Real-time monitoring and early warning method and apparatus for state of power distribution cabinet

Abstract

The present invention relates to the field of state monitoring of power distribution cabinets. Disclosed are a real-time monitoring and early warning method and apparatus for the state of a power distribution cabinet. The method comprises: receiving multi-angle images of a power distribution cabinet that are collected by power distribution cabinet monitoring cameras; analyzing the multi-angle images of the power distribution cabinet by means of a spatial positioning channel to obtain control mode information of a control button; receiving the input voltage of the power distribution cabinet; inputting the control mode information of the control button and the input voltage of the power distribution cabinet into an electrical-state prediction channel of the power distribution cabinet, and outputting an electrical-state prediction value matrix; obtaining an electrical-state monitoring value matrix by means of an electrical-state sensor; and when the electrical-state prediction value matrix is inconsistent with the electrical-state monitoring value matrix, generating an abnormal operation early warning signal of the power distribution cabinet, and sending the abnormal operation early warning signal to a control terminal of the power distribution cabinet for early warning. The technical problem of low early warning accuracy and efficiency that is present in real-time state monitoring and early warning of existing power distribution cabinets is solved, and a technical effect of improving the accuracy and efficiency of an early warning system is achieved.

Description

A method and device for real-time monitoring and early warning of power distribution cabinet status Technical Field

[0001] This application relates to the field of power distribution cabinet status monitoring, and in particular to a method and device for real-time monitoring and early warning of power distribution cabinet status. Background Technology

[0002] In power systems, switchgear is a key device for power distribution and control, and its operating status directly affects the stability and security of the power system. Traditional switchgear status early warning methods typically require complex spatial modeling to analyze the button status of the switchgear. This method suffers from high computational load, complexity, susceptibility to errors, and low efficiency. These problems not only affect the accuracy and reliability of the early warning system but also limit its widespread application in power systems.

[0003] Currently, the relevant technologies for real-time monitoring and early warning of power distribution cabinet status suffer from technical problems such as low accuracy and efficiency of early warning. Summary of the Invention

[0004] This application provides a method and device for real-time monitoring and early warning of the status of a power distribution cabinet. It employs a method of receiving multi-angle images captured by a monitoring camera of the power distribution cabinet, using a spatial positioning channel to analyze the images to obtain control key mode information, and combining this information with the input voltage of the power distribution cabinet. This information is then input into the electrical status prediction channel of the power distribution cabinet. Through a trained prediction channel, an electrical status prediction matrix is ​​output. Simultaneously, an electrical status monitoring matrix is ​​obtained through an electrical status sensor. The predicted value matrix is ​​compared with the monitoring value matrix. When the two are inconsistent, an abnormal operation early warning signal for the power distribution cabinet is generated and sent to the power distribution cabinet control terminal for execution of the early warning. These technical means improve the accuracy and efficiency of the early warning system.

[0005] This application provides a method for real-time monitoring and early warning of the status of a power distribution cabinet, including:

[0006] The system receives multi-angle images of the power distribution cabinet captured by a monitoring camera; it analyzes these images via a spatial positioning channel to obtain control key mode information; it receives the power distribution cabinet input voltage; it inputs the control key mode information and the power distribution cabinet input voltage into the power distribution cabinet electrical status prediction channel and outputs an electrical status prediction matrix; it obtains an electrical status monitoring matrix via an electrical status sensor; and when the electrical status prediction matrix and the electrical status monitoring matrix are inconsistent, it generates an abnormal operation warning signal for the power distribution cabinet and sends it to the power distribution cabinet control terminal for warning execution.

[0007] In one possible implementation, multi-angle images of the power distribution cabinet captured by the monitoring camera are received, and the following processing is performed:

[0008] A front image of the power distribution cabinet is captured using a binocular camera on the first power distribution cabinet; a first side image of the power distribution cabinet is captured using a binocular camera on the second power distribution cabinet; a second side image of the power distribution cabinet is captured using a binocular camera on the third power distribution cabinet; and a rear image of the power distribution cabinet is captured using a binocular camera on the fourth power distribution cabinet. The front image, the first side image, the second side image, and the rear image of the power distribution cabinet are then added to the multi-angle image of the power distribution cabinet.

[0009] In one possible implementation, the control key control mode information is obtained by parsing multi-angle images of the power distribution cabinet through a spatial positioning channel, and the following processing is performed:

[0010] The multi-angle image of the power distribution cabinet has image acquisition parameter identifiers; through the spatial positioning channel, the image acquisition parameter identifiers and the multi-angle image of the power distribution cabinet are processed to obtain a spatial positioning label for the multi-angle image of the power distribution cabinet; control key identifiers are applied to the multi-angle image of the power distribution cabinet to obtain a control key area identifier for the multi-angle image of the power distribution cabinet; based on the spatial positioning label of the multi-angle image of the power distribution cabinet and the control key area identifier, spatial positioning information of the control key is obtained; based on the spatial positioning information of the control key, the control mode information of the control key is matched according to the control key preset mode library.

[0011] In a possible implementation, the image acquisition parameter identifier and the multi-angle image of the power distribution cabinet are processed through the spatial positioning channel to obtain a spatial positioning tag for the multi-angle image of the power distribution cabinet, and the following processing is performed:

[0012] A set of multi-angle images of the power distribution cabinet and a dataset of image acquisition parameter records are obtained; a spatial coordinate system for the power distribution cabinet is constructed; based on the spatial coordinate system of the power distribution cabinet, the multi-angle images of the power distribution cabinet are positioned and identified to obtain a dataset of multi-angle image positioning identifiers of the power distribution cabinet; the multi-angle images of the power distribution cabinet and the dataset of image acquisition parameter records are input into a convolutional neural network model, and the spatial positioning channel is trained using the dataset of multi-angle image positioning identifiers of the power distribution cabinet as supervision.

[0013] In a possible implementation, control key markings are performed on the multi-angle images of the power distribution cabinet to obtain control key area markings for the multi-angle images of the power distribution cabinet, and the following processing is performed:

[0014] A first control key reference image is obtained, wherein the first control key reference image includes a front image of the first control key and a side image of the first control key; based on the first control key front image, semantic segmentation is performed on the front image and back image of the power distribution cabinet in the multi-angle image of the power distribution cabinet to obtain a first control key front region identifier; based on the first control key side image, semantic segmentation is performed on the first side image and second side image of the power distribution cabinet in the multi-angle image of the power distribution cabinet to obtain a first control key side region identifier; the first control key front region identifier and the first control key side region identifier are stored as a first control key region identifier and added to the control key region identifier of the multi-angle image of the power distribution cabinet.

[0015] In a possible implementation, the control key control mode information and the power distribution cabinet input voltage are input into the power distribution cabinet electrical state prediction channel, and an electrical state prediction value matrix is ​​output, and the following processing is performed:

[0016] A 3D model of the distribution cabinet is obtained and meshed to obtain a 3D mesh model of the distribution cabinet. Based on the 3D mesh model of the distribution cabinet, control key control recording modes, input recording voltage, and 3D mesh electrical index recording status values ​​are collected. Based on the 3D mesh model of the distribution cabinet, a distribution cabinet graph neural network topology is constructed, wherein the input nodes of the distribution cabinet graph neural network topology are the control key distribution area and the distribution cabinet voltage input area, and the output nodes of the distribution cabinet graph neural network topology are the mesh distribution area. The control key control recording modes and the input recording voltage are input to the input nodes, and the 3D mesh electrical index recording status values ​​are used as the supervision values ​​of the output nodes to train the electrical state prediction channel of the distribution cabinet.

[0017] In a possible implementation, the control key controls the recording mode and the input recording voltage are input to the input node, and the three-dimensional mesh electrical index recording state value is used as the supervision value of the output node to train the electrical state prediction channel of the distribution cabinet, and the following processing is performed:

[0018] Obtain first grid output data, wherein the first grid output data has a first-dimensional electrical state prediction value up to the Mth-dimensional electrical state prediction value; obtain first grid supervision data of the first grid output data, wherein the first grid supervision data has a first-dimensional electrical state supervision value up to the Mth-dimensional electrical state supervision value; calculate twice the product of the first-dimensional electrical state prediction value and the first-dimensional electrical state supervision value, and store it as a first loss parameter; calculate the sum of the squares of the first-dimensional electrical state prediction value and the squares of the first-dimensional electrical state supervision value, and store it as a second loss parameter; calculate the ratio of the second loss parameter to the first loss parameter, and set it as a first-dimensional loss coefficient; until the Mth-dimensional loss coefficient is obtained, calculate the mean of the first-dimensional loss coefficient up to the Mth-dimensional loss coefficient, and set it as the first grid training loss value; until the Qth grid training loss value of the same training data is obtained, when the first grid training loss value up to the Qth grid training loss value is less than or equal to the convergence loss threshold, it is considered that the training has converged; when the percentage of training convergence times is greater than or equal to the percentage of convergence times threshold after several consecutive trainings, the electrical state prediction channel of the distribution cabinet is generated.

[0019] This application also provides a real-time monitoring and early warning device for the status of a power distribution cabinet, including:

[0020] The system includes: a multi-angle image receiving module for receiving multi-angle images of the power distribution cabinet captured by a monitoring camera; a control key control mode information acquisition module for analyzing the multi-angle images of the power distribution cabinet through a spatial positioning channel to obtain control key control mode information; a power distribution cabinet input voltage receiving module for receiving the input voltage of the power distribution cabinet; an electrical state prediction value matrix output module for inputting the control key control mode information and the input voltage of the power distribution cabinet into the electrical state prediction channel of the power distribution cabinet and outputting an electrical state prediction value matrix; an electrical state monitoring value matrix acquisition module for obtaining an electrical state monitoring value matrix through an electrical state sensor; and an early warning module for generating an abnormal operation early warning signal for the power distribution cabinet when the electrical state prediction value matrix and the electrical state monitoring value matrix are inconsistent, and sending the signal to the power distribution cabinet control terminal for execution of the early warning.

[0021] The proposed method and device for real-time monitoring and early warning of distribution cabinet status, as described in this application, first receives multi-angle images of the distribution cabinet captured by a monitoring camera. Then, through a spatial positioning channel, the multi-angle images are analyzed to obtain control key mode information. Next, the input voltage of the distribution cabinet is received, and the control key mode information and the input voltage are input into the electrical status prediction channel of the distribution cabinet, outputting an electrical status prediction matrix. Then, through an electrical status sensor, an electrical status monitoring matrix is ​​obtained. When the electrical status prediction matrix and the electrical status monitoring matrix are inconsistent, an abnormal operation early warning signal is generated and sent to the distribution cabinet control terminal for execution. This achieves the technical effect of improving the accuracy and efficiency of the early warning system. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the apparatus according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0023] Figure 1 is a flowchart illustrating a real-time monitoring and early warning method for the status of a power distribution cabinet provided in an embodiment of this application.

[0024] Figure 2 is a schematic diagram of the structure of a real-time monitoring and early warning device for the status of a power distribution cabinet provided in an embodiment of this application.

[0025] Explanation of reference numerals in the attached figures: 10 for multi-angle image receiving module of distribution cabinet, 20 for control key mode information acquisition module, 30 for input voltage receiving module of distribution cabinet, 40 for electrical condition prediction matrix output module, 50 for electrical condition monitoring matrix acquisition module, and 60 for early warning module. Detailed Implementation

[0026] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or apparatuses. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0029] This application provides a method for real-time monitoring and early warning of the status of a power distribution cabinet, as shown in Figure 1. The method includes:

[0030] Step S100: Receive multi-angle images of the power distribution cabinet captured by the monitoring cameras. Specifically, multiple monitoring cameras (camera devices for capturing multi-angle images of the power distribution cabinet) are deployed at key locations within the cabinet. These cameras simultaneously capture images from multiple perspectives, including the front, side, and back of the cabinet, providing stereoscopic visual information. Each camera acquires multi-angle images of the power distribution cabinet according to preset shooting parameters (such as exposure time and white balance) and triggering mechanisms (such as timed triggering and event triggering), for spatial positioning and extraction of control key modal information.

[0031] In one possible implementation, receiving multi-angle images of the power distribution cabinet collected by the power distribution cabinet monitoring camera, step S100 further includes step S110, acquiring a front image of the power distribution cabinet using a first binocular camera; step S120, acquiring a first side image of the power distribution cabinet using a second binocular camera; step S130, acquiring a second side image of the power distribution cabinet using a third binocular camera; step S140, acquiring a rear image of the power distribution cabinet using a fourth binocular camera; and step S150, adding the front image, the first side image, the second side image, and the rear image of the power distribution cabinet to the multi-angle image of the power distribution cabinet.

[0032] Specifically, the binocular camera for the power distribution cabinet is a camera system capable of simultaneously capturing images from two perspectives. Through stereoscopic vision technology, it can provide accurate depth information and spatial positioning. The binocular camera is activated, and images of the front of the power distribution cabinet (including the front panel and its control keys, indicator lights, etc.), the first side image (including the control keys and wiring ports on the side of the cabinet), the second side image (the side opposite the first side), and the rear image (including the rear panel and its wiring, heat dissipation devices, etc.) are acquired according to the set parameters. The images acquired in steps S110-S140 are integrated to form a complete multi-angle image set of the power distribution cabinet. This implementation method, by acquiring images of the front, side, and rear of the power distribution cabinet, allows for comprehensive observation of all parts of the cabinet, including control keys, indicator lights, wiring ports, etc., thereby achieving comprehensive monitoring of the cabinet's status and providing comprehensive visual information support for early warning of the power distribution cabinet.

[0033] Step S200: The multi-angle image of the power distribution cabinet is analyzed through a spatial positioning channel to obtain control key modal information. Specifically, the received multi-angle image of the power distribution cabinet is preprocessed, including noise reduction and contrast enhancement, to improve image quality. The spatial positioning channel is used to match and locate feature points in the image, generating spatial positioning labels for the multi-angle image of the power distribution cabinet. The control keys in the image are identified and labeled, generating control key area identifiers for the multi-angle image of the power distribution cabinet. Based on the spatial positioning labels and control key area identifiers, and combined with a preset control key modal library, the control modal information of the control keys is matched. The spatial positioning channel is a model used to process image spatial information and can generate spatial positioning labels for the image. The control key modal information describes the current state of the control key (e.g., on / off, adjustment position, etc.) and possible behaviors (e.g., about to be pressed, about to be adjusted, etc.).

[0034] In one possible implementation, the multi-angle images of the power distribution cabinet are analyzed through a spatial positioning channel to obtain control key control mode information. Step S200 further includes step S210, where the multi-angle images of the power distribution cabinet have image acquisition parameter identifiers. Specifically, when acquiring multi-angle images of the power distribution cabinet, the camera device automatically adds image acquisition parameter identifiers, which include, but are not limited to, shooting time, camera position, angle, model, focal length, aperture size, shutter speed, etc. Step S220 involves processing the image acquisition parameter identifiers and the multi-angle images of the power distribution cabinet through the spatial positioning channel to obtain a spatial positioning label for the multi-angle images of the power distribution cabinet. Specifically, the received multi-angle images of the power distribution cabinet are first preprocessed, including noise reduction and contrast enhancement, to improve image quality. Feature points in the images are extracted using the model in the spatial positioning channel, including corner points, edges, textures, etc. Image matching and positioning are performed based on the image acquisition parameter identifiers (such as camera position, focal length, etc.) and the extracted feature points. Based on the matching and positioning results, a spatial positioning label is generated for each multi-angle image of the power distribution cabinet. This label describes the position and orientation of the image in three-dimensional space. Step S230: Control key identification is performed on the multi-angle images of the power distribution cabinet to obtain control key area identifiers. Specifically, a template image of the control keys is prepared, and a template matching algorithm is used to find areas similar to the control key template in the multi-angle images of the power distribution cabinet. Based on the template matching results, the multi-angle images of the power distribution cabinet are divided into control key areas and non-control key areas. A unique identifier is generated for each control key area, including information such as the position, size, and shape of the control key. Step S240: Spatial positioning information of the control keys is obtained by combining the spatial positioning labels of the multi-angle images of the power distribution cabinet with the control key area identifiers. Specifically, the spatial positioning labels generated in step S220 are combined with the control key area identifiers generated in step S230, and the control key area identifiers are converted into control key position information in three-dimensional space based on the information in the spatial positioning labels. Step S250: Based on the spatial positioning information of the control keys, and using a preset control key modality library, match the control key control mode information. Specifically, the preset control key modality library is a database containing various control keys and their possible control modes. This library includes the shape, color, position of the control keys, and their corresponding control modes (such as on / off, adjustment, etc.). Based on the spatial positioning information of the control keys obtained in step S240, search for the most matching control key and its control mode in the preset modality library, and output the matched control key control mode information. This implementation method, through the combination of the spatial positioning channel and the preset control key modality library, accurately obtains the position and state information of the control keys, improving the accuracy of electrical state prediction.

[0035] In one possible implementation, the image acquisition parameter identifiers and the multi-angle images of the power distribution cabinet are processed through the spatial positioning channel to obtain spatial positioning labels for the multi-angle images of the power distribution cabinet. Step S220 further includes step S221, obtaining a set of multi-angle images of the power distribution cabinet and an image acquisition parameter recording dataset; step S222, constructing a spatial coordinate system for the power distribution cabinet. Specifically, the set of multi-angle images of the power distribution cabinet is used for spatial positioning, the image acquisition parameter recording dataset is used to provide the correspondence between the images and the actual physical space, and the spatial coordinate system of the power distribution cabinet is a three-dimensional coordinate system used to represent the actual spatial position of the power distribution cabinet and its internal components. Constructing the spatial coordinate system of the power distribution cabinet is used to convert the pixel positions in the images into positions in the actual physical space, thereby achieving precise positioning of the internal components of the power distribution cabinet. Step S223, based on the spatial coordinate system of the power distribution cabinet, the multi-angle images of the power distribution cabinet are labeled with positioning information to obtain a multi-angle image positioning identifier dataset for the power distribution cabinet. Specifically, the pixel positions in the multi-angle images of the power distribution cabinet are correlated with the positions in the spatial coordinate system of the power distribution cabinet to obtain the actual spatial position of each pixel in each image. Specifically, a binocular camera system is used to simultaneously capture images of the power distribution cabinet from two different angles. In the two captured images, a feature matching algorithm is used to extract identical feature points (such as corners and edges). Using the geometric relationships of the binocular cameras (such as baseline distance and focal length) and the positions of the matched feature points, the three-dimensional coordinates of each feature point are calculated. The calculated three-dimensional coordinates of the feature points are then correlated with the spatial coordinate system of the power distribution cabinet, thereby achieving spatial positioning of the power distribution cabinet and its internal components. Based on the spatial positioning results, each pixel in the multi-angle images of the power distribution cabinet is assigned its position information in the spatial coordinate system of the power distribution cabinet. That is, the multi-angle image positioning identifier dataset of the power distribution cabinet refers to the set of multi-angle images of the power distribution cabinet after positioning and identification, where each pixel corresponds to its position in the spatial coordinate system of the power distribution cabinet.

[0036] Step S224: Input the multi-angle image set of the power distribution cabinet and the image acquisition parameter recording dataset into the convolutional neural network model, and train the spatial positioning channel using the multi-angle image positioning identifier dataset of the power distribution cabinet as supervision. Specifically, the multi-angle image set of the power distribution cabinet, the image acquisition parameter recording dataset, and the multi-angle image positioning identifier dataset of the power distribution cabinet are input into the convolutional neural network model for supervised training. This allows the model to learn to predict the spatial location information of each pixel in the image from the input images and parameters. After training, the spatial positioning channel is obtained. This implementation method uses a convolutional neural network model for supervised training, which can automatically learn the features in the image and achieve efficient and accurate spatial positioning of multi-angle images of the power distribution cabinet, improving computational efficiency and enhancing the robustness and accuracy of positioning.

[0037] In one possible implementation, control key identification is performed on the multi-angle images of the distribution cabinet to obtain control key area identification for the multi-angle images of the distribution cabinet. Step S230 further includes step S231, obtaining a first control key reference image, wherein the first control key reference image includes a front image and a side image of the first control key. Specifically, high-resolution images of all control keys on the distribution cabinet are collected, including front and side views of the control keys. For each control key, the front and side images that best represent its appearance features (such as shape, color, label, etc.) are selected as reference images, that is, the first control key reference image refers to the front and side high-resolution images of a specific control key on the distribution cabinet, which are used as a reference for image matching and semantic segmentation. Step S232, based on the first control key front image, semantic segmentation is performed on the front and back images of the distribution cabinet in the multi-angle images of the distribution cabinet to obtain the first control key front area identification. Specifically, a deep learning model is used to perform semantic segmentation on the front and back images of the power distribution cabinet in multi-angle images. During the segmentation process, image similarity matching (including comparison of features such as shape, color, and texture) is used to identify regions similar to the front image of the first control key. When a matching region is found, an anchor box is used to select that region as the front region identifier of the first control key. Step S233: Based on the side image of the first control key, semantic segmentation is performed on the first side image and the second side image of the power distribution cabinet in the multi-angle images to obtain the side region identifier of the first control key. Specifically, similar to step 232, semantic segmentation is performed on the first side and second side images of the power distribution cabinet. Using the side image of the first control key as a reference, regions similar to the side image are identified and marked as the side regions of the first control key. An anchor box is also used to select these regions as the side region identifiers of the first control key. Step S234: The front region identifier and the side region identifier of the first control key are stored as the first control key region identifier and added to the control key region identifier of the multi-angle images of the power distribution cabinet. Specifically, the front and side area identifiers of the first control key obtained in steps 232 and 233 are merged to form a first control key area identifier. This identifier is then added to the database of control key area identifiers for multi-angle images of the distribution cabinet. This database stores the area identifiers of all control keys in the multi-angle images of the distribution cabinet. This implementation considers not only the front image of the control key but also the side image, monitoring the status of the control key from multiple angles, thus improving the comprehensiveness and accuracy of the monitoring.

[0038] Step S300: Receive the input voltage of the distribution cabinet. Specifically, the input voltage of the distribution cabinet, i.e., the power supply voltage received by the distribution cabinet, is acquired through a voltage sensor or measuring instrument.

[0039] Step S400: Input the control key control mode information and the power distribution cabinet input voltage into the power distribution cabinet electrical state prediction channel, and output the electrical state prediction value matrix. Specifically, the control key control mode information and the power distribution cabinet input voltage are integrated into input data, and the power distribution cabinet electrical state prediction channel is used to process the input data to output the electrical state prediction value matrix. The power distribution cabinet electrical state prediction channel is a model used to predict the electrical state of the power distribution cabinet, and the electrical state prediction value matrix is ​​a matrix containing the predicted values ​​of each electrical state of the power distribution cabinet, used for comparison with the monitored values.

[0040] In one possible implementation, the control key control mode information and the power distribution cabinet input voltage are input into the power distribution cabinet electrical state prediction channel, and an electrical state prediction value matrix is ​​output. Step S400 further includes step S410, obtaining a three-dimensional model of the power distribution cabinet and performing meshing processing to obtain a three-dimensional mesh model of the power distribution cabinet. Specifically, the three-dimensional model of the power distribution cabinet is obtained using computer-aided design (CAD) software, and the three-dimensional model is meshed, that is, the continuous three-dimensional space is divided into a series of small, interconnected mesh units (such as triangles) for numerical analysis and calculation. Each mesh unit contains information about its position, shape, and size. Finally, the three-dimensional mesh model of the power distribution cabinet is output. Step S420, based on the three-dimensional mesh model of the power distribution cabinet, the control key control recording mode, the input recording voltage, and the three-dimensional mesh electrical index recording state values ​​are collected. Specifically, the positions of the control key and the voltage input point are determined according to the three-dimensional mesh model of the power distribution cabinet. The control recording mode of the control key is collected, that is, a series of state changes of the control key during operation, including information such as the key being pressed, released, the pressing time, and the releasing time. The system collects the input recorded voltage (voltage value and its change record) of the distribution cabinet, and simultaneously collects the electrical index record status values ​​of each grid cell in the three-dimensional mesh, such as current, voltage, and resistance. Step S430: Based on the three-dimensional mesh model of the distribution cabinet, a graph neural network topology for the distribution cabinet is constructed. The input nodes of the graph neural network topology are the control key distribution area and the distribution cabinet voltage input area, and the output nodes are the grid distribution area. Specifically, the input and output nodes of the graph neural network topology are determined, where the input nodes correspond to the control key distribution area and the distribution cabinet voltage input area, and the output nodes correspond to the grid distribution area. Based on the connection relationships of the three-dimensional mesh model of the distribution cabinet, edges and nodes are constructed in the graph neural network. Edges represent the connection relationships between grid cells, and nodes represent the grid cell itself. Step S440: The control key control recording mode and the input recorded voltage are input to the input node, and the three-dimensional mesh electrical index record status values ​​are used as the supervision values ​​of the output node to train the electrical state prediction channel of the distribution cabinet. Specifically, the collected control key control modes and input voltage are used as input data for the input nodes, and the collected 3D mesh electrical index records are used as the supervisory values ​​(i.e., target values) for the output nodes. A backpropagation algorithm is used to train a graph neural network to predict the electrical state of the distribution cabinet. During training, the parameters of the graph neural network are continuously adjusted to minimize the difference between the predicted and actual values. After training, the electrical state prediction channel for the distribution cabinet is obtained. This implementation uses a graph neural network topology to construct the electrical state prediction channel for the distribution cabinet. The graph neural network can capture the connection relationships and mutual influences between various components in the distribution cabinet, thereby improving the accuracy of electrical state prediction.

[0041] In one possible implementation, the control key control recording mode and the input recording voltage are input to the input node, and the three-dimensional grid electrical index recording state value is used as the supervision value of the output node to train the electrical state prediction channel of the distribution cabinet. Step S440 further includes step S441, obtaining first grid output data, wherein the first grid output data has a first-dimensional electrical state prediction value up to the M-dimensional electrical state prediction value. Specifically, during the forward propagation of the graph neural network, the control key control recording mode and the input recording voltage are input to the input node. After the network calculation, the output data of the first grid is obtained from the output node. This output data is a multi-dimensional vector, where each dimension represents a predicted value of an electrical state, from the first-dimensional electrical state prediction value to the M-dimensional electrical state prediction value. Step S442, obtaining first grid supervision data of the first grid output data, wherein the first grid supervision data has a first-dimensional electrical state supervision value up to the M-dimensional electrical state supervision value. Specifically, for the first grid, the corresponding supervision data is obtained from the state values ​​recorded by the electrical index of the three-dimensional grid. The supervision data of the first grid is the real electrical state data used to compare with the network output data during the training process. It is also a multi-dimensional vector, from the first-dimensional electrical state supervision value to the M-dimensional electrical state supervision value.

[0042] Step S443: Calculate twice the product of the first-dimensional electrical state prediction value and the first-dimensional electrical state monitoring value, and store it as the first loss parameter; Step S444: Calculate the sum of the squares of the first-dimensional electrical state prediction value and the squares of the first-dimensional electrical state monitoring value, and store it as the second loss parameter; Step S445: Calculate the ratio of the second loss parameter to the first loss parameter, and set it as the first-dimensional loss coefficient. Specifically, calculate twice the product of the first-dimensional electrical state prediction value and the monitoring value, and use it as the first loss parameter. Calculate the sum of the squares of the first-dimensional electrical state prediction value and the squares of the monitoring value, and use it as the second loss parameter. Use the ratio of the second loss parameter to the first loss parameter as the first-dimensional loss coefficient. Here, the loss parameter is used to measure the difference between the network's predicted value and the actual value, and the loss coefficient is used to evaluate the network's performance in predicting a certain dimension of electrical state.

[0043] Step S446: Until the Mth dimension loss coefficient is obtained, the average of the first dimension loss coefficient up to the Mth dimension loss coefficient is calculated and set as the first grid training loss value. Specifically, the loss coefficients of each dimension are statistically analyzed, and their average is calculated to obtain the training loss value of the first grid. This loss value reflects the overall prediction performance of the graph neural network on the first grid. Step S447: Until the Qth grid training loss value of the same training data is obtained, when the training loss values ​​of the first grid up to the Qth grid are all less than or equal to the convergence loss threshold, the training is considered to have converged. Specifically, for the same training data, the above steps are repeated to obtain the training loss values ​​of multiple grids. If the training loss values ​​of all grids are less than or equal to the set convergence loss threshold (the threshold used to determine whether the training has converged), the training is considered to have converged, and the graph neural network has stable performance during training, with the loss value no longer decreasing significantly. Step S448: When the percentage of training convergence times is greater than or equal to the percentage of convergence times threshold after several consecutive training iterations, the electrical state prediction channel of the distribution cabinet is generated. Specifically, if the percentage of convergence times during multiple consecutive training sessions is greater than or equal to a set convergence percentage threshold (a threshold used to determine the stability of the graph neural network), the graph neural network is considered stable, and a power distribution cabinet electrical state prediction channel is generated. This implementation utilizes multidimensional loss coefficients for training, which can capture the complex relationships between the electrical states of the power distribution cabinet. Through multiple training sessions and convergence checks, it ensures that the graph neural network can work stably under various conditions, improving prediction accuracy and enhancing the robustness of the graph neural network.

[0044] Step S500: Obtain an electrical condition monitoring value matrix using electrical condition sensors. Specifically, electrical condition sensors are deployed on key electrical components of the distribution cabinet. These sensors collect electrical condition data in real time, such as current, voltage, and temperature. The collected data is then organized into an electrical condition monitoring value matrix, which contains the monitoring values ​​of each electrical condition in the distribution cabinet and is used for comparison with predicted values.

[0045] Step S600: When the electrical condition prediction matrix and the electrical condition monitoring matrix are inconsistent, an abnormal operation warning signal for the distribution cabinet is generated and sent to the distribution cabinet control terminal for execution of the warning. Specifically, the electrical condition prediction matrix and the electrical condition monitoring matrix are compared to determine if there is an inconsistency. When an inconsistency is found, an abnormal operation warning signal for the distribution cabinet is generated to indicate that the distribution cabinet may be operating abnormally. The abnormal operation warning signal is sent to the distribution cabinet control terminal to trigger a warning mechanism, such as an audible and visual alarm or SMS notification. This application embodiment uses multi-angle images captured by a monitoring camera of the distribution cabinet, analyzes the images using a spatial positioning channel to obtain control key control mode information, and combines this information with the input voltage of the distribution cabinet. This information is then input into the electrical state prediction channel of the distribution cabinet. Through the trained prediction channel, an electrical state prediction value matrix is ​​output. Simultaneously, an electrical state monitoring value matrix is ​​obtained through an electrical state sensor. The predicted value matrix is ​​compared with the monitoring value matrix. When the two are inconsistent, an abnormal operation warning signal of the distribution cabinet is generated and sent to the distribution cabinet control terminal for warning execution. These technical means achieve the technical effect of improving the accuracy and efficiency of the warning system.

[0046] In the preceding text, a method for real-time monitoring and early warning of the status of a power distribution cabinet according to an embodiment of the present invention was described in detail with reference to FIG1. ​​Next, a device for real-time monitoring and early warning of the status of a power distribution cabinet according to an embodiment of the present invention will be described with reference to FIG2.

[0047] A real-time monitoring and early warning device for distribution cabinet status according to an embodiment of the present invention addresses the technical problems of low accuracy and efficiency in existing real-time monitoring and early warning systems for distribution cabinets, thereby improving the accuracy and efficiency of the early warning system. The device comprises: a multi-angle image receiving module 10 for the distribution cabinet, a control key control mode information acquisition module 20, an input voltage receiving module 30 for the distribution cabinet, an electrical status prediction value matrix output module 40, an electrical status monitoring value matrix acquisition module 50, and an early warning module 60.

[0048] The distribution cabinet multi-angle image receiving module 10 is used to receive multi-angle images of the distribution cabinet collected by the distribution cabinet monitoring camera; the control key control mode information acquisition module 20 is used to parse the multi-angle images of the distribution cabinet through the spatial positioning channel to obtain control key control mode information; the distribution cabinet input voltage receiving module 30 is used to receive the distribution cabinet input voltage; the electrical state prediction value matrix output module 40 is used to input the control key control mode information and the distribution cabinet input voltage into the distribution cabinet electrical state prediction channel and output the electrical state prediction value matrix; the electrical state monitoring value matrix acquisition module 50 is used to obtain the electrical state monitoring value matrix through the electrical state sensor; the early warning module 60 is used to generate an abnormal operation early warning signal of the distribution cabinet when the electrical state prediction value matrix is ​​inconsistent with the electrical state monitoring value matrix, and send it to the distribution cabinet control terminal to execute the early warning.

[0049] The specific configuration of the power distribution cabinet multi-angle image receiving module 10 will be described in detail below. As mentioned above, the power distribution cabinet multi-angle image receiving module 10 receives multi-angle images of the power distribution cabinet collected by the power distribution cabinet monitoring camera. The power distribution cabinet multi-angle image receiving module 10 may further include: a power distribution cabinet front image acquisition unit for acquiring a front image of the power distribution cabinet through a first power distribution cabinet binocular camera; a power distribution cabinet first side image acquisition unit for acquiring a first side image of the power distribution cabinet through a second power distribution cabinet binocular camera; a power distribution cabinet second side image acquisition unit for acquiring a second side image of the power distribution cabinet through a third power distribution cabinet binocular camera; a power distribution cabinet rear image acquisition unit for acquiring a rear image of the power distribution cabinet through a fourth power distribution cabinet binocular camera; and a power distribution cabinet multi-angle image acquisition unit for adding the power distribution cabinet front image, the power distribution cabinet first side image, the power distribution cabinet second side image, and the power distribution cabinet rear image into the power distribution cabinet multi-angle image.

[0050] The specific configuration of the control key control modality information acquisition module 20 will be described in detail below. As mentioned above, the control key control modality information is obtained by parsing the multi-angle image of the power distribution cabinet through the spatial positioning channel. The control key control modality information acquisition module 20 may further include: a spatial positioning unit for processing the image acquisition parameter identifier of the multi-angle image of the power distribution cabinet through the spatial positioning channel to obtain a spatial positioning label of the multi-angle image of the power distribution cabinet; a control key identification unit for identifying the control keys in the multi-angle image of the power distribution cabinet to obtain a control key area identifier of the multi-angle image of the power distribution cabinet; a control key spatial positioning information acquisition unit for obtaining control key spatial positioning information based on the spatial positioning label of the multi-angle image of the power distribution cabinet and the control key area identifier of the multi-angle image of the power distribution cabinet; and a control key control modality information matching unit for matching the control key control modality information based on the control key spatial positioning information and a preset control key modality library.

[0051] Specifically, the spatial positioning unit processes the image acquisition parameter identifiers and the multi-angle images of the power distribution cabinet through the spatial positioning channel to obtain spatial positioning labels for the multi-angle images of the power distribution cabinet. The spatial positioning unit may further include: an image information acquisition subunit for obtaining a set of multi-angle images of the power distribution cabinet and an image acquisition parameter recording dataset; a power distribution cabinet spatial coordinate system construction subunit for constructing a spatial coordinate system of the power distribution cabinet; a power distribution cabinet multi-angle image positioning identifier dataset acquisition subunit for performing positioning identifiers on the set of multi-angle images of the power distribution cabinet based on the spatial coordinate system of the power distribution cabinet to obtain a multi-angle image positioning identifier dataset of the power distribution cabinet; and a spatial positioning channel training subunit for inputting the set of multi-angle images of the power distribution cabinet and the image acquisition parameter recording dataset into a convolutional neural network model, and training the spatial positioning channel using the multi-angle image positioning identifier dataset of the power distribution cabinet as supervision.

[0052] The control key identification unit further includes: a first control key reference image acquisition subunit for acquiring a first control key reference image, wherein the first control key reference image includes a first control key front image and a first control key side image; a semantic segmentation subunit for performing semantic segmentation on the front and back images of the distribution cabinet in the multi-angle images of the distribution cabinet based on the first control key front image to obtain a first control key front area identifier, and performing semantic segmentation on the first side image of the distribution cabinet in the multi-angle images of the distribution cabinet based on the first control key side image to obtain a first control key side area identifier; and a first control key area identifier generation subunit for storing the first control key front area identifier and the first control key side area identifier as a first control key area identifier and adding it to the control key area identifier of the multi-angle images of the distribution cabinet.

[0053] The specific configuration of the electrical state prediction matrix output module 40 will be described in detail below. As described above, the control key control mode information and the power distribution cabinet input voltage are input into the power distribution cabinet electrical state prediction channel, and an electrical state prediction value matrix is ​​output. The electrical state prediction value matrix output module 40 may further include: a power distribution cabinet three-dimensional mesh model acquisition unit for obtaining a power distribution cabinet three-dimensional model and performing meshing processing to obtain a power distribution cabinet three-dimensional mesh model; a data acquisition unit for acquiring control key control recording modes, input recording voltage, and three-dimensional mesh electrical index recording state values ​​based on the power distribution cabinet three-dimensional mesh model; a power distribution cabinet graph neural network topology construction unit for constructing a power distribution cabinet graph neural network topology based on the power distribution cabinet three-dimensional mesh model, wherein the input nodes of the power distribution cabinet graph neural network topology are the control key distribution area and the power distribution cabinet voltage input area, and the output nodes of the power distribution cabinet graph neural network topology are the mesh distribution area; and a power distribution cabinet electrical state prediction channel training unit for inputting the control key control recording modes and the input recording voltage into the input nodes, using the three-dimensional mesh electrical index recording state values ​​as the supervision values ​​of the output nodes, and training the power distribution cabinet electrical state prediction channel.

[0054] The process involves inputting the control key control recording mode and the input recording voltage into the input node, using the three-dimensional mesh electrical index recording state value as the supervision value of the output node, and training the electrical state prediction channel of the distribution cabinet. The electrical state prediction channel training unit may further include: a first mesh output data acquisition subunit for acquiring first mesh output data, wherein the first mesh output data has a first-dimensional electrical state prediction value up to the Mth-dimensional electrical state prediction value; a first mesh supervision data acquisition subunit for acquiring first mesh supervision data of the first mesh output data, wherein the first mesh supervision data has a first-dimensional electrical state supervision value up to the Mth-dimensional electrical state supervision value; a first loss parameter acquisition subunit for calculating twice the product of the first-dimensional electrical state prediction value and the first-dimensional electrical state supervision value, storing it as a first loss parameter; and a second loss parameter acquisition subunit. The parameter acquisition subunit is used to calculate the sum of the square of the first-dimensional electrical state prediction value and the square of the first-dimensional electrical state supervision value, and store it as the second loss parameter; the first-dimensional loss coefficient acquisition subunit is used to calculate the ratio of the second loss parameter to the first loss parameter, and set it as the first-dimensional loss coefficient; the first grid training loss value acquisition subunit is used to obtain the Mth-dimensional loss coefficient, and calculate the mean of the first-dimensional loss coefficient up to the Mth-dimensional loss coefficient, and set it as the first grid training loss value; the distribution cabinet electrical state prediction channel generation subunit is used to obtain the Qth grid training loss value of the same training data. When the first grid training loss value up to the Qth grid training loss value is less than or equal to the convergence loss threshold, it is considered as training convergence. When the training is carried out for a number of consecutive times, and the proportion of training convergence times is greater than or equal to the proportion of convergence times threshold, the distribution cabinet electrical state prediction channel is generated.

[0055] The real-time monitoring and early warning device for the status of a power distribution cabinet provided in this embodiment of the invention can execute the real-time monitoring and early warning method for the status of a power distribution cabinet provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0056] Although this application makes various references to certain modules in the apparatus according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not intended to limit the scope of protection of this invention.

[0057] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A power distribution cabinet state real-time monitoring and early warning method, characterized in that, include: Receive multi-angle images of the power distribution cabinet captured by the monitoring camera; By analyzing multi-angle images of the power distribution cabinet through a spatial positioning channel, control key control mode information can be obtained. Receive the input voltage from the distribution cabinet; The control key control mode information and the power distribution cabinet input voltage are input into the power distribution cabinet electrical state prediction channel, and the electrical state prediction value matrix is ​​output. An electrical condition monitoring value matrix is ​​obtained through electrical condition sensors; When the electrical condition prediction matrix is ​​inconsistent with the electrical condition monitoring matrix, an abnormal operation warning signal for the distribution cabinet is generated and sent to the distribution cabinet control terminal to execute the warning.

2. The real-time monitoring and early warning method for power distribution cabinet states according to claim 1, characterized in that, Receive multi-angle images of the power distribution cabinet captured by the monitoring camera, including: The front image of the power distribution cabinet is captured using a binocular camera on the first power distribution cabinet. The first side image of the power distribution cabinet was captured using a binocular camera in the second power distribution cabinet. The second side image of the power distribution cabinet is captured using a binocular camera in the third power distribution cabinet. The image of the back of the power distribution cabinet was captured using a binocular camera in the fourth power distribution cabinet. Add the front image of the power distribution cabinet, the first side image of the power distribution cabinet, the second side image of the power distribution cabinet, and the rear image of the power distribution cabinet to the multi-angle image of the power distribution cabinet.

3. The real-time monitoring and early warning method for power distribution cabinet states according to claim 1, characterized in that, By analyzing multi-angle images of the power distribution cabinet through a spatial positioning channel, control key control mode information is obtained, including: The multi-angle images of the power distribution cabinet have image acquisition parameter identifiers; The image acquisition parameter identifier and the multi-angle image of the power distribution cabinet are processed through the spatial positioning channel to obtain the spatial positioning label of the multi-angle image of the power distribution cabinet. Control key markings are performed on the multi-angle images of the power distribution cabinet to obtain control key area markings for the multi-angle images of the power distribution cabinet. Based on the spatial positioning label of the multi-angle image of the power distribution cabinet and the control key area identifier of the multi-angle image of the power distribution cabinet, spatial positioning information of the control key is obtained. Based on the spatial positioning information of the control keys, and using the preset modality library of control keys, the control modality information of the control keys is matched.

4. The power distribution cabinet state real-time monitoring and early warning method of claim 3, wherein, Through the spatial positioning channel, the image acquisition parameter identifier and the multi-angle image of the power distribution cabinet are processed to obtain a multi-angle image spatial positioning tag for the power distribution cabinet, including: Obtain a set of multi-angle images of the power distribution cabinet and a dataset of image acquisition parameter records; Construct a spatial coordinate system for the power distribution cabinet; Based on the spatial coordinate system of the power distribution cabinet, the multi-angle image set of the power distribution cabinet is positioned and labeled to obtain a multi-angle image positioning and label dataset of the power distribution cabinet. The set of multi-angle images of the power distribution cabinet and the dataset of image acquisition parameters are input into a convolutional neural network model, and the spatial positioning channel is trained using the dataset of multi-angle image positioning identifiers of the power distribution cabinet as supervision.

5. The power distribution cabinet state real-time monitoring and early warning method of claim 3, wherein, The control key area of ​​the multi-angle image of the power distribution cabinet is marked by the control keys, including: Obtain a first control key reference image, wherein the first control key reference image includes a front image of the first control key and a side image of the first control key; Based on the front image of the first control key, semantic segmentation is performed on the front and back images of the power distribution cabinet from multiple angles to obtain the front area identifier of the first control key. Based on the first control key side image, semantic segmentation is performed on the first side image and the second side image of the power distribution cabinet from multiple angles to obtain the first control key side area identifier. Store the front area identifier and the side area identifier of the first control key as the first control key area identifier, and add them to the multi-angle image control key area identifier of the power distribution cabinet.

6. The real-time monitoring and early warning method for power distribution cabinet states according to claim 1, characterized in that, The control key control mode information and the power distribution cabinet input voltage are input into the power distribution cabinet electrical state prediction channel, and the electrical state prediction value matrix is ​​output, including: The 3D model of the power distribution cabinet is obtained and then meshed to obtain a 3D mesh model of the power distribution cabinet. Based on the three-dimensional mesh model of the power distribution cabinet, the control key is used to record the mode, input and record the voltage, and record the status values ​​of the three-dimensional mesh electrical index. Based on the three-dimensional mesh model of the distribution cabinet, a distribution cabinet graph neural network topology is constructed, wherein the input nodes of the distribution cabinet graph neural network topology are the control key distribution area and the distribution cabinet voltage input area, and the output nodes of the distribution cabinet graph neural network topology are the mesh distribution area; The control key controls the recording mode and the input recording voltage are input to the input node, and the three-dimensional mesh electrical index recording state value is used as the supervision value of the output node to train the electrical state prediction channel of the distribution cabinet.

7. The power distribution cabinet state real-time monitoring and early warning method of claim 6, wherein, The control key controls the recording mode and the input recording voltage are input to the input node, and the three-dimensional mesh electrical index recording state value is used as the supervision value of the output node to train the electrical state prediction channel of the distribution cabinet, including: Obtain first grid output data, wherein the first grid output data has a first-dimensional electrical state prediction value up to the M-dimensional electrical state prediction value; First grid supervision data is obtained from the first grid output data, wherein the first grid supervision data has a first-dimensional electrical state supervision value up to the M-th-dimensional electrical state supervision value; Calculate twice the product of the first-dimensional electrical state prediction value and the first-dimensional electrical state monitoring value, and store it as the first loss parameter; Calculate the sum of the squares of the predicted values ​​of the first-dimensional electrical state and the squares of the supervised values ​​of the first-dimensional electrical state, and store it as the second loss parameter; Calculate the ratio of the second loss parameter to the first loss parameter, and set it as the first dimension loss coefficient; Until the Mth dimension loss coefficient is obtained, the average of the first dimension loss coefficient up to the Mth dimension loss coefficient is calculated and set as the first grid training loss value; Training convergence is considered to occur when the training loss value of the first grid up to the Q-th grid is less than or equal to the convergence loss threshold, until the Q-th grid training loss value of the same training data is obtained. When the training is performed several times in a row and the percentage of training convergence times is greater than or equal to the threshold of the percentage of convergence times, the electrical state prediction channel of the distribution cabinet is generated.

8. A power distribution cabinet state real-time monitoring and early warning device, characterized in that, The device is used to implement the real-time monitoring and early warning method for the status of a power distribution cabinet as described in any one of claims 1-7, and the device comprises: A power distribution cabinet multi-angle image receiving module, which is used to receive multi-angle images of the power distribution cabinet collected by the power distribution cabinet monitoring camera; The control key control mode information acquisition module is used to analyze multi-angle images of the power distribution cabinet through a spatial positioning channel to obtain control key control mode information. A power distribution cabinet input voltage receiving module, wherein the power distribution cabinet input voltage receiving module is used to receive the input voltage of the power distribution cabinet; An electrical condition prediction matrix output module is used to input the control key control mode information and the power distribution cabinet input voltage into the power distribution cabinet electrical condition prediction channel and output an electrical condition prediction matrix. An electrical condition monitoring value matrix acquisition module is used to obtain an electrical condition monitoring value matrix through an electrical condition sensor. The early warning module is used to generate an early warning signal for abnormal operation of the distribution cabinet when the electrical state prediction value matrix is ​​inconsistent with the electrical state monitoring value matrix, and send it to the distribution cabinet control terminal to execute the early warning.

Images