A Time-Effect Intelligent Evaluation Method for Multi-Source Measurement Data in Low-Voltage Distribution Networks

Abstract

This invention relates to the field of intelligent processing and data quality evaluation of power big data, specifically to an intelligent timeliness evaluation method for multi-source measurement data in low-voltage distribution networks. This method first performs unified analysis, windowed aggregation, and preprocessing of multi-source heterogeneous data to generate basic timeliness behavior feature vectors. Then, based on the transformer area topology, it constructs a multi-source data mutual verification relationship graph, extracting two types of mutual verification evidence: events and physical constraints. Next, it calculates edge consistency and node local consistency, combining global consensus time, event propagation compliance, and historical behavior baselines to generate multi-dimensional timeliness indicators. Finally, it obtains extended conflict indicators through multi-indicator fusion, completes the timeliness credibility level judgment, and outputs a full-link evidence chain. This invention can achieve accurate and intelligent evaluation of the timeliness of multi-source data, effectively identifying regular low timeliness and collusive high-risk anomaly patterns.

Description

Technical Field

[0001] This invention relates to the field of intelligent processing and data quality evaluation technology for power big data, and more specifically, to a time-sensitive intelligent evaluation method for multi-source measurement data of low-voltage distribution networks. Background Technology

[0002] Low-voltage distribution networks are the core link connecting power systems to end users. Massive heterogeneous multi-source measurement data are the core foundation for distribution network status perception, lean operation and maintenance, rapid fault handling, and digital management and control. The timeliness and reliability of the data directly determine the accuracy and reliability of distribution network operation decisions.

[0003] Existing technologies for evaluating the quality of distribution network measurement data primarily focus on routine verification of data integrity and accuracy. Evaluations of timeliness are often limited to fixed thresholds for single-source statistical indicators such as transmission delay and packet loss rate, making them ill-suited for the complex scenarios of low-voltage distribution networks, where multi-source data exhibits significant differences in sampling periods, diverse transmission modes, and widespread batch data replenishment. Furthermore, existing technologies lack a mutual verification mechanism based on the electrical topology and physical laws of the distribution network. They can only identify routine timeliness anomalies in single-source data, failing to identify falsely fresh data created through multi-source collusion. Moreover, the evaluation results lack traceable, end-to-end evidence, making it difficult to meet the verification requirements for data timeliness reliability in high-security distribution network operations.

[0004] To address the aforementioned shortcomings of existing technologies, this invention proposes a time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks includes the following steps: S1. Perform unified parsing and windowed aggregation of heterogeneous multi-source measurement data in the low-voltage distribution network. After completing the preprocessing of duplicate frame removal, missing segment marking, out-of-order frame detection and data arrival interval statistics, generate a basic timeliness behavior feature vector containing data arrival behavior, missing status, out-of-order degree and transmission mode identifier. S2. Based on the topology mapping relationship of the low-voltage distribution network area, each data source is mapped to the corresponding electrical node. The mutual verification edge is established based on the degree of electrical coupling between nodes to form a mutual verification relationship diagram. A constraint template is configured for each mutual verification edge. The event alignment window is adaptively adjusted according to the topology distance between the nodes at both ends of the mutual verification edge. At the same time, event evidence and physical constraint evidence for mutual verification are extracted from the measurement sequence of each data source. S3. Based on the mutual verification relationship graph and the extracted mutual verification evidence, calculate the event mutual verification consistency and physical mutual verification consistency of each mutual verification edge, take the lower bound of the two as the edge consistency, and then generate the local consistency of each data source node by weighted aggregation based on the edge consistency of adjacent edges. S4. Based on the event aggregation within the window, generate the global consensus moment, calculate the global consistency, local global consistency deviation and event propagation compliance of each data source node, and calculate the historical behavior deviation of the current behavior based on the historical trusted behavior baseline of the node, and update the historical baseline only when the node's timeliness indicator meets the trusted conditions. S5. After robustly normalizing the aforementioned multi-dimensional timeliness indicators, the extended conflict indicators containing local and global deviation anomalies are obtained. Based on the extended conflict indicators, the timeliness credibility level of each data source is determined, and a full-link evidence chain containing mutual evidence conflict edges, global deviation event clusters, propagation non-compliant event clusters, and historical deviation components is output.

[0007] Furthermore, in step S1, the batch replenishment transmission mode is determined based on the robust center and robust scale of the data arrival interval sequence within the window. When valid data with an arrival interval smaller than the corresponding threshold of the robust center and robust scale appears continuously and the cumulative number exceeds the corresponding proportion of the total amount of valid data from the source within the window, it is marked as a batch replenishment mode and included in the basic timeliness behavior feature vector.

[0008] Furthermore, in step S2, the calculation formula for the event alignment window is: where is the median absolute deviation of the time deviation of all events within the window, is the distance coefficient, and is the topological distance between the two data sources.

[0009] Furthermore, in step S2, the event detection threshold is adaptively determined based on the median absolute deviation of the first-order difference sequence of the measurement sequence, abrupt changes exceeding the threshold are identified as events, and the type, magnitude, and time information of the event are extracted as event evidence.

[0010] Furthermore, in step S3, the event mutual verification consistency is a weighted fusion value of the event matching rate and the relative displacement stability, wherein the event matching rate is the normalized ratio of the number of matching pairs to the total number of events at both ends, and the relative displacement stability is the complementary quantity of the dispersion of the time difference of the matching pair events after normalization on a robust scale.

[0011] Furthermore, in step S3, the physical mutual verification consistency is obtained by mapping the relative improvement rate of the residuals, which is the difference between the minimum value of the single-ended constraint residuals and the joint constraint residuals divided by the minimum value of the single-ended constraint residuals.

[0012] Furthermore, in step S3, the edge weight of adjacent edges is negatively correlated with the topological distance and positively correlated with the electrical coupling strength of the mutual verification relationship. The edge weight associated with the multi-dimensional measurement of the same device has the highest basic weight, while the edge weight associated with the shared communication path has the lowest basic weight.

[0013] Furthermore, in step S4, the global consensus time is obtained through the following steps: calculating the initial median of all event times within the event cluster, calculating the median absolute deviation relative to the initial median as the robustness scale, removing outlier events that deviate from the robustness scale by more than a multiple of the corresponding number of times, and calculating the median of the remaining event times as the final consensus time.

[0014] Furthermore, in step S4, before calculating the compliance of event propagation, the location of the event source is estimated, and node events with global consistency higher than the corresponding threshold are selected as valid samples. The locations of neighboring nodes of the candidate event source are traversed, and the location with the smallest total propagation residual is selected as the final event source location.

[0015] Furthermore, in step S5, the timeliness credibility level is divided into three categories: timeliness credibility, normal low timeliness, and collusion high risk. The classification threshold is adaptively determined based on the median and median absolute deviation of the network node expansion conflict index. The abnormal contribution of the collusion high risk mode mainly comes from the local global deviation anomaly item.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention constructs a basic timeliness behavior feature vector by uniformly parsing and windowed aggregation preprocessing of heterogeneous multi-source data. This vector includes data arrival behavior, missing data, disorder level, and transmission mode identifier. Simultaneously, it constructs a multi-source data mutual verification relationship graph based on the transformer area topology mapping relationship, configuring a constraint template for each mutual verification edge that adaptively adjusts with topology distance. This achieves high adaptability to heterogeneous multi-source and diverse transmission mode scenarios in low-voltage distribution networks. This solution overcomes the limitations of traditional single fixed threshold verification, constructing a mutual verification system with electrical coupling relationships as the core. It can adapt to data sources with different sampling periods and transmission modes, completing a refined characterization and consistency verification of the timeliness characteristics of multi-source data. 2. This invention integrates event-based mutual verification consistency and physical mutual verification consistency to generate edge consistency and node local consistency. Combined with global event consensus, electrical propagation compliance, and historical credible behavior baselines, it constructs a multi-dimensional timeliness evaluation index system. Through robust normalization and fusion, it generates extended conflict indicators, enabling adaptive hierarchical judgments for three modes: reliable timeliness, conventional low timeliness, and high-risk collusion. This scheme, through local-global structural deviation verification and electrical propagation law constraints, can effectively identify abnormal behavior of multi-source synchronous collusion forgery. Simultaneously, it can output a full-link evidence chain supporting the judgment based on the degree of abnormal contribution, ensuring the interpretability and traceability of the evaluation results. Attached Figure Description

[0017] Figure 1 A flowchart for a time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks; Figure 2 This is a flowchart illustrating the implementation steps of the joint calculation of global consensus, propagation compliance, and historical behavior deviation in this invention. Figure 3 This is a flowchart illustrating the implementation steps of the extended conflict fusion, timeliness level judgment, and evidence chain output of this invention. Detailed Implementation

[0018] Example, refer to Figure 1 The time-efficiency intelligent evaluation method for multi-source measurement data in low-voltage distribution networks in this embodiment specifically includes the following steps: Step 1: Data access, windowed preprocessing, and generation of basic timeliness behavior features.

[0019] This step unifies heterogeneous multi-source data into a computable sequence within a window and generates the basic statistics for subsequent mutual verification and historical deviation calculations. The output of this step is the ordered sequence, missing and out-of-order markers, and basic timeliness behavior feature vectors of each data source within the window.

[0020] S11. Parse each submitted data entry to obtain: Data source identifier. Measurement type, source timestamp (If missing, leave blank), Platform receiving time , Sequence number or frame number (if missing, use arrival number instead), measurement value sequence; S12. Within the window, aggregate the data sources to form a sequence, and perform the following cleaning process: Duplicate frame removal: For uploaded data with completely identical sequence numbers or frame numbers within the same data source, only the first data to arrive is retained, and the rest are removed. Missing segment marking: Determine the expected arrival time according to the nominal sampling period of the data source. For periods of time when no valid data is received for consecutive periods exceeding the expected arrival interval, mark them as missing segments. Out-of-order detection: For sequences from the same data source, check them in ascending order of sequence number or frame number. If the sequence number or frame number of the later arriving data is less than the sequence number or frame number of the already received data, it is marked as an out-of-order frame. Arrival interval statistics: For valid data sequences from the same data source, calculate the difference in platform reception time between two adjacent data sets to obtain the arrival interval sequence; Batch replenishment mode indicator: The batch replenishment mode is determined by the abrupt changes in the distribution of the arrival interval sequence and the clustering of short-term dense arrivals. The discrimination threshold is adaptively determined by the robustness scale of the source arrival interval. Specifically, the discrimination method is as follows: calculate the robust center and robustness scale of the arrival interval sequence. When valid data with arrival intervals smaller than the thresholds corresponding to the robust center and robustness scale appear consecutively within the window, and the cumulative number of this group of data exceeds the corresponding proportion of the total amount of valid data from this source within the window, it is marked as batch replenishment mode. S13. Construct a basic timeliness behavior feature vector for each data source. Its components include at least the robust center reaching the interval, the robust scale reaching the jitter, the proportion of missing segments, the disorder intensity, and the batch imputation indicator; the above feature components are calculated as follows: Robust center of reaching interval The median of the arrival interval sequence of the source within the window is used to characterize the typical arrival interval of the source data. Reaching the robustness scale of jitter The median absolute deviation of the arrival interval sequence within the window is used to characterize the fluctuation of the arrival interval. The calculation formula is as follows: ; in, For data source The The platform's reception time for each valid data entry. This is the sequence number of the valid data within the window, and its value is a positive integer. This is a function for calculating the median; percentage of missing segments The total duration of segments marked as missing within the window is equal to the total length of the window. The ratio, ranging from 0 to 1; Disorder strength , is the ratio of the number of out-of-order frames in the window to the total number of valid data from that source in the window, and its value ranges from 0 to 1; Batch replenishment indication quantity , is a binary indicator, which takes the value of 1 when the source in the window is marked as batch replenishment mode, and takes the value of 0 otherwise; The final constructed basic time-sensitive behavior feature vector is: ; Each component of the above feature vector is a definite value that can be directly calculated from the valid data within the window, without any empirically dependent parameters.

[0021] Step 2: Generation of mutual verification relationship diagram and extraction of mutual verification evidence.

[0022] This step organizes multi-source relationships into a mutual verification graph and transforms the original measurements into mutually verifiable event and physical evidence to support time-sensitive mutual verification under heterogeneous sampling rate conditions. This step outputs the mutual verification graph. Event sets and event time sequences, as well as constraint instances required for calculating physical consistency residuals.

[0023] S21. Obtain the topology mapping relationship of the transformer area, map each data source to the location of an electrical node or branch, and determine the topology distance. The calculation method; the topology distance can be determined through the transformer area topology parameters or operation and maintenance configuration, and can be implemented using either of the following two methods: Method 1: Jump count method, which uses the minimum number of branches between corresponding nodes of two data sources in the topology as the topology distance, and takes a positive integer value; Method 2: Impedance method, which sums the impedance magnitudes of all series branches between the corresponding nodes of the two data sources in the topology as the topological distance, with the dimension in ohms; Both methods can achieve a correspondence between topological distance and electrical coupling degree, with smaller topological distances for neighboring nodes and larger topological distances for distant nodes; S22. Construct a mutual verification relationship diagram , where the set of nodes Each node in the set corresponds to an independent data source within the transformer area, and the edge set Each edge in the diagram corresponds to a mutual verification relationship between two data sources. When two data sources meet one of the following conditions: proximity within the same transformer area, upstream / downstream electrical correlation, multi-dimensional measurement correlation within the same equipment, or shared communication transmission path, then... China establishes border ,in , This serves as the identifier for the two mutually verified data sources; Configure a constraint template for each edge, and the constraint template must contain at least the allowed event alignment windows. With the set of verifiable physical constraint types; Event Alignment Window The value is the topological distance between the two data sources. The function is a monotonically non-decreasing function, combined with robust scaling adaptive adjustment of event time deviation within the window, to make the alignment window of neighboring sources narrower and the alignment window of distant sources wider. The specific calculation formula is as follows: ; in, The base alignment window is used, and the value is the median absolute deviation of the time deviation of all events within the window, with the unit being time. This is a distance coefficient, which can be determined by the topology of the transformer area and the characteristics of event propagation. Its dimension is time per unit distance. The topological distance between two data sources is measured in units that match the distance calculation method. The above formula ensures that the larger the topological distance, the wider the alignment window, which conforms to the propagation characteristics of electrical events. There are no fixed empirical parameters, and it can be adaptively updated with the data within the window. S23. Event Evidence Extraction: Construct a first-order difference sequence for the measurement sequence of each data source. The differential sequence is used to characterize the degree of abrupt change in the measured value. The formula for calculating the differential sequence is: ; in, For data source The The measurement value of each valid data point, This is the sequence number of the valid data within the window, and its value is a positive integer greater than or equal to 2; The event detection threshold is adaptively determined by the robust scale of the differential sequence. The robust scale is calculated using the median absolute deviation of the differential sequence. The detection threshold can be set as a corresponding multiple of the robust scale to adapt to the fluctuation characteristics of different measurement sequences.

[0024] Accumulated sum change detection or sliding window abrupt change detection methods are used to identify abrupt changes in the differential sequence that exceed a detection threshold, thus obtaining an event set. For each identified event, the event type, event amplitude characteristics, and event time are output. Event types can be classified into positive and negative mutation events based on the direction and type of the abrupt change in the measured value; the event amplitude characteristic is the absolute value of the difference sequence corresponding to the mutation point; the event time... This refers to the sampling time of the data corresponding to the mutation point; To ensure feasibility, the event time should be taken primarily from the source timestamp. The corresponding sampling time; when In case of missing or unstable timing, take the platform's receiving time. It serves as the initial event time marker, and is subject to binding screening by global consistency and propagation compliance in subsequent step four; S24. Physical Evidence Extraction: Based on the topology of the transformer area and the availability of measurement, instantiate physical constraints, construct a generalized physical constraint residual calculation paradigm, and adapt to the mutual verification requirements of different electrical scenarios. The general physical constraint residual is calculated as follows: for mutually verified edges For the corresponding physical constraints, measurements from two data sources are obtained at the same time scale. The absolute value of the residuals is calculated according to the electrical constraint relationship. Then, the residual statistics within the window are obtained through a robust aggregation function. The general calculation formula is: ; in, For mutual proof edge The physical constraint residual statistics have dimensions that match those of the measured values; Data sources At the same time scale The measured values ​​below; The electrical parameters of the transformer branch corresponding to the mutual verification edge can be obtained through the transformer topology parameters or operation and maintenance configuration. For electrical constraint functions, adaptable to different scenarios: For scenarios where node currents are conserved, the electrical constraint function is the algebraic sum of the currents in adjacent branches; For power balance scenarios, the electrical constraint function is the difference between the power on the distribution transformer side and the total power on the user side; For upstream and downstream voltage constraint scenarios, the electrical constraint function is the difference between the voltage difference between upstream and downstream nodes and the calculated value of the branch voltage drop; Residuals are aligned to the window by events within the window. The residual statistics obtained from the calculation of the limited time-scaled set are used as input for subsequent edge physical mutual verification consistency calculation.

[0025] Step 3: Calculate edge consistency and local consistency.

[0026] This step calculates edge consistency on the mutual verification graph and aggregates it into local consistency, which is used to characterize whether the data source and its neighborhood are consistent. This step outputs the edge consistency. Local consistency with nodes This provides a foundation for subsequent identification of collusion, including local and global biases and the fusion of extended conflicts.

[0027] S31. Event Mutual Authentication Consistency Calculation: For Edges In the event alignment window Under certain conditions, bidirectional event matching is performed; the matching strategy is: for the source... Each event In the source Find the event of the same type with the smallest time difference as a candidate, and the candidate must satisfy the following conditions: ; To the source Each event performs the same matching operation, ultimately resulting in a set of unique bidirectional matching pairs, denoted as . ; Based on the set of matching pairs, the event matching rate and relative displacement stability are calculated, thus obtaining the mutual consistency of events: Event Matching Rate To match the normalized ratio of the number of pairs to the total number of events at both ends, the formula is: ; in, To match the set The number of elements, Data sources within the window The total number of events, with the event matching rate ranging from 0 to 1; Relative displacement stability To match the complementary quantity of the dispersion of the difference between event time points after robust scaling normalization, the calculation formula is as follows: ; in, To match the median absolute deviation of the time difference values ​​of events within the set, used to characterize the dispersion of the time difference values, with the dimension being time; The event alignment window for this edge is in the dimension of time; the relative displacement stability ranges from 0 to 1, and the smaller the dispersion of the difference between event times, the higher the relative displacement stability. Event mutual verification consistency , is a weighted fusion value of event matching rate and relative displacement stability, normalized to 0~1, and calculated using the following formula: ; in, To integrate weights, satisfy The weights can be determined by the reliability of event detection and the business's requirements for timing accuracy. In scenarios without special requirements, equal weights can be set. S32. Physical mutual verification consistency calculation: for edges Within each set of physical constraint types, residual statistics are calculated. To avoid relying on empirical thresholds, the relative improvement of the residuals is used as the consistency criterion, and the specific calculation method is as follows: Calculate separately using only the data source The constrained residual statistic obtained from the measured values Use only the data source The constrained residual statistic obtained from the measured values Joint use of data sources and The constrained residual statistic obtained from the measured values The above residual statistics are all obtained through the general residual calculation paradigm in step S24; Calculate the relative improvement rate of residuals This is used to characterize the degree of residual reduction after using two data sources together, and the calculation formula is: ; in, The function is used to calculate the minimum value. The relative improvement rate of the residuals ranges from 0 to 1. The more significant the reduction of the joint residuals relative to the single-ended residuals, the higher the improvement rate. Physical mutual verification consistency It is obtained by mapping the relative improvement rate of the residuals, normalized to 0~1, and the calculation formula is: ; When both the single-ended residual and the joint residual are 0, the physical mutual verification consistency is directly set to 1, indicating that the measurement values ​​of the two data sources fully satisfy the physical constraint relationship. S33. Edge Consistency Fusion: To avoid bypassability caused by simple superposition, this implementation takes the lower bound of the two types of consistency as edge consistency: ; in, To ensure consistency in event verification, For physical mutual verification consistency, both are dimensionless quantities of 0 to 1; using lower bound fusion can ensure edge consistency is constrained by both event laws and physical laws, reducing the possibility of collusion to circumvent physical constraints by synchronously forging events, or to circumvent mutual verification by satisfying physical constraints but forging event sequence. S34. Node Local Consistency Calculation: ; in, For nodes The set of adjacent nodes in the mutual verification graph; The edge weight is used to reflect the mutual verification strength between two data sources. Its value ranges from 0 to 1 and is normalized within the neighborhood. The edge consistency obtained in step S33 is a dimensionless quantity of 0 to 1; edge weight The determination method is as follows: determine the basic weights based on the mutual verification relationship type and topological distance, and then at the node... Normalization is performed within the neighborhood. The basic weight is negatively correlated with the topological distance and positively correlated with the electrical coupling strength of the mutual verification relationship. Edges associated with multi-dimensional measurements of the same device have the highest basic weight, followed by edges with upstream and downstream electrical connections, and edges with shared communication paths have the lowest basic weight.

[0028] Step 4: Joint calculation of global consensus, dissemination compliance, and deviation from historical behavior.

[0029] This step introduces global scale and propagation law constraints to specifically identify localized clustering patterns of multi-source synchronized collusion and forgery, and utilizes online historical baselines to identify abrupt unnatural behavior. This step outputs global consistency. Local-to-global consistency deviation , dissemination of compliance Historical behavioral deviation This is used for step five, the fusion judgment. Specific steps are as follows: Figure 2 As shown: S41. Event Cluster Construction and Global Consensus: Aggregate events by event type within a window, and construct event clusters using a connectivity clustering method based on mutual verification edge constraints. The specific process is as follows: Treat each event as an independent initial cluster, traverse all mutual verification edges, and if the two events at both ends of an edge come from different data sources, have the same event type, and their time difference is less than that of the edge, align the event window. If the clusters to which the two events belong are merged, each connected cluster obtained is an event cluster. For each event cluster The consensus moment of the event is obtained using robust statistical methods. The specific calculation process is as follows: Step 1: Calculate the initial median of all events within the event cluster. ; Step 2: Calculate the median absolute deviation of all event times within the event cluster relative to the initial median. As a robustness metric; Step 3: Remove outliers within the event cluster whose absolute deviation from the initial median exceeds a multiple of the robustness scale. Step 4: Calculate the median of the remaining event times after removing outliers as the final consensus time. ; The above process can effectively reduce the interference of abnormal forged events on the consensus time, and the obtained consensus time can characterize the actual occurrence time of the electrical event within the transformer area; S42. Global Consistency Calculation: For each node Calculate the set of event clusters that participate in the matching. The deviation is defined as the median deviation of the node event time from the consensus time. Then map it to global consistency: ; in, For nodes The deviation of the event from the time of the event, measured in time; The scale parameter, with time as its dimension, can be obtained by measuring the deviation of event times across all nodes in the network. The robustness scale is determined, and the specific scale can be taken from the entire network. The median absolute deviation is measured and updated adaptively in different windows to adapt to the event propagation speed and noise level in different areas. It is a natural exponential function, with global consistency. The value range is 0 to 1, it is dimensionless, and the smaller the deviation at the moment of the event, the higher the global consistency. S43, Local-to-Global Consistency Deviation: Calculation and The absolute value of the difference is calculated using the following formula: ; in, For the node local consistency obtained in step three, For the global consistency of nodes obtained in this step, both are dimensionless quantities between 0 and 1. The value ranges from 0 to 1; this quantity is used to identify structural anomalies with high local consistency but low global consistency, and is one of the key discriminative dimensions for multi-source collusion forgery. S44. Propagation Compliance Calculation: The propagation of electrical events in the transformer substation topology follows a time-series pattern positively correlated with topological distance. The time of event occurrence should increase with the topological distance from the event source. Propagation compliance is used to characterize whether the timing of node events conforms to this propagation pattern. The specific calculation process is as follows: Step 1: For each event cluster Estimate the location of the event source The robust fitting process is used to achieve this: (1) Sample selection: from event clusters Internal selection for global consistency Node events exceeding the corresponding threshold are considered valid samples, while outlier samples with low global consistency are removed to ensure the reliability of the fitted samples. (2) Initial selection of candidate sources: Select the node with the earliest event time from the valid samples as the candidate event source; (3) Determining the optimal source: Within the neighborhood of the candidate event source, traverse all reachable node positions, calculate the sum of propagation residuals when each position is the event source, and select the position with the smallest sum of propagation residuals as the final event source position. ; (4) Insufficient sample degradation strategy: When the number of valid samples in an event cluster is less than the corresponding threshold, the consensus time of that event cluster is used directly. This serves as the propagation fitting moment for all nodes, at which point subsequent fitting calculations cease. Step 2: Event Clusters Each node within Calculate the propagation fitting time The propagation fitting time is the event occurrence time of the event source plus the propagation delay matched with the topological distance. The propagation delay can be obtained by linearly fitting the time of the effective samples within the event cluster with the topological distance. Step 3: Event Clusters Each node within The propagation residual is calculated as the absolute value of the difference between the event time and the propagation fitting time, i.e. ; Nodes within the window The median of the propagation residuals of all participating event clusters is used to obtain the node. Integrated propagation residual ; Step 4: Map the integrated communication residuals to communication compliance: ; in, The scale parameter, with the dimension of time, is the combined propagation residual of all nodes in the network. The robustness scale is determined, and the specific scale can be taken from the entire network. The magnitude of the median absolute deviation, and updated adaptively with the window; propagation compliance. The value ranges from 0 to 1, is dimensionless, and the smaller the overall propagation residual, the higher the propagation compliance. The lower the propagation compliance, the less the event timing conforms to the propagation law corresponding to electrical distance, and multi-source synchronous forgery usually cannot simultaneously satisfy this law; S45. Calculation and Online Update of Historical Behavior Deviation: S451. Constructing the node comprehensive feature vector Its components are composed of the basic timeliness behavior feature vector. Local consistency of nodes Global consistency of nodes Local-to-global consistency deviation , dissemination of compliance Together they form the following: ; All components of the comprehensive feature vector are definite values ​​that can be directly calculated within the window. Dimensionless and dimensional components are processed separately to ensure the effectiveness of subsequent statistical calculations. S452. Recursive statistics of the historical reliable state of online maintenance nodes, including the feature mean vector. With covariance matrix ; The statistical update is triggered when a node simultaneously meets the following conditions in the current window: global consistency is higher than the corresponding threshold, propagation compliance is higher than the corresponding threshold, and local global consistency deviation is lower than the corresponding threshold. This ensures that the update is driven only by a trusted window, avoiding abnormal data from polluting the historical statistical baseline. The statistical update method can be implemented using exponential recursion, and the recursion formula is as follows: ; ; in, The current window number. The mean vector and covariance matrix are the updated values ​​for the current window. This is the historical statistics from the previous window. This is the combined feature vector of the nodes in the current window; The forgetting factor has a value range of 0 to 1 and can be determined by the business's requirements for the length of memory of historical behavior. The closer the value is to 1, the longer the length of memory of historical data. S453. Calculate the deviation of historical behavior using Mahalanobis distance. The calculation formula is as follows: ; in, This is the combined feature vector of the nodes in the current window. This is the historical mean vector of the current window; both are vectors of the same dimension. The historical covariance matrix of the current window; It is the inverse of the covariance matrix; It is a dimensionless quantity used to characterize the degree of deviation between the node behavior of the current window and the historical reliable behavior. The larger the value, the greater the degree of deviation. To ensure the covariance matrix It is invertible. It is obtained by superimposing a regularization term on its diagonal and then inverting it. The value of the regularization term is the corresponding proportion of the historical variance of each feature component. It can be updated adaptively with the window to avoid calculation failure caused by matrix singularity.

[0030] Step 5: Expand conflict fusion, time limit level judgment and evidence chain output.

[0031] This step unifies neighborhood mutual verification, local-to-global structural deviations, propagation patterns, and historical stability into an extended conflict index, and outputs an interpretable chain of evidence, forming a closed loop for intelligent timeliness evaluation oriented towards collusive forgery. This step outputs the timeliness credibility level, main anomaly types, and the set of mutually corroborating evidence supporting the judgment for each data source. Specific steps are as follows: Figure 3 As shown: S51, Indicator Normalization: For The indicators are robustly normalized to eliminate dimensional differences and uniformly map them to the 0~1 interval; the specific process of robust normalization is as follows: S511. For the target indicator, calculate its distribution statistic across all nodes in the current window, and take the low quantile of the distribution as the lower bound for normalization. The high quantiles of the distribution are taken as the upper bound of normalization. The values ​​of quantiles can be determined by the distribution characteristics of the index to cover the value range of the vast majority of normal samples. S512. Take values ​​for each node of the target indicator. After truncating the upper and lower bounds, the truncated values ​​are: ; S513. Perform a linear mapping on the truncated values ​​to obtain the normalized values. : ; For indicators with higher values, the higher the degree of abnormality, such as The above normalization method is directly adopted; for indicators with higher values ​​and higher reliability, such as... After normalization, take 1 and subtract the normalized value to ensure that the normalization results of all indicators satisfy the condition that the higher the value, the higher the degree of anomaly, which facilitates subsequent fusion calculation. S52. Construction of Extended Conflict Indicators: Defining a Set of Abnormal Indicators The set contains at least the normalized results of local consistency anomalies, local-to-global deviation anomalies, propagation compliance anomalies, and historical behavior deviation anomalies, where: Local consistency anomalies are The higher the value of the normalized value, the worse the consistency of mutual verification between the node and its neighbors. Local and global deviation anomalies are The higher the value of the normalized value, the greater the degree to which local clustering of nodes deviates from the global consensus; The dissemination of compliance anomalies is The higher the value of the normalized value, the greater the degree to which the timing of node events does not conform to the propagation law; Historical behavioral deviations are anomalies The higher the value of the normalization factor, the greater the degree to which the node's current behavior deviates from its historical reliable behavior; Based on the above set of abnormal indicators, an extended conflict indicator is constructed. This is used to comprehensively characterize the degree of anomaly in the timeliness of node behavior, and the calculation formula is as follows: ; in, Output the robust normalization function for the corresponding index in step S51; The weighting coefficients for each abnormal indicator satisfy the following conditions: ; Weighting coefficients can be determined according to business scenario requirements: for businesses with strong time-series accuracy requirements, the weight of compliance anomalies can be increased; for businesses sensitive to data forgery risks, the weight of local and global deviation anomalies and historical behavior deviation anomalies can be increased; in general scenarios without special business requirements, equal weighting can be used to reduce parameter dependence; and extended conflict indicators can be implemented. The value ranges from 0 to 1, is dimensionless, and the higher the value, the higher the degree of abnormality in the node's timeliness behavior and the greater the risk of forgery. S53, Timeliness Level Judgment: Extended Conflict Indicator for All Nodes in the Network Calculate its robustness center and robustness scale. The robustness center adopts the whole network. The median, robustness scale adopts the whole network The median absolute deviation is used to determine the adaptive grading threshold, and the timeliness reliability level and anomaly pattern of each data source are output. The judgment must distinguish at least the following three modes, and the judgment conditions for each mode are as follows: Time-sensitive trustworthy model: Node expansion conflict index The data is below the threshold corresponding to the robust center, and the normalized values ​​of all abnormal indicators are below the corresponding thresholds. This indicates that the timeliness behavior of the node is normal in terms of neighborhood mutual verification, global consensus, propagation rules and historical stability, and the data timeliness is highly reliable. Conventional low-timeliness mode: Node expansion conflict indicators The values ​​are higher than the threshold corresponding to the robust center, and the abnormal contributions mainly come from local consistency anomalies. The remaining abnormal indicators are all within the normal range. The timeliness anomalies of the character nodes are mainly caused by unstable, missing, or out-of-order data transmission, with no risk of collusion forgery. High-risk collusion mode: node expansion conflict indicators The anomaly threshold is higher than that determined by the robust center and the robust scale, and the anomaly contribution mainly comes from local global deviation anomalies. At the same time, it is accompanied by a significant increase in the propagation compliance anomalies or historical behavior deviation anomalies, indicating that the nodes have a high risk of local collusion and synchronous forgery of pseudo-fresh data. The grading threshold can be updated adaptively with the window to adapt to changes in the operating status of the transformer area, the size of the data source, and the noise level, without the need for fixed empirical thresholds; S54. Evidence Chain Output: For each data source judged as having low timeliness or high risk of collusion, output the indicators leading to expanded conflicts in descending order of abnormal contribution. The elevated dominant anomaly component, and the complete set of evidence supporting the judgment, which includes at least: Mutual verification of the consistency between the set of conflicting edges and the corresponding edges The breakdown of components clarifies the proportion of abnormal contributions between event-based mutual verification consistency and physical mutual verification consistency. A set of event clusters that deviate from the global consensus, including the consensus time of the event clusters and the deviation value between the node event time and the consensus time; The set of non-compliant event clusters includes the location of the event source in the event cluster, the deviation of the node propagation residual from the fitting time; Deviation from historical behavior in the comprehensive feature vector Indices of components that make significant contributions and their corresponding deviations.

[0032] Through the detailed description of the above embodiments, the intelligent timeliness evaluation method for multi-source measurement data in low-voltage distribution networks of the present invention completes the standardized extraction of basic timeliness behavior characteristics by uniformly parsing and windowing preprocessing heterogeneous multi-source measurement data in low-voltage distribution networks; it constructs a multi-source data mutual verification relationship graph based on the electrical topology of the transformer area, and integrates dual-dimensional evidence of event mutation law and electrical physical constraints to realize mutual verification consistency of neighboring node data; it constructs a multi-dimensional timeliness evaluation system without experience dependence by combining the global event consensus time, electrical event propagation law and historical reliable behavior baseline; and finally realizes intelligent hierarchical judgment of data timeliness credibility and full-link evidence output through robust fusion of multiple indicators. The present invention breaks through the bottlenecks of poor adaptability of traditional data timeliness evaluation scenarios and insufficient collusion anomaly identification capabilities, and can provide highly reliable data timeliness verification support for core businesses such as low-voltage distribution network status perception and operation control, ensuring the decision-making accuracy and operational safety of digital control of distribution networks.

[0033] The preset parameters in the above formulas shall be set by those skilled in the art according to the actual situation.

[0034] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0035] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0036] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0037] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0038] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0039] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0040] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks, characterized in that, The method flow is as follows: S1. Perform unified parsing and windowed aggregation of heterogeneous multi-source measurement data in the low-voltage distribution network. After completing the preprocessing of duplicate frame removal, missing segment marking, out-of-order frame detection and data arrival interval statistics, generate a basic timeliness behavior feature vector containing data arrival behavior, missing status, out-of-order degree and transmission mode identifier. S2. Based on the topology mapping relationship of the low-voltage distribution network area, each data source is mapped to the corresponding electrical node. The mutual verification edge is established based on the degree of electrical coupling between nodes to form a mutual verification relationship diagram. A constraint template is configured for each mutual verification edge. The event alignment window is adaptively adjusted according to the topology distance between the nodes at both ends of the mutual verification edge. At the same time, event evidence and physical constraint evidence for mutual verification are extracted from the measurement sequence of each data source. S3. Based on the mutual verification relationship graph and the extracted mutual verification evidence, calculate the event mutual verification consistency and physical mutual verification consistency of each mutual verification edge, take the lower bound of the two as the edge consistency, and then generate the local consistency of each data source node by weighted aggregation based on the edge consistency of adjacent edges. S4. Based on the event aggregation within the window, generate the global consensus moment, calculate the global consistency, local global consistency deviation and event propagation compliance of each data source node, and calculate the historical behavior deviation of the current behavior based on the historical trusted behavior baseline of the node, and update the historical baseline only when the node's timeliness indicator meets the trusted conditions. S5. After robustly normalizing the aforementioned multi-dimensional timeliness indicators, the extended conflict indicators containing local and global deviation anomalies are obtained. Based on the extended conflict indicators, the timeliness credibility level of each data source is determined, and a full-link evidence chain containing mutual evidence conflict edges, global deviation event clusters, propagation non-compliant event clusters, and historical deviation components is output.

2. The time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks according to claim 1, characterized in that, In step S1, the batch replenishment transmission mode is determined based on the robust center and robust scale of the data arrival interval sequence within the window. When valid data with an arrival interval smaller than the corresponding threshold of the robust center and robust scale appears continuously and the cumulative number exceeds the corresponding proportion of the total amount of valid data from the source within the window, it is marked as a batch replenishment mode and included in the basic timeliness behavior feature vector.

3. The time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks according to claim 1, characterized in that, In step S2, the formula for calculating the event alignment window is: ,in This represents the median absolute deviation of the time offset of all events within the window. This is the distance coefficient. This represents the topological distance between the two data sources.

4. The time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks according to claim 1, characterized in that, In step S2, the event detection threshold is adaptively determined based on the median absolute deviation of the first-order difference sequence of the measurement sequence, and abrupt changes exceeding the threshold are identified as events. The type, magnitude, and time information of the event are extracted as event evidence.

5. The time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks according to claim 1, characterized in that, In step S3, the event mutual verification consistency is a weighted fusion value of the event matching rate and the relative displacement stability, wherein the event matching rate is the normalized ratio of the number of matching pairs to the total number of events at both ends, and the relative displacement stability is the complementary quantity of the dispersion of the event time difference of the matching pair after normalization on the robust scale.

6. The time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks according to claim 1, characterized in that, In step S3, the physical mutual verification consistency is obtained by mapping the relative improvement rate of the residuals. The relative improvement rate of the residuals is the difference between the minimum value of the single-ended constraint residuals and the joint constraint residuals divided by the minimum value of the single-ended constraint residuals.

7. The time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks according to claim 1, characterized in that, In step S3, the edge weight of adjacent edges is negatively correlated with the topological distance and positively correlated with the electrical coupling strength of the mutual verification relationship. The edge weight associated with the multi-dimensional measurement of the same device is the highest, and the edge weight associated with the shared communication path is the lowest.

8. The time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks according to claim 1, characterized in that, In step S4, the global consensus time is obtained through the following steps: calculate the initial median of all event times within the event cluster, calculate the median absolute deviation relative to the initial median as the robustness scale, remove outlier events that deviate from the robustness scale by more than a multiple of the corresponding number, and calculate the median of the remaining event times as the final consensus time.

9. The time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks according to claim 1, characterized in that, In step S4, before calculating the compliance of event propagation, the location of the event source is estimated, and node events with global consistency higher than the corresponding threshold are selected as valid samples. The locations of neighboring nodes of the candidate event source are traversed, and the location with the smallest total propagation residual is selected as the final event source location.

10. The time-sensitive intelligent evaluation method for multi-source measurement data in low-voltage distribution networks according to claim 1, characterized in that, In step S5, the timeliness credibility level is divided into three categories: timeliness credibility, normal low timeliness, and collusion high risk. The classification threshold is adaptively determined based on the median and median absolute deviation of the network node expansion conflict index. The abnormal contribution of the collusion high risk mode mainly comes from the local global deviation anomaly item.

Images