A method and system for monitoring and fault isolation of low-voltage cable branch boxes

By deducting the contribution of sealing leakage in the monitoring of low-voltage cable branch boxes, adaptively adjusting the threshold for judging loose contact, and combining it with blind spot verification during inspections, the problem of distinguishing between sealing leakage and grounding anomalies is solved, improving the accuracy of fault identification and the safety of handling, ensuring thorough isolation, and reducing the impact on power supply.

CN123073635APending Publication Date: 2026-11-10SHANDONG TONGCHANG ELECTRIC POWER EQUIP TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611332709.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing monitoring methods for low-voltage cable branch boxes are insufficient to accurately distinguish between leakage current caused by sealing leaks and grounding abnormalities. Intermittent faults are difficult to trace to their source, and complete disconnection cannot be verified after isolation. Cross-box commands are prone to conflict, and historical records lack verification.

Method used

By deducting the contribution of sealing leakage, adaptively adjusting the contact loosening judgment threshold, combining inspection blind spot verification, verifying hidden passages after isolation, and uniformly handling cross-container conflicts and legacy records, the accuracy of fault identification and the safety of handling are improved.

Benefits of technology

It enables clear differentiation between sealing leaks and grounding anomalies, improves the efficiency of locating intermittent faults, ensures thorough isolation and reduces the impact on power supply continuity, and avoids command conflicts and legacy issues.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123073635A_ABST
    Figure CN123073635A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for monitoring and isolating faults in low-voltage cable branch boxes. It collects branch current and box environment data to identify insulation drops, and uses condensation cycles to eliminate interference and determine sealing degradation indicators. Then, it subtracts the contribution of sealing leakage from the zero-sequence signal, screens for potential dual-point grounding hazards in excessively stable intervals, and calculates a floating discrimination threshold based on the degree of looseness. Based on this threshold, it identifies abnormal states, traces the root cause of intermittent faults through pre-stabilization duration and concurrent common-cause analysis, and outputs confirmed potential abnormal signals after blind spot inspection. It then generates isolation commands to switch to healthy branches, verifies hidden conductive bypasses through residual voltage attenuation characteristics, and finally collects handling data along the tracing trajectory, completing cross-box command coordination and historical branch verification to form a closed-loop isolation command. This improves the accuracy of identifying sealing and grounding anomalies and the efficiency of locating intermittent faults, ensuring the safety and integrity of isolation handling.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power equipment fault monitoring technology, and in particular to a method and system for monitoring and isolating faults in low-voltage cable branch boxes. Background Technology

[0002] Low-voltage cable distribution boxes, as critical nodes for load distribution in power distribution networks, operate long-term in underground or semi-enclosed environments. The sealing condition of the box and the insulation level of the cables directly affect the operational safety and power supply continuity of the power distribution system. Problems such as sealing aging and condensation leakage in the box can easily induce localized insulation deterioration. Furthermore, when the cable sheath forms an additional grounding path beyond the normal single-point grounding, grounding loop currents may also occur. If neither of these issues is detected and addressed in a timely manner, they can further evolve into serious faults affecting power supply reliability. In addition, fixed threshold values ​​do not account for differences in grounding point loosening, and uneven inspection coverage can easily lead to repeated false alarms.

[0003] However, most existing branch box monitoring methods are designed independently for single types of abnormal signals. When leakage current caused by sealing leakage and grounding abnormalities are manifested in the same set of electrical quantities, they are prone to mutual interference, making it difficult to accurately distinguish the source of the fault. Intermittent and poor contact faults often manifest as simultaneous fluctuations or intermittent interruptions in multiple branches, and relying on a single abnormality judgment can easily lead to misjudgment or omission. In addition, after the faulty branch is isolated, whether a complete electrical disconnection is truly achieved, whether there are conflicts between different box instructions during the handling process, and whether historical unverified records can be reviewed in a timely manner are all issues that currently lack a unified closed-loop processing mechanism. Summary of the Invention

[0004] This invention discloses a monitoring and fault isolation method and system for low-voltage cable branch boxes, aiming to solve problems such as mutual interference between sealing leakage and grounding abnormalities, difficulty in tracing the source of intermittent faults, inability to verify complete disconnection after isolation, cross-box command conflicts, and lack of historical record verification. By deducting leakage contribution before grounding discrimination, adaptively calculating the discrimination threshold according to the degree of contact looseness, tracing the source based on the pre-stabilization time and concurrent common causes, supplementing the verification with inspection blind spots, verifying hidden paths after isolation and switching, and uniformly handling cross-box conflicts and legacy verification, the accuracy of fault identification and the safety and reliability of the handling process are improved.

[0005] The first aspect of this invention proposes a method for monitoring and fault isolation of low-voltage cable branch boxes, comprising: Real-time data of the branch current and the enclosure environment are collected. Based on the real-time data of the branch, insulation drop is identified to form insulation drop period. Through the insulation drop period, the seal failure of the misjudgment interference condensation cycle is screened to obtain the seal deterioration mark. Based on the sealing degradation indicator and the real-time data of the branch, the associated zero sequence is extracted to generate a zero sequence cancellation segment. In the zero sequence cancellation segment, the grounding screening of the overly stable interval is carried out to obtain the double-point grounding hazard. The attenuation characteristic analysis is performed on the double-point grounding hazard to determine the floating discrimination threshold. Based on the floating discrimination threshold, abnormal states are identified and abnormal discrimination results are constructed. The abnormal discrimination results are subjected to non-reproducible fault time sequence analysis to obtain the location of the fault branch. The location of the fault branch is used to screen for false alarm blind spots and output potential abnormal signals. Based on the potential abnormal signal, the faulty branch is isolated to obtain the branch isolation command. Based on the branch isolation command, the healthy branch is switched to form the switching and handling result. The residual pressure hysteresis hidden bypass is screened through the switching and handling result to generate the bypass hidden indicator. Based on the bypass concealment flag, the data is traced and aggregated along the repeated fading process to generate scheduling and reporting data. Based on the scheduling and reporting data, cross-box instruction verification is performed to form a set of handling tasks. For the set of handling tasks, long-term unverified branch verification is carried out to output isolation closed-loop instructions.

[0006] A second aspect of this invention provides a low-voltage cable branch box monitoring and fault isolation system, comprising: The data acquisition module is used to collect real-time data of the branch current and the enclosure environment. Based on the real-time data of the branch, it identifies insulation drop periods and uses these insulation drop periods to screen for misjudgment interference condensation cycle seal failure and obtain seal deterioration indicators. The grounding determination module is used to extract the associated zero sequence based on the sealing deterioration mark and the real-time data of the branch to generate a zero sequence cancellation segment, perform grounding screening in the over-stable interval of the zero sequence cancellation segment to obtain the double-point grounding hazard, and perform attenuation characteristic analysis on the double-point grounding hazard to determine the floating discrimination threshold. The fault location module is used to identify abnormal states based on the floating discrimination threshold, construct abnormal discrimination results, perform non-reproducible fault time sequence analysis on the abnormal discrimination results to obtain the location of the fault branch, and screen for false alarm blind spots and output potential abnormal signals based on the location of the fault branch. The isolation and handling module is used to isolate faulty branches based on the potential abnormal signals to obtain branch isolation instructions, switch healthy branches according to the branch isolation instructions to form a switching and handling result, and perform residual pressure hysteresis concealed bypass screening to generate bypass concealment indicators through the switching and handling result; The scheduling and handling module is used to trace and collect data along the repeated fading process based on the bypass concealment flag to generate scheduling and reporting data, perform cross-box instruction verification based on the scheduling and reporting data to form a handling task set, and perform long-term unverified branch verification and output isolation closed-loop instructions for the handling task set.

[0007] The beneficial effects of this invention are reflected in the following points: First, the manifestations of two types of faults, namely sealing aging and grounding anomalies, in the zero-sequence signal are easily masked by each other. Traditional monitoring methods often misjudge leakage current caused by sealing leakage as part of grounding anomalies, or conversely, ignore the actual grounding hazards. This invention, by deducting the inherent contribution of sealing leakage to the zero-sequence signal before grounding anomaly identification, separates the criteria for the two types of faults from sharing the same signal quantity into their own independent criterion chains. This makes the identification of grounding anomalies no longer affected by fluctuations in the sealing state, the physical meaning of the criteria is clearer, and the source of misjudgment is easier to trace. Secondly, addressing the characteristic of intermittent anomalies caused by poor contact repeatedly occurring in the same branch and often overlapping with short-term fluctuations in adjacent branches, this invention uses a joint analysis of two dimensions—pre-stabilization duration and concurrent common causes—to link scattered multiple anomalies according to their common root cause. It then uses recurrence weighting to pinpoint the responsible branch, while also including insufficiently covered areas in the verification scope, reducing omissions caused by blind spots. Branches with different degrees of loose contact are assigned their respective judgment thresholds, avoiding misjudgments and omissions due to the same fixed threshold under different loose states, thus improving the efficiency of intermittent fault location. Finally, this invention incorporates three easily overlooked steps into the same process: while isolating, power is transferred to backup branches according to priority, reducing the impact on power supply continuity; after switching, residual voltage bypass verification is added to avoid incomplete isolation despite the switch being off; and before issuing instructions, cross-box conflicts are checked, and long-unverified historical records are reviewed to prevent the backlog of old problems. Attached Figure Description

[0008] Figure 1 This is a flowchart of a low-voltage cable branch box monitoring and fault isolation method according to the present invention.

[0009] Figure 2 This is a schematic diagram of the fault branch isolation and healthy load switching topology of the present invention.

[0010] Figure 3 This is a structural block diagram of a low-voltage cable branch box monitoring and fault isolation system according to the present invention.

[0011] Wherein: 1-Faulted branch; 2-Isolating switch; 3-Tie switch; 4-Backup branch; 5-Load switching path; 6-High priority load identifier; 7-Low priority load identifier; 8-Insufficient capacity identifier; 9-Loads retained in the original branch. 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] 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.

[0014] The technical solutions of the embodiments of this application are described below.

[0015] like Figure 1 As shown, this embodiment of the invention provides a method for monitoring and isolating faults in low-voltage cable branch boxes, including the following steps S101-S105. The evaluation period described herein is 7 days. Unless otherwise stated, the thresholds, coefficients, grading limits, and baseline values ​​appearing in the following steps are all taken from the type test records, factory calibration data, and operation and maintenance procedures of the power supply area for this type of branch box, and their sources are not individually indicated.

[0016] Step S101: Real-time data of the branch current and the enclosure environment are collected. Based on the real-time data of the branch, insulation drop is identified to form insulation drop period. Through the insulation drop period, the seal failure is screened for misjudgment interference condensation cycle to obtain the seal deterioration mark.

[0017] Specifically, real-time data for the branch current and the enclosure environment are collected. Branch current is synchronously collected by three-phase current transformers installed at the beginning of each branch. The enclosure environment includes overall temperature and humidity within the enclosure and segmented temperature and humidity at the joints: the former is measured by a single probe installed in the middle of the inner wall of the enclosure, avoiding the inlet and outlet air ducts, representing the overall environment rather than localized airflow; the latter is measured at 200mm intervals along the enclosure joints, with temperature and humidity values ​​collected independently for each joint segment, and the results of each segment are independent of each other. All three signals share the same clock reference to trigger sampling, with a sampling period of 1 minute and timestamp alignment accuracy to the second. The branch real-time data records the branch current, enclosure temperature, and enclosure humidity for each sampling moment, and archives them in chronological order of acquisition time. If a signal is missing due to a momentary communication interruption, the missing measurement point is marked in the record and the system exits subsequent insulation sag detection and zero-sequence calculation. Entire records with missing measurements exceeding three sampling cycles are skipped without interpolation. This recording method ensures that the zero-sequence current can be aligned and compared with the synchronous temperature and humidity during subsequent insulation sag detection. The zero-sequence component is additionally calculated in the branch current channel of the branch real-time data; the sum of the instantaneous values ​​of the three-phase branch currents is the zero-sequence current. When the three-phase currents are in balance, they sum up and cancel each other out, and this amount remains at a low level for a long time. After the enclosure gets damp, such as after a rainstorm when water seeps into the cable inlet hole, or when the temperature difference between day and night causes condensation inside the enclosure, the water film creates a leakage path between the live parts and the enclosure. Some current flows into the ground without returning through the load side, and the three phases no longer cancel each other out. The zero-sequence current then jumps up for a short time, and the jump amplitude is limited to the upper limit of the zero-sequence current fluctuation under normal operating conditions.

[0018] Insulation drop periods are identified based on real-time branch data. The determination is made point-by-point on the zero-sequence current channel of the branch's real-time data: if the single sampling level of the zero-sequence current exceeds the aforementioned fluctuation limit, it is marked as the start time of the drop; if the zero-sequence current falls back and remains stable below the fluctuation limit for five consecutive sampling points, it is marked as the end time of the drop. The time between the start and end times constitutes the insulation drop period. The insulation level itself is not directly measurable during operation. Once the insulation inside the enclosure becomes damp and deteriorates, the leakage current increases accordingly. Therefore, the rise in the zero-sequence current becomes a substitute observation of the insulation state; the greater the rise, the greater the corresponding insulation margin loss. If the interval between two insulation drop periods on the same branch is less than 10 minutes, it is determined as sampling jitter of the same drop and merged into a single record with connected start and end times. If the interval exceeds 10 minutes, it is treated as two independent drops, and the sampling points between the two segments are not included in any insulation drop period. The box temperature and humidity during the specified time period are captured synchronously from the branch's real-time data and saved along with the segment for quick reference during on-site verification. In the next sub-step, when comparing the condensation phase, the complete day-night record is retrieved from the branch's real-time data, independent of this segment snapshot. The criteria for the start and end of the insulation drop are intentionally asymmetrical: the start only requires a single horizontal exceedance to trigger, while the end requires five consecutive sampling points to fall below the limit. This avoids prematurely closing the insulation drop period before the fluctuation has subsided and also prevents the same degradation from being split into multiple segments. Multiple insulation drop periods are allowed to be generated within a single day and night for a single branch, each independently numbered. The retained insulation drop periods are archived according to the branch number, and each segment records the start and end times of the drop, the peak zero-sequence current at trigger time, and the extreme values ​​of temperature and humidity during the same period.

[0019] In some embodiments, the step of screening for seal failure during the insulation drop period to detect misjudgment interference condensation cycle includes: tracking the diurnal temperature difference phase of the condensation cycle during the insulation drop period to obtain the condensation-evaporation phase difference; calculating the elastic recovery hysteresis of the sealing ring based on the condensation-evaporation phase difference to determine the elastic hysteresis amount; identifying local micro-leakage paths based on the elastic hysteresis amount to form a micro-leakage path distribution; and evaluating the sealing level of the continuous leakage section along the box joint for the micro-leakage path distribution to obtain the seal deterioration indicator.

[0020] The condensation-evaporation phase difference is obtained by tracking the diurnal temperature difference phase of the condensation cycle during the insulation drop period. The temperature sequence used for comparison is retrieved from the real-time data of the branch, taking into account the complete diurnal record of the drop, instead of relying on the segments saved during the insulation drop period. The diurnal temperature difference phase is characterized by the crossing time of the enclosure temperature against the dew point temperature: the moment when the temperature drops below the dew point from high to low is recorded as the start of condensation, and the moment when it rises above the dew point is recorded as the completion of evaporation. The dew point temperature is calculated from the synchronously collected enclosure humidity according to the saturated vapor pressure relationship. Multiple pairs are allowed within the diurnal cycle. The condensation-evaporation phase difference is calculated using the formula Δφ=t_evap-t_cond, where t_cond is the condensation start time, t_evap is the evaporation completion time, and Δφ is the time span of this condensation cycle, in minutes. Condensation is a recurring process of water vapor precipitating and evaporating on a cold surface. The diurnal temperature range causes the metal wall of the enclosure to repeatedly fall below and rise above the dew point. During this period, the water film also lowers insulation and raises the zero-sequence value, making it difficult to distinguish from actual seal failure in terms of electrical quantities. Therefore, it must be separated by phase. The condensation-evaporation phase difference is calculated only for the diurnal cycle including periods of sudden insulation drop. Δφ is extracted sequentially from each pair, and arranged in sequence according to the time of occurrence to reflect the speed of the seal ring response. When the start and end times of a sudden insulation drop fall between the start of condensation and the completion of evaporation, the drop is considered to be in phase with the condensation cycle, recorded as condensation interference, and removed from the seal failure criteria. If the drop does not fall within any condensation-evaporation window, it is considered a continuous insulation drop unrelated to condensation, retained as a candidate for seal failure, and the Δφ of the pair closest to its time and with a temperature change rate difference not exceeding one time is taken as the phase difference value corresponding to the drop. When all pairing rates exceed the limit, the drop is only recorded for the time period itself and is not included in the elastic hysteresis calculation.

[0021] The elastic hysteresis is determined by measuring the condensation-evaporation phase difference to assess the elastic recovery hysteresis of the sealing ring. The condensation-evaporation phase difference reflects the rhythm of water vapor precipitation and evaporation. The expansion and contraction of the sealing ring follow the same rhythm but with a response delay, which increases with material aging. The factory test report for this model of sealing ring provides the nominal hysteresis phase φ0 in the unaged state. The hysteresis is determined based on the average phase hysteresis of the sampled tests for this batch, with the unit being minutes, the same as Δφ. The elastic hysteresis is calculated using the formula ΔH = Δφ - φ0, where Δφ is the condensation-evaporation phase difference associated with the sudden drop, φ0 is the nominal hysteresis phase, and ΔH is the calculated elastic hysteresis, retaining the positive or negative sign, with the same unit as Δφ. When |ΔH| exceeds the upper limit Δ_u of the uncertainty interval corresponding to φ0 (determined by calibration repeatability error), the sealing ring material is considered to have entered the measurable aging stage, and the elastic hysteresis is marked with an aging indicator; if |ΔH| falls within the uncertainty interval, only the value is recorded. Aging indicators are categorized as follows: a positive indicator shows overall material hardening, slower rubber rebound, and inability to keep up with the expansion and contraction rhythm due to temperature changes, with Δφ consistently greater than φ0; a negative indicator shows localized permanent deformation, where the area has been compressed and no longer fluctuates with temperature changes, with Δφ consistently less than φ0. The former indicates leakage throughout the circumference, while the latter is concentrated at the indentation point; their causes are different and they are not two descriptions of the same failure. The response delay of the sealing ring varies with the rate of temperature change, therefore Δφ and φ0 must be taken from similar temperature change rate ranges. Periods with rates differing by more than double are not included in this calculation to avoid mistaking the rate of temperature change for material aging. When the aging indicator direction obtained from three consecutive calculations for the same branch is consistent, a direction confirmation mark is added to the sealing ring file for that branch, containing the aging direction and the number of consecutively compliant calculations. This mark is used as a priority reference in the next sub-step to determine the distribution of micro-leakage paths, reducing misjudgments caused by occasional single aging indicators.

[0022] The microleakage path distribution is formed by identifying local microleakage paths based on elastic hysteresis. The joint of the enclosure is divided into grid segments with a 200mm interval along its direction. This interval is consistent with the conventional point density of the sealing ring measurement points, with typically only one point within a segment. Each segment is associated with a corresponding elastic hysteresis value, derived from the independent calculations of temperature and humidity measurement points within that segment using the aforementioned method. When there are more than one measurement point within the same segment, the average of all measurement points is taken. When the difference in elastic hysteresis between two adjacent segments exceeds a set gradient (taken as 1.5 times the standard deviation of elastic hysteresis across all segments, and not less than Δ_u), the segment coordinates at the corresponding location are included in the microleakage path distribution. At the point of abrupt gradient change, the seal ring and the joint surface are least tightly fitted, allowing moisture to seep in preferentially. When all segments |ΔH| exceed this upper limit, the coordinates of the entire joint segment are included regardless of whether the gradient exceeds the limit. When a segment lacks measured points and therefore has no elastic hysteresis value, it is temporarily filled by linear interpolation using the values ​​of the nearest valid segments on both sides. Segments without valid segments on both sides are not filled and are recorded as empty segments. The interpolated segments are only for reference and are not used as the start and end points of continuous leakage segments to avoid interpolation noise being mistakenly included in the classification as the actual leakage boundary. Each coordinate included in the micro-leakage path distribution carries the gradient difference at the time of triggering and the segment number to which it belongs, and the two are stored in pairs in the same record. If no adjacent segment difference on the entire joint exceeds the set gradient and all segments |ΔH| do not exceed the upper limit, the current micro-leakage path distribution is empty, the elasticity of the sealing ring is judged to be intact in this evaluation cycle, the empty result is retained and marked with an intact judgment mark.

[0023] The sealing level of continuous leakage sections along the tank joints is assessed based on the distribution of micro-leakage paths to obtain sealing degradation indicators. The continuous span between adjacent included coordinates in the micro-leakage path distribution along the tank joint direction is statistically analyzed. Coordinates with a spacing of less than or equal to one grid unit (i.e., 200 mm) are considered the same continuous leakage section, while those exceeding one grid unit are considered the boundaries of two independent leakage sections. This yields the length of each continuous leakage section. Single-point leakage is often just a localized indentation; only when it connects along the joint does it indicate that the sealing surface has completely lost its adhesion. Therefore, continuous length, rather than the number of points, is used as the grading basis. A common scenario in the field is that after a localized aging and sealing ring at a tank joint dies, initially only moisture seeps from that point, but over time it spreads along the joint to both sides, increasing the length of the continuous leakage section. In contrast, single-point scratches or indentations left during transportation or installation usually do not expand outwards and remain as isolated coordinates over time. Sealing levels are categorized into three grades based on the length of the continuous leakage segment: Slightly severe (length not exceeding 2 grid units); Moderately severe (length exceeding 2 but not exceeding 5 grid units); and Severely severe (length exceeding 5 grid units). The sealing degradation indicator is determined by the current sealing level of the branch and the maximum length of the corresponding continuous leakage segment. When the sealing level reaches moderate or higher, the joint location interval triggering this level is also recorded, indicated by start and end segment numbers, for direct location during on-site resealing operations. If multiple continuous leakage segments are identified within the same branch's micro-leakage path distribution during this assessment period, the sealing degradation indicator is the one with the highest sealing level. The lengths and locations of the remaining segments are separately added to the branch's joint log, which is updated periodically according to the joint direction.

[0024] Step S102: Based on the sealing deterioration mark and the real-time data of the branch, the associated zero sequence is extracted to generate a zero sequence cancellation segment. In the zero sequence cancellation segment, the grounding screening of the overly stable interval is carried out to obtain the double-point grounding hazard. The attenuation characteristic analysis is performed on the double-point grounding hazard to determine the floating discrimination threshold.

[0025] Specifically, zero-sequence current is generated by extracting associated zero-sequence current based on sealing degradation indicators and real-time branch data. Once a leakage path is formed, its conductance to ground remains essentially constant within an assessment period, exhibiting a stable bias rather than random fluctuations in the zero-sequence current. This can be deducted from the zero-sequence current calculated from the real-time branch data. The equivalent leakage current coefficient is then calculated based on the maximum length of the continuous leakage section carried by the sealing degradation indicator and the sealing level. The coefficients for light, moderate, and severe conditions are 0.5, 1.2, and 2.0 mA per grid unit, respectively, denoted as k_s. I_leak is calculated using the formula I_leak=k_s×L_s, where L_s is the maximum length of the continuous leakage segment recorded in the seal deterioration indicator, in grid units, and I_leak is the zero-sequence current associated with the seal leakage, in milliamperes. The residual after cancellation is calculated using the formula I_res(t)=I_0(t)-I_leak, where I_0(t) is the instantaneous value of the zero-sequence current at time t in the branch's real-time data, in milliamperes, and I_res(t) is the residual after cancellation. If I_leak exceeds the average value of I_0(t) for that period, the conversion is deemed invalid, and a fixed value of 80% of the average is used instead. If there is currently no seal deterioration indicator for a branch, I_leak is treated as zero. The period in which multiple consecutive sampling points of the residual after cancellation do not contain strong disturbances such as sudden drops or switching is designated as the zero-sequence cancellation segment. Strong disturbances are excluded when the jump in I_res exceeds three times the standard deviation of the historical period without grounding faults in that branch. When multiple zero-sequence cancellation segments appear successively on the same branch, segments with an interval of less than 5 minutes are merged into one segment. The start and end times of the zero-sequence cancellation segments are recorded separately according to the branch number; zero-sequence cancellation segments that have been marked after the flag is updated are not retroactively recalculated, and only the real-time data of the newly collected branch after the update takes effect.

[0026] In some embodiments, the step of performing grounding screening in the excessively stable interval of the zero-sequence cancellation section to obtain potential two-point grounding hazards includes: calculating the zero-sequence residual fluctuation in the excessively stable interval based on the zero-sequence cancellation section to generate a residual stability index; locating the phase difference based on the residual stability index to form the location of a suspected grounding segment; determining the confidence level of the two-point grounding by measuring abnormally high resistance based on the location of the suspected grounding segment; and determining the grounding level of the continuously exceeding threshold high confidence segment based on the confidence level of the two-point grounding to obtain potential two-point grounding hazards.

[0027] The residual stability index is generated based on the zero-sequence residual fluctuation in the transitional stable interval calculated according to the zero-sequence cancellation segment. The residual after cancellation within the zero-sequence cancellation segment is divided into fixed-window sliding segments with a window width of 30 minutes and an overlap of 15 minutes between adjacent windows. Each window has at least 30 sampling points, and the standard deviation of the residual is calculated window by window. The residual stability index is calculated using the formula η = σ_window / σ_baseline, where σ_window is the standard deviation of the residual after cancellation in this window, σ_baseline is the baseline standard deviation under the historical normal single-point grounding condition of this branch, and η is a dimensionless quantity that typically fluctuates between 0.8 and 1.2 during normal single-point grounding. After the formation of the second grounding point, the sheathing is directly connected back to the ground, forming a shunt loop. The zero-sequence residual, which should fluctuate with the load, is continuously clamped and flattened, and the readings become abnormally flat instead of chaotic, thus η is significantly lower than 1. Before calculating σ_window, outliers within the window are removed. The criterion is to deviate from the window mean by 5 times the standard deviation; windows with outliers exceeding 10% of the window's points are discarded. σ_baseline is uniformly calibrated based on the records of the first year of operation of the branch with confirmed no grounding anomalies, and is reviewed annually. For branches that have been in operation for less than one year, the average value of other branches of the same model and batch is used as a temporary benchmark. When the actual duration of the zero-sequence cancellation segment within a certain window is less than 30 minutes, σ_window is calculated using the residual points after all segments are cancelled, and the window width is marked as a short window. Branches with η consistently below 0.3 for 24 consecutive hours (48 windows) are recorded as continuously low stability; this mark is only for record-keeping and does not participate in the data delineation for subsequent screening. Windows with η exceeding 2.0 in a given instance are considered strong disturbance residues and are excluded from this screening. Windows are kept overlapping to avoid the actual disturbance being diluted at the boundary between two windows, resulting in a falsely low residual stability index.

[0028] For example, the step of locating the suspected grounding segment location based on the residual stationary index to determine the phase difference includes: performing multi-measurement point phase difference reversal analysis based on the residual stationary index to generate a phase difference distribution map; calculating the zero-sequence loop impedance difference according to the phase difference distribution map to obtain the impedance difference distribution; using the impedance difference distribution to locate the grounding point triangle of the symmetrical superposition area to obtain the grounding point coordinates; and mapping the box drawing position using the grounding point coordinates to form the suspected grounding segment location.

[0029] Phase difference distribution map is generated based on multi-point phase difference reversal analysis using the residual stationarity index. The multi-point acquisition in this step uses the time period selected by the residual stationarity index as the standard. The phase difference distribution map is obtained from multiple current measurement points arranged along the wiring of the branch box. Each current measurement point is equipped with an independent cycle-level synchronous clock and voltage transformer, with a spacing of approximately 1.5 meters between adjacent branch cabinets. Phase measurement uses the current transformer angle difference comparison method, with the phase of the first current measurement point as the reference zero point, and the measurement error is controlled within ±3 degrees. Under normal single-point grounding conditions, the zero-sequence current phase direction of all current measurement points is consistent, and the phase difference between adjacent current measurement points should be close to 0 degrees. When double-point grounding occurs, the current direction of the current measurement points on both sides of the grounding point is reversed through an additional grounding loop, and the phase difference between adjacent current measurement points approaches 180 degrees. The value at the corresponding position on the phase difference distribution map jumps from close to 0 degrees to close to 180 degrees, arranged into a curve according to the current measurement point numbering order. Only readings corresponding to periods when the residual stationarity index is below a set threshold (the 10th percentile of the historical distribution of η for that branch) are included in the phase difference distribution map. During periods not below this threshold, the flow measurement point remains silent and does not collect data to avoid irrelevant data diluting the inversion characteristics. The window division of this index is consistent with the sampling window of the flow measurement point to avoid misalignment of the two sets of data. If a flow measurement point loses a reading due to a communication failure, the phase difference is given by linear interpolation of the two adjacent flow measurement points, and the interpolation position is recorded as the missing measurement. Readings with a measurement error exceeding ±3 degrees are marked as low-confidence readings, still included in the phase difference distribution map but listed separately from normal readings. The installation location of the flow measurement point is preferably selected at both ends of the branch entering and exiting the enclosure, determined along with the enclosure's factory drawings, and generally will not be changed after commissioning.

[0030] The impedance difference distribution is obtained by calculating the zero-sequence loop impedance difference using the phase difference distribution diagram. The currents on both sides of the grounding point flow to their respective nearest grounding paths in opposite directions. The point where the phase difference crosses 90 degrees on the phase difference distribution diagram marks the boundary between the two currents; at this point, the equivalent impedance of the zero-sequence loop switches from one path to the other. If neither end of this point has a low-confidence reading, it is considered a candidate interval for impedance abrupt change. For each candidate interval, the zero-sequence voltage and zero-sequence current are synchronously collected from the current measurement points at both ends. The zero-sequence voltage is obtained by summing the three-phase-to-ground voltages from the voltage transformers installed at each current measurement point. Both share an established clock reference and do not require resynchronization. The zero-sequence voltage and zero-sequence current must be sampled within the same cycle. If the time scale deviation between the two exceeds half a cycle, Z_k for that candidate interval is not calculated, because phase misalignment will cause a falsely high impedance estimate. The zero-sequence loop impedance is calculated using the formula Z_k = U_k / I_k, where U_k is the effective value of the zero-sequence voltage at the starting point of the k-th candidate interval (in volts), I_k is the effective value of the zero-sequence current at the same current measurement point (in amperes), and Z_k is the estimated impedance of the corresponding loop (in ohms). When two adjacent candidate intervals share the same current measurement point as their endpoint, Z_k at that current measurement point is calculated only once, and the same value is used for both intervals. The impedance difference is calculated using the formula ΔZ_k = Z_k - Z_(k-1). The impedance difference distribution is formed by arranging ΔZ_k of all candidate intervals in the order of the current measurement points. The positive and negative signs retain the original calculation results, reflecting whether the impedance increases or decreases sequentially along the current measurement points. For branches where the phase difference distribution diagram does not generate a reversal judgment curve, the impedance difference distribution degenerates into a single difference between the two current measurement points. Candidate intervals where the absolute value of ΔZ_k exceeds the upper limit of the impedance baseline fluctuation of the branch are marked as exceeding the limit in the impedance difference distribution.

[0031] The grounding point coordinates are obtained by using impedance difference distribution to locate the symmetrical superposition region grounding point triangle. When the ΔZ_k amplitude of two adjacent over-limit intervals in the impedance difference distribution shows a symmetrical distribution with decreasing amplitude as distance, this is determined to be a symmetrical superposition region, and the grounding point is located within the symmetrical superposition region. The symmetry criterion is that the ratio of the attenuation rates on both sides falls within the range of 0.7 to 1.3. The attenuation rate is calculated based on the number of sampling points in each interval from the over-limit point back to the baseline. The number of sampling points used for calculation is not less than 3; intervals with fewer than 3 sampling points are considered insufficient and are skipped. If the back-falling process from the over-limit point is interrupted by an adjacent disturbance, the attenuation rate is recalculated based on the number of effective sampling points before the interruption; if both sides are interrupted, the symmetrical superposition region is not determined in this case, and the data is retrieved again after the next disturbance. The amplitude of ΔZ_k decreases with increasing distance from the grounding point, so the end with the larger amplitude is closer to the grounding point. Let the distance between the flow measurement points at both ends of the interval be L, and let the larger of the amplitudes of ΔZ_k at both ends be a and the smaller be b. The distance from the grounding point to end a is taken as d = L·b / (a+b), where L and d are both in meters. The interval number of the flow measurement point where the symmetrical superposition area is located and the estimated distance d together constitute the grounding point coordinates. The two together uniquely determine a physical location, which is considered the final location result for this evaluation cycle once generated. When multiple symmetrical superposition areas appear simultaneously in the current evaluation of the same branch, the grounding point coordinates are recorded separately for each symmetrical superposition area, arranged in the order of the flow measurement point interval number. When the over-limit interval in the impedance difference distribution does not constitute a symmetrical attenuation relationship, it is not judged as a symmetrical superposition area. This over-limit interval is marked as a single-sided anomaly, and the grounding point coordinate calculation skips this interval. In the special case where the values ​​of a and b are equal, d is directly taken as half of L, and the grounding point coordinates fall at the midpoint of the interval.

[0032] Suspected grounding sections are identified by mapping the grounding point coordinates to the enclosure drawings. The enclosure drawings pre-mark the corresponding physical coordinate ranges according to the current measurement point intervals, including the positions of cable heads, supports, and terminals within that interval. The drawing section identification follows a unified enclosure numbering system, consisting of the enclosure number plus the section number. Within the same enclosure, the section number is sequentially numbered starting from the incoming line end. Subsequent grounding point coordinate location results are mapped to this numbering system. The current measurement point interval number in the grounding point coordinates is used to locate the specific physical section on the drawing. The precise location within the section is determined by linear interpolation along the cable route of the section, based on the estimated distance 'd' in the grounding point coordinates. Electrical measurements provide the distance along the line, while field operations rely on support and terminal numbers. A conversion from the drawing is necessary to determine the verifiable physical location. The interpolation result is accurate to the nearest cable support number, with a positioning error not exceeding half the distance between adjacent supports, typically between 0.3 and 0.8 meters, meeting the needs of field verification. The corresponding drawing section identification and the nearest support number are included in the suspected grounding section location. If the previous sub-step did not generate grounding point coordinates, or if only one-sided abnormal branches are marked in the out-of-limit section, the location of the suspected grounding segment should be changed to the entire current measurement interval covered by the impedance difference distribution as the segment range, and the start and end current measurement point numbers should be recorded instead of the support numbers. If the drawing calibration information is missing for a segment, only the current measurement point interval number should be recorded for the location of the suspected grounding segment corresponding to the missing segment, and the drawing segment identifier should be left blank and manually entered after the measurement is completed. The drawing segment identifier and support number remain unchanged throughout the entire life cycle of the enclosure. After the enclosure is modified, the drawings must be updated synchronously and the physical coordinate range of each interval must be recalibrated.

[0033] The confidence level of dual-point grounding is determined by measuring abnormally high resistance at the location of suspected grounding segments. An online loop resistance measurement device is connected to the nearest support corresponding to the suspected grounding segment to measure the local sheath-to-ground loop resistance. The sampling period is 5 minutes, and the measurement is compared with the baseline loop resistance under historical normal single-point grounding conditions for that support. After the second grounding point is formed, a continuous circulating current will appear in the sheath (e.g., due to accidental contact with the nearest grounding electrode after the outer sheath is cut, or due to repeated grounding of the sheath grounding wire during construction). Its direction is exactly opposite to the test current injected by the online device. The superposition of the two causes the loop resistance measured by the voltage-current ratio to be higher than the true value, i.e., a false high resistance. The magnitude of the false high resistance is calculated using the formula ξ=(R_t-R_base) / R_base, where R_t is the measured loop resistance, R_base is the historical normal operating baseline for that support, and ξ is dimensionless and expressed as a percentage. The confidence level for dual-point grounding is calculated using the formula C=max(0,min(1,ξ / ξ_ref)), where ξ_ref is the average false height amplitude confirmed during the dual-point grounding simulation test of this type of enclosure, and C is a dimensionless quantity ranging from 0 to 1. R_base is calibrated separately for each bracket. Different brackets have different grounding grid connection methods and burial depths, and baselines cannot be borrowed from each other. For newly added brackets, the initial value is temporarily used based on manual measurements. When the bracket corresponding to a suspected grounding section does not have online resistance measurement conditions, the resistance is periodically checked manually to replace online data, and the confidence level for dual-point grounding in this case is noted as being obtained from manual checks. When ξ is negative, C is automatically set to 0 according to the formula, and the original ξ value is retained; when ξ exceeds ξ_ref, C is automatically capped at 1 according to the formula. In this case, ξ itself reflects the grounding strength and is recorded for future reference. The results of successive dual-point grounding confidence level measurements are recorded in the order of bracket number, without using a moving average.

[0034] To address the issue of identifying potential dual-point grounding hazards by determining the grounding level during periods of sustained high confidence exceeding the threshold, a high-confidence segment is defined as follows: When the dual-point grounding confidence level C exceeds a set threshold (0.6, derived from the upper limit of confidence of spurious grounding samples in a dual-point grounding simulation test for this type of enclosure) for three consecutive sampling periods, it is marked as the start of a high-confidence segment; when C falls below the threshold and does not exceed it again for three consecutive sampling periods, it is marked as the end of the high-confidence segment. A high-confidence segment lasting no more than 30 minutes is considered an occasional fluctuation in resistance measurement. Once a true grounding occurs, it persists. Even after a lightning strike seals the grounding point with an electric arc, or construction machinery cuts through the cable sheath, creating a metallic path, the grounding state will not automatically disappear. Measurement deviations and environmental interference are unpredictable, such as brief induced coupling caused by nearby heavy load switching. Therefore, the duration of these interferences is more effective than the confidence amplitude of a single double-point grounding event in distinguishing between the two. For high-confidence segments lasting more than 30 minutes, the grounding level is categorized by duration: greater than 30 minutes but not exceeding 120 minutes is considered the initial stage, and more than 120 minutes is the confirmed stage. The categorization boundary is derived from the confirmation rate of samples with different durations in historical statistics. The proportion of confirmed stage grounding ultimately confirmed as true double-point grounding exceeds 85%, while the initial stage is approximately 40%. Double-point grounding hazards are determined by the current grounding level of the branch and the start and end times of the corresponding high-confidence segment. Once the record is generated, the grounding level remains unchanged within the current evaluation period and does not escalate with newly added sampling points exceeding the threshold within the period. When multiple high-confidence segments meeting the criteria appear on the same branch within the current assessment period, the segment with the highest grounding level is taken as the representative for a two-point grounding hazard, and the remaining segments are separately included in the grounding anomaly log for that branch. When a two-point grounding hazard occurs repeatedly on the same branch and remains at the confirmed level, the log will additionally record the cumulative number of confirmed levels for that branch.

[0035] In some embodiments, the step of determining the floating discrimination threshold by analyzing the attenuation characteristics of the two-point grounding hazard includes: extracting the attenuation tail of the grounding disturbance based on the two-point grounding hazard to obtain attenuation characteristic values; using the attenuation characteristic values ​​to perform a loosening degree classification of the grounding disturbance waveform symmetry deviation to form a loosening classification result; checking the loosening classification result for long-term high confidence segment stagnation and misjudgment to obtain an effective loosening level; and determining the floating discrimination threshold by converting the upper and lower limits of the threshold range according to the effective loosening level.

[0036] The attenuation characteristic value is obtained by extracting the tail of the grounding disturbance attenuation based on the double-point grounding hazard. Within the high-confidence segment corresponding to the double-point grounding hazard, the moment when the load undergoes a step change within this segment is selected as the disturbance start point. From the disturbance start point, the online device reduces the sampling period of the loop resistance from 5 minutes to 1 second, and the resistance time-series curve after the disturbance start point is obtained. After the disturbance start point, the resistance first shows a significant jump, then gradually attenuates and tends to stabilize. The jump amplitude is the difference between the average resistance values ​​of the three 5-minute sampling points before the disturbance start point and the three 1-second sampling points after the start point. The moment when the attenuation curve falls below 10% of the jump amplitude is recorded as the end of the attenuation body. The interval from then until the curve stabilizes within the normal operating resistance fluctuation range (3 times the standard deviation range) for 5 consecutive sampling points is recorded as the tail segment. The attenuation characteristic value is denoted as F = T_tail·σ_R, where T_tail is the tail segment duration (seconds), σ_R is the standard deviation of the resistance within the tail segment (ohms), and F is in ohm-seconds; the tail segment duration of metallic stable grounding is usually no more than 2 minutes. Metallic grounding has a constant contact area, and its resistance stabilizes immediately after a disturbance. Loose contacts, however, experience repeated switching on and off under the electrodynamic vibration of a large load current (e.g., intermittent separation of the contact surface after the pre-tightening force of the grounding terminal bolts decays due to thermal expansion and contraction). After a disturbance, the resistance continuously jumps within a small range, thus lengthening the trailing segment, while the amplitude of the jump itself is not significantly different. If a high-confidence segment corresponding to a double-point grounding hazard spans one enclosure inspection and maintenance operation, the trailing segments are extracted from the resistance curves before and after maintenance. If no load step change occurs within the high-confidence segment of this branch during the current assessment period, the operational disturbance generated by the most recent commissioning or shutdown operation is used instead. The calculation methods for the attenuation characteristic values ​​of the two sources are completely consistent, only the source identification differs. When multiple usable disturbances occur within the same high-confidence segment, the maximum value among them is taken as the representative value for that segment.

[0037] The attenuation characteristic value is used to classify the looseness of the grounding disturbance waveform symmetry deviation, resulting in a looseness classification. The attenuation characteristic value describes the length of the tail, and this step simultaneously introduces the deviation value D_sym, which describes the waveform symmetry, to jointly classify it. The resistance curve of the main disturbance stage is folded in half at the moment of the jump peak. The absolute values ​​of the differences between the corresponding positions of the rising and falling segments are summed and divided by the number of differences involved in pairing to obtain the average symmetry deviation D_sym, in ohms. The folding window takes the sampling interval of equal length before and after the jump peak. If the lengths on both sides are unequal, the shorter side is symmetrically truncated. If the number of sampling points is odd, causing the center point to be unable to be paired, the center point is discarded separately. Grounding points with good contact follow the same path before and after the disturbance, and the waveform rise and fall are approximately mirror images. When the contact is loose, the on / off state changes midway through the disturbance, and the two sides are no longer symmetrical. The D_sym of a stable metallic grounding is usually less than 0.05 ohms. The attenuation characteristic value and D_sym are combined and looked up according to a preset grading table: if both values ​​are below their respective grading limits, it is considered slightly loose; if either value reaches the middle grading limit but not the higher grading limit, it is considered moderately loose; and if either value reaches the higher grading limit, it is considered severely loose. The grading table is determined by statistical analysis of samples with different degrees of looseness in the loosening tests of bolts and grounding terminals of this type of bracket. The current grading level of this branch is the core content of the loosening grading result, used for subsequent threshold conversion, and archived by branch for direct reference when comparing the grading levels of multiple high-confidence segments. If the grading level obtained by the same branch in this assessment is lower than its historical highest grading record, the loosening grading result will still record the current grading and set a grade decline indicator, while the historical highest grading will be kept as a separate record.

[0038] The loosening classification results are used to identify and eliminate false positives for stuck loosening sections that have remained unchanged for a long time due to high confidence levels. If the loosening classification result for the same suspected grounding section on the same branch remains unchanged for six consecutive assessment periods, and all sections are classified as moderate loosening or above, then that section is considered a candidate for false positive due to stuck loosening. The criterion for stuck loosening is based on the historical statistical patterns of structural grounding points within the enclosure. The classification results of structural grounding points are usually constant over a long period, while the actual loosening progress often shows fluctuations in classification within a few months. The judgment window of six consecutive assessment periods is longer than this time span, which avoids mistakenly eliminating slowly progressing loosening sections. A common situation in the field is that the grounding point at the fixed support of the enclosure remains at the same level for a long time due to structural reasons, and its loosening classification result remains almost unchanged period by period. This is significantly different from the trajectory of actual loosening gradually developing and the classification level increasing accordingly. This is a typical source of candidate candidates for false positive due to stuck loosening. Further verification is conducted on candidates for misjudged grounding to determine if they correspond to the design grounding point locations marked on the enclosure drawings. The threshold for location matching is that the nearest bracket number between the design grounding point marked on the drawings and the suspected grounding segment location should not exceed one bracket. The verification scope covers all design grounding points marked on the drawings for this branch, not just the one initially located in this high-confidence segment. Locations matching are removed from this assessment, marked only with an exclusion label; locations not matching but with a consistently high classification are judged as persistent loosening, recorded according to the original classification, and marked with a long-term unchanged label. The effective loosening level is the highest classification level among all high-confidence segments that were not removed during this assessment cycle for this branch. When all high-confidence segments are removed, it is recorded as no effective loosening, and this judgment result is also included in the assessment records of this branch in all previous assessments. The effective loosening level is retained for each assessment cycle, with each linked to the high-confidence segment number participating in this assessment.

[0039] The floating judgment threshold is determined based on the upper and lower limits of the threshold range converted from the effective loosening level. When the effective loosening level is no effective loosening, the threshold range follows the fixed benchmark set by the factory for that branch, without any floating correction. When the level is slight, moderate, or severe, correction coefficients of 0.1, 0.2, and 0.4 are added to the fixed benchmark, respectively. The looser the contact (the contact resistance may experience short-term jumps of tens of milliohms under wind vibration or alternating load current stress), the greater the fluctuation of the measured value under the same grounding condition. Using the original threshold would misjudge such fluctuations as changes in state. Therefore, the judgment range is widened according to the degree of loosening. The correction coefficients do not increase linearly in the three levels. The increase from moderate to severe is greater than that from slight to moderate, because the randomness of contact continuity increases faster after loosening enters severe. The values ​​of the three levels are fixed through field regression and are not fine-tuned with single measurements. The upper limit of the threshold interval is calculated using the formula T_high = T_0 × (1 + k_T), and the lower limit is calculated using the formula T_low = T_0 × (1 - k_T). T_0 is the fixed benchmark set at the factory, which is the judgment threshold of 0.6 for the confidence level of dual-point grounding mentioned above. k_T is the correction coefficient for the corresponding level. Both are dimensionless. T_high is capped at 1 (T_high / T_low / T_0 are dimensionless thresholds, not time quantities, and are different from T_tail mentioned above). The floating discrimination threshold is composed of the current T_high, T_low of the branch and the corresponding effective loosening level. The benchmark T_0 and the correction coefficient k_T are included together. Each conversion of the floating discrimination threshold covers the previous value, and only the latest version of the interval is retained for the same branch. When the difference between the monthly average temperature of the current month and the monthly average temperature of the month specified by the factory T_0 exceeds 15 degrees Celsius, T_0 is first adjusted according to the seasonal correction table before being substituted into the formula. The correction table is reviewed every three years. The correction factor k_T is calibrated based on the branch rather than the entire enclosure.

[0040] Step S103: Based on the floating discrimination threshold, identify abnormal states and construct abnormal discrimination results. Perform non-reproducible fault timing analysis on the abnormal discrimination results to obtain the location of the fault branch. Screen for false alarm blind spots and output potential abnormal signals for the location of the fault branch.

[0041] Specifically, an abnormal state identification result is constructed by identifying abnormal states based on a floating discrimination threshold. The T_high and T_low interval given by the floating discrimination threshold is used to compare with the newly measured two-point grounding confidence C of the branch each time thereafter. The measurement period of C is consistent with the on-loop resistance online measurement period, which is once every 5 minutes. A high C indicates that the conduction of the grounding path is enhanced and the circulating current is increased, while a low C indicates that the path changes from continuous conduction to intermittent conduction, both of which deviate from the stable state; when C exceeds T_high, or C directly drops from above T_high to below T_low within a single period, it is determined as an abnormal state. The abnormal state is calculated according to the formula flag(t)=1 (when C(t)>T_high, or C(t)<T_low and C(t-1)≥T_high), where flag(t) takes the value 0 or 1, all time points with a value of 1 are retained, and time points with a value of 0 are not retained separately. The comparison between C and T_high or T_low is performed independently at each 5-minute sampling moment, and no sliding average is performed across points. Real intermittent degradation often only lasts for one or two sampling periods, and sliding average will smooth out such short-time over-limit conditions, which will cause intermittent faults to be missed at this step. The abnormal discrimination result is formed by summarizing all time points with flag(t)=1 within the current evaluation period of the branch in chronological order of occurrence. Each time point is associated with a corresponding trigger direction, which is divided into two categories: high deviation and low deviation, and is transmitted to the subsequent time sequence analysis together with the time point. When the floating discrimination threshold is updated in the current period, the old interval is used for the comparison before the update, and the new interval is adopted after the update. Abnormal states generated by the two versions of intervals are allowed to be mixed in the same abnormal discrimination result. A branch with an empty abnormal discrimination result is regarded as operating normally in the current period.

[0042] In some embodiments, performing non-reproducible fault time sequence analysis on the abnormal discrimination result to obtain a fault branch position includes: analyzing the pre-stable duration of the intermittent fault time sequence through the abnormal discrimination result to form a working condition feature label; performing simultaneous no-alarm common cause analysis according to the working condition feature label to generate a homologous fault group; using the homologous fault group to assign a recurrence count traceability weight to obtain a traceability weighting result; and determining the branch position with the highest weight based on the traceability weighting result to obtain the fault branch position.

[0043] The pre-stabilization duration of intermittent faults is analyzed based on the anomaly detection results to form a condition characteristic label. For each anomalous state time point in the anomaly detection results, the original C value sequence of that branch is traced back from the dual-point grounding confidence archive. The tracing granularity is consistent with the C measurement period, and the total tracing length is 4 hours, covering the complete alternation of peak and trough loads, ensuring that the pre-stabilization duration is not affected by a single load. Starting from the tracing point, a window of 6 sampling points is slid along the time axis point by point. The rolling standard deviation of the C value within the window is calculated at each step. The longest continuous duration for which the standard deviation is continuously lower than a set threshold (1.2 times the historical normal fluctuation standard deviation of C for that branch, calibrated in the same way as the residual stationarity index σ_baseline) is recorded as the pre-stabilization duration of the anomalous state. The pre-stabilization duration is categorized as follows: less than 10 minutes is short-term stability, greater than 10 minutes but not exceeding 60 minutes is medium-term stability, and more than 60 minutes is long-term stability. Short-term stability indicates that the anomaly's confidence level fluctuates, with disturbances primarily originating from external instantaneous impacts (such as the starting of a nearby branch motor or lightning strikes coupled through the grounding grid). Medium-term stability indicates that the contact state deteriorates only after a brief change, still a gradual process but progressing rapidly. Long-term stability indicates that the contact state deteriorates suddenly only after a long period of stability, consistent with the gradual loosening process of bolt preload slowly decaying to a critical point. The classification result corresponding to this anomaly state is recorded in the operating condition feature label. When multiple anomaly states on the same branch have inconsistent classifications, the operating condition feature labels are recorded separately without merging. For anomalies with a low trigger direction in the anomaly identification results, the pre-stability duration is calculated using the same method, and the operating condition feature label additionally records the trigger direction, listed separately from the labels for high trigger directions. When the traceable record is insufficient for the required window duration, it is calculated based on the actual traceable duration and associated with the insufficient window indicator; this classification is for reference only.

[0044] Based on the operating condition characteristic labels, a common-source fault group is generated through synchronous no-alarm common-cause analysis. The operating condition characteristic labels are categorized into short-stable abnormal states. The status of other branches within the same enclosure is checked within the same time window (5 minutes before and after): both branches exhibiting the same short-stable abnormality and branches without alarms but with C synchronous rise not exceeding the limit are counted. The 5-minute window width is selected based on the duration of common load change events for this type of enclosure. If the total of both types of branches meets the standard, it is determined that the abnormality was triggered by the same external factor. When 3 or more branches within the same enclosure hit the same time window, all hitting branches are grouped into the same common-source fault group and the entire group is labeled with the external common cause. If there are branches with historical confirmation records within the group, proceed to the next sub-step to assign tracing weights; if there are no historical confirmation records for any branches within the group, the branch with the largest original C value in the current time is directly taken as the fault branch location. Because the scope of the inspection is limited to the same enclosure, and all branches within the same enclosure share a common incoming line and the same auxiliary busbar, external factors such as sudden load changes and incoming line voltage drops are simultaneously transmitted to all branches via the busbar. Different enclosures have different incoming lines and do not have this shared channel. Branches with historical confirmation records within the hit window are still included in the group for judgment. If the same branch falls into different same-source fault groups within the same assessment cycle, the two groupings are retained separately and not merged into one group. Same-source fault groups are only judged under the short-term stability category. Abnormal states with operating condition characteristic labels of medium-term stability and long-term stability directly constitute a same-source fault group containing only one branch. Gradual loosening itself does not have a physical mechanism to be triggered by external factors to multiple branches simultaneously. When the total number of branches in the same enclosure is less than 3, the same-source fault group under short-term stability will not be judged for that enclosure in the current assessment cycle, while medium-term and long-term stability will still be grouped separately according to the aforementioned rules.

[0045] The source-weighted result is obtained by assigning recurrence count weights to fault groups based on their origin. For each branch within a fault group, the weight is determined by the cumulative number of times a fault has been ultimately confirmed as a true two-point grounding fault in its historical records. The statistical scope includes confirmation records from the most recent 24 months; records past this period are not included. Higher recurrence counts result in higher weights. The weights are calculated using the formula w_i = n_i / Σn_j, where n_i is the historical confirmation count of the i-th branch in the group, j iterates through all branches in the group, Σn_j is the sum of their confirmation counts, and w_i is a dimensionless quantity. The sum of weights within the group is always 1. When all branches in the group have zero historical confirmation counts, the weights are reassigned based on the relative magnitude of the original C value for each branch in the current fault. Higher C values ​​result in higher weights. The formula structure remains the same, only n_i is replaced with the corresponding C value. The 24-month statistical window corresponds to the major overhaul cycle of the enclosure. A window that is too short will allow occasional, one-off confirmations to dominate the weighting, while a window that is too long will re-include historical issues that have already been rectified. Both deviations will cause the source tracing weighting results to deviate from the current actual state. The fault group number, the branch numbers within the group, and the corresponding source weight w_i are all included in the source tracing weighting results, arranged from highest to lowest weight. Fault groups with only a single branch have a weight of 1. If there is more than one branch with a weight difference less than 0.05 in the source tracing weighting results, all parallel branches are retained in the results. The statistical source of n_i is the number of records in the branch's historical grounding anomaly log that were marked as "confirmed" and confirmed as genuine two-point grounding after on-site re-inspection. Only "confirmed" records are counted; those confirmed as genuine two-point grounding after on-site re-inspection are included, regardless of the system's judgment level. The weights of each branch in the source tracing weighting results are displayed to two decimal places, and sorting and comparison are based on the unrounded original values.

[0046] The location of the faulty branch is determined based on the highest weighted branch location in the source tracing weighting results. The branch number with the highest weight in the source tracing weighting results is directly used as the location of the faulty branch. The corresponding cabinet number and drawing section identifier are also included as the physical location basis for on-site verification. The drawing section identifier is taken from the corresponding field already recorded in the suspected grounding section location; if this field is blank, the current measurement point interval number is replaced. If there are more than one branch with a weight difference less than 0.05 in the source tracing weighting results, the location of the faulty branch is recorded for all parallel branches simultaneously, without mandatory selection. During on-site handling, the verification order is determined by the actual reachability of each parallel branch, and the reachability judgment is performed by on-site personnel. After the faulty branch location is generated, it is associated with the number of the fault group to which it belongs. The recorded content includes all branches within the group and their weight distribution. The location of the faulty branch corresponding to a fault group containing only a single branch is the location of that single branch itself; the determination process does not involve weight comparison. If the same branch is identified as a faulty branch multiple times within the same assessment cycle, each identification is retained separately, and the grounding anomaly log is accumulated according to the number of identifications. The on-site accessibility criteria for parallel branches include whether the enclosure maintenance window is open and whether there are other conflicting operations on site. Branches meeting both criteria are prioritized for verification; those meeting only one are second; and those not meeting both are postponed. The branch with the highest weight does not necessarily coincide with the actual faulty branch. The faulty branch location output in this step only serves as the starting point for on-site verification; the final conclusion is subject to re-inspection. Once the faulty branch location, along with its corresponding fault group number, is identified, it is passed down the handling process.

[0047] In some embodiments, the step of screening for false alarm blind spots and outputting potential abnormal signals at the location of the faulty branch includes: determining a false alarm feature set by analyzing the false alarm interval of the inspection false alarm blind spot based on the location of the faulty branch; classifying the blind spot level according to the inspection coverage and dwell time of the abnormal branch based on the false alarm feature set to form a blind spot classification result; planning the seasonal recurrence frequency to supplement inspection points according to the blind spot classification result to obtain a supplementary verification set; and conducting on-site re-inspection of the supplementary verification set to output potential abnormal signals.

[0048] False alarm feature set is determined by analyzing the false alarm blind zone and false alarm interval based on the location of the faulty branch. The nearest support to the suspected grounding section corresponding to the faulty branch location is examined. All historical inspection records of that support are reviewed, and all entries marked as false alarms are included in the extraction scope. The criteria for determining a false alarm are: no abnormally high loop resistance detected during on-site re-inspection, no visible signs of loose or oxidized terminals, and no further anomalies triggered at that location for the next three consecutive evaluation cycles. All three criteria must be met for an entry to be included in the extraction scope. Extracted false alarm entries are arranged chronologically, and the time interval between adjacent false alarms is recorded. If a valid confirmation record is interspersed between two false alarms, this confirmation is used as a segmentation point, and the interval between the two segments is calculated independently. The number of segmentation points is archived with the false alarm feature set to determine whether there has been a history of a valid fault at that location. The false alarm interval is calculated using the mean and standard deviation: the mean reflects the average occurrence cycle of false alarms at that location, and the standard deviation reflects the stability of the false alarm occurrence pattern. A smaller standard deviation indicates that false alarms reproduce according to a fixed rhythm, and the triggering source is mostly periodic environmental changes rather than random faults. For example, condensation during the day and night forms a temporary parallel path in the measurement circuit, causing regular distortion of the resistance reading. This is precisely the source that the false alarm feature set attempts to capture. The total number of false alarms, the mean interval, and the standard deviation interval for the corresponding location of the branch are all included in the false alarm feature set. If the number of historical false alarms is less than 3, only the total number is recorded, and the mean and standard deviation are left blank. If the location of the faulty branch corresponds to multiple historical inspection locations, the false alarm feature set is statistically analyzed separately for each location, without merging across locations. Locations in the false alarm feature set where both the mean and standard deviation are valid and the standard deviation is less than 20% of the mean are marked with a regular false alarm label.

[0049] Blind spot classification results are generated by dividing the inspection coverage dwell time into levels based on the false alarm feature set. The false alarm feature set provides the pattern of false alarm occurrences, and this step focuses on examining the coverage quality of the inspection itself. For the corresponding location on the branch, all dwell time data for that location are retrieved from the inspection check-in records. The dwell time is automatically calculated based on the difference between the arrival and departure times of the inspector, without relying on manual entry. Only continuous dwell times between arrival and departure are counted; turnbacks are considered two independent dwell times and counted separately. Signs such as terminal oxidation and slightly loose bolts require dwell time to measure resistance and tighten each bolt individually to detect; shorter dwell times are more likely to be missed due to visual inspection. Therefore, the average dwell time of all instances is taken as the inspection coverage dwell time, serving as one of the main bases for determining the blind spot classification results. Blind spot levels are categorized by dwell time: an average dwell time of less than 2 minutes is classified as a severe blind spot; more than 2 minutes but less than 5 minutes is classified as a moderate blind spot; and more than 5 minutes is classified as a mild blind spot. The categorization thresholds are based on the minimum effective time required for single-point verification in the inspection standards for this type of enclosure. For locations with a consistent pattern of false alarms, the blind spot level is increased by one level from the original categorization; blind spots already classified as severe remain unchanged. Consistent false alarms indicate that the current inspection method has not yet identified the source of interference, requiring a longer on-site verification time for accurate identification. The blind spot level for the corresponding location on this branch is thus determined as the blind spot classification result. If there are no historical inspection records for this location, the blind spot level is directly classified as severe, and the blind spot classification result is further annotated with "no historical record." A total of zero false alarms indicates that the location has been inspected but never falsely alarmed; this is not the same as never having been inspected. The two are determined based on different criteria and are associated with their respective original situation indicators.

[0050] Based on the blind zone classification results, supplementary inspection points were planned to account for seasonal recurrence frequency, resulting in a supplementary verification set. For locations with blind zone classification results of moderate or higher, the monthly recurrence frequency distribution in the branch's grounding anomaly log was checked. Locations where the number of recurrences within the month before and after the current month accounted for more than half of the total annual recurrences were considered to have a seasonal pattern and were included in this supplementary inspection point. Locations with fewer than 6 months of grounding anomaly log entries were temporarily deferred for seasonality and included according to the default rules for moderate or higher blind zones, without being excluded due to insufficient sample size. The diurnal temperature range is greatest before and after seasonal changes, condensation is most frequent inside the enclosure, and the expansion and contraction of metal parts and bolt preload fluctuations are most pronounced. Therefore, the recurrence peaks for sealing failures and grounding loosening issues fall within similar time periods, and the same window width is used. Locations with blind zone classification results of mild blind zones maintained their original inspection frequency. The supplementary verification set records all included locations, with each location associated with its corresponding blind zone level and the seasonal recurrence frequency value that triggered its inclusion. Locations temporarily deferred for seasonality were marked with a separate deferred flag. Locations identified as severely blind spots in the blind spot grading results will be included in the supplementary verification set, regardless of whether seasonal recurrences are concentrated. The arrangement of supplementary inspection points will not change the overall frequency of the original inspection routes; only additional stops and verification items will be added to the established routes to avoid organizing separate vehicle deployments for single-point verifications. If the same location was already included in a previous assessment cycle and confirmed as abnormal in the current re-inspection, it will be reassessed for inclusion under the current conditions. When the supplementary verification set is empty, the judgment results corresponding to the faulty branch location will directly enter the on-site re-inspection stage.

[0051] For the supplementary verification set, on-site re-inspection is conducted to output potential abnormal signals. Each location in the supplementary verification set undergoes on-site manual re-inspection. The re-inspection includes visually inspecting the oxidation of wiring terminals, manually measuring loop resistance, and verifying the enclosure's sealing status. If no abnormalities are found in any of the three items, it is recorded as confirmed normal; if any item shows signs of abnormality, it is recorded as confirmed abnormal. The judgment results for each of the three items are recorded independently. If a certain item requires manual re-inspection, it is marked as pending re-inspection. For other items with confirmed abnormalities, the conclusion remains confirmed abnormal. If all remaining items are normal, no conclusion is drawn yet; items pending re-inspection are separately marked in the supplementary verification set record. The measuring tools used for on-site re-inspection are the same model as the online measuring device to ensure that the re-inspection readings are consistent with the historical data used by the system for judgment. For locations with disputed results, on-site personnel can request a second person to review them. If the two conclusions differ, the more serious one prevails to avoid overlooking minor abnormalities due to subjective judgment by a single person. Locations with a confirmed abnormality conclusion, along with the corresponding fault branch location, the corresponding fault group number, and the specific phenomena found during the re-inspection, are summarized as potential abnormal signals. If all locations in the supplementary verification set show a confirmed normal result during re-inspection, this round of judgment is recorded as a pending false alarm and written back into the historical inspection record of that location. Only after meeting the three criteria for false alarm judgment will it be included in the total number of false alarms and interval statistics. If multiple locations on the same branch show confirmed abnormalities during re-inspection, the specific phenomena for each location are listed separately, not combined into a single general description. Re-inspection tools are uniformly calibrated before each round of verification, and re-inspection is prioritized for locations with the most severe blind spots. Locations where verification cannot be completed due to on-site conditions are marked as having a delayed re-inspection.

[0052] Step S104: Isolate the faulty branch according to the potential abnormal signal to obtain the branch isolation command, switch the healthy branch according to the branch isolation command to form the switching and handling result, and perform residual pressure hysteresis concealed bypass screening to generate bypass concealment mark through the switching and handling result.

[0053] Specifically, a branch isolation command is obtained by isolating the faulty branch based on a potential abnormal signal. The branch number and location information carried in the potential abnormal signal are mapped to the corresponding branch switch number. The mapping relationship is retrieved from the branch-to-switch lookup table established when the enclosure was put into operation. Each branch corresponds to a unique branch switch, and there is no many-to-one situation. Before generating the command, the pre-planned backup branch allocation table is checked to see if there is an available transfer path for the branch. If there is no available path, manual confirmation is required before deciding whether to force isolation, to avoid the healthy load losing power along with the faulty branch. After the verification is passed, if... Figure 2The branch isolation command is generated according to the disconnector switch number 2 corresponding to the faulty branch 1. The command content includes three items: switch number, target action, and issuance time, and is accompanied by a unique command number for subsequent cross-box verification and traceability. After the branch isolation command is issued, the local control unit of the switch executes the trip and sends back the result. If the switch execution result feedback times out (more than 10 seconds without feedback), the branch isolation command is marked as unconfirmed, and the system reissues it once every 30 seconds, with a maximum of 3 retries. If it is still unconfirmed after 3 retries, the branch isolation command is switched to manual intervention and remains in transit. The system stops automatic retries, and the switch status is manually confirmed by on-site personnel before being cancelled. The command issuance channel and the data acquisition channel are physically separated. Command feedback and data transmission do not occupy bandwidth, and the isolation action itself does not change the operation mode of the backup branch. When the same branch corresponds to multiple locations in the potential abnormal signal, only one branch isolation command is generated; the other locations in the branch that are still marked as re-inspection delayed and not yet included in the potential abnormal signal will be judged after the re-inspection is completed.

[0054] Based on the branch isolation command, the system switches to a healthy branch to generate a switching result. After the branch isolation command is confirmed, residual voltage sampling is triggered synchronously with the tripping signal. The system simultaneously checks the list of healthy loads originally connected to the branch, and switches the connection of each load along the load switching path 5 according to the preset standby branch allocation table. The switching action is issued in the form of a tie switch closing command. The command content includes the switch number, target action, and issuance time, similar to the branch isolation command. This is achieved by closing the tie switch 3 corresponding to the standby branch 4. The standby branch allocation table is established according to the load transfer relationship pre-planned when the enclosure is put into operation. Each branch corresponds to one or two standby branches 4, and the transfer relationship remains fixed in the long term as long as the enclosure wiring remains unchanged. The switching is executed item by item according to the load. After each switching is completed, the actual position of the tie switch 3 is read back to confirm that it is closed before switching to the next item, avoiding instantaneous overload of standby branch 4 caused by the simultaneous transfer of multiple loads. The load list for this switchover, the standby branch number to which each load was switched, and the switchover completion time together constitute the switchover handling result. The switchover completion time is the time of confirmation of the last load readback, used as a time reference for retrieving the residual voltage curve. When standby capacity is insufficient, overall optimal allocation is not performed. Since power transfer must be completed within a second window after the branch isolation command is confirmed and the circuit breaker is tripped, calculating each item individually would prolong the outage time. Therefore, loads with high priority (6) are switched sequentially from high to low until full. Loads with low priority (7) are not transferred; that is, load 9, which remains on the original branch in the diagram, is de-energized along with the faulty branch. The switchover handling result is correspondingly marked with an insufficient capacity indicator (8), indicating that this round of power transfer did not cover all loads. When the branch isolation command is in the manual intervention indicator state, the switchover handling result is temporarily suspended, awaiting manual confirmation of the switch status before execution.

[0055] In some embodiments, the step of screening for hidden bypasses with residual voltage hysteresis based on the switching and switching results to generate a hidden bypass marker includes: identifying residual voltage attenuation hysteresis step-like pauses based on the switching and switching results to establish residual voltage attenuation characteristics; performing bypass impedance inversion analysis based on the residual voltage attenuation characteristics to generate a bypass impedance estimate; using the bypass impedance estimate to locate the hidden path in an abnormal segment below the normal insulation level to determine the location of the hidden path; and verifying bypass connectivity based on the location of the hidden path to generate a hidden bypass marker.

[0056] Based on the switching results, a sluggish, step-like pause in residual voltage attenuation is identified to establish residual voltage attenuation characteristics. The residual voltage monitoring point starts sampling synchronously with the disconnecting switch opening signal. Using the switching completion time recorded in the switching results as a reference, the monitoring point, composed of a capacitive voltage divider, is installed on the load side of the disconnecting switch for that branch. The sampling period is 10 milliseconds, recording the complete attenuation curve of the branch's residual voltage to ground from the opening moment. Under normal, complete isolation, the residual voltage attenuation conforms to the law of capacitor discharge, with the curve exhibiting an approximately exponential decay. The time constant is determined by the branch's capacitance to ground and insulation resistance, typically ranging from hundreds of milliseconds to several seconds. If there is still a conductive path connected to the energized part, the charge that should have been lost is quietly replenished by this path. The discharge is repeatedly interrupted and resumed, causing the curve to stagnate, resulting in a near-stagnant step-like decline. Common causes of concealed bypasses in the field include weak connections formed between the cable insulation layer and adjacent metal components due to localized aging, or the failure to remove temporary grounding clamps mistakenly left during maintenance. Both situations will exhibit the same step-like stagnation characteristic on the residual voltage curve. If the measured voltage drop within 10 consecutive sampling points on the curve is less than 30% of the theoretical voltage drop of the same period's exponential decay, and this segment is not located in the flat area at the end of the curve, it is judged as a stepped pause. The theoretical voltage drop is calculated based on the branch's own decay time constant. The time of occurrence of the stepped pause is recorded in the residual voltage decay characteristic. For branches that do not experience stepped pauses throughout the entire process, the residual voltage decay characteristic is recorded as having no abnormal pause, and this isolation is judged as complete isolation. If a pause occurs more than once, the residual voltage decay characteristic records the value of each pause separately. For branches with insufficient capacity indicators associated with the switching treatment results, residual voltage monitoring is still performed normally.

[0057] For example, the step of generating a bypass impedance estimate by performing bypass impedance inversion analysis based on the residual voltage attenuation characteristics includes: performing high-frequency residual frequency domain separation analysis on the residual voltage attenuation characteristics to obtain a frequency-divided attenuation sequence; calculating equivalent capacitance and resistance parameters based on the frequency-divided attenuation sequence to determine equivalent bypass parameters; performing strength rating on components deviating from the normal insulation range according to the equivalent bypass parameters to form a bypass strength level; and using the bypass strength level to correct the equivalent bypass parameters to generate a bypass impedance estimate.

[0058] Frequency-domain separation analysis of the residual voltage attenuation characteristics yields a frequency-divided attenuation sequence. Based on the pause times recorded in the residual voltage attenuation characteristics, a 200-millisecond interval before and after the pause is taken from the original sampling records of the residual voltage monitoring points and subjected to Fourier transform. This 400-millisecond interval corresponds to 40 sampling points at a 10-millisecond sampling period, satisfying the minimum window length required for frequency domain analysis and avoiding insufficient frequency resolution and confusion between true bypass coupling and random noise due to an excessively short window. The charging process of the bypass path is much faster than the branch's own discharge; the two are located at high and low segments in the frequency domain, and signals that are difficult to distinguish in the time domain curve can be directly separated in the frequency domain. The transformed spectrum is divided into low-frequency and high-frequency parts according to the frequency cutoff point (taken as 5 times the reciprocal of the branch capacitor discharge time constant, and not exceeding half of the upper limit corresponding to the sampling rate): the low-frequency component corresponds to the main attenuation trend of the branch's own capacitor discharge, and the high-frequency component corresponds to the rapid fluctuation residual caused by the coupling path. Spectral lines above the cutoff point are individually retained, while the rest are set to zero and then subjected to an inverse transform to restore the amplitude sequence of high-frequency components over time, i.e., the frequency-division attenuation sequence. When the amplitude of a high-frequency component exceeds three standard deviations of the background level outside the pause interval, it is determined that the pause is indeed related to high-frequency coupling, and the frequency-division attenuation sequence is marked with a high-frequency prominent label; those that do not exceed this limit still generate a frequency-division attenuation sequence but without this label. A transition region is added to both ends of the Fourier transform window, with a width of 10% of the total window length, to prevent spectral leakage introduced by boundary truncation from being misjudged as high-frequency remnants. For branches with only one pause in the entire curve, only the sequence corresponding to that pause is generated; branches with step-like pauses not marked in the residual voltage attenuation feature are skipped in subsequent equivalent parameter fitting steps.

[0059] The equivalent bypass parameters are determined by calculating the equivalent capacitance and resistance parameters based on the frequency-division decay sequence. For frequency-division decay sequences with significant high-frequency indicators, the decay law of their high-frequency component amplitudes over time is fitted using a single exponential decay model. Where t is the time interval between the sampling point and the initial sampling point of the fitting, in seconds; A_0 is the initial amplitude of the high-frequency component, in volts; and τ is the time constant obtained from the fitting, in seconds. The fitting adopts the least squares method, taking each sampling point of the frequency-division attenuation sequence as the granularity, calculating the sum of squares of the differences between the measured amplitude and the predicted value of the single exponential model, and taking the set with the smallest sum of squared residuals as the fitting result. The fitting interval includes all samples after the first point of the frequency-division attenuation sequence. The first point often has a higher initial positioning error due to the pause, and including it in the fitting will lengthen τ and systematically increase R_eq. The high-frequency component attenuates through the loop formed by the bypass path and the stray capacitance to ground of this branch. The equivalent resistance is calculated according to the formula R_eq=τ / C_s, where C_s is the typical value of the stray capacitance of this branch, taken from the factory capacitance calibration data of the same cross-section and length section of this type of cable, in farads; and R_eq is the equivalent resistance of the bypass path, in ohms. The fitted τ and R_eq, together with the fitting residuals, constitute the equivalent bypass parameters. The residuals are calculated by dividing the sum of squared residuals by the mean square error obtained from the number of samples in the frequency-division decay sequence. If the fitted τ is negative or significantly deviates from the physically reasonable range, the fitting is considered divergent, the equivalent bypass parameters are marked as fitting failure, and the R_eq field is set to empty. When the fitting residuals exceed the set upper limit, the equivalent bypass parameters are marked with a low confidence flag. When the number of samples in the frequency-division decay sequence is less than 10, a two-point estimation method is used (taking the amplitude and time interval of the first and last points of the sequence, and calculating τ and A_0 inversely according to the single exponential decay model). The equivalent bypass parameters are marked as simplified estimations. This explanation, along with the low confidence warning, is passed down to lower levels, ensuring that the bypass impedance estimate must be verified and confirmed on-site.

[0060] The bypass strength level is determined by classifying the components deviating from the normal insulation range according to the equivalent bypass parameters. The equivalent resistance R_eq in the equivalent bypass parameters is compared with the typical lower limit of the resistance to ground under normal insulation conditions for this type of cable. This typical lower limit is taken from the lowest qualified value given in the cable's factory insulation performance test report, and the same value is used for subsequent segmented insulation measurements. The greater the deviation of R_eq from this lower limit, the stronger the conductivity of the bypass coupling path and the more serious the hidden danger. The typical lower limit used for comparison is taken according to the cable cross-section; the larger the cross-section, the greater the capacitance to ground and the lower the normal insulation resistance. Using a uniform lower limit would systematically classify large-section branches as severely affected. The bypass strength is graded according to the proportion of R_eq deviating from the normal lower limit: deviation within 30% is considered minor; deviation exceeding 30% but not exceeding 60% is considered moderate; and deviation exceeding 60% is considered severe. The grading proportions refer to the measured deviation proportions corresponding to different bypass conductivity levels in the insulation aging simulation test of this type of cable. The grading results constitute the bypass strength level. When the equivalent bypass parameter is marked as "fitting failed," the bypass strength level cannot be calculated using the formula and is directly marked as "unable to be graded," with a "fitting failed" flag attached, and processed separately from the normal three levels. When the equivalent bypass parameter carries a low-confidence flag, the bypass strength level is still calculated and recorded using the original formula, and the flag is transmitted along with it. If multiple step-like pauses occur within the same branch during this evaluation period, each corresponding to different equivalent bypass parameters, after each pause completes its own strength grading process (deviation ratio calculation, grade lookup table, low-confidence judgment), the highest grade among them is taken as the representative level of that branch in this evaluation. The R_eq deviation ratio is calculated as (lower limit - R_eq) / lower limit, which is a dimensionless quantity expressed as a percentage. A negative value indicates that it is not lower than the normal lower limit and is marked as "no bypass."

[0061] Bypass impedance estimates are generated by correcting equivalent bypass parameters using bypass strength levels. Bypass strength levels are categorized into three grades: slight, moderate, and severe, corresponding to correction factors of 0.9, 1.0, and 1.15, respectively. These correction factors are determined statistically by comparing the ratio of the back-calculated impedance to the measured impedance at different strength levels during insulation aging simulation tests of this cable model. The bypass impedance estimate is calculated using the formula Z_bypass = R_eq × k_Z, where R_eq is the original value of the equivalent resistance in the equivalent bypass parameters, k_Z is the correction factor for the corresponding strength level (dimensionless), and Z_bypass is the generated bypass impedance estimate in ohms. The corrected Z_bypass value, the correction factor k_Z, and the corresponding bypass strength level are all stored in the bypass impedance estimate. If a branch's bypass strength level is marked as unclassifiable, the bypass impedance estimate cannot be generated; instead, a fitting failure flag is recorded, and the process proceeds to manual on-site verification without numerical estimation. When a bypass strength level is transmitted with a low-confidence flag, the bypass impedance estimate is also associated with this flag. When multiple bypass strength levels exist within the same branch during the current assessment period, the light, medium, and heavy levels each generate independent bypass impedance estimates. The results of the three levels are stored side-by-side, with the highest-scoring level taken as representative, ensuring that the representative level and representative impedance originate from the same pause. The bypass impedance estimate also records the original R_eq value from the equivalent bypass parameters substituted in this assessment, and saves it side-by-side with the corrected Z_bypass value. The strength range covered by the three correction factors is independent of the cable insulation class classification in the factory inspection report; the two systems are not forcibly mapped to avoid the existing insulation class classification interfering with the correction factor values ​​in this step.

[0062] By using bypass impedance estimation, concealed paths with insulation levels below normal are located, thus determining the location of concealed paths. Bypass impedance estimation identifies branches with potential bypass hazards. These branches are then divided into grid segments at 200mm intervals, with Z_bypass used as the expected value of the parallel path impedance at that point. Insulation resistance is measured segment by segment, with the segment spacing consistent with the joint grid spacing used in seal deterioration investigation. The measurement principle is changed from phase analysis of elastic hysteresis to direct measurement of insulation resistance. A portable insulation resistance meter is used, contacting the cable outer sheath measurement points segment by segment. Initially, a coarse measurement is performed in 10-grid units to identify abnormal segments, followed by detailed measurements within each segment to avoid uniform density measurements across the entire line. Segments with insulation resistance below the lower limit of normal insulation levels and on the same order of magnitude as Z_bypass are marked as abnormal segments. Adjacent abnormal segments are merged into continuous abnormal intervals, the interval length being the actual distance between the start and end segment numbers of the abnormal segment. The start and end segment numbers of the abnormal interval, along with the lowest insulation resistance value within the interval, are included in the location of the concealed path. If no abnormal sections are found after the entire branch circuit measurement is completed, the location of the concealed path is marked as unlocated. The potential hazard assessment given by the bypass impedance estimation is retained, but the physical location is temporarily missing, and the entire line is then manually inspected. If there are multiple discontinuous abnormal sections in the same branch circuit, the location of the concealed path is recorded for each section, and the section with the lowest insulation resistance value is taken as the most likely actual bypass location. The segmented measurement sequence is fixed, starting from one end of the branch circuit entering the enclosure and proceeding segment by segment to the other end. The measurement sequence does not affect the judgment result, only the time taken; the coordinates of abnormal sections are represented by the segment number followed by a hyphen.

[0063] Bypass connectivity is verified based on the location of the concealed path, generating a bypass concealment marker. For the abnormal section corresponding to the concealed path location, connectivity is verified on-site using a low-voltage detection method: a low-voltage test signal is applied to the isolated branch. The test signal is a 50Hz AC, extra-low voltage with an amplitude not exceeding 36 volts. The amplitude and phase of the response current are measured between the two ends of the abnormal section and the nearest grounding grid. If the resistive component of the response current in phase with the test voltage exceeds a set threshold, an electrically connected bypass is determined to exist at that location. The threshold is twice the normal capacitive current of the same cross-section and length section of the cable of that type. A single capacitive component indicates capacitive coupling, not a true bypass. Low insulation resistance only indicates insufficient insulation margin at that location and does not necessarily mean a continuous conductive path has been formed; therefore, final confirmation must be based on the actual current response. An abnormal section where no response current was detected by low-voltage detection is identified as localized insulation degradation without forming an actual bypass. No bypass concealment marker is generated; the section is simply recorded in the insulation weakness log for that branch. This log is stored separately from the bypass concealment marker for subsequent manual inspection. If a bypass is confirmed, the coordinates of the abnormal section corresponding to the concealed path location and the amplitude of the response current measured during this verification are both stored in the bypass concealment marker. If the concealed path location is recorded as an unlocated branch, a low-voltage test signal is applied to the entire branch, and the response current is measured segment by segment. Sections confirmed to have a response are also recorded in the bypass concealment marker. The upper limit of the low-voltage test signal amplitude is fixed at 36 volts, regardless of the branch voltage level, and is completely isolated from the normal operating voltage of the branch. When multiple bypasses are detected on the same branch, the bypass concealment marker is recorded separately for each location. After the bypass concealment marker is generated, the corresponding branch remains isolated; a separate closing command is required to restore power.

[0064] Step S105: Based on the bypass concealment flag, trace and collect data along the repeated fading process to generate scheduling and reporting data. Based on the scheduling and reporting data, perform cross-box instruction verification to form a set of handling tasks. For the set of handling tasks, perform long-term unverified branch verification and output isolation closed-loop instructions.

[0065] Specifically, based on the bypass concealed markers, the system traces and aggregates data along the repeated fading process to generate scheduling and reporting data. Each record in the branch grounding anomaly log includes six items: generation time, anomaly type, corresponding level, handling instruction number, elimination confirmation time, and on-site re-inspection confirmation mark. Sealing deterioration markers and double-point grounding hazards are recorded in this format, as are bypass concealed markers, with the corresponding level item left blank. After the bypass concealed marker is confirmed, the system searches backward along the timeline for all records of the bypass concealed markers corresponding to the same anomaly interval coordinates since the branch was put into operation. If the bypass concealed marker corresponding to the same coordinate appears two or more times in the records, it is judged as a repeated fading process, indicating that the previous handling failed to eradicate the root cause of the problem. If the branch was judged to have been completely repaired on-site between the two occurrences, it is also judged as a repeated fading, with the trajectory marked as repaired but recurring, and reported with priority. The repeated fading process is aggregated into a traceable trajectory in chronological order of occurrence, with each step recording the judgment time, the issued branch isolation instruction number, and the interval between the reappearance after fading. The dispatch report data consists of the current bypass concealment flag and tracing trajectory packaged according to the dispatch system's receiving format. The format includes four mandatory fields: branch number, fault type, recurrence count, and current handling recommendation. Seal deterioration flags and double-point grounding hazards are also generated independently when generated separately, with the fault type being the flag itself. When all three coexist, the leakage path risk of the bypass hazard is more urgent, so the latter takes priority. The handling recommendation is the switch number and target action of the current branch isolation command. When the abnormal interval coordinates first appear, the initial value of the recurrence count field in the dispatch report data is counted as 1. Subsequent triggers at the same coordinates increment the count. A report is initiated immediately after packaging.

[0066] Based on the dispatch report data, cross-container instruction verification is performed to form a task set for handling. The branch number carried in the dispatch report data is used to check whether there are any in-transit instructions that have not yet been completed in other branches within the same container and in adjacent containers. In-transit instructions include two types: branch isolation instructions and tie switch closing instructions. Instruction conflicts only occur between containers that can transfer power to each other. If tie switches are directly connected, they will compete for the same spare capacity. Adjacent containers without a tie path do not affect each other. Therefore, adjacent containers are determined according to the power supply topology diagram, not by physical geographical distance. When the target spare branch involved in any in-transit instruction coincides with the spare branch number recorded in the current switching handling result, the target actions of the two are compared item by item: if the target actions of the same switch are opposite, it is determined to be a state conflict; if the target actions are the same but the total load exceeds the capacity limit of the spare branch, it is determined to be a capacity contention. The capacity limit is the difference between the factory design capacity and the currently connected load. Both types of conflicts are decided based solely on the priority of the dispatching data associated with their respective instructions, without directly comparing the three flags themselves: sealing degradation, double-point grounding, and bypass concealment. In the dispatching data, fault type field values ​​of "bypass concealment" take precedence over "double-point grounding," and "double-point grounding" takes precedence over "sealing degradation." This order is consistent with the order in which the fault type field was determined when the dispatching data was generated. For fields with the same value, the number of recurrences is compared, with higher recurrences taking precedence. If still the same, the order of dispatch time determines the priority, and the decision criteria are recorded along with the handling task set for later review. Capacity contention does not involve a binary choice; capacity is allocated sequentially based on load priority identifiers within the above ranking. The handling task set consists of the instructions determined to be executed first after the decision, along with the handling schemes suggested in this dispatching data. The number, type, and target branch of each executable instruction are recorded, arranged from highest to lowest execution priority.

[0067] For the task set, long-unverified branches are reviewed and isolated in a closed-loop manner. After identifying the enclosures involved in the task set, the system checks the grounding anomaly log of all branches within the enclosure, identifying branches whose last record was generated more than 12 months ago and for which there has been no subsequent elimination confirmation record. The 12-month period is taken as the longest verification interval stipulated in the operation and maintenance procedures. For the branches included in the scope, the sealing deterioration mark, double-point grounding hazard, and bypass concealment mark generated in the latest assessment cycle are retrieved one by one. The levels recorded in the sealing deterioration mark, double-point grounding hazard, and grounding anomaly log are compared item by item. The bypass concealment mark does not have a level field; it is only compared whether it is still generated. If it is still generated and has a level of Class 1 or higher than the original record, it is classified as still having a problem. If it is no longer generated, or if the level has decreased after leakage rebound healing or oxidation circuit breaking, it is classified as having naturally resolved the problem. If it has never appeared in any cycle after the original judgment, lacks on-site re-inspection support, and no isolation rectification has been implemented during the period, it is classified as a misjudgment in the original judgment. Three types of results are recorded in the isolation closed-loop command verification record. Branches classified as misjudged are further investigated to determine if the misjudgment stems from an inappropriate threshold. If it does, the adjustment range is calculated by reverse calculation based on the deviation ratio between the measured value and the threshold at the time, ensuring the misjudgment is not triggered again. A single adjustment cannot exceed 20% of the original threshold. Adjustment objects are categorized by flag type: double-point grounding adjusts T_0, sealing degradation adjusts the sealing level threshold, and bypass concealment adjusts the strength threshold. The results are recorded in the corresponding branch's handling record within the handling task set. The isolation closed-loop command consists of the original command from this handling task set, along with the verification and threshold correction results mentioned above. Each command's type, execution time, and execution result are recorded by branch number—for example, the type is recorded as double-point grounding isolation, the execution time as 14:32:07, and the execution result as tripped and backup branch 4 supplying power normally.

[0068] To execute the methods described in the above embodiments and achieve the corresponding functions and technical effects, this application also provides the following system. See [link to system]. Figure 3 , Figure 3 This paper shows a structural block diagram of a low-voltage cable branch box monitoring and fault isolation system 300 provided in an embodiment of this application, including: Data acquisition module 301 is used to acquire real-time data of branch current and enclosure environment, identify insulation drop period based on the real-time data of branch, and obtain sealing deterioration mark by screening for misjudgment interference condensation cycle seal failure through the insulation drop period. Grounding determination module 302 is used to extract the associated zero sequence based on the sealing deterioration mark and the branch real-time data to generate a zero sequence cancellation segment, perform grounding screening in the overly stable interval of the zero sequence cancellation segment to obtain the double-point grounding hazard, and perform attenuation characteristic analysis on the double-point grounding hazard to determine the floating discrimination threshold. The fault location module 303 is used to identify abnormal states based on the floating discrimination threshold, construct abnormal discrimination results, perform non-reproducible fault time sequence analysis on the abnormal discrimination results to obtain the location of the fault branch, and screen for false alarm blind spots and output potential abnormal signals for the location of the fault branch. The isolation and handling module 304 is used to isolate the faulty branch according to the potential abnormal signal to obtain the branch isolation command, switch the healthy branch according to the branch isolation command to form the switching and handling result, and perform residual pressure hysteresis concealed bypass screening to generate bypass concealment mark through the switching and handling result; The scheduling and handling module 305 is used to trace and collect data along the repeated fading process based on the bypass concealment flag to generate scheduling and reporting data, perform cross-box instruction verification based on the scheduling and reporting data to form a handling task set, and perform long-term unverified branch verification and output isolation closed-loop instructions for the handling task set.

[0069] The low-voltage cable branch box monitoring and fault isolation system 300 described above can implement a low-voltage cable branch box monitoring and fault isolation method according to the above 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.

[0070] 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 monitoring and fault isolation of low-voltage cable branch boxes, characterized in that, include: Real-time data of the branch current and the enclosure environment are collected. Based on the real-time data of the branch, insulation drop is identified to form insulation drop period. Through the insulation drop period, the seal failure of the misjudgment interference condensation cycle is screened to obtain the seal deterioration mark. Based on the sealing degradation indicator and the real-time data of the branch, the associated zero sequence is extracted to generate a zero sequence cancellation segment. In the zero sequence cancellation segment, the grounding screening of the overly stable interval is carried out to obtain the double-point grounding hazard. The attenuation characteristic analysis is performed on the double-point grounding hazard to determine the floating discrimination threshold. Based on the floating discrimination threshold, abnormal states are identified and abnormal discrimination results are constructed. Non-reproducible fault time sequence analysis is performed on the abnormal discrimination results to obtain the location of the fault branch. False alarm blind spot hazards are screened for the location of the fault branch and potential abnormal signals are output. Based on the potential abnormal signal, the faulty branch is isolated to obtain the branch isolation command. Based on the branch isolation command, the healthy branch is switched to form the switching and handling result. The residual pressure hysteresis hidden bypass is screened through the switching and handling result to generate the bypass hidden indicator. Based on the bypass concealment flag, the data is traced and aggregated along the repeated fading process to generate scheduling and reporting data. Based on the scheduling and reporting data, cross-box instruction verification is performed to form a set of handling tasks. For the set of handling tasks, long-term unverified branch verification is carried out to output isolation closed-loop instructions.

2. The method according to claim 1, characterized in that, The method of screening for seal deterioration indicators by misjudging interference condensation cycle seal failure during the insulation drop period includes: Based on the insulation drop period, the condensation-evaporation phase difference is obtained by tracking the diurnal temperature difference phase of the condensation cycle. The elastic hysteresis is determined by calculating the elastic recovery hysteresis of the sealing ring based on the condensation-evaporation phase difference. Based on the elastic hysteresis, local microleakage paths are identified, and a microleakage path distribution is formed. The sealing level of the continuous leakage section along the joint of the box is evaluated based on the distribution of the micro-leakage path to obtain the sealing deterioration mark.

3. The method according to claim 1, characterized in that, The method for identifying potential two-point grounding hazards during the overly stable interval grounding screening in the zero-sequence cancellation section includes: Based on the zero-sequence offset segment, the zero-sequence residual fluctuation in the excessively stable interval is calculated to generate the residual stability index; The location of a suspected grounding segment is determined based on the phase difference determined by the residual stationary index. The confidence level of the two-point grounding was determined by measuring the abnormally high resistance at the location of the suspected grounding segment. The risk of dual-point grounding is identified by determining the grounding level of the high-confidence segment that continuously exceeds the threshold for the dual-point grounding confidence level.

4. The method according to claim 1, characterized in that, The process of determining the floating discrimination threshold through attenuation characteristic analysis for the aforementioned dual-point grounding hazard includes: Based on the aforementioned dual-point grounding hazard, the attenuation tail of the grounding disturbance attenuation is extracted to obtain attenuation characteristic values; The attenuation characteristic value is used to perform a loosening degree classification of the grounding disturbance waveform symmetry deviation, resulting in a loosening classification result; The loosening grading results were used to identify and correct misjudgments of the long-term high confidence segment that remained unchanged, thus obtaining an effective loosening grade. The floating discrimination threshold is determined based on the upper and lower limits of the effective loosening level conversion threshold range.

5. The method according to claim 1, characterized in that, The step of performing non-reproducible fault timing analysis on the anomaly identification result to obtain the location of the fault branch includes: The anomaly detection results are used to analyze the pre-stability duration of intermittent fault timing to form operating condition feature labels; Based on the aforementioned operating condition feature labels, a common cause analysis of simultaneous alarm-free faults is performed to generate a group of faults originating from the same source. The source-weighted result is obtained by assigning recurrence count weights to the same fault group. Based on the source tracing weighting results, the location of the faulty branch is obtained by determining the location of the branch with the highest weight.

6. The method according to claim 1, characterized in that, The step of screening for false alarm blind spots and outputting potential abnormal signals for the location of the faulty branch includes: Based on the location of the faulty branch, the false alarm blind spot and false alarm interval analysis are used to determine the false alarm feature set; The blind zone classification results are formed by classifying blind zone levels according to the duration of abnormal branch inspection coverage based on the false alarm feature set. Based on the blind zone classification results, supplementary inspection points were planned to supplement the seasonal recurrence frequency, thus obtaining a supplementary verification set; On-site re-inspection is conducted on the supplementary verification set to output potential abnormal signals.

7. The method according to claim 1, characterized in that, The step of generating bypass concealment indicators by screening for residual pressure hysteresis based on the switching and disposal results includes: Based on the results of the switching process, identify the sluggish, step-like pauses in residual pressure attenuation and establish residual pressure attenuation characteristics; Bypass impedance estimation is generated by inverting and analyzing the residual voltage attenuation characteristics. The location of the concealed path is determined by using the bypass impedance estimation to locate the abnormal section with insulation levels below the normal level. Based on the location of the concealed path, the bypass connectivity is verified and a bypass concealment flag is generated.

8. The method according to claim 3, characterized in that, The method of locating the suspected grounding segment position based on the residual stationary index includes: Based on the residual stationarity index, multi-point phase difference reversal analysis is performed to generate a phase difference distribution map; The impedance difference distribution is obtained by calculating the zero-sequence loop impedance difference according to the phase difference distribution diagram. The grounding point coordinates are obtained by using the impedance difference distribution to locate the grounding point triangle in the symmetrical superimposed area. The location of the suspected grounding segment is determined by mapping the box drawing position using the grounding point coordinates.

9. The method according to claim 7, characterized in that, The process of generating a bypass impedance estimate by performing bypass impedance inversion analysis based on the residual voltage attenuation characteristics includes: The residual voltage attenuation characteristics are subjected to high-frequency residual frequency domain separation analysis to obtain a frequency-divided attenuation sequence; The equivalent bypass parameters are determined by calculating the equivalent capacitance and resistance parameters based on the frequency division attenuation sequence. Based on the equivalent bypass parameters, the components deviating from the normal insulation range are subjected to strength rating to form a bypass strength level; The equivalent bypass parameters are corrected using the bypass strength level to generate a bypass impedance estimate.

10. A monitoring and fault isolation system for low-voltage cable branch boxes, characterized in that, include: The data acquisition module is used to collect real-time data of the branch current and the enclosure environment. Based on the real-time data of the branch, it identifies insulation drop periods and uses these insulation drop periods to screen for misjudgment interference condensation cycle seal failure and obtain seal deterioration indicators. The grounding determination module is used to extract the associated zero sequence based on the sealing deterioration mark and the real-time data of the branch to generate a zero sequence cancellation segment, perform grounding screening in the over-stable interval of the zero sequence cancellation segment to obtain the double-point grounding hazard, and perform attenuation characteristic analysis on the double-point grounding hazard to determine the floating discrimination threshold. The fault location module is used to identify abnormal states based on the floating discrimination threshold, construct abnormal discrimination results, perform non-reproducible fault time sequence analysis on the abnormal discrimination results to obtain the location of the fault branch, and screen for false alarm blind spots and output potential abnormal signals for the location of the fault branch. The isolation and handling module is used to isolate faulty branches based on the potential abnormal signals to obtain branch isolation instructions, switch healthy branches according to the branch isolation instructions to form a switching and handling result, and perform residual pressure hysteresis concealed bypass screening to generate bypass concealment indicators through the switching and handling result; The scheduling and handling module is used to trace and collect data along the repeated fading process based on the bypass concealment flag to generate scheduling and reporting data, perform cross-box instruction verification based on the scheduling and reporting data to form a handling task set, and perform long-term unverified branch verification and output isolation closed-loop instructions for the handling task set.