A highway micro-grid fault positioning method and system

Abstract

The application discloses a highway micro-grid fault positioning method and system, and relates to the field of relay protection of power systems.The method comprises the following steps: obtaining the reactive power amplitude and the reactive power direction of each measuring point on each line section in the highway micro-grid; calculating the fault feature vector of each measuring point according to the reactive power amplitude and the reactive power direction of each measuring point; the fault feature vectors of multiple measuring points constitute a to-be-tested data group; inputting the to-be-tested data group into a micro-grid fault detection model to obtain the fault line section on the highway micro-grid; the micro-grid fault detection model is obtained by training and optimizing a graph attention network based on a fault sample set by using multiple kernel functions; the fault sample in the fault sample set comprises fault simulation feature vectors of multiple measuring points on each line section in the highway micro-grid and corresponding fault positions.The application realizes rapid and accurate positioning of faults in the highway micro-grid.

Description

A method and system for fault location in highway microgrids Technical Field

[0001] This invention relates to the field of power system relay protection technology, and in particular to a method and system for fault location in highway microgrids based on multi-core graph attention networks. Background Technology

[0002] Highways require substantial electrical energy to power electric vehicles and monitoring / maintenance loads. The highway power grid integrates large-scale photovoltaic (PV) and wind power, forming active microgrids. The integration of distributed power sources transforms the traditional "passive" distribution network into an "active" one. Highway microgrids differ from other microgrids in that they exhibit a long, chain-like structure, sometimes reaching tens of kilometers in length. Numerous branches and PV power sources are integrated into the main chain, resulting in localized weak loops and radial structures. Due to their long spans, numerous devices, and harsh environments, highway microgrids suffer from high failure rates. The large-scale integration of PV and wind power significantly alters the fault characteristics within the microgrid, making fault location primarily dependent on the control strategies of the new energy sources. Therefore, rapid and accurate fault location in highway microgrids is crucial, providing effective guidance for maintenance personnel, significantly shortening fault handling and power restoration time, and improving power supply reliability and security.

[0003] Data-driven fault location methods for highway microgrids utilize a large amount of fault data to train a location model. When a fault occurs, real-time measured fault data is input into the trained model to locate the faulty section. However, the microgrid topology frequently changes during operation, especially with the frequent commissioning and decommissioning of distributed photovoltaic power sources and the unpredictable control strategies and parameters, leading to insufficient generalization ability of commonly used fault location models. When the topology of a highway microgrid changes, the location accuracy of commonly used data-driven fault location models also decreases significantly, making it difficult to meet the fault location requirements of highway microgrids. With the increasing complexity of the power grid structure and changes in the operating environment, accurate fault location in highway microgrids faces enormous challenges. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for fault location in highway microgrids, enabling rapid and accurate fault location in highway microgrids.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for fault location in a highway microgrid includes:

[0007] Acquire the reactive power amplitude and reactive power direction at measurement points on various line sections in a highway microgrid;

[0008] Based on the reactive power amplitude and reactive power direction at each measurement point, calculate the fault feature vector for each measurement point; the fault feature vectors of multiple measurement points constitute the data set to be measured.

[0009] The test data set is input into the microgrid fault detection model to obtain the faulty line sections on the highway microgrid. The microgrid fault detection model is obtained by training and optimizing the graph attention network using multiple kernel functions based on the fault sample set. The fault samples in the fault sample set include the fault simulation feature vectors of multiple measurement points on each line section of the highway microgrid and the corresponding fault locations.

[0010] Optionally, the data set to be tested is input into the microgrid fault detection model to obtain the faulty line sections on the highway microgrid, specifically including:

[0011] The data set to be tested is input into the microgrid fault detection model, and the operating status of the measurement points on each line section of the highway microgrid output by the microgrid fault detection model is obtained; the operating status is a value of 0-1.

[0012] For any line segment, if the operating state of the line segment is 0, and the operating state of the line segment connected to the input terminal of the line segment is 1, then the line segment is characterized as a faulty line segment.

[0013] Optionally, the construction process of the microgrid fault detection model specifically includes:

[0014] The attention coefficients of a graph attention network are calculated based on multiple kernel functions.

[0015] The fault sample set and the attention coefficients are input into a preset classifier for training to obtain the optimal classifier; the optimal classifier is a microgrid fault detection model.

[0016] Optionally, the process of determining the fault sample set includes:

[0017] Establish a power system topology model for the aforementioned highway microgrid system;

[0018] Based on the different power systems connected to each line section in the highway microgrid system, the power system topology model is simulated to obtain the original power grid fault dataset and the fault location corresponding to each original power grid fault data; the original power grid fault data includes the reactive power amplitude and reactive power direction at each measurement point corresponding to the power system topology model.

[0019] Based on the reactive power amplitude and reactive power direction at each measurement point corresponding to the power system topology model, the fault simulation feature vector of each measurement point is determined; the fault simulation feature vectors of multiple measurement points and the corresponding fault locations constitute a fault sample; multiple fault samples constitute a fault sample set.

[0020] To achieve the above objectives, the present invention also provides the following technical solutions:

[0021] A fault location system for a highway microgrid includes:

[0022] The measurement data acquisition module is used to acquire the reactive power amplitude and reactive power direction at measurement points on various line sections in the highway microgrid.

[0023] The feature vector calculation module is used to calculate the fault feature vector of each measurement point based on the reactive power amplitude and reactive power direction at each measurement point; the fault feature vectors of multiple measurement points constitute the data set to be measured;

[0024] The fault location module is used to input the data set to be tested into the microgrid fault detection model to obtain the faulty line section on the highway microgrid. The microgrid fault detection model is obtained by training and optimizing the graph attention network based on the fault sample set using multiple kernel functions. The fault samples in the fault sample set include the fault simulation feature vectors of multiple measurement points on each line section of the highway microgrid and the corresponding fault locations.

[0025] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0026] This invention discloses a method and system for fault location in highway microgrids. It pre-acquires a fault sample set when a fault occurs in the highway microgrid. Based on this sample set, a graph attention network is trained and optimized using various kernel functions to obtain a microgrid fault detection model. During actual fault detection, the reactive power amplitude and direction at measurement points on each line section of the highway microgrid are obtained, and a fault feature vector is further calculated. This vector is then input into the pre-trained microgrid fault detection model to identify the faulty line section in the highway microgrid, thus achieving fault location. The microgrid fault detection model constructed in this invention improves the adaptability of the fault location model to topology changes and fully leverages the advantages of strong global information mining capabilities and fast computation speed of artificial intelligence algorithms. It requires no manual tuning and has strong adaptability, solving the problem of weak generalization ability in traditional data-driven highway microgrid fault location methods. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 is a flowchart illustrating the fault location method for highway microgrids of the present invention;

[0029] Figure 2 is a flowchart of the fault location scheme for highway microgrids in a specific embodiment of the present invention.

[0030] Figure 3 is a graph attention network structure diagram according to an embodiment of the present invention;

[0031] Figure 4 is a diagram of the multi-core graph attention network structure according to an embodiment of the present invention;

[0032] Figure 5 is a flowchart illustrating the fault location system for highway microgrids of the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] The purpose of this invention is to provide a method and system for fault location in highway microgrids. Drawing on the good application results of measurement point classification and AI technology in power systems, this invention starts from the topology of highway microgrids and uses a multi-core graph attention network to locate faults in highway microgrids. It breaks through the bottleneck of traditional data-driven fault location in highway microgrids and solves the problem of insufficient model generalization ability. When the network topology changes, the model still has high reliability and excellent generalization ability, and still has excellent adaptability when facing complex changes in highway microgrids.

[0035] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] Example 1

[0037] Given that measurement points on various sections of a highway microgrid contain a wealth of information when a fault occurs, utilizing the reactive power and voltage at these measurement points can better pinpoint the fault. Drawing on measurement point classification and the successful application of AI technology in power systems, this invention proposes a fault location method for highway microgrids, as shown in Figures 1 and 2. The method includes:

[0038] Step 100: Obtain the reactive power amplitude and direction at measurement points on each line section of the highway microgrid. Generally, the measurement points are located at the beginning of each line section and the outlet of the distributed power source. In practical applications, it is also possible to collect the voltage amplitude and phase angle information of the measurement points of the highway microgrid in real time.

[0039] In real-time fault location applications, real-time measurement information provided by measurement points on each section of the microgrid is collected in real time; when the start-up criterion is activated (protection is activated), the reactive power amplitude and reactive power direction on each line section are collected.

[0040] Step 200: Calculate the fault feature vector of each measurement point based on the reactive power amplitude and reactive power direction at each measurement point; the fault feature vectors of multiple measurement points constitute the test data set.

[0041] Specifically, for a given highway microgrid, there are N sections with corresponding N measurement points. The formula for calculating the fault feature vector of each measurement point is as follows:

[0042] ;

[0043] Among them, Q i V represents the reactive power amplitude at the i-th measurement point. i h represents the direction of reactive power at the i-th measurement point. i This represents the fault feature vector of the i-th measurement point.

[0044] Step 300: Input the data set to be tested into the microgrid fault detection model to obtain the faulty line section on the highway microgrid; the microgrid fault detection model is obtained by training and optimizing the graph attention network based on the fault sample set using multiple kernel functions; the fault samples in the fault sample set include the fault simulation feature vectors of multiple measurement points on each line section of the highway microgrid and the corresponding fault locations.

[0045] The fault sample set and the microgrid fault detection model can be constructed offline. After the test data set is obtained through real-time online acquisition and calculation, it is input into the pre-constructed microgrid fault detection model. Based on the multi-core graph attention network, the fault location is adaptively completed, realizing the fault diagnosis of highway microgrid.

[0046] (1) The process of determining the fault sample set includes:

[0047] 11) Establish a power system topology model of the highway microgrid system; specifically, establish a power system topology model based on PSCAD / EMTDC simulation software.

[0048] 12) Based on the different power systems connected to each line section in the highway microgrid system, the power system topology model is simulated to obtain the original power grid fault dataset and the fault location corresponding to each original power grid fault data; the original power grid fault data includes the reactive power amplitude and reactive power direction at each measurement point corresponding to the power system topology model.

[0049] Specifically, considering the integration of power systems with varying strengths and the use of hybrid wind and solar power systems, phase-to-phase short-circuit fault conditions are simulated under different operating conditions. Short-circuit faults are set at measurement points, and fault data under various operating conditions are collected to establish a fault sample library with a total of 6000 samples. These samples are then divided into training and test sets at a ratio of 4:1.

[0050] 13) Based on the reactive power amplitude and reactive power direction at each measurement point corresponding to the power system topology model, determine the fault simulation feature vector of each measurement point; the fault simulation feature vectors of multiple measurement points and the corresponding fault locations constitute a fault sample; multiple fault samples constitute a fault sample set.

[0051] Among them, (2) the construction process of the microgrid fault detection model specifically includes:

[0052] 21) Calculate the attention coefficients of graph attention networks based on multiple kernel functions.

[0053] Specifically, as shown in Figure 3, the formula for calculating the attention coefficient of a graph attention network is as follows:

[0054]

[0055] Where W is a shared weight coefficient matrix, which transforms the original measurement point features from F dimensions to... Dimension, then through a function This is mapped to an attention weight. This attention weight e ij This represents the importance of measurement point j relative to i. In graph attention networks, a single-layer feedforward neural network and a Leaky ReLU are typically chosen as non-linear activation functions to calculate e. ij :

[0056]

[0057] Where || denotes concatenation, and a is a vector parameter. To preserve the original graph's structural information, similar to message network propagation, only the attention of measurement point i's neighboring measurement point j is calculated, followed by normalization and fusion. The normalized attention weights are:

[0058]

[0059] e above ij Substituting the calculation formula into the normalized formula for attention weights, we obtain:

[0060]

[0061] Based on this attention weight, by fusing information from all neighboring measurement points, the updated measurement point features can be obtained:

[0062]

[0063]

[0064] Furthermore, as shown in Figure 4, the formulas for calculating the attention coefficient using various kernel functions are as follows:

[0065]

[0066] Here, ker represents the set of kernel functions. Automatic hyperparameter tuning is performed based on a hypothesis substitution method. It is assumed that the optimal kernel function parameters lie within a pre-given set of candidate kernel functions, and the corresponding constraints can be written using this assumption method:

[0067]

[0068] in, By setting the candidate kernel functions to polynomial, Gaussian, and Sigmoid kernels, and defining the set of candidate kernel functions, the formula for calculating the attention coefficients of the graph attention network is expressed as:

[0069]

[0070] Among them, a ij Softmax represents the attention coefficients of a graph attention network. j() indicates normalization, e ij The attention weight represents the importance of the j-th measurement point relative to the i-th measurement point, h. i h represents the fault feature vector input at the i-th measurement point. j Ker represents the fault feature vector input at the j-th measurement point. t () represents the set of t kernel functions, N i Let i represent the set of adjacent measurement points of the i-th measurement point.

[0071] 22) Input the fault sample set and the attention coefficient into a preset classifier for training to obtain the optimal classifier; the optimal classifier is a microgrid fault detection model.

[0072] The calculation formula for the operating status of measurement points on each line section in the highway microgrid, output by the microgrid fault detection model, is as follows:

[0073]

[0074] Among them, h i ' represents the operating state of the i-th measurement point, and σ represents the activation function, typically the ReLU function. Indicates a ij Normalized values ​​among K different attention mechanisms, a ij N represents the attention coefficients of a graph attention network. i Let W represent the set of neighboring measurement points of the i-th measurement point, where K is the number of neighboring measurement points; k Let represent the shared weight coefficient matrix under the k-th attention mechanism.

[0075] Step 300 specifically includes:

[0076] 1) Input the data set to be tested into the microgrid fault detection model, and obtain the operating status of the measurement points on each line section of the highway microgrid output by the microgrid fault detection model; the operating status is a 0-1 value.

[0077] Specifically, for an actual fault in a highway microgrid, the reactive power amplitude and direction information of each measurement point are input into the microgrid fault detection model, and the microgrid fault detection model outputs the operating status of each measurement point. , It is a one-dimensional vector, where each element is either 0 or 1.

[0078] 2) For any line segment, if the operating state of the line segment is 0, and the operating state of the line segment connected to the input terminal of the line segment is 1, then the line segment is considered a faulty line segment. That is, when the state information of the front-end measurement point of a line segment is 1 and the state information of the rear-end measurement point is 0, the segment that changes from 1 to 0 is the faulty segment, thus completing the fault segment location of the highway microgrid.

[0079] In a specific practical application, taking the fault in section 18 as an example, when the fault occurs in section 18, the output results of each measurement point are shown in Table 1.

[0080] Table 1

[0081]

[0082] As shown in Table 1, the status number of measurement point 17 is 1, and the status number of measurement point 18 is 0, indicating that the fault occurs in section 18. Using the method of this invention to locate the faulty section, the sample classification accuracy can reach 100%.

[0083] Preferably, the method further includes: evaluating the model performance based on the test set, selecting accuracy (ACC), recall (REC), and precision (PRE) as the basic evaluation metrics, and further selecting the F1-score, which is a weighted harmonic mean that takes into account both REC and PRE, with a value range of [0,1]. Specifically, it includes the following steps:

[0084] 1) Input the preset test sample set into the microgrid fault detection model to determine the fault detection result set.

[0085] 2) Verify the one-to-one correspondence between the fault detection result set and the preset test sample set to obtain the confusion matrix corresponding to the microgrid fault detection model. The confusion matrix includes TP, FN, TN, and FP, where TP is the number of samples that are actually fault measurement points but are identified as fault measurement points, FN is the number of samples that are actually fault measurement points but are identified as non-fault measurement points, TN is the number of samples that are actually non-fault measurement points but are identified as non-fault measurement points, and FP is the number of samples that are actually non-fault measurement points but are identified as fault measurement points.

[0086] 3) Based on the confusion matrix corresponding to the microgrid fault detection model, calculate the accuracy (ACC), recall (REC), and precision (PRE) of the confusion matrix corresponding to the microgrid fault detection model. The calculation formulas are as follows:

[0087] ;

[0088] ;

[0089] .

[0090] 4) Calculate the F1 score based on the recall and precision; both the F1 score and precision are used to characterize the fault detection quality of the microgrid fault detection model. The formula for calculating the F1 score is as follows:

[0091] .

[0092] Table 2 below shows the average accuracy under different fault conditions in a specific example.

[0093] Table 2. Mean accuracy under different fault conditions

[0094]

[0095] To better demonstrate the superiority of the proposed fault location scheme, tests were conducted with data from two measurement points lost. The accuracy rates of the fault location method based on GAT and the fault location method proposed in this invention are shown in Table 3.

[0096] Table 3

[0097]

[0098] As shown in Table 3, the fault location accuracy of the kernel-based GKAT model is higher than that of GAT, reaching 99.4%. Compared with the traditional fault location scheme based on neural networks for highway microgrids, the fault location scheme proposed in this invention has greatly improved accuracy.

[0099] In summary, this invention provides real-time measurement information from real-time measurement points. When protection is activated, it extracts the reactive power amplitude and direction, as well as the voltage amplitude and phase angle of each measurement point; calculates feature vectors, and inputs them into the stored GKAT model to complete fault location of the highway microgrid using adaptive measurement point classification based on graph attention networks. Furthermore, in real-time applications, after each action decision is completed, the results are verified and fed back to the sample database, enabling periodic offline learning and thus improving model performance.

[0100] This invention maps electrical measurement points and lines of highway microgrids to vertices and edges in a graph attention network. It calculates attention coefficients based on the similarity of fault features between adjacent vertices, better integrating the correlation between vertex features into the fault location model. This improves the fault location model's adaptability to topology changes and fully leverages the advantages of artificial intelligence algorithms, such as strong global information mining capabilities and fast computation speed. It requires no manual tuning and has strong adaptability, solving the problem of weak generalization ability in traditional data-driven fault location methods for highway microgrids.

[0101] Example 2

[0102] As shown in Figure 5, in order to implement the technical solution in Embodiment 1, this embodiment provides a highway microgrid fault location system, including:

[0103] The measurement data acquisition module 101 is used to acquire the reactive power amplitude and reactive power direction at measurement points on various line sections in the highway microgrid.

[0104] The feature vector calculation module 201 is used to calculate the fault feature vector of each measurement point based on the reactive power amplitude and reactive power direction at each measurement point; the fault feature vectors of multiple measurement points constitute the test data group.

[0105] The fault location module 301 is used to input the data set to be tested into the microgrid fault detection model to obtain the faulty line section on the highway microgrid. The microgrid fault detection model is obtained by training and optimizing the graph attention network based on the fault sample set using multiple kernel functions. The fault samples in the fault sample set include the fault simulation feature vectors of multiple measurement points on each line section of the highway microgrid and the corresponding fault locations.

[0106] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0107] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for fault location in a highway microgrid, characterized in that, The method includes: acquiring the reactive power amplitude and direction at measurement points on each line section of a highway microgrid; the measurement points are located at the beginning of each line section and the outlet of the distributed power source; calculating the fault feature vector of each measurement point based on the reactive power amplitude and direction; the fault feature vectors of multiple measurement points constitute a test data set; inputting the test data set into a microgrid fault detection model to obtain the faulty line section on the highway microgrid; the microgrid fault detection model is obtained by training and optimizing a graph attention network based on a fault sample set using multiple kernel functions; the fault samples in the fault sample set include the fault simulation feature vectors of multiple measurement points on each line section of the highway microgrid and the corresponding fault locations; the... The process of determining the fault sample set includes: establishing a power system topology model of the highway microgrid system; simulating the power system topology model based on different power systems connected to each line section of the highway microgrid system to obtain the original power grid fault dataset and the fault location corresponding to each original power grid fault data; the original power grid fault data includes the reactive power amplitude and reactive power direction at each measurement point corresponding to the power system topology model; determining the fault simulation feature vector of each measurement point according to the reactive power amplitude and reactive power direction at each measurement point corresponding to the power system topology model; the fault simulation feature vectors of multiple measurement points and the corresponding fault locations constitute a fault sample; multiple fault samples constitute a fault sample set.

2. The method for fault location in a highway microgrid according to claim 1, characterized in that, The test data set is input into the microgrid fault detection model to obtain the faulty line sections on the highway microgrid. Specifically, this includes: inputting the test data set into the microgrid fault detection model and obtaining the operating status of the measurement points on each line section of the highway microgrid output by the microgrid fault detection model; the operating status is a value of 0-1; for any line section, if the operating status of the line section is 0 and the operating status of the line section connected to the input terminal of the line section is 1, then the line section is characterized as a faulty line section.

3. The method for fault location in a highway microgrid according to claim 2, characterized in that, The construction process of the microgrid fault detection model specifically includes: calculating the attention coefficients of the graph attention network based on multiple kernel functions; inputting the fault sample set and the attention coefficients into a preset classifier for training to obtain the optimal classifier; the optimal classifier is the microgrid fault detection model.

4. The method for fault location in a highway microgrid according to claim 3, characterized in that, The calculation of attention coefficients for graph attention networks based on multiple kernel functions specifically includes: according to the formula Calculate the attention coefficients of the graph attention network; where a ij Softmax represents the attention coefficients of a graph attention network. j () indicates normalization, e ij The attention weight represents the importance of the j-th measurement point relative to the i-th measurement point, h. i h represents the fault feature vector input at the i-th measurement point. j This represents the fault feature vector input at the j-th measurement point. ker t () represents the set of t kernel functions, N i Let i represent the set of adjacent measurement points of the i-th measurement point.

5. The method for fault location in a highway microgrid according to claim 3, characterized in that, The calculation formula for the operating status of measurement points on each line section in the highway microgrid, output by the microgrid fault detection model, is as follows: Among them, h i ' represents the operating state of the i-th measurement point, and σ represents the activation function. Indicates a ij Normalized values ​​among K different attention mechanisms, a ij N represents the attention coefficients of a graph attention network. i Let W represent the set of neighboring measurement points of the i-th measurement point, where K is the number of neighboring measurement points; k Let represent the shared weight coefficient matrix under the k-th attention mechanism.

6. The method for fault location in a highway microgrid according to claim 1, characterized in that, The process of determining the fault feature vector specifically includes: according to the formula Determine the fault feature vector; where Q i V represents the reactive power amplitude at the i-th measurement point. i h represents the direction of reactive power at the i-th measurement point. i This represents the fault feature vector of the i-th measurement point.

7. The method for fault location in a highway microgrid according to claim 1, characterized in that, The method further includes: inputting a preset test sample set into the microgrid fault detection model to determine a fault detection result set; verifying the one-to-one correspondence between the fault detection result set and the preset test sample set to obtain a confusion matrix corresponding to the microgrid fault detection model; calculating the accuracy, recall, and precision of the confusion matrix corresponding to the microgrid fault detection model based on the confusion matrix; calculating the F1 score based on the recall and precision; and using both the F1 score and the accuracy to characterize the fault detection quality of the microgrid fault detection model.

8. A fault location system for a highway microgrid, employing the fault location method for a highway microgrid as described in any one of claims 1-7, characterized in that, The system includes: a measurement data acquisition module for acquiring the reactive power amplitude and direction at measurement points on various line sections in the highway microgrid; a feature vector calculation module for calculating the fault feature vector of each measurement point based on the reactive power amplitude and direction at each measurement point; the fault feature vectors of multiple measurement points constitute a set of data to be measured; and a fault location module for inputting the set of data to be measured into a microgrid fault detection model to obtain the faulty line sections on the highway microgrid; the microgrid fault detection model is obtained by training and optimizing a graph attention network based on a fault sample set using multiple kernel functions; the fault samples in the fault sample set include the fault simulation feature vectors of multiple measurement points on various line sections in the highway microgrid and the corresponding fault locations.

Images