Fault diagnosis method and system for power distribution system

Abstract

The invention belongs to the field of power distribution system fault diagnosis, discloses a power distribution system fault diagnosis method and system, and solves the problem of inaccurate diagnosis caused by difficult cross-regional data collaboration and inconsistent time in the prior art through edge node collaboration, data snapshot acquisition, timestamp alignment and cross-regional data association analysis. Rapid and accurate cross-regional fault diagnosis can be realized, and the power supply reliability of a power distribution system is improved.

Description

Technical Field

[0001] The present application relates to the field of power distribution system fault diagnosis, and more specifically, to a power distribution system fault diagnosis method and system. Background Art

[0002] Traditionally, fault diagnosis in power distribution systems relies primarily on centralized control systems to analyze data from various locations. However, distribution systems cover a wide geographic area, encompassing a vast number of equipment units and connected users, and the impact of faults can span multiple regions. To improve the efficiency and timeliness of fault diagnosis, edge computing nodes with sufficient computing and storage capabilities have been deployed at key locations within the distribution system in recent years. These edge nodes collect regional operational data in real time and at high frequency, utilizing local computing power for preliminary processing and analysis.

[0003] However, the topology of power distribution systems is complex, with numerous and interconnected devices. Faults can occur anywhere in the system, and their impact is often not limited to the monitoring area of a single edge node but can quickly spread across the monitoring ranges of multiple edge nodes. Different types of faults, especially complex or hidden ones, can exhibit significantly different characteristics when detected by sensors in different locations. Due to the limitations of their connectivity, a single edge node can only capture data from its local area, lacking a complete, real-time view of the fault's global impact, its propagation path within the system, and the potential chain reactions it may trigger.

[0004] In order to accurately determine the type of fault, precisely locate the fault point, assess the true scope of the fault's impact, and identify whether there are other faults caused by or related to the initial fault, efficient and timely information coordination is required between multiple edge nodes and between edge nodes and the central system. Although the central control system receives information from multiple edge nodes, due to data transmission delays and possible time inconsistencies in information from different nodes, it takes a long time to aggregate, align, and analyze this massive, dispersed information. In cases where a fault develops rapidly or involves the interaction of multiple links, information delays or inconsistencies may make it impossible for the central system to make an accurate judgment within the specified time, thereby missing the optimal isolation opportunity or taking incorrect isolation measures, leading to an expansion of the power outage and seriously affecting power supply reliability.

[0005] Therefore, in an environment like the power distribution system, which has a wide geographical distribution, complex topology, and limited data acquisition range of edge nodes, how to achieve effective collaboration between edge nodes to quickly and accurately determine the area involved in the fault and identify related faults after a fault occurs is a key technical challenge facing current distribution system fault diagnosis.

[0006] In view of the above problems, the existing technology is in urgent need of improvement. Summary of the Invention

[0007] The purpose of this application is to provide a distribution system fault diagnosis method and system, which can realize fast and accurate cross-regional fault diagnosis and improve the power supply reliability of the distribution system.

[0008] In a first aspect, the present application provides a power distribution system fault diagnosis method, which is applied to an edge node of the power distribution system to perform fault diagnosis. The method comprises the following steps: A1. Extract electrical quantity and switch status data collected by this edge node to identify fault-related abnormal signals; A2. When a fault-related abnormal signal is identified, an abnormal event report is generated containing the abnormality type, abnormality occurrence time, and abnormality location; A3. According to the abnormal event report and the distribution system topology, determine the collaborative node group associated with the abnormal event; the distribution system topology includes the connection relationship between the distribution equipment and the connection relationship between each edge node and each distribution equipment; A4. Based on the anomaly type and occurrence time in the anomaly event report, a data snapshot request containing the corresponding data snapshot time range is sent to the collaborative node group. This enables the collaborative node group to collect a data snapshot of local data within the data snapshot time range, append a high-precision local timestamp to the data snapshot, and then send it to the aggregation node for cross-regional data correlation analysis. The aggregation node is the local edge node or another pre-defined edge node. A5. If a data snapshot is received from a collaborative node group, the node data streams are aligned based on the high-precision local timestamp attached to the data snapshot, cross-regional data correlation analysis is performed, and a fault diagnosis report is generated; the fault diagnosis report includes at least one item from the list of fault type, fault location, fault-affected area, associated faults, and chain events.

[0009] In a second aspect, the present application provides a distribution system fault diagnosis system, the system comprising a plurality of edge nodes, the plurality of edge nodes comprising general nodes and aggregation nodes; the edge nodes being deployed in different areas of the distribution system; the edge nodes being provided with a fault diagnosis program, the fault diagnosis program of the general nodes being configured with an abnormality identification module, an abnormal event report generation module, a collaborative node determination module, a data snapshot request module, and a data snapshot response module, and the fault diagnosis program of the aggregation node being configured with an abnormality identification module, an abnormal event report generation module, a collaborative node determination module, a data snapshot request module, a data snapshot response module, and an association analysis module; The anomaly identification module is used to extract the electrical quantity and switch status data collected by the edge node to identify fault-related abnormal signals; The abnormal event report generation module is used to generate an abnormal event report containing the abnormal type, abnormal occurrence time and abnormal location when a fault-related abnormal signal is identified; The collaborative node determination module is used to determine the collaborative node group associated with the abnormal event based on the abnormal event report and the topological relationship of the power distribution system; the topological relationship of the power distribution system includes the connection relationship between each distribution device and the connection relationship between each edge node and each distribution device; The data snapshot request module is used to send a data snapshot request containing a corresponding data snapshot time range to the collaborative node group according to the abnormality type and abnormality occurrence time in the abnormal event report; The data snapshot response module is used to respond to the data snapshot request upon receiving the data snapshot request, collect a data snapshot of local data within the data snapshot time range, attach a high-precision local timestamp to the data snapshot, and send it to the aggregation node for cross-region data association analysis; The association analysis module is used to align the node data stream according to the high-precision local timestamp attached to the data snapshot when receiving the data snapshot, perform cross-regional data association analysis, and generate a fault diagnosis report; the fault diagnosis report includes at least one item from the fault type, fault location, fault-related area, associated faults and chain event list.

[0010] Beneficial effects: The present application provides a distribution system fault diagnosis method and system, which solves the problems of difficulty in cross-regional data collaboration and inaccurate diagnosis caused by time inconsistency in the existing technology through edge node collaboration, data snapshot collection and timestamp alignment, and cross-regional data correlation analysis. It can achieve fast and accurate cross-regional fault diagnosis and improve the power supply reliability of the distribution system. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a flow chart of a power distribution system fault diagnosis method provided in an embodiment of the present application.

[0012] Figure 2 Configure a diagram for the troubleshooting procedure for a general node.

[0013] Figure 3 Configure a diagram for the troubleshooting procedure of the sink node.

[0014] Explanation of the numbers: 1. Abnormal identification module; 2. Abnormal event report generation module; 3. Collaborative node determination module; 4. Data snapshot request module; 5. Data snapshot response module; 6. Correlation analysis module. DETAILED DESCRIPTION

[0015] The technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. The components of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of this application.

[0016] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0017] refer to Figure 1 This application proposes a power distribution system fault diagnosis method, which is applied to the edge nodes of the power distribution system to perform fault diagnosis. The steps of the method include: A1. Extract electrical quantity and switch status data collected by this edge node to identify fault-related abnormal signals; A2. When a fault-related abnormal signal is identified, an abnormal event report is generated containing the abnormality type, abnormality occurrence time, and abnormality location; A3. According to the abnormal event report and the distribution system topology, determine the collaborative node group associated with the abnormal event; the distribution system topology includes the connection relationship between the distribution equipment and the connection relationship between each edge node and each distribution equipment; A4. Based on the anomaly type and occurrence time in the anomaly event report, a data snapshot request containing the corresponding data snapshot time range is sent to the collaborative node group. This enables the collaborative node group to collect a data snapshot of local data within the data snapshot time range, append a high-precision local timestamp to the data snapshot, and then send it to the aggregation node for cross-regional data correlation analysis. The aggregation node is the local edge node or another pre-defined edge node. A5. If a data snapshot is received from a collaborative node group, the node data streams are aligned based on the high-precision local timestamp attached to the data snapshot, cross-regional data correlation analysis is performed, and a fault diagnosis report is generated; the fault diagnosis report includes at least one item from the list of fault type, fault location, fault-affected area, associated faults, and chain events.

[0018] Among them, electrical quantities and switch status data are physical quantities and equipment status information that indicate the operating status of the distribution system. Electrical quantities may include voltage, current, frequency, etc. Switch status data may include the opening and closing status of equipment such as circuit breakers and disconnectors, which are used to reflect the real-time status of the system operation.

[0019] Among them, fault-related abnormal signals are abnormal changes in electrical quantities or switch status data that indicate a possible fault in the distribution system. They can manifest as sudden changes or exceeding limits in electrical quantities, or abnormal jumps in switch status. They are used to preliminarily determine whether there is an abnormality in the system.

[0020] Among them, the abnormal event report is a structured information carrier that encapsulates the key information of the abnormal event initially identified, including the abnormal type, abnormal occurrence time and abnormal location.

[0021] Among them, the distribution system topology describes the connection method between each device in the distribution system and the connection relationship between edge nodes and these devices. It is used to understand the propagation path and impact range of abnormal events in the system, and to identify the devices and areas related to the abnormalities.

[0022] Among them, the collaborative node group is a group of edge nodes that may be related to specific abnormal events and need to participate in data collection and information sharing. It is determined based on the abnormal location and system topology and is used to provide more comprehensive, cross-regional fault-related data.

[0023] The data snapshot time range is a specific time window during which data needs to be collected, determined based on the type and occurrence time of the abnormal event. It is used to ensure that the collected data is closely related to the abnormal event.

[0024] Among them, the data snapshot is a collection of local operating data collected within a specific time range, including data such as electrical quantities and switch status. It is used to provide detailed data information before and after the abnormality occurs.

[0025] Among them, the high-precision local timestamp is a precise time mark attached to the collected data by the local clock of the edge node. The local clock is synchronized with the global time source through a high-precision synchronization protocol. It is used to solve the time alignment problem of data from different nodes in a distributed system.

[0026] A sink node is a designated edge node responsible for receiving data snapshots from a collaborative node group and performing cross-regional data correlation analysis. This node can be the node initiating the exception report or another pre-defined node, centrally processing and analyzing cross-regional data. Among the edge nodes in the power distribution system, some are designated sink nodes, responsible for performing cross-regional data correlation analysis on data snapshots from other nearby edge nodes. Generally, edge nodes have stronger data processing capabilities than regular nodes.

[0027] Among them, aligning node data streams means adjusting the data sequences from different nodes based on the high-precision local timestamps attached to the data snapshots so that they are synchronized on the time axis. Methods such as interpolation, resampling, or filtering can be used to eliminate the impact of data time inconsistency on association analysis.

[0028] Among them, cross-regional data association analysis refers to the comprehensive analysis of time-aligned data from different collaborative nodes to identify the relationships and patterns between data. Methods such as Bayesian networks, machine learning models, or expert systems can be used to determine the fault type, location, impact range, and related faults.

[0029] Among them, the fault diagnosis report refers to a structured report that encapsulates the results of cross-regional data correlation analysis, including information such as fault type, fault location, fault-involved area, related faults and chain event lists, which is used to guide fault isolation and recovery operations.

[0030] The core innovation of this application lies in utilizing the local anomaly perception capabilities of edge nodes in the edge node environment of the distribution system, combining the topological relationship of the distribution system to intelligently determine the collaborative node group, and realizing the time synchronization convergence of cross-node data through a data snapshot mechanism with attached high-precision local timestamps, and finally performing cross-regional data correlation analysis at the aggregation node, thereby overcoming the limitations of the field of view of a single edge node and the problems of diagnostic delays and data time inconsistency of the traditional central system, realizing fast, accurate and comprehensive cross-regional fault diagnosis, and being able to identify related faults and chain events.

[0031] Specifically, this method is applied to edge nodes in a power distribution system. First, edge nodes continuously extract locally collected electrical quantity and switch status data to identify fault-related abnormal signals. Once an abnormal signal is identified, the edge node generates an abnormal event report containing the abnormality type, occurrence time, and location. This report serves as the basis for triggering collaboration. Next, based on the abnormality location information in the abnormal event report and the pre-set distribution system topology, a collaborative node group associated with the abnormal event is determined. This collaborative node group includes edge nodes in areas potentially affected by or related to the fault. Then, based on the abnormality type and occurrence time in the abnormal event report, a data snapshot time range for data collection is determined, and a data snapshot request is sent to the collaborative node group. Upon receiving the request, the edge nodes in the collaborative node group collect a data snapshot of local data within the specified time range, attach a high-precision local timestamp to the data snapshot, and then send the timestamped data snapshot to a designated sink node. After receiving the data snapshot from the collaborative node group, the sink node uses the high-precision local timestamp attached to the data snapshot to align data streams from different nodes, solving the time synchronization problem of distributed data. Finally, based on the aligned cross-regional data streams, cross-regional data correlation analysis is performed to comprehensively determine the fault situation and generate a fault diagnosis report containing information such as fault type, fault location, affected regions, and a list of related faults and chain reactions. This entire process, through collaboration among edge nodes and data aggregation analysis, enables rapid and accurate diagnosis of cross-regional faults.

[0032] Through the above scheme, the present application improves the timeliness of fault perception by performing preliminary anomaly identification locally at the edge node. By intelligently determining the collaborative node group based on abnormal events and system topology, unnecessary communication and computing overheads are avoided, and collaborative efficiency is improved. By attaching high-precision local timestamps to data snapshots and aligning them, the problem of data time asynchrony in distributed systems is effectively solved, laying the foundation for accurate analysis. By performing cross-regional data correlation analysis at the aggregation node, the limitations of the field of view of a single edge node are overcome, and the data characteristics of multiple related areas can be comprehensively analyzed, thereby improving the comprehensiveness and accuracy of fault diagnosis, and being able to identify cross-regional correlated faults and chain events. Overall, this scheme realizes fast, accurate and comprehensive cross-regional fault diagnosis in the edge node environment of the distribution system.

[0033] In some embodiments, step A1 comprises: A101. Extract electrical quantities and switch status data collected by this edge node; the electrical quantities include voltage, current, and frequency; the switch status data includes status information of circuit breakers and disconnectors; A102. Using a wavelet threshold denoising method to filter out high-frequency noise interference in the electrical quantity, thereby obtaining a denoised electrical quantity; A103. Based on the denoised electrical quantity and the switch state data, a characteristic signal is obtained; the characteristic signal includes the electrical quantity mutation amount, the number of switch state jumps, and the duration of the electrical quantity exceeding the limit; A104. If at least one of the characteristic signals exceeds the corresponding preset threshold range, it is determined that a fault-related abnormal signal is identified, and the corresponding characteristic signal is determined to be an abnormal signal.

[0034] Wavelet threshold denoising is a signal processing technique that uses wavelet transforms to decompose signals into different scales, performs threshold processing on the noise components in the wavelet domain, and then reconstructs the signal. This can be achieved using either hard or soft thresholding. High-frequency noise interference refers to unwanted signals with high frequency components introduced by sensors, communication links, or environmental factors. When superimposed on the original electrical signal, it can mask or distort the true fault characteristics.

[0035] Among them, characteristic signals refer to key indicators extracted from raw or preprocessed data that can effectively characterize the operating status or potential faults of the distribution system. Their function is to extract key information closely related to the fault mode. The electrical quantity mutation refers to the rapid change of the electrical quantity in a short period of time, reflecting a transient fault. It can be calculated as the absolute value of the difference between adjacent sampling points. The number of switch state jumps refers to the number of times the state of the switching device changes within a period of time, reflecting the operation or abnormality of the device. It can be calculated as the state change count within a unit time window. The electrical quantity over-limit duration refers to the length of time that the electrical quantity exceeds the normal range and continues, reflecting a steady-state abnormality. It can be calculated as the length of time that the electrical quantity continuously exceeds the threshold.

[0036] The preset threshold range refers to the standard limit for determining abnormalities based on normal operating data or simulations. This can be determined through statistical analysis of historical normal operating data or based on simulation models. Each electrical quantity's sudden change and duration of over-limit fluctuations, as well as the number of switch state data transitions, are assigned a corresponding preset threshold range. When determining abnormal signals, these are compared against the corresponding preset threshold range. Fault-related abnormal signals refer to abnormal signal behavior that may be caused by a distribution system fault. Abnormal signals are specific characteristic signals determined to be outside the normal range.

[0037] This method accurately identifies fault-related abnormal signals by processing electrical quantity and switch state data collected by edge nodes. First, local electrical quantity and switch state data are collected at the edge node; these data serve as the basis for abnormality determination. Considering that raw electrical quantity data may contain high-frequency noise, direct use can lead to misjudgment. Therefore, wavelet threshold denoising is used to preprocess the electrical quantities, effectively filtering out noise interference and producing smoother, more realistic electrical quantity signals. Based on the denoised electrical quantities and combined with the original switch state data, characteristic signals closely related to the fault phenomenon are extracted, including the magnitude of electrical quantity mutations, the number of switch state transitions, and the duration of electrical quantity violations. These characteristic signals refine the raw data and more effectively reflect system anomalies. Finally, the extracted characteristic signals are compared with preset normal range thresholds. If any characteristic signal exceeds the threshold, a fault-related abnormal signal is identified and the corresponding characteristic signal is marked as abnormal. This processing flow ensures accurate and robust anomaly determination, providing reliable input for subsequent fault diagnosis. In this way, this method can quickly and accurately capture potential fault signs on the edge node side, laying a solid foundation for the entire distribution system fault diagnosis process.

[0038] In some embodiments, step A2 comprises: A201. Determine the abnormality type according to the abnormal signal; the abnormality types include mutation abnormality, switch jump abnormality and over-limit abnormality; A202. Record the moment when the fault-related abnormal signal is identified as the abnormality occurrence time; A203. Based on the topology of the power distribution system, determine the power distribution equipment where the abnormal signal is detected as the abnormal location; A204. Encapsulate the determined abnormality type, the abnormality occurrence time, and the abnormality location into an abnormal event report.

[0039] In the above steps, determining the anomaly type based on the abnormal signal is a key step. The abnormal signal itself is an abstraction of the underlying data features, such as the sudden change in an electrical quantity, the number of switch state transitions, or the duration of an electrical quantity exceeding a limit. Mapping these abstract signals to physically meaningful anomaly types (such as sudden change, switch transition, and limit violation) requires a set of defined rules or algorithms. This can be achieved through preset threshold comparisons, pattern matching, or rule-based reasoning. For example, if the detected abnormal signal is a rapid, large change in voltage or current, it can be classified as a sudden change; if the abnormal signal is an unexpected change in the state of a circuit breaker or disconnector, it can be classified as a switch transition; and if the abnormal signal is a voltage or current that persists outside the normal operating range, it can be classified as a limit violation. This identification process gives the original abnormal signal a concrete physical meaning, facilitating subsequent analysis of the nature and potential causes of the anomaly.

[0040] Another feature requiring explanation is the ability to locate anomalies based on the distribution system topology. When an anomaly signal is identified, it is typically acquired by a sensor or monitoring unit connected to a specific distribution device. The distribution system topology records the connections between various devices in the system (such as transformers, switches, and line segments), as well as the relationships between edge nodes or sensors and these devices. The process of locating anomalies utilizes this topological information to trace back to the specific distribution device to which the sensor that acquired the anomaly signal is connected. For example, if the anomaly signal originates from a monitoring unit connected to a feeder switch, the anomaly can be located at that feeder switch or its vicinity. This requires the system to maintain an accurate, real-time distribution system topology database and be able to quickly retrieve corresponding device information based on the sensor or edge node identifier.

[0041] This application refines the steps of generating abnormal event reports and specifically provides how to accurately determine the abnormal type, abnormal occurrence time and abnormal location based on the identified abnormal signal, and encapsulate this information into a structured abnormal event report. First, according to the characteristics of the abnormal signal, such as its morphology, duration or associated device state changes, it is classified into one of the preset abnormal types. This makes the initial understanding of the abnormality more specific. At the same time, accurately recording the moment when the abnormal signal is identified provides a basis for subsequent time synchronization and data alignment. Then, using the topological information of the distribution system, the edge node or sensor that detects the abnormal signal is associated with the specific distribution equipment in the system to determine the spatial location of the abnormality. Finally, the key information such as the abnormal type, abnormal occurrence time and abnormal location determined are integrated into a structured abnormal event report.

[0042] Through the above method, the present application can accurately and in detail determine the type, time, and location of anomalies based on identified fault-related abnormal signals, and structure this key information into an abnormal event report. This provides accurate and reliable basic information for subsequent collaborative node determination, cross-regional data collection, and correlation analysis, effectively solving the problem of inaccurate subsequent diagnosis caused by merely identifying an abnormal signal but lacking detailed contextual information, significantly improving the efficiency and accuracy of distribution system fault diagnosis.

[0043] In some embodiments, step A3 comprises: A301. According to the abnormal location and distribution system topology in the abnormal event report, obtain a list of distribution equipment directly connected to the abnormal location, and use the distribution equipment in the list of distribution equipment as a primary associated distribution equipment; A302. For each level-associated distribution device, according to the distribution system topology, obtain the other distribution devices directly connected to the level-associated distribution device, remove the distribution devices already in the level-associated distribution device list, and use the remaining distribution devices as secondary-associated distribution devices; A303. The edge nodes connected by the primary associated power distribution equipment and the secondary associated power distribution equipment constitute the initial collaborative node group; A304. Evaluate the electrical distance between each edge node in the initial collaborative node group and the abnormal location; the electrical distance between the edge node and the abnormal location is: the number of distribution devices passed from the edge node to the abnormal location; A305. Eliminate edge nodes whose electrical distance is greater than a preset distance threshold from the initial collaborative node group to obtain a final collaborative node group.

[0044] The first-level associated distribution equipment refers to the distribution equipment directly connected to the abnormal location in the distribution system topology, which can be obtained by querying the topological relationship data and finding all device nodes directly connected to the abnormal location node.

[0045] Among them, the second-level associated power distribution equipment refers to other power distribution equipment that is directly connected to the first-level associated power distribution equipment but does not belong to the first-level associated power distribution equipment. It can be obtained by traversing the neighbor nodes of the first-level associated power distribution equipment and excluding the identified first-level associated power distribution equipment.

[0046] The initial collaborative node group refers to a set consisting of all edge nodes connected to the first-level associated power distribution equipment and the second-level associated power distribution equipment, which can be constructed by querying the topological relationship data, finding all edge nodes associated with these devices and merging the sets.

[0047] Among them, electrical distance usually refers to the number of devices passed from one device to another in the distribution system topology; therefore, the electrical distance between the edge node and the abnormal location refers to the number of distribution devices passed from the edge node to the abnormal location; it can be calculated on the distribution equipment connection diagram using the shortest path algorithm in graph theory.

[0048] The preset distance threshold refers to an upper limit of the electrical distance used to screen collaborative nodes, which can be set based on factors such as the scale of the power distribution system, fault propagation characteristics, diagnostic accuracy requirements, or historical experience.

[0049] This solution uses a step-by-step expansion and screening approach to determine the collaborative node group associated with the abnormal event based on the known abnormal location and distribution system topology in the abnormal event report. First, based on the abnormal location and topological relationship, directly connected first-level associated distribution devices are identified. Next, based on the first-level associated devices, second-level associated distribution devices are further identified. Finally, the edge nodes connected to the first- and second-level associated devices are aggregated to form an initial collaborative node group. To quantify the degree of association between these nodes and the abnormal location, the concept of electrical distance is introduced, and the electrical distance between each initial collaborative node and the abnormal location is calculated. Finally, by setting a preset distance threshold, nodes with excessive electrical distances are eliminated, resulting in the final collaborative node group. This series of steps is organically integrated, starting from the abnormal point, gradually expanding outward, and performing electrical distance-based screening. This ensures that the selected collaborative nodes are closely electrically connected to the abnormal event, providing a necessary and relatively streamlined data source for subsequent cross-regional data correlation analysis. In this way, this solution concretely implements the step of determining the collaborative node group, overcoming the data redundancy or insufficient coverage issues associated with simple node selection, thereby improving the efficiency and accuracy of overall fault diagnosis.

[0050] By gradually expanding the scope of associated devices and filtering based on electrical distance, this solution can specifically identify the group of collaborative nodes associated with the abnormal event. This ensures that the selected nodes cover the critical areas potentially affected by the fault, while avoiding the involvement of irrelevant nodes, reducing data redundancy and processing burden. By providing a necessary and relatively streamlined set of data sources, this solution improves the efficiency and accuracy of subsequent cross-regional data correlation analysis, thereby more effectively diagnosing distribution system faults.

[0051] Preferably, step A305 may include: Determining a distance correction coefficient for each edge node in the initial collaborative node group based on the device type and historical failure rate of the power distribution equipment connected to each edge node; Correcting the electrical distance corresponding to each edge node in the initial collaborative node group using the distance correction coefficient; The edge nodes whose corrected electrical distances are greater than a preset distance threshold are eliminated from the initial collaborative node group to obtain the final collaborative node group.

[0052] The distance correction factor is a numerical factor used to adjust the original electrical distance. This numerical factor reflects the potential contribution of the type and historical failure rate of the power distribution equipment connected to the edge node to the propagation or impact range of the fault. It can be determined using a table lookup method based on preset rules, an assignment method based on expert experience, or a machine learning model based on historical data. The distance correction factor is a value in the range [0, 1]. Generally, the more important the edge node, the smaller its distance correction factor, so that the corrected electrical distance is smaller, allowing important edge nodes to be included in the collaborative node group. The importance of the edge node is determined by the type and historical failure rate of the power distribution equipment connected to it. The higher the importance level of the connected power distribution equipment and the higher the historical failure rate, the more important the edge node is, and the corresponding distance correction factor is smaller.

[0053] When the distance correction coefficient is used to correct the electrical distance corresponding to each edge node in the initial cooperative node group, the distance correction coefficient may be directly multiplied by the original electrical distance.

[0054] This solution addresses the shortcomings of simple electrical distance measurement when determining collaborative node groups. It optimizes electrical distance calculation by introducing a distance correction factor, thereby more accurately identifying collaborative node groups associated with abnormal events. Specifically, after obtaining the initial collaborative node group, instead of simply using the number of devices as the electrical distance for screening, a distance correction factor is first determined for each edge node in the initial collaborative node group based on the device type and historical failure rate of the power distribution equipment connected to each edge node. This factor takes into account the device's inherent properties. For example, certain types of equipment may play a more critical role in fault propagation, or devices with higher historical failure rates may be more likely to become fault sources or weak links in the propagation path. By incorporating these factors, the distance correction factor more precisely quantifies the impact of device properties on electrical correlation. This distance correction factor is then used to correct the original electrical distance to obtain a corrected electrical distance. This corrected electrical distance incorporates multiple factors, such as the number of devices, device type, and historical failure rate, to more accurately reflect the electrical connection between the edge node and the abnormal location and its potential importance in fault diagnosis. Finally, based on this corrected electrical distance, screening is performed to eliminate edge nodes whose corrected electrical distance is greater than the preset distance threshold, thereby obtaining the final collaborative node group. In this way, the screened collaborative node group not only takes into account the physical electrical connection "distance" (number of devices), but also incorporates the "electrical importance" or "fault correlation" brought by the device attributes, so that the final collaborative node group can more accurately include those edge nodes that are critical to fault diagnosis, providing a higher-quality data source for subsequent cross-regional data correlation analysis, thereby improving the accuracy and efficiency of fault diagnosis. This optimization of the collaborative node group screening process is an effective improvement to the simple electrical distance screening method in the aforementioned solution, allowing the entire fault diagnosis process to more accurately focus on the areas and data that are truly relevant to abnormal events.

[0055] By determining the distance correction coefficient of each edge node based on the equipment type and historical failure rate of the distribution equipment connected to each edge node in the initial collaborative node group, and using the distance correction coefficient to correct the electrical distance corresponding to each edge node in the initial collaborative node group, and finally eliminating the edge nodes whose corrected electrical distance is greater than the preset distance threshold, the final collaborative node group is obtained. This scheme can overcome the shortcomings of simple electrical distance measurement methods and more accurately evaluate the degree of electrical correlation between edge nodes and abnormal locations, thereby screening out collaborative nodes that are more closely associated with faults and more valuable for fault diagnosis, improving the accuracy of collaborative node group determination, and thereby improving the accuracy and efficiency of subsequent cross-regional data association analysis.

[0056] Furthermore, after step A305, the following steps may also be included: A306. Based on the abnormal location in the abnormal event report and the current operating mode of the power distribution system, a preset regional interconnection information database is queried to identify other areas electrically connected to the area where the abnormal location is located. Edge nodes within these areas that are not part of the final coordinated node group are selected as candidate supplementary edge nodes. The regional interconnection information database records the electrical connection relationships between various areas in the power distribution system under different operating modes. A307. Evaluate the electrical accessibility between each candidate supplementary edge node and the abnormal location based on the distribution system topology. Electrical accessibility indicates whether there is an effective electrical path between the candidate supplementary edge node and the abnormal location, or whether there is a mutual influence or correlation between the electrical state changes between the candidate supplementary edge node and the abnormal location. A308. Add the candidate supplementary edge nodes whose electrical reachability is reachable to the final cooperative node group.

[0057] The current operating mode of the power distribution system refers to the operating state of the power distribution system at a specific moment, such as normal operation, specific load distribution or maintenance operation, which will affect the electrical connection paths between different areas in the system.

[0058] Among them, the preset regional interconnection information database refers to a storage structure that records the electrical connection relationship between different areas in the distribution system under various possible working modes, such as which areas form electrical paths through interconnecting switches or transformers in a certain mode. Other areas that are electrically connected to the area where the abnormal location is located refer to other geographical or logical areas that form effective electrical connections with the area where the abnormality occurs through electrical equipment under the current working mode of the distribution system. The edge nodes in these areas refer to edge computing nodes deployed in these other areas with electrical connections. Edge nodes that do not belong to the final collaborative node group refer to edge nodes that have not been included in the current collaborative node group after previous screening based on topological proximity and electrical distance. Candidate supplementary edge nodes refer to edge nodes that have been screened from other areas that are electrically connected to the area where the abnormal location is located and have not yet been included in the current collaborative node group. They are potential objects that need further evaluation to determine whether they should be added to the collaborative node group.

[0059] Evaluating the electrical reachability between each candidate supplementary edge node and the abnormal location refers to determining whether there is an effective electrical path between the candidate supplementary edge node and the abnormal location, or whether there is a mutual influence or correlation between their electrical state changes. This can be achieved by analyzing the system topology, device status, or electrical parameter correlation. Electrical reachability of reachable means that the evaluation results indicate that there is an effective electrical connection or mutual influence between the candidate supplementary edge node and the abnormal location.

[0060] Based on the above technical features, the overall working principle of the method for determining a collaborative node group in the present application can be explained as follows: After determining a collaborative node group based on the abnormal location and the distribution system topology, to address the issue of insufficient consideration of cross-regional electrical connectivity within the distribution system under different operating modes, the system queries a pre-defined regional interconnection information database based on the abnormal location in the abnormal event report and the current actual operating mode of the distribution system. This database details the electrical connectivity status of each region in the distribution system under different operating modes. Through this query, the system can identify other regions that are electrically connected to the region where the abnormal location is located under the current operating mode. These regions may not be directly adjacent to the abnormal location geographically or topologically, but may have significant electrical connections under specific operating conditions. The system then selects edge nodes within these identified associated regions that have not yet been included in the final collaborative node group and identifies them as potential supplementary nodes, namely candidate supplementary edge nodes. This process ensures that even cross-regional, topologically non-adjacent, but electrically related nodes can be preliminarily identified. The system then evaluates the electrical reachability between these candidate supplementary edge nodes and the abnormal location based on the current topology of the distribution system. This evaluation is designed to confirm whether these candidate nodes are effectively electrically connected to or influence the abnormal location, thereby eliminating nodes that are located in the associated area but have weak electrical connections. Finally, candidate supplementary edge nodes that are evaluated as electrically reachable are added to the final collaborative node group.

[0061] Through this series of steps, the present application is able to identify and include edge nodes that have cross-regional electrical connections with the fault area under a specific working mode, even if these nodes are not topologically adjacent. This enables the final collaborative node group to more comprehensively cover areas affected by or related to the fault, providing a more complete and representative data basis for subsequent cross-regional data association analysis. Compared with methods that rely solely on local topological proximity and electrical distance, this solution can more accurately identify the entire range of nodes related to the fault by introducing working modes and cross-regional interconnection information, combined with electrical reachability assessment, especially when dealing with complex faults involving cross-regional electrical connections or influences, significantly improving the comprehensiveness and accuracy of fault diagnosis.

[0062] In some embodiments, step A4 comprises: A401. Determine the data snapshot time range based on the abnormality type and abnormality occurrence time reported in the abnormal event report; A402. If this edge node is one of the preset sink nodes, then the edge node is set as the target sink node; if this edge node is not one of the preset sink nodes, then based on the network topology distance between the abnormal location and each preset sink node, select the preset sink node with the smallest network topology distance as the target sink node; A403. Add the data snapshot time range and the ID of the target aggregation node to the data snapshot request and send it to the collaborative node group, so that the collaborative node group collects the data snapshot of the local data within the data snapshot time range, and attaches a high-precision local timestamp to the data snapshot, and then sends it to the target aggregation node for cross-regional data association analysis.

[0063] Different time window lengths can be preset based on different anomaly types (for example, for sudden anomalies, the time window can be set from tens of milliseconds before the anomaly to over a hundred milliseconds after it occurs; for over-limit anomalies, the time window can be set to the duration of the anomaly plus a certain margin before and after). These preset time windows can be recorded in a time window lookup table, and when the data snapshot time range needs to be determined, the table can be consulted to determine it. This ensures that the subsequently collected data is closely related to the anomaly event, providing an effective and necessary data foundation for subsequent correlation analysis.

[0064] Among them, the preset aggregation nodes refer to a group of edge nodes pre-designated when the system is deployed. These nodes have strong computing, storage and communication capabilities and can serve as the center for data aggregation and correlation analysis.

[0065] Among them, the network topology distance refers to the number of network hops required to travel from one node to another in the distribution system communication network, or a comprehensive distance measurement calculated based on factors such as network bandwidth and delay.

[0066] The data snapshot request method of the present application operates according to the following principles: First, when an edge node detects a fault-related anomaly and generates an anomaly event report, the edge node dynamically determines a specific data snapshot time range based on the anomaly type and occurrence time in the report. Different types of anomalies manifest differently in the temporal dimension. For example, a transient fault may only require data from the very short period before and after the anomaly occurs, while a persistent fault may require a longer data window. By determining the time range based on the anomaly's characteristics, the collected data is ensured to be the most relevant and valuable data for the current anomaly event, avoiding the burden of collecting excessive irrelevant data and preventing the omission of data from critical time periods. Next, the edge node determines a target sink node to receive the data snapshot from the collaborative node group. If the edge node is already one of the pre-set sink nodes, it directly sets itself as the target sink node, leveraging local aggregation and computing capabilities and avoiding data transmission delays. If the edge node is not a pre-set sink node, the pre-set sink node with the smallest network topological distance from the anomaly location to each pre-set sink node is selected as the target sink node. Selecting the node with the smallest network topology distance as the aggregation point usually means the shortest data transmission path, which can effectively reduce the transmission delay when the collaborative node sends the data snapshot to the aggregation node, thereby improving the efficiency and timeliness of data aggregation. Finally, the edge node adds the determined data snapshot time range and the ID of the target aggregation node to the data snapshot request, and sends the request to the collaborative node group. After receiving the request, the collaborative node group collects a data snapshot of the local data according to the time range specified in the request, and attaches a high-precision local timestamp to the data snapshot. Attaching a high-precision local timestamp is a key technical means to ensure that data from different collaborative nodes can be accurately aligned when the target aggregation node performs association analysis, solving the problem of data time synchronization in a distributed system. Finally, the collaborative node group sends the data snapshot with a high-precision local timestamp to the target aggregation node for cross-regional data association analysis. By optimizing the time range of the data snapshot and the selection of the aggregation node, this application can provide a more relevant and timely data basis for subsequent cross-regional data association analysis, thereby improving the efficiency and accuracy of overall fault diagnosis.

[0067] By adopting the above technical solution, the present application can achieve the following technical effects: determining the data snapshot time range based on the abnormality type and abnormality occurrence time reported by the abnormal event, ensuring that data related to fault diagnosis is collected, and improving the pertinence and effectiveness of data collection. Selecting the target aggregation node based on the network topology distance between the abnormal location and the preset aggregation node reduces data transmission delays and improves the efficiency and timeliness of data aggregation. It provides a high-quality, timely, and easy-to-align data foundation for subsequent cross-regional data correlation analysis, thereby improving the efficiency and accuracy of fault diagnosis.

[0068] In some embodiments, step A5 comprises: A501. Use a Kalman filter algorithm to fuse data snapshots from the collaborative node group and align the node data streams. The data snapshots include voltage, current, and switch status data. The state variables of the Kalman filter algorithm include voltage amplitude, current amplitude, and switch status. The observed variables are the sampled values of voltage, current, and switch status reported by each collaborative node. A502. Based on the aligned node data streams, a Bayesian network is used to perform cross-regional data association analysis to determine the fault type, fault location, and affected area. The nodes of the Bayesian network include fault type, fault location, affected area, voltage amplitude, current amplitude, and switch status. The edges of the Bayesian network represent dependencies between nodes, which are determined based on the distribution system topology and historical fault data. A503. Based on the fault type and fault location determined, query the preset fault knowledge base to obtain a list of associated faults and chain events related to the fault type and fault location; the fault knowledge base records the correspondence between various fault types, fault locations, associated faults and chain events in the power distribution system; A504. Encapsulate the determined fault type, fault location, fault-involved area, and the obtained list of associated faults and chain events into a fault diagnosis report.

[0069] The Kalman filter algorithm refers to an optimal estimation algorithm, which can be implemented by using a linear Kalman filter or an extended Kalman filter.

[0070] The Bayesian network is a probabilistic graphical model that can be constructed using methods based on structure learning and parameter learning. The fault knowledge base is a database that stores fault-related information and can be implemented using a relational database or a graph database.

[0071] In this solution, after receiving data snapshots from a coordinated node group, these data may be temporally inconsistent due to transmission delays, minor node clock deviations, and other factors. Step A501 uses the Kalman filter algorithm to fuse these multi-source data snapshots and align the node data streams. The Kalman filter algorithm can process time series data with noise and uncertainty. By establishing a system state model (including voltage amplitude, current amplitude, and switch state) and an observation model, it uses sampled values from different nodes as observation variables to estimate and predict the system state in real time. This process not only filters out measurement noise but, more importantly, accurately synchronizes and fuses data from different sources based on data timestamps and the system dynamic model. This results in a set of node data streams that are highly temporally aligned and provide more accurate state estimates. This data fusion and state estimation capability of the Kalman filter provides a high-quality, temporally consistent data foundation for subsequent correlation analysis. The Kalman filter algorithm is state-of-the-art, and its specific process is not described in detail here.

[0072] Next, based on the Kalman filter-aligned node data streams, step A502 uses a Bayesian network to perform cross-regional data correlation analysis. Bayesian networks are powerful probabilistic reasoning tools that can represent and process probabilistic dependencies between variables. A Bayesian network is constructed, whose nodes represent potential fault information (fault type, fault location, and fault-affected area) as well as observable electrical quantities and switch states. Edges between nodes are established based on the actual topology of the distribution system and causal or correlation relationships determined by historical fault data. Using the aligned node data streams as evidence input to the Bayesian network, Bayesian inference algorithms (such as belief propagation and Markov Chain Monte Carlo) can be used to calculate the posterior probabilities of various fault hypotheses. Because the input data is precisely aligned and fused, the Bayesian network can more accurately capture the correlations between different regions and device states, thereby robustly determining the most likely fault type, location, and fault-affected area in complex cross-regional scenarios. This probabilistic model-based analysis method can effectively process multi-source heterogeneous data, improving the accuracy and reliability of fault diagnosis.

[0073] Again, after determining the type and location of the initial fault, step A503 obtains a list of associated faults and chain events related to the fault by querying the preset fault knowledge base. The fault knowledge base stores the historical operation and fault experience of the distribution system, and records other faults or events that may be caused by specific fault types and locations. By querying the knowledge base, the current diagnostic results can be linked to known system vulnerabilities or fault propagation paths to predict potential secondary faults or chain reactions. This makes fault diagnosis no longer limited to the initial fault itself, but can provide a more comprehensive fault impact assessment, provide early warning information to operation and maintenance personnel, and support the formulation of more comprehensive fault handling and risk control strategies.

[0074] Finally, step A504 packages the diagnostic results and knowledge base query results into a complete fault diagnosis report. This report not only includes the core fault type, location, and affected areas, but also provides information on potential related faults and chain reactions. This comprehensive and accurate report can be quickly transmitted to operations and maintenance personnel or upper-level systems to provide decision support and guide fault isolation, repair, and restoration efforts, thereby shortening outages and improving power supply reliability. This entire process, through collaboration between edge nodes and intelligent analysis at the convergence node, enables rapid, accurate, and comprehensive diagnosis of complex cross-regional faults.

[0075] By employing a Kalman filter algorithm to precisely align and fuse multi-source data, this solution effectively overcomes temporal bias and measurement noise, improving input data quality. Cross-regional correlation analysis using a Bayesian network based on the aligned data accurately captures the complex probabilistic relationships between device states in different regions, improving the accuracy and robustness of diagnosis regarding fault type, location, and affected regions. Combined with a fault knowledge base query, it provides information on potential associated faults and chain reactions, making fault diagnosis reports more comprehensive and providing strong support for fault resolution and risk assessment.

[0076] Preferably, step A502 may include: Constructing a Bayesian network, wherein the nodes of the Bayesian network include fault type, fault location, fault-affected area, voltage amplitude, current amplitude, and switch status, and the edges of the Bayesian network represent dependencies between nodes, wherein the dependencies are determined based on the topology of the power distribution system and historical fault data; Based on the aligned node data stream, the conditional probability of each node state in the Bayesian network is calculated; The maximum a posteriori probability estimation method is adopted to infer the fault type, fault location and fault-involved area with the maximum a posteriori probability according to the conditional probability.

[0077] The nodes of a Bayesian network are random variables within the network, specifically key information and observational data relevant to distribution system fault diagnosis. These include the attributes of the fault to be diagnosed (fault type, location, and affected area), as well as electrical quantities and switch status data collected from various locations in the system (voltage amplitude, current amplitude, and switch status). The edges of a Bayesian network represent dependencies between nodes, referring to the impact of a change in the state of one node on the probability distribution of the state of another. These dependencies are determined based on the actual physical connections between equipment and lines in the distribution system, as well as data patterns and associations learned from past fault events.

[0078] Aligned node data streams refer to time-synchronized sequences of system status data from different edge nodes that can be used for unified analysis. Calculating the conditional probability of each node state in a Bayesian network involves using the Bayesian network's inference algorithm to calculate the probabilities of various possible states for other unknown nodes (fault information nodes), given the specific values (evidence) of some nodes (observation data nodes). The maximum a posteriori probability estimation method is a statistical inference method that, after calculating the conditional probabilities of various possible states for an unknown variable (fault type, location, affected area), selects the state with the highest probability as the final estimate.

[0079] The proposed solution integrates fault information and observation data into a unified probabilistic model by constructing a Bayesian network that reflects the actual characteristics and historical experience of the distribution system. First, the structure and parameters of the Bayesian network are determined based on the distribution system's topology and historical fault data. This enables the network to accurately capture the fault propagation path within the system and the typical data characteristics that different fault types may exhibit at different locations, thereby establishing a complex probabilistic association between the fault and the observation data. Next, the time-aligned multi-source observation data is used as evidence input into the Bayesian network. Leveraging the probabilistic reasoning capabilities of the Bayesian network, the conditional probability distribution of various possible fault types, fault locations, and fault-affected regions is calculated based on these observation data. This process converts actual system operating state information into probabilistic information and quantifies the likelihood of different fault hypotheses. Finally, a maximum a posteriori probability estimation method is used to select the fault state with the highest probability from the calculated conditional probability distribution as the final diagnosis result. This probabilistic model-based reasoning process fully utilizes the aligned cross-regional multi-source data and can select the most likely fault explanation in data with uncertainty or noise, thereby improving the accuracy and reliability of the diagnostic results. In this way, the solution of the present application can effectively extract key fault information from multi-source data across regions, overcome the limitations of simple rules or local analysis, and achieve accurate inference of fault type, fault location and fault-involved area.

[0080] By constructing a Bayesian network that reflects the characteristics of the distribution system and historical experience, the solution of this application can accurately capture the complex probabilistic associations between faults and cross-regional multi-source observation data. Probabilistic reasoning based on aligned node data streams can quantify the likelihood of different fault hypotheses under the current system state. The maximum a posteriori probability estimation method is used to select the most likely fault explanation in data with uncertainty or noise. The combination of these technical means improves the accuracy and reliability of fault diagnosis and can accurately infer the fault type, fault location, and fault-affected area.

[0081] refer to Figure 2 、 Figure 3 The present application provides a power distribution system fault diagnosis system, the system comprising a plurality of edge nodes, the plurality of edge nodes comprising general nodes and aggregation nodes; the edge nodes being deployed in different areas of the power distribution system; the edge nodes being provided with a fault diagnosis program, the fault diagnosis program of the general nodes being configured with an abnormality identification module 1, an abnormal event report generation module 2, a collaborative node determination module 3, a data snapshot request module 4, and a data snapshot response module 5, and the fault diagnosis program of the aggregation node being configured with an abnormality identification module 1, an abnormal event report generation module 2, a collaborative node determination module 3, a data snapshot request module 4, a data snapshot response module 5, and a correlation analysis module 6; The abnormality identification module 1 is used to extract the electrical quantity and switch status data collected by the edge node to identify fault-related abnormal signals (for the specific process, refer to step A1 above); The abnormal event report generating module 2 is used to generate an abnormal event report containing the abnormality type, abnormality occurrence time and abnormality location when a fault-related abnormal signal is identified (for the specific process, refer to step A2 above); The collaborative node determination module 3 is configured to determine a collaborative node group associated with the abnormal event based on the abnormal event report and the topology of the power distribution system; the topology of the power distribution system includes the connection relationship between each power distribution device and the connection relationship between each edge node and each power distribution device (for the specific process, refer to step A3 above); The data snapshot request module 4 is used to send a data snapshot request containing a corresponding data snapshot time range to the coordination node group according to the abnormality type and abnormality occurrence time in the abnormal event report (for the specific process, refer to step A4 above); The data snapshot response module 5 is configured to respond to the data snapshot request upon receiving the data snapshot request, collect a data snapshot of the local data within the data snapshot time range, attach a high-precision local timestamp to the data snapshot, and then send the data snapshot to the aggregation node for cross-regional data association analysis (for the specific process, refer to step A4 above); The association analysis module 6 is used to align the node data stream according to the high-precision local timestamp attached to the data snapshot when receiving the data snapshot, perform cross-regional data association analysis, and generate a fault diagnosis report; the fault diagnosis report includes at least one item from the list of fault type, fault location, fault-related area, associated faults, and chain events (for the specific process, refer to step A5 above).

[0082] An edge node is a device unit deployed on-site in a power distribution system that possesses certain computing, storage, and communication capabilities. It can be implemented using industrial computers, embedded systems, or smart terminals. General nodes and aggregation nodes are types of edge nodes, categorized by their roles in the fault diagnosis process. Aggregation nodes typically possess greater computing and storage capabilities and are used to receive and process data from multiple general nodes. A fault diagnosis program is a software application running on an edge node. It can be implemented using a modular software architecture and comprise multiple functional modules.

[0083] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A power distribution system fault diagnosis method, applied to an edge node of a power distribution system to perform fault diagnosis, characterized in that: The steps of the method include: A1. Extract electrical quantity and switch status data collected by this edge node to identify fault-related abnormal signals; A2. When a fault-related abnormal signal is identified, an abnormal event report is generated containing the abnormality type, abnormality occurrence time, and abnormality location; A3. According to the abnormal event report and the distribution system topology, determine the collaborative node group associated with the abnormal event; the distribution system topology includes the connection relationship between the distribution equipment and the connection relationship between each edge node and each distribution equipment; A4. Based on the anomaly type and occurrence time in the anomaly event report, a data snapshot request containing the corresponding data snapshot time range is sent to the collaborative node group. This enables the collaborative node group to collect a data snapshot of local data within the data snapshot time range, append a high-precision local timestamp to the data snapshot, and then send it to the aggregation node for cross-regional data correlation analysis. The aggregation node is the local edge node or another pre-defined edge node. A5. If a data snapshot is received from a collaborative node group, the node data streams are aligned based on the high-precision local timestamp attached to the data snapshot, cross-regional data correlation analysis is performed, and a fault diagnosis report is generated; the fault diagnosis report includes at least one item from the list of fault type, fault location, fault-affected area, associated faults, and chain events.

2. A power distribution system fault diagnosis method according to claim 1, characterized in that: Step A1 includes: A101. Extract electrical quantities and switch status data collected by this edge node; the electrical quantities include voltage, current, and frequency; the switch status data includes status information of circuit breakers and disconnectors; A102. Using a wavelet threshold denoising method to filter out high-frequency noise interference in the electrical quantity, thereby obtaining a denoised electrical quantity; A103. Based on the denoised electrical quantity and the switch state data, a characteristic signal is obtained; the characteristic signal includes the electrical quantity mutation amount, the number of switch state jumps, and the duration of the electrical quantity exceeding the limit; A104. If at least one of the characteristic signals exceeds the corresponding preset threshold range, it is determined that a fault-related abnormal signal is identified, and the corresponding characteristic signal is determined to be an abnormal signal.

3. A power distribution system fault diagnosis method according to claim 2, characterized in that: Step A2 includes: A201. Determine the abnormality type according to the abnormal signal; the abnormality types include mutation abnormality, switch jump abnormality and over-limit abnormality; A202. Record the moment when the fault-related abnormal signal is identified as the abnormality occurrence time; A203. Based on the topology of the power distribution system, determine the power distribution equipment where the abnormal signal is detected as the abnormal location; A204. Encapsulate the determined abnormality type, the abnormality occurrence time, and the abnormality location into an abnormal event report.

4. A power distribution system fault diagnosis method according to claim 3, characterized in that: Step A3 includes: A301. According to the abnormal location and distribution system topology in the abnormal event report, obtain a list of distribution equipment directly connected to the abnormal location, and use the distribution equipment in the list of distribution equipment as a primary associated distribution equipment; A302. For each level-associated distribution device, according to the distribution system topology, obtain the other distribution devices directly connected to the level-associated distribution device, remove the distribution devices already in the level-associated distribution device list, and use the remaining distribution devices as secondary-associated distribution devices; A303. The edge nodes connected by the primary associated power distribution equipment and the secondary associated power distribution equipment constitute the initial collaborative node group; A304. Evaluate the electrical distance between each edge node in the initial collaborative node group and the abnormal location; the electrical distance between the edge node and the abnormal location is: the number of distribution devices passed from the edge node to the abnormal location; A305. Eliminate edge nodes whose electrical distance is greater than a preset distance threshold from the initial collaborative node group to obtain a final collaborative node group.

5. A power distribution system fault diagnosis method according to claim 4, characterized in that: Step A305 includes: Determining a distance correction coefficient for each edge node in the initial collaborative node group based on the device type and historical failure rate of the power distribution equipment connected to each edge node; Correcting the electrical distance corresponding to each edge node in the initial collaborative node group using the distance correction coefficient; The edge nodes whose corrected electrical distances are greater than a preset distance threshold are eliminated from the initial collaborative node group to obtain the final collaborative node group.

6. A power distribution system fault diagnosis method according to claim 4, characterized in that: After step A305, the following steps are also included: A306. Based on the abnormal location in the abnormal event report and the current operating mode of the power distribution system, a preset regional interconnection information database is queried to identify other areas that are electrically connected to the area where the abnormal location is located. Edge nodes within these areas that are not part of the final coordinated node group are selected as candidate supplementary edge nodes. The regional interconnection information database records the electrical connection relationships between various areas in the power distribution system under different operating modes. A307. Evaluate the electrical accessibility between each candidate supplementary edge node and the abnormal location based on the distribution system topology. Electrical accessibility indicates whether there is an effective electrical path between the candidate supplementary edge node and the abnormal location, or whether there is a mutual influence or correlation between the electrical state changes between the candidate supplementary edge node and the abnormal location. A308. Add the candidate supplementary edge nodes whose electrical reachability is reachable to the final cooperative node group.

7. A power distribution system fault diagnosis method according to claim 1, characterized in that: Step A4 includes: A401. Determine the data snapshot time range based on the abnormality type and abnormality occurrence time reported in the abnormal event report; A402. If this edge node is one of the preset sink nodes, then the edge node is set as the target sink node; if this edge node is not one of the preset sink nodes, then based on the network topology distance between the abnormal location and each preset sink node, select the preset sink node with the smallest network topology distance as the target sink node; A403. Add the data snapshot time range and the ID of the target aggregation node to the data snapshot request and send it to the collaborative node group, so that the collaborative node group collects the data snapshot of the local data within the data snapshot time range, and attaches a high-precision local timestamp to the data snapshot, and then sends it to the target aggregation node for cross-regional data association analysis.

8. A power distribution system fault diagnosis method according to claim 7, characterized in that: Step A5 includes: A501. Use a Kalman filter algorithm to fuse data snapshots from the collaborative node group and align the node data streams. The data snapshots include voltage, current, and switch status data. The state variables of the Kalman filter algorithm include voltage amplitude, current amplitude, and switch status. The observed variables are the sampled values of voltage, current, and switch status reported by each collaborative node. A502. Based on the aligned node data streams, a Bayesian network is used to perform cross-regional data association analysis to determine the fault type, fault location, and affected area. The nodes of the Bayesian network include fault type, fault location, affected area, voltage amplitude, current amplitude, and switch status. The edges of the Bayesian network represent dependencies between nodes, which are determined based on the distribution system topology and historical fault data. A503. Based on the fault type and fault location determined, query the preset fault knowledge base to obtain a list of associated faults and chain events related to the fault type and fault location; the fault knowledge base records the correspondence between various fault types, fault locations, associated faults and chain events in the power distribution system; A504. Encapsulate the determined fault type, fault location, fault-involved area, and the obtained list of associated faults and chain events into a fault diagnosis report.

9. A power distribution system fault diagnosis method according to claim 8, characterized in that: Step A502 includes: Constructing a Bayesian network, wherein the nodes of the Bayesian network include fault type, fault location, fault-affected area, voltage amplitude, current amplitude, and switch status, and the edges of the Bayesian network represent dependencies between nodes, wherein the dependencies are determined based on the topology of the power distribution system and historical fault data; Based on the aligned node data stream, the conditional probability of each node state in the Bayesian network is calculated; The maximum a posteriori probability estimation method is adopted to infer the fault type, fault location and fault-involved area with the maximum a posteriori probability according to the conditional probability.

10. A power distribution system fault diagnosis system, characterized in that: The system includes multiple edge nodes, which include general nodes and aggregation nodes; each edge node is deployed in a different area of the power distribution system; the edge nodes are provided with a fault diagnosis program, the fault diagnosis program of the general node is configured with an abnormality identification module, an abnormal event report generation module, a collaborative node determination module, a data snapshot request module and a data snapshot response module, and the fault diagnosis program of the aggregation node is configured with an abnormality identification module, an abnormal event report generation module, a collaborative node determination module, a data snapshot request module, a data snapshot response module and a correlation analysis module; The anomaly identification module is used to extract the electrical quantity and switch status data collected by the edge node to identify fault-related abnormal signals; The abnormal event report generation module is used to generate an abnormal event report containing the abnormal type, abnormal occurrence time and abnormal location when a fault-related abnormal signal is identified; The collaborative node determination module is used to determine the collaborative node group associated with the abnormal event based on the abnormal event report and the topological relationship of the power distribution system; the topological relationship of the power distribution system includes the connection relationship between each distribution device and the connection relationship between each edge node and each distribution device; The data snapshot request module is used to send a data snapshot request containing a corresponding data snapshot time range to the collaborative node group according to the abnormality type and abnormality occurrence time in the abnormal event report; The data snapshot response module is used to respond to the data snapshot request upon receiving the data snapshot request, collect a data snapshot of local data within the data snapshot time range, attach a high-precision local timestamp to the data snapshot, and send it to the aggregation node for cross-region data association analysis; The association analysis module is used to align the node data stream according to the high-precision local timestamp attached to the data snapshot when receiving the data snapshot, perform cross-regional data association analysis, and generate a fault diagnosis report; the fault diagnosis report includes at least one item from the fault type, fault location, fault-related area, associated faults and chain event list.

Images