises the following steps: according to first time series measurement data, performing double-layer disturbance splitting to generate second time series measurement data with disturbance labels; according to non-exception intervals, performing node response differential analysis to generate first correlation data; extracting voltage sudden drop slope, sudden drop starting point, recovery speed and node response sequence, forming a transient feature vector, and correcting the first correlation data according to the transient feature vector to generate second correlation data; checking the periodicity of voltage fluctuations in the micro-perturbation section, identifying the response offset across branches, and adjusting the second correlation data to generate final correlation data; reconstructing the branch structure, and determining the subordinate relationship among the main nodes, branch nodes and end nodes through node sequence consistency and correlation strength. The application improves the autonomy and accuracy of low-voltage transformer area topology intelligent identification.

Description

Technical Field

[0001] This invention relates to the field of electrical monitoring technology, and in particular to a method and system for intelligent identification of low-voltage distribution area topology. Background Technology

[0002] In existing low-voltage distribution substations, commonly used topology identification techniques mainly rely on line carrier signal injection or rule-based matching of user-end measurement data to determine the connection relationship between user meters and substation transformers. The carrier injection method typically injects a signal of a specific frequency band onto the low-voltage bus, and then collects the attenuation characteristics at each user meter to determine the branch to which the meter belongs. Another type of measurement-based topology identification technique relies on the synchronous acquisition of voltage and current curves within the substation, comparing characteristics such as voltage drop amplitude and fluctuation correlation to infer the topology of each branch. Theoretically, these methods can provide basic topology information for substation condition assessment, line loss analysis, and fault location.

[0003] However, in scenarios where intermittent high-power pulse loads (such as fast charging stations for electric vehicles) are connected at the end of residential areas, the aforementioned technologies are prone to identification biases. Taking the carrier injection method as an example, when a pulse load causes drastic fluctuations in local line impedance within a short period, the carrier signal received at the user's meter will experience unstable attenuation, causing the algorithm to misclassify users originally located on the same branch as different branches. Furthermore, the voltage correlation-based identification method, when a pulse load causes local voltage transient dips at certain nodes, can generate "spurious correlation" signals from user nodes not belonging to that branch, thus disrupting the correlation calculation and leading to incorrect topology inference. Such misidentifications directly affect subsequent business processes such as line loss allocation, transformer load prediction, and the location of electricity theft users. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for intelligent identification of low-voltage distribution area topology, which aims to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, a method for intelligent identification of low-voltage transformer area topology, the method comprising:

[0007] Voltage and current measurement data of each user node are collected from the low-voltage distribution area and normalized and calibrated. The first time series measurement data is generated by unifying the sampling interval, amplitude range and offset.

[0008] A two-layer perturbation decomposition is performed on the first time series measurement data. By calculating the change gradient and stationarity, the intervals that meet the pulse load condition are marked as the first abnormal intervals, and the intervals that meet the perturbation condition are marked as the second abnormal intervals. Local abrupt change points in the first abnormal intervals are stripped to generate the second time series measurement data with perturbation labels.

[0009] Based on the non-abnormal intervals of the second time series measurement data, node response differential analysis is performed. First correlation data is generated by calculating voltage direction consistency, current increment comparison and differential stability. The first correlation data includes correlation strength to represent the degree of correlation between user nodes.

[0010] Based on the first abnormal interval of the second time series measurement data, the voltage drop slope, drop start point, recovery speed and node response sequence are extracted to form a transient feature vector. The first correlation data is then corrected based on the transient feature vector to generate the second correlation data.

[0011] Based on the second abnormal interval of the second time series measurement data, the periodicity of voltage fluctuation in the perturbation segment is verified, the cross-branch response offset is identified, and the second correlation data is adjusted to generate the final correlation data.

[0012] The branch structure is reconstructed based on the final correlation data. The subordinate relationships of the main nodes, branch nodes and terminal nodes are determined by the consistency of node order and the correlation strength, and the low-voltage transformer area topology identification results are generated.

[0013] Preferably, a two-layer perturbation decomposition is performed on the first time series measurement data. By calculating the change gradient and stationarity, the intervals that meet the impulse load condition are marked as the first abnormal intervals, and the intervals that meet the perturbation condition are marked as the second abnormal intervals. Local abrupt changes within the first abnormal intervals are then removed to generate second time series measurement data with perturbation labels, including:

[0014] The amplitude and direction of change of adjacent sampling points of the first time series measurement data are calculated to generate change gradient data for disturbance determination.

[0015] The average value of the change amplitude, the consistency of the change direction, and the uniformity of the change rate of the first time series measurement data within the continuous sampling window are analyzed to generate stationarity data to characterize stability.

[0016] Based on the changing gradient data, the sampling interval with a continuous sudden increase in the change amplitude is determined as the candidate interval for pulse disturbance;

[0017] Based on the stationarity data, sampling intervals with small change amplitudes and high directional consistency are identified as candidate intervals for perturbations.

[0018] The candidate intervals for impulse perturbations are marked as the first anomalous intervals, and the candidate intervals for minor perturbations are marked as the second anomalous intervals, generating initial interval data with anomalous interval labels;

[0019] Local mutation detection is performed on the gradient data in the first abnormal interval, and sampling points that meet the preset mutation characteristics are extracted as independent mutation point data;

[0020] The initial interval data is integrated with the independent mutation point data to generate a second time series measurement data with perturbation labels.

[0021] Preferably, based on the first abnormal interval of the second time series measurement data, the voltage sag slope, sag initiation point, recovery rate, and node response sequence are extracted to form a transient feature vector. The first correlation data is then corrected based on the transient feature vector to generate second correlation data, including:

[0022] Voltage change data of each user node within the first abnormal interval are extracted from the second time series measurement data to generate transient voltage data for sequence determination.

[0023] Based on the transient voltage data, the voltage sag slope, voltage sag initiation point, and voltage recovery speed of each user node are determined to form corresponding transient characteristic data;

[0024] Based on the order of voltage drop start times in the transient voltage data, the order of node response of each user node is determined, node sequence data is generated, and the transient feature data and node sequence data are combined to form a transient feature vector.

[0025] Based on the node response sequence and electrical position relationship reflected by the transient feature vector, the correlation strength between nodes in the first correlation data is adjusted item by item to generate the second correlation data.

[0026] Preferably, based on the second abnormal interval of the second time series measurement data, the periodicity of voltage fluctuations in the perturbation segment is verified, cross-branch response offsets are identified, and the second correlation data is adjusted to generate the final correlation data, including:

[0027] Voltage fluctuation data of each user node within the second abnormal interval are extracted from the second time series measurement data to generate perturbation fluctuation data for periodic analysis.

[0028] The repetition time interval of voltage fluctuations at each user node is calculated based on the perturbation fluctuation data to generate voltage fluctuation periodic data.

[0029] Based on the voltage fluctuation period data, the period consistency of different user nodes is compared, nodes with period differences exceeding the preset difference range are identified, and cross-branch response offset node data is formed.

[0030] The node association strength corresponding to the cross-branch response offset node data is down-processed in the second association data to generate the down-processed association data.

[0031] The node association strength corresponding to nodes with consistent voltage fluctuation periods is adjusted upward in the second association data to generate the adjusted association data.

[0032] The adjusted relationship data is merged with the adjusted relationship data to generate the final relationship data.

[0033] Preferably, the branch structure is reconstructed based on the final association data, and the subordinate relationships of trunk nodes, branch nodes, and terminal nodes are determined through node order consistency and association strength, generating low-voltage transformer area topology identification results, including:

[0034] Based on the final association data, extract the node combination with the highest association strength to generate candidate data for backbone nodes;

[0035] Generate backbone node sequence data based on the order of node responses in the backbone node candidate data;

[0036] Branch nodes are identified based on the decreasing trend of the correlation strength between nodes in the main node sequence data, and branch node data is generated.

[0037] Generate terminal node data based on the correlation strength between branch node data and its subordinate nodes;

[0038] The data of the main nodes, branch nodes, and terminal nodes are reconstructed into a branch structure according to the node hierarchy, generating the low-voltage transformer area topology identification results.

[0039] Preferably, based on the chronological order of the voltage drop start times in the transient voltage data, the node response sequence of each user node is determined, node sequence data is generated, and the transient feature data is combined with the node sequence data to form a transient feature vector, including:

[0040] Extract the voltage drop start time of each user node from the transient voltage data to generate voltage drop start time data;

[0041] Based on the descent start time data, user nodes are sorted chronologically to generate initial response sequence data for the nodes;

[0042] Based on the node group whose descent start time difference is lower than the preset time difference range in the initial response sequence data, the voltage descent slope data of the corresponding node is extracted, and the node response sequence data after differentiation is formed by comparing the voltage descent slope.

[0043] Based on the differentiated node response sequence data and transient feature data, the transient feature vector is generated by combining them according to the preset field order.

[0044] Preferably, based on the node response sequence and electrical position relationship reflected by the transient feature vector, the correlation strength between nodes in the first correlation data is adjusted item by item to generate the second correlation data, including:

[0045] The response sequence of each user node is analyzed based on the transient feature vector to generate node response sequence data;

[0046] Based on the node response sequence data, the relative position of each user node in the electrical path is inferred to form node position relationship data;

[0047] Based on the node position relationship data, the node association strength in the first association relationship data is adjusted item by item. The association strength of nodes with the same response order is increased, and the association strength of nodes with opposite response order is decreased, thus generating the adjusted association relationship data.

[0048] Based on the adjusted association data, inconsistencies in the association strength between the detected nodes are addressed, and the rules are further modified to generate the second association data.

[0049] Secondly, a low-voltage distribution area topology intelligent identification system, the system comprising:

[0050] The data acquisition and calibration module is used to collect voltage and current measurement data from each user node in the low-voltage distribution area and perform normalization calibration processing. By unifying the sampling interval, amplitude range and offset, it generates the first time series measurement data.

[0051] The two-layer perturbation decomposition module is used to perform two-layer perturbation decomposition on the first time series measurement data. By calculating the change gradient and stationarity, the intervals that meet the pulse load condition are marked as the first abnormal intervals, and the intervals that meet the perturbation condition are marked as the second abnormal intervals. The local abrupt change points in the first abnormal intervals are stripped to generate the second time series measurement data with perturbation labels.

[0052] The node response differential analysis module is used to perform node response differential analysis based on the non-abnormal interval of the second time series measurement data. It generates first correlation data by calculating voltage direction consistency, current increment comparison and differential stability. The first correlation data includes correlation strength to represent the degree of correlation between user nodes.

[0053] The transient feature correction module is used to extract the voltage sag slope, sag initiation point, recovery rate and node response sequence from the first abnormal interval of the second time series measurement data to form a transient feature vector, and to correct the first correlation data based on the transient feature vector to generate the second correlation data.

[0054] The perturbation period verification module is used to verify the periodicity of voltage fluctuations in the perturbation segment based on the second abnormal interval of the second time series measurement data, identify cross-branch response offsets, adjust the second correlation data, and generate the final correlation data.

[0055] The topology reconstruction module is used to reconstruct the branch structure based on the final association data. It determines the subordinate relationships of trunk nodes, branch nodes, and terminal nodes through node order consistency and association strength, and generates the low-voltage transformer area topology recognition results.

[0056] The above-described solution of the present invention has at least the following beneficial effects:

[0057] First, by normalizing and calibrating the voltage and current measurement data collected from the low-voltage distribution area, the measurement curves of different user nodes are kept consistent in terms of sampling interval, amplitude range, and offset. This avoids the impact of differences in measurement equipment accuracy and inconsistent data scales on correlation analysis, making the basic data for subsequent disturbance identification and inter-node response comparison more reliable.

[0058] Furthermore, this invention, by performing a two-layer perturbation decomposition on the normalized time-series measurement data, can effectively distinguish between severe perturbations caused by pulsed loads and minor perturbations caused by end-load fluctuations, and can also isolate local abrupt changes within the pulsed perturbation interval. This processing method can reduce the interference of strong perturbations such as pulsed loads on the overall trend identification, avoid the pulse noise misjudgment problem existing in carrier injection methods and traditional correlation methods, thereby making the perturbation pattern label more accurate.

[0059] Based on this, by performing differential analysis of node responses in non-abnormal intervals, this invention can comprehensively extract information such as voltage direction consistency, current increment variation relationship, and node response stability to establish a preliminary node correlation model. This model, based on the cross-node response patterns under steady-state conditions, overcomes the problem of existing technologies relying solely on overall fluctuation amplitude while ignoring detailed response trends, thus improving the effectiveness of correlation judgment.

[0060] Subsequently, this invention extracts the voltage descent slope, descent start point, recovery rate, and node response sequence within the first abnormal interval. These features are combined into a transient feature vector, which is used to correct the preliminary correlation. This step utilizes the propagation characteristics of pulsed load disturbances between nodes to perform reverse verification of node position relationships, helping to eliminate "spurious correlations" caused by pulsed loads, thereby further improving the reliability of the correlation.

[0061] Furthermore, this invention utilizes the periodicity of perturbations within the second abnormal interval to determine the consistency of responses between nodes under small voltage fluctuations, thereby identifying periodically offset nodes across branch paths and adjusting their correlation strength. This process can further filter node relationships with weak coupling but high transient similarity, making the correlation relationships closer to the actual line structure.

[0062] Through the multi-stage data processing and correlation analysis described above, this invention can ultimately reconstruct the main path, branch paths, and terminal nodes based on the correlation strength and node sequence relationship, outputting a complete topology identification result. Compared with traditional methods that rely solely on carrier signals or overall voltage correlation, this invention can maintain high topology identification stability in actual transformer substation environments with pulse loads, transient disturbances, or minor noise. Attached Figure Description

[0063] Figure 1 This is a flowchart of a low-voltage distribution area topology intelligent identification method provided by an embodiment of the present invention. Detailed Implementation

[0064] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0065] like Figure 1 As shown, an embodiment of the present invention proposes a method for intelligent identification of low-voltage distribution area topology, the method comprising:

[0066] Voltage and current measurement data of each user node are collected from the low-voltage distribution area and normalized and calibrated. The first time series measurement data is generated by unifying the sampling interval, amplitude range and offset.

[0067] A two-layer perturbation decomposition is performed on the first time series measurement data. By calculating the change gradient and stationarity, the intervals that meet the pulse load condition are marked as the first abnormal intervals, and the intervals that meet the perturbation condition are marked as the second abnormal intervals. Local abrupt change points in the first abnormal intervals are stripped to generate the second time series measurement data with perturbation labels.

[0068] Based on the non-abnormal intervals of the second time series measurement data, node response differential analysis is performed. First correlation data is generated by calculating voltage direction consistency, current increment comparison and differential stability. The first correlation data includes correlation strength to represent the degree of correlation between user nodes.

[0069] Based on the first abnormal interval of the second time series measurement data, the voltage drop slope, drop start point, recovery speed and node response sequence are extracted to form a transient feature vector. The first correlation data is then corrected based on the transient feature vector to generate the second correlation data.

[0070] Based on the second abnormal interval of the second time series measurement data, the periodicity of voltage fluctuation in the perturbation segment is verified, the cross-branch response offset is identified, and the second correlation data is adjusted to generate the final correlation data.

[0071] The branch structure is reconstructed based on the final correlation data. The subordinate relationships of the main nodes, branch nodes and terminal nodes are determined by the consistency of node order and the correlation strength, and the low-voltage transformer area topology identification results are generated.

[0072] In this embodiment of the invention, by normalizing and calibrating the voltage and current measurement data collected from the low-voltage distribution area, the data from different user nodes are brought to a uniform scale. This helps to reduce the impact of acquisition noise, calibration differences between different devices, and inconsistent sampling step sizes on subsequent analysis, thereby providing a more stable data foundation for subsequent disturbance identification and topology inference.

[0073] A two-level perturbation decomposition is performed on the normalized time series measurement data. Through comprehensive analysis of change gradients and stationarity, it is possible to distinguish between drastic perturbation regions affected by pulsed loads and minor perturbation regions caused by terminal load fluctuations. By stripping away local abrupt change points within anomalous regions, the abrupt change components in anomalous regions can be prevented from interfering with subsequent structural judgments, making the time series used for topology analysis more representative.

[0074] After identifying non-abnormal intervals, node response differential analysis can be used to extract information such as the consistency of voltage direction changes among user nodes, the comparison of current increments caused by load changes, and load response stability, under conditions free from strong disturbances. This yields preliminary correlation strengths between user nodes. These correlation strengths reflect the degree of mutual influence between nodes, providing foundational data for further topology inference.

[0075] After identifying the first abnormal interval, a transient feature vector is formed by extracting the voltage sag slope, voltage sag initiation point, recovery rate, and the sequence of node responses. This vector characterizes the disturbance propagation properties. When the transient feature vector is used to correct the initial correlation strength, it can compensate for node relationships that steady-state differential analysis cannot accurately distinguish, making the correlation strength more consistent with the sequence and propagation characteristics in the actual electrical path, thereby improving the reliability of node relationships.

[0076] After identifying the second abnormal interval, the response characteristics of nodes under slight disturbances can be obtained by detecting the periodicity of voltage fluctuations in the perturbation segment. By comparing the consistency of the periods among nodes, offset nodes that do not belong to the same branch can be identified, and the correlation strength can be adjusted accordingly to further improve the ability to distinguish node relationships. After this step, the correlation between nodes is more stable and can better reflect the actual connection status of the distribution line.

[0077] After obtaining the final association relationship, the subordinate relationship between nodes is reconstructed. The order and hierarchy of trunk nodes, branch nodes and terminal nodes are determined by the consistency of node order and the strength of association. This can yield a topology result that is closer to the real line structure, providing effective input information for load analysis, anomaly detection, line loss assessment and operation and maintenance decision-making in the transformer area.

[0078] For example, in a low-voltage distribution area of ​​a residential community, during peak electricity consumption periods and in scenarios with frequent electric vehicle charging, this method can utilize voltage and current measurement data acquired by conventional data acquisition terminals. Through a combination of disturbance identification, steady-state analysis, transient correction, and micro-disturbance verification, it can identify the actual connection topology between user meters and distribution transformers. For instance, when a voltage drop occurs on a branch, this method can determine the node location relationship based on the order of voltage drop and recovery characteristics. Then, combined with periodic consistency under micro-disturbance, it can screen out misidentified nodes and ultimately determine the connection distribution of each user's meters on the actual line. This provides a reference for subsequent maintenance personnel to conduct line segment inspections or load management.

[0079] In a preferred embodiment of the present invention, node response differential analysis is performed based on the non-abnormal interval of the second time series measurement data. First correlation data is generated by calculating voltage direction consistency, current increment comparison, and differential stability. Specifically, this includes:

[0080] The time periods that were not marked as the first or second abnormal intervals were selected from the second time series measurement data to form non-abnormal interval data for differential analysis.

[0081] Based on the voltage change trend of adjacent sampling points in the non-abnormal interval, the voltage change direction of each user node is determined. When the voltage change direction of two user nodes is consistent at the same sampling time, it is recorded as a consistent direction event. The cumulative number of consistent direction events between each node pair is counted to form voltage direction consistency data.

[0082] Based on the current changes of each user node within the non-abnormal range, the ratio of the current increment of one node to the current increment of another node is calculated. The similarity of the ratio is used to evaluate the consistency of the response of the two nodes under the same load change, thus forming current increment comparison data.

[0083] Based on the stability of voltage direction consistency data and current increment comparison data over time, the reliability of the node pair response relationship under non-abnormal conditions is evaluated, and differential stability data is generated.

[0084] The voltage direction consistency data, current increment comparison data, and differential stability data are weighted and integrated to obtain the first correlation data that characterizes the strength of the response relationship between each node pair. The degree of correlation between nodes is then used to form the correlation strength for subsequent topology inference.

[0085] In a preferred embodiment of the present invention, a two-layer perturbation decomposition is performed on the first time series measurement data. By calculating the change gradient and stationarity, the intervals that meet the impulse load condition are marked as the first abnormal intervals, and the intervals that meet the perturbation condition are marked as the second abnormal intervals. Local abrupt changes within the first abnormal intervals are stripped to generate second time series measurement data with perturbation labels, including:

[0086] The amplitude and direction of change of adjacent sampling points of the first time series measurement data are calculated to generate change gradient data for disturbance determination.

[0087] The average value of the change amplitude, the consistency of the change direction, and the uniformity of the change rate of the first time series measurement data within the continuous sampling window are analyzed to generate stationarity data to characterize stability.

[0088] Based on the changing gradient data, the sampling interval with a continuous sudden increase in the change amplitude is determined as the candidate interval for pulse disturbance;

[0089] Based on the stationarity data, sampling intervals with small change amplitudes and high directional consistency are identified as candidate intervals for perturbations.

[0090] The candidate intervals for impulse perturbations are marked as the first anomalous intervals, and the candidate intervals for minor perturbations are marked as the second anomalous intervals, generating initial interval data with anomalous interval labels;

[0091] Local mutation detection is performed on the gradient data in the first abnormal interval, and sampling points that meet the preset mutation characteristics are extracted as independent mutation point data;

[0092] The initial interval data is integrated with the independent mutation point data to generate a second time series measurement data with perturbation labels.

[0093] In this embodiment of the invention, by analyzing the amplitude and direction of change in the normalized time-series measurement data, gradient data for distinguishing between significant and minor disturbances can be obtained. Further utilizing the stability detection of a continuous sampling window, segments with stable characteristics in load fluctuations can be distinguished from segments affected by pulse impacts. Pulse disturbance candidate intervals formed based on the gradient data and micro-disturbance candidate intervals formed based on the stability data can provide a preliminary classification of different disturbance sources, which helps to improve the targeting of subsequent processing. By identifying local abrupt change points in the first abnormal interval and separating them from the overall disturbance interval, the abrupt change information can avoid biasing the overall disturbance trend judgment, allowing the second time-series measurement data to more completely reflect the disturbance evolution law, thereby providing relatively accurate input data for steady-state differential analysis and transient feature extraction.

[0094] In a preferred embodiment of the present invention, a preset mutation feature is used to determine whether a sampling point has a sudden and instantaneous change attribute, so as to separate it from the overall pulse perturbation curve, specifically including:

[0095] First, the magnitude of the change exhibits abnormally abrupt jumps within a short period of time.

[0096] By analyzing the change amplitude of sampling points, when the change amplitude of a certain sampling point relative to multiple sampling points before and after it increases significantly, and the increase exceeds the amplitude difference range used for mutation identification, it can be preliminarily considered that the sampling point has mutation characteristics.

[0097] The amplitude difference range is determined based on the measurement noise level, sampling accuracy, and normal fluctuation range of the load in the transformer area, enabling those skilled in the art to set it through normal data statistics.

[0098] Second, the direction of change may exhibit a sharp reversal or instability.

[0099] When the direction of change of a certain sampling point is inconsistent with the direction of change of multiple sampling points before and after it, and the magnitude of the reverse change is large, it can be regarded as a sudden change;

[0100] Or, if the direction of change of the sampling point shows a continuous abnormal jump, such as suddenly changing from rising to falling, then it meets the characteristics of a directional mutation.

[0101] Third, the rate of change deviates significantly from the trend of neighboring sampling points.

[0102] By comparing the rate of change of a sampling point within a small time interval with the rate of change of its adjacent sampling points, a sudden change can be considered when the rate of change exceeds the tolerance range of the rate deviation obtained from the load operation statistics of the distribution area.

[0103] The deviation range is set based on the equipment sampling rate, normal fluctuation range, and load response characteristics.

[0104] Fourth, the mutations are extremely short-lived and not continuous.

[0105] When a significant jump occurs at a sampling point, and no similar trend is observed at the sampling points before and after, it can be determined that the jump is an instantaneous disturbance rather than a continuous disturbance.

[0106] Such short-term fluctuations are usually related to voltage or current spikes, instantaneous interference during equipment startup, etc., and need to be extracted separately and cannot be mixed with the overall interval for processing.

[0107] Based on the above multiple judgment conditions, when a sampling point simultaneously meets any combination of conditions such as abnormal change amplitude, abrupt change in change direction, or deviation in change speed, it can be determined that it meets the preset mutation characteristics.

[0108] This embodiment enables the system to accurately separate local spike changes by setting the above-mentioned mutation characteristics, avoiding such instantaneous interference from misleading subsequent disturbance pattern analysis, and helping to improve the reliability of overall transformer area topology identification.

[0109] In a preferred embodiment of the present invention, the change amplitude and change direction of adjacent sampling points of the first time series measurement data are calculated to generate change gradient data for disturbance determination, specifically including:

[0110] The voltage and current values ​​of adjacent sampling points are read sequentially from the first time series measurement data. The change between the previous sampling point and the next sampling point is calculated, and the absolute magnitude of the change is recorded as the change amplitude.

[0111] The direction of change is determined by the positive or negative trend of the change amplitude. When the value of the subsequent sampling point is higher than that of the previous sampling point, it is recorded as an upward direction, and when it is lower than that of the previous sampling point, it is recorded as a downward direction.

[0112] Based on the combination of change amplitude and change direction, change gradient data is formed to reflect the sensitivity to disturbance. When the change amplitude is large and the change direction is obvious, it can be initially regarded as input data for areas where pulse disturbances may exist.

[0113] In a preferred embodiment of the present invention, the average value of the change amplitude, the consistency of the change direction, and the uniformity of the change rate of the first time series measurement data within a continuous sampling window are analyzed to generate stationarity data, specifically including:

[0114] The first time series measurement data is divided into multiple continuous fixed-length sampling windows. The average value of the voltage and current changes within each window is calculated to reflect the overall degree of change within the window.

[0115] The direction of change of continuous sampling points within the window is statistically analyzed. When the direction of change of most sampling points is consistent, the window is considered to have high directional consistency.

[0116] Analyze whether the magnitude of the change within the window is stable. When the difference in the increase or decrease of adjacent changes is small, it indicates that the rate of change is relatively uniform.

[0117] Based on characteristics such as small average amplitude of change, high directional consistency, and uniform rate of change, it is determined whether the window has stationary features. The above results are recorded as stationarity data in numerical or grade form for subsequent identification of perturbation candidate intervals.

[0118] In a preferred embodiment of the present invention, sampling intervals with continuously increasing amplitudes are determined as candidate intervals for pulse perturbations based on gradient change data; sampling intervals with smaller amplitudes and higher directional consistency are determined as candidate intervals for micro-perturbations based on stability data, specifically including:

[0119] The sampling segment with a significant increase in continuous change amplitude is selected from the changing gradient data. When the change amplitude of multiple sampling points exceeds the set amplitude change threshold and the change direction changes abruptly in a short period of time, the sampling segment is divided into the impulse disturbance candidate interval.

[0120] Identify sampling segments from stationarity data that have small amplitude changes, highly consistent directions of change, and uniform rates of change, and classify these sampling segments as perturbation candidate intervals.

[0121] The candidate regions for impulse perturbation and micro-perturbation are marked for subsequent steps of abnormal region division, mutation point removal, and transient feature identification.

[0122] In a preferred embodiment of the present invention, the initial interval data and independent mutation point data are integrated to generate second time series measurement data with perturbation labels, specifically including:

[0123] Read the first and second abnormal interval labels from the initial interval data and match them with each sampling point of the first time series measurement data to give each sampling point an abnormal attribute identifier.

[0124] Write the aforementioned independent mutation point data into the corresponding sampling point location, and add a mutation label to the sampling point to characterize the sudden change characteristics of the sampling point;

[0125] During the label integration process, if an independent mutation point is located within the first abnormal interval, the mutation label is superimposed as an additional attribute onto the sampling point without changing the interval label, for fine-tuning in the subsequent transient feature extraction stage.

[0126] After integration, the output second time series measurement data not only includes the original measurement data, but also anomaly type labels and mutation labels, enabling subsequent transient feature identification and perturbation period analysis to be performed based on a more accurate data structure.

[0127] In a preferred embodiment of the present invention, based on the first abnormal interval of the second time series measurement data, the voltage sag slope, sag initiation point, recovery rate, and node response sequence are extracted to form a transient feature vector. The first correlation data is then corrected based on the transient feature vector to generate second correlation data, including:

[0128] Voltage change data of each user node within the first abnormal interval are extracted from the second time series measurement data to generate transient voltage data for sequence determination.

[0129] Based on the transient voltage data, the voltage sag slope, voltage sag initiation point, and voltage recovery speed of each user node are determined to form corresponding transient characteristic data;

[0130] Based on the order of voltage drop start times in the transient voltage data, the order of node response of each user node is determined, node sequence data is generated, and the transient feature data and node sequence data are combined to form a transient feature vector.

[0131] Based on the node response sequence and electrical position relationship reflected by the transient feature vector, the correlation strength between nodes in the first correlation data is adjusted item by item to generate the second correlation data.

[0132] In this embodiment of the invention, by extracting features such as voltage descent slope, voltage descent start point, and recovery speed from the first abnormal interval, the response strength and recovery characteristics of different nodes during disturbance propagation can be reflected. By identifying the order of voltage descent start times, the sequential relationship of user nodes on the transient propagation path can be obtained. For nodes with relatively close descent start times, comparing the voltage descent slope can supplement the judgment of differences between nodes, making the response order more refined. The transient feature vector formed by combining the above features can accurately reflect the relative positional relationship between nodes. When using the transient feature vector to correct the initial correlation strength, it can compensate for the deficiency of steady-state analysis in identifying the sequential path of nodes, making the correlation relationship between nodes more consistent with the actual electrical connection sequence, and providing support for the accuracy of subsequent topology inference.

[0133] In a preferred embodiment of the present invention, the voltage sag slope, voltage sag initiation point, and voltage recovery rate of each user node are determined based on transient voltage data to form corresponding transient characteristic data, specifically including:

[0134] The voltage change sequence marked as the first abnormal interval is extracted from the second time series measurement data and used as transient voltage data to analyze the disturbance propagation characteristics;

[0135] In transient voltage data, identify the sampling point where the voltage begins to drop significantly, and use the time index of this sampling point as the starting point of the voltage drop, recording it as the node's drop start marker;

[0136] After the voltage sag initiation point, the change in voltage drop amplitude is detected according to the sampling sequence. When the voltage value shows a decreasing trend between consecutive sampling points, the rate of voltage drop is determined by comparing the change ratio of the drop amplitude over time. The drop rate is used as the voltage sag slope to describe the instantaneous sensitivity of the node to disturbances.

[0137] After the voltage reaches its lowest value, its recovery process is monitored. Based on the rate at which the voltage gradually rises from its lowest value to its steady-state value, the recovery rate is determined by comparing the relationship between the recovery amplitude and the recovery time required, which is used to reflect the recovery characteristics of the node.

[0138] The voltage sag initiation point, voltage sag slope, and recovery speed are combined into transient feature data, providing input for subsequent determination of the order of node responses and correction of correlation strength.

[0139] In a preferred embodiment of the present invention, based on the second abnormal interval of the second time series measurement data, the periodicity of voltage fluctuations in the perturbation segment is verified, cross-branch response offsets are identified, and the second correlation data is adjusted to generate final correlation data, including:

[0140] Voltage fluctuation data of each user node within the second abnormal interval is extracted from the second time series measurement data to generate perturbation fluctuation data for periodic analysis. The voltage fluctuation refers to the change in voltage between consecutive sampling points not exceeding the preset amplitude difference range determined based on sampling accuracy, measurement noise and steady-state operation characteristics of the transformer area. When the change falls within this range, the voltage can be considered to be in a perturbation state.

[0141] The repetition time interval of voltage fluctuations at each user node is calculated based on the perturbation fluctuation data to generate voltage fluctuation periodic data.

[0142] Based on the voltage fluctuation period data, the period consistency of different user nodes is compared, nodes with period differences exceeding the preset difference range are identified, and cross-branch response offset node data is formed.

[0143] The node association strength corresponding to the cross-branch response offset node data is down-processed in the second association data to generate the down-processed association data.

[0144] The node association strength corresponding to nodes with consistent voltage fluctuation periods is adjusted upward in the second association data to generate the adjusted association data.

[0145] The adjusted relationship data is merged with the adjusted relationship data to generate the final relationship data.

[0146] In this embodiment of the invention, by extracting micro-amplitude voltage fluctuation data from the second abnormal interval, the periodic variation law of nodes under slight disturbances can be obtained. Different nodes, due to their different positions on the line, will exhibit differences in their disturbance propagation characteristics, showing either consistent or offset periods. The periodic consistency between nodes can be judged using the disturbance fluctuation period data. When the period offset of a node exceeds a preset difference range, it can be considered that it does not belong to the same branch path, thus forming a cross-branch response offset node. Downgrading the association strength of offset nodes can reduce their interference with topology inference; upgrading nodes with consistent periods can improve their stability in the association relationship. By merging the association relationships again, a final association relationship closer to the actual line structure can be obtained, providing suitable input for subsequent topology reconstruction.

[0147] In a preferred embodiment of the present invention, the node association strength corresponding to the cross-branch response offset node data is down-adjusted in the second association data to generate down-adjusted association data, specifically including:

[0148] Nodes whose period differences exceed the preset difference range are selected from the results of the perturbation fluctuation period consistency analysis and recorded as cross-branch response offset nodes.

[0149] In the second correlation data, find the correlation strength between the cross-branch response offset node and other nodes. Based on the degree of period difference reflected by the offset node, reduce the corresponding correlation strength according to a preset ratio. The reduction range can be divided into levels according to the degree of offset. When the degree of offset is greater, a higher reduction ratio can be used.

[0150] In the downsizing process, the original structure of the node pairs is kept unchanged, and only the association strength value is adjusted, so that the influence of the offset node on the topology judgment is reduced, and it should no longer be regarded as a direct association object of the same branch.

[0151] The reduced association strengths of each node will form a new set of associations, providing a more stable data foundation for subsequent association merging.

[0152] In a preferred embodiment of the present invention, the node association strength corresponding to nodes with consistent voltage fluctuation periods is adjusted in the second association data to generate adjusted association data, specifically including:

[0153] Extract node pairs with high period consistency from the results of perturbation fluctuation period analysis. When the period difference between two nodes is within a preset difference range, they are considered to be nodes with consistent periods.

[0154] In the second set of relational data, the correlation strength between these periodically consistent nodes is located, and their correlation strength is increased according to different consistency levels. When the periodic consistency is higher, a higher increase can be given so that it can better reflect the stable correlation between the two nodes.

[0155] During the enhancement process, the node-to-relationship structure remains unchanged, and only the numerical values ​​are amplified to increase their weight in the final integration of relational data.

[0156] The adjusted node pair association strengths are integrated into adjusted association relationship data, which serves as an important input for the next step of node relationship merging and topology identification.

[0157] In a preferred embodiment of the present invention, a preset difference range is used to determine whether the periodic differences of each user node belong to the same perturbation response feature, specifically including:

[0158] Based on the sampling frequency, data noise distribution, and average periodic variation of voltage fluctuations under normal conditions of the acquisition device, the periodic fluctuations of the nodes under no significant load changes are statistically analyzed to obtain the periodic reference values ​​of multiple nodes under steady-state conditions.

[0159] Based on the statistically obtained periodic benchmark value, the possible range of period offset of nodes under normal conditions is calculated. When the period offset value is within this range, the period difference between nodes is considered to meet the consistency requirements.

[0160] Based on parameters such as power grid structure scale, line length, end load variation characteristics, and transformer capacity, the cycle deviation tolerance is adjusted. When the distribution of transformer substations is relatively dispersed and the response delay is large, the range of difference can be appropriately increased.

[0161] Ultimately, the difference range was set to an interval based on the actual operating environment of the transformer area and the measurement accuracy, which was used to determine whether the cycle was consistent or offset, so that the perturbation cycle determination could be reliably executed under actual operating conditions.

[0162] In a preferred embodiment of the present invention, the branch structure is reconstructed based on the final association data, and the subordinate relationships of the main nodes, branch nodes, and terminal nodes are determined through node order consistency and association strength, generating a low-voltage transformer area topology identification result, including:

[0163] Based on the final association data, extract the node combination with the highest association strength to generate candidate data for backbone nodes;

[0164] Generate backbone node sequence data based on the order of node responses in the backbone node candidate data;

[0165] Branch nodes are identified based on the decreasing trend of the correlation strength between nodes in the main node sequence data, and branch node data is generated.

[0166] Generate terminal node data based on the correlation strength between branch node data and its subordinate nodes;

[0167] The data of the main nodes, branch nodes, and terminal nodes are reconstructed into a branch structure according to the node hierarchy, generating the low-voltage transformer area topology identification results.

[0168] In this embodiment of the invention, by analyzing the final correlation data, the node combinations most likely to be located on the trunk path in the distribution area can be determined based on the correlation strength distribution between nodes, thus forming candidate data for trunk nodes. Further combining the node response sequence, the arrangement order of the trunk path can be determined, making the subordinate relationships between trunk nodes more consistent with electrical propagation characteristics. Based on the trunk node sequence data and the decreasing trend of correlation strength, branch nodes separated from the trunk nodes can be identified, and their relative positions within the branches can be distinguished. By continuing to analyze the correlation strength between branch nodes and their subordinate nodes, terminal nodes can be determined, gradually clarifying the hierarchical relationship of the topology. When the trunk node data, branch node data, and terminal node data are combined into a complete structure, complete low-voltage distribution area topology information can be obtained, allowing the actual line connection relationships to be restored, providing a structured reference for subsequent load analysis, anomaly location, and equipment inspection.

[0169] In a preferred embodiment of the present invention, extracting the node combination with the highest association strength based on the final association relationship data to generate candidate backbone node data specifically includes:

[0170] Read the association strength values ​​between all node pairs from the final association data, sort the node pairs, and select the node pairs with higher association strength as the focus of analysis.

[0171] Identify nodes that appear repeatedly in multiple high-strength node pairs. For example, when a node shows a high correlation strength with different nodes, it indicates that the node may play the role of connecting multiple branches in the electrical path.

[0172] The above-mentioned high-frequency nodes are combined with their adjacent nodes with the highest correlation strength to form a preliminary backbone node candidate set, which is used to represent the nodes that are more likely to constitute the main path of power transmission in actual lines.

[0173] The candidate set is checked for continuity. When multiple nodes form a stable chain structure according to their association strength, they are recorded as candidate data for the backbone node, providing a basis for subsequent path order inference.

[0174] In a preferred embodiment of the present invention, generating terminal node data based on the correlation strength between branch node data and its subordinate nodes specifically includes:

[0175] Read the list of subordinate nodes for each branch node from the identified branch node data. These subordinate nodes typically appear in the final association data with low association strength or few connections.

[0176] For each branch node, sort the association strength between its corresponding subordinate nodes. When a subordinate node no longer connects to other lower-level nodes in the structure, or when its association strength is only high with the branch node and weak with other nodes, the node can be determined as an end node.

[0177] By checking whether there are further connecting chains between lower-level nodes, a node is classified as an end node when it no longer shows an upward or downward trend in all association strengths and is at the end of the tree structure.

[0178] The terminal nodes corresponding to each branch node are aggregated to generate terminal node data, providing clear endpoint information for topology reconstruction.

[0179] In a preferred embodiment of the present invention, the trunk node data, branch node data, and terminal node data are reconstructed into a branch structure according to the node hierarchy relationship to form a topology structure identification result, specifically including:

[0180] Based on the association order of nodes in the candidate data of the backbone nodes, the backbone nodes are connected sequentially to form a backbone path with a continuous relationship.

[0181] For each main node, read its corresponding branch node data. When the association strength between the branch node and the main node meets the branch judgment condition, add the branch node as the starting point of the branch to the topology.

[0182] Continue searching downwards along the lower-level nodes of the branch node, using the correlation strength trend to determine the hierarchical relationship between nodes, and connect the nodes level by level to form a complete branch path;

[0183] For each branch path, place the previously identified end node at the end of the path, so that the topology structure is tree-like or chain-like hierarchical;

[0184] By combining the main path with each branch path, the connection relationships between nodes are uniformly organized to generate a topology identification result covering all user nodes in the transformer area, enabling it to accurately represent the actual connection layout of the line.

[0185] In a preferred embodiment of the present invention, the node response sequence of each user node is determined according to the chronological order of the voltage drop start times in the transient voltage data, node sequence data is generated, and the transient feature data and node sequence data are combined to form a transient feature vector, including:

[0186] Extract the voltage drop start time of each user node from the transient voltage data to generate voltage drop start time data;

[0187] Based on the descent start time data, user nodes are sorted chronologically to generate initial response sequence data for the nodes;

[0188] Based on the node group whose descent start time difference is lower than the preset time difference range in the initial response sequence data, the voltage descent slope data of the corresponding node is extracted, and the node response sequence data after differentiation is formed by comparing the voltage descent slope.

[0189] Based on the differentiated node response sequence data and transient feature data, the transient feature vector is generated by combining them according to the preset field order.

[0190] In this embodiment of the invention, by extracting the voltage drop start time of each user node, the response order of the nodes during disturbance propagation can be quantified into comparable data. Sort the drop start times to form a basic response order structure. For nodes whose drop time differences are within a preset time difference range and are difficult to distinguish directly, comparing their voltage drop slopes can further refine the priority order of nodes during disturbance propagation, making the determination of node response order more reliable. Combining the calibrated response order with the corresponding transient feature data forms a transient feature vector with propagation order characteristics, making the relative path relationship between nodes clearer. This feature vector can serve as an important basis for subsequent correlation strength correction, helping to correct node position relationships that cannot be accurately identified in steady-state analysis and improving the accuracy of node correlation judgment.

[0191] In a preferred embodiment of the present invention, a preset time difference range is used to determine whether the onset time of voltage drop between nodes is difficult to distinguish directly, thereby triggering a further comparison of the voltage sag slope, specifically including:

[0192] The fall start time of multiple user nodes in the same pulse disturbance event is statistically analyzed from transient voltage data and sorted to analyze the natural distribution differences in fall time among nodes;

[0193] Based on the sampling interval, data refresh cycle, and voltage measurement noise of the sampling device, the possible time error between adjacent sampling points is determined, and this time error is used as the minimum time difference reference value that must not be lower than.

[0194] Based on factors such as the overall rate of voltage drop at nodes, line length in transformer substations, differences in electrical distance, and load response inertia, a reasonable tolerance range for the drop time difference is derived. When the drop start time difference between two nodes is less than this range, it is considered that the sampling error may affect the direct sorting, and the slope should be used to further distinguish them.

[0195] The time difference range obtained by combining the sampling error range, node response characteristics, and voltage change trend is set as the preset time difference range. This is used to screen node groups whose descent start time is difficult to distinguish directly, so that the triggering conditions for subsequent slope comparison have a clear basis.

[0196] In a preferred embodiment of the present invention, a preset field order is used to combine multiple transient features into a transient feature vector in a fixed format, so that it can be uniformly parsed in the subsequent correlation strength correction process, specifically including:

[0197] Based on the formation order of transient characteristics, identify the characteristic fields involved in node response analysis, including characteristic elements such as voltage sag initiation point, voltage sag slope, voltage recovery speed, and node response sequence.

[0198] Based on the importance of each feature in determining the propagation order and positional relationship between nodes, the order of node responses is placed in the starting field of the combined vector, so that subsequent processing can prioritize the use of node order to adjust the association strength.

[0199] By setting the voltage sag initiation point as the second field, the propagation relationship of transient disturbances in the time dimension can work together with the node sequence.

[0200] By setting the voltage descent slope and voltage recovery speed as the third and fourth fields respectively, the transient change patterns of the nodes have a fixed and analyzable arrangement order in the combined vector.

[0201] The above fields are concatenated in a fixed order, and a uniform data bit is reserved for each field to construct a transient feature vector that can be directly recognized by the processing module. This allows the transient features of different nodes to be parsed consistently, improving the stability of subsequent node association correction steps.

[0202] In a preferred embodiment of the present invention, based on the node response order and electrical position relationship reflected by the transient feature vector, the correlation strength between nodes in the first correlation data is adjusted item by item to generate second correlation data, including:

[0203] The response sequence of each user node is analyzed based on the transient feature vector to generate node response sequence data;

[0204] Based on the node response sequence data, the relative position of each user node in the electrical path is inferred to form node position relationship data;

[0205] Based on the node position relationship data, the node association strength in the first association relationship data is adjusted item by item. The association strength of nodes with the same response order is increased, and the association strength of nodes with opposite response order is decreased, thus generating the adjusted association relationship data.

[0206] Based on the adjusted association data, inconsistencies in the association strength between the detected nodes are addressed, and the rules are further modified to generate the second association data.

[0207] In this embodiment of the invention, by extracting the voltage drop start time of each user node, the response order of the nodes during disturbance propagation can be quantified into comparable data. Sort the drop start times to form a basic response order structure. For nodes whose drop time differences are within a preset time difference range and are difficult to distinguish directly, comparing their voltage drop slopes can further refine the priority order of nodes during disturbance propagation, making the determination of node response order more reliable. Combining the calibrated response order with the corresponding transient feature data forms a transient feature vector with propagation order characteristics, making the relative path relationship between nodes clearer. This feature vector can serve as an important basis for subsequent correlation strength correction, helping to correct node position relationships that cannot be accurately identified in steady-state analysis and improving the accuracy of node correlation judgment.

[0208] In a preferred embodiment of the present invention, the relative positions of each user node in the electrical path are inferred based on node response sequence data to form node position relationship data, specifically including:

[0209] Extract the response sequence of each user node from the transient feature vector. When the voltage drop start time of node A is earlier than that of node B, it is determined that the transmission position of node A on the electrical path is closer to the disturbance source.

[0210] In the sequence set composed of multiple nodes, the node that responds earlier is regarded as the superior node, and the node that responds later is regarded as the subordinate node, thus forming a preliminary hierarchical order.

[0211] In cases where multiple nodes have short response time intervals, the priority relationship between nodes is further fine-tuned based on the voltage drop slope. When a node's drop rate is significantly faster than another node, its position on the path is moved forward.

[0212] Based on the sorted and fine-tuned node order, a relative position identifier is assigned to each node to clarify its subordinate relationship on the electrical path. For example, nodes with consecutive responses can be regarded as adjacent nodes on the same branch chain.

[0213] The determined relative position information is recorded as node position relationship data, which will provide a structural basis for the next step of adjusting the association strength.

[0214] In a preferred embodiment of the present invention, the node association strength in the first association relationship data is adjusted item by item according to the node position relationship data, specifically including:

[0215] Read whether node A and node B are in the same propagation order relationship from the node position relationship data. When the response order of the two nodes is continuous or the difference is small, it indicates that their electrical path is strongly related.

[0216] For nodes with the same sequence, their association strength in the first association data is increased by a preset ratio to make them more consistent with the connectivity of the actual electrical path.

[0217] For nodes whose response sequences contradict each other or whose sequences differ significantly, their association strength is reduced proportionally to reflect that the two nodes may not belong to the same branch on the actual route.

[0218] When increasing or decreasing the correlation strength, each node pair is processed sequentially to avoid deviations caused by overall compensation;

[0219] The adjusted set of association strengths is aggregated into a new set of node relationships, which serves as the adjusted association data for contradiction detection in the next segment.

[0220] In a preferred embodiment of the present invention, contradictions in the correlation strength between nodes are detected based on the adjusted correlation data, and correction rules are applied to them, specifically including:

[0221] Read the association strength between each pair of nodes from the adjusted association data. When the hierarchical order of a node is inconsistent in multiple node pairs, for example, node A is in a superior position in the relationship with node B, but in a subordinate position in the relationship with node C, it can be determined that there is an order contradiction.

[0222] For the aforementioned contradictory node pairs, the relative positions between the nodes are reassessed based on the response order in the transient feature vector and the electrical path inference results.

[0223] When it is determined that the trend of the association strength between a node and multiple nodes does not conform to the progressive law of the same path link, the association strength of that node is adjusted again to make it consistent with the overall path direction.

[0224] When multiple contradictory nodes are identified, the values ​​with larger deviations can be adjusted according to the overall trend of the correlation strength, so that the node relationships present an interpretable path structure.

[0225] The set of node association strengths after rule correction is recorded as the second association data, so that the final node relationships no longer conflict, which is suitable for the subsequent topology reconstruction process.

[0226] Embodiments of the present invention also provide a low-voltage distribution area topology intelligent identification system, the system comprising:

[0227] The data acquisition and calibration module is used to collect voltage and current measurement data from each user node in the low-voltage distribution area and perform normalization calibration processing. By unifying the sampling interval, amplitude range and offset, it generates the first time series measurement data.

[0228] The two-layer perturbation decomposition module is used to perform two-layer perturbation decomposition on the first time series measurement data. By calculating the change gradient and stationarity, the intervals that meet the pulse load condition are marked as the first abnormal intervals, and the intervals that meet the perturbation condition are marked as the second abnormal intervals. The local abrupt change points in the first abnormal intervals are stripped to generate the second time series measurement data with perturbation labels.

[0229] The node response differential analysis module is used to perform node response differential analysis based on the non-abnormal interval of the second time series measurement data. It generates first correlation data by calculating voltage direction consistency, current increment comparison and differential stability. The first correlation data includes correlation strength to represent the degree of correlation between user nodes.

[0230] The transient feature correction module is used to extract the voltage sag slope, sag initiation point, recovery rate and node response sequence from the first abnormal interval of the second time series measurement data to form a transient feature vector, and to correct the first correlation data based on the transient feature vector to generate the second correlation data.

[0231] The perturbation period verification module is used to verify the periodicity of voltage fluctuations in the perturbation segment based on the second abnormal interval of the second time series measurement data, identify cross-branch response offsets, adjust the second correlation data, and generate the final correlation data.

[0232] The topology reconstruction module is used to reconstruct the branch structure based on the final association data. It determines the subordinate relationships of trunk nodes, branch nodes, and terminal nodes through node order consistency and association strength, and generates the low-voltage transformer area topology recognition results.

[0233] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0234] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0235] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0236] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A low-voltage transformer area topology intelligent identification method, characterized in that, The method comprises: Collecting voltage measurement data and current measurement data of each user node from a low-voltage transformer area, and performing normalization calibration processing to generate first time series measurement data by unifying sampling interval, amplitude range and offset; Performing double-layer disturbance decomposition on the first time series measurement data, marking intervals meeting the pulse load condition as first abnormal intervals and intervals meeting the perturbation condition as second abnormal intervals by calculating change gradient and stationarity, and performing local mutation point stripping processing in the first abnormal intervals to generate second time series measurement data with disturbance labels; Performing node response difference analysis according to the non-abnormal intervals of the second time series measurement data, generating first correlation relationship data by calculating voltage direction consistency, current increment ratio and difference stability, wherein the first correlation relationship data comprises correlation strength for representing the correlation degree between user nodes; According to the first abnormal interval of the second time series measurement data, extracting the voltage drop slope, the drop starting point, the recovery speed and the node response sequence, forming a transient feature vector, and correcting the first correlation relationship data according to the transient feature vector to generate second correlation relationship data; According to the second abnormal interval of the second time series measurement data, checking the periodicity of the voltage fluctuation of the perturbation section, identifying the cross-branch response offset, and adjusting the second correlation relationship data to generate final correlation relationship data; According to the final correlation relationship data, the branch structure is reconstructed, the subordinate relationship of the backbone node, the branch node and the terminal node is determined through the node sequence consistency and the correlation strength, and the low-voltage transformer area topology structure identification result is generated. 2.The low-voltage transformer area topology intelligent identification method according to claim 1, characterized in that, Performing double-layer disturbance decomposition on the first time series measurement data, marking intervals meeting the pulse load condition as first abnormal intervals and intervals meeting the perturbation condition as second abnormal intervals by calculating change gradient and stationarity, and performing local mutation point stripping processing in the first abnormal intervals to generate second time series measurement data with disturbance labels, comprising: Performing change amplitude and change direction calculation on adjacent sampling points of the first time series measurement data to generate change gradient data for disturbance determination; Performing analysis on the average value of the change amplitude, the consistency of the change direction and the uniformity of the change speed of the first time series measurement data in the continuous sampling window to generate stationarity data for characterizing stability; According to the change gradient data, the sampling interval with continuous sudden increase of change amplitude is determined as the pulse disturbance candidate interval; According to the stationarity data, the sampling interval with smaller change amplitude and higher direction consistency is determined as the perturbation candidate interval; Marking the pulse disturbance candidate interval as the first abnormal interval and the perturbation candidate interval as the second abnormal interval to generate initial interval data with abnormal interval labels; Performing local mutation detection on the change gradient data in the first abnormal interval, and extracting sampling points meeting the preset mutation characteristics as independent mutation point data; Integrating the initial interval data and the independent mutation point data to generate second time series measurement data with disturbance labels. 3.The low-voltage transformer area topology intelligent identification method according to claim 1, characterized in that, According to the first abnormal interval of the second time sequence measurement data, the voltage sag slope, the sag starting point, the recovery speed and the node response sequence are extracted to form a transient feature vector, and the first correlation data is corrected according to the transient feature vector to generate second correlation data, including: From the second time sequence measurement data, the voltage change data of each user node in the first abnormal interval is extracted to generate transient voltage data for sequence determination; According to the transient voltage data, the voltage sag slope, the voltage sag starting point and the voltage recovery speed of each user node are determined to form corresponding transient feature data; According to the order of the voltage drop start time in the transient voltage data, the node response sequence of each user node is determined to generate node sequence data, and the transient feature data and the node sequence data are combined to form a transient feature vector; According to the node response sequence reflected by the transient feature vector and the electrical position relationship, the correlation strength between nodes in the first correlation data is adjusted item by item to generate second correlation data.

4. The low-voltage transformer area topology intelligent identification method according to claim 1, characterized in that, According to the second abnormal interval of the second time sequence measurement data, the voltage fluctuation periodicity of the perturbation section is verified, the cross-branch response offset is identified, and the second correlation data is adjusted to generate final correlation data, including: From the second time sequence measurement data, the voltage micro-fluctuation data of each user node in the second abnormal interval is extracted to generate perturbation fluctuation data for period analysis; According to the perturbation fluctuation data, the repeated time interval of the voltage fluctuation of each user node is calculated to generate voltage fluctuation period data; According to the voltage fluctuation period data, the consistency of the periods of different user nodes is compared, and the nodes with period difference exceeding the preset difference range are identified to form cross-branch response offset node data; The node correlation strength corresponding to the cross-branch response offset node data is down-regulated in the second correlation data to generate down-regulated correlation data; The node correlation strength corresponding to the nodes with consistent voltage fluctuation period is up-regulated in the second correlation data to generate up-regulated correlation data; The down-regulated correlation data and the up-regulated correlation data are merged to generate final correlation data.

5. The low-voltage transformer area topology intelligent identification method according to claim 1, characterized in that, According to the final correlation data, the branch structure is reconstructed, the subordinate relationship of the backbone node, the branch node and the terminal node is determined through the node sequence consistency and the correlation strength, and the low-voltage area topology structure identification result is generated, including: According to the final correlation data, the node combination with the highest correlation strength is extracted to generate backbone node candidate data; According to the node response sequence in the backbone node candidate data, backbone node sequence data is generated; According to the correlation strength between nodes in the backbone node sequence data, branch nodes are identified to generate branch node data; According to the correlation strength between the branch node data and its subordinate nodes, terminal node data is generated; The backbone node data, the branch node data and the terminal node data are reconstructed into a branch structure according to the node subordinate relationship to generate a low-voltage area topology structure identification result.

6. The low-voltage transformer area topology intelligent identification method according to claim 3, characterized in that, According to the order of the voltage drop start time in the transient voltage data, the node response order of each user node is determined, node order data is generated, and the transient feature data and the node order data are combined to form a transient feature vector, including: Extracting the voltage drop start time of each user node from the transient voltage data to generate drop start time data; According to the drop start time data, the user nodes are sorted in time order to generate initial response order data of the nodes; According to the node group in the initial response order data whose drop start time difference is lower than the preset time difference range, the voltage step drop slope data of the corresponding nodes is extracted, and the differentiated node response order data is formed by comparing the voltage step drop slope; According to the differentiated node response order data and the transient feature data, the combination processing is performed according to the preset field order to generate the transient feature vector.

7. The low-voltage transformer area topology intelligent identification method according to claim 3, characterized in that, According to the node response order reflected by the transient feature vector and the electrical position relationship, the association strength between the nodes in the first association relationship data is adjusted item by item to generate the second association relationship data, including: According to the response order of each user node analyzed by the transient feature vector, node response order data is generated; According to the node response order data, the relative position of each user node in the electrical path is inferred to form node position relationship data; According to the node position relationship data, the node association strength in the first association relationship data is adjusted item by item, the association strength of the nodes with the same response order is improved, and the association strength of the nodes with opposite response order is reduced to generate the adjusted association relationship data; According to the adjusted association relationship data, the contradictory items between the node association strengths are detected, and the contradictory items are processed again according to the correction rule to generate the second association relationship data.

8. A low-voltage transformer area topology intelligent identification system, characterized in that, The system is applied to the method of any one of claims 1 to 7, and the system comprises: A data acquisition calibration module is configured to acquire voltage measurement data and current measurement data of each user node from a low-voltage area, and perform normalization calibration processing to generate first time sequence measurement data by unifying sampling interval, amplitude range and offset; A double-layer disturbance splitting module is configured to perform double-layer disturbance splitting on the first time sequence measurement data, mark intervals meeting impulse load conditions as first abnormal intervals and mark intervals meeting micro-perturbation conditions as second abnormal intervals by calculating change gradient and stationarity, and perform stripping processing on local mutation points in the first abnormal intervals to generate second time sequence measurement data with disturbance labels; A node response difference analysis module is configured to perform node response difference analysis on non-abnormal intervals of the second time sequence measurement data to generate first association relationship data by calculating voltage direction consistency, current increment ratio and difference stability, wherein the first association relationship data comprises association strength for indicating the association degree between user nodes; A transient feature correction module is configured to extract voltage step drop slope, step drop start point, recovery speed and node response order from the first abnormal intervals of the second time sequence measurement data to form a transient feature vector, and correct the first association relationship data according to the transient feature vector to generate second association relationship data. The perturbation period checking module is configured to check the perturbation segment voltage fluctuation periodicity according to the second abnormal interval of the second time sequence measurement data, identify the cross-branch response deviation, and adjust the second correlation relationship data to generate final correlation relationship data. The topology structure reconstruction module is configured to reconstruct the branch structure according to the final correlation relationship data, determine the master-slave relationship of the backbone node, the branch node and the terminal node through node sequence consistency and correlation strength, and generate a low-voltage transformer area topology structure identification result.

9. A computing device, comprising: The method comprises: one or more processors; a storage device configured to store one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a program, and the program is executed by the processor to implement the method as claimed in any one of claims 1 to 7.

Images