An intelligent distribution network fault locating and early warning system based on multi-source perception cooperation

Abstract

The application discloses a kind of intelligent distribution network fault location and early warning system based on multi-source perception cooperation, it is related to fault location and early warning technical field.The system of the present application includes: edge perception node module, self-organizing networking module, data sharing module, collaborative research module and decision output module;Edge perception node module collects distribution network multi-source signal, triggers distributed collaborative decision when abnormal;Self-organizing networking module completes node scanning, multi-dimensional core node screening, forms and monitors temporary collaborative consensus group;Data sharing module realizes in-group feature data sharing;Collaborative research module forms research results, triggers secondary collection verification when not consensus;Decision output module pushes early warning information, generates and uploads fault location result and disposal scheme.The present application realizes edge side autonomous collaborative decision, breaks through main station dependence, shortens fault response delay, reduces misjudgment and missing rate, and improves distribution network power supply reliability.

Description

Technical Field

[0001] This invention relates to the field of fault location and early warning technology, specifically to an intelligent distribution network fault location and early warning system based on multi-source sensing collaboration. Background Technology

[0002] As a crucial link connecting users in the power system, the efficiency of fault location and early warning in the distribution network directly affects the reliability of power supply. Currently, distribution network fault decision-making generally adopts a centralized processing model. Terminals are only responsible for data collection, uploading the data to the master station, which then performs data analysis and generates decision instructions before issuing disposal commands to the terminals.

[0003] This model has significant technical limitations: First, the decision-making process involves multiple stages such as data transmission and master station analysis, resulting in decision delays typically reaching the second level, making it difficult to meet the requirements for rapid fault response and thus extending the duration of power outages for users; Second, the system operation is highly dependent on the master station, and once the master station fails or communication is interrupted, the terminal will lose its decision-making basis, and the entire fault handling system will be paralyzed; Third, the terminal lacks independent judgment capabilities, the data collected by a single node is easily affected by electromagnetic interference and environmental noise, and there is no cross-verification mechanism for multi-node data, leading to prominent problems of misjudgment and missed judgment of faults.

[0004] Furthermore, existing technologies lack an autonomous collaboration mechanism between edge nodes, hindering self-organized networking and distributed analysis among devices, further limiting the timeliness and accuracy of fault handling. These issues result in low distribution network operation and maintenance efficiency and difficulty in improving power supply reliability, necessitating a technical solution that breaks through the dependence on the master station and enables rapid decision-making through autonomous collaboration at the edge. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent power distribution network fault location and early warning system based on multi-source sensing collaboration, so as to solve the problems mentioned in the background art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A smart distribution network fault location and early warning system based on multi-source sensing collaboration includes: an edge sensing node module, a self-organizing network module, a data sharing module, a collaborative judgment module, and a decision output module; Edge sensing node modules are deployed on the distribution network lines to collect distribution network current, electric field, and vibration signals in real time. When any node detects an abnormal signal that meets the preset single-dimensional threshold and has passed preliminary validity verification, it triggers a distributed collaborative decision-making process. The self-organizing networking module initiates self-organizing networking, scans neighboring edge sensing nodes, selects core nodes through a filtering strategy that correlates signal quality with spatial distance, and builds a temporary collaborative consensus group. The data sharing module is used to enable the sharing of pre-processed feature data of each edge sensing node within the group. The feature data includes current amplitude, vibration frequency and electric field distortion rate. The collaborative analysis module, based on the feature data shared by the data sharing module, forms a consensus analysis result through multi-node feature consistency verification and priority judgment of key abnormal features. When more than half of the nodes detect consistent feature data without contradictions, it is determined that there is a potential fault in the distribution network. When any node detects key abnormal feature data that characterizes a sudden fault in the distribution network, it is directly confirmed that a fault has occurred in the distribution network. If no consensus analysis result is reached, a secondary data collection and feature refinement verification process is triggered. The decision output module is used to push early warning information when a potential fault is detected, generate fault location results and handling plans after a fault is confirmed, and simultaneously upload the early warning information, fault location results and handling plans to the distribution network master station.

[0007] Furthermore, the edge sensing node module includes a multi-source signal acquisition and preprocessing unit and an anomaly detection and triggering unit; The multi-source signal acquisition and preprocessing unit acquires power distribution network current, electric field, and vibration signals in real time, and performs noise reduction, format standardization, and preliminary validity verification on the raw signals to eliminate invalid data caused by environmental interference. The anomaly detection and triggering unit compares the preprocessed signals with preset single-dimensional thresholds to determine whether there are any anomalies. When an anomaly signal that meets the threshold conditions and has been verified is detected, a trigger command is generated to start the distributed collaborative decision-making process.

[0008] Furthermore, the self-organizing networking module includes a node scanning unit, a core node screening unit, and a temporary collaborative consensus group construction unit; The node scanning unit initiates a self-organizing network process, scanning neighboring edge sensing nodes, identifying the communication status and location information of each node, and initially marking communicable nodes to form a candidate node list. The core node screening unit, based on the candidate node list, adopts a screening strategy that correlates signal quality and spatial distance to comprehensively evaluate the candidate nodes, thereby obtaining the core nodes. The temporary collaborative consensus group construction unit, based on the screened core nodes, forms a temporary collaborative consensus group, establishes dedicated communication links between nodes within the group, and monitors the online status and communication stability of each core node within the group in real time.

[0009] Furthermore, the node scanning unit initiates a broadcast detection signal based on a preset communication protocol, with the detection range covering adjacent edge sensing nodes within the monitoring area of ​​the distribution network line, and records the transmission time of the detection signal; it receives the response signal from each edge sensing node, extracts the unique identifier, real-time communication status parameters and location coordinate information of each response node, and records the reception time of the response signal; For each responding node, the communication response time is obtained by the difference between the transmission time and the reception time. A communication response time threshold is set, and the communication response time of each responding node is compared with the threshold. Nodes whose communication response time does not exceed the threshold are judged to have good communication status and are marked into the candidate node list. Nodes whose communication response time exceeds the threshold are judged to have excessive communication latency and are removed. Finally, a candidate node list containing the identifier, location and communication status of communicable nodes is formed.

[0010] Furthermore, the screening process for the core node screening unit is as follows: The basic parameters of candidate nodes and the current distribution network scenario information are extracted. All parameters are directly collected or preset by the node's built-in module, without the need for complex derivation and calculation. The basic parameters of candidate nodes include the received signal strength and the straight-line distance between the candidate node and the source node. The received signal strength is directly read by the node's radio frequency module, and the straight-line distance between the candidate node and the source node is directly calculated based on the preset latitude and longitude coordinates when the node is deployed. The distribution network scenario information includes the line topology type and the electromagnetic interference level. The line topology type is preset as a straight line or a branch line, with a corresponding preset topology correction coefficient. The electromagnetic interference level is directly output by the node's interference detection module as level one, level two, and level three results. For received signal strength, the minimum and maximum received signal strength values ​​in the candidate node set are statistically analyzed and normalized. The normalization results are then adjusted based on the preset correction value corresponding to the electromagnetic interference level. For the straight-line distance between the candidate node and the source node, quantification is performed using the preset maximum effective communication distance and the preset topology correction coefficient for the corresponding line topology type. Different preset topology correction coefficients are configured for straight lines and branch lines. The basic evaluation value of the candidate node is calculated using a weighted approach based on electromagnetic interference level. At low interference levels, the signal quality weight and spatial adaptation weight are equal; at medium interference levels, the signal quality weight is greater than the spatial adaptation weight; and at high interference levels, the signal quality weight is further increased compared to the medium interference level, with the sum of the signal quality weight and the spatial adaptation weight being one. Candidate nodes are sorted from largest to smallest based on their baseline evaluation values. A predetermined number of nodes at the top of the list are selected to form a preliminary core node set, with the predetermined number being appropriate for the scale of collaborative analysis. The straight-line distance between any two nodes in the preliminary core node set is calculated and compared with a predetermined conflict distance threshold. If there is a situation where the distance between two nodes is less than the predetermined conflict distance threshold, the node with the higher baseline evaluation value is retained and the other node is removed. The node with the next highest baseline evaluation value is selected from the candidate node list and added to the preliminary set. The above verification process is repeated until the distance between all nodes in the set is not less than the predetermined conflict distance threshold, thus forming the final core node set.

[0011] Furthermore, the temporary collaborative consensus group construction unit sends a multicast networking instruction containing the unique identifier of the temporary collaborative consensus group, preset communication link parameters, and data transmission protocol to all nodes in the core node set. After receiving and responding to the confirmation instruction, the core nodes complete the construction of the temporary collaborative consensus group, establish a point-to-point exclusive communication link between any two core nodes in the group based on the preset encrypted communication protocol, and allocate an independent communication channel for each link. Heartbeat probe packets are periodically sent to each communication link within the group. The link communication delay is obtained by recording the sending time of the heartbeat probe packets and the receiving time of the response heartbeat packets. The link data packet loss rate is obtained by statistically analyzing the total number of heartbeat probe packets sent and the number of successfully received response heartbeat packets within a preset time period. The link communication delay is standardized by combining it with the preset maximum allowable communication delay. The link stability factor is obtained by combining the standardized result with the link data packet loss rate. The system counts the number of consecutive heartbeat responses from each core node within the group, sets an offline threshold, and determines that a core node is offline when the number of consecutive non-response counts reaches this threshold. It also sets a link stability threshold, triggering a link reconnection mechanism when the stability factor of a communication link falls below this threshold. Offline nodes are removed from the temporary collaborative consensus group, and nodes ranked immediately after the original core node are selected from the candidate node list output by the core node screening unit to be added to the group, and a new point-to-point dedicated communication link is established. This monitoring process is repeated until all nodes in the temporary collaborative consensus group remain online and the stability factor of all communication links is not lower than the link stability threshold.

[0012] Furthermore, the collaborative analysis module includes a multi-node feature consistency verification unit and a key anomaly feature priority determination unit; The multi-node feature consistency verification unit performs consistency comparison and verification on the feature data of each edge sensing node in the temporary collaborative consensus group based on the feature data shared by the data sharing module, counts the number of nodes with consistent feature data, and checks whether there is contradictory feature data in the group; when the number of nodes with consistent feature data exceeds half and there is no contradictory data, a preliminary judgment conclusion that there is a potential fault in the distribution network is output. The key anomaly feature priority judgment unit identifies whether the feature data collected by each node in the group contains key anomaly feature data that characterizes a sudden fault in the distribution network. If any node is found to have such key anomaly feature data, a definitive judgment conclusion that the distribution network has failed is directly output. In the case where no consensus judgment result is reached after the feature consistency verification of multiple nodes, a secondary feature data collection and feature refinement verification process is triggered.

[0013] Furthermore, the multi-node feature consistency verification unit specifically includes: The feature data of all edge sensing nodes in the temporary collaborative consensus group are normalized to obtain the normalized current amplitude, vibration frequency and electric field distortion rate. The core node in the group is used as the benchmark node. The comprehensive feature similarity between the remaining nodes and the benchmark node is calculated by combining the feature weights. The feature consistency between a single node and the benchmark node is quantified. The feature weights are set based on the priority of distribution network fault judgment and the sum of the weights is one. A feature similarity threshold is set based on the distribution network experimental data. When the comprehensive feature similarity between a node and the benchmark node is not lower than the threshold, the node is determined to be consistent with the benchmark node. The total number of nodes with consistent features within the group is counted, and the ratio of this total number to the total number of nodes within the temporary collaborative consensus group is calculated as the consistency ratio. If the comprehensive feature similarity between a node and the baseline node is lower than the preset threshold for contradictory data, it is determined that there is contradictory data in the group; if there is no such node, it is determined that there is no contradictory data. When the consistency ratio exceeds one-half and there is no contradictory data, a preliminary conclusion that there is a potential fault in the distribution network is output, that is, a consensus is reached. When the consistency ratio does not exceed one-half or there is contradictory data, it is determined that no consensus has been reached, and the verification result is transferred to the key abnormal feature priority judgment unit for processing.

[0014] Furthermore, the key anomaly feature priority determination unit specifically includes: Based on distribution network industry standards and fault cases, key abnormal feature thresholds for sudden faults in distribution networks are set, and the original feature data collected from each node are verified one by one to directly determine whether a sudden fault exists. Among them, current amplitude, vibration frequency, and electric field distortion rate correspond to preset key thresholds for sudden faults, and each threshold is set with appropriate values ​​based on distribution network industry standards and fault cases. If any node detects any of the above-mentioned key abnormal features, a definitive conclusion that a distribution network fault has occurred is directly output without subsequent verification. When the multi-node feature consistency verification fails to reach a consensus and no node detects a key abnormal feature, a secondary data acquisition and feature refinement verification process is triggered. First, all nodes in the group are controlled to improve the acquisition accuracy and shorten the acquisition cycle, and the three types of feature data are re-acquired and the above normalization process is repeated. Then, the feature deviation rate between the secondary acquisition data and the original feature data acquired in the first acquisition is calculated using the same feature weights as the multi-node feature consistency verification unit to eliminate the influence of random errors. A preset deviation rate threshold is set, and abnormal node data with a feature deviation rate exceeding the threshold are removed. Based on the remaining valid data, the feature similarity and consistency ratio are recalculated until a consensus is reached or it is confirmed that there is no clear fault and the corresponding conclusion is output.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention, by constructing a self-organizing network of edge nodes and a multi-node collaborative judgment system, and combining a multi-dimensional core node screening strategy to build a reliable collaborative entity, breaks through the high dependence of traditional centralized fault decision-making on the main station. Even if the main station fails or communication is interrupted, the edge side can still autonomously complete fault judgment and handling, significantly improving the anti-interference capability and operational stability of the distribution network fault handling system. Furthermore, this invention, through the design of a key anomaly feature priority judgment mechanism, allows any edge node to directly output a deterministic conclusion upon detecting a key feature of a sudden fault, eliminating the need for data transmission and analysis from the main station. This significantly shortens the fault decision-making link and response delay, effectively reducing user power outage time and meeting [the requirements of]... This invention addresses the need for rapid fault handling in distribution networks. It utilizes a multi-node feature consistency verification unit to achieve weighted multi-feature fusion similarity calculation and contradictory data verification. By relying on multi-node data cross-validation to replace single-node analysis, it effectively eliminates invalid and extreme anomaly data caused by electromagnetic interference and environmental noise, significantly reducing fault misjudgment and missed judgment rates, and improving the accuracy of fault location and early warning. Furthermore, through secondary data acquisition and feature deviation rate elimination mechanisms in scenarios where consensus is not reached, this invention forms a complete analysis process of "rapid judgment - accurate verification - closed-loop optimization." Simultaneously, it relies on edge-side autonomous collaboration to achieve distributed decision-making, eliminating the need for centralized control by the main station, greatly improving distribution network operation and maintenance efficiency, ensuring power supply reliability, and reducing distribution network operation and maintenance costs. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of a module of an intelligent power distribution network fault location and early warning system based on multi-source sensing collaboration according to the present invention. Detailed Implementation

[0017] 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, and 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.

[0018] Please see Figure 1 The present invention provides the following technical solution: A smart distribution network fault location and early warning system based on multi-source sensing collaboration includes: an edge sensing node module, a self-organizing network module, a data sharing module, a collaborative judgment module, and a decision output module; Edge sensing node modules are deployed on the distribution network lines to collect distribution network current, electric field, and vibration signals in real time. When any node detects an abnormal signal that meets the preset single-dimensional threshold and has passed preliminary validity verification, it triggers a distributed collaborative decision-making process. The self-organizing networking module initiates self-organizing networking, scans neighboring edge sensing nodes, selects core nodes through a filtering strategy that correlates signal quality with spatial distance, and builds a temporary collaborative consensus group. The data sharing module is used to enable the sharing of pre-processed feature data of each edge sensing node within the group. The feature data includes current amplitude, vibration frequency and electric field distortion rate. The collaborative analysis module, based on the feature data shared by the data sharing module, forms a consensus analysis result through multi-node feature consistency verification and priority judgment of key abnormal features. When more than half of the nodes detect consistent feature data without contradictions, it is determined that there is a potential fault in the distribution network. When any node detects key abnormal feature data that characterizes a sudden fault in the distribution network, it is directly confirmed that a fault has occurred in the distribution network. If no consensus analysis result is reached, a secondary data collection and feature refinement verification process is triggered. The decision output module is used to push early warning information when a potential fault is detected, generate fault location results and handling plans after a fault is confirmed, and simultaneously upload the early warning information, fault location results and handling plans to the distribution network master station.

[0019] The edge sensing node module includes a multi-source signal acquisition and preprocessing unit and an anomaly detection and triggering unit; The multi-source signal acquisition and preprocessing unit acquires power distribution network current, electric field, and vibration signals in real time, and performs noise reduction, format standardization, and preliminary validity verification on the raw signals to eliminate invalid data caused by environmental interference. The anomaly detection and triggering unit compares the preprocessed signals with preset single-dimensional thresholds to determine whether there are any anomalies. When an anomaly signal that meets the threshold conditions and has been verified is detected, a trigger command is generated to start the distributed collaborative decision-making process.

[0020] In this embodiment, the multi-source signal acquisition and preprocessing unit uses a synchronous sampling method to acquire signals in real time. It is assumed that the sampling frequency of current and electric field signals is set to 100Hz and the sampling frequency of vibration signal is set to 200Hz. During acquisition, the three types of signals are time-stamped to ensure spatiotemporal consistency. The original signal is denoised using a weighted moving average method to eliminate environmental interference and sensor noise. The calculation formula is: X1 i =(∑ k∈[0,n-1] wk·X i-k ) / n, where X1 i Let be the denoised signal value, n be the sliding window length, wk be the weighting coefficient, and ∑ k∈[0,n-1] wk=1; X i-kThe original signal value is used; after noise reduction, the signal is mapped to the [0,1] interval using the min-max normalization method to unify the judgment benchmark, thereby obtaining the standardized signal value X2. i ; According to Xmin≤X2 i ≤Xmax is used for validity verification, where Xmin and Xmax are the minimum and maximum values ​​of the validity threshold range, respectively. Invalid data is removed and then transferred to the next unit.

[0021] In this embodiment, the anomaly detection and triggering unit sets a single-dimensional anomaly interval threshold [Xt] for each of the three types of standardized signals. min ,Xt max If the signal value exceeds this range, it is marked as an abnormal signal; The system employs a logic of "any abnormal signal determines a node to be abnormal," comparing preprocessed signals one by one with corresponding thresholds to determine whether a node has detected an abnormal signal. The process triggering requires two conditions: first, the node must detect an abnormal signal; second, the signal must have passed preliminary validity verification. If both conditions are met, a trigger command is generated, and abnormal feature data is synchronized to the self-organizing network module, initiating the subsequent process. If only one condition is met, the node is deemed an invalid anomaly, and the event is only recorded for future reference.

[0022] The self-organizing network module includes a node scanning unit, a core node screening unit, and a temporary collaborative consensus group construction unit; The node scanning unit initiates a self-organizing network process, scanning neighboring edge sensing nodes, identifying the communication status and location information of each node, and initially marking communicable nodes to form a candidate node list. The core node screening unit, based on the candidate node list, adopts a screening strategy that correlates signal quality and spatial distance to comprehensively evaluate the candidate nodes, thereby obtaining the core nodes. The temporary collaborative consensus group construction unit, based on the screened core nodes, forms a temporary collaborative consensus group, establishes dedicated communication links between nodes within the group, and monitors the online status and communication stability of each core node within the group in real time.

[0023] The node scanning unit initiates a broadcast detection signal based on a preset communication protocol. The detection range covers adjacent edge sensing nodes within the monitoring area of ​​the distribution network line, and records the transmission time of the detection signal. It receives the response signals from each edge sensing node, extracts the unique identifier, real-time communication status parameters, and location coordinate information of each response node, and records the reception time of the response signal. For each responding node, the communication response time is obtained by the difference between the transmission time and the reception time. A communication response time threshold is set, and the communication response time of each responding node is compared with the threshold. Nodes whose communication response time does not exceed the threshold are judged to have good communication status and are marked into the candidate node list. Nodes whose communication response time exceeds the threshold are judged to have excessive communication latency and are removed. Finally, a candidate node list containing the identifier, location and communication status of communicable nodes is formed.

[0024] In this embodiment, the node scanning unit initiates a broadcast detection signal based on a preset communication protocol. The detection range covers adjacent edge sensing nodes within the monitoring area of ​​the distribution network line, and records the transmission timestamp t0 of the detection signal. It receives the response signals from each edge sensing node, extracts the unique identifier ID, real-time communication status parameters, and location coordinate information of each response node, and records the response signal reception timestamp t1. For each responding node, its communication response time Δt is calculated to initially determine the node's communication status. The calculation formula is: Δt = t1 - t0, where Δt is the node's communication response time. A communication response time threshold ΔT is set. When Δt ≤ ΔT, the node is considered to have a good communication status, is marked, and included in the candidate node list. When Δt > ΔT, the node is considered to have excessive communication latency and is removed from the candidate range. Finally, a complete candidate node list containing the identifier, location, and communication status of communicable nodes is formed.

[0025] The filtering process for the core node filtering unit is as follows: The basic parameters of candidate nodes and the current distribution network scenario information are extracted. All parameters are directly collected or preset by the node's built-in module, without the need for complex derivation and calculation. The basic parameters of candidate nodes include the received signal strength and the straight-line distance between the candidate node and the source node. The received signal strength is directly read by the node's radio frequency module, and the straight-line distance between the candidate node and the source node is directly calculated based on the preset latitude and longitude coordinates when the node is deployed. The distribution network scenario information includes the line topology type and the electromagnetic interference level. The line topology type is preset as a straight line or a branch line, with a corresponding preset topology correction coefficient. The electromagnetic interference level is directly output by the node's interference detection module as level one, level two, and level three results. For received signal strength, the minimum and maximum received signal strength values ​​in the candidate node set are statistically analyzed and normalized. The normalization results are then adjusted based on the preset correction value corresponding to the electromagnetic interference level. For the straight-line distance between the candidate node and the source node, quantification is performed using the preset maximum effective communication distance and the preset topology correction coefficient for the corresponding line topology type. Different preset topology correction coefficients are configured for straight lines and branch lines. The basic evaluation value of the candidate node is calculated using a weighted approach based on electromagnetic interference level. At low interference levels, the signal quality weight and spatial adaptation weight are equal; at medium interference levels, the signal quality weight is greater than the spatial adaptation weight; and at high interference levels, the signal quality weight is further increased compared to the medium interference level, with the sum of the signal quality weight and the spatial adaptation weight being one. Candidate nodes are sorted from largest to smallest based on their baseline evaluation values. A predetermined number of nodes at the top of the list are selected to form a preliminary core node set, with the predetermined number being appropriate for the scale of collaborative analysis. The straight-line distance between any two nodes in the preliminary core node set is calculated and compared with a predetermined conflict distance threshold. If there is a situation where the distance between two nodes is less than the predetermined conflict distance threshold, the node with the higher baseline evaluation value is retained and the other node is removed. The node with the next highest baseline evaluation value is selected from the candidate node list and added to the preliminary set. The above verification process is repeated until the distance between all nodes in the set is not less than the predetermined conflict distance threshold, thus forming the final core node set.

[0026] In this implementation, the core node screening unit extracts the basic parameters of candidate nodes and the current distribution network scenario information. All parameters are directly collected or preset by the node's built-in module, without the need for complex derivation and calculation. The basic parameters of the candidate nodes include the received signal strength Pj, which is directly read by the node's radio frequency module and is used to characterize the communication link quality; the straight-line distance dj between the candidate node and the source node, which is directly calculated based on the latitude and longitude coordinates preset during node deployment; the distribution network scenario information includes the line topology type, which is preset to a straight line or a branch line, with a corresponding topology correction coefficient γ; and the electromagnetic interference level, which is directly output by the node interference detection module as level three, level two, and level one results, quantized as I=0.2, 0.5, and 0.8 respectively, to adapt to dynamic interference scenarios. The core evaluation factors are quantified, and the quantification process incorporates distribution network scenario adaptation logic, which simplifies calculations and ensures evaluation accuracy. The signal quality factor Qj is calculated by normalizing the received signal strength Pj and introducing an electromagnetic interference level correction coefficient to offset the bias in signal quality assessment caused by different interference scenarios. The calculation formula is: Qj=(Pj-Pmin) / (Pmax-Pmin)×(1+0.2×(1-I)), where Pmin and Pmax are the minimum and maximum received signal strengths in the candidate node set, respectively, which can be obtained by real-time statistical analysis of candidate node parameters; (1+0.2×(1-I)) is the interference adaptation coefficient, whose core function is to moderately improve signal quality in low-interference scenarios. The signal quality factor weight weakens the impact of interference on signal evaluation in high-interference scenarios, ensuring the objectivity of signal quality evaluation in different scenarios. The spatial adaptation factor Dj is quantified based on the straight-line distance dj between the candidate node and the source node, and a topology correction coefficient γ is introduced to adapt to the characteristics of the distribution network line topology. For example, γ=1.0 is configured for straight lines and γ=0.9 is configured for branch lines to avoid distance evaluation distortion caused by branch lines being blocked. The calculation formula is: Dj=1-[dj / (Lmax×γ)], where Lmax is the preset maximum effective communication distance, for example, 500m in the distribution network scenario. The value range of Dj is [0,1]. The closer the value is to 1, the more the node's spatial location is adapted to the collaborative judgment requirements. A scenario-based dynamic weight allocation strategy is adopted to calculate the basic evaluation value Sj of candidate nodes, abandoning the traditional fixed weight mode. Based on the preset electromagnetic interference level, the weights are adapted accordingly, eliminating the need for complex real-time calculations. This simplifies the logic and conforms to the dynamic changes in distribution network interference. For example, the specific weight allocation rules are as follows: at the third level (I=0.2), the signal quality weight α=0.5 and the spatial adaptation weight β=0.5; at the second level (I=0.5), the signal quality weight α=0.6 and the spatial adaptation weight β=0.4; at the first level (I=0.8), the signal quality weight α=0.7 and the spatial adaptation weight β=0.3, and α+β=1 is satisfied. The core of this weight allocation logic is that the stronger the interference, the higher the priority is given to ensuring communication stability. The basic evaluation value is calculated as follows: Sj = α·Qj + β·Dj. Candidate nodes are sorted from largest to smallest according to Sj, and the top m nodes are selected to form a preliminary core node set, where m is the preset number of core nodes. For example, in the distribution network scenario, 3-5 nodes are configured to adapt to the scale requirements of collaborative assessment.

[0027] Calculate the straight-line distance djk between any two nodes in the initial core node set, and preset the conflict distance threshold d0. For example, in the distribution network scenario, d0 is set to 50m to adapt to the node deployment density requirements. If there is a case where djk < d0, it is determined that there is a communication conflict between the two nodes. The node with the higher Sj value is retained, the other node is removed, and the next node with a high Sj value is selected from the candidate node list to be added to the initial set. Repeat the above verification process until the distance between all nodes in the set satisfies djk ≥ d0, and finally a core node set without communication conflicts and adapted to the current distribution network scenario is formed.

[0028] The temporary collaborative consensus group construction unit sends a multicast networking instruction containing the unique identifier of the temporary collaborative consensus group, preset communication link parameters, and data transmission protocol to all nodes in the core node set. After receiving and responding to the confirmation instruction, the core nodes complete the construction of the temporary collaborative consensus group, establish a point-to-point exclusive communication link between any two core nodes in the group based on the preset encrypted communication protocol, and allocate an independent communication channel for each link. Heartbeat probe packets are periodically sent to each communication link within the group. The link communication delay is obtained by recording the sending time of the heartbeat probe packets and the receiving time of the response heartbeat packets. The link data packet loss rate is obtained by statistically analyzing the total number of heartbeat probe packets sent and the number of successfully received response heartbeat packets within a preset time period. The link communication delay is standardized by combining it with the preset maximum allowable communication delay. The link stability factor is obtained by combining the standardized result with the link data packet loss rate. The system counts the number of consecutive heartbeat responses from each core node within the group, sets an offline threshold, and determines that a core node is offline when the number of consecutive non-response counts reaches this threshold. It also sets a link stability threshold, triggering a link reconnection mechanism when the stability factor of a communication link falls below this threshold. Offline nodes are removed from the temporary collaborative consensus group, and nodes ranked immediately after the original core node are selected from the candidate node list output by the core node screening unit to be added to the group, and a new point-to-point dedicated communication link is established. This monitoring process is repeated until all nodes in the temporary collaborative consensus group remain online and the stability factor of all communication links is not lower than the link stability threshold.

[0029] In this embodiment, the temporary collaborative consensus group construction unit sends a multicast networking instruction to all nodes in the core node set. The networking instruction includes the unique identifier of the temporary collaborative consensus group, preset communication link parameters, and data transmission protocol. After the core nodes receive and respond with the confirmation instruction, the temporary collaborative consensus group is established. Based on the preset encrypted communication protocol, a point-to-point dedicated communication link is established between any two core nodes in the group, and an independent communication channel is allocated to each link to avoid signal interference with external communication links. For each communication link within the group, heartbeat probe packets are periodically sent to the peer node. The sending timestamp ts of the heartbeat probe packets and the receiving timestamp tr of the response heartbeat packets are recorded. The link communication delay τ is calculated using the formula: τ = tr - ts. The total number of heartbeat probe packets sent Nt and the number of successfully received response heartbeat packets Ns within a preset time period T are counted. The link data packet loss rate L is calculated using the formula: L = 1 - (Ns / Nt). Based on the link communication delay τ and the data packet loss rate L, the link stability factor K is calculated, and K = 1 / (1 + τ' + L), where τ' is the standardized result of the link communication delay, calculated from τ and the preset maximum allowable communication delay τmax, i.e., τ' = τ / τmax. The value range of τ' is [0,1], and the value range of the link stability factor K is also [0,1]. The closer the value is to 1, the stronger the link stability. For the online status of each core node in the group, the number of consecutive times Nw that a node fails to respond to a heartbeat packet is counted, and an offline judgment threshold N0 is set. When Nw ≥ N0, the core node is judged to be offline. A link stability threshold K0 is set. When the stability factor K of a certain communication link is < K0, the stability of the link is judged to be unsatisfactory, and the link reconnection mechanism is immediately triggered. For offline nodes, they are removed from the temporary collaborative consensus group, and based on the candidate node list output by the core node screening unit, the node ranked immediately after the original core node is selected to be added to the group. A new point-to-point dedicated communication link is re-established, and the above monitoring process is repeated until all nodes in the temporary collaborative consensus group are online and the stability factor of all communication links is not less than the link stability threshold K0, thereby maintaining the stable operation of the temporary collaborative consensus group.

[0030] The collaborative analysis module includes a multi-node feature consistency verification unit and a key anomaly feature priority determination unit; The multi-node feature consistency verification unit performs consistency comparison and verification on the feature data of each edge sensing node in the temporary collaborative consensus group based on the feature data shared by the data sharing module, counts the number of nodes with consistent feature data, and checks whether there is contradictory feature data in the group; when the number of nodes with consistent feature data exceeds half and there is no contradictory data, a preliminary judgment conclusion that there is a potential fault in the distribution network is output. The key anomaly feature priority judgment unit identifies whether the feature data collected by each node in the group contains key anomaly feature data that characterizes a sudden fault in the distribution network. If any node is found to have such key anomaly feature data, a definitive judgment conclusion that the distribution network has failed is directly output. In the case where no consensus judgment result is reached after the feature consistency verification of multiple nodes, a secondary feature data collection and feature refinement verification process is triggered.

[0031] The multi-node feature consistency verification unit specifically includes: The feature data of all edge sensing nodes in the temporary collaborative consensus group are normalized to obtain the normalized current amplitude, vibration frequency and electric field distortion rate. The core node in the group is used as the benchmark node. The comprehensive feature similarity between the remaining nodes and the benchmark node is calculated by combining the feature weights. The feature consistency between a single node and the benchmark node is quantified. The feature weights are set based on the priority of distribution network fault judgment and the sum of the weights is one. A feature similarity threshold is set based on the distribution network experimental data. When the comprehensive feature similarity between a node and the benchmark node is not lower than the threshold, the node is determined to be consistent with the benchmark node. The total number of nodes with consistent features within the group is counted, and the ratio of this total number to the total number of nodes within the temporary collaborative consensus group is calculated as the consistency ratio. If the comprehensive feature similarity between a node and the baseline node is lower than the preset threshold for contradictory data, it is determined that there is contradictory data in the group; if there is no such node, it is determined that there is no contradictory data. When the consistency ratio exceeds one-half and there is no contradictory data, a preliminary conclusion that there is a potential fault in the distribution network is output, that is, a consensus is reached. When the consistency ratio does not exceed one-half or there is contradictory data, it is determined that no consensus has been reached, and the verification result is transferred to the key abnormal feature priority judgment unit for processing.

[0032] In this embodiment, the specific steps of the multi-node feature consistency verification unit are as follows: The feature data of all edge sensing nodes in the temporary collaborative consensus group are normalized to eliminate the influence of differences in different dimensions and numerical ranges on the verification results and ensure that the weights of each feature are balanced; thus, the normalized current amplitude Ai, vibration frequency Fi and electric field distortion rate Di are obtained. Using the core node within the group as the baseline node, calculate the comprehensive feature similarity between the remaining nodes and the baseline node to quantify the feature consistency between a single node and the baseline node. The calculation formula is as follows: Zi = wA·|1-|Ai-A0||+wF·|1-|Fi-F0||+wD·|1-|Di-D0||, Where Zi is the comprehensive feature similarity between the i-th node and the baseline node, with a value range of [0,1]. The closer the value is to 1, the stronger the feature consistency. A0, F0, and D0 are the normalized results of the three types of features of the baseline node, respectively. wA, wF, and wD are feature weights, set based on the priority of distribution network fault assessment and satisfying wA+wF+wD=1, for example, wA=0.4, wF=0.3, and wD=0.3. A feature similarity threshold Z is set, preset based on distribution network experimental data. Assuming Zt=0.8, when Zi≥Z, Determine that the i-th node has the same characteristics as the baseline node, count the total number of nodes with the same characteristics in the group Nc, calculate the consistency ratio R, and R = Nc / N, where N is the total number of nodes in the temporary collaborative consensus group; if there are nodes Zi < 0.5, it is determined that there is contradictory data in the group, otherwise it is determined that there is no contradictory data. When R > 0.5 and there is no contradictory data, output the preliminary conclusion that there is a potential fault in the distribution network, that is, consensus is reached; if R ≤ 0.5 or there is contradictory data, it is determined that consensus has not been reached and the case is transferred to the key abnormal feature priority determination unit for processing.

[0033] The key anomaly feature priority determination unit specifically includes: Based on distribution network industry standards and fault cases, key abnormal feature thresholds for sudden faults in distribution networks are set, and the original feature data collected from each node are verified one by one to directly determine whether a sudden fault exists. Among them, current amplitude, vibration frequency, and electric field distortion rate correspond to preset key thresholds for sudden faults, and each threshold is set with appropriate values ​​based on distribution network industry standards and fault cases. If any node detects any of the above-mentioned key abnormal features, a definitive conclusion that a distribution network fault has occurred is directly output without subsequent verification. When the multi-node feature consistency verification fails to reach a consensus and no node detects a key abnormal feature, a secondary data acquisition and feature refinement verification process is triggered. First, all nodes in the group are controlled to improve the acquisition accuracy and shorten the acquisition cycle, and the three types of feature data are re-acquired and the above normalization process is repeated. Then, the feature deviation rate between the secondary acquisition data and the original feature data acquired in the first acquisition is calculated using the same feature weights as the multi-node feature consistency verification unit to eliminate the influence of random errors. A preset deviation rate threshold is set, and abnormal node data with a feature deviation rate exceeding the threshold are removed. Based on the remaining valid data, the feature similarity and consistency ratio are recalculated until a consensus is reached or it is confirmed that there is no clear fault and the corresponding conclusion is output.

[0034] In this implementation, the key anomaly feature priority determination unit specifically includes: Based on industry standards and fault cases in the distribution network, key abnormal characteristic thresholds for sudden faults in the distribution network are set. The original characteristic data of each node are verified one by one to directly determine whether a sudden fault exists. For example, the key abnormal judgment standard for current amplitude is that the current amplitude Ai collected by the node is ≥ Aw, where Aw is the preset sudden fault current threshold, which is 1.5 times the rated current of the distribution network. The key abnormal judgment standard for vibration frequency is that the vibration frequency Fi collected by the node is ≥ Fw, where Fw is the preset sudden fault vibration threshold, which is 2 times the upper limit of the normal vibration frequency of the distribution network. The key abnormal judgment standard for electric field distortion rate is that the electric field distortion rate Di collected by the node is ≥ Dw, where Dw is the preset sudden fault electric field threshold, which is 3 times the upper limit of the normal electric field distortion rate of the distribution network. As long as any node detects any of the above key abnormal characteristics, a definitive conclusion that a fault has occurred in the distribution network is directly output without subsequent verification. When multi-node feature consistency verification fails to reach a consensus and no node detects key abnormal features, a secondary data acquisition and feature refinement verification process is triggered. First, all nodes within the group are controlled to increase acquisition accuracy and shorten the acquisition cycle to half of the original cycle. The three types of feature data are re-acquired, and the above normalization process is repeated. Then, the feature deviation rate between the secondary and primary acquisition data is calculated to eliminate the influence of random errors. The calculation formula is as follows: ΔCi=wA·|Ai2-Ai1| / Ai1+wF·|Fi2-Fi1| / Fi1+wD·|Di2-Di1| / Di1, Where ΔCi is the feature deviation rate of the i-th node, Ai1, Fi1, and Di1 are the original feature data collected for the first time, and Ai2, Fi2, and Di2 are the original feature data collected for the second time. The weights wA, wF, and wD are consistent with the multi-node feature consistency verification unit. A deviation rate threshold ΔC0 is set, for example, ΔC0=0.2. Abnormal node data with ΔCi>ΔC0 are removed. Based on the remaining valid data, the feature similarity and consistency ratio are recalculated until a consensus is reached or it is confirmed that there is no obvious fault and the corresponding conclusion is output.

[0035] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0036] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart distribution network fault location and early warning system based on multi-source sensing collaboration, characterized in that: The system includes: an edge sensing node module, a self-organizing network module, a data sharing module, a collaborative analysis module, and a decision output module; Edge sensing node modules are deployed on the distribution network lines to collect distribution network current, electric field, and vibration signals in real time. When any node detects an abnormal signal that meets the preset single-dimensional threshold and has passed preliminary validity verification, it triggers a distributed collaborative decision-making process. The self-organizing networking module initiates self-organizing networking, scans neighboring edge sensing nodes, selects core nodes through a filtering strategy that correlates signal quality with spatial distance, and builds a temporary collaborative consensus group. The data sharing module is used to enable the sharing of pre-processed feature data of each edge sensing node within the group. The feature data includes current amplitude, vibration frequency and electric field distortion rate. The collaborative analysis module, based on the feature data shared by the data sharing module, forms a consensus analysis result through multi-node feature consistency verification and priority judgment of key abnormal features. When more than half of the nodes detect consistent feature data without contradictions, it is determined that there is a potential fault in the distribution network. When any node detects key abnormal feature data that characterizes a sudden fault in the distribution network, it is directly confirmed that a fault has occurred in the distribution network. If no consensus analysis result is reached, a secondary data collection and feature refinement verification process is triggered. The decision output module is used to push early warning information when a potential fault is identified, generate fault location results and handling plans after a fault is confirmed, and simultaneously upload the early warning information, fault location results and handling plans to the distribution network master station. The self-organizing network module includes a node scanning unit, a core node screening unit, and a temporary collaborative consensus group construction unit. The core node filtering unit specifically includes: The basic parameters of candidate nodes and the current distribution network scenario information are extracted. All parameters are directly collected or preset by the node's built-in module, without the need for complex derivation and calculation. The basic parameters of candidate nodes include the received signal strength and the straight-line distance between the candidate node and the source node. The received signal strength is directly read by the node's radio frequency module, and the straight-line distance between the candidate node and the source node is directly calculated based on the preset latitude and longitude coordinates when the node is deployed. The distribution network scenario information includes the line topology type and the electromagnetic interference level. The line topology type is preset as a straight line or a branch line, with a corresponding preset topology correction coefficient. The electromagnetic interference level is directly output by the node's interference detection module as level one, level two, and level three results. For received signal strength, the minimum and maximum received signal strength values ​​in the candidate node set are statistically analyzed and normalized. The normalization results are then adjusted based on the preset correction values ​​corresponding to the electromagnetic interference level. For the straight-line distance between the candidate node and the source node, the preset maximum effective communication distance and the preset topology correction coefficient for the corresponding line topology type are used for quantification. Different preset topology correction coefficients are configured for straight lines and branch lines. The basic evaluation value of the candidate node is calculated by configuring weights based on the electromagnetic interference level. Candidate nodes are sorted from largest to smallest based on their baseline evaluation values. A predetermined number of nodes at the top of the list are selected to form a preliminary core node set, with the predetermined number being appropriate for the scale of collaborative analysis. The straight-line distance between any two nodes in the preliminary core node set is calculated and compared with a predetermined conflict distance threshold. If there is a situation where the distance between two nodes is less than the predetermined conflict distance threshold, the node with the higher baseline evaluation value is retained and the other node is removed. The node with the next highest baseline evaluation value is selected from the candidate node list and added to the preliminary set. The above verification process is repeated until the distance between all nodes in the set is not less than the predetermined conflict distance threshold, thus forming the core node set.

2. The intelligent distribution network fault location and early warning system based on multi-source sensing collaboration according to claim 1, characterized in that: The edge sensing node module includes a multi-source signal acquisition and preprocessing unit and an anomaly detection and triggering unit; The multi-source signal acquisition and preprocessing unit acquires power distribution network current, electric field, and vibration signals in real time, and performs noise reduction, format standardization, and preliminary validity verification on the original signals to remove invalid data caused by environmental interference; the anomaly detection and triggering unit compares the preprocessed signal with a preset single-dimensional threshold to determine whether there is an anomaly. When an abnormal signal that meets the threshold condition and has been verified is detected, a trigger command is generated to start the distributed collaborative decision-making process.

3. The intelligent distribution network fault location and early warning system based on multi-source sensing collaboration according to claim 1, characterized in that: The self-organizing network module includes a node scanning unit, a core node screening unit, and a temporary collaborative consensus group construction unit. The node scanning unit initiates a self-organizing networking process, scans adjacent edge sensing nodes, identifies the communication status and location information of each node, initially marks communicable nodes, and forms a candidate node list. The core node screening unit, based on the candidate node list, adopts a screening strategy that correlates signal quality with spatial distance to comprehensively evaluate the candidate nodes, thereby obtaining the core nodes. The temporary collaborative consensus group construction unit forms a temporary collaborative consensus group based on the selected core nodes, establishes a dedicated communication link between nodes within the group, and monitors the online status and communication stability of each core node within the group in real time.

4. The intelligent distribution network fault location and early warning system based on multi-source sensing collaboration according to claim 3, characterized in that: The node scanning unit specifically includes: Based on a preset communication protocol, a broadcast detection signal is initiated, covering adjacent edge sensing nodes within the monitoring area of ​​the distribution network line, and the transmission time of the detection signal is recorded; the response signals fed back by each edge sensing node are received, and the unique identifier, real-time communication status parameters and location coordinate information of each response node are extracted, and the reception time of the response signal is recorded; For each responding node, the communication response time is obtained by the difference between the transmission time and the reception time. A communication response time threshold is set, and the communication response time of each responding node is compared with the threshold. Nodes whose communication response time does not exceed the threshold are judged to have good communication status and are marked into the candidate node list. Nodes whose communication response time exceeds the threshold are judged to have excessive communication latency and are removed. Finally, a candidate node list containing the identifier, location and communication status of communicable nodes is formed.

5. The intelligent distribution network fault location and early warning system based on multi-source sensing collaboration according to claim 3, characterized in that: The temporary collaborative consensus group construction unit specifically includes: A multicast networking instruction containing a unique identifier for the temporary collaborative consensus group, preset communication link parameters, and data transmission protocol is sent to all nodes in the core node set. After receiving and responding to the confirmation instruction, the core nodes complete the formation of the temporary collaborative consensus group, establish a point-to-point dedicated communication link between any two core nodes in the group based on the preset encrypted communication protocol, and allocate an independent communication channel for each link. Heartbeat probe packets are periodically sent to each communication link within the group. The link communication delay is obtained by recording the sending time of the heartbeat probe packets and the receiving time of the response heartbeat packets. The link data packet loss rate is obtained by statistically analyzing the total number of heartbeat probe packets sent and the number of successfully received response heartbeat packets within a preset time period. The link communication delay is standardized by combining it with the preset maximum allowable communication delay. The link stability factor is obtained by combining the standardized result with the link data packet loss rate. The system counts the number of consecutive heartbeat responses that each core node in the group fails to send. An offline threshold is set, and when the number of consecutive failures reaches this threshold, the core node is considered offline. A link stability threshold is also set; when the stability factor of a communication link falls below this threshold, a link reconnection mechanism is triggered. Offline nodes are removed from the temporary collaborative consensus group. Based on the candidate node list output by the core node screening unit, nodes ranked immediately after the original core node are added to the group, and a new point-to-point dedicated communication link is established. This monitoring process is repeated until all nodes in the temporary collaborative consensus group remain online, and the stability factor of all communication links is not lower than the link stability threshold.

6. The intelligent distribution network fault location and early warning system based on multi-source sensing collaboration according to claim 1, characterized in that: The collaborative analysis module includes a multi-node feature consistency verification unit and a key anomaly feature priority determination unit; The multi-node feature consistency verification unit, based on the feature data shared by the data sharing module, performs consistency comparison and verification on the feature data of each edge sensing node in the temporary collaborative consensus group, counts the number of nodes with consistent feature data, and checks whether there is contradictory feature data in the group. When the number of nodes with consistent feature data exceeds half and there is no contradictory data, a preliminary judgment conclusion is output that there is a potential fault in the distribution network. The key anomaly feature priority judgment unit identifies whether the feature data collected by each node in the group contains key anomaly feature data that characterizes a sudden fault in the distribution network. If any node is found to have such key anomaly feature data, a definitive judgment conclusion that the distribution network has failed is directly output. In the case where no consensus judgment result is reached after the feature consistency verification of multiple nodes, a secondary feature data collection and feature refinement verification process is triggered.

7. The intelligent distribution network fault location and early warning system based on multi-source sensing collaboration according to claim 6, characterized in that: The multi-node feature consistency verification unit specifically includes: The feature data of all edge sensing nodes in the temporary collaborative consensus group are normalized to obtain the normalized current amplitude, vibration frequency and electric field distortion rate. The core node in the group is used as the benchmark node. The comprehensive feature similarity between the remaining nodes and the benchmark node is calculated by combining the feature weights. The feature consistency between a single node and the benchmark node is quantified. The feature weights are set based on the priority of distribution network fault judgment and the sum of the weights is one. A feature similarity threshold is set based on the distribution network experimental data. When the comprehensive feature similarity between a node and the benchmark node is not lower than the threshold, the node is determined to be consistent with the benchmark node. The total number of nodes with consistent features within the group is counted, and the ratio of this total number to the total number of nodes within the temporary collaborative consensus group is calculated as the consistency ratio. If the comprehensive feature similarity between a node and the baseline node is lower than the preset threshold for contradictory data, it is determined that there is contradictory data in the group; if there is no such node, it is determined that there is no contradictory data. When the consistency ratio exceeds one-half and there is no contradictory data, a preliminary conclusion that there is a potential fault in the distribution network is output. When the consistency ratio does not exceed one-half or there is contradictory data, it is determined that no consensus has been reached, and the verification result is transferred to the key abnormal feature priority judgment unit for processing.

8. The intelligent distribution network fault location and early warning system based on multi-source sensing collaboration according to claim 6, characterized in that: The key anomaly feature priority determination unit specifically includes: Based on distribution network industry standards and fault cases, key abnormal feature thresholds for sudden faults in distribution networks are set, and the original feature data collected from each node are verified one by one to directly determine whether a sudden fault exists. Among them, current amplitude, vibration frequency, and electric field distortion rate correspond to the key thresholds for sudden faults preset based on distribution network industry standards and fault cases. If any node detects any of the above-mentioned key abnormal features, a definitive conclusion that a distribution network fault has occurred is directly output without subsequent verification. When the multi-node feature consistency verification fails to reach a consensus and no node detects a key abnormal feature, a secondary data acquisition and feature refinement verification process is triggered. First, all nodes in the group are controlled to improve the acquisition accuracy and shorten the acquisition cycle, and the three types of feature data are re-acquired and the above normalization process is repeated. Then, the feature deviation rate between the secondary acquisition data and the original feature data acquired in the first acquisition is calculated using the same feature weights as the multi-node feature consistency verification unit to eliminate the influence of random errors. A preset deviation rate threshold is set, and abnormal node data with a feature deviation rate exceeding the threshold are removed. Based on the remaining valid data, the feature similarity and consistency ratio are recalculated until a consensus is reached or it is confirmed that there is no clear fault and the corresponding conclusion is output.

Images