Fault identification and diagnosis system for edge calculation of power distribution equipment

Abstract

The invention discloses a fault identification and diagnosis system for power distribution equipment edge calculation, and relates to the technical field of power distribution equipment management. Constructing a chained fault propagation path composed of at least two observable fault feature nodes arranged in sequence and an allowable time delay interval, and issuing and storing the chained fault propagation path in an edge node; the evidence generation module is used for collecting operation parameters in real time and extracting features; and when the features have stable deviation and meet a preset change rate or acceleration threshold, constructing a fault evidence node containing the type, the timestamp and the amplitude. According to the method, a chain fault propagation path is constructed, so that edge nodes can execute closed-loop inspection according to an equipment operation mechanism and a characteristic evolution rule; when the appearance sequence and the time difference of a plurality of fault evidence nodes are consistent with the chained propagation path, the propagation logic of the internal fault of the equipment can be accurately identified.

Description

Technical Field

[0001] This invention relates to the field of power distribution equipment management technology, specifically a fault identification and diagnosis system for edge computing of power distribution equipment. Background Technology

[0002] During actual operation, power distribution equipment may be affected by factors such as periodic load fluctuations, insulation material aging, electromagnetic interference, changes in ambient temperature, and poor local contact, leading to phenomena such as deviations in operating parameters, waveform distortion, increased partial discharge, or abnormal power factors. At present, fault identification of power distribution equipment usually relies on single-point monitoring or periodic inspections. If abnormal characteristics only appear in a short period of time, they may not be fully captured by traditional monitoring methods. At the same time, if there is a chain effect or disturbance propagation between multiple devices, conventional independent monitoring methods are difficult to identify the fault evolution pattern from an overall perspective. As edge computing capabilities are gradually deployed in power distribution systems, more and more monitoring devices are equipped with on-site computing and local decision-making capabilities, enabling them to process operational data in real time and thus reduce the pressure on the central side. However, most edge nodes only perform local judgments and lack cross-device and cross-time correlation analysis. Once there is a slight disturbance, intermittent anomaly, or cross-device coupled fault propagation, single-node judgment may result in misjudgment or omission. In addition, the sampling frequency, buffer period, and communication latency of different devices vary, which may lead to time axis misalignment when performing cross-node anomaly comparison, affecting the accuracy of fault identification. When fault disturbances exhibit chain propagation characteristics, accurately recording the sequential relationship between fault evidence, ensuring the reliability of the graph data, reducing the impact of random noise, and avoiding omissions due to sampling bias during cross-node analysis have become technical directions that urgently need improvement in the field of power distribution equipment fault identification. Therefore, this application proposes a fault identification and diagnosis system for edge computing of power distribution equipment. Summary of the Invention

[0003] The purpose of this invention is to provide a fault identification and diagnosis system for edge computing of power distribution equipment, so as to solve the problems mentioned in the background art.

[0004] The present invention can be achieved through the following technical solution: a fault identification and diagnosis system for edge computing of power distribution equipment, including a path storage module, an evidence generation module, a graph recording module, a closed-loop verification module, a mutual verification instruction module, and a consensus positioning module; The path storage module constructs a chain-like fault propagation path for each power distribution device within the edge computing range, based on the device's structural characteristics and operating mechanism. The chain-like fault propagation path consists of at least two sequentially arranged observable fault feature nodes, with an allowable time delay interval between adjacent fault feature nodes, which is used to constrain the fault propagation from the preceding feature to the following feature within a preset time range within the power distribution equipment. Before distributing the chained fault propagation path to the corresponding edge node, the path storage module calculates the minimum and maximum propagation times of the corresponding operating parameters of any two adjacent fault feature nodes in the chained fault propagation path in the historical samples, and uses the minimum and maximum propagation times as the upper and lower limits of the allowable delay interval to ensure that the allowable delay interval matches the actual operating characteristics of the power distribution equipment.

[0005] The path storage module distributes the chain-like fault propagation path and the allowable latency interval to the corresponding edge nodes and stores them locally for subsequent verification and retrieval. During the operation of the edge node, the evidence generation module collects the operating parameters of the corresponding power distribution equipment in real time and extracts features from the operating parameters. When any feature exhibits a stable deviation within a continuous time window, and the trend of the feature's change meets a preset change rate threshold or a preset change acceleration threshold, the evidence generation module uses the feature as fault evidence and constructs a fault evidence node based on the fault evidence type, occurrence timestamp, and feature amplitude to reflect the fault evolution state. The graph recording module is used to receive the fault evidence nodes constructed by the evidence generation module and write the fault evidence nodes into the local fault evidence graph, so that the fault evidence graph presents a multi-node graph structure that evolves over time. When a fault evidence node exceeds the set retention time and is not used in the closed-loop verification or mutual verification process, the graph recording module cleans up the fault evidence node to ensure the validity of the fault evidence graph data. The closed-loop verification module is used to check the order of occurrence and the time difference between adjacent fault evidence nodes when there are multiple fault evidence nodes in the fault evidence graph. If there is at least one set of fault evidence nodes whose order of occurrence is consistent with the chain fault propagation path, and the time difference of all adjacent fault evidence nodes falls within the corresponding allowable delay interval, then the closed-loop verification module will mark the corresponding power distribution equipment as a suspected fault state. If the order is inconsistent or the time difference does not meet the requirements, the downgrade or elimination operation will be performed automatically to avoid misjudgment.

[0006] The mutual verification instruction module is used to generate verifiable feature instructions and broadcast them to other edge nodes in the same area when power distribution equipment is marked as a suspected fault state. Verifiable feature instructions include the feature type, frequency band range, energy threshold identifier, and occurrence time period index of the target fault evidence node; The mutual verification instruction module receives mutual verification results from other edge nodes to provide cross-device evidence consistency analysis.

[0007] The consensus positioning module is used to count the mutual verification results returned by multiple edge nodes within a set time period; If the number of identical fault evidence or the consistency ratio in the mutual verification results reach the preset judgment threshold, the suspected fault state is confirmed as a real fault state. If the preset judgment threshold is not reached, the suspected fault status is revoked, and the graph recording module is instructed to remove the corresponding fault evidence chain.

[0008] After confirming the actual fault status, the consensus positioning module constructs an arrival time difference matrix based on the timestamps of the fault evidence returned by all participating edge nodes. The arrival time difference matrix is ​​used to calculate the time difference of fault disturbance propagation between any two edge nodes participating in mutual verification. When the deviation between the time difference of fault disturbance propagation and the allowable delay interval in the chain fault propagation path does not exceed the preset deviation threshold, the consensus positioning module outputs the edge node where the fault evidence first appears as the fault source edge node to improve the accuracy of fault positioning.

[0009] A further technical improvement of the present invention is that: when constructing a chain-like fault propagation path, the path storage module uses multiple observable fault feature nodes as the initial node set, and for any one preceding feature node, establishes a branch chain structure formed by at least two candidate subsequent feature nodes. During the operation of the target power distribution equipment, when the edge node collects a newly emerging fault evidence node, the path storage module determines whether the fault evidence node matches any candidate branch based on the feature type, occurrence timestamp and corresponding allowable delay interval of the fault evidence node. If a match is found, the candidate branch is retained as a valid branch. If there is no match, the candidate branch is removed from the chain failure propagation path, and a filtered chain failure propagation path is formed. The selected chain-like fault propagation paths and their allowable delay intervals are sent to the corresponding edge nodes and stored locally for use by the subsequent closed-loop verification module.

[0010] A further technical improvement of the present invention is that the step of feature extraction of the running parameters by the evidence generation module includes: A1. Establish a first time window and a second time window for the same operating parameter, and set the time width of the second time window to be greater than that of the first time window, in order to extract change features from different time scales; A2. Write the collected operating parameters into the first time window, calculate the numerical difference of the operating parameters between adjacent sampling points to form a difference sequence, and perform an average calculation on the difference sequence. Use the obtained average change as the change feature within the first time window. A3. The collected operating parameters are written into the second time window, and the operating parameters are processed by moving average to obtain the moving average value in multiple consecutive moving windows; the evidence generation module calculates the numerical difference between adjacent moving average values ​​and uses the numerical difference as the slow variable feature in the second time window; A4. The evidence generation module performs weighted calculations on the change feature and the slow variable feature according to preset weights to form a fused feature value; A5. The evidence generation module calculates the changes in fused feature values ​​over multiple consecutive time windows and writes the fused feature values ​​into the running parameter feature sequence for subsequent fault analysis.

[0011] A further technical improvement of the present invention is that the graph recording module includes the following steps when constructing the fault evidence graph and performing fault evidence node cleanup: Z1. After receiving the fault evidence node constructed by the evidence generation module, the graph recording module writes the fault evidence node into the local fault evidence graph according to the occurrence timestamp of the fault evidence node, and establishes a forward-to-back directed edge between the fault evidence node and the previous fault evidence node whose occurrence timestamp is earlier than the fault evidence node and whose adjacent time difference falls within the corresponding allowable delay interval, so that the fault evidence graph presents a time-evolved directed edge multi-node graph structure. Z2. When recording fault evidence nodes, the graph recording module sets a reference count value for each fault evidence node to count the number of calls. When the fault evidence node is called once by any module, the corresponding reference count value increases by one. Z3. When a fault evidence node exceeds the set retention time and its reference count value is zero, the graph recording module obtains the predecessor node and successor node corresponding to the fault evidence node in the fault evidence graph, and determines that the appearance timestamp of the predecessor node is earlier than the appearance timestamp of the successor node, and the time difference between the two falls into the corresponding allowable delay interval. Z4. When the timestamp of the predecessor node appears earlier than the timestamp of the successor node, and the time difference between the predecessor node and the successor node falls within the allowable delay interval, the graph recording module establishes a forward-backward directed edge connection between the predecessor node and the successor node, so that the fault evidence graph maintains a continuous multi-node directed edge structure. Z5. After completing the directed edge connection, the graph recording module deletes faulty evidence nodes that have exceeded the set retention time and have a reference count of zero, in order to ensure the validity of the graph data and avoid the breaking of the evidence chain.

[0012] A further technical improvement of the present invention is that: after the mutual verification instruction module receives the mutual verification results from other edge nodes, if the number of returned evidence of the same source failure or the consistency ratio does not reach the preset judgment threshold, the mutual verification instruction module expands the retrieval parameters of the original verifiable feature instruction to generate an enhanced verification instruction. The enhanced verification command expands the original frequency band range to both sides by a preset ratio or extends the original occurrence time period index forward or backward by a preset duration. When generating enhanced verification instructions, the mutual verification instruction module adds a time alignment identifier to the enhanced verification instructions. The time alignment identifier is used to instruct the receiving node to perform offset correction on the occurrence timestamp of the fault evidence node recorded locally according to the unified reference time, so that the evidence timelines of different edge nodes can be compared under the same time reference. The mutual verification instruction module rebroadcasts the enhanced verification instruction within the set waiting period and re-executes consistency statistics based on the mutual verification results returned by other edge nodes.

[0013] A further technical improvement of the present invention is that: before the path storage module distributes the chain-like fault propagation path to the corresponding edge node, it calculates the minimum propagation time and the maximum propagation time of the corresponding operating parameters of any two adjacent fault feature nodes in the chain-like fault propagation path in the historical samples, and uses the minimum propagation time and the maximum propagation time as the upper and lower limits of the allowable delay interval, so as to ensure that the allowable delay interval matches the actual operating characteristics of the power distribution equipment.

[0014] A further technical improvement of the present invention is that: after confirming the real fault state, the consensus positioning module constructs an arrival time difference matrix for the timestamps of the fault evidence returned by all edge nodes participating in mutual authentication; The arrival time difference matrix is ​​used to calculate the time difference of fault disturbance propagation between any two edge nodes participating in mutual verification. When the deviation between the time difference of fault disturbance propagation and the allowable delay interval in the chain fault propagation path does not exceed the preset deviation threshold, the consensus positioning module outputs the edge node where the fault evidence first appears as the fault source edge node to improve the accuracy of fault positioning.

[0015] A further technical improvement of the present invention is that, in step Z5, when the graph recording module performs cleanup on the fault evidence node, it first compares the difference between the occurrence timestamp of the fault evidence node and the current time. If the difference exceeds the set retention time and the fault evidence node is not called by the closed-loop verification module or the mutual verification instruction module, then the fault evidence node is deleted to ensure the data validity of the fault evidence graph.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention constructs a chain-like fault propagation path, enabling edge nodes to perform closed-loop verification based on the equipment's operating mechanism and characteristic evolution. When the order and time difference of multiple fault evidence nodes are consistent with the chain-like propagation path, the propagation logic of internal equipment faults can be accurately identified, thereby avoiding misjudgments caused by transient disturbances or local jitter and improving the reliability of fault identification. Furthermore, the present invention employs a directed edge structure and a reference counting cleanup mechanism in the graph recording module, enabling fault evidence to form a traceable multi-node graph structure in chronological order; when cleaning up expired and unused nodes, the graph chain is kept intact through directed edge reconstruction, which maintains data validity and avoids evidence loss due to node deletion, thus facilitating subsequent diagnostic analysis and parameter tuning. On the other hand, this invention proposes an enhanced verification instruction mechanism in the cross-device mutual verification process. By expanding the frequency band range, extending the time period index, and introducing time alignment identifiers, it effectively solves the problem of missed detection caused by sampling offset, buffer delay, or network transmission between edge nodes. The consensus positioning module performs propagation comparison based on the fault evidence timestamps returned by multiple nodes and can reverse the calculation of the fault source edge node to achieve more accurate fault positioning and diagnosis. Attached Figure Description

[0017] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0018] Figure 1 This is a schematic diagram of the system logic of the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0020] Please see Figure 1 As shown, the present invention provides a fault identification and diagnosis system for edge computing of power distribution equipment, including a path storage module, an evidence generation module, a graph recording module, a closed-loop verification module, a mutual verification instruction module, a consensus positioning module, and a result archiving module; The path storage module constructs a chain-like fault propagation path for each power distribution device within the edge computing range, based on the device's structural characteristics and operating mechanism. When constructing a chain-like fault propagation path, the path storage module uses multiple observable fault feature nodes as the initial node set, and for any preceding feature node, establishes a branch chain structure formed by at least two candidate subsequent feature nodes. During the operation of the target power distribution equipment, when the edge node collects a newly emerging fault evidence node, the path storage module determines whether the fault evidence node matches any candidate branch based on the feature type, occurrence timestamp and corresponding allowable delay interval of the fault evidence node. If a match is found, the candidate branch is retained as a valid branch. If there is no match, the candidate branch is removed from the chain failure propagation path, and a filtered chain failure propagation path is formed. The selected chain-like fault propagation paths and their allowable delay intervals are sent to the corresponding edge nodes and stored locally for use by the subsequent closed-loop verification module.

[0021] Specifically, when constructing a chain-like fault propagation path, the path storage module first determines a set of multiple observable fault features as the initial feature nodes based on the fault features that can be collected from the target power distribution equipment. The initial feature node set includes at least the following types of feature nodes: current waveform distortion, temperature rise rate change, partial discharge energy, insulation leakage current, and harmonic amplitude change. The path storage module takes any one of the preceding feature nodes in the initial feature node set as the starting point, sets at least two candidate following feature nodes for that preceding feature node, and records each propagation relationship from the preceding feature node to the candidate following feature node as a candidate branch, thereby constructing a branch chain structure containing multiple candidate branches. Each candidate branch corresponds to an allowable time delay interval, which is used to represent the maximum and minimum propagation time difference that may occur between the preceding and following feature nodes under normal fault evolution conditions. After construction, the path storage module stores the branch chain structure and its corresponding allowable time delay interval in the local storage area and distributes it to the corresponding edge nodes for subsequent use.

[0022] During the actual operation of the target power distribution equipment, the evidence generation module generates multiple fault evidence nodes based on the collected operating parameters. Each fault evidence node includes at least the fault evidence type, the occurrence timestamp, and the characteristic amplitude. After receiving a fault evidence node, the path storage module uses the fault evidence type of the fault evidence node as the judgment criterion to search for candidate subsequent feature nodes with the same fault evidence type in the branch chain structure, and determines the preceding feature nodes connected to the candidate subsequent feature nodes. Subsequently, the path storage module calculates the time difference based on the occurrence timestamp of the preceding feature nodes and compares the time difference with the allowable delay interval corresponding to the candidate subsequent feature nodes. If the time difference is greater than or equal to the lower limit of the allowed delay interval and less than or equal to the upper limit of the allowed delay interval, the candidate branch is determined to satisfy the propagation law; the path storage module increments the matching count of the candidate branch by one, clears the consecutive mismatch count of the candidate branch to zero, and increases the weight value of the candidate branch according to the preset weight increment; if the time difference is less than the lower limit of the allowed delay interval or greater than the upper limit of the allowed delay interval, the candidate branch is determined to not conform to the propagation law; the path storage module increments the consecutive mismatch count of the candidate branch by one and decreases the weight value of the candidate branch according to the preset weight decrease amount.

[0023] When the consecutive mismatch count of any candidate branch is greater than or equal to the set upper limit, the path storage module removes the candidate branch; or, when the weight value of any candidate branch is less than or equal to the set removal threshold, the path storage module also removes the candidate branch; as the system running time increases, the path storage module repeatedly executes the above steps of matching, counting, incrementing, decrementing and removing based on the continuous input of fault evidence nodes, so that candidate branches that do not conform to the actual propagation law of the device are gradually eliminated.

[0024] At any fan-out position, if only one candidate branch remains unremoved, the path storage module sets that candidate branch as the only subsequent branch at that fan-out position and deletes other candidate branches. In this way, the original branch chain structure containing multiple candidate paths gradually converges into a filtered chain fault propagation path containing only the actual propagation path, so that the propagation path can accurately reflect the fault evolution sequence and propagation law of the target power distribution equipment in actual operation.

[0025] After forming the filtered chain-like fault propagation path, the path storage module sends the propagation path and its corresponding allowable delay interval to the corresponding edge node processing unit and writes it into the local storage area of ​​the edge node. When the closed-loop verification module performs sequential verification and time difference verification of fault evidence nodes, it can directly read the filtered chain-like fault propagation path from the local storage to make a logical judgment on whether the fault conforms to the chain-like propagation law.

[0026] During the operation of the edge node, the evidence generation module collects the operating parameters of the corresponding power distribution equipment in real time and extracts features from the operating parameters. The evidence generation module performs feature extraction on the running parameters, including the following steps: A1. Establish a first time window and a second time window for the same operating parameter, and set the time width of the second time window to be greater than that of the first time window, in order to extract change features from different time scales; A2. Write the collected operating parameters into the first time window, calculate the numerical difference of the operating parameters between adjacent sampling points to form a difference sequence, and perform an average calculation on the difference sequence. Use the obtained average change as the change feature within the first time window. A3. The collected operating parameters are written into the second time window, and the operating parameters are processed by moving average to obtain the moving average value in multiple consecutive moving windows; the evidence generation module calculates the numerical difference between adjacent moving average values ​​and uses the numerical difference as the slow variable feature in the second time window; A4. The evidence generation module performs weighted calculations on the change characteristics and slow variable characteristics according to preset weights to form a fused feature value, which is used to simultaneously characterize the rapid fluctuation characteristics and slow change characteristics of the operating parameters. A5. The evidence generation module calculates the changes in the fusion feature value within multiple consecutive time windows and writes the fusion feature value into the running parameter feature sequence for subsequent fault analysis calls. When any feature exhibits a stable deviation within a continuous time window, and the trend of the feature's change meets a preset change rate threshold or a preset change acceleration threshold, the evidence generation module uses the feature as fault evidence and constructs a fault evidence node based on the fault evidence type, occurrence timestamp, and feature amplitude to reflect the fault evolution state. The graph recording module is used to receive the fault evidence nodes constructed by the evidence generation module and write the fault evidence nodes into the local fault evidence graph, so that the fault evidence graph presents a multi-node graph structure that evolves over time. The fault evidence recording module includes the following steps when constructing the fault evidence graph and performing fault evidence node cleanup: Z1. After receiving the fault evidence node constructed by the evidence generation module, the graph recording module writes the fault evidence node into the local fault evidence graph according to the occurrence timestamp of the fault evidence node, and establishes a forward-to-back directed edge between the fault evidence node and the previous fault evidence node whose occurrence timestamp is earlier than the fault evidence node and whose adjacent time difference falls within the corresponding allowable delay interval, so that the fault evidence graph presents a time-evolved directed edge multi-node graph structure. Z2. When recording fault evidence nodes, the graph recording module sets a reference count value for each fault evidence node to count the number of calls. When the fault evidence node is called once by any module, the corresponding reference count value increases by one. Z3. When a fault evidence node exceeds the set retention time and its reference count value is zero, the graph recording module obtains the predecessor node and successor node corresponding to the fault evidence node in the fault evidence graph, and determines that the appearance timestamp of the predecessor node is earlier than the appearance timestamp of the successor node, and the time difference between the two falls into the corresponding allowable delay interval. Z4. When the timestamp of the predecessor node appears earlier than the timestamp of the successor node, and the time difference between the predecessor node and the successor node falls within the allowable delay interval, the graph recording module establishes a forward-backward directed edge connection between the predecessor node and the successor node, so that the fault evidence graph maintains a continuous multi-node directed edge structure. Z5. After completing the directed edge connection, the graph recording module deletes faulty evidence nodes that have exceeded the set retention time and have a reference count of zero, in order to ensure the validity of the graph data and avoid the breaking of the evidence chain. Specifically, when the graph recording module cleans up faulty evidence nodes, it first compares the difference between the timestamp of the faulty evidence node and the current time. If the difference exceeds the set retention time and the faulty evidence node has not been called by the closed-loop verification module or the mutual verification instruction module, then the faulty evidence node is deleted to ensure the data validity of the faulty evidence graph.

[0027] The closed-loop verification module is used to check the order of occurrence and the time difference between adjacent fault evidence nodes when there are multiple fault evidence nodes in the fault evidence graph. If there is at least one set of fault evidence nodes whose order of occurrence is consistent with the chain fault propagation path, and the time difference of all adjacent fault evidence nodes falls within the corresponding allowable delay interval, then the closed-loop verification module will mark the corresponding power distribution equipment as a suspected fault state. If the order is inconsistent or the time difference does not meet the requirements, the downgrade or elimination operation will be performed automatically to avoid misjudgment.

[0028] The mutual verification instruction module is used to generate verifiable feature instructions and broadcast them to other edge nodes in the same area when power distribution equipment is marked as a suspected fault state. Verifiable feature instructions include the feature type, frequency band range, energy threshold identifier, and occurrence time period index of the target fault evidence node; The mutual verification instruction module receives mutual verification results from other edge nodes to provide cross-device evidence consistency analysis.

[0029] Furthermore, after receiving the mutual verification results from other edge nodes, if the number of returned evidence of the same source failure or the consistency ratio does not reach the preset judgment threshold, the mutual verification instruction module will expand the retrieval parameters of the original verifiable feature instruction to generate an enhanced verification instruction. The enhanced verification command expands the original frequency band range to both sides by a preset ratio or extends the original occurrence time period index forward and backward by a preset duration to improve the coverage of different edge nodes in searching for the same source fault evidence nodes in the local fault evidence map. When generating enhanced verification instructions, the mutual verification instruction module adds a time alignment identifier to the enhanced verification instructions. The time alignment identifier is used to instruct the receiving node to perform offset correction on the occurrence timestamp of the fault evidence node recorded locally according to the unified reference time, so that the evidence timelines of different edge nodes can be compared under the same time reference. The mutual verification instruction module rebroadcasts the enhanced verification instruction within the set waiting period and re-executes consistency statistics based on the mutual verification results returned by other edge nodes to solve the problem of missing evidence of the same source failure due to sampling offset, cache delay or network transmission.

[0030] Specifically, after receiving the mutual verification results from other edge nodes, the mutual verification instruction module first counts the number of returned evidence of the same source failure and the consistency ratio. If the number of returned items or the consistency ratio does not reach the preset judgment threshold, the mutual verification instruction module does not directly cancel the suspected fault status, but expands the retrieval parameters in the original verifiable feature instruction to generate an enhanced verification instruction. When generating enhanced verification commands, the mutual verification command module expands the original frequency band range upwards and downwards by preset proportions, or extends the original occurrence time period index forwards and backwards by preset durations, enabling edge nodes to cover a wider frequency band or a longer time span when performing mutual verification searches. Through these expanded search parameters, the coverage of searching for nodes with the same source of fault evidence in the local fault evidence map by different edge nodes can be improved, reducing the risk of missed detections due to differences in sampling accuracy or inconsistent cache periods among edge nodes.

[0031] To avoid time offsets in fault evidence records from different edge nodes, the mutual verification instruction module adds a time alignment identifier to the enhanced verification instruction. After parsing the time alignment identifier, the receiving node uses the unified reference time defined in the enhanced verification instruction as a benchmark to perform offset correction on the occurrence timestamps of the locally recorded fault evidence nodes. This ensures that the evidence timelines of different nodes are compared under the same benchmark, thereby avoiding the impact of time errors caused by clock deviations, sampling delays, or network transmission on the accuracy of mutual verification.

[0032] After constructing the enhanced verification command, the mutual verification command module rebroadcasts the enhanced verification command within a set waiting period. The receiving node returns the corresponding information after completing the retrieval and generating a new mutual verification result. The mutual verification command module re-executes consistency statistics based on the mutual verification results returned by multiple edge nodes to determine whether the number or consistency ratio of evidence of shared faults meets a preset judgment threshold. This extended verification process solves the problem of missed evidence of shared faults due to sampling offset, cache latency, or network transmission, improving the reliability of cross-device mutual verification judgment.

[0033] The consensus positioning module is used to count the mutual verification results returned by multiple edge nodes within a set time period; If the number of identical fault evidence or the consistency ratio in the mutual verification results reach the preset judgment threshold, the suspected fault state is confirmed as a real fault state. If the preset judgment threshold is not reached, the suspected fault status is revoked, and the graph recording module is instructed to remove the corresponding fault evidence chain.

[0034] After confirming the actual fault state, the consensus location module calculates the arrival time difference of the fault disturbance based on the timestamps of the fault evidence returned by all participating edge nodes, and combines the allowed delay interval of the chain fault propagation path to reverse-engineer the edge node where the fault evidence first appears, and outputs the corresponding fault source location.

[0035] The results archiving module is used to generate fault diagnosis results after the location results are output, including the fault source location, fault type, and fault evidence chain structure, and push them to the upper-level monitoring system; at the same time, it archives all fault evidence nodes, the sequential relationships between nodes, the mutual verification process, and the location basis, generating traceable fault evidence records to support auditing, analysis, or parameter tuning.

[0036] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0037] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A fault identification diagnostic system for power distribution equipment edge computing, characterized by, The method comprises the following steps: A path storage module, for each power distribution equipment within the edge computing range, constructs a chain fault propagation path composed of at least two sequentially arranged observable fault feature nodes and an allowed time delay interval according to the structure and operation mechanism of the power distribution equipment, and issues and stores it in the edge node; An evidence generation module, which collects operating parameters in real time and extracts features; when any feature deviates stably within a continuous time window and meets the preset change rate or acceleration threshold, the corresponding fault evidence node is constructed; A graph recording module receives the fault evidence node and constructs a local fault evidence graph that evolves over time; when the fault evidence node exceeds the set retention time and has not been closed-loop verified or mutually verified, it is cleaned up; A closed-loop verification module calls the chain fault propagation path to collate the appearance order and adjacent time difference of multiple fault evidence nodes in the graph; if they are consistent with the chain fault propagation path and the time difference falls within the allowed time delay interval, the suspected fault state is marked; A mutual verification instruction module generates a verifiable feature instruction when the power distribution equipment is marked as a suspected fault state and broadcasts it to other edge nodes in the same area, and receives mutual verification results from other edge nodes; A consensus positioning module counts the mutual verification results and confirms the fault and outputs the fault source edge node when the consistency meets the standard.

2. The fault identification diagnostic system for power distribution edge computing of claim 1, wherein, When constructing the chain fault propagation path, the path storage module takes multiple observable fault feature nodes as the initial node set, and for any previous feature node, establishes a branch chain structure formed by at least two candidate next-level feature nodes; During the operation of the target power distribution equipment, when the edge node collects a newly appeared fault evidence node, the path storage module judges whether the fault evidence node matches any candidate branch according to the feature type, appearance timestamp and corresponding allowed time delay interval of the fault evidence node; If it matches, the candidate branch is retained as an effective branch; If it does not match, the candidate branch is removed from the chain fault propagation path, and a filtered chain fault propagation path is formed; The filtered chain fault propagation path and its allowed time delay interval are issued to the corresponding edge node and stored locally for subsequent closed-loop verification module calls.

3. The fault identification diagnostic system for power distribution edge computing of claim 2, wherein, The evidence generation module performs feature extraction on the operating parameters, including: A1, for the same operating parameter, a first time window and a second time window are established respectively, and the time width of the second time window is set to be greater than that of the first time window, which is used to extract change features from different time scales; A2, write the collected operating parameters into the first time window, calculate the numerical difference between adjacent sampling points, form a difference sequence, and perform average calculation on the difference sequence, taking the average change as the change feature in the first time window; A3, write the collected operating parameters into the second time window, and perform moving average processing on the operating parameters to obtain moving average values in multiple consecutive moving windows; the evidence generation module calculates the numerical difference between adjacent moving average values, and takes the numerical difference as the slowly varying variable feature in the second time window. A4. The evidence generation module performs weighted calculations on the change feature and the slow variable feature according to preset weights to form a fused feature value; A5. The evidence generation module calculates the changes in the fusion feature value within multiple consecutive time windows and writes the fusion feature value into the running parameter feature sequence.

4. The fault identification diagnostic system for power distribution edge computing of claim 3, wherein, The graph recording module includes the following steps when constructing the fault evidence graph and performing fault evidence node cleanup: Z1. After receiving the fault evidence node constructed by the evidence generation module, the graph recording module writes the fault evidence node into the local fault evidence graph according to the occurrence timestamp of the fault evidence node, and establishes a forward-to-back directed edge between the fault evidence node and the previous fault evidence node whose occurrence timestamp is earlier than the fault evidence node and whose adjacent time difference falls within the corresponding allowable delay interval, so that the fault evidence graph presents a time-evolved directed edge multi-node graph structure. Z2. When recording fault evidence nodes, the graph recording module sets a reference count value for each fault evidence node to count the number of calls. When the fault evidence node is called once by any module, the corresponding reference count value increases by one. Z3. When a fault evidence node exceeds the set retention time and its reference count value is zero, the graph recording module obtains the predecessor node and successor node corresponding to the fault evidence node in the fault evidence graph, and determines that the appearance timestamp of the predecessor node is earlier than the appearance timestamp of the successor node, and the time difference between the two falls into the corresponding allowable delay interval. Z4. When the timestamp of the predecessor node appears earlier than the timestamp of the successor node, and the time difference between the predecessor node and the successor node falls within the allowable delay interval, the graph recording module establishes a forward-backward directed edge connection between the predecessor node and the successor node, so that the fault evidence graph maintains a continuous multi-node directed edge structure. Z5. After completing the directed edge connection, the graph recording module deletes faulty evidence nodes that have exceeded the set retention time and have a reference count of zero, in order to ensure the validity of the graph data and avoid the breaking of the evidence chain.

5. The fault identification diagnostic system for power distribution edge computing of claim 1, wherein, After receiving the mutual verification results from other edge nodes, if the number of returned evidence of the same source failure or the consistency ratio does not reach the preset judgment threshold, the mutual verification instruction module will expand the retrieval parameters of the original verifiable feature instruction to generate an enhanced verification instruction. The enhanced verification command expands the original frequency band to both sides by a preset ratio or extends the original occurrence time period index forward or backward by a preset duration. When generating enhanced verification instructions, the mutual verification instruction module adds a time alignment identifier to the enhanced verification instructions. The time alignment identifier is used to instruct the receiving node to perform offset correction on the occurrence timestamp of the fault evidence node recorded locally according to the unified reference time, so that the evidence timelines of different edge nodes can be compared under the same time reference. The mutual verification instruction module rebroadcasts the enhanced verification instruction within the set waiting period and re-executes consistency statistics based on the mutual verification results returned by other edge nodes.

6. The fault identification diagnostic system for power distribution edge computing of claim 1, wherein, The path storage module calculates the minimum propagation time and the maximum propagation time of the running parameters of any adjacent fault feature nodes in the chain fault propagation path in the historical samples before issuing the chain fault propagation path to the corresponding edge node, and takes the minimum propagation time and the maximum propagation time as the upper and lower limit parameters of the allowed time delay interval, so as to ensure that the allowed time delay interval matches the actual running characteristics of the power distribution equipment.

7. The fault identification diagnostic system for power distribution edge computing of claim 1, wherein, The consensus positioning module constructs a time difference matrix after confirming the real fault state; The time difference matrix is used to calculate the fault disturbance propagation time difference between any two edge nodes participating in mutual verification. When the deviation between the fault disturbance propagation time difference and the allowed time delay interval in the chain fault propagation path does not exceed the preset deviation threshold, the consensus positioning module outputs the edge node that first appears the fault evidence as the fault source edge node.

8. The fault identification diagnostic system for power distribution edge computing of claim 4, wherein, In the Z5 step, the graph recording module compares the difference between the appearance time stamp of the fault evidence node and the current time before performing cleaning on the fault evidence node. If the difference exceeds the set retention time and the fault evidence node is not called by the closed-loop verification module or the mutual verification instruction module, the fault evidence node is deleted, so as to ensure the data validity of the fault evidence graph.

Images