Online monitoring method for partial discharge fault of cable branch box

Abstract

The invention relates to the technical field of power equipment on-line monitoring, and discloses a cable branch box partial discharge fault on-line monitoring method, which is used for solving the problems of high monitoring result false alarm rate, insufficient fault positioning accuracy and poor maintenance directivity in a traditional method. The method comprises the following steps: firstly, establishing a monitoring session of a target cable branch box, collecting a partial discharge pulse record, an external reference record and an accompanying monitoring record, carrying out event merging on each point record under a unified time caliber, and carrying out interference checking and working condition freezing re-checking on the external reference record and an operation working condition; carrying out undertaking judgment and true judgment on abnormal events, and confirming branch affiliation and fault parts by utilizing a branch propagation sequence relation, a branch adjacency relation and an attachment position relation; and performing rechecking, state updating and monitoring conclusion output on the affiliated abnormal event, thereby realizing identification of the partial discharge abnormity of the cable branch box and determination of a fault branch and a fault part.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology for power equipment, specifically to an online monitoring method for partial discharge faults in cable branch boxes. Background Technology

[0002] In power distribution systems, cable branch boxes are used for cable branching, connection, and operational isolation. Cable branch boxes typically contain multiple branches, cable accessories, and a metal enclosure. During operation, partial discharge can easily cause insulation aging and joint deterioration, and in severe cases, may even lead to the expansion of the fault. Therefore, uninterrupted online monitoring of cable branch boxes is of practical significance. For example, the published invention patent application CN101710166B discloses an online monitoring method for partial discharge at power cable joints. This method uses a high-frequency sensor to collect partial discharge signals and combines pulse polarity identification to eliminate interference. Another example is the published invention patent application CN120414907B, which discloses a state recognition-based method for monitoring the state of power distribution cable branch boxes. This method collects multi-dimensional operational data and performs state evolution analysis to achieve dynamic monitoring of the operating state of the cable branch boxes. The above-mentioned prior art has improved upon partial discharge detection and branch box state monitoring, respectively, and plays a positive role in enhancing the online sensing capabilities of power distribution equipment. However, the above technologies still cannot completely solve the problems of verifying the authenticity of partial discharge anomalies in cable branch boxes operating without power interruption and accurately determining the faulty branch and location. This is because cable branch boxes contain multiple branches and various cable accessories, and are located in a metal-enclosed environment, making partial discharge signals susceptible to multi-path propagation, reflection superposition, and coupling. Furthermore, external electromagnetic interference, pulse coupling from surrounding equipment, load changes, and variations in temperature and humidity can easily introduce interference pulses similar to partial discharge. Therefore, while existing technologies can detect abnormal signals or monitor the operation of branch boxes, they cannot definitively determine whether the anomaly originates from the branch box. The propagation of actual partial discharge pulses inside the cable branch box makes it impossible to accurately identify the actual branch and fault location within the box. Furthermore, it is difficult to provide a stable and unified discrimination method suitable for online operation, resulting in false alarms, misjudgments, and inaccuracies. This lack of directional maintenance capability fails to meet the needs for accurate online early warning and targeted maintenance of partial discharge faults in cable branch boxes. Therefore, an online monitoring technology for partial discharge faults in cable branch boxes is needed to address the problems in existing technologies, such as difficulty in identifying actual partial discharges and accurately determining faulty branches and locations due to signal propagation aliasing, external interference coupling, and changes in operating conditions. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an online monitoring method for partial discharge faults in cable branch boxes, which solves the problems of high false alarm rate, insufficient fault location accuracy, and weak maintenance directionality in traditional methods.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for online monitoring of partial discharge faults in cable branch boxes includes: S1. Establish a monitoring session for the target cable branch box, read the branch box branch relationship, accessory distribution, measurement point correspondence and current operating conditions, and simultaneously acquire the partial discharge pulse record, external reference record and accompanying monitoring record of each monitoring point; S2. Merge the records of each monitoring point based on a unified time standard, and combine external reference records and current operating conditions to investigate interference and verify the freezing of operating conditions for abnormal events. S3. Make continuous acceptance judgments on abnormal events that have passed the working condition freeze verification, and make authenticity judgments based on the response sequence, recurrence relationship and regional concentration of abnormal events among various monitoring points. S4. For abnormal events that pass the authenticity determination, determine the branch attribution based on the branch propagation order, branch adjacency, and attachment position, and confirm the fault location along the attachment chain of the corresponding branch. S5. Continuously review and update the status of abnormal events that have completed the attribution determination, and output monitoring conclusions based on the attribution status after review.

[0005] Preferably, a monitoring session is established for the target cable branch box, reading the branch relationships, accessory distribution, measurement point correspondence, and current operating conditions of the branch box, and simultaneously acquiring the partial discharge pulse records, external reference records, and accompanying monitoring records of each monitoring point, including: When establishing a monitoring session, an object file containing the branch attachment chain order and the measurement point layout relationship is invoked to generate a monitoring session record containing a working condition snapshot, a measurement point validity table, an original record cache area, and a subsequent record area. Call the branch mapping table, attachment chain list and measurement point mapping table, receive partial discharge pulse records, external reference records and accompanying monitoring records according to the unified timestamp caliber and write them into the unified original acquisition record table, and perform measurement point access validity judgment. When the measuring point is abnormal, it enters a frozen state awaiting recovery; under operational disturbance conditions, it enters a restricted monitoring state.

[0006] Preferably, events are merged based on a unified time caliber from records at each monitoring point, including: The original acquisition records in the current monitoring session are sorted according to a unified time caliber, and candidate pulse entries are generated for records with trigger markers and that pass the validity judgment of the measurement points. Based on the external reference synchronization relationship and accompanying record matching relationship within the nearest time window, the entries are associated, and within the merging time window, they are merged according to the time sequence relationship within the same monitoring session, the same working condition snapshot, and the same branch, adjacent branches, or the total reference association range, and abnormal events are generated. When multiple measuring points cannot be distinguished in order, the relevant measuring points are jointly marked as the first measuring point in parallel.

[0007] Preferably, and in conjunction with external reference records and current operating conditions, interference investigation and operating condition freeze verification are performed on abnormal events, including: For abnormal events resulting from the merger, conduct external disturbance investigation, internal spatial distribution investigation, and operational condition freeze verification. Based on the sequential relationship between external reference points and first-reach measurement points, the cross-branch distribution of internal trigger measurement points, and the concentration of accompanying monitoring areas, the interference state of abnormal events is determined. The operation events, changes in door status, load changes, and environmental fluctuations are reviewed in conjunction with the operating condition freeze window before and after the abnormal event. Based on the review results, the abnormal event is transferred to the freeze observation state or subsequent continuous acceptance judgment.

[0008] Preferably, the abnormal events that pass the operational condition freeze review are subject to continuous acceptance determination, including: An event acceptance table is established for abnormal events that pass the working condition freeze verification, and event set matching is performed based on the historical sequence of the first measurement point, the historical sequence of the branch candidate, the historical sequence of the power frequency phase section, the historical sequence of the accompanying monitoring area, and the historical sequence of the external reference state. When the first measurement point is the same or located within the allowable drift range, the power frequency phase segment is the same or adjacent, the accompanying monitoring area corresponds to the same area or the same adjacent segment, the branch candidate set has an intersection and the external reference state has not turned into continuous dominance, the new abnormal event will be merged into the existing event set. If the matching conditions are not met, create a new event set.

[0009] Preferably, the authenticity is determined based on the sequence of responses, recurrence, and regional concentration of abnormal events among monitoring points, including: The event set is judged based on the stability of the response sequence, the persistence of the recurrence relationship, and the consistency of the regional concentration. When at least two decision dimensions meet the stability condition, the event set is transformed into a suspected real state; When the external reference state is not dominant, the first measurement point remains stable, and the interval between recurrences does not exceed the interruption threshold, the event set is converted to the formal real state. Under conditions of parallel first arrival, weak intermittent anomalies, and high disturbance background, the results are verified by combining the accompanying monitoring area, operating conditions, and external reference conditions to form an authenticity judgment result table.

[0010] Preferably, for abnormal events that pass the authenticity determination, branch attribution is determined based on branch propagation order, branch adjacency, and attachment location, including: The branch propagation sequence template, which is established based on the branch mapping relationship, the measuring point mapping relationship, the positional relationship of the attachment chain, and the stable triggering sequence under normal working conditions, is invoked. The consistency of the historical sequence of the first measuring point, the historical sequence of the triggering measuring point set, and the historical sequence of the accompanying monitoring area in multiple observation rounds of the event set is compared to determine the main candidate branch. When adjacent candidate branches coexist, conflict is determined based on the difference in the number of first arrivals, the difference in the number of times the accompanying area is concentrated, and the difference in the number of times the historical continuous recurrence occurs. When the event set reaches the formal real state and the main candidate branch is consistent in multiple consecutive observation rounds, the formal branch assignment is formed.

[0011] Preferably, the fault location is identified along the accessory chain of the corresponding branch, including: Based on the positional relationship of the attachment segments in the object file, the accompanying monitoring sensitive areas, and the relative distance relationship with each measuring point, abnormal events are located hierarchically along the attachment chain of the affiliated branch; By combining the concentrated location of the accompanying monitoring area, the arrival order of the first measuring point and the total reference point, and the continuous concentrated location of historical abnormal events within the branch road, the attachment section is compared, and an attachment section-level location conclusion is formed when at least two positioning bases point to the same attachment section. When only a single location is clearly defined, the lower-level location conclusions are retained; when there is no clear segment, the branch-level conclusions are retained; and when the attached segment is clearly defined and the accompanying monitoring area and first arrival sequence within the same attached segment are further stabilized, the location conclusions are refined to the specific component level.

[0012] Preferably, the abnormal events for which attribution determination has been completed are continuously reviewed and their status updated, including: Establish a continuous review form, and set observation windows and recovery observation thresholds according to risk levels; Based on the consistency of event set reproduction within the observation window, duration of continuous missing events, external reference status, operating status, branch attribution stability, and fault location stability, continuous verification and status updates are performed. When abnormal events continue to recur, the original attribution conclusion is retained and the risk level is adjusted; when abnormal events stop recurring, the observation period is resumed. When external references become dominant, branch lines frequently drift, fault locations change across sections, or operating conditions become abnormal for a long period, a rollback review is performed, and the risk level adjustment within a single observation window is limited to one level.

[0013] Preferably, the monitoring conclusion is output based on the verified attribution status, including: Based on the authenticity status after continuous verification, branch attribution, fault location, risk level, current status description, recommended verification time and recommended handling time limit, a formal conclusion record is generated and a monitoring conclusion table is formed; When the risk level changes progressively, the review requirements and handling recommendations shall be adjusted in accordance with the corresponding handling time limits; When the data is under the following conditions: recovery observation, downgrade review, multiple branches to be decomposed, multiple parts to be decomposed, or data is incomplete, the corresponding status information will be written into the formal conclusion record and monitoring conclusion table.

[0014] Compared with the prior art, the present invention provides an online monitoring method for partial discharge faults in cable branch boxes, which has the following beneficial effects: 1. This invention establishes a unified monitoring session to organize object files, operating condition snapshots, measurement point mappings, partial discharge pulse records, external reference records, and accompanying monitoring records in a consistent manner. It also merges abnormal events under a unified time caliber. Furthermore, by combining the sequence of external references, internal spatial distribution, and operating condition freeze verification, it identifies false anomalies caused by external interference, operational procedures, and environmental fluctuations. Further, through cross-cycle event handling, it incorporates changes in the first measurement point, recurring patterns, regional concentration, and external reference status into a continuous judgment process to verify the authenticity of abnormal events. Stable identification; based on this, combined with the branch propagation sequence relationship, branch adjacency relationship and attachment position relationship, the branch attribution of abnormal events is determined and the fault location is confirmed level by level. In subsequent continuous verification, the risk level and conclusion status are dynamically adjusted by combining the recurrence frequency, attribution stability, operating condition status and reverse elimination information, thereby reducing false alarms, false attribution and coarse location, and improving the accuracy of determining the actual partial discharge, fault branch and fault location inside the cable branch box under the condition of uninterrupted power supply. Finally, it realizes accurate online early warning and targeted operation and maintenance handling of partial discharge anomalies in cable branch boxes.

[0015] 2. This invention incorporates the determination of the authenticity of abnormal events, branch attribution, confirmation of fault location, continuous verification, and status rollback into a unified processing chain. It also continuously manages statuses such as recovery observation, frozen observation, multiple branches to be decomposed, multiple locations to be decomposed, and historical archiving. This ensures that partial discharge anomalies can be continuously received and updated according to a unified judgment standard under different monitoring cycles and operating conditions. This avoids problems such as inconsistent judgment standards, loss of abnormal statuses midway, confusion of abnormal chains before and after maintenance, and difficulty in tracing conclusions. As a result, it improves the continuity, verifiability, and traceability of conclusions in the partial discharge monitoring process of cable branch boxes. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the online monitoring method for partial discharge faults in cable branch boxes according to the present invention; Figure 2This is a schematic diagram of the monitoring objects and measuring point arrangement of the cable branch box of the present invention; Figure 3 This is a flowchart of the abnormal event merging, interference investigation, and operating condition freeze verification process of the present invention; Figure 4 This is a state diagram illustrating the continuous reception and authenticity determination of abnormal events in this invention. Figure 5 This is a schematic diagram illustrating the branch attribution and fault location confirmation of the accessory chain in this invention; Figure 6 This diagram illustrates the continuous review, status update, and conclusion output of abnormal events in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: Figures 1-6 A method for online monitoring of partial discharge faults in cable branch boxes is presented, including: S1. Establish a monitoring session for the target cable branch box, read the branch box branch relationship, accessory distribution, measurement point correspondence and current operating conditions, and simultaneously acquire the partial discharge pulse record, external reference record and accompanying monitoring record of each monitoring point; S2. Merge the records of each monitoring point based on a unified time standard, and combine external reference records and current operating conditions to investigate interference and verify the freezing of operating conditions for abnormal events. S3. Make continuous acceptance judgments on abnormal events that have passed the working condition freeze verification, and make authenticity judgments based on the response sequence, recurrence relationship and regional concentration of abnormal events among various monitoring points. S4. For abnormal events that pass the authenticity determination, determine the branch attribution based on the branch propagation order, branch adjacency, and attachment position, and confirm the fault location along the attachment chain of the corresponding branch. S5. Continuously review and update the status of abnormal events that have completed the attribution determination, and output monitoring conclusions based on the attribution status after review. Specifically, such as Figure 2As shown: First, a monitoring session is established for the target cable branch box. Under the same monitoring session, the static information of the object is retrieved, the current operating condition is written, the access status of the monitoring point is confirmed, and the records of each monitoring point are received synchronously. The target cable branch box can be a medium-voltage distribution branch box of 10kV to 35kV, and the number of branches can be 2 to 6, with a typical scenario of 4 branches. The above voltage range is determined according to the common application range of cable branch boxes in medium-voltage distribution networks. The number of branches is determined according to the commonly used two- to six-branch lead-out structures on site, among which the four-branch structure is more common, can reflect the propagation coupling characteristics under multi-branch conditions, and has good representativeness. Before establishing a monitoring session, the object file is accessed. The object file can be generated during equipment commissioning and archiving based on design data, on-site wiring relationships, and measurement point installation relationships, and is saved in a structured record format. The object file includes at least the following fields: branch box number, number of branches, phase information of each branch, direction of each branch, sequence of accessory chains for each branch, box structure dimensions, measurement point layout relationship, measurement point type, and current parameter version. The sequence of accessory chains for each branch can be selected to be registered according to the spatial order from the busbar side to the cable lead-out side, such as wiring transition area, stress control accessory area, joint area, and near-end area of ​​lead-out cable. The use of accessory chain sequence recording is to determine not only the type of accessory but also the front and rear positions of each structural segment on the same branch during the subsequent branch attribution and fault location confirmation process, thereby supporting segment-by-segment judgment along the branch direction. After the object file is successfully retrieved, a monitoring session record should be established. The monitoring session should include at least the session number, start time, object number, current parameter version, current operating condition snapshot, measurement point validity table, original record cache area, and subsequent record area. The session number can be a unique code composed of the object number and time code. The current operating condition snapshot should include at least the following fields: load status of each branch, total load status, door status, operation permission status, ambient temperature, ambient humidity, most recent operation time, and communication link status. Writing the current operating condition snapshot into the monitoring session is to ensure that the merging, investigation, and verification of subsequent abnormal events are carried out in the same time and operating condition caliber, preventing the drift of judgment conditions between different time slices and ensuring the consistent judgment of abnormal events. After the monitoring session is established, the basic session period is set. The basic session period can be selected from 200 milliseconds to 2 seconds, with a typical value of 500 milliseconds. This range is designed to ensure that the same partial discharge event cluster can be continuously received and fully processed in time, while avoiding excessive mixing of abnormal pulses from different sources within the same processing period. If the basic session period is too short, the same partial discharge event cluster is easily divided into multiple discontinuous recording units, weakening the integrity of subsequent event merging. If the basic session period is too long, abnormal pulses from different sources and of different natures are more likely to enter the same time slice, reducing the ability to distinguish the order of responses. The 500 millisecond setting balances recording continuity and judgment resolution, making it suitable for the common processing rhythm of online monitoring of cable branch boxes. After the monitoring session is established, the branch mapping table, attachment list, and measuring point mapping table in the object file are invoked. The branch mapping table includes at least the following fields: branch number, phase, corresponding measuring point number set, adjacent branch number set, and shared structure segment marker. The adjacent branch number set is used to represent the spatially adjacent branch ranges where monitoring signals may interfere with or transmit to each other. The attachment list includes at least the following fields: branch number, segment number, segment type, relative position, distance level from the measuring point, corresponding accompanying sensitive area number, and historical anomaly marker. The measuring point mapping table includes at least the following fields: measuring point number, measuring point type, corresponding branch, corresponding area, synchronization status, and most recent self-test result. The measurement point type can be selected as branch near-end partial discharge pulse acquisition point, total reference point, external reference point, and accompanying monitoring point. The branch near-end partial discharge pulse acquisition point can be set according to the number of branches and is used to receive partial discharge pulse records related to each branch. The total reference point is used to receive records of overall propagation changes inside the enclosure. The external reference point is used to receive records of spatial disturbances outside the enclosure. The accompanying monitoring point is used to collect records of ultrasonic, mechanical vibration, or other auxiliary abnormal monitoring. The use of multiple types of measurement points in a joint access method is to incorporate the records inside the enclosure, the reference records outside the enclosure, and the accompanying monitoring records into the same monitoring session, providing a unified data foundation for subsequent differentiation between real and pseudo partial discharge. After the measurement point mapping relationship is invoked, the records of each monitoring point are acquired synchronously. To ensure the feasibility of subsequent event merging and response sequence determination, a unified timestamp caliber is preferred for each monitoring point record. The timestamp accuracy can be selected from 1 microsecond to 100 microseconds, with typical values ​​such as 10 microseconds or 20 microseconds. This range is set because the arrival time difference of abnormal events between monitoring points inside the cable branch box is usually short, and the timestamp accuracy needs to be sufficient to support the sequence determination under short-distance propagation conditions. At the same time, it should also match the common implementation capabilities of the front-end synchronization device to avoid deviating from the field deployment conditions. When the front-end synchronization capability is weak, a 100-microsecond time caliber can be used; when the front-end synchronization capability is strong, a 10-microsecond or 20-microsecond time caliber can be used to improve the resolution capability of subsequent branch propagation sequence determination. Monitoring point records are written into a unified original acquisition record table according to the fields; partial discharge pulses should at least include the acquisition timestamp, measurement point number, trigger marker, pulse strength, pulse duration range, power frequency phase segment, number of repetitions, and record integrity marker; external reference records should at least include the external reference point number, trigger timestamp, spatial disturbance level, number of consecutive triggers, and whether it is currently in a high disturbance state; accompanying monitoring records should at least include the accompanying monitoring point number, trigger timestamp, trigger area, accompanying anomaly level, and duration range; the spatial disturbance level, accompanying anomaly level, and record integrity marker can be generated by the corresponding acquisition unit according to preset grading rules or self-testing rules and then written into the unified original acquisition record table; the pulse strength level is preferably divided into five levels, each... The levels are divided according to the amplitude range after the acquisition unit is calibrated, and can be selected as weak, weak-medium, medium, medium-strong, and strong. Five levels are used because this classification method can cover the common engineering variation range of partial discharge pulses from weak to strong, while avoiding the significant increase in the complexity of on-site calibration and calling due to overly fine classification. The pulse duration range can be selected as short pulse, standard pulse, and extended pulse, which are used for time feature comparison when merging subsequent abnormal events and identifying pseudo pulses. The power frequency phase segment is preferably divided into 12 segments, each segment corresponding to 30 degrees. The 12-segment division is used because this granularity can reflect the repetitive correlation between partial discharge anomalies and power frequency phase, and will not reduce the on-site statistical stability due to overly fine segmentation. It is a phase segmentation method commonly used in this field and easy to implement. While synchronously receiving records from each monitoring point, a valid measurement point access judgment is performed. The valid measurement point access judgment checks at least whether the synchronization status exceeds the limit, whether the continuous packet loss ratio exceeds the limit, whether the background noise status is abnormal, and whether the most recent self-test has passed. The synchronization deviation is preferably no more than 50 microseconds. This value matches the uniform timestamp accuracy and the propagation time difference between typical monitoring points, and is used to ensure that the subsequent response order of multiple measurement points is distinguishable. If the synchronization deviation exceeds this range, the order judgment between different measurement points may lose its practical meaning. The continuous packet loss ratio is preferably no more than 5%. This value serves as the lower limit setting for access validity and is used to ensure the basic record integrity required for event merging and continuous reception within a single monitoring session. If the continuous packet loss ratio is too high, the same abnormal event may cause record breaks. The duration of abnormal background noise is preferably no more than 3 session basic cycles. This duration is used to distinguish between short-term noise rise and continuous measurement point abnormality. Abnormalities with a duration less than this range can be preferentially judged as short-term disturbances, and abnormalities with a duration exceeding this range can be preferentially judged as measurement point status abnormalities. If a branch test point fails the access validity judgment in the current monitoring session, the entire monitoring session will not be terminated directly. Instead, the branch will be marked as a degraded access state, and the degraded information will be written into the test point validity table. While in a degraded access state, the branch record is still allowed to enter the subsequent processing flow, but its attribution determination level cannot be directly upgraded to the highest level. If the total reference point and a critical branch test point fail simultaneously, the current monitoring session will enter a frozen pending recovery state. In the frozen pending recovery state, the original records will continue to be cached, but no formal abnormal event will be generated to avoid outputting highly deterministic conclusions under the condition of missing critical references. After the object file is retrieved, the monitoring session is established, the operating condition snapshot is written, the measurement point mapping is retrieved, and the unified original acquisition record table is established, the output structure of the current processing stage is generated. The output structure preferably includes an object session table, a measurement point validity table, and a unified original acquisition record table, which are used for subsequent abnormal event handling, measurement point status constraints, and original record retrieval, respectively. If the monitoring session establishment stage detects that the box door is open, the maintenance authorization is valid, or the recent operation time has not exceeded the preset freeze window, the current monitoring session is started in a restricted monitoring state. The freeze window is preferably set to 5 to 20 seconds according to the operation type. For example, 5 to 10 seconds can be used for short-term general operations, 10 to 20 seconds for switching operations, and 10 seconds for typical operations. This setting is because the duration of the impact of different operation behaviors on the monitoring records in the branch box varies. An adjustable freeze window can cover the common duration range of short-term disturbances after the operation ends. In the restricted monitoring state, it is allowed to continue caching and merging of original records, but abnormal events can only enter the authenticity upgrade process after at least one additional observation round to avoid operation interference directly entering the formal fault chain. The above processing forms a unified input basis for subsequent abnormal event merging, interference investigation, and operational condition freeze verification.

[0019] Specifically, such as Figure 3 As shown: Based on the object session table, the measurement point validity table, and the unified original acquisition record table, the records of each monitoring point within the current monitoring session are organized according to a unified time caliber. Abnormal events are merged, interference is investigated, and the operating condition is frozen and verified in conjunction with external reference records and the current operating conditions. After sorting all original acquisition records within the current monitoring session in chronological order, each triggered record is not directly treated as an independent abnormal event. Instead, candidate pulse entries are generated for original acquisition records with trigger markers and that pass the measurement point validity judgment. Candidate pulse entries must include at least the following fields: entry number, trigger time, triggering measurement point, corresponding branch identifier, pulse strength level, duration interval, power frequency phase segment, external reference synchronization status, and accompanying record matching status. The external reference synchronization status indicates whether an external reference point responds within the nearest time window before and after the current entry is triggered. If an external reference record appears within the nearest time window, it is considered synchronized; otherwise, it is considered unsynchronized. The accompanying record matching status indicates whether an accompanying monitoring record exists in the same area within the same nearest time window. If an accompanying monitoring record appears within the nearest time window, and the triggering area is the same as or adjacent to the area corresponding to the current entry, it is considered matched; otherwise, it is considered unmatched. The nearest time window is preferably between 50 microseconds and 1 millisecond, typically 200 microseconds. This range is used to complete the entry-level association judgment, and its setting considers two factors: on the one hand, the propagation of partial discharge pulses and accompanying anomalies in the box are usually concentrated in a short period of time, and a time window that is too small will miss the nearest reference record and accompanying record related to the current entry; on the other hand, a time window that is too large is prone to misjudging records that occur simultaneously but are not related as related records. A typical value of 200 microseconds can balance matching sensitivity and mismatch control, and is suitable for common record registration needs under short-distance propagation conditions in medium-voltage cable branch boxes. After candidate pulse entries are formed, event merging is performed. Event merging is used to combine multiple candidate pulse entries within the same time cluster, satisfying spatial correlation and located within the same operating condition boundary into one abnormal event. Before merging, a merging time window is set. The merging time window is preferably 100 microseconds to 5 milliseconds, typically 500 microseconds. This range needs to be determined in conjunction with the branch box volume, number of branches, measurement point spacing, and field synchronization accuracy. If the merging time window is too small, the following triggers and short-term repetitive pulses of the same abnormal event at different measurement points may be split into multiple independent events. If the merging time window is too large, abnormal records from different sources are easily mistakenly merged into the same event. For four-branch branch boxes with small box size and close branch spacing, the merging time window can be, for example, 300 microseconds to 800 microseconds. For branch boxes with larger enclosures or more complex multi-branch structures, the upper limit can be appropriately relaxed as long as it does not cross obvious operating condition boundaries. The 5-millisecond upper limit is mainly used to cover the hysteresis-following triggering of the same abnormal event at multiple measurement points and the requirement for short-term repetitive pulses to be incorporated. When merging, the following rules can be followed: within the same merging time window, if multiple candidate pulse entries belong to the same monitoring session, correspond to the same operating condition snapshot, and the triggering measurement points belong to the same branch, adjacent branches, or the total reference association range, then these entries will be merged into the same candidate abnormal event. If multiple entries are close in time but are located before or after obviously different operating condition boundaries, such as one entry being located before the load transition and another entry being located after the load transition, then they will not be merged to avoid different types of records being incorrectly incorporated into the same abnormal event. The merged abnormal events are preferably recorded in a structured manner, including at least the following fields: event number, event start time, event end time, first-reaching measuring point, set of triggering measuring points, set of branch candidate points, number of repetitions within the event, set of accompanying monitoring areas, set of external reference triggers, and current operating condition label. The first-reaching measuring point can be understood as the measuring point that is the first to trigger in the abnormal event and passes the measuring point validity judgment. If multiple measuring points cannot be distinguished in sequence within the current synchronization accuracy range, these measuring points are jointly marked as parallel first-reaching measuring points, and the certainty of the abnormal event's attribution is reduced in subsequent attribution determination. The number of repetitions within the event is used to indicate the number of times the same type of pulse occurs repeatedly within the same abnormal event, and its statistical scope can be selected as the same branch, the same power frequency phase segment, or adjacent power frequency phase segments. The set of accompanying monitoring areas is used to reflect whether the accompanying anomalies are concentrated in a certain local area. The set of external reference triggers is used to record whether external reference points are triggered before and after the event, the triggering sequence, and the triggering coverage. By recording the above fields, a structured abnormal event that simultaneously includes time relationship, spatial relationship, branch candidate relationship, and external reference relationship can be formed, providing a unified input for subsequent authenticity determination and branch attribution. After an abnormal event occurs, interference investigation is first performed. Interference investigation can be divided into external disturbance investigation and internal spatial distribution investigation. During external disturbance investigation, the external reference trigger set in the abnormal event is called to check whether the external reference point is triggered before the first arrival measurement point and whether the internal trigger measurement point crosses multiple non-adjacent branches. If the external reference point is triggered before the first arrival measurement point, and the internal trigger measurement points in the abnormal event are simultaneously distributed across multiple non-adjacent branches, and the monitoring area is not concentrated, then the abnormal event is marked as an external interference priority event. If the external reference point is triggered but later than the internal first arrival measurement point, or only briefly follows the internal anomaly after its formation, it is not directly excluded, but recorded as an external reference existence state and transferred to subsequent verification. This processing method is adopted because the abnormal signal generated by the actual partial discharge in the box may also be transmitted to the external reference point along the box or the surrounding space, thereby causing the external reference point to respond. Therefore, it is not possible to make a exclusion judgment based solely on whether the external reference point is triggered. It is still necessary to combine the trigger sequence and spatial coverage for comprehensive identification. During internal spatial distribution investigation, check whether the distribution of abnormal events on the measurement points within the box conforms to the branch adjacency relationship and the overall reference correlation relationship. If the same abnormal event is triggered almost synchronously on two or more non-adjacent branch measurement points and continues to repeat, and there is no first-reach measurement point that can maintain stability, it is marked as an internal mixed event. Here, "almost synchronous triggering" means that the triggering time difference between multiple non-adjacent branch measurement points does not exceed a preset multiple of the current synchronization accuracy limit, such as 3 to 10 times. "Continuously repeating" means that the distribution state occurs more than twice in the current session, or reappears in an adjacent session. Internal mixed events do not directly enter the real partial discharge candidate chain, but are transferred to the decomposition state to reduce the interference of random crosstalk and diffusion-induced abnormal events on subsequent branch determination and fault location confirmation. After interference investigation is completed, a condition freeze verification is performed. The condition freeze verification is used to isolate abnormal records introduced by load transitions, circuit breaker opening and closing, grounding switching, enclosure door opening, maintenance actions, and short-term environmental fluctuations from the real partial discharge candidate chain. During the verification, the condition snapshot of the current monitoring session is called, and a condition freeze window is set before and after the abnormal event. The freeze window is preferably 1 to 10 seconds before the event and 1 to 30 seconds after the event. The time ranges before and after are asymmetrical because the front window is mainly used to capture the induced operation or significant disturbance before the abnormal event occurs, while the back window is mainly used to cover the residual disturbance attenuation process after the operation ends. Therefore, the back window is usually longer than the front window. For circuit breaker opening and closing and grounding switching, for example, 5 seconds before and 20 seconds after can be selected. For ordinary load changes, for example, 1 second before and 5 seconds after can be selected. If an operation event, enclosure door status change, or load change exceeding the threshold occurs within the freeze window, the abnormal event is marked as a condition freeze event. The load change threshold can be set to 10% to 30% of the rated load, typically 15%. The rated load can be selected as the rated operating load of the current branch or the rated operating load of the current object. This range is set to distinguish between normal load fluctuations and significant load changes that may cause transient pulse disturbances. A typical value of 15% is suitable for judging load fluctuations in common distribution branches. Temperature and humidity fluctuations can be judged based on short-term changes. For example, a temperature change of 3°C or a relative humidity change of 8% or more within a preset environmental observation window can be considered a significant environmental change. The preset environmental observation window can be 5 to 15 minutes, typically 10 minutes. The humidity change threshold is mainly used to identify significant environmental fluctuations that drift slowly beyond the normal environment in a short period of time, thereby determining whether environmental factors may have a short-term impact on monitoring records. The above-mentioned environmental changes are only used as conditions for identifying operating condition disturbances and are not used as a basis for independent fault judgment. Operating condition frozen events are not directly deleted, but the event structure is retained and the event is transferred to the frozen observation state. If the same type of event does not reappear after the operating condition returns to stability, it is archived as an operating condition disturbance event. If the same type of event continues to reappear after the operating condition returns to stability, according to the same branch candidate, the same or adjacent power frequency phase segment, and the same accompanying monitoring area, the freeze is lifted and the event enters the subsequent continuous acceptance judgment. To ensure closed-loop processing, anomaly and boundary handling rules can be set. If an external reference point fails when an anomaly occurs, the external disturbance investigation only records the missing reference state and does not use it as a basis for exclusion. In this case, the entry threshold for condition freeze verification and subsequent continuous acceptance judgment is correspondingly increased. If multiple parallel first-reaching measurement points appear simultaneously after event merging, the parallel first-reaching marker is retained, and the events are not forcibly split to avoid introducing erroneous branches when evidence is insufficient. If the number of anomaly events in a single monitoring session exceeds the preset limit, they are sorted according to event persistence, branch concentration, and the stability of the accompanying monitoring area. The anomaly events with higher ranking are retained to enter the next stage, while the remaining events retain their original records but do not enter the current round of continuous acceptance judgment. This limit can be set to, for example, 200 groups. This setting is because a large number of instantaneous pulse events may occur under high interference conditions on site. If all of them enter the subsequent link, it will significantly increase the processing burden and weaken the judgment stability. For deployment conditions with different processing capabilities, this limit can also be adjusted according to the on-site computing resources and communication capabilities. After completing the merging, interference investigation, and operating condition freeze verification, a merging and verification result table is generated. The merging and verification result table includes at least the following fields: abnormal event number, first reaching measurement point, set of triggering measurement points, priority marker for external interference, marker for internal mixing, whether operating condition freeze is triggered, current branch candidate set, and status code for entering the next stage. The status code can be selected as entering continuous acceptance judgment, entering pending decomposition observation, entering operating condition freeze observation, entering external interference archiving, or entering downgrade verification. After the above processing, the original scattered monitoring records can be transformed into a set of abnormal events after being merged with unified time caliber and operating condition constraints, thus providing a structured basis for subsequent authenticity determination, branch attribution, and fault location confirmation.

[0020] Specifically, such as Figure 4 As shown: Based on the merged and verified result table, abnormal events that have passed the working condition freeze verification are continuously accepted for judgment, and a authenticity status is formed; the authenticity judgment is continuously carried out in multiple monitoring session cycles; for this purpose, an event acceptance table is first established; the event acceptance table is used to associate abnormal events of the same source or suspected same source across cycles to form a continuously tracked event set; the event acceptance table includes at least the following fields: event set number, first occurrence time, most recent occurrence time, first measurement point historical sequence, branch candidate historical sequence, power frequency phase segment historical sequence, accompanying monitoring area historical sequence, external reference status historical sequence, current authenticity status, and most recent status update time; the authenticity status can be divided into the observation status, suspected authenticity status, and formal authenticity status; the above-mentioned hierarchical status is adopted to adapt to the characteristics of the early discrete and intermittent occurrence of partial discharge anomalies in cable branch boxes, so that the authenticity judgment can gradually converge with the continuous acceptance results; After establishing the event acceptance table, event set matching is performed on abnormal events that can proceed to the next stage. During event set matching, newly entered abnormal events are not required to be completely identical to historical abnormal events; instead, their inclusion in an existing event set is determined based on acceptance criteria. Acceptance criteria include at least the following: first-reaching measurement point relationship, power frequency phase segment relationship, accompanying monitoring area relationship, branch candidate relationship, and external reference state relationship. The first-reaching measurement point relationship can be determined based on whether the first-reaching measurement points are identical or within the allowable drift range. The allowable drift range can be determined by the measurement point mapping relationship and branch mapping relationship in the object file, and is set between the near-end measurement point of the same branch and the total reference point to accommodate slight changes between neighboring measurement points and exclude unordered jumps at distant points. Power frequency phase... Segment relationships can be determined based on identical or adjacent segments; accompanying monitoring area relationships can be determined based on adjacent areas within the same area or the same attachment segment; branch candidate relationships can be determined based on the existence of intersections in the branch candidate sets; external reference status relationships can be determined based on the fact that the external reference status has not changed from non-dominant to continuously dominant; if at least three of the above conditions are met, the new abnormal event will be merged into the existing event set; if not, a new event set will be created; if the same new abnormal event simultaneously meets the merging conditions of multiple existing event sets, the event set with the most matching items will be merged first; if the number of matching items is the same, the event set with the most recent occurrence time will be merged first; if it is still impossible to distinguish, it will be temporarily marked as pending decomposition and retained for the next round of review; After completing the event set matching, a continuous acceptance judgment is performed. This judgment focuses on the stability of response sequence, the persistence of recurring relationships, and the consistency of regional concentration. The stability of response sequence is used to determine whether the first-reaching measurement point remains stable during continuous observation, or whether it remains confined to the same local measurement point group. The observation rounds can range from 3 to 10 basic session cycles, typically 5 cycles. This setting balances recognition speed and stability confirmation requirements. Too few observation rounds make it difficult to distinguish between single, occasional anomalies and persistent true anomalies; too many observation rounds prolong the time required for authenticity determination. The persistence of recurring relationships is used to determine whether the same event set recurs within continuous observation rounds, or whether, although... It does not appear in every round, but it does not completely disappear within the preset observation window; the observation window can be set from 5 minutes to 24 hours, with 2 hours being typical; this wide range is designed to accommodate abnormal events of different intensities and development stages, with shorter observation windows suitable for rapidly recurring abnormal events and longer observation windows suitable for the continuous confirmation of weak intermittent abnormalities; the observation window can also be set in layers according to the event intensity level or recurrence frequency; the consistency of regional concentration is used to determine whether the accompanying monitoring area and the triggering area within the box are concentrated in the same branch or the same adjacent section for a long period of time, rather than randomly jumping between multiple non-adjacent areas; only when at least two of the above three aspects reach the stable condition can the event set be upgraded from the observation state to the suspected real state; The transition between real and unreal states follows state transition rules. In the observation state, if the event set exhibits at least two valid abnormal events within the last two observation rounds, and the first-reaching measuring point has not left the same branch's local measuring point group, while no cross-branch jumps occur in the monitoring area, then the event set is upgraded to a suspected real state. In the suspected real state, if, within at least two subsequent observation rounds, the external reference state remains dominant, the first-reaching measuring point remains stable, and the recurrence interval does not exceed the preset interruption threshold, then the event set is upgraded to a formal real state. The interruption threshold can be set from 30 minutes to 24 hours, typically 4 hours. The observation window is used to determine whether the event set still exhibits continuous occurrence characteristics within a certain period, and the interruption threshold is used to determine whether the continuous continuity relationship has been interrupted. The interruption threshold is set as described above. Within this range, it is designed to accommodate both weak intermittent anomalies and continuous anomalies. If the duration is too short, weak intermittent anomalies are easily misjudged as interruptions; if the duration is too long, independent anomalies from different sources may be incorrectly linked together. If, under the suspected true state, the external reference state continues to dominate, branch candidates change frequently, or the region disappears in a concentrated manner, the event set is rolled back to the observation state. The official true state is not permanently fixed; if it does not reappear for a long time, it can be moved to the recovery observation state. If obvious conflicting evidence appears later, it can be rolled back to the suspected true state or the decomposition state. Under the recovery observation state, if the continuous missing duration exceeds the preset time limit, it is moved to the historical anomaly archiving state. If, during the recovery observation period, a similar event that meets the acceptance conditions reappears, the original continuous acceptance chain is restored and the review continues. In the process of determining authenticity, boundary conditions also need to be addressed. For parallel first-arrival cases, if two parallel first-arrival measurement points exist for a long period of time and these two measurement points belong to the same branch's local measurement point group, then the case can continue to be accepted and is not considered unstable. If the parallel first-arrival measurement points belong to different branches, then the threshold for upgrading the authenticity status is raised, and the verification requirements for the consistency of the accompanying monitoring area and operating conditions are added. For weak intermittent anomalies, if the intensity of a single event set is not high, but it is reproduced multiple times within a long observation window according to the same branch and the same power frequency phase segment, then it is allowed to upgrade from the pending observation state to the suspected authenticity state. However, it must not be directly upgraded to the official real state, but should be confirmed through at least one more observation round; for high-disturbance background anomalies, if the external reference point is in a high-disturbance state for a period of time, all newly entered event sets during that period should be registered as pending observation state with a restricted tag attached; the high-disturbance state can be determined according to the high-disturbance state tag in the external reference record; the continued recurrence after the high-disturbance state is lifted can be judged by its recurrence within at least one observation round; only event sets that continue to recur after the external reference high-disturbance state is lifted can resume the normal upgrade path; To ensure that the authenticity determination results can directly enter the subsequent branch assignment process, input and output structures are also set up. The input structure includes at least the following fields: event set number, event sequence, first arrival measurement point sequence, accompanying area sequence, external reference sequence, and operating condition sequence. The output structure includes at least the following fields: authenticity status, status update time, next review time, whether branch assignment is allowed, whether extended observation is required, and current constraint description. The current constraint description is used to record the reasons for restricting status escalation, such as unstable external reference, undecomposed parallel first arrivals, unconcentrated accompanying areas, or branch... Candidates exhibit significant fluctuations; the next review time can be set tiered according to the authenticity status; the pending observation status can be reviewed after 1 to 3 basic session cycles; the suspected true status can be reviewed after 1 observation round; the official true status can be continuously reviewed within a longer monitoring window; this setting is because different authenticity statuses correspond to different levels of stability, and the review interval should match the confirmation requirements of the current status; among them, the review interval for the pending observation status is shorter, used to quickly determine whether occasional anomalies continue to reproduce, and the suspected true status is reviewed according to the observation round, used to confirm the upgrade conditions in a more complete and continuous context; After completing the continuous acceptance judgment, an authenticity judgment result table is generated. The authenticity judgment result table includes at least the event set number, current authenticity status, stability mark of the first measurement point, repeat occurrence mark, regional central consistency mark, external reference dominance mark, next review time, whether access to branch attribution is allowed, and current constraint description. After the above processing, the authenticity judgment results required for subsequent branch attribution and fault location confirmation can be generated.

[0021] Specifically, such as Figure 5 As shown: Based on the authenticity judgment result table, the branch attribution is determined for abnormal events that meet the attribution conditions, and the fault location is confirmed along the accessory chain of the corresponding branch. To ensure that the branch attribution has a clear basis for judgment, a branch propagation sequence template is first established. The branch propagation sequence template can be determined based on the branch mapping relationship, measurement point mapping relationship, accessory chain position relationship and stable triggering sequence formed under normal operating conditions in the object file, and is associated with and saved with the object file. Each branch corresponds to a branch propagation sequence template. The branch propagation sequence template includes at least the following fields: priority trigger measurement point, set of allowed follow measurement points, set of adjacent propagable branches, set of far-end non-priority trigger measurement points, and corresponding accessory segment order. The priority trigger measurement point can be the near-end partial discharge pulse acquisition point of the branch, the set of allowed follow measurement points can be the total reference point and the near-end measurement points of adjacent branches, and the set of far-end non-priority trigger measurement points can be the measurement points of non-adjacent branches. If the branch box structure is symmetrical, resulting in local similarity of the propagation sequence templates of two branches, the accessory position relationship and the accompanying monitoring area are combined to further distinguish them to avoid branch misjudgment. When determining branch attribution, first read the historical sequence of the first arrival measuring point, the historical sequence of the trigger measuring point set, and the historical sequence of the accompanying monitoring area in the most recent rounds of the event set, and then compare them with the consistency of the propagation sequence template of each branch. Branch attribution determination mainly examines whether the first arrival measuring point falls within the priority trigger measuring point or the allowable local drift range of the target branch, whether the follow-up triggering mainly occurs within the allowable follow-up measuring point set of the target branch, whether the far-end non-priority trigger measuring point has not been continuously arriving first, and whether the accompanying monitoring area is consistent with the adjacent sensitive area corresponding to the target branch; the allowable local drift range can be determined by the same... The mapping relationship between the near-end measuring points of a branch and the overall reference point is determined to accommodate small changes between neighboring measuring points and exclude disordered jumps in distant, unrelated branches. If a branch continuously meets the above conditions in the most recent two or more observation rounds, it is identified as the main candidate branch. The observation rounds are set to two or more because single-round matching is easily affected by occasional anomalies, while two or more consecutive rounds can balance the stability of the attribution and the response speed of online monitoring. If the main candidate branch is unique, the attribution confirmation is initiated. If multiple candidate branches exist simultaneously, adjacent branch conflict judgment is performed. When determining conflicts between adjacent branches, three differences are compared: the difference in the number of first arrivals, the difference in the number of times the accompanying area is concentrated, and the difference in the number of times the historical continuous recurrence occurs. These differences can be formed based on the statistical results of the most recent two or more observation rounds. If at least two of the three differences point to the same branch, then the branch is identified as the primary candidate branch, and the other branch is recorded as the secondary candidate branch. Using at least two consistent differences as the judgment threshold is because relying on a single difference can easily lead to incorrect attribution, which is not conducive to the stable distinction between adjacent branches. If the three differences are insufficient to point to the same branch, then the current event set is marked as a multi-branch decomposition state, and a single-branch conclusion is not directly output. Branch assignment is linked to the authenticity status. If the event set has reached the formal authenticity status and the main candidate branch is consistent in two consecutive observation rounds, a formal branch assignment is formed. If the event set is still in the suspected authenticity status, even if the current main candidate branch is clear, only a temporary branch assignment is formed, and it will continue to be verified in subsequent rounds. This setting is because early weak anomalies may be temporarily misaligned between adjacent branches due to signal crosstalk and propagation. Only when the authenticity status and branch assignment results remain stable at the same time can the branch conclusion have a high degree of credibility. After assigning the branch to its corresponding section, the fault location is confirmed along the accessory chain of that section. The accessory chain can be divided into 3 to 6 segments according to the order from the electrical connection transition section to the near end of the lead cable within the branch. Typically, it can be divided into the wiring transition section, stress control section, joint body section, insulating sleeve section, and near end of the lead cable. Dividing the accessory chain into 3 to 6 segments is to balance the granularity of fault location and the ability to distinguish on-site measurement points. If there are too few segments, the fault location will be too coarse and it will be difficult to reflect the differences between different structural sections within the branch. If there are too many segments, it will be difficult to distinguish the measurement point response and accompanying monitoring characteristics between adjacent segments. Each accessory segment should at least have a corresponding location description, accompanying monitoring sensitive area, and relative distance level from each measurement point in the object file. When confirming the fault location, first determine whether the accompanying monitoring area is stably concentrated in the sensitive area corresponding to a certain attachment segment. Then, determine whether the response sequence of the first measuring point and the total reference point conforms to the near-end or far-end characteristics of that attachment segment. Next, determine whether the concentration of historical abnormal events within the branch is continuously concentrated in the same attachment segment. The near-end or far-end characteristics can be determined according to the relative position of the attachment segment in the branch and the arrival sequence of the corresponding measuring points. If two or more of the three judgments point to the same attachment segment, then that attachment segment is identified as the formal fault location. Using two or more consistent judgments as the confirmation threshold is to avoid directly forming a location conclusion based on a single piece of evidence, thereby improving the stability of fault location confirmation. If only one judgment is clear, then the lower-level location confirmation result is retained. If none of the three judgments can form a clear indication, then only the branch-level conclusion is retained. To facilitate tiered output, fault location identification can be categorized by granularity into branch level, accessory section level, and specific component level. Branch level indicates that the anomaly can only be identified as belonging to a specific branch. Accessory section level indicates that the anomaly is located in a specific accessory section. Specific component level indicates that, based on the identified accessory section, the identification is further refined to the specific joint area, specific stress control accessory area, or other specific component area. Specific component level identification requires that the previous level has been identified, and that the accompanying monitoring area and first arrival sequence within the same accessory section remain stable before refining from the accessory section level to the specific component level. If the resolution of the field measurement points is insufficient, it is sufficient to remain at the accessory section level, without forcibly refining to the specific component level. During the process of branch attribution and fault location confirmation, it is also necessary to handle abnormal and boundary situations. If the branch propagation sequence template fails due to modification, maintenance, or measurement point relocation, the current monitoring session will not output the formal branch attribution, but will instead enter the template update pending confirmation state. If all monitoring points fail, the fault location confirmation will be based solely on the measurement point response sequence and historical concentrated location, with the output granularity decreasing by one level. For example, if it could originally reach the specific component level, it will be reduced to the accessory segment level; if it could originally reach the accessory segment level, it will be reduced to the branch level. If the same branch points to two different accessory segments in different observation rounds, but these two accessory segments belong to the same continuous chain segment, such as the boundary between the stress control segment and the joint body segment, they can be temporarily merged into a composite accessory segment and used as a transitional confirmation state before subsequent verification. If the two accessory segments are completely separated in spatial location, the process will enter the multi-part decomposition state, and a single fault location conclusion will not be directly formed. After confirming the branch attribution and fault location, a branch attribution and fault location confirmation table is generated. This table includes at least the following fields: event set number, authenticity status, primary candidate branch, secondary candidate branch, current branch attribution level, current fault location level, fault location name, adjacency conflict status, and subsequent review suggestions. The branch attribution level can be divided into temporary branch attribution and formal branch attribution; the fault location level can be divided into suspected location, formal segment, and formal specific location; the adjacency conflict status is used to record whether there are branch conflicts, multi-location conflicts, or template mismatches. After the above processing, the branch attribution and fault location confirmation results required for subsequent continuous review, status updates, and conclusion output are generated.

[0022] Specifically, such as Figure 6As shown: Based on the branch attribution and fault location confirmation table, abnormal events for which attribution determination has been completed are continuously reviewed and their status updated, and monitoring conclusions are output. To this end, a continuous review table is established. This table continuously records the recurrence, attribution stability, risk changes, and handling recommendations for the same event set during subsequent observations. The continuous review table includes at least the following fields: event set number, current authenticity status, current branch attribution, current fault location, current risk level, most recent occurrence time, most recent operating condition status, most recent external reference status, cumulative recurrence count, duration of continuous missing data, recommended review time limit, and recommended handling time limit. The cumulative recurrence count is used to characterize... The continuous development trend of abnormal events and the duration of continuous absence are used to characterize the degree of decline of abnormal events; risk levels can be divided into alert level, attention level, warning level, and alarm level to adapt to management needs under different risk levels; the initial risk level can be determined by primarily the actual status, secondarily by the branch attribution level and the fault location level, and by using the cumulative number of recurrences for adjustment at the same level; if an abnormal event has reached a formal actual status and formed a formal branch attribution, but the fault location is still at the suspected level, it can be determined as attention level; if a formal branch attribution has been formed and the fault location has reached the formal attachment segment level, it can be determined as warning level; if it recurs frequently in a short period of time and external references do not dominate for a long period of time, it can be further upgraded to alarm level; When conducting continuous verification, first check the consistency of the reproduction of the same event set within subsequent observation windows. Observation windows can be set in layers according to risk level; the higher the risk level, the shorter the observation window. The observation window for alert level can be 24 hours to 7 days, for attention level it can be 4 hours to 72 hours, for warning level it can be 30 minutes to 24 hours, and for alarm level it can be 10 minutes to 2 hours. Observation windows for each level can be selected within the corresponding range based on on-site risk tolerance, operation and maintenance response capabilities, and anomaly reproduction density. The above range is adopted to ensure that low-risk anomalies have sufficient continuous observation time, while allowing high-risk anomalies to be reconfirmed in a shorter time. If, within the corresponding observation window, the same event set continuously reproduces along the same branch and the same or adjacent accessory segments, and the external reference status and working conditions are consistent, the observation window should be considered valid. If no evidence of reverse exclusion is provided in the current situation, the original attribution conclusion is retained, and the risk level update process begins. If the event no longer recurs within the observation window, the event set is transferred to the recovery observation state. In the recovery observation state, the existing conclusion is not immediately revoked, but the duration of continuous absence is recorded. The recovery observation threshold can be set incrementally according to the risk level; for example, 24 hours can be used for the alert level, 72 hours for the attention level, and 7 days for the warning level and above. The incremental setting is to avoid premature revocation of true early partial discharges due to short-term disappearance, while also preserving more sufficient confirmation time for high-risk anomalies. If the duration of continuous absence exceeds the corresponding recovery observation threshold, the conclusion is downgraded to a historical anomaly state. If a similar event that meets the acceptance conditions reappears during the recovery observation period, the original continuous acceptance chain is restored and the review continues. During status updates, a reverse exclusion review is also performed. The reverse exclusion review is used to check whether evidence of conflict with the current attribution status has appeared during the continuous review. Conflict evidence includes at least the following situations: the external reference point is reached and dominates in more than half of the reproducible events in the current observation window; the main candidate branch has changed more than twice in the last two or more observation rounds; the fault location frequently jumps across segments; the operating condition freeze state persists for a long time; the door is in the maintenance authorized open state for a long time. If conflict evidence appears, the current event set will not be directly upgraded along the original conclusion, but will be rolled back according to the conflict type. If the conflict only involves the fault location jumping but the branch remains stable, it will roll back to the branch-level formal attribution status. If the conflict involves frequent branch drift, it will roll back to the suspected real state or the multi-branch decomposition state. If the conflict involves the external reference being dominated for a long time, it will roll back to the observation state and add a pseudo partial discharge mark. Risk level updates can be based on a comprehensive assessment of recurrence frequency, attribution stability, and operational stability. If the recurrence frequency increases within a short period, and the attribution branch, fault location, and operational status are stable, the risk level should be adjusted upwards. An increase in recurrence frequency can be defined as the number of recurrences within a single observation window reaching 1.5 to 3 times that of the previous observation window, typically 2 times. Using a relative multiple rather than a fixed increase is to accommodate abnormal events with different baseline recurrence frequencies and avoid deviations when using the same absolute increment for low-baseline and high-baseline events. Lower multiples are suitable for abnormal events with fewer baseline recurrences, while higher multiples are suitable for abnormal events with more baseline recurrences. If the recurrence frequency remains unchanged but attribution stability increases, such as from the branch level to the adjacent segment level, the risk level can also be appropriately adjusted upwards. If the operational status is unstable, or the external reference is in a state of high disturbance for a long period, the risk level should not be adjusted upwards, but only maintained or lowered. Risk level adjustments within a single observation window can be limited to one level to avoid short-term abnormal fluctuations causing a leap in risk level. When outputting monitoring conclusions, a formal conclusion record is generated. The formal conclusion record includes at least the following fields: object number, session number, event set number, monitoring time, authenticity status, branch attribution, fault location, risk level, current status description, suggested review time, and suggested handling time limit. The current status description records whether the current conclusion is in a state of recovery observation, downgrade review, multiple branches awaiting decomposition, or multiple locations awaiting decomposition. The suggested handling time limit is set progressively according to risk level to balance the needs of maintenance resource allocation and timely handling of high-risk anomalies. For example, a prompt level can be reviewed within 7 days, a concern level within 72 hours, a warning level within 24 hours for on-site inspection, and an alarm level for timely arrangement of special retesting or power outage maintenance preparation. The above time limit settings follow the principle that the higher the risk, the faster the handling. The formal conclusion record can be output using a fixed field format and written to the maintenance platform database for subsequent manual retesting, maintenance decisions, and historical record traceability. To ensure continuous operation under complex conditions, exception and resource constraint handling rules can be set. If the number of event sets within a certain observation window exceeds a preset limit, the event sets are sorted by risk level and most recent reproduction time, with higher-level event sets receiving priority for continuous review, while lower-level event sets are retained for basic observation. The statistical object here is the number of event sets within the observation window, not the number of exception events within a single session. This limit can be set to, for example, 100 sets. This setting is because if too many event sets within the observation window are all subject to continuous review in the same round, it will significantly increase the processing burden and reduce the stability of the judgment. This limit can also be adjusted based on on-site computing resources, observation window length, and communication capabilities. If the communication link is interrupted but the front-end cache is valid, the front-end continues to cache and retransmits the data after the link is restored. The original timestamp and event set mapping relationship are retained, and the conclusion time is not reset. If the link is interrupted for longer than the cache retention time, such as more than 24 hours, the current event set enters the data incomplete state, and the conclusion will no longer be upgraded, but only the existing level will be maintained. This duration is matched with the front-end cache retention capacity to avoid further upgrading the risk level in the case of long-term data loss. After the data integrity is restored and the continuous review conditions are met again, the normal level update path will be restored. If the accessory of a branch is replaced after maintenance, the maintenance completion time is marked in the object file, and the original event set of that branch is transferred to the archive state. New anomalies after maintenance should not directly inherit the old event set number, but should re-establish the event set, while retaining the comparison relationship with the historical record to avoid confusion between the anomaly chain before and after maintenance. After completing continuous review, status updates, and conclusion output, a monitoring conclusion table is generated. The monitoring conclusion table is used to summarize the current valid conclusions and includes at least the following fields: event set number, current risk level, current authenticity status, current branch attribution, current fault location, current status description, suggested review time, suggested handling time limit, whether escalation is allowed, and whether there is evidence for reverse exclusion. The formal conclusion record is used to save the results of a single formal output, while the monitoring conclusion table is used to summarize the current valid status. The two correspond to single records and continuous summaries, respectively. Through the above processing, the monitoring conclusion results required for subsequent status tracking, operation and maintenance handling, and historical record tracing can be generated.

[0023] Example 2: Based on Example 1, the specific application process of an online monitoring method for partial discharge faults in cable branch boxes is further explained: A 10kV cable branch box in normal energized operation can be selected as the monitoring object. This cable branch box includes, for example, four outgoing branches, referred to as the first branch, the second branch, the third branch, and the fourth branch. Inside the box, along each branch from the busbar side to the cable lead-out side, there are sequentially distributed wiring transition area, stress control accessory area, joint area, and cable lead-out near-end area. Partial discharge pulse acquisition points are set at the near end of each branch. A general reference point is set inside the box, and an external reference point is set outside the box. Accompanying monitoring points are set in the accessory concentration area of ​​each branch. Accompanying monitoring points can be used to collect ultrasonic, vibration, or other auxiliary anomaly records. The object file pre-stores information such as branch mapping relationship, measuring point mapping relationship, accessory chain position relationship, measuring point installation position, adjacent branch relationship, and stable triggering sequence under normal operating conditions to support subsequent monitoring and judgment. During operation, a monitoring session for the target cable branch box is first established, and a snapshot of the current operating condition is written. The snapshot includes at least the following fields: load status of each branch, total load status, box door status, operation permission status, ambient temperature, ambient humidity, the time of the most recent operation, and communication link status. After the monitoring session is established, records from each measuring point are received according to a unified timestamp standard, and a unified original acquisition record table is formed. Assuming that during a certain monitoring period, a short-term pulse record first appears at the near-end partial discharge pulse acquisition point of the second branch, followed by a record from the total reference point, and a local ultrasonic anomaly appears synchronously at the accompanying monitoring point corresponding to the stress control attachment area of ​​the second branch, while the external reference point is not triggered in the nearest time window, the relevant trigger records can be written into the unified original acquisition record table, and the records are sorted and organized according to the session base cycle, timestamp accuracy, and measuring point validity results. Based on this, the original records in the current monitoring session are merged. Records with trigger markers and valid measurement point judgments can be organized into candidate pulse entries. If the records of the near-end partial discharge pulse acquisition point of the second branch, the total reference point, and the corresponding accompanying monitoring point of the second branch appear within the merging time window and are all under the same working condition snapshot, and the trigger measurement point falls within the second branch or the adjacent range of the second branch, these records can be merged into the same abnormal event. If the fourth branch measurement point does not have a continuous first arrival record at this time, and the external reference point is not triggered before the first arrival measurement point, then the abnormal event is not judged as an external interference priority event. If, within the same monitoring session, the box door status does not change, the opening and closing and grounding switching records are empty, the load change does not reach the significant disturbance threshold, and the environmental temperature and humidity change does not exceed the significant fluctuation range within the preset environmental observation window, then the abnormal event is reviewed by the working condition freeze and enters the subsequent continuous acceptance judgment chain. In subsequent monitoring sessions, abnormal events verified by the operating condition freeze are continuously received. Assuming the near-end partial discharge pulse acquisition point of the second branch repeatedly appears as the first measurement point in subsequent observation rounds, the total reference point continues to trigger, the monitoring area remains concentrated near the stress control area of ​​the second branch, the power frequency phase segment remains within the same or adjacent segments, and the external reference state never becomes continuously dominant, then these cross-cycle abnormal events can be merged into the same event set. As the observation rounds progress, if this event set appears in the last two observation rounds... If two or fewer valid abnormal events occur, and the first-reaching measuring point never leaves the local measuring point group of the second branch, and no cross-branch jump occurs in the monitoring area, then the event set is upgraded from the observation state to the suspected real state. After continued observation, if the first-reaching measuring point remains stable in at least two subsequent observation rounds, the recurrence interval does not exceed the interruption threshold, and the external reference state is still not dominant, then the event set is further upgraded to the formal real state. Through the above processing, early discrete and intermittent abnormal records can be filtered from ordinary abnormalities into an abnormal event set with real partial discharge characteristics. Once the event set meets the attribution criteria, branch attribution is determined. Based on the pre-established branch propagation sequence template in the object file, the priority triggering measurement point for the second branch is the near-end partial discharge pulse acquisition point of the second branch. The set of following measurement points is allowed to include the overall reference point and near-end measurement points of branches adjacent to the second branch. The set of far-end triggering measurement points should not include measurement points of far-end branches not adjacent to the second branch. The historical sequence of the first-reaching measurement point, the historical sequence of the triggering measurement point set, and the historical sequence of the accompanying monitoring area in the most recent rounds of the current event set are compared with the second branch propagation sequence template. If the first-reaching measurement point consistently falls on the second branch... If the primary triggering measuring point or its local drift range consistently shows the total reference point as the following measuring point, and the primary triggering measuring point at a distance does not consistently arrive first, and the accompanying monitoring area is consistent with the sensitive area of ​​the stress control attachment area of ​​the second branch, then the second branch can be identified as the primary candidate branch. If the first branch also shows a local similar response in individual rounds, but at least two of its differences in the number of first arrivals, the number of times the accompanying area is concentrated, and the number of times it has a continuous historical recurrence are weaker than those of the second branch, then the first branch is recorded as the secondary candidate branch, and the second branch forms the formal branch affiliation. In this way, abnormal events can be located from the background of multi-branch interference to specific branches. After identifying the second branch as the faulty branch, the fault location is confirmed along the accessory chain of the second branch. Assume the accessory chain of the second branch is sequentially divided into a wiring transition section, a stress control section, a joint body section, an insulating sleeve section, and a lead-out proximal section. When confirming the location, first determine if the accompanying monitoring area is stably concentrated in the sensitive area corresponding to the stress control section of the second branch. Then determine if the response sequence of the first reaching measurement point and the total reference point conforms to the proximal characteristic of the stress control section being located in the middle to front of the branch. Next, determine if the concentrated location of historical abnormal events within the second branch is consistently concentrated in the stress control section. If at least two of the above judgments point to the stress control section of the second branch, then that section is identified as the formal fault location. If the subsequent accompanying monitoring area further concentrates within a localized area of ​​the stress control accessory area, and the first reaching sequence becomes more stable over multiple observation cycles, then the conclusion at the accessory section level can be further refined to a specific component level conclusion for the specific stress control accessory area. If the on-site measurement point resolution is insufficient to support further refinement, then the accessory section level location conclusion is maintained, and the granularity is not forcibly increased. After confirming the branch attribution and fault location, the event set is continuously reviewed and its status updated. If the abnormal events in the stress control section of the second branch continue to reproduce in subsequent observation windows as the same branch and the same adjacent section, and the operating condition is stable and the external reference status does not provide evidence to reverse the exclusion, the original attribution conclusion is retained, and the risk level is updated according to the recurrence frequency, attribution stability, and degree of operating condition stability. Assuming the initial risk level of the event set is "attention level," and the number of recurrences in a subsequent observation window reaches twice that of the previous observation window, while the attribution of the second branch remains stable... If the fault location remains locked in the stress control section, the risk level of the event set can be upgraded from the attention level to the warning level. If it continues to recur frequently in a shorter period of time and the external reference state is still not dominant, it can be further upgraded to the alarm level. Conversely, if conflict evidence appears later, such as the external reference point being reached and dominant in most recurring events, or the main candidate branch changing frequently in multiple observation rounds, the current event set will no longer be upgraded along the original level, but will regress to a lower authenticity state, a lower attribution level, or a state to be decomposed according to the conflict type. If the event set no longer reproduces within subsequent observation windows, it is transferred to the recovery observation state, and the duration of continuous missing events continues to be recorded. If the duration of continuous missing events exceeds the corresponding recovery observation threshold, such as reaching the recovery observation threshold corresponding to the warning level, the original conclusion is downgraded to a historical abnormal state. If, during the recovery observation period, the abnormal event in the stress control section of the second branch reappears and meets the acceptance conditions again, the original continuous acceptance chain is restored and the review continues. If the accessories of the second branch are subsequently repaired or replaced, the repair completion time is recorded in the object file, and the event set formed before the repair is transferred to the archive state. Abnormal events that reappear after the repair are completed should not directly inherit the original event set number, but should be re-established, while retaining the correspondence with the historical records, so as to determine whether the events before and after the repair belong to the continuation of the same fault chain or a new source of abnormality. When outputting monitoring conclusions, a formal conclusion record and a monitoring conclusion table are generated. The formal conclusion record includes at least the following fields: object number, monitoring time, event set number, authenticity status, branch attribution, fault location, risk level, current status description, recommended review time, and recommended handling time limit. In this example, the following conclusion can be drawn: There is a continuously recurring formal real-state abnormal event in the second branch of the target cable branch box. The branch is assigned to the second branch, the fault location is the stress control section, and the risk level is warning level. It is recommended to arrange on-site inspection within 24 hours, and decide whether to carry out power outage maintenance preparation based on the subsequent retest results. The monitoring conclusion table is used to summarize the current valid status for subsequent operation and maintenance, status tracking, and historical record tracing.

[0024] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0025] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above embodiments can be implemented in whole or in part by a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions of the embodiments of this application are implemented in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted wirelessly or wiredly from one website, computer, server, or data center to another website, computer, server, or data center. Wired methods include optical fiber, twisted pair, coaxial cable, etc. Wireless methods include infrared, microwave, etc. Available media include any available media that can be accessed by a computer or data storage devices such as servers and data centers that contain one or more sets of available media. Available media can be magnetic media (floppy disks, hard disks, magnetic tapes), optical media (DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0026] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0027] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for online monitoring of partial discharge faults in cable branch boxes, characterized in that, include: S1. Establish a monitoring session for the target cable branch box, read the branch box branch relationship, accessory distribution, measurement point correspondence and current operating conditions, and simultaneously acquire the partial discharge pulse record, external reference record and accompanying monitoring record of each monitoring point; S2. Merge the records of each monitoring point based on a unified time caliber, and combine external reference records and current operating conditions to investigate interference and verify the freezing of operating conditions for abnormal events; S3. Make continuous acceptance judgments on abnormal events that have passed the working condition freeze verification, and make authenticity judgments based on the response sequence, recurrence relationship and regional concentration of abnormal events among various monitoring points. S4. For abnormal events that pass the authenticity determination, determine the branch attribution based on the branch propagation order, branch adjacency, and attachment position, and confirm the fault location along the attachment chain of the corresponding branch. S5. Continuously review and update the status of abnormal events that have completed the attribution determination, and output monitoring conclusions based on the attribution status after review.

2. The method for online monitoring of partial discharge faults in cable branch boxes according to claim 1, characterized in that, Establish a monitoring session for the target cable branch box, read the branch box branch relationships, accessory distribution, measurement point correspondence, and current operating conditions, and simultaneously acquire partial discharge pulse records, external reference records, and accompanying monitoring records from each monitoring point, including: When establishing a monitoring session, an object file containing the branch attachment chain order and the measurement point layout relationship is invoked to generate a monitoring session record containing a working condition snapshot, a measurement point validity table, an original record cache area, and a subsequent record area. Call the branch mapping table, attachment chain list and measurement point mapping table, receive partial discharge pulse records, external reference records and accompanying monitoring records according to the unified timestamp caliber and write them into the unified original acquisition record table, and perform measurement point access validity judgment. When the measuring point is abnormal, it enters a frozen state awaiting recovery; under operational disturbance conditions, it enters a restricted monitoring state.

3. The online monitoring method for partial discharge faults in cable branch boxes according to claim 1, characterized in that, Event merging based on a unified time metric was performed on records from various monitoring points, including: The original acquisition records in the current monitoring session are sorted according to a unified time caliber, and candidate pulse entries are generated for records with trigger markers and that pass the validity judgment of the measurement points. Based on the external reference synchronization relationship and accompanying record matching relationship within the nearest time window, the entries are associated, and within the merging time window, they are merged according to the time sequence relationship within the same monitoring session, the same working condition snapshot, and the same branch, adjacent branches, or the total reference association range, and abnormal events are generated. When multiple measuring points cannot be distinguished in order, the relevant measuring points are jointly marked as the first measuring point in parallel.

4. The method for online monitoring of partial discharge faults in cable branch boxes according to claim 1, characterized in that, In conjunction with external reference records and current operating conditions, interference investigation and operational freeze verification are conducted for abnormal events, including: For abnormal events resulting from the merger, conduct external disturbance investigation, internal spatial distribution investigation, and operational condition freeze verification. Based on the sequential relationship between external reference points and first-reach measurement points, the cross-branch distribution of internal trigger measurement points, and the concentration of accompanying monitoring areas, the interference state of abnormal events is determined. The operation events, changes in door status, load changes, and environmental fluctuations are reviewed in conjunction with the operating condition freeze window before and after the abnormal event. Based on the review results, the abnormal event is transferred to the freeze observation state or subsequent continuous acceptance judgment.

5. The online monitoring method for partial discharge faults in cable branch boxes according to claim 1, characterized in that, Continuous acceptance judgment is performed on abnormal events that pass the operational condition freeze review, including: An event acceptance table is established for abnormal events that pass the working condition freeze verification, and event set matching is performed based on the historical sequence of the first measurement point, the historical sequence of the branch candidate, the historical sequence of the power frequency phase section, the historical sequence of the accompanying monitoring area, and the historical sequence of the external reference state. When the first measurement point is the same or located within the allowable drift range, the power frequency phase segment is the same or adjacent, the accompanying monitoring area corresponds to the same area or the same adjacent segment, the branch candidate set has an intersection and the external reference state has not turned into continuous dominance, the new abnormal event will be merged into the existing event set. If the matching conditions are not met, create a new event set.

6. The online monitoring method for partial discharge faults in cable branch boxes according to claim 1, characterized in that, The authenticity of the event is determined based on the sequence of responses, recurrence, and regional concentration of abnormal events among monitoring points, including: The event set is judged based on the stability of the response sequence, the persistence of the recurrence relationship, and the consistency of the regional concentration. When at least two decision dimensions meet the stability condition, the event set is transformed into a suspected real state; When the external reference state is not dominant, the first measurement point remains stable, and the recurrence interval does not exceed the interruption threshold, the event set is converted to the formal real state. Under conditions of parallel first arrival, weak intermittent anomalies, and high disturbance background, the results are verified by combining the accompanying monitoring area, operating conditions, and external reference conditions to form an authenticity judgment result table.

7. The method for online monitoring of partial discharge faults in cable branch boxes according to claim 1, characterized in that, For abnormal events that pass the authenticity check, branch attribution is determined based on branch propagation order, branch adjacency, and attachment location, including: The branch propagation sequence template, which is established based on the branch mapping relationship, the measuring point mapping relationship, the positional relationship of the attachment chain, and the stable triggering sequence under normal working conditions, is invoked. The consistency of the historical sequence of the first measuring point, the historical sequence of the triggering measuring point set, and the historical sequence of the accompanying monitoring area in multiple observation rounds of the event set is compared to determine the main candidate branch. When adjacent candidate branches coexist, conflict is determined based on the difference in the number of first arrivals, the difference in the number of times the accompanying area is concentrated, and the difference in the number of times the historical continuous recurrence occurs. When the event set reaches the formal real state and the main candidate branch is consistent in multiple consecutive observation rounds, the formal branch assignment is formed.

8. The method for online monitoring of partial discharge faults in cable branch boxes according to claim 1, characterized in that, And confirm the location of the fault along the corresponding branch's accessory chain, including: Based on the positional relationship of the attachment segments in the object file, the accompanying monitoring sensitive areas, and the relative distance relationship with each measuring point, abnormal events are located hierarchically along the attachment chain of the belonging branch; By combining the concentrated location of the accompanying monitoring area, the arrival order of the first measuring point and the total reference point, and the continuous concentrated location of historical abnormal events within the branch road, the attachment section is compared, and an attachment section-level location conclusion is formed when at least two positioning bases point to the same attachment section. When only a single location is clearly defined, the lower-level location conclusions are retained; when there is no clear segment, the branch-level conclusions are retained; and when the attached segment is clearly defined and the accompanying monitoring area and first arrival sequence within the same attached segment are further stabilized, the location conclusions are refined to the specific component level.

9. The online monitoring method for partial discharge faults in cable branch boxes according to claim 1, characterized in that, Continuously review and update the status of abnormal events that have completed attribution determination, including: Establish a continuous review form, and set observation windows and recovery observation thresholds according to risk levels; Based on the consistency of event set reproduction within the observation window, duration of continuous missing events, external reference status, operating status, branch attribution stability, and fault location stability, continuous verification and status updates are performed. When abnormal events continue to recur, the original attribution conclusion is retained and the risk level is adjusted; when abnormal events stop recurring, the observation period is resumed. When external references become dominant, branch lines frequently drift, fault locations change across sections, or operating conditions become abnormal for a long period, a rollback review is performed, and the risk level adjustment within a single observation window is limited to one level.

10. The method for online monitoring of partial discharge faults in cable branch boxes according to claim 1, characterized in that, Based on the verified attribution status, the monitoring conclusions are output, including: Based on the authenticity status after continuous verification, branch attribution, fault location, risk level, current status description, recommended verification time and recommended handling time limit, a formal conclusion record is generated and a monitoring conclusion table is formed; When the risk level changes progressively, the review requirements and handling recommendations shall be adjusted in accordance with the corresponding handling time limits; When the data is under the following conditions: recovery observation, downgrade review, multiple branches to be decomposed, multiple parts to be decomposed, or data is incomplete, the corresponding status information will be written into the formal conclusion record and monitoring conclusion table.

Images