A distribution network self-healing recovery system based on outdoor primary and secondary fusion circuit breaker

Abstract

The application belongs to the technical field of power distribution network power supply recovery and self-healing control, and particularly relates to a distribution network self-healing recovery system based on outdoor primary-secondary fusion circuit breakers. The system establishes a directed feeder topology according to the electrical connection relationship of each primary-secondary fusion circuit breaker, obtains voltage, current, zero sequence quantity, protection start, tripping and switch position time sequence data before and after the fault, forms a fault event sequence through cross-node time correction, and constructs a fault propagation fingerprint in combination with upstream and downstream adjacency relationships; the application generates an anti-factual fault propagation state for a candidate fault section, matches the anti-factual fault propagation state with an actual propagation fingerprint for consistency, jointly determines a fault boundary and a suspected distorted switch topology state, and forms self-healing recovery topology constraint data. The application can verify the fault propagation relationship and the topology state under the conditions of event disordering, local observation missing and switch remote signaling distortion, and reduce the situations of fault section deviation, healthy section expansion isolation and selection of an untrustworthy recovery path.

Description

Technical Field

[0001] This invention belongs to the field of power distribution network restoration and self-healing control technology, specifically a power distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker. Background Technology

[0002] In existing distribution networks, feeder self-healing recovery is typically based on the feeder topology maintained by the distribution master station, combined with voltage, current, zero-sequence quantities, protection start-up, trip signals, and switch position remote signaling transmitted from outdoor primary and secondary integrated circuit breakers for fault location. After a fault occurs, the master station determines the range of circuit breakers through which the fault current passes according to the preset upstream and downstream connection relationships, and then determines the suspected fault section based on the protection action sequence, undervoltage section, and switch opening and closing status, and issues isolation and recovery control commands to adjacent section circuit breakers and tie circuit breakers. To handle communication delays and data gaps, existing solutions typically set over-limit criteria for telemetry measurements, set allowable time windows for remote signaling events, and filter abnormal records through time sorting or simple majority consensus rules. When a circuit breaker has not uploaded a complete event, the system often supplements the judgment based on the status of adjacent nodes; when the switch remote signaling is consistent with the master station topology model, the system directly continues the fault section search and recovery path calculation along the established topology. Therefore, the existing positioning process essentially relies on the premise that the topology is correct and uses the fault information collected by each circuit breaker as a one-way verification basis for the given topology.

[0003] Outdoor integrated primary and secondary circuit breakers operate in a decentralized manner after distribution feeders for extended periods. The timing of protection actions, telemetry sampling, and communication arrival are not always consistent. Furthermore, individual switch positions may exhibit temporary deviations from the actual on-site continuity due to telemetry delays, unupdated status, or out-of-order communication messages. In such cases, if the current topology of the master station is still used as the sole constraint, the true propagation path of the fault current will be mapped to incorrect upstream and downstream relationships; the absence of abnormal electrical quantities at a certain node may also be interpreted as the termination of fault propagation at that point. Because existing location methods lack a mechanism to reverse-engineer the switch topology status using the entire fault event propagation relationship, once erroneous telemetry enters the topology model, the subsequent protection action sequence, abnormal coverage area, and fault isolation boundary may continuously shift along the erroneous connection relationship, causing inconsistencies between the actual fault section and the calculated section. This further leads to the expansion of healthy sections outside the fault boundary into isolation, or the misinformation of sections with fault risk into the recovery path. The challenge lies in how to simultaneously determine the true fault boundary and the reliable topology status under conditions of event timing deviations, missing local observations, and potentially distorted switch topology status. Summary of the Invention

[0004] The purpose of this invention is to provide a distribution network self-healing and recovery system based on an outdoor primary and secondary integrated circuit breaker, which can solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A distribution network self-healing and recovery system based on outdoor primary and secondary integrated circuit breakers includes: establishing a directed feeder topology based on the electrical connection relationship of each primary and secondary integrated circuit breaker in the distribution feeder;

[0007] Acquire the timing data of voltage, current, zero-sequence quantity, protection start, trip and switch position of the primary and secondary fusion circuit breakers before and after the fault, and form a fault event sequence according to a unified event time scale;

[0008] Based on the directed feeder topology, the propagation correlation of the fault event sequence between adjacent circuit breakers is determined, forming a fault propagation fingerprint;

[0009] For each candidate fault segment, a corresponding counterfactual fault propagation state is constructed. The counterfactual fault propagation state is matched with the fault propagation fingerprint for consistency, and the fault boundary and the switch topology state that is inconsistent with the actual propagation relationship are jointly determined.

[0010] Preferably, when performing event-based processing on the timing data of voltage, current, zero-sequence quantity, protection start, trip, and switch position, the protection start event is used as a local time reference point, and the electrical quantity change segment and switch state transition segment before and after the reference point are extracted respectively.

[0011] For data from different primary and secondary integrated circuit breakers with communication arrival time differences, a cross-node time correspondence is established based on the common electrical disturbances detected by adjacent circuit breakers, and the logical order of each event is corrected based on the cross-node time correspondence.

[0012] The corrected events are encoded according to the circuit breaker identifier, event category, occurrence order, and preceding and following states to form a unified event sequence used for constructing the fault propagation fingerprint.

[0013] Preferably, the fault propagation fingerprint is constructed according to the upstream and downstream adjacency relationship in the directed feeder topology, including the initial set of circuit breakers that determine the abnormal electrical quantity, the set of circuit breakers that continuously exhibit abnormalities along the feeder direction, the relative order of protection activation and tripping, and the set of sections where the abnormal electrical quantity disappears after tripping.

[0014] For a propagation chain with missing events between adjacent circuit breakers, the propagation chain is not directly cut off. Instead, a propagation connection to be verified is generated based on the continuity of events on both sides of the missing node in terms of time sequence, abnormal polarity, and topological direction. The determined propagation connection and the propagation connection to be verified are written together into the fault propagation fingerprint.

[0015] Preferably, for each candidate fault segment in the directed feeder topology, the candidate fault segment is set as a fault source, and the set of circuit breakers theoretically affected by the fault, the theoretical protection action sequence, and the abnormal disappearance range after the corresponding isolation action are derived along the current switch conduction relationship to form a counterfactual fault propagation state.

[0016] The counterfactual fault propagation state is associated with the actual anomaly coverage, actual action sequence, and actual anomaly disappearance range in the fault propagation fingerprint. Candidate results that can only satisfy propagation consistency after changing at least one switch logic state are given a topology state correction flag for joint determination of fault segment and switch topology state.

[0017] Preferably, the cross-node time correspondence uses the current sudden change edge, zero-sequence quantity change segment, or protection start-up associated segment that occur together in adjacent circuit breakers as the anchor point of the same source event;

[0018] For the communication arrival deviation of the same circuit breaker at multiple consecutive anchor points of the same source event, a deviation sequence is established, and the part that can be explained by the unified offset relationship is used as the node time offset, while the discrete deviation that cannot be explained by the unified offset relationship is retained as the event timing uncertainty.

[0019] When forming a unified event sequence, the event sorting is corrected according to the node time offset, and the event timing uncertainty is added to the corresponding event so that subsequent propagation associations retain multiple candidate orders when there is overlap in time order.

[0020] Preferably, for the propagation connection to be verified, two types of candidate connections are established: one propagating from the upstream node to the downstream node and the other reversing from the downstream node to the upstream node.

[0021] The abnormal electrical quantity change direction, protection status change, and intermediate switch conduction status at both ends of the candidate connection are combined with the directed feeder topology for verification. When the candidate connection needs to cross a switch that is in the open state, the corresponding connection is converted into a topology suspected distortion connection.

[0022] When the candidate connection has continuous electrical quantity changes but discontinuous protection events, the missing protection events are marked as observation gaps, and the topological continuity of the propagation chain is maintained, so that topological distortion and event missing are entered into subsequent counterfactual matching using different propagation labels.

[0023] Preferably, when constructing the counterfactual fault propagation state, each candidate fault segment is taken as the starting segment, and the theoretically affected nodes are expanded step by step according to the switch conduction relationship. The propagation in the corresponding direction is terminated when the first boundary switch, theoretical isolation switch or feeder terminal in the open state is encountered.

[0024] For feeders with branch structures, the theoretical anomaly coverage range on each branch is recorded separately, and the protection events at the common upstream node are used as branch propagation merging constraints.

[0025] Simultaneously, a theoretical action partial order relationship is generated based on the electrical upstream and downstream sequence between the candidate fault section and each protection node, without pre-limiting a fixed action duration, so that the differences in protection responses of different circuit breakers can be compared in the same counterfactual propagation state.

[0026] Preferably, the topology state correction flag is represented as a candidate logic state variable corresponding to a specific switch, and candidate topology combinations are generated for the same candidate fault segment, including those that do not modify the topology, modify a single switch state, and modify multiple switch states with propagation correlation.

[0027] For each candidate topology combination, the corresponding counterfactual fault propagation state is regenerated and jointly matched with the fault propagation fingerprint containing suspected topological distorted connections, observation gaps and event timing uncertainties.

[0028] A candidate explanation set is constructed based on inconsistencies in fault propagation, topology state modifications, and unexplainable events. From the candidate explanation set, fault segment-topology state combinations that do not have contradictory propagation relationships are retained.

[0029] Preferably, for the multiple candidate event sequences formed by the event timing uncertainty, they are respectively combined with the candidate topology to form propagation interpretation branches, and in each propagation interpretation branch, the closure relationship between the initial anomaly location, adjacent propagation continuity, protection action partial order, and the anomaly disappearance boundary after isolation is checked along the directed feeder topology.

[0030] If the same switch state modification can simultaneously eliminate contradictions in multiple independent propagation chains, then the switch state modification is retained as a shared topology correction term.

[0031] If a state modification can only explain a single isolated event and causes new contradictions in other propagation chains, then the state modification is removed from the corresponding propagation explanation branch to form a joint explanation result with the minimum set of topological corrections.

[0032] Preferably, based on the joint interpretation results, the upstream fault boundary and downstream fault boundary corresponding to the actual propagation termination position on both sides of the candidate segment where the fault source is located are extracted, and the segment located outside the fault boundary and whose power supply path does not contain the shared topology correction term is marked as the topology determination area;

[0033] For power supply paths passing through sections where multiple candidate interpretations still exist in the topology state, these sections are retained as topology unconfirmed areas.

[0034] The fault boundary, the topology determination area, the topology unconfirmed area, and the candidate logic states of the corresponding switches are combined to form self-healing recovery topology constraint data, which is used by the subsequent distribution network self-healing recovery process to select isolation objects and recoverable paths for non-fault power loss sections.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] 1. By uniformly encoding the voltage, current, zero-sequence quantity, protection initiation, tripping, and switch position changes of each primary and secondary integrated circuit breaker into an event sequence, and using common disturbances occurring at adjacent nodes to correct cross-node time offsets, the order of communication arrival and the order of on-site events can be separated, reducing interference from out-of-order messages on the judgment of protection action sequence. Based on this, the initial location of the anomaly, the continuous range of the anomaly along the feeder, the protection action offset, and the range of anomaly disappearance after tripping are organized into a fault propagation fingerprint. Then, a corresponding counterfactual propagation state is generated for each candidate segment, allowing the entire propagation chain to simultaneously examine the candidate fault location and switch conduction relationship. When a candidate segment can only explain multiple interrelated events after modifying a specific switch logic state, the switch is included in the topology correction candidate, instead of continuing to expand the fault range along the erroneous topology; when a local event is missing but its propagation relationship on both sides remains continuous, the observation gap will not be directly determined as the propagation termination location. This forms a joint constraint result of fault location and topology state, ensuring that the upstream and downstream boundaries of the fault are consistent with the actual electrical propagation relationship, and reducing fault segment offsets caused by erroneous remote signaling.

[0037] 2. During the joint judgment process, multiple candidate sequences are retained for events with uncertain time order, and these sequences are combined with different candidate topologies to form propagation interpretation branches. This avoids the single time correction result directly determining the fault conclusion. Abnormal continuity relationships spanning disconnect switches are marked as suspected topology distortion connections, and situations where electrical quantities are continuous but protection events are missing are marked as observation gaps, allowing different constraint methods to be used for the two types of anomalies in subsequent matching. Furthermore, the minimum topology correction set is used to filter out state modifications that can only explain isolated events but introduce contradictory information in other propagation chains, which can limit the scope of topology correction and maintain the original connection relationships of unaffected feeder sections. After the fault boundary is determined, power supply paths that do not go through shared topology correction terms are included in the topology determination area, and paths that still have multiple candidate interpretations are included in the topology confirmation area. These two types of areas, together with candidate logic states, form recovery topology constraint data, which allows subsequent isolation objects and recovery paths for non-faulty power-loss sections to be selected based on verified propagation relationships, reducing the probability of excessive isolation of healthy sections and direct inclusion of unconfirmed topology sections in the recovery sequence. Attached Figure Description

[0038] Figure 1Flowchart for fault event timing correction, propagation fingerprint construction, and counterfactual fault segment determination;

[0039] Figure 2 Plotting node time offsets and remaining errors;

[0040] Figure 3 Cost map of structural differences in candidate fault sections;

[0041] Figure 4 Generate flowcharts for candidate topology correction, propagation interpretation branch screening, and self-healing recovery of topology constraints;

[0042] Figure 5 The cost graph for propagating the interpretation of branch constraints. Detailed Implementation

[0043] In one embodiment, a distribution network self-healing system based on outdoor primary and secondary integrated circuit breakers includes: establishing a directed feeder topology based on the electrical connection relationship of each primary and secondary integrated circuit breaker in the distribution feeder; acquiring the timing data of voltage, current, zero-sequence quantity, protection start, trip, and switch position corresponding to each primary and secondary integrated circuit breaker before and after the fault, and forming a fault event sequence according to a unified event time scale; determining the propagation correlation of the fault event sequence between adjacent circuit breakers based on the directed feeder topology, forming a fault propagation fingerprint; constructing a corresponding counterfactual fault propagation state for each candidate fault section, performing consistency matching between the counterfactual fault propagation state and the fault propagation fingerprint, and jointly determining the fault boundary and the switch topology state that is inconsistent with the actual propagation relationship.

[0044] In this embodiment, the directed feeder topology uses the feeder section where the primary and secondary fusion circuit breakers are located as the electrical connection boundary, the connection relationship between adjacent circuit breakers that can form a continuous power supply path as the topology edge, and the reference direction of the topology edge is determined by the normal power supply direction. The circuit breaker's closed, open, and pending status states are mapped to conducting, blocked, and pending verification topology connection states, respectively. This makes the feeder structure maintained by the master station no longer just a node relationship for static querying, but a set of directed relationships that can be directly calculated for subsequent fault propagation reasoning. Specifically, for feeders with trunk and branch structures... Multiple downstream sections connected to the same upstream circuit breaker are recorded as independent branches. The branches converge only through the common upstream node. For connections that are disconnected when a fault occurs, only their network position is retained and they are not included in the current fault propagation path. For circuit breakers whose switch position data changes within the fault time window, the connection state before and after the action is retained respectively, and the time of the event is used as the state switching boundary. This allows the pre-fault propagation, protection action process, and the range of anomaly disappearance after fault isolation to be interpreted in the same topology in chronological order.

[0045] Specifically, the time-series data before and after the fault continuously enters the fault event buffer according to the circuit breaker identification and acquisition time. Voltage, current, and zero-sequence quantities do not directly participate in all combination calculations as continuous sampling sequences. Instead, based on the direction of change, duration of change, and time correlation with the protection start event between adjacent samples, continuous quantities are converted into event expressions such as start-up, duration, decline, and recovery to stability. Protection start-up, tripping, and switch position changes are directly written as discrete state events into the same event structure. Each event is associated with at least the circuit breaker from which the event originated, the event category, the original occurrence time, the corrected logic time, the state before the event, the state after the event, and the feeder to which it belongs. Location and time reliability attributes are used to distinguish between events that have been corrected by multiple disturbances from the same source and events that still have uncertain order. The fault event buffer is sorted according to logical time to form a unified event sequence. Similar events that occur repeatedly in a very short time for the same circuit breaker without forming a state reversal are merged into a continuous event to avoid repeated communication transmissions being interpreted as multiple fault propagations. If protection has been activated but the corresponding electrical quantity does not exist due to missing local records, the protection event itself is retained and the missing electrical quantity is not fabricated, so that subsequent propagation analysis can distinguish data gaps from the actual termination of electrical propagation.

[0046] The fault propagation fingerprint is composed of spatial propagation relationships and temporal action relationships. The spatial propagation relationship is based on the actual set of nodes and continuous segments in the directed feeder topology where abnormal electrical quantities actually occur. For each primary and secondary fusion circuit breaker that experiences a sudden current change, zero-sequence quantity abnormality, or voltage abnormal change, its upstream and downstream adjacent nodes are determined. Based on the logical time of the event, it is determined whether the abnormality occurs continuously along the reference power supply direction, converges from the branch line to the common upstream, or stops at a certain switch position. The temporal action relationship describes the order of response of different circuit breakers to the same fault disturbance by the partial sequence relationship between protection initiation and tripping, without requiring all nodes to meet a fixed time difference. For the data after the tripping event, the abnormal quantities of which nodes in the original abnormal node set have disappeared, which nodes have maintained normal power supply, and which nodes have turned to an undervoltage state are re-examined. This forms the abnormal disappearance segment corresponding to the isolation boundary. The initial abnormal location, the continuous range of the abnormality, the partial sequence of protection actions, the range of abnormal disappearance after tripping, and the observation gaps therein constitute the propagation fingerprint of a single fault, so that subsequent judgment no longer depends on a single over-limit point, but on the mutual constraints of multiple nodes in time and topology.

[0047] For each candidate line segment that may contain a fault source, the counterfactual fault propagation state starts from the current directed feeder topology, setting the candidate segment as the only fault source to be tested and propagating the theoretical fault impact to the upstream and downstream nodes electrically connected to it. During propagation, if a switch that was already in an open state when the fault occurred is encountered, the propagation stops in the corresponding direction. When a circuit breaker that actually tripped is encountered, the propagation process is divided into two states according to the tripping time: before and after the action. This obtains the theoretical anomaly coverage set during the fault occurrence stage and the theoretical anomaly disappearance set after isolation is completed. The propagation is then performed according to the candidate fault... The electrical distance relationship between the fault section and each protection node generates a theoretical protection partial order, thereby obtaining a counterfactual state that can be compared item by item with the actual fault propagation fingerprint. When the abnormal node, the order of protection actions, and the boundary of abnormal disappearance in the actual fingerprint can be explained by the candidate section under the current topology, the corresponding candidate section is retained as a consistent candidate. When the consistency of the entire propagation chain can only be restored by changing the logic opening and closing state of a certain switch from the current sending state to the opposite state, the switch and the modified logic state are bound and recorded with the candidate fault section, so that the fault section determination and topology state verification are completed in the same interpretation process.

[0048] In one embodiment, please refer to Figure 1 When processing the timing data of voltage, current, zero-sequence quantities, protection start-up, tripping, and switch position as events, the protection start-up event is used as a local time reference point. The electrical quantity change segments and switch state transition segments before and after the reference point are extracted respectively. For data from different primary and secondary integrated circuit breakers with communication arrival time differences, cross-node time correspondence is established based on the common electrical disturbances detected by adjacent circuit breakers. The logical occurrence order of each event is corrected based on the cross-node time correspondence. The corrected events are encoded according to the circuit breaker identifier, event category, occurrence order, and before and after states to form a unified event sequence for fault propagation fingerprint construction.

[0049] Furthermore, selecting a protection initiation event as a local reference point does not mean that all nodes must initiate protection simultaneously. Instead, a local event cluster is established using circuit breakers that have already initiated protection. Before protection initiation, the system searches forward for initiation positions that have a direct abrupt change relationship with current, zero-sequence quantity, or voltage changes. After protection initiation, the system searches backward for tripping, switch position changes, and abnormal quantity fallback positions. These initiations and state transitions are collectively bound to the same local disturbance. When adjacent circuit breakers simultaneously detect current abrupt change edges or zero-sequence quantity initiation segments with similar morphologies, they are identified as cross-node homogeneous event anchor points. The criteria for homogeneity include consistent change direction, overlapping continuous intervals, and continuous conduction relationship in the topology. Equal amplitudes are not required because load shunting and fault locations in adjacent sections can cause differences in measured amplitudes. After multiple homogeneous event anchor points are formed, one circuit breaker with a stable time record is used as a local time reference. The relative time offset is calculated for other nodes, and the relative time offset is subtracted from the original arrival time, thus separating the stabilization delay caused by the communication link from the sequential relationship of fault propagation itself.

[0050] In this embodiment, the node time offset is determined using the criterion of minimizing the absolute deviation of multiple anchor points of the same source event, and its mathematical expression is:

[0051]

[0052]

[0053] ;

[0054] Where i represents the primary and secondary fusion circuit breaker to be corrected. This represents the set of source event anchor points where circuit breaker i can correspond to the reference circuit breaker, and r represents the anchor point index. This indicates the original recording time of anchor point r on circuit breaker i. This indicates the original recording time of the same anchor point on the reference circuit breaker. Indicates the candidate time offset. This represents the stable time offset obtained from multiple anchor points. This represents the original time of any event e to be corrected on circuit breaker i. Indicates the corrected logic occurrence time. This represents the remaining timing error at the r-th anchor point that cannot be explained by the uniform time offset. When the arrival differences of the three anchor points from the same source event relative to the reference node are 8 milliseconds, 10 milliseconds, and 9 milliseconds, respectively, the absolute deviation and the time offset are... The minimum value is obtained near milliseconds. After subtracting 9 milliseconds from the original recorded time, the remaining errors of the three anchor points are negative deviation, positive deviation and zero deviation, respectively. The first two remaining errors continue to be retained as event timing uncertainties. Therefore, an abnormal communication jitter will not directly change the order of the entire propagation chain, and the persistent uniform delay can be separated from the fault event time.

[0055] Please refer to Figure 2 , Figure 2 The horizontal axis uses a simplified Chinese sequence from point one to point twelve to represent the time-calibration samples after the anchor points of the same source event are expanded. The vertical axis represents the millisecond-level time offset. The black solid line represents the arrival difference of each anchor point relative to the reference circuit breaker, and the gray dashed line represents the time offset calculated from the formula. The defined steady-state time offset is represented by the gray dotted line. The corresponding remaining timing error. Figure 2 Used to The observation differences of multiple source anchor points, the uniform offset, and the discrete errors that cannot be uniformly explained are displayed separately.

[0056] from Figure 2 As can be seen, the arrival differences for the first three samples corresponding to B104 are approximately 8 milliseconds, 10 milliseconds, and 9 milliseconds, respectively, with a stable offset remaining around 9 milliseconds and the remaining error fluctuating slightly around zero. The stable offset for the sample corresponding to B105 is approximately 13.5 milliseconds, for the sample corresponding to B402 it is approximately 20 milliseconds, and for the sample corresponding to B801 it is approximately 30 milliseconds. The curves show a stepped rise, indicating that there are stable communication arrival offsets between different circuit breakers and the reference node, while the residuals of multiple anchor points within the same circuit breaker are relatively small.

[0057] Combination The meaning, Figure 2 This reflects that the system does not directly determine the propagation order based on the original record time, but first deducts the stable offset of each node, and then retains the remaining error as a timing uncertainty. The residual amplitude at B801 reaches about 2 milliseconds, which is higher than that of other nodes, indicating that there is more obvious communication jitter at the corresponding anchor point of this node, and a wider time tolerance should be retained when constructing the propagation fingerprint and determining the candidate order.

[0058] In a preferred embodiment, the cross-node time correspondence uses the current mutation edge, zero-sequence quantity initiation segment, or protection start-up associated segment that occur together with adjacent circuit breakers as the anchor point of the same source event; a deviation sequence is established for the communication arrival deviation of the same circuit breaker at multiple consecutive anchor points of the same source event, and the part that can be explained by the unified offset relationship is used as the node time offset, while the discrete deviation that cannot be explained by the unified offset relationship is retained as the event timing uncertainty; when forming the unified event sequence, the event order is corrected according to the node time offset, and the event timing uncertainty is added to the corresponding event, so that the subsequent propagation association retains multiple candidate orders when the time order overlaps.

[0059] Specifically, when a circuit breaker forms anchor points at multiple adjacent fault disturbances or multiple common changing positions in the same fault process, the deviation sequence is saved according to the actual occurrence order of the anchor points. The maximum and minimum deviations are not directly clipped. Instead, it is determined whether each deviation can be explained by the same time shift. If multiple deviations are distributed around the same shift, the shift is used to adjust the logical time of all events of the corresponding node. Individual records that deviate from this shift relationship are given a time-series uncertainty attribute. For example, when two adjacent protection events partially overlap in the corrected time interval, their unique order is not forcibly determined. Instead, candidate partial orders such as event A preceding event B and event B preceding event A are generated respectively. Only when the subsequent topology propagation relationship can exclude one of the candidate partial orders is the actual interpretation determined. This avoids communication jitter being solidified into the conclusion of the order of protection actions during the event sequencing stage. If a node lacks enough common event anchor points, the node does not use the time offset of other feeders or non-adjacent nodes to replace it. Instead, the original time is retained and the expression range of its time-series uncertainty attribute is expanded to keep the data source and correction basis traceable.

[0060] In one embodiment, the fault propagation fingerprint is constructed according to the upstream and downstream adjacency relationship in the directed feeder topology, including the set of the first circuit breakers to determine the abnormal electrical quantity, the set of circuit breakers that continuously exhibit abnormalities along the feeder direction, the relative order of protection activation and tripping, and the set of sections where the abnormal electrical quantity disappears after tripping. For propagation chains with missing events between adjacent circuit breakers, the propagation chain is not directly cut off. Instead, a propagation connection to be verified is generated based on the continuity relationship of events on both sides of the missing node in terms of time sequence, abnormal polarity, and topological direction. The determined propagation connection and the propagation connection to be verified are jointly written into the fault propagation fingerprint.

[0061] In this embodiment, the initial set of circuit breakers is not simply selected from the node with the smallest timestamp. Instead, all nodes in the corrected event sequence that are within the same temporal uncertainty range and have no explainable upstream prior anomalies are selected as initial candidates. If the abnormal initiation time of a branch circuit breaker and its upstream circuit breaker overlaps within the uncertainty interval, the partial order candidates of the two are retained by combining the abnormal current change direction and topological position. The set of continuous abnormal circuit breakers is checked edge by edge along the directed feeder topology. When adjacent nodes have corresponding abnormal events and the occurrence time meets the current candidate partial order, a deterministic propagation connection is established between the two. When an intermediate node lacks an event record but its upstream and downstream nodes have abnormal changes in the same direction, and the intermediate switch is in the conducting state when the fault occurs, the propagation chain is not cut off at the missing node. Instead, a propagation connection to be verified is established and the missing node is recorded as an observation gap. The abnormal polarity is used here to indicate whether the direction of current change, the direction of zero-sequence quantity initiation, or the direction of voltage drop can form a consistent propagation relationship, so that the missing data of a single node will not be automatically converted into an electrical break.

[0062] Furthermore, the relative order of protection activation and tripping is represented by a partial order set. Each partial order relationship only expresses the order that can be confirmed by existing records. For example, the downstream protection activation being earlier than the upstream tripping can form a clear constraint. However, when the time intervals of the two activation events overlap, no mandatory order is established. After the tripping, the abnormal disappearance section is recalculated with the tripping event as the boundary. For nodes that were in an abnormal state before tripping but returned to normal or lost fault current after tripping, the abnormal disappearance set is formed. For nodes that remain abnormal, they are retained in the propagation path. When the actual abnormal disappearance boundary coincides with the logical disconnection position of a circuit breaker, a definite isolation boundary is formed. When there is a missing data node between the two, a candidate isolation boundary with an observation gap is formed. The propagation fingerprint is ultimately represented by the node set, the definite propagation edge, the propagation edge to be verified, the protection partial order relationship, and the abnormal disappearance set. This avoids compressing the complex fault process into a single fault code and preserves sufficient structural information for subsequent counterfactual matching.

[0063] In a preferred embodiment, for the propagation connection to be verified, two types of candidate connections are established: one propagating from the upstream node to the downstream node and the other propagating from the downstream node to the upstream node in reverse. The direction of abnormal electrical quantity changes, protection status changes, and the conduction status of intermediate switches at both ends of the candidate connection are combined with the directed feeder topology for verification. When the candidate connection needs to cross a switch that is in the open state, the corresponding connection is transformed into a topology suspected distortion connection. When the candidate connection has continuous electrical quantity changes but discontinuous protection events, the missing protection events are marked as observation gaps, and the topological continuity of the propagation chain is maintained, so that topological distortion and event missing are entered into subsequent counterfactual matching using different propagation labels.

[0064] Specifically, for each propagation connection to be verified, both the forward propagation hypothesis and the reverse correlation hypothesis are retained. The forward propagation hypothesis requires that the two ends of the connection can form a continuous electrical path along the current power supply direction, and that the abnormal changes in the upstream and downstream can form an acceptable propagation sequence in logical time. The reverse correlation hypothesis is used to describe the subsequent response of the upstream protection when the branch fault converges to the common upstream node or the fault source is located downstream. Therefore, the reverse correlation does not change the power supply direction, but indicates that the causal relationship of the event points from the downstream fault location to the upstream protection node. When a candidate connection must pass through a switch recorded as tripped on the topology, but continuous abnormalities from the same source occur at both ends of the connection, the system does not directly discard the abnormal correlation, but marks the edge as a suspected topology distortion connection and adds the relevant switch to the set of states to be verified. When the electrical quantity changes are continuous at both ends of the connection and are logically compatible, but the intermediate node lacks protection start or trip records, the missing content is only marked as an observation gap. This makes the missing communication record and the distortion of the switch state enter different data fields, avoiding the substitution of the two causes in the subsequent joint solution.

[0065] In one embodiment, for each candidate fault segment in the directed feeder topology, the candidate fault segment is set as the fault source. The theoretical set of circuit breakers affected by the fault, the theoretical protection action sequence, and the abnormal disappearance range after the corresponding isolation action are derived along the current switch conduction relationship to form a counterfactual fault propagation state. The counterfactual fault propagation state is associated with the actual abnormal coverage range, actual action sequence, and actual abnormal disappearance range in the fault propagation fingerprint item by item. Candidate results that can only satisfy propagation consistency after changing at least one switch logic state are given a topology state correction mark for joint determination of the fault segment and switch topology state.

[0066] In this embodiment, candidate fault sections are generated from all line sections within the current feeder electrical range that can establish a connection with at least one abnormal event. Sections located entirely within branches without abnormalities and without topological connections to the abnormal propagation chain are not included in the counterfactual calculation. For a candidate section, the theoretical set of affected circuit breakers is obtained by extending along the conduction topology to the power supply side and adjacent branches that can be affected by the same fault disturbance. The theoretical protection action sequence is constructed based on the upstream and downstream positions of the candidate section and each protection node to form a partial order constraint. The action delay is not fixed to a certain value. It only requires that the protection nodes close to the candidate fault section, the upper-level protection nodes, and the isolation boundary maintain a hierarchical relationship that can explain the actual action records. The theoretical anomaly disappearance range is re-cut off from the counterfactual topology based on the actual tripping position, and it is calculated which nodes among the nodes affected by the fault source should lose the fault anomaly after the corresponding tripping. This forms a data object with the same fingerprint structure as the actual one. The matching process compares the anomaly coverage set, the protection partial order relationship, and the anomaly disappearance set separately, rather than compressing multiple pieces of information into a black-box judgment result with a preset weight.

[0067] The structural difference between the counterfactual fault propagation state and the actual fault propagation fingerprint is quantified using the following cost expression:

[0068] ;

[0069] Where k represents the candidate fault segment, and q represents the candidate topology state combination associated with the candidate fault segment k. This represents the cost of the structural difference between the counterfactual propagation state corresponding to the candidate fault segment q and the actual fault propagation fingerprint. This represents the set of anomalous overlay nodes obtained from the actual propagation fingerprint. This represents the theoretical set of anomalous coverage nodes derived under the candidate fault segment and candidate topology combination, denoted by [symbol]. To represent the symmetric difference between two sets, This represents the number of contradictory relationships between theoretically protected partial orders and practically verifiable partial orders. This represents the set of actual anomaly disappearance segments. This represents the set of anomalous disappearance segments derived from counterfactual reasoning. This indicates the number of events that cannot be explained by observation gaps or timing uncertainties. , , , All comparison coefficients are positive and remain unchanged within the same fault calculation. For example, candidate A generates 1 anomaly coverage difference, 0 partial order contradictions, 0 anomaly disappearance differences, and 1 unexplained event, while candidate B generates 0 anomaly coverage differences, 2 partial order contradictions, 0 anomaly disappearance differences, and 0 unexplained events. When set... When the cost is higher than that corresponding to the difference of a single abnormal node, candidate A can be ranked ahead of candidate B. The value expresses the engineering constraint that the causal contradiction of the protection action cannot be easily offset by the lack of local measurement. Therefore, the candidate ranking is controlled by the entire propagation structure.

[0070] Please refer to Figure 3 The horizontal axis represents candidate fault segments and topological state combinations sorted by structural difference cost from one to eleven, while the vertical axis represents cost and its components. Figure 3 The solid black line in the middle is The overall cost, with the dashed line representing the set of anomaly coverage. and The differences in coverage between them, represented by dotted lines, indicate the protection of partial order contradictions. The dotted-dash line represents the set of abnormal disappearances. and The differences between them need to be cleared.

[0071] from Figure 3It can be seen that the overall cost of candidate 1, corresponding to segment S104, is approximately 11.4, which is the lowest in the current sequence. Its abnormal coverage difference is 6, its clearance difference is 3, and its protection order conflict is 0, indicating that this candidate can maintain the actual protection action sequence well. The overall costs of candidates 2 through 4 increase sequentially, mainly due to the increase in abnormal coverage differences and unexplained events. Although the abnormal coverage difference of candidate 5 decreases to 2, its protection order conflict rises to 3, making its overall cost higher than the preceding candidates, reflecting... The punitive effect in engineering constraints.

[0072] Figure 3 The curve changes illustrate that The goal is not simply to minimize the number of anomalous nodes, but to compare structural consistency among anomalous coverage, protected partial order, anomalous disappearance range, and unexplained events. The combined cost of candidates 10 and 11 rises to over 21, mainly due to the high difference in anomalous coverage and the large number of unexplained events, indicating that even if these candidates may have connections in local segments, it is difficult to explain the entire fault propagation chain.

[0073] In one embodiment, when constructing the counterfactual fault propagation state, each candidate fault segment is used as the starting segment, and the theoretically affected nodes are expanded step by step according to the switch conduction relationship. The propagation in the corresponding direction is terminated when the first boundary switch, theoretical isolating switch, or feeder terminal in the open state is encountered. For feeders with branch structures, the theoretical abnormal coverage range on each branch is recorded separately, and the protection events at the common upstream node are used as branch propagation merging constraints. At the same time, the theoretical action partial order relationship is generated according to the electrical upstream and downstream order between the candidate fault segment and each protection node, without pre-limiting the fixed action duration, so that the differences in the protection responses of different circuit breakers can be compared in the same counterfactual propagation state.

[0074] Specifically, the propagation from the candidate fault section adopts a topological adjacency recursive method. Each time an adjacent section in the conducting state is visited, the corresponding circuit breaker is added to the theoretically affected node set, and the switch state version corresponding to the node at the time of the fault is read to continue the propagation. When a boundary switch that has already been tripped before the fault is encountered, the propagation along that edge is stopped immediately. When a disconnecting switch that trips after the fault is encountered, the propagation connection before the trip is retained and the connection is cut off in the post-trip version. When the feeder terminal is reached, the recursion in this direction ends. For nodes on the main trunk that connect multiple branches, a theoretical abnormality coverage subset of each branch is established. Abnormalities missing in one branch are not automatically propagated to another branch. The common upstream protection node serves as the convergence point between multiple subsets. Its protection start-up and trip events must be able to explain the propagation state of each branch simultaneously. If a candidate fault section requires the common upstream node to act in advance before an abnormality in a branch arrives, the corresponding candidate partial order will conflict. Thus, the branch structure is used to constrain the fault location.

[0075] Furthermore, the theoretical action partial order only limits the electrical hierarchy relationship and does not limit the specific millisecond value. For example, when a candidate fault section is located after a downstream sectional circuit breaker, if the downstream node records protection activation and the upstream node subsequently trips, a partial order constraint is established that the downstream activation precedes the upstream trip. If the upstream protection activation and the downstream protection activation overlap in the time uncertainty interval, the state in which the two have no definite order is retained, and the overlapping event is not forcibly interpreted as a step-by-step action. When a candidate topology combination changes the switch conduction relationship, the theoretical action partial order is regenerated with the new electrical upstream and downstream relationship, so that the protection order judgment is no longer fixed on the topology before the fault. At the same time, the anomaly disappearance range is also recalculated according to the new isolation boundary. Therefore, the same candidate fault section can form different counterfactual states under different topology interpretations. The subsequent joint solution compares the complete fault location and topology state combination, rather than determining the fault location first and then checking the remote signaling separately.

[0076] In one embodiment, please refer to Figure 4 The topology state correction marker is represented as a candidate logic state variable corresponding to a specific switch. For the same candidate fault segment, candidate topology combinations are generated, including those that do not modify the topology, modify a single switch state, and modify multiple switch states with propagation correlation. For each candidate topology combination, the corresponding counterfactual fault propagation state is regenerated and jointly matched with the fault propagation fingerprint containing suspected topology distortion connections, observation gaps, and event timing uncertainties. A candidate interpretation set is constructed according to fault propagation inconsistencies, topology state modification items, and unexplainable event items. From the candidate interpretation set, fault segment-topology state combinations that do not have contradictory propagation relationships are retained.

[0077] In this embodiment, candidate logic state variables are only established for switches that are directly related to the fault propagation contradiction. Arbitrary flipping of all switches along the entire feeder is not performed. Specifically, the generation method involves reading the switches crossed by the suspected distorted topology connection and the switches near the anomaly disappearance boundary whose states are inconsistent with the measured propagation. These switches form a local set to be verified. After generating the original topology combination based on the current remote signaling state, a single state reversal combination and a combination consisting of multiple suspected switches on the same propagation chain are generated. If two suspected switches are located on unrelated, anomaly-free branches, they are not forced into the same correction scheme due to combination enumeration. After each candidate topology combination is generated, the derivation of theoretical anomaly coverage, protection partial order, and the anomaly disappearance range after isolation is re-executed. This ensures that switch state modifications must be able to explain the actual propagation structure to be retained, rather than solely determining whether to modify the topology based on the remote signaling anomaly probability. When a modified state eliminates the continuous anomaly contradiction across the disconnected switch but simultaneously causes the upstream protection action to be inexplicable, the combination is recorded as having a new propagation conflict and removed from the consistent candidates.

[0078] In a preferred embodiment, each element in the candidate interpretation set simultaneously stores the candidate fault segment, the modified set of switches, the candidate logic state of each switch, the adopted event timing candidate order, the location of the event considered as an observation gap, and the propagation contradiction that is still unexplainable. During matching, the observation gap is allowed to cover the data location that has been confirmed as missing in the event acquisition stage, but it is not allowed to re-mark records that are opposite to existing events as missing. For example, if a circuit breaker has clearly recorded the tripping state but the actual abnormal propagation must cross the location, the candidate interpretation can only be corrected by the topology state, and the switch state contradiction cannot be replaced by the missing protection event. Similarly, the timing uncertainty is only allowed to adjust the order of events with overlapping time intervals, and cannot change the action order that already has a clear time interval. Through this field constraint, each type of uncertainty source can only explain its corresponding data conflict, avoiding multiple uncertainty conditions compensating for each other and causing any candidate to obtain a consistent result.

[0079] In one embodiment, for multiple candidate event sequences formed by event timing uncertainties, they are combined with candidate topologies to form propagation interpretation branches. In each propagation interpretation branch, the closure relationship between the initial anomaly location, adjacent propagation continuity, protection action partial order, and anomaly disappearance boundary after isolation is checked along the directed feeder topology. If the same switch state modification can simultaneously eliminate contradictions in multiple independent propagation chains, the switch state modification is retained as a shared topology correction term. If a state modification can only explain a single isolated event and causes new contradictions in other propagation chains, the state modification is removed from the corresponding propagation interpretation branch to form a joint interpretation result with the minimum set of topology corrections.

[0080] Specifically, each propagation interpretation branch starts from a defined candidate fault segment and fixes the current event candidate order and topology state combination. Then, it checks four types of closure conditions along the fault propagation diagram. The initial anomaly location must be able to propagate outward from the candidate fault segment without requiring the fault to start outside the already disconnected electrical boundary. Adjacent propagation continuity requires that all determined propagation connections can be explained by the current topology conduction relationship, and the propagation connection to be verified can only be supplemented by the already registered observation gap or the topology suspected distortion connection. Protection action partial order requires that the current event sequence does not have any action relationship that directly conflicts with the electrical upstream and downstream relationship. The anomaly disappearance boundary after isolation requires that the theoretical cut-off position can be compatible with the actual anomaly disappearance set. When the same switch state reversal can close the main propagation chain and a certain branch propagation chain at the same time, it is regarded as a shared topology correction item and is only recorded once in the corresponding joint interpretation. When a state reversal can only explain an isolated node, but cuts off other originally continuous propagation connections, the reversal does not enter the minimum correction set.

[0081] The selection of joint interpretation results is based on the following constraint cost expression:

[0082] ;

[0083] Where h represents a complete propagation interpretation branch, This represents the constraint cost of the h-th propagation interpretation branch. This indicates the number of propagation closure conflicts that still exist in the branch. Indicates the number of switch logic states that need to be modified. This indicates the number of time-determined choices that were adopted. This indicates that continued observation gaps are still necessary to maintain a continuous number of propagation connections. , , , These are the positive coefficients of the corresponding constraint terms and remain fixed during the same fault judgment process. The coefficient for propagating a closed-loop conflict is set to a relationship that cannot be offset by adding a single topology modification term. This ensures that branches with clear causal contradictions are not preferentially retained simply because one switch is modified less. For example, explanation branch A has no closed-loop conflict but requires modifying one switch and using one timing candidate; explanation branch B retains one closed-loop conflict and does not modify the switch. Greater than When the cost of branch A is lower than that of branch B, the value is chosen so that the joint interpretation prioritizes satisfying the verifiable propagation closure relation, and then the result with fewer modification switches and less dependence on uncertain information is selected from the closure candidates.

[0084] Please refer to Figure 5 , Figure 5 The horizontal axis represents the propagation interpretation branches retained for joint interpretation screening, with branches one through eleven on the horizontal axis. The vertical axis represents the cost of branch constraints. The solid black line represents... The comprehensive constraint value, the dashed line corresponds to the propagation closure conflict. The number of times the switch logic state is modified corresponds to the number of dots and lines. The dotted line corresponds to the observation gap dependency. . Figure 5 around , , and The constraint weights they represent indicate the reasons why different branches are retained or downgraded.

[0085] Figure 5In the middle, the comprehensive constraint values ​​of branches one through seven are all approximately 3.0, and the closure conflict and switch modification are both 0. The main cost comes from the dependence on three observation gaps, indicating that although these branches need to wait for the missing measurement locations to be supplemented and verified, the fault propagation closure relationship has not been broken. The comprehensive constraint value of branches eight through ten is approximately 4.5, slightly higher than the previous group, because one temporal uncertainty choice has been added, indicating that these branches can still be retained, but the interpretation of the event sequence needs to be more carefully reviewed.

[0086] Furthermore, if multiple propagation interpretation branches have the same minimum topology correction set and differ only in the order of overlapping events at a certain time, these branches are retained as temporal sub-interpretations under the same joint interpretation result, without expanding the fault segment range due to the inability to determine the strict order of individual events. If multiple branches require modification of different switches but can close all propagation relationships, then no single switch state is temporarily determined as the unique true state. Instead, these switches are maintained as candidate logic states, and the power supply path passing through these switches is marked as the topology unconfirmed area in the subsequent fault boundary output stage. If different branches modify different switches but derive the same upstream and downstream fault boundaries, then the fault boundary can be directly output from the shared part. This separates the already determined fault segment from the multiple solutions still available for individual non-boundary switch states, ensuring that the uncertainty of the topology state is not unconditionally expanded to the entire feeder.

[0087] In one embodiment, the upstream and downstream fault boundaries corresponding to the actual propagation termination positions on both sides of the candidate segment where the fault source is located are extracted based on the joint interpretation results. The segments located outside the fault boundaries and whose power supply paths do not contain shared topology correction terms are marked as topology-determined areas. For segments where the power supply path still has multiple candidate interpretations after passing through the topology state, they are kept as topology-unconfirmed areas. The fault boundaries, topology-determined areas, topology-unconfirmed areas, and the candidate logic states of the corresponding switches are combined to form self-healing topology constraint data, which is used by the subsequent distribution network self-healing recovery process to select isolation objects and recoverable paths for non-fault power loss segments.

[0088] In this embodiment, the upstream fault boundary is traced back to the power supply side from the candidate fault segments in the joint interpretation results. The nearest boundary position is selected so that the actual fault propagation stops after the isolation action and its upstream node no longer contains fault anomalies. The downstream fault boundary checks the last segment that still maintains a continuous abnormal association with the candidate fault source and its subsequent conduction state along the fault propagation direction. In the case of branches, branch downstream boundaries are formed and combined with the common upstream boundary. This allows the fault segment expression to cover the propagation structure at the junction of the trunk and branches. For cases where the fault boundaries are consistent between the joint interpretation results but there are multiple solutions for some non-boundary switch states, the boundary is still output according to the common result. Each segment outside the boundary is then traced for power supply path. If there is no shared topology correction term and no switch with multiple candidate logic states in the path from the current energized node to the segment, it is marked as a topology determined area. If the power supply path must pass through a switch whose candidate state has not yet been uniquely determined, it is marked as a topology unconfirmed area, and the candidate logic state set of the corresponding switch is kept for subsequent recovery decision reading.

[0089] The self-healing recovery topology constraint data uses the same feeder node identifiers as those used in fault event analysis to establish an index. The fault boundary field stores the circuit breaker and section relationships corresponding to the upstream boundary and each downstream boundary. The topology determined area field stores non-faulty sections that can be uniquely supported by the current propagation interpretation. The topology unconfirmed area field stores sections that are not directly entered into the recovery path due to multiple solutions for candidate switch states. The candidate logic state field stores the possible states of each unconfirmed switch in all retained joint interpretations, along with the propagation interpretation identifier that forms the candidate state. When generating isolation objects, subsequent self-healing recovery calculations must use the fault boundary as an insurmountable constraint. When searching for recovery paths for non-faulty power loss sections, the connection relationships in the topology determined area are used first. For candidate paths that must pass through the topology unconfirmed area, they are not considered as equally reliable connections as determined paths, but are kept as unconfirmed until a new field switch state, subsequent electrical quantity changes, or control feedback can exclude some of the candidate logic states.

[0090] Preferably, during the normal operation phase before a fault occurs, the system continuously maintains the event relationships of the primary and secondary integrated circuit breakers during the most recent period of stable operation. This is used to determine which adjacent nodes can repeatedly observe the common changes caused by load fluctuations from the same source and normal switching operations. Historical relationships are only used to establish the structural basis for cross-node events and are not directly used as the current fault classification template. When a current fault occurs, the event processing logic still regenerates the propagation fingerprint based on the actual voltage, current, zero-sequence quantity, protection start, trip, and switch position records before and after the current fault. For example, if the communication time of an upstream circuit breaker is consistently later than that of adjacent nodes, but its deviation remains stable across multiple common events, then the stable deviation can be corrected by the node time offset. However, if an additional single message suddenly arrives late during a fault, its remaining difference is recorded as a timing uncertainty rather than revising the entire node time base. This processing allows long-term stable offsets and single communication disturbances to have different representations at the data level.

[0091] In a preferred embodiment, when multiple protection activation events occur during a fault but only some circuit breakers form a clear trip record, the system establishes two topology versions before and after the confirmed trip node. For non-tripped nodes, it only records the protection status change without fabricating their switching actions, and checks the abnormal coverage area before the confirmed trip and the abnormal disappearance area after the trip. If the fault abnormality of a certain branch still exists after the trip, the branch remains in the propagation diagram and continues to participate in the derivation of candidate fault sections. If the abnormality disappears from multiple nodes simultaneously after the trip, it is determined whether they belong to the same isolation range based on the topology connection between these nodes and the tripped circuit breaker. If there is a disconnected switch between an abnormal node and the tripped circuit breaker on the current remote signaling topology, a topology suspected distortion connection is generated and enters the candidate logic state combination, instead of directly assuming that the abnormal disappearance event is wrong. In this way, it is possible to use which positions simultaneously lose fault characteristics after the action to reverse verify the switch conduction state recorded before the fault.

[0092] Furthermore, when there is a short-term missing measurement of voltage, current, or zero-sequence quantity in the primary and secondary integrated circuit breakers, the event-based processing does not perform numerical interpolation on the missing samples before using them as actual fault quantities. Instead, it checks whether an event boundary with the same trend can be formed before and after the missing measurement interval. If an abnormal change has occurred before the missing measurement and the abnormal state remains after the missing measurement, the missing measurement interval is regarded as an unknown internal segment of the same continuous abnormal event. If the situation is normal before the missing measurement but an abnormality occurs directly after the missing measurement, the abnormality start time is represented as a time interval and assigned a timing uncertainty attribute. If the protection start event is located within this time interval, a compatible relationship between the electrical abnormality and the protection event is established. Subsequent propagation matching only uses the state transition and time interval relationship that can be determined, and does not generate specific sample values ​​that do not exist. This makes the missing observation appear as a constraint widening rather than falsified data in the algorithm, so that it can still form a propagation connection to be verified with the determined events of adjacent circuit breakers.

[0093] In this embodiment, when multiple branches of the feeder are simultaneously affected by the same fault disturbance, the propagation fingerprint stores the abnormal chain of each branch separately with the common upstream node as the relationship convergence point. Instead of simply sorting all the abnormal nodes of multiple branches into a linear sequence by time, for example, when the candidate fault source is located on the main trunk and its downstream connects to two branches, the abnormal events in the two branches can occur in parallel within the time uncertainty range, as long as the common upstream node and the entrance of each branch satisfy the topological continuity relationship. The counterfactual propagation state also generates a theoretical abnormal set according to the same branch structure. If the candidate fault source is assumed to be located inside one of the branches, whether the other branch should have the same source abnormality is determined by the electrical connection state at the common upstream node. Thus, the actual existence or non-existence of the abnormality in the other branch can be used to eliminate the candidate fault location that cannot be explained by branch propagation. Furthermore, when the common upstream switch state is sent incorrectly, the propagation contradictions that occur simultaneously in multiple branches can all point to the same shared topology correction term.

[0094] Specifically, candidate fault sections and candidate switch logic states are stored in a binding relationship. Each time a counterfactual propagation state is regenerated, it starts from the corresponding candidate topology copy without directly modifying the original topology data of the master station. The candidate topology copy only changes the logical conduction state of the suspected switch during the reasoning process and records the source of modification, associated propagation connection, and eliminated or added propagation contradictions. After the joint interpretation is completed, only the retained candidate state is output as self-healing recovery topology constraint data, instead of directly overwriting the original remote signaling during the fault reasoning stage. This avoids writing the temporary candidate state back to the basic topology and affecting other candidate branches when a fault interpretation is not yet completed. At the same time, it allows the same switch to maintain coexistence if it has both closing and opening candidate states in different interpretations, until the propagation closure relationship or a new field event provides further constraints.

[0095] Preferably, during the self-healing recovery preparation phase, the determination of the isolation object directly reads the upstream and downstream fault boundaries and checks whether the corresponding boundary switches belong to the shared topology correction item or still have multiple candidate logical states. If the boundary itself has multiple solutions, a certain candidate state is not directly solidified as the basis for determining isolation. Instead, the outer safety boundary recognized by all candidate branches is found along the joint interpretation results and a constraint to be executed is formed. If the boundary state has been jointly confirmed by the anomaly disappearance range and the protection action relationship, the boundary can be directly used as the isolation relationship between the fault section and the non-fault section. The recovery path search of the non-fault power loss section is limited to outside the fault boundary, and the search process only uses the topology-determined area as the determined connected edge and the topology-unconfirmed area as the restricted connected edge, so that the subsequent recovery path will not automatically cross the switch state that has not been verified by fault propagation because the master station has a static connection relationship.

[0096] In a preferred embodiment, when a new switch position record arrives after a fault event ends or a new stable voltage and current state appears during the recovery preparation process, the newly added record is mapped back to the retained joint interpretation result according to the same circuit breaker identifier as the original fault event. This is used to delete candidate logical states that directly contradict the newly added record. For example, if a certain switch to be confirmed is a candidate for being on and off in two joint interpretations, and the subsequent stable voltage relationship indicates that both sides of the switch maintain the same power supply state and there are no other connection paths, the candidate state that cannot explain the continuity of power supply can be deleted. After the candidate set is reduced, the topology determination area and the topology confirmation area passing through the switch are re-divided. However, the fault propagation fingerprint, the original event time, and the already determined fault boundary are still recorded in the original record. The data of the fault occurrence stage is not rewritten due to subsequent state confirmation, thereby forming a continuous data chain from fault occurrence, event interpretation, topology verification to recovery constraint update.

[0097] In this embodiment, the entire fault handling process is based on the same directed feeder topology identification system, which runs through data acquisition, eventification, time correction, propagation fingerprint construction, counterfactual state generation, candidate topology combination, joint interpretation, and recovery constraint output. Circuit breaker electrical quantity events and switch state events are associated with specific topology nodes before entering inference. Cross-node time correction only modifies the logical time of the event without modifying the event source. Propagation fingerprints are generated only based on established topology adjacency and verifiable event relationships. Counterfactual calculation only changes the switch logical state with propagation contradiction evidence in the candidate topology copy. Joint interpretation only allows timing uncertainty, observation gaps, and suspected topology distortion connections to explain their corresponding uncertainty sources. The fault boundary extracts common results that can be supported by the actual abnormal propagation termination position from multiple reserved interpretations. Therefore, even if message out-of-order, local data loss, and individual switch remote signaling distortion occur simultaneously during the fault process, the joint derivation of the fault section and switch topology state can be completed without fabricating measurement data or arbitrarily modifying the feeder structure, and topology constraint data can be directly used in the subsequent distribution network self-healing recovery process.

Claims

1. A distribution network self-healing and recovery system based on an outdoor primary and secondary integrated circuit breaker, characterized in that, include: A directed feeder topology is established based on the electrical connection relationship of the primary and secondary integrated circuit breakers in the power distribution feeder; Acquire the timing data of voltage, current, zero-sequence quantity, protection start, trip and switch position of the primary and secondary fusion circuit breakers before and after the fault, and form a fault event sequence according to a unified event time scale; Based on the directed feeder topology, the propagation correlation of the fault event sequence between adjacent circuit breakers is determined, forming a fault propagation fingerprint; For each candidate fault segment, a corresponding counterfactual fault propagation state is constructed. The counterfactual fault propagation state is matched with the fault propagation fingerprint for consistency, and the fault boundary and the switch topology state that is inconsistent with the actual propagation relationship are jointly determined.

2. The distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker according to claim 1, characterized in that, When performing event-based processing on the timing data of voltage, current, zero-sequence quantity, protection start, trip, and switch position, the protection start event is used as a local time reference point, and the electrical quantity change segment and switch state transition segment before and after the reference point are extracted respectively. For data from different primary and secondary integrated circuit breakers with communication arrival time differences, a cross-node time correspondence is established based on the common electrical disturbances detected by adjacent circuit breakers, and the logical order of each event is corrected based on the cross-node time correspondence. The corrected events are encoded according to the circuit breaker identifier, event category, occurrence order, and preceding and following states to form a unified event sequence used for constructing the fault propagation fingerprint.

3. The distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker according to claim 1, characterized in that, The fault propagation fingerprint is constructed according to the upstream and downstream adjacency relationship in the directed feeder topology, including the initial set of circuit breakers that determine the abnormal electrical quantity, the set of circuit breakers that continuously exhibit abnormalities along the feeder direction, the relative order of protection activation and tripping, and the set of sections where the abnormal electrical quantity disappears after tripping. For a propagation chain with missing events between adjacent circuit breakers, the propagation chain is not directly cut off. Instead, a propagation connection to be verified is generated based on the continuity of events on both sides of the missing node in terms of time sequence, abnormal polarity, and topological direction. The determined propagation connection and the propagation connection to be verified are written together into the fault propagation fingerprint.

4. The distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker according to claim 1, characterized in that, For each candidate fault segment in the directed feeder topology, the candidate fault segment is set as the fault source. The set of circuit breakers theoretically affected by the fault, the theoretical protection action sequence, and the abnormal disappearance range after the corresponding isolation action are derived along the current switch conduction relationship to form a counterfactual fault propagation state. The counterfactual fault propagation state is associated with the actual anomaly coverage, actual action sequence, and actual anomaly disappearance range in the fault propagation fingerprint. Candidate results that can only satisfy propagation consistency after changing at least one switch logic state are given a topology state correction flag for joint determination of fault segment and switch topology state.

5. A distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker according to claim 2, characterized in that, The cross-node time correspondence uses the current sudden change edge, zero-sequence quantity change segment, or protection start-up associated segment that occur together with adjacent circuit breakers as the anchor point of the same source event. For the communication arrival deviation of the same circuit breaker at multiple consecutive anchor points of the same source event, a deviation sequence is established, and the part that can be explained by the unified offset relationship is used as the node time offset, while the discrete deviation that cannot be explained by the unified offset relationship is retained as the event timing uncertainty. When forming a unified event sequence, the event sorting is corrected according to the node time offset, and the event timing uncertainty is added to the corresponding event so that subsequent propagation associations retain multiple candidate orders when there is overlap in time order.

6. A distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker according to claim 3, characterized in that, For the propagation connection to be verified, two types of candidate connections are established: one propagating from the upstream node to the downstream node and the other reverse association from the downstream node to the upstream node. The abnormal electrical quantity change direction, protection status change, and intermediate switch conduction status at both ends of the candidate connection are combined with the directed feeder topology for verification. When the candidate connection needs to cross a switch that is in the open state, the corresponding connection is converted into a topology suspected distortion connection. When the candidate connection has continuous electrical quantity changes but discontinuous protection events, the missing protection events are marked as observation gaps, and the topological continuity of the propagation chain is maintained, so that topological distortion and event missing are entered into subsequent counterfactual matching using different propagation labels.

7. A distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker according to claim 4, characterized in that, When constructing the counterfactual fault propagation state, each candidate fault segment is taken as the starting segment, and the theoretically affected nodes are expanded step by step according to the switch conduction relationship. The propagation in the corresponding direction is terminated when the first boundary switch, theoretical isolation switch or feeder terminal in the open state is encountered. For feeders with branch structures, the theoretical anomaly coverage range on each branch is recorded separately, and the protection events at the common upstream node are used as branch propagation merging constraints. Simultaneously, a theoretical action partial order relationship is generated based on the electrical upstream and downstream sequence between the candidate fault section and each protection node, without pre-limiting a fixed action duration, so that the differences in protection responses of different circuit breakers can be compared in the same counterfactual propagation state.

8. A distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker according to claim 4, characterized in that, The topology state correction flag is represented as a candidate logic state variable corresponding to a specific switch, and candidate topology combinations are generated for the same candidate fault segment, namely, no topology modification, modification of a single switch state, and modification of multiple switch states with propagation correlation. For each candidate topology combination, the corresponding counterfactual fault propagation state is regenerated and jointly matched with the fault propagation fingerprint containing suspected topological distorted connections, observation gaps and event timing uncertainties. A candidate explanation set is constructed based on inconsistencies in fault propagation, topology state modifications, and unexplainable events. From the candidate explanation set, fault segment-topology state combinations that do not have contradictory propagation relationships are retained.

9. A distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker according to claim 8, characterized in that, For the multiple candidate event sequences formed by the event timing uncertainty, they are combined with the candidate topology to form propagation interpretation branches. In each propagation interpretation branch, the closure relationship between the initial anomaly location, adjacent propagation continuity, protection action partial order, and anomaly disappearance boundary after isolation is checked along the directed feeder topology. If the same switch state modification can simultaneously eliminate contradictions in multiple independent propagation chains, then the switch state modification is retained as a shared topology correction term. If a state modification can only explain a single isolated event and causes new contradictions in other propagation chains, then the state modification is removed from the corresponding propagation explanation branch to form a joint explanation result with the minimum set of topological corrections.

10. A distribution network self-healing and restoration system based on an outdoor primary and secondary integrated circuit breaker according to claim 9, characterized in that, Based on the joint interpretation results, the upstream and downstream fault boundaries corresponding to the actual propagation termination positions on both sides of the candidate segment where the fault source is located are extracted, and the segments located outside the fault boundaries and whose power supply paths do not include the shared topology correction term are marked as topology determination areas. For power supply paths passing through sections where multiple candidate interpretations still exist in the topology state, these sections are retained as topology unconfirmed areas. The fault boundary, the topology determination area, the topology unconfirmed area, and the candidate logic states of the corresponding switches are combined to form self-healing recovery topology constraint data, which is used by the subsequent distribution network self-healing recovery process to select isolation objects and recoverable paths for non-fault power loss sections.

Images