A method and system for locating power transmission line faults based on transfer learning

Abstract

The present invention provides a transmission line fault location method and system based on transfer learning. By collecting real-time monitoring data of the target transmission line, combining meteorological information and satellite remote sensing images, analyzing changes in the surrounding environment and performing anomaly detection, an abnormal data set is obtained, and a comprehensive fault feature space is constructed; an adversarial domain adaptation algorithm is used to optimize the comprehensive fault feature space, and a long short-term memory network, an attention mechanism, and a graph convolutional network are used for data processing to generate an enhanced fault feature representation; an integrated learning framework based on Bayesian optimization is constructed, and a plurality of preliminary basic models are used to predict the target transmission line, and all prediction results are integrated through an adaptive weighting strategy to generate an initial fault probability distribution; and geographic information system technology and a topological sorting algorithm are used to determine fault location information. The present invention provides an efficient and accurate transmission line fault location method, which provides strong support for the safe operation of the power system.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of power systems, and in particular to a method and system for locating power transmission line faults based on transfer learning. Background Art

[0002] As the power system becomes increasingly complex, the operation status of transmission lines, as an important part of the power system, is directly related to the safety and stability of the entire power grid. In severe weather or unexpected situations, transmission lines may suffer from short circuits, disconnections and other faults, which will not only cause large-scale power outages and affect people's normal lives, but also damage power equipment and increase maintenance costs. Therefore, quickly and accurately locating the fault location of the transmission line is of great significance for timely restoration of power supply and reduction of economic losses.

[0003] At present, the fault location of transmission lines mainly relies on traditional technologies such as the traveling wave method, the impedance method and the signal injection method. The traveling wave method uses the propagation characteristics of the transient voltage or current waveform generated by the fault to determine the fault distance; the impedance method calculates the equivalent impedance by measuring the voltage and current at both ends of the line, and then deduce the location of the fault point; and the signal injection method injects a test signal of a specific frequency into the line and determines the fault location by analyzing the response signal. The traveling wave method and the impedance method are sensitive to changes in line parameters. Factors such as temperature, humidity, and line aging will affect the accuracy of the waveform and impedance value, resulting in increased positioning errors; on long-distance transmission lines, due to the huge amount of data, traditional methods require a lot of calculations, which takes a long time and is difficult to meet real-time requirements; traditional methods are usually designed for specific types of faults. When encountering different types of faults or new fault modes, the algorithm parameters often need to be readjusted, lacking sufficient flexibility and adaptability. Summary of the invention

[0004] The embodiment of the present invention provides a transmission line fault location method and system based on transfer learning, which is used to solve the problems in the prior art of large interference from environmental factors, large positioning errors, difficulty in meeting real-time requirements, and lack of sufficient flexibility and adaptability.

[0005] In a first aspect, an embodiment of the present invention provides a method for locating a transmission line fault based on transfer learning, comprising:

[0006] Collect real-time monitoring data of the target transmission line, combine it with meteorological information and satellite remote sensing images, use deep neural network and image recognition technology to analyze environmental changes around the target transmission line, obtain environmental change information, use variational autoencoder to perform anomaly detection on the environmental change information, obtain anomaly data set, and construct a comprehensive fault feature space based on the anomaly data set;

[0007] Based on the comprehensive fault feature space, an adversarial domain adaptation algorithm is used to reduce the distribution gap between the existing fault data set and the abnormal data set to obtain an optimized fault feature space;

[0008] Based on the optimized fault feature space, the trained long short-term memory network is used to analyze the long-term dependency between the fault features in the abnormal data set, and the key features are selected from the fault features in combination with the attention mechanism algorithm, and the influence of the network structure of the target transmission line is analyzed by using the graph convolution network to obtain the structural influence result, and based on the long-term dependency, the key features and the structural influence result, an enhanced fault feature representation is generated;

[0009] Constructing an integrated learning framework based on Bayesian optimization, using the enhanced fault feature representation to train the basic learner in the integrated learning framework to obtain multiple preliminary basic models, using each of the preliminary basic models to predict the target transmission line respectively, obtaining corresponding prediction results, integrating the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy, and generating an initial fault probability distribution;

[0010] The fault location information of the target transmission line is determined based on the initial fault probability distribution by using geographic information system technology and a topological sorting algorithm.

[0011] Optionally, the determining the fault location information of the target transmission line based on the initial fault probability distribution by using geographic information system technology and a topological sorting algorithm includes:

[0012] Analyzing the network structure of the target transmission line by using a topological sorting algorithm to determine the topological relationship between nodes in the network structure and obtain a target topological sorting result, wherein the topological sorting algorithm includes a Kahn algorithm;

[0013] Optimizing the initial fault probability distribution based on the target topological sorting result to obtain a target fault probability distribution to narrow the scope of fault location;

[0014] By using geographic information system technology, the geographical location coordinates corresponding to the target probability value in the target fault probability distribution in the target transmission line map are determined as the fault location information of the target transmission line, and the target probability value is a fault probability value higher than a preset probability value.

[0015] Optionally, the network structure of the target power transmission line is analyzed by using a topological sorting algorithm to determine the topological relationship between nodes in the network structure to obtain a target topological sorting result, and the topological sorting algorithm includes a Kahn algorithm, including:

[0016] Applying a graph traversal algorithm to analyze the network structure of the target transmission line, extracting information of all nodes and line segments, and obtaining an information list of nodes and line segments, wherein the nodes represent network connection points in the target transmission line, and the nodes include transmission towers, substations, and switch stations, and the line segments represent the transmission line section connecting two nodes, and the line segments include a section of wire or cable;

[0017] Based on the information list of the nodes and line segments, a network graph model is constructed;

[0018] Based on the network graph model, the nodes in the network structure are topologically sorted using the Kahn algorithm to obtain an initial topological sorting result, and the initial topological sorting result is verified. If the verification passes, an intermediate topological sorting result is obtained, and the condition for passing the verification is that the in-degree values ​​of all nodes are 0;

[0019] Matching the intermediate topological sorting result with the network structure of the target transmission line so that the node order in the intermediate topological sorting result is consistent with the topological relationship between the nodes in the network structure; after the matching is successful, clustering the nodes in the network structure based on the intermediate topological sorting result to obtain a clustering result;

[0020] The intermediate topological sorting result is optimized in combination with the clustering result to generate a target topological sorting result.

[0021] Optionally, based on the network graph model, the nodes in the network structure are topologically sorted using the Kahn algorithm to obtain an initial topological sorting result, and the initial topological sorting result is verified. An intermediate topological sorting result is obtained after the verification is passed, and the condition for passing the verification is that the in-degree values ​​of all nodes are 0, including:

[0022] Traversing the edges in the network graph model, counting the in-degree value of each node in the edge, adding all nodes with an in-degree value of 0 to a queue, and obtaining an initialized in-degree table and a queue, wherein the in-degree value represents the number of edges pointing to the node, and the node with an in-degree value of 0 is used as the starting point of the topological sorting;

[0023] Based on the initialized in-degree table and queue, using the Kahn algorithm, a node is randomly extracted as the target node, and the in-degree value of the neighbor node of the target node is reduced. If the in-degree value of the neighbor node is 0, the neighbor node is added to the initialized in-degree table and queue, and the operation of reducing the in-degree value of the neighbor node is repeated until the initialized in-degree table and queue are empty, and an initial topological sorting result is generated. The initial topological sorting result includes the target node and all neighbor nodes with in-degree values ​​of 0. The initialization in-degree table and queue are empty, indicating that the in-degree values ​​of all nodes are 0;

[0024] Based on the initial topological sorting result, verify whether the topological sorting process is finished. If the in-degree values ​​of all nodes in the topological sorting result are 0, it is determined that the topological sorting process is finished. If the initialized in-degree table and queue are empty, but there are nodes with in-degree values ​​not equal to 0, it is determined that there is a loop in the network graph model.

[0025] When a loop exists in the network graph model, a loop detection algorithm is used to confirm the location of the loop, and corresponding measures are taken to eliminate the loop to obtain a processed network graph model. Based on the processed network graph model, the Kahn algorithm is repeatedly executed until the in-degree values ​​of all nodes are 0, thereby obtaining an intermediate topological sorting result.

[0026] Optionally, the loop detection algorithm is used to confirm the location of the loop, and corresponding measures are taken to eliminate the loop to obtain a processed network graph model. Based on the processed network graph model, the Kahn algorithm is repeatedly executed until the in-degree values ​​of all nodes are 0, and an intermediate topological sorting result is obtained, including:

[0027] Based on the network graph model, selecting a depth-first search algorithm in a cycle detection algorithm;

[0028] Based on the depth-first search algorithm, an initial node is randomly selected from the network graph model, and a depth-first search is performed starting from the initial node. The access status of each node in the depth-first search process is recorded. If a node is detected to have an access status of being accessed, it is determined that a loop exists, and location information of the loop is generated. The access status includes not accessed, being accessed, and visited.

[0029] Based on the location information of the loop, taking corresponding measures to disconnect the loop to obtain a processed network graph model, wherein the corresponding measures include deleting edges or nodes in the loop and changing the direction of edges in the loop;

[0030] Based on the processed network graph model, the Kahn algorithm is repeatedly executed until the in-degree values ​​of all nodes are 0, and the intermediate topological sorting result is obtained.

[0031] Optionally, the construction of an integrated learning framework based on Bayesian optimization, using the enhanced fault feature representation to train a basic learner in the integrated learning framework to obtain a plurality of preliminary basic models, using each of the preliminary basic models to predict the target transmission line respectively to obtain a corresponding prediction result, integrating the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy to generate an initial fault probability distribution, including:

[0032] Based on the enhanced fault feature representation, an initial integrated learning framework is selected, and the initial integrated learning framework is optimized using a Bayesian optimization method to obtain an integrated learning framework based on Bayesian optimization;

[0033] Dividing the enhanced fault feature representation into a training set and a validation set, and using the training set to train multiple basic learners in the Bayesian optimization-based integrated learning framework to obtain multiple preliminary basic models;

[0034] Inputting the enhanced fault feature representation into the preliminary basic model to obtain a plurality of prediction results corresponding to the preliminary basic model, wherein the prediction results represent the probability of the target transmission line failing;

[0035] Using an adaptive weighting strategy, the weight of each preliminary basic model is calculated according to the performance of the preliminary basic model on the validation set, the weights of all the preliminary basic models are multiplied by the corresponding prediction results to obtain multiple product results, the multiple product results are added together to obtain a final prediction result, and based on the final prediction result, an initial fault probability distribution is generated.

[0036] Optionally, based on the optimized fault feature space, the trained long short-term memory network is used to analyze the long-term dependency between the fault features in the abnormal data set, and the key features are selected from the fault features in combination with the attention mechanism algorithm, and the graph convolution network is used to analyze the influence of the network structure of the target transmission line to obtain the structural influence result, and based on the long-term dependency, the key features and the structural influence result, the enhanced fault feature representation is generated, including:

[0037] Based on the optimized fault feature space, a long short-term memory network model is constructed, and the long short-term memory network model is trained using the time series data in the optimized fault feature space to obtain a trained long short-term memory network model, and the long short-term memory network model is used to analyze the long-term dependency between the fault features in the abnormal data set;

[0038] Based on the long-term dependency, an attention mechanism is introduced to calculate the importance weight of each feature in the trained long short-term memory network model at different time steps, so as to select key features from the fault features;

[0039] Constructing a graph convolutional network model, and using the graph convolutional network model to analyze the impact of the network structure of the target transmission line on the fault characteristics, and obtaining a structural impact result;

[0040] The long-term dependencies, the key features, and the structural impact results are integrated to generate an enhanced fault feature representation.

[0041] In a second aspect, an embodiment of the present invention provides a transmission line fault location system based on transfer learning, comprising:

[0042] An acquisition module is used to collect real-time monitoring data of a target transmission line, and analyze the environmental changes around the target transmission line using deep neural network and image recognition technology in combination with meteorological information and satellite remote sensing images to obtain environmental change information, and perform anomaly detection on the environmental change information using a variational autoencoder to obtain an abnormal data set, and construct a comprehensive fault feature space based on the abnormal data set;

[0043] An optimization module, for reducing the distribution gap between the existing fault data set and the abnormal data set by using an adversarial domain adaptation algorithm based on a comprehensive fault feature space to obtain an optimized fault feature space;

[0044] A generation module is used to analyze the long-term dependency between the fault features in the abnormal data set based on the optimized fault feature space by using the trained long short-term memory network, select key features from the fault features in combination with the attention mechanism algorithm, analyze the influence of the network structure of the target transmission line by using the graph convolution network, obtain the structural influence result, and generate an enhanced fault feature representation based on the long-term dependency, the key features and the structural influence result;

[0045] A construction module is used to construct an integrated learning framework based on Bayesian optimization, train the basic learner in the integrated learning framework using the enhanced fault feature representation to obtain multiple preliminary basic models, use each of the preliminary basic models to predict the target transmission line respectively, obtain corresponding prediction results, integrate the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy, and generate an initial fault probability distribution;

[0046] The determination module is used to determine the fault location information of the target transmission line based on the initial fault probability distribution by using geographic information system technology and topological sorting algorithm.

[0047] In a third aspect, an embodiment of the present invention provides a computing device, comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a transmission line fault location method based on transfer learning as described in any one of the first aspects.

[0048] In a fourth aspect, an embodiment of the present invention provides a computer storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement a transmission line fault location method based on transfer learning as described in any one of the first aspects.

[0049] In an embodiment of the present invention, real-time monitoring data of a target transmission line is collected, and combined with meteorological information and satellite remote sensing images, a deep neural network and image recognition technology are used to analyze environmental changes around the target transmission line to obtain environmental change information, and a variational autoencoder is used to perform anomaly detection on the environmental change information to obtain an abnormal data set, and a comprehensive fault feature space is constructed based on the abnormal data set; based on the comprehensive fault feature space, an adversarial domain adaptation algorithm is used to reduce the distribution gap between the existing fault data set and the abnormal data set to obtain an optimized fault feature space; based on the optimized fault feature space, a trained long short-term memory network is used to analyze the long-term dependency between fault features in the abnormal data set, and an attention mechanism algorithm is used to select key features from the fault features, and the optimization of the fault feature space is realized by using the training data set. Use a graph convolutional network to analyze the impact of the network structure of the target transmission line, obtain the structural impact results, and generate an enhanced fault feature representation based on the long-term dependency, the key features and the structural impact results; construct an integrated learning framework based on Bayesian optimization, use the enhanced fault feature representation to train the basic learner in the integrated learning framework, and obtain multiple preliminary basic models, use each of the preliminary basic models to predict the target transmission line respectively, and obtain the corresponding prediction results, and integrate the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy to generate an initial fault probability distribution; use geographic information system technology and topological sorting algorithm to determine the fault location information of the target transmission line based on the initial fault probability distribution. The technical solution provided by the present invention provides an efficient, accurate and reliable transmission line fault location method, which can accurately provide specific fault location information and provide strong support for the safe operation of the power system.

[0050] These and other aspects of the present invention will become more apparent from the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 A flow chart of a transmission line fault location method based on transfer learning provided by an embodiment of the present invention;

[0053] Figure 2 A schematic diagram of the structure of a transmission line fault location system based on transfer learning provided by an embodiment of the present invention;

[0054] Figure 3 A schematic diagram of the structure of a computing device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0056] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0058] Figure 1 A flow chart of a method for locating a power transmission line fault based on transfer learning is provided in an embodiment of the present invention, such as Figure 1 As shown, the method includes:

[0059] Based on this, the present invention provides a transmission line fault location method based on transfer learning, such as Figure 1 ,include:

[0060] Step 101: collect real-time monitoring data of the target transmission line, combine meteorological information and satellite remote sensing images, use deep neural network and image recognition technology to analyze environmental changes around the target transmission line, obtain environmental change information, use variational autoencoder to perform anomaly detection on the environmental change information, obtain anomaly data set, and construct a comprehensive fault feature space based on the anomaly data set;

[0061] In this step, real-time monitoring data refers to sensor data including current, voltage, temperature, humidity, etc. of the transmission line, which is used to monitor the status of the transmission line in real time. Meteorological information refers to meteorological data including temperature, humidity, wind speed, rainfall, etc., which is used to analyze the impact of weather on the transmission line. Satellite remote sensing images refer to high-resolution images taken by satellites, which are used to analyze changes in the environment around the transmission line, such as vegetation growth, terrain changes, etc. Deep neural network refers to processing and analyzing multi-source data to extract useful features. Image recognition technology refers to analyzing satellite remote sensing images to identify potential environmental changes. Variational autoencoder (VAE) refers to detecting anomalies in environmental change information and identifying abnormal data. Anomaly data set refers to data points that are identified as abnormal, which may indicate potential faults. Comprehensive fault feature space refers to constructing a space containing multiple fault features by extracting and analyzing features of abnormal data sets.

[0062] This step collects real-time monitoring data from sensors installed along the transmission lines, obtains meteorological information from meteorological stations, and obtains remote sensing images from satellites; uses deep neural networks to analyze real-time monitoring data and meteorological information and extract features; uses image recognition technology to analyze satellite remote sensing images and identify environmental changes; uses variational autoencoders to perform anomaly detection on the extracted features, identifies abnormal data points, and forms an abnormal data set; based on the abnormal data set, extracts multiple fault features, such as temperature changes, current fluctuations, vegetation coverage, etc., to construct a comprehensive fault feature space.

[0063] Step 102: Based on the comprehensive fault feature space, an adversarial domain adaptation algorithm is used to reduce the distribution gap between the existing fault data set and the abnormal data set to obtain an optimized fault feature space;

[0064] In this step, the adversarial domain adaptation algorithm is used to reduce the distribution gap between the existing fault data set and the newly collected abnormal data set, and improve the generalization ability of the model. The existing fault data set refers to the historical fault data used for training the model.

[0065] This step collects historical fault data to form an existing fault data set; uses adversarial domain adaptation algorithms, such as DANN and MCD, to process the existing fault data set and the abnormal data set to reduce the distribution gap between the two; generates an optimized fault feature space for subsequent feature analysis and model training.

[0066] Step 103: Based on the optimized fault feature space, the trained long short-term memory network is used to analyze the long-term dependency between the fault features in the abnormal data set, and the key features are selected from the fault features in combination with the attention mechanism algorithm, and the influence of the network structure of the target transmission line is analyzed by using a graph convolutional network to obtain a structural influence result, and based on the long-term dependency, the key features and the structural influence result, an enhanced fault feature representation is generated;

[0067] In this step, the long short-term memory network (LSTM) refers to the method used to analyze the long-term dependencies in time series data and capture the time evolution of fault features. The attention mechanism algorithm refers to the method used to select key features from fault features and improve the accuracy of feature selection. The graph convolutional network (GCN) refers to the method used to analyze the influence of the transmission line network structure and consider the influence of network topology on fault propagation. The enhanced fault feature representation refers to the method that combines the long-term dependencies, key features and network structure influence results to generate a richer fault feature representation.

[0068] In this step, the long short-term memory network is used to analyze the time series data in the optimized fault feature space to capture the long-term dependencies of the fault features. The attention mechanism algorithm is combined to select key features from the features extracted by the long short-term memory network to improve the accuracy of feature selection. The graph convolutional network is used to analyze the network structure of the transmission line, considering the impact of network topology on fault propagation and generating structural impact results. The long-term dependencies extracted by the long short-term memory network, the key features selected by the attention mechanism and the structural impact results generated by the graph convolutional network are combined to generate an enhanced fault feature representation.

[0069] Step 104: construct an integrated learning framework based on Bayesian optimization, use the enhanced fault feature representation to train the basic learner in the integrated learning framework to obtain multiple preliminary basic models, use each of the preliminary basic models to predict the target transmission line respectively, obtain corresponding prediction results, integrate the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy, and generate an initial fault probability distribution;

[0070] In this step, Bayesian optimization refers to optimizing the hyperparameters in the ensemble learning framework to improve the performance of the model. The ensemble learning framework refers to combining multiple basic learners, such as decision trees, random forests, and SVMs, to improve the stability and accuracy of the model. The adaptive weighting strategy refers to dynamically adjusting the weights of different basic models to optimize the final prediction results.

[0071] In this step, an integrated learning framework based on Bayesian optimization is constructed, and multiple basic learners are selected, such as decision trees, random forests, and SVMs. The basic learners are trained using enhanced fault feature representations to generate multiple preliminary basic models. Each preliminary basic model is used to predict the target transmission line to obtain the corresponding prediction results. The prediction results of all preliminary basic models are integrated through an adaptive weighting strategy to generate an initial fault probability distribution.

[0072] Step 105: Determine the fault location information of the target transmission line based on the initial fault probability distribution by using geographic information system technology and topological sorting algorithm;

[0073] In this step, geographic information system technology (GIS) refers to the technology used to convert the fault probability distribution into specific geographic location coordinates to provide intuitive fault location information. Topological sorting algorithm refers to a graph theory algorithm used to sort the nodes in a directed acyclic graph (DAG) so that for each directed edge (u, v), node u appears before node v. Fault location information refers to the final determined fault location information, including specific geographic location coordinates.

[0074] In this step, the Kahn algorithm is used to topologically sort the network structure of the transmission line, determine the topological relationship between nodes, and obtain the topological sorting result; based on the topological sorting result, the initial fault probability distribution is optimized to narrow the scope of fault location; based on the optimized fault probability distribution, geographic information system technology is used to determine the specific fault location information, including geographic location coordinates.

[0075] The embodiments of the present invention can achieve the following beneficial effects through the above steps:

[0076] The accuracy and reliability of fault detection have been significantly improved through technologies such as multi-source data fusion, anomaly detection, domain adaptation and feature optimization;

[0077] Combining long short-term memory networks, attention mechanisms, graph convolutional networks, and ensemble learning methods, the accuracy and robustness of fault location are effectively improved;

[0078] The use of geographic information system technology and topological sorting algorithms can provide specific fault location information, making it easier for operation and maintenance personnel to quickly respond to and handle faults.

[0079] The present invention provides a specific embodiment, wherein the step 105 uses geographic information system technology and a topological sorting algorithm to determine the fault location information of the target transmission line based on the initial fault probability distribution, and specifically includes the following steps:

[0080] Step 501: Analyze the network structure of the target transmission line using a topological sorting algorithm to determine the topological relationship between nodes in the network structure and obtain a target topological sorting result, wherein the topological sorting algorithm includes a Kahn algorithm;

[0081] In this step, the topological relationship refers to the connection relationship between nodes and describes the structure of the network.

[0082] This step executes the Kahn algorithm to obtain the topological sorting result, matches the topological sorting result with the network structure of the transmission line, ensures that the node order in the topological sorting result is consistent with the topological relationship between the nodes in the network structure; clusters the nodes in the network structure, optimizes the topological sorting result, and generates the target topological sorting result.

[0083] Step 502: Optimizing the initial fault probability distribution based on the target topological sorting result to obtain a target fault probability distribution to narrow the scope of fault location;

[0084] This step optimizes and adjusts the initial fault probability distribution based on the target topological sorting result to generate the target fault probability distribution to narrow the scope of fault location;

[0085] Step 503: using geographic information system technology, determining the geographical location coordinates corresponding to the target probability value in the target fault probability distribution in the target transmission line map as the fault location information of the target transmission line, wherein the target probability value is a fault probability value higher than a preset probability value;

[0086] This step uses geographic information system technology to correspond the probability values ​​in the target fault probability distribution to the geographical coordinates of the transmission line; preset a probability threshold, such as 0.8, to filter out the target probability values ​​above the threshold; and determine the geographical location coordinates corresponding to the filtered target probability values ​​as fault location information.

[0087] The embodiments of the present invention can achieve the following beneficial effects through the above steps:

[0088] The network structure of the transmission line is analyzed through the topological sorting algorithm, the topological relationship between nodes is determined, the fault probability distribution is optimized, the scope of fault location is reduced, and the accuracy of fault location is improved;

[0089] Combined with geographic information system technology, the fault probability distribution is converted into specific geographical location coordinates, providing intuitive fault location information and enhancing the reliability of fault detection;

[0090] Through multi-step optimization and screening, only high-probability fault points are marked as fault locations, reducing the possibility of false positives and false negatives;

[0091] It provides accurate fault location information to help operation and maintenance personnel quickly locate and repair faults, improves the operation and maintenance efficiency of the power system, and ensures the safe operation of the power system.

[0092] The present invention provides a specific embodiment, in which step 501 uses a topological sorting algorithm to analyze the network structure of the target transmission line to determine the topological relationship between nodes in the network structure and obtain a target topological sorting result, wherein the topological sorting algorithm includes a Kahn algorithm, and specifically includes the following steps:

[0093] Step 511: applying a graph traversal algorithm to analyze the network structure of the target transmission line, extracting information of all nodes and line segments, and obtaining a list of information of nodes and line segments, wherein the nodes represent network connection points in the target transmission line, and the nodes include transmission towers, substations, and switch stations, and the line segments represent the transmission line portion connecting two nodes, and the line segments include a section of wire or cable;

[0094] In this step, the graph traversal algorithm refers to an algorithm used to traverse all nodes in the graph, common ones are depth-first search (DFS) and breadth-first search (BFS).

[0095] This step extracts detailed information of all nodes, such as transmission towers, substations, and switch stations, from the database of the transmission line, including node ID, location coordinates, type, etc.; extracts detailed information of all line segments, such as wires or cables, including line segment ID, starting node ID, end node ID, length, etc.; generates information lists, including node information lists and line segment information lists, node information list, [node ID, location coordinates, type], line segment information list, [line segment ID, starting node ID, end node ID, length]; selects depth-first search (DFS) or breadth-first search (BFS) algorithm; starting from any node, traverses the entire network structure using the selected graph traversal algorithm to ensure that all nodes and line segments are visited; records the order of nodes and line segments visited during the traversal process, and generates information lists of nodes and line segments.

[0096] Step 512: constructing a network graph model based on the information list of the nodes and line segments;

[0097] In this step, the network graph model refers to representing the network structure of the transmission line in the form of a graph, wherein a node represents a connection point of the transmission line, and an edge represents a line segment connecting two nodes.

[0098] In this step, an object of each node is created according to the node information list, including properties such as node ID, location coordinates, and type; all node objects are stored in a collection for the convenience of subsequent operations; an object of each line segment is created according to the line segment information list, including properties such as line segment ID, starting node ID, end node ID, and length; all line segment objects are stored in a collection for the convenience of subsequent operations; a network graph model is constructed based on node objects and line segment objects; for example, an adjacency list is used to represent the network graph model, in which each node corresponds to a list containing all edges directly connected to the node; an adjacency matrix is ​​used to represent the network graph model, in which the rows and columns of the matrix represent nodes respectively, and the elements in the matrix represent the connection relationship between nodes and the weight of the edges.

[0099] Step 513: Based on the network graph model, use the Kahn algorithm to topologically sort the nodes in the network structure to obtain an initial topological sorting result, verify the initial topological sorting result, and obtain an intermediate topological sorting result if the verification passes. The condition for passing the verification is that the in-degree values ​​of all nodes are 0;

[0100] In this step, Kahn's algorithm refers to a topological sorting algorithm used to sort nodes in a directed acyclic graph (DAG). The node in-degree value refers to the number of edges pointing to a node.

[0101] This step uses the Kahn algorithm to topologically sort the nodes in the network structure, obtains the initial topological sorting result, and checks whether the in-degree values ​​of all nodes in the topological sorting result are 0. If so, the topological sorting is successful and the initial topological sorting result is obtained; otherwise, there is a loop in the network graph model, and a loop detection algorithm, such as a depth-first search algorithm, is required to confirm the loop location and eliminate the loop, and re-execute the Kahn algorithm until the in-degree values ​​of all nodes are 0, and the intermediate topological sorting result is obtained.

[0102] Step 514: matching the intermediate topological sorting result with the network structure of the target transmission line so that the node order in the intermediate topological sorting result is consistent with the topological relationship between the nodes in the network structure; after the matching is successful, clustering the nodes in the network structure based on the intermediate topological sorting result to obtain a clustering result;

[0103] This step matches the node order in the intermediate topological sorting result with the topological relationship between the nodes in the actual network structure to ensure that the node order in the sorting result is consistent with the topological relationship between the nodes in the network structure; if the match is successful, proceed to the next step; if the match fails, re-execute the Kahn algorithm to ensure the correctness of the topological sorting result; select a suitable clustering algorithm, such as K-means, DBSCAN, etc.; based on the intermediate topological sorting result, cluster the nodes in the network structure to obtain the clustering result;

[0104] Step 515: optimizing the intermediate topological sorting result in combination with the clustering result to generate a target topological sorting result;

[0105] This step combines the clustering results with the intermediate topological sorting results to optimize the sorting order of the nodes; according to the clustering results, adjust the node order in the intermediate topological sorting results to ensure that the node order within the same clustering group is reasonable; generate the final target topological sorting results to ensure that the node order is consistent with the topological relationship in the network structure, and the optimized sorting results are more reasonable.

[0106] The embodiments of the present invention can achieve the following beneficial effects through the above steps:

[0107] The network structure is topologically sorted using the Kahn algorithm to ensure that the node order is consistent with the topological relationship in the network structure, thus improving the accuracy of fault location.

[0108] The topological sorting results are optimized in combination with the clustering results to ensure that the order of nodes in the same clustering group is reasonable, thus enhancing the reliability of fault detection;

[0109] Through multi-step verification and optimization, the correctness of the topological sorting results is ensured, reducing the possibility of false positives and false negatives;

[0110] It provides accurate topological sorting results to help operation and maintenance personnel quickly locate and repair faults, improve the operation and maintenance efficiency of the power system, and ensure the safe operation of the power system.

[0111] The present invention provides a specific embodiment, wherein the step 513, based on the network graph model, uses the Kahn algorithm to perform topological sorting on the nodes in the network structure to obtain an initial topological sorting result, verifies the initial topological sorting result, and obtains an intermediate topological sorting result after the verification, wherein the condition for passing the verification is that the in-degree values ​​of all nodes are 0, specifically comprises the following steps:

[0112] Step 521: traverse the edges in the network graph model, count the in-degree value of each node in the edge, add all nodes with an in-degree value of 0 to the queue, and obtain an initialized in-degree table and queue, wherein the in-degree value represents the number of edges pointing to the node, and the node with an in-degree value of 0 is used as the starting point of the topological sorting;

[0113] In this step, the queue refers to a first-in-first-out (FIFO) data structure used to store nodes with in-degree value 0.

[0114] This step creates a dictionary with the key as the node ID and the value as the in-degree value of the node, and the initial value is 0; traverse each edge in the network graph model, and for each edge (a, b), add 1 to the in-degree value of node c; after the traversal is completed, the in-degree value table records the in-degree value of each node; create an empty queue; traverse the in-degree value table, add all nodes with in-degree values ​​of 0 to the queue, and obtain the initialized in-degree table and queue.

[0115] Step 522: Based on the initialized in-degree table and queue, use the Kahn algorithm to randomly extract a node as the target node, reduce the in-degree value of the neighbor node of the target node, and if the in-degree value of the neighbor node is 0, add the neighbor node to the initialized in-degree table and queue, and repeat the operation of reducing the in-degree value of the neighbor node until the initialized in-degree table and queue are empty, and generate an initial topological sorting result, which includes the target node and all neighbor nodes with in-degree values ​​of 0. The empty initialized in-degree table and queue indicate that the in-degree values ​​of all nodes are 0;

[0116] In this step, the target node refers to a node randomly selected from the queue in the Kahn algorithm. The neighbor node refers to the node directly connected to the target node.

[0117] This step creates an empty list to store the topological sorting results; takes a target node from the queue, and reduces the in-degree value of each neighbor node of the node; if the in-degree value of the neighbor node becomes 0, it is added to the initialized in-degree table and queue; repeats the above steps until the initialized in-degree table and queue are empty, that is, all nodes have been processed, and the initial topological sorting result is generated.

[0118] Step 523: Based on the initial topological sorting result, verify whether the topological sorting process is finished. If the in-degree values ​​of all nodes in the topological sorting result are 0, it is determined that the topological sorting process is finished. If the initialized in-degree table and queue are empty, but there are nodes with in-degree values ​​not equal to 0, it is determined that there is a loop in the network graph model.

[0119] In this step, a loop refers to a situation in the network graph model where one or more nodes can return to themselves through an edge.

[0120] This step checks whether the initial topological sorting result contains all nodes; if the in-degree values ​​of all nodes in the initial topological sorting result are 0 or the length of the result is equal to the number of nodes in the network graph model, the topological sorting is successful and the in-degree values ​​of all nodes are 0; if the length of the initial topological sorting result is less than the number of nodes in the network graph model, it means that there are nodes with in-degree values ​​not equal to 0, then it is determined that there is a loop in the network graph model, or the initialized in-degree table and queue are empty, but there are nodes with in-degree values ​​not equal to 0, then it is determined that there is a loop in the network graph model.

[0121] Step 524: when a loop exists in the network graph model, a loop detection algorithm is used to confirm the location of the loop, and corresponding measures are taken to eliminate the loop to obtain a processed network graph model. Based on the processed network graph model, the Kahn algorithm is repeatedly executed until the in-degree values ​​of all nodes are 0, thereby obtaining an intermediate topological sorting result.

[0122] In this step, a depth-first search (DFS) or breadth-first search (BFS) algorithm is selected to detect loops; starting from any node, the network graph model is traversed using the selected cycle detection algorithm, the visited nodes and paths are recorded, and the location of the loop is detected; the edges in the loop are found, and these edges are removed to ensure that the network graph model becomes a directed acyclic graph (DAG); the adjacency list or adjacency matrix is ​​updated to reflect the network graph model after the loop is removed; the in-degree values ​​of all nodes are recalculated, and all nodes with in-degree values ​​of 0 are added to the queue; according to the method of step 522, the Kahn algorithm is re-executed until the in-degree values ​​of all nodes are 0, and an intermediate topological sorting result is generated.

[0123] The embodiments of the present invention can achieve the following beneficial effects through the above steps:

[0124] The network structure is topologically sorted using the Kahn algorithm to ensure that the node order is consistent with the topological relationship in the network structure, thus improving the accuracy of fault location.

[0125] Combined with the loop detection algorithm, it ensures that there are no loops in the network graph model, which enhances the reliability of fault detection;

[0126] Through multi-step verification and optimization, the correctness of the topological sorting results is ensured, reducing the possibility of false positives and false negatives;

[0127] It provides accurate topological sorting results to help operation and maintenance personnel quickly locate and repair faults, improve the operation and maintenance efficiency of the power system, and ensure the safe operation of the power system.

[0128] The present invention provides a specific embodiment, wherein step 104 constructs an integrated learning framework based on Bayesian optimization, uses the enhanced fault feature representation to train the basic learner in the integrated learning framework to obtain multiple preliminary basic models, uses each of the preliminary basic models to predict the target transmission line respectively, obtains corresponding prediction results, integrates the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy, and generates an initial fault probability distribution, specifically comprising the following steps:

[0129] Step 401: Based on the enhanced fault feature representation, an initial integrated learning framework is selected, and the initial integrated learning framework is optimized using a Bayesian optimization method to obtain an integrated learning framework based on Bayesian optimization;

[0130] In this step, multiple basic learners are selected, such as decision trees, random forests, support vector machines (SVMs), neural networks, etc.; multiple basic learners are combined into an integrated learning framework, such as bagging, boosting, stacking, etc. The initial integrated learning framework is optimized using the Bayesian optimization method. First, the hyperparameter space of each basic learner is defined, such as the maximum depth of the decision tree, the number of trees in the random forest, the kernel function and penalty parameter of the SVM, etc.; the Bayesian optimization model is constructed using Gaussian process regression (GPR) or other probability models; the optimal hyperparameter combination is searched in the hyperparameter space through the Bayesian optimization method to maximize the performance indicators of the model on the validation set, such as accuracy, F1 score, etc.; the optimal hyperparameter combination is used to generate an integrated learning framework based on Bayesian optimization.

[0131] Step 402: dividing the enhanced fault feature representation into a training set and a validation set, and using the training set to train multiple basic learners in the Bayesian optimization-based integrated learning framework to obtain multiple preliminary basic models;

[0132] In this step, the data set represented by the enhanced fault features is divided into a training set and a validation set, usually in a ratio of 80% of the data for the training set and 20% for the validation set for the validation set; the training set and the validation set are loaded from the data set; each base learner is trained using the training set to generate multiple preliminary base models.

[0133] Step 403: inputting the enhanced fault feature representation into the preliminary basic model to obtain a plurality of prediction results corresponding to the preliminary basic model, wherein the prediction results represent the probability of the target transmission line failing;

[0134] This step inputs the data represented by the enhanced fault characteristics into each preliminary base model, and each preliminary base model outputs a prediction result that characterizes the probability of a fault occurring in the target transmission line.

[0135] Step 404: using an adaptive weighting strategy, according to the performance of the preliminary basic model on the validation set, calculating the weight of each of the preliminary basic models, multiplying the weights of all the preliminary basic models by the corresponding prediction results to obtain a plurality of product results, adding the plurality of product results to obtain a final prediction result, and generating an initial fault probability distribution based on the final prediction result;

[0136] This step evaluates the performance of each preliminary basic model on the validation set and calculates the weight of each model; dynamically adjusts the weight of each model based on the performance of the model on the validation set; multiplies the prediction results of all preliminary basic models by the weights and then adds them together to obtain the final prediction result; and generates the initial fault probability distribution based on the final prediction result.

[0137] The embodiments of the present invention can achieve the following beneficial effects through the above steps:

[0138] The integrated learning framework is optimized through the Bayesian optimization method to ensure the optimal performance of the model on the validation set and improve the accuracy of fault detection;

[0139] The integrated learning framework composed of multiple basic learners improves the stability and reliability of the model and reduces the limitations of a single model.

[0140] Through the adaptive weighting strategy, the weights of each model are dynamically adjusted to optimize the final prediction results and reduce the possibility of false positives and false negatives;

[0141] It provides accurate fault probability distribution, helps operation and maintenance personnel to quickly locate and repair faults, improves the operation and maintenance efficiency of the power system, and ensures the safe operation of the power system.

[0142] The present invention provides a specific embodiment, wherein step 103, based on the optimized fault feature space, applies the trained long short-term memory network to analyze the long-term dependency between the fault features in the abnormal data set, selects key features from the fault features in combination with the attention mechanism algorithm, and uses the graph convolutional network to analyze the influence of the network structure of the target transmission line to obtain the structural influence result, and generates an enhanced fault feature representation based on the long-term dependency, the key features and the structural influence result, specifically comprising the following steps:

[0143] Step 301: constructing a long short-term memory network model based on the optimized fault feature space, and using the time series data in the optimized fault feature space to train the long short-term memory network model to obtain a trained long short-term memory network model, and using the trained long short-term memory network model to analyze the long-term dependency between the fault features in the abnormal data set;

[0144] In this step, time series data refers to data that changes over time, such as sensor data such as current, voltage, and temperature. Long-term dependencies refer to the correlation between different time steps in time series data, and the long short-term memory network model can capture this relationship.

[0145] This step defines the structure of the LSTM network model, including the input layer, hidden layer, and output layer. The hidden layer can have multiple layers of LSTM network units; set the model's hyperparameters, such as the number of units in the hidden layer, the learning rate, and the batch size; extract time series data from the optimized fault feature space, such as sensor data such as current, voltage, and temperature; perform preprocessing operations such as normalization and standardization on the time series data to ensure that the data meets the requirements of the model input; compile the LSTM network model, select the loss function and optimizer; train the LSTM network model using the time series data in the optimized fault feature space until the model converges or reaches the predetermined number of training rounds; save the trained LSTM network model for subsequent use; input the fault features in the abnormal data set into the trained LSTM network model to analyze the long-term dependencies between the fault features.

[0146] Step 302: Based on the long-term dependency, an attention mechanism is introduced to calculate the importance weight of each feature in the trained long short-term memory network model at different time steps to select key features from fault features;

[0147] This step uses the attention mechanism to calculate the importance weight of each feature at different time steps; the weight of each time step is calculated through the attention mechanism to highlight important features; the weight of each feature is extracted from the attention mechanism; key features are selected according to the weight size, features with higher weights are retained, and features with lower weights are discarded.

[0148] Step 303: constructing a graph convolutional network model, and using the graph convolutional network model to analyze the impact of the network structure of the target transmission line on the fault characteristics, and obtaining a structural impact result;

[0149] This step uses a graph convolutional network (GCN) to build a model, which can effectively analyze the node relationship in graph structure data; define the number of layers of the graph convolutional network, the number of hidden units in each layer, the activation function and other parameters; build an adjacency matrix based on the network structure of the transmission line to represent the connection relationship between nodes; convert the feature data in the optimized fault feature space into a feature matrix; define the loss function and optimizer; use graph data to train the graph convolutional network model, and adjust the model parameters through the validation set data to prevent overfitting; use the trained graph convolutional network model to extract features from the network structure of the transmission line and analyze the impact of the network structure on the fault characteristics.

[0150] Step 304: Fusing the long-term dependency, the key features, and the structural impact results to generate an enhanced fault feature representation;

[0151] This step combines the long-term dependencies extracted by the LSTM network model, the key features selected by the attention mechanism, and the structural impact results of the graph convolutional network analysis to generate an enhanced fault feature representation.

[0152] The embodiments of the present invention can achieve the following beneficial effects through the above steps:

[0153] The accuracy of fault detection is improved by using the long short-term memory network model to capture long-term dependencies, the attention mechanism to select key features, and the graph convolutional network model to analyze the impact of network structure;

[0154] Comprehensive use of time series data, key features and network structure information improves the robustness and reliability of the model;

[0155] Through multi-step optimization and fusion, the possibility of false positives and false negatives is reduced, and the sensitivity and accuracy of fault detection are improved;

[0156] It provides accurate fault feature representation to help operation and maintenance personnel quickly locate and repair faults, improves the operation and maintenance efficiency of the power system, and ensures the safe operation of the power system.

[0157] Figure 2 A schematic diagram of a transmission line fault location system based on transfer learning is provided for an embodiment of the present invention. Figure 2 As shown, the system includes:

[0158] The acquisition module 21 is used to collect real-time monitoring data of the target transmission line, combine meteorological information and satellite remote sensing images, use deep neural network and image recognition technology to analyze environmental changes around the target transmission line, obtain environmental change information, use variational autoencoder to perform anomaly detection on the environmental change information, obtain anomaly data set, and construct a comprehensive fault feature space based on the anomaly data set;

[0159] An optimization module 22 is used to reduce the distribution gap between the existing fault data set and the abnormal data set by using an adversarial domain adaptation algorithm based on the comprehensive fault feature space to obtain an optimized fault feature space;

[0160] A generation module 23 is used to analyze the long-term dependency between the fault features in the abnormal data set based on the optimized fault feature space by using the trained long short-term memory network, select key features from the fault features in combination with the attention mechanism algorithm, and use a graph convolutional network to analyze the impact of the network structure of the target transmission line to obtain a structural impact result, and generate an enhanced fault feature representation based on the long-term dependency, the key features and the structural impact result;

[0161] A construction module 24 is used to construct an integrated learning framework based on Bayesian optimization, train the basic learner in the integrated learning framework using the enhanced fault feature representation to obtain multiple preliminary basic models, use each of the preliminary basic models to predict the target transmission line respectively, obtain corresponding prediction results, integrate the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy, and generate an initial fault probability distribution;

[0162] The determination module 25 is used to determine the fault location information of the target transmission line based on the initial fault probability distribution by using geographic information system technology and topological sorting algorithm.

[0163] Figure 2 The transmission line fault location system based on transfer learning can be executed Figure 1 The implementation principle and technical effect of the transmission line fault location method based on transfer learning described in the illustrated embodiment will not be repeated. The specific manner in which each module and unit performs operations in the transmission line fault location system based on transfer learning in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0164] In one possible design, Figure 2 A transmission line fault location system based on transfer learning in the illustrated embodiment can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0165] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0166] The processing component 32 is used to: collect real-time monitoring data of the target transmission line, combine meteorological information and satellite remote sensing images, use deep neural networks and image recognition technology to analyze environmental changes around the target transmission line, obtain environmental change information, use variational autoencoders to perform anomaly detection on the environmental change information, obtain an abnormal data set, and construct a comprehensive fault feature space based on the abnormal data set; based on the comprehensive fault feature space, use an adversarial domain adaptation algorithm to reduce the distribution gap between the existing fault data set and the abnormal data set to obtain an optimized fault feature space; based on the optimized fault feature space, use a trained long short-term memory network to analyze the long-term dependency between fault features in the abnormal data set, and select key features from the fault features in combination with the attention mechanism algorithm , using a graph convolutional network to analyze the impact of the network structure of the target transmission line, obtaining the structural impact results, and generating an enhanced fault feature representation based on the long-term dependency, the key features and the structural impact results; constructing an integrated learning framework based on Bayesian optimization, using the enhanced fault feature representation to train the basic learner in the integrated learning framework to obtain multiple preliminary basic models, using each of the preliminary basic models to predict the target transmission line respectively, and obtaining the corresponding prediction results, integrating the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy to generate an initial fault probability distribution; using geographic information system technology and a topological sorting algorithm, determining the fault location information of the target transmission line based on the initial fault probability distribution.

[0167] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0168] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0169] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0170] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0171] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0172] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0173] The embodiment of the present invention further provides a computer storage medium storing a computer program, which can achieve the above-mentioned Figure 1 The illustrated embodiment is a method for locating power transmission line faults based on transfer learning.

[0174] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0175] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0176] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0177] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for locating power transmission line faults based on transfer learning, characterized in that: include: Collect real-time monitoring data of the target transmission line, combine it with meteorological information and satellite remote sensing images, use deep neural network and image recognition technology to analyze environmental changes around the target transmission line, obtain environmental change information, use variational autoencoder to perform anomaly detection on the environmental change information, obtain anomaly data set, and construct a comprehensive fault feature space based on the anomaly data set; Based on the comprehensive fault feature space, an adversarial domain adaptation algorithm is used to reduce the distribution gap between the existing fault data set and the abnormal data set to obtain an optimized fault feature space; Based on the optimized fault feature space, the trained long short-term memory network is used to analyze the long-term dependency between the fault features in the abnormal data set, and the key features are selected from the fault features in combination with the attention mechanism algorithm, and the influence of the network structure of the target transmission line is analyzed by using the graph convolution network to obtain the structural influence result, and based on the long-term dependency, the key features and the structural influence result, an enhanced fault feature representation is generated; Constructing an integrated learning framework based on Bayesian optimization, using the enhanced fault feature representation to train the basic learner in the integrated learning framework to obtain multiple preliminary basic models, using each of the preliminary basic models to predict the target transmission line respectively, obtaining corresponding prediction results, integrating the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy, and generating an initial fault probability distribution; The fault location information of the target transmission line is determined based on the initial fault probability distribution by using geographic information system technology and a topological sorting algorithm.

2. The method according to claim 1, characterized in that The method of using geographic information system technology and topological sorting algorithm to determine the fault location information of the target transmission line based on the initial fault probability distribution includes: Analyzing the network structure of the target transmission line by using a topological sorting algorithm to determine the topological relationship between nodes in the network structure and obtain a target topological sorting result, wherein the topological sorting algorithm includes a Kahn algorithm; Optimizing the initial fault probability distribution based on the target topological sorting result to obtain a target fault probability distribution to narrow the scope of fault location; By using geographic information system technology, the geographical location coordinates corresponding to the target probability value in the target fault probability distribution in the target transmission line map are determined as the fault location information of the target transmission line, and the target probability value is a fault probability value higher than a preset probability value.

3. The method according to claim 2, characterized in that The network structure of the target power transmission line is analyzed by using a topological sorting algorithm to determine the topological relationship between nodes in the network structure and obtain a target topological sorting result. The topological sorting algorithm includes a Kahn algorithm, including: Applying a graph traversal algorithm to analyze the network structure of the target transmission line, extracting information of all nodes and line segments, and obtaining an information list of nodes and line segments, wherein the nodes represent network connection points in the target transmission line, and the nodes include transmission towers, substations, and switch stations, and the line segments represent the transmission line section connecting two nodes, and the line segments include a section of wire or cable; Based on the information list of the nodes and line segments, a network graph model is constructed; Based on the network graph model, the nodes in the network structure are topologically sorted using the Kahn algorithm to obtain an initial topological sorting result, and the initial topological sorting result is verified. If the verification passes, an intermediate topological sorting result is obtained, and the condition for passing the verification is that the in-degree values ​​of all nodes are 0; Matching the intermediate topological sorting result with the network structure of the target transmission line so that the node order in the intermediate topological sorting result is consistent with the topological relationship between the nodes in the network structure; after the matching is successful, clustering the nodes in the network structure based on the intermediate topological sorting result to obtain a clustering result; The intermediate topological sorting result is optimized in combination with the clustering result to generate a target topological sorting result.

4. The method according to claim 3, characterized in that Based on the network graph model, the Kahn algorithm is used to topologically sort the nodes in the network structure to obtain an initial topological sorting result, and the initial topological sorting result is verified. After the verification, an intermediate topological sorting result is obtained. The condition for passing the verification is that the in-degree values ​​of all nodes are 0, including: Traversing the edges in the network graph model, counting the in-degree value of each node in the edge, adding all nodes with an in-degree value of 0 to a queue, and obtaining an initialized in-degree table and a queue, wherein the in-degree value represents the number of edges pointing to the node, and the node with an in-degree value of 0 is used as the starting point of the topological sorting; Based on the initialized in-degree table and queue, using the Kahn algorithm, a node is randomly extracted as the target node, and the in-degree value of the neighbor node of the target node is reduced. If the in-degree value of the neighbor node is 0, the neighbor node is added to the initialized in-degree table and queue, and the operation of reducing the in-degree value of the neighbor node is repeated until the initialized in-degree table and queue are empty, and an initial topological sorting result is generated. The initial topological sorting result includes the target node and all neighbor nodes with in-degree values ​​of 0. The initialization in-degree table and queue are empty, indicating that the in-degree values ​​of all nodes are 0; Based on the initial topological sorting result, verify whether the topological sorting process is finished. If the in-degree values ​​of all nodes in the topological sorting result are 0, it is determined that the topological sorting process is finished. If the initialized in-degree table and queue are empty, but there are nodes with in-degree values ​​not equal to 0, it is determined that there is a loop in the network graph model. When a loop exists in the network graph model, a loop detection algorithm is used to confirm the location of the loop, and corresponding measures are taken to eliminate the loop to obtain a processed network graph model. Based on the processed network graph model, the Kahn algorithm is repeatedly executed until the in-degree values ​​of all nodes are 0, thereby obtaining an intermediate topological sorting result.

5. The method according to claim 4, characterized in that The loop detection algorithm is used to confirm the location of the loop, and corresponding measures are taken to eliminate the loop to obtain a processed network graph model. Based on the processed network graph model, the Kahn algorithm is repeatedly executed until the in-degree values ​​of all nodes are 0, and an intermediate topological sorting result is obtained, including: Based on the network graph model, selecting a depth-first search algorithm in a cycle detection algorithm; Based on the depth-first search algorithm, an initial node is randomly selected from the network graph model, and a depth-first search is performed starting from the initial node. The access status of each node in the depth-first search process is recorded. If a node is detected to have an access status of being accessed, it is determined that a loop exists, and location information of the loop is generated. The access status includes not accessed, being accessed, and visited. Based on the location information of the loop, taking corresponding measures to disconnect the loop to obtain a processed network graph model, wherein the corresponding measures include deleting edges or nodes in the loop and changing the direction of edges in the loop; Based on the processed network graph model, the Kahn algorithm is repeatedly executed until the in-degree values ​​of all nodes are 0, and the intermediate topological sorting result is obtained.

6. The method according to claim 1, characterized in that The integrated learning framework based on Bayesian optimization is constructed, and the basic learner in the integrated learning framework is trained using the enhanced fault feature representation to obtain multiple preliminary basic models. Each of the preliminary basic models is used to predict the target transmission line to obtain corresponding prediction results, and the prediction results corresponding to all the preliminary basic models are integrated through an adaptive weighting strategy to generate an initial fault probability distribution, including: Based on the enhanced fault feature representation, an initial integrated learning framework is selected, and the initial integrated learning framework is optimized using a Bayesian optimization method to obtain an integrated learning framework based on Bayesian optimization; Dividing the enhanced fault feature representation into a training set and a validation set, and using the training set to train multiple basic learners in the Bayesian optimization-based integrated learning framework to obtain multiple preliminary basic models; Inputting the enhanced fault feature representation into the preliminary basic model to obtain a plurality of prediction results corresponding to the preliminary basic model, wherein the prediction results represent the probability of the target transmission line failing; Using an adaptive weighting strategy, the weight of each preliminary basic model is calculated according to the performance of the preliminary basic model on the validation set, the weights of all the preliminary basic models are multiplied by the corresponding prediction results to obtain multiple product results, the multiple product results are added together to obtain a final prediction result, and based on the final prediction result, an initial fault probability distribution is generated.

7. The method according to claim 1, characterized in that Based on the optimized fault feature space, the trained long short-term memory network is used to analyze the long-term dependency between the fault features in the abnormal data set, and the key features are selected from the fault features in combination with the attention mechanism algorithm, and the graph convolution network is used to analyze the influence of the network structure of the target transmission line to obtain the structural influence result, and based on the long-term dependency, the key features and the structural influence result, an enhanced fault feature representation is generated, including: Based on the optimized fault feature space, a long short-term memory network model is constructed, and the long short-term memory network model is trained using the time series data in the optimized fault feature space to obtain a trained long short-term memory network model, and the long short-term memory network model is used to analyze the long-term dependency between the fault features in the abnormal data set; Based on the long-term dependency, an attention mechanism is introduced to calculate the importance weight of each feature in the trained long short-term memory network model at different time steps, so as to select key features from the fault features; Constructing a graph convolutional network model, and using the graph convolutional network model to analyze the impact of the network structure of the target transmission line on the fault characteristics, and obtaining a structural impact result; The long-term dependencies, the key features, and the structural impact results are integrated to generate an enhanced fault feature representation.

8. A transmission line fault location system based on transfer learning, characterized in that: include: An acquisition module is used to collect real-time monitoring data of a target transmission line, and analyze the environmental changes around the target transmission line using deep neural network and image recognition technology in combination with meteorological information and satellite remote sensing images to obtain environmental change information, and perform anomaly detection on the environmental change information using a variational autoencoder to obtain an abnormal data set, and construct a comprehensive fault feature space based on the abnormal data set; An optimization module, for reducing the distribution gap between the existing fault data set and the abnormal data set by using an adversarial domain adaptation algorithm based on a comprehensive fault feature space to obtain an optimized fault feature space; A generation module is used to analyze the long-term dependency between the fault features in the abnormal data set based on the optimized fault feature space by using the trained long short-term memory network, select key features from the fault features in combination with the attention mechanism algorithm, analyze the influence of the network structure of the target transmission line by using the graph convolution network, obtain the structural influence result, and generate an enhanced fault feature representation based on the long-term dependency, the key features and the structural influence result; A construction module is used to construct an integrated learning framework based on Bayesian optimization, train the basic learner in the integrated learning framework using the enhanced fault feature representation to obtain multiple preliminary basic models, use each of the preliminary basic models to predict the target transmission line respectively, obtain corresponding prediction results, integrate the prediction results corresponding to all the preliminary basic models through an adaptive weighting strategy, and generate an initial fault probability distribution; The determination module is used to determine the fault location information of the target transmission line based on the initial fault probability distribution by using geographic information system technology and topological sorting algorithm.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a transmission line fault location method based on transfer learning as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a transmission line fault location method based on transfer learning as described in any one of claims 1 to 7 is implemented.

Images