A circuit breaker state diagnosis method based on multi-source time sequence fusion

Abstract

The application discloses a circuit breaker state diagnosis method based on multi-source time sequence fusion and relates to the technical field of power equipment pre-diagnosis, which comprises the following steps: according to the action mechanism of the electrical drive of a circuit breaker to the mechanical execution transmission, projecting a multi-source action event set to a preset action stage, constructing an action event matching graph with partial order constraints, and screening events meeting the partial order relationship and the action time delay constraints as action nodes, connecting the action nodes according to the action transmission relationship to generate a circuit breaker action causal graph; performing evidence attribute analysis on the circuit breaker action causal graph, extracting electrical action evidence and mechanical response evidence, and calculating the time sequence residual error of the actual time sequence interval between the two types of evidence relative to the action time delay constraint, determining the evidence conflict polarity according to the time sequence residual error direction, and forming a cross-source evidence conflict relationship. The application realizes early implicit performance degradation identification and state degradation starting action stage positioning, and improves the reliability of equipment whole life cycle state management and control.

Description

Technical Field

[0001] This invention relates to the field of power equipment pre-diagnosis technology, and in particular to a circuit breaker condition diagnosis method based on multi-source time-series fusion. Background Technology

[0002] Circuit breakers are the core execution equipment for the on / off control of power transmission and distribution networks. Currently, the mainstream solutions in the field of power equipment pre-diagnosis and health management are mostly based on the characteristic threshold comparison of single-source time-series signals such as coil current and vibration to carry out state assessment. They rely on algorithms such as Fourier transform and wavelet analysis to extract characteristic quantities such as peak value and time delay in the action process, and combine historical operation and maintenance samples to calibrate health criteria. This has been applied on a large scale in the state perception scenario of power transmission and transformation equipment, providing basic data support for the formulation of operation and maintenance strategies.

[0003] However, conventional methods often process heterogeneous timing signals in the electrical and mechanical domains independently, without considering the partial order constraints and time delay matching relationship of the electrical drive of the circuit breaker to the mechanical execution, making it difficult to explore the coupling and correlation characteristics of cross-domain signals; they mostly determine the equipment status based on the static characteristics of a single action cycle, lacking the ability to continuously track the evolution of the degradation process and failing to identify early hidden performance degradation signs. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a circuit breaker condition diagnosis method based on multi-source timing fusion to solve the problems of difficulty in characterizing the coupling relationship between electrical and mechanical domain timing signals and difficulty in tracking the early degradation and evolution process.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a circuit breaker condition diagnosis method based on multi-source time-series fusion, comprising: performing segmented change detection on control voltage and coil current in multi-source operation time-series data of the circuit breaker to identify electrical operation events; performing multi-scale detection on response waveforms generated by the operation of the circuit breaker mechanism to identify mechanical response events; and performing time-series aggregation of various events according to the time of event occurrence to generate a multi-source operation event set; based on the action mechanism of the circuit breaker's electrical drive to mechanical execution, projecting the multi-source operation event set onto a preset action stage to construct a partially ordered constraint action event matching graph, and selecting events that satisfy the partially ordered relationship and action delay constraint as action nodes, connecting them according to the action transmission relationship to generate a circuit breaker action cause-effect graph; and performing evidence attribute analysis on the circuit breaker action cause-effect graph to improve... Evidence of electrical action and mechanical response is collected, and the temporal residual between the actual time interval of the two types of evidence relative to the action delay constraint is calculated. The polarity of the evidence conflict is determined according to the direction of the temporal residual, forming a cross-source evidence conflict relationship. The cross-source evidence conflict relationship is causally traced along the action transmission direction of the circuit breaker action causal diagram. Cross-source evidence conflict relationships with consistent evidence conflict polarity and continuous accumulation of temporal residuals are aggregated into a decoupling propagation chain. The stable starting point is determined by the residual drift verification of continuous similar actions, generating an electromechanical decoupling precursor path. The electromechanical decoupling precursor path is backtracked with the least interpretation path, and the first action node that continuously generates the same polarity conflict is searched in reverse to locate the initial action stage of state degradation. The health status level of the circuit breaker is evaluated according to the degree of accumulation of the path temporal residual.

[0007] As a preferred embodiment of the circuit breaker condition diagnosis method based on multi-source timing fusion described in this invention, the steps for generating the multi-source action event set are as follows: Using the control voltage change time in the multi-source operation timing data of the circuit breaker as the electrical test benchmark, a sliding window cumulative deviation analysis is performed on the control voltage and coil current to divide the operation segment, and the boundary of the operation segment is corrected by the waveform turning position to generate electrical operation events. Based on the detection time window of the mechanism action response defined by the electrical action event, the response waveform generated by the circuit breaker mechanism action within the detection time window is subjected to multi-scale envelope energy tracking, and the response mutation position that is stably reproduced across scales is screened to generate mechanical response events. The mechanical response events are mapped to the time axis of the electrical action events. Cross-source event pairings are established within the detection time window according to the principle of minimum event time difference, and unpaired events and event source identifiers are retained. Multi-source action event sets are generated by merging events according to their occurrence time.

[0008] As a preferred embodiment of the circuit breaker condition diagnosis method based on multi-source timing fusion described in this invention, the steps for constructing the action event matching graph with partial order constraints are as follows: Based on the action mechanism of the circuit breaker's electrical drive to mechanical execution, an action stage template is established. A dynamic time warping algorithm is used to align the multi-source action event set with the action stage template in order constraint. The action stage to which each action event belongs is determined according to the minimum cumulative time sequence cost, and an action stage projection sequence is generated. Based on the sequential dependencies between action stages, the projected sequence of action stages is constructed into a directed acyclic candidate graph, and the absolute deviation between the event occurrence time difference and the center value of the action delay constraint interval is used as the connection weight to generate a partially ordered action event matching graph.

[0009] As a preferred embodiment of the circuit breaker condition diagnosis method based on multi-source timing fusion described in this invention, the steps for generating the circuit breaker operation cause-effect graph are as follows: Dynamic programming path search is performed on the action event matching graph with partial order constraints. Candidate connection edges that do not satisfy the partial order relationship and action delay constraints are eliminated. The optimal action path is determined according to the principle of minimum cumulative connection weight, and action nodes and action transmission edges are obtained. Connect the action nodes and action transmission edges according to the transmission direction from electrical drive to mechanical execution, and write the actual action delay into the action transmission edge to generate a cause-effect graph of circuit breaker action.

[0010] As a preferred embodiment of the circuit breaker condition diagnosis method based on multi-source time-series fusion described in this invention, the steps for extracting electrical action evidence and mechanical response evidence are as follows: Evidence attributes of each action node are analyzed along the action transmission direction of the circuit breaker action cause-effect diagram. The driving side action nodes are classified into electrical action evidence, and the execution side response nodes are classified into mechanical response evidence, generating a hierarchical evidence node sequence. Based on the stage connection relationship of action nodes in the hierarchical evidence node sequence, causal adjacency pairing is performed on electrical action evidence and mechanical response evidence, and action delay constraints that match the action transmission edges connecting the two types of evidence are obtained to generate a cross-source evidence matching set.

[0011] As a preferred embodiment of the circuit breaker condition diagnosis method based on multi-source time-series fusion described in this invention, the steps for forming cross-source evidence conflict relationships are as follows: Directed difference is performed on the evidence occurrence time in the cross-source evidence matching pair set to obtain the actual time interval. The actual time interval is compared with the action delay constraint at the interval boundary. For the actual time interval that exceeds the action delay constraint, the signed deviation is calculated. For the actual time interval that is within the action delay constraint, the deviation is set to zero. The evidence time residual of each cross-source evidence matching pair is obtained. Cross-source evidence matching pairs with non-zero temporal residuals are selected, the polarity of evidence conflict is determined according to the sign direction of the temporal residuals, and the degree of conflict is characterized by the residual amplitude, thus forming cross-source evidence conflict relationships.

[0012] As a preferred embodiment of the circuit breaker condition diagnosis method based on multi-source temporal fusion described in this invention, the step of aggregating cross-source evidence conflict relationships with consistent evidence conflict polarity and continuous accumulation of temporal residuals into a decoupled propagation chain is as follows: Using cross-source evidence conflict relationships as the starting point for tracing, a directed graph traversal is performed along the action transmission direction of the circuit breaker action causal graph to associate cross-source evidence conflict relationships with the action nodes to which the mechanical response evidence belongs, and cross-source evidence conflict relationships with causal reachability are connected according to action stages to generate a conflict causal tracing chain. A polarity consistency test is performed on the conflict relationship between adjacent cross-source evidence in the conflict causal tracing chain. Conflict causal tracing chains with consistent evidence conflict polarity and whose absolute value of the signed deviation increases step by step along the action transmission direction are selected. The cumulative residual of the link is calculated according to the same-direction increment of the signed deviation to generate the decoupled propagation chain.

[0013] As a preferred embodiment of the circuit breaker condition diagnosis method based on multi-source timing fusion described in this invention, the steps for generating electromechanical decoupling precursor paths are as follows: The decoupling propagation chains corresponding to the continuous operation of the same type of circuit breaker are aligned according to the operation stage. The sliding window least squares method is used to fit the residual drift slope of the signed deviation associated with each operation node as the number of circuit breaker operations changes. The foremost action node with the consistent sign of the residual drift slope and the smallest slope dispersion is selected as the stable starting point. The same polarity residual extension path is retained along the decoupling propagation chain from the stable starting point to generate the electromechanical decoupling precursor path.

[0014] As a preferred embodiment of the circuit breaker condition diagnosis method based on multi-source timing fusion described in this invention, the steps of the initial action stage for locating the state degradation are as follows: Using the terminal action node of the electromechanical decoupling precursor path as the backtracking starting point, the backtracking is performed along the action transmission edge, and the interpretation cost is set for the backtracking edge according to the consistency of the conflict polarity of adjacent action nodes to generate a reverse interpretation graph. Perform a constrained shortest path search on the reverse interpretation graph, and determine the backtracking path that can continuously cover the precursor paths of electromechanical decoupling and has the minimum cumulative interpretation cost as the minimum interpretation path; Along the least interpretation path, the signed deviations associated with each action node under consecutive circuit breaker operations of the same type are aligned according to the number of circuit breaker operations. The position where the signed deviations begin to drift in the same direction is identified. The action node that first maintains the same polarity conflict is taken as the state degradation initiation action node, and the state degradation initiation action stage is determined.

[0015] As a preferred embodiment of the circuit breaker condition diagnosis method based on multi-source timing fusion described in this invention, the step of assessing the health status level of the circuit breaker based on the accumulation degree of path timing residuals refers to performing path-weighted accumulation of the signed deviations associated with each action node along the minimum interpretation path from the starting action node of state degradation, obtaining the accumulated value of path timing residuals, and comparing the accumulated value of path timing residuals with a preset circuit breaker health status range to assess the health status level of the circuit breaker.

[0016] The beneficial effects of this invention are as follows: by generating a causal graph of circuit breaker operation and organizing multi-source operation events according to partial order relationships and operation delay constraints, the causal connection between electrical operation evidence and mechanical response evidence is realized, providing a unified analytical basis for identifying cross-source evidence conflicts; by generating electromechanical decoupling precursor paths and implementing minimum interpretation path backtracking, early latent performance degradation identification and state degradation initiation stage positioning are realized, improving the reliability of equipment life cycle state management. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a circuit breaker condition diagnosis method based on multi-source timing fusion.

[0019] Figure 2 A flowchart for assessing the health status level of a circuit breaker.

[0020] Figure 3 This is a comparative data chart showing the ability to identify early latent performance degradation under different diagnostic methods.

[0021] Figure 4 A comparative chart showing the accuracy of localization during the initial action phase of state degradation under different localization methods. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a circuit breaker condition diagnosis method based on multi-source timing fusion, including the following steps: S1. Perform segmented change detection on the control voltage and coil current in the multi-source operation timing data of the circuit breaker to identify electrical operation events. Perform multi-scale detection on the response waveform generated by the operation of the circuit breaker mechanism to identify mechanical response events. Then, perform time-series aggregation on various events according to the time of event occurrence to generate a multi-source operation event set.

[0026] S1.1: Using the control voltage change time in the multi-source operation timing data of the circuit breaker as the electrical test benchmark, a sliding window cumulative deviation analysis is performed on the control voltage and coil current to divide the operation segment, and the boundary of the operation segment is corrected by the waveform turning position to generate electrical operation events.

[0027] Specifically, the control voltage change time in the multi-source operation timing data of the circuit breaker is used as the electrical test benchmark. Sliding windows are set before and after the control voltage change time. The mean deviation of the control voltage and coil current between adjacent sliding windows is calculated respectively. The mean deviation is accumulated along the time axis. The operation segment is divided according to the position where the direction of the accumulated deviation change changes. The difference between adjacent samples of control voltage and coil current is calculated, and the position where the sign of the adjacent sample difference changes is determined as the waveform inflection point. Boundary correction windows are set on both sides of the start and end boundaries of the action section. The first-order difference of control voltage and coil current and the change of adjacent first-order difference within the boundary correction windows are calculated. Within the start boundary correction window, the position where the waveform changes from steady fluctuation to continuous unidirectional change and the change of adjacent first-order difference is the largest is selected as the corrected start boundary of the action section. Within the end boundary correction window, the position where the waveform returns from continuous unidirectional change to steady fluctuation outside the action section is selected as the corrected end boundary of the action section. The time corresponding to the start boundary of the corrected action section is taken as the event occurrence time, and the waveform change characteristics are determined according to the direction of control voltage change and the cumulative change amplitude of coil current within the corrected action section to generate an electrical action event.

[0028] S1.2: Based on the electrical action event, the detection time window of the mechanism action response is defined. Multi-scale envelope energy tracking is performed on the response waveform generated by the circuit breaker mechanism action within the detection time window. The response mutation positions that are stably reproduced across scales are screened to generate mechanical response events.

[0029] Specifically, the start time of the electrical action event is taken as the starting point of the detection time window, and the end point of the detection time window is determined by combining the duration of the action segment and the historical action response delay. The response waveform generated by the circuit breaker mechanism action within the detection time window is decomposed into wavelet multi-scale decomposition. The response waveform envelope is extracted by Hilbert transform at each scale, and the squared value of the response waveform envelope is accumulated by sliding to form a multi-scale envelope energy sequence. The energy difference between adjacent sliding windows in the multi-scale envelope energy sequence is calculated. The time corresponding to the extreme value of the energy difference is marked as the candidate response mutation position. Then, the candidate response mutation positions at different scales are aligned according to the temporal proximity, and the position that repeats across scales and has the smallest time offset is retained as the response mutation position. The time corresponding to the response mutation position is taken as the occurrence time of the mechanical response event, and the change amplitude of the envelope energy before and after the response mutation position is taken as the response feature of the mechanical response event, thus generating the mechanical response event.

[0030] It should be noted that wavelet multiscale decomposition selects wavelet bases that match the abrupt change characteristics of the response waveform, performs low-frequency approximation and high-frequency detail decomposition on the response waveform within the detection window layer by layer, obtains wavelet coefficients at different scales by downsampling, and then reconstructs the response waveform components at the corresponding scales of the wavelet coefficients at each scale. The principle of Hilbert transform is to convolve the response waveform with a kernel function to obtain orthogonal components that have the same amplitude change as the original response waveform but are 90 degrees out of phase. The original response waveform and the orthogonal components are then combined to form an analytic signal, so that the magnitude of the analytic signal can characterize the instantaneous envelope of the response waveform. In this embodiment, Hilbert transform is performed on the response waveform components at each scale to obtain orthogonal components that are 90 degrees out of phase with the response waveform components. The response waveform components and the orthogonal components are then combined to form an analytic signal. The real and imaginary parts of the analytic signal are squared and then squared to obtain the response waveform envelope that reflects the instantaneous amplitude change at each moment.

[0031] S1.3: Map mechanical response events to the time axis of electrical action events, establish cross-source event pairings within the detection window according to the principle of minimum event time difference, retain unpaired events and event source identifiers, and merge them according to the event occurrence time to generate a multi-source action event set.

[0032] Specifically, using the start point of the detection time window as a unified timing benchmark, the occurrence time of mechanical response events is converted to the time axis of electrical action events according to the sampling interval. Within each detection time window, the event time difference between mechanical response events and each electrical action event is calculated. Cross-source event pairings are established in ascending order of event time difference. Paired electrical action events and mechanical response events are marked as occupied to avoid duplicate pairings of the same event. The pairing result with the smallest event time difference is recorded as the cross-source event pairing. For electrical action events and mechanical response events that have not formed cross-source event pairings, the event occurrence time and event source identifier are retained respectively. Then, the cross-source event pairings and unpaired events are sorted and merged according to the event occurrence time to generate a multi-source action event set.

[0033] S2. Based on the action mechanism of the circuit breaker's electrical drive to mechanical execution, project the multi-source action event set to the preset action stage, construct a partial order constraint action event matching diagram, and select events that satisfy the partial order relationship and action delay constraint as action nodes, connect them according to the action transmission relationship, and generate a circuit breaker action cause-effect diagram.

[0034] S2.1: Based on the action mechanism of the circuit breaker's electrical drive to mechanical execution, an action stage template is established. The dynamic time warping algorithm is used to align the multi-source action event set with the action stage template in order constraint, and the action stage to which each action event belongs is determined according to the minimum cumulative time sequence cost, generating an action stage projection sequence.

[0035] It should be explained that the operating mechanism of the circuit breaker from electrical drive to mechanical execution is as follows: after the control voltage changes, the opening or closing coil establishes a coil current. The electromagnetic driving force generated by the coil current drives the iron core to move. The movement of the iron core releases the holding state of the operating mechanism. The operating mechanism releases its stored energy and drives the contacts to move through the transmission structure until the contacts complete the opening or closing and enter a stable position. Based on the above energy transfer and action connection process, the electrical triggering process corresponding to the change in control voltage is taken as the starting point of the action stage, and the electromagnetic drive process corresponding to the establishment of the coil current is taken as the subsequent action stage. The mechanical execution is divided according to the release time of the mechanism and the movement time of the contacts reflected in the mechanical response event. The process is as follows: the endpoint of the action stage is determined by the position where the mechanical response waveform returns to stability; the sequential constraints between action stages are recorded according to the chronological relationship of each process; a matching event source and a time range relative to the start of the action stage are configured for each action stage; the time range of each action stage is set according to the historical normal operation samples of the same type of circuit breaker; the occurrence time difference of the associated action events of each action stage relative to the moment of control voltage change is statistically analyzed; after removing the abnormal occurrence time difference, the lower quantile (5th percentile) and upper quantile (95th percentile) of the remaining occurrence time difference are used as the lower limit and upper limit of the time range of the corresponding action stage, and an action stage template is generated.

[0036] Specifically, the occurrence time and source of each action event are extracted from the multi-source action event set. The event source matching cost between each action event and each action stage in the action stage template is calculated (0 is taken when the event source of the action event matches the event source configured in the action stage, and 1 is taken when they do not match). The time cost of the event occurrence time deviating from the stage time sequence range is also calculated. The event source matching cost and the time cost are merged into a local time sequence cost. A dynamic time warping algorithm is used to construct a cumulative time sequence cost matrix according to the occurrence order of action events and the order of action stages. The minimum cumulative time sequence cost is selected from the predecessor positions that satisfy the order constraints and merged with the local time sequence cost of the current position. Unreachable marks are set for positions that do not satisfy the order constraints. The order constraint alignment path is determined by searching backward from the end position of the cumulative time sequence cost matrix along the direction of the minimum cumulative time sequence cost. Each action event is assigned to the action stage with the minimum cumulative time sequence cost according to the order constraint alignment path. The action stages to which each action event belongs are arranged according to the occurrence time of the event to generate the action stage projection sequence.

[0037] It should be noted that the time cost is calculated based on the distance between the event occurrence time and the stage time range. When the event occurrence time is within the stage time range, it is 0. When it is earlier than the lower limit of the stage time range, it is the difference between the lower limit and the event occurrence time. When it is later than the upper limit of the stage time range, it is the difference between the event occurrence time and the upper limit. The time cost is normalized using the width of the stage time range.

[0038] S2.2: Based on the sequential dependencies between action stages, construct a directed acyclic candidate graph from the projected sequence of action stages, and use the absolute deviation between the event occurrence time difference and the center value of the action delay constraint interval as the connection weight to generate a partially ordered action event matching graph.

[0039] Specifically, each action event in the action phase projection sequence is set as a candidate node. According to the sequential dependency between action phases, candidate connection edges are established only from action events in the preceding action phase to action events in the following action phase. Candidate connection edges that violate the action phase order are excluded, so that the candidate nodes and candidate connection edges form a directed acyclic candidate graph. For each candidate connection edge, the event occurrence time difference between the action events at both ends of the connection edge is calculated. The average of the upper limit and the lower limit of the action delay constraint interval is used as the center value of the action delay constraint interval. The absolute deviation between the event occurrence time difference and the center value of the action delay constraint interval is calculated and written into the corresponding candidate connection edge as the connection weight, generating a partially ordered action event matching graph.

[0040] It should be noted that the occurrence time difference of adjacent action events is extracted from historical normal action samples, abnormal occurrence time differences that exceed the interquartile range are removed, and the lower quartile value (such as the 5th percentile) and upper quartile value (such as the 95th percentile) of the remaining occurrence time difference are used as the lower limit and upper limit of the action delay constraint, respectively, to form the action delay constraint. The partially ordered action event matching graph filters out erroneous cross-source pairings by using the sequential relationship of action stages and action delay constraints, providing a reliable event association basis for constructing the causal graph of circuit breaker actions.

[0041] S2.3: Perform dynamic programming path search on the action event matching graph with partial order constraints, eliminate candidate connection edges that do not satisfy the partial order relationship and action delay constraints, and determine the optimal action path according to the principle of minimum cumulative connection weight, thereby filtering out action nodes and action transmission edges.

[0042] Specifically, taking the candidate node of the first action stage in the partially ordered action event matching graph as the path starting point, the cumulative connection weight of the path starting point is initialized to zero, and the candidate nodes are topologically arranged according to the order of action stages. The candidate connection edges are checked layer by layer along the topological arrangement direction, comparing whether the action stages of the action events at both ends of the candidate connection edge satisfy the partial order relationship (e.g., control voltage change event precedes coil current establishment event, coil current establishment event precedes mechanical response event, and the candidate association direction is consistent with the above order, i.e., satisfying the partial order relationship), and comparing whether the time difference of event occurrence is within the action delay constraint. Candidate connection edges that do not meet any condition are eliminated; for the remaining candidate connection edges, the minimum cumulative connection weight of the preceding candidate node is accumulated with the connection weight of the current candidate connection edge to obtain the cumulative connection weight of the current candidate node, and the preceding candidate node that generates the minimum cumulative connection weight is recorded, until the path recursion of the final action stage is completed; from the final action stage, the candidate node with the minimum cumulative connection weight is selected, and the optimal action path is formed by backtracking along the preceding candidate node, the candidate nodes on the optimal action path are determined as action nodes, and the candidate connection edges connecting adjacent action nodes are determined as action transmission edges.

[0043] S2.4: Connect the action node and the action transmission edge according to the transmission direction from electrical drive to mechanical execution, and write the actual action delay into the action transmission edge to generate the circuit breaker action cause-effect graph.

[0044] Specifically, according to the transmission direction from electrical drive to mechanical execution, the action nodes in the optimal action path are sorted sequentially, and directed connections are established from the preceding action node to the following action node using action transmission edges. For each action transmission edge, the difference between the event occurrence time of the following action node and the event occurrence time of the preceding action node is calculated as the actual action delay and written into the corresponding action transmission edge. The action stage and event source associated with the action node are written into the action node attributes, so that the directed connections between action nodes can represent the action transmission process of the circuit breaker from electrical drive to mechanical execution, and generate a circuit breaker action cause-effect graph.

[0045] It should be noted that the circuit breaker action cause-effect diagram is a directed cause-effect structure that organizes electrical action events and mechanical response events according to the action stage and actual action delay. It realizes a cross-source unified representation of the electrical drive process and mechanical execution process of the circuit breaker, and can reveal the temporal coupling state and decoupling position in the action transmission relationship.

[0046] S3. Perform evidence attribute analysis on the circuit breaker action cause-effect diagram, extract electrical action evidence and mechanical response evidence, and calculate the time residual between the actual time interval of the two types of evidence and the action delay constraint. Determine the polarity of evidence conflict according to the direction of the time residual to form a cross-source evidence conflict relationship.

[0047] S3.1: Perform evidence attribute analysis on each action node along the action transmission direction of the circuit breaker action cause-effect diagram, classify the driving side action nodes into electrical action evidence, classify the execution side response nodes into mechanical response evidence, and generate a hierarchical evidence node sequence.

[0048] Specifically, the action nodes are traversed level by level along the action transmission direction of the circuit breaker action cause-effect diagram, the event sources associated with the action nodes are extracted, and the action nodes originating from the control voltage sampling channel and the coil current sampling channel are classified as electrical action evidence, and the action nodes originating from the mechanism action response waveform are classified as mechanical response evidence. The hierarchical connection relationship is determined according to the starting and ending action nodes of the action transmission edge: when the starting point of the action transmission edge belongs to electrical action evidence and the ending point belongs to mechanical response evidence, the layer where the electrical action evidence is located is set as the predecessor evidence layer, and the layer where the mechanical response evidence is located is set as the successor evidence layer; when there are action transmission edges between similar evidence, the adjacent evidence layers are determined according to the order of the action stages, and arranged sequentially along the direction of the action transmission edge. After completing the hierarchical connection, each evidence layer is arranged in the order of the preceding evidence layer to the subsequent evidence layer, and the action nodes within each evidence layer are arranged in the order of the action stages. When there are multiple action nodes in the same action stage, they are sorted according to the time of the event to generate a hierarchical evidence node sequence.

[0049] S3.2: Based on the stage connection relationship of action nodes in the hierarchical evidence node sequence, perform causal adjacency pairing of electrical action evidence and mechanical response evidence, obtain action delay constraints that match the action transmission edges connecting the two types of evidence, and generate cross-source evidence matching sets.

[0050] Specifically, starting from the electrical action evidence in the preceding evidence layer, the mechanical response evidence that is directly accessible in the subsequent evidence layer is searched along the action transmission edge. The electrical action evidence with the same action transmission direction and adjacent action stages are paired with the mechanical response evidence. When there are multiple accessible mechanical response evidences for the same electrical action evidence, the mechanical response evidence with the smallest action stage span is selected first, and the pairing results are determined according to the time difference of the event occurrence from small to large. Based on the action stages spanned by the action transmission edge connecting electrical action evidence and mechanical response evidence, action delay constraints are extracted from the action delay constraints corresponding to the same action stage combination, and the action delay constraints are written into causal adjacency pairing to generate a cross-source evidence matching set.

[0051] S3.3: Perform directed difference on the evidence occurrence time in the cross-source evidence matching pair set to obtain the actual time interval, and compare the actual time interval with the action delay constraint at the interval boundary. Calculate the signed deviation for the actual time interval that exceeds the action delay constraint, and set the deviation of the actual time interval that is within the action delay constraint to zero. Obtain the evidence time residual for each cross-source evidence matching pair.

[0052] Specifically, the actual time interval from the occurrence of electrical drive to the occurrence of mechanical response is obtained by subtracting the occurrence time of electrical action evidence from the occurrence time of mechanical response evidence. The lower limit and upper limit of action delay constraints associated with cross-source evidence matching pairs are extracted. The actual time interval is compared with the lower limit and upper limit of action delay constraints in turn: when the actual time interval is less than the lower limit of action delay constraints, the difference between the actual time interval and the lower limit of action delay constraints is calculated to obtain a negative signed deviation. When the actual time interval is greater than the upper limit of action delay constraints, the difference between the actual time interval and the upper limit of action delay constraints is calculated to obtain a positive signed deviation. When the actual time interval is within the action delay constraints, the actual time interval deviation is set to zero, and the obtained signed deviation is recorded as the evidence time residual of each cross-source evidence matching pair.

[0053] S3.4: Screen cross-source evidence matching pairs with non-zero temporal residuals, determine the polarity of evidence conflict according to the sign direction of the temporal residuals, and characterize the degree of conflict with the residual amplitude to form cross-source evidence conflict relationships.

[0054] Specifically, cross-source evidence matching pairs with non-zero temporal residuals are selected from the cross-source evidence matching pair set, and the polarity of evidence conflict is determined according to the sign direction of the temporal residuals: when the temporal residuals are positive, mechanical response evidence lagging behind the action delay constraint is marked as a positive conflict; when the temporal residuals are negative, mechanical response evidence advancing relative to the action delay constraint is marked as a negative conflict. The absolute value of the temporal residual of the evidence is calculated as the residual amplitude, and the residual amplitude is normalized using the width of the action delay constraint interval to obtain the degree of conflict. The polarity and degree of conflict of the evidence are written into the corresponding cross-source evidence matching pair, and the action transmission relationship from electrical action evidence to mechanical response evidence is retained to form a cross-source evidence conflict relationship.

[0055] It should be noted that the cross-source evidence conflict relationship transforms the temporal deviation between electrical action evidence and mechanical response evidence that exceeds the action delay constraint into a unified representation with conflict polarity and conflict degree, thereby realizing the quantitative identification of the direction and degree of cross-domain action mismatch.

[0056] S4. Follow the action transmission direction of the circuit breaker action causal diagram to trace the cross-source evidence conflict relationship, aggregate the cross-source evidence conflict relationship with consistent polarity and continuous accumulation of time sequence residuals into a decoupling propagation chain, and determine the stable starting point through residual drift verification of continuous similar actions to generate electromechanical decoupling precursor path.

[0057] S4.1: Using cross-source evidence conflict relationships as the starting point for tracing, perform a directed graph traversal along the action transmission direction of the circuit breaker action causal graph, associate cross-source evidence conflict relationships with the action nodes to which the mechanical response evidence belongs, and connect cross-source evidence conflict relationships with causal reachability according to action stages to generate a conflict causal tracing chain.

[0058] Specifically, the action node to which the mechanical response evidence belongs in each cross-source evidence conflict relationship is used as the tracking node, and the evidence conflict polarity and conflict degree are marked to the tracking node; along the action transmission direction of the circuit breaker action causal graph, the search proceeds backward from the current tracking node. When a subsequent action node is associated with another cross-source evidence conflict relationship, a causal reachable connection is established from the current cross-source evidence conflict relationship to the subsequent cross-source evidence conflict relationship; cross-source evidence conflict relationships with causal reachable relationships are connected sequentially according to the order of action stages, and links are formed for different action transmission branches to generate a conflict causal tracking chain.

[0059] S4.2: Perform a polarity consistency test on the conflict relationship between adjacent cross-source evidence in the conflict causal tracing chain, screen the conflict causal tracing chains with consistent evidence conflict polarity and whose absolute value of signed deviation increases step by step along the action transmission direction, and calculate the cumulative residual of the link according to the same direction increment of the signed deviation to generate the decoupled propagation chain.

[0060] Specifically, the evidence conflict polarity of adjacent cross-source evidence conflict relationships is compared sequentially along the action transmission direction of the conflict causality tracing chain. When the evidence conflict polarity of adjacent cross-source evidence conflict relationships is the same, the increase in the absolute value of the subsequent signed deviation relative to the absolute value of the preceding signed deviation is calculated, and adjacent cross-source evidence conflict relationships with an increase greater than zero are continuously retained. When the evidence conflict polarity changes or the increase is not greater than zero, the current chain segment is terminated and re-examined from the subsequent cross-source evidence conflict relationships. The increase in each retained chain segment is accumulated along the action transmission direction to obtain the link cumulative residual, and the retained chain segments with consistent evidence conflict polarity and progressively increasing absolute values ​​of signed deviation are determined as decoupling propagation chains.

[0061] It should be noted that the decoupling propagation chain is a degenerate propagation structure formed by continuously organizing cross-source evidence conflict relationships with consistent polarity of evidence conflict and progressively increasing sign deviation along the direction of action transmission. It realizes the chain-like characterization of the degree of decoupling between the electrical drive and mechanical execution of the circuit breaker.

[0062] S4.3: Align the decoupling propagation chains corresponding to the continuous operation of circuit breakers of the same type according to the operation stage, and use the sliding window least squares method to fit the residual drift slope of the signed deviation associated with each operation node as the number of circuit breaker operations changes.

[0063] Specifically, decoupling propagation chains corresponding to consecutive circuit breaker actions of the same type are collected according to the type and sequence of circuit breaker actions. The action stage is used as the alignment index to arrange the action nodes in the same action stage of different circuit breaker actions to the same position. The action sequence number is set for each decoupling propagation chain according to the number of circuit breaker actions. The signed deviation associated with each action node under the same action stage is extracted to form a time sequence of the signed deviation as the number of circuit breaker actions changes. A fixed-length sliding window is set on the time series and moved sequentially according to the single circuit breaker operation. The number of circuit breaker operations within the window is used as the independent variable, and the corresponding signed deviation is used as the dependent variable. The linear fitting slope is obtained by minimizing the squared error between the fitted value and the signed deviation. The linear fitting slope is determined as the residual drift slope of the signed deviation associated with each operation node as a function of the number of circuit breaker operations. The expression is: ; in, Indicates the first The residual drift slope of each action node in the Wth sliding window; N represents the number of consecutive circuit breaker actions of the same type contained in the sliding window, N is an integer not less than 2, the number of actions of each circuit breaker in the sliding window is set in a continuously increasing sequence to make the number of actions of each action not completely the same, and to ensure that the denominator of the residual drift slope expression is greater than zero. When the number of effective consecutive circuit breaker actions of the same type is less than 2, the residual drift slope is not calculated. The fitting calculation is performed after the number of effective actions meets the condition. This represents the number of actions performed by the circuit breaker during the i-th operation within the sliding window. This represents the average number of actions within the W-th sliding window; Indicates the first The signed deviation associated with each action node in the i-th circuit breaker action; Indicates the first The average signed deviation of each action node within the Wth sliding window; W is the sliding window number. is the action node sequence number; i is the circuit breaker action sequence number.

[0064] It should be noted that circuit breaker actions of the same type refer to continuous circuit breaker actions with the same action category, the same action command, and comparable operating conditions. For example, multiple closing actions performed continuously under the same rated voltage and load conditions are considered circuit breaker actions of the same type, while closing actions and opening actions are not considered circuit breaker actions of the same type.

[0065] S4.4: Select the foremost action node with the same sign of residual drift slope and the smallest slope dispersion as the stable starting point, and retain the same polarity residual extension path along the decoupling propagation chain from the stable starting point to generate the electromechanical decoupling precursor path.

[0066] Specifically, for each action node in the decoupling propagation chain, the residual drift slope within a fixed number (e.g., 3) of consecutive sliding windows is extracted. When all residual drift slopes are positive or negative, it is determined that the signs of the residual drift slopes are consistent. The standard deviation of each residual drift slope relative to the average residual drift slope is calculated as the slope dispersion. The action node with the smallest slope dispersion and the earliest action stage is selected as the stable starting point. From the stable starting point, the subsequent cross-source evidence conflict relationship is examined level by level along the action transmission direction of the decoupling propagation chain. Action nodes with the same evidence conflict polarity and whose absolute value of the signed deviation increases relative to the preceding action node, as well as action transmission edges, are retained until the evidence conflict polarity changes or the absolute value of the signed deviation no longer increases, thus generating the electromechanical decoupling precursor path.

[0067] It should be noted that when the cross-source evidence conflict relationship fails to form an electromechanical decoupling precursor path through residual increment tracing and residual drift verification, the cumulative value of the path timing residual is recorded as zero, and the circuit breaker health status level assessment is directly entered. The early signs of electromechanical decoupling are characterized by aggregating cross-source evidence conflicts with consistent polarity and continuously increasing residuals, and by screening stable starting points based on residual drift during continuous operation. This enables continuous characterization of the early decoupling trend between the electrical drive and mechanical execution of circuit breakers.

[0068] Figure 3 This paper demonstrates the changing trends of normalized diagnostic indices for different diagnostic methods as the number of circuit breaker operations increases during the continuous operation of the same type of circuit breaker. The horizontal axis represents the number of circuit breaker operations, and the vertical axis represents the normalized diagnostic indices. The normalization process is only used to convert the diagnostic quantities output by different methods to a unified display scale. The original diagnostic quantities represented by the three curves are not the same: the single-source feature threshold method uses the deviation of the coil current peak and mechanical response energy from their respective judgment thresholds as diagnostic quantities; the single cross-source time delay judgment method uses the evidence time sequence residual formed after the actual time interval between electrical action evidence and mechanical response evidence exceeds the action time delay constraint as diagnostic quantities; and the method of this invention uses the cumulative value of the path time sequence residual obtained by weighting and accumulating the signed deviations associated with each action node in the electromechanical decoupling precursor path as diagnostic quantities. It can be seen that the single-source characteristic threshold method requires waiting for a significant change in the coil current or mechanical response waveform before gradually entering the high-value region, making it difficult to form a clear diagnostic result in the early stages of performance degradation. The single-event cross-source delay determination method can detect the phenomenon of the actual timing interval deviating from the action delay constraint earlier, but the single-event evidence timing residual is easily affected by the fluctuation of the event time. The method of this invention organizes cross-source action events through the circuit breaker action causal graph and performs causal tracking on the cross-source evidence conflict relationship with consistent evidence conflict polarity and continuous accumulation of timing residuals, so that the cumulative value of the path timing residual appears to rise continuously earlier and enters the stable high-value region first. Therefore, Figure 3 This invention verifies that it can identify early latent performance degradation based on the continuously expanding temporal decoupling relationship between electrical action evidence and mechanical response evidence, even before the single-source action characteristics have reached a significant abnormality level, thereby improving the foresight and stability of circuit breaker condition diagnosis.

[0069] S5. Perform minimum interpretation path backtracking on the electromechanical decoupling precursor path, reverse search for the first action node that continuously generates the same polarity conflict, locate the initial action stage of state degradation, and evaluate the circuit breaker health status level according to the degree of accumulation of path timing residuals.

[0070] S5.1: Using the terminal action node of the electromechanical decoupling precursor path as the backtracking starting point, perform reverse traversal along the action transmission edge, and set the interpretation cost for the backtracking edge according to the consistency of the conflict polarity of adjacent action nodes to generate a reverse interpretation graph.

[0071] Specifically, taking the end action node of the electromechanical decoupling precursor path as the backtracking starting point, the action transmission edge pointing to the current action node in the circuit breaker action cause-effect graph is converted into a backtracking edge pointing from the current action node to the preceding driving action node, and the preceding driving action node is searched level by level along the backtracking edge. For each backtracking edge, read the cross-source evidence conflict relationships associated with the action nodes at both ends of the backtracking edge; when both action nodes are associated with cross-source evidence conflict relationships and the evidence conflict polarity is the same, record the conflict polarity consistency as 1 and set the interpretation cost to 0; when the conditions of both action nodes being associated with cross-source evidence conflict relationships and the evidence conflict polarity being the same are not met simultaneously, record the conflict polarity consistency as 0 and set the interpretation cost to 1; connect the action nodes and backtracking edges according to the backtracking direction, and write the interpretation cost into the backtracking edge to generate a reverse interpretation graph.

[0072] S5.2: Perform a constrained shortest path search on the reverse interpretation graph, and determine the backtracking path that can continuously cover the electromechanical decoupling precursor path and has the minimum cumulative interpretation cost as the minimum interpretation path.

[0073] Specifically, the search starts from the terminal action node of the electromechanical decoupling precursor path. The forward action node is extended step by step along the backtracking edge in the reverse interpretation graph. The action stages are connected continuously from back to front, and no action stage is skipped or the same action node is repeated during the backtracking process as path constraints. The interpretation cost of the backtracking edge in each candidate backtracking path is accumulated and summarized to obtain the cumulative interpretation cost. When multiple candidate backtracking paths reach the same forward action node, the candidate backtracking path with the smaller cumulative interpretation cost is retained and continues to extend forward. Each candidate backtracking path is checked against its corresponding action nodes. The candidate backtracking path must reach the front-end action node of the electromechanical decoupling precursor path from the end action node. It must pass through each action stage contained in the electromechanical decoupling precursor path between the end action node and the front-end action node. The action nodes in adjacent action stages are directly connected by backtracking edges. Candidate backtracking paths that are missing action stages, skip action stages, or do not reach the front-end action node are excluded. The cumulative interpretation cost of the candidate backtracking paths that meet the path constraints is compared, and the backtracking path with the minimum cumulative interpretation cost is determined as the minimum interpretation path.

[0074] S5.3: Along the least interpretation path, align the signed deviations associated with each action node under consecutive circuit breaker operations of the same type according to the number of circuit breaker operations, identify the position where the signed deviations begin to drift in the same direction, take the earliest action node that maintains the same polarity conflict as the state degradation start action node, and determine the state degradation start action stage.

[0075] Specifically, the action nodes corresponding to consecutive circuit breaker actions of the same type are collected along the least interpretation path, and the same action nodes in different circuit breaker actions are arranged to the same position according to the action stage. Then, the number of circuit breaker actions is set according to the order of the circuit breaker actions. For each action node, the signed deviation associated with each number of circuit breaker actions is extracted and a deviation sequence is formed. The signed deviation of adjacent circuit breaker action counts is compared one by one. When the signed deviation from a certain number of circuit breaker actions maintains the same sign and the absolute value increases continuously with the subsequent number of circuit breaker actions, the corresponding number of circuit breaker actions is determined as the position where the signed deviation begins to drift in the same direction. Each action node is examined according to the action transmission order of the least explanation path. The action node with the earliest action stage is selected from the action nodes that maintain the same polarity conflict as the state degradation start action node, and the action stage to which the state degradation start action node belongs is determined as the state degradation start action stage.

[0076] Figure 4The horizontal axis represents the location methods for the initial action stage of degradation in different states, and the vertical axis represents the initial action stage of degradation in the actual state. The accuracy of location for the initial action stage of degradation is shown by the brightness of color blocks, where light-colored blocks indicate lower accuracy and dark-colored blocks indicate higher accuracy. The initial action stage of degradation in the actual state includes the coil current establishment stage, the mechanism release stage, the contact movement stage, and the action stabilization stage. Location methods include the maximum evidence time-series residual location method, the first non-zero evidence time-series residual location method, and the minimum interpretation path backtracking method of this invention. The maximum evidence time-series residual location method directly uses the action stage with the largest evidence time-series residual as the starting point of degradation, which easily leads to the amplification of residuals after degradation propagation. The continuation action phase was mistakenly identified as the starting action phase of state degradation; the first non-zero evidence temporal residual localization method takes the action phase with the first non-zero evidence temporal residual as the starting point of degradation, which is easily affected by occasional temporal fluctuations and local evidence conflicts; the minimum interpretation path backtracking method of this invention searches backward along the electromechanical decoupling precursor path, and sets the interpretation cost according to the consistency of the conflict polarity of adjacent action nodes. It determines the minimum interpretation path with the minimum cumulative interpretation cost from the backtracking paths that satisfy the continuous connection of action phases, and then locates the first action node that continuously generates the same polarity conflict; as can be seen from the overall color block distribution, the method of this invention maintains a high and relatively stable localization capability under different real state degradation starting action phases, therefore... Figure 4 The study verified that minimal explanation path backtracking can effectively distinguish between the degradation propagation result and the degradation initiation position, thereby improving the accuracy and reliability of the localization of the state degradation initiation action phase.

[0077] S5.4: Starting from the state degradation start node, perform path-weighted accumulation of the signed deviation associated with each action node along the minimum interpretation path to obtain the path time sequence residual accumulation value, and compare the path time sequence residual accumulation value with the preset circuit breaker health level range to evaluate the circuit breaker health status level.

[0078] Specifically, starting from the initial action node of state degradation, the signed deviations associated with each action node are extracted sequentially according to the action transmission order of the minimum interpretation path, and the path weight is obtained based on the action transmission position of each action node in the minimum interpretation path. The ratio of the absolute value of the signed deviation to the width of the action delay constraint interval associated with the corresponding action node is normalized (normalization results greater than 1 are counted as 1). The normalized absolute values ​​of the signed deviations of each action node are accumulated and summarized according to the path weight to obtain the cumulative value of the path time sequence residual. The cumulative value of the path time sequence residual is compared with the circuit breaker health level interval. When the cumulative value of the path time sequence residual is lower than the lower limit of the circuit breaker health level interval, the circuit breaker health status is rated as healthy. When the cumulative value of the path time sequence residual is not lower than the lower limit of the circuit breaker health level interval and not higher than the upper limit of the circuit breaker health level interval, the circuit breaker health status is rated as slightly degraded. When the cumulative value of the path time sequence residual is higher than the upper limit of the circuit breaker health level interval, the circuit breaker health status is rated as severely degraded.

[0079] It should be noted that the path weights are set according to the action transmission position of the action node in the minimum interpretation path. The state degradation start action node to the end action node are assigned an incrementing number in sequence, and the normalization is performed by using the proportion of each incrementing number to the sum of all incrementing numbers, so that the path weights closer to the end action node are larger and the sum of all path weights is 1. The circuit breaker health level range is set based on the distribution of cumulative path timing residual values ​​under normal operating conditions of the same type of circuit breaker. The cumulative path timing residual values ​​are extracted from historical normal operation samples. After removing outliers that exceed the interquartile range, the 95th percentile of the remaining cumulative path timing residual values ​​is taken as the lower limit of the circuit breaker health level range, and the 99th percentile is taken as the upper limit of the circuit breaker health level range. An example circuit breaker health level range is 0.25~0.75.

[0080] In summary, this invention achieves causal connection between electrical action evidence and mechanical response evidence by generating a circuit breaker action causal graph and organizing multi-source action events according to partial order relationships and action delay constraints, thus providing a unified analytical basis for identifying cross-source evidence conflicts. By generating electromechanical decoupling precursor paths and implementing minimum interpretation path backtracking, it achieves early latent performance degradation identification and the location of the initial action stage of state degradation, thereby improving the reliability of equipment state management throughout its entire life cycle.

[0081] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A circuit breaker condition diagnosis method based on multi-source timing fusion, characterized in that, include: The control voltage and coil current in the multi-source operation timing data of the circuit breaker are segmented and changed to identify electrical operation events. The response waveform generated by the operation of the circuit breaker mechanism is detected on multiple scales to identify mechanical response events. The events are then time-series aggregated according to their occurrence time to generate a multi-source operation event set. Based on the action mechanism of the circuit breaker's electrical drive to mechanical execution, the multi-source action event set is projected to the preset action stage, a partial order constraint action event matching diagram is constructed, and events that satisfy the partial order relationship and action delay constraint are selected as action nodes, connected according to the action transmission relationship, and a circuit breaker action cause-effect diagram is generated. Evidence attributes are analyzed on the cause-and-effect diagram of circuit breaker action to extract electrical action evidence and mechanical response evidence. The time residual between the actual time interval of the two types of evidence and the action delay constraint is calculated. The polarity of evidence conflict is determined according to the direction of the time residual to form cross-source evidence conflict relationship. The cause-and-effect relationship of cross-source evidence conflict is traced along the action transmission direction of the circuit breaker action cause-and-effect diagram. Cross-source evidence conflict relationships with consistent polarity and continuous accumulation of temporal residuals are aggregated into decoupling propagation chains. The stable starting point is determined by residual drift verification of continuous similar actions, and electromechanical decoupling precursor paths are generated. The minimum interpretation path backtracking is performed on the electromechanical decoupling precursor path to search in reverse for the first action node that continuously generates the same polarity conflict, locate the initial action stage of state degradation, and assess the health status level of the circuit breaker based on the degree of accumulation of path timing residuals.

2. The circuit breaker condition diagnosis method based on multi-source timing fusion as described in claim 1, characterized in that, The steps for generating the multi-source action event set are as follows: Using the control voltage change time in the multi-source operation timing data of the circuit breaker as the electrical test benchmark, a sliding window cumulative deviation analysis is performed on the control voltage and coil current to divide the operation segment, and the boundary of the operation segment is corrected by the waveform turning position to generate electrical operation events. Based on the detection time window of the mechanism action response defined by the electrical action event, the response waveform generated by the circuit breaker mechanism action within the detection time window is subjected to multi-scale envelope energy tracking, and the response mutation position that is stably reproduced across scales is screened to generate mechanical response events. The mechanical response events are mapped to the time axis of the electrical action events. Cross-source event pairings are established within the detection time window according to the principle of minimum event time difference, and unpaired events and event source identifiers are retained. Multi-source action event sets are generated by merging events according to their occurrence time.

3. The circuit breaker condition diagnosis method based on multi-source timing fusion as described in claim 1 or 2, characterized in that, The steps for constructing the action-event matching graph with partial order constraints are as follows: Based on the action mechanism of the circuit breaker's electrical drive to mechanical execution, an action stage template is established. A dynamic time warping algorithm is used to align the multi-source action event set with the action stage template in order constraint. The action stage to which each action event belongs is determined according to the minimum cumulative time sequence cost, and an action stage projection sequence is generated. Based on the sequential dependencies between action stages, the projected sequence of action stages is constructed into a directed acyclic candidate graph, and the absolute deviation between the event occurrence time difference and the center value of the action delay constraint interval is used as the connection weight to generate a partially ordered action event matching graph.

4. The circuit breaker condition diagnosis method based on multi-source timing fusion as described in claim 3, characterized in that, The steps for generating the cause-effect graph of the circuit breaker operation are as follows: Dynamic programming path search is performed on the action event matching graph with partial order constraints. Candidate connection edges that do not satisfy the partial order relationship and action delay constraints are eliminated. The optimal action path is determined according to the principle of minimum cumulative connection weight, and action nodes and action transmission edges are obtained. Connect the action nodes and action transmission edges according to the transmission direction from electrical drive to mechanical execution, and write the actual action delay into the action transmission edge to generate a cause-effect graph of circuit breaker action.

5. The circuit breaker condition diagnosis method based on multi-source timing fusion as described in claim 1, characterized in that, The steps for extracting electrical action evidence and mechanical response evidence are as follows: Evidence attributes of each action node are analyzed along the action transmission direction of the circuit breaker action cause-effect diagram. The driving side action nodes are classified into electrical action evidence, and the execution side response nodes are classified into mechanical response evidence, generating a hierarchical evidence node sequence. Based on the stage connection relationship of action nodes in the hierarchical evidence node sequence, causal adjacency pairing is performed on electrical action evidence and mechanical response evidence, and action delay constraints that match the action transmission edges connecting the two types of evidence are obtained to generate a cross-source evidence matching set.

6. The circuit breaker condition diagnosis method based on multi-source timing fusion as described in claim 5, characterized in that, The steps to establish a cross-source evidence conflict relationship are as follows: Directed difference is performed on the evidence occurrence time in the cross-source evidence matching pair set to obtain the actual time interval. The actual time interval is compared with the action delay constraint at the interval boundary. For the actual time interval that exceeds the action delay constraint, the signed deviation is calculated. For the actual time interval that is within the action delay constraint, the deviation is set to zero. The evidence time residual of each cross-source evidence matching pair is obtained. Cross-source evidence matching pairs with non-zero temporal residuals are selected, the polarity of evidence conflict is determined according to the sign direction of the temporal residuals, and the degree of conflict is characterized by the residual amplitude, thus forming cross-source evidence conflict relationships.

7. The circuit breaker condition diagnosis method based on multi-source timing fusion as described in claim 6, characterized in that, The steps for aggregating cross-source evidence conflict relationships with consistent polarity and continuously accumulating temporal residuals into a decoupled propagation chain are as follows: Using cross-source evidence conflict relationships as the starting point for tracing, a directed graph traversal is performed along the action transmission direction of the circuit breaker action causal graph to associate cross-source evidence conflict relationships with the action nodes to which the mechanical response evidence belongs, and cross-source evidence conflict relationships with causal reachability are connected according to action stages to generate a conflict causal tracing chain. A polarity consistency test is performed on the conflict relationship between adjacent cross-source evidence in the conflict causal tracing chain. Conflict causal tracing chains with consistent evidence conflict polarity and whose absolute value of the signed deviation increases step by step along the action transmission direction are selected. The cumulative residual of the link is calculated according to the same-direction increment of the signed deviation to generate the decoupled propagation chain.

8. The circuit breaker condition diagnosis method based on multi-source timing fusion as described in claim 7, characterized in that, The steps for generating the precursor path of electromechanical decoupling are as follows: The decoupling propagation chains corresponding to the continuous operation of the same type of circuit breaker are aligned according to the operation stage. The sliding window least squares method is used to fit the residual drift slope of the signed deviation associated with each operation node as the number of circuit breaker operations changes. The foremost action node with the consistent sign of the residual drift slope and the smallest slope dispersion is selected as the stable starting point. The same polarity residual extension path is retained along the decoupling propagation chain from the stable starting point to generate the electromechanical decoupling precursor path.

9. The circuit breaker condition diagnosis method based on multi-source timing fusion as described in claim 8, characterized in that, The steps of the initial action phase of the positioning state degradation are as follows: Using the terminal action node of the electromechanical decoupling precursor path as the backtracking starting point, the backtracking is performed along the action transmission edge, and the interpretation cost is set for the backtracking edge according to the consistency of the conflict polarity of adjacent action nodes to generate a reverse interpretation graph. Perform a constrained shortest path search on the reverse interpretation graph, and determine the backtracking path that can continuously cover the precursor paths of electromechanical decoupling and has the minimum cumulative interpretation cost as the minimum interpretation path; Along the least interpretation path, the signed deviations associated with each action node under consecutive circuit breaker operations of the same type are aligned according to the number of circuit breaker operations. The position where the signed deviations begin to drift in the same direction is identified. The action node that first maintains the same polarity conflict is taken as the state degradation initiation action node, and the state degradation initiation action stage is determined.

10. The circuit breaker condition diagnosis method based on multi-source timing fusion as described in claim 9, characterized in that, The method of assessing the health status level of a circuit breaker based on the cumulative degree of path timing residuals refers to performing path-weighted accumulation of the signed deviations associated with each action node along the minimum interpretation path from the starting point of state degradation, obtaining the cumulative value of path timing residuals, and comparing the cumulative value of path timing residuals with a preset circuit breaker health status range to assess the health status level of the circuit breaker.

Images