A load monitoring system of an electric energy metering box

Abstract

This invention discloses a load monitoring system for an electricity metering box, belonging to the field of power system monitoring technology. It includes modules for electrical mapping construction, anomaly detection and location, and fault identification. The system collects main and branch current data, filters steady-state periods and divides time windows, determines effective acquisition channels using Pearson correlation coefficients, generates a main-branch correlation mapping table, and periodically verifies and automatically updates failure relationships. It monitors and marks abnormal main electrical parameters, distinguishes between main and branch conduction anomalies, fuses linear and nonlinear correlations to determine faulty circuits, and decouples multiple fault signal output circuits and phase line identifiers. It constructs a fault feature library, matches features to confidence levels through dynamic time warping, triggers secondary verification through time-series evolution after initial judgment with dual thresholds, and pushes fault types to terminals according to different scenarios. This system can improve the problems of data deviation, difficulty in location, and insufficient reliability in judgment, thereby enhancing monitoring and maintenance management efficiency.

Description

Technical Field

[0001] This invention belongs to the field of power system monitoring technology, and specifically relates to a load monitoring system for an electricity metering box. Background Technology

[0002] Electricity metering boxes are critical equipment at the end of power distribution, and load monitoring directly impacts the efficiency of electricity management and maintenance. Currently, industry monitoring largely relies on manual configuration and simple parameter monitoring, making it susceptible to fluctuations in electricity consumption and line adjustments. Data stability, anomaly location, and fault diagnosis are limited, failing to meet the needs of refined operation and maintenance. Therefore, the following technical problems urgently need to be addressed:

[0003] The correspondence between the main circuit and the individual circuits is mostly sorted out manually, making it difficult to automatically select stable data collection channels. It is greatly affected by power fluctuations and line modifications. Once the correlation fails, it is difficult to update in a timely manner, which can easily lead to deviations in the basic monitoring data.

[0004] Conventional monitoring can only detect electrical anomalies in the main circuit, and it is difficult to distinguish whether the anomaly originates from the main circuit itself or is conducted through the branch circuits. When multiple circuits are abnormal at the same time, it is difficult to separate mixed fault signals, and the location of the fault is somewhat difficult.

[0005] Fault identification often relies on static feature comparison, lacks verification of fault time-series changes, and is prone to identification bias when features are similar. It is not adapted to different abnormal scenarios, and the reliability of fault identification needs to be improved. To address this, we propose a load monitoring system for power metering boxes. Summary of the Invention

[0006] The purpose of this invention is to provide a load monitoring system for an electricity metering box to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a load monitoring system for an electricity metering box, comprising:

[0008] Electrical mapping construction module: Collects total and individual current data from the electricity metering box, filters steady-state operation segments and divides time windows, determines valid acquisition channels through correlation coefficients, and generates a total-individual electrical quantity correlation mapping table; periodically verifies the validity of the mapping relationship, and automatically triggers a relearning process to update the mapping table when it fails;

[0009] Anomaly detection and localization module: Monitors the main circuit electrical parameters, marks the main circuit anomaly event when it exceeds the limit and continues to time out, extracts the electrical feature sequence of the main circuit and mapped branches before and after the anomaly; distinguishes between the main circuit itself and the branch circuit conducted anomalies by the absolute value of the correlation coefficient, calculates the comprehensive correlation degree when the branch circuit is abnormal, determines the unique fault or decouples multiple faults, and outputs the circuit identifier and phase line according to the correlation degree.

[0010] Fault identification output module: Constructs a typical fault feature library, calculates the dynamic time warping distance of the electrical feature sequence of the main circuit or fault circuit in different scenarios and converts it into confidence level. If the standard is met, the fault type is directly output. If the standard is not met, a second verification of time sequence evolution is triggered. Finally, the results are output in different scenarios and pushed to the operation and maintenance terminal.

[0011] Preferably, the specific process of collecting the main circuit and individual phase current data of the electricity metering box, screening the steady-state operation period and dividing the time window, and determining the effective acquisition channel through the correlation coefficient is as follows:

[0012] Collect current data from the main power acquisition device and all individual household and phase acquisition devices at the same time within a preset number of complete power consumption cycles, and filter out the steady-state operation data segments.

[0013] Divide the steady-state operating data segment into several consecutive and non-overlapping time windows;

[0014] For each time window, calculate the Pearson correlation coefficient between the current data sequence of each household and phase acquisition channel and the total current data sequence within that window;

[0015] The number of windows in which the absolute value of the Pearson correlation coefficient of each household phase acquisition channel is greater than or equal to the preset correlation threshold within a preset number of time windows is counted.

[0016] If the number of qualified windows accounts for a preset number of time windows and is not less than a preset percentage, it is determined to be a valid data collection channel.

[0017] If the proportion of qualified windows is lower than the preset proportion, or if the absolute value of the Pearson correlation coefficient is less than the preset correlation threshold in all time windows, the data collection channel is determined to be invalid.

[0018] Preferably, a total-to-part electrical quantity correlation mapping table is generated; the validity of the mapping relationship is periodically checked, and a relearning process is automatically triggered to update the mapping table when it fails. The specific process is as follows:

[0019] The relevant information of all the data collected as valid acquisition channels and their mapping relationship with the stable data of the current total circuit are structured and organized to generate a total-to-part electrical quantity association mapping table;

[0020] A preset mapping relationship verification period is set, and current data at multiple consecutive sampling times are acquired at a preset sampling frequency within each mapping relationship verification period.

[0021] For each sampling time, the current data output by all valid acquisition channels at that time is summed and compared with the current data output by the total acquisition device at that time to calculate the current deviation rate.

[0022] The number of sampling moments in this verification period where the current deviation rate is greater than the preset deviation threshold is counted and recorded as the number of failed samplings.

[0023] The failure rate is obtained by calculating the ratio of the number of failed samples to the total number of samples in the current verification period.

[0024] If the failure rate is less than or equal to the preset failure rate threshold, the current total-to-part electrical quantity association mapping relationship is determined to be valid, and the total-to-part electrical quantity association mapping table remains unchanged.

[0025] If the failure rate is greater than the preset failure rate threshold, the mapping relationship is determined to be invalid, and the relearning process is automatically triggered to repeat the corresponding steps to update the total-to-part electrical quantity association mapping table.

[0026] Preferably, the specific process for distinguishing between the main circuit's own and branch circuit conduction anomalies is as follows:

[0027] When any electrical parameter of the main circuit exceeds its corresponding normal operating range, and the duration of the abnormal state is greater than the preset abnormal duration threshold, it is marked as a main circuit abnormal event.

[0028] Extract the electrical characteristic sequence of all mapped household and phase circuits under the main circuit within a preset time range before and after the occurrence of the main circuit abnormal event;

[0029] For each mapped individual household phase circuit, calculate the Pearson correlation coefficient between its electrical characteristic sequence and the total electrical characteristic sequence;

[0030] If the absolute value of the Pearson correlation coefficient of all mapped individual household phase circuits is greater than or equal to the preset synchronization correlation threshold, it is determined that the total circuit itself is abnormal.

[0031] If the absolute value of the Pearson correlation coefficient of a mapped individual household phase circuit below a preset ratio is greater than or equal to a preset synchronous correlation threshold, it is determined to be an abnormality in the transmission from the branch circuit to the main circuit.

[0032] Preferably, the specific process for calculating the overall correlation degree and determining a unique fault or decoupling multiple faults when a branch circuit is abnormal is as follows:

[0033] Mapped individual household phase loops whose absolute value of the Pearson correlation coefficient is greater than or equal to the preset synchronous correlation threshold are marked as suspected abnormal loops.

[0034] Extract the electrical characteristic sequence of all suspected abnormal circuits within a preset time range before and after the occurrence of the main circuit abnormal event;

[0035] Determine the core time window of the main circuit abnormal event and extract the abnormal electrical feature sequence of the main circuit within the window;

[0036] Calculate the linear and nonlinear correlation between each suspected abnormal loop and the total path. The linear correlation is calculated using the absolute value of the Pearson correlation coefficient, and the nonlinear correlation is calculated using the maximum information coefficient. The two are weighted and fused to obtain the comprehensive correlation.

[0037] Traverse all suspected abnormal loops to obtain their overall correlation degree, and determine the unique faulty loop or trigger the multi-fault decoupling process based on the difference in overall correlation degree.

[0038] Preferably, the multi-fault decoupling process is triggered, and the specific process is as follows:

[0039] The physical circuit that can independently cause abnormal changes in the electrical parameters of the main circuit in this abnormal event is defined as the fault source, and the electrical characteristic sequence of the main circuit within the core time window is input.

[0040] Calculate the average mutual information between any two suspected anomalous loops to determine whether all suspected anomalous loops are approximately independent pairwise.

[0041] When the approximate independence condition is met, the independent component analysis algorithm is used for decomposition; when the condition is not met, a fault separation method based on timing difference is used to obtain independent fault source signals equal to the number of suspected abnormal loops.

[0042] Calculate the overall correlation between each fault source signal and the suspected abnormal circuit, and match the fault source signal with the suspected abnormal circuit with the highest correlation to obtain multiple fault circuits; sort them in descending order of overall correlation, and output the unique identifier of the fault circuit and the identifier of its corresponding phase line in sequence.

[0043] Preferably, the specific process of constructing a typical fault feature library and calculating the dynamic time warping distance of the electrical feature sequence of the main circuit or fault circuit under different scenarios is as follows:

[0044] Construct a typical fault feature library, and pre-set corresponding standard electrical feature sequences and standard electrical feature evolution trend sequences for each typical fault type;

[0045] Based on the fault location results, the fault type is initially screened according to the scenario. When the main circuit itself is abnormal, the electrical feature sequence within the core time window of the main circuit is obtained. When the only faulty circuit is abnormal, the electrical feature sequence within the preset time range before and after the fault of the circuit is obtained. When multiple circuits are abnormal at the same time, the electrical feature sequence within the preset time range before and after the fault of each faulty circuit is obtained in order of comprehensive correlation from high to low.

[0046] Substitute the above electrical feature sequence into the feature-weighted dynamic time warping algorithm to calculate the corresponding dynamic time warping distance.

[0047] Preferably, the specific process of converting dynamic time-warped distance into confidence level and directly outputting the fault type upon meeting the criteria is as follows:

[0048] The dynamic time warping distance between each electrical feature sequence and the corresponding standard electrical feature sequence is converted into the confidence level for fault root cause identification. The confidence level is inversely proportional to the dynamic time warping distance.

[0049] When the highest confidence level is higher than the preset minimum confidence level threshold, and the difference between the highest confidence level and the second highest confidence level is greater than the preset confidence difference threshold, the fault type corresponding to the highest confidence level is directly output.

[0050] If the preferred result is not met, a second verification based on the time sequence evolution will be triggered. The specific process of finally outputting the results according to the scenario and pushing them to the operation and maintenance terminal is as follows:

[0051] When the highest confidence level is lower than the preset minimum confidence level threshold, or the difference between the highest confidence level and the second highest confidence level is less than the preset confidence difference threshold, the time series evolution feature secondary verification process is triggered and executed according to the fault scenario classification.

[0052] Electrical data of the corresponding circuits within a preset time period after a fault occurs is collected in different scenarios, and the evolution trend sequence of electrical features is extracted.

[0053] The standard electrical feature evolution trend sequence corresponding to the top number of suspected fault types with confidence in the typical fault feature library is compared one by one. The trend consistency coefficient of each group of sequences is calculated by using cosine similarity, and the category with the highest coefficient is selected as the final fault type.

[0054] Output the fault type according to the scenario, and send the results to the operation and maintenance terminal after completing the fault location and source tracing.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] (1) The load monitoring system of the power metering box collects the current data of the main line and the individual households and phases, filters the steady-state operation section and divides multiple time windows, determines the effective acquisition channel based on the Pearson correlation coefficient, automatically generates the total-individual electrical quantity correlation mapping table, and periodically checks the current deviation rate. After the mapping relationship fails, it can automatically trigger the relearning process to complete the update, replacing the manual configuration method, reducing the deviation of basic data caused by power fluctuations and line changes, and keeping the monitoring basic data stable.

[0057] (2) The load monitoring system of the power metering box marks abnormal events by combining the over-limit of the main circuit electrical parameters with the duration of the event. It distinguishes between the main circuit's own abnormality and the branch circuit's conducted abnormality by relying on the Pearson correlation coefficient. It calculates the comprehensive correlation degree by integrating linear and nonlinear correlation degrees to determine the fault circuit. Then, it decouples the mixed signals of multiple faults by using an adaptive signal separation method to determine the fault circuit and its corresponding phase line, which helps to reduce the difficulty of anomaly tracing and fault location.

[0058] (3) The load monitoring system of the power metering box constructs a typical fault feature library, uses a feature weighted dynamic time warping algorithm to match electrical feature sequences and convert them into fault confidence, completes the initial fault judgment based on dual thresholds, triggers secondary verification of time evolution trend when the initial screening fails, optimizes the judgment result by combining cosine similarity comparison, adapts the recognition logic to different scenarios, reduces the judgment deviation when features are similar, and improves the reliability of fault type judgment and on-site operation and maintenance efficiency. Attached Figure Description

[0059] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example 1;

[0062] Please see Figure 1 The present invention provides a load monitoring system for an electricity metering box, comprising: an electrical mapping construction module, an anomaly detection and location module, and a fault identification and output module;

[0063] The electrical mapping construction module collects total and individual phase current data from the electricity metering box, filters steady-state operating segments and divides time windows, determines valid acquisition channels through correlation coefficients, and generates a total-individual electrical quantity correlation mapping table. It periodically verifies the validity of the mapping relationship; if it fails, it automatically triggers a relearning process to update the mapping table. The specific process is as follows:

[0064] The system collects current data from the main power metering device and all individual household and phase-specific power metering devices at the same time within a preset number of complete power consumption cycles, and then selects the steady-state operation data segment from it.

[0065] The steady-state operation data segment is defined as: a data segment in which the current fluctuation amplitude is less than the normal fluctuation range obtained based on historical data statistics within a continuous preset time period. The normal fluctuation range can be determined using the 3σ principle.

[0066] The selected steady-state operating data segments are divided into several continuous and non-overlapping time windows;

[0067] For each time window, calculate the Pearson correlation coefficient between the current data sequence of each household and phase acquisition channel and the total current data sequence within that time window. The calculation formula is as follows:

[0068] ,in, The Pearson correlation coefficient between the individual household phase circuit current sequence X and the total circuit current sequence Y within the current time window. Let covariance be the variance of the two sequences. These are the standard deviations of the two sequences;

[0069] For each household-phase data acquisition channel, count the number of windows in which the absolute value of the Pearson correlation coefficient between the current data sequence of the household-phase data acquisition channel and the total current data sequence is greater than or equal to a preset correlation threshold within a preset number of time windows.

[0070] If the number of qualified windows accounts for a proportion of the preset number of time windows not less than the preset proportion, then the acquisition channel is determined to be a valid acquisition channel, and its corresponding physical loop has a stable data mapping relationship with the current total path.

[0071] If the proportion of the number of qualified windows to the preset number of time windows is lower than the preset proportion, or if the absolute value of the Pearson correlation coefficient between the current data sequence of the household and phase acquisition channel and the total current data sequence is less than the preset correlation threshold in all time windows, then the acquisition channel is determined to be an invalid acquisition channel and will not be included in the subsequent construction of the total-to-partial electrical quantity correlation mapping relationship and fault location process.

[0072] All relevant information of the valid acquisition channels and their mapping relationship with the current total circuit are structured and organized to generate a total-to-part electrical quantity association mapping table. For each valid acquisition channel, the table stores the corresponding total circuit unique identifier, valid acquisition channel number, physical circuit unique identifier, phase line identifier of the circuit, correlation verification result, and effective time of the mapping relationship.

[0073] A preset mapping relationship verification period is set, and current data at multiple consecutive sampling times are acquired at a preset sampling frequency within each mapping relationship verification period.

[0074] For each sampling time, the current data output by all valid acquisition channels at that time is summed and compared with the current data output by the total acquisition device at that time to calculate the current deviation rate.

[0075] The number of sampling moments in this verification period where the current deviation rate is greater than the preset deviation threshold is counted and recorded as the number of failed samplings.

[0076] The failure rate is obtained by calculating the ratio of the number of failed samples to the total number of samples in the current verification period.

[0077] If the failure rate is less than or equal to the preset failure rate threshold, the current total-to-part electrical quantity association mapping relationship is determined to be valid, and the mapping table remains unchanged.

[0078] If the failure rate is greater than the preset failure rate threshold, the current total-to-part electrical quantity association mapping relationship is determined to be invalid, and the mapping table relearning process is automatically triggered. The above data acquisition, correlation calculation, effective channel determination and mapping table generation steps are repeated to update the corresponding fields in the total-to-part electrical quantity association mapping table.

[0079] It should be noted that using a combination of steady-state operation data segment screening and multi-time window correlation statistics to determine the effective acquisition channels can effectively reduce the interference of transient fluctuations such as power consumption peaks and equipment start-ups and shutdowns on the channel correlation calculation results.

[0080] By screening steady-state current data and statistically analyzing the correlation compliance rate within multiple consecutive time windows, compared to the single time window judgment method, the misjudgment of effective channels caused by accidental factors can be reduced, and the stability of the effective acquisition channel identification results can be improved.

[0081] The generated structured total-to-branch electrical quantity correlation mapping table fully stores the correspondence between the total circuit and each effective acquisition channel, including key information such as the unique identifier of the physical circuit and the phase line to which it belongs. This provides core data basis for the subsequent anomaly detection and localization module to distinguish between the total circuit's own anomalies and the branch circuit's conducted anomalies. In the case of branch circuit anomalies, the electrical feature sequence of the corresponding suspected abnormal circuit can be quickly extracted based on this mapping table, laying the foundation for comprehensive correlation calculation, fault circuit localization, and phase line identification output, and avoiding fault analysis errors caused by chaotic circuit correspondence.

[0082] The established periodic mapping relationship verification and automatic relearning process can adapt to the actual operating changes of the power metering box.

[0083] When the original mapping relationship fails due to factors such as adjustment of the individual household circuit, replacement of the acquisition device, or changes in wiring, the failure status can be automatically identified through current deviation rate statistics, and a relearning process can be triggered to update the mapping table fields.

[0084] This mechanism can maintain the validity of the mapping relationship without human intervention, reduce the subsequent fault location deviation caused by the outdated mapping relationship, and improve the adaptability of the system in long-term operation.

[0085] Anomaly detection and location module: Monitors the main circuit electrical parameters, marks the main circuit anomaly event when it exceeds limits and continues to time out, extracts the electrical feature sequences of the main circuit and mapped branches before and after the anomaly; distinguishes between the main circuit's own anomalies and branch-borne anomalies by the absolute value of the correlation coefficient; when a branch circuit is abnormal, calculates the comprehensive correlation degree to determine a unique fault or decouple multiple faults, and outputs the circuit identifier and phase line according to the correlation degree. The specific process is as follows:

[0086] When any electrical parameter of the main circuit exceeds its corresponding normal operating range, and the duration of the abnormal state is greater than the preset abnormal duration threshold, it is marked as a main circuit abnormal event.

[0087] Extract the electrical characteristic sequence of all mapped household and phase circuits under the main circuit within a preset time range before and after the occurrence of the main circuit abnormal event;

[0088] The electrical characteristic sequence includes: current, voltage, active power, reactive power, power factor, harmonic content, current change rate, and voltage fluctuation amplitude.

[0089] For each mapped individual household phase circuit, calculate the Pearson correlation coefficient between its electrical characteristic sequence and the total electrical characteristic sequence;

[0090] If the absolute value of the Pearson correlation coefficient of all mapped individual phase circuits is greater than or equal to the preset synchronous correlation threshold, then it is determined that the electrical parameters of all mapped individual phase circuits have synchronously shown abnormal changes consistent with the main circuit, and it is determined that the main circuit itself is abnormal.

[0091] If the absolute value of the Pearson correlation coefficient of a portion of the mapped individual household phase circuits below a preset ratio is greater than or equal to the preset synchronous correlation threshold, it is determined that only a portion of the electrical parameters of the mapped individual household phase circuits have abnormal changes, and it is determined that the abnormality of the branch circuit is transmitted to the main circuit.

[0092] When it is determined that a branch circuit anomaly has propagated to the main circuit, the following procedure is executed:

[0093] Mapped individual household phase loops whose absolute value of the Pearson correlation coefficient is greater than or equal to the preset synchronous correlation threshold are marked as suspected abnormal loops.

[0094] Extract the electrical characteristic sequence of all suspected abnormal circuits within a preset time range before and after the occurrence of the abnormal event;

[0095] The core time window of the abnormal event of the total circuit is determined by taking the moment when all electrical parameters of the total circuit first exceed the normal operating range as the starting point and the moment when all electrical parameters of the total circuit finally return to the normal operating range or reach the preset maximum abnormal duration as the ending point.

[0096] Extract the abnormal electrical feature sequence of the main circuit within this core time window;

[0097] For each suspected abnormal loop, calculate its linear correlation and nonlinear correlation with the total loop within the same core time window. The linear correlation is calculated using the absolute value of the Pearson correlation coefficient.

[0098] The nonlinear correlation degree is calculated using the maximum information coefficient, and the calculation formula is as follows:

[0099] ,

[0100] in, The nonlinear correlation between the total circuit abnormal electrical characteristic sequence X and the suspected abnormal circuit electrical characteristic sequence Y within the same core time window is defined as follows: 'a' represents the number of grid columns used to divide the numerical range of the total circuit abnormal electrical characteristic sequence X into continuous intervals, and 'b' represents the number of grid rows used to divide the numerical range of the suspected abnormal circuit electrical characteristic sequence Y into continuous intervals. The interval divisions of columns 'a' and rows 'b' together constitute a two-dimensional grid representing the scatter distribution of the corresponding points of the two sequences. The mutual information of the two sequences under the a×b grid partitioning. The maximum size of the grid.

[0101] The linear and nonlinear correlation degrees of the current suspected anomalous loops are weighted and fused to obtain the comprehensive correlation degree of the current suspected anomalous loops. The fusion formula is as follows:

[0102] ,

[0103] Where S represents the overall correlation degree of the currently suspected abnormal loop. The linear correlation (Pearson correlation coefficient). These are preset weighting coefficients, and the sum of the two is one.

[0104] Traverse all suspected abnormal loops to obtain the overall correlation degree of all suspected abnormal loops.

[0105] When the overall correlation of a single suspected abnormal loop is greater than the difference between the overall correlation of the loop and that of every other suspected abnormal loop, the suspected abnormal loop is determined to be the only faulty loop.

[0106] When there are a preset number or more suspected abnormal loops, and the difference between the comprehensive correlation degree of any two suspected abnormal loops is less than the preset difference threshold, it is determined that multiple loops are abnormal at the same time, triggering the multi-fault decoupling process.

[0107] The multi-fault decoupling process is triggered, and the specific process is as follows:

[0108] The physical circuit that can independently cause abnormal changes in the electrical parameters of the main circuit in this abnormal circuit event is defined as the fault source;

[0109] The electrical characteristic sequence of the total circuit within the core time window is used as input;

[0110] For any two suspected abnormal loops, calculate the mutual information of their corresponding electrical characteristic sequences within the same core time window, and take the average value as the final mutual information value of the two suspected abnormal loops.

[0111] When the final mutual information values ​​of any two suspected abnormal loops are both less than the preset independent judgment threshold, all suspected abnormal loops are determined to be approximately independent pairwise.

[0112] When the approximate independence condition is met, the independent component analysis algorithm is used to decompose the total electrical feature sequence. The decomposition dimension is set to the number of suspected abnormal loops: first, the total electrical feature sequence is whitened to remove linear correlation, and then the separation matrix is ​​estimated by maximizing non-Gaussianity. Finally, the number of independent fault source signals equal to the number of suspected abnormal loops is obtained. This algorithm belongs to the conventional signal separation technology in this field and will not be elaborated on here.

[0113] When the approximate independence condition is not met, a fault separation method based on time difference is adopted: the moment when all electrical parameters of each suspected abnormal circuit first exceed the normal operating range is determined. According to the order of these moments, the total electrical characteristic sequence is segmented by time with each abnormal start moment as the dividing point, and the fault source component corresponding to the time starting point is decomposed. Each fault source component corresponds to an independent fault source signal. If the first abnormal moment of multiple suspected abnormal circuits is the same, the circuits are assigned to the fault source component with the same time starting point. This method belongs to the conventional signal separation technology in this field and will not be elaborated on here.

[0114] Each separated fault source signal is compared with the electrical characteristic sequence of each suspected abnormal circuit within the same core time window to calculate the comprehensive correlation degree. The comprehensive correlation degree is calculated by a combination of linear and nonlinear correlation degree.

[0115] Each fault source signal is matched one by one with the suspected abnormal loop with the highest comprehensive correlation to obtain multiple fault loops.

[0116] Multiple faulty circuits are prioritized and sorted from largest to smallest based on their comprehensive correlation value. The unique identifier of each faulty circuit (i.e., the unique identifier of the physical circuit stored in the total-to-part electrical quantity correlation mapping table) and the identifier of its corresponding phase line (i.e., A phase, B phase, C phase identifier) ​​are output sequentially.

[0117] It should be noted that using the duration of exceeding the limits of the total electrical parameters as the basis for judging abnormal events can effectively filter out non-fault interference signals such as instantaneous voltage fluctuations and short-term current spikes, reduce the probability of false triggering of abnormal events, and make the abnormal marking more consistent with the actual operating state. At the same time, extracting multi-dimensional electrical feature sequences such as current, voltage, power, and harmonics can completely restore the electrical changes before and after the occurrence of the abnormality, providing sufficient data support for subsequent abnormality type differentiation and fault location.

[0118] Based on the total-to-branch correlation mapping relationship generated by the electrical mapping construction module, the Pearson correlation coefficient can be used to quickly distinguish between the total circuit's own anomalies and the branch circuit's conducted anomalies. This can directly identify the type of anomaly source, avoid misjudging branch circuit conducted faults as total circuit faults, provide a clear direction for subsequent fault handling and analysis, and reduce ineffective troubleshooting steps.

[0119] In branch circuit anomaly determination, a weighted fusion of linear and nonlinear correlation degrees is used to calculate the comprehensive correlation degree. Compared with single correlation analysis, it can more comprehensively capture the correlation characteristics between suspected abnormal circuits and main circuit anomalies, and adapt to complex electrical changes in power consumption scenarios. At the same time, by automatically identifying single and multiple fault scenarios through the comprehensive correlation degree difference, it can adapt to different fault occurrence modes of metering boxes and improve the scenario adaptability of fault determination.

[0120] An adaptive decoupling process is designed for concurrent multi-fault scenarios. Based on the mutual information between loops, an independent component analysis or time difference separation method is selected to decompose the mixed abnormal signal of the total circuit into independent fault source signals, thereby completing the accurate matching and location of multiple fault loops. Finally, the fault loops and phase line identifiers are output in order of correlation, which can not only provide reliable input for the fault identification output module, but also enable maintenance personnel to quickly locate the fault and shorten the time spent on on-site handling.

[0121] Fault identification output module: Constructs a typical fault feature library, calculates the dynamic time warping distance of the electrical feature sequence of the main circuit or faulty circuit according to different scenarios and converts it into confidence level. If the standard is met, the fault type is directly output; otherwise, a secondary verification of time sequence evolution is triggered. Finally, the results are output according to different scenarios and pushed to the operation and maintenance terminal. The specific process is as follows:

[0122] Construct a typical fault feature library, in which a corresponding standard electrical feature sequence and a standard electrical feature evolution trend sequence are preset for each typical fault type;

[0123] Furthermore, this feature library can be generated by labeling and training historical fault data, which is a conventional technical method in this field and will not be elaborated on here.

[0124] Based on the different fault location results, conduct initial screening of fault types according to different scenarios:

[0125] When the fault location result is that the main circuit itself is abnormal, the electrical feature sequence of the main circuit within the core time window is obtained, and it is substituted with each standard electrical feature sequence in the typical fault feature library into the feature weighted dynamic time warping algorithm to calculate the dynamic time warping distance.

[0126] When the fault location result is that the only faulty circuit is abnormal, the electrical feature sequence of the located faulty circuit within a preset time range before and after the fault occurs is obtained, and it is substituted with each standard electrical feature sequence in the typical fault feature library into the feature weighted dynamic time warping algorithm to calculate the dynamic time warping distance.

[0127] When the fault location result is that multiple circuits are abnormal at the same time, each located fault circuit is processed in turn according to the priority order of comprehensive correlation from high to low: obtain the electrical feature sequence of the fault circuit within a preset time range before and after the fault occurs, and substitute it with each standard electrical feature sequence in the typical fault feature library into the feature weighted dynamic time warping algorithm to calculate the dynamic time warping distance.

[0128] The calculation formula for the feature-weighted dynamic time warping algorithm is as follows:

[0129] ,

[0130] in, Let be the dynamic time warp distance between electrical characteristic sequence A and standard electrical characteristic sequence B. This is a time-warping path (i.e., a mapping relationship that achieves optimal alignment of sequence elements). It is the distance metric between corresponding elements in the sequence (i.e., the Euclidean distance between the i-th element in the electrical feature sequence A and the corresponding mapped element in the standard electrical feature sequence B). Let be the weight of the k-th electrical feature, i be the index of the sequence element, and n be the total length of the sequence.

[0131] The dynamic time-warped distance between the electrical feature sequence and each standard electrical feature sequence is converted into a confidence level for fault root cause identification. The confidence level is inversely proportional to the distance, and the conversion formula is as follows: ,

[0132] in, Let be the confidence level of the j-th type of typical fault. is the dynamic time warping distance between the real-time fault feature sequence and the standard electrical feature sequence of the j-th type of typical fault, and m is the total number of fault types in the typical fault feature library;

[0133] When the highest confidence level is higher than the preset minimum confidence level threshold, and the difference between the highest confidence level and the second highest confidence level is greater than the preset confidence level difference threshold, the fault type corresponding to the highest confidence level is directly output.

[0134] When the highest confidence level is lower than the preset minimum confidence level threshold, or the difference between the highest and second-highest confidence levels is less than the preset confidence difference threshold, the secondary verification process for time-series evolution features is triggered and executed according to the fault scenario classification:

[0135] If the fault is caused by the main circuit itself, collect the electrical data of the main circuit within a preset time after the fault occurs, and extract the evolution trend sequence of the main circuit's electrical characteristics. The evolution trend sequence of electrical characteristics is a time sequence of the continuous changes of all electrical characteristics of the circuit, such as current, voltage, and power, over time within a preset time after the fault occurs.

[0136] If the fault is a single circuit, collect the electrical data of the fault circuit within a preset time after the fault occurs, and extract the corresponding electrical feature evolution trend sequence.

[0137] If multiple circuits are simultaneously abnormal, electrical data of each faulty circuit is collected in order of priority from high to low comprehensive correlation, and the corresponding electrical feature evolution trend sequence is extracted.

[0138] The extracted electrical feature evolution trend sequence is compared one by one with the standard electrical feature evolution trend sequences corresponding to the top-ranked number of suspected fault types in the typical fault feature library. The trend consistency coefficient of each group of sequences is calculated using cosine similarity. The category with the highest trend consistency coefficient is selected as the final fault type.

[0139] The formula for calculating the trend consistency coefficient using cosine similarity is as follows:

[0140] ,

[0141] in, Here, represents the trend consistency coefficient, U represents the real-time electrical characteristic evolution trend sequence, and V represents the standard electrical characteristic evolution trend sequence corresponding to a typical fault. Let L be the L2 norm of the two sequences.

[0142] Differentiate execution output based on scenario: In scenarios where the main circuit itself is abnormal, only the fault type of the main circuit will be output;

[0143] In a single fault loop exception scenario, only the fault type of that loop is output.

[0144] In the scenario of simultaneous failure of multiple circuits, the fault types of each circuit are output in descending order of comprehensive correlation.

[0145] This completes the fault location and tracing process;

[0146] The fault type result is sent to the operation and maintenance terminal. Based on the fault type information, the operation and maintenance terminal carries out targeted operation and maintenance and equipment maintenance for the corresponding abnormal object (total circuit or faulty loop).

[0147] It should be noted that a typical fault feature library is constructed based on historical fault data, storing standard electrical feature sequences and evolution trend sequences to provide a unified comparison benchmark for fault type identification. The results of the anomaly detection and localization module are processed according to different scenarios, adapting to three working conditions: total circuit anomaly, single circuit anomaly, and multi-circuit anomaly. Then, the sequence distance is calculated through the feature weighted dynamic time warping algorithm, which can take into account the influence weight of different electrical features and better reflect the fault feature change patterns in actual power consumption scenarios.

[0148] By converting dynamic time-normalized distance into fault confidence and determining fault type through dual threshold conditions, results can be directly output when the feature matching degree is high, simplifying the identification process and improving response speed. This method can filter out invalid fault types with low matching degree, reduce the possibility of fault type misjudgment, and make the preliminary identification results more reliable.

[0149] For scenarios where the initial screening fails, a secondary verification based on temporal evolution is added. By extracting the evolution trend sequence of electrical features after the fault occurs and comparing the trend consistency with cosine similarity, the shortcomings of static feature matching can be made up for, and the fault type can be identified more accurately. Especially in scenarios where the feature similarity is close and it is difficult to make a direct judgment, the reliability of the fault identification results can be further improved.

[0150] Finally, the results are output according to the fault scenario and pushed to the operation and maintenance terminal, which can clearly identify the fault type and the corresponding abnormal object, providing operation and maintenance personnel with a clear basis for handling. The entire identification process is seamlessly connected with the front-end positioning link, forming a complete closed loop of fault location, identification and output, which helps to shorten the fault investigation and handling time and improve the overall efficiency of load monitoring and operation and maintenance management of power metering boxes.

[0151] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A load monitoring system for an electricity metering box, characterized in that, include: Electrical mapping construction module: Collects the total circuit and individual phase current data of the power metering box, filters the steady-state operation segment and divides the time window, determines the effective acquisition channel through the correlation coefficient, and generates a total-individual electrical quantity correlation mapping table; The validity of the mapping relationship is periodically checked, and the relearning process is automatically triggered to update the mapping table when it fails. Anomaly detection and localization module: Monitors the main circuit electrical parameters, marks the main circuit anomaly event when it exceeds the limit and continues to time out, and extracts the electrical feature sequence of the main circuit and mapped branches before and after the anomaly; The absolute value of the correlation coefficient is used to distinguish between the main circuit itself and the branch circuit conduction anomalies. When a branch circuit is abnormal, the comprehensive correlation degree is calculated to determine whether it is a unique fault or multiple faults are decoupled. The output circuit identifier and phase line are sorted according to the correlation degree. Fault identification output module: Constructs a typical fault feature library, calculates the dynamic time warping distance of the electrical feature sequence of the main circuit or fault circuit in different scenarios and converts it into confidence level. If the standard is met, the fault type is directly output. If the standard is not met, a second verification of time sequence evolution is triggered. Finally, the results are output in different scenarios and pushed to the operation and maintenance terminal.

2. The load monitoring system for an electricity metering box according to claim 1, characterized in that: The specific process of collecting main circuit and individual phase current data from the electricity metering box, screening steady-state operation periods and dividing time windows, and determining effective data acquisition channels through correlation coefficients is as follows: Collect current data from the main power acquisition device and all individual household and phase acquisition devices at the same time within a preset number of complete power consumption cycles, and filter out the steady-state operation data segments. Divide the steady-state operating data segment into several consecutive and non-overlapping time windows; For each time window, calculate the Pearson correlation coefficient between the current data sequence of each household and phase acquisition channel and the total current data sequence within that window; The number of windows in which the absolute value of the Pearson correlation coefficient of each household phase acquisition channel is greater than or equal to the preset correlation threshold within a preset number of time windows is counted. If the number of qualified windows accounts for a preset number of time windows and is not less than a preset percentage, it is determined to be a valid data collection channel. If the proportion of qualified windows is lower than the preset proportion, or if the absolute value of the Pearson correlation coefficient is less than the preset correlation threshold in all time windows, the data collection channel is determined to be invalid.

3. The load monitoring system for an electricity metering box according to claim 2, characterized in that: The process of generating a total-to-part electrical quantity correlation mapping table and periodically verifying the validity of the mapping relationship, automatically triggering a relearning process to update the mapping table when it fails, is as follows: The relevant information of all the data collected as valid acquisition channels and their mapping relationship with the stable data of the current total circuit are structured and organized to generate a total-to-part electrical quantity association mapping table; A preset mapping relationship verification period is set, and current data at multiple consecutive sampling times are acquired at a preset sampling frequency within each mapping relationship verification period. For each sampling time, the current data output by all valid acquisition channels at that time is summed and compared with the current data output by the total acquisition device at that time to calculate the current deviation rate. The number of sampling moments in this verification period where the current deviation rate is greater than the preset deviation threshold is counted and recorded as the number of failed samplings. The failure rate is obtained by calculating the ratio of the number of failed samples to the total number of samples in the current verification period. If the failure rate is less than or equal to the preset failure rate threshold, the current total-to-part electrical quantity association mapping relationship is determined to be valid, and the total-to-part electrical quantity association mapping table remains unchanged. If the failure rate is greater than the preset failure rate threshold, the mapping relationship is determined to be invalid, and the relearning process is automatically triggered to repeat the corresponding steps to update the total-to-part electrical quantity association mapping table.

4. The load monitoring system for an electricity metering box according to claim 3, characterized in that: The specific process for distinguishing between the main circuit's own and branch circuit's transmission anomalies is as follows: When any electrical parameter of the main circuit exceeds its corresponding normal operating range, and the duration of the abnormal state is greater than the preset abnormal duration threshold, it is marked as a main circuit abnormal event. Extract the electrical characteristic sequence of all mapped household and phase circuits under the main circuit within a preset time range before and after the occurrence of the main circuit abnormal event; For each mapped individual household phase circuit, calculate the Pearson correlation coefficient between its electrical characteristic sequence and the total electrical characteristic sequence; If the absolute value of the Pearson correlation coefficient of all mapped individual household phase circuits is greater than or equal to the preset synchronization correlation threshold, it is determined that the total circuit itself is abnormal. If the absolute value of the Pearson correlation coefficient of a mapped individual household phase circuit below a preset ratio is greater than or equal to a preset synchronous correlation threshold, it is determined to be an abnormality in the transmission from the branch circuit to the main circuit.

5. The load monitoring system for an electricity metering box according to claim 4, characterized in that: When a branch circuit fails, the specific process for calculating the overall correlation degree and determining whether there is a unique fault or decoupling multiple faults is as follows: Mapped individual household phase loops whose absolute value of the Pearson correlation coefficient is greater than or equal to the preset synchronous correlation threshold are marked as suspected abnormal loops. Extract the electrical characteristic sequence of all suspected abnormal circuits within a preset time range before and after the occurrence of the main circuit abnormal event; Determine the core time window of the main circuit abnormal event and extract the abnormal electrical feature sequence of the main circuit within the window; Calculate the linear and nonlinear correlation between each suspected abnormal loop and the total path. The linear correlation is calculated using the absolute value of the Pearson correlation coefficient, and the nonlinear correlation is calculated using the maximum information coefficient. The two are weighted and fused to obtain the comprehensive correlation. Traverse all suspected abnormal loops to obtain their overall correlation degree, and determine the unique faulty loop or trigger the multi-fault decoupling process based on the difference in overall correlation degree.

6. The load monitoring system for an electricity metering box according to claim 5, characterized in that: The multi-fault decoupling process is triggered, and the specific process is as follows: The physical circuit that can independently cause abnormal changes in the electrical parameters of the main circuit in this abnormal event is defined as the fault source, and the electrical characteristic sequence of the main circuit within the core time window is input. Calculate the average mutual information of any two suspected anomalous loops. When the average mutual information of any two suspected anomalous loops is less than the preset independent judgment threshold, all suspected anomalous loops are determined to be approximately independent pairwise. When the approximate independence condition is met, the independent component analysis algorithm is used to decompose the total electrical characteristic sequence; when the condition is not met, a fault separation method based on timing difference is used to obtain independent fault source signals equal to the number of suspected abnormal circuits. Calculate the overall correlation between each fault source signal and the suspected abnormal circuit, and match the fault source signal with the suspected abnormal circuit with the highest correlation to obtain multiple fault circuits; sort them in descending order of overall correlation, and output the unique identifier of the fault circuit and the identifier of its corresponding phase line in sequence.

7. The load monitoring system for an electricity metering box according to claim 6, characterized in that: The specific process of constructing a typical fault feature library and calculating the dynamic time warping distance of the electrical feature sequences of the main circuit or fault circuit under different scenarios is as follows: Construct a typical fault feature library, and pre-set corresponding standard electrical feature sequences and standard electrical feature evolution trend sequences for each typical fault type; Based on the fault location results, the fault type is initially screened according to the scenario. When the main circuit itself is abnormal, the electrical feature sequence within the core time window of the main circuit is obtained. When the only faulty circuit is abnormal, the electrical feature sequence within the preset time range before and after the fault of the circuit is obtained. When multiple circuits are abnormal at the same time, the electrical feature sequence within the preset time range before and after the fault of each faulty circuit is obtained in order of comprehensive correlation from high to low. Substitute the above electrical feature sequence into the feature-weighted dynamic time warping algorithm to calculate the corresponding dynamic time warping distance.

8. The load monitoring system for an electricity metering box according to claim 7, characterized in that: The specific process of converting dynamic time-warped distance into confidence level and directly outputting the fault type upon meeting the criteria is as follows: The dynamic time warping distance between each electrical feature sequence and the corresponding standard electrical feature sequence is converted into the confidence level for fault root cause identification. The confidence level is inversely proportional to the dynamic time warping distance. When the highest confidence level is higher than the preset minimum confidence level threshold, and the difference between the highest confidence level and the second highest confidence level is greater than the preset confidence difference threshold, the fault type corresponding to the highest confidence level is directly output.

9. The load monitoring system for an electricity metering box according to claim 8, characterized in that: If the standard is not met, a second verification of the time-series evolution will be triggered. The specific process of finally outputting the results according to the scenario and pushing them to the operation and maintenance terminal is as follows: When the highest confidence level is lower than the preset minimum confidence level threshold, or the difference between the highest confidence level and the second highest confidence level is less than the preset confidence difference threshold, the time series evolution feature secondary verification process is triggered and executed according to the fault scenario classification. Electrical data of the corresponding circuits within a preset time period after a fault occurs is collected in different scenarios, and the evolution trend sequence of electrical features is extracted. The standard electrical feature evolution trend sequence corresponding to the top number of suspected fault types with confidence in the typical fault feature library is compared one by one. The trend consistency coefficient of each group of sequences is calculated by using cosine similarity, and the category with the highest coefficient is selected as the final fault type. Output the fault type according to the scenario, and send the results to the operation and maintenance terminal after completing the fault location and source tracing.

Images