A primary and secondary fusion integrated ring main unit line protection data analysis method

Abstract

This invention relates to the field of power line protection data processing technology, specifically a method for analyzing protection data of a primary and secondary integrated ring network box line. The method acquires primary electrical quantities, secondary protection events, and switch state sequences, converting them into event atoms carrying attributes of data source, time interval, and electrical direction. It constructs a dynamic line topology map based on switch state changes and generates topology boundary migration markers. Based on fault electrical propagation constraints and protection action logic constraints, it constructs a physical timing diagram and a control timing diagram, forming a dual-time-series causal graph with topology boundary migration markers as coupling nodes. Candidate causal chains are obtained through interval constraint reasoning to determine the fault section, power supply direction, first-action protection, backup protection, and abnormal action links. This invention can distinguish between upload delay, action delay, and location feedback delay when faults and switch switching overlap or data timescales are inconsistent, avoiding erroneous merging of events before and after switching.

Description

Technical Field

[0001] This invention relates to the field of power line protection data processing technology, specifically a method for analyzing primary and secondary integrated ring network box line protection data. Background Technology

[0002] Integrated primary and secondary ring main units typically upload primary-side sampled data such as current, voltage, and zero-sequence components, along with secondary-side events such as protection initiation, protection output, opening and closing commands, position feedback, reclosing results, and blocking status, to the station or main station analysis platform. Current line protection data analysis generally involves first converting and arranging the data from different sources according to their time scales, then determining the fault occurrence time, protection initiation time, and fault clearing time based on preset protection settings, action time limits, and switch positions. For primary-side data, characteristics such as amplitude abrupt changes, zero-sequence component exceeding limits, voltage drops, and current direction changes are commonly used to identify line disturbances. For secondary-side data, action records are typically generated in a fixed order: protection initiation, output commands, switch opening, and position feedback. The analysis platform summarizes these records into an event list and evaluates protection action delay, switch response time, and the coordination between primary and backup protection based on the time difference between adjacent events. When the clocks of all acquisition modules are synchronized, the communication link is stable, and the line operation mode remains unchanged, the resulting event list can reflect the main state changes before and after the fault, providing operators with a basis for fault querying and action tracing.

[0003] Existing technologies also establish a line topology model based on the primary wiring relationship of the ring main unit, mapping the switch's open / closed state to the branch connection or disconnection state, and then projecting the detected fault electrical quantities and protection action records to the corresponding line sections. During fault location, current surges, voltage drops, and zero-sequence component changes of adjacent nodes are typically compared along a fixed power supply direction, and fault sections are divided based on the location where fault characteristics appear and disappear. During protection action analysis, the protection at the first exit point is determined as the primary protection, and the protection at subsequent exit points is determined as backup protection, based on the preset protection range. Some schemes can recalculate the topology after receiving a switch position change, but usually the position feedback time is directly used as the topology switching time, and data analysis is performed before and after the switching. For timescale deviations between different data sources, a unified offset correction or supplementation according to the received time is often used; for missing, duplicate, or sequentially conflicting events, the process is often completed by deleting duplicate records, linearly supplementing points, or manually selecting the event sequence. The above processing methods rely on determining the time, topology, and fixed action sequence, and do not jointly express the transient connection state during switch switching and the correlation between the primary electrical propagation process and the secondary control process.

[0004] When line faults and switchovers overlap in a primary and secondary integrated ring main unit, primary electrical quantities, protection records, control commands, and position feedback are generated via different sampling and uploading links. The recording timestamps are affected by sampling periods, module clock deviations, buffer waits, and communication transmission, resulting in multiple events corresponding to the same physical process only being able to be determined to occur within a certain time interval. Simultaneously, there is a process delay from receiving a command to the change in contact state and then to the transmission of position feedback. If the line topology is switched directly according to the position feedback time, the actual connection relationship to which the fault current propagation belongs may be incorrectly classified as the state before or after the switchover. Existing analysis methods using unified time sequencing and fixed topology projection cannot establish a consistent causal correspondence between the fault electrical propagation chain and the protection initiation, output, tripping, and feedback chains when both timing uncertainty and topology boundary migration exist simultaneously. This easily leads to misinterpreting data upload delays as protection action delays, status feedback lags as switch response anomalies, or mixing events before and after topology switching into the same action chain, making it difficult to reliably determine the fault section, the actual power supply direction, and the action attribution of the main and backup protection. Summary of the Invention

[0005] The purpose of this invention is to provide a data analysis method for integrated primary and secondary ring network box line protection, which can solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for analyzing data of integrated primary and secondary ring network box line protection includes: acquiring the primary electrical quantity sequence, secondary protection event sequence and switch status sequence of the line, and converting each sequence into event atoms carrying data source, time uncertainty interval and electrical direction attributes; A dynamic line topology map is constructed based on the changes in switch states, and topology boundary migration markers representing the connection relationships before switching, the connection relationships after switching, and the connection relationships to be confirmed are generated in the state change interval. A physical timing diagram is constructed based on the electrical propagation constraints of the fault, and a control timing diagram is constructed based on the logical constraints between protection start-up, output commands, and position feedback. Using the topology boundary migration marker as the coupling node, the physical timing diagram and the control timing diagram are combined into a dual-time causal graph; Perform interval constraint reasoning on the dual-temporal causal graph to obtain candidate causal chains that satisfy electrical propagation constraints, protection logic constraints, and topology migration constraints; The fault section, power supply direction, first-action protection, backup protection, and abnormal action links are determined based on each candidate causal chain.

[0007] Preferably, the generation of the event atom includes: dividing a primary event segment into segments based on the abrupt change start point, direction reversal point, voltage drop start point, zero-sequence component duration boundary, and fault disappearance point in a primary electrical quantity sequence, and incorporating the sampling clock error and upload delay interval into the corresponding time uncertainty interval; The secondary event segment is formed by the start record, criterion confirmation record, output record, opening and closing command, position feedback and blocking record in the secondary protection event sequence, and the corresponding time uncertainty interval is determined according to the local timestamp and the received timestamp of the same record. The electrical direction attribute is generated based on the current direction relationship, phase sequence relationship, and zero-sequence polarity of adjacent event segments; The data source identifier, event category, time uncertainty interval, electrical direction attribute, and associated switch identifier are encapsulated into the event atom.

[0008] Preferably, the generation of the topology boundary migration marker includes: extracting the control command time, position feedback time, and corresponding branch current continuity change time for the switch that undergoes a state change, and converting the three types of time into their respective time intervals; The starting point of the control command interval forms the termination candidate boundary of the connection relationship before the switch, and the ending point of the position feedback interval forms the starting candidate boundary of the connection relationship after the switch. The two types of candidate boundaries are cross-constrained by the interval where the branch current changes from continuous to interrupted or from interrupted to continuous. Within the cross-constraint interval, the topology subgraph before switching, the topology subgraph after switching, and the transition topology subgraph containing the connection edges to be confirmed are retained simultaneously, and the node mapping relationship of the same electrical node between the three topology subgraphs is established.

[0009] Preferably, the generation of the dual-time-series causal graph includes: in the physical time-series graph, connecting events that are adjacent in line topology and conform to the fault current propagation direction, voltage drop sequence, and fault quantity continuity into physical causal edges; In the control timing diagram, the start event, criterion confirmation event, exit event, switch command event and position feedback event belonging to the same protection action chain are connected as control causal edges; Based on the pre-switch topology subgraph, transition topology subgraph, and post-switch topology subgraph corresponding to the topology boundary migration marker, a topology stage identifier is added to the cross-graph event atom; Only when there is an overlapping relationship that satisfies the topological stage identifier within the uncertain interval of event time, a coupled causal edge connecting the physical timing diagram and the control timing diagram is established.

[0010] Preferably, the determination of the time uncertainty interval includes: extracting paired anchor points that can characterize the same line disturbance from the primary event segment and the secondary event segment. The paired anchor points include fault quantity change and protection start, output record and branch current interruption, reclosing command and voltage recovery. Segmented time difference constraints are formed based on the time scale difference of each pair of anchor points in different data sources, and the segmented time difference constraints are propagated to adjacent event atoms according to the continuous interval of the data source time scale. When the time difference constraints of adjacent segments overlap, their intersection is taken to correct the corresponding time uncertainty interval; when there is no intersection, the branch time interval is retained and the source of conflict is recorded. Write the corrected time uncertainty interval back to the event atom, while maintaining the correspondence between the original timescale and the corrected interval.

[0011] Preferably, the construction of the transition topology subgraph includes: extracting a local topology region centered on the switch where the state change occurs, which includes the upstream protection node, the downstream protection node, and the adjacent connection node; Write the connection edge before switching, the connection edge after switching, and the connection edge to be confirmed into the local topology region respectively to generate candidate topology slices corresponding to different connection combinations; For each candidate topology slice, a consistency check is performed based on the continuity of branch currents, the correlation of node voltages, the order of control commands and location feedback, and the range of fault disappearance. Candidate topology slices that cannot simultaneously accommodate primary event segments and secondary event segments are deleted. The remaining candidate topology slices are connected to the pre-switch topology subgraph and the post-switch topology subgraph according to the node mapping relationship to form a transition topology subgraph with topology migration path.

[0012] Preferably, the interval constraint reasoning includes: converting each causal edge into the corresponding constraint type among the preceding interval endpoint not later than the following interval start point, intervals allowing overlap, or intervals being mutually exclusive, and grouping causal edges that share event atoms but have incompatible constraint types into a conflict set; Extract a closed conflict subgraph from each conflict set that simultaneously involves the physical causal edge, the control causal edge, and the topological migration constraint; Using constraints such as protecting the integrity of the action chain, the continuity of electrical fault propagation, and unidirectional migration of the topology phase, candidate resolution branches are generated to retain, disconnect, or reconnect the causal edges in the closed conflict subgraph. The interval satisfiability test is re-executed for each candidate resolution branch, and the branch that does not have a time cycle and all event atoms have reachable predecessors or reachable successors is determined as the candidate causal chain.

[0013] Preferably, the screening of the candidate causal chain includes: determining the earliest event atom in each candidate causal chain that can simultaneously connect the current direction change, voltage drop and protection start-up events, and taking the topology node where the earliest event atom is located as a candidate fault start point; Tracee the fault quantity along the fault causal chain to the corresponding protection node in each exit event, and record the protection zone boundary corresponding to each exit event, the protection node corresponding to the first exit event, the protection node corresponding to subsequent exit events, and the fault disappearance boundary after the switch state changes. The fault section is determined based on the minimum connected section between the candidate fault starting point and the fault quantity disappearance boundary. The power supply direction is determined based on the directed connection relationship of the propagation path. The first-action protection, backup protection and abnormal action links are marked based on the continuous relationship between the output event and the corresponding position feedback on the control causal edge.

[0014] Preferably, it also includes: when the corrected time uncertainty interval still causes event duplication or event gap in the same protection action chain, constructing an event equivalence group according to the data source identifier, event category, associated switch identifier and time interval inclusion relationship, and merging duplicate records with the same preceding event and the same succeeding event into a composite event atom; For the event gap, control causal constraint and physical causal constraint events are extracted from the event atoms on both sides of the gap, respectively, and placeholder compensation atoms carrying allowed event categories, allowed topology nodes and allowed time intervals are generated; When the placeholder compensation atom can simultaneously connect to the previous secondary event segment and the corresponding topology migration path, it is written into the dual-time-series causal graph, and the original gap branch and the placeholder compensation branch are retained side by side.

[0015] Preferably, the processing of the candidate causal chain further includes: merging multiple candidate causal chains into equivalent chains according to the consistency relationship of event atom sets, topology migration paths, and protection action chains, and extracting causal edges that appear in common and have the same connection order from each equivalent group to form a common causal skeleton; Subchains connected to the common causal skeleton but differing in the topological stage identifier, the time uncertainty interval of the event, or the attribution of the exit event are identified as divergent subchains; For each branch subchain, backtrack to the corresponding topology boundary migration marker, branch time interval, or conflict resolution branch to establish a mapping between the source of the branch and the fault section, power supply direction, and protection action attribution; When the common causal skeleton of all equivalence groups points to the same result, a unified analysis record is output; otherwise, a branch analysis record containing the common part and each divergent subchain is output.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The primary electrical quantity sequence, secondary protection event sequence, and switch state sequence are converted into event atoms carrying time uncertainty intervals, electrical direction attributes, and data sources, so that records formed by different sampling periods and transmission paths are no longer forcibly compressed into a single moment. A dynamic line topology is established based on switch state changes, and the connection relationships before switching, after switching, and pending confirmation are simultaneously retained within the state change interval. This allows the transition process between command issuance, contact changes, and position feedback to be incorporated into the topology analysis. The physical timing diagram connects primary events according to the direction of fault current, the order of voltage drops, and the continuity of fault quantities. The control timing diagram connects secondary events according to protection initiation, criterion confirmation, output commands, and position feedback. The two types of timing are then coupled using topology boundary migration markers. Interval constraint reasoning retains only candidate causal chains that simultaneously satisfy electrical propagation, protection logic, and topology migration conditions. This allows for the differentiation of record upload delays, protection action delays, and position feedback delays when switch switching and line faults overlap, avoiding the incorrect classification of events before and after switching into the same link. Based on this, the fault section, power supply direction, first-action protection, backup protection, and abnormal action links can be determined.

[0017] 2. The time uncertainty intervals in event atoms can be segmented and corrected using paired anchor points such as fault quantity mutations and protection activation, output records and branch current interruptions, and reclosing commands and voltage recovery. This allows the time difference between different data sources to propagate according to their respective continuous intervals, rather than using a uniform offset to process all records. The transition topology subgraph centers on the switch where the state change occurs, combining the pre-switching connection edges, post-switching connection edges, and connection edges to be confirmed into candidate topology slices. Incompatible connection combinations are eliminated through branch current continuity, node voltage correlation, control command and position feedback sequence, and fault quantity disappearance range. For events with sequence conflicts, the closed conflict subgraph can retain, disconnect, or reconnect the corresponding causal edges, and generate branch causal chains through interval satisfiability checks. For duplicate records and event gaps, composite event atoms and constrained placeholder event atoms can be formed respectively. The parallel recording of the common causal skeleton and divergent subchains ensures that the analysis results retain common facts, alternative paths, and corresponding conflict sources, avoiding the forced output of a unique conclusion when evidence is insufficient, and providing traceable data correlations for subsequent verification. Attached Figure Description

[0018] Figure 1 A flowchart of the overall data analysis method for integrated primary and secondary ring network box line protection; Figure 2 This is a contraction plot of the event's time interval. Figure 3 This is a map showing the distribution of physical correlation strength. Figure 4 The control relationships can satisfy a matrix diagram; Figure 5Flowchart for event atom generation and time uncertainty interval correction; Figure 6 Flowchart for topology boundary migration and bi-temporal cause-effect graph construction; Figure 7 This is a graph showing the cost of topological state consistency. Figure 8 This is a segmented time-scaled correction diagram; Figure 9 Conflict resolution branch cost graph; Figure 10 A ranking graph of candidate chain completeness. Detailed Implementation

[0019] In one embodiment, the line protection data analysis method operates in the protection data processing environment of a primary and secondary integrated ring network box. The input data includes primary electrical quantity records, secondary protection action records, switch control records, position feedback records, and topology records that can characterize the line connection relationship. The primary electrical quantity records include at least the time-series values ​​of current, voltage, and zero-sequence components. The secondary protection action records include at least the protection start-up, criterion confirmation, protection output, reclosing, and blocking status. The switch control records and position feedback records reflect the process of command generation and switch status confirmation on the control side, respectively. All types of data retain the original time stamp, the received time stamp, the source identifier, and the associated line identifier. During the data processing, it is not assumed that the clocks of different data sources are completely consistent, nor is the position feedback time directly identified as the actual switching time of the line topology. Instead, the fault electrical propagation process, the protection control process, and the topology change process are correlated and modeled.

[0020] In one embodiment, please refer to Figure 1 This paper provides a data analysis method for integrated primary and secondary ring network enclosure line protection, including: acquiring the primary electrical quantity sequence, secondary protection event sequence, and switch state sequence of the line; converting each sequence into event atoms carrying data source, time uncertainty interval, and electrical direction attributes; constructing a dynamic line topology map based on switch state changes, and generating topology boundary migration markers representing connection relationships before switching, connection relationships after switching, and connection relationships to be confirmed within the state change interval; constructing a physical timing diagram based on fault electrical propagation constraints, and constructing a control timing diagram based on the logical constraints between protection start-up, output commands, and position feedback; using the topology boundary migration markers as coupling nodes, merging the physical timing diagram and the control timing diagram into a dual-time-series causal graph; performing interval constraint reasoning on the dual-time-series causal graph to obtain candidate causal chains that satisfy electrical propagation constraints, protection logic constraints, and topology migration constraints; and determining the fault section, power supply direction, first-action protection, backup protection, and abnormal action links based on the candidate causal chains.

[0021] Specifically, the data acquisition process establishes a unified index based on line identifiers, ring network box identifiers, switch identifiers, and protection module identifiers. Data with the same line association relationship are written into the same event buffer, and the arrangement position of each record in the original data source is retained. For continuously sampled primary electrical quantities, ordered data segments are formed according to the sampling time scale. For discrete secondary protection events and switch status events, sorting results are established according to the original time scale and the received time scale, respectively. When the two sorting results are inconsistent, no record is deleted. Instead, a time scale conflict mark is written on the corresponding record. Subsequently, time scale conflicts are resolved through time uncertainty intervals and causal constraints to avoid fixing incorrect event sequence relationships in the data preprocessing stage.

[0022] As the basic node of the dual-time-series causal graph, the event atom's data content includes the event atom identifier, data source, associated line, associated switch or protection node, event category, original time stamp, received time stamp, time uncertainty interval, electrical direction attribute, topology stage attribute, and original data reference address. The event category is used to distinguish events such as current surge, voltage drop, zero-sequence component persistence, protection activation, criterion confirmation, protection output, switch control, position feedback, and fault quantity disappearance. The original data reference address is used to backtrack the corresponding sampling segment or discrete record after the causal analysis is completed. The topology stage attribute can be null when the event atom is initially generated and is written to the pre-switching, transition, or post-switching state after the topology boundary migration analysis.

[0023] For any event atom i, its time uncertainty interval is expressed as follows: ; in, This indicates the time interval that event atom i is allowed to occupy. This indicates the reference timescale selected from the original record. This represents the maximum amount by which an event can be shifted forward relative to a reference time scale. This represents the maximum amount that an event may shift backward relative to the reference time scale. The forward offset is limited by the sampling period boundary, the data buffer pre-time, and the local clock offset. The backward offset is limited by the sampling period boundary, the transmission wait time, and the reception processing delay. For example, if the reference time scale of a protection start record is 10, its forward offset is determined to be 1 based on the same source continuous records, and its backward offset is determined to be 3 based on the reception time scale difference, then the time uncertainty interval is [9,13]. Subsequent inference only requires that the relevant events have a satisfactory sequential position within the interval, without forcing the events to occur at the reference time scale of 10.

[0024] Please refer to Figure 2 , Figure 2The horizontal axis represents the position of events arranged according to a uniform time scale. The horizontal axis scale uses Chinese labels such as start event, preceding event, middle event, following event, and end event, which can be understood from left to right as the order in which events enter the analysis chain. The vertical axis represents the width of the time interval that an event can occupy, with the value in milliseconds. The solid line corresponds to the original envelope width formed by the sampling boundary, reception delay, and local clock offset, while the dashed line corresponds to the feasible width retained after pruning by paired anchor points, physical causal edges, and control sequence constraints.

[0025] from Figure 2 As can be seen, the original envelope width fluctuates between 12 ms and 70 ms, with local peaks formed by events such as switch position feedback, protection exit, and large upload / receive delays. The average retained width after constrained inference is approximately 28.125 ms, lower than the original average width of 31.372 ms, indicating that some events have been effectively tightened; while events that still retain a wider range correspond to records with time-scale conflicts, sampling gaps, or topology transitions that have not yet been fully confirmed.

[0026] Figure 2 Intuitively, the time uncertainty interval is not a simple enlargement of the original time scale, but rather a gradual contraction based on subsequent satisfiability checks, while allowing for forward and backward offsets. Figure 2 The distance between the midline segments reflects the intensity of interval pruning and is also the basic criterion for whether candidate causal chains can continue to connect.

[0027] The dynamic circuit topology diagram uses nodes and connecting edges to represent the circuit connection status. Nodes correspond to bus connection points, line segment endpoints, protection nodes, and connection points on both sides of the switch. Connecting edges correspond to the circuit connection relationships that can transmit electrical influence under a given switch state. For switches whose state has not changed, stable connecting edges are determined based on multiple consecutive state records and the continuity of the corresponding branch current. For switches whose state has changed, control commands, position feedback, branch current changes, and adjacent node voltage changes are extracted simultaneously to form a state change interval. Within the state change interval, connecting edges before switching, connecting edges after switching, and connecting edges whose connectivity cannot yet be determined by existing evidence are retained.

[0028] The topology boundary migration marker is not a single-moment marker, but a composite data object attached to the state change interval and its associated switch. The composite data object includes the topology subgraph index before switching, the transition topology subgraph index, the post-switching topology subgraph index, the control command time interval, the position feedback time interval, the branch current continuous change interval, and the node mapping relationship. The node mapping relationship is used to indicate the corresponding node of the same physical connection point in different topology subgraphs. Therefore, fault events that occur during switch switching can be projected into the topology states before switching, transition, and post-switching respectively, and the topology stage to which the event belongs can be determined through subsequent constraints.

[0029] The physical time sequence diagram uses event atoms formed by a single electrical quantity as nodes. Directed physical causal edges are established according to the line connection relationship and the electrical propagation characteristics of the fault. When a sudden current change propagates from the upstream node to the downstream node along the actual power supply direction, a causal edge with the same direction is established. When a voltage drop occurs between electrical adjacent nodes, overlapping or sequential constraints are established based on the time uncertainty interval. For zero-sequence component continuous events, continuous edges are established based on the polarity, duration interval, and changes before and after fault clearing. For fault quantity disappearance events, termination edges are established based on the influence range after the switch state change. Candidate edges that do not meet the current topology connection relationship are not deleted temporarily, but are given corresponding topology stage conditions.

[0030] The physical correlation between any two primary event atoms a and b is calculated according to the following formula: ; in, This represents the physical correlation quantity that event atom a points to event atom b. This indicates whether two events satisfy the line topology adjacency relationship; a value of 1 indicates that the relationship is satisfied, and a value of 0 indicates that the relationship is not satisfied. This indicates whether the current direction is consistent with the candidate propagation direction. This indicates whether the voltage variation range has an allowable propagation order. This indicates whether the continuity of zero-sequence components or fault quantities is compatible. to For association weights that sum to 1, for example, take , , , When all four relations are satisfied When the line topologies are not adjacent and the current directions are inconsistent, even if the voltage change intervals overlap, the obtained physical correlation quantities are insufficient to establish a physical causal edge independently.

[0031] Please refer to Figure 3 , Figure 3 The horizontal axis represents candidate electrical connections sorted by overall correlation strength. The horizontal axis scale uses Chinese labels such as highest correlation connection, high correlation connection, stable connection, critical connection, and low correlation connection. The vertical axis represents the normalized physical correlation strength, with values ​​ranging from 0 to 1. Solid lines represent overall correlation strength, and horizontal dashed lines represent the retention threshold for candidate physical edges to enter the main propagation path. The legend is uniformly placed above the coordinate graph and does not cover any curve segments.

[0032] from Figure 3It is evident that the overall correlation strength of the candidate edges in the first segment remained around 1.0 for a long time, indicating that these edges simultaneously met the requirements of topological adjacency, consistent current direction, compatible voltage change sequence, and continuity of fault quantities, and could be directly used as the backbone edges of the electrical propagation path of the fault. As the electrical connection order shifted to the later stages, the curve showed a step-like decline, reaching a minimum of approximately 0.59, which is below the retention threshold of 0.68, indicating that although such candidate edges may have local voltage responses, they lack sufficient topological or directional support.

[0033] Figure 3 Corresponding to the calculation logic of physical related quantities: Figure 3 The above only displays the overall results and retention threshold, while the contribution weights of each component are retained in the exported data to trace the specific reasons why a physical edge is retained or excluded. This representation shows that physical causal edges are not determined by a single measurement, but are derived from multiple electrical propagation conditions.

[0034] The control timing diagram uses protection start-up, criterion confirmation, protection output, switch command, position feedback, reclosing, and blocking event atoms as nodes. The control causal edges are established based on the action state transition relationship within the same protection node and the coordination relationship between different protection nodes. For the same action chain, protection start-up must have a satisfyable time position before the corresponding output. There is an interval formed by equipment action and data transmission between the output command and the position feedback. The blocking event and the output event of the blocked protection are constrained according to the mutual exclusion relationship. The reclosing event and the preceding tripping, subsequent voltage recovery, or re-fault event are connected according to the state transition sequence. Records that cannot satisfy the control sequence are retained as conflict edges and enter the subsequent resolution process.

[0035] The control edge between any two quadratic event atoms c and d can satisfy the following formula: ; in, This indicates whether the control edge from event atom c to event atom d is satisfiable. This indicates whether two events belong to the same protection action chain or have a preset primary / backup relationship; a value of 1 indicates a match, and a value of 0 indicates otherwise. This indicates whether there exists a combination of moments within the uncertain time interval of two events that satisfies the control sequence relationship; a value of 1 indicates existence, and a value of 0 indicates otherwise. This indicates whether the states of two events satisfy the allowed transition from start to exit, exit to command, or command to feedback. A value of 1 indicates a satisfying transition, and 0 indicates otherwise. For example, if the protection start and exit records belong to the same protection node and there are satisfying preceding and following positions within the time interval, then if the state transition relationship is simultaneously satisfied, then... If any condition is not met The corresponding record enters the conflict set instead of directly forming a control causal edge.

[0036] Please refer to Figure 4 , Figure 4 The horizontal axis represents the protection control relationship categories and their summary items. The horizontal axis scale is labeled in Chinese with relationships summary, start-up criterion relationship, exit command relationship, feedback relationship and detailed segment. The vertical axis represents the number of establishments or detailed judgment values ​​under the corresponding control conditions. Figure 4 The multiple groups of bars correspond to the statistical results of action chain attribution, temporal sequence, state transition, and simultaneous occurrence of all three. The details in the latter part are used to show the satisfiability of each control edge.

[0037] from Figure 4 As can be seen, the number of occurrences of both action chain attribution and state transition is 18, indicating that most protection start-up, output, switch commands, and position feedback can be found in the semantics of action. The number of occurrences of timing sequence and simultaneous occurrence of all three is 12, which is lower than the first two, indicating that although some records belong to the correct action chain, they cannot form a value combination that satisfies the sequence relationship of protection control within the time uncertainty interval.

[0038] Figure 4 The engineering significance lies in displaying the three types of criteria for control edges in a hierarchical manner, which makes it easier to distinguish between two different problems: records belonging to the same action chain and action chains that can truly satisfy two different problems in terms of timing. Figure 4 The bar value where all three terms are true is the key input for controlling the entry of causal edges into the bi-temporal causal graph. Detailed edges that are not true at the same time are entered into the conflict set for subsequent resolution.

[0039] The dual-time causal graph connects the physical timing graph and the control timing graph through topology boundary migration markers. Specifically, protection output events, switch command events, and position feedback events are paired with the current interruption, voltage change, and fault disappearance events on both sides of the corresponding switch. If the time uncertainty intervals of the paired events have an allowed overlap relationship, and the events have a consistent line connection meaning in at least one candidate topology stage, then a coupled causal edge is established, and the applicable topology stage set is written into the coupled causal edge. If the same pair of events is only compatible in the transition topology subgraph, then the corresponding coupled causal edge is limited to a transition stage edge and cannot participate in causal chain generation in the topology before or after the switch.

[0040] Interval constraint reasoning transforms each physical causal edge, control causal edge, and coupled causal edge into relational conditions between event time intervals. These relational conditions include the existence of a combination of values ​​where the preceding event is earlier than the subsequent event, the two events being allowed to overlap in certain intervals, the two events not being allowed to belong to the same topological stage simultaneously, and the events being located before or after a certain topological migration path. The reasoning process uses constraint propagation to update the feasible time intervals of each event atom. When an edge causes a time interval to be empty, the edge and its associated topological stage conditions are written into the conflict set, and candidate disconnection, reconnection, or topological stage transformation schemes are formed from the conflict set.

[0041] Candidate causal chains start from current surges, voltage drops, or persistent zero-sequence components that characterize the initiation of line disturbances. They trace the propagation of fault effects along physical causal edges, then enter the control chain formed by protection activation, output, switching commands, and position feedback via coupled causal edges. After the switching state changes, they return to the physical timing diagram to trace the range where the fault quantity disappears. Each candidate causal chain saves the event atoms, causal edges, topological stages, interval value conditions, and conflict resolution records it has passed through. Candidate chains that lack protection actions but have complete electrical propagation chains can be used to represent the situation where protection does not output. Candidate chains that have protection outputs but lack corresponding electrical propagation support are retained as abnormal action candidates.

[0042] The fault section is determined by the topology node containing the earliest event atom in the candidate causal chain that can simultaneously explain the current direction, voltage drop, and protection activation, as well as the boundary node where the fault quantity stops propagating after the switch action. The power supply direction is determined by the directed connection relationship of the physical causal edge and the current direction attribute. The first-action protection is determined by the protection node that forms the earliest complete activation to exit chain in the candidate causal chain. The backup protection is determined by the exit event located after the first-action protection and whose protection range covers the same fault propagation path. The abnormal action link is determined by the break position between the physical causal chain and the control causal chain, and is divided into exit records lacking electrical quantity support, switch response anomalies with exits but no position feedback, state record anomalies with position feedback but incompatible topology migration, and protection action gaps where the fault quantity persists but no exit chain has been formed.

[0043] In this embodiment, the analysis results are saved in the form of causal chain records. Each record includes the fault section, power supply direction, first-action protection, backup protection, abnormal action link, common event set, branch event set, and original data reference. When outputting the analysis results, the system simultaneously outputs the time interval values ​​and topology migration paths used by the candidate causal chains, so that the review process can return from the analysis conclusion to the specific electrical quantity segment, protection record, and switch status record, avoiding the output of a single fault label that cannot trace the formation process.

[0044] In one embodiment, please refer to Figure 5The generation of event atoms includes: dividing primary event segments based on the start point of abrupt changes, direction reversal points, voltage drop start points, zero-sequence component duration boundaries, and fault disappearance points in the primary electrical quantity sequence, and incorporating the sampling clock error and upload delay interval into the corresponding time uncertainty interval; forming secondary event segments based on the start record, criterion confirmation record, output record, opening and closing command, position feedback, and interlocking record in the secondary protection event sequence, and determining the corresponding time uncertainty interval based on the local and received timestamps of the same record; generating electrical direction attributes based on the current direction relationship, phase sequence relationship, and zero-sequence polarity of adjacent primary event segments; and encapsulating the data source identifier, event category, time uncertainty interval, electrical direction attribute, and associated switch identifier into event atoms.

[0045] The division of an event segment is based on the characteristic boundaries in the continuous sampling sequence. The starting point of the mutation is determined by comparing the mean difference and the direction of change between the current sampling window and the previous stable window. The direction reversal point is determined by the sign change of the current vector in the adjacent window. The starting point of the voltage drop is determined by the first boundary of the continuous decreasing interval. The boundary of the continuous segment of the zero-sequence component is determined by the set of sampling points that continuously satisfy the same direction of change. The point where the fault quantity disappears is determined by the first stable interval where the current, voltage and zero-sequence component jointly return to the non-fault evolution state. The segment obtained by division retains the original sampling values ​​before and after the boundary so as to determine whether the boundary is formed by sampling jitter, communication frame loss or real fault change.

[0046] Secondary event fragments are merged according to the protection node and action chain identifier. The start record and the criterion confirmation record establish candidate before-after relationships. The exit record and the opening and closing command are associated according to the mapping relationship from the protection node to the switch object. The position feedback forms candidate pairings according to the switch identifier and the corresponding command. The blocking record is associated with both the blocked protection and the blocking release record. If multiple events of the same type occur in the same action chain, each record is retained and multiple candidate event atoms are formed according to the local time stamp, the received time stamp and the before-after event relationship. Duplicate events are not directly deleted during the encapsulation stage.

[0047] The electrical direction attribute is formed by the phase relationship between adjacent event segments, the vector change of current, and the zero-sequence polarity. Specifically, it can be represented as positive, negative, direction undetermined, or direction conflict. Positive means that the event change is consistent with the current candidate power supply direction, negative means that it is opposite to the current candidate power supply direction, direction undetermined means that the sampled segment is insufficient to determine the propagation direction, and direction conflict means that the phase relationship and the zero-sequence polarity point in different directions. Event atoms with direction undetermined and direction conflict can enter the physical time series graph, but the corresponding physical causal edge is subject to additional directional branch conditions, which are subsequently constrained by the topological connection relationship and adjacent node events.

[0048] In a preferred embodiment, the event atom also stores fragment integrity and source continuity. Fragment integrity is determined based on whether the original samples before and after the event boundary are continuous, and source continuity is determined based on whether there are timescale jumps or sequence number gaps in adjacent records from the same data source. Fragment integrity and source continuity do not directly replace the event content, but are used to determine whether to prioritize retaining the original continuous records or use the associated records from other data sources during conflict resolution. When a primary event fragment has a sampling gap but the secondary protection action chain is complete, the gap position still enters the dual-time-series causal graph in the form of a time-determined interval.

[0049] In one embodiment, please refer to Figure 6 The generation of topology boundary migration markers includes: extracting the control command time, position feedback time, and corresponding branch current continuity change time for switches that undergo state changes, and converting the three types of time into their respective time intervals; forming the termination candidate boundary of the connection relationship before switching with the starting point of the control command interval, forming the starting candidate boundary of the connection relationship after switching with the ending point of the position feedback interval, and applying cross constraints to the two types of candidate boundaries with the interval where the branch current changes from continuous to interrupted or from interrupted to continuous; simultaneously retaining the topology subgraph before switching, the topology subgraph after switching, and the transition topology subgraph containing the connection edge to be confirmed within the cross constraint interval, and establishing the node mapping relationship of the same electrical node among the three topology subgraphs.

[0050] When the control command time is converted to the control command time interval, the interval boundary is determined based on the command generation record, local transmission record, and received confirmation record. When the position feedback time is converted to the position feedback time interval, the interval boundary is determined based on the first change record of switch state, continuous state confirmation record, and uploaded and received record. When the branch current continuous change time is converted to the electrical change time interval, the interval range is determined based on the sampling segment boundary where the current changes from continuous to interrupted or from interrupted to continuous. The three types of intervals do not require the same start and end times, but rather form topology migration constraints by whether there is compatible overlap and allowed sequence relationship.

[0051] The termination candidate boundary of the connection relationship before switching is used to limit electrical events that occur before the boundary to be directly projected onto the topology subgraph before switching. The starting candidate boundary of the connection relationship after switching is used to limit electrical events that occur after the boundary to be directly projected onto the topology subgraph after switching. Events located within the intersection range of the two types of candidate boundaries simultaneously enter the transition topology subgraph. If a control command has been issued but the branch current remains continuous and the position feedback has not yet been confirmed, the corresponding connection edge of the switch is marked as a connection edge to be confirmed in the transition topology subgraph. If the branch current has been interrupted but the position feedback remains in its original state, the connection edge to be confirmed is subject to an electrical disconnection condition.

[0052] The consensus cost of candidate topology state h is calculated according to the following formula: ; in, This indicates the degree of inconsistency between the candidate topological state h and the current event set. This indicates the number of conflicts between the continuity of branch currents and the candidate connection edges. This represents the number of conflicts between the voltage variation relationships between adjacent nodes and the candidate topology adjacency relationships. This indicates the number of conflicts between the control command and position feedback sequence and the candidate switch states. This indicates the number of conflicts between the range where the fault disappears and the candidate disconnected edges. to For cost weights that sum to 1, for example, take , , , When a candidate topology state causes two current continuity conflicts and one control sequence conflict at the same time, its consistency cost is higher than that of a candidate state with only one position feedback conflict. The latter will enter the topology migration path first but will still retain the conflict mark.

[0053] Please refer to Figure 7 , Figure 7 The horizontal axis represents the category of topology judgment schemes to be compared. The scale of the horizontal axis uses Chinese labels such as transition scheme, scheme before switching, scheme after switching, retained scheme, and verification scheme. The vertical axis represents the topology consistency cost. The larger the value, the more conflicts there are between the candidate state and the primary electrical quantity and secondary control record. Figure 7 The middle column represents the total cost of transition, the total cost before switching, the total cost after switching, as well as sub-items such as branch current conflict, node voltage conflict, control sequence conflict, and fault range conflict.

[0054] from Figure 7 As can be seen, the total cost of the transition topology states all reached 1.0, significantly higher than the approximately 0.25 of the pre-switching state and the 0 to 0.2 range of the post-switching state. This indicates that during the switching process, unconfirmed connection edges are prone to simultaneously triggering three types of conflicts: branch current, adjacent node voltage, and control sequence. The post-switching candidate state has the lowest total cost, indicating that it is more compatible with the fault disappearance boundary and location feedback records.

[0055] Figure 7 The necessity of topological boundary migration markers is further explained: relying solely on feedback moments from a single location can easily lead to misclassifying high-cost transitional states as stable connections. By displaying each conflict quantity separately, the reasons for retaining, excluding, or transferring candidate topological slices to manual review can be traced, and quantitative evidence can be provided for subsequent candidate causal chains crossing topological stages.

[0056] The node mapping relationship is established according to the stable identifier of the electrical node. The same line endpoint, bus connection point and protection node use a unified basic identifier in the topology sub-diagram before switching, transition and after switching, and a topology stage suffix is ​​added. When branch merging or line transfer occurs, the relationship between one pre-switching node and multiple transition nodes or multiple transition nodes and one post-switching node is recorded through the mapping table. When the causal chain crosses the topology stage, it must be transferred along the node mapping relationship. It cannot directly cross from a node without a mapping relationship to another topology stage.

[0057] In one embodiment, the generation of a dual-time-series causal graph includes: in the physical time-series graph, connecting event atoms that are adjacent in line topology and conform to the fault current propagation direction, voltage drop sequence, and fault quantity continuity as physical causal edges; in the control time-series graph, connecting start events, criterion confirmation events, exit events, switch command events, and position feedback events belonging to the same protection action chain as control causal edges; adding topology stage identifiers to cross-graph event atoms based on the pre-switching topology subgraph, transition topology subgraph, and post-switching topology subgraph corresponding to the topology boundary migration marker; and establishing coupled causal edges connecting the physical time-series graph and the control time-series graph only when there is an overlapping relationship that satisfies the topology stage identifier and the topology stage identifier is compatible within the uncertain interval of event time.

[0058] When establishing physical causal edges, a combination of local topological adjacency and global propagation direction is used. Local topological adjacency is used to limit whether there is a candidate electrical connection between the nodes corresponding to two primary event atoms. Global propagation direction is used to determine whether the extension from the candidate fault starting point is compatible with the current direction of each branch. When local topological adjacency is adjacent but global propagation direction conflicts, the physical causal edge enters the directional branch set. When the voltage drop sequence and fault quantity continuity simultaneously support a certain propagation direction, the directional branch set is reduced to a single directional edge.

[0059] When establishing the causal edge control, the internal action chain of the protection node and the coordination chain between protection nodes are processed separately. The internal action chain of the node requires that there is a satisfyable sequence between the start-up, criterion confirmation, output, switch command and position feedback. The coordination chain between protection nodes determines the primary protection candidate and backup protection candidate based on the protection range coverage relationship and the event time interval. When the start intervals of multiple protection nodes overlap, the order is not directly determined by the reference time scale, but the satisfyable action chain sequence is determined after the respective output, switch feedback and fault quantity disappearance event are added.

[0060] Coupled causal edges include at least electrical triggering edges, action clearing edges, and state verification edges. Electrical triggering edges connect current surges, voltage drops, or zero-sequence component persistence events with protection start events. Action clearing edges connect protection output or switch command events with branch current interruption or fault quantity disappearance events. State verification edges connect position feedback events with corresponding topology boundary migration markers. Coupled causal edges all store the applicable topology stage set and time interval relationship. Any coupled causal edge can enter the candidate causal chain only if it is true under at least one topology stage and at least one combination of time values.

[0061] In a preferred embodiment, the dual temporal causal graph is stored using a hierarchical adjacency list. Event atoms are stored in partitions according to the physical layer, control layer, and topology migration layer. Physical causal edges, control causal edges, and coupled causal edges are identified using different edge types. Interval constraint inference is recalculated only in the local subgraph affected by new events or new topology states. Unaffected common causal skeletons retain their original results, avoiding the need to reconstruct all event relationships each time supplementary records are received.

[0062] In one embodiment, determining the time uncertainty interval includes: extracting paired anchor points that can characterize the same line disturbance from primary and secondary event segments, where the paired anchor points include fault quantity abrupt change and protection start-up, output record and branch current interruption, reclosing command and voltage recovery; forming segmented time difference constraints based on the time scale difference of each paired anchor point in different data sources, and propagating the segmented time difference constraints to adjacent event atoms according to the continuous time scale interval of the data source; when adjacent segmented time difference constraints overlap, taking their intersection to correct the corresponding time uncertainty interval, and when there is no intersection, retaining the branch time interval and recording the source of conflict; writing the corrected time uncertainty interval back to the event atom, and maintaining the correspondence between the original time scale and the corrected interval.

[0063] The extraction of paired anchor points does not require that the two types of events have exactly the same recording time, but requires that they have a stable correspondence in the physical process or control process. The sudden change of fault quantity and the start of protection form a trigger anchor point, the output record and the branch current interruption form a cut-off anchor point, and the reclosing command and voltage recovery form a recovery anchor point. Each anchor point saves the original time difference, the received time difference and the allowed sequential direction of the two data sources. When the same disturbance forms multiple anchor points of the same type, candidate pairs are established according to the line identifier, protection node and switch identifier respectively.

[0064] The segmented time difference correction for data source v within a continuous interval m is determined by the following formula: ; in, This represents the time difference correction of data source v relative to the reference data source within a continuous interval m. This indicates taking the median of the set. to This represents the time scale difference obtained by K pairs of anchor points within a continuous interval. For example, the time scale differences corresponding to 3 pairs of anchor points are 2, 3 and 8 respectively, so the median is 3. The larger deviation of 8 does not directly determine the correction amount for the entire interval, but is retained as a source of local conflict. The time uncertainty interval of adjacent event atoms within the continuous interval is shifted with the correction amount of 3 as the center and re-intersected by combining with the original offset boundary.

[0065] Please refer to Figure 8 , Figure 8 The horizontal axis represents a continuous data period formed by the same data source. The horizontal axis scale uses Chinese labels such as first segment data, preceding segment data, middle segment data, subsequent segment data, and last segment data. The vertical axis represents the time scale correction relative to the reference data source, in milliseconds. Figure 8 The solid line represents the recommended correction amount for the continuous interval, and the discrete point represents the single time difference calculated from paired anchor points such as fault quantity change and protection start-up, output record and branch current interruption.

[0066] from Figure 8 As can be seen, the recommended correction amount ranges from 8 milliseconds to 58 milliseconds, with an average of approximately 20.167 milliseconds, reflecting inconsistent uplink delays among different primary sampling sources, secondary protection sources, and switch status sources. The anchor point time difference in most intervals largely coincides with the recommended correction amount, indicating that the time scale difference within the interval is stable; locally higher correction amounts correspond to sudden changes in received order or buffered data segments with longer waiting times.

[0067] Figure 8 This reflects the engineering reason for using the median instead of the arithmetic mean for segmented time difference corrections. When individual anchor points deviate significantly due to communication delays or recording gaps, the median ensures that the correction result for the entire continuous interval is not skewed by single-point anomalies. Therefore, Figure 8 The recommended correction line serves as the direct basis for subsequent time interval shifting, intersection calculations, and branch generation.

[0068] Segmented time difference constraint propagation proceeds according to the continuity of records from the same data source. If there are no sequence number gaps, time scale backs, or sudden changes in the receiving order within a continuous interval, the correction amount obtained from the anchor point can be propagated to adjacent event atoms. When the propagation distance crosses a new anchor point, the new anchor point re-constrains the data. If a time scale back or a long-term gap occurs in the record, a new continuous interval is segmented at the abnormal position, and the time difference correction amount of the previous interval does not continue to propagate, thus avoiding the spread of local clock jumps to all subsequent events.

[0069] When the correction intervals of adjacent segment time difference constraints intersect, the time uncertainty interval of the event atom is updated with the intersection. When there is no intersection, multiple time branches are generated for the same event atom. Each branch records the source anchor point, applicable continuous interval, and conflict anchor point. Multiple time branches share the same original data reference in the dual-time-series causal graph, but have different interval identifiers. Subsequent candidate causal chains can only select the same time branch to participate in the complete link and cannot use conflicting time correction results simultaneously in a single causal chain.

[0070] In one embodiment, the construction of the transition topology subgraph includes: extracting a local topology region centered on the switch where the state change occurs, containing upstream protection nodes, downstream protection nodes, and adjacent interconnection nodes; writing the pre-switching connection edge, post-switching connection edge, and connection edge to be confirmed into the local topology region to generate candidate topology slices corresponding to different connection combinations; performing consistency verification on each candidate topology slice according to the continuity of branch current, the correlation of node voltage, the order of control commands and position feedback, and the range of fault disappearance, and deleting candidate topology slices that cannot simultaneously accommodate primary event segments and secondary event segments; and connecting the remaining candidate topology slices to the pre-switching topology subgraph and the post-switching topology subgraph according to the node mapping relationship to form a transition topology subgraph with a topology migration path.

[0071] The boundary of the local topology region is extended upstream and downstream according to the switch-associated branch where the state change occurs, until it can cover the nearest protection node, adjacent sectionalizing switch and tie node. If the fault quantity propagation has crossed the local region, it continues to extend along the physical causal edge to the fault quantity disappearance boundary. Remote nodes that did not participate in this state change and do not have fault electrical quantity propagation remain in the global topology graph and do not enter the candidate topology slice combination process.

[0072] Candidate topology slices are formed by assigning connected, disconnected, or partially connected states to the edges to be confirmed. A partially connected state indicates that the connection relationship is only established for a portion of the transition period. Each candidate topology slice also saves its applicable time interval. Candidate slices that disconnect the corresponding edge throughout the entire process are excluded when the branch current remains continuous. Candidate slices that lack any electrical connection path are excluded when the node voltage is still strongly correlated. Candidate slices that have entered the post-switching state prematurely are excluded when the control command has not yet appeared. Candidate slices that still maintain the fault propagation path connection are excluded when the fault quantity has disappeared.

[0073] Consistency verification uses item-by-item constraint propagation instead of single-point scoring. Branch current continuity is used to check whether candidate connection edges allow current propagation. Node voltage correlation is used to check whether candidate nodes are in the same electrical influence range. Control command and position feedback sequence is used to limit the state transition direction. Fault quantity disappearance range is used to determine the location where the disconnected edge should be on the fault propagation path. Any candidate topology slice is deleted from the transition topology subgraph if it causes an event atom to have no mapping node, causes the same switch to be in a mutually exclusive state at the same time, or causes the fault propagation path to cross a confirmed disconnected connection edge.

[0074] When the remaining candidate topology slices are connected according to the node mapping relationship, the stable nodes in the topology subgraph before the switch are connected to the corresponding nodes in the transition topology slice, and the nodes in the transition topology slice are then connected to the corresponding nodes in the topology subgraph after the switch. The connecting edge records the allowed time relationship and triggering event of the topology stage transition. When the candidate causal chain crosses the connecting edge, it must contain at least one of the following: corresponding control command, position feedback or branch current change evidence, and satisfy the time uncertainty interval constraint of the corresponding event.

[0075] In one embodiment, interval constraint reasoning includes: converting each causal edge into a corresponding constraint type among the following: the endpoint of the preceding interval is no later than the starting point of the subsequent interval; intervals are allowed to overlap; or intervals are mutually exclusive; and grouping causal edges that share event atoms but have incompatible constraint types into conflict sets; extracting closed conflict subgraphs from each conflict set that simultaneously involve physical causal edges, control causal edges, and topology migration constraints; using constraints such as protecting the integrity of the action chain, the continuity of fault electrical propagation, and unidirectional migration of topology stages as constraints, generating candidate resolution branches that retain, disconnect, or reconnect causal edges in the closed conflict subgraph; re-performing the interval satisfiability test on each candidate resolution branch, and determining the branches that do not have temporal loops and whose event atoms all have reachable predecessors or reachable successors as candidate causal chains.

[0076] The preceding interval endpoint is no later than the following interval start point to indicate actions with a definite sequential relationship, such as protection start-up to exit. Interval overlap is allowed to indicate electrical changes and control events that may occur simultaneously under sampling and recording delays, such as fault quantity mutation and protection start-up. Interval mutual exclusion is used to indicate events that cannot occupy the same topology stage at the same time, such as the confirmed disconnected branch current continuous event and the topology state after switching. After causal edge transformation, a constraint table is formed, which stores the preceding event, following event, relationship type, topology stage, and source evidence.

[0077] The closed conflict subgraph expands forward and backward along the shared event atoms from the conflict set until it contains the minimum closed range that can form a complete physical propagation path, control action chain, and topology migration path. For example, if the position feedback event is recorded before the branch current interruption, but there is still a compatibility relationship between the control command, protection exit, and current interruption, the closed conflict subgraph must contain at least the protection exit, control command, position feedback, current interruption, corresponding switch, and its topology boundary migration marker, in order to determine whether the conflict source is a position feedback timescale deviation or a topology stage assignment error.

[0078] The constraint cost of candidate resolution branch f is determined by the following formula: ; in, This represents the constraint cost of the candidate resolution branch f. This represents the set of causal edges that are broken or reconnected in the branch. Indicates causal edge The source consistency cost corresponding to the change. This represents the set of event atoms that are being adjusted in a time branch or topology phase. Represents event atoms The record continuity cost of branch adjustments. This indicates whether a branch introduces a new timing loop or topology reversal. It takes 0 when no new loop is introduced and an unacceptable positive value when it is introduced. For example, if a branch only adjusts one position feedback edge and does not change the physical propagation chain, its cost is lower than a branch that simultaneously disconnects the current propagation edge and the protection exit edge. The interval satisfiability test is performed on the branch with the lower cost, but other feasible branches are not omitted.

[0079] Please refer to Figure 9 , Figure 9 The horizontal axis represents the conflict resolution scheme, and the scale of the horizontal axis is labeled in Chinese as low cost scheme, relatively low cost scheme, medium cost scheme, relatively high cost scheme and high cost scheme. The vertical axis represents the constraint cost. The lower the cost, the less the branch disturbs the original causal edge, time branch and topology migration path. Figure 9 The central column displays the cost by adjusting the edge, the cost of the time branch, the penalty term, and the total cost of the branch, which is used to decompose an overall score into verifiable source components.

[0080] from Figure 9 As can be seen, the edge adjustment cost of each candidate branch is approximately 0.12, the time branch cost is approximately 0.05, the penalty term is 0, and the total branch cost is approximately 0.17. This numerical distribution indicates that none of the candidate branches in the current closed conflict subgraph introduce new temporal loops or topological reverse migrations; the main differences come from minor adjustments to position feedback edges, event time branches, or local control edges.

[0081] Figure 9 Its role is not to solely identify the final fault segment, but rather to quantitatively rank conflict resolution strategies before candidate causal chains are generated. When the total branch costs are similar, the system also needs to further filter based on the completeness of the link, the continuity of the original data, and the unidirectional migration results of the topology phase. Figure 9 Visualizing each cost component can explain why a certain resolution branch is prioritized for testing, and also facilitates a post-event review of the conflict record processing.

[0082] The integrity of the protection action chain requires an interpretable state transition from protection initiation to exit or blockage. The continuity of fault electrical propagation requires a coherent physical path from the fault initiation candidate to the fault quantity disappearance boundary. The unidirectional migration of the topology stage requires that the causal chain can only enter the transition state from the pre-switching state and then enter the post-switching state, and cannot return from the post-switching state to the pre-switching state. If a candidate branch violates any constraint, it will not enter the candidate causal chain set. If the constraint is satisfied but there are still unresolved records, the unresolved records will be saved as a branch sub-chain.

[0083] The interval satisfiability test is completed by iteratively shrinking the allowed time intervals of event atoms. In each round, the relevant intervals are updated according to the preceding, overlapping and mutually exclusive constraints until all intervals no longer change or a certain interval becomes empty. When all intervals retain selectable values ​​and there are no closed loops in the graph consisting of strict sequential relationships, the candidate branch passes the test. If an event atom has only a predecessor and no successor but belongs to the event of the disappearance of fault quantity or the termination of action chain, it is considered to have a valid endpoint. If an event atom has only a successor and no predecessor but belongs to the candidate of the fault starting point, it is considered to have a valid starting point.

[0084] In one embodiment, the screening of candidate causal chains includes: determining the earliest event atom in each candidate causal chain that can simultaneously connect current direction change, voltage drop, and protection activation events, and using the topology node where the earliest event atom is located as a candidate fault start point; tracing the propagation path of the fault quantity in each topology stage along the candidate causal chain, and recording the protection section boundary crossed by the propagation path, the protection node corresponding to the first exit event, the protection node corresponding to subsequent exit events, and the fault quantity disappearance boundary after the switch state change; determining the fault section based on the minimum connected section between the candidate fault start point and the fault quantity disappearance boundary, determining the power supply direction based on the directed connection relationship of the propagation path, and marking the first-action protection, backup protection, and abnormal action links based on the continuous relationship between the exit event and the corresponding position feedback on the control causal edge.

[0085] The earliest event atom is determined by a combination of time lower bound and correlation integrity. In the same candidate causal chain, nodes that can be connected to current direction change and voltage drop events through physical causal edges and protection start events through coupling causal edges are given priority as fault start point candidates. If the time uncertainty intervals of multiple nodes overlap and all meet the above conditions, multiple fault start point candidates are retained and traced along the corresponding physical path to the fault quantity disappearance boundary.

[0086] When the fault propagation path crosses the boundary of the protected area, the protection nodes on both sides of the boundary, the corresponding topological stage, and the event atoms that pass through the boundary are recorded. The first exit event is determined according to the time order that can be satisfied within the candidate causal chain, and the smallest hour scale of the original record is not directly used. Subsequent exit events are marked as backup protection candidates only if their protection range covers the same fault propagation path and is located after the first exit event. If the subsequent exits correspond to different fault propagation branches, they form parallel action chains and are not classified into the same primary backup relationship.

[0087] The link completeness of candidate causal chain j is expressed by the following formula: ; in, This represents the link completeness of candidate causal chain j. Indicates the completeness of the electrical propagation path of a fault, with a value ranging from 0 to 1. This indicates the completeness of the protection start-up to switch feedback control chain, with a value ranging from 0 to 1. This indicates the completeness of the migration path in the topology phase, with a value ranging from 0 to 1. This represents the amount of unresolved conflicts, ranging from 0 to 1. For example, it represents the number of unresolved conflicts when the physical path, control chain, and topology migration path of a candidate causal chain are all complete and there are no unresolved conflicts. If the control chain lacks position feedback, If there is a conflict occupancy of 0.1, the link integrity will be reduced accordingly. The analysis results will retain the missing positions and conflict sources.

[0088] Please refer to Figure 10 , Figure 10 The horizontal axis represents the candidate analysis chains sorted by backend completeness. The horizontal axis scale uses Chinese labels such as preferred analysis chain, priority analysis chain, alternative analysis chain, analysis chain to be reviewed, and last analysis chain. The vertical axis represents the completeness of the candidate chain and its sub-judgment values, with values ​​ranging from 0 to 1. Figure 10 The solid line represents the integrity of the candidate chain, and the other lines represent the electrical propagation path, protection and control chain, topology migration path, and unresolved occupancy items, respectively. The legend above uniformly labels all the lines to facilitate comparison of changes in each item.

[0089] from Figure 10 As can be seen, the component values ​​of the electrical propagation path vary between 0.75 and 1.0, indicating that most candidate chains can explain the propagation process of fault quantities in line connections. The components of the protection control chain and topology migration path are close to 0 in the current figure, while the unresolved component occupies approximately 0.45, resulting in a low completeness of the synthesized candidate chains. This suggests that there are still gaps in control records or topology stage evidence in the current candidate chains, and it is inappropriate to draw a single conclusion based solely on the physical propagation path.

[0090] Figure 10 Corresponding to the calculation logic of the complete link quantity, its engineering value lies in comparing fault electrical propagation, protection and control response, topology phase migration, and unresolved conflicts on the same coordinate system. When a candidate link has a high electrical component but lacks a control or topology component, Figure 10 The system will display obvious sub-fractures, guiding it to output abnormal action links or branch analysis records, instead of outputting untraceable fault labels.

[0091] The fault section is the smallest connected region between the candidate fault starting point and the boundary where the fault disappears, which can cover all fault propagation events. When the candidate path crosses the topology migration, the smallest connected region is spliced ​​according to the node mapping relationship of each topology stage. The power supply direction is the dominant directional direction of the physical propagation path. If the power supply direction changes in different topology stages, the direction before switching and the direction after switching are recorded respectively. Abnormal action links are determined by finding the position where the link integrity decreases, including the lack of electrical trigger before startup, the lack of switch command after exit, the lack of position feedback after command, the failure of the fault to disappear after position feedback, and the inconsistency between the topology migration path and the electrical propagation path.

[0092] In one embodiment, when the corrected time uncertainty interval still causes event duplication or event gaps in the same protection action chain, an event equivalence group is constructed according to the data source identifier, event category, associated switch identifier, and time interval inclusion relationship. Duplicate records with the same preceding event and the same succeeding event are merged into composite event atoms. For event gaps, control causal constraints and physical causal constraints are extracted from the event atoms on both sides of the gap to generate placeholder event atoms carrying allowed event categories, allowed topology nodes, and allowed time intervals. Only when the placeholder event atom can simultaneously connect a primary event segment, a secondary event segment, and the corresponding topology migration path is it written into the dual-time-series causal graph, and the original gap branch and the placeholder completion branch are retained side by side.

[0093] The establishment of event equivalence groups is based on the premise that the data sources are different but the event content points to the same physical or control process. The data source identifier is used to distinguish local records, summary records and forwarded records. The event category is used to limit the duplicate records to have the same action semantics. The associated switch identifier is used to exclude the same type of records that act on different switches. The time interval inclusion relationship is used to determine whether multiple records may correspond to the same occurrence process. Only when the current event set and the subsequent event set are consistent at the same time are multiple duplicate records merged into a composite event atom. The composite event atom retains all original record references and their respective time intervals.

[0094] The time uncertainty interval of the composite event atom is taken as the intersection of the time intervals of each member record. When the intersection is empty, the merging is not performed and the event conflict is retained. The source attribute record of the composite event atom is the set of member data sources. The control causal edge and the coupled causal edge are connected to the composite event atom respectively and no longer connected to each member record. When the analysis results are backtracked, the composite event atom can be expanded and each original record can be displayed to avoid duplicate records forming false multiple starts, multiple exits or multiple feedbacks in the candidate causal chain.

[0095] The placeholder event atom does not fill in a definite time or state that is not supported by records. Instead, it saves the allowed event category, allowed topology node, allowed time interval, predecessor constraint and successor constraint. For example, when there is no switch command record between the protection output and the branch current interruption, the allowed event category of the placeholder event atom is limited to the opening command of the corresponding switch, the allowed topology node is limited to the protection output associated switch, and the allowed time interval is limited to after the output event and before the current interruption event. If the placeholder event can simultaneously continue the control action chain and the physical cut-off chain, then it enters the placeholder completion branch.

[0096] The original gap branch does not contain placeholder event atoms and is used to indicate that the existing records are insufficient to determine the missing event. The placeholder completion branch contains constrained placeholder event atoms and is used to indicate the minimum completion path that can exist under the existing physical and control constraints. Both types of branches share the same original record and topology migration marker. When outputting, they indicate whether placeholder event atoms and placeholder conditions are used respectively. When a new switch command record is received, the new record is matched with the placeholder conditions. If the match is successful, the placeholder event atom is replaced with the real event atom. If the match is unsuccessful, the local conflict resolution is re-executed.

[0097] In one embodiment, the processing of candidate causal chains further includes: equivalence merging of multiple candidate causal chains according to the consistency of event atom sets, topology migration paths, and protection action chains; extracting causal edges that appear in common and have the same connection order from each equivalence group to form a common causal skeleton; identifying sub-chains connected to the common causal skeleton but differing in topology stage identifiers, event time uncertainty intervals, or exit event attribution as divergent sub-chains; for each divergent sub-chain, backtracking to the corresponding topology boundary migration marker, branch time interval, or conflict resolution branch to establish a mapping between the divergence source and the fault section, power supply direction, and protection action attribution; outputting a unified analysis record when the common causal skeleton of each equivalence group points to the same result, otherwise outputting a branch analysis record containing the common part and each divergent sub-chain.

[0098] When merging candidate causal chains into equivalent groups, candidate chains with the same event atom set, the same topology migration path, and the same protection start-up to exit main chain are grouped into the same equivalent group. Candidate chains with different values ​​within the time interval but without changing the order of events are still considered equivalent. Candidate chains that change the topology stage or exit affiliation due to different time branches for a certain event are entered into different equivalent groups. Equivalent merging does not delete the original candidate chains, but instead creates an equivalent group index and saves the interval values ​​and conflict resolution records of each chain.

[0099] The common causal skeleton consists of event atoms and causal edges that are common to all candidate causal chains within the equivalence group. During extraction, it is required that the event atom identifiers are consistent, the causal edge directions are consistent, and the topology stage migration order is consistent. The common causal skeleton usually includes the occurrence of fault electrical quantities, protection activation, at least one output or blocking result, switch state changes, and subsequent states of fault quantities. Events that cannot occur in all candidate chains enter the branch sub-chain. The branch sub-chain is connected to the common causal skeleton through the branching start point and the confluence point.

[0100] The divergence source mapping uses topology boundary migration markers, branch time intervals, and conflict resolution branches as source nodes, and fault sections, power supply directions, first-action protection, backup protection, and abnormal action links as result nodes. If a divergence is caused by a location feedback time branch, the mapping records the time interval used and the topology stage changed as a result. If a divergence is caused by a candidate topology slice, the mapping records the state of the connection edge to be confirmed and the fault propagation path changed as a result. If a divergence is caused by the merging of repeated events or placeholder completion, the mapping records the participation position of the composite event atom or placeholder event atom.

[0101] When the common causal framework of each equivalent group points to the same fault section, power supply direction, and protection action attribution, the unified analysis record saves the common results and retains the divergent sub-chains as supplementary records. When different equivalent groups point to different fault sections or different first-action protections, the branch analysis record saves their respective common parts, divergent sub-chains, time branches, topology migration paths, and conflict sources. The output process does not force the merging of branches based on any single reference time scale, so that the analysis results under the case of overlapping faults and switch switching remain consistent with the existing data evidence.

[0102] Furthermore, newly received primary electrical quantities, secondary protection events, or switch status records are located to the relevant equivalent groups through original data references and line indexes. If the new record can eliminate a certain branch time interval, confirm the connection edge to be confirmed, or supplement the placeholder event, the corresponding branch subchain is locally updated and the common causal skeleton is re-extracted. Event atoms and causal edges unaffected by the new record remain unchanged. The updated unified analysis record or branch analysis record continues to save the version association before and after the update so as to continuously verify the line protection data analysis process.

Claims

1. A data analysis method for integrated primary and secondary ring network box line protection, characterized in that, include: The primary electrical quantity sequence, secondary protection event sequence, and switch status sequence of the line are obtained, and each sequence is converted into event atoms carrying data source, time uncertainty interval, and electrical direction attributes; A dynamic line topology map is constructed based on the changes in switch states, and topology boundary migration markers representing the connection relationships before switching, the connection relationships after switching, and the connection relationships to be confirmed are generated in the state change interval. A physical timing diagram is constructed based on the electrical propagation constraints of the fault, and a control timing diagram is constructed based on the logical constraints between protection start-up, output commands, and position feedback. Using the topology boundary migration marker as the coupling node, the physical timing diagram and the control timing diagram are combined into a dual-time causal graph; Perform interval constraint reasoning on the dual-temporal causal graph to obtain candidate causal chains that satisfy electrical propagation constraints, protection logic constraints, and topology migration constraints; The fault section, power supply direction, first-action protection, backup protection, and abnormal action links are determined based on each candidate causal chain.

2. The data analysis method for primary and secondary integrated ring network box line protection according to claim 1, characterized in that, The generation of the event atom includes: dividing the event segment into segments based on the abrupt change start point, direction reversal point, voltage drop start point, zero-sequence component duration boundary, and fault disappearance point in the electrical quantity sequence, and incorporating the sampling clock error and upload delay interval into the corresponding time uncertainty interval; The secondary event segment is formed by the start record, criterion confirmation record, output record, opening and closing command, position feedback and blocking record in the secondary protection event sequence, and the corresponding time uncertainty interval is determined according to the local timestamp and the received timestamp of the same record. The electrical direction attribute is generated based on the current direction relationship, phase sequence relationship, and zero-sequence polarity of adjacent event segments; The data source identifier, event category, time uncertainty interval, electrical direction attribute, and associated switch identifier are encapsulated into the event atom.

3. The data analysis method for primary and secondary integrated ring network box line protection according to claim 1, characterized in that, The generation of the topology boundary migration marker includes: extracting the control command time, position feedback time, and corresponding branch current continuity change time for the switch that undergoes a state change, and converting the three types of time into their respective time intervals; The starting point of the control command interval forms the termination candidate boundary of the connection relationship before the switch, and the ending point of the position feedback interval forms the starting candidate boundary of the connection relationship after the switch. The two types of candidate boundaries are cross-constrained by the interval where the branch current changes from continuous to interrupted or from interrupted to continuous. Within the cross-constraint interval, the topology subgraph before switching, the topology subgraph after switching, and the transition topology subgraph containing the connection edges to be confirmed are retained simultaneously, and the node mapping relationship of the same electrical node between the three topology subgraphs is established.

4. The data analysis method for primary and secondary integrated ring network box line protection according to claim 1, characterized in that, The generation of the dual-time-series causal graph includes: in the physical time-series graph, connecting events that are adjacent in line topology and conform to the fault current propagation direction, voltage drop sequence, and fault quantity continuity into physical causal edges. In the control timing diagram, the start event, criterion confirmation event, exit event, switch command event and position feedback event belonging to the same protection action chain are connected as control causal edges; Based on the pre-switch topology subgraph, transition topology subgraph, and post-switch topology subgraph corresponding to the topology boundary migration marker, a topology stage identifier is added to the cross-graph event atom; Only when there is an overlapping relationship that satisfies the topological stage identifier within the uncertain interval of event time, a coupled causal edge connecting the physical timing diagram and the control timing diagram is established.

5. The data analysis method for primary and secondary integrated ring network box line protection according to claim 2, characterized in that, The determination of the time uncertainty interval includes: extracting paired anchor points that can characterize the same line disturbance from the primary event segment and the secondary event segment. The paired anchor points include fault quantity change and protection start, output record and branch current interruption, reclosing command and voltage recovery. Segmented time difference constraints are formed based on the time scale difference of each pair of anchor points in different data sources, and the segmented time difference constraints are propagated to adjacent event atoms according to the continuous interval of the data source time scale. When the time difference constraints of adjacent segments overlap, their intersection is taken to correct the corresponding time uncertainty interval; when there is no intersection, the branch time interval is retained and the source of conflict is recorded. Write the corrected time uncertainty interval back to the event atom, while maintaining the correspondence between the original timescale and the corrected interval.

6. The data analysis method for primary and secondary integrated ring network box line protection according to claim 3, characterized in that, The construction of the transition topology subgraph includes: taking the switch whose state has changed as the center and extracting a local topology region containing the upstream protection node, the downstream protection node and the adjacent connection node; Write the connection edge before switching, the connection edge after switching, and the connection edge to be confirmed into the local topology region respectively to generate candidate topology slices corresponding to different connection combinations; For each candidate topology slice, a consistency check is performed based on the continuity of branch currents, the correlation of node voltages, the order of control commands and location feedback, and the range of fault disappearance. Candidate topology slices that cannot simultaneously accommodate primary event segments and secondary event segments are deleted. The remaining candidate topology slices are connected to the pre-switch topology subgraph and the post-switch topology subgraph according to the node mapping relationship to form a transition topology subgraph with topology migration path.

7. The data analysis method for primary and secondary integrated ring network box line protection according to claim 4, characterized in that, The interval constraint reasoning includes: converting each causal edge into the corresponding constraint type in the following categories: the endpoint of the preceding interval is no later than the starting point of the following interval, the intervals are allowed to overlap, or the intervals are mutually exclusive; and grouping causal edges that share event atoms but have incompatible constraint types into a conflict set. Extract a closed conflict subgraph from each conflict set that simultaneously involves the physical causal edge, the control causal edge, and the topological migration constraint; Using constraints such as protecting the integrity of the action chain, the continuity of electrical fault propagation, and unidirectional migration of the topology phase, candidate resolution branches are generated to retain, disconnect, or reconnect the causal edges in the closed conflict subgraph. The interval satisfiability test is re-executed for each candidate resolution branch, and the branch that does not have a time cycle and all event atoms have reachable predecessors or reachable successors is determined as the candidate causal chain.

8. The data analysis method for primary and secondary integrated ring network box line protection according to claim 7, characterized in that, The screening of the candidate causal chain includes: determining the earliest event atom in each candidate causal chain that can simultaneously connect current direction change, voltage drop and protection start event, and taking the topology node where the earliest event atom is located as a candidate fault start point. Tracee the fault quantity along the fault causal chain to the corresponding protection node in each exit event, and record the protection zone boundary corresponding to each exit event, the protection node corresponding to the first exit event, the protection node corresponding to subsequent exit events, and the fault disappearance boundary after the switch state changes. The fault section is determined based on the minimum connected section between the candidate fault starting point and the fault quantity disappearance boundary. The power supply direction is determined based on the directed connection relationship of the propagation path. The first-action protection, backup protection and abnormal action links are marked based on the continuous relationship between the output event and the corresponding position feedback on the control causal edge.

9. The data analysis method for primary and secondary integrated ring network box line protection according to claim 5, characterized in that, Also includes: When the corrected time uncertainty interval still causes event duplication or event gap in the same protection action chain, an event equivalence group is constructed according to the data source identifier, event category, associated switch identifier and time interval inclusion relationship, and duplicate records with the same preceding event and the same succeeding event are merged into composite event atoms. For the event gap, control causal constraint and physical causal constraint events are extracted from the event atoms on both sides of the gap, respectively, and placeholder compensation atoms carrying allowed event categories, allowed topology nodes and allowed time intervals are generated; When the placeholder compensation atom can simultaneously connect to the previous secondary event segment and the corresponding topology migration path, it is written into the dual-time-series causal graph, and the original gap branch and the placeholder compensation branch are retained side by side.

10. The data analysis method for primary and secondary integrated ring network box line protection according to claim 8, characterized in that, The processing of the candidate causal chains also includes: merging multiple candidate causal chains into equivalent chains according to the consistency of event atom sets, topology migration paths, and protection action chains, and extracting causal edges that appear in common and have the same connection order from each equivalent group to form a common causal skeleton. Subchains connected to the common causal skeleton but differing in the topological stage identifier, the time uncertainty interval of the event, or the attribution of the exit event are identified as divergent subchains; For each branch subchain, backtrack to the corresponding topology boundary migration marker, branch time interval, or conflict resolution branch to establish a mapping between the source of the branch and the fault section, power supply direction, and protection action attribution; When the common causal skeleton of all equivalence groups points to the same result, a unified analysis record is output; otherwise, a branch analysis record containing the common part and each divergent subchain is output.

Images