A Fast Fault Location Method for Distribution Networks Based on Artificial Intelligence

Abstract

The present invention relates to the field of distribution network fault detection, and specifically to a fast distribution network fault location method based on artificial intelligence. This method establishes a double-layer star topology to effectively connect the dispatching center with substations and distribution substations, uses multi-modal sensors to collect primary and secondary distribution network data, and performs fault prediction through a federated learning framework; combines graph convolutional networks to analyze topological relevance, generates a set of candidate fault distribution substations, and determines the fault type through cosine similarity verification and current waveform correlation analysis; uses traveling wave ranging technology to accurately locate the fault distribution network point. When an unknown fault is detected, the system triggers an incremental learning mechanism to update the global model parameters and continuously optimize the fault diagnosis ability. Based on the above method, fast and accurate distribution network fault detection and location are finally realized.

Description

Technical Field

[0001] The present invention relates to the field of distribution network fault detection, and specifically to a method for rapid fault location of distribution network based on artificial intelligence. Background Art

[0002] The distribution network is an important part of the power system, responsible for distributing the electric energy transmitted by the substation to each user. Due to the complex structure of the distribution network and the wide distribution of equipment, a large-scale power outage may occur after a fault, affecting the social and economic operation. Therefore, rapid fault detection and accurate location are crucial for improving power supply reliability and reducing power outage time. Realizing the rapid diagnosis of distribution network faults through intelligent technology can not only improve the automation level of the power system, but also reduce the operation and maintenance costs and improve the safety and stability of the power grid.

[0003] However, there are still many drawbacks at the data level in the current distribution network fault detection. On the one hand, traditional monitoring means mainly rely on SCADA (Supervisory Control and Data Acquisition) and fault recorders, and the data sampling rate is relatively low, making it difficult to accurately capture fault characteristics. On the other hand, due to the wide distribution of the distribution network and the limited density of sensor deployment, the fault data is incomplete, affecting the accuracy of fault diagnosis. In addition, the distribution network environment is complex, and external interference factors (such as load fluctuations, weather changes, etc.) are easy to introduce noise, making it difficult to further improve the fault location accuracy based on traditional methods. Therefore, how to use artificial intelligence and multi-source data fusion technology to overcome the data quality problems and achieve efficient and accurate fault location is the current research focus.

[0004] For this reason, a method for rapid fault location of distribution network based on artificial intelligence is proposed. Summary of the Invention

[0005] The present invention provides a method for rapid fault location of distribution network based on artificial intelligence. This method constructs a two-level star topology, with the dispatching center and the substation as the core, collects distribution network data through multi-modal sensors, and conducts fault prediction based on federated learning. Analyze the topological association using time alignment and graph convolutional network, screen candidate fault areas, and accurately identify the fault type through cosine similarity, current waveform correlation and dynamic time warping. Finally, combine traveling wave ranging to locate the fault distribution point, and trigger incremental learning when an unknown fault is detected to dynamically optimize the global model and improve the accuracy and adaptability of fault detection.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for rapid fault location of distribution network based on artificial intelligence, comprising:

[0008] Establish a first-level star topology with the dispatching center as the first-level central node and substations as the first-level edge nodes, and establish a second-level star topology with substations as the second-level central nodes and distribution transformer areas as the second-level edge nodes;

[0009] Collect first-level distribution network data through multi-modal sensors deployed in substations. Use the local server in the substation as the edge node and the dispatching center as the federated aggregation center to establish a federated learning framework, and use the federated learning framework to predict the first-level distribution network data to obtain faulty first-level edge nodes;

[0010] Collect second-level distribution network data through multi-modal sensors deployed in distribution transformer areas, align the first-level distribution network data and the second-level distribution network data in time series, and construct a cross-level feature matrix; Based on the cross-level feature matrix, analyze the topological correlation through a pre-trained graph convolutional network to generate a set of candidate faulty transformer areas;

[0011] Verify the set of candidate faulty transformer areas through cosine similarity and eliminate misjudged data, and further determine the type of distribution network fault by combining current waveform correlation analysis and dynamic time warping;

[0012] Obtain the coordinates of the faulty distribution network point through traveling wave ranging to complete the positioning of the distribution network fault.

[0013] Furthermore, the federated learning framework includes edge nodes and a federated aggregation center. The edge nodes use the LSTM sub-model to train the first-level distribution network data, and the federated aggregation center aggregates the parameters of the LSTM sub-model and generates a global model.

[0014] Furthermore, the training of the federated learning framework includes:

[0015] The federated aggregation center initializes the global model and distributes the global model to the edge nodes;

[0016] The edge nodes use the LSTM sub-model to train the first-level distribution network data and upload the trained model parameters to the federated aggregation center;

[0017] The federated aggregation center performs FedAvg aggregation on the model parameters, updates the global model and distributes it to the edge nodes.

[0018] Furthermore, constructing a cross-level feature matrix includes:

[0019] Obtain the first-level distribution network data and the second-level distribution network data;

[0020] Unify the timestamp formats of the first-level distribution network data and the second-level distribution network data, find the common timestamps of the first-level distribution network data and the second-level distribution network data, and use linear interpolation to complement sparse data to align the two sets of data;

[0021] Merge the aligned data to obtain the cross - level feature matrix.

[0022] Furthermore, the verification process of the cosine similarity includes:

[0023] Obtain the secondary distribution network data during normal operation of the distribution network and calculate its mean data;

[0024] Calculate the cosine similarity between the secondary distribution network data of each distribution sub - area in the candidate fault sub - area set and the mean data;

[0025] If the cosine similarity is greater than the similarity threshold, it is determined as misjudged data and removed from the candidate fault sub - area set; otherwise, it is determined as non - misjudged data and retained.

[0026] Furthermore, determining the distribution network fault type includes:

[0027] Obtain the current waveform data before and after the fault and perform pre - processing;

[0028] Use the dynamic time warping method to compare the waveforms before and after the fault, and calculate the optimal alignment path and DTW distance;

[0029] Calculate the Pearson correlation coefficient of the current waveform to judge the waveform change trend;

[0030] Combine the DTW distance and the Pearson correlation coefficient to determine the distribution network fault type.

[0031] Furthermore, obtaining the coordinates of the fault distribution point includes:

[0032] Record the time difference of the traveling wave arrival between the fault distribution point and the fault sub - area, and calculate the coordinates of the fault distribution point in combination with the wave velocity.

[0033] Furthermore, the method further includes:

[0034] When an unknown fault is detected, trigger the incremental learning mechanism to update the global model parameters.

[0035] Furthermore, triggering the incremental learning mechanism to update the global model parameters includes:

[0036] Calculate the similarity between the current fault feature and the known fault library. If it is lower than the set threshold, it is determined as an unknown fault;

[0037] Cache the unknown fault data, and perform preliminary training at the edge node to learn new patterns, and upload the trained model parameters to the dispatching center through each edge node for global model update;

[0038] Adopt FedAvg for parameter aggregation, and ensure the compatibility of new and old fault patterns through the knowledge distillation and replay mechanism;

[0039] After confirming the new fault mode, add it to the fault database and deploy the updated model to the edge nodes.

[0040] The beneficial effects of the present invention are as follows:

[0041] 1. By constructing a two-layer star topology, the data stream is efficiently organized under the hierarchical architecture of the dispatching center - substation - distribution substation area, making fault detection more accurate and efficient. The multi-modal sensors in the substation are used to collect primary distribution network data, and preprocessing is carried out on the local server relying on edge computing, reducing data transmission delay and improving real-time response ability. Combining with the federated learning framework, direct transmission of raw data is avoided, which not only ensures data privacy and security but also makes full use of the computing power of each edge node to improve the generalization ability of the fault prediction model. Finally, this method can quickly locate the fault at the primary edge node, providing a basis for subsequent accurate fault location and improving the intelligent operation and maintenance level of the distribution network.

[0042] 2. By deploying multi-modal sensors in the distribution substation area, secondary distribution network data is finely collected, and time series alignment is carried out in combination with the primary distribution network data to construct a unified cross-layer feature matrix, thereby enhancing data integrity and temporal consistency. Using the pre-trained graph convolutional network to mine the topological correlation of the distribution network can effectively model the complex power network structure and improve the understanding ability of the fault propagation path. Finally, this method can accurately screen the candidate fault substation area set, reduce misjudgment and missed judgment, and lay a solid foundation for subsequent accurate fault location and type analysis, improving the intelligent level of distribution network fault diagnosis.

[0043] 3. By using cosine similarity verification to screen out misjudged data, the accuracy of fault recognition is improved. At the same time, combining current waveform correlation analysis and dynamic time warping to accurately match the fault mode, thereby accurately determining the type of distribution network fault. Further, using the traveling wave ranging technology, combined with the time difference of the fault signal propagation to calculate the fault point coordinates, high-precision fault location is realized. This method integrates a variety of intelligent analysis means, which can not only effectively reduce misjudgment and missed judgment but also improve the accuracy of fault recognition and location, providing strong support for the intelligent operation and maintenance and rapid emergency repair of the distribution network.

[0044] 4. To achieve the timeliness of fault type update, this method also detects unknown faults through similarity determination, triggers incremental learning, optimizes the global model through federated learning after preliminary training at the edge node, and uses the knowledge distillation and replay mechanism to ensure compatibility between new and old faults. Finally, the dynamic update of the model is realized, improving the accuracy and adaptability of distribution network fault recognition. Description of the Drawings

[0045] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the accompanying drawings:

[0046] Figure 1 is a flowchart of a method for rapid fault location of a distribution network based on artificial intelligence provided by the present invention;

[0047] Figure 2 is another flowchart of a method for rapid fault location of a distribution network based on artificial intelligence provided by the present invention. Detailed implementation manners

[0048] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0049] Embodiment 1

[0050] A method for rapid fault location of a distribution network based on artificial intelligence, as Figure 1 shown, includes:

[0051] S110: Establish a first-level star topology with the dispatching center as the first-level central node and the substations as the first-level edge nodes, and establish a second-level star topology with the substations as the second-level central nodes and the distribution substations as the second-level edge nodes;

[0052] S120: Collect first-level distribution network data through multi-modal sensors deployed in the substations. Establish a federated learning framework with the local servers of the substations as the edge nodes and the dispatching center as the federated aggregation center, and use the federated learning framework to predict the first-level distribution network data to obtain the fault first-level edge nodes;

[0053] Further, the federated learning framework includes edge nodes and a federated aggregation center. The edge nodes use the LSTM sub-model to train the first-level distribution network data, and the federated aggregation center aggregates the parameters of the LSTM sub-model and generates a global model.

[0054] Based on the distributed LSTM sub-model training, the federated aggregation center can integrate the information of each edge node, improve the generalization ability of the global model, make it more suitable for fault prediction in different scenarios, and thus enhance the intelligent diagnosis ability and overall operation stability of the distribution network system.

[0055] Further, the training of the federated learning framework includes:

[0056] The federated aggregation center initializes the global model and distributes the global model to the edge nodes;

[0057] The edge node uses the LSTM sub - model to train the primary distribution network data and uploads the trained model parameters to the federated aggregation center;

[0058] The federated aggregation center performs FedAvg aggregation on the model parameters, updates the global model and distributes it to the edge nodes.

[0059] Specifically, the federated aggregation center initializes a global LSTM model parameter , and distributes the initial global model parameter to each edge node; each edge node uses the local primary distribution network data to train the LSTM sub - model. After training, each edge node sends the updated model parameter to the federated aggregation center; the federated aggregation center performs FedAvg aggregation, and the aggregation formula can be expressed as:

[0060] ;

[0061] where, represents the global model, represents the local model parameter of the th edge node, represents the number of data samples used by this node, represents the total amount of data of all edge nodes participating in training; the federated aggregation center distributes the updated global model back to each edge node, and each edge node continues to use the local data to perform the next round of training based on the new global model, iterating the above process until the global model converges or reaches the preset number of training rounds.

[0062] The federated learning framework improves the intelligence and efficiency of distribution network fault detection through a distributed training method. The federated aggregation center initializes the global model and distributes it to the edge nodes, enabling each substation to independently train the LSTM sub - model using local data, thereby reducing data transmission, protecting privacy, and at the same time reducing the communication burden. The trained model parameters are aggregated by the FedAvg method, effectively integrating the feature information of multiple substations and enhancing the generalization ability and adaptability of the model. Finally, the optimized global model is distributed back to the edge nodes again to achieve continuous optimization and improve the accuracy and response speed of distribution network fault location.

[0063] S130: Collect secondary distribution network data through multi - modal sensors deployed in the distribution area, align the primary distribution network data and the secondary distribution network data in time series, and construct a cross - layer feature matrix; based on the cross - layer feature matrix, analyze the topological correlation through a pre - trained graph convolutional network to generate a set of candidate fault areas;

[0064] Furthermore, constructing a cross - level feature matrix includes:

[0065] Obtain primary distribution network data and secondary distribution network data;

[0066] Unify the timestamp formats of the primary distribution network data and the secondary distribution network data, find the common timestamps of the primary distribution network data and the secondary distribution network data, and use linear interpolation to complete the sparse data to align the two sets of data;

[0067] Merge the aligned data to obtain the cross - level feature matrix.

[0068] Specifically, the primary distribution network data is collected by multimodal sensors deployed in substations, including at least current, voltage, and traveling wave signals. The secondary distribution network data is collected by multimodal sensors deployed in distribution substations, including at least load information, current waveforms, and short - circuit faults. Both types of data are accompanied by timestamps; convert all timestamps to a unified format (such as Unix timestamp or ISO8601 format), extract all timestamp sets of the primary distribution network data and the secondary distribution network data to calculate the common timestamp set , and use linear interpolation to complete the sparse data to align the two sets of data. In a feasible embodiment, and are common timestamps, and are common timestamps, then is sparse data, and linear interpolation is performed on the data between and . The interpolation formula is:

[0069] ;

[0070] where and represent adjacent timestamps where the data exists, and represent the corresponding measured values, represents the interpolated measured value; merge the aligned data to obtain the cross - level feature matrix , where each column corresponds to all distribution network data at the same timestamp.

[0071] Through time alignment and data completion, ensure that the primary distribution network data and the secondary distribution network data are analyzed on the same time scale, improve data consistency and integrity. Using a unified timestamp format and linear interpolation to complete sparse data can avoid information deviation caused by data loss, thereby enhancing the accuracy of cross - level data fusion.

[0072] S140: Verify the candidate faulty substation area set through cosine similarity, eliminate misjudged data, and further determine the distribution network fault type by combining current waveform correlation analysis and dynamic time warping.

[0073] Further, the verification process of the cosine similarity includes:

[0074] Obtain the secondary distribution network data during normal operation of the distribution network and calculate its mean data.

[0075] Calculate the cosine similarity between the secondary distribution network data of each distribution substation area in the candidate faulty substation area set and the mean data.

[0076] If the cosine similarity is greater than the similarity threshold, it is determined as misjudged data and eliminated from the candidate faulty substation area set; otherwise, it is determined as non-misjudged data and retained.

[0077] Specifically, collect the secondary distribution network data of the distribution substation area under normal operation conditions to construct a historical normal data set , let be the secondary distribution network feature data of the th sampling, then the mean data . For each distribution substation area in the candidate faulty substation area set, obtain its current secondary distribution network data , and the process of calculating the cosine similarity is:

[0078] ;

[0079] where represents the cosine similarity, represents the dot product of two vectors, represents the Euclidean norm of two vectors; set the similarity threshold . If the cosine similarity is greater than the similarity threshold, it is determined as misjudged data and eliminated from the candidate faulty substation area set; otherwise, it is determined as non-misjudged data and retained.

[0080] Verify the candidate faulty substation area data using cosine similarity. By comparing with the historical mean data under normal operation conditions, misjudged data can be effectively identified and eliminated. By setting the similarity threshold, normal substations and truly faulty substations can be accurately distinguished, reducing the false alarm rate and improving the accuracy and reliability of fault location.

[0081] Further, determining the distribution network fault type includes:

[0082] Obtain the current waveform data before and after the fault and perform preprocessing.

[0083] Use the dynamic time warping method to compare the waveforms before and after the fault and calculate the optimal alignment path and DTW distance.

[0084] Calculate the Pearson correlation coefficient of the current waveform to determine the waveform change trend;

[0085] The distribution network fault type is determined by combining the DTW distance and the Pearson correlation coefficient.

[0086] Specifically, sensors deployed in the distribution area are used to obtain After the failure The current waveform data is obtained and data preprocessing is performed, including normalization, noise reduction and time window selection. The optimal alignment path between the pre-fault waveform and the post-fault waveform is calculated, and nonlinear time deformation is allowed to find the best match between the two waveforms and calculate the DTW distance:

[0087] ;

[0088] in, The current waveform data before the fault occurs is shown in The value at a time point, Indicates the current waveform data after the fault occurs. The Pearson correlation coefficient of the current waveform is calculated as:

[0089] ;

[0090] in, represents the Pearson correlation coefficient, represents the mean value of the waveform before the fault, Represents the mean of the waveform after the fault; combines the DTW distance and Pearson correlation coefficient, as well as the distance threshold and coefficient threshold , determine the distribution network fault type. In a feasible embodiment, the distribution network fault type determination rule is shown in Table 1.

[0091] Table 1. Rules for determining distribution network fault types

[0092]

[0093] By obtaining the current waveform data before and after the fault, and using dynamic time warping to calculate the optimal alignment path and DTW distance, the problem of inconsistent waveform time scales before and after the fault is overcome; at the same time, the Pearson correlation coefficient is introduced to analyze the waveform change trend and improve the ability to identify fault characteristics. Finally, combined with DTW distance and correlation analysis, the type of distribution network fault can be accurately determined, improving the accuracy and robustness of fault diagnosis.

[0094] Furthermore, obtaining the coordinates of the fault distribution point includes:

[0095] Record the time difference of the traveling wave arrival between the faulty distribution point and the faulty substation area, and calculate the coordinates of the faulty distribution point in combination with the wave velocity.

[0096] Utilize the traveling wave ranging technology to calculate the coordinates of the faulty distribution point by recording the time difference of the traveling wave arrival and combining with the wave velocity, realize the accurate and rapid positioning of the fault, and improve the timeliness and accuracy of the distribution network fault diagnosis.

[0097] S150: Obtain the coordinates of the faulty distribution point through traveling wave ranging to complete the distribution network fault positioning.

[0098] Embodiment 2

[0099] As another optional embodiment of the present application, refer to Figure 2 , which is a flowchart of Embodiment 2 of a method for rapid positioning of distribution network faults based on artificial intelligence provided by the present application. This embodiment is mainly an extended solution to the fault rapid positioning method described in the above Embodiment 1. As Figure 2 shown, the method includes:

[0100] S210: Establish a first-level star topology with the dispatching center as the first-level central node and the substations as the first-level edge nodes, and establish a second-level star topology with the substations as the second-level central nodes and the distribution substations as the second-level edge nodes;

[0101] S220: Collect first-level distribution network data through multimodal sensors deployed in the substations, establish a federated learning framework with the local servers of the substations as edge nodes and the dispatching center as the federated aggregation center, and use the federated learning framework to predict the first-level distribution network data to obtain faulty first-level edge nodes;

[0102] S230: Collect second-level distribution network data through multimodal sensors deployed in the distribution substations, align the first-level distribution network data and the second-level distribution network data in time series, and construct a cross-level feature matrix; based on the cross-level feature matrix, analyze the topological correlation through a pre-trained graph convolutional network to generate a set of candidate faulty substation areas;

[0103] S240: Verify the set of candidate faulty substation areas through cosine similarity and eliminate misjudged data, and further determine the distribution network fault type in combination with current waveform correlation analysis and dynamic time warping;

[0104] S250: Obtain the coordinates of the faulty distribution point through traveling wave ranging to complete the distribution network fault positioning;

[0105] S260: When an unknown fault is detected, trigger an incremental learning mechanism to update the global model parameters.

[0106] For the detailed processes of steps S210 - S250, reference can be made to the relevant descriptions of steps S110 - S150 in Embodiment 1, which will not be elaborated here.

[0107] Further, referring to Figure 2 S260 in

[0108] Calculate the similarity between the current fault feature and the known fault database. If it is lower than the set threshold, it is determined as an unknown fault.

[0109] Cache the unknown fault data, and conduct preliminary training at the edge node to learn new patterns, and upload the trained model parameters to the scheduling center through each edge node for global model update.

[0110] Use FedAvg for parameter aggregation, and ensure the compatibility of new and old fault patterns through knowledge distillation and replay mechanisms.

[0111] After confirming the new fault pattern, add it to the fault database and deploy the updated model to the edge node.

[0112] By calculating the similarity between the current fault feature and the known fault database, unknown faults are automatically identified and incremental learning is triggered. Through preliminary training at the edge node and uploading model parameters to the scheduling center, global model update is performed using FedAvg to ensure the compatibility of new and old fault patterns. Knowledge distillation and replay mechanisms are introduced to optimize the stability and accuracy of the model. Finally, the new fault pattern is added to the database and updated to the edge node to improve the intelligence and adaptability of distribution network fault diagnosis.

[0113] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A rapid fault location method for distribution network based on artificial intelligence, characterized in that, Including: Establish a first-level star topology with the dispatching center as the first-level central node and substations as the first-level edge nodes, and establish a second-level star topology with substations as the second-level central nodes and distribution transformer areas as the second-level edge nodes; Collect first-level distribution network data through multimodal sensors deployed in substations, establish a federated learning framework with the local server of the substation as the edge node and the dispatching center as the federated aggregation center, and use the federated learning framework to predict the first-level distribution network data to obtain faulty first-level edge nodes; Collect second-level distribution network data through multimodal sensors deployed in distribution transformer areas, align the first-level distribution network data and the second-level distribution network data in time series, and construct a cross-level feature matrix; Based on the cross-level feature matrix, analyze the topological correlation through a pre-trained graph convolutional network to generate a set of candidate faulty transformer areas; Verify the set of candidate faulty transformer areas through cosine similarity and eliminate misjudged data, and further determine the distribution network fault type by combining current waveform correlation analysis and dynamic time warping; Obtain the coordinates of the faulty distribution network point through traveling wave ranging to complete the distribution network fault location; Among them, constructing the cross-level feature matrix includes: Obtain the first-level distribution network data and the second-level distribution network data; Unify the timestamp formats of the first-level distribution network data and the second-level distribution network data, find the common timestamps of the first-level distribution network data and the second-level distribution network data, and use linear interpolation to complement sparse data to align the two sets of data; Merge the aligned data to obtain the cross-level feature matrix.

2. The fast fault location method for distribution network based on artificial intelligence according to claim 1, wherein, The federated learning framework includes edge nodes and a federated aggregation center. The edge nodes use the LSTM sub-model to train the first-level distribution network data, and the federated aggregation center aggregates the parameters of the LSTM sub-model and generates a global model.

3. The rapid fault location method for distribution network based on artificial intelligence according to claim 1, characterized in that, The training of the federated learning framework includes: The federated aggregation center initializes the global model and distributes the global model to the edge nodes; The edge nodes use the LSTM sub-model to train the first-level distribution network data and upload the trained model parameters to the federated aggregation center; The federated aggregation center performs FedAvg aggregation on the model parameters, updates the global model and distributes it to the edge nodes.

4. A method for quickly locating distribution network faults based on artificial intelligence according to claim 1, characterized in that, The verification process of the cosine similarity includes: Obtain the second-level distribution network data during normal operation of the distribution network and calculate its mean data; Calculate the cosine similarity between the second-level distribution network data of each distribution transformer area in the set of candidate faulty transformer areas and the mean data; If the cosine similarity is greater than the similarity threshold, it is determined as misjudged data and eliminated from the set of candidate faulty transformer areas; otherwise, it is determined as non-misjudged data and retained.

5. A rapid fault location method for distribution network based on artificial intelligence according to claim 1, characterized in that, Determining the distribution network fault type includes: Obtain the current waveform data before and after the fault and perform preprocessing; Use the dynamic time warping method to compare the waveforms before and after the fault, and calculate the optimal alignment path and DTW distance; Calculate the Pearson correlation coefficient of the current waveform to judge the waveform change trend; Combine the DTW distance and the Pearson correlation coefficient to determine the distribution network fault type.

6. The rapid fault location method for distribution network based on artificial intelligence according to claim 1, characterized in that Obtaining the coordinates of the faulty distribution network point includes: Record the time difference of the traveling wave arrival between the faulty distribution network point and the faulty transformer area, and calculate the coordinates of the faulty distribution network point in combination with the wave velocity.

7. A method for quickly locating distribution network faults based on artificial intelligence according to claim 1, characterized in that, The method further includes: When an unknown fault is detected, trigger the incremental learning mechanism to update the global model parameters.

8. A rapid fault location method for distribution network based on artificial intelligence according to claim 1, characterized in that Triggering the incremental learning mechanism to update the global model parameters includes: Calculate the similarity between the current fault feature and the known fault library. If it is lower than the set threshold, it is determined as an unknown fault; Cache the unknown fault data, and perform preliminary training at the edge node to learn the new pattern, and upload the trained model parameters to the scheduling center through each edge node for global model update; Use FedAvg for parameter aggregation, and ensure the compatibility of the new and old fault patterns through the knowledge distillation and replay mechanisms; After confirming the new fault pattern, add it to the fault database and deploy the updated model to the edge node.

Images