A low-voltage cable branch box protection grade dynamic evaluation method and system

Abstract

The application discloses a low-voltage cable branch box protection level dynamic evaluation method and system, collects box body environment data and power supply time sequence data, forms a reliability feature set through joint degradation trend analysis, identifies a power failure early warning mark from a power supply time sequence history, and restores and constructs a power supply level evolution diagram through degradation path backtracking; in combination with the power failure early warning mark, a power supply evaluation section is delimited; a power failure threshold is determined through correlation analysis of the power supply evaluation section; node inspection rules are established according to a power failure diffusion priority; the power supply evaluation sequence and the emergency power recovery sequence are split; a power recovery priority sequence is generated by evaluating the power recovery attenuation trend; a power supply deviation amount is obtained by detecting stability; a power supply recovery path is determined by matching the power supply degradation coefficient with the power recovery priority sequence; a power supply cooperative recovery network is formed through breakpoint connection and repair; a protection level dynamic evaluation result is output according to the box body environment data features; and continuous dynamic evaluation of equipment protection level and improvement of emergency transfer reliability are realized.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance management technology for power distribution network equipment, and in particular to a method and system for dynamic evaluation of the protection level of low-voltage cable branch boxes. Background Technology

[0002] As the load distribution hub at the end of the distribution network, low-voltage cable distribution boxes are subjected to long-term outdoor temperature and humidity fluctuations, underground moisture immersion, and dust intrusion. Under multiple stresses, their outer casing seals and internal insulation media gradually deteriorate year by year. Current operation and maintenance generally adopts a management model of monthly or quarterly regular inspections plus completion acceptance records. The assessment of protection performance is mostly based on type test data and on-site visual inspection at the time of commissioning, which is difficult to reflect the true condition of the box inside the equipment after many years of operation. The periods of most rapid deterioration often happen to fall between two inspection cycles and go unnoticed.

[0003] With the expansion of urban power distribution, the number of branch boxes is enormous while the maintenance teams are relatively fixed. Inspection plans are often arranged according to logbook order or the year of commissioning, easily leading to long-term neglect of equipment whose external appearance is still good but whose incoming circuit insulation is nearing its critical point. In addition, during the switching operation after a circuit goes out, the on-site personnel often decide which backup circuit to connect based on the immediate signal of whether there is power, with little consideration for the health condition of the backup circuit itself. As a result, circuits that have just had power restored often trip again within a short period of time, leading to frequent emergency repairs. Summary of the Invention

[0004] This invention discloses a method and system for dynamic evaluation of the protection level of low-voltage cable branch boxes. It collects environmental data and power supply time-series data of the box, and combines this with degradation trend analysis and early anomaly detection to construct a reliability feature set and power outage warning indicators. Through degradation path backtracking, it constructs a power supply level evolution diagram and defines power supply evaluation segments. Based on historical power outage patterns between circuits, it determines an adaptive power outage threshold, thereby generating node inspection rules that match the degree of degradation. It incorporates the degradation rate into emergency power transfer matching, and through breakpoint connection repair, forms a power supply collaborative recovery network. Combining the characteristics of the box's environmental data, it outputs dynamic evaluation results of the protection level, improving the timeliness of protection assessment.

[0005] The first aspect of this invention proposes a method for dynamically evaluating the protection level of low-voltage cable branch boxes, comprising the following steps: Collect enclosure environmental data and power supply time sequence data, and perform joint degradation trend analysis to form a reliability feature set. Identify early power failure characteristics from the power supply time sequence data to obtain power failure early warning indicators. A power supply level evolution diagram is constructed by performing degradation path backtracking on the reliability feature set, and a power supply evaluation segment is determined based on the power supply level evolution diagram and the power failure early warning indicator constraint. The power supply assessment section is subjected to inter-loop correlation analysis to determine the power failure threshold. Based on the power failure threshold and the power failure propagation priority weight allocation of the power supply assessment section, node inspection rules are established. Based on the node inspection rules, the circuit is divided into a power supply assessment sequence and an emergency power restoration sequence. The power restoration success rate decay trend assessment is performed on the emergency power restoration sequence to generate a power restoration priority sequence. The power supply assessment sequence is subjected to level stability detection to obtain the power supply deviation. The power supply degradation coefficient is determined by analyzing the deviation rate based on the power supply deviation amount. The power supply degradation coefficient and the power restoration priority sequence are matched to determine the power supply recovery path by the degradation acceleration stage. The power supply recovery path and the power supply assessment sequence are connected and repaired at the maximum degradation rate to form a power supply collaborative recovery network. Based on the power supply collaborative recovery network and the enclosure environment data, the protection status is assessed and the dynamic assessment result of the protection level is output.

[0006] A second aspect of this invention provides a dynamic evaluation system for the protection level of low-voltage cable branch boxes, comprising: The data acquisition module is used to collect environmental data and power supply time sequence data of the enclosure and perform joint degradation trend analysis to form a reliability feature set. It identifies early power failure characteristics from the power supply time sequence data to obtain a power failure early warning indicator. The evolution reconstruction module is used to perform degradation path backtracking and reconstruction on the reliability feature set to construct a power supply level evolution map, and to determine the power supply evaluation segment based on the power supply level evolution map and the power failure early warning indicator constraint. The rule generation module is used to perform inter-loop correlation analysis on the power supply assessment segment to determine the power outage threshold, and to establish node inspection rules based on the power outage threshold and the power outage propagation priority weight allocation of the power supply assessment segment. The circuit restoration module is used to divide the circuit into a power supply assessment sequence and an emergency power restoration sequence according to the node inspection rules. It performs a power restoration success rate decay trend assessment on the emergency power restoration sequence to generate a power restoration priority sequence, and performs a level stability detection on the power supply assessment sequence to obtain the power supply deviation. The result output module is used to determine the power supply degradation coefficient by performing deviation growth rate analysis based on the power supply deviation amount, match the power supply degradation coefficient with the power restoration priority sequence to determine the power supply recovery path by performing degradation acceleration stage transfer, perform maximum degradation rate breakpoint connection repair on the power supply recovery path and the power supply assessment sequence to form a power supply collaborative recovery network, and perform protection status assessment based on the power supply collaborative recovery network and the enclosure environment data to output the protection level dynamic assessment result.

[0007] The beneficial effects of this invention are reflected in the following points: 1. The voltage anomaly distribution pattern before power outage in the historical power supply sequence is the core judgment basis for the protection status perception level of this invention. Traditional solutions trigger assessment based on the power outage event itself, while this invention identifies the density dispersion and periodic alternation characteristics of the abnormal interval before and after the re-inspection date as the early power outage signal, so that the response node of protection level degradation is shifted from after the power outage to the observable precursor stage. The power supply level evolution diagram fills the gaps in historical data by retrospectively restoring the degradation path, and the power supply level change trajectory of the entire operation cycle of the enclosure can be continuously presented. The starting position of the degradation acceleration stage can be accurately located in time sequence rather than relying on post-event manual tracing. 2. The power outage threshold incorporates the seasonal stratification and the trend of successively shortening power outage intervals into the calculation, distinguishing the differences in the causes between short-cycle transient power outages and long-cycle continuous power outages. The threshold is adaptively updated according to the actual operation law of the circuit rather than using a fixed empirical value. The node inspection rules use the historical frequency of no early warning reports near the threshold as the weighting basis, dynamically binding inspection density with the cumulative risk level. Maintenance resources are tilted towards truly high-risk nodes rather than being evenly distributed based on equipment age or appearance. 3. The selection of emergency power transfer candidate circuits is based on the current single degradation value as a tolerance, prioritizing circuits with low degradation rates to undertake power transfer, reducing the probability of power transfer circuits losing power again in the short term after restoration. The power supply collaborative restoration network performs connection repair at the breakpoint with the highest degradation rate, with the repair location precisely corresponding to the weakest node in each restoration path. The control priority of each node in the output power supply guarantee control command is updated in real time according to the repair quality, and nodes with low repair reliability are not prematurely removed from the monitoring queue due to a temporary increase in weight. Attached Figure Description

[0008] Figure 1 This is a flowchart of a dynamic evaluation method for the protection level of a low-voltage cable branch box according to the present invention.

[0009] Figure 2 This is a schematic diagram of the power supply level evolution diagram and degradation path tracing and restoration of the present invention.

[0010] Figure 3 This is a structural block diagram of a dynamic evaluation system for the protection level of a low-voltage cable branch box according to the present invention.

[0011] Among them: 1-Original grade discrete sequence; 2-Grade drop start point; 3-Brief recovery section; 4-High grade difference missing section; 5-Forward neighbor mean benchmark; 6-Backward neighbor mean benchmark; 7-Grade interpolation completion section; 8-Power supply grade evolution curve; 9-Deterioration acceleration trend label; 10-Grade step breakpoint. Detailed Implementation

[0012] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0013] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0014] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0015] The technical solutions of the embodiments of this application will be described below.

[0016] like Figure 1 As shown, this embodiment of the invention provides a method for dynamic evaluation of the protection level of a low-voltage cable branch box, including the following steps S101-S105: Step S101: Collect enclosure environmental data and power supply time sequence data, and perform joint degradation trend analysis to form a reliability feature set. Identify early power failure characteristics from the power supply time sequence data to obtain power failure early warning indicators.

[0017] Specifically, environmental data and power supply timing data of the enclosure are collected. The enclosure's environmental data is collected concurrently by temperature and humidity sensors, water accumulation detection contacts, and insulation aging monitoring modules at differentiated sampling rates. Temperature and humidity readings are continuous quantities on the order of seconds, while water accumulation status is a discrete signal indicating contact opening and closing. The timestamp accuracy of these two types of signals when they reach the convergence layer is inconsistent. They need to be re-aligned using a unified time base and archived together with the corresponding re-inspection and patrol dates. Samples with alignment deviations exceeding the tolerance are recorded separately for traceability. Lightning-induced overvoltage will cause a momentary interruption of power supply to the sensors inside the enclosure, resulting in a short-term frame loss in the enclosure's environmental data, which then recovers quickly. These missing frames are distinguished according to the source of lightning interference and are not included in the degradation trend baseline. Occasional lightning disturbances and normal environmental degradation are strictly distinguished in terms of the source of the missing data. When the missing data from the same lightning source occurs consecutively and exceeds the upper limit, it is upgraded to a hardware damage pending verification item and registered separately. Power supply timing data is continuously collected by the voltage transformer on the incoming side of the branch box at a fixed sampling period. The sampling period is preset at the factory. Points where the absolute value of the voltage difference between adjacent batches exceeds the threshold are identified as voltage mutation points. Periods with a high density of mutation points are weighted less in subsequent analyses. If the voltage recovery of batches after historical power outages deviates from the normal power restoration curve, the corresponding weight is reduced according to the deviation. Poor heat dissipation at the sensor installation location may cause the ambient temperature readings of the enclosure to be systematically higher during high load periods. After cross-verification with the concurrent load current, batches with verification differences exceeding the limit are not included in the absolute temperature threshold comparison. During planned power outages for maintenance, the voltage of the power supply timing data returning to zero is considered normal operation and maintenance. It is listed separately according to the source of the planned power outage and is not included in the power outage feature identification. Planned power outages and spontaneous power outages of equipment are strictly distinguished in terms of source.

[0018] A reliability feature set is formed by jointly analyzing the degradation trend of enclosure environmental data and power supply time-series data. The sliding window slope k = ΔX / Δt for each parameter batch mean, where ΔX is the parameter mean within the window and Δt is the corresponding time span. Parameters with continuously negative k are considered degradation direction parameters. All monitoring indicators of enclosure environmental data are included in the degradation direction determination. The current mean of each degradation direction parameter is used as the degradation amplitude score based on its proximity to the degradation threshold after normalization. The slope is used as the degradation rate score after normalization based on the historical maximum degradation rate. The power supply level score g of the node is obtained by weighted summation of the two scores. The weights are determined by the precursor prediction accuracy of the two scores in historical power outage events. The reliability feature set is the node reliability vector composed of the current mean of each degradation direction parameter, the slope, and the power supply level score g. When the degradation direction of the enclosure environment data and the power supply time series data is consistent, the joint slope is the weighted average of the two slopes. Nodes with degradation of only a single data source parameter have limited confidence in their scores and are marked with a single parameter label. When consecutive batches exceed the threshold, cross-validation is performed to check for hidden degradation in the enclosure environment data. In the power supply time series data, the voltage of power outage batches returning to zero does not constitute a degradation signal. The reliability feature set estimates the degradation slope on both sides of the power outage batch. Nodes with slope differences exceeding the threshold on both sides are archived independently with segmented slopes on both sides. Batches deviating from the normal power restoration curve are substituting their slope estimates with a reduced weight based on the deviation magnitude. The impact of the reduced weight on the normal slope baseline is controlled within the set deviation range. Nodes in the reliability feature set where the number of continuously deteriorating batches monotonically increases within consecutive batches are classified as degradation acceleration and marked with an acceleration label. This label is used to adjust the sudden drop threshold when identifying subsequent level drops.

[0019] In some embodiments, the step of identifying early power outage characteristics from the power supply time sequence data to obtain a power outage early warning identifier includes: verifying the power supply time sequence data segment by segment from the last maintenance to the first commissioning to determine historical voltage anomaly nodes; generating an anomaly interval distribution map by performing anomaly time and re-inspection date interval analysis on the historical voltage anomaly nodes; quantifying the early fluctuation density dispersion of the anomaly interval distribution map to form a fluctuation weight distribution; and identifying periodic alternating fluctuation segments based on the fluctuation weight distribution to obtain a power outage early warning identifier.

[0020] Historical voltage anomaly nodes are identified by segment-by-segment verification of power supply timing data from the last maintenance to the first commissioning. Starting from the last maintenance time and proceeding towards the first commissioning, power supply timing data is scanned sequentially. Segments where voltage readings deviate from the normal operating bandwidth and continuously cross the anomaly confirmation window are identified as anomaly segments. The start time and deviation magnitude of each anomaly segment are archived to form historical voltage anomaly nodes. Larger deviations indicate more severe voltage quality deviations during that period. The normal operating bandwidth of historical voltage anomaly nodes is defined as the average voltage of each batch after the feeder has stabilized after commissioning, plus or minus a certain number of standard deviations. Segments exceeding this bandwidth and continuously crossing the anomaly confirmation window are identified as anomaly segments. During the initial commissioning and wiring debugging phases, voltage fluctuations far exceed this bandwidth; these are listed separately according to the debugging markers and are not included in the bandwidth statistics. The end time of debugging is taken as the effective timing start point. If metering devices are replaced during the retrospective period, the voltage readings before and after the replacement have systematic deviations due to the adjustment of the metering transformer ratio, which are not actual voltage changes. Therefore, independent voltage baselines are established on both sides of the transformer ratio replacement point for historical voltage anomaly nodes to isolate metering deviations from actual anomalies. The seasonal load differences cause the normal operating bandwidth of power supply timing data to shift systematically with the season. The anomaly confirmation window must be executed after seasonal factor correction. Anomaly segments identified after seasonal correction are listed separately for their seasonal source and weighted differently from anomalies caused by non-seasonal factors. When the number of anomaly nodes with seasonal sources exceeds that with non-seasonal sources within the same re-inspection interval, it indicates that the current anomaly density of that node is mainly driven by seasonal load characteristics rather than equipment degradation. The width of the anomaly confirmation window is set according to the noise level of the feeder voltage measurement. If the window is too narrow, measurement spikes will be included in the anomaly count; if it is too wide, short-term anomalies will be missed.

[0021] An anomaly spacing distribution map is generated by analyzing the time interval between the anomaly execution and the re-inspection date of historical voltage anomaly nodes. The difference between the start time of each anomaly segment and the most recent re-inspection date is Δd = t_anomaly - t_reinspect, where t_anomaly is the start time of the anomaly segment and t_reinspect is the most recent re-inspection completion time in the archived records. A positive and small Δd indicates that the anomaly appeared early after the re-inspection and the equipment condition deteriorated rapidly after the re-inspection. The Δd of each historical voltage anomaly node is divided into several equally wide intervals according to the spacing value. The count of historical voltage anomaly nodes within each interval constitutes the density value of that interval segment. The density values ​​of all intervals, arranged from smallest to largest spacing, constitute the main frequency distribution of the anomaly spacing distribution map. The density peaks in the anomaly spacing distribution map are concentrated in nodes with small spacing intervals, indicating that the equipment's anomalies mostly occurred shortly after the re-inspection, and the re-inspection's effect on suppressing deterioration has a limited duration. Historical voltage anomaly nodes that have continuously experienced voltage anomalies in several batches after the annual routine re-inspection have smaller and denser Δd values; the weight of such densely distributed segments is correspondingly increased. If multiple abnormal nodes with similar Δd values ​​exist within the same re-inspection interval, a lower variance threshold is applied to identify nodes with higher weighting for that period. For nodes originating from the commissioning period, the date of the first formal re-inspection after commissioning is used instead of the temporary verification date before the end of commissioning when calculating Δd. The temporary verification date and the formal re-inspection date are strictly distinguished in the spacing calculation to ensure consistency in the statistical caliber of Δd. For nodes originating from ratio adjustments, the spacing analysis uses the re-inspection date after the ratio adjustment as the benchmark. The spacing before and after the adjustment are archived independently and are not merged into continuous sequences. The spacing sequences on both sides of the adjustment point are not merged into continuous sequences. Each Δd value is also archived with a dual index according to its season and re-inspection batch, facilitating horizontal comparison of density within the same season and re-inspection cycle.

[0022] An early fluctuation density dispersion measure is quantified to form a fluctuation weight distribution for the anomaly spacing distribution map. The dispersion is taken as the batch-to-batch variance of the node spacing values ​​in the anomaly spacing distribution map within each re-inspection interval. A smaller variance indicates that early anomalies are concentrated in fixed intervals and have strong regularity. The inverse of the variance of each interval is normalized and aggregated in time sequence to obtain the fluctuation weight distribution. The normalization is based on the sum of the inverses of the variances of all intervals to calculate the relative proportion of each segment. Intervals with smaller variances in the fluctuation weight distribution have higher weights. When periodically alternating identification, high-weight segments are given priority as references. During the flood season, the rise in groundwater level increases the water inflow rate of the tank. Nodes with smaller Δd appear densely in the anomaly spacing distribution map within the corresponding batch, and the variance of each interval is systematically low. The variance of this flood season batch does not participate in the update of the normal dispersion baseline for the non-flood season. The normal dispersion baseline is maintained separately for non-flood season batches. Nodes with rapid degradation sources after re-inspection have their weights increased in variance calculation. When these nodes appear in clusters, their impact on the fluctuation weight distribution is more pronounced, reflecting their higher contribution to the risk of early power outages compared to general abnormal nodes. When the cumulative weight of rapidly deteriorating nodes in consecutive batches exceeds the upper limit, the corresponding segment is pre-classified as high-risk, and its validity verification threshold is relaxed accordingly. For nodes with variable ratio adjustment sources, the dispersion is calculated separately for the interval values ​​archived before and after the adjustment. The fluctuation weight distribution is independently weighted before and after the adjustment, and the baseline difference introduced by the adjustment does not dilute the true dispersion evolution. Before variance reciprocal normalization, the sample size of each interval is screened out using a lower limit. Intervals with insufficient sample size are linearly imputed with the variance of adjacent intervals to prevent the accidental underestimation of the variance of small samples from artificially inflating their weights.

[0023] Power outage warning indicators are obtained by identifying periodically alternating fluctuation segments based on the fluctuation weight distribution. Time segments in the fluctuation weight distribution where high-weight and low-weight segments alternate are identified as candidates for periodic alternation. A candidate segment is confirmed as a valid periodic fluctuation segment when the alternation frequency exceeds the minimum number of cycles and the alternation amplitude exceeds a threshold. The start and end times and alternation frequencies of each valid segment are archived to form the power outage warning indicator. During shift changes and maintenance shutdowns, sudden load changes cause regular alternations in the incoming line voltage between high and low levels. If the alternation frequency closely matches the known production shift cycle, the alternation in this segment is attributed to the load cycle rather than equipment degradation. The load cycle source is marked and retained for exclusion during subsequent segment division evaluation. In the fluctuation weight distribution, the absolute value of the weight difference between adjacent high and low weight segments constitutes an alternation amplitude sequence. The moving average of this sequence is compared with the alternation amplitude threshold. A valid alternation is confirmed only when the number of consecutive batches exceeding the threshold reaches the minimum number of cycles. High-density batches during the flood season are first removed from the weight sequence according to their seasonal markers before participating in the above comparison. Those still meeting the standard after removal are written into the power outage warning flag. For alternation segments in the fluctuation weight distribution where seasonal source nodes are concentrated, a relaxation factor is applied to the alternation amplitude threshold to retain valid early power outage signals. The higher the alternation frequency in the power outage warning flag, the stronger the early power outage regularity. Nodes with frequencies exceeding the high-frequency threshold are classified as high-frequency early power outages, and their level is correspondingly downgraded when updating the reliability feature set in the next collection cycle. The alternation amplitude of valid periodic fluctuation segments is measured by the average weight difference between high and low weight segments, and the alternation frequency is counted as the number of high and low weight switching times per unit time; both are included in the segment record.

[0024] Step S102: Perform degradation path backtracking and reconstruction on the reliability feature set to construct a power supply level evolution diagram, and determine the power supply assessment segment based on the power supply level evolution diagram and power failure early warning label constraints.

[0025] In some embodiments, the step of constructing a power supply level evolution map by performing degradation path backtracking and restoration on the reliability feature set includes: generating a level difference set by performing statistical analysis of the level differences of distributed power supply nodes based on the reliability feature set; determining a group of nodes with missing high-level differences based on the brief recovery period after a sudden drop in level in the level difference set; forming a level interpolation distribution by performing neighboring node interpolation to complete the level interpolation distribution; and constructing a power supply level evolution map by performing power supply level sequence restoration on the level interpolation distribution.

[0026] A grade difference set is generated by statistically analyzing the grade differences of distributed power supply nodes based on the reliability feature set. The grade dimension difference of the reliability vectors of adjacent batches is Δg = g_b - g_{b-1}, where g_b and g_{b-1} are the power supply grade scores of the current batch and the previous batch, respectively. The grade difference of each node is calculated sequentially from this. A negative Δg indicates that the power supply grade of the node has decreased compared to the previous batch, while a positive Δg indicates an increase. The direction and magnitude of Δg together characterize the evolution characteristics of the power supply grade of the node. The grade difference set is then grouped and archived by node according to the time sequence of Δg for each node. When the load current increases in a stepwise manner, the power supply grade score of the corresponding batch in the reliability feature set shows a stepwise decrease, and Δg appears as a pulse negative value in the step batch. The Δg of such stepwise decreasing batches is not included in the estimation of the slope of the normal degradation trend. The load step effect and equipment degradation are distinguished by the source of Δg. For nodes in the reliability feature set with sources of accelerated degradation, the absolute value of the acceleration segment Δg is consistently large. The grade difference set separately lists the source of this type of accelerated segment Δg, allowing it to be triggered with a lower drop threshold when identifying nodes with missing high-grade differences. For nodes in the reliability feature set with a single parameter source, the confidence level of the grade score is limited, and the corresponding Δg calculation result is recorded as low confidence. Low-confidence batches are treated with a conservative threshold when identifying grade drops. After processing with a conservative threshold, the interference of single parameter estimation error on the drop identification result is controlled within an acceptable range. Before calculation, the Δg time series is first removed from batches with missing samples and then linearly filled in with valid batches before and after, ensuring that grade differences do not experience breaks due to missing measurements of individual batches.

[0027] Based on the rank difference set, identify the high-rank difference missing node group in the short-term rebound segment after a sudden drop in rank. For example... Figure 2As shown, the grade difference set is built upon the original grade discrete sequence 1. The time range covered by the brief rebound segment 3 after the grade drop starting point 2 is defined as the high-grade difference missing interval 4, and the high-grade difference missing node group is determined accordingly. The grade drop starting point 2 is marked by the first batch where the cumulative negative value of Δg in consecutive batches exceeds the drop threshold. The brief rebound segment 3 is defined by the subsequent batches where Δg turns positive and the continuous batches are below the upper limit of the brief rebound. The high-grade difference missing interval 4 is defined by the batch corresponding to the grade drop starting point 2 as the left boundary. After the box seal fails, moisture evaporates outward during the high-temperature period, causing a brief grade rebound in the grade difference set during the high-temperature batches. This rebound is due to the improvement of the ambient temperature rather than the actual recovery of the equipment. Therefore, the rebound batches that overlap with the high-temperature period are marked as thermal interference, and their rebound is not considered as true grade recovery. The duration of the drop at this node is re-evaluated after excluding the thermal interference batches. The Δg values ​​of each node in the grade difference set are accumulated in batches to form a cumulative negative value sequence. The first batch whose cumulative negative value exceeds the sudden drop threshold is marked as the sudden drop starting point. The load step effect batches will artificially inflate the accumulated amount due to their pulsed negative values. Therefore, these batches are first removed according to the step mark before accumulation. The sudden drop starting point that still exceeds the threshold after recalculation is included in the high grade difference missing node group. When the distance between two adjacent short recovery segments is less than the fusion window, they are merged into a single sudden drop event. After merging, the recovery amplitude is the average of the recovery amplitudes of each segment. In the high grade difference missing node group, the node whose sudden drop amplitude exceeds the depth threshold and whose recovery amplitude is lower than the lower limit is defined as a deep sudden drop. If the grade cannot be actually recovered after permanent insulation breakdown, a more conservative completion parameter is used when interpolating.

[0028] The grade interpolation distribution is formed by interpolating neighboring nodes to complete the high-grade difference missing node group and grade difference set. The forward neighbor mean benchmark 5 is taken as the grade mean g_prev of several batches with stable Δg direction before the high-grade difference missing interval 4. The backward neighbor mean benchmark 6 is taken as the grade mean g_next of several batches after the interval. The grade interpolation completion segment 7 is estimated by the weighted mean g_fill=w_prev×g_prev+w_next×g_next. w_prev and w_next are the weights of the forward and backward neighbor segments, respectively, and are taken according to w=1 / d, where d is the distance of the neighbor segment from the starting batch of the sudden drop. The closer the neighbor segment, the greater the weight. The grade interpolation distribution is formed by merging the interpolation results of each node with the original grade sequence. In high-level difference missing node groups, nodes with long-term water accumulation leading to insulation aging exhibit lower average grade values ​​in adjacent segments before and after a sudden drop in batch, resulting in overall lower interpolation results. These nodes are classified as bilaterally low baselines, limiting their interpolation accuracy, and a conservative estimate is used when restoring the grade sequence. Before interpolation, the consistency of the direction of the grade difference set at both ends of the missing interval is checked; if the directions at both ends are opposite, a direction conflict warning is added to the interpolation results. When the direction of the grade difference set Δg in the forward adjacent segment frequently reverses, the forward mean is insufficiently representative. The weight of forward unstable nodes is reduced in the grade interpolation distribution, with the reduction measured by the ratio of the frequency of forward Δg reversals to the number of batches. After reduction, the weights are concentrated in the backward adjacent segments. When the duration of the forward adjacent segment is insufficient to meet the minimum requirement, the average grade value of the same type of nodes in the entire network during the same period is used instead of the forward mean, and the borrowing source is listed separately for each batch. For nodes with thermal interference sources in the high-level difference missing node group, the thermal interference batch is excluded first during interpolation before the neighboring segment is redefined. The short-term rebound corresponding to thermal interference is not used as the backward neighbor mean benchmark 6. The interpolation benchmark is the backward mean after exclusion.

[0029] Power supply level evolution diagrams are constructed by restoring the power supply level sequence from the interpolation distribution. The interpolation completion segment 7 of the interpolation distribution is merged with the original level sequence of each node according to batch time sequence. The batch containing the completion segment retains its interpolation source marker to distinguish between measured and estimated values. After smoothing the merged sequence using a sliding window mean, it forms the main body of the level time sequence for each node in the power supply level evolution curve 8. After smoothing, the level value of each batch is labeled with the proportion of the measured batch as the confidence level of that batch; the confidence level is correspondingly reduced for periods with a higher proportion of interpolated batches. When continuous construction vibrations from surrounding areas affect the acquisition accuracy, the interpolation distribution will exhibit abnormal fluctuations in interpolation results in construction batches. The level time sequence of these construction batches is replaced by a linear extension of adjacent normal batches, and the replacement segment is recorded separately for identification during subsequent power supply assessment segment delineation. For nodes in the interpolation distribution with low baseline sources on both sides, the smoothing window is widened to a set multiple of the conventional window. Widening the smoothing window provides stronger suppression of random errors in batches with limited interpolation accuracy. When there is a significant horizontal difference on both sides of the deep drop node in the power supply level evolution curve 8, the power supply level evolution diagram marks a level step breakpoint 10 at the drop batch. Level step breakpoint 10 serves as the natural segment boundary when delineating the power supply assessment segment, and the assessment segments before and after the step are archived independently. For power supply level evolution curve 8 where the slope between batches of each node is continuously negative and the absolute value of the slope gradually increases, a degradation acceleration trend label 9 is marked. Nodes with degradation acceleration trend label 9 are considered valid power supply assessment segments with a lower level stability threshold when delineating assessment segments under the power outage warning indicator constraint. The window width of the sliding window with mean smoothing is adaptively adjusted according to the batch noise amplitude of each node's level sequence. Nodes with high noise have larger window widths to more strongly suppress random jitter. The window width and the number of batches used are retained along with the curve.

[0030] Power supply assessment segments are determined based on the power supply level evolution chart and power outage warning indicators. In the power supply level evolution chart, segments where the level difference between batches is continuously lower than a set bandwidth are considered stable segments in the level sequence. These stable segments intersect with the early fluctuation periods indicated by the power outage warning indicators in time sequence. Segments where the number of intersecting batches exceeds the shortest assessment length are considered valid power supply assessment segments, aggregated by the start and end batch numbers and the average level value within the segment. During peak production seasons and high electricity consumption periods, multiple batches in the power supply level evolution chart continuously decline in level values. These declining segments highly overlap with the high-frequency early power outage periods in the power outage warning indicators. These overlapping segments constitute the main body of the power supply assessment segments. A high rate of level decline within a segment indicates deeper equipment degradation; batches with a decline rate exceeding a threshold are considered to have accelerated degradation. For periods originating from load cycles in the power outage warning indicators, level fluctuations are driven by load cycles rather than equipment degradation; therefore, the corresponding overlapping batches are not included in the equipment degradation assessment baseline. In the power supply level evolution diagram, batches of levels that suddenly drop and then quickly rebound, with the drop exceeding a threshold, are considered transient level anomalies. Transient anomaly batches do not constitute the boundary of an effective power supply assessment segment; their transient nature is recorded separately. Stable segments before and after the transient anomaly batch each form independent power supply assessment segments. When the difference in the average level within either segment exceeds a threshold, it indicates a jump in overall power supply quality before and after a brief fluctuation. The average level value within an effective power supply assessment segment is calculated as the duration-weighted average of the levels of each batch within the segment. The weight is taken from the duration of each batch, ensuring that the contribution of longer batches to the segment average is proportional to their duration. Isolated stable segments shorter than the shortest assessment length are merged into adjacent assessment segments instead of forming separate segments.

[0031] Step S103: Perform inter-loop correlation analysis on the power supply assessment section to determine the power outage threshold, and establish node inspection rules based on the power outage threshold and the power outage propagation priority weight allocation of the power supply assessment section.

[0032] In some embodiments, the step of performing inter-loop correlation analysis on the power supply assessment segment to determine the power outage threshold includes: dividing the power supply assessment segment into high-risk power outage circuits and power outage interval fluctuation distributions according to the power outage risk level; establishing a threshold candidate parameter set based on the high-risk power outage circuits through periodic hierarchical seasonal integration; matching the threshold candidate parameter set with the power outage interval fluctuation distribution to form a threshold correction distribution by successively shortening the interval trend; and using the threshold correction distribution to screen effective threshold intervals to determine the power outage threshold.

[0033] The power supply assessment section is divided into high-risk power outage circuits and power outage interval fluctuation distributions according to the power outage risk level. The risk level is defined by the frequency of power outage events for each circuit within the power supply assessment section, F = N_event / T_eval, where N_event is the total number of power outage events within the assessment section and T_eval is the duration of the assessment section. The start and end batch numbers archived in the power supply assessment section and the power outage event records within the section jointly provide the basis for frequency statistics. Circuits whose F exceeds the high-risk threshold are classified as high-risk power outage circuits. The inter-batch standard deviation of the interval sequence of adjacent power outage events for each circuit forms the power outage interval fluctuation distribution according to the circuit archive. During periods with high frequency of seepage in underground cable trenches, multiple circuits may simultaneously exceed the threshold F. These synchronously selected circuits are marked as flood season clusters, and their F values ​​do not participate in the update of the normal high-risk baseline. The normal high-risk baseline is maintained independently using the F values ​​of non-flood season batches. In the power supply assessment segment, for batches of load disturbance sources, power outage events caused by load disturbances are excluded when calculating F. Those whose F still exceeds the threshold after exclusion are retained as high-risk events; those that do not exceed the threshold are retained for risk monitoring but are not included in the high-risk power outage circuits. Circuits with a continuously increasing standard deviation in the power outage interval fluctuation distribution indicate that the regularity of their power outage intervals is weakening and their power outage behavior is becoming increasingly unpredictable. Circuits with a standard deviation growth rate exceeding the threshold are marked as experiencing increased volatility and trigger an independent parameter estimation branch. In high-risk power outage circuits, if the F of a certain circuit monotonically increases within consecutive batches, it is marked as having increasing frequency, and the growth rate is used to correct the seasonal layer parameters when generating threshold candidates. The statistical window for the power outage event frequency F is strictly aligned with the start and end of the power supply assessment segment. Cross-segment events are counted separately according to the segment to which their start time belongs and are not repeatedly accumulated.

[0034] For example, the step of establishing a threshold candidate parameter set by periodic hierarchical seasonal integration based on the high-risk power failure circuit includes: dividing the high-risk power failure circuit into a short-cycle power failure set and a long-cycle power failure set according to the duration of power failure; performing power failure cause difference type identification based on the short-cycle power failure set to generate a short-cycle cause classification set; integrating the short-cycle cause classification set and the long-cycle power failure set into a seasonal power failure set segment priority interval to determine the threshold constraint interval; and performing interval parameter segmentation analysis on the threshold constraint interval to establish a threshold candidate parameter set.

[0035] Based on high-risk power outage circuits, power outages are segmented into short-cycle and long-cycle sets according to their duration. The duration is calculated from the start of each power outage event to the number of batches from which power is restored. Power outage events with a duration below the upper threshold of the short-cycle set are assigned to the short-cycle set, while those exceeding the lower threshold of the long-cycle set are assigned to the long-cycle set. Power outage events between these two thresholds are recorded separately as intermediate events and are not forcibly merged into either set. Transient overcurrents caused by lightning strikes will trip circuit breakers, and power outages are usually restored by automatic reclosing within minutes; these events mostly fall into the short-cycle set. Grounding faults caused by aging cable insulation require manual repairs, resulting in power outages lasting several hours; these events mostly fall into the long-cycle set. The duration distribution of these two types of power outages differs significantly, and separate archiving provides a time-stratified basis for subsequent cause identification. For high-risk power outage circuits that already include power outage events originating from the flood season, the flood season origin is retained during segmentation. For both short-cycle and long-cycle power outage sets, flood season subsets are separately defined for events originating from the flood season. During seasonal integration, the flood season subsets are processed independently and not mixed with similar events outside the flood season for parameter estimation. Power outage events whose duration falls close to the upper threshold of the short-cycle period are sensitive to the threshold. For high-risk power outage circuits, events within the boundary tolerance are classified according to the dominant type of the most recent historical batch to maintain consistency in causal classification. The thresholds for dividing short-cycle and long-cycle events are set separately according to the voltage level of the circuit and the historical fault spectrum. The event counts and duration distributions of both sets are retained for causal identification.

[0036] Based on short-cycle power outage sets, a short-cycle cause classification set is generated by identifying the differences in power outage causes. The cause difference type is determined by combining the preceding voltage fluctuation characteristics of each event in the short-cycle power outage set with concurrent enclosure environmental data. Events with a rapid recovery after a sudden voltage drop in the preceding batch are classified as contact power outages, while sudden power outages without obvious preceding signs are classified as impulse power outages. The preceding voltage drop in contact power outages is usually less than 15% of the rated voltage and occurs in more batches than in impulse power outages. Impulsive power outages are characterized by a sudden drop in voltage to zero without warning. The differences in the preceding characteristics of these two types of events constitute the core criterion for identification. The cause type of each event is merged with the original event records to establish the short-cycle cause classification set. For circuits where terminals have become loose due to long-term vibration, the voltage fluctuation amplitude gradually increases several batches before the power outage, unlike the sudden preceding impulse power outage. This gradual fluctuation preceding event triggers a lower power outage interval identification threshold, indicating that the circuit is already in a contact degradation path. When the proportion of a certain type of event in a short-cycle power outage cluster continuously increases within consecutive batches, it is marked as a dominant type evolution, indicating that the cause of the power outage at that node is concentrating in a certain direction. During seasonal integration, the parameter weight of the batches concentrated with the dominant type is increased. Events with insufficient basis for cause determination are recorded as pending verification types. The type of the event pending verification is substituted by the type of the nearest adjacent determined event, and its type is borrowed. A high proportion of type borrowing in the short-cycle cause classification cluster indicates that the quality of the historical power outage data at that node is low, and the confidence level of the node's parameters is downgraded when generating threshold candidates. The distinction between contact power outages and impulse power outages is determined by a combination of two indicators: the slope of the preceding voltage fluctuation and the number of continuous batches. Events that meet one indicator but fail the other are temporarily suspended pending verification and not forcibly classified.

[0037] Threshold constraint intervals are determined by integrating the priority intervals of seasonal power outage concentrated segments using short-cycle cause classification sets and long-cycle power outage sets. Concentrated segments are defined based on consecutive batches of power outage events exceeding a density threshold within each seasonal layer. After identifying concentrated segments, both the short-cycle cause classification set and the long-cycle power outage set cross-reference them by temporal overlap. Seasonal segments with overlap exceeding a threshold are considered dual-type concentrated segments, and the batch range covered by these dual-type concentrated segments serves as the priority integration boundary for the threshold constraint interval. During the rainy season, when the enclosure is flooded, both short-cycle power outages (short-circuit tripping) and long-cycle power outages (insulation penetration) are triggered simultaneously within a short period. The short-cycle cause classification set and the long-cycle power outage set form highly overlapping dual-type concentrated segments during this period. The threshold constraint interval is marked with these dual-type segments for the flood season, and the integration parameters, corrected for historical averages during the flood season, are independent of the normal constraint boundary during the non-flood season. For batches of short-cycle cause-related concentrated types, their weight is reduced during concentrated segment density calculation. Concentrated segments whose density still exceeds the threshold after weight reduction are retained. Concentrated segments whose boundaries shrink due to weight reduction are marked as boundary shrinkage and treated with extended tolerance during segment analysis. When the dominant type evolution batch of short-cycle cause-related concentrated types overlaps with the concentrated segments of long-cycle power outage concentrated types, the threshold constraint interval is marked as cause-related evolution concentrated, indicating that the synchronous concentration of different power outage types during this period may be driven by common upstream factors such as overall line aging. This type of interval is estimated separately during segment analysis. The integration boundary of dual-type concentrated segments extends to both sides based on the intersection of the two types of power outage density curves. The extension width is constrained by the upper limit of the seasonal layer duration and does not expand across the seasonal boundary. After extension, the overlapping area of ​​the two types is still the priority verification segment.

[0038] A threshold candidate parameter set is established by segmenting and analyzing the parameters within the threshold constraint interval. Each constraint segment of the threshold constraint interval is estimated segment by segment using the mean power outage interval μ_c and the standard deviation of fluctuation σ_c. μ_c is the mean of the power outage interval sequence within the segment, and σ_c is the standard deviation of the batch difference within the segment. A smaller μ_c indicates a more frequent power outage event within the constraint segment, while a larger σ_c indicates a worse regularity of the power outage interval. μ_c and σ_c constitute the candidate parameter pair for that segment. The threshold candidate parameter set is archived by segment number, with parameter pairs for each circuit and constraint segment. Seasonal shutdowns of large users within the power supply range may cause a sudden increase in intervals and a systematically high μ_c in the threshold constraint interval during shutdown batches. The μ_c of these shutdown batches is not included in the baseline update of normal constraint segments to isolate non-equipment factors. For constraint segments within the threshold constraint interval with dual sources during the flood season, μ_c is adjusted and substituted using the historical correction factor for the flood season. The correction factor is determined based on the ratio of the mean intervals during the flood season to the mean intervals during the non-flood season for similar enclosures. The adjusted parameters are retained with the flood season correction mark and the original parameters as auxiliary parameters. When a seasonal parameter in a threshold constraint interval is estimated using samples from adjacent seasons due to insufficient effective samples in the current season, it is marked as sample borrowing. The sample-borrowed parameters in the threshold candidate parameter set are corrected for trend matching using features from the borrowed season. During segmented analysis, constraint segments with boundary contraction sources are extended to both sides with several batches of re-estimation of μ_c and σ_c using an expanded tolerance. Segments where the difference between the expanded result and the original segment exceeds the threshold are replaced by the expanded result. When the difference in μ_c between two adjacent constraint segments exceeds the step threshold, they are not merged but remain independent segments. The step is marked as an inter-segment step. During trend matching, the shortening trend is matched separately on both sides of the step without merging across steps.

[0039] A threshold correction distribution is formed by matching the threshold candidate parameter set with the power outage interval fluctuation distribution to form a successive shortening trend. Based on batch segments where the interval difference between adjacent power outage events is continuously negative in the power outage interval fluctuation distribution, a successive shortening trend is identified. Batch segments with consecutive negative differences exceeding the trend confirmation window are considered shortening trend segments. The mean power outage interval μ_c of each circuit's seasonal parameter pair in the threshold candidate parameter set is corrected using the shortening rate and denoted as μ_s; the corresponding standard deviation σ_c is simultaneously corrected and denoted as σ_s. The corrected parameter pairs are written into the threshold correction distribution. During regional capacity expansion, the addition of large-scale electrical equipment will cause a continuous shortening segment in the power outage interval fluctuation distribution, and the corresponding circuit μ_s will be tightened. The correction results of the expansion batches are given priority in valid threshold verification, and the systemic shortening caused by capacity expansion is distinguished from equipment degradation in terms of source. For seasonal parameter pairs with borrowed samples in the threshold candidate parameter set, the shortening rate of the borrowed season is used instead of the current season rate during trend matching, and the corresponding batch is marked as the borrowed rate, which has a lower confidence level than the normal seasonal estimate. In the power outage interval fluctuation distribution, loops with sources of increased fluctuation are identified by shortening trends and matched individually in batches of accelerated segments. The corresponding correction results are marked as accelerated corrections, and stricter tightening limits are applied during effective threshold verification. In the threshold correction distribution, when multiple loops shorten simultaneously in similar batches, the correction magnitude is adjusted based on upstream common cause analysis results. Synchronous shortening of multiple loops corresponds to common external load pressure rather than independent equipment degradation. The shortening rate is estimated by the regression slope of the difference sequence between adjacent intervals in the power outage interval fluctuation distribution. When the slope is negative and significant, it is linearly tightened to the mean power outage interval μ_c according to its absolute value to obtain μ_s. The tightening ratio is proportional to the absolute value of the slope. For segments with insufficient slope significance, μ_c remains unchanged and is included in the threshold correction distribution.

[0040] The effective threshold interval is determined by screening effective threshold intervals using a threshold correction distribution. The effective threshold interval is defined by the lower bound of the correction parameter pair in the threshold correction distribution: μ_s - z × σ_s. μ_s is the mean of the corrected seasonal power outage interval for each circuit, σ_s is the standard deviation of the corrected seasonal power outage interval, and z is the confidence coefficient determined based on historical power outage warning accuracy. This lower bound represents the statistically reliable shortest normal power outage interval at a given confidence level. Intervals shorter than this value are considered abnormally dense power outages. A larger z value results in a looser lower bound, reducing missed alarms but increasing false alarms. Therefore, z is calibrated using the value at the highest warning accuracy. A smaller lower bound value indicates a tighter minimum allowable power outage interval for the circuit and a more sensitive warning trigger. A valid interval is defined as the number of consecutive stable batches exceeding the effective window. The power outage threshold is determined by the mean of the lower bounds of the effective intervals for each circuit. After the planned maintenance window is completed, the power supply quality of the circuit will briefly recover. The threshold correction distribution will revert to its trend after a brief increase in the correction parameter in the post-maintenance batch. The correction parameter of the post-maintenance batch will not be included in the stability judgment of the effective interval to isolate the transient effect of maintenance. For parameters with accelerated correction sources in the threshold correction distribution, the lower bound is tightened to the upper quartile of the accelerated correction amplitude. Those exceeding the limit are truncated at the upper limit, and the corresponding circuits are marked as accelerated truncated. Circuits where accelerated truncation and increased volatility occur simultaneously are classified as dual risks. When generating node inspection rules, the highest inspection density level is prioritized for dual-risk circuits. The length of the effective window is adaptively adjusted according to the seasonal sample density of the circuit. For circuits with sparse samples, the window is appropriately widened to ensure the statistical robustness of the lower bound estimate. Buffer batches are reserved at both ends of the window to mitigate the estimate jump at the boundary.

[0041] In some embodiments, establishing node inspection rules based on the power outage threshold and the power supply assessment segment by performing power outage propagation priority weight allocation includes: using the power outage threshold to perform power outage proximity period statistics for each node in the power supply assessment segment to generate a cumulative risk value sequence; performing a correspondence conversion between the cumulative risk value sequence and the power supply assessment segment based on the duration of no warning to determine the inspection density weight coefficient; standardizing and calibrating the inspection density weight coefficient to form a dynamic inspection density distribution; and establishing node inspection rules based on the dynamic inspection density distribution by performing density-level time-series node allocation.

[0042] Using power outage thresholds, a cumulative risk value sequence is generated for each node within the power supply assessment section by performing statistical analysis of the period before a power outage without warning. Historical batches where the power supply level score of each node falls below the power outage threshold are recorded as power outage events. A few batches preceding the start time of each power outage event are taken as a defining window. Within this window, batches for which no warnings are reported are categorized as periods before a warning. The cumulative number of batches in these periods constitutes the cumulative risk value for that node. The cumulative risk values ​​of each node are then entered into a cumulative risk value sequence in chronological order. In areas where power distribution facilities are lagging behind in upgrades, nodes experience continuous insulation aging and low on-site maintenance frequency. These nodes often have no warnings reported for extended periods before a power outage, and the power outage threshold is triggered frequently. Nodes whose consecutive periods before a warning exceed the batch threshold are marked as continuously without warnings, and these nodes are prioritized for triggering the highest density level. In the power supply assessment section, batches of load interference sources are excluded when there is no early warning in the immediate vicinity. After exclusion, the cumulative risk value of the remaining batches within the window that are still judged to be without early warning increases normally, and the exclusion operation is marked as load exclusion. If the load exclusion ratio is too high, it indicates that the early warning statistics of this node are greatly affected by load interference and the reliability of the cumulative risk value sequence is limited. When determining the inspection density weighting coefficient, a conservative estimate is used. The number of batches before the window is defined by the set ratio of the average power outage interval of each circuit. Batches in the power supply assessment section within the window that have no early warning reports are counted as early warning-free batches, and their cumulative number is the cumulative risk value. The circuit window with dual risk sources is expanded proportionally according to its risk level to cover a longer early period to include earlier early warning exposures.

[0043] The cumulative risk value sequence and power supply assessment section are converted to correspondence of no-warning duration to determine the inspection density weighting coefficient. No-warning duration is calculated by multiplying the cumulative risk value of each node in the cumulative risk value sequence by the duration of a single batch. The ratio of the converted duration to the average power outage interval within the power supply assessment section quantifies the degree of no-warning exposure for that node. A higher ratio indicates a larger proportion of no-warning duration in the power outage interval. This ratio is normalized and aggregated by node to form the inspection density weighting coefficient. When the number of inspection personnel in the jurisdiction is limited and high-risk nodes are concentrated in the same section, inspection resources are insufficient for full coverage. Multiple nodes in the cumulative risk value sequence of that section have simultaneously high ratios. Nodes with concurrent high ratios in the same section are marked as resource competition, and their density is capped by the overall resource constraints of the section. A high ratio for a single node will not exceed the section capacity. Nodes in the cumulative risk value sequence with a high proportion of load removal are calculated using a conservative lower limit value and marked as low-confidence weights. Their density can only be increased after several consecutive batches of ratios have stabilized; a sudden increase in the ratio of a single batch will not be immediately adjusted. For batches with accelerated degradation sources in the power supply assessment section, the duration without warning is adjusted using an acceleration coefficient. The adjusted ratio is marked as the accelerated correction. The inspection density weighting coefficient is tightened during standardization, shifting its relative position forward in the normalization space. The ratio of no-warning exposure is truncated before normalization. Extreme ratios exceeding the upper quantile are capped at the upper quantile value. Individual nodes with excessively long periods without warning do not exclusively occupy high-density slots but instead encroach on the inspection resources that other nodes in the same section should receive. The capping threshold is dynamically adjusted according to the total capacity of the section.

[0044] A dynamic inspection density distribution is formed through standardized calibration using inspection density weighting coefficients. Standardization maps the quantiles of the network-wide inspection density weighting coefficients, sorted by node, to density levels. Nodes with weighting coefficients in the top quartile are mapped to the highest density level, and those in the bottom quartile to the lowest density level. Level boundaries are pre-set based on the total inspection resources and allocable inspection frequencies. Within each level, the product of the inspection density weighting coefficient and the standard inspection interval constitutes the node's dynamic inspection interval; shorter intervals indicate higher inspection frequencies. The dynamic inspection intervals of each node and its corresponding level form a dynamic inspection density distribution based on node inclusion. In power supply areas where new and old communities coexist, equipment aging varies significantly, and the inspection density weighting coefficients are extremely unevenly distributed between new and old nodes. Direct quantile mapping would cause older nodes to accumulate high-density levels, leading to single-level overload. Therefore, nodes with service life exceeding a threshold are corrected using a weighted average based on their service life. Older nodes are weighted towards higher density levels with the same weight. The age difference of equipment is not smoothed out by quantile mapping, resulting in relatively insufficient density for older nodes. For node groups with resource competition sources in the inspection density weighting coefficient, the labeling is constrained by the upper limit of segment resources. Nodes within the group are sorted in descending order of weight and then tiered accordingly. Nodes exceeding the upper limit are forcibly downgraded and marked as downgraded. Downgraded nodes are prioritized for verification of whether they meet the recovery conditions in the next batch of updates. For nodes with accelerated correction sources in the inspection density weighting coefficient, the normalization range is tightened, and the dynamic inspection density distribution moves them forward, marking them as accelerated forward moves to give them priority in the node inspection rule generation. Nodes with low-confidence weight sources maintain their current tier and their affiliation is updated only after the confidence level is restored.

[0045] Based on the dynamic inspection density distribution, a density-level time-series node allocation is performed to establish node inspection rules. Density-level grouping categorizes nodes in the dynamic inspection density distribution by their respective tiers. Within the same tier, nodes are sorted from shortest to longest dynamic inspection interval. Inspection time sequences are allocated node-by-node on a calendar timeline with the dynamic inspection interval as the step size. The interval between inspection times of adjacent nodes is no less than the minimum interval to ensure operational feasibility. The inspection time sequence of each node and its corresponding tier are archived to form node inspection rules. During the year-end concentrated defect elimination period, available inspection manpower is temporarily reduced, and the number of nodes in high-density tiers of the dynamic inspection density distribution may exceed the available manpower coverage. Batches during this period are marked as manpower-constrained, and high-density tier nodes are forcibly postponed to the next available time-series window. Nodes postponed are given priority to restore their original tier time-series in the next batch update. Nodes in the dynamic inspection density distribution with a tier downgrade origin are included in the downgraded time-series at the downgraded interval. Batches meeting the recovery conditions are automatically upgraded and their subsequent time-series adjusted. In the dynamic inspection density distribution, nodes with accelerated forward movement sources are prioritized in time allocation over other nodes in the same gear. The initial inspection time is tightened at the standard gear interval with an acceleration coefficient. In the node inspection rules, when the inspection interval between adjacent time-series nodes is continuously compressed within consecutive batches, it is marked as a risk concentration, and the resource allocation priority of the corresponding section is simultaneously increased in subsequent batches. The shortest inspection distance between adjacent nodes is calculated based on the average working hours and round-trip distance of a single on-site operation. The shortest distance is correspondingly increased for sections with remote distances to ensure that the schedule can be implemented under the constraints of manpower and vehicles.

[0046] Step S104: Based on the node inspection rules, the circuit is divided into a power supply assessment sequence and an emergency power restoration sequence. The power restoration success rate attenuation trend assessment is performed on the emergency power restoration sequence to generate a power restoration priority sequence. The power supply assessment sequence is subjected to level stability detection to obtain the power supply deviation.

[0047] Specifically, based on the node inspection rules, circuits are categorized into power supply assessment sequences and emergency power restoration sequences. The categorization is performed according to the grade level of each circuit in the node inspection rules. High-density grade circuits are assigned to the emergency power restoration sequence, while low-density grade circuits are assigned to the power supply assessment sequence. The grade boundaries are determined by comparing the cumulative risk value sequence of each circuit with a set segmentation threshold. During regional power grid maintenance, multiple circuits simultaneously enter a high-risk state. The number of high-density grade circuits in the node inspection rules may surge briefly during maintenance batches. If all of them were assigned to the emergency power restoration sequence, it would exceed the emergency response resource limit. Therefore, circuits exceeding the limit are marked as resource over-limit and placed in a waiting queue according to their priority. Resource over-limit circuits are not subject to a reduced priority assessment weight for subsequent power restoration due to their lower ranking. For circuits with downgrade sources in the node inspection rules, the downgraded grade is assigned to the corresponding sequence during segmentation. If the downgraded grade falls close to the segmentation threshold, the power supply assessment sequence classifies it as critical segmentation. Critical segmentation circuits are subject to stricter deviation thresholds during stability level testing. When the node inspection rules are refreshed in batches, the circuits on both sides of the critical level may migrate in sequence. The migrated batches are marked as dynamic migrations. For the power supply assessment sequence and the emergency power restoration sequence, the migration time is used as the statistical starting point for each dynamically migrated circuit, and the history before migration does not continue across sequences. Circuits with accelerated forward migration sources in the node inspection rules are given priority to be included in the emergency power restoration sequence and are not subject to the current level boundary constraints, indicating that the urgency of their power restoration has exceeded the risk level reflected by the normal level. A buffer zone is set at the boundary between the two levels for the split threshold. Circuits falling into the buffer zone are assigned to the sequence only after several consecutive batches have the same level, avoiding repeated fluctuations in level at the boundary that could cause frequent sequence migrations.

[0048] In some embodiments, the step of assessing the power restoration success rate decay trend of the emergency power restoration sequence to generate a power restoration priority sequence includes: calculating the transfer adaptability of each power failure circuit based on the historical power restoration parameters of the emergency power restoration sequence to form a transfer adaptability sequence; identifying the accelerated decay segment after a brief rebound in the transfer adaptability sequence to obtain a power restoration risk score; comprehensively ranking the power restoration risk score and the emergency power restoration sequence based on transfer reliability to obtain a circuit power restoration priority table; and selecting effective power restoration circuits through the circuit power restoration priority table to generate a power restoration priority sequence.

[0049] Based on the emergency power restoration sequence, the transfer adaptability is calculated using the historical power restoration parameters of each power-loss circuit to form a transfer adaptability sequence. The historical power restoration parameters for each power-loss circuit are taken from the transfer logs archived in the operation and maintenance management system. The adaptability is calculated as A_r = N_succ / N_total, where N_succ is the number of successful power transfers in historical power restoration events, N_total is the total number of power restoration attempts, and A_r is a dimensionless ratio ranging from 0 to 1. A_r equal to 1 indicates that all historical power transfers were successful, and A_r approaching 0 indicates that almost all historical power transfers failed. The transfer adaptability sequence is archived chronologically by circuit number, using the A_r of each circuit in the emergency power restoration sequence and the corresponding historical sample size. In densely populated residential areas, power transfer operations are frequent during concentrated power outages in the plum rain season. Emergency power restoration sequences accumulate a large amount of historical power restoration parameters during this period. However, load fluctuations and line dampness jointly lower the success rate of power transfer, resulting in a low A_r (equipment performance rate) that does not reflect normal equipment conditions. Therefore, the A_r for this period is marked for its flood season operating conditions and is not included in the normal adaptability baseline update. For circuits in the emergency power restoration sequence with insufficient historical power restoration parameter sample size to the minimum threshold, the A_r estimation error is too large, and they are marked as low samples. Their A_r is replaced with the historical average of the same type of enclosure across the entire network as a conservative baseline, and the weight is reduced during attenuation segment identification after replacement. When a circuit in the emergency power restoration sequence attempts power transfer multiple times within the same batch, N_total is counted according to the actual number of attempts. Even if multiple attempts are successful, it is still counted as one success. Success after multiple attempts is not overestimated as high adaptability. In the power transfer adaptability sequence, circuits with continuously decreasing A_r between batches are marked as adaptability attenuation. Under attenuation background, the acceleration segment determination requires a lower slope change threshold.

[0050] A power restoration risk score is obtained by identifying accelerated attenuation segments after a brief rebound in the power supply adaptability sequence. Accelerated attenuation segments are identified by the sequence where A_r continuously decreases for several batches, then experiences a brief positive rebound, and the absolute value of the decline slope after the rebound exceeds the absolute value of the decline slope before the rebound. The slope is estimated using the mean difference of a sliding window. Batch segments meeting both conditions are considered candidates for accelerated attenuation. The power restoration risk score is E_raw = γ_1 × R_slope + γ_2 × T_dur, where R_slope is the ratio of the absolute values ​​of the slopes before and after the rebound, T_dur is the number of batches of accelerated attenuation, and γ_1 and γ_2 are weights. Since T_dur has batch dimensions while R_slope is dimensionless, both are normalized to the 0-1 range before being weighted. A higher E_raw indicates more severe and longer-lasting attenuation. The power restoration risk score is obtained by archiving the E_raw values ​​of each circuit. After a brief cooling period following a high temperature, the insulation performance of the batch briefly rebounds. The supply adaptability sequence exhibits a brief rebound in A_r followed by a rapid decline, similar to accelerated decay. The power restoration risk score uses temperature correlation verification to distinguish between genuine equipment rebound and pseudo-rebound caused by temperature effects. Batches with significant temperature correlation rebounds do not trigger accelerated decay. For batches in the supply adaptability sequence originating from flood season conditions, the slope is reassessed after removing flood season batches during accelerated decay identification. Accelerated decay is retained only if both conditions are met after removal. For circuits in the supply adaptability sequence originating from low-sample sources, the batch variance of A_r is large, and the brief rebound may be caused by sampling errors rather than genuine improvement. The power restoration risk score increases the requirement for sustained rebound batches of low-sample circuits to a normal multiple, ensuring sampling noise does not trigger misjudgments. For circuits originating from adaptability decay sources, the accelerated decay identification threshold is tightened proportionally to the number of batches with sustained decay. The power restoration risk score increases the weight of accelerated segments under long-term decay backgrounds by the magnitude of slope abrupt changes. The slope abrupt change of the accelerated decay candidate is measured by the ratio of the slopes before and after the rebound. A larger ratio indicates a more severe acceleration of decay. The ratio and the number of continuous batches together determine the magnitude of the candidate in the power restoration risk score.

[0051] A circuit restoration priority table is obtained by comprehensively ranking the power supply reliability based on the power restoration risk score and the emergency power restoration sequence. The comprehensive score S = α × (1 - E_r) + β × A_r, where E_r is the dimensionless value of the power restoration risk score E_raw after normalization to its maximum and minimum values, A_r is the corresponding circuit's power supply adaptability, and α and β are weighting coefficients with α + β = 1. The score is jointly calibrated based on the prediction accuracy of the two types of indicators for power restoration results in historical power restoration events. A higher S indicates stronger current power supply reliability for the circuit. The circuit restoration priority table is archived with each circuit arranged in descending order of S. Multiple important production lines experience significant hourly losses after power outages, resulting in higher weightings for these circuit loads in the emergency power restoration sequence. The circuit restoration priority table prioritizes high-importance load circuits under the same S value. Importance is derived from user classification information entered into the operation and maintenance system. If a user classification has not been updated within a set period, it is reduced to a general weight, and the ranking is not increased based on expired classifications. In the power restoration risk assessment, for batches experiencing a false rebound due to temperature effects, E_r is substituted with the temperature-corrected value. The false rebound does not lower E_r but overestimates the true reliability of the circuit. For circuits with resource overruns in the emergency power restoration sequence, S is calculated normally, but they are listed separately in the waiting queue in the circuit power restoration priority table, not mixed with normally ordered circuits. After resource release, they are added according to S order. When multiple circuits have similar S values, the circuit power restoration priority table uses the absolute value of the power restoration risk score as the secondary ranking basis. Under the same S value, those with lower risk are prioritized to ensure the overall power restoration success rate. The weighting coefficients α and β are recalibrated in each assessment period using backtesting of historical power restoration results. Their weights are adjusted when the marginal contribution of the two types of indicators to the success rate shifts, ensuring that the comprehensive score continuously reflects the current power transfer situation.

[0052] A power restoration priority sequence is generated by filtering valid power restoration circuits through a power restoration priority table. A dual admission criterion is that the S value in the power restoration priority table exceeds the effective lower limit threshold, and the historical power restoration parameter sample size reaches the minimum confidence batch. A circuit is considered valid only if both conditions are met simultaneously. The power restoration priority sequence is entered into the database in descending order of S value, with each circuit recorded by its recognized batch number at the time of aggregation. After substation maintenance, frequent power transfers within several batches may cause some circuits in the power restoration priority table to have artificially high S values ​​due to a temporary increase in adaptability after maintenance. The power restoration priority sequence uses an additional stable and continuous batch constraint to ensure the validity of batches after maintenance. If the number of batches with continuously exceeding the threshold after maintenance is insufficient to reach a stable window, the circuit is temporarily not written in until S value stabilizes. For circuits in the power restoration priority table with sources from the waiting queue, when resources are released and added to the main list, it is re-verified whether S value still exceeds the effective lower limit. Circuits with S value falling below the threshold during the waiting period are removed from the waiting queue and not written into the power restoration priority sequence. For circuits with low sample sources in the circuit restoration priority table, borrowed batches are included in the valid samples when determining the sample size. If the borrowing ratio is too high, the confidence condition is tightened and marked as limited confidence. For circuits in the restoration priority sequence where the S batches continuously decrease, a reliability decay warning is issued. During the accelerated degradation phase, when transferring and matching, priority is given to verifying whether they still meet the effective lower limit; if not, they are automatically removed and re-evaluated. The minimum confidence batch threshold is set according to the completeness of the circuit's transfer ledger. Circuits with more missing ledgers have a correspondingly higher threshold, and circuits with low sample quality are not mistakenly included in the valid set based on a few inflated scores.

[0053] Power supply deviation is obtained by performing graded stability testing on the power supply assessment sequence. Grade stability is estimated using the batch-to-batch power supply grade standard deviation σ_q = std(g_1,g_2,…,g_n), where g_i is the power supply grade score for each batch and n is the total number of batches within the statistical window. Circuits whose σ_q continuously exceeds the stability upper limit are judged as unstable. The power supply deviation is quantified by the difference between the mean σ_q of unstable circuits and the historical stable baseline. Aging of distribution line insulation causes frequent fluctuations in grounding resistance between batches. Circuits in the power supply assessment sequence with significant jumps in power supply grade between adjacent batches and consecutive σ_q exceeding limits in multiple batches are marked as continuously exceeding limits. Continuous exceeding limits triggers an accelerated phase for independent estimation; the more batches of continuously exceeding limits, the deeper the line degradation. For circuits in the power supply assessment sequence with critical split sources, the σ_q judgment threshold is tightened to a set proportion of the normal threshold. Power supply deviations that still exceed limits after tightening are recorded as critical exceeding limits, and their deviation growth rate is estimated using a conservative growth rate during deviation growth analysis. Seasonal load variations can cause the power supply assessment sequence to be systematically lower and σ_q higher in batches during peak seasons. The residuals after subtracting the historical average for the same season from each batch's level are recalculated. The difference between σ_q before and after correction quantifies the impact of seasonal effects on stability. Circuits that still exceed limits after correction are recorded as non-seasonal deviations, and these non-seasonal deviations trigger higher degradation weights. For batches in the power supply assessment sequence with accelerated degradation sources, σ_q is processed in an independent window for the accelerated segment. The power supply deviations for the accelerated and normal segments are output separately. The extent to which the accelerated segment deviation exceeds the normal value directly quantifies the additional contribution of accelerated degradation to the overall deviation.

[0054] Step S105: Analyze the deviation growth rate based on the power supply deviation to determine the power supply degradation coefficient. Match the power supply degradation coefficient with the power restoration priority sequence to determine the power supply recovery path during the degradation acceleration phase. Perform maximum degradation rate breakpoint connection repair on the power supply recovery path and the power supply assessment sequence to form a power supply collaborative recovery network. Based on the power supply collaborative recovery network and the enclosure environment data, perform a protection status assessment and output the dynamic assessment result of the protection level.

[0055] Specifically, the power supply degradation coefficient is determined by analyzing the rate of increase of power supply deviation. The rate of increase of deviation is calculated sequentially based on the difference between adjacent batches of power supply deviation, ΔD = D_b - D_{b-1}, where D_b and D_{b-1} are the power supply deviations of the current batch and the previous batch, respectively. Batches with consistently positive ΔD are classified as deviation expansion stages, and the power supply degradation coefficient quantifies the current rate of degradation progression using the average ΔD within the expansion stage. During centralized meter replacement, the replacement of metering devices causes a systematic jump in the power supply deviation in the replaced batches. This jump originates from the switching of the metering baseline rather than actual degradation. The ΔD of the baseline-switched batches is not included in the deviation rate of increase statistics to isolate changes in the metering baseline. For circuits with non-seasonal deviation sources in the power supply deviation, the corrected deviation is substituted into ΔD. The difference between ΔD before and after correction quantifies the seasonal effect and eliminates its actual impact on the rate of increase estimation. Batches with consistently positive ΔD after correction are classified as confirmed after correction, and their rate of increase reliability is higher than that of uncorrected batches. For batches with accelerated deviation sources in the power supply deviation, ΔD is estimated using an independent window for the accelerated segment. The power supply degradation coefficient of the accelerated segment and the normal segment are output separately. A larger ratio between the two indicates significant acceleration. If the ratio exceeds a set multiple when the insulation enters the rapid degradation period before breakdown, it is marked as accelerated dominance. During power supply matching, a stricter power supply deviation tolerance is applied to the accelerated dominance circuit. For circuits with continuous over-limit sources, the growth rate estimation window is narrowed to the shortest continuous batch of the accelerated segment. The difference between the mean and the normal window after narrowing is used to quantify the impact of the window selection on the accuracy of the power supply degradation coefficient estimation. The start and end of the deviation expansion stage are defined by the zero-crossing point where ΔD changes from negative to positive and from positive to negative. A dead zone filter is set near the zero-crossing point. Batches with an absolute value of ΔD lower than the dead zone width do not trigger stage segmentation.

[0056] In some embodiments, the step of matching the power supply degradation coefficient with the power restoration priority sequence during the degradation acceleration phase to determine the power supply recovery path includes: parsing the single degradation value and the cross-circuit cumulative offset based on the power supply degradation coefficient; filtering the power restoration priority sequence according to the single degradation value to form a low-degradation circuit set; evaluating the cross-circuit transfer coverage rate based on the low-degradation circuit set and the cross-circuit cumulative offset to determine the transfer priority sequence; and sorting and integrating the low-degradation circuit set step by step using the transfer priority sequence to determine the power supply recovery path.

[0057] The single degradation value and cross-circuit cumulative offset are analyzed based on the power supply degradation coefficient. The average value of the deviation increment of a single batch within the accelerated phase of the power supply degradation coefficient is defined as the single degradation value, quantifying the single-step advancement of the power supply deviation of each batch of equipment. The historical deviation increments of each expansion phase are successively accumulated and summed to form the cross-circuit cumulative offset. Both constitute the core parameters for subsequent transfer tolerance screening. When low temperature causes increased condensation inside the enclosure, the power supply degradation coefficient is temporarily higher in the low-temperature batches, and the corresponding single degradation value is briefly artificially high before recovering as the temperature rises. The single degradation value of the low-temperature batches is adjusted with a seasonal correction factor. The correction factor is determined based on the historical low-temperature batch offset average of the same type of equipment and is marked as low-temperature correction after adjustment. It is not included in the slope estimation of the normal accelerated phase. For the batches from which the baseline is switched in the power supply degradation coefficient, the switched batches are excluded when calculating the single degradation value and cross-circuit cumulative offset. They are accumulated independently before and after the switch. The cross-circuit cumulative offset is based on the batch after the switch, and the historical data before the switch is retained as a reference. When the accelerated and normal phases alternate, the single degradation value is represented by the average value of the most recent accelerated phase and the average value of the historical normal phase as the baseline. A larger ratio between the two indicates a significant acceleration. When the cumulative offset across circuits accounts for a high proportion of the accelerated phase, it is marked as accelerated dominance. When it exceeds the 90ths of the historical values ​​for the entire network, it is marked as a high accumulation warning. When the number of equipment batches put into operation is less than the minimum sample size, the single degradation value is replaced by the average value of the same type of equipment in the same period across the entire network. The single degradation value is taken as the arithmetic mean of the deviation increment of the power supply degradation coefficient in the accelerated phase, reflecting the instantaneous rate. The cumulative offset across circuits is taken as the successive sum of the historical deviation increments in each expansion phase, reflecting historical accumulation. The former is substituted into the current tolerance threshold for comparison, and the latter is substituted into V_c for coverage scoring.

[0058] The power restoration priority sequence is filtered based on single-instance degradation values ​​to form a low-degradation circuit set. Each circuit in the power restoration priority sequence has a tolerance threshold based on the maximum acceptable single-instance degradation value corresponding to its overall score; circuits with higher scores have lower thresholds. Circuits in the current batch whose single-instance degradation values ​​do not exceed the corresponding thresholds are included in the low-degradation circuit set and are listed jointly based on their score and degradation value. During emergency power restoration after a typhoon, it is necessary to quickly identify transferable circuits. Some circuits in the power restoration priority sequence are ranked high but have high single-instance degradation values. The low-degradation circuit set applies a relaxed tolerance coefficient to emergency mode batches, expanding the relaxed threshold to a normal multiple. Emergency mode circuits are treated with reduced reliability tolerance during coverage assessment. Circuits in the power restoration priority sequence with confidence-limited sources have larger single-instance degradation value estimation errors. The low-degradation circuit set uses historical quartiles as a conservative upper limit for their screening, and the weight of conservative circuits is reduced during coverage assessment. When the low-deterioration circuit set is empty, it indicates that all candidate circuits have exceeded the limit for a single degradation value. The screening is automatically relaxed to prioritize the circuits with the lowest degradation values ​​among all circuits and mark them as degraded circuits. If multiple batches trigger degraded circuit selection, they are marked as continuous degraded circuits with warnings. For circuits in the power restoration priority sequence with large fluctuations in single degradation values ​​between batches, the low-deterioration circuit set requires that several consecutive batches meet the standard to be considered valid. A single batch that occasionally meets the standard is not archived. The slope of the tolerance threshold, which decreases monotonically with the comprehensive score, is calibrated according to the fitting relationship between degradation value and failure rate in historical failure cases, so that high-reliability circuits bear stricter degradation constraints and low-reliability circuits are given appropriate leniency.

[0059] For example, the step of evaluating the cross-loop power transfer coverage rate and determining the power transfer priority sequence based on the low-deterioration circuit set and the cross-loop cumulative offset includes: generating a deviation coverage score sequence by quantifying the deviation based on the low-deterioration circuit set and the cross-loop cumulative offset; determining a stability correction coefficient by statistically analyzing the cross-loop adaptation score fluctuation of each power failure circuit based on the deviation coverage score sequence; applying a power transfer failure segment priority weight reduction correction to the deviation coverage score sequence using the stability correction coefficient to form a power transfer level distribution; and determining the power transfer priority sequence based on the distribution concentration intensity of the power transfer level distribution.

[0060] A deviation coverage scoring sequence is generated by quantifying the deviation based on the set of low-deterioration circuits and the cumulative offset across circuits. The transfer coverage rate C_r is calculated locally as the proportion of the lost load that can be connected by the carrying capacity of each candidate circuit in the low-deterioration circuit set to the total load of that lost circuit. The carrying capacity of the candidate circuit is taken from its circuit ledger rated value. C_r equal to 1 indicates that it can be fully connected, and less than 1 indicates that it can only be partially connected. The joint score Q of the coverage rate of each scheme in the low-deterioration circuit set and the normalized value of the cumulative offset across circuits is Q=C_r×(1-V_c / V_max), where V_c is the cumulative offset across circuits of the corresponding circuit, and V_max is the maximum historical cumulative offset of the entire network. A high Q indicates that the scheme has sufficient coverage and a small historical accumulation of deterioration. The deviation coverage scoring sequence is archived according to the circuit combination time sequence based on Q of each scheme. After the delivery and occupancy of newly built residential areas, the load increases rapidly. The V_c of the low-deterioration circuit set is higher due to the rapid accumulation of recent batches, while the Q is systematically lower. For newly added load batches, V_c is adjusted using a load growth correction factor to reflect the true deterioration level after construction. For schemes involving circuits from the low-deterioration circuit set that have been selected for downgrade, Q is substituted with the downgraded capacity and marked as downgraded; its weight is reduced in the stability correction factor calculation. In the deviation coverage scoring sequence, for schemes where multiple circuits jointly bear the same power outage load, Q is calculated for each participating circuit, and the capacity-weighted average is taken. Circuits with a weighted Q lower than the lowest Q of a single circuit are marked as joint weak points, and their weights are further reduced. When V_max is updated due to historical extreme values ​​in new batches, Q for archived batches in the deviation coverage scoring sequence is recalculated using the new V_max.

[0061] For deviation coverage scoring sequences, stability correction coefficients are determined by statistically analyzing the cross-circuit adaptation score fluctuations of each power outage circuit. The batch-to-batch fluctuation of scheme Q in the deviation coverage scoring sequence is captured by the standard deviation σ_Q. A larger σ_Q indicates weaker score stability. The stability correction coefficient is normalized using the reciprocal of σ_Q and archived by scheme number. A higher normalized value indicates stronger cross-batch stability for that scheme. During the construction phase, power load fluctuations are frequent, and the deviation coverage scoring sequence shows significantly larger fluctuations in Q and higher σ_Q in construction batches. Construction interference is identified for these batches, and their σ_Q is not included in the baseline update for normal fluctuation amplitude or in the reduction benchmark. Schemes in the deviation coverage scoring sequence with sources of downgrade reduction have already had their σ_Q further increased due to coverage downgrade. The stability correction coefficient re-evaluates σ_Q for these schemes after excluding the downgrade effect, quantifying the actual impact of the downgrade operation on the stability assessment by eliminating the difference before and after the downgrade. For schemes with joint short-board sources, σ_Q is increased by the short-board loop. The stability correction coefficient is re-estimated using the Q sequence after removing the short-board loop. Schemes with a difference exceeding the threshold before and after removal are marked as short-board drags. When generating the supply level distribution, the weight reduction of schemes with short-board drags is increased. When the number of historical participation batches of a certain scheme is less than the minimum estimated sample size, σ_Q is replaced by the fluctuation range of similar coverage schemes in the same period of the entire network. The confidence of the corresponding stability correction coefficient is limited, and a conservative value is used when the weight is reduced. The statistical window of σ_Q slides synchronously with the archived batches of the deviation coverage score sequence. When there are insufficient effective samples in the window, it is postponed until the sample accumulation reaches the standard before output. The high variance of the standard deviation estimate under the short window will not misjudge the stability of the scheme.

[0062] A transfer-to-supply grade distribution is formed by reducing the priority weight of the transfer failure segment in the deviation coverage scoring sequence using a stability correction coefficient. The initial weight of each scheme is taken as its score Q in the deviation coverage scoring sequence. The reduction magnitude of the transfer-to-supply grade distribution's weight is jointly determined by the stability correction coefficient and the deviation degree of the failure segment: the deviation ratio is calculated by dividing the difference between the current Q and the mean Q within the failure segment by the standard deviation within the segment. The transfer-to-supply grade distribution reduces the initial weight of each scheme by a reduction coefficient = deviation ratio × (1 - stability correction coefficient). After reduction, the weight is not lower than the set lower limit of the initial weight to retain basic participation eligibility. The corrected weight of each scheme is written into the transfer-to-supply grade distribution according to the scheme number for subsequent weight-based grading. The transfer failure segment is defined by batches in the deviation coverage scoring sequence where each scheme's Q is continuously below the failure threshold. Within each segment, the cumulative reduction amount and cumulative value of the scheme increase with the number of consecutive batches in the failure segment, causing a deeper weight reduction for schemes that have been in the failure segment for a long time. Batch failures caused by equipment annual inspections, construction interference, etc., may result in a brief, sharp drop in Q followed by a rapid rebound, resembling a true failure pattern. These batches are excluded from the failure history based on their source markers. After this exclusion, batches with insufficient confirmation windows are not considered valid failure segments. For schemes with sources dragging down by short circuits, a short circuit penalty coefficient is added to the basic decrease in failure rate. This penalty coefficient is calculated by normalizing the historical failure frequency of the short circuit loop. The final weight of each scheme after the failure segment decrease and short circuit penalty is written into the transfer level distribution according to the scheme number. Schemes with repeated failure segments within consecutive batches and continuously narrowing intervals are directly placed in the lowest weight tier. A brief rebound in Q in a single batch does not trigger a re-ranking. The transfer level distribution outputs the tiered weights of each scheme based on this for subsequent concentrated intensity statistics.

[0063] The priority sequence for power transfer is determined based on the concentration intensity of the power transfer level distribution. The concentration intensity is the proportion of the top quartile schemes in the power transfer level distribution to all schemes. A high proportion indicates that high-weight schemes are concentrated in a few circuit combinations, resulting in a high dependence on power transfer resources; a low proportion indicates that available schemes are dispersed, offering greater flexibility in power transfer. The power transfer priority sequence is arranged in tiers based on the concentration intensity of high-weight schemes, with the tier boundaries determined by the quartiles of the weight distribution of historical successful power restoration events. When multiple branch boxes lose power simultaneously, high-weight schemes in the power transfer level distribution are concentrated on a few main lines, indicating extremely high concentration intensity. This suggests excessive pressure on a single point on the main lines. When generating power restoration paths, maximum capacity constraints are used to limit the scale of single-batch acceptance to avoid overloading the main lines and causing cascading power outages. When schemes with cumulative penalty sources in the power transfer level distribution are concentrated in the high-weight range, a more conservative coverage threshold is used when generating power restoration paths. When the weight distribution of power transfer levels is extremely dispersed, the concentration intensity is low. The power transfer priority sequence extends to both sides based on the median weight of all schemes. When generating power restoration paths, this type of batch is given priority for manual verification. When the concentration intensity of the power transfer level distribution alternates rapidly between consecutive batches, the power transfer priority sequence continues the previous stable batch's priority for oscillating batches. When the oscillating batch exceeds the upper limit, it is transferred to manual assessment. The concentration intensity is quantified by the difference between the proportions of the top quartile and the bottom quartile of the weight distribution in the power transfer level distribution. The larger the difference, the smaller the gap is to differentiate in dense high-weight areas; the smaller the difference, the larger the gap is to avoid excessive division of dispersed weights. Each scheme is classified accordingly and written into the power transfer priority sequence in descending order of concentration intensity.

[0064] Power restoration paths are determined by sequentially sorting and integrating low-deterioration circuit sets through a power transfer priority sequence. The coverage scheme with the highest priority in descending order is written into the power restoration path first. Capacity allocation for each participating circuit in the low-deterioration circuit set is recorded synchronously, and the load undertaken by each circuit does not exceed its current corrected power transfer margin. Multiple schemes within the same priority are sorted a second time based on joint power transfer. In joint schemes, circuits are arranged from largest to smallest capacity to ensure the primary receiving circuit is clearly identified. Power restoration paths are entered into the database according to circuit number, capacity allocation, and coverage rate, in the order of restoration. During annual overhauls, multiple feeders experience simultaneous power outages. If the available coverage schemes in the power transfer priority sequence are insufficient to restore all circuits at once, a phased restoration sequence is determined sequentially under the capacity constraints of the low-deterioration circuit set. Subsequent batches recalculate the remaining demand and available capacity after the previous batch is restored. If the power transfer priority sequence includes emergency mode source schemes, these are marked as emergency paths when written into the power restoration path. Emergency paths have higher priority for breakpoint connection than non-emergency paths when the power supply collaborative restoration network is generated. In schemes involving circuits from low-deterioration circuits that are conservatively selected as sources, the power restoration path is marked with a conservative coverage rate during writing. When reconnecting breakpoints, the breakpoint location is determined for conservative path segments using a lower degradation rate threshold. When multiple power restoration paths are written simultaneously in similar batches, the capacity of overlapping nodes is determined by the highest priority; nodes with insufficient capacity to meet the minimum transfer requirements do not participate in the capacity allocation for that path. The writing order of power restoration paths is the same as the field execution order. The transfer capacity occupied by the preceding path is deducted before the subsequent path is scheduled, ensuring that the available capacity of each path always reflects the real-time remaining capacity rather than the static total.

[0065] A collaborative power restoration network is formed by connecting and repairing breakpoints at the maximum degradation rate in the power restoration path and power assessment sequence. The maximum degradation rate breakpoint is located in the batch with the largest absolute value of the slope of the power supply level in the power assessment sequence among the nodes of each restoration path. The slope is estimated using the average difference of the sliding window; a larger absolute value of the slope indicates that the power supply level of that node has deteriorated most severely in the vicinity of that batch, such as the section where cable joint oxidation is accelerated, which is the weakest point in the restoration path structure. The path segments on both sides of the breakpoint are independently connected in the collaborative power restoration network. The connection uses the power supply level value at the breakpoint as the common constraint boundary for the two path segments. After repair, the difference in levels between the two sides at the breakpoint does not exceed a set smoothing tolerance. After a large-scale power outage, multiple restoration paths simultaneously enter the breakpoint repair phase. Repair resources need to be allocated among multiple paths. Paths with higher priority in the power restoration path complete breakpoint location first. Breakpoint connection within parallel batches is executed one by one from high to low priority to avoid resource contention caused by parallel processing. For path segments marked as emergency paths in the power restoration path, breakpoint repair has a higher priority than non-emergency paths. The power supply collaborative restoration network prioritizes batches of emergency path breakpoint connections for connectivity verification. For path segments marked as conservative coverage in the power restoration path, breakpoint location is identified using a lower degradation rate threshold, and the connection results supplement coverage by expanding the verification range. For loops with dynamic migration sources in the power supply assessment sequence, breakpoint location uses the batch after migration as the effective statistical starting point, and the slopes before and after migration are estimated independently. The power supply collaborative restoration network marks the migration nodes as migration breakpoints. Path segments on both sides of the breakpoint are repaired independently, using the batch with the highest degradation rate as the dividing point. Connection is based on the power supply level value at the breakpoint as the common constraint boundary for both sides. After repair, the difference between the levels on both sides at the breakpoint does not exceed the set smoothing tolerance. Those exceeding the tolerance are reworked until connectivity verification is passed.

[0066] The protection status assessment is based on power supply collaborative restoration network and enclosure environmental data, outputting dynamic assessment results of protection level. The protection status score is obtained by weighted fusion of the recovery weakness of the power supply collaborative restoration network and the sealing degradation, internal water accumulation, and insulation degradation characteristics of the enclosure environmental data. The recovery weakness is quantified by the network interruption repair density and the node missing completion ratio. The enclosure environmental data characteristics are normalized to the current degradation magnitude of each environmental parameter. The dynamic assessment results of protection level are aggregated by node, with the protection level mapped by the fused score and the corresponding confidence level. Nodes with a cross value of zero in the power supply collaborative restoration network connection matrix are judged as missing. The recovery weakness is quantified by the product of the interruption repair density and the node missing completion ratio; the higher both are, the greater the weakness. Missing nodes are filled by the shortest path of the restored nodes on both sides of the gap. Nodes temporarily withdrawn during the distribution network renovation period are excluded according to the renovation mark and are not included in the weakness calculation. Based on this, the dynamic assessment results of protection level are output. The physical protection of nodes in the enclosure environment, where sealing deterioration and internal water accumulation are simultaneously worsening, has been significantly weakened. The dynamic assessment of protection level results downgrades the protection level based on environmental degradation as the dominant factor. Even if the power supply coordination and network recovery capabilities are still acceptable, it cannot mask the degradation of physical protection. Stricter connectivity thresholds are applied to path segments with dense breakpoint repairs, and nodes with high vulnerability are marked for priority review. Nodes with path stability or fusion confidence below the lower limit must be manually confirmed before release and will not be removed from the monitoring queue. When the protection level is confirmed to be consistently low, node isolation is triggered, the reason for isolation and the expected recovery batch are recorded, and power supply guarantee control instructions are generated and issued simultaneously based on the dynamic assessment results of protection level.

[0067] To implement the above-described method embodiment, a dynamic evaluation method for the protection level of a low-voltage cable branch box is provided to achieve the corresponding functional and technical effects. See also... Figure 3 , Figure 3 This paper presents a structural block diagram of a dynamic evaluation system 300 for the protection level of a low-voltage cable branch box according to an embodiment of this application, including: Data acquisition module 301 is used to collect enclosure environmental data and power supply time sequence data and perform joint degradation trend analysis to form a reliability feature set, and to identify early power failure characteristics from the power supply time sequence data to obtain a power failure early warning identifier; Evolution reconstruction module 302 is used to perform degradation path backtracking and reconstruction on the reliability feature set to construct a power supply level evolution map, and to determine the power supply evaluation segment based on the power supply level evolution map and the power failure early warning flag constraint. Rule generation module 303 is used to perform inter-loop correlation analysis on the power supply assessment section to determine the power failure threshold, and to establish node inspection rules based on the power failure threshold and the power failure propagation priority weight allocation of the power supply assessment section. The circuit restoration module 304 is used to divide the circuit into a power supply assessment sequence and an emergency restoration sequence according to the node inspection rules, perform a restoration success rate decay trend assessment on the emergency restoration sequence to generate a restoration priority sequence, and perform a level stability detection on the power supply assessment sequence to obtain the power supply deviation. The result output module 305 is used to determine the power supply degradation coefficient by performing deviation growth rate analysis based on the power supply deviation amount, match the power supply degradation coefficient with the power restoration priority sequence to determine the power supply recovery path by performing degradation acceleration stage transfer, perform maximum degradation rate breakpoint connection repair on the power supply recovery path and the power supply assessment sequence to form a power supply collaborative recovery network, and perform protection status assessment based on the power supply collaborative recovery network and the enclosure environment data to output the protection level dynamic assessment result.

[0068] The aforementioned dynamic evaluation system 300 for the protection level of a low-voltage cable branch box can implement one of the dynamic evaluation methods for the protection level of a low-voltage cable branch box as described in the above-described method embodiments. The options in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining contents of this application embodiment can be referred to the contents of the above method embodiments, and will not be repeated in this embodiment.

[0069] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

Claims

1. A method for dynamic evaluation of the protection level of a low-voltage cable branch box, characterized in that, include: Collect enclosure environmental data and power supply time sequence data, and perform joint degradation trend analysis to form a reliability feature set. Identify early power failure characteristics from the power supply time sequence data to obtain power failure early warning indicators. A power supply level evolution diagram is constructed by performing degradation path backtracking on the reliability feature set, and a power supply evaluation segment is determined based on the power supply level evolution diagram and the power failure early warning indicator constraint. The power supply assessment section is subjected to inter-loop correlation analysis to determine the power failure threshold. Based on the power failure threshold and the power failure propagation priority weight allocation of the power supply assessment section, node inspection rules are established. Based on the node inspection rules, the circuit is divided into a power supply assessment sequence and an emergency power restoration sequence. The power restoration success rate decay trend assessment is performed on the emergency power restoration sequence to generate a power restoration priority sequence. The power supply assessment sequence is subjected to level stability detection to obtain the power supply deviation. The power supply degradation coefficient is determined by analyzing the deviation rate based on the power supply deviation amount. The power supply degradation coefficient and the power restoration priority sequence are matched to determine the power supply recovery path by the degradation acceleration stage. The power supply recovery path and the power supply assessment sequence are connected and repaired at the maximum degradation rate to form a power supply collaborative recovery network. Based on the power supply collaborative recovery network and the enclosure environment data, the protection status is assessed and the dynamic assessment result of the protection level is output.

2. The method according to claim 1, characterized in that, The step of identifying early power outage characteristics from the power supply timing data to obtain a power outage warning identifier includes: The power supply timing data was checked segment by segment from the last maintenance to the first commissioning to determine the historical voltage anomaly nodes; An anomaly spacing distribution map is generated by analyzing the interval between the anomaly time and the re-inspection date of the historical voltage anomaly nodes. The abnormal spacing distribution map is quantified by the initial fluctuation density dispersion measure to form a fluctuation weight distribution; Based on the fluctuation weight distribution, periodic alternating fluctuation segments are identified to obtain power outage warning indicators.

3. The method according to claim 1, characterized in that, The process of constructing a power supply level evolution map by backtracking the degradation path of the reliability feature set includes: Based on the aforementioned reliability feature set, perform statistical analysis of the grade differences of distributed power supply nodes to generate a grade difference set; Based on the short-term rebound segment after a sudden drop in the level difference set, a group of nodes with missing high-level differences is identified. A grade interpolation distribution is formed by interpolating neighboring nodes to complete the high-grade difference missing node group and the grade difference set; The power supply level evolution diagram is constructed by performing power supply level sequence restoration on the level interpolation distribution.

4. The method according to claim 1, characterized in that, The step of performing inter-loop correlation analysis on the power supply assessment section to determine the power outage threshold includes: The power supply assessment section is divided into high-risk power outage circuits and power outage interval fluctuation distributions according to the power outage risk level; A threshold candidate parameter set is established by periodically hierarchically and seasonally integrating the high-risk power failure circuits. The threshold candidate parameter set is matched with the power outage interval fluctuation distribution to form a threshold correction distribution by successively shortening the interval trend. The power loss threshold is determined by filtering the effective threshold interval using the threshold correction distribution.

5. The method according to claim 1, characterized in that, The node inspection rules established based on the power outage threshold and the power supply assessment segment's power outage propagation priority weight allocation include: Using the power outage threshold, a cumulative risk value sequence is generated by performing statistical analysis on the power outage period approaching each node within the power supply assessment segment without early warning. The cumulative risk value sequence and the power supply assessment segment are converted to correspondence of no-early warning duration to determine the inspection density weighting coefficient; A dynamic inspection density distribution is formed by standardizing and calibrating the inspection density weighting coefficient. Based on the dynamic inspection density distribution, a density-level time-series node allocation is performed to establish node inspection rules.

6. The method according to claim 1, characterized in that, The step of evaluating the power restoration success rate decay trend of the emergency power restoration sequence to generate a power restoration priority sequence includes: Based on the emergency power restoration sequence, the power transfer adaptability is calculated using the historical power restoration parameters of each power failure circuit to form a power transfer adaptability sequence; A power restoration risk score is obtained by identifying the accelerated attenuation segment after a brief rebound in the power supply adaptability sequence. The circuit power restoration priority table is obtained by comprehensively ranking the power restoration risk score and the emergency power restoration sequence based on the reliability of power transfer. The power restoration priority table is used to filter valid power restoration circuits and generate a power restoration priority sequence.

7. The method according to claim 1, characterized in that, The step of matching the power supply degradation coefficient with the power restoration priority sequence during the degradation acceleration phase to determine the power restoration path includes: Based on the power supply degradation coefficient, the single degradation value and the cross-circuit cumulative offset are analyzed; The power restoration priority sequence is screened according to the single degradation value to form a low degradation loop set; A cross-loop transfer coverage assessment is performed on the low-deterioration loop set and the cross-loop cumulative offset to determine the transfer priority sequence; The power restoration path is determined by sequentially sorting and integrating the set of low-deterioration circuits using the power transfer priority sequence.

8. The method according to claim 4, characterized in that, The establishment of a threshold candidate parameter set based on the periodic hierarchical seasonal integration of the high-risk power failure circuits includes: Based on the aforementioned high-risk power failure circuits, the power failure duration is segmented into short-cycle power failure sets and long-cycle power failure sets. Based on the short-cycle power loss set, a power loss cause difference type identification is performed to generate a short-cycle cause classification set; The threshold constraint interval is determined by integrating the short-period cause classification set and the long-period power loss set into the priority interval of the seasonal power loss concentration segment. A threshold candidate parameter set is established by performing interval parameter segmentation analysis on the threshold constraint interval.

9. The method according to claim 7, characterized in that, The step of determining the transfer priority sequence by evaluating the cross-loop transfer coverage based on the low-deterioration loop set and the cross-loop cumulative offset includes: A deviation coverage score sequence is generated by performing deviation quantization based on the set of low-deterioration loops and the cross-loop cumulative offset; For the deviation coverage scoring sequence, the stability correction coefficient is determined by statistically analyzing the fluctuation amplitude of the cross-loop adaptation score of each power failure loop. The deviation coverage scoring sequence is corrected by decreasing the priority weight of the failed transfer segment using the stability correction coefficient to form a transfer level distribution. The priority sequence for power transfer is determined based on the intensity of the distribution concentration of the power transfer level distribution.

10. A dynamic evaluation system for the protection level of a low-voltage cable branch box, characterized in that, include: The data acquisition module is used to collect environmental data and power supply time sequence data of the enclosure and perform joint degradation trend analysis to form a reliability feature set. It identifies early power failure characteristics from the power supply time sequence data to obtain a power failure early warning indicator. The evolution reconstruction module is used to perform degradation path backtracking and reconstruction on the reliability feature set to construct a power supply level evolution map, and to determine the power supply evaluation segment based on the power supply level evolution map and the power failure early warning indicator constraint. The rule generation module is used to perform inter-loop correlation analysis on the power supply assessment segment to determine the power outage threshold, and to establish node inspection rules based on the power outage threshold and the power outage propagation priority weight allocation of the power supply assessment segment. The circuit restoration module is used to divide the circuit into a power supply assessment sequence and an emergency power restoration sequence according to the node inspection rules. It performs a power restoration success rate decay trend assessment on the emergency power restoration sequence to generate a power restoration priority sequence, and performs a level stability detection on the power supply assessment sequence to obtain the power supply deviation. The result output module is used to determine the power supply degradation coefficient by performing deviation growth rate analysis based on the power supply deviation amount, match the power supply degradation coefficient with the power restoration priority sequence to determine the power supply recovery path by performing degradation acceleration stage transfer, perform maximum degradation rate breakpoint connection repair on the power supply recovery path and the power supply assessment sequence to form a power supply collaborative recovery network, and perform protection status assessment based on the power supply collaborative recovery network and the enclosure environment data to output the protection level dynamic assessment result.

Images