A power distribution network ground fault automatic identification method and device

Abstract

This invention relates to the field of distribution network fault technology, and discloses an automatic identification method and device for distribution network grounding faults. The method includes: normalizing the format of multi-source heterogeneous data to obtain a standardized fault feature sequence; performing cluster analysis on historical fault features to obtain a feature pattern library; comparing and correlating the standardized fault feature sequence with the feature pattern library, and cross-validating the comparison results to obtain a preliminary fault type inference; performing deep alignment on the standardized fault feature sequence and the feature pattern library, and comprehensively judging the results of the deep alignment based on the line connection relationship to obtain the final fault type; performing spatiotemporal fusion deduction on the standardized fault feature sequence to obtain the fault propagation path and fault impact range; and tracing the fault propagation path backward to obtain the fault occurrence point. This invention can improve the accuracy of automatic identification of distribution network grounding faults.

Description

Technical Field

[0001] This invention relates to the field of distribution network fault technology, and in particular to an automatic identification method and device for grounding faults in distribution networks. Background Technology

[0002] In the current process of identifying grounding faults in distribution networks, there is a lack of efficient and unified standardized solutions for processing multi-source heterogeneous data. The differences in format and time-space synchronization of different types of data lead to insufficient accuracy in fault feature extraction, making it difficult to form an effective feature sequence that can accurately reflect the fault state, thus affecting the basic reliability of subsequent fault identification. Traditional methods lack dynamic correlation and cross-validation of real-time changes in the distribution network state when matching fault features and judging the type. They rely solely on single-dimensional feature comparison, which is prone to matching deviations and cannot quickly and accurately lock the initial fault type.

[0003] In the fault type confirmation and fault source tracing stages, existing technologies have failed to fully integrate the line connection relationships and topology of the distribution network, and the deep alignment of fault characteristics is not comprehensive enough, resulting in a lack of rigorous logical support for the final fault type judgment. Moreover, the deduction of fault propagation paths and the location of fault occurrence points mostly rely on single features or local data, and the spatiotemporal fusion analysis of standardized fault feature sequences has not been achieved. This makes the definition of the fault impact range not accurate enough, and the accuracy and efficiency of reverse source tracing are low, making it difficult to meet the actual needs of rapid handling and recovery of distribution network faults. Therefore, how to improve the accuracy of distribution network fault identification has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides an automatic identification method and apparatus for grounding faults in power distribution networks to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an automatic identification method for grounding faults in a power distribution network, comprising:

[0006] S1. Obtain multi-source heterogeneous data of the distribution network, perform format normalization processing on the multi-source heterogeneous data, and obtain the standardized fault feature sequence of the distribution network.

[0007] S2. Perform cluster analysis on the historical fault characteristics of the distribution network to obtain a feature pattern library of the distribution network;

[0008] S3. The standardized fault feature sequence is correlated and compared with the feature pattern library, and the comparison results are cross-validated based on the state changes of the distribution network to obtain a preliminary fault type inference of the distribution network.

[0009] S4. Based on the preliminary fault type inference, perform deep alignment between the standardized fault feature sequence and the feature pattern library, and comprehensively evaluate the results of the deep alignment according to the line connection relationship of the distribution network to obtain the final fault type of the distribution network.

[0010] S5. Based on the final fault type and the line connection relationship, perform spatiotemporal fusion simulation on the standardized fault feature sequence to obtain the fault propagation path and fault impact range of the distribution network;

[0011] S6. Based on the line connection relationship and the fault impact range, the fault propagation path is traced in reverse to obtain the fault occurrence point of the distribution network.

[0012] In a preferred embodiment, the step of performing cluster analysis on the historical fault characteristics of the distribution network to obtain a feature pattern library of the distribution network includes:

[0013] Cluster analysis is performed on the historical fault cases of the power distribution network to obtain the historical fault types of the power distribution network;

[0014] The characteristics of historical faults in the historical fault cases are summarized and analyzed to obtain the historical fault groups of the distribution network.

[0015] By associating and mapping the historical fault types and the historical fault groups, the historical fault characteristic patterns of the distribution network are obtained.

[0016] The historical fault feature patterns are structured and stored to obtain the feature pattern library of the power distribution network.

[0017] In a preferred embodiment, the step of comparing and associating the standardized fault feature sequence with the feature pattern library includes:

[0018] Based on the standardized fault feature sequence, pattern matching is performed on the feature pattern library to obtain the candidate feature pattern set of the distribution network;

[0019] The standardized fault feature sequence is decomposed into multidimensional features to obtain the morphological feature subset, temporal feature subset, and event feature subset of the distribution network;

[0020] The morphological feature subset, the temporal feature subset, and the event feature subset are compared item by item with the candidate feature pattern set to obtain the local matching degree of the distribution network.

[0021] The local matching degree is integrated to obtain the comprehensive matching degree of the distribution network, and the comprehensive matching degree is used as the comparison result between the standardized fault feature sequence and the feature pattern library.

[0022] In a preferred embodiment, the step of cross-validating the results based on the state changes of the distribution network to obtain a preliminary fault type inference for the distribution network includes:

[0023] Real-time monitoring of switch position change events and event occurrence times in the distribution network is used to obtain the state change sequence of the distribution network.

[0024] The temporal consistency of the distribution network is obtained by temporally correlating the state change sequence with the comparison result.

[0025] Based on the aforementioned time-series consistency, the local matching degree of the distribution network is adjusted in a targeted manner to obtain the optimized matching degree of the distribution network;

[0026] By performing event consistency verification on the event feature subset of the distribution network and the state change sequence, the causal relationship of the distribution network can be obtained.

[0027] Based on the causal relationship and the optimized matching degree, the candidate feature pattern set of the distribution network is sorted and filtered to obtain the preliminary fault type inference of the distribution network.

[0028] In a preferred embodiment, the step of performing deep alignment between the standardized fault feature sequence and the feature pattern library based on the preliminary fault type inference, and comprehensively evaluating the results of the deep alignment according to the line connection relationships of the distribution network to obtain the final fault type of the distribution network, includes:

[0029] Based on the preliminary fault type inference, feature slicing is performed on the standardized fault feature sequence to obtain the target feature segment of the distribution network.

[0030] Based on the preliminary fault type inference, the feature pattern library is matched and searched to obtain the similar feature pattern set of the distribution network;

[0031] The similar feature pattern set and the target feature fragment are aligned in a multimodal manner to obtain the fault alignment result of the distribution network;

[0032] Based on the line connection relationship of the distribution network, the fault alignment result is topology verified to obtain the fault propagation logic of the distribution network.

[0033] By associating and coupling the fault propagation logic and the fault alignment result, the final fault type of the distribution network is obtained.

[0034] In a preferred embodiment, the step of performing multimodal alignment between the similar feature pattern set and the target feature fragment to obtain the fault alignment result of the distribution network includes:

[0035] Tensor synthesis is performed on the target feature fragment and the set of similar feature patterns to obtain the target feature vector and the similar feature vector of the target feature fragment;

[0036] By associating the target feature vector with the similar feature vector, the feature group of the power distribution network is obtained;

[0037] Perform a spatial distance metric on the feature group to obtain the Euclidean distance of the feature group;

[0038] Based on the physical characteristics of the line connection relationship and the preliminary fault type inference, the feature group is correlated and quantified to obtain the degree of cross-influence of the feature group;

[0039] Based on the Euclidean distance and the degree of cross-influence, the comprehensive dissimilarity of the feature group is calculated, wherein the formula for calculating the comprehensive dissimilarity is:

[0040] ;

[0041] in, This indicates the overall degree of difference. Indicates the number of the feature groups. Indicates the first The weight coefficients of the feature groups Represents the Euclidean distance of the feature set. Indicates the first The first feature group and the first Cross-influence coefficients among feature groups This represents the preset smallest positive number. This represents the function that takes the maximum value.

[0042] The overall difference is used as the fault alignment result of the distribution network.

[0043] In a preferred embodiment, the step of performing spatiotemporal fusion simulation on the standardized fault feature sequence based on the final fault type and the line connection relationship to obtain the fault propagation path and fault impact range of the distribution network includes:

[0044] Based on the final fault type, a subset of key features of the standardized fault feature sequence is selected;

[0045] Path fitting is performed on the key feature subset to obtain the fault evolution trajectory of the distribution network;

[0046] The spatial topology network of the power distribution network is obtained by constructing the topology of the line connections.

[0047] The fault evolution trajectory is mapped onto the spatial topology network to obtain the potential fault propagation path of the distribution network;

[0048] The confidence level of the potential fault propagation paths is assessed to obtain the fault propagation paths of the distribution network.

[0049] The range of the fault propagation path is combined into the fault impact range of the distribution network.

[0050] In a preferred embodiment, the step of tracing the fault propagation path backward based on the line connection relationship and the fault impact range to obtain the fault occurrence point of the distribution network includes:

[0051] Based on the fault's impact range, an impact diffusion simulation is performed on the nodes along the fault propagation path to obtain the characteristic intensity change trend of the nodes.

[0052] The node with the most significant trend in characteristic intensity change is selected as the starting point for tracing the source of the distribution network.

[0053] Using the line connection relationship as a constraint network, the fault propagation path is reversed from the source of the fault to obtain the reverse source path of the distribution network.

[0054] The reverse tracing path is overlapped with the spatial boundary of the fault impact range to obtain the candidate tracing path of the distribution network.

[0055] Perform endpoint convergence analysis on the candidate tracing paths to obtain the endpoint nodes of the candidate tracing paths;

[0056] By comprehensively evaluating the electrical location and connection relationships of the endpoint nodes, the fault location of the distribution network can be obtained.

[0057] To address the above problems, the present invention also provides an automatic identification device for grounding faults in power distribution networks, the device comprising:

[0058] The fault feature processing module is used to acquire multi-source heterogeneous data of the distribution network, perform format normalization processing on the multi-source heterogeneous data, and obtain a standardized fault feature sequence of the distribution network.

[0059] The fault mode library construction module is used to perform cluster analysis on the historical fault characteristics of the distribution network to obtain the feature mode library of the distribution network.

[0060] The fault type inference module is used to correlate and compare the standardized fault feature sequence with the feature pattern library, and to perform cross-validation of the comparison results based on the state changes of the distribution network to obtain a preliminary fault type inference of the distribution network.

[0061] The fault type determination module is used to perform deep alignment between the standardized fault feature sequence and the feature pattern library based on the preliminary fault type inference, and to comprehensively evaluate the result of the deep alignment according to the line connection relationship of the distribution network to obtain the final fault type of the distribution network.

[0062] The fault impact analysis module is used to perform spatiotemporal fusion simulation of the standardized fault feature sequence based on the final fault type and the line connection relationship, so as to obtain the fault propagation path and fault impact range of the distribution network.

[0063] The fault origin tracing module is used to trace the fault propagation path in reverse based on the line connection relationship and the fault impact range to obtain the fault occurrence point of the distribution network.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. This invention improves the accuracy and efficiency of fault type identification by normalizing the format of multi-source heterogeneous data from the distribution network, constructing a feature pattern library based on cluster analysis of historical fault characteristics, and then performing precise processing through multiple stages such as correlation comparison, cross-validation, deep alignment, and comprehensive judgment. The entire process forms a complete closed loop from data processing to fault inference, significantly shortening the response time for fault identification and ensuring that core fault information can be quickly identified.

[0066] 2. This invention clarifies the fault propagation path and impact range through spatiotemporal fusion simulation, and accurately locates the fault occurrence point by relying on reverse tracing. This not only provides clear guidance for fault investigation but also effectively reduces the blindness and workload of fault investigation. Its systematic processing logic and precise analysis capabilities significantly improve the overall efficiency of grounding fault handling in distribution networks, providing reliable technical support for the stable operation of distribution networks. Attached Figure Description

[0067] Figure 1 This is a flowchart illustrating an automatic identification method for grounding faults in a power distribution network according to an embodiment of the present invention.

[0068] Figure 2 A functional block diagram of an automatic grounding fault identification device for power distribution networks provided in an embodiment of the present invention;

[0069] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0070] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0071] This application provides an automatic identification method for grounding faults in a power distribution network. The executing entity of this automatic identification method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the automatic identification method for grounding faults in a power distribution network can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0072] Reference Figure 1 The diagram shown is a flowchart illustrating an automatic grounding fault identification method for a distribution network according to an embodiment of the present invention. In this embodiment, the automatic grounding fault identification method for a distribution network includes:

[0073] S1. Obtain multi-source heterogeneous data of the distribution network, perform format normalization processing on the multi-source heterogeneous data, and obtain the standardized fault feature sequence of the distribution network.

[0074] In this embodiment of the invention, the step of acquiring multi-source heterogeneous data of the distribution network and performing format normalization processing on the multi-source heterogeneous data to obtain a standardized fault feature sequence of the distribution network includes:

[0075] Acquire electrical characteristic data and event signal streams of the distribution network to obtain multi-source heterogeneous data of the distribution network;

[0076] The multi-source heterogeneous data is spatiotemporally aligned to obtain the synchronization data of the power distribution network;

[0077] The synchronization data is formatted to obtain intermediate data for the power distribution network;

[0078] Based on historical fault cases of the distribution network, the intermediate data is correlated and matched to obtain the fault characteristics of the distribution network;

[0079] The fault characteristics are segmented by a sliding window to obtain a standardized fault characteristic sequence of the distribution network.

[0080] By deploying various sensors, monitoring terminals, and data acquisition equipment in the distribution network, electrical characteristic data and event signal streams during the operation of the distribution network are collected comprehensively. The electrical characteristic data includes parameters such as voltage, current, resistance, and power that reflect the electrical operating status of the distribution network. The event signal stream includes various event data related to the operation of the distribution network, such as equipment start-up and shutdown signals, fault alarm signals, and operation command signals. These data from different sources and of different types are integrated and aggregated to form multi-source heterogeneous data of the distribution network.

[0081] Based on the unified time reference and spatial location identifier of the distribution network, the collected multi-source heterogeneous data is spatiotemporally aligned. In terms of time dimension, according to the preset unified time interval, the data from different collection time points are adjusted to the same time node to ensure that the data remains synchronized in the time dimension. In terms of spatial dimension, according to the physical location distribution of distribution network equipment and line connection relationship, the corresponding spatial location information is labeled for each data, so that the data can correspond to the actual spatial layout of the distribution network, and finally the synchronized data of the distribution network is obtained.

[0082] Establish a unified data format standard, clarify the requirements for data storage structure, field types, encoding methods, etc., and perform unified format processing on the synchronized data after time and space alignment. Convert all heterogeneous format data corresponding to different acquisition devices and different data types into a standardized format that conforms to the unified standard, eliminate the impact of data format differences, and form intermediate data for the power distribution network.

[0083] The system retrieves the historical fault case database stored in the distribution network. This database contains complete information such as fault scenarios, fault manifestations, and related data characteristics corresponding to various types of faults that have occurred in the past. The processed intermediate data is compared and correlated with the data in the historical fault cases one by one. Information that matches the data characteristics of the historical fault cases and can reflect the fault-related attributes is selected. From this, key indicators and characteristic manifestations that can characterize the fault state of the distribution network are extracted to obtain the fault characteristics of the distribution network.

[0084] A fixed-length sliding window is set and continuously slides over the extracted fault feature data sequence. After each slide, the window covers a fixed number of fault feature data. The fault feature data within the window is extracted as an independent feature unit. By continuously sliding the window until all fault feature data is covered, a series of standardized fault feature sequences of distribution networks with uniform structure and consistent length are finally obtained.

[0085] The beneficial effects are that the systematic steps of collecting and standardizing multi-source heterogeneous data ensure the integrity, synchronization and standardization of fault feature data. The resulting standardized fault feature sequence can accurately reflect the fault-related attributes of the distribution network, providing high-quality and usable basic data support for subsequent links such as correlation and comparison with feature pattern library and fault type inference, effectively ensuring the smooth progress of automatic identification of grounding faults in distribution network.

[0086] S2. Perform cluster analysis on the historical fault characteristics of the distribution network to obtain a feature pattern library of the distribution network;

[0087] In this embodiment of the invention, the step of performing cluster analysis on the historical fault characteristics of the distribution network to obtain a feature pattern library of the distribution network includes:

[0088] Cluster analysis is performed on the historical fault cases of the power distribution network to obtain the historical fault types of the power distribution network;

[0089] The characteristics of historical faults in the historical fault cases are summarized and analyzed to obtain the historical fault groups of the distribution network.

[0090] By associating and mapping the historical fault types and the historical fault groups, the historical fault characteristic patterns of the distribution network are obtained.

[0091] The historical fault feature patterns are structured and stored to obtain the feature pattern library of the power distribution network.

[0092] Retrieve all historical fault cases stored in the distribution network. These cases contain complete information such as the fault manifestations, related data, and processing records for each fault occurrence. Classify and organize them according to the core attributes and characteristics of the faults, grouping cases with the same or similar fault essence into the same category, thereby clarifying the historical fault types of the distribution network.

[0093] Each historical fault case is reviewed to identify its characteristics. These characteristics cover key information such as changes in electrical parameters, signal characteristics, and event triggering sequence at the time of the fault. The characteristics are grouped and integrated according to their similarity and correlation. Historical fault characteristics that show consistent patterns are grouped together to form historical fault groups for the distribution network.

[0094] Establish the correspondence between historical fault types and historical fault groups, clarify the typical historical fault groups corresponding to each historical fault type, and determine the historical fault type to which each historical fault group belongs, so that historical fault types and historical fault groups are interconnected and correspond to each other, forming a historical fault characteristic pattern of the distribution network that can fully reflect the nature and characteristic correlation of faults.

[0095] The design standard storage structure can clearly present the components and interrelationships of historical fault feature patterns. All the collected historical fault feature patterns are stored in an orderly manner according to the preset storage structure, ensuring that the stored historical fault feature patterns are easy to retrieve, compare and call later, and finally form a feature pattern library of the distribution network.

[0096] The beneficial effect is that the feature pattern library constructed through the systematic clustering, induction, association mapping and structured storage process can completely retain the core features and type association information of historical faults in the distribution network, providing a comprehensive, standardized and usable reference for the subsequent association comparison of standardized fault feature sequences, and effectively ensuring the accuracy and efficiency of fault type inference.

[0097] S3. The standardized fault feature sequence is correlated and compared with the feature pattern library, and the comparison results are cross-validated based on the state changes of the distribution network to obtain a preliminary fault type inference of the distribution network.

[0098] In this embodiment of the invention, the step of comparing and associating the standardized fault feature sequence with the feature pattern library includes:

[0099] Based on the standardized fault feature sequence, pattern matching is performed on the feature pattern library to obtain the candidate feature pattern set of the distribution network;

[0100] The standardized fault feature sequence is decomposed into multidimensional features to obtain the morphological feature subset, temporal feature subset, and event feature subset of the distribution network;

[0101] The morphological feature subset, the temporal feature subset, and the event feature subset are compared item by item with the candidate feature pattern set to obtain the local matching degree of the distribution network.

[0102] The local matching degree is integrated to obtain the comprehensive matching degree of the distribution network, and the comprehensive matching degree is used as the comparison result between the standardized fault feature sequence and the feature pattern library.

[0103] The preliminary fault type inference of the distribution network is obtained by cross-validating the results based on the state changes of the distribution network, including:

[0104] Real-time monitoring of switch position change events and event occurrence times in the distribution network is used to obtain the state change sequence of the distribution network.

[0105] The temporal consistency of the distribution network is obtained by temporally correlating the state change sequence with the comparison result.

[0106] Based on the aforementioned time-series consistency, the local matching degree of the distribution network is adjusted in a targeted manner to obtain the optimized matching degree of the distribution network;

[0107] By performing event consistency verification on the event feature subset of the distribution network and the state change sequence, the causal relationship of the distribution network can be obtained.

[0108] Based on the causal relationship and the optimized matching degree, the candidate feature pattern set of the distribution network is sorted and filtered to obtain the preliminary fault type inference of the distribution network.

[0109] Using standardized fault feature sequences as a reference, all historical fault feature patterns stored in the feature pattern library are traversed to select feature patterns that are similar to the standardized fault feature sequences in terms of core features. These selected feature patterns are then summarized and integrated to form a candidate feature pattern set for the distribution network.

[0110] The standardized fault feature sequence is decomposed and analyzed from different dimensions to extract relevant features that reflect the physical morphological attributes of the fault, forming a subset of the morphological features of the distribution network; relevant features that reflect the time-varying pattern of the fault are extracted to form a subset of the temporal features of the distribution network; and various event triggering and development features related to the fault are extracted to form a subset of the event features of the distribution network.

[0111] Each feature in the morphological feature subset, temporal feature subset, and event feature subset is compared with each corresponding feature in the candidate feature pattern set to determine the degree of fit between each feature. The corresponding value is determined based on the degree of fit, and these values ​​are used as the local matching degree of the distribution network.

[0112] All obtained local matching degrees are summarized and integrated. By comprehensively considering the importance of each local matching degree in the overall comparison, a comprehensive value that can fully reflect the matching situation between the standardized fault feature sequence and the feature pattern library is calculated. This value is the comprehensive matching degree of the distribution network, which is used as the comparison result between the standardized fault feature sequence and the feature pattern library.

[0113] The monitoring equipment deployed in the distribution network captures changes in switch positions in real time, records each switch position change event and the specific time of the event, and arranges this information in chronological order of the events to form a sequence of distribution network state changes.

[0114] The occurrence time of each event in the state change sequence is correlated with the corresponding feature matching time node in the comparison result. The consistency of the two in the time development order is checked, and the timing consistency of the distribution network is obtained based on the consistency judgment result.

[0115] Based on the judgment results of time sequence consistency, the previously obtained local matching degrees are adjusted accordingly. For local matching degrees that are consistent with time sequence consistency, their values ​​are maintained or appropriately increased, while for local matching degrees that are inconsistent with time sequence consistency, their values ​​are decreased, and finally the optimized matching degree of the distribution network is obtained.

[0116] By comparing the fault-related events recorded in the event feature subset with the switch position change events in the state change sequence, we can analyze whether there is a logical relationship between the two types of events, clarify the causal relationship between different events, and obtain the causal relationship of the distribution network.

[0117] By combining the two indicators of causal relationship and optimal matching degree, the candidate feature pattern set is sorted and screened. Feature patterns with clear causal relationship and high optimal matching degree are retained first, while feature patterns with unclear causal relationship or low optimal matching degree are eliminated. From the screened feature patterns, the pattern that best matches the current fault situation is determined, and the preliminary fault type inference of the distribution network is obtained.

[0118] The beneficial effects are that the comprehensiveness and accuracy of the correlation comparison are ensured by multi-dimensional decomposition and item-by-item comparison, and the matching results are further corrected by cross-validation of the distribution network status changes. This makes the preliminary fault type inference based on sufficient feature matching basis and in line with the actual operating status of the distribution network, effectively improving the reliability and accuracy of the preliminary fault type inference, and laying a solid foundation for the determination of the final fault type.

[0119] S4. Based on the preliminary fault type inference, perform deep alignment between the standardized fault feature sequence and the feature pattern library, and comprehensively evaluate the results of the deep alignment according to the line connection relationship of the distribution network to obtain the final fault type of the distribution network.

[0120] In this embodiment of the invention, the step of performing deep alignment between the standardized fault feature sequence and the feature pattern library based on the preliminary fault type inference, and comprehensively evaluating the results of the deep alignment according to the line connection relationship of the distribution network to obtain the final fault type of the distribution network, includes:

[0121] Based on the preliminary fault type inference, feature slicing is performed on the standardized fault feature sequence to obtain the target feature segment of the distribution network.

[0122] Based on the preliminary fault type inference, the feature pattern library is matched and searched to obtain the similar feature pattern set of the distribution network;

[0123] The similar feature pattern set and the target feature fragment are aligned in a multimodal manner to obtain the fault alignment result of the distribution network;

[0124] Based on the line connection relationship of the distribution network, the fault alignment result is topology verified to obtain the fault propagation logic of the distribution network.

[0125] By associating and coupling the fault propagation logic and the fault alignment result, the final fault type of the distribution network is obtained.

[0126] The step of performing multimodal alignment between the similar feature pattern set and the target feature fragment to obtain the fault alignment result of the distribution network includes:

[0127] Tensor synthesis is performed on the target feature fragment and the set of similar feature patterns to obtain the target feature vector and the similar feature vector of the target feature fragment;

[0128] By associating the target feature vector with the similar feature vector, the feature group of the power distribution network is obtained;

[0129] Perform a spatial distance metric on the feature group to obtain the Euclidean distance of the feature group;

[0130] Based on the physical characteristics of the line connection relationship and the preliminary fault type inference, the feature group is correlated and quantified to obtain the degree of cross-influence of the feature group;

[0131] Based on the Euclidean distance and the degree of cross-influence, the comprehensive dissimilarity of the feature group is calculated, wherein the formula for calculating the comprehensive dissimilarity is:

[0132] ;

[0133] in, This indicates the overall degree of difference. Indicates the number of the feature groups. Indicates the first The weight coefficients of the feature groups Represents the Euclidean distance of the feature set. Indicates the first The first feature group and the first Cross-influence coefficients among feature groups This represents the preset smallest positive number. This represents the function that takes the maximum value.

[0134] The overall difference is used as the fault alignment result of the distribution network.

[0135] Guided by the preliminary fault type inference, this study focuses on the key feature parts in the standardized fault feature sequence that are related to the preliminary inference. According to the logical correlation of fault features and the data distribution pattern, the study performs precise segmentation and extracts feature segments that can directly reflect the core attributes of the preliminary fault type, thus obtaining the target feature segments of the distribution network.

[0136] Based on the specific content inferred from the preliminary fault type, a targeted search is conducted in the feature pattern library to select historical fault feature patterns that are highly correlated with the preliminary fault type in terms of fault nature and characteristic manifestation. These selected feature patterns are then integrated and aggregated to form a set of similar feature patterns for the distribution network.

[0137] Data integration and dimensional normalization are performed on the target feature fragment and the similar feature pattern set respectively, and the multi-dimensional feature information contained in each is fused into a vector form with a unified dimension, so as to obtain the target feature vector corresponding to the target feature fragment and the similar feature vector corresponding to the similar feature pattern set respectively.

[0138] Based on the attribute categories and logical correspondences of the features, the target feature vector is associated one-to-one with the vector elements representing the same fault attribute in similar feature vectors, so that each group of corresponding vector elements forms an independent combination unit, thus obtaining the feature group of the distribution network.

[0139] By measuring the spatial positional difference between the target feature vector and similar feature vectors in a feature group, the degree of proximity between the two sets of vectors in spatial distribution is determined. The result of this measurement is the Euclidean distance of the feature group.

[0140] Based on the equipment association determined by the line connection relationship of the distribution network and the physical action law of the fault corresponding to the preliminary fault type inference, the degree of interaction and mutual influence between different vector elements in the feature group is analyzed, and the degree of influence is specifically quantified to obtain the degree of cross-influence of the feature group.

[0141] Taking into account both the Euclidean distance and the degree of cross-influence of the feature group, the quantitative results of the two are integrated and calculated to fully reflect the overall difference between the target feature segment and the set of similar feature patterns. This calculation result is the comprehensive difference degree of the feature group, which is used as the fault alignment result of the distribution network.

[0142] Based on the line connection relationship of the distribution network, a corresponding power grid topology is constructed, and the fault alignment results are mapped onto this topology. The fault characteristics reflected by the fault alignment results are checked to see if they match the connection logic of the equipment and the energy transmission path in the power grid topology. The rationality and consistency of the fault characteristics in the power grid topology are verified, and the fault propagation logic of the distribution network is obtained.

[0143] By deeply integrating the fault development path and function law reflected by the fault propagation logic with the characteristic differences reflected by the fault alignment results, we can mutually verify, supplement and improve each other, comprehensively judge the essential attributes and specific types of faults, and finally determine the final fault type of the distribution network.

[0144] The weighting coefficients of the feature groups are derived from the analysis of the physical characteristics of the distribution network line connection relationships and preliminary fault type inferences. The specific values ​​are determined by quantifying the importance of different feature groups in fault identification.

[0145] The cross-influence coefficient between characteristic groups is the result obtained by quantifying the degree of interaction and mutual influence between two characteristic groups based on the physical characteristics inferred from the connection relationship of distribution network lines and the preliminary fault type.

[0146] The Euclidean distance of a feature group is calculated by synthesizing the target feature vector and similar feature vector by tensor synthesis of the target feature fragment and similar feature pattern set, associating the two to form a feature group, and then using a spatial distance metric.

[0147] The preset minimum normal number is a fixed value set in advance to avoid the denominator being zero during the calculation process, based on the actual application scenario and calculation accuracy requirements of distribution network fault identification.

[0148] The number of feature groups is the total number of feature groups formed after associating the target feature vector with similar feature vectors.

[0149] The calculation of the overall difference first considers the Euclidean distance of each feature group and combines it with its corresponding weight coefficient. The individual influence value of each feature group is obtained through multiplication. Then, the individual influence values ​​of all feature groups are summed.

[0150] For each pair of different feature groups, calculate the product of their Euclidean distances. Simultaneously, calculate the maximum value of the Euclidean distance between the two feature groups and add it to the preset minimum positive constant to obtain the denominator. Divide the product of the Euclidean distances by the denominator to obtain the basic value of the interaction effect between the two feature groups. Multiply the basic value by the cross-influence coefficients corresponding to the two feature groups to obtain the interaction effect value between the two feature groups. Finally, sum the interaction effect values ​​of all pairwise feature groups.

[0151] The summation results of the two parts are combined to obtain comprehensive data that fully reflects the degree of difference between the target feature fragment and the set of similar feature patterns.

[0152] As the Euclidean distance of feature groups increases, the individual influence value of each feature group will increase accordingly, and the basic value of the interaction influence between pairs of feature groups will also increase, leading to an increasing trend in the overall difference.

[0153] When the weight coefficient of a feature group increases, the individual influence value of that feature group will increase accordingly. If other conditions remain unchanged, the overall difference will also increase.

[0154] When the cross-influence coefficient between feature groups increases, the interaction influence value between the two feature groups will increase accordingly. If other conditions remain unchanged, the overall difference will increase accordingly.

[0155] The preset minimum positive constant is a fixed value, and its size will not change with other factors. It only serves to avoid the denominator being zero during the calculation process and does not affect the trend of the overall difference.

[0156] The beneficial effects are that by using feature slicing, targeted retrieval and multimodal alignment based on preliminary fault type inference, the standardized fault feature sequence and feature pattern library are accurately matched. Combined with topology verification of line connection relationships, the rationality of fault analysis is further ensured. Finally, the final fault type obtained through correlation coupling has both accuracy and reliability, providing accurate core basis for subsequent fault propagation path analysis and fault location.

[0157] S5. Based on the final fault type and the line connection relationship, perform spatiotemporal fusion simulation on the standardized fault feature sequence to obtain the fault propagation path and fault impact range of the distribution network;

[0158] In this embodiment of the invention, the step of performing spatiotemporal fusion simulation on the standardized fault feature sequence based on the final fault type and the line connection relationship to obtain the fault propagation path and fault impact range of the distribution network includes:

[0159] Based on the final fault type, a subset of key features of the standardized fault feature sequence is selected;

[0160] Path fitting is performed on the key feature subset to obtain the fault evolution trajectory of the distribution network;

[0161] The spatial topology network of the power distribution network is obtained by constructing the topology of the line connections.

[0162] The fault evolution trajectory is mapped onto the spatial topology network to obtain the potential fault propagation path of the distribution network;

[0163] The confidence level of the potential fault propagation paths is assessed to obtain the fault propagation paths of the distribution network.

[0164] The range of the fault propagation path is combined into the fault impact range of the distribution network.

[0165] Based on the core attributes and characteristics of the fault as defined by the final fault type, feature data that can directly reflect the development and changes of this type of fault and play a decisive role in fault propagation analysis are precisely selected from the standardized fault feature sequence. These selected feature data together constitute the key feature subset of the distribution network.

[0166] Based on a subset of key features, and according to the order of change and logical relationship of feature data over time, a trajectory curve that can fully present the process of fault occurrence and gradual development is constructed through continuous fitting. This trajectory curve is the fault evolution trajectory of the distribution network.

[0167] By sorting out the connection methods and topological relationships of all lines and equipment in the distribution network, clarifying the connection sequence, connection ports, and the direction and distribution of lines between each device, a network model that can intuitively reflect the physical structure of the distribution network is constructed according to the actual connection logic, thus obtaining the spatial topology network of the distribution network.

[0168] The constructed fault evolution trajectory is mapped one by one to the corresponding nodes and lines in the spatial topology network according to the correspondence between time dimension and spatial location. This combines the fault evolution process with the physical structure of the distribution network to form a preliminary path scheme that can reflect the possible propagation path of the fault in the power grid, thus obtaining the potential fault propagation path of the distribution network.

[0169] By combining the characteristics of the final fault type, the reliability of line connections, and the matching degree of key feature subsets, a comprehensive evaluation is conducted on each potential fault propagation path. The probability of each path matching the actual fault propagation situation is determined, paths with high matching probabilities are retained, and paths with low matching probabilities are eliminated to obtain the fault propagation paths of the distribution network.

[0170] By summarizing all the distribution network lines, equipment and areas covered by all fault propagation paths, removing the overlapping parts between different paths, and integrating and summarizing the scope involved by all paths, a complete scope definition covering all affected areas is formed, thus obtaining the fault impact range of the distribution network.

[0171] The beneficial effects are that by selectively screening key features and combining topology for trajectory mapping and confidence assessment, the fault propagation path can be accurately predicted. At the same time, by comprehensively integrating the path range, a clear fault impact definition can be obtained, providing clear path guidance and range reference for fault handling. This effectively improves the pertinence and efficiency of fault handling and reduces the continuous impact of faults on the operation of the distribution network.

[0172] S6. Based on the line connection relationship and the fault impact range, the fault propagation path is traced in reverse to obtain the fault occurrence point of the distribution network.

[0173] In this embodiment of the invention, the step of tracing the fault propagation path backward based on the line connection relationship and the fault impact range to obtain the fault occurrence point of the distribution network includes:

[0174] Based on the fault's impact range, an impact diffusion simulation is performed on the nodes along the fault propagation path to obtain the characteristic intensity change trend of the nodes.

[0175] The node with the most significant trend in characteristic intensity change is selected as the starting point for tracing the source of the distribution network.

[0176] Using the line connection relationship as a constraint network, the fault propagation path is reversed from the source of the fault to obtain the reverse source path of the distribution network.

[0177] The reverse tracing path is overlapped with the spatial boundary of the fault impact range to obtain the candidate tracing path of the distribution network.

[0178] Perform endpoint convergence analysis on the candidate tracing paths to obtain the endpoint nodes of the candidate tracing paths;

[0179] By comprehensively evaluating the electrical location and connection relationships of the endpoint nodes, the fault location of the distribution network can be obtained.

[0180] Based on the defined scope of the fault's impact, the study focuses on all nodes along the fault propagation path, simulates the process of the fault spreading from each node to the surrounding area, and continuously tracks the changes in the strength of the characteristic indicators reflecting the fault attributes of each node during the diffusion process. This results in a trend of characteristic intensity changes of nodes that can clearly show the fluctuation pattern of the characteristic indicators of each node with the diffusion process.

[0181] The characteristic intensity change trends of all nodes are analyzed one by one. By comparing the key performance indicators such as the fluctuation range and rate of change of the characteristic intensity of different nodes, the node with the most obvious characteristic intensity fluctuation and the fastest rate of change is selected and determined as the source starting point of the distribution network.

[0182] Based on the line connection relationship of the distribution network, a constraint network is constructed to clarify the connection rules, transmission path restrictions and association logic between each node. Starting from the selected source, the possible reverse development path of the fault is deduced step by step along the line connection relationship in the order opposite to the fault propagation direction. The process of the fault extending from the source to the initial occurrence location is fully restored, and the reverse source path of the distribution network is obtained.

[0183] The obtained reverse tracing path is superimposed and compared with the spatial boundary of the fault impact range to check whether the reverse tracing path is completely within the fault impact range. At the same time, it is verified whether the overlapping area of ​​the path and the spatial boundary conforms to the logical law of fault propagation. Paths that fall completely within the fault impact range and have reasonable overlap logic are selected to obtain candidate tracing paths for the distribution network.

[0184] For each candidate tracing path, perform endpoint analysis to observe whether there is a clear endpoint convergence during the path extension process, determine whether the endpoint of each path is unique and stable, eliminate paths with scattered endpoints and no clear convergence direction, retain paths with clear endpoints and stable convergence, and determine the endpoints of these paths as the endpoint nodes of the candidate tracing paths.

[0185] Collect detailed electrical location information of the endpoint node, including its specific coordinates in the distribution network and the line section it belongs to. At the same time, sort out the connection relationship between the node and surrounding equipment and lines, such as connection method, connection strength and transmission characteristics. Comprehensively evaluate whether the electrical location of the node meets the conditions for the initial occurrence of the fault, and whether its connection relationship conforms to the law of fault propagation from the node. Finally, determine the fault location of the distribution network based on the evaluation results.

[0186] The beneficial effects are that, through node simulation based on the scope of fault impact, directional reverse inference, and multi-round verification analysis, the fault location was accurately determined. The entire tracing process strictly followed the line connection relationship and fault propagation logic, ensuring the accuracy and reliability of fault location and providing precise guidance for staff to quickly carry out fault repair and reduce fault losses.

[0187] like Figure 2 The diagram shown is a functional block diagram of an automatic grounding fault identification device for a power distribution network provided in an embodiment of the present invention.

[0188] The automatic grounding fault identification device 100 for power distribution networks described in this invention can be installed in electronic equipment. Depending on the functions implemented, the automatic grounding fault identification device 100 may include a fault feature processing module 101, a fault mode library construction module 102, a fault type inference module 103, a fault type determination module 104, a fault impact analysis module 105, and a fault origin tracing module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0189] In this embodiment, the functions of each module / unit are as follows:

[0190] The fault feature processing module 101 is used to acquire multi-source heterogeneous data of the distribution network, perform format normalization processing on the multi-source heterogeneous data, and obtain a standardized fault feature sequence of the distribution network.

[0191] The fault mode library construction module 102 is used to perform cluster analysis on the historical fault characteristics of the distribution network to obtain the feature mode library of the distribution network.

[0192] The fault type inference module 103 is used to correlate and compare the standardized fault feature sequence with the feature pattern library, and to perform cross-validation of the comparison results based on the state changes of the distribution network to obtain a preliminary fault type inference of the distribution network.

[0193] The fault type determination module 104 is used to perform deep alignment of the standardized fault feature sequence and the feature pattern library based on the preliminary fault type inference, and to comprehensively evaluate the result of the deep alignment according to the line connection relationship of the distribution network to obtain the final fault type of the distribution network.

[0194] The fault impact analysis module 105 is used to perform spatiotemporal fusion simulation of the standardized fault feature sequence based on the final fault type and the line connection relationship, so as to obtain the fault propagation path and fault impact range of the distribution network.

[0195] The fault origin tracing module 106 is used to trace the fault propagation path in reverse based on the line connection relationship and the fault impact range to obtain the fault occurrence point of the distribution network.

[0196] In the several embodiments provided by this invention, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0197] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0198] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0199] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0200] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application device that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0201] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An automatic identification method for grounding faults in a power distribution network, characterized in that, The method includes: S1. Obtain multi-source heterogeneous data of the distribution network, perform format normalization processing on the multi-source heterogeneous data, and obtain the standardized fault feature sequence of the distribution network. S2. Perform cluster analysis on the historical fault characteristics of the distribution network to obtain a feature pattern library of the distribution network; S3. The standardized fault feature sequence is correlated and compared with the feature pattern library, and the comparison results are cross-validated based on the state changes of the distribution network to obtain a preliminary fault type inference of the distribution network. S4. Based on the preliminary fault type inference, perform deep alignment between the standardized fault feature sequence and the feature pattern library, and comprehensively evaluate the results of the deep alignment according to the line connection relationship of the distribution network to obtain the final fault type of the distribution network, including: Based on the preliminary fault type inference, feature slicing is performed on the standardized fault feature sequence to obtain the target feature segment of the distribution network. Based on the preliminary fault type inference, the feature pattern library is matched and searched to obtain the similar feature pattern set of the distribution network; The similar feature pattern set and the target feature fragment are aligned in a multimodal manner to obtain the fault alignment result of the distribution network; Based on the line connection relationship of the distribution network, the fault alignment result is topology verified to obtain the fault propagation logic of the distribution network. By associating and coupling the fault propagation logic and the fault alignment result, the final fault type of the distribution network is obtained; S5. Based on the final fault type and the line connection relationship, perform spatiotemporal fusion simulation on the standardized fault feature sequence to obtain the fault propagation path and fault impact range of the distribution network; S6. Based on the line connection relationship and the fault impact range, the fault propagation path is traced in reverse to obtain the fault occurrence point of the distribution network.

2. The automatic identification method for grounding faults in a distribution network as described in claim 1, characterized in that, The process of acquiring multi-source heterogeneous data of the distribution network, performing format normalization on the multi-source heterogeneous data, and obtaining a standardized fault feature sequence of the distribution network includes: Acquire electrical characteristic data and event signal streams of the distribution network to obtain multi-source heterogeneous data of the distribution network; The multi-source heterogeneous data is spatiotemporally aligned to obtain the synchronization data of the power distribution network; The synchronization data is formatted to obtain intermediate data for the power distribution network; Based on historical fault cases of the distribution network, the intermediate data is correlated and matched to obtain the fault characteristics of the distribution network; The fault characteristics are segmented by a sliding window to obtain a standardized fault characteristic sequence of the distribution network.

3. The automatic identification method for grounding faults in a distribution network as described in claim 1, characterized in that, The clustering analysis of historical fault characteristics of the distribution network to obtain a feature pattern library of the distribution network includes: Cluster analysis is performed on the historical fault cases of the power distribution network to obtain the historical fault types of the power distribution network; The characteristics of historical faults in the historical fault cases are summarized and analyzed to obtain the historical fault groups of the distribution network. By associating and mapping the historical fault types and the historical fault groups, the historical fault characteristic patterns of the distribution network are obtained. The historical fault feature patterns are structured and stored to obtain the feature pattern library of the power distribution network.

4. The automatic identification method for grounding faults in a distribution network as described in claim 1, characterized in that, The step of comparing and associating the standardized fault feature sequence with the feature pattern library includes: Based on the standardized fault feature sequence, pattern matching is performed on the feature pattern library to obtain the candidate feature pattern set of the distribution network; The standardized fault feature sequence is decomposed into multidimensional features to obtain the morphological feature subset, temporal feature subset, and event feature subset of the distribution network; The morphological feature subset, the temporal feature subset, and the event feature subset are compared item by item with the candidate feature pattern set to obtain the local matching degree of the distribution network. The local matching degree is integrated to obtain the comprehensive matching degree of the distribution network, and the comprehensive matching degree is used as the comparison result between the standardized fault feature sequence and the feature pattern library.

5. The automatic identification method for grounding faults in a distribution network as described in claim 4, characterized in that, The preliminary fault type inference of the distribution network is obtained by cross-validating the results based on the state changes of the distribution network, including: Real-time monitoring of switch position change events and event occurrence times in the distribution network is used to obtain the state change sequence of the distribution network. The temporal consistency of the distribution network is obtained by temporally correlating the state change sequence with the comparison result. Based on the aforementioned time-series consistency, the local matching degree of the distribution network is adjusted in a targeted manner to obtain the optimized matching degree of the distribution network; By performing event consistency verification on the event feature subset of the distribution network and the state change sequence, the causal relationship of the distribution network can be obtained. Based on the causal relationship and the optimized matching degree, the candidate feature pattern set of the distribution network is sorted and filtered to obtain the preliminary fault type inference of the distribution network.

6. The automatic identification method for grounding faults in a distribution network as described in claim 1, characterized in that, The step of performing multimodal alignment between the similar feature pattern set and the target feature fragment to obtain the fault alignment result of the distribution network includes: Tensor synthesis is performed on the target feature fragment and the set of similar feature patterns to obtain the target feature vector and the similar feature vector of the target feature fragment; By associating the target feature vector with the similar feature vector, the feature group of the power distribution network is obtained; Perform a spatial distance metric on the feature group to obtain the Euclidean distance of the feature group; Based on the physical characteristics of the line connection relationship and the preliminary fault type inference, the feature group is correlated and quantified to obtain the degree of cross-influence of the feature group; Based on the Euclidean distance and the degree of cross-influence, the comprehensive dissimilarity of the feature group is calculated, wherein the formula for calculating the comprehensive dissimilarity is: ； in, This indicates the overall degree of difference. Indicates the number of the feature groups. Indicates the first The weight coefficients of the feature groups Represents the Euclidean distance of the feature set. Indicates the first The first feature group and the first Cross-influence coefficients among feature groups This represents the preset smallest positive number. This represents the function that takes the maximum value. The overall difference is used as the fault alignment result of the distribution network.

7. The automatic identification method for grounding faults in a distribution network as described in claim 1, characterized in that, The process of performing spatiotemporal fusion simulation on the standardized fault feature sequence based on the final fault type and the line connection relationship to obtain the fault propagation path and fault impact range of the distribution network includes: Based on the final fault type, a subset of key features of the standardized fault feature sequence is selected; Path fitting is performed on the key feature subset to obtain the fault evolution trajectory of the distribution network; The spatial topology network of the power distribution network is obtained by constructing the topology of the line connections. The fault evolution trajectory is mapped onto the spatial topology network to obtain the potential fault propagation path of the distribution network; The confidence level of the potential fault propagation paths is assessed to obtain the fault propagation paths of the distribution network. The range of the fault propagation path is combined into the fault impact range of the distribution network.

8. The automatic identification method for grounding faults in a distribution network as described in claim 1, characterized in that, The method of tracing the fault propagation path backward based on the line connection relationship and the fault impact range to obtain the fault occurrence point of the distribution network includes: Based on the fault's impact range, an impact diffusion simulation is performed on the nodes along the fault propagation path to obtain the characteristic intensity change trend of the nodes. The node with the most significant trend in characteristic intensity change is selected as the starting point for tracing the source of the distribution network. Using the line connection relationship as a constraint network, the fault propagation path is reversed from the source of the fault to obtain the reverse source path of the distribution network. The reverse tracing path is overlapped with the spatial boundary of the fault impact range to obtain the candidate tracing path of the distribution network. Perform endpoint convergence analysis on the candidate tracing paths to obtain the endpoint nodes of the candidate tracing paths; By comprehensively evaluating the electrical location and connection relationships of the endpoint nodes, the fault location of the distribution network can be obtained.

9. An automatic identification device for grounding faults in a power distribution network, characterized in that, The apparatus for implementing the automatic identification method for ground faults in a power distribution network as described in claim 1 includes: The fault feature processing module is used to acquire multi-source heterogeneous data of the distribution network, perform format normalization processing on the multi-source heterogeneous data, and obtain a standardized fault feature sequence of the distribution network. The fault mode library construction module is used to perform cluster analysis on the historical fault characteristics of the distribution network to obtain the feature mode library of the distribution network. The fault type inference module is used to correlate and compare the standardized fault feature sequence with the feature pattern library, and to perform cross-validation of the comparison results based on the state changes of the distribution network to obtain a preliminary fault type inference of the distribution network. The fault type determination module is used to perform deep alignment between the standardized fault feature sequence and the feature pattern library based on the preliminary fault type inference, and to comprehensively evaluate the results of the deep alignment according to the line connection relationship of the distribution network to obtain the final fault type of the distribution network, including: Based on the preliminary fault type inference, feature slicing is performed on the standardized fault feature sequence to obtain the target feature segment of the distribution network. Based on the preliminary fault type inference, the feature pattern library is matched and searched to obtain the similar feature pattern set of the distribution network; The similar feature pattern set and the target feature fragment are aligned in a multimodal manner to obtain the fault alignment result of the distribution network; Based on the line connection relationship of the distribution network, the fault alignment result is topology verified to obtain the fault propagation logic of the distribution network. By associating and coupling the fault propagation logic and the fault alignment result, the final fault type of the distribution network is obtained; The fault impact analysis module is used to perform spatiotemporal fusion simulation of the standardized fault feature sequence based on the final fault type and the line connection relationship, so as to obtain the fault propagation path and fault impact range of the distribution network. The fault origin tracing module is used to trace the fault propagation path in reverse based on the line connection relationship and the fault impact range to obtain the fault occurrence point of the distribution network.

Images