Line fault positioning method and system based on neural network

Abstract

The invention discloses a line fault positioning method and system based on a neural network, and relates to the technical field of line fault positioning, and the method comprises the steps: collecting power distribution network data, carrying out the preprocessing, and extracting fault features; constructing and training a fault positioning model based on the fault features; and carrying out fault positioning by using the trained fault positioning model. According to the method, the fault positioning model is designed by integrating the multi-layer graph convolutional network and the multi-head attention mechanism, multi-layer features are deeply mined and modeled in consideration of multi-layer data, multi-source data and deep feature learning can be integrated, and the accuracy of fault positioning and the robustness of the system are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of line fault location, and in particular to a line fault location method and system based on a neural network. Background Art

[0002] As an important part of the power system, the operation stability and fault diagnosis capability of the distribution network directly affect the reliability and safety of power supply. Traditional distribution network fault location technology mainly relies on electrical quantity measurement and analysis methods, such as impedance method, overcurrent method and time difference method, which estimate the location of the fault point by measuring current, voltage and fault waveform. However, with the complexity of the distribution network structure and the diversification of load conditions, traditional methods have gradually exposed their limitations when dealing with large-scale power grids and complex fault modes. Therefore, it is gradually turning to data-driven methods, using advanced data analysis and machine learning technologies to improve the accuracy and response speed of fault location.

[0003] Although data-driven fault location methods have high potential in theory, they still have some significant deficiencies in practical applications. Most existing methods are still limited to shallow neural networks or single model construction, usually ignoring the complex multi-dimensional data characteristics in the distribution network, lacking in-depth mining and modeling of multi-layer features, and unable to fully mine important features such as node attributes and edge existence in the network, resulting in insufficient accuracy and robustness of fault location. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: how to conduct in-depth mining and modeling of multi-layer features based on multi-dimensional data features to improve the accuracy and robustness of fault location.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a line fault location method based on a neural network, comprising: collecting distribution network data for preprocessing, and then extracting fault characteristics; constructing and training a fault location model based on the fault characteristics; and locating the fault using the trained fault location model.

[0007] As a preferred solution of the neural network-based line fault location method described in the present invention, the distribution network data includes geographic data, meteorological data, electrical data and historical fault data, the geographic data includes the geographic coordinates of the node; the meteorological data includes temperature, wind speed, and humidity; the historical fault data includes the frequency and type of historical faults; the electrical data includes voltage, current, power, line load, and phase angle of three-phase current.

[0008] As a preferred solution of the neural network-based line fault location method described in the present invention, the preprocessing includes: converting the geographic data into a unified coordinate system and standardizing the numerical range; removing missing values ​​and outliers from the meteorological data and then standardizing it; processing the missing values, noise and outliers from the electrical data and then standardizing it; converting historical fault records into a coding format suitable for model processing.

[0009] As a preferred solution of the line fault location method based on neural network described in the present invention, wherein: the fault location model includes an input layer, a hidden layer, a learning layer and an output layer. When the fault feature is input, the model will pre-determine whether the feature includes an edge feature. If it does, the input layer is activated; otherwise, the input layer is skipped and the fault feature is input to the hidden layer;

[0010] When the input layer is enabled, the edge features are combined with the features of the adjacent nodes to generate new node features, that is, the features of each edge are averaged or weightedly distributed to the two nodes connected to the edge, the new node features are fused with the original node features, and a certain weight is given to the edge features through an adaptive weighting mechanism before being added to the node features, and the fused node features are input into the hidden layer;

[0011] The hidden layer includes a multi-layer graph convolutional network and a multi-head attention mechanism. The multi-layer graph convolutional network first maps the input features to the hidden layer through a linear layer. The features of each node exchange information with its neighboring nodes through a graph convolution operation to extract local and global information. Then, the perception range is dynamically adjusted according to the connection density of the node. If the node is connected to multiple lines, the perception range is expanded to capture load fluctuations at a longer distance. Otherwise, the perception range is narrowed to only focus on the features of nearby lines. The multi-head attention mechanism assigns different attention weights to the features output by the multi-layer graph convolutional network and automatically selects the most important features for fault location.

[0012] The learning layer identifies feature differences between normal lines and faulty lines through positive and negative samples;

[0013] The output layer gives the prediction result of fault location.

[0014] As a preferred solution of the line fault location method based on neural network described in the present invention, wherein: the feature difference identification includes generating negative samples by perturbing edge features, generating positive samples by retaining original features, learning the feature differences between normal lines and faulty lines, and performing self-supervised learning node attribute prediction and edge existence prediction;

[0015] The node attribute prediction includes masking some attributes of the node, allowing the fault location model to predict the masked value based on the characteristics of the neighboring nodes; the edge existence prediction includes masking the line connection, predicting the existence of the line, and thus inferring the propagation path of the fault.

[0016] As a preferred solution of the line fault location method based on neural network described in the present invention, the training of the location model includes defining a loss function L to train the fault location model until the loss function value is minimized and then stopping the training, wherein the loss function includes a first loss L 1 , the second loss L 2 、The third loss L 3 and the fourth loss L 4 , expressed as:

[0017] L=L 1 +λ 1 L 2 +λ 2 L 3 +λ 3 L 4

[0018] Among them, λ 1 is the weight coefficient of the first loss, λ 2 is the weight coefficient of the second loss, λ 3 is the weight coefficient of the third loss.

[0019] As a preferred solution of the line fault location method based on neural network described in the present invention, the first loss is the classification loss L 1 It is expressed as:

[0020]

[0021] The second loss is the contrast loss L 2 It is expressed as:

[0022]

[0023] The third loss is the node attribute prediction loss L 3 It is expressed as:

[0024]

[0025] The fourth loss is the edge existence prediction loss L 4 It is expressed as:

[0026]

[0027] in, is the fault prediction result, true faultsis the target label; similarity(x i ,x j ) is the node x i and x j The cosine similarity, similarity(x i ,x k ) is the node x i and x k The cosine similarity of , τ is a control parameter used to control the smoothness of the similarity distribution; N is the number of negative samples, is the predicted node attribute, true attributes is the real node attribute, is the predicted edge existence, true edges For the real edge existence.

[0028] In a second aspect, another object of the present invention is to provide a line fault location system based on a neural network, comprising: a collection module, a data processing module, a model building module and a fault location module;

[0029] The acquisition module is used to collect distribution network data;

[0030] The data processing module is used to pre-process the distribution network data and extract fault characteristics;

[0031] The model building module is used to build and train a fault location model based on fault features;

[0032] The fault location module is used to perform fault location using a trained fault location model.

[0033] In a third aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned neural network-based line fault location method when executing the computer program.

[0034] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the neural network-based line fault location method as described above.

[0035] Beneficial effects of the present invention: The neural network-based line fault location method and system provided by the present invention designs a fault location model by integrating a multi-layer graph convolutional network and a multi-head attention mechanism. Taking into account multi-source data, multi-layer features are deeply mined and modeled, and can integrate multi-source data and deep-level feature learning, thereby greatly improving the accuracy of fault location and the robustness of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0037] Figure 1 A method flow chart of a line fault location method and system based on a neural network provided by an embodiment of the present invention;

[0038] Figure 2 A schematic diagram of a process flow of processing edge features of a line fault location method and system based on a neural network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0041] Example 1

[0042] Reference Figure 1-2 , as an embodiment of the present invention, provides a line fault location method based on a neural network, comprising:

[0043] S1: After collecting distribution network data and preprocessing it, fault characteristics are extracted;

[0044] Furthermore, the distribution network data includes geographic data, meteorological data, electrical data and historical fault data. Geographic data includes the geographic coordinates of the nodes; meteorological data includes temperature, wind speed, and humidity; historical fault data includes the frequency and type of historical faults; and electrical data includes voltage, current, power, line load, and the phase angle of three-phase current.

[0045] S2: constructing and training a fault location model based on the fault characteristics;

[0046] Furthermore, geographic data, meteorological data, electrical data and historical fault data are cleaned, standardized and time synchronized.

[0047] It should be noted that (1) preprocessing of geographic data

[0048] Normalization: Convert geographic data into a uniform coordinate system and standardize the numerical range.

[0049] Feature extraction: Extract geographic location and map it to node features.

[0050] (2) Preprocessing of meteorological data

[0051] Data cleaning: remove missing values ​​and outliers.

[0052] Standardization: Meteorological data is standardized to eliminate the impact of different dimensions.

[0053] Time synchronization: Ensure that weather data is synchronized with other data sources.

[0054] (3) Preprocessing of electrical data

[0055] Data cleaning: dealing with missing values, noise, and outliers.

[0056] Feature extraction: Extract important electrical features (voltage, current, power) and standardize them.

[0057] Time alignment: Ensure electrical data is synchronized with other data.

[0058] (4) Preprocessing of historical fault data

[0059] Label encoding: Convert historical fault records into an encoding format suitable for model processing (such as category encoding).

[0060] Fault features are extracted from preprocessed geographic data, meteorological data, electrical data and historical fault data.

[0061] Furthermore, the fault location model includes an input layer, a hidden layer, a learning layer and an output layer, and the input layer, the hidden layer, the learning layer and the output layer are connected in sequence; wherein, referring to Figure 2 ,When the fault feature is input, the model will pre-judge whether the feature includes the edge feature. If it is included, the input layer will be activated; otherwise, the input layer will be skipped and the fault feature will be input to the hidden layer;

[0062] When the input layer is enabled, the edge features are combined with the features of the adjacent nodes to generate new node features; that is, the features of each edge are averaged or weighted to the two nodes connected to the edge;

[0063] The new node features are merged with the original node features; through the adaptive weighting mechanism, a certain weight is given to the edge features and then added to the node features.

[0064] The fused node features are input into the hidden layer.

[0065] The hidden layer includes a multi-layer graph convolutional network and a multi-head attention mechanism; the multi-layer graph convolutional network includes three layers of graph convolutional networks, and the output of each layer of the graph convolutional network is set with a residual block to maintain the detailed information of low-level features and combine high-level global features. The multi-layer graph convolutional network first maps the input features to the hidden layer through a linear layer. The features of each node are exchanged with its neighboring nodes through a graph convolution operation to extract local and global information. Then, the perception range is dynamically adjusted according to the connection density of the node. If the node is connected to multiple lines, the perception range is expanded to capture load fluctuations at a longer distance. Otherwise, the perception range is narrowed and only the features of nearby lines are focused. The multi-head attention mechanism assigns different attention weights to the features output by the multi-layer graph convolutional network, and automatically selects the most important features for fault location. Through the multi-head mechanism, the model can learn features in different subspaces, enhancing the model's ability to capture the relationship between different modes.

[0066] In addition, in the hidden layer, the node features are also updated, and each node is updated according to the features of the edges connected to it and the features of the neighboring nodes.

[0067] Furthermore, the learning layer is used to identify the feature differences between normal lines and faulty lines through positive and negative samples. Specifically, negative samples are generated by perturbing edge features and positive samples are generated by retaining original features, so as to learn the feature differences between normal lines and faulty lines and perform self-supervised learning, including node attribute prediction and edge existence prediction.

[0068] Among them, node attribute prediction includes covering some attributes of the node, allowing the fault location model to predict the covered value based on the characteristics of the neighboring nodes to better capture the global dependency of nodes and edges. For example, assuming that node n2 is a distribution station and its current attribute is covered, the model needs to predict the current of n2 based on the voltage, current and power flow of the surrounding nodes n1, n3, n4; edge existence prediction includes covering the line connection and predicting the existence of the line, so as to infer the propagation path of the fault. For example, select line e3, 5 and cover its connection relationship. The model predicts whether e3, 5 should exist based on the information of the nodes n3 and n5 at both ends and the surrounding lines, helping the model learn the difference between normal lines and faulty lines. Preferably, by introducing self-supervised learning, when a fault occurs in a certain area, the model can automatically infer how the power is redistributed in the local grid based on the topological structure of the global graph, so as to predict potential secondary faults in advance.

[0069] The prediction result of fault location is given through the output layer.

[0070] Furthermore, a loss function is defined to train the fault location model until the loss function value is minimized and the training is stopped. The steps are as follows:

[0071] Define the loss function L, which includes the first loss L 1 , the second loss L 2 、The third loss L 3 and the fourth loss L 4 ;

[0072] Among them, the loss function L is:

[0073] L=L 1 +λ 1 L 2 +λ 2 L 3 +λ 3 L 4

[0074] The first loss is the classification loss L 1 :

[0075]

[0076] The second loss is the contrast loss L 2 :

[0077]

[0078] The third loss is the node attribute prediction loss L 3 :

[0079]

[0080] The fourth loss is the edge existence prediction loss L 4 :

[0081]

[0082] In the formula, λ 1 is the weight coefficient of the first loss, λ 2 is the weight coefficient of the second loss, λ 3 is the weight coefficient of the third loss, which is used to control the weight of different loss items in the total loss, and its value range is (0, 1); is the fault prediction result, true faults is the target label; similarity(x i ,x j ) is the node x i and x jThe cosine similarity, similarity(x i ,x k ) is the node x i and x k The cosine similarity of , τ is a control parameter used to control the smoothness of the similarity distribution, and N is the number of negative samples; is the predicted node attribute, true attributes is the real node attribute; is the predicted edge existence, true edges For the real edge existence.

[0083] Furthermore, the gradient descent algorithm is used to calculate the gradient of the loss function with respect to each parameter of the fault localization model;

[0084] Update the model parameters based on the gradient using the AdamW optimizer

[0085] S3: Utilize the trained fault location model to perform fault location.

[0086] Furthermore, the geographic data, meteorological data, and electrical data collected in real time are input into the trained fault location model for prediction to determine the location of the fault.

[0087] Example 2

[0088] As an embodiment of the present invention, a digital construction system for typical planning scenarios of county distribution networks is provided, including:

[0089] It includes acquisition module, data processing module, model building module and fault location module;

[0090] The acquisition module is used to collect distribution network data;

[0091] The data processing module is used to pre-process the distribution network data and extract fault characteristics;

[0092] The model building module is used to build and train a fault location model based on fault features;

[0093] The fault location module is used to locate faults using the trained fault location model.

[0094] The above-mentioned unit modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above-mentioned modules.

[0095] This embodiment also provides a computer device, which can be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a line fault location method based on a neural network is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0096] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0097] Obtaining the number of electric vehicles, the number of agents and parameter information of electric vehicles participating in the V2G plan within the target range, and establishing V2G-based electric vehicle response capability constraint conditions based on the parameter information;

[0098] The parameter information includes the time when the electric vehicles participating in the V2G program connect to and leave the grid, the load status of the electric vehicle, the battery capacity of the electric vehicle, and the charging and discharging power of the electric vehicle;

[0099] The V2G-based electric vehicle response capability constraints include the electric vehicle access grid time constraint, the electric vehicle's own charge state constraint, the V2G response capacity constraint, and the charging and discharging power constraint;

[0100] Establishing the rules for obtaining the responsiveness value of electric vehicles under the responsiveness constraint conditions of electric vehicles based on V2G;

[0101] Establish a two-level optimization model for electric vehicle charging and discharging strategy that includes the EV’s response capability value;

[0102] According to the improved multi-population genetic optimization algorithm, the double-layer optimization model of electric vehicle charging and discharging strategy is solved to obtain the optimal charging power in each time period, thereby obtaining the final electric vehicle charging and discharging strategy.

[0103] Example 3

[0104] An embodiment of the present invention is different from the first two embodiments in that:

[0105] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0106] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0107] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0108] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0109] Example 4

[0110] An embodiment of the present invention provides a line fault location method based on a neural network. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0111] The specific experimental data collected are as follows:

[0112] 1. Fault location of urban distribution network

[0113] Environmental conditions: urban areas with dense distribution lines.

[0114] Dataset: Contains geographic, meteorological, electrical and historical fault data.

[0115] Fault type: short circuit, ground fault, and fault caused by equipment aging.

[0116] Sample size: 1000 failure instances.

[0117] Evaluation indicators: accuracy, processing time, false alarm rate, user feedback.

[0118] 2. Rural distribution network fault location

[0119] Environmental conditions: Rural areas with sparse distribution lines.

[0120] Dataset: mainly includes meteorological data and historical fault records.

[0121] Failure type: Branch contact caused by storm, animal damage.

[0122] Sample size: 500 failure instances.

[0123] Evaluation indicators: fault detection rate, processing time, and user satisfaction.

[0124] 3. Fault location during high load period

[0125] Environmental conditions: During the high-load period in summer, the distribution network is under extreme load. Dataset: Contains electrical data and environmental factors during the high-load period.

[0126] Fault type: overload, equipment failure.

[0127] Sample size: 800 failure instances.

[0128] Evaluation metrics: accuracy, system robustness score, processing time.

[0129] 4. Fault location under extreme weather conditions

[0130] Environmental conditions: typhoons, blizzards and other extreme weather conditions.

[0131] Dataset: Combining real-time meteorological data with historical fault records.

[0132] Fault type: line break, equipment failure.

[0133] Sample size: 600 failure instances.

[0134] Evaluation indicators: accuracy, false alarm rate, user feedback.

[0135] 5. Equipment aging and maintenance record fault location

[0136] Environmental conditions: Distribution network with maintenance records for multiple devices.

[0137] Dataset: includes historical failures, maintenance records and electrical data of the equipment. Failure type: equipment failure, preventive maintenance failure.

[0138] Sample size: 700 failure instances.

[0139] Evaluation indicators: fault detection rate, processing time, and user satisfaction.

[0140] 6. Historical failure mode identification

[0141] Environmental Conditions: Analyze historical data to identify failure modes.

[0142] Dataset: Contains many years of historical fault records and environmental data.

[0143] Failure Type: Different types of failure modes.

[0144] Sample size: 1500 failure instances.

[0145] Evaluation indicators: feature recognition accuracy, system robustness score, user feedback The final test results are shown in the following table:

[0146]

[0147] 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 preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A line fault location method based on neural network, characterized in that: include: After collecting distribution network data and preprocessing it, fault features are extracted; Building and training a fault location model based on the fault characteristics; The trained fault location model is used to perform fault location.

2. The line fault location method based on neural network according to claim 1, characterized in that: The distribution network data includes geographic data, meteorological data, electrical data and historical fault data. The geographic data includes the geographic coordinates of the nodes; the meteorological data includes temperature, wind speed and humidity; the historical fault data includes the frequency and type of historical faults; the electrical data includes voltage, current, power, line load and the phase angle of three-phase current.

3. The line fault location method based on neural network according to claim 2, characterized in that: The preprocessing includes converting the geographic data into a unified coordinate system and standardizing the numerical range; removing missing values ​​and outliers from the meteorological data and then standardizing it; processing missing values, noise and outliers from the electrical data and then standardizing it; and converting historical fault records into a coding format suitable for model processing.

4. The line fault location method based on neural network as claimed in claim 3, characterized in that: The fault location model includes an input layer, a hidden layer, a learning layer and an output layer. When a fault feature is input, the model will pre-determine whether the feature includes an edge feature, and if so, the input layer will be activated; Otherwise, the input layer is skipped and the fault features are input into the hidden layer; When the input layer is enabled, the edge features are combined with the features of the adjacent nodes to generate new node features, that is, the features of each edge are averaged or weightedly distributed to the two nodes connected to the edge, the new node features are fused with the original node features, and a certain weight is given to the edge features through an adaptive weighting mechanism before being added to the node features, and the fused node features are input into the hidden layer; The hidden layer includes a multi-layer graph convolutional network and a multi-head attention mechanism. The multi-layer graph convolutional network first maps the input features to the hidden layer through a linear layer. The features of each node exchange information with its neighboring nodes through a graph convolution operation to extract local and global information. Then, the perception range is dynamically adjusted according to the connection density of the node. If the node is connected to multiple lines, the perception range is expanded to capture load fluctuations at a longer distance. Otherwise, the perception range is narrowed to only focus on the features of nearby lines. The multi-head attention mechanism assigns different attention weights to the features output by the multi-layer graph convolutional network and automatically selects the most important features for fault location. The learning layer identifies feature differences between normal lines and faulty lines through positive and negative samples; The output layer gives the prediction result of fault location.

5. The line fault location method based on neural network according to claim 4, characterized in that: The feature difference identification includes generating negative samples by perturbing edge features, retaining original features to generate positive samples, learning the feature differences between normal lines and faulty lines, and performing self-supervised learning node attribute prediction and edge existence prediction; The node attribute prediction includes masking some attributes of the node, allowing the fault location model to predict the masked value based on the characteristics of the neighboring nodes; the edge existence prediction includes masking the line connection, predicting the existence of the line, and thus inferring the propagation path of the fault.

6. The line fault location method based on neural network according to claim 5, characterized in that: The training of the positioning model includes defining a loss function L to train the fault positioning model until the loss function value is minimized and then stopping the training. The loss function includes a first loss L1, a second loss L2, a third loss L3 and a fourth loss L4, which are expressed as: L=L1+λ1L2+λ2L3+λ3L4 Among them, λ1 is the weight coefficient of the first loss, λ2 is the weight coefficient of the second loss, and λ3 is the weight coefficient of the third loss.

7. The line fault location method based on neural network according to claim 6, characterized in that: The first loss is the classification loss L1, which is expressed as: The second loss is the contrast loss L2 and is expressed as: The third loss is the node attribute prediction loss L3, which is expressed as: The fourth loss is the edge existence prediction loss L4, which is expressed as: in, is the fault prediction result, true faults is the target label; similarity(x i ,x j ) is the node x i and x j The cosine similarity, similarity(x i ,x k ) is the node x i and x k The cosine similarity of , τ is a control parameter used to control the smoothness of the similarity distribution; N is the number of negative samples, is the predicted node attribute, true attributes is the real node attribute, is the predicted edge existence, true edges For the real edge existence.

8. A system using the line fault location method based on a neural network as claimed in any one of claims 1 to 7, characterized in that: It includes acquisition module, data processing module, model building module and fault location module; The acquisition module is used to collect distribution network data; The data processing module is used to pre-process the distribution network data and extract fault characteristics; The model building module is used to build and train a fault location model based on fault features; The fault location module is used to perform fault location using a trained fault location model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the line fault location method based on a neural network as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the line fault location method based on a neural network as described in any one of claims 1 to 7 are implemented.

Images