Cable branch box fault location method based on circuit breaker and time series data

Abstract

This application discloses a fault location method for cable branch boxes based on circuit breakers and timing data. The method includes: establishing a power supply topology model by reading branch box configuration data; continuously collecting circuit breaker action records and operational timing data such as branch current, bus voltage, and branch power; and performing preprocessing. When a circuit breaker trips or abnormal trips are detected, a circuit breaker action chain is constructed. Combined with power supply topology analysis, near-synchronous power outages and secondary abnormal behaviors after restoration are analyzed to obtain a candidate branch set. Furthermore, operational characteristics are extracted around the power restoration process and compared with historical healthy baselines to calculate the abnormal deviation. Collaborative reasoning is then performed based on the power supply path topology relationship to achieve fault location. This method can achieve fault location based on existing monitoring systems.

Description

Technical Field

[0001] This application relates to the field of power distribution monitoring technology, and more specifically, to a method for fault location of cable branch boxes based on circuit breaker and time-series data. Background Technology

[0002] In communication equipment rooms, communication base stations, and various sites with high continuous power supply requirements, cable distribution boxes typically serve as important power distribution nodes in low-voltage or DC power supply systems. They are used to distribute upstream power to various loads such as rectifiers, main communication equipment, air conditioning equipment, and battery management equipment. A distribution box generally houses an upstream power supply circuit breaker, a main incoming circuit breaker, and multiple branch circuit breakers. Each branch is connected to its corresponding load via cables or terminals. When a fault occurs in a branch cable, connector, or terminal, it often causes the circuit breaker to trip or trip abnormally, potentially accompanied by power outages or voltage fluctuations in multiple branches, thus affecting the stable operation of the site equipment.

[0003] In existing operation and maintenance systems, fault location in cable branch boxes typically relies on manual inspections, segment-by-segment checks, or experience-based judgment based on circuit breaker trip records. When multiple branches experience simultaneous power outages or upstream power supply disturbances, multiple circuit breakers may trip sequentially within a short period, leading to a discrepancy between the tripping sequence and the actual fault location. Relying solely on the tripping information of a single circuit breaker is often insufficient to accurately pinpoint the fault section. Furthermore, some connection-related faults, such as loose joints, poor contact, or localized insulation degradation, may not manifest as obvious electrical anomalies during steady-state operation, but may exhibit dynamic characteristics such as delayed current recovery, short-term voltage drops, or power fluctuations during power restoration or load re-establishment. These types of faults are easily overlooked by traditional monitoring methods if a systematic analysis approach is lacking.

[0004] On the other hand, some existing fault location methods rely on the installation of additional dedicated sensors or online monitoring equipment to acquire more electrical parameters to assist in fault location by adding hardware. However, in many existing sites, on-site equipment space is limited, and the cost of adding new hardware is high, while maintenance complexity also increases. Therefore, under the existing monitoring system, using only circuit breaker operation records and existing operating data such as branch current, bus voltage, and branch power to achieve fault location has become an important requirement in operation and maintenance practice. Summary of the Invention

[0005] This application provides a method for fault location of cable branch boxes based on circuit breakers and timing data, so as to at least solve the technical problems existing in the related technologies described above.

[0006] According to a first aspect of the embodiments of this application, a method for fault location of a cable branch box based on circuit breaker and timing data is provided, including: Read the branch box configuration data, establish a power supply topology model that includes the upstream power supply circuit breaker, the main incoming circuit breaker, the branch circuit breaker, and the correspondence between branches and loads, and generate downstream coverage sets and branch power supply paths for each circuit breaker; Continuously collect the action records of each circuit breaker and the runtime sequence data corresponding to each branch, and perform preprocessing on the action records and runtime sequence data; Based on preprocessed action records and runtime sequence data, when a circuit breaker trips or abnormal trips are detected, an action chain analysis window is constructed at the trigger time, all action records in the window are extracted, and a circuit breaker action chain is formed in chronological order. Based on the power supply topology model, the near-synchronous power outage, upstream associated action, and secondary tripping after recovery in the circuit breaker action chain are analyzed to obtain a set of candidate branches after protection coordination distortion correction. For each candidate branch, a power supply event window is constructed based on the time when the power supply is restored. Branch current, bus voltage and branch power are extracted to generate event feature vectors. A health response baseline is established based on historical health events. The abnormal deviation of the current event feature vector of each candidate branch relative to the health response baseline is calculated. Collaborative reasoning is performed by combining the candidate branch set, the abnormal deviation, and the common topological constraints of the branch power supply path to output the fault location result.

[0007] As an optional approach, establishing the power supply topology model includes: reading the upstream power supply circuit breaker number, main incoming circuit breaker number, branch circuit breaker number, branch number, load type, and connection relationship from the branch box configuration record; constructing a directed hierarchical structure according to the power supply direction; determining the set of direct downstream nodes for each circuit breaker based on the directed hierarchical structure, and generating a downstream coverage set according to the path relationship between the circuit breaker and the load; for each branch record, the branch power supply path is composed of the upstream power supply circuit breaker, the main incoming circuit breaker, and the corresponding branch circuit breaker in sequence; when the configuration data changes, reloading the configuration data and rebuilding the corresponding power supply topology model.

[0008] As an optional approach, the formation of the circuit breaker action chain includes: establishing an action chain analysis window that backtracks forward and tracks backward around the triggering time; extracting all circuit breaker action records within the action chain analysis window and arranging them in ascending order according to the action timestamp to form an action sequence; and attaching a topology level identifier and downstream coverage set in the power supply topology model to each action record.

[0009] As an optional approach, the process of obtaining the candidate branch set after protection coordination distortion correction specifically involves: comparing the time differences between the power-off actions of each branch circuit breaker, and classifying power-offs with time differences less than the synchronization judgment threshold as near-synchronous power-offs; when multiple branches experience near-synchronous power-offs and there are tripping, opening, or undervoltage associated records of the upstream power supply circuit breaker or the main incoming line circuit breaker during the corresponding time period, the event is marked as a common power supply disturbance event; branches that independently trip again after power restoration or exhibit independent abnormal recovery behavior in the power supply event window are incorporated into the candidate branch set.

[0010] As an optional approach, for each candidate branch, a power supply event window is constructed based on its power restoration time. Branch current, bus voltage, and branch power are extracted to generate an event feature vector. This includes: using the corresponding aligned time among the reclosing time, power restoration time, or bus voltage restoration time of the candidate branch as the event center, constructing a power supply event window containing the observation period before closing and the observation period after closing; extracting the branch current, bus voltage, and branch power of the corresponding candidate branch within the power supply event window and forming an event data matrix; performing short-window smoothing and local denoising processing on the time-series signals in the event data matrix; and calculating the current recovery slope, recovery stabilization time, short-time voltage drop characteristics, and power recovery deviation based on the processed time-series signals to constitute the event feature vector of the candidate branch.

[0011] As an optional approach, calculating the recovery stabilization time includes: constructing a normalized current change rate based on the branch current, wherein the normalized current change rate is determined by the ratio of the current change at adjacent sampling times to the rated current of the branch; searching within the power supply event window for the earliest time when multiple consecutive sampling points satisfy the normalized current change rate not exceeding the stability judgment threshold, and determining the earliest time as the stabilization time; determining the time difference between the stabilization time and the event center as the recovery stabilization time; and determining the short-time voltage drop characteristic by the difference between the bus reference voltage before closing and the minimum bus voltage within the event window after closing, or their normalized ratio.

[0012] As an optional approach, establishing a health response baseline based on historical health events includes: selecting historical health events from the historical database that have not experienced circuit breaker tripping or tripping, have consistent power supply modes, similar load levels, and have not undergone external maintenance switching; extracting the event feature vectors corresponding to each historical health event to form a health sample set; calculating the health mean vector and covariance matrix based on the health sample set; removing outliers that deviate from the health sample group and recalculating the health mean vector and covariance matrix; and adding a diagonal stabilizing term to the covariance matrix when the inversion of the covariance matrix is ​​unstable to obtain a regularized covariance matrix.

[0013] As an optional approach, calculating the anomaly deviation includes: using the current event feature vector of the candidate branch, the health mean vector, and the covariance matrix or the regularized covariance matrix, calculating the Mahalanobis distance of the candidate branch as the anomaly deviation; determining an anomaly judgment threshold based on the distance distribution of each historical health event relative to the health mean vector; and comparing the anomaly deviation of the current candidate branch with the anomaly judgment threshold to obtain the anomaly judgment result of the candidate branch.

[0014] As an optional approach, the output fault location result includes: when the candidate branch set contains only one branch and the abnormal deviation of that branch exceeds the abnormal judgment threshold, the end segment of the power supply path corresponding to that branch is determined as the fault segment; when the abnormal deviation of multiple candidate branches exceeds the abnormal judgment threshold, the branch power supply path of each abnormal branch is obtained and the common intersection of the paths is calculated. When the common intersection of the paths is not empty and contains the upstream power supply circuit breaker or the main incoming line circuit breaker, the corresponding common power supply path is determined as the fault segment.

[0015] According to a second aspect of the embodiments of this application, a cable branch box fault location system based on circuit breaker and timing data is also provided for implementing the above method, including: The topology modeling module is used to read the branch box configuration data, establish a power supply topology model that includes the upstream power supply circuit breaker, the main incoming circuit breaker, the branch circuit breaker, and the correspondence between the branches and the load, and generate the downstream coverage set and branch power supply path for each circuit breaker. The data acquisition and preprocessing module is used to continuously acquire the action records of each circuit breaker and the runtime sequence data corresponding to each branch, and to perform preprocessing on the action records and runtime sequence data. The action chain construction module is used to construct an action chain analysis window based on pre-processed action records and runtime sequence data when a circuit breaker trips or abnormal trips are detected. It extracts all action records within the window and forms a circuit breaker action chain in chronological order. The candidate branch analysis module is used to analyze the near-synchronous power outage, upstream associated action and secondary tripping after recovery in the circuit breaker action chain based on the power supply topology model, and obtain a set of candidate branches after protection coordination distortion correction. The event feature generation module is used to construct a power supply event window for each candidate branch based on the time when the power supply is restored, extract the branch current, bus voltage and branch power, and generate an event feature vector. The anomaly assessment and collaborative reasoning module is used to establish a health response baseline based on historical health events, calculate the anomaly deviation of the current event feature vector of each candidate branch relative to the health response baseline, and perform collaborative reasoning by combining the candidate branch set, the anomaly deviation, and the common topology constraints of the branch power supply path to output the fault location result.

[0016] This application constructs a power supply topology model that includes the relationships between upstream power supply circuit breakers, main incoming circuit breakers, and branch circuit breakers. It then performs joint analysis of circuit breaker action records with operational sequence data such as branch current, bus voltage, and branch power. After detecting a tripping or abnormal tripping event, a circuit breaker action chain is formed. Furthermore, by combining near-synchronous power outage relationships, upstream associated actions, and secondary abnormal behaviors after recovery, candidate branches are screened. This corrects the traditional method of locating circuit breakers that relies solely on single tripping information.

[0017] Meanwhile, by constructing a power supply event window around the power restoration process, features such as current recovery slope, recovery stabilization time, and short-term voltage drop are extracted and compared with a baseline established from historical health events, effectively identifying abnormal responses during the recovery phase. Furthermore, by combining collaborative reasoning with the common topology relationships between branch power supply paths, the fault section can be accurately determined in cases of multi-branch disturbances or cascading actions. This method requires no additional field sensors and can be deployed on top of existing monitoring systems, not only improving the accuracy and reliability of cable branch box fault location but also helping to reduce the workload of manual segment-by-segment troubleshooting and improving operation and maintenance efficiency.

[0018] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Furthermore, no embodiment in this disclosure is required to achieve all the effects described above. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0020] Figure 1 This is a schematic diagram of a cable branch box fault location method based on circuit breaker and timing data, provided in an embodiment of this disclosure.

[0021] Figure 2 This is a schematic diagram of the circuit breaker action chain formation process provided in an embodiment of the present disclosure.

[0022] Figure 3 This is a schematic diagram of the process for forming a candidate branch set according to an embodiment of the present disclosure.

[0023] Figure 4 This is a schematic diagram of the event feature vector generation process provided in an embodiment of this disclosure.

[0024] Figure 5 This is a schematic diagram of a cable branch box fault location system based on circuit breaker and timing data, provided in an embodiment of this disclosure. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0026] In this embodiment, the method is applicable to cable branch boxes in communication equipment rooms, communication base stations, access nodes, edge sites, or other scenarios with high continuous power supply requirements. An external power supply is connected upstream of the branch box, and the box contains an upstream power circuit breaker, a main incoming circuit breaker, and multiple branch circuit breakers. Each branch is connected to a rectifier, main communication equipment, air conditioning equipment, battery management equipment, or auxiliary loads. The executing entity can be a branch box monitoring terminal, a station-side acquisition unit, a power distribution monitoring server, or an analysis and processing unit in an operation and maintenance platform. This method utilizes existing circuit breaker action records and operational sequence data such as branch current, bus voltage, and branch power to locate fault sections, without relying on additional hardware sensors, making it suitable for direct deployment within existing monitoring systems.

[0027] First, some nouns or terms that appear in the description of the embodiments of this application shall be interpreted as follows: In some embodiments, the branch power supply path in the method refers to a path sequence consisting of the upstream power supply circuit breaker, the main incoming circuit breaker, and the corresponding branch circuit breakers arranged in the power supply direction. The downstream coverage set refers to the set of all downstream branches or load objects that would theoretically lose power after a circuit breaker trips. The action chain analysis window refers to a time interval that traces backward and forward around the trigger moment. The action cluster refers to a group of adjacent action records with a time interval less than a preset action cluster division threshold.

[0028] The common power supply disturbance event refers to an event in which multiple branches experience near-synchronous power outages while simultaneously experiencing associated actions of the upstream power supply circuit breaker or the main incoming circuit breaker. The power supply event window refers to the observation interval established around the time the candidate branch restores power. The health response baseline refers to a statistical reference of features extracted from historical health events under the same power supply mode and similar load conditions. The anomaly deviation refers to the degree of deviation of the current event feature vector from the health response baseline. The common topology constraint means that the power supply paths of multiple abnormal branches must have an interpretable common upstream segment in the power supply structure.

[0029] The implementation process of the method described in this application will be described in detail below with reference to specific embodiments. It should be noted that this embodiment is only used to explain this application and is not intended to limit the scope of protection of this application. Conventional adjustments or substitutions of each step by those skilled in the art without departing from the concept of this application should be included in the scope of protection of this application.

[0030] Figure 1 This is a flowchart of a cable branch box fault location method based on circuit breaker and timing data according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes steps S1-S6: In step S1, the branch box configuration data is read, a power supply topology model is established including the upstream power supply circuit breaker, the main incoming circuit breaker, the branch circuit breaker, and the correspondence between branches and loads, and a downstream coverage set and branch power supply path are generated for each circuit breaker.

[0031] In this embodiment, the branch box configuration data is read first. The configuration data can originate from a ledger database, maintenance configuration files, edge terminal cache files, or the configuration interface of the power distribution master station. The read content includes at least the upstream power supply circuit breaker number, the main incoming circuit breaker number, the branch circuit breaker number, the branch number, the load type, and the connection relationship. The branch number uniquely identifies a specific output branch within the branch box, and the load type characterizes the main power supply object category of the branch, such as a rectifier load, communication equipment load, air conditioning load, or a mixed load.

[0032] Specifically, the system performs field integrity checks on the original configuration records. When a missing circuit breaker number, duplicate branch number, broken connection relationship, or circular connection description is found, the corresponding record is entered into the configuration anomaly log and the system does not proceed to the formal topology construction process. Subsequent action chain analysis, candidate branch screening, and common path intersection all rely on a clear directed power supply structure. If there are breaks or loops at the configuration level, it will cause circular dependencies in subsequent logic.

[0033] In one implementation, the system constructs a directed hierarchical structure according to the power supply direction. The root layer is the upstream power supply circuit breaker, the second layer is the main incoming line circuit breaker, the third layer is multiple branch circuit breakers, and the fourth layer is the load object or load category identifier corresponding to the branch. If there are intermediate protection nodes on site, such as busbar sectionalizing circuit breakers, area distribution circuit breakers, or standby switching nodes, these nodes are inserted into the corresponding layers according to the actual power supply sequence.

[0034] Furthermore, the system determines the set of direct downstream nodes for each circuit breaker. Let the circuit breaker node be denoted as . Its direct downstream node set is denoted as If a node has no directly downstream circuit breakers but corresponds to one or more load objects, then that load object is considered the node's final downstream member. Based on direct downstream relationships, the system recursively generates the downstream coverage set for each circuit breaker. Let the circuit breaker... The downstream coverage set is ,but It consists of all successor branches of the node and the load objects. Once generated, this set will not change its semantics and will only be rebuilt as a whole when the configuration changes.

[0035] For each branch record, the system simultaneously generates the branch power supply path. Let a certain branch be denoted as... Its power supply path is denoted as In the most basic scenario, It consists of the corresponding upstream power supply circuit breaker, the main incoming circuit breaker, and the branch circuit breaker. If there are additional intermediate nodes on site, they are inserted according to the power supply sequence. The branch power supply path is used in subsequent collaborative reasoning to find the common path intersection of multiple abnormal branches.

[0036] The topology model can be stored using a dual-table approach: a node table and a path table. The node table stores circuit breaker numbers, node types, hierarchical identifiers, sets of direct downstream nodes, and sets of downstream coverage. The path table stores branch numbers, load types, corresponding branch circuit breaker numbers, and branch power supply paths. The node table primarily serves action chain analysis, while the path table primarily serves candidate branch inference and fault segment output. Both share the same topology version number. Whenever configuration data changes, such as circuit breaker replacement, branch migration, branch expansion, or load adjustment, the system reloads the configuration and generates a new version of the topology model. Older versions are only used for historical event tracing and do not participate in new event inference.

[0037] In step S2, the operation records of each circuit breaker and the runtime sequence data corresponding to each branch are continuously collected, and the operation records and runtime sequence data are preprocessed.

[0038] After the power supply topology model is established, the system continuously collects circuit breaker action records and runtime sequence data corresponding to each branch. Action records include at least the circuit breaker number, action type, and action timestamp. The action type is preferably standardized using standard enumerated values ​​such as closing, opening, tripping, reclosing, and undervoltage-related actions. Runtime sequence data includes at least the timestamp, branch number, branch current, bus voltage, and branch power. In some embodiments, power supply mode identifiers, maintenance status identifiers, ambient temperature, or site load level identifiers may also be added; however, this additional information is only used as filtering conditions and does not change the core analysis link.

[0039] Since action logs and timing data may come from different acquisition sources, the system first performs time alignment processing. Specifically, the action timestamp and sampling timestamp are mapped to a unified clock source. The unified clock source can be from station-level unified time synchronization, network time synchronization, or a time base issued by the master station.

[0040] Subsequently, missing data processing is performed. For single-point missing data, if valid sample values ​​exist before and after the missing point, adjacent valid sample points are used to fill the gap. Preferably, linear interpolation is used for current and power, and proximity hold or linear interpolation is used for bus voltage.

[0041] The system also performs deduplication on duplicate records. For action records, a joint determination is made based on the circuit breaker number, action type, and timestamp. If all three are identical, the record is considered a duplicate and only one is retained. For time-series data, if multiple records exist for the same branch at the same sampling time, the record with the normal acquisition status and higher source identifier priority is retained first. The source identifier priority is preset based on a combination of device protocol stability, acquisition link integrity, and historical effectiveness, and can be pre-set in the device management table during actual implementation.

[0042] For state validity verification, the system focuses on checking for situations where the same circuit breaker simultaneously exhibits mutually exclusive states. For example, if both closing and tripping occur within the same time slice, or both opening and closing confirmation occur simultaneously. For such conflicting records, they are first sorted chronologically, and then corrected according to equipment protocol priority. If the state transition logic is still not satisfied after correction, the record is discarded and not included in subsequent event analysis.

[0043] In step S3, based on the pre-processed action records and runtime sequence data, when a circuit breaker trips or abnormal trips are detected, an action chain analysis window is constructed at the trigger time, all action records in the window are extracted, and a circuit breaker action chain is formed in chronological order.

[0044] Figure 2 A schematic diagram illustrating the circuit breaker action chain formation process provided in an embodiment of this disclosure is shown. Figure 2 As shown, in step S201, the triggering time is determined by detecting the tripping or abnormal opening of any circuit breaker, and an action chain analysis window is constructed.

[0045] During continuous monitoring, once any circuit breaker trips or abnormally trips, the system uses the moment of occurrence of this action as the trigger time. An abnormal trip refers to a tripping action that does not correspond to planned maintenance, manual remote control, or a registered switching operation. The system can filter planned actions by querying the maintenance plan, control command log, and switching permission records. Only after confirming that it is not a planned operation will the system proceed to the fault event analysis process.

[0046] Let the trigger time be denoted as The system revolves around Construct an action chain analysis window. Let the forward backtracking duration be... The backward tracking time is The action chain analysis window is denoted as ; in, Used to cover potential voltage collapse, pilot tripping, or undervoltage triggering actions that may occur before a fault occurs. This is used to cover cascaded tripping, reclosing, power restoration, and secondary tripping after restoration. Both parameters are time-based. Their values ​​can be configured according to the sampling granularity and site protection coordination characteristics. If the sampling granularity is fine and the circuit breaker action chain is short, it can be appropriately shortened; if delayed reclosing or manual reset is common in the field, it can be appropriately lengthened. Parameter updates are only completed at the system configuration layer and are not dynamically changed in a single event processing to maintain comparability between similar events.

[0047] In step S202, based on all circuit breaker operation records, the operation sequences are formed by sorting them in ascending order of operation timestamps. The system then extracts the operation sequence from the operation database. All circuit breaker operation records are stored and arranged in ascending order by operation timestamp to form an operation sequence. Each action record At a minimum, it includes the circuit breaker number, action type, and action time. Furthermore, a topology level identifier and the downstream coverage set of the circuit breaker to which it belongs are added to each action record. With the level identifier, the system can distinguish between upstream actions and branch actions; with the downstream coverage set, the system can determine which branches a given action might theoretically affect.

[0048] Optionally, to form timing units more suitable for subsequent protection and distortion identification, the system continues to calculate the time interval between adjacent action records. Let the first... Action and the first The time interval between each action is: ; in, For the first The timestamp of the action. If Less than the action cluster segmentation threshold If so, these two actions will be grouped into the same action cluster. Action cluster classification threshold. This is an important time parameter, representing the maximum permissible time interval between a set of actions that can be considered as nearly simultaneous electrical responses. It can be determined by pre-setting or by statistical analysis of historical action chains. If a statistical method is used, the distribution of time differences between adjacent actions can be extracted from historical normal switching and typical fault actions, and a quantile value that balances the risks of erroneous and missed connections can be selected as the time difference. .

[0049] Once action clusters are constructed, each action cluster corresponds to a more stable timing unit. In some events, upstream circuit breaker tripping and multiple branch power outage records may arrive consecutively within a very short period. Directly processing individual records is easily affected by communication delays. By using action clusters, cluster-level relationships can be identified first, and then record-level details can be preserved within the cluster, thereby improving the robustness of subsequent analysis.

[0050] In step S4, based on the power supply topology model, the near-synchronous power outage, upstream associated action, and secondary tripping after recovery in the circuit breaker action chain are analyzed to obtain a set of candidate branches corrected for protection coordination distortion.

[0051] Figure 3 A schematic diagram illustrating the candidate branch set formation process provided in an embodiment of this disclosure is shown. Figure 3 As shown, in step S301, the power-off actions of each branch circuit breaker are extracted from the action chain, and synchronization is determined. In this embodiment of the disclosure, relying solely on the first action or a single tripped circuit breaker is insufficient to accurately characterize the actual fault section. In communication power supply scenarios, upstream voltage disturbances, cascade tripping, recovery impacts, and power switching delays are common, which can easily lead to distortions in protection coordination.

[0052] Protection coordination distortion refers to a discrepancy between the apparent sequence of actions in the action chain and the actual location of the fault. For example, the fault may actually be located in a branch connection section, but the first tripping occurs at the main incoming circuit breaker; or a short-term disturbance upstream may cause multiple branches to lose power almost simultaneously, with only some branches showing abnormalities again during the recovery phase.

[0053] To correct the aforementioned distortion, the system first extracts the power-off actions of each branch circuit breaker from the action chain. A power-off action can be either tripping or an abnormal tripping that causes a power outage downstream without planned operation. The time difference between any two branch power-off actions is denoted as: ; in, and These represent the times when the two branch circuit breakers tripped due to power failure. If... If so, it is determined that the two branches have a near-synchronous power-down relationship. Parameter The synchronization threshold represents the maximum time difference used to determine whether power outages in different branches can be attributed to the same common disturbance. This parameter is related to the action cluster segmentation threshold. Different meanings It acts on adjacent record groups within the action chain. This function is used to determine the synchronization between branch power failure events.

[0054] In step S302, the system searches for tripping, opening, or undervoltage-related records of the upstream power supply circuit breaker or the main incoming circuit breaker within the same time period. After detecting near-synchronous power outages in multiple branches, the system continues to search for tripping, opening, or undervoltage-related records of the upstream power supply circuit breaker or the main incoming circuit breaker within the same time period. If such records exist, the event is marked as a common power supply disturbance event. Undervoltage-related records refer to status records reported by the device side, such as undervoltage tripping triggers, bus undervoltage alarms, or status records corresponding to upstream power supply drops. The common power supply disturbance event serves as a constraint condition for candidate branch correction, indicating that the current multi-branch power outage may initially be caused by disturbances in the common upstream power supply segment.

[0055] In step S303, the system identifies secondary tripping or independent abnormal recovery behavior after recovery. Based on this, the system further identifies secondary tripping or independent abnormal recovery behavior after recovery. Secondary tripping after recovery refers to a branch tripping or abnormally tripping again independently after reclosing, power supply restoration, or the bus returning to normal voltage, while other branches do not experience the same level of abnormality. Independent abnormal recovery behavior refers to a branch that, although not tripping again, exhibits recovery hysteresis, recovery jitter, abnormal local voltage drop, or abnormal power build-up that is clearly different from other branches in the power supply event window. Both indicate that the branch retains local abnormal characteristics after experiencing a common disturbance and should be retained as a key candidate branch.

[0056] In step S304, branches that satisfy the synthesis conditions are merged into the candidate branch set. Thus, the system forms a candidate branch set. The criteria for selecting candidate branches to enter the set are based on a combination of the following conditions: the existence of a secondary tripping after recovery, or the existence of independent abnormal recovery behavior, or although neither of the first two conditions is met, the branch is clearly distinguishable from common disturbances in the topology and has sufficient evidence of the action chain. This allows for the separation of general power outages caused by global common causes from persistent anomalies caused by local branch defects, thereby narrowing the scope of the fault search.

[0057] In some embodiments, to facilitate subsequent joint determination, the system may also generate an action chain confidence score for each candidate branch. Let the action chain confidence score of a candidate branch be denoted as . The value ranges from 0 to 1. This parameter represents the confidence weight derived from a comprehensive assessment of the independence of action sequences, secondary actions after recovery, whether the branch is removed from the common disturbance source, and the integrity of related actions. It can be determined using rule-based labeling. For example, if a branch independently detriggers again after recovery, it is assigned a higher confidence weight. If the power outage is only passive during a common disturbance without any subsequent independent anomalies, then a lower rating is assigned. This weight is only used as supplementary evidence in collaborative reasoning and does not replace the determination of abnormal deviation. If the system does not enable joint scoring, it is reserved. However, it does not participate in the final numerical calculation and does not affect the implementation of the main process.

[0058] In step S5, for each candidate branch, a power supply event window is constructed based on the time when the power supply is restored, and the branch current, bus voltage and branch power are extracted to generate an event feature vector.

[0059] Figure 4 A schematic diagram of the event feature vector generation process provided in an embodiment of this disclosure is shown. For example... Figure 4 As shown, in step S401, for each branch in the candidate branch set, the system constructs a power supply event window around its power restoration behavior. The power restoration time refers to the reclosing time of that branch, which is used preferentially; if there is no clear reclosing record for the event, the branch power restoration time is used; if the branch-side restoration time cannot be directly determined, the bus voltage restoration time is used as the corresponding alignment time. The priority among the three is fixed: first the reclosing time, then the branch power restoration time, and finally the bus voltage restoration time, to ensure that the source of the event center is consistent in different events.

[0060] Let the event-centric moment of a candidate branch be denoted as . The system constructs a power supply event window that includes observation periods before and after power-on. ; in, The duration of the initial observation. The duration of subsequent observation. Both are time quantities used to extract the baseline state before recovery and the transient-to-steady-state response after recovery within the same time frame. The value of should be sufficient to cover the stable reference segment before recovery. The value should be sufficient to cover the main process of transitioning from transient to stable operation after recovery. Both can be configured according to site load characteristics, but it is preferable to keep them consistent for similar events at the same site.

[0061] In step S402, the branch current, bus voltage, and branch power of candidate branches within the power supply event window are extracted to form an event data matrix. The system then extracts... The branch current, bus voltage, and branch power of the candidate branch are used to form an event data matrix. The rows of the matrix correspond to the sampling time, and the columns correspond to the three types of core signals.

[0062] In step S403, in generating the matrix Subsequently, the system performs short-window smoothing and local denoising on the time-series signals in the matrix. Short-window smoothing can employ moving average or median filtering to suppress sampling spikes and random jitter. Local denoising focuses on the minimum bus voltage search interval and the current recovery segment to prevent single-point noise from misleading feature calculations. The smoothing window should not be too long, otherwise it will mask the true transient changes. For event data matrices with sudden gaps or large missing segments, the data incompleteness marker is maintained, and the confidence limit is reflected in the subsequent output.

[0063] In step S404, an event feature vector containing current recovery slope, recovery settling time, short-time voltage drop characteristics, and power recovery deviation is extracted based on the processed matrix.

[0064] The dynamic response of candidate branches during power restoration reflects their local impedance state, contact state, and load rebuilding state. For issues such as high-resistance contacts, loose joints, and local insulation degradation, although they may not immediately manifest as continuous over-limit conditions in the steady-state phase, they often exhibit slow current recovery, jitter during the restoration process, increased bus voltage drop, or deviations in power recovery during the power restoration and load build-up phases. Based on this, the system revolves around... Extract the event feature vector, which includes the current recovery slope, recovery settling time, short-time voltage drop characteristics, and power recovery deviation.

[0065] Specifically, the current recovery slope is calculated as follows: let the branch current corresponding to the event center time be... The branch current corresponding to the steady-state moment is Stable moments are recorded as The current recovery slope It can be represented as: ; This characteristic reflects the speed at which the branch current returns to a stable level from the recovery point. For health events involving similar loads and power supply modes... They are usually concentrated in one area. If there is poor contact or abnormal impedance in the branches, the current build-up after power is supplied may be significantly slower, thus... reduce.

[0066] The extraction of the recovery settling time specifically includes, to determine the settling moment, the system first constructs a normalized rate of change of current based on the branch currents. Let the first... The current value at each sampling point is The current value corresponding to the previous sampling point is The rated current of the branch is denoted as Then the normalized rate of change of current Recorded as: ; in, This is an important parameter, representing the rated current value that the branch is designed to operate at for extended periods. It is derived from the circuit breaker's rated parameters, branch design records, or standard configuration tables.

[0067] The system searches within the power supply event window and finds that multiple consecutive sampling points meet the requirements. The earliest moment, of which To determine the stability threshold. Parameters This represents the proportion of adjacent sampled current changes relative to the upper limit of the allowable fluctuation of the rated current. It can be determined by pre-setting or by statistical analysis of historical health samples. If a statistical method is used, it is extracted from the steady-state phase of the health event. Distribution, and select the high quantile as In determining Then, the system finds the earliest moment that satisfies the continuous stability condition. Then, the recovery time can be calculated and expressed as: ; in, The longer the recovery time, the greater the time required for the branch to transition from a transient state to stable operation, which usually reflects additional obstruction or disturbance in the recovery process of that branch.

[0068] The extraction of short-time voltage drop characteristics specifically includes, assuming the reference bus voltage before closing is denoted as... The minimum bus voltage within the event window after closing is denoted as The rated voltage of the busbar is denoted as The short-term pressure drop Ratio of normalized pressure drop They are respectively ; in, For voltage, As a proportion. The average bus voltage of the reference interval before the event center is preferred to reduce the impact of single-point jitter. If A significant increase indicates that the branch will cause a stronger disturbance to the bus voltage when it is restored to the line, which is common in scenarios with local high resistance, impulsive load coupling, or abnormal contact.

[0069] Specifically, the power recovery deviation is defined as follows: let the mean steady-state power of the candidate branch over a period of time after recovery be denoted as... The corresponding healthy baseline reference power is denoted as Then the power recovery deviation Represented as: ; in, Based on the statistical results of historical health samples under the same working conditions from the subsequent health response baseline, the current event can be temporarily stored in the implementation order. The result will be obtained after the health baseline is generated. .

[0070] Based on the above characteristics, the event feature vector of the candidate branch formed by the system can be represented as: ; in, This is a column vector, which will be used for health baseline statistics and abnormal deviation calculation. Considering that the physical dimensions and numerical ranges of the current recovery slope, recovery settling time, short-time voltage drop characteristics, and power recovery deviation may differ, it is preferable to perform a dimension-by-dimensional standardization process on the event feature vector based on historical health sample statistical parameters, such as Z-score standardization, before constructing the health response baseline and calculating the abnormal deviation.

[0071] If a stable power reference is not readily available at a particular site, a vector can be temporarily constructed using only the first three dimensions of the features. However, once the feature dimensions used are determined, they should remain unchanged within the same model version. The design of this feature vector covers not only the recovery speed and stabilization process but also voltage drop and power deviation, providing a relatively complete reflection of the dynamic behavior during the power restoration phase. This compresses the anomaly representations, originally scattered throughout the post-action timeline, into a feature set with a clear structure, well-defined origin, and statistical comparability.

[0072] In step S6, a health response baseline is established based on historical health events, and the abnormal deviation of the current event feature vector of each candidate branch relative to the health response baseline is calculated. Collaborative reasoning is performed by combining the candidate branch set, the abnormal deviation, and the common topology constraints of the branch power supply path to output the fault location result.

[0073] To determine whether the current event feature vector is abnormal, a health response baseline corresponding to that branch or type of branch needs to be constructed. The health response baseline is a statistical representation of historical health events that meet specific screening criteria in the feature space, and its role is to provide a normal reference range under similar operating conditions for the current candidate branch.

[0074] The system first filters health events from the historical database. The filtering criteria include at least the following: First, no circuit breaker tripping or tripping occurred within the corresponding event window; second, the power supply mode is consistent; third, the load levels are similar; and fourth, no external maintenance switching or known manual switching occurred. Consistent power supply mode means that historical events and current events are powered by the same source and operating state, such as both after mains power restoration, both under normal direct mains power supply, or both after similar backup switching. Similar load levels can be filtered by the steady-state current range, steady-state power range, or load rate level after restoration. The absence of external maintenance switching is to prevent manual switching actions from distorting the dynamic behavior of healthy samples into abnormal patterns.

[0075] Let the set of feature vectors of the healthy samples obtained through screening be . Considering the potential differences in dimensions and numerical ranges among different features, the system preferentially performs dimension-wise standardization on the health sample feature vectors based on the mean and standard deviation of each feature in the health sample set, obtaining a standardized health sample feature vector set. The event feature vectors of the current candidate branch are then standardized using the same parameters to ensure consistency in the comparison space; a standardization method such as Z-score standardization can be used. Based on the standardized health sample feature vector set, the system calculates the health mean vector: ; And calculate the covariance matrix: ; in, Used to characterize the average level of various features in a healthy state. This is used to characterize the joint fluctuation relationships among various features under healthy conditions. Since there may be correlations among current recovery slope, recovery settling time, normalized voltage drop ratio, and power recovery deviation, relying solely on a single feature threshold can easily miss coupled anomalies. Using a covariance matrix allows for a comprehensive reflection of the health distribution across multiple dimensions.

[0076] Optionally, to improve the quality of the healthy baseline, the system further performs outlier removal. Specifically, it can first be based on the initial... and Calculate the deviation of each healthy sample from the current health center. If some samples are significantly far from the main population, it indicates that although these samples did not trigger protective actions, they may contain potential anomalies or sampling noise. These samples are then removed and the calculation is recalculated. and This can make the baseline of the health response more stable. Outlier removal only applies to the set of healthy samples, not to the current event to be judged.

[0077] When the number of healthy samples is small or the feature dimension is high, the covariance matrix may exhibit inversion instability. To address this, the system adds a diagonal stabilizing term to the covariance matrix when necessary, resulting in a regularized covariance matrix: ; in, It is the identity matrix. is the regularization coefficient. The non-negative parameter representing the enhancement of the numerical stability of the covariance matrix can be determined by pre-setting or by statistically obtaining it based on the variance level of the healthy sample. If a statistical method is used, it can be set as follows: It is proportional to the average variance of each dimension of the feature.

[0078] Meanwhile, the set of samples from the aforementioned healthy samples that best matches the current candidate branch operating condition can also be used to determine the reference power. Preferably, Take the average of the power mean of the corresponding healthy samples within the stable period, or take the median, to enhance the ability to resist outliers.

[0079] After obtaining the standardized event feature vector of the current candidate branch Health mean vector and covariance matrix Or after regularization Then, the system calculates the anomaly deviation. Preferably, Mahalanobis distance is used as the anomaly deviation. If a regularized covariance matrix is ​​used, the anomaly deviation... The calculation is as follows: ; If no regularization is needed, then use Alternative Perform the same calculation. It can simultaneously consider differences in feature scales and correlations. Compared to simple Euclidean distance, It can better reflect the degree of abnormality of the current event in the distribution of health characteristics.

[0080] To form an anomaly determination, an anomaly determination threshold needs to be established. The system determines the threshold based on the distance distribution of each healthy sample relative to the healthy mean vector. Let the... The distance corresponding to each healthy sample is Then the distance set can be obtained. The system selects the high quantile value of this set as the anomaly detection threshold. . This represents the upper limit of acceptable deviations under healthy conditions, derived from historical healthy sample statistics, rather than being set based on temporary experience. Therefore, the threshold can adaptively change according to the actual operational characteristics of different branch types and different sites.

[0081] When the current candidate branch satisfies If so, it is determined that the candidate branch has an abnormal response in this power restoration event. Conversely, if This indicates that the recovery process remains within the acceptable range of the healthy response baseline. Therefore, evaluating branch recovery behavior using a multi-dimensional feature-based approach is more suitable for identifying hidden faults such as high resistance, poor contact, and intermittent abnormalities.

[0082] Optionally, when joint scoring is required, the system can also map abnormal deviations to abnormal scores. Let the abnormal score be denoted as... Then it can be defined as: ; in, This indicates the degree to which the current deviation exceeds the abnormal threshold. If If the threshold is not exceeded, then If it is zero; if it exceeds the threshold, then The value increases with the degree of exceeding the limit. This preserves the significance of the threshold boundary while providing a continuous metric for subsequent multi-evidence fusion. If joint scoring is not enabled, only one output is provided. The result will be the corresponding anomaly assessment.

[0083] After obtaining the candidate branch set, the abnormal deviation of each candidate branch, and the power supply path of the branches, the system performs collaborative reasoning. This step, based on action chain evidence, temporal anomaly evidence, and topology constraints, ultimately outputs the faulty segment rather than just a list of abnormal branches.

[0084] First, the single-branch candidate case. If the candidate branch set contains only one branch. And the abnormal deviation corresponding to this branch Exceeding the anomaly detection threshold If the fault is located at the end of the power supply path corresponding to the branch, the system will identify the faulty section as the terminal segment. The terminal segment refers to the actual power supply section from downstream of the branch circuit breaker to upstream of the corresponding load, which typically includes the branch connection terminal, branch cable joint, the near-end segment of the branch cable body, or a combination thereof. Locating the fault to the terminal segment of the path does not only lock the branch circuit breaker body, because it is based on the dynamic response of power restoration and topology coverage, and can reliably identify the range of the faulty section, rather than the failure point inside a single device.

[0085] Secondly, the multi-branch anomaly. If multiple candidate branches all satisfy... The system then obtains the power supply path for each abnormal branch and finds the common intersection of these paths. Let the paths corresponding to the abnormal branch set be as follows: The intersection of common paths is denoted as: ; like Not empty, and If the faulty branches include a common upstream segment containing the upstream power supply circuit breaker or the main incoming circuit breaker, it indicates that these faulty branches share the same upstream power supply segment, and the fault is more likely located in the common power supply path than multiple branches failing simultaneously and independently. In this case, the system will identify the corresponding common power supply path as the faulty segment. Conversely, if... If the node is empty, or remains at the end node without any actual common cause, then each branch is considered an independent candidate and is not directly merged into a common power supply segment fault.

[0086] In some embodiments, the system may further introduce joint scoring to refine the ranking. Let the action chain confidence of the candidate branch be... The abnormal score is The fusion weight coefficient is denoted as Then the overall score It can be defined as: ; in, The value ranges from 0 to 1, representing the relative weight between action chain evidence and temporal anomaly evidence. It can be determined by pre-setting or by calibration using historical known location result samples. If the on-site protection action information is reliable, it can improve... If the on-site action recording has a delay but the timing characteristics are more stable, the time limit can be appropriately reduced. This parameter is typically set during the model deployment phase and should not be changed arbitrarily in a single event. If joint scoring is used, the candidate branches are ranked according to... Sorting is used to help distinguish between the main fault branch and the accompanying disturbed branch.

[0087] For joint scoring of common power supply segments, scores of outlier branches falling into the same common path intersection can be aggregated. Aggregation can be done using either the average value or a weighted average. If the average value is used, the common segment score is recorded as: ; in, For the first The overall score of each abnormal branch. If... Satisfying common topology constraints and If the set output conditions are met, output a common power supply section fault; if a single branch... If the score is significantly higher than the common segment score, and there is sufficient evidence of a secondary trip after recovery, then the fault at the end of that branch should be output first. This collaborative reasoning approach effectively distinguishes between two scenarios: one where an anomaly in the upstream common power supply path causes multiple branches to be simultaneously disturbed; and another where only a few branches expose local defects again after a common disturbance. The former should output the common power supply segment, and the latter should output the corresponding branch end segment. This logic avoids misjudging a common anomaly as a single-branch fault by only considering the sequence of actions, and also avoids ignoring the true local fault object due to multiple branches losing power simultaneously.

[0088] After collaborative reasoning is completed, the system generates fault location results. The fault location results include at least the fault object identifier, fault segment type, anomaly deviation degree, and event time. The fault object identifier is the branch number in a single-branch scenario, and the common path identifier or common affected branch set identifier in a common power supply segment scenario. The fault segment type is at least distinguished between branch end segment faults and common power supply path faults. The anomaly deviation degree is as described above. The event time is preferably recorded as the trigger moment. or the moment at the center of the event The specific time field used is fixed in the system version to avoid semantic inconsistencies in the result fields of different events.

[0089] In some embodiments, the result record may further include auxiliary information such as action chain summary, candidate branch set, recovery stabilization time, normalized voltage drop ratio, and comprehensive score, to facilitate traceability and location by operations and maintenance personnel. These auxiliary fields do not change the core output logic, but only enhance the interpretability of the results. After output, the results are written to the operations and maintenance database, event table, or alarm center, and simultaneously trigger visualization or maintenance work order recommendations.

[0090] Optionally, the system can link and archive the current location result with subsequent manual maintenance results. If the maintenance result confirms that the fault is located at the branch terminal, connector, or common power supply path, the event can be included in the subsequent threshold verification and parameter calibration samples. If there is a deviation between the maintenance result and the location, the event is transferred to the review library and is not directly merged into the healthy samples, nor is it immediately used for parameter recalibration. This setting can prevent mislabeled samples from contaminating the health response baseline and weight parameters.

[0091] In an exemplary scenario, the system first detects an unplanned tripping of the main incoming circuit breaker. Based on the trigger time, the system constructs an action chain analysis window and extracts multiple branches that simultaneously lost power within a short period. Further investigation reveals an undervoltage correlation record upstream, thus the event is marked as a public power supply disturbance event. If, after power is restored, one of the branches trips again independently, and this branch exhibits a decreased current recovery slope, a prolonged recovery stabilization time, and an increased normalized voltage drop ratio within the power supply event window, then this branch is retained in the candidate branch set.

[0092] Subsequently, the system compares the event feature vector of this branch with the health response baseline established from historical health events under the same operating conditions, and finds that the abnormal deviation exceeds the threshold. Since the evidence from a single branch is significantly stronger than the evidence from a common disturbance at this point, the system ultimately outputs a fault in the terminal section of that branch. Thus, even if the initial action occurs at the main incoming line level, the system can still trace back to the actual local fault section through protection coordination distortion correction and recovery stage feature analysis.

[0093] In another example, multiple branches experienced near-synchronous power outages. After power was restored, all branches exhibited abnormal voltage drops and recovery lags. Furthermore, the common intersection of the power supply paths of these branches included a common upstream segment between the upstream power supply circuit breaker and the main incoming circuit breaker. Based on this, the system outputs the segment corresponding to the common path intersection as the fault segment, rather than treating each branch as an independent fault point. This avoids incorrectly splitting a common power supply path problem into multiple branch problems, facilitating on-site maintenance to prioritize addressing the true common-cause segment.

[0094] In summary, the embodiments of this disclosure first establish a power supply topology model, then perform unified preprocessing on the action records and runtime sequence data, construct an action chain analysis window and form an action chain after detecting a trip or abnormal trip, and then analyze near-synchronous power outages, upstream related actions, and secondary trips after recovery to obtain a set of candidate branches corrected for protection coordination distortion; subsequently, a power supply event window is constructed around the power supply recovery time of the candidate branches, and the current recovery slope, recovery stabilization time, short-time voltage drop characteristics, and power recovery deviation are extracted to establish an event feature vector; A health response baseline is established based on historical health events, and anomaly deviation is calculated. Finally, collaborative reasoning is performed by combining the candidate branch set, anomaly deviation, and common topology constraints of the branch power supply paths to output fault location results. This can improve the accuracy of identifying branch connection faults, common power supply section faults, and hidden anomalies during the recovery phase without increasing the number of field sensors, reduce the cost of manual segment-by-segment troubleshooting, and enhance the consistency between fault location results and the actual power supply topology.

[0095] Please see Figure 5 , Figure 5This is a schematic diagram of a cable branch box fault location system based on circuit breaker and timing data, provided in an embodiment of this application. As shown in the figure, the system includes: The topology modeling module 501 is used to read the branch box configuration data, establish a power supply topology model that includes the upstream power supply circuit breaker, the main incoming circuit breaker, the branch circuit breaker, and the correspondence between the branch and the load, and generate the downstream coverage set and branch power supply path for each circuit breaker. The data acquisition and preprocessing module 502 is used to continuously acquire the action records of each circuit breaker and the runtime sequence data corresponding to each branch, and to perform preprocessing on the action records and runtime sequence data. The action chain construction module 503 is used to construct an action chain analysis window based on pre-processed action records and runtime sequence data when a circuit breaker trips or abnormal trips are detected, extract all action records in the window and form a circuit breaker action chain in chronological order. The candidate branch analysis module 504 is used to analyze the near-synchronous power outage, upstream associated action and secondary tripping after recovery in the circuit breaker action chain based on the power supply topology model, and obtain a set of candidate branches after protection coordination distortion correction. The event feature generation module 505 is used to construct a power supply event window for each candidate branch based on the power supply recovery time, extract branch current, bus voltage and branch power, and generate event feature vectors. The anomaly assessment and collaborative reasoning module 506 is used to establish a health response baseline based on historical health events, calculate the anomaly deviation of the current event feature vector of each candidate branch relative to the health response baseline, and perform collaborative reasoning by combining the candidate branch set, the anomaly deviation, and the common topology constraints of the branch power supply path to output the fault location result.

[0096] Each processing unit and / or module in the embodiments of this application can be implemented by an analog circuit that implements the functions described in the embodiments of this application, or by software that executes the functions described in the embodiments of this application.

[0097] In the above embodiments, the descriptions of each embodiment have different focuses. Parts not described in detail in a certain embodiment can be referred to in the relevant descriptions of other embodiments. The above descriptions are merely preferred embodiments of this application and explanations of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above technical features, but should also cover other technical solutions formed by arbitrary combinations of the above technical features or their equivalent features without departing from the inventive concept.

Claims

1. A method for fault location of cable branch boxes based on circuit breaker and timing data, characterized in that, include: Read the branch box configuration data, establish a power supply topology model that includes the upstream power supply circuit breaker, the main incoming circuit breaker, the branch circuit breaker, and the correspondence between branches and loads, and generate downstream coverage sets and branch power supply paths for each circuit breaker; Continuously collect the action records of each circuit breaker and the runtime sequence data corresponding to each branch, and perform preprocessing on the action records and runtime sequence data; Based on preprocessed action records and runtime sequence data, when a circuit breaker trips or abnormal trips are detected, an action chain analysis window is constructed at the trigger time, all action records in the window are extracted, and a circuit breaker action chain is formed in chronological order. Based on the power supply topology model, the near-synchronous power outage, upstream associated actions, and secondary tripping after recovery in the circuit breaker action chain are analyzed to obtain a candidate branch set corrected for protection coordination distortion. Specifically, the time difference between the power outage actions of each branch circuit breaker is compared, and power outages with a time difference less than the synchronization threshold are classified as near-synchronous power outages. When multiple branches experience near-synchronous power outages and there are tripping, opening, or undervoltage associated records of upstream power supply circuit breakers or main incoming line circuit breakers during the corresponding time period, the event is marked as a common power supply disturbance event. Branches that independently trip again after power restoration or exhibit recovery hysteresis, recovery jitter, abnormal local voltage drop, or abnormal power establishment that are significantly different from other branches in the power supply event window are incorporated into the candidate branch set. For each candidate branch, a power supply event window is constructed based on the power supply recovery time. Branch current, bus voltage, and branch power are extracted to generate an event feature vector. The event feature vector includes current recovery slope, recovery stabilization time, short-time voltage drop characteristics, and power recovery deviation. The recovery stabilization time is determined based on the moment when the normalized current change rate first continuously meets the stability judgment threshold within the power supply event window. A health response baseline is established based on historical health events. The abnormal deviation of the current event feature vector of each candidate branch relative to the health response baseline is calculated. Collaborative reasoning is performed by combining the candidate branch set, the abnormal deviation, and the common topological constraints of the branch power supply path to output the fault location result.

2. The method according to claim 1, characterized in that, Establishing the power supply topology model includes: Read the upstream power supply circuit breaker number, main incoming circuit breaker number, branch circuit breaker number, branch number, load type, and connection relationship from the branch box configuration record; construct a directed hierarchical structure according to the power supply direction; determine the set of direct downstream nodes for each circuit breaker based on the directed hierarchical structure, and generate a downstream coverage set according to the path relationship between the circuit breaker and the load; for each branch record, the branch power supply path is composed of the upstream power supply circuit breaker, the main incoming circuit breaker, and the corresponding branch circuit breaker in sequence; when the configuration data changes, reload the configuration data and rebuild the corresponding power supply topology model.

3. The method according to claim 1, characterized in that, The formation of the circuit breaker action chain includes: An action chain analysis window is established around the trigger time, allowing for forward backtracking and backward tracking. All circuit breaker action records within the action chain analysis window are extracted and arranged in ascending order by action timestamp to form an action sequence. Each action record is appended with its topology level identifier and downstream coverage set in the power supply topology model.

4. The method according to claim 3, characterized in that, For each candidate branch, a power supply event window is constructed based on the time of power restoration. Branch current, bus voltage, and branch power are extracted to generate an event feature vector, including: Using the corresponding alignment time among the reclosing time, power restoration time, or bus voltage restoration time of the candidate branch as the event center, a power supply event window containing the observation period before closing and the observation period after closing is constructed; the branch current, bus voltage, and branch power of the corresponding candidate branch within the power supply event window are extracted and an event data matrix is ​​formed; short-window smoothing and local denoising processing are performed on the timing signals in the event data matrix.

5. The method according to claim 4, characterized in that, Calculating the recovery stabilization time includes: The normalized current change rate is constructed based on the branch current, and the normalized current change rate is determined by the ratio of the current change at adjacent sampling times to the rated current of the branch. Within the power supply event window, the earliest time when multiple consecutive sampling points satisfy the normalized current change rate not greater than the stability judgment threshold is searched, and this earliest time is determined as the stable time. The time difference between the stable time and the event center is determined as the recovery stabilization time. The short-time voltage drop characteristic is determined by the difference between the bus reference voltage before closing and the minimum bus voltage within the event window after closing, or their normalized ratio.

6. The method according to claim 5, characterized in that, The establishment of a health response baseline based on historical health events includes: Historical health events are selected from the historical database that have not experienced circuit breaker tripping or tripping, have consistent power supply modes, similar load levels, and have not undergone external maintenance switching. The event feature vectors corresponding to each historical health event are extracted to form a health sample set. The health mean vector and covariance matrix are calculated based on the health sample set. Outliers that deviate from the health sample group are removed, and the health mean vector and covariance matrix are recalculated. When the inversion of the covariance matrix is ​​unstable, a diagonal stabilizing term is added to the covariance matrix to obtain a regularized covariance matrix.

7. The method according to claim 6, characterized in that, Calculating the abnormal deviation includes: Using the current event feature vector of the candidate branch, the health mean vector, and the covariance matrix or the regularized covariance matrix, the Mahalanobis distance of the candidate branch is calculated as the anomaly deviation degree; based on the distance distribution of each historical health event relative to the health mean vector, an anomaly judgment threshold is determined; the anomaly deviation degree of the current candidate branch is compared with the anomaly judgment threshold to obtain the anomaly judgment result of the candidate branch.

8. The method according to claim 7, characterized in that, The output fault location results include: When the candidate branch set contains only one branch and the abnormal deviation of that branch exceeds the abnormal judgment threshold, the end section of the power supply path corresponding to that branch is determined as a fault section; when the abnormal deviation of multiple candidate branches exceeds the abnormal judgment threshold, the branch power supply path of each abnormal branch is obtained and the common intersection of the paths is calculated. When the common intersection of the paths is not empty and contains the upstream power supply circuit breaker or the main incoming line circuit breaker, the corresponding common power supply path is determined as a fault section.

9. A cable branch box fault location system based on circuit breaker and timing data, characterized in that, For implementing the method as described in any one of claims 1-8, comprising: The topology modeling module is used to read the branch box configuration data, establish a power supply topology model that includes the upstream power supply circuit breaker, the main incoming circuit breaker, the branch circuit breaker, and the correspondence between the branches and the load, and generate the downstream coverage set and branch power supply path for each circuit breaker. The data acquisition and preprocessing module is used to continuously acquire the action records of each circuit breaker and the runtime sequence data corresponding to each branch, and to perform preprocessing on the action records and runtime sequence data. The action chain construction module is used to construct an action chain analysis window based on pre-processed action records and runtime sequence data when a circuit breaker trips or abnormal trips are detected. It extracts all action records within the window and forms a circuit breaker action chain in chronological order. The candidate branch analysis module is used to analyze near-synchronous power outages, upstream associated actions, and secondary tripping after recovery in the circuit breaker action chain based on the power supply topology model, and to obtain a set of candidate branches after protection coordination distortion correction. Specifically, the time difference between the power outage actions of each branch circuit breaker is compared, and power outages with a time difference less than the synchronization judgment threshold are classified as near-synchronous power outages. When multiple branches experience near-synchronous power outages and there are tripping, opening, or undervoltage associated records of the upstream power supply circuit breaker or the main incoming line circuit breaker during the corresponding time period, the event is marked as a common power supply disturbance event. Branches that independently trip again after power restoration or exhibit recovery hysteresis, recovery jitter, abnormal local voltage drop, or abnormal power establishment that are significantly different from other branches in the power supply event window are added to the candidate branch set. The event feature generation module is used to construct a power supply event window for each candidate branch based on its power supply recovery time, extract branch current, bus voltage and branch power, and generate an event feature vector. The event feature vector includes current recovery slope, recovery stabilization time, short-time voltage drop characteristics and power recovery deviation, wherein the recovery stabilization time is determined based on the moment when the normalized current change rate first continuously meets the stability judgment threshold within the power supply event window. The anomaly assessment and collaborative reasoning module is used to establish a health response baseline based on historical health events, calculate the anomaly deviation of the current event feature vector of each candidate branch relative to the health response baseline, and perform collaborative reasoning by combining the candidate branch set, the anomaly deviation, and the common topology constraints of the branch power supply path to output the fault location result.

Images