Power distribution network fault inspection method and device based on unmanned aerial vehicle, and storage medium

Abstract

The invention discloses a power distribution network fault inspection method and device based on an unmanned aerial vehicle, and a storage medium, relates to the technical field of power grid detection, and mainly aims to improve the inspection efficiency and the inspection accuracy of a power distribution network. The method comprises the steps that a fault detection signal of a target power distribution network is responded, an inspection path of an unmanned aerial vehicle for inspecting the target power distribution network is acquired, and an image acquisition device is installed on the unmanned aerial vehicle; controlling the unmanned aerial vehicle to fly according to the inspection path, and controlling the unmanned aerial vehicle to use the image acquisition device to acquire a distribution network image of the target distribution network in real time in the flight process; and inputting the distribution network image into a preset fault identification model for fault identification, and obtaining a fault identification result of the target distribution network. The method is suitable for the application scene of fault detection of the power distribution network.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid detection, and particularly to a method, device and storage medium for fault inspection of a distribution network based on an unmanned aerial vehicle (UAV). Background Art

[0002] With the continuous development of the power system, the scale and complexity of the distribution network are increasing day by day. In order to ensure the stable operation of the distribution network, it is necessary to detect faults in the distribution network.

[0003] Currently, faults in the distribution network are detected manually. However, this manual detection method is greatly affected by human subjective factors, and there may be situations of missed detection or detection errors. At the same time, manual detection is time-consuming and laborious. Summary of the Invention

[0004] The present invention provides a method, device and storage medium for fault inspection of a distribution network based on an unmanned aerial vehicle, mainly aiming at improving the inspection efficiency and inspection accuracy of the distribution network.

[0005] According to the first aspect of the present invention, there is provided a method for fault inspection of a distribution network based on an unmanned aerial vehicle, including:

[0006] In response to a fault detection signal for a target distribution network, obtaining an inspection path of an unmanned aerial vehicle for inspecting the target distribution network, wherein an image acquisition device is installed on the unmanned aerial vehicle;

[0007] Controlling the unmanned aerial vehicle to fly according to the inspection path, and controlling the unmanned aerial vehicle to use the image acquisition device to collect real-time distribution network images of the target distribution network during flight;

[0008] Inputting the distribution network images into a preset fault recognition model for fault recognition to obtain a fault recognition result of the target distribution network.

[0009] Optionally, an obstacle recognition device is installed on the unmanned aerial vehicle;

[0010] The controlling the unmanned aerial vehicle to fly according to the inspection path includes:

[0011] During the process of controlling the unmanned aerial vehicle to fly according to the inspection path, using the obstacle recognition device to collect obstacles in the pre-flight direction of the unmanned aerial vehicle in real time, and based on the position information of the obstacles, optimizing the inspection path in real time to obtain the optimized inspection path;

[0012] Controlling the unmanned aerial vehicle to fly according to the optimized inspection path.

[0013] Optionally, the obtaining an inspection path of an unmanned aerial vehicle for inspecting the target distribution network includes:

[0014] Obtain the geographical layout information, historical fault information, and inspection requirement information of the target distribution network;

[0015] Based on the geographical layout information, historical fault information, and inspection requirement information, determine the inspection start point and inspection end point for inspecting the target distribution network, and determine the geographical attribute information between the inspection start point and the inspection end point;

[0016] Based on the inspection start point, the inspection end point, and the geographical attribute information, plan the inspection path of the unmanned aerial vehicle for inspecting the target distribution network.

[0017] Optionally, the inputting the distribution network image into a preset fault recognition model for fault recognition to obtain the fault recognition result of the target distribution network includes:

[0018] Obtain a preset fault recognition model, where the preset fault recognition model includes an input layer for inputting an image, a convolutional layer for extracting image features, a pooling layer for downsampling the image features, and a fully connected layer for fault recognition;

[0019] Input the distribution network image into the preset fault recognition model, input the distribution network image into the convolutional layer through the input layer, extract the distribution network image features through the convolutional layer, obtain the downsampled image features by downsampling the distribution network image features through the pooling layer, and perform fault recognition on the downsampled image features through the fully connected layer to obtain the fault recognition result of the target distribution network.

[0020] Optionally, before obtaining the fault recognition prompt information, the method further includes:

[0021] Obtain at least one initial fault recognition model and obtain a sample data set, where the sample data set includes sample distribution network images with fault annotation information;

[0022] Based on the number of models of the initial fault recognition model, divide the sample data set into multiple groups of training data and multiple groups of test data, train the corresponding initial fault recognition models using each group of training data, test the corresponding trained initial fault recognition models using each group of test data, and use the trained initial fault recognition models that meet the test conditions as the preset fault recognition models.

[0023] Optionally, the method further includes:

[0024] Obtain fault recognition prompt information;

[0025] Determine the prompt feature vector corresponding to the fault identification prompt information and the image feature vector corresponding to the distribution network image respectively;

[0026] Perform cross-processing on the prompt feature vector and the image feature vector to obtain a fault cross vector. Among them, the method for performing cross-processing on the prompt feature vector and the image feature vector includes:

[0027] Perform feature-level cross-processing on the prompt feature vector and the image feature vector to obtain a feature cross vector; perform element-level cross-processing on the prompt feature vector and the image feature vector to obtain an element cross vector; perform low-order cross-processing on the prompt feature vector and the image feature vector to obtain a low-order cross vector;

[0028] Use a preset transformation function to perform transformation processing on the feature cross vector, element cross vector, and low-order cross vector to obtain the fault cross vector;

[0029] Input the fault cross vector into the preset fault identification model for fault identification to obtain the fault identification result of the target distribution network.

[0030] Optionally, after inputting the distribution network image into the preset fault identification model for fault identification to obtain the fault identification result of the target distribution network, the method further includes:

[0031] Generate an inspection report for inspecting the target distribution network based on the fault identification result, and send the inspection report to the client.

[0032] According to a second aspect of the present invention, there is provided a distribution network fault inspection device based on an unmanned aerial vehicle, including:

[0033] An acquisition unit, configured to respond to a fault detection signal of a target distribution network, and acquire an inspection path of an unmanned aerial vehicle for inspecting the target distribution network, wherein an image acquisition device is installed on the unmanned aerial vehicle;

[0034] An image acquisition unit, configured to control the unmanned aerial vehicle to fly according to the inspection path, and control the unmanned aerial vehicle to use the image acquisition device to collect real-time distribution network images of the target distribution network during flight;

[0035] A fault identification unit, configured to input the distribution network image into a preset fault identification model for fault identification to obtain the fault identification result of the target distribution network.

[0036] According to a third aspect of the present invention, there is provided a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned distribution network fault inspection method based on an unmanned aerial vehicle is implemented.

[0037] According to a fourth aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the above-mentioned method for fault inspection of a distribution network based on an unmanned aerial vehicle is implemented.

[0038] A method, device, and storage medium for fault inspection of a distribution network based on an unmanned aerial vehicle provided by the present invention, compared with the current method of manually detecting faults in a distribution network, the present invention obtains an inspection path of an unmanned aerial vehicle for inspecting a target distribution network in response to a fault detection signal for the target distribution network, wherein an image acquisition device is installed on the unmanned aerial vehicle; then controls the unmanned aerial vehicle to fly according to the inspection path, and controls the unmanned aerial vehicle to use the image acquisition device to collect distribution network images of the target distribution network in real time during flight; finally inputs the distribution network images into a preset fault recognition model for fault recognition to obtain a fault recognition result of the target distribution network. Thus, by using an unmanned aerial vehicle equipped with an image acquisition device to collect distribution network images in real time and performing intelligent analysis on the distribution network images to identify faults in the distribution network, it can avoid the time and effort consumed by manual on-site inspections. At the same time, it can also avoid missed inspections and detection errors caused by the negligence of inspection personnel and uneven technical levels. Therefore, the present invention can improve the inspection accuracy and inspection efficiency of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0040] Figure 1 A flowchart of a method for fault inspection of a distribution network based on an unmanned aerial vehicle provided by an embodiment of the present invention is shown;

[0041] Figure 2 A flowchart of another method for fault inspection of a distribution network based on an unmanned aerial vehicle provided by an embodiment of the present invention is shown;

[0042] Figure 3 A schematic structural diagram of a device for fault inspection of a distribution network based on an unmanned aerial vehicle provided by an embodiment of the present invention is shown;

[0043] Figure 4 A schematic structural diagram of another device for fault inspection of a distribution network based on an unmanned aerial vehicle provided by an embodiment of the present invention is shown;

[0044] Figure 5 A schematic physical structure diagram of a computer device provided by an embodiment of the present invention is shown. Detailed Embodiments

[0045] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments may be combined with each other.

[0046] Currently, the method of manually detecting faults in the distribution network is greatly affected by human subjective factors, and there may be situations of missed detection or detection errors. At the same time, manual detection is time-consuming and laborious.

[0047] To solve the above problems, an embodiment of the present invention provides a method for fault inspection of a distribution network based on an unmanned aerial vehicle (UAV), as Figure 1 shown, the method includes:

[0048] 101. In response to a fault detection signal for a target distribution network, obtain the inspection path of a UAV for inspecting the target distribution network, wherein an image acquisition device is installed on the UAV.

[0049] The embodiment of the present invention performs fault identification of the distribution network based on a fault identification system. The fault identification system includes a ground control center and a UAV platform, and the ground control center is electrically connected (i.e., wirelessly signal-connected) to the UAV platform. The UAV is used to collect distribution network images in real time and transmit the images collected in real time to the ground control center, and the ground control information analyzes the images to identify faults in the distribution network. At the same time, the inspection path of the UAV is also planned by the ground control center.

[0050] Among them, the image acquisition device can be a high-definition camera such as an infrared camera, various image acquisition sensors, etc. The image acquisition device can be located at the nose part and the belly part of the UAV, etc.; an obstacle identification device such as a lidar is also installed on the UAV, and the obstacle identification device can be located on both sides of the UAV body.

[0051] For the embodiment of the present invention, when receiving a fault detection signal for a target distribution network, it is first necessary to specify the inspection path of the UAV, that is, the flight path. Based on this, step 101 specifically includes: obtaining the geographical layout information, historical fault information, and inspection requirement information of the target distribution network; based on the geographical layout information, historical fault information, and inspection requirement information, determining the inspection start point and the inspection end point for inspecting the target distribution network, and determining the geographical attribute information between the inspection start point and the inspection end point; based on the inspection start point, the inspection end point, and the geographical attribute information, planning the inspection path of the UAV for inspecting the target distribution network.

[0052] Among them, the geographical layout information includes the physical location of the distribution network, the line direction, the equipment distribution, etc., which can be obtained through devices such as geographic information systems; the historical fault information refers to information such as the fault type, the fault equipment type, the fault frequency, and the fault location of the target distribution network in the past; the inspection demand information refers to the specific fault detection types for the target distribution network, such as line fault detection, equipment fault detection, etc., and demand information such as the equipment and areas to be focused on for inspection; the geographical attribute information refers to factors such as the terrain, obstacles, wind speed, and temperature between the inspection starting point and the inspection ending point that may affect the flight and inspection effect of the UAV.

[0053] Specifically, a preset inspection starting point prediction model is pre-constructed. The specific construction method includes: obtaining an initial prediction model and a sample data set. The sample data set includes sample data with starting point annotation information, and the sample data is the geographical layout information, historical fault information, and inspection demand information of the sample distribution network; dividing the sample data set into training data and test data, training the initial prediction model with the training data, then testing the trained initial prediction model with the test data, and finally taking the trained initial prediction model that meets the test conditions as the preset inspection starting point prediction model. Further, input the geographical layout information, historical fault information, and inspection demand information of the target distribution network into the preset inspection starting point prediction model for starting point prediction to obtain the inspection starting point and inspection end point for inspecting the target distribution network. Further, based on the inspection starting point, inspection end point, and geographical attribute information, formulate the inspection path of the unmanned aerial vehicle for inspecting the target distribution network. The specific method for formulating the inspection path includes: respectively determining the starting point feature vector corresponding to the inspection starting point, the end point feature vector corresponding to the inspection end point, and the geographical feature vector corresponding to the geographical attribute information; performing feature fusion on the starting point feature vector and the end point feature vector to obtain a fused feature vector; determining the principal component feature vector between the fused feature vector and the geographical feature vector; inputting the principal component feature vector into a preset path prediction model for path prediction to obtain the inspection path of the unmanned aerial vehicle for inspecting the target distribution network. Among them, the method for determining the principal component feature vector includes: constructing a feature matrix based on the fused feature vector and the geographical feature vector; determining the element mean of each element in the feature matrix, and subtracting each element in the feature matrix from the element mean to obtain a centralized feature matrix; determining the covariance matrix corresponding to the centralized feature matrix; performing eigenvalue decomposition on the covariance matrix to obtain matrix eigenvalues and matrix eigenvectors; based on the magnitudes of the matrix eigenvalues, selecting a preset number of matrix eigenvectors from the matrix eigenvectors and determining the preset number of matrix eigenvectors as the principal component feature vectors. The preset path prediction model is pre-trained based on a sample data set, and the sample data set includes sample data with path label information (the sample data is the inspection starting point, inspection end point, and geographical attribute information between the inspection starting point and inspection end point of the distribution network that has successfully completed the inspection).

[0054] Among them, the preset number is a value set according to actual needs (experience). Specifically, if the fused feature vector is [11 2 4 2] and the geographical feature vector is [1 3 3 4 4], the feature matrix formed by the fused feature vector and the geographical feature vector is as follows:

[0055]

[0056] Among them, T represents the feature matrix. Then, the element mean corresponding to the elements in the first row of the feature matrix is determined, and each element in the first row is subtracted from this mean to obtain the subtraction results corresponding to each element in the first row. At the same time, the element mean corresponding to each element in the second row is determined, and each element in the second row is subtracted from the corresponding element mean to obtain the subtraction results corresponding to each element in the second row. Finally, the subtraction results corresponding to each element in the first row and the subtraction results corresponding to each element in the second row are used to form the centralized feature matrix as follows:

[0057]

[0058] Among them, T z represents the centralized feature matrix. Then, the covariance between each row of elements in the centralized feature matrix is calculated. For example, the covariance between the elements in the first row Cov(X1, X1) = [(-1) 2 + (-1) 2 + (0) 2 + (2) 2 + (0) 2 / (5 - 1) = 1.5, and the covariance between the elements in the first row and the elements in the second row Cov(X1, X2) = [(-1)×(-2) + (-1)×0 + 0×0 + 2×1 + 0×1] / (5 - 1) = 1. Thus, the covariance between all rows of elements can be calculated. Then, a covariance matrix is formed by all the covariances, and eigenvalue decomposition is performed on this covariance matrix. The specific decomposition method is as follows: First, the eigenvector group of the covariance matrix is determined, and the corresponding eigenvalues are calculated according to the eigenvector group. For example, if the obtained eigenvector group is k groups, then the corresponding eigenvalues are k, and each eigenvalue has its corresponding eigenvector. Then, the k eigenvalues are sorted in descending order to obtain the sorted eigenvalues. Then, the top n (preset quantity) eigenvalues are selected from the sorted eigenvalues, and the eigenvectors corresponding to the top n eigenvalues are determined as the principal component eigenvectors. Through determining the principal component eigenvectors between the geographical feature vectors and the fusion feature vectors in the embodiments of the present invention, an effective dimensionality reduction technology is realized, which can convert the original high-dimensional data into a few linearly independent principal components. These principal components can reflect most of the important information of the original data, while greatly reducing the dimension of the data. In the path prediction model, reducing the dimension of the input data can simplify the model structure, reduce the calculation amount, thereby improving the operation efficiency of the model, and further improving the path prediction efficiency. At the same time, by comprehensively analyzing the geographical layout information, historical fault information, inspection demand information, and geographical attribute information between the inspection starting point and the inspection ending point of the distribution network in the embodiments of the present invention to formulate the inspection path, the accuracy of formulating the inspection path can be improved, and further the fault detection accuracy of the distribution network can be improved.

[0059] 102. Control the drone to fly according to the inspection path, and control the drone to use the image acquisition device to collect the power grid images of the target power distribution network in real time during the flight.

[0060] For the embodiments of the present invention, after specifying the inspection path of the drone, control the drone to fly according to the inspection path. During the flight, control the image acquisition device on the drone to collect the power grid images of the target power distribution network in real time, and then identify the faults of the target power distribution network in real time by intelligently analyzing the power grid images. The embodiments of the present invention collect power grid images through the drone and perform intelligent analysis on the power grid images to identify the faults of the power distribution network, which can avoid the time and energy consumed by manual on-site inspections. At the same time, it can also avoid the situations of missed inspections and detection errors caused by the negligence of inspection personnel and the uneven technical levels. Therefore, the present invention can improve the inspection accuracy and efficiency of the power distribution network.

[0061] 103. Input the power grid image into a preset fault identification model for fault identification to obtain the fault identification result of the target power distribution network.

[0062] For the embodiments of the present invention, in order to improve the identification accuracy of the preset fault identification model, it is necessary to pre-train and construct the preset fault identification model. Based on this, the method includes: obtaining at least one initial fault identification model and obtaining a sample data set, where the sample data set includes sample power grid images with fault annotation information; dividing the sample data set into multiple groups of training data and multiple groups of test data based on the number of models of the initial fault identification model, training the corresponding initial fault identification model with each group of training data, and testing the corresponding trained initial fault identification model with each group of test data, and using the trained initial fault identification model that meets the test conditions as the preset fault identification model.

[0063] Among them, the model structures of each initial fault identification model can be the same or different; the sample power grid images for constructing the sample data set can be from the power distribution networks of multiple regions, and each sample power grid image is marked with fault information.

[0064] Specifically, during the model training process, the sample data set can be first divided into multiple groups of training data and multiple groups of test data according to the number of initial fault recognition models. Then, multiple data pairs in each group of training data are used to train each initial fault recognition model respectively. Among them, the sample distribution network image is used as the input data, and the fault information is used as the output data. After that, multiple data pairs in each group of test data are used to test the trained initial fault recognition models respectively. Finally, the trained initial fault recognition model that meets the test conditions is used as the preset fault recognition model. Among them, meeting the test conditions can be that the number of training times reaches the requirement or the fault recognition accuracy of the trained initial fault recognition model meets specific requirements.

[0065] For example, in the embodiments of the present invention, an initial fault recognition model 1 and an initial fault recognition model 2 are respectively constructed. At the same time, the sample data set can also be randomly divided into training data 1 and training sub-data 2, as well as test data 1 and test data 2. Further, the initial fault recognition model 1 can be trained using training data 1, and the initial fault recognition model 2 can be trained using training data 1. After that, the trained initial fault recognition model 1 is tested using test data 1, and the trained initial fault recognition model 2 is tested using test data 1. Finally, the model that meets the test conditions is selected from the trained initial fault recognition model 1 and the initial fault recognition model 2 as the preset fault recognition model. In the method provided by the embodiments of the present invention, by constructing and training multiple initial fault recognition models, calculating and comparing the accuracy of each initial fault recognition model among the multiple initial fault recognition models, and using the initial fault recognition model with the highest accuracy as the preset fault recognition model for actual use finally, the subsequent recognition of distribution network faults can be made more accurate.

[0066] Further, after the preset fault identification model is constructed, it is necessary to use the preset fault identification model to identify faults in the target distribution network. Based on this, step 103 specifically includes: obtaining a fault identification prompt message; respectively determining a prompt feature vector corresponding to the fault identification prompt message and an image feature vector corresponding to the distribution network image; performing cross-processing on the prompt feature vector and the image feature vector to obtain a fault cross vector; inputting the fault cross vector into the preset fault identification model for fault identification to obtain a fault identification result of the target distribution network. Among them, the method of performing cross-processing on the prompt feature vector and the image feature vector includes: performing feature-level cross-processing on the prompt feature vector and the image feature vector to obtain a feature cross vector; performing element-level cross-processing on the prompt feature vector and the image feature vector to obtain an element cross vector; performing low-order cross-processing on the prompt feature vector and the image feature vector to obtain a low-order cross vector; using a preset transformation function to perform transformation processing on the feature cross vector, the element cross vector, and the low-order cross vector to obtain the fault cross vector.

[0067] Specifically, since the fault identification prompt information and the distribution network image are in different fields, the processing method is to embed the data of the two fields into vectors of the same dimension. For example, if the prompt feature vector is (a1, a2) and the image feature vector is (b1, b2): perform cross at the feature level between different feature vectors, that is, after performing the Hadamard product on all elements between the vectors, perform a convolution transformation under a certain weight to obtain the feature cross vector as f(w*(a1*b1, a2*b2, a3*b3)); at the same time, perform cross at the element level on all feature vector data, that is, after performing the Hadamard product on each element between the vectors, assign different weight values to each product result, and then perform a linear transformation to obtain the element cross vector as f(w1*a1*b1, w2*a2*b2); in addition, perform low-order cross processing on all feature vectors, then assign a weight coefficient to the result of the cross processing, and then perform a linear transformation to obtain the low-order cross vector as f(w2(a1, a2, b1, b2)); finally, combine the above feature cross vector, element cross vector, and low-order cross vector together, and perform a transformation process using a preset function to obtain the fault cross vector. The preset function here can be set according to the actual situation, and this embodiment does not limit it. It should be noted that the above examples are only illustrative and do not limit the embodiments of the present application. Thus, by performing cross processing on the prompt feature vector and the image feature vector, different features can be automatically or explicitly combined to generate new feature combinations, and these combined features may contain complex non-linear relationships between the original features, enabling the model to capture more refined and rich information in the data, that is, being able to make full use of the relationships between various data, extract more implicit features, and take into account both high-order and low-order processing, making the data utilization more sufficient, and the subsequent fault identification result more accurate, meeting the requirements of the actual application scenario.

[0068] Further, after the fault is identified, in order to facilitate subsequent analysis, it is also necessary to generate an inspection report. Based on this, the method includes: generating an inspection report for inspecting the target distribution network based on the fault identification result, and sending the inspection report to the client.

[0069] Specifically, based on the fault identification result, determine the fault type, the location and identification of the faulty device or line, and the fault phenomenon, and generate an inspection report based on the fault type, the location and identification of the faulty device or line, and the fault phenomenon. Then send the inspection report to the client. In another embodiment of the present invention, it is also possible to analyze the cause of the fault based on the fault type, the location and identification of the faulty device or line, and the fault phenomenon, and determine a fault correction strategy based on the cause of the fault. Finally, generate an inspection report based on the fault type, the location and identification of the faulty device or line, the fault phenomenon, the cause of the fault, and the fault correction strategy, and send the inspection report to the client so that professionals on the client side can verify the inspection report. If the verification is correct, directly correct the corresponding fault of the distribution network based on the fault correction strategy in the inspection report. By generating an inspection report in the embodiment of the present invention, it is possible to facilitate the subsequent analysis of distribution network faults. At the same time, by directly generating a fault correction strategy, it is possible to avoid the time wasted by staff in formulating the fault correction strategy and the situation of negligence in formulation, thereby improving the correction efficiency and accuracy of distribution network faults.

[0070] According to a method for inspecting faults in a distribution network based on an unmanned aerial vehicle provided by the present invention, compared with the current method of manually detecting faults in a distribution network, the present invention obtains the inspection path of an unmanned aerial vehicle for inspecting a target distribution network in response to a fault detection signal for the target distribution network, wherein an image acquisition device is installed on the unmanned aerial vehicle; then controls the unmanned aerial vehicle to fly along the inspection path, and controls the unmanned aerial vehicle to use the image acquisition device to collect real-time distribution network images of the target distribution network during flight; finally, inputs the distribution network images into a preset fault identification model for fault identification to obtain a fault identification result of the target distribution network. Thus, by using an unmanned aerial vehicle equipped with an image acquisition device to collect distribution network images in real time and performing intelligent analysis on the distribution network images for fault identification, it is possible to avoid the time and effort consumed by manual on-site inspections. At the same time, it is also possible to avoid the situations of missed inspections and detection errors caused by the negligence of inspection personnel and uneven technical levels. Therefore, the present invention can improve the inspection accuracy and inspection efficiency of the distribution network.

[0071] Further, to better illustrate the process of inspecting faults in the above-mentioned distribution network, as a refinement and extension of the above embodiment, the embodiment of the present invention provides another method for inspecting faults in a distribution network based on an unmanned aerial vehicle, as Figure 2 shown, the method includes:

[0072] 201. In response to a fault detection signal for a target distribution network, obtain the inspection path of an unmanned aerial vehicle for inspecting the target distribution network, wherein an image acquisition device is installed on the unmanned aerial vehicle.

[0073] Specifically, when a fault detection signal for the target distribution network is received, first obtain the inspection path of the unmanned aerial vehicle (UAV) for inspecting the target distribution network, and then control the UAV to fly according to this inspection path. During the flight, control the image acquisition device on the UAV to collect the distribution network images of the target distribution network in real time, and finally identify the faults in the target distribution network by analyzing the distribution network images.

[0074] 202. Control the UAV to fly according to the inspection path, and control the UAV to use the image acquisition device to collect the distribution network images of the target distribution network in real time during the flight.

[0075] Wherein, an obstacle recognition device is also installed on the UAV. For the embodiments of the present invention, after the inspection path is formulated, it is necessary to control the UAV to fly according to this inspection path. Since the obstacles during the UAV flight cannot be accurately predicted during the formulation of the inspection path, based on this, it is necessary to continuously optimize the inspection path in real time. Based on this, step 202 specifically includes: during the process of controlling the UAV to fly according to the inspection path, use the obstacle recognition device to collect the obstacles in the pre-flight direction of the UAV in real time, and based on the position information of the obstacles, optimize the inspection path in real time to obtain the optimized inspection path; control the UAV to fly according to the optimized inspection path.

[0076] Wherein, the obstacles can be: birds, buildings, trees, etc. Specifically, during the flight of the UAV along the inspection path, use the obstacle recognition device such as lidar on the UAV to collect the obstacles in the pre-flight direction of the UAV in real time. When an obstacle is detected, in order to avoid the obstacle, at this time, it is necessary to re-plan the inspection path based on the position information of the obstacle to avoid the obstacle and obtain the optimized inspection path, and finally control the UAV to continue flying according to the optimized inspection path. The embodiments of the present invention can avoid the collision between the obstacle and the UAV and avoid the damage of the obstacle and the UAV by optimizing the inspection path in real time. Further, during the flight of the UAV, control the sensors, cameras and other devices on the UAV to collect the distribution network images of the target distribution network in real time, and finally transmit the distribution network images to the ground control center in real time. The computing device of the ground control center analyzes the distribution network images in real time to identify the faults in the target distribution network.

[0077] 203. Obtain a preset fault recognition model, wherein the preset fault recognition model includes an input layer for inputting images, a convolutional layer for extracting image features, a pooling layer for downsampling the image features, and a fully connected layer for fault recognition.

[0078] 204. Input the distribution network image into a preset fault recognition model. Input the distribution network image into the convolutional layer through the input layer, extract features from the distribution network image through the convolutional layer to obtain distribution network image features, perform downsampling on the distribution network image features through the pooling layer to obtain downsampled image features, and perform fault recognition on the downsampled image features through the fully connected layer to obtain the fault recognition result of the target distribution network.

[0079] Among them, the preset fault recognition model includes: an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. Specifically, input the distribution network image into the convolutional layer through the input layer, extract features from the distribution network image using the convolutional layer to obtain distribution network image features, then input the distribution network image features into the pooling layer for downsampling processing, output the downsampled image features through the pooling layer, input the downsampled image features into the fully connected layer for fault recognition to obtain the fault recognition result of the target distribution network, and finally output the fault recognition result through the output layer and display the fault recognition result. In the embodiment of the present invention, the preset fault recognition model is used to identify faults in the distribution network, avoiding the problems of time-consuming and laborious manual inspections and negligence, thereby improving the detection efficiency and detection accuracy of distribution network faults.

[0080] According to another method for inspecting faults in a distribution network based on an unmanned aerial vehicle provided by the present invention, compared with the current method of manually detecting faults in the distribution network, the present invention obtains the inspection path of the unmanned aerial vehicle for inspecting the target distribution network in response to a fault detection signal for the target distribution network, wherein an image acquisition device is installed on the unmanned aerial vehicle; then control the unmanned aerial vehicle to fly according to the inspection path, and control the unmanned aerial vehicle to use the image acquisition device to collect the distribution network image of the target distribution network in real time during flight; finally, input the distribution network image into a preset fault recognition model for fault recognition to obtain the fault recognition result of the target distribution network. Thus, the unmanned aerial vehicle equipped with an image acquisition device is used to collect the distribution network image in real time, and the distribution network is fault-recognized by intelligent analysis of the distribution network image, which can avoid the time and effort consumed by manual on-site inspections. At the same time, it can also avoid the situation of missed inspections and detection errors caused by the negligence of inspection personnel and uneven technical levels. Therefore, the present invention can improve the inspection accuracy and inspection efficiency of the distribution network.

[0081] Further, as Figure 1 a specific implementation, the embodiment of the present invention provides a device for inspecting faults in a distribution network based on an unmanned aerial vehicle. As shown in Figure 3 , the device includes: an acquisition unit 31, an image acquisition unit 32, and a fault recognition unit 33.

[0082] The obtaining unit 31 can be used to obtain the inspection path of a drone for inspecting the target distribution network in response to a fault detection signal of the target distribution network, where an image acquisition device is installed on the drone.

[0083] The image acquisition unit 32 can be used to control the drone to fly according to the inspection path, and control the drone to use the image acquisition device to collect the distribution network images of the target distribution network in real time during flight.

[0084] The fault identification unit 33 can be used to input the distribution network images into a preset fault identification model for fault identification to obtain the fault identification result of the target distribution network.

[0085] In a specific application scenario, an obstacle identification device is installed on the drone; as Figure 4 shown, in order to control the drone to fly according to the inspection path, the acquisition unit 32 includes a path optimization module 321 and a flight control module 322.

[0086] The path optimization module 321 can be used to, during the process of controlling the drone to fly according to the inspection path, use the obstacle identification device to collect the obstacles in the pre-flight direction of the drone in real time, and based on the position information of the obstacles, optimize the inspection path in real time to obtain the optimized inspection path.

[0087] The flight control module 322 can be used to control the drone to fly according to the optimized inspection path.

[0088] In a specific application scenario, in order to obtain the inspection path of a drone for inspecting the target distribution network, the obtaining unit 31 includes a first obtaining module 311, a first determining module 312, and a path planning module 313.

[0089] The first obtaining module 311 can be used to obtain the geographical layout information, historical fault information, and inspection requirement information of the target distribution network.

[0090] The first determining module 312 can be used to determine the inspection start point and inspection end point for inspecting the target distribution network based on the geographical layout information, historical fault information, and inspection requirement information, and determine the geographical attribute information between the inspection start point and the inspection end point.

[0091] The path planning module 313 can be used to plan the inspection path of a drone for inspecting the target distribution network based on the inspection start point, the inspection end point, and the geographical attribute information.

[0092] In a specific application scenario, for fault identification of distribution network images, the fault identification unit 33 includes a second acquisition module 331 and a fault identification module 332.

[0093] The second acquisition module 331 can be used to obtain a preset fault identification model for fault identification of distribution network images in a specific application scenario. The preset fault identification model includes an input layer for inputting images, a convolutional layer for extracting image features, a pooling layer for downsampling the image features, and a fully connected layer for fault identification.

[0094] The fault identification module 332 can be used to input the distribution network image into the preset fault identification model, input the distribution network image into the convolutional layer through the input layer, extract the distribution network image features through the convolutional layer, obtain downsampled image features by downsampling the distribution network image features through the pooling layer, and perform fault identification on the downsampled image features through the fully connected layer to obtain the fault identification result of the target distribution network.

[0095] In a specific application scenario, for training and constructing a preset fault identification model, the device further includes a construction unit 34.

[0096] The construction unit 34 can be used to obtain at least one initial fault identification model and obtain a sample data set. The sample data set includes sample distribution network images with fault annotation information. Based on the number of models of the initial fault identification model, the sample data set is divided into multiple groups of training data and multiple groups of test data. Each group of training data is used to train the corresponding initial fault identification model, and each group of test data is used to test the corresponding trained initial fault identification model. The trained initial fault identification model that meets the test conditions is used as the preset fault identification model.

[0097] In a specific application scenario, for fault identification of a distribution network, the fault identification unit 33 further includes a second determination module 333 and a feature cross module 334.

[0098] The second acquisition module 331 can also be used to obtain fault identification prompt information.

[0099] The second determination module 333 can be used to respectively determine a prompt feature vector corresponding to the fault identification prompt information and an image feature vector corresponding to the distribution network image.

[0100] The feature cross module 334 can be used to perform cross processing on the prompt feature vector and the image feature vector to obtain a fault cross vector. Among them, the method for performing cross processing on the prompt feature vector and the image feature vector includes: performing feature-level cross processing on the prompt feature vector and the image feature vector to obtain a feature cross vector; performing element-level cross processing on the prompt feature vector and the image feature vector to obtain an element cross vector; performing low-order cross processing on the prompt feature vector and the image feature vector to obtain a low-order cross vector; and using a preset transformation function to perform transformation processing on the feature cross vector, element cross vector, and low-order cross vector to obtain the fault cross vector.

[0101] The fault identification module 332 can be used to input the fault cross vector into the preset fault identification model for fault identification to obtain the fault identification result of the target distribution network.

[0102] In a specific application scenario, in order to generate an inspection report, the device further includes a report generation unit 35.

[0103] The report generation unit 35 can be used to generate an inspection report for inspecting the target distribution network based on the fault identification result and send the inspection report to the client.

[0104] It should be noted that for other corresponding descriptions of each functional module involved in a distribution network fault inspection device based on an unmanned aerial vehicle provided in an embodiment of the present invention, reference can be made to Figure 1 the corresponding description of the method shown, which will not be elaborated here.

[0105] Based on the above as Figure 1 shown in the method, correspondingly, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the following steps are implemented: in response to a fault detection signal of a target distribution network, obtain the inspection path of an unmanned aerial vehicle for inspecting the target distribution network, where an image acquisition device is installed on the unmanned aerial vehicle; control the unmanned aerial vehicle to fly according to the inspection path, and control the unmanned aerial vehicle to use the image acquisition device to collect the distribution network image of the target distribution network in real time during flight; and input the distribution network image into a preset fault identification model for fault identification to obtain the fault identification result of the target distribution network.

[0106] Based on the above as Figure 1 shown in the method and as Figure 3 shown in the embodiment of the device, an embodiment of the present invention further provides an entity structure diagram of a computer device, as Figure 5As shown in the figure, the computer device includes: a processor 41, a memory 42, and a computer program stored on the memory 42 and executable on the processor. Both the memory 42 and the processor 41 are provided on a bus 43. When the processor 41 executes the program, the following steps are implemented: in response to a fault detection signal of a target distribution network, obtain an inspection path of a drone for inspecting the target distribution network, where an image acquisition device is installed on the drone; control the drone to fly according to the inspection path, and control the drone to use the image acquisition device to collect real-time distribution network images of the target distribution network during flight; input the distribution network images into a preset fault recognition model for fault recognition to obtain a fault recognition result of the target distribution network.

[0107] Through the technical solution of the present invention, the present invention obtains an inspection path of a drone for inspecting the target distribution network in response to a fault detection signal of the target distribution network, where an image acquisition device is installed on the drone; then controls the drone to fly according to the inspection path, and controls the drone to use the image acquisition device to collect real-time distribution network images of the target distribution network during flight; finally, inputs the distribution network images into a preset fault recognition model for fault recognition to obtain a fault recognition result of the target distribution network. Thus, by using a drone equipped with an image acquisition device to collect distribution network images in real time and performing intelligent analysis on the distribution network images to identify faults in the distribution network, it can avoid the time and effort consumed by manual on-site inspections. At the same time, it can also avoid missed inspections and detection errors caused by the negligence of inspection personnel and uneven technical levels. Therefore, the present invention can improve the inspection accuracy and efficiency of the distribution network.

[0108] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a different order than here, or they can be separately made into individual integrated circuit modules, or multiple modules or steps among them can be made into a single integrated circuit module to implement. In this way, the present invention is not limited to any specific combination of hardware and software.

[0109] The above is only the preferred embodiment of the present invention and is not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A distribution network fault inspection method based on drones, characterized in that: include: In response to a fault detection signal of a target distribution network, an inspection path of a drone for inspecting the target distribution network is acquired, wherein an image acquisition device is installed on the drone; Controlling the UAV to fly along the inspection path, and controlling the UAV to use the image acquisition device to acquire a distribution network image of the target distribution network in real time during the flight; The distribution network image is input into a preset fault identification model for fault identification to obtain a fault identification result of the target distribution network.

2. The method according to claim 1, characterized in that An obstacle identification device is installed on the drone; The controlling the drone to fly according to the inspection path includes: In the process of controlling the UAV to fly along the inspection path, the obstacle recognition device is used to collect obstacles in the pre-flight direction of the UAV in real time, and based on the position information of the obstacles, the inspection path is optimized in real time to obtain the optimized inspection path; The UAV is controlled to fly according to the optimized inspection path.

3. The method according to claim 1, characterized in that The obtaining of the inspection path of the drone for inspecting the target distribution network includes: Obtaining geographical layout information, historical fault information, and inspection demand information of the target distribution network; Based on the geographic layout information, historical fault information, and inspection demand information, determine an inspection starting point and an inspection end point for inspecting the target distribution network, and determine geographic attribute information between the inspection starting point and the inspection end point; Based on the inspection starting point, the inspection end point, and the geographic attribute information, an inspection path of a drone for inspecting the target distribution network is planned.

4. The method according to claim 1, characterized in that: The step of inputting the distribution network image into a preset fault identification model for fault identification to obtain a fault identification result of the target distribution network includes: Obtaining a preset fault recognition model, wherein the preset fault recognition model includes an input layer for inputting an image, a convolution layer for extracting image features, a pooling layer for downsampling the image features, and a fully connected layer for fault recognition; The distribution network image is input into the preset fault identification model, the distribution network image is input into the convolution layer through the input layer, the distribution network image is feature extracted through the convolution layer to obtain distribution network image features, the distribution network image features are downsampled through the pooling layer to obtain downsampled image features, fault identification is performed on the downsampled image features through the fully connected layer to obtain the fault identification result of the target distribution network.

5. The method according to claim 1, characterized in that Before inputting the distribution network image into a preset fault identification model for fault identification to obtain a fault identification result of the target distribution network, the method further includes: Acquire at least one initial fault identification model and acquire a sample data set, wherein the sample data set includes a sample distribution network image with fault annotation information; Based on the number of models of the initial fault identification model, the sample data set is divided into multiple groups of training data and multiple groups of test data, each group of training data is used to train the corresponding initial fault identification model, and each group of test data is used to test the corresponding trained initial fault identification model, and the trained initial fault identification model that meets the test conditions is used as the preset fault identification model.

6. The method according to claim 1, characterized in that The step of inputting the distribution network image into a preset fault identification model for fault identification to obtain a fault identification result of the target distribution network includes: Obtain fault identification prompt information; Respectively determining a prompt feature vector corresponding to the fault identification prompt information and an image feature vector corresponding to the distribution network image; The prompt feature vector and the image feature vector are cross-processed to obtain a fault cross-vector, wherein the method of cross-processing the prompt feature vector and the image feature vector comprises: Performing feature-level cross processing on the prompt feature vector and the image feature vector to obtain a feature cross vector; performing element-level cross processing on the prompt feature vector and the image feature vector to obtain an element cross vector; performing low-order cross processing on the prompt feature vector and the image feature vector to obtain a low-order cross vector; The characteristic cross vector, the element cross vector, and the low-order cross vector are transformed by using a preset transformation function to obtain the fault cross vector; The fault cross vector is input into the preset fault identification model for fault identification to obtain a fault identification result of the target distribution network.

7. The method according to claim 1, characterized in that After inputting the distribution network image into a preset fault identification model for fault identification and obtaining a fault identification result of the target distribution network, the method further includes: Based on the fault identification result, an inspection report for inspecting the target distribution network is generated, and the inspection report is sent to a client.

8. A distribution network fault inspection device based on drone, characterized in that: include: an acquisition unit, configured to acquire, in response to a fault detection signal of a target distribution network, an inspection path of a drone for inspecting the target distribution network, wherein an image acquisition device is installed on the drone; An image acquisition unit, used to control the UAV to fly along the inspection path, and to control the UAV to use the image acquisition device to acquire a distribution network image of the target distribution network in real time during the flight; The fault identification unit is used to input the distribution network image into a preset fault identification model to perform fault identification and obtain a fault identification result of the target distribution network.

9. 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 method according to any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images