Low-voltage power distribution cabinet fault self-diagnosis method and system based on edge computing

Abstract

The application discloses a low-voltage power distribution cabinet fault self-diagnosis method and system based on edge computing and relates to the technical field of power distribution cabinets. The method comprises the following steps: deploying a high-frequency pickup unit in the circuit topology structure of a low-voltage power distribution cabinet; driving the high-frequency pickup unit to detect full-channel waveforms according to the protection action trigger of a protection circuit by an edge controller; determining time-varying impedance trajectories according to the full-channel waveforms, and measuring impedance deviation data based on quantitative indicators; inputting the impedance deviation data into a fault decision maker preset on the edge side to determine transient fault data, and determining periodic fault data based on the digital modeling of the protection action event and the time sequence causal reverse positioning of the node protection action; and performing instant fault management and testability fault management on the low-voltage power distribution cabinet according to the transient fault data and the periodic fault data. The technical problem of low fault diagnosis efficiency of the low-voltage power distribution cabinet in the prior art is solved, and the technical effect of improving the fault diagnosis efficiency is achieved.

Description

Technical Field

[0001] This invention relates to the field of power distribution cabinet technology, and more specifically to a method and system for self-diagnosis of faults in low-voltage power distribution cabinets based on edge computing. Background Technology

[0002] Low-voltage switchgear, as a key piece of equipment for power distribution and consumption control at the end of the power system, is widely used in industrial plants, commercial buildings, and public infrastructure, undertaking important functions such as power distribution, line protection, and operation monitoring. During actual operation, due to factors such as load fluctuations, aging electrical components, poor contact, short-circuit impacts, and changes in ambient temperature and humidity, transient or periodic latent faults are prone to occur inside low-voltage switchgear. In existing technologies, fault diagnosis typically relies on the operating signals of circuit breakers and relay protection devices or simple current and voltage sampling data for judgment, often using manual inspection or centralized back-end analysis to complete fault diagnosis. This approach suffers from problems such as low data sampling frequency, limited information dimensions, and a lack of effective modeling capabilities for time-varying impedance characteristics, making it difficult to finely characterize the fault evolution process. Especially for latent faults such as transient arcs and intermittent poor contact, accurate identification and rapid location are often impossible, leading to delayed fault response and affecting the safety and continuity of power supply in the distribution system. Summary of the Invention

[0003] This application provides a method and system for self-diagnosis of faults in low-voltage distribution cabinets based on edge computing, which solves the technical problem of low efficiency in fault diagnosis of low-voltage distribution cabinets in the prior art.

[0004] The first aspect of this application provides a self-diagnosis method for low-voltage distribution cabinet faults based on edge computing, the method comprising: A high-frequency pickup unit is deployed in the circuit topology of the low-voltage distribution cabinet. The edge controller is triggered by the protection action of the protection circuit and drives the high-frequency pickup unit to detect the waveform of the entire channel. The deployment position is at the input and output terminals of each protection circuit. The time-varying impedance trajectory is determined based on the waveform of the entire channel, and the impedance deviation data based on quantitative indicators is measured. The impedance deviation data is input into the fault decision unit preset on the edge side to determine the transient fault data. Based on the digital modeling of protection action events and the time-series causal reverse location of node protection actions, the periodic fault data is determined. Based on the transient fault data and the periodic fault data, real-time fault management and test fault management are performed on the low-voltage distribution cabinet.

[0005] A second aspect of this application provides a self-diagnostic system for low-voltage switchgear faults based on edge computing, the system comprising: Waveform Detection Module: A high-frequency pickup unit is deployed in the circuit topology of the low-voltage distribution cabinet. The edge controller triggers the high-frequency pickup unit to detect the waveform of the entire channel based on the protection action of the protection circuit. The deployment location is at the input and output terminals of each protection circuit. Impedance Trajectory Determination Module: The time-varying impedance trajectory is determined based on the waveform of the entire channel, and the impedance deviation data based on quantitative indicators is measured. Fault Data Determination Module: The impedance deviation data is input into the fault decision unit preset on the edge side to determine transient fault data. Based on the digital modeling of protection action events and the time-series causal reverse positioning of node protection actions, periodic fault data is determined. Fault Management Module: Based on the transient fault data and periodic fault data, the low-voltage distribution cabinet performs real-time fault management and test fault management.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, high-frequency pickup units are deployed in the circuit topology of the low-voltage distribution cabinet. The edge controller, triggered by the protection actions of the protection circuits, drives the high-frequency pickup units to detect waveforms across all channels. These pickup units are deployed at the input and output terminals of each protection circuit. Next, the time-varying impedance trajectory is determined based on the full-channel waveforms, and impedance deviation data based on quantified indicators is measured. Then, the impedance deviation data is input into a pre-set fault decision unit on the edge side to determine transient fault data. Periodic fault data is determined by digital modeling of protection action events and temporal causal reverse localization of node protection actions. Finally, based on the transient and periodic fault data, real-time fault management and test fault management are performed on the low-voltage distribution cabinet. This solves the technical problem of low fault diagnosis efficiency in existing low-voltage distribution cabinet technologies, achieving rapid fault identification and localization through edge computing and real-time analysis of full-channel waveforms, thereby improving fault diagnosis efficiency. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 A schematic diagram of the fault self-diagnosis method for low-voltage distribution cabinets based on edge computing provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a low-voltage distribution cabinet fault self-diagnosis system based on edge computing, provided in an embodiment of this application.

[0009] Explanation of reference numerals in the attached diagram: Waveform detection module 11, Impedance trajectory determination module 12, Fault data determination module 13, Fault management module 14. Detailed Implementation

[0010] To further illustrate the technical means and effects of the present invention in achieving its 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 below.

[0011] Example 1, as Figure 1 As shown, this application provides a fault self-diagnosis method for low-voltage distribution cabinets based on edge computing, wherein the method includes: In the circuit topology of the low-voltage distribution cabinet, a high-frequency pickup unit is deployed. The edge controller is triggered by the protection action of the protection circuit and drives the high-frequency pickup unit to detect the waveform of the entire channel. The deployment position is at the input and output terminals of each protection circuit.

[0012] First, based on the primary system wiring diagram and actual wiring structure of the low-voltage distribution cabinet, an electrical topology mapping table is constructed, including incoming circuits, busbar sections, branch circuits, and nodes of each protection circuit. The incoming and outgoing terminals of each protection circuit are then numbered and identified. Subsequently, high-frequency voltage pickup probes and high-frequency current pickup sensors are installed at the incoming and outgoing terminals of each protection circuit. The voltage pickup probes utilize high-bandwidth isolated sampling modules, while the current pickup sensors employ high-frequency current transformers or Hall effect current sensors. The sampling bandwidth is no less than 100kHz, and the sampling accuracy is no less than 12 bits to ensure the ability to capture transient arcs, current surges, and high-frequency oscillation components. Each high-frequency pickup unit is connected to the edge controller via shielded twisted-pair cables or industrial Ethernet, and a channel mapping table corresponding one-to-one with the node numbers is established within the edge controller.

[0013] The edge controller has a pre-set protection action trigger monitoring module. This module is electrically connected to the circuit breaker auxiliary contacts, contactor status feedback terminals, thermal relay alarm output terminals, and surge protector status indicator terminals, and collects the action status signals of each protection circuit in real time. When any protection action signal experiences a level transition, the edge controller records the current timestamp and sends an activation command to the high-frequency pickup unit of the corresponding node within a trigger delay of no more than 1ms, causing it to enter a high-speed continuous sampling mode. Within a preset time window (including the pre-stored time period before triggering and the extended time period after triggering), it synchronously collects voltage and current waveform data from all channels. The pre-stored time period is implemented through a cyclic buffer, with a length of 10ms to 50ms, and the extended time period after triggering is 20ms to 200ms, to fully cover the transient process before and after the protection action. Finally, all channels are aligned under the same time reference to form a full-channel waveform data set, providing a unified data foundation for subsequent time-varying impedance trajectory calculation and fault determination.

[0014] Furthermore, protection actions based on the protection circuit are set, wherein the protection actions include circuit breaker tripping, contactor closing, thermal relay operation, and surge protector conduction; according to the edge controller, protection action trigger detection is performed, and when the protection action is triggered, an activation command is sent to the high-frequency pickup unit to perform transient electrical waveform pickup of the entire channel at the moment of the protection action, and the waveform of the entire channel is determined.

[0015] In this embodiment, the protection actions based on the protection circuit are uniformly modeled and parameterized. Specifically, circuit breaker tripping, contactor closing, thermal relay operation, and surge protector activation are defined as four types of standard protection events, and a unique event code and action judgment logic are assigned to each type of protection event. Specifically, circuit breaker tripping is triggered by the level change of the auxiliary contact from closed to open; contactor closing is triggered by the main contact feedback signal changing from open to closed after the coil is energized; thermal relay operation is triggered by the normally closed contact opening signal; and surge protector activation is triggered by the alarm level or transient overvoltage response signal output from its status monitoring terminal. The edge controller polls the status signals of the above protection actions in real time through a digital input module or high-speed acquisition interface, and uses de-jitter filtering and a minimum holding time determination mechanism (e.g., holding time ≥ 5ms) to avoid false triggering.

[0016] When the edge controller detects that any protection action meets the preset trigger criteria, it immediately generates a trigger instruction data packet containing an event code, node number, and timestamp. Within a preset maximum response delay (e.g., within 1ms), it synchronously sends activation instructions to all deployed high-frequency pickup units. After receiving the activation instructions, each high-frequency pickup unit switches to high-speed transient sampling mode, synchronously acquires the voltage and current signals of each channel according to a unified clock reference, and extracts the pre-stored time window data before triggering from its local cyclic buffer. Simultaneously, it continuously acquires the extended time window data after triggering, thereby forming a full-channel electrical waveform data set covering the complete transient process before and after the moment of protection action. Finally, the edge controller performs time alignment and data encapsulation on the waveforms of each channel to determine the full-channel waveform, providing a unified and complete transient electrical data foundation for subsequent time-varying impedance trajectory calculation and fault feature extraction.

[0017] The time-varying impedance trajectory is determined based on the full-channel waveform, and the impedance deviation data is measured based on quantization indicators.

[0018] Furthermore, the impedance deviation data is measured based on quantitative indicators, including: Based on the full-channel waveform, the time-varying impedance trajectory is determined by the voltage and current waveforms. Based on the time-varying impedance trajectory, quantitative indicators are extracted, and the impedance deviation data is determined by calculating the deviation from the baseline trajectory at the initial stage of commissioning. The quantitative indicators include the abrupt change slope, overshoot amplitude, settling time, oscillation frequency, and zero-rest characteristics.

[0019] First, synchronize the sampling points of the time-aligned full-channel voltage and current waveforms. Let the voltage value at the i-th sampling moment be... The current value is Under the condition that the current amplitude is greater than the preset minimum effective threshold (to avoid amplification error in the zero current range), according to Calculate the instantaneous impedance value to obtain the discrete impedance sequence that varies with time. The time-varying impedance trajectory is formed by combining the time axis with the time curve. For a three-phase circuit, the impedance trajectory of each phase is calculated separately, and the phase imbalance is further calculated as an auxiliary characteristic. Subsequently, the time-varying impedance trajectory is preprocessed, including sliding window smoothing filtering, abnormal peak removal, and separation of fundamental and high-frequency components, to obtain a standardized impedance trajectory for fault characteristic analysis. Next, multi-dimensional quantitative indicators are extracted, including but not limited to: impedance jump slope (the maximum value of the impedance change rate per unit time), overshoot amplitude (the difference between the impedance peak value and the steady-state average value), settling time (the time required for the impedance to recover to the steady-state allowable error band), oscillation frequency (obtained by performing a fast Fourier transform on the impedance sequence), and zero-rest feature (the duration for which the impedance approaches zero or a maximum value in a short period of time). The reference impedance trajectory and its reference indicator values ​​for the corresponding circuit are pre-stored during the initial commissioning phase or under healthy conditions. Finally, the current quantitative indicators are compared with the reference indicators item by item using difference or relative deviation calculation to construct an impedance deviation vector, and a comprehensive impedance deviation value is generated according to weighting rules to form impedance deviation data. The impedance deviation data is used to characterize the degree of deviation of the electrical characteristics of the circuit corresponding to the current protection action from the normal state, providing a quantitative input basis for subsequent fault decision-making and type determination.

[0020] The impedance deviation data is input into a fault decision-maker preset on the edge side to determine transient fault data. Based on digital modeling of protection action events and temporal causal reverse localization of node protection actions, periodic fault data is determined.

[0021] Furthermore, the impedance deviation data is input into a fault decision unit preset on the edge side to determine transient fault data, including: A pre-set fault physical model library is provided, wherein the fault physical model library consists of the physical mapping relationship between fault types, transient impedance response waveforms and loop distribution parameters; a fault decision-maker is constructed based on the fault physical model library, wherein the fault decision-maker takes impedance trajectory data class as input, trajectory feature-degradation parameter response surface as likelihood function, and degradation parameter probability distribution as input; the impedance deviation data is input into the fault decision-maker to determine transient fault data.

[0022] First, a fault physical model library is pre-configured in the edge controller. This library is based on typical fault mechanisms of low-voltage distribution cabinets and includes a set of fault types, corresponding transient impedance response waveform templates, and physical mapping relationships between circuit distribution parameters. The fault types include at least increased contact resistance, partial short circuit, arc discharge, overload thermal degradation, and loose busbar connections. The transient impedance response waveforms are obtained by fusing simulation modeling with measured samples, describing the impedance mutation characteristics, oscillation frequency distribution, and attenuation curve morphology under different circuit parameters (such as line resistance, inductance, and capacitance distribution parameters). The physical mapping relationships are used to characterize the corresponding functional relationship between fault types and changes in circuit distribution parameters, thus forming a ternary mapping structure of "fault type - degradation parameter - impedance response characteristics". Secondly, a fault decision-maker is constructed based on a fault physics model library. This decision-maker takes impedance trajectory data as input, including impedance jump slope, overshoot amplitude, settling time, oscillation frequency, zero-rest features, and an overall trajectory deviation vector. In terms of model structure, the response surface constructed between trajectory features and degradation parameters is used as a likelihood function to calculate the probability of the current trajectory feature occurring under given degradation parameters. Simultaneously, a prior probability distribution for each degradation parameter is preset, derived from historical operational data statistics or expert experience parameter settings. Using Bayesian inference or maximum likelihood estimation methods, the posterior probability of degradation parameters corresponding to different fault types is calculated, thereby obtaining confidence scores for various fault types. Finally, impedance deviation data is input into the fault decision-maker. Based on the likelihood function and degradation parameter probability distribution, a joint probability calculation is performed. The fault type with the highest posterior probability exceeding a preset threshold is selected as the current judgment result, and transient fault data including fault type identifier, corresponding degradation parameter estimate, occurrence node number, and confidence level is output.

[0023] Furthermore, by using digital modeling based on protection action events and temporal causal reverse localization of node protection actions, periodic fault data is determined, including: Each protection circuit is mapped to a unique node in the circuit topology, and protection action events are digitally modeled to serve as the state database of the distribution cabinet. The modeling elements are the timestamp, node location, action type, and impedance trajectory fingerprint hash value of each protection action event. Based on the state database, the fault source is reversed by using the dynamic temporal causality of protection action events between different nodes to determine periodic fault data.

[0024] First, each protection circuit is mapped one-to-one with its physical connection to the circuit topology, assigning a unique node number to each incoming circuit, branch circuit, and busbar section, thus constructing a node mapping table. When any protection action occurs, the edge controller generates a structured event data record and writes it into the state database. The digital modeling elements include: a high-precision timestamp (accuracy not less than 1ms) of the protection action, the corresponding node number and physical location identifier, the action type code (circuit breaker opening, contactor closing, etc.), and a fingerprint value generated by a hash algorithm from the impedance trajectory feature vector associated with the event, used to identify the unique characteristics of the transient response. By continuously accumulating the above structured event data, a state database reflecting the operating history of the distribution cabinet is formed. Second, based on the state database, dynamic temporal causal analysis is performed on protection action events between different nodes. Specifically, the action time series of each node is extracted within a preset statistical period. The time interval distribution and condition trigger probability of actions between nodes are calculated, i.e., the probability value of the second node taking action within a preset time window after the first node takes a protective action is statistically analyzed. Combining time-series correlation coefficients, delay correlation analysis, or propagation entropy indicators, the temporal dependency strength between different nodes is quantified, and a directed temporal causal relationship graph is constructed accordingly. In this graph, the direction of the edges represents the potential fault propagation direction, and the edge weight represents the dependency strength. In the temporal causal graph, nodes that repeatedly act as upstream triggering sources and have high out-degrees within multiple statistical periods are identified as potential fault source nodes. Backtracking analysis is performed on the propagation path starting from this source node to extract its periodic triggering pattern and propagation path stability, generating periodic fault data including the fault starting node number, typical propagation path sequence, periodic occurrence frequency, and node confidence level. Through the above digital modeling and temporal causal reverse localization process, the identification and location of hidden and recurring fault sources are achieved, providing data support for predictive fault management.

[0025] Furthermore, fault source reverse localization is performed based on the dynamic temporal causality of protection action events between different nodes, including: According to a preset period, the temporal dependency strength between protection actions of different nodes in the state database is calculated using the temporal Pearson correlation coefficient and propagation entropy. Directed edges are established based on the temporal dependency strength to generate a temporal causal graph. A constraint condition is established that the probability of the second node performing a protection action within a first preset time period after the first node performs a protection action is higher than a preset threshold. By identifying the source nodes in the temporal causal graph, the circuit location corresponding to the source node is taken as the fault initiation point to determine periodic fault data. The fault data includes the fault initiation point coordinates, propagation path tree, and confidence scores of each node. The source node is determined by the absence of incoming edges or the strength of incoming edges being lower than that of outgoing edges.

[0026] First, within a preset statistical period (e.g., 7 days, 30 days, or an analysis period defined by a preset number of cumulative actions), the protection action time series of each node are extracted from the state database. The action events of each node are transformed into discrete or binary time series. The time axis is discretized according to a fixed sampling interval; a node performing a protection action within its corresponding time slice is recorded as 1, otherwise as 0. For any two nodes A and B, the temporal Pearson correlation coefficient between their action time series is calculated to measure the degree of linear correlation. Simultaneously, the transfer entropy value from node A to node B is calculated to measure the degree to which the historical action information of node A reduces the uncertainty of node B's future actions, thereby quantifying the information flow strength in the causal direction. The obtained Pearson correlation coefficient and transfer entropy are normalized and weighted and fused to form a temporal dependency strength index between node pairs. Second, directed edges are established based on the temporal dependency strength index to generate a temporal causal graph. Specifically, when node A performs a protection action, if the conditional probability of node B performing a protection action within a first preset time period (e.g., 100ms~5s, set according to loop characteristics) is higher than a preset threshold (e.g., 0.6 or an adaptive threshold determined by historical statistics), and the propagation entropy value from node A to node B exceeds the minimum causal strength threshold, a directed edge from A to B is established between node A and node B, and the temporal dependency strength is used as the edge weight. This judgment is performed on all node pairs to form a temporal causal graph containing a set of nodes and a weighted set of directed edges. Finally, the source node in the temporal causal graph is identified to achieve reverse fault location. The determination rule for the source node is: a node with no incoming edges in the graph structure, or whose sum of incoming edge weights is significantly lower than the sum of outgoing edge weights, and which is repeatedly identified as an upstream trigger node in multiple analysis cycles. The physical location of the circuit corresponding to the source node is used as the fault starting point, and a propagation path tree is generated along its outgoing edge direction. The frequency of occurrence and edge weight of each path are statistically analyzed, and the confidence score of each node is calculated. The final output of periodic fault data includes the circuit coordinates of the fault initiation point, the fault propagation path tree structure, and the confidence level of each node, which are used to characterize the periodic evolution and propagation characteristics of the fault in the distribution cabinet topology.

[0027] Based on the transient fault data and periodic fault data, real-time fault management and test fault management are performed on the low-voltage distribution cabinet.

[0028] Real-time fault management is implemented for transient fault data. After acquiring transient fault data, the edge controller parses the fault type, occurrence node number, severity level, and confidence parameters, and matches them with a pre-configured fault handling strategy library on the edge side. This strategy library establishes hierarchical response rules according to "fault type—node location—severity level," including control commands such as dynamic adjustment of protection settings, selective tripping control, rapid load switching, backup circuit activation, audible and visual alarms, and remote maintenance notifications. When the fault confidence level is higher than a first threshold, the corresponding control command is automatically executed and the execution result is recorded; when the confidence level is within the warning range, only an alarm is triggered and maintenance suggestions are generated. Simultaneously, after the handling is completed, subsequent impedance trajectory data continues to be collected, and the impedance recovery rate and degradation parameter changes before and after handling are compared to verify the effectiveness of real-time intervention.

[0029] Predictive fault management is implemented for periodic fault data. The edge controller performs a system-wide risk assessment based on the fault initiation node, propagation path tree, and node confidence level contained in the periodic fault data, combined with the circuit topology. This assessment constructs a fault evolution trend model to predict the probability of protection actions occurring at critical nodes within a preset timeframe and the possible propagation paths. For high-risk nodes with confidence levels exceeding a second threshold, maintenance work orders or inspection plans are generated in advance, and the sampling resource allocation strategy of the high-frequency pickup unit is dynamically optimized to increase the waveform sampling frequency and data storage priority of high-risk nodes. Simultaneously, load balancing or preventative replacement recommendations are implemented for loops exhibiting continuous degradation trends, forming a predictive management strategy that includes risk level, recommended handling time window, and priority ranking.

[0030] By coordinating the above-mentioned real-time fault management and predictive fault management, we can achieve rapid response to sudden faults in low-voltage distribution cabinets and early intervention for hidden periodic faults, thereby improving system operation safety and maintenance initiative.

[0031] Furthermore, the low-voltage distribution cabinet is subjected to real-time fault management based on the transient fault data and predictive fault management based on the periodic fault data; the secondary action waveform of the protection circuit during the execution of the fault contingency plan is collected and differentially compared with the transient fault data to calculate the recovery degree of the impedance trajectory and the amount of regression of the fault degradation parameters, and to quantitatively evaluate the intervention effect of the contingency plan.

[0032] First, after matching and executing the corresponding fault contingency plan based on transient fault data, the edge controller continuously monitors the status of relevant protection circuits within a preset evaluation time window (e.g., 5 minutes, 30 minutes, or one operating cycle after the contingency plan is executed). Upon the occurrence of a secondary protection action or active trigger detection, it again drives the high-frequency pickup unit to acquire transient electrical waveforms across all channels, forming secondary action waveform data. The secondary action waveform generates a new time-varying impedance trajectory and corresponding quantitative indicators according to the same calculation process as the first fault. Second, the impedance trajectory corresponding to the secondary action is compared with the impedance trajectory in the original transient fault data using differential analysis. Specifically, the difference sequence between the two impedance trajectories under the same time base is calculated, and the peak difference, mean square error, and main oscillation frequency offset are obtained. Simultaneously, the changes in quantitative indicators such as the abrupt change slope, overshoot amplitude, settling time, and oscillation frequency are compared. Based on the above comparison results, an impedance trajectory recovery index is defined, for example, using the decrease ratio of the current impedance deviation relative to the initial deviation as the recovery value. The regression amount of the fault degradation parameters is also calculated, i.e., based on the difference between the estimated degradation parameters output by the fault decision-maker and the values ​​before and after, the degree of physical degradation is assessed to determine whether it has been reduced. Finally, the impedance trajectory recovery degree and the fault degradation parameter regression amount are fused according to preset weights to generate a plan intervention effect score. When the score is higher than a preset effective threshold, the current fault plan intervention is deemed effective; when the score is lower than the threshold, the plan priority is automatically adjusted or a replacement plan is rematched, and the evaluation results are recorded and written into the status database.

[0033] Furthermore, real-time fault management based on the transient fault data includes: For the transient fault data, a fault plan is determined by interacting with a lightweight contingency plan library stored at the edge. The management mode includes dynamic adjustment of protection settings, load transfer, or operation and maintenance early warning. Historical fault data of low-voltage distribution cabinets of the same model are retrieved, and the confidence level of the fault plan and the suggested handling priority are generated by similarity weighted voting to perform real-time fault management of the low-voltage distribution cabinets.

[0034] After receiving transient fault data, the edge controller parses the fault type identifier, the occurrence node number, the impedance anomaly level, the estimated value of the degradation parameter, and the confidence index, and constructs a fault feature vector from these parameters. It then interacts with a lightweight contingency plan library stored at the edge. This lightweight contingency plan library establishes a multi-level mapping relationship according to "fault type-node location-severity level." Each contingency plan entry includes at least the trigger condition range, the corresponding management mode, and the set of execution instructions. The management modes include dynamic adjustment of protection settings (such as adjusting overcurrent settings or delay parameters), load transfer (such as switching the load of an abnormal branch to a backup circuit), and maintenance early warning (such as sending alarm information to the upper-level system or maintenance terminal). The edge controller calculates the matching degree between the current fault feature vector and the contingency plan trigger conditions, and filters candidate fault contingency plans that meet the threshold conditions.

[0035] Historical fault data of low-voltage distribution cabinets of the same model are retrieved, and candidate contingency plans are evaluated using a similarity-weighted approach. Specifically, fault samples with the same or similar fault types as the current fault are extracted from the historical database. Similarity indices (including Euclidean distance, cosine similarity, or normalized deviation coefficient) between the current fault feature vector and the historical sample feature vectors are calculated, and the historical samples are weighted according to the similarity. The actual handling plans and their effectiveness scores corresponding to each historical sample are statistically analyzed, and the overall confidence level of each candidate contingency plan is calculated through a weighted voting method. Simultaneously, a suggested handling priority ranking is generated based on factors such as historical handling success rate, fault severity, and execution cost. The edge controller selects the fault contingency plan with the highest confidence and highest priority for execution, and writes the execution results and subsequent monitoring data into the status database, achieving closed-loop control for real-time fault management.

[0036] Furthermore, predictive fault management based on the aforementioned periodic fault data includes: For periodic fault data, the overall system evolution is predicted by combining the circuit topology, and a fault evolution trend map is generated. The prediction direction includes the propagation path and the time limit of critical nodes. Based on the fault evolution trend map, high-risk branches and vulnerable nodes are marked, and the sampling resource allocation of the high-frequency pickup unit by the edge controller is dynamically optimized. The risk is positively correlated with the waveform pickup frequency.

[0037] First, a fault evolution prediction model is constructed based on the fault initiation node, propagation path tree, node confidence level, and periodic occurrence frequency parameters contained in the periodic fault data, combined with the circuit topology model of the low-voltage distribution cabinet. Specifically, the confidence value of each node is used as the initial risk weight, and the edge weights in the propagation path are used as risk transmission coefficients. Risk iterative propagation calculation is performed in the topology graph. Within a preset prediction time window, based on the historical periodic repetition frequency and the current degradation parameter growth trend, the risk growth curve of each node on the future time axis is estimated, and the time point when key nodes reach the protection action threshold or impedance over-limit threshold is predicted, thereby generating a fault evolution trend map containing the fault propagation path and the over-limit time of key nodes. Second, based on the fault evolution trend map, the risk levels of each branch and node are classified. Specifically, nodes with high risk growth rates, short expected over-limit times, or those in critical relay positions in the propagation path are marked as high-risk branches or vulnerable nodes, and assigned higher risk weight levels. Subsequently, the edge controller dynamically adjusts the sampling resource allocation strategy of the high-frequency pickup unit based on the risk weight of each node, increasing the sampling frequency, data buffer capacity, and analysis priority of the channels corresponding to high-risk nodes, while reducing the sampling frequency of low-risk nodes, thus achieving adaptive resource scheduling. Notably, node risk is positively correlated with waveform pickup frequency; that is, the higher the risk weight, the greater the allocated sampling frequency and processing resource ratio. Through this overall system evolution prediction and dynamic resource optimization process, early intervention for periodic potential hazards and enhanced monitoring of key nodes are achieved, improving the predictive maintenance capabilities and operational safety level of the low-voltage distribution cabinet.

[0038] In summary, the embodiments of this application have at least the following technical effects: First, high-frequency pickup units are deployed in the circuit topology of the low-voltage distribution cabinet. The edge controller, triggered by the protection actions of the protection circuits, drives the high-frequency pickup units to detect waveforms across all channels. These pickup units are deployed at the input and output terminals of each protection circuit. Next, the time-varying impedance trajectory is determined based on the full-channel waveforms, and impedance deviation data based on quantified indicators is measured. Then, the impedance deviation data is input into a pre-set fault decision unit on the edge side to determine transient fault data. Periodic fault data is determined by digital modeling of protection action events and temporal causal reverse localization of node protection actions. Finally, based on the transient and periodic fault data, real-time fault management and test fault management are performed on the low-voltage distribution cabinet. This solves the technical problem of low fault diagnosis efficiency in existing low-voltage distribution cabinet technologies, achieving rapid fault identification and localization through edge computing and real-time analysis of full-channel waveforms, thereby improving fault diagnosis efficiency.

[0039] Example 2 is based on the same inventive concept as the edge computing-based low-voltage distribution cabinet fault self-diagnosis method in the previous examples, such as... Figure 2As shown, this application provides a low-voltage switchgear fault self-diagnosis system based on edge computing, wherein the system includes: Waveform detection module 11: Deploys a high-frequency pickup unit in the circuit topology of the low-voltage distribution cabinet. The edge controller triggers the high-frequency pickup unit to detect the waveform of the entire channel according to the protection action of the protection circuit. The deployment position is the input and output terminals of each protection circuit. Impedance trajectory determination module 12: Determines the time-varying impedance trajectory based on the waveform of the entire channel and measures the impedance deviation data based on quantitative indicators. Fault data determination module 13: Inputs the impedance deviation data into the fault decision unit preset on the edge side to determine transient fault data. Based on the digital modeling of protection action events and the time-series causal reverse positioning of node protection actions, periodic fault data is determined. Fault management module 14: Performs real-time fault management and test fault management on the low-voltage distribution cabinet according to the transient fault data and periodic fault data.

[0040] Furthermore, the waveform detection module 11 is used to perform the following method: The protection actions based on the protection circuit are set, wherein the protection actions include circuit breaker tripping, contactor closing, thermal relay operation, and surge protector conduction; according to the edge controller, the protection action trigger detection is performed, and when the protection action is triggered, an activation command is sent to the high-frequency pickup unit to perform transient electrical waveform pickup of the entire channel at the moment of the protection action, and the waveform of the entire channel is determined.

[0041] Furthermore, the impedance trajectory determination module 12 is used to perform the following method: Based on the full-channel waveform, the time-varying impedance trajectory is determined by the voltage and current waveforms. Based on the time-varying impedance trajectory, quantitative indicators are extracted, and the impedance deviation data is determined by calculating the deviation from the baseline trajectory at the initial stage of commissioning. The quantitative indicators include the abrupt change slope, overshoot amplitude, settling time, oscillation frequency, and zero-rest characteristics.

[0042] Furthermore, the fault data determination module 13 is used to perform the following method: A pre-set fault physical model library is provided, wherein the fault physical model library consists of the physical mapping relationship between fault types, transient impedance response waveforms and loop distribution parameters; a fault decision-maker is constructed based on the fault physical model library, wherein the fault decision-maker takes impedance trajectory data class as input, trajectory feature-degradation parameter response surface as likelihood function, and degradation parameter probability distribution as input; the impedance deviation data is input into the fault decision-maker to determine transient fault data.

[0043] Furthermore, the fault data determination module 13 is used to perform the following method: Each protection circuit is mapped to a unique node in the circuit topology, and protection action events are digitally modeled to serve as the state database of the distribution cabinet. The modeling elements are the timestamp, node location, action type, and impedance trajectory fingerprint hash value of each protection action event. Based on the state database, the fault source is reversed by using the dynamic temporal causality of protection action events between different nodes to determine periodic fault data.

[0044] Furthermore, the fault data determination module 13 is used to perform the following method: According to a preset period, the temporal dependency strength between protection actions of different nodes in the state database is calculated using the temporal Pearson correlation coefficient and propagation entropy. Directed edges are established based on the temporal dependency strength to generate a temporal causal graph. A constraint condition is established that the probability of the second node performing a protection action within a first preset time period after the first node performs a protection action is higher than a preset threshold. By identifying the source nodes in the temporal causal graph, the circuit location corresponding to the source node is taken as the fault initiation point to determine periodic fault data. The fault data includes the fault initiation point coordinates, propagation path tree, and confidence scores of each node. The source node is determined by the absence of incoming edges or the strength of incoming edges being lower than that of outgoing edges.

[0045] Furthermore, the fault management module 14 is used to perform the following methods: The low-voltage distribution cabinet is subjected to real-time fault management based on transient fault data and predictive fault management based on periodic fault data; the secondary action waveform of the protection circuit during the execution of the fault contingency plan is collected and differentially compared with the transient fault data to calculate the recovery degree of impedance trajectory and the amount of fallback of fault degradation parameters, and to quantitatively evaluate the intervention effect of the contingency plan.

[0046] Furthermore, the fault management module 14 is used to perform the following methods: For the transient fault data, a fault plan is determined by interacting with a lightweight contingency plan library stored at the edge. The management mode includes dynamic adjustment of protection settings, load transfer, or operation and maintenance early warning. Historical fault data of low-voltage distribution cabinets of the same model are retrieved, and the confidence level of the fault plan and the suggested handling priority are generated by similarity weighted voting to perform real-time fault management of the low-voltage distribution cabinets.

[0047] Furthermore, the fault management module 14 is used to perform the following methods: For periodic fault data, the overall system evolution is predicted by combining the circuit topology, and a fault evolution trend map is generated. The prediction direction includes the propagation path and the time limit of critical nodes. Based on the fault evolution trend map, high-risk branches and vulnerable nodes are marked, and the sampling resource allocation of the high-frequency pickup unit by the edge controller is dynamically optimized. The risk is positively correlated with the waveform pickup frequency.

[0048] 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 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 self-diagnosis method for low-voltage distribution cabinet faults based on edge computing, characterized in that, The method includes: In the circuit topology of the low-voltage distribution cabinet, a high-frequency pickup unit is deployed. The edge controller is triggered by the protection action of the protection circuit and drives the high-frequency pickup unit to detect the waveform of the entire channel. The deployment position is the input and output terminals of each protection circuit. The time-varying impedance trajectory is determined based on the full-channel waveform, and the impedance deviation data is measured based on quantization indicators. The impedance deviation data is input into the fault decision-maker preset on the edge side to determine transient fault data. Based on the digital modeling of protection action events and the temporal causal reverse location of node protection actions, periodic fault data is determined. Based on the transient fault data and periodic fault data, real-time fault management and test fault management are performed on the low-voltage distribution cabinet.

2. The fault self-diagnosis method for low-voltage distribution cabinets based on edge computing as described in claim 1, characterized in that, The protection actions based on the protection circuit are set, wherein the protection actions include circuit breaker tripping, contactor closing, thermal relay operation, and surge protector activation; According to the edge controller, a protection action trigger detection is performed. When a protection action is triggered, an activation command is sent to the high-frequency pickup unit to perform transient electrical waveform pickup of the entire channel at the moment of the protection action and determine the waveform of the entire channel.

3. The low-voltage distribution cabinet fault self-diagnosis method based on edge computing as described in claim 1, characterized in that, The measurement is based on impedance deviation data using quantitative metrics, including: Based on the full-channel waveform, the time-varying impedance trajectory is determined by comparing the voltage and current waveforms. Based on the time-varying impedance trajectory, quantitative indicators are extracted. By calculating the deviation from the baseline trajectory at the initial stage of commissioning, the impedance deviation data is determined. The quantitative indicators include abrupt change slope, overshoot amplitude, settling time, oscillation frequency, and zero-rest characteristics.

4. The low-voltage distribution cabinet fault self-diagnosis method based on edge computing as described in claim 1, characterized in that, The impedance deviation data is input into a pre-set fault decision unit on the edge side to determine transient fault data, including: A pre-set fault physical model library, wherein the fault physical model library consists of the physical mapping relationship between fault type, transient impedance response waveform and loop distributed parameters; Based on the fault physics model library, a fault decision-maker is constructed, wherein the fault decision-maker takes impedance trajectory data class as input, trajectory feature-degradation parameter response surface as likelihood function, and degradation parameter probability distribution as input; The impedance deviation data is input into the fault decision unit to determine the transient fault data.

5. The low-voltage distribution cabinet fault self-diagnosis method based on edge computing as described in claim 1, characterized in that, By using digital modeling based on protection action events and temporal causal reverse localization of node protection actions, periodic fault data is determined, including: Each protection circuit is mapped to a unique node in the circuit topology, and protection action events are digitally modeled to serve as the state database of the distribution cabinet. The timestamp, node location, action type, and impedance trajectory fingerprint hash value of each protection action event are used as modeling elements. Based on the state database, the fault source is reversed and the periodic fault data is determined by the dynamic temporal causality of protection action events between different nodes.

6. The low-voltage distribution cabinet fault self-diagnosis method based on edge computing as described in claim 5, characterized in that, Fault source reverse localization is performed based on the dynamic temporal causality of protection action events between different nodes, including: Based on a preset period, the temporal dependency strength between protection actions of different nodes in the state database is calculated using the temporal Pearson correlation coefficient and the transfer entropy. Directed edges are established based on the temporal dependency strength to generate a temporal causal graph. The constraint condition is that the probability of the second node performing a protection action within a first preset time period after the first node performs a protection action is higher than a preset threshold. By identifying the source nodes in the time-series cause-effect graph, the circuit location corresponding to the source node is taken as the fault initiation point to determine the periodic fault data. The fault data includes the coordinates of the fault initiation point, the propagation path tree, and the confidence of each node. The source node is judged by the absence of incoming edges or the strength of the incoming edge being lower than that of the outgoing edge.

7. The low-voltage distribution cabinet fault self-diagnosis method based on edge computing as described in claim 1, characterized in that, The low-voltage distribution cabinet is subjected to real-time fault management based on the transient fault data and predictive fault management based on the periodic fault data. The secondary action waveforms of the protection circuit during the execution of the fault contingency plan are collected and compared differentially with transient fault data to calculate the recovery degree of the impedance trajectory and the amount of regression of the fault degradation parameters, thereby quantitatively evaluating the intervention effect of the contingency plan.

8. The low-voltage distribution cabinet fault self-diagnosis method based on edge computing as described in claim 7, characterized in that, Real-time fault management based on the transient fault data includes: For the transient fault data, a fault plan is determined by interacting with a lightweight contingency plan library stored at the edge. The management mode includes dynamic adjustment of protection settings, load transfer, or operation and maintenance early warning. Historical fault data of low-voltage distribution cabinets of the same model are retrieved, and the confidence level of the fault plan and the recommended handling priority are generated by weighted voting based on similarity, so as to carry out real-time fault management of low-voltage distribution cabinets.

9. The fault self-diagnosis method for low-voltage distribution cabinets based on edge computing as described in claim 7, characterized in that, Predictive fault management based on the aforementioned periodic fault data includes: For periodic fault data, the system evolution is predicted by combining the circuit topology, and a fault evolution trend map is generated. The prediction direction includes the propagation path and the time limit of critical nodes. Based on the fault evolution trend map, high-risk branches and vulnerable nodes are marked, and the sampling resource allocation of the high-frequency pickup unit by the edge controller is dynamically optimized. Among them, the risk is positively correlated with the waveform pickup frequency.

10. A low-voltage distribution cabinet fault self-diagnosis system based on edge computing, characterized in that, For implementing the edge computing-based low-voltage distribution cabinet fault self-diagnosis method according to any one of claims 1-9, the system comprises: Waveform detection module: A high-frequency pickup unit is deployed in the circuit topology of the low-voltage distribution cabinet. The edge controller is triggered by the protection action of the protection circuit to drive the high-frequency pickup unit to detect the waveform of the entire channel. The deployment position is at the input and output terminals of each protection circuit. Impedance trajectory determination module: Determines the time-varying impedance trajectory based on the full-channel waveform and measures the impedance deviation data based on quantization indicators; Fault data determination module: Input the impedance deviation data into the fault decision unit preset on the edge side to determine transient fault data, and determine periodic fault data based on digital modeling of protection action events and temporal causal reverse positioning of node protection actions; Fault Management Module: Based on the transient fault data and periodic fault data, performs real-time fault management and test fault management on the low-voltage distribution cabinet.

Images