An edge computing and working condition self-adaptive based low-voltage distribution network fault intelligent positioning and self-healing method and system and storage medium

Abstract

The application discloses a kind of based on edge computing and working condition self-adaptive low-voltage distribution network fault intelligent positioning and self-healing method, system and storage medium, belong to distribution network technical field.The method aims at solving the problems of high false alarm rate, rough positioning accuracy, long response delay and unnecessary power outage in the prior art.The method of the present application comprises: based on load fluctuation, photovoltaic output and other working condition factors, the fault determination threshold is calculated and updated by dynamic threshold model;Real-time fault characteristics are matched with local sample library to determine fault type.The present application can reduce the dependence of fault handling on real-time interaction of remote distribution master station by completing abnormal event detection, fault type identification, fault section positioning and differentiated control on the edge side, and help to narrow the isolation range of permanent fault and reduce unnecessary isolation actions for transient faults.

Description

Technical Field

[0001] This invention relates to the field of power distribution network technology, specifically to a method, system, and storage medium for intelligent fault location and self-healing in low-voltage power distribution networks based on edge computing and adaptive operating conditions. Background Technology

[0002] As the last level of power supply network facing end users, the low-voltage distribution network's operational stability and power supply reliability directly affect users' electricity experience. With the increasing proportion of distributed photovoltaic and other new energy sources being integrated, and with dynamic changes in electricity load, the operating conditions of the low-voltage distribution network are becoming increasingly complex, placing higher demands on the timeliness and accuracy of fault detection, location, and handling.

[0003] In current technologies, troubleshooting low-voltage distribution network faults still relies to some extent on manual inspections. When a fault occurs, maintenance personnel typically need to go to the site to check each section, which is not only inefficient but may also prolong the power outage time for affected users.

[0004] To improve the automation level of fault handling, existing technologies have proposed using edge computing devices for fault detection and location. For example, Chinese invention patent application CN110824300A, entitled "An Edge Computing System for Fault Detection and Location in Distribution Networks," discloses a technical solution that uses an edge computing system to collect and process distribution network fault data and send fault alarm information and location results to the distribution automation master station or distribution IoT cloud platform.

[0005] Existing technologies have also proposed self-healing control schemes for fault isolation and power restoration in distribution networks. For example, Chinese invention patent application CN114825627A, entitled "A Self-Healing Method for Distribution Network Line Faults," discloses a technical solution in which the distribution network master station determines the fault area based on data from relevant equipment of the distribution line and performs fault isolation and power restoration according to the corresponding self-healing strategy.

[0006] The above-mentioned technical solutions can improve the automation level of distribution network fault handling to a certain extent. However, from the perspective of existing technologies as a whole, there are still the following shortcomings: On the one hand, some fault detection methods use pre-set fixed thresholds, which are difficult to adapt to complex and ever-changing operating conditions such as load fluctuations, distributed photovoltaic power output changes, harmonic interference, and three-phase imbalance, and are prone to false alarms or missed alarms; on the other hand, some fault judgment and self-healing decisions still rely on remote distribution master stations, and the processing link formed by data uploading, master station analysis, and control command issuance is relatively long, which may affect the timeliness of fault handling.

[0007] Furthermore, the fault location range of some existing technologies is still relatively large, making it difficult to accurately determine the fault range to the line section where selective isolation can be implemented by combining the timing, amplitude, and topological connection relationships of electrical parameter changes from multiple monitoring nodes. At the same time, some existing technologies cannot effectively distinguish between transient faults that can recover on their own and permanent faults that require isolation, which may easily lead to unnecessary isolation actions for transient faults and cause unnecessary power outages.

[0008] Therefore, existing low-voltage distribution network fault handling technologies still have problems such as fault judgment conditions being difficult to adapt to real-time operating conditions, long fault analysis and handling links, large fault location range, and lack of differentiation in handling methods for different types of faults. Summary of the Invention

[0009] This invention aims to solve the technical problems in the existing low-voltage distribution network fault handling process, such as the difficulty of using fixed fault judgment thresholds to adapt to complex operating conditions such as load fluctuations, distributed photovoltaic power output changes, harmonic interference and three-phase imbalance, which easily leads to misjudgment or omission, as well as the reliance on remote master stations for fault analysis, the large fault location range, and the lack of differentiation in the handling methods for instantaneous and permanent faults, which in turn lead to delays in fault handling and unnecessary power outages.

[0010] To address the aforementioned technical problems, this invention provides a method for intelligent fault location and self-healing in low-voltage distribution networks based on edge computing and adaptive operating conditions. The method is executed by at least one edge computing device deployed on the low-voltage distribution network side, and includes: Acquire real-time data from multiple monitoring nodes and at least one operating condition factor characterizing the real-time operating status of the low-voltage distribution network, and update the dynamic threshold for abnormal event detection based on the operating condition factor. In response to detecting an abnormal event based on the real-time collected data and the dynamic threshold, fault features are extracted from the real-time collected data corresponding to the abnormal event, and the fault features are matched with a fault feature sample library pre-installed on the edge computing device to determine the fault type corresponding to the abnormal event. Based on a pre-stored distribution network topology model, the timing and magnitude of changes in the electrical parameters of the multiple monitoring nodes are analyzed to determine the fault section corresponding to the abnormal event. Based on the changes in electrical parameters during and after the occurrence of the abnormal event, a fault nature determination result is obtained, which is used to characterize whether the abnormal event is a transient fault or a permanent fault. A differentiated control strategy is generated and executed based on the faulty section and the fault nature determination result. Specifically, when the fault nature determination result indicates that the abnormal event is a permanent fault, a control strategy is executed to isolate the faulty section and maintain power supply to the non-faulty section. When the fault nature determination result indicates that the abnormal event is a transient fault, no isolation action is executed for the faulty section, and the corresponding event log is recorded.

[0011] The edge computing device can be a low-voltage intelligent converged terminal, edge gateway, intelligent power distribution terminal, or other edge-side device with data processing and control functions deployed on the low-voltage distribution area side. The at least one edge computing device is used to perform dynamic threshold updates, abnormal event detection, fault type identification, fault segment location, fault nature determination, and differentiated control strategy generation near the data source, thereby reducing the reliance of the core fault assessment process on real-time interaction with the remote power distribution master station.

[0012] Optionally, the operating condition factors include at least one of load fluctuation, distributed photovoltaic output, environmental factors, harmonic distortion, and three-phase imbalance. The environmental factors may include ambient temperature, ambient humidity, seasonal conditions, and other environmental parameters that can affect the operating characteristics of low-voltage distribution lines.

[0013] The step of updating the dynamic threshold for abnormal event detection based on the operating condition factors may include: performing dimensionless processing on each acquired operating condition factor; performing weighted correction on a preset initial threshold according to the weight coefficients corresponding to each operating condition factor to obtain a corrected threshold; and limiting the corrected threshold within a preset threshold range determined according to the initial threshold to obtain the dynamic threshold.

[0014] The dynamic threshold can be a single threshold used to determine whether an electrical parameter is abnormal, or it can be a set of thresholds consisting of multiple thresholds corresponding to different electrical parameters, different fault characteristics, or different fault types. By updating the dynamic threshold according to the real-time operating conditions of the low-voltage distribution network, the triggering conditions for abnormal events can be adjusted with changes in the normal operating baseline, reducing the possibility of misjudging fluctuations in normal operating conditions as faults or missing real faults.

[0015] Furthermore, when at least one electrical parameter in the real-time acquired data or at least one feature parameter generated from the real-time acquired data reaches the corresponding dynamic threshold, an abnormal event is determined to have been detected, and corresponding fault features are extracted from the real-time acquired data before, during, and after the occurrence of the abnormal event.

[0016] The fault characteristics may include at least one of the following: electrical parameter amplitude variation characteristics, harmonic distortion characteristics, zero-sequence electrical quantity characteristics, phase sequence characteristics, duration characteristics, rate of change characteristics, and time-sequence variation characteristics. The electrical parameters may include at least one of the following: three-phase voltage, three-phase current, zero-sequence current, zero-sequence voltage, power factor, active power, reactive power, harmonic distortion rate, and three-phase unbalance.

[0017] Optionally, the fault feature sample library contains multiple standard feature vectors and corresponding fault types for each standard feature vector. The fault types may include at least one of short-circuit faults, grounding faults, open-circuit faults, leakage faults, overvoltage faults, undervoltage faults, three-phase imbalance faults, and transient voltage disturbances.

[0018] The step of matching the fault features with the fault feature sample library may include: normalizing the fault features and multiple standard feature vectors in the fault feature sample library according to the same normalization rule; calculating the similarity between the normalized fault features and each standard feature vector; and determining the fault type represented by the standard feature vector corresponding to the highest similarity as the fault type corresponding to the abnormal event when the highest similarity reaches a preset judgment threshold.

[0019] The similarity can be calculated directly using a similarity measurement method, or the distance between the fault feature and the standard feature vector can be calculated first, and then the distance can be converted into a similarity according to a preset conversion relationship. This way, the larger the similarity value, the closer the two feature vectors are, avoiding confusion between the judgment direction of the distance value and the similarity value.

[0020] Optionally, the distribution network topology model is used to represent the connection relationships between various monitoring nodes, line sections, electrical loads, distributed power sources, and controllable switches in the low-voltage distribution network. The multiple monitoring nodes can be distributed at at least two levels among the low-voltage distribution area main node, trunk line node, branch line node, and user terminal node.

[0021] The step of analyzing the timing and magnitude of changes in the electrical parameters of the multiple monitoring nodes to determine the fault section corresponding to the fault may include: comparing the multiple monitoring nodes step by step according to the upstream and downstream connection relationships between the monitoring nodes in the distribution network topology model; selecting candidate monitoring nodes from the multiple monitoring nodes based on the time when the electrical parameters of each monitoring node change; and when multiple candidate monitoring nodes exist, determining the fault section by combining the magnitude of the changes in the electrical parameters of each candidate monitoring node and the topological connection relationships between the candidate monitoring nodes.

[0022] When the time difference of electrical parameter changes among multiple monitoring nodes is less than a preset synchronization error limit, all of the multiple monitoring nodes can be retained as candidate monitoring nodes. Furthermore, the fault section can be determined based on at least one of the following: change amplitude, upstream and downstream relationship, fault type, and parameter differences between adjacent nodes, so as to avoid locating the fault based solely on the earliest changing node or the node with the largest change amplitude.

[0023] The faulty section can be a line section between two adjacent monitoring nodes, a branch line connected to a candidate monitoring node, a user terminal section downstream of a candidate monitoring node, or other isolable sections divided by the distribution network topology model. Therefore, this invention does not need to limit the fault location result to a specific physical point that is difficult to measure directly, but rather determines the fault range to a faulty section that can be selectively isolated by a controllable switch.

[0024] Optionally, acquiring real-time data from multiple monitoring nodes includes: performing high-frequency synchronous acquisition of at least one of voltage data, current data, and zero-sequence electrical quantity data from the multiple monitoring nodes according to a unified time reference; and performing at least one of time-series alignment, abnormal data removal, waveform noise reduction, and data standardization on the acquired data to obtain the real-time data.

[0025] The unified time reference can be provided by satellite time synchronization, network clock synchronization, power grid reference phase synchronization, or a unified time stamp issued by edge computing devices. For data whose acquisition time difference exceeds the preset synchronization error limit, time compensation, realignment, or removal from the current positioning calculation can be performed.

[0026] Furthermore, obtaining the fault nature determination result based on the changes in electrical parameters during and after the occurrence of the abnormal event may include: determining the duration of the abnormal electrical parameter corresponding to the abnormal event; determining the recovery deviation of the electrical parameters after the occurrence of the abnormal event relative to a preset normal operating range; when the duration of the abnormality is less than a preset time threshold and the recovery deviation is not greater than a preset recovery threshold, determining that the fault nature determination result indicates that the abnormal event is an instantaneous fault; and when the duration of the abnormality is not less than the preset time threshold, or the recovery deviation is greater than the preset recovery threshold, determining that the fault nature determination result indicates that the abnormal event is a permanent fault.

[0027] The recovery deviation can be determined based on the difference between the electrical parameters after the fault and the baseline electrical parameters before the fault, or it can be determined based on the degree to which the electrical parameters after the fault deviate from the preset normal operating range. By combining the duration of the abnormality and the recovery status after the fault, misjudgments caused by distinguishing between transient and permanent faults based solely on a single duration threshold can be reduced.

[0028] After determining that the fault is permanent, the edge computing device determines the controllable switch corresponding to the fault segment and generates a control command for isolating the fault segment. Optionally, before sending the control command, at least one of the following can be verified: the distribution network topology model, the current state of the controllable switch, the coordination relationship between upper and lower level protection systems, and the operating status of distributed power sources.

[0029] When the verification result meets the preset control conditions, the control command is sent to the controllable switch to isolate the faulty section and ensure that the non-faulty sections located outside the faulty section continue to be powered. For low-voltage distribution networks equipped with tie switches, backup power supplies, or reconfigurable power supply paths, after isolating the faulty section, the power restoration path for the non-faulty sections can be determined based on the updated distribution network topology model, and the corresponding switches can be controlled to perform power restoration operations.

[0030] After determining that the fault is a transient fault, no isolation action is performed on the faulty segment, and the event log corresponding to the abnormal event is recorded locally on the edge computing device to reduce unnecessary power outages caused by transient faults that can recover on their own.

[0031] Furthermore, based on the event log, the number of transient faults occurring in the same fault segment within a preset time window can be counted. When the number of transient faults reaches a preset threshold, an alarm message is generated for the fault segment; when the number of transient faults does not reach the preset threshold, the power supply to the fault segment is maintained. Thus, while ensuring power supply continuity, it provides alerts for recurring transient faults that may be caused by intermittent poor contact, deteriorated insulation performance, or other potential defects.

[0032] Optionally, the method further includes locally storing full-cycle data on the edge computing device before, during, and after each abnormal event. The full-cycle data may include at least one of real-time acquired data, fault characteristics, fault type, fault segment, fault nature determination results, and execution results of differentiated control strategies.

[0033] Based on preset abnormal waveform identification rules and the correspondence between the full-cycle data and external operation and maintenance records, valid fault samples are screened from the retained full-cycle data. Specifically, data generated by high-frequency electromagnetic interference, abnormal data acquisition, manual maintenance tests, or equipment debugging operations can be identified as invalid samples and removed before being added to the fault feature sample library.

[0034] Furthermore, the selected valid fault samples are added to the fault feature sample library, and based on the valid fault samples, at least one of the parameters used to update the dynamic threshold and the parameters used to match the fault features is adjusted.

[0035] Before applying the adjusted parameters to actual fault handling, the adjusted parameters can be verified using historical retained samples, independent verification samples, or fault samples confirmed by maintenance personnel. The preset verification conditions may include at least one of the following: parameter values ​​meet preset constraints, the false positive rate after adjustment is not higher than the false positive rate before adjustment, the false negative rate after adjustment is not higher than the false negative rate before adjustment, and the fault location results meet preset consistency requirements.

[0036] When the adjusted parameters meet the preset verification conditions, the corresponding parameters on the edge computing device are updated using the adjusted parameters; when the adjusted parameters do not meet the preset verification conditions, the parameters before adjustment are continued to be used. This improves the model's adaptability to different operating conditions while reducing the possibility that unverified parameter updates will affect fault diagnosis and control safety.

[0037] The present invention also provides a low-voltage distribution network fault intelligent location and self-healing system based on edge computing and operating condition adaptation, including multiple monitoring terminals, at least one edge computing device and at least one controllable switch.

[0038] The multiple monitoring terminals are respectively installed at multiple monitoring nodes of the low-voltage distribution network to collect real-time data from the corresponding monitoring nodes.

[0039] The edge computing device is communicatively connected to the plurality of monitoring terminals and the at least one controllable switch, and is used to acquire real-time data collected from the plurality of monitoring nodes and at least one operating condition factor, update the dynamic threshold for abnormal event detection, identify the fault type corresponding to the abnormal event, determine the fault section, determine whether the fault is a transient fault or a permanent fault, and generate a differentiated control strategy based on the fault section and the fault determination result.

[0040] The at least one controllable switch is used to respond to control commands sent by the edge computing device to perform isolation actions on the fault section corresponding to a permanent fault, or to maintain the power supply state of the corresponding fault section when the fault is determined to be a transient fault.

[0041] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the above-mentioned method for intelligent fault location and self-healing of low-voltage distribution networks based on edge computing and adaptive operating conditions.

[0042] Compared with the prior art, the present invention has at least the following beneficial effects: First, the present invention updates the dynamic threshold used for abnormal event detection based on the real-time operating conditions of the low-voltage distribution network, so that the fault judgment conditions can be adjusted with load fluctuations, distributed photovoltaic power output and power quality status changes, which helps to reduce misjudgments and missed judgments caused by fixed thresholds under complex operating conditions.

[0043] Secondly, this invention completes abnormal event detection, fault type identification, fault section location and control strategy generation in the edge computing equipment on the low-voltage distribution network side, which can shorten the processing link formed by uploading monitoring data to the remote master station and returning control commands, and improve the timeliness of fault handling.

[0044] Third, this invention determines the fault section by integrating the timing, magnitude, and topological connection relationships of electrical parameter changes from multiple monitoring nodes. This avoids relying solely on a single monitoring node or a single electrical parameter for location, allowing the location results to directly correspond to isolable line sections.

[0045] Fourth, this invention combines the duration of the abnormality during the fault occurrence with the recovery status of electrical parameters after the fault occurs to determine whether the fault is a transient or permanent fault, and performs differentiated control based on the fault section and the fault determination result. This is beneficial to isolate permanent faults in a timely manner while reducing unnecessary isolation actions for transient faults that can recover on their own.

[0046] Fifth, the present invention selectively isolates the faulty section when a permanent fault occurs, while maintaining power supply to the non-faulty section, thereby reducing the scope of the power outage and improving the power supply reliability of the low-voltage distribution network.

[0047] Sixth, by retaining full-cycle data of abnormal events, screening valid fault samples, and updating model parameters when preset verification conditions are met, this invention enables the edge-side fault judgment model to be continuously optimized as the operating characteristics of the low-voltage distribution network change, while reducing the adverse effects of invalid samples or unverified parameter updates on the fault judgment results. Attached Figure Description

[0048] Figure 1 This is a flowchart illustrating the overall system architecture and edge computing execution process of the present invention.

[0049] Figure 2 This is a schematic diagram of the four-level topology nodes and fault layering location of the transformer area according to the present invention.

[0050] Figure 3 This is a flowchart of the dynamic threshold model operation of the present invention.

[0051] Figure 4 This is a flowchart illustrating the differentiated self-healing control process for transient and permanent faults in this invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0053] This invention aims to overcome the problems of high false positive rates, coarse fault location accuracy, slow response speed, and inability to effectively distinguish fault nature caused by the use of fixed thresholds in existing low-voltage distribution network fault detection methods. To address these issues, this invention proposes an intelligent method based on edge computing. This method is executed on edge devices (such as low-voltage intelligent fusion terminals) close to the data source, enabling rapid fault analysis and self-healing. Edge computing here refers to the process of data processing, analysis, and decision-making at the network edge close to the data source, without real-time interaction with a remote center (such as a distribution master station), thus effectively shortening response latency. This invention introduces adaptive dynamic thresholds to adapt to the complex and ever-changing operating states of the distribution network; utilizes a refined distribution network topology model to determine the fault section corresponding to the fault; and distinguishes between transient and permanent faults based on the transient characteristics of the fault, executing differentiated control strategies to achieve edge-side fault assessment and differentiated control.

[0054] In one aspect of the present invention, a method for intelligent fault location and self-healing in low-voltage distribution networks based on edge computing and adaptive operating conditions is provided. For example... Figure 1 As shown, in a complete processing cycle, this method first acquires real-time operating data of the power grid through a high-frequency data acquisition and preprocessing module, and then proceeds to a fault analysis and location module. In this module, the core steps of the method unfold sequentially: First, dynamic thresholds are calculated based on operating condition factors characterizing the real-time operating state of the low-voltage distribution network; next, real-time fault characteristics are matched with a local fault characteristic sample library to determine the fault type; then, source tracing and location are performed based on the distribution network topology model; finally, the method distinguishes between transient and permanent faults. Based on the distinction results, a differentiated self-healing control module executes corresponding control strategies. The entire process forms a closed loop, and the system performance is continuously improved through a sample self-iterative optimization module.

[0055] Specifically, this method first calculates and updates dynamic thresholds for fault determination based on at least one operating condition factor characterizing the real-time operating status of the low-voltage distribution network, using a pre-set dynamic threshold model. This approach abandons the rigid mode of traditional fixed thresholds, allowing fault determination criteria to be dynamically adjusted according to actual operating conditions such as grid load fluctuations and changes in renewable energy output, thereby effectively reducing false alarms and missed alarms under complex operating conditions. After obtaining the dynamic thresholds, the method matches fault features generated from the acquired real-time data with a pre-set fault feature sample library on the edge computing device to determine the fault type corresponding to the abnormal event. This sample library contains standard features of various typical faults. By comparing the similarity between real-time fault features and standard features, it can accurately identify whether the fault is a short circuit, grounding, or other type, providing a basis for subsequent accurate processing.

[0056] After determining the fault type, the method, based on a pre-stored distribution network topology model, analyzes the changes in electrical parameters of multiple nodes in the topology model and determines the location of the fault within the topology model according to preset source tracing and location rules. This distribution network topology model accurately describes the electrical connection relationships of each node in the network, making fault location no longer a vague regional judgment, but rather able to gradually narrow the fault range from the transformer substation level to the corresponding fault section, providing a basis for subsequent selection of the corresponding controllable switch.

[0057] Next, based on the transient characteristics of electrical parameters during the fault, the method classifies the fault into transient or permanent faults. Transient faults are usually caused by lightning flashes, instantaneous load surges, etc., are short-lived, and can recover on their own; while permanent faults are caused by equipment damage, line breaks, etc., are long-lasting, and cannot recover on their own. By effectively distinguishing between these two types of faults, unnecessary power outages can be avoided.

[0058] Finally, the method generates and executes differentiated control strategies based on the determined faulty sections and fault determination results. When the fault is determined to be a permanent fault, the edge computing device determines the corresponding controllable switch based on the faulty section and controls the controllable switch to isolate the faulty section, allowing the non-faulty sections to maintain power supply; when the fault is determined to be a transient fault, no isolation action is performed for the faulty section, and the corresponding event log is recorded locally on the edge computing device.

[0059] Furthermore, to achieve adaptive operating conditions, the dynamic threshold model can be an algorithm that weights and calculates based on multiple operating condition factors. In this implementation, the operating condition factors may include at least one of load fluctuations, photovoltaic output, environmental factors, harmonic interference, and three-phase imbalance. By comprehensively considering these key factors affecting the grid's operating state, the current operating conditions can be more accurately characterized, and a more adaptive dynamic threshold can be calculated. In one implementation, dynamic thresholds can be set separately for electrical parameters or fault characteristics that require anomaly detection. For the first... The dynamic threshold of an electrical parameter or fault characteristic to be judged. Determined according to the following formula: ; in, For the first Initial threshold corresponding to each electrical parameter or fault characteristic; For the first Dimensionless correction coefficients corresponding to each working condition factor; For the first The first working condition factor affects the first Weighting coefficients for each dynamic threshold; The number of operating condition factors participating in dynamic threshold correction, and: Correction coefficient Using 1 as the baseline value. When the first... The first operating condition factor makes the first When the normal operating fluctuation range of an electrical parameter or fault characteristic increases, the correction coefficient is used to increase the corresponding dynamic threshold; when the operating condition factor decreases the normal operating fluctuation range, the correction coefficient is used to decrease the corresponding dynamic threshold. The value and adjustment direction of the correction coefficient can be determined based on the line parameters, equipment operating characteristics, historical operating data, or on-site setting results of the low-voltage distribution area.

[0060] To avoid excessive shifts in the dynamic threshold due to abnormal operating conditions or abnormal data acquisition, the calculated dynamic threshold will be... Limited to the corresponding preset threshold lower limit and preset threshold upper limit The calculation result is between the preset lower threshold and the preset upper threshold. When the calculation result is lower than the preset lower threshold, the preset lower threshold is used; when the calculation result is higher than the preset upper threshold, the preset upper threshold is used.

[0061] In one specific implementation, the operating condition factors include at least one of load fluctuation, distributed photovoltaic output, environmental factors, harmonic distortion, and three-phase imbalance. The correction coefficient, weighting coefficient, threshold update cycle, and upper and lower limits of each operating condition factor can be preset based on the historical normal operation data, equipment parameters, and protection setting requirements of the corresponding low-voltage distribution area, and stored in the edge computing device.

[0062] In terms of fault type matching, as a preferred implementation, the step of matching fault features with a sample library specifically includes: calculating the similarity between the fault features and each standard feature vector contained in the sample library; when the calculated highest similarity reaches a preset judgment threshold, the fault type corresponding to the current abnormal event is determined as the fault type corresponding to the highest similarity.

[0063] The fault feature matching can employ a similarity calculation method based on feature vector distance. Following the same normalization rule, the real-time fault feature vector and each standard feature vector in the fault feature sample library are normalized, and the feature distance between the normalized real-time fault feature vector and each standard feature vector is calculated.

[0064] For the Each standard feature vector has a corresponding similarity. Determined according to the following formula: ; in, The normalized real-time fault feature vector and the first The feature distance between standard feature vectors. The smaller the feature distance, the greater the similarity.

[0065] When the highest similarity reaches the preset judgment threshold, the fault type corresponding to the highest similarity is determined as the fault type of the current abnormal event; when the highest similarity does not reach the preset judgment threshold, the current abnormal event is marked as an abnormal event of undetermined type, and an alarm is issued or further analysis is performed.

[0066] In terms of fault location, such as Figure 2 As shown, the distribution network topology model may include low-voltage distribution area master nodes, main line nodes, branch line nodes, and user terminal nodes. Figure 2In the topology shown, the low-voltage distribution area's main node is Node 1, the trunk line node is Node 2, and the branch line nodes 1#, 2#, 3#, 4#, 5#, 6#, 7#, and 8# are respectively Node 3, 4#, 5#, 6#, 7#, and 8#. Branch line user 1 and 2# are Node 9 and 10#, respectively. Node 1 is connected to Node 2, Node 2 is connected to Nodes 3 through 8, and Node 5 is connected to Nodes 9 and 10. Each branch line node connects downwards to its corresponding user terminal node, thus forming a multi-level topology that reflects the upstream and downstream connections between the monitoring nodes.

[0067] After a fault is triggered, the edge computing device compares the times when electrical parameters of each monitoring node change according to the upstream and downstream connection relationships between the monitoring nodes. Using the earliest change time among all monitoring nodes as a benchmark, the monitoring nodes whose time difference with the earliest change time is not greater than a preset synchronization error limit are determined as candidate monitoring nodes.

[0068] When there is only one candidate monitoring node, the corresponding candidate fault section is determined based on the topological connection relationship between the candidate monitoring node and its upstream and downstream adjacent monitoring nodes.

[0069] When multiple candidate monitoring nodes exist, the magnitude of electrical parameter changes of each candidate monitoring node is further compared. Combined with the upstream and downstream connection relationship between candidate monitoring nodes, fault type, and parameter differences between adjacent nodes, the fault segment consistent with the fault propagation characteristics is determined from multiple candidate fault segments.

[0070] When multiple candidate fault segments cannot be ruled out based on the change timing, change amplitude, and topological connectivity, the least common isolable segment corresponding to the multiple candidate fault segments is determined as the fault segment to be confirmed, and the reliability level of the location result is reduced. When the reliability level of the location result does not meet the preset control conditions, a fault alarm and the fault segment to be confirmed are output, and automatic isolation action is not directly executed.

[0071] To distinguish between transient and permanent faults, as a specific implementation method, the edge computing device determines the duration of the abnormal electrical parameters corresponding to the fault, and determines the recovery deviation of the electrical parameters after the fault occurs relative to a preset normal operating range. The nature of the fault is determined based on the duration of the abnormality and the recovery deviation.

[0072] The preset time threshold and preset recovery threshold can be determined based on the line parameters of the low-voltage distribution area, the characteristics of the protection device, historical operating data, or on-site setting results. When the duration of the abnormal electrical parameter is less than the preset time threshold, and the recovery deviation of the electrical parameter after the fault occurs relative to the preset normal operating range is not greater than the preset recovery threshold, the fault is determined to be an instantaneous fault.

[0073] When the duration of abnormal electrical parameters is not less than the preset time threshold, or the recovery deviation is greater than the preset recovery threshold, the fault is determined to be a permanent fault.

[0074] To obtain multi-node time-series data for fault diagnosis and fault location, monitoring terminals can be set up at at least two levels among the low-voltage distribution area main node, trunk line node, branch line node, and user terminal node. Each monitoring terminal includes a data acquisition unit and a time synchronization unit, used to acquire electrical waveform data of the corresponding monitoring node according to a unified time reference, and add sampling time stamps to the acquired data.

[0075] The unified time reference can be provided by satellite time synchronization, network clock synchronization, power grid reference phase synchronization, or a unified time stamp issued by the edge computing device. After detecting an abnormal event, the edge computing device uses the trigger time of the abnormal event as a reference to extract data within the same analysis time window from the data collected by each monitoring terminal, and performs time sequence alignment according to the sampling time stamp.

[0076] For data whose sampling time deviation between different monitoring nodes is no greater than the preset synchronization error limit, it is included in the fault analysis of the same abnormal event; for data whose sampling time deviation is greater than the preset synchronization error limit, time correction is performed first, and if correction cannot be completed, it is not used for this fault segment location based on changing time sequence.

[0077] When data collected by some monitoring nodes is missing, the edge computing device can continue fault analysis based on the data collected by the remaining valid monitoring nodes and the distribution network topology model. When there are not enough valid monitoring nodes to determine the fault section, or when data from key monitoring nodes at the boundary of the fault section is missing, the device outputs the scope of the fault to be confirmed and an abnormal alarm, without directly executing automatic isolation actions.

[0078] Preprocessing of the collected data may include at least one of the following: outlier removal, time alignment, waveform noise reduction, and data standardization.

[0079] In order to ensure both power supply continuity and safety, the differentiated control strategy for handling transient faults may further include: recording the corresponding event log locally on the edge computing device, and generating alarm information for the fault segment when the number of times the transient fault recurs within a preset time window reaches a preset threshold.

[0080] Specifically, when the duration of an abnormal event is less than a preset time threshold, and the recovery deviation of the electrical parameters after the abnormal event ends relative to a preset normal operating range is not greater than a preset recovery threshold, the abnormal event is determined to be a transient fault. For the transient fault, the edge computing device does not perform isolation actions for the faulty segment and records the corresponding event log locally; the event log can also be reported to the operation and maintenance platform.

[0081] When a transient fault in the same faulty section recurs within a preset time window, and the number of occurrences reaches a preset threshold, the edge computing device generates an alarm message for the faulty section and increases its monitoring level to prompt maintenance personnel to inspect it. Isolation of the faulty section is not performed until the conditions for determining a permanent fault are met. Therefore, while maintaining power continuity, it is possible to provide alerts for recurring transient faults.

[0082] To adapt the fault feature sample library and related model parameters to changes in the operating characteristics of low-voltage distribution networks, edge computing devices locally retain full-cycle data before, during, and after anomaly events. This full-cycle data includes at least one of the following: electrical waveform data, fault characteristics, fault type determination results, fault section location results, fault nature determination results, and control strategy execution results.

[0083] The edge computing device correlates the full-cycle data with protection device action records, controllable switch status records, maintenance work orders, on-site fault handling records, or maintenance personnel confirmation results in time. When the fault type, fault segment, and fault nature corresponding to the abnormal event can be determined based on the records, a corresponding sample label is added to the full-cycle data, and it is marked as a candidate valid fault sample.

[0084] Data generated by manual maintenance and testing, equipment debugging, abnormal data acquisition, or electromagnetic interference will not be added to the fault characteristic sample database as valid fault samples. Data whose automatic judgment results differ from maintenance records will be reviewed by maintenance personnel before being considered as valid fault samples.

[0085] Before adjusting the dynamic threshold model parameters or fault feature matching parameters using newly added valid fault samples, the edge computing device retains the currently used parameter version and generates candidate parameters based on the newly added valid fault samples. The candidate parameters can be determined using a preset parameter optimization algorithm; this invention does not limit the specific type of the parameter optimization algorithm.

[0086] The candidate parameters are validated using validation samples independent of the candidate parameter generation process. If a candidate parameter satisfies preset parameter constraints, and the fault determination result and fault segment location result corresponding to the candidate parameter are not inferior to the result corresponding to the current parameter, the candidate parameter is written as an update parameter into the edge computing device; otherwise, the current parameter continues to be used.

[0087] After the parameters are updated, the edge computing device continues to record the fault assessment results corresponding to the updated parameters. When the updated parameters do not meet the preset operation verification conditions, or when the abnormal judgment results exceed the preset range, the edge computing device will be restored to the parameter version retained before the parameter update.

[0088] To more clearly illustrate the core concept of the invention, a basic embodiment can be described. This embodiment describes a fault handling method executed on an edge computing device (such as a low-voltage smart converged terminal) deployed in a low-voltage distribution network, aiming to achieve rapid fault location and self-healing. Its core processes include dynamic threshold calculation, fault type matching, topology location, fault nature differentiation, and differentiated control.

[0089] In this basic embodiment, the method is initiated first. First, the system calculates a dynamic threshold for fault determination based on one or more general operating condition factors characterizing the current power grid operating state, such as the current load level (heavy load, normal, light load). Compared to using the same fixed threshold throughout the year, this approach better adapts to normal fluctuations in voltage and current baselines caused by load changes, thereby reducing false alarms. Second, when an abnormal electrical parameter is detected, exceeding the newly calculated dynamic threshold, the system matches the fault features generated in real time with a pre-set sample library containing at least two typical fault types stored locally on the device, to preliminarily determine which preset type the currently occurring fault belongs to.

[0090] The third step involves locating the fault section corresponding to the fault on a pre-stored distribution network topology model, which includes at least two node levels. In a simplified embodiment, the distribution network topology model may include transformer substation nodes, branch line nodes, and end-user nodes; in a preferred embodiment, the distribution network topology model includes a low-voltage transformer substation main node, a trunk line node, branch line nodes, and end-user nodes. By analyzing the timing and magnitude of changes in electrical parameters at different levels of monitoring nodes, and combining this with the topological connections between the monitoring nodes, the system narrows the fault range from the transformer substation level to the corresponding fault section.

[0091] The fourth step is for the system to determine the duration of the abnormal electrical parameters corresponding to the fault and the recovery deviation of the electrical parameters after the fault occurs, and to determine the fault as a transient fault or a permanent fault based on the duration of the abnormality and the recovery deviation.

[0092] Fifth, the system executes differentiated control strategies based on the determined faulty section and the fault determination result. When the fault is determined to be a permanent fault, the system isolates the faulty section; when the fault is determined to be a transient fault, the system does not perform isolation actions for the faulty section and records the event log locally.

[0093] By sequentially executing the above five core steps, this embodiment achieves the core inventive concept of fault location and self-healing without relying on complex algorithms and external conditions.

[0094] The following detailed description of a preferred embodiment of the present invention, in conjunction with the accompanying drawings and specific application scenarios, illustrates this embodiment. The application scenario is a low-voltage distribution area for urban residents with 30% distributed photovoltaic (PV) access. The network structure of this area includes one main line and six branch lines, supplying power to 128 end-users. This distribution area exhibits typical characteristics of significant diurnal load fluctuations and a certain degree of harmonic and three-phase imbalance phenomena.

[0095] like Figure 1 As shown, the method in this embodiment is executed by a low-voltage intelligent converged terminal deployed in the distribution area. First, the high-frequency data acquisition and preprocessing module on the terminal simultaneously acquires three-phase voltage, three-phase current, and zero-sequence current data of the main line, six branch lines, and all user terminals in the distribution area at a sampling frequency of 12.8kHz, and calculates derived characteristics such as power factor, harmonic distortion rate, three-phase imbalance, and real-time photovoltaic output in real time. After preprocessing, the acquired raw data is constructed into a refined time-series dataset, providing high-quality input for the subsequent fault diagnosis and location module.

[0096] like Figure 3As shown, since this distribution area is a high-PV grid connection area, the weight configuration of the dynamic threshold model is as follows: load fluctuation weight. Photovoltaic power output weight Environmental factor weights Harmonic interference weight Three-phase unbalanced weights The terminal executes once every 30 seconds. Figure 3 The process is as follows: First, collect multi-factor data; then, calculate the weighted coefficients; and finally, substitute them into the weighted calculation formula to calculate the dynamic threshold. The system then executes threshold upper and lower limit constraints. Subsequently, the system enters a pre-set waiting phase (e.g., waiting for a 30-second iteration cycle). After the timeout, the system performs a continuous operation status judgment: if a continuous operation command is received or the equipment is in normal service, the system returns to the start step (collecting multi-factor data) to start the next round of factor collection and threshold calculation; if a shutdown command is received or the equipment enters maintenance mode, the algorithm stops. For example, when the output of distributed photovoltaic power changes, the edge computing device corrects the corresponding dynamic threshold based on the impact of the distributed photovoltaic output on the normal operating baseline of the target electrical parameters, ensuring that the threshold used for abnormal event detection is adapted to the current operating conditions.

[0097] At 14:22 on a certain day, a single-phase-to-ground fault occurred on branch line 3 of the transformer substation, triggering the fault feature matching and analysis step. The terminal collected the feature vector at the time of the fault occurrence and performed Euclidean distance similarity matching with the standard feature vector in the local fault sample library. The calculation results showed that the current fault feature had the highest similarity to the standard feature vector corresponding to the single-phase-to-ground fault, and the highest similarity reached the preset judgment threshold. Therefore, the current abnormal event was judged as a single-phase-to-ground fault.

[0098] Once the fault type is determined, the fourth-level topology hierarchical localization step is immediately initiated. For example... Figure 2 As shown, the terminal retrieves a pre-stored four-level topology node encoding table, which details the hierarchy and connection relationships from the main node of the distribution area (node ​​1) to each user terminal node (such as node 9 and node 10). The system compares the timing and magnitude of electrical parameter changes at each monitoring node level by level, based on the upstream and downstream topology connections. According to the comparison results, the node corresponding to branch line 3 (node ​​5) is identified as a candidate monitoring node. Combining this candidate monitoring node (node ​​5) with the topology connections between it and adjacent monitoring nodes (such as upstream mainline node 2, and downstream nodes 9 and 10), the fault range is determined to be the fault section corresponding to branch line 3.

[0099] Subsequently, the method proceeds to the step of implementing a differentiated self-healing strategy. For example... Figure 4As shown, the system first determines the duration of the abnormal electrical parameters corresponding to the current fault and the deviation of the electrical parameters from normal operation after the fault. Upon testing, the duration of the abnormal electrical parameters for the current fault reaches a preset time threshold, and the electrical parameters have not recovered to the preset normal operating range after the fault occurred. Therefore, the current fault is classified as a permanent fault.

[0100] The differentiated control module determines the target controllable switch corresponding to the identified fault section. Before sending the isolation control command, the edge computing device verifies at least one of the following: the current distribution network topology, the current state of the target controllable switch, the correspondence between the target controllable switch and the fault section, the coordination relationship between upper and lower level protection systems, and the operating status of distributed power sources.

[0101] When the fault is determined to be a permanent fault, the reliability level of the location result reaches the preset control conditions, and the target controllable switch is located at the isolation boundary of the fault section, an isolation control command is sent to the target controllable switch.

[0102] After sending the isolation control command, the edge computing device reads the switch position status fed back by the target controllable switch. When it receives status feedback indicating that the isolation action is completed within a preset confirmation time, it records the isolation result and determines the power supply status of the non-faulty section based on the updated distribution network topology.

[0103] If no valid status feedback is received within the preset confirmation time, or if the feedback switch position status indicates that the target controllable switch has not completed the isolation action, the target controllable switch is marked as a non-operation state, and a switch non-operation alarm is generated.

[0104] With a pre-stored backup isolation scheme, the edge computing device determines the backup controllable switch based on the updated switch status and distribution network topology. When the backup controllable switch meets the preset safety verification conditions, it sends an isolation control command to the backup controllable switch. If there is no backup controllable switch that meets the safety verification conditions, it stops subsequent automatic control and outputs a manual handling prompt.

[0105] To demonstrate the handling of another type of fault, this embodiment adds a scenario. At 16:10 on the same day, a transformer area experiences a momentary voltage flicker due to a lightning strike. The system detects that the duration of the current abnormal event is less than a preset time threshold, and the electrical parameters return to the preset normal operating range after the abnormal event ends. Therefore, the current fault is determined to be a momentary fault. The system does not perform isolation actions for the faulty segment and records the corresponding event log locally on the edge computing device. The system can also count the number of times a momentary fault occurs in the same faulty segment within a preset time window. When the number reaches a preset threshold, an alarm message is generated for the faulty segment. It should be noted that, whether the local isolation control strategy for permanent faults is executed or the non-tripping and recording / alarm actions for momentary faults are executed, the system will locally store and record the fault characteristic data and handling results of this occurrence, reset the monitoring status, and wait for the next round of fault detection.

[0106] Finally, the system performs sample screening and model parameter updates. The sample self-iterative optimization module associates the full-cycle data of this abnormal event with protection action records, controllable switch status records, and operation and maintenance confirmation results. After confirming the corresponding fault type, fault section, and fault nature, it adds sample tags to the full-cycle data and supplements it to the fault feature sample library.

[0107] When a newly added valid fault sample meets the preset update conditions, the system generates candidate parameters based on the newly added valid fault sample, and verifies the candidate parameters using independent verification samples while retaining the current parameter version. If the candidate parameter meets the preset verification conditions, the parameter is updated; if it does not meet the preset verification conditions, the current parameter continues to be used. If the updated parameter does not meet the preset operation verification conditions, the system reverts to the parameter version before the parameter update. Through the above steps, this embodiment illustrates how edge computing devices combine adaptive dynamic thresholds, fault feature matching, fault segment location, fault nature determination, and differentiated control to handle abnormal events in low-voltage distribution networks. By completing fault assessment and control strategy generation at the edge, it helps reduce reliance on real-time interaction with remote distribution master stations; by using different control methods for permanent and transient faults, it helps reduce unnecessary isolation actions for transient faults and narrows the isolation scope of permanent faults.

Claims

1. A method for intelligent fault location and self-healing in low-voltage distribution networks based on edge computing and adaptive operating conditions, characterized in that, The method is performed by at least one edge computing device deployed on the low-voltage distribution network side, including: Acquire real-time data from multiple monitoring nodes and at least one operating condition factor characterizing the real-time operating status of the low-voltage distribution network, and update the dynamic threshold for abnormal event detection based on the operating condition factor. In response to detecting an abnormal event based on the real-time collected data and the dynamic threshold, fault features are extracted from the real-time collected data corresponding to the abnormal event, and the fault features are matched with a fault feature sample library pre-installed on the edge computing device to determine the fault type corresponding to the abnormal event. Based on a pre-stored distribution network topology model, the timing and magnitude of changes in the electrical parameters of the multiple monitoring nodes are analyzed to determine the fault section corresponding to the abnormal event. Based on the changes in electrical parameters during and after the occurrence of the abnormal event, a fault nature determination result is obtained, which is used to characterize whether the abnormal event is a transient fault or a permanent fault. A differentiated control strategy is generated and executed based on the faulty section and the fault nature determination result. Specifically, when the fault nature determination result indicates that the abnormal event is a permanent fault, a control strategy is executed to isolate the faulty section and maintain power supply to the non-faulty section. When the fault nature determination result indicates that the abnormal event is a transient fault, no isolation action is executed for the faulty section, and the corresponding event log is recorded.

2. The method according to claim 1, characterized in that, The operating condition factors include at least one of load fluctuation, distributed photovoltaic output, environmental factors, harmonic distortion, and three-phase imbalance. The updating of the dynamic threshold for abnormal event detection based on the operating condition factor includes: The obtained working condition factors are dimensionless. Based on the weighting coefficients corresponding to each working condition factor, the preset initial threshold is weighted and corrected to obtain the corrected threshold; and the corrected threshold is limited to a preset threshold range determined according to the initial threshold to obtain the dynamic threshold.

3. The method according to claim 1, characterized in that, The fault characteristics include at least one of the following: electrical parameter amplitude variation characteristics, harmonic distortion characteristics, zero-sequence electrical quantity characteristics, and time-sequence variation characteristics; Matching the fault features with the fault feature sample library includes: The fault features and multiple standard feature vectors in the fault feature sample library are normalized according to the same normalization rule. The similarity between the normalized fault features and each standard feature vector is calculated respectively; and when the highest similarity reaches a preset judgment threshold, the fault type represented by the standard feature vector corresponding to the highest similarity is determined as the fault type corresponding to the abnormal event.

4. The method according to claim 1, characterized in that, The analysis of the timing and magnitude of changes in the electrical parameters of the multiple monitoring nodes to determine the fault section corresponding to the abnormal event includes: Based on the upstream and downstream connection relationships between each monitoring node in the power distribution network topology model, the multiple monitoring nodes are compared level by level. Candidate monitoring nodes are selected from the plurality of monitoring nodes based on the time when the electrical parameters of each monitoring node change; and when there are multiple candidate monitoring nodes, the fault section is determined by combining the magnitude of the electrical parameter changes of each candidate monitoring node and the topological connection relationship between the candidate monitoring nodes.

5. The method according to claim 1, characterized in that, The method of obtaining a fault nature determination result based on changes in electrical parameters during and after the occurrence of the abnormal event includes: Determine the duration of the abnormal electrical parameters corresponding to the abnormal event; Determine the recovery deviation of the electrical parameters relative to the preset normal operating range after the occurrence of the abnormal event; When the duration of the anomaly is less than a preset time threshold and the recovery deviation is not greater than a preset recovery threshold, the fault nature determination result indicates that the anomaly event is an instantaneous fault; and when the duration of the anomaly is not less than the preset time threshold, or the recovery deviation is greater than the preset recovery threshold, the fault nature determination result indicates that the anomaly event is a permanent fault.

6. The method according to claim 1, characterized in that, The multiple monitoring nodes are distributed across at least two levels of low-voltage distribution area main nodes, trunk line nodes, branch line nodes, and user terminal nodes. The acquisition of real-time data from multiple monitoring nodes includes: According to a unified time reference, at least one of the voltage data, current data, and zero-sequence electrical quantity data of the multiple monitoring nodes is collected synchronously at high frequency. Furthermore, at least one of the following processes is performed on the collected data: time alignment, outlier removal, waveform noise reduction, and data standardization, to obtain the real-time collected data.

7. The method according to claim 1, characterized in that, The differentiated control strategy implemented for the transient fault also includes: Based on the event log, the number of instantaneous faults occurring in the same fault segment within a preset time window is counted; When the number of instantaneous faults reaches a preset threshold, an alarm message is generated for the faulty section; and when the number of instantaneous faults does not reach the preset threshold, the power supply to the faulty section is maintained.

8. The method according to claim 1, characterized in that, The method further includes: The edge computing device locally retains full-cycle data before, during, and after each abnormal event, including at least one of real-time acquired data, fault characteristics, fault type, fault segment, and execution results of differentiated control strategies. Based on the preset abnormal waveform identification rules and the correspondence between the full-cycle data and external operation and maintenance records, valid fault samples are screened from the retained full-cycle data. The effective fault samples are added to the fault feature sample library; and, based on the effective fault samples, at least one of the parameters used to update the dynamic threshold and the parameters used to match the fault features is adjusted, and when the adjusted parameters meet the preset verification conditions, the corresponding parameters on the local edge computing device are updated using the adjusted parameters.

9. A low-voltage distribution network fault intelligent location and self-healing system based on edge computing and operating condition adaptation, characterized in that, It includes multiple monitoring terminals, at least one edge computing device, and at least one controllable switch; The multiple monitoring terminals are respectively installed at multiple monitoring nodes of the low-voltage distribution network to collect real-time data from the corresponding monitoring nodes. The edge computing device is communicatively connected to the plurality of monitoring terminals and the at least one controllable switch, and is used to execute the method of any one of claims 1 to 8 to determine the fault segment corresponding to the abnormal event and send a control command to the target controllable switch of the at least one controllable switch corresponding to the fault segment. The target controllable switch is used to respond to the control command to perform isolation action on the faulty section or maintain the power supply state of the faulty section.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 8.

Images