Power distribution terminal device exception checking and responding method and system based on edge computing

Abstract

This invention discloses a method and system for anomaly verification and response of power distribution terminal equipment based on edge computing, belonging to the field of power distribution terminal status monitoring technology. The method acquires the operating status data of the power distribution terminal equipment and triggers abnormal events based on preset rules; extracts the temporal evolution trajectory of abnormal parameters to generate anomaly formation features; retrieves associated status data to generate physical closed-loop results; extracts the temporal disturbance features and sampling discrete features of the acquisition link before and after the anomaly occurs to generate link offset results; performs hierarchical correlation analysis on the anomaly formation features, physical closed-loop results, and link offset results to generate anomaly authenticity results, corrects the rule credibility level, further determines the validity of the authenticity verification of existing edge rules, and performs delayed response processing in cases of invalidity. This invention can effectively distinguish between real faults and pollution anomalies, reduce the probability of false action by edge rules, and improve the reliability of power distribution terminal equipment status monitoring.

Description

Technical Field

[0001] This invention relates to the field of power distribution terminal status monitoring technology, specifically to a method and system for anomaly verification and response of power distribution terminal equipment based on edge computing. Background Technology

[0002] As the intelligence level of power distribution networks continues to improve, more and more power distribution terminal equipment is adopting edge computing to monitor its operational status. In existing technologies, edge nodes typically collect real-time data on the operational status of power distribution terminal equipment, such as voltage, current, temperature, vibration, and partial discharge. Abnormal states are then identified based on preset fault rules to achieve fault alarms, status assessments, and coordinated control. Compared to traditional centralized cloud processing, edge computing reduces data transmission latency and improves on-site response efficiency, and therefore has been widely applied in smart power distribution scenarios.

[0003] However, most existing edge rule monitoring methods rely on fixed thresholds, empirical rules, or single parameter exceedance judgments. Essentially, they only identify abnormal results and lack the ability to analyze the authenticity of the abnormal data formation process. In complex power distribution environments, sensors are susceptible to strong electromagnetic interference, communication fluctuations, instantaneous pulse noise, or abnormal data injection, leading to distorted data collected by edge nodes. Since existing technologies typically assume the collected data is accurate and reliable, when abnormal data meets preset rules, the edge rule system can easily trigger alarms, trips, or shutdowns, resulting in false alarms, malfunctions, or even regional power supply anomalies.

[0004] Furthermore, existing technologies for verifying abnormal data are mostly limited to simple filtering, redundancy comparison, or fixed logic verification. They lack the comprehensive analytical capabilities to consider the temporal evolution of abnormal states, the energy coupling relationships of associated states, and the discrete disturbance characteristics of the acquisition link, making it difficult to effectively distinguish between real faults and pollution anomalies. When power distribution terminal equipment is in a situation involving new energy access, high-frequency load fluctuations, or complex electromagnetic environments, traditional rule-based systems are more prone to authenticity verification failures, thereby affecting the stability and security of the power distribution system. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for anomaly verification and response of power distribution terminal equipment based on edge computing, so as to solve the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for anomaly verification and response of power distribution terminal equipment based on edge computing, comprising: Acquire the operating status data of the power distribution terminal equipment and trigger abnormal events based on preset rules; Extract the temporal evolution trajectory of the abnormal parameters corresponding to the abnormal event, and generate abnormal formation features based on the continuity of change, change inertia and evolution convergence between adjacent sampling intervals; Based on the anomaly formation characteristics, retrieve the associated state data that has an energy coupling relationship with the anomaly parameters, and analyze the consistency of synchronous response and propagation association direction among the associated state data to generate physical closed-loop results. Based on the data source information corresponding to the abnormal event, extract the temporal disturbance characteristics and sampling discrete characteristics of the acquisition link before and after the abnormality occurs, and generate the link offset result; The anomaly formation characteristics, physical closed-loop results, and link offset results are analyzed in a hierarchical manner to generate anomaly authenticity results, and the credibility level of the corresponding rules is corrected based on the anomaly authenticity results. Obtain the execution results of the existing edge rule system for abnormal events, and determine the validity of the authenticity verification of the existing edge rule system by combining the revised rule credibility level; When the authenticity verification is invalid, a delayed response is performed on the corresponding abnormal event.

[0007] Preferably, the system acquires the operating status data of the power distribution terminal equipment and triggers abnormal events based on preset rules, including: The running state data within a continuous sampling period is reconstructed in time to obtain the interval representative value of each sampling interval, and the interval representative value forms the state change trajectory. Based on the state change trajectory, the perturbation continuous features between adjacent sampling intervals are extracted, and the perturbation continuous features are checked for directional consistency. Based on the verified continuous perturbation features, state transition features are generated. When the state transition features meet the gradual triggering conditions in the preset rules, abnormal event information is generated.

[0008] Preferably, the temporal evolution trajectory of the abnormal parameters corresponding to the abnormal event is extracted, including: Based on the abnormal parameter name and abnormal trigger time in the abnormal event information, continuous sampling data within a preset traceability time range is retrieved and arranged progressively according to the sampling interval to form a time-series evolution trajectory. Based on the temporal evolution trajectory, the change continuation state and change stagnation state are extracted, and inertial extension features are generated according to the degree of preservation of the change direction in continuous sampling intervals. The degree of consistency of the offset of the inertial extension feature is analyzed to obtain the convergent extension state of the abnormal parameters in the time-series evolution process.

[0009] Preferably, the generated anomaly formation features include: When the convergence expansion state is established, the change continuation state is established, and the change stagnation state does not have a swingback, the anomaly formation characteristic is determined to be continuous gradual change formation; When the convergence extension state is established and the change stagnation state has a swingback, the anomaly formation characteristic is determined to be the formation of a disturbed gradual change. When the convergence extension state does not hold and the change continuation state does not hold, the abnormal formation characteristic is identified as discontinuous mutation formation.

[0010] Preferably, generating a physical closed-loop result includes: Based on the abnormal parameter name and abnormal formation characteristics, the state diffusion path is determined, and the associated state data that have an energy coupling relationship with the abnormal parameter are retrieved along the state diffusion path. The temporal mapping is performed according to the starting position of the change of each associated state data in the continuous sampling interval to form a state response chain. The starting positions of the changes of adjacent associated state data in the state response chain are progressively compared to generate a state transmission sequence. Based on the state transmission sequence, the direction of change between adjacent associated state data is extracted, and the cumulative state of offset caused by reverse offset and transmission loss is determined; When the offset accumulation state forms a reverse abrupt change segment in the continuous sampling interval, and the continuous expansion range of the reverse abrupt change segment exceeds the preset continuous interval, it is determined that there is a reverse interruption region. When there is no reverse interruption region and the state changes are continuously propagated, a physical closed-loop result is generated that establishes the closed loop.

[0011] Preferably, the generated link offset result includes: Based on the data source information in the abnormal event information, extract the data arrival order and sampling interval change status of the continuous sampling interval before and after the abnormality, and form the link time sequence trajectory; Based on the link time sequence trajectory, data stagnation sections, sudden aggregation sections, and the number of sequence anomalies are determined to obtain time sequence disturbance characteristics; The discrete distribution of sampled data within each sampling interval is progressively compared to obtain the discrete sampling features. Link offset results are then generated based on the temporal disturbance features and the temporal overlap of the discrete sampling features.

[0012] Preferably, generating anomaly authenticity results and correcting the credibility level of corresponding rules includes: determining the degree of continuous expansion of the state change corresponding to the anomaly event based on the anomaly formation characteristics; determining the degree of transmission integrity of the associated state data based on the physical closed-loop results; determining the offset interference range of the data source based on the link offset results; determining the authenticity maintenance state in the progressive order of continuous expansion, transmission integrity, and offset interference range; increasing the credibility level of the corresponding rule when the authenticity maintenance state continuously strengthens; decreasing the credibility level of the corresponding rule when the authenticity maintenance state continuously weakens; and generating anomaly authenticity results.

[0013] Preferably, determining the validity of the authenticity verification of the existing edge rule processing flow includes: Obtain the execution status and execution sequence corresponding to abnormal events to form a rule execution trajectory; Based on the revised rule credibility level analysis, the execution continuation state and execution rollback state in the rule execution trajectory are analyzed to determine the degree of credibility matching; When there is a reverse execution segment in the rule execution trajectory that deviates from the abnormal authenticity result, and the reverse execution segment continues to expand in the continuous execution interval, the authenticity verification is determined to be invalid.

[0014] Preferably, delayed response processing is performed on the corresponding abnormal event, including: After the authenticity verification is invalid, the execution status of the corresponding lockout, trip, or shutdown of the abnormal event is suspended, and the change and expansion trajectory of the abnormal event is continuously recorded within the preset delay interval. Based on the trajectory of change, determine the degree of continuation and retraction of the abnormal parameters, and determine the true stable state accordingly; When the associated status data has a synchronous decay segment within the preset delay interval, the response restriction state of the corresponding abnormal event is released; when there is no synchronous decay segment, the delayed response processing state of the corresponding abnormal event is maintained.

[0015] This invention also provides an edge computing-based power distribution terminal equipment anomaly verification and response system, including: Data acquisition module: acquires the operating status data of power distribution terminal equipment and triggers abnormal events based on preset rules; Temporal evolution analysis module: Extracts the temporal evolution trajectory of abnormal parameters corresponding to abnormal events, and generates abnormal formation characteristics based on the continuity of change, change inertia and evolution convergence between adjacent sampling intervals; Physical closed-loop analysis module: Based on the anomaly formation characteristics, retrieve the associated state data that has an energy coupling relationship with the anomaly parameters, and analyze the consistency of synchronous response and propagation association direction among the associated state data to generate physical closed-loop results; Link offset analysis module: Extracts the temporal disturbance characteristics and sampling discrete characteristics of the acquisition link before and after the occurrence of the anomaly based on the data source information corresponding to the anomaly event, and generates link offset results; The authenticity hierarchical correlation module performs hierarchical correlation analysis on anomaly formation characteristics, physical closed-loop results, and link offset results to generate anomaly authenticity results, and corrects the credibility level of the corresponding rules based on the anomaly authenticity results. Verification validity determination module: Obtain the execution results of the existing edge rule system for abnormal events, and determine the authenticity verification validity of the existing edge rule system in combination with the corrected rule credibility level; Delayed response processing module: When the authenticity verification is invalid, a delayed response processing is performed on the corresponding abnormal event.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention departs from relying solely on a single threshold or fixed rules to directly determine abnormal data. Instead, it verifies the authenticity of anomalies layer by layer from three dimensions: the anomaly formation process, the physical transmission process, and the data acquisition process. Specifically, by extracting the continuity, inertia, and convergence of changes in abnormal parameters within a continuous sampling interval, it can identify whether abnormal data conforms to the natural state migration patterns of equipment. Furthermore, by combining the synchronous response relationship and propagation direction between associated state data, it determines whether the anomaly has a basis in real physical energy transfer. Simultaneously, it further verifies whether there is contamination interference in the data source by combining the temporal disturbance characteristics and sampling discrete characteristics in the acquisition link. Compared to existing technologies that trigger alarms or actions based solely on instantaneous abnormal values, this invention can effectively distinguish between real faults and instantaneous pulses, communication disturbances, and forged data, fundamentally reducing the probability of false anomaly judgments.

[0017] This invention further employs a technique of dynamically correcting rule credibility levels and handling delayed responses to invalidity issues to achieve adaptive authenticity verification during the execution of edge rules. Unlike existing technologies where rules are fixed for extended periods and tripping or shutting down immediately upon anomaly triggering, this invention, upon detecting invalidity in the authenticity verification, first maintains continuous monitoring of the abnormal event and then, based on the synchronous decay of subsequent associated state data, determines whether the anomaly still has a physical basis for expansion before deciding whether to lift the response restriction. Because it introduces a joint verification process involving the anomaly change trajectory and synchronous decay segments, it avoids data contamination that could induce erroneous actions in edge rules without affecting continuous monitoring of actual faults, thereby improving the operational stability of distribution terminal equipment and the reliability of edge monitoring. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the method of the present invention.

[0020] Figure 2 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1, please refer to Figure 1 As shown in this embodiment, the anomaly verification and response method for power distribution terminal equipment based on edge computing includes the following steps: Acquire the operating status data of the power distribution terminal equipment and trigger abnormal events based on preset rules.

[0023] In this invention, the edge computing side acquires the operating status data of the power distribution terminal equipment according to a preset sampling period. The operating status data includes at least one of voltage, current, temperature, vibration, partial discharge, and switch status. The sampling period is preferably 1 second. When the monitored object is a temperature parameter, the sampling period can be set to 5 seconds; when the monitored object is a current or voltage parameter, the sampling period can be set to 0.2 seconds. The edge computing side uses 10 consecutive sampling periods as a sampling interval and 6 consecutive sampling intervals as a trigger judgment window, ensuring sufficient continuous data support before an abnormal event is triggered.

[0024] It should be noted that in this invention, the threshold parameters are not fixed constants, but are dynamically determined based on the operating characteristics of the power distribution terminal equipment, the sampling period, the rated parameters of the equipment, and historical operating data.

[0025] After obtaining the operational status data, the data within the continuous sampling period is first reconstructed temporally. Temporal reconstruction refers to arranging the data of the same operational parameter in chronological order of sampling time and extracting representative values ​​from the data within each sampling interval. The representative value is obtained by using the arithmetic mean of all sampled values ​​within the interval, that is, adding all sampled values ​​within the interval and dividing by the number of sampled values ​​to obtain the representative value of the interval. The representative values ​​of multiple sampling intervals are arranged in chronological order to form a state change trajectory. Through this processing, it is possible to avoid a single sampling point directly participating in rule triggering and reduce the impact of instantaneous pulse data on anomaly identification.

[0026] After obtaining the state change trajectory, the perturbation continuity features between adjacent sampling intervals are extracted. These features include the interval change amount, the interval change direction, and the interval change hold state. The interval change amount is the value obtained by subtracting the representative value of the previous sampling interval from the representative value of the subsequent sampling interval. The interval change direction is determined based on the interval change amount: an interval change amount greater than 0 indicates an upward direction, an interval change amount less than 0 indicates a downward direction, and an interval change amount equal to 0 indicates a hold direction. To avoid misjudgment of direction due to zero-point fluctuations, a direction dead zone threshold is set. This threshold is used to filter out measurement noise and sampling fluctuations. It is determined as follows: Dth = max(3σ, Dn × 0.3%); where σ is the standard deviation of the sampling data during normal operation; Dn is the rated range of the corresponding parameter. The 3σ principle is used when the equipment noise level is high; the rated range ratio principle is used when the equipment is operating stably. The 0.5% in this example is only a preferred value; when the absolute value of the interval change amount does not exceed the direction dead zone threshold, the interval change direction is recorded as a hold direction.

[0027] Directional consistency verification is used to determine whether the perturbation continuity feature originates from continuous state changes. The edge computing side reads the interval change direction of five consecutive adjacent sampling intervals within the judgment window. If at least four of these intervals have the same change direction, and there is no single interval change exceeding three times the average of the changes of the two adjacent intervals, the perturbation continuity feature passes the directional consistency verification. If the interval change of a certain sampling interval exceeds three times the average of the changes of the two adjacent intervals, and the change directions of the intervals before and after that sampling interval are opposite, then the data corresponding to that sampling interval is identified as instantaneous isolated jump data and removed from the state change trajectory. After removal, the average of the representative values ​​of the preceding and following sampling intervals is used to fill in the gaps, resulting in the verified perturbation continuity feature.

[0028] Subsequently, state transition features are generated based on the verified perturbation continuity features. The state transition features include the transition direction, the number of transition duration intervals, and the cumulative transition magnitude. The transition direction is determined by the dominant change direction after passing the direction consistency verification; the number of transition duration intervals is the number of sampling intervals where the dominant change direction is continuously maintained. The migration duration interval number is used to determine whether the anomaly has a continuous evolution trend. It is determined as: N = Tresponse / Tsample; where: Tresponse is the shortest response time for fault formation; Tsample is the sampling period. Preferably, N ≥ 60% × (Tresponse / Tsample); in this example, four sampling intervals are used only as an optimal value for a temperature monitoring scenario. The cumulative migration amplitude is the value obtained by summing the changes in each interval within the judgment window according to the dominant change direction.

[0029] The gradual triggering conditions include direction conditions, duration conditions, and amplitude conditions. The direction condition requires that the migration direction in the state transition feature is consistent with the fault direction corresponding to the preset rule; the duration condition requires that the number of migration duration intervals reach the preset number of duration intervals, preferably 4 sampling intervals; the amplitude condition requires that the cumulative migration amplitude reach 10% to 30% of the corresponding warning threshold, preferably 15% for temperature parameters and 20% for current parameters. When the state transition feature simultaneously satisfies the direction condition, duration condition, and amplitude condition, the state change is determined not to be an instantaneous, isolated jump, but rather possesses continuous gradual change attributes, thereby generating corresponding abnormal event information.

[0030] The warning threshold is determined based on the rated operating range and historical normal operating data of the power distribution terminal equipment. Its calculation method is: W = α × (Lmax - Lavg); where W is the warning threshold, Lmax is the rated upper limit of the corresponding monitoring parameter, Lavg is the average normal operating value within the most recent preset time period without any abnormal events, and α is the warning coefficient, which can be taken as 0.5 to 0.8 depending on the sensitivity requirements of the monitored object. The average normal operating value is preferably obtained by averaging the operating status data without any abnormal events within the most recent 24 hours; when there is insufficient valid normal data within 24 hours, it can be extended to valid normal operating data within the most recent 72 hours.

[0031] The cumulative migration amplitude threshold is: A = β × W; where A is the cumulative migration amplitude required to enter the gradual triggering condition matching, β is the gradual triggering coefficient, and β can be taken as 0.1 to 0.3 according to the false triggering suppression requirements.

[0032] In a preferred embodiment, α is set to 0.7 and β to 0.1, meaning the warning threshold is preferably 70% of the difference between the rated upper limit and the normal operating average. When the cumulative migration amplitude is not less than 10% of the corresponding warning threshold, the gradual trigger condition matching is allowed. These values ​​are used to filter out single-point noise and short-term disturbances while ensuring early identification capability for abnormal gradual changes, and are not intended to limit the scope of protection of this invention.

[0033] The abnormal event information includes the abnormal parameter name, abnormal trigger time, state change trajectory, verified continuous perturbation characteristics, state transition characteristics, matched gradual triggering conditions, and corresponding distribution terminal equipment number. This abnormal event information serves as the input for subsequent extraction of the abnormal parameter time-series evolution trajectory, generation of physical closed-loop results, and generation of link offset results. This ensures that subsequent authenticity judgments can directly follow the aforementioned triggering process, guaranteeing that the abnormal event is generated by continuous state transitions, rather than being directly triggered by single-point distorted data.

[0034] Extract the temporal evolution trajectory of the abnormal parameters corresponding to the abnormal event, and generate abnormal formation features based on the continuity of change, change inertia and evolution convergence between adjacent sampling intervals.

[0035] In this invention, after the abnormal event information is generated, the edge computing side extracts continuous sampling data of the abnormal parameters within a preset tracing time range based on the abnormal parameter name, abnormal trigger time, and state change trajectory in the abnormal event information.

[0036] The preset traceability time range is set based on the aforementioned trigger judgment window, preferably five times the length of the trigger judgment window. When the abnormal parameter is a temperature parameter, the preset traceability time range is preferably 30 minutes; when the abnormal parameter is a current or voltage parameter, the preset traceability time range is preferably 60 seconds. This setting can cover the continuous change process before the anomaly occurs, avoiding the need to judge the authenticity of the anomaly solely based on data near the anomaly trigger time.

[0037] Continuous sampling data are arranged sequentially according to sampling time, and representative values ​​for each sampling interval are obtained according to the aforementioned sampling interval division method. The representative values ​​of adjacent sampling intervals are then connected sequentially to form a time-series evolution trajectory. The time-series evolution trajectory not only includes the numerical changes of anomalous parameters but also retains the temporal position of each sampling interval, enabling subsequent determination of whether anomalous parameters have a continuous formation process.

[0038] After forming the temporal evolution trajectory, the continuity and stagnation states of change between adjacent sampling intervals are extracted. The continuity state is used to characterize whether the anomalous parameter continues to advance along the same direction of change in adjacent sampling intervals. During the determination, the difference between the representative value of the later sampling interval and the representative value of the earlier sampling interval is calculated, and the direction of change is determined based on the difference; when the absolute value of the difference does not exceed the direction dead zone threshold, the direction of change is recorded as the stabilization direction. If three or more consecutive adjacent sampling intervals have the same non-stabilization direction, the corresponding interval forms a continuity state; if fewer than three adjacent sampling intervals have the same non-stabilization direction, no continuity state is formed.

[0039] The change-dwelling state is used to characterize whether an abnormal parameter experiences a brief pause after energy accumulation before an anomaly is triggered. During the determination, the interval change value in a continuous sampling interval is read. If the interval change value of two or more consecutive sampling intervals does not exceed the directional dead zone threshold, and the change direction before and after the consecutive intervals remains consistent, the consecutive interval is determined to have a change-dwelling state. If the change direction before and after the consecutive intervals remains opposite, it is considered a swing back and is not included in subsequent processing as a change-dwelling state. This limitation allows for the differentiation between staged dwellings in a real gradual change process and repetitive jumps caused by noise.

[0040] After obtaining the states of continued change and stagnant change, an inertial tendency feature is generated based on the degree to which the change direction is maintained within continuous sampling intervals. The generation method is as follows: Statistically count the number of consecutively maintained intervals of the dominant change direction within a preset tracing time range, the longest consecutively maintained interval length, and the recovery direction after being interrupted by a stagnant change state. The inertial tendency feature is considered valid when the number of consecutively maintained intervals of the dominant change direction is no less than 70% of the total number of adjacent sampling intervals, the longest consecutively maintained interval length is no less than 4 sampling intervals, and the recovery direction after the stagnant change state is consistent with the direction before stagnant change; otherwise, the inertial tendency feature is considered invalid. The dominant change direction is the non-maintained direction that appears most frequently within the preset tracing time range.

[0041] Correlation analysis is performed on the consistency of the inertial extension characteristics across multiple consecutive sampling intervals to determine the convergence expansion state. The consistency of the offset is determined by the deviation of the change direction and the deviation of the change magnitude in each consecutive sampling interval. The change direction deviation is whether the change direction of the current sampling interval is consistent with the dominant change direction; the change magnitude deviation is the absolute value of the difference between the change amount of the current interval and the change amount of the previous interval in the same direction. If the number of sampling intervals with consistent directions is not less than 80% of the number of intervals involved in the judgment, and the change magnitude deviation does not exceed 50% of the absolute value of the change amount of the previous interval in the same direction for three consecutive times, then the abnormal parameter is determined to have formed a convergence expansion state during the time-series evolution. If two consecutive sampling intervals have opposite directions, or if the change amount of a single interval exceeds three times the average change amount of the adjacent intervals, then a convergence expansion state is not formed.

[0042] The convergence expansion state, change continuation state, and change stagnation state are combined to generate anomaly formation characteristics. The generation rules are as follows: when the convergence expansion state and the change continuation state are both valid, and the change stagnation state does not exhibit a reversal, the anomaly formation characteristic is recorded as continuous gradual change formation; when the convergence expansion state is valid but the change stagnation state exhibits a reversal, the anomaly formation characteristic is recorded as disturbed gradual change formation; when the convergence expansion state is invalid, and no change continuation state meets the conditions, the anomaly formation characteristic is recorded as discontinuous abrupt change formation. This anomaly formation characteristic is used to subsequently retrieve associated state data with energy coupling relationships and serves as the basis for determining whether the anomaly data conforms to the natural state migration law of the equipment.

[0043] Based on the anomaly formation characteristics, retrieve the associated state data that has an energy coupling relationship with the anomaly parameters, and analyze the consistency of synchronous response and propagation direction among the associated state data to generate physical closed-loop results.

[0044] In this invention, after obtaining the anomaly formation characteristics, the edge computing side determines the state diffusion path based on the anomaly parameter name and the anomaly formation characteristics. The state diffusion path is used to characterize the transmission order of the state changes corresponding to the anomaly parameter within the power distribution terminal equipment and between adjacent measurement points. The edge computing side pre-stores an energy coupling relationship table, which is established according to the energy transfer relationship between the anomaly parameter and the associated state data. When the anomaly parameter is temperature, the associated state data preferentially includes current, load rate, and adjacent temperatures; when the anomaly parameter is current, the associated state data preferentially includes voltage, power factor, and temperature; when the anomaly parameter is vibration, the associated state data preferentially includes switch action time, contact temperature, and adjacent vibration. When the anomaly formation characteristics are formed by continuous gradual change, the state diffusion path is determined according to the order of energy input data, energy response data, and spatial accompanying data; when the anomaly formation characteristics are formed by disturbance gradual change, spatial accompanying data is retained in this order; when the anomaly formation characteristics are formed by discontinuous abrupt change, only the associated state data within the same sampling period is retrieved as the basis for subsequent reverse interruption analysis.

[0045] After retrieving associated state data along the state diffusion path, representative values ​​are extracted for each type of associated state data according to the sampling interval, and the change between adjacent sampling intervals is calculated. The change is the value obtained by subtracting the representative value of the previous sampling interval from the representative value of the later sampling interval. To determine whether the associated state data generates a synchronous response to the abnormal parameters, a response threshold is set, preferably 0.8% of the rated range of the associated state data. When the absolute value of the change in the associated state data in a certain sampling interval reaches the response threshold twice consecutively, and the direction of change is consistent with the preset response direction in the energy coupling relationship table, the sampling interval that first reaches the response threshold is recorded as the change start position. If there is no change start position within 6 sampling intervals before and after the abnormal trigger time, the associated state data is recorded as no-response data.

[0046] On the edge computing side, time-series mapping is performed according to the chronological order of the starting positions of changes in each associated state data, forming a state response chain. Each node in the state response chain corresponds to a type of associated state data, and the node content includes the associated state data name, the starting position of the change, the direction of change, and the number of continuous response intervals. The number of continuous response intervals is the number of consecutive sampling intervals where the direction of change remains consistent and the absolute value of the change reaches the response threshold. If two associated state data have the same starting position of change, they are arranged according to the order of energy transfer in the energy coupling relationship table; if a certain associated state data is non-responsive, it is placed at the end of the state response chain and marked as having a missing response.

[0047] Subsequently, a progressive comparison is performed on the starting positions of changes in adjacent associated state data within the state response chain to form a state transmission sequence. The progressive comparison process is as follows: the starting positions of changes in the preceding and following associated state data are read, and the number of sampling intervals between them is calculated. When the number of sampling intervals is 0 to 3, it is determined that the two have a synchronous transmission relationship; when the number of sampling intervals is 4 to 6, it is determined that the two have a delayed transmission relationship; when the number of sampling intervals exceeds 6, or the following associated state data has a missing response, it is determined that the transmission between the two is broken. The aforementioned interval ranges of 0 to 3 and 4 to 6 can be adjusted according to the sampling period, preferably using the integer obtained by dividing the physical response time of the abnormal parameter by the sampling period as the upper limit.

[0048] After the state propagation sequence is formed, the direction of change continuation between adjacent associated state data is extracted. The direction of change continuation is determined by the combination of the direction of change of the previous associated state data and the direction of change of the next associated state data; when both conform to the preset response direction in the energy coupling relationship table, the direction of change continuation is recorded as positive continuation; when the direction of change of the next associated state data is opposite to the preset response direction, the direction of change continuation is recorded as reverse offset; when the next associated state data has no change start position, the direction of change continuation is recorded as propagation missing. The offset accumulation state is determined by the number of occurrences of reverse offset and propagation missing within a continuous sampling interval. Reverse offset is recorded as 1 offset, and propagation missing is recorded as 2 offsets; when the cumulative number of offsets reaches 3 within 3 consecutive sampling intervals, the offset accumulation state is determined to be valid.

[0049] The offset cumulative state analysis determines whether a reverse abrupt change segment exists in the state response chain. A reverse abrupt change segment is a continuous sampling interval that continues in the opposite direction to the preceding change and exhibits a persistent reverse offset or transmission gap. The continuous expansion range is determined by the number of intervals between the starting and ending sampling intervals of the reverse abrupt change segment. The preset continuous interval is preferably 3 sampling intervals; when the continuous expansion range exceeds 3 sampling intervals, it is determined that the state change corresponding to the abnormal parameter has a reverse interruption region. If the continuous expansion range does not exceed 3 sampling intervals, and subsequent sampling intervals resume positive continuation, then the reverse abrupt change segment is recorded as a short-term disturbance and is not considered a reverse interruption region.

[0050] When there is no reverse interruption region in the state response chain, and there is at least a synchronous or delayed transmission relationship between adjacent associated state data, the physical closed-loop result generated by the edge computing side is considered a valid closed loop. When there is a short-term disturbance but the reverse interruption region determination condition is not met, the physical closed-loop result is considered a disturbed closed loop. When there is a reverse interruption region, or when two consecutive adjacent nodes in the state response chain are both disconnected, the physical closed-loop result is considered a closed loop interrupted. This physical closed-loop result is used for subsequent hierarchical correlation analysis with link offset results and anomaly formation characteristics to determine whether the state changes corresponding to the anomaly parameters have a real physical transmission basis.

[0051] Based on the data source information corresponding to the abnormal event, extract the temporal disturbance characteristics and sampling discrete characteristics of the acquisition link before and after the abnormality occurs, and generate the link offset result.

[0052] In this invention, after generating abnormal event information, the edge computing side extracts the data arrival order and sampling interval changes within consecutive sampling intervals before and after the abnormality based on the data source information in the abnormal event information. The data source information includes the acquisition terminal number, communication channel number, data arrival time, sampling time, data sequence number, and abnormality trigger time. Preferably, the consecutive sampling intervals before and after the abnormality trigger time are the 6 sampling intervals before and after the abnormality trigger time; when the sampling period is 1 second, the analysis range is 60 seconds before and after the abnormality trigger time.

[0053] The link timing trajectory is formed by the data arrival order and the sampling interval change status. The data arrival order is arranged according to the data sequence number and the data arrival time. When the subsequent data sequence number is equal to the previous data sequence number plus 1, and the subsequent data arrival time is later than the previous data arrival time, the order is considered continuous; when the data sequence number is missing, duplicated, or the arrival time is reversed, the order is considered abnormal. The sampling interval change status is obtained by comparing the actual arrival interval with the rated sampling period. The actual arrival interval is the current data arrival time minus the previous data arrival time, and the interval deviation is the absolute value of the actual arrival interval minus the rated sampling period. The interval deviation threshold is preferably 20% of the rated sampling period; when the interval deviation exceeds the interval deviation threshold, the sampling interval is recorded as a disturbance interval.

[0054] The edge computing side analyzes data stagnation and burst clustering segments based on link timing trajectories. A data stagnation segment is a period of continuous data arrival without any data; a stagnation segment is defined as a time when the period without data arrival is three times the rated sampling period. A burst clustering segment is a period of concentrated arrival of multiple data points within a short time; a burst clustering segment is defined as a time when three or more data points arrive within one rated sampling period, and their corresponding sequence numbers are consecutive. Timing disturbance characteristics include the number of data stagnation segments, the number of burst clustering segments, the number of disturbance intervals, and the number of sequence anomalies. Preferably, within 12 consecutive sampling intervals, if the number of data stagnation segments reaches two, or the number of burst clustering segments reaches two, or the number of sequence anomalies reaches one, the timing disturbance characteristics are considered valid.

[0055] Subsequently, a progressive comparison is performed on the discrete distribution of the sampled data before and after the anomaly. Within each sampling interval, the sample mean is first calculated by summing all sampled values ​​within the interval and dividing by the number of sampled values. Then, the discrete value is calculated by summing the absolute values ​​of the differences between each sampled value and the sample mean and dividing by the number of sampled values. The degree of discrete diffusion in adjacent sampling intervals is the discrete value of the subsequent interval minus the discrete value of the preceding interval. Discrete diffusion is considered valid when the degree of discrete diffusion is positive and two consecutive sampling intervals exceed the discrete threshold. The degree of discrete shrinkage is the discrete value of the preceding interval minus the discrete value of the subsequent interval. Discrete shrinkage is considered valid when the degree of discrete shrinkage is positive and two consecutive sampling intervals exceed the discrete threshold. The discrete threshold is preferably 1% of the rated range of the anomaly parameter, or it can be twice the average discrete value of data from the 24 hours prior to the anomaly trigger.

[0056] The discrete characteristics of the sampling are jointly determined by the number of times discrete diffusion occurs, the number of times discrete contraction occurs, and whether discrete contraction occurs immediately after discrete diffusion. When discrete contraction occurs within 1 to 2 sampling intervals after discrete diffusion occurs, it is determined that discrete contraction exists; when discrete contraction occurs more than 2 times in consecutive sampling intervals before and after the anomaly occurs, the discrete characteristics of the sampling are recorded as abnormal discreteness.

[0057] Finally, the offset expansion state of the data source within the continuous sampling interval is analyzed based on the temporal disturbance characteristics and sampling discrete characteristics. If the temporal disturbance characteristics are valid and the sampling discrete characteristics are anomalously discrete, and there is an overlap in the sampling intervals where the two occur, the offset expansion state is recorded as a continuous offset; if only the temporal disturbance characteristics are valid or only the sampling discrete characteristics are anomalously discrete, the offset expansion state is recorded as a local offset; if neither of these conditions is met, the offset expansion state is recorded as no offset. The link offset result is generated based on the offset expansion state: continuous offset corresponds to a decrease in link reliability, local offset corresponds to a link offset result where link reliability needs verification, and no offset corresponds to a link offset result where link reliability remains unchanged. This link offset result is used for subsequent hierarchical correlation analysis with anomaly formation characteristics and physical loop closure results.

[0058] The anomaly formation characteristics, physical closed-loop results, and link offset results are analyzed in a hierarchical manner to generate anomaly authenticity results, and the credibility level of the corresponding rules is adjusted based on the anomaly authenticity results.

[0059] In this invention, after obtaining the anomaly formation characteristics, physical loop closure results, and link offset results, the edge computing side first determines the degree of sustained expansion of the state change corresponding to the anomaly event based on the anomaly formation characteristics. The degree of sustained expansion is jointly determined by the proportion of the dominant change direction, the proportion of continuous gradual change intervals, and the consistency of recovery after stagnation. The proportion of the dominant change direction is the number of adjacent sampling intervals where the dominant change direction appears divided by the total number of adjacent sampling intervals participating in the judgment; the proportion of continuous gradual change intervals is the number of sampling intervals that satisfy the change continuation state divided by the total number of sampling intervals participating in the judgment; the consistency of recovery after stagnation is the number of times the recovery direction is consistent with the direction before stagnation after the change stagnation state ends divided by the total number of change stagnation states. When all three reach 0.75, the degree of sustained expansion is recorded as Level 1 sustained; when any two reach 0.75, it is recorded as Level 2 sustained; when less than two reach 0.75, it is recorded as Level 3 sustained. When the degree of sustained expansion is Level 1 or Level 2 sustained, the corresponding physical loop closure results are retrieved for subsequent hierarchical association comparison; when the degree of sustained expansion is Level 3 sustained, the physical loop closure results are still retrieved, but the anomaly event enters the authenticity decay judgment.

[0060] After retrieving the physical loop closure results, the degree of transmission integrity of abnormal events in the associated state data is analyzed. The degree of transmission integrity is determined by the number of effective transmission nodes in the state response chain, the closed-loop status, and the reverse interruption region. The number of effective transmission nodes is the number of nodes with synchronous or delayed transmission relationships; the transmission integrity ratio is the number of effective transmission nodes divided by the total number of nodes in the state response chain. When the physical loop closure result is a closed loop, and the transmission integrity ratio is not less than 0.8, the transmission integrity is recorded as complete transmission; when the physical loop closure result is a disturbed closed loop, and the transmission integrity ratio is not less than 0.6, the transmission integrity is recorded as disturbed transmission; when the physical loop closure result is a closed loop interruption, or the transmission integrity ratio is less than 0.6, the transmission integrity is recorded as broken transmission.

[0061] The offset interference range of the data source corresponding to the abnormal event is determined by combining the link offset results. The offset interference range is determined based on the number of consecutive sampling intervals before and after the anomaly occurs. When the link offset result indicates a decrease in link reliability and the continuous offset covers more than 4 consecutive sampling intervals, the offset interference range is recorded as full-segment interference; when the link offset result indicates that the link reliability is yet to be verified, or the continuous offset covers 2 to 4 consecutive sampling intervals, the offset interference range is recorded as local interference; when the link offset result indicates that the link reliability is maintained and there is no overlap between the temporal disturbance characteristics and sampling discrete characteristics of 2 consecutive sampling intervals, the offset interference range is recorded as no continuous interference.

[0062] The hierarchical correlation comparison is performed in the order of persistence expansion degree, transmission integrity degree, and offset interference range. Persistence expansion degree is used to determine whether the anomaly has a temporal basis; transmission integrity degree is used to determine whether the anomaly has a physical transmission basis; offset interference range is used to determine whether the anomaly has a risk of data source contamination. The hierarchical relationship among the three is: first confirm the temporal basis, then confirm the physical transmission basis, and finally eliminate the risk of data source contamination. If the persistence expansion degree is Level 1 persistence, the transmission integrity degree is complete transmission, and the offset interference range is no persistent interference, then the authenticity maintenance state is recorded as persistent enhancement. If the persistence expansion degree is Level 2 persistence, the transmission integrity degree is disturbed transmission, and the offset interference range is local interference, then the authenticity maintenance state is recorded as pending stabilization. If the persistence expansion degree is Level 3 persistence, or the transmission integrity degree is broken transmission, or the offset interference range is full-segment interference, then the authenticity maintenance state is recorded as persistent decay.

[0063] The rule's credibility level is set at 5 levels, with an initial credibility level preferably of level 3. When the authenticity status is continuously enhanced twice consecutively, the credibility level of the corresponding rule is increased by 1 level, up to a maximum of level 5; when the authenticity status is continuously diminished twice consecutively, the credibility level of the corresponding rule is decreased by 1 level, down to a minimum of level 1; when the authenticity status is pending stabilization, the credibility level remains unchanged. Abnormal authenticity results are generated based on the corrected credibility level: when the credibility level is 4 or 5, the abnormal authenticity result is recorded as truly credible; when the credibility level is 3, the abnormal authenticity result is recorded as pending verification; when the credibility level is 1 or 2, the abnormal authenticity result is recorded as suspected contamination. This abnormal authenticity result is used to subsequently combine with the execution results of abnormal events in the existing edge rule system to determine the validity of authenticity verification.

[0064] Obtain the execution results of the existing edge rule system for abnormal events, and determine the validity of the authenticity verification of the existing edge rule system by combining the revised rule credibility level.

[0065] After obtaining the anomaly authenticity results and the corrected rule credibility level, the edge computing side reads the execution status and corresponding execution sequence of the existing edge rule processing flow for the same anomaly event. Execution status includes one of the following: no action, alarm, delay, lockout, trip, or shutdown; the corresponding execution sequence is the chronological order in which each execution status occurs. To facilitate progressive analysis, execution statuses are assigned levels according to action intensity: no action is recorded as level 0, alarm as level 1, delay as level 2, lockout as level 3, trip as level 4, and shutdown as level 5. The edge computing side selects six consecutive execution intervals as the judgment range, starting from the anomaly trigger time, with each execution interval preferably equal to the length of the aforementioned sampling interval.

[0066] The rule execution trajectory is formed by arranging the execution state levels and execution times sequentially within each execution interval. If multiple execution states occur within the same execution interval, the execution state with the highest action intensity is designated as the representative execution state for that interval. When the representative execution state levels of adjacent execution intervals are the same or increase, it is recorded as an execution continuation state; when the representative execution state levels of adjacent execution intervals decrease, it is recorded as an execution rollback state. Execution continuation states indicate that abnormal events are continuously accepted, while execution rollback states indicate that abnormal events are weakened or withdrawn.

[0067] The reliability matching degree is calculated based on the revised rule reliability level and the rule execution trajectory. First, the allowable execution range corresponding to the reliability level is determined: for reliability level 5, the allowable execution range is from alarm to shutdown; for reliability level 4, the allowable execution range is from alarm to trip; for reliability level 3, the allowable execution range is from alarm to delay; for reliability level 2, the allowable execution range is from no action to delay; and for reliability level 1, the allowable execution range is from no action to alarm. Then, the number of intervals representing execution states falling within the allowable execution range in the continuous execution intervals is counted. This number is divided by the total number of execution intervals involved in the judgment, and the result is used as the reliability matching degree. Preferably, when the reliability matching degree is not lower than 0.8, the execution trajectory is determined to match the revised rule reliability level; when the reliability matching degree is lower than 0.8, response convergence state analysis is initiated.

[0068] The convergence status is used to characterize whether the direction of change between the rule execution trajectory and the anomaly authenticity result is consistent. When the anomaly authenticity result is true and reliable, the execution status level in the rule execution trajectory remains or increases, which is recorded as positive convergence; when the anomaly authenticity result is suspected contamination, the execution status level in the rule execution trajectory remains within the range of no action, alarm, or delay, or regresses from a high-level action to the above range, which is recorded as positive convergence; when the anomaly authenticity result is pending verification, the execution status level in the rule execution trajectory remains within the range of alarm or delay, which is recorded as positive convergence. All other cases are recorded as deviation from convergence.

[0069] The reverse execution segment is determined based on the consecutive occurrence of deviations from convergence. If the anomaly authenticity result is suspected contamination, and the rule execution trajectory experiences blocking, tripping, or shutdown in two consecutive execution intervals, then that consecutive interval is recorded as a reverse execution segment. If the anomaly authenticity result is credible, and the rule execution trajectory remains inactive for two consecutive execution intervals, or reverts from alarm to inactive, it is also recorded as a reverse execution segment. If the anomaly authenticity result is pending verification, and the rule execution trajectory experiences tripping or shutdown in two consecutive execution intervals, it is also recorded as a reverse execution segment.

[0070] The continuous expansion range of the reverse execution segment is the number of consecutive intervals from its starting execution interval to its ending execution interval.

[0071] The preset continuous execution interval is preferably three execution intervals. When the continuous expansion range of the reverse execution segment reaches or exceeds three execution intervals, and the confidence matching degree is less than 0.8, the authenticity verification is determined to be invalid. When the continuous expansion range of the reverse execution segment is less than three execution intervals, and the subsequent two execution intervals return to positive convergence, the reverse execution segment is recorded as a short-term deviation and is not determined to be invalid. When there is no reverse execution segment, and the confidence matching degree is not less than 0.8, the authenticity verification is determined to be valid. Therefore, it is possible to identify situations where the existing edge rule processing flow continues to strengthen actions under suspected contaminated data, or continues to weaken actions under truly reliable anomalies, providing a clear basis for subsequent delayed response processing.

[0072] When the authenticity verification is invalid, a delayed response is performed on the corresponding abnormal event.

[0073] After the authenticity verification is determined to be invalid, the edge computing side suspends the high-level execution state caused by the corresponding abnormal event. High-level execution states include locking, tripping, and shutdown, while retaining alarms and data records. The preset delay interval is calculated from the time of invalidation determination, preferably 6 execution intervals; when the abnormal parameter is temperature, the preset delay interval is preferably 10 minutes; when the abnormal parameter is current or voltage, the preset delay interval is preferably 30 seconds. The setting rule for the preset delay interval is: take twice the time required to complete one physical response after the abnormal parameter reaches the warning threshold, and round up to obtain the complete execution interval.

[0074] Within a preset delay interval, the edge computing side continuously collects abnormal parameters and forms a change and expansion trajectory according to the sampling intervals. The change and expansion trajectory is formed by arranging the representative value of each sampling interval, the amount of change in the interval, and the direction of change in chronological order. The amount of change in the interval is the value obtained by subtracting the representative value of the previous sampling interval from the representative value of the next sampling interval; when the absolute value of the amount of change in the interval does not exceed 0.5% of the rated range of the abnormal parameter, the direction of change is recorded as the holding direction; when it exceeds this value, the upward or downward direction is determined by positive or negative.

[0075] The continuation and regression degrees of the abnormal event are analyzed based on the trajectory of the change. The continuation degree is determined by counting the number of sampling intervals within the delay interval that align with the direction of the anomaly trigger, and dividing this number by the total number of sampling intervals involved in the judgment within the delay interval; simultaneously, the cumulative change in that direction is calculated as the sum of the absolute values ​​of the changes in each interval in the same direction. When the continuation degree reaches 0.7 and the cumulative change reaches 10% of the warning threshold, the abnormal event is considered to have continued within the delay interval. The regression degree is determined by counting the number of sampling intervals opposite to the direction of the anomaly trigger, and dividing this number by the total number of sampling intervals involved in the judgment; simultaneously, the cumulative change in the opposite direction is calculated. When the regression degree reaches 0.6 and the cumulative change in the opposite direction reaches 8% of the warning threshold, the abnormal event is considered to have regressed within the delay interval.

[0076] The true stable state is determined based on the degree of state continuation and the degree of retreat. When the abnormal event continues and the degree of retreat does not reach 0.6, the true stable state is recorded as continuously stable; when the abnormal event retreats and the degree of state continuation does not reach 0.7, the true stable state is recorded as decaying stable; when neither the degree of state continuation nor the degree of retreat meets the corresponding conditions, the true stable state is recorded as fluctuating and pending verification. This result is used to determine whether to continue retrieving subsequent response changes from the associated state data.

[0077] When the actual stable state is either decaying and stable or fluctuating and pending verification, the edge computing side retrieves the subsequent response changes of each associated state data in the state response chain within a preset delay interval. Subsequent response changes include the interval representative value, direction of change, and response start position of each associated state data. The determination rule for a synchronous decay segment is as follows: at least two types of associated state data change in the opposite direction to the abnormal trigger within three consecutive sampling intervals, and the cumulative reverse change of each reaches 5% of the corresponding warning threshold; if the number of associated state data exceeds three types, then at least 60% of the associated state data meets the aforementioned condition. When this condition is met, a synchronous decay segment is determined to exist.

[0078] When a synchronous decay segment exists in the associated state data, it indicates that the abnormal event did not maintain physical transmission within the delay interval. The edge computing side removes the response restriction state, cancels the waiting queue of the high-level execution state, and marks the abnormal event as decaying after delay. When there is no synchronous decay segment and the true stable state is continuously stable, the delayed response processing state is maintained, and the corresponding abnormal event is submitted for subsequent manual confirmation or review by the superior edge computing side. When there is no synchronous decay segment and the true stable state is fluctuating and pending verification, the delay interval is extended by one preset interval, preferably no more than two times. This process can avoid data contamination triggering erroneous actions, while preserving the conditions for subsequent handling when the real anomaly continues to expand.

[0079] Example 2, please refer to Figure 2As shown in this embodiment, the power distribution terminal equipment anomaly verification and response system based on edge computing includes: Data acquisition module: acquires the operating status data of power distribution terminal equipment and triggers abnormal events based on preset rules; Temporal evolution analysis module: Extracts the temporal evolution trajectory of abnormal parameters corresponding to abnormal events, and generates abnormal formation characteristics based on the continuity of change, change inertia and evolution convergence between adjacent sampling intervals; Physical closed-loop analysis module: Based on the anomaly formation characteristics, retrieve the associated state data that has an energy coupling relationship with the anomaly parameters, and analyze the consistency of synchronous response and propagation association direction among the associated state data to generate physical closed-loop results; Link offset analysis module: Extracts the temporal disturbance characteristics and sampling discrete characteristics of the acquisition link before and after the occurrence of the anomaly based on the data source information corresponding to the anomaly event, and generates link offset results; The authenticity hierarchical correlation module performs hierarchical correlation analysis on anomaly formation characteristics, physical closed-loop results, and link offset results to generate anomaly authenticity results, and corrects the credibility level of the corresponding rules based on the anomaly authenticity results. Verification validity determination module: Obtain the execution results of the existing edge rule system for abnormal events, and determine the authenticity verification validity of the existing edge rule system in combination with the corrected rule credibility level; Delayed response processing module: When the authenticity verification is invalid, a delayed response processing is performed on the corresponding abnormal event.

[0080] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An edge computing-based power distribution terminal device anomaly verification and response method, characterized in that, include: Acquire the operating status data of the power distribution terminal equipment and trigger abnormal events based on preset rules; Extract the temporal evolution trajectory of the abnormal parameters corresponding to the abnormal event, and generate abnormal formation features based on the continuity of change, change inertia and evolution convergence between adjacent sampling intervals; Based on the anomaly formation characteristics, retrieve the associated state data that has an energy coupling relationship with the anomaly parameters, and analyze the consistency of synchronous response and propagation association direction among the associated state data to generate physical closed-loop results. Based on the data source information corresponding to the abnormal event, extract the temporal disturbance characteristics and sampling discrete characteristics of the acquisition link before and after the abnormality occurs, and generate the link offset result; The anomaly formation characteristics, physical closed-loop results, and link offset results are analyzed in a hierarchical manner to generate anomaly authenticity results, and the credibility level of the corresponding rules is corrected based on the anomaly authenticity results. Obtain the execution results of the existing edge rule system for abnormal events, and determine the validity of the authenticity verification of the existing edge rule system by combining the revised rule credibility level; When the authenticity verification is invalid, a delayed response is performed on the corresponding abnormal event.

2. The edge computing based power distribution terminal device anomaly checking and responding method of claim 1, wherein, Acquire operational status data of power distribution terminal equipment and trigger abnormal events based on preset rules, including: The running state data within a continuous sampling period is reconstructed in time to obtain the interval representative value of each sampling interval, and the interval representative value forms the state change trajectory. Based on the state change trajectory, the perturbation continuous features between adjacent sampling intervals are extracted, and the perturbation continuous features are checked for directional consistency. Based on the verified continuous perturbation features, state transition features are generated. When the state transition features meet the gradual triggering conditions in the preset rules, abnormal event information is generated.

3. The edge computing based power distribution terminal device anomaly checking and responding method of claim 2, wherein, Extract the temporal evolution trajectory of the abnormal parameters corresponding to the abnormal event, including: Based on the abnormal parameter name and abnormal trigger time in the abnormal event information, continuous sampling data within a preset traceability time range is retrieved and arranged progressively according to the sampling interval to form a time-series evolution trajectory. Based on the temporal evolution trajectory, the change continuation state and change stagnation state are extracted, and inertial extension features are generated according to the degree of preservation of the change direction in continuous sampling intervals. The degree of consistency of the offset of the inertial extension feature is analyzed to obtain the convergent extension state of the abnormal parameters in the time-series evolution process.

4. The edge computing based power distribution terminal device anomaly checking and responding method of claim 3, wherein, Generate anomaly formation features, including: When the convergence expansion state is established, the change continuation state is established, and the change stagnation state does not have a swingback, the anomaly formation characteristic is determined to be continuous gradual change formation; When the convergence extension state is established and the change stagnation state has a swingback, the anomaly formation characteristic is determined to be the formation of a disturbed gradual change. When the convergence extension state does not hold and the change continuation state does not hold, the abnormal formation characteristic is identified as discontinuous mutation formation.

5. The edge computing based power distribution terminal device anomaly checking and responding method of claim 1, wherein, Generate physical closed-loop results, including: Based on the abnormal parameter name and abnormal formation characteristics, the state diffusion path is determined, and the associated state data that have an energy coupling relationship with the abnormal parameter are retrieved along the state diffusion path. The temporal mapping is performed according to the starting position of the change of each associated state data in the continuous sampling interval to form a state response chain. The starting positions of the changes of adjacent associated state data in the state response chain are progressively compared to generate a state transmission sequence. Based on the state transmission sequence, the direction of change between adjacent associated state data is extracted, and the cumulative state of offset caused by reverse offset and transmission loss is determined; When the offset accumulation state forms a reverse abrupt change segment in the continuous sampling interval, and the continuous expansion range of the reverse abrupt change segment exceeds the preset continuous interval, it is determined that there is a reverse interruption region. When there is no reverse interruption region and the state changes are continuously propagated, a physical closed-loop result is generated that establishes the closed loop.

6. The edge computing based power distribution terminal device anomaly checking and responding method of claim 5, wherein, Generate link offset results, including: Based on the data source information in the abnormal event information, extract the data arrival order and sampling interval change status of the continuous sampling interval before and after the abnormality, and form the link time sequence trajectory; Based on the link time sequence trajectory, data stagnation sections, sudden aggregation sections, and the number of sequence anomalies are determined to obtain time sequence disturbance characteristics; The discrete distribution of sampled data within each sampling interval is progressively compared to obtain the discrete sampling features. Link offset results are then generated based on the temporal disturbance features and the temporal overlap of the discrete sampling features.

7. The edge computing based power distribution terminal device anomaly checking and responding method of claim 6, wherein, The process of generating anomaly authenticity results and correcting the credibility level of corresponding rules includes: determining the degree of continuous expansion of the state change corresponding to the anomaly event based on the anomaly formation characteristics; determining the degree of transmission integrity of the associated state data based on the physical closed-loop results; determining the offset interference range of the data source based on the link offset results; determining the authenticity maintenance state in the progressive order of continuous expansion, transmission integrity, and offset interference range; increasing the credibility level of the corresponding rule when the authenticity maintenance state continuously strengthens; decreasing the credibility level of the corresponding rule when the authenticity maintenance state continuously weakens; and generating anomaly authenticity results.

8. The method for anomaly verification and response of power distribution terminal equipment based on edge computing according to claim 7, characterized in that, Determine the validity of the authenticity verification of the existing edge rule processing flow, including: Obtain the execution status and execution sequence corresponding to abnormal events to form a rule execution trajectory; Based on the revised rule credibility level analysis, the execution continuation state and execution rollback state in the rule execution trajectory are analyzed to determine the degree of credibility matching; When there is a reverse execution segment in the rule execution trajectory that deviates from the abnormal authenticity result, and the reverse execution segment continues to expand in the continuous execution interval, the authenticity verification is determined to be invalid.

9. The method for anomaly verification and response of power distribution terminal equipment based on edge computing according to claim 8, characterized in that, Perform delayed response processing on the corresponding abnormal events, including: After the authenticity verification is invalid, the execution status of the corresponding lockout, trip, or shutdown of the abnormal event is suspended, and the change and expansion trajectory of the abnormal event is continuously recorded within the preset delay interval. Based on the trajectory of change, determine the degree of continuation and retraction of the abnormal parameters, and determine the true stable state accordingly; When the associated status data has a synchronous decay segment within the preset delay interval, the response restriction state of the corresponding abnormal event is released; when there is no synchronous decay segment, the delayed response processing state of the corresponding abnormal event is maintained.

10. An edge computing-based power distribution terminal equipment anomaly verification and response system, used to implement the edge computing-based power distribution terminal equipment anomaly verification and response method according to any one of claims 1-9, characterized in that, include: Data acquisition module: acquires the operating status data of power distribution terminal equipment and triggers abnormal events based on preset rules; Temporal evolution analysis module: Extracts the temporal evolution trajectory of abnormal parameters corresponding to abnormal events, and generates abnormal formation characteristics based on the continuity of change, change inertia and evolution convergence between adjacent sampling intervals; Physical closed-loop analysis module: Based on the anomaly formation characteristics, retrieve the associated state data that has an energy coupling relationship with the anomaly parameters, and analyze the consistency of synchronous response and propagation association direction among the associated state data to generate physical closed-loop results; Link offset analysis module: Extracts the temporal disturbance characteristics and sampling discrete characteristics of the acquisition link before and after the occurrence of the anomaly based on the data source information corresponding to the anomaly event, and generates link offset results; The authenticity hierarchical correlation module performs hierarchical correlation analysis on anomaly formation characteristics, physical closed-loop results, and link offset results to generate anomaly authenticity results, and corrects the credibility level of the corresponding rules based on the anomaly authenticity results. Verification validity determination module: Obtain the execution results of the existing edge rule system for abnormal events, and determine the authenticity verification validity of the existing edge rule system in combination with the corrected rule credibility level; Delayed response processing module: When the authenticity verification is invalid, a delayed response processing is performed on the corresponding abnormal event.

Images