AI-based power distribution network distributed fault processing method and system

Abstract

The application discloses an AI-based power distribution network distributed fault processing method and system, relates to the power distribution operation and maintenance technical field, and comprises the following steps: dividing a power distribution network into multiple fault processing partitions, collecting electrical quantity operation data of nodes corresponding to the partitions by each monitoring node, performing node-level fault analysis and processing based on the collected electrical quantity data to generate node fault feature data, gathering node fault feature data of all nodes in each partition, integrating and analyzing the fault features of each node based on the topological connection relationship between nodes in the partition to generate partition fault integration results of each partition, and performing boundary consistency verification and fusion on the partition fault integration results of each partition according to the topological connection relationship between the partitions to generate a global fault processing result of the power distribution network. The application solves the technical problems of low fault positioning accuracy and low processing efficiency of the power distribution network in the prior art, and achieves the technical effects of improving the fault positioning accuracy and the fault processing efficiency of the power distribution network.

Description

Technical Field

[0001] This invention relates to the field of power distribution operation and maintenance technology, specifically to an AI-based method and system for handling distributed faults in power distribution networks. Background Technology

[0002] As a crucial link in the power system supplying power to the user side, the distribution network has numerous line branches, a large number of nodes, and a complex operating topology. After a fault occurs, changes in electrical quantities propagate along the lines and adjacent nodes, causing fault characteristics to exhibit correlations across multiple monitoring points. Traditional fault handling methods typically rely on single-point monitoring data or centralized fault analysis, making it difficult to timely combine the distribution network topology to collaboratively determine the fault characteristics of each node. Especially in scenarios with multiple zones, multiple nodes, and boundary correlations, inaccurate fault segment identification and time-consuming analysis processes can easily occur, affecting the accuracy and efficiency of distribution network fault handling. Summary of the Invention

[0003] This application provides an AI-based method and system for handling distributed faults in power distribution networks, which addresses the technical problems of low accuracy and low efficiency in fault location and handling in existing technologies.

[0004] In view of the above problems, this application provides an AI-based method and system for handling distributed faults in power distribution networks.

[0005] The first aspect of this application provides an AI-based distributed fault handling method for power distribution networks, the method comprising: Based on the operating topology of the distribution network, the distribution network is divided into multiple fault handling zones, and the node range covered by each zone and the topological connection relationship between nodes are determined. Based on multiple fault handling zones, each monitoring node collects electrical quantity operation data of the corresponding nodes in its zone, and performs node-level fault analysis and processing based on the collected electrical quantity data to generate node fault feature data. Taking each zone as a unit, the node fault feature data of all nodes in the zone are aggregated, and the fault features of each node are integrated and analyzed based on the topological connection relationship between nodes in the zone to generate the zone fault integration result of each zone. The zone fault integration results of each zone are then checked for boundary consistency and fused according to the topological connection relationship between the zones to generate the global fault handling result of the distribution network.

[0006] A second aspect of this application provides an AI-based distributed fault handling system for power distribution networks, the system comprising: The distribution network is divided into several modules: a partitioning module and a data acquisition module. The former is responsible for dividing the distribution network into multiple fault handling zones based on its operational topology, determining the node coverage of each zone and the topological connections between nodes. The latter is responsible for collecting electrical quantity operation data from the corresponding nodes within each zone, performing node-level fault analysis based on the collected data, and generating node fault characteristic data. The former is responsible for aggregating node fault characteristic data from all nodes within each zone, performing integrated analysis on the fault characteristics of each node based on the topological connections between nodes within the zone, and generating integrated fault results for each zone. The latter is responsible for verifying and fusing the integrated fault results of each zone according to the topological connections between zones, generating a global fault handling result for the distribution network.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application, based on the operational topology of a distribution network, divides the distribution network into multiple fault handling zones, determining the node range covered by each zone and the topological connections between nodes. Based on these multiple fault handling zones, each monitoring node collects electrical quantity operation data of the corresponding nodes within its zone. Node-level fault analysis is performed based on the collected electrical quantity data to generate node fault feature data. Taking each zone as a unit, the node fault feature data of all nodes within that zone are aggregated. The fault features of each node are integrated and analyzed based on the topological connections between nodes within the zone, generating a zone-wide fault integration result. The zone-wide fault integration results are then checked for boundary consistency and fused according to the topological connections between zones to generate a global fault handling result for the distribution network. This invention solves the technical problems of low fault location accuracy and low processing efficiency in existing distribution network technologies. By performing zoned node fault analysis and fusion of zoned fault results based on the operational topology, it achieves the technical effect of improving the accuracy of fault location and the efficiency of fault handling in distribution networks. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A schematic diagram of the AI-based distributed fault handling method for power distribution networks provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of an AI-based distributed fault handling system for power distribution networks, provided in an embodiment of this application.

[0010] Figure labeling: Module 11, Acquisition Module 12, Integration and Analysis Module 13, Verification and Fusion Module 14. Detailed Implementation

[0011] This application provides an AI-based distributed fault handling method and system for distribution networks, addressing the technical problems of low fault location accuracy and low processing efficiency in existing technologies. By performing partitioned node fault analysis based on the operating topology and fusing the partitioned fault results, it achieves the technical effect of improving the accuracy of fault location and the efficiency of fault handling in distribution networks.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0013] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0014] Example 1, as Figure 1 As shown, this application provides an AI-based distributed fault handling method for power distribution networks, the method comprising: Step S100: Based on the operating topology of the distribution network, the distribution network is divided into multiple fault handling zones, and the node range covered by each zone and the topological connection relationship between the nodes are determined.

[0015] In this embodiment, based on the physical connection relationships between feeders, buses, switchgear, branch lines, monitoring nodes, and load nodes in the distribution network, and combined with the switch open / closed status, tie switch status, and power supply path status, an operational topology representing the current operational connection relationships of the distribution network is generated. Based on the operational topology, and considering the electrical propagation characteristics and impact range corresponding to different fault types, an operational sensing parameter set is determined to represent the fault-dependent response relationship. Then, according to the response mapping relationship between fault types and the operational sensing parameter set, the distribution network topology is partitioned into fault handling partitions, resulting in multiple fault handling partitions. Subsequently, based on the topology node range covered by each fault handling partition and the topology connection relationships between nodes, a partitioned topology engine is established to identify the fault topology connection relationships within and between each fault handling partition during the fault handling process.

[0016] Furthermore, the method provided in the application embodiments, which divides the distribution network into multiple fault handling zones and determines the node range covered by each zone and the topological connection relationship between nodes, also includes: Based on the electrical propagation characteristics and impact range of different fault types in the distribution network, a set of operation sensing parameters for characterizing fault-dependent responses is determined. According to the response mapping relationship between fault type and operation sensing parameter set, the distribution network topology is partitioned to obtain multiple fault handling partitions. Based on the multiple fault handling partitions, the range of topology nodes covered, and the topology connection relationships between nodes, a partitioned topology engine is established for the fault handling process to call and identify the topology connection relationships of each fault.

[0017] In this embodiment, when determining the set of operational sensing parameters to characterize fault-dependent response based on the electrical propagation characteristics and impact range of different fault types in the distribution network, the electrical quantity data of the distribution network under normal operation and fault operation states are first acquired, and the data under fault operation states are classified according to fault type. Then, for each fault type, the three-phase voltage difference, three-phase current difference, zero-sequence voltage difference, active power difference, and reactive power difference of each monitoring node before and after the fault occurs are calculated, and each difference is compared with the corresponding preset parameter threshold. Among them, electrical propagation characteristics refer to the laws of numerical and directional changes of electrical quantities such as voltage, current, and power along the distribution network lines and nodes after the fault occurs, and impact range refers to the range of monitoring nodes where the electrical quantity difference exceeds the corresponding preset parameter threshold. Then, based on the topological position of each monitoring node in the fault node, upstream node, downstream node, and adjacent nodes, the parameters whose electrical quantity difference exceeds the corresponding preset parameter threshold are retained, and the retained parameters are summarized according to the fault type and the corresponding monitoring node range to obtain the set of operational sensing parameters to characterize fault-dependent response.

[0018] Next, the distribution network topology is partitioned according to the response mapping relationship between fault types and operational sensing parameter sets. In this process, based on the operational sensing parameter sets corresponding to different fault types, the range of monitoring nodes relied upon by each type of fault in the fault determination process is determined. Then, according to the correspondence between the monitoring node range and the distribution network topology, the response mapping relationship is hierarchically divided. Specifically, when fault determination relies only on the operational sensing parameters of a single monitoring node, the corresponding response mapping relationship is divided into node-level response mapping; when fault determination relies on the combination of operational sensing parameters of multiple monitoring nodes within the same topological region, the corresponding response mapping relationship is divided into partition-level response mapping; when fault determination relies on the combination of operational sensing parameters between monitoring nodes at the boundary of adjacent topological regions, the corresponding response mapping relationship is divided into cross-partition interactive response mapping. Based on this, distribution network fault handling partitions matching the node-level response mapping, partition-level response mapping, and cross-partition interactive response mapping are formed.

[0019] Finally, based on the multiple fault handling zones, the covered topology node range, and the topology connection relationships between nodes, a zone topology engine is established. Specifically, the starting node, ending node, and boundary node corresponding to each fault handling zone are first determined according to the operating topology of the distribution network. Then, along the feeder connection path between the starting node and the ending node, the monitoring nodes, switch nodes, feeder nodes, and load nodes located on this connection path are extracted as the topology node range covered by the fault handling zone. For fault handling zones with branch lines, the monitoring nodes, switch nodes, and load nodes on the branch lines are further extracted based on the connection nodes between the branch lines and the main feeder, and then incorporated into the topology node range of the corresponding fault handling zone. Subsequently, the nodes contained in each fault handling partition are identified according to their respective fault handling partitions, and the node number, node type, partition to which each node belongs, and adjacent connected nodes are recorded. Then, based on the actual connection relationship between nodes, the node connection order in each fault handling partition is formed, and the boundary nodes and boundary connection relationships between adjacent fault handling partitions are determined. Finally, the topology node range, topology connection relationship between nodes, topology connection relationship between sub-partitions, and boundary node correspondence of each fault handling partition are associated and stored to form a partition topology engine. During the fault handling process, the partition topology engine calls the corresponding fault topology connection relationship according to the fault node, fault segment, or partition identifier.

[0020] Furthermore, in the method provided in the application embodiments, partitioning the distribution network topology according to the response mapping relationship between fault type and operational sensing parameter set further includes: Based on the response mapping relationship between fault type and operational perception parameter set, the range of monitoring nodes required for each type of fault is determined; according to the range of monitoring nodes required for each type of fault, the response mapping relationship is divided into node-level response mapping, partition-level response mapping, and cross-partition interactive response mapping.

[0021] Furthermore, the method provided in the application embodiments also includes: The node-level response mapping establishes a mapping relationship between the operational perception parameters of a single monitoring node and whether the node has failed; the partition-level response mapping establishes a mapping relationship between the combination of operational perception parameters of all monitoring nodes within a partition and whether a faulty section exists within the corresponding partition; the cross-partition interactive response mapping establishes a mapping relationship between the combination of operational perception parameters of boundary monitoring nodes of adjacent partitions and whether a cross-partition fault exists at the corresponding boundary.

[0022] In this embodiment, when determining the range of monitoring nodes required for various faults based on the response mapping relationship between fault types and operation sensing parameter sets, the operation sensing parameter sets corresponding to different fault types are first obtained. The operation sensing parameter sets include the three-phase voltage, three-phase current, zero-sequence voltage, active power, reactive power, and fault direction corresponding to each monitoring node. Then, the operation sensing parameters corresponding to each fault type are determined, and the fault type, operation sensing parameters, monitoring nodes to which the operation sensing parameters belong, fault processing zones to which the monitoring nodes belong, and connection positions of the monitoring nodes in the distribution network topology are recorded to form a response mapping relationship between fault types and operation sensing parameter sets. The response mapping relationship represents the correspondence between different fault types and the operation sensing parameter sets participating in the response analysis of that fault type.

[0023] In the response mapping relationship, the corresponding operation perception parameters are extracted by indexing the fault type. Based on the node number of the monitoring node to which the operation perception parameter belongs, the monitoring nodes participating in the response analysis of the fault type are aggregated. After removing duplicate node numbers, the range of monitoring nodes required by this type of fault is obtained. The range of monitoring nodes represents the set of monitoring nodes that are associated with the operation perception parameters of the corresponding fault type.

[0024] Based on the range of monitoring nodes required by various faults, the response mapping relationship is divided into node-level response mapping, zone-level response mapping, and cross-zone interactive response mapping. The range of monitoring nodes required by various faults is matched with the node affiliation relationship in the distribution network topology, and the classification type of response mapping relationship is determined according to the distribution location of the monitoring node range in the distribution network topology.

[0025] When the monitoring node range corresponds to a single monitoring node, the corresponding response mapping relationship is divided into node-level response mapping, which establishes a mapping relationship between the operational perception parameters of a single monitoring node and whether the node has experienced a fault. When the monitoring node range corresponds to all monitoring nodes within the same fault handling partition, the corresponding response mapping relationship is divided into partition-level response mapping, which establishes a mapping relationship between the combination of operational perception parameters of all monitoring nodes within the partition and whether a faulty segment exists within the corresponding partition. When the monitoring node range corresponds to boundary monitoring nodes of adjacent partitions, the corresponding response mapping relationship is divided into cross-partition interactive response mapping, which establishes a mapping relationship between the combination of operational perception parameters of boundary monitoring nodes of adjacent partitions and whether a cross-partition fault exists at the corresponding boundary.

[0026] Step S200: Based on multiple fault handling zones, each monitoring node collects electrical quantity operation data of the corresponding node in its zone, performs node-level fault analysis and processing based on the collected electrical quantity data, and generates node fault characteristic data.

[0027] In this embodiment, based on multiple fault handling partitions, each monitoring node collects electrical quantity operation data of the corresponding node according to its fault handling partition, and obtains the three-phase voltage, three-phase current, zero-sequence voltage, active power and reactive power of the node at a set sampling frequency. Subsequently, each monitoring node preprocesses the collected raw electrical quantity data to form a standardized input data sequence, and performs node-level fault analysis processing based on the standardized input data sequence. The sensing parameters in the electrical quantity operation data are compared with preset fault judgment parameter thresholds, and the presence of fault characteristics of the node is determined according to the comparison results. When the node has fault characteristics, node fault characteristic data containing fault type, fault direction and fault probability is generated, and the node fault characteristic data is uploaded to the edge computing node corresponding to the partition.

[0028] Furthermore, in the method provided in the application embodiments, based on multiple fault handling partitions, each monitoring node collects electrical quantity operation data of the corresponding node in its partition, performs node-level fault analysis and processing based on the collected electrical quantity data, and generates node fault feature data, further comprising: Each monitoring node collects its own electrical quantity operation data at a set sampling frequency. The electrical quantity operation data includes three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power. Each monitoring node preprocesses the collected raw electrical quantity data to form a standardized input data sequence. The standardized input data sequence is then subjected to node-level fault analysis. The node-level fault analysis includes comparing each sensing parameter in the collected electrical quantity operation data with preset fault judgment parameter thresholds. Based on the comparison results, it is determined whether the node has fault characteristics. If fault characteristics are found, node fault characteristic data containing fault type, fault direction, and fault probability is generated. Each monitoring node uploads the node fault characteristic data to the edge computing node corresponding to its partition.

[0029] In this embodiment, when each monitoring node collects its electrical quantity operation data at a set sampling frequency, it first determines the fault handling zone to which it belongs based on the fault handling zone results, and records the node number, zone identifier, and sampling time of the monitoring node in relation to the collected data. Subsequently, each monitoring node continuously collects its three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power at the set sampling frequency, forming electrical quantity operation data arranged in chronological order of sampling time. The set sampling frequency is a fixed frequency at which each monitoring node collects electrical quantity operation data, and the electrical quantity operation data are the three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power reflecting the operating status of the node.

[0030] Next, each monitoring node preprocesses the collected electrical quantity operation data. First, it organizes the three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power according to the sampling time, ensuring that the electrical quantity operation data under the same sampling time are grouped into the same data record. Then, it processes missing data items, duplicate data items, and data items exceeding the range of the acquisition equipment. Missing data items are filled in using data from adjacent sampling times, duplicate data items are retained only once, and data items exceeding the range of the acquisition equipment are not included in subsequent analysis. Finally, the three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power are processed according to their corresponding rated values ​​and arranged in order of sampling time to form a standardized input data sequence. The standardized input data sequence includes the node number, the zone identifier, the sampling time, and the preprocessed three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power.

[0031] Subsequently, node-level fault analysis is performed on the standardized input data sequence. Each monitoring node reads the preset fault judgment parameter thresholds corresponding to its own node. The preset fault judgment parameter thresholds include thresholds corresponding to three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power. Then, the three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power in the standardized input data sequence are used as sensing parameters, and each sensing parameter is compared with its corresponding preset fault judgment parameter threshold to obtain the comparison result of each sensing parameter. The sensing parameters are the three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power participating in the node-level fault analysis process. The node-level fault analysis process is the process by which each monitoring node identifies fault features based on its own standardized input data sequence.

[0032] When determining whether a node exhibits fault characteristics based on the comparison results, if at least one sensing parameter reaches the corresponding preset fault determination parameter threshold, then the node is determined to have fault characteristics; if none of the sensing parameters reach the corresponding preset fault determination parameter threshold, then the node is determined not to have fault characteristics. If the node exhibits fault characteristics, the fault type is determined based on the category of sensing parameters that reach the preset fault determination parameter threshold. Specifically, a ground fault type is generated when the zero-sequence voltage reaches the corresponding threshold, a short-circuit fault type is generated when the three-phase current reaches the corresponding threshold, and a voltage-current composite fault type is generated when both the three-phase voltage and three-phase current reach the corresponding threshold.

[0033] When generating fault direction, each monitoring node determines its fault direction based on the direction information corresponding to the three-phase current, active power, and reactive power. If the direction information of the three-phase current, active power, or reactive power is consistent with the preset positive direction of the node, the fault direction is marked as positive; if the direction information of the three-phase current, active power, or reactive power is opposite to the preset positive direction of the node, the fault direction is marked as negative. When generating fault probability, each monitoring node counts the number of sensing parameters that reach the preset fault judgment parameter threshold and the total number of sensing parameters participating in the comparison. The ratio of the number of sensing parameters reaching the preset fault judgment parameter threshold to the total number of sensing parameters participating in the comparison is used as the fault probability. Subsequently, node fault feature data containing fault type, fault direction, and fault probability is generated, and the node fault feature data, along with the node number, the partition identifier, and the sampling time, is uploaded to the edge computing node corresponding to the partition.

[0034] Step S300: For each partition, aggregate the node fault characteristic data of all nodes within the partition, integrate and analyze the fault characteristics of each node based on the topological connection relationship between nodes within the partition, and generate the partition fault integration result for each partition.

[0035] In this embodiment, when aggregating node fault feature data of all nodes within a partition, each partition's edge computing node receives node fault feature data uploaded by all monitoring nodes within that partition. Based on the partition's topological connection, the node fault feature data of each node is arranged in topological order to construct a fault feature chain with nodes as the sequence. Subsequently, the edge computing node traverses the fault direction information of each node along the fault feature chain and identifies fault segments based on the combination relationship of fault directions between adjacent nodes. When the fault directions of two adjacent nodes form a combination of opposing directions, the fault is determined to be located in the segment between the two nodes. When the fault direction of a node does not form an opposing combination with the fault directions of the preceding and following nodes, and the node itself detects a fault feature, the fault is determined to be located at that node. Finally, the edge computing node summarizes all identified fault segments and fault nodes to generate a partition fault integration result for that partition. The partition fault integration result includes the location of fault segments and fault type within the partition.

[0036] Furthermore, in the method provided in the application embodiments, taking each partition as a unit, the node fault characteristic data of all nodes within the partition are aggregated, and the fault characteristics of each node are integrated and analyzed based on the topological connection relationship between nodes within the partition to generate the partition fault integration result of each partition, which also includes: Each edge computing node in a partition receives node fault feature data uploaded by all monitoring nodes within that partition. Following the partition's topological connections, it arranges the fault feature data of each node in topological order, constructing a fault feature chain with nodes as the sequence. The edge computing node traverses the fault direction information of each node along the fault feature chain, identifying fault segments based on the combination relationship of fault directions between adjacent nodes. If the fault directions of two adjacent nodes form a combination of opposing directions, the fault is determined to be located in the segment between the two nodes. If the fault direction of a node does not form an opposing combination with the fault directions of the preceding and following nodes, and the node itself detects a fault feature, the fault is determined to be located at that node. The edge computing node summarizes all identified fault segments and fault nodes to generate a partition fault integration result for that partition. The partition fault integration result includes the location and type of fault segments within the partition.

[0037] In this embodiment, after receiving node fault feature data uploaded by all monitoring nodes within its partition, the edge computing nodes of each partition match the node fault feature data of each node with the corresponding node within its partition based on the partition identifier and node number in the node fault feature data, thereby obtaining the fault type, fault direction, and fault probability corresponding to each node within its partition. The edge computing nodes read the topology connection relationship of their partition, determine the upstream and downstream connection relationship, adjacent connection relationship, and branch connection relationship between each node within its partition, and arrange the fault feature data of each node along the feeder connection direction, using the starting node within its partition as the starting point. Within this partition, there are... For branch lines, the edge computing nodes connect the nodes on the branch lines to their corresponding positions according to their connection with the main line nodes, so that the fault feature data of each node is consistent with the topological connection relationship between nodes in this partition. After completing the arrangement of the fault feature data of each node, the edge computing nodes write the node number, fault type, fault direction, fault probability and the connection relationship of adjacent nodes of each node into the fault feature chain in topological order, and construct the fault feature chain with nodes as the sequence. The topological order is the connection order of each node in the topological connection relationship of this partition, and the fault feature chain is the chain arrangement result formed according to the topological order of the nodes in this partition.

[0038] Next, the edge computing node traverses the fault direction information of each node along the fault feature chain, and reads the fault direction information of the current node and its adjacent nodes in sequence according to the node arrangement in the fault feature chain. The fault directions between adjacent nodes are combined and compared. If the fault directions of two adjacent nodes point to the line segment between the two adjacent nodes, the edge computing node determines that the fault directions of the two adjacent nodes constitute a combination feature with opposite directions, and determines that the fault is located in the segment between the two nodes. At the same time, it records the node numbers at both ends of the fault segment, the fault processing partition to which it belongs, and the corresponding fault type. The fault direction information is the information in the node fault feature data that indicates the direction of the fault. The combination feature with opposite directions is the combination relationship between the fault directions of two adjacent nodes pointing to the segment between the two nodes.

[0039] The edge computing node continues to read the fault direction information of the current node, the previous neighboring node, and the next neighboring node along the fault feature chain, and compares the fault direction combination relationship between the current node and the previous neighboring node, and between the current node and the next neighboring node. If the fault direction of the current node does not form a direction-facing combination with the fault direction of the previous neighboring node, and the fault direction of the current node does not form a direction-facing combination with the fault direction of the next neighboring node, and the current node itself detects the fault feature, the edge computing node determines that the fault is located at this node, and records the node number, the fault processing partition to which it belongs, and the corresponding fault type.

[0040] After the edge computing node completes the fault feature chain traversal, it summarizes all the identified fault segments and fault nodes, and associates the node numbers of the two ends of the fault segment, the node number of the fault node, the fault type, and the fault processing partition to which it belongs, to generate the partition fault integration result of this partition; the partition fault integration result includes the location of the fault segment and the fault type within the partition.

[0041] Step S400: The partition fault integration results of each partition are checked and merged according to the topological connection relationship between the partitions to generate the global fault handling results of the distribution network.

[0042] In this embodiment, when the partition fault integration results of each partition are checked and fused according to the topological connection relationship between the partitions, the cloud intelligent layer receives the partition fault integration results reported by the edge computing nodes of each partition, and arranges and splices the partition fault integration results along the partition topology order according to the topological adjacency relationship between each partition to establish a fault information sequence covering the entire network. Subsequently, the cloud intelligent layer checks the consistency of the fault status information of the boundary nodes at the boundary connection of adjacent partitions to determine whether the fault judgment results of adjacent partitions on the shared boundary nodes are consistent. When the fault judgment results are consistent, the partition fault integration results of each partition are fused to generate the global fault handling result of the distribution network. When the fault judgment results are inconsistent, the boundary joint re-judgment mechanism is triggered, and the boundary nodes are re-judged and checked again based on the original operation perception parameters of the boundary nodes and their upstream and downstream nodes. The results of consistent verification are fused to generate the global fault handling result of the distribution network.

[0043] Furthermore, the method provided in the application embodiment, which integrates the partition faults of each partition according to the topological connection relationship between the partitions, performs boundary consistency verification and fusion to generate a global fault handling result for the distribution network, also includes: The cloud-based intelligent layer receives the partition fault integration results reported by the edge computing nodes of each partition. Based on the topological adjacency relationship between each partition, it arranges and splices the partition fault integration results of each partition along the partition topological order to establish a fault information sequence covering the entire network. The cloud-based intelligent layer performs consistency verification on the fault status information of the boundary nodes at the boundary junctions of adjacent partitions to determine whether the fault judgment results of adjacent partitions on the shared boundary nodes are consistent. If consistent, the partition fault integration results of each partition are merged to generate a global fault handling result for the distribution network. If inconsistent, a boundary joint re-judgment mechanism is triggered. Based on the original operating perception parameters of the boundary nodes and their upstream and downstream nodes, the fault judgment of the boundary nodes is re-performed and verified again. The results of the verification are merged to generate a global fault handling result for the distribution network.

[0044] In this embodiment, after receiving the partition fault integration results reported by the edge computing nodes of each partition, the cloud-based intelligent layer extracts the fault handling partition, the location of the fault segment within the partition, and the fault type recorded in the fault integration results of each partition, and reads the topological adjacency relationship between each fault handling partition; the topological adjacency relationship is the connection relationship formed between adjacent fault handling partitions through boundary nodes or boundary lines. The cloud-based intelligent layer determines the connection order of each fault handling partition in the distribution network operation topology based on the topological adjacency relationship, and uses the boundary connection of adjacent partitions as the splicing position to arrange and splice the partition fault integration results of each partition along the partition topological order to establish a fault information sequence covering the entire network; the fault information sequence includes the partition fault integration results arranged according to the partition topological order, the location of the fault segment within the partition, the fault type, and the fault status information of the boundary nodes of adjacent partitions.

[0045] After establishing a fault information sequence covering the entire network, the cloud-based intelligent layer sequentially reads the boundary nodes at the boundary junctions of adjacent partitions along the fault information sequence, and extracts the fault status information of each boundary node from the partition fault integration results of the adjacent partitions. Boundary nodes are monitoring nodes shared or interconnected at the topological connection points of adjacent fault-handling partitions. The fault status information includes whether the boundary node exhibits fault characteristics, fault type, fault direction, and the location of the corresponding fault segment. The cloud-based intelligent layer performs consistency verification on the fault judgment results of adjacent partitions at the shared boundary nodes. Specifically, it compares the fault characteristic judgment results, fault type, fault direction, and fault segment location of the same shared boundary node in adjacent partitions item by item. If all the comparison results are the same, the fault judgment results of adjacent partitions at the shared boundary nodes are determined to be consistent; if any comparison result is different, the fault judgment results of adjacent partitions at the shared boundary nodes are determined to be inconsistent.

[0046] If adjacent partitions have consistent fault determination results at their shared boundary nodes, the cloud-based intelligent layer merges the partition fault integration results of each partition according to the fault information sequence covering the entire network, and merges the fault status information repeatedly recorded at the shared boundary nodes of adjacent partitions. During the merging process, the cloud-based intelligent layer retains the already determined fault segment locations and fault types within each partition, and writes the consistent fault status information at the boundary junctions of adjacent partitions as the fault status information of the same boundary node into the merging result, forming the global fault handling result of the distribution network according to the partition topology order. The global fault handling result of the distribution network includes the fault segment locations, fault types, and corresponding fault handling partitions throughout the entire network.

[0047] If the fault determination results of adjacent zones at the shared boundary node are inconsistent, the cloud-based intelligent layer triggers a boundary joint re-determination mechanism and identifies the inconsistent boundary node and its upstream and downstream nodes based on the topological connection relationship between zones. Subsequently, the cloud-based intelligent layer retrieves the original operational sensing parameters of the boundary node and its upstream and downstream nodes, including three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power. The cloud-based intelligent layer arranges the original operational sensing parameters of the boundary node and its upstream and downstream nodes according to the topological connection relationship between the nodes and compares each original operational sensing parameter with the corresponding fault determination parameter threshold to obtain the fault status information of the boundary node again. Then, the cloud-based intelligent layer performs consistency verification again with the fault determination results of adjacent zones at the shared boundary node. If the verification is consistent, the original inconsistent fault status information is replaced with the newly obtained boundary node fault status information, and the zone fault integration results of each zone are merged to generate the global fault handling result of the distribution network.

[0048] In summary, the embodiments of this application have at least the following technical effects: This application, based on the operational topology of a distribution network, divides the distribution network into multiple fault handling zones, determining the node range covered by each zone and the topological connections between nodes. Based on these multiple fault handling zones, each monitoring node collects electrical quantity operation data of the corresponding nodes within its zone. Node-level fault analysis is performed based on the collected electrical quantity data to generate node fault feature data. Taking each zone as a unit, the node fault feature data of all nodes within that zone are aggregated. The fault features of each node are integrated and analyzed based on the topological connections between nodes within the zone, generating a zone-wide fault integration result. The zone-wide fault integration results are then checked for boundary consistency and fused according to the topological connections between zones to generate a global fault handling result for the distribution network. This invention solves the technical problems of low fault location accuracy and low processing efficiency in existing distribution network technologies. By performing zoned node fault analysis and fusion of zoned fault results based on the operational topology, it achieves the technical effect of improving the accuracy of fault location and the efficiency of fault handling in distribution networks.

[0049] Example 2, based on the same inventive concept as the AI-based distributed fault handling method for distribution networks in the previous examples, such as... Figure 2 As shown, this application provides an AI-based distributed fault handling system for power distribution networks. The system and method embodiments in this application are based on the same inventive concept. The system includes: The partitioning module 11 is used to divide the distribution network into multiple fault handling zones based on the operating topology of the distribution network, and determine the node range covered by each zone and the topological connection relationship between the nodes; the acquisition module 12 is used to collect electrical quantity operation data of the corresponding nodes in the multiple fault handling zones by each monitoring node, and perform node-level fault analysis and processing based on the collected electrical quantity data to generate node fault feature data; the integration and analysis module 13 is used to aggregate the node fault feature data of all nodes in each zone, and perform integrated analysis of the fault features of each node based on the topological connection relationship between the nodes in the zone to generate the partition fault integration result of each zone; the verification and fusion module 14 is used to perform boundary consistency verification and fusion of the partition fault integration results of each zone according to the topological connection relationship between the zones to generate the global fault handling result of the distribution network.

[0050] Furthermore, the system is also used to implement the following functions: Based on the electrical propagation characteristics and impact range of different fault types in the distribution network, a set of operation sensing parameters for characterizing fault-dependent responses is determined. According to the response mapping relationship between fault type and operation sensing parameter set, the distribution network topology is partitioned to obtain multiple fault handling partitions. Based on the multiple fault handling partitions, the range of topology nodes covered, and the topology connection relationships between nodes, a partitioned topology engine is established for the fault handling process to call and identify the topology connection relationships of each fault.

[0051] Furthermore, the system is also used to implement the following functions: Based on the response mapping relationship between fault type and operational perception parameter set, the range of monitoring nodes required for each type of fault is determined; according to the range of monitoring nodes required for each type of fault, the response mapping relationship is divided into node-level response mapping, partition-level response mapping, and cross-partition interactive response mapping.

[0052] Furthermore, the system is also used to implement the following functions: The node-level response mapping establishes a mapping relationship between the operational perception parameters of a single monitoring node and whether the node has failed; the partition-level response mapping establishes a mapping relationship between the combination of operational perception parameters of all monitoring nodes within a partition and whether a faulty section exists within the corresponding partition; the cross-partition interactive response mapping establishes a mapping relationship between the combination of operational perception parameters of boundary monitoring nodes of adjacent partitions and whether a cross-partition fault exists at the corresponding boundary.

[0053] Furthermore, the system is also used to implement the following functions: Each monitoring node collects its own electrical quantity operation data at a set sampling frequency. The electrical quantity operation data includes three-phase voltage, three-phase current, zero-sequence voltage, active power, and reactive power. Each monitoring node preprocesses the collected raw electrical quantity data to form a standardized input data sequence. The standardized input data sequence is then subjected to node-level fault analysis. The node-level fault analysis includes comparing each sensing parameter in the collected electrical quantity operation data with preset fault judgment parameter thresholds. Based on the comparison results, it is determined whether the node has fault characteristics. If fault characteristics are found, node fault characteristic data containing fault type, fault direction, and fault probability is generated. Each monitoring node uploads the node fault characteristic data to the edge computing node corresponding to its partition.

[0054] Furthermore, the system is also used to implement the following functions: Each edge computing node in a partition receives node fault feature data uploaded by all monitoring nodes within that partition. Following the partition's topological connections, it arranges the fault feature data of each node in topological order, constructing a fault feature chain with nodes as the sequence. The edge computing node traverses the fault direction information of each node along the fault feature chain, identifying fault segments based on the combination relationship of fault directions between adjacent nodes. If the fault directions of two adjacent nodes form a combination of opposing directions, the fault is determined to be located in the segment between the two nodes. If the fault direction of a node does not form an opposing combination with the fault directions of the preceding and following nodes, and the node itself detects a fault feature, the fault is determined to be located at that node. The edge computing node summarizes all identified fault segments and fault nodes to generate a partition fault integration result for that partition. The partition fault integration result includes the location and type of fault segments within the partition.

[0055] Furthermore, the system is also used to implement the following functions: The cloud-based intelligent layer receives the partition fault integration results reported by the edge computing nodes of each partition. Based on the topological adjacency relationship between each partition, it arranges and splices the partition fault integration results of each partition along the partition topological order to establish a fault information sequence covering the entire network. The cloud-based intelligent layer performs consistency verification on the fault status information of the boundary nodes at the boundary junctions of adjacent partitions to determine whether the fault judgment results of adjacent partitions on the shared boundary nodes are consistent. If consistent, the partition fault integration results of each partition are merged to generate a global fault handling result for the distribution network. If inconsistent, a boundary joint re-judgment mechanism is triggered. Based on the original operating perception parameters of the boundary nodes and their upstream and downstream nodes, the fault judgment of the boundary nodes is re-performed and verified again. The results of the verification are merged to generate a global fault handling result for the distribution network.

[0056] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An AI-based distributed fault handling method for power distribution networks, characterized in that, include: Based on the operating topology of the distribution network, the distribution network is divided into multiple fault handling zones, and the node range covered by each zone and the topological connection relationship between the nodes are determined. Based on multiple fault handling zones, each monitoring node collects electrical quantity operation data of the corresponding node in its zone, performs node-level fault analysis and processing based on the collected electrical quantity data, and generates node fault characteristic data. Taking each partition as a unit, the node fault characteristic data of all nodes within the partition are aggregated. Based on the topological connection relationship between nodes within the partition, the fault characteristics of each node are integrated and analyzed to generate the partition fault integration result of each partition. The fault integration results of each partition are checked and merged according to the topological connection relationship between the partitions to generate the global fault handling results of the distribution network.

2. The AI-based distributed fault handling method for distribution networks according to claim 1, characterized in that, The distribution network is divided into multiple fault handling zones, and the node coverage area and topological connections between nodes in each zone are determined, including: Based on the electrical propagation characteristics and impact range of different fault types in the distribution network, a set of operation sensing parameters for characterizing fault-dependent responses is determined. Based on the response mapping relationship between fault type and operation sensing parameter set, the distribution network topology is partitioned to obtain multiple fault handling partitions; Based on the multiple fault handling partitions, the range of topology nodes covered, and the topology connection relationships between nodes, a partition topology engine is established to identify the topology connection relationships of each fault during the fault handling process.

3. The AI-based distributed fault handling method for distribution networks according to claim 2, characterized in that, Based on the mapping relationship between fault types and the set of operational sensing parameters, the distribution network topology is partitioned, including: Based on the response mapping relationship between fault type and operational perception parameter set, the range of monitoring nodes required for each type of fault is determined. Based on the range of monitoring nodes required for various types of faults, the response mapping relationship is divided into node-level response mapping, partition-level response mapping, and cross-partition interactive response mapping.

4. The AI-based distributed fault handling method for distribution networks according to claim 3, characterized in that, The node-level response mapping establishes a mapping relationship between the operational sensing parameters of a single monitoring node and whether the node has failed. The partition-level response mapping establishes a mapping relationship between the combination of operational sensing parameters of all monitoring nodes within a partition and whether a faulty section exists within the corresponding partition. The cross-zone interactive response mapping establishes a mapping relationship between the combination of operational sensing parameters of the boundary monitoring nodes of adjacent zones and whether there is a cross-zone fault at the corresponding boundary.

5. The AI-based distributed fault handling method for distribution networks according to claim 1, characterized in that, Based on multiple fault handling zones, each monitoring node collects electrical quantity operation data of the corresponding node in its zone. Node-level fault analysis and processing are performed based on the collected electrical quantity data to generate node fault characteristic data, including: Each monitoring node collects electrical quantity operation data of its own node at a set sampling frequency. The electrical quantity operation data includes three-phase voltage, three-phase current, zero-sequence voltage, active power and reactive power. Each monitoring node preprocesses the collected raw electrical quantity data to form a standardized input data sequence. The standardized input data sequence is then subjected to node-level fault analysis. The node-level fault analysis includes comparing each sensing parameter in the collected electrical quantity operation data with a preset fault judgment parameter threshold. Based on the comparison result, it is determined whether the node has fault characteristics. If fault characteristics exist, node fault characteristic data containing fault type, fault direction, and fault probability is generated. Each monitoring node uploads the node fault characteristic data to the edge computing node corresponding to its partition.

6. The AI-based distributed fault handling method for distribution networks according to claim 1, characterized in that, Each partition is treated as a unit, and node fault characteristic data of all nodes within that partition are aggregated. Based on the topological connections between nodes within the partition, the fault characteristics of each node are integrated and analyzed to generate partition fault integration results for each partition, including: Each edge computing node in a partition receives node fault feature data uploaded by all monitoring nodes within that partition. According to the topological connection relationship of the partition, the fault feature data of each node is arranged in the topological order to construct a fault feature chain with nodes as the sequence. The edge computing node traverses the fault direction information of each node along the fault feature chain and identifies the fault segment based on the combination relationship of fault directions between adjacent nodes. If the fault directions of two adjacent nodes form a combination feature of opposite directions, the fault is determined to be located in the segment between the two nodes. If the fault direction of a node does not form a combination of opposite directions with the fault directions of the nodes before and after it and the node itself detects the fault feature, the fault is determined to be located at the node itself. The edge computing node will summarize all identified faulty segments and faulty nodes to generate a partition fault integration result for this partition. The partition fault integration result includes the location of the faulty segment and the fault type within the partition.

7. The AI-based distributed fault handling method for distribution networks according to claim 6, characterized in that, The fault integration results of each partition are performed according to the topological connection relationship between the partitions, and boundary consistency verification and fusion are carried out to generate the global fault handling results of the distribution network, including: The cloud-based intelligent layer receives the partition fault integration results reported by the edge computing nodes of each partition. Based on the topological adjacency relationship between each partition, it arranges and splices the partition fault integration results of each partition along the partition topological order to establish a fault information sequence covering the entire network. The cloud-based intelligent layer performs consistency verification on the fault status information of the boundary nodes at the boundary junction of adjacent partitions, and determines whether the fault judgment results of adjacent partitions on the shared boundary nodes are consistent. If the fault integration results of each partition are consistently merged, a global fault handling result for the distribution network will be generated. If there is a discrepancy, the boundary joint re-judgment mechanism is triggered. Based on the original operational perception parameters of the boundary node and its upstream and downstream nodes, the fault judgment of the boundary node is re-evaluated and then verified again. The results of the verification are merged to generate the global fault handling result of the distribution network.

8. An AI-based distributed fault handling system for power distribution networks, characterized in that: The system is used to execute the AI-based distributed fault handling method for distribution networks as described in any one of claims 1-7, and the system includes: The partitioning module is used to divide the distribution network into multiple fault handling zones based on the operation topology of the distribution network, and to determine the node range covered by each zone and the topological connection relationship between the nodes. The data acquisition module is used to collect electrical quantity operation data of the corresponding nodes in a fault handling zone based on multiple fault handling zones. Based on the collected electrical quantity data, node-level fault analysis and processing are performed to generate node fault characteristic data. The integration and analysis module is used to aggregate the node fault characteristic data of all nodes within each partition, and to perform integrated analysis on the fault characteristics of each node based on the topological connection relationship between nodes within the partition, generating the partition fault integration results for each partition. The verification and fusion module is used to perform boundary consistency verification and fusion of the partition fault integration results of each partition according to the topological connection relationship between the partitions, and generate the global fault handling results of the distribution network.

Images