Power equipment image monitoring method and device

Abstract

The application provides a power equipment image monitoring method and device, the method comprises the following steps: using multiple monitors to collect monitoring images of different power equipment in real time; inputting each monitoring image and the corresponding safety image into a preset power equipment monitoring model based on a twin network, so that the power equipment monitoring model outputs a feature similarity value corresponding to each monitoring image; if there is a feature similarity value less than a preset safety threshold of the corresponding power equipment in each feature similarity value, a fault report corresponding to the feature similarity value is generated and transmitted to the control layer. The application can effectively improve the communication efficiency of the monitor and the control layer in the power system and reduce the communication energy consumption.

Description

Technical Field

[0001] This application relates to the field of power system monitoring, and in particular to a method and apparatus for image monitoring of power equipment. Background Technology

[0002] In many sparsely populated and inaccessible areas, while power distribution operation has been somewhat resolved, distribution maintenance remains a major pain point in the power industry. This is because the energy consumption of cameras used to monitor these power devices in harsh geographical environments is difficult to address. Most monitoring equipment relies on solar panels for power, meaning that limited energy must be used to monitor power equipment in real time to ensure distribution safety. Traditional distribution operation and maintenance management measures are no longer sufficient to meet the requirements of power grid development. Integrated distribution operation and maintenance is an effective way to solve the distribution maintenance problems of my country's power grid and is highly valued by the power sector. Therefore, transmitting as much effective information as possible under limited energy conditions is the development direction of distribution operation and maintenance.

[0003] Existing technologies, on the one hand, do not utilize knowledge from the field of artificial intelligence; they simply rely on monitors to track various parameter values ​​and return them to the control layer for analysis, or rely on manual periodic checks and real-time monitoring, which is inefficient. On the other hand, the monitors need to maintain constant communication with the control layer, resulting in high energy consumption. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method and apparatus for image monitoring of power equipment to eliminate or improve one or more defects existing in the prior art.

[0005] The first aspect of this application provides a method for image monitoring of power equipment, the method comprising:

[0006] Multiple monitors are used to collect real-time monitoring images of different power equipment;

[0007] Each of the monitoring images and their corresponding security images are input into a preset power equipment monitoring model based on twin networks, so that the power equipment monitoring model outputs feature similarity values ​​corresponding to the monitoring images of each of the power equipment.

[0008] If any of the aforementioned feature similarity values ​​is less than the preset safety threshold of the corresponding power equipment, a fault report corresponding to that feature similarity value is generated and transmitted to the control layer.

[0009] In some embodiments of this application, before the step of using multiple monitors to acquire real-time monitoring images of different power devices, the method further includes:

[0010] The historical monitoring datasets of each of the aforementioned power devices are preprocessed to standardize their format.

[0011] The twin network is used to extract features from each historical monitoring image and the corresponding historical security image in the historical monitoring dataset to obtain the historical monitoring features and the historical security features corresponding to each historical monitoring image and the historical security image, respectively.

[0012] Each of the historical monitoring features and its corresponding historical security features is compared and calculated to obtain the historical feature similarity value;

[0013] The error corresponding to each historical feature similarity value is calculated based on the similarity value of each historical feature and its corresponding expected value.

[0014] Calculate the gradient of multiple weights in the Siamese network based on each of the aforementioned errors;

[0015] The optimization algorithm based on the gradient updates each weight to obtain the corresponding optimized weight, and the hyperparameters in the Siamese network are updated according to the validation set in the historical monitoring dataset to obtain the validation parameters.

[0016] The preset image comparison model is trained using the historical feature similarity values, the optimization weights, and the verification parameters to obtain the corresponding general model for power equipment monitoring.

[0017] The power equipment monitoring model includes: a general power equipment monitoring model.

[0018] In some embodiments of this application, before performing preprocessing to standardize the format of the historical monitoring datasets of each of the power devices, the method further includes:

[0019] A knowledge graph is constructed by acquiring normal historical image data and fault historical image data of each of the aforementioned power devices through different channels.

[0020] Extract the historical monitoring dataset from the knowledge graph.

[0021] In some embodiments of this application, after training a preset image comparison model using the various historical feature similarity values, the various optimization weights, and the various verification parameters to obtain a corresponding general power equipment monitoring model, the method further includes:

[0022] The weights corresponding to the general power equipment monitoring model are adjusted according to each of the power equipment to obtain the power equipment monitoring sub-models corresponding to each of the power equipment.

[0023] Correspondingly, the power equipment monitoring model also includes: power equipment monitoring sub-models corresponding to each of the power equipment.

[0024] In some embodiments of this application, the step of comparing and calculating the historical monitoring features and their corresponding historical security features to obtain historical feature similarity values ​​includes:

[0025] The historical similarity features corresponding to each historical monitoring feature are obtained by subtracting each of the historical monitoring features from its corresponding historical security features.

[0026] Each of the historical similar features is obtained by using the fully connected layers in the Siamese network to obtain the similarity value of each historical feature.

[0027] In some embodiments of this application, before inputting each of the monitoring images and their corresponding security images into a preset twin network-based power equipment monitoring model, the method further includes:

[0028] Each of the aforementioned monitoring images undergoes preprocessing to standardize its format.

[0029] In some embodiments of this application, each of the monitors and the power equipment monitoring model is set on a local node.

[0030] A second aspect of this application provides an image monitoring device for power equipment, the device comprising:

[0031] The data acquisition module is used to acquire real-time monitoring images of different power equipment using multiple monitors;

[0032] The feature similarity value calculation module is used to input each of the monitoring images and their corresponding security images into a preset power equipment monitoring model based on twin networks, so that the power equipment monitoring model outputs feature similarity values ​​corresponding to the monitoring images of each of the power equipment.

[0033] The fault feedback module is used to generate a fault report corresponding to the feature similarity value if there is a feature similarity value that is less than the preset safety threshold of the corresponding power equipment, and transmit it to the control layer.

[0034] A third aspect of this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the power equipment image monitoring method described in the first aspect.

[0035] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the power equipment image monitoring method described in the first aspect above.

[0036] This application provides a method and apparatus for image monitoring of power equipment. The method includes: using multiple monitors to acquire monitoring images of different power equipment in real time; inputting each monitoring image and its corresponding safety image into a preset power equipment monitoring model based on a twin network, so that the power equipment monitoring model outputs feature similarity values ​​corresponding to the monitoring images of each power equipment; if there is a feature similarity value among the feature similarity values ​​that is less than a preset safety threshold of the corresponding power equipment, then generating a fault report corresponding to the feature similarity value and transmitting it to the control layer. This application can effectively improve the communication efficiency between the monitor and the control layer in the power system and reduce communication energy consumption. Additional advantages, objectives, and features of this application will be set forth in part in the description which follows, and will in part become apparent to those skilled in the art upon review of the description, or may be learned by practice of this application. The objectives and other advantages of this application can be realized and obtained by means of the structures specifically pointed out in the specification and drawings.

[0037] Those skilled in the art will understand that the purposes and advantages that can be achieved with this application are not limited to those specifically described above, and that the above and other purposes that this application can achieve will be more clearly understood from the following detailed description. Attached Figure Description

[0038] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, do not constitute a limitation thereof. The components in the drawings are not drawn to scale but are merely for illustrating the principles of this application. For ease of illustration and description of certain parts of this application, corresponding portions in the drawings may be enlarged, i.e., may appear larger relative to other components in an exemplary device actually manufactured according to this application. In the drawings:

[0039] Figure 1 This is a flowchart illustrating a power equipment image monitoring method according to one embodiment of this application.

[0040] Figure 2 This is a schematic diagram of the structure of a power equipment image monitoring device according to another embodiment of this application.

[0041] Figure 3(a) is a schematic diagram of a knowledge graph of insulators for power equipment in another embodiment of this application.

[0042] Figure 3(b) is a schematic diagram of the insulator normal operation threshold calculation in another embodiment of this application.

[0043] Figure 3(c) is a schematic diagram of inputting a monitoring image into a local node containing a twin network for processing in another embodiment of this application.

[0044] Figure 3(d) is a schematic diagram of a comparison calculation between the normal screen of the twin network insulator and the monitoring screen of the insulator in another embodiment of this application.

[0045] Figure 3(e) is a schematic diagram of determining whether the similarity of the insulator image is higher than the normal operating threshold of the insulator in another embodiment of this application. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and their descriptions are used to explain this application, but are not intended to limit it.

[0047] It should also be noted that, in order to avoid obscuring this application with unnecessary details, only the structures and / or processing steps closely related to the solution according to this application are shown in the accompanying drawings, while other details that are not closely related to this application are omitted.

[0048] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0049] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.

[0050] In the following description, embodiments of the present application will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0051] The following examples will provide a detailed description.

[0052] This application provides a power equipment image monitoring method that can be executed by a power equipment image monitoring device. See [link to relevant documentation]. Figure 1 The power equipment image monitoring method specifically includes the following:

[0053] Step 110: Use multiple monitors to collect real-time monitoring images of different power equipment.

[0054] Step 120: Input each of the monitoring images and their corresponding security images into a preset twin network-based power equipment monitoring model, so that the power equipment monitoring model outputs feature similarity values ​​corresponding to the monitoring images of each of the power equipment.

[0055] Step 130: If there is a feature similarity value among the various feature similarity values ​​that is less than the preset safety threshold of the corresponding power equipment, then generate a fault report corresponding to the feature similarity value and transmit it to the control layer.

[0056] Specifically, a local node is used to input the monitoring images of different power devices collected in real time by multiple monitors and their corresponding safety images into a preset power device monitoring model based on twin networks. This allows the power device monitoring model to output feature similarity values ​​corresponding to the monitoring images of each power device. If any feature similarity value is less than the preset safety threshold of the corresponding power device, a fault report corresponding to that feature similarity value is generated and transmitted to the control layer through the local node. This effectively improves the communication efficiency between the monitor and the control layer in the power system and reduces communication energy consumption.

[0057] The various power equipment includes: insulators, generators, motors, transformers; circuit breakers, contactors, fuses; busbars, towers, power cables, surge arresters, and instrument transformers. The control layer represents the upper-level server that communicates with the monitor. Fault reports include: specific numerical values, the capture time of the monitored image, and the device number corresponding to the monitored image. It should be noted that the power equipment image monitoring method is applicable not only to wired communication power systems but also to wireless communication power systems. Preset safety thresholds are determined manually or obtained during the training of the power equipment monitoring model.

[0058] In one embodiment of this application, referring to Figure 3(a), the image data of insulators in the monitored area are collected in real time by a monitor according to the power equipment image monitoring method, forming an insulator knowledge graph for the monitored area. Referring to Figure 3(b), the threshold for normal operation of the insulator is calculated based on the main indicators in the insulator knowledge graph, such as creepage distance and clearance, which affect the function of the insulator. For example, the creepage ratio is calculated. Referring to Figure 3(c), the insulator images collected in real time by the monitor are input into a local node with a power equipment monitoring model based on a twin network for processing. Referring to Figure 3(d), the power equipment monitoring model based on a twin network compares the normal images and the insulator images in the insulator knowledge graph to obtain the image similarity. In Figure 3(e), it is determined whether the image similarity is greater than the threshold for normal operation. If it is greater, the insulator is operating normally; if it is less than or equal to, the local node generates an insulator fault report and sends it to the upper layer, i.e., the control layer.

[0059] To further improve the communication efficiency between the monitor and the control layer in the power system, before step 110, the following is also included:

[0060] Step 210: Perform preprocessing to standardize the format of the historical monitoring datasets of each of the power devices;

[0061] Step 220: Use the twin network to extract features from each historical monitoring image and the corresponding historical security image in the historical monitoring dataset to obtain the historical monitoring features and the historical security features corresponding to each historical monitoring image and the corresponding historical security image, respectively.

[0062] Step 230: Compare and calculate the historical monitoring features and their corresponding historical security features to obtain the historical feature similarity value;

[0063] Step 240: Calculate the error corresponding to each of the historical feature similarity values ​​and their corresponding expected values.

[0064] Here, the expected value is the label corresponding to the historical monitoring image. The label includes 0 and 1, where 0 represents normal and 1 represents fault. For example, if a normal image is input into the Siamese network during training, the expected value is 0×1=0, and conversely, the expected value is 1×1=1. However, the Siamese network itself does not know whether the image is normal or faulty. When a normal historical monitoring image is input, the Siamese network may determine that the probability of it being normal is 0.8 and the probability of it being faulty is 0.2. In this case, the historical feature similarity value output by the Siamese network would be 0.8×0+0.2×1=0.2, which has an error of 0.2 from the correct expected value of 0.

[0065] Step 250: Calculate the gradient of multiple weights in the Siamese network based on each of the errors;

[0066] The method and steps for calculating the weight gradient are as follows:

[0067] 1. A Siamese network consists of two identical convolutional neural networks. The structures in a convolutional neural network that need to update parameters are called gates. For each gate, there is a set of weights and biases.

[0068] 2. According to the forward propagation process of a convolutional neural network, after passing through each gate, a cost function D is finally obtained. 3. Perform backward differentiation, that is, differentiate the cost function D with respect to each gate in turn. For example, if there are three gates A, B, and C, then according to the chain rule, we should calculate the derivative of D with respect to C, the derivative of C with respect to B, and the derivative of B with respect to A, so as to obtain the gradient of each gate.

[0069] Step 260: Update each weight based on the gradient optimization algorithm to obtain the corresponding optimized weight, and update the hyperparameters in the Siamese network according to the validation set in the historical monitoring dataset to obtain the validation parameters;

[0070] The optimization algorithm involves multiplying the weight gradient obtained in step 250 by a preset parameter to obtain new weights.

[0071] Meanwhile, it's understandable that hyperparameter settings are not fixed. Common hyperparameters include learning rate and number of training epochs. Hyperparameters are generally set manually, not trained, because these parameters are difficult to train. The validation set serves to test and determine a good set of hyperparameters, improving the training performance of the power system fault prediction model. The method for updating hyperparameters based on the validation set is as follows: for example, to update the number of training epochs, different numbers of epochs are trained on the training set to obtain different models. These models are then tested on the validation set, and the epoch corresponding to the best-performing model should be selected.

[0072] Step 270: Train the preset image comparison model using the historical feature similarity values, the optimized weights, and the verification parameters to obtain the corresponding general model for power equipment monitoring;

[0073] The power equipment monitoring model includes: a general power equipment monitoring model.

[0074] Specifically, a Siamese network is used to extract features from each historical monitoring image and its corresponding historical safety image in the historical monitoring dataset to obtain historical monitoring features and historical safety features. The historical monitoring features and their corresponding historical safety features are then compared to obtain historical feature similarity values. Errors are calculated based on these similarity values ​​and their corresponding expected values. Gradients of multiple weights in the Siamese network are calculated based on these errors. A gradient-based optimization algorithm updates the weights to obtain optimized weights. Hyperparameters in the Siamese network are updated using the validation set in the historical monitoring dataset to obtain validation parameters. A preset image comparison model is trained using the historical feature similarity values, optimized weights, and validation parameters to obtain a corresponding general model for power equipment monitoring, thereby further improving the communication efficiency between the monitor and control layer in the power system.

[0075] The historical dataset includes multiple historical monitoring images of various power equipment and corresponding historical safety images for each power equipment; the validation set is a partial historical monitoring dataset; the power equipment monitoring model includes a general power equipment monitoring model. The specific preprocessing method is as follows: historical monitoring images are either videos or images. If it is a video, the video is first decomposed into individual frames for processing. The preprocessing results of the images must meet the input requirements of the twin network. For example, the input requirement is that both input images are 256×256×3 RGB images, while the original images collected by monitoring are 512×512×3 RGB images. Therefore, the images need to be cropped or scaled. The input requirements for the twin network are manually set, and the image processing rules are also manually set.

[0076] To obtain a large amount of training data to improve the model's accuracy, the following steps are included before step 210:

[0077] A knowledge graph is constructed by acquiring normal historical image data and fault historical image data of each of the aforementioned power devices through different channels.

[0078] Extract the historical monitoring dataset from the knowledge graph.

[0079] Specifically, normal historical image data and fault historical image data of various power equipment are obtained through different channels to construct a knowledge graph; the historical monitoring dataset is extracted from the knowledge graph, thereby obtaining a large amount of training data to improve the accuracy of the model.

[0080] These materials were obtained from various sources, including the internet, materials provided by power equipment manufacturers, and self-filmed footage.

[0081] Furthermore, for example, the preset safety threshold of a certain power equipment in a power system is 0.9. The output of the power equipment image monitoring model for the monitoring image of the power equipment during normal operation is generally 0.95. However, there is no data for the power equipment in snowy weather in the knowledge graph. In snowy weather, the output of the power equipment monitoring model becomes 0.93, which is higher than 0.9. At this time, it can be determined that the equipment is still operating normally. The historical monitoring images of the power equipment in snowy weather are added to the knowledge graph. When snowy weather occurs again, the output result of the power equipment monitoring model for the historical feature similarity value of the monitoring images of the power equipment in snowy weather will be improved.

[0082] To effectively improve the accuracy of model application, after step 270, the following steps are also included:

[0083] The weights of each power device are adjusted according to each power device to obtain the power device monitoring sub-model for each power device.

[0084] Correspondingly, the power equipment monitoring model also includes: power equipment monitoring sub-models corresponding to each of the power equipment.

[0085] The specific adjustment process is as follows: Based on the general power equipment monitoring model, specialized training is performed for a specific type of power equipment. For example, training the general power equipment monitoring model using only historical monitoring image datasets of surge arresters will give the general model greater accuracy in judging surge arrester images. This results in a unique power equipment monitoring sub-model corresponding to each power device, effectively improving the accuracy of model application.

[0086] Therefore, the power equipment monitoring model includes: a general power equipment monitoring model and power equipment monitoring sub-models corresponding to each of the aforementioned power equipment.

[0087] To further solve for historical similarity features, step 230 includes:

[0088] The historical similarity features corresponding to each historical monitoring feature are obtained by subtracting each of the historical monitoring features from its corresponding historical security features.

[0089]

[0090] The subtraction calculation process is the loss function calculation process. Commonly used loss functions in Siamese networks include the Contrastive Loss Function (see formula (1)). Here, W is the network weight; Y is the pair label. If samples X1 and X2 belong to the same class, Y = 0; otherwise, Y = 1. Dw is the Euclidean distance between X1 and X2 in the latent variable space. When Y = 0, the parameters are adjusted to minimize the distance between X1 and X2. When Y = 1, if the distance between X1 and X2 is greater than m, no optimization is performed (saving time and effort); if the distance between X1 and X2 is less than m, the distance is increased to m. In short, the goal is to find a suitable method to measure or judge the similarity between the historical monitoring features and historical security features. This patent does not provide a specific loss function.

[0091] Each of the historical similar features is obtained by using the fully connected layers in the Siamese network to obtain the similarity value of each historical feature.

[0092] Specifically, fully connected layers are generally used to transform the two-dimensional feature map output by convolution into a one-dimensional vector. Each node in a fully connected layer is connected to each node in the previous layer, thus combining the output features of the previous layer. The role of the fully connected layer network is to stretch the feature map obtained from the last convolutional layer into a vector, multiply this vector, and finally reduce its dimensionality.

[0093] For example, after several convolutional kernel pooling operations, it will output 10 12×12 matrices. This means that 10 neurons were used, and each neuron performed convolutions on the image. The 12×12 matrix output by each neuron represents the neuron's understanding of a feature of the image. Next, we reach the fully connected layer. Suppose the fully connected layer outputs a 1×100 matrix, which is actually the result of 100 convolutional kernels of size 20×12×12. For each input image, a kernel of the same size as the image is used for convolution. In this way, the entire image becomes a number. Since the thickness is 20, after convolution with 20 kernels, the results are summed to get a number. There are a total of 100 convolutional kernels, so we get 100 numbers. Each number represents a highly refined feature. These one hundred features can be further reduced in dimensionality using a fully connected layer. For example, if the final output is a binary classification, i.e., normal or faulty, then another fully connected layer with an output of 1×2 can be added to calculate the final output result from the input of the upper layer of 1×100.

[0094] To ensure data consistency, the following steps are included before step 120:

[0095] Each of the aforementioned monitoring images undergoes preprocessing to standardize its format.

[0096] The specific preprocessing method is as follows: The data collected by the monitor in real time is video or images. For example, in a power transmission scenario, the video monitoring content includes line towers, conductors, and insulators, and the collected data is the video monitoring footage of the aforementioned power transmission equipment. If the data collected by the monitor in real time is video, the video can first be decomposed into each frame of images for processing. The preprocessing results of the images must meet the input requirements of the twin network. For example, the input requirement is that both input images are 256×256×3 RGB images, while the original image collected by the monitor is a 512×512×3 RGB image. Therefore, the image needs to be cropped or scaled. The input requirements for the twin network are set manually, and the image processing rules are also set manually.

[0097] To further improve the communication efficiency between the monitor and the control layer, each monitor and power equipment monitoring model in the aforementioned steps is set on a local node, which can improve the communication efficiency between the monitor and the control layer.

[0098] From a software perspective, this application also provides a power equipment image monitoring device for performing all or part of the aforementioned power equipment image monitoring method, see [link to relevant documentation]. Figure 2 The aforementioned power equipment image monitoring device specifically includes the following components:

[0099] Data acquisition module 10 is used to acquire monitoring images of different power equipment in real time using multiple monitors;

[0100] Feature similarity calculation module 20 is used to input each of the monitoring images and their corresponding security images into a preset power equipment monitoring model based on twin networks, so that the power equipment monitoring model outputs feature similarity values ​​corresponding to the monitoring images of each of the power equipment.

[0101] The fault feedback module 30 is used to generate a fault report corresponding to the feature similarity value if there is a feature similarity value among the various feature similarity values ​​that is less than the preset safety threshold of the corresponding power equipment, and transmit it to the control layer.

[0102] The embodiments of the power equipment image monitoring device provided in this application can be used to execute the processing flow of the power equipment image monitoring method embodiments described above. Its functions will not be repeated here, but can be referred to the detailed description of the power equipment image monitoring method embodiments described above.

[0103] This application provides a power equipment device that uses multiple monitors to acquire real-time monitoring images of different power equipment. Each monitoring image and its corresponding safety image are input into a preset power equipment monitoring model based on a twin network. The monitoring model then outputs feature similarity values ​​corresponding to the monitoring images of each power equipment. If any of these feature similarity values ​​is less than a preset safety threshold for the corresponding power equipment, a fault report corresponding to that feature similarity value is generated and transmitted to the control layer. This application effectively improves the communication efficiency between the monitors and the control layer in a power system and reduces communication energy consumption.

[0104] This application also provides an electronic device, such as a central server, which may include a processor, a memory, a receiver, and a transmitter. The processor is used to execute the power equipment image monitoring method mentioned in the above embodiments. The processor and memory can be connected via a bus or other means, taking a bus connection as an example. The receiver can be connected to the processor and memory via wired or wireless means.

[0105] The processor can be a central processing unit (CPU). The processor can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0106] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the power equipment image monitoring method in the embodiments of this application. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory, thereby implementing the power equipment image monitoring method in the above method embodiments.

[0107] The memory may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0108] The one or more modules are stored in the memory, and when executed by the processor, they perform the power equipment image monitoring method in the embodiment.

[0109] In some embodiments of this application, the user equipment may include a processor, a memory, and a transceiver unit. The transceiver unit may include a receiver and a transmitter. The processor, memory, receiver, and transmitter may be connected via a bus system. The memory is used to store computer instructions, and the processor is used to execute the computer instructions stored in the memory to control the transceiver unit to send and receive signals.

[0110] As one implementation method, the functions of the receiver and transmitter in this application can be implemented by transceiver circuits or dedicated transceiver chips, and the processor can be implemented by dedicated processing chips, processing circuits or general-purpose chips.

[0111] As another implementation approach, the server provided in this application embodiment can be implemented using a general-purpose computer. That is, the program code implementing the processor, receiver, and transmitter functions is stored in memory, and the general-purpose processor implements the processor, receiver, and transmitter functions by executing the code in memory.

[0112] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned power equipment image monitoring method. The computer-readable storage medium can be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disks, removable storage disks, CD-ROMs, or any other form of storage medium known in the art.

[0113] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave.

[0114] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0115] In this application, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0116] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to the embodiments of this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for image monitoring of power equipment, characterized in that, include: A knowledge graph is constructed by acquiring normal historical image data and fault historical image data of each of the aforementioned power devices through different channels. Extract historical surveillance datasets from the knowledge graph; The historical monitoring datasets of each of the aforementioned power devices are preprocessed to standardize their format. A twin network is used to extract features from each historical monitoring image and the corresponding historical security image in the historical monitoring dataset to obtain the historical monitoring features and the historical security features corresponding to each historical monitoring image and the historical security image, respectively. Each of the historical monitoring features and its corresponding historical security features is compared and calculated to obtain the historical feature similarity value; The error corresponding to each historical feature similarity value is calculated based on the similarity value of each historical feature and its corresponding expected value; the gradient of multiple weights in the Siamese network is calculated based on the error; the weights are updated based on the gradient optimization algorithm to obtain the corresponding optimized weights; and the hyperparameters in the Siamese network are updated based on the validation set in the historical monitoring dataset to obtain the validation parameters. The preset image comparison model is trained using the historical feature similarity values, the optimization weights, and the verification parameters to obtain the corresponding general power equipment monitoring model; wherein, each monitor and the power equipment monitoring model are set on a local node; Multiple monitors are used to collect real-time monitoring images of different power equipment; Each of the monitoring images and its corresponding security image is input into a preset power equipment monitoring model based on the twin network, so that the power equipment monitoring model outputs feature similarity values ​​corresponding to the monitoring images of each power equipment; wherein, the power equipment monitoring model includes: a general power equipment monitoring model; If any of the aforementioned feature similarity values ​​is less than the preset safety threshold of the corresponding power equipment, a fault report corresponding to that feature similarity value is generated and transmitted to the control layer.

2. The power equipment image monitoring method according to claim 1, characterized in that, After training the preset image comparison model using the historical feature similarity values, optimization weights, and verification parameters to obtain the corresponding general power equipment monitoring model, the method further includes: The weights corresponding to the general power equipment monitoring model are adjusted according to each of the power equipment to obtain the power equipment monitoring sub-models corresponding to each of the power equipment. Correspondingly, the power equipment monitoring model also includes: power equipment monitoring sub-models corresponding to each of the power equipment.

3. The power equipment image monitoring method according to claim 1, characterized in that, The step of comparing and calculating the historical feature similarity value by comparing each of the historical monitoring features and their corresponding historical security features includes: The historical similarity features corresponding to each historical monitoring feature are obtained by subtracting each of the historical monitoring features from its corresponding historical security features. Each of the historical similar features is obtained by using the fully connected layers in the Siamese network to obtain the similarity value of each historical feature.

4. The power equipment image monitoring method according to claim 1, characterized in that, Before inputting each of the monitoring images and their corresponding security images into a preset power equipment monitoring model based on the twin network, the method further includes: Each of the aforementioned monitoring images undergoes preprocessing to standardize its format.

5. A power equipment image monitoring device, characterized in that, include: The data acquisition module is used to acquire normal historical image data and fault historical image data of each of the power devices through different channels to construct a knowledge graph; Historical monitoring datasets are extracted from the knowledge graph; the historical monitoring datasets of each power device are preprocessed to unify their format. A Siamese network is used to extract features from each historical monitoring image and its corresponding historical safety image in the historical monitoring dataset to obtain historical monitoring features and historical safety features. The historical monitoring features and their corresponding historical safety features are then compared to obtain historical feature similarity values. Errors are calculated based on the historical feature similarity values ​​and their corresponding expected values. Gradients of multiple weights in the Siamese network are calculated based on these errors. An optimization algorithm based on these gradients updates the weights to obtain corresponding optimized weights. Hyperparameters in the Siamese network are updated based on the validation set in the historical monitoring dataset to obtain validation parameters. A preset image comparison model is trained using the historical feature similarity values, optimized weights, and validation parameters to obtain a corresponding general power equipment monitoring model. Each monitor and the power equipment monitoring model are hosted on a local node. Multiple monitors are used to collect monitoring images of different power equipment in real time. The feature similarity value calculation module is used to input each of the monitoring images and their corresponding security images into a preset power equipment monitoring model based on the twin network, so that the power equipment monitoring model outputs feature similarity values ​​corresponding to the monitoring images of each power equipment; wherein, the power equipment monitoring model includes: a general power equipment monitoring model; The fault feedback module is used to generate a fault report corresponding to the feature similarity value if there is a feature similarity value that is less than the preset safety threshold of the corresponding power equipment, and transmit it to the control layer.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the power equipment image monitoring method as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the power equipment image monitoring method as described in any one of claims 1 to 4.

Images