CN115619999A - Real-time monitoring method and device for power equipment, electronic equipment and readable medium - Google Patents

Abstract

The present disclosure relates to a real-time monitoring method, device, electronic equipment and computer-readable medium of electric equipment. The method includes: acquiring a real-time image of the power equipment through the Internet of Things; preprocessing the real-time image to generate image data; inputting the image data into a target detection model to generate a target detection result, wherein the target detection model It is a deep neural network model based on Newton's method to optimize the loss function for compression; based on the target detection result, the state of the electric equipment is monitored in real time. The real-time monitoring method, device, electronic device and computer-readable medium of the power equipment involved in the present disclosure can arrange high-precision and low-complexity calculation models in the edge computing equipment, so as to monitor the power equipment in real time and ensure the safe operation of the power grid While reducing the data pressure of the Internet of Things.

Description

Power equipment real-time monitoring method, device, electronic equipment and readable medium

technical field

The present disclosure relates to the field of power equipment inspection, and in particular, to a real-time monitoring method and device for power equipment, electronic equipment, and a computer-readable medium.

Background technique

The substation equipment in my country's power grid has the characteristics of many types, wide distribution, and different structural parameters. In the process of long-term operation, power equipment failures are almost inevitable. The causes of failures include equipment defects left in the manufacturing process, installation and maintenance Problems in maintenance, insulation aging and structural deterioration after long-term operation. At present, the inspection service in the substation is mostly carried out by using intelligent inspection equipment, such as drones, robots, smart helmets, etc., which greatly reduces the workload of operation and maintenance staff and reduces the cost of daily operation and maintenance of power equipment. Real-time shooting will generate a large amount of equipment image data, and it is necessary to manually judge whether the equipment status is normal based on the image. This process increases the workload of the operation and maintenance staff and takes up a lot of manpower. Therefore, the intelligent recognition technology is used to analyze and process the data. It is very necessary to obtain key information and then automate data processing.

Edge computing enables more computing tasks of monitoring applications to be performed on distributed nodes at the edge of the network, so these edge devices can be used to reduce time delays and enable real-time online decision-making. In the field of power equipment operation, maintenance and detection, real-time status monitoring of power equipment in the power Internet of Things environment is required. Since the power Internet of Things deployed by power companies is composed of a large number of power terminal devices, power monitoring applications based on edge computing are generally used. The system analyzes and processes the collected real-time images of power equipment to monitor whether the equipment is in good condition. However, complex network models have resource constraints when deploying terminals or edge devices.

The above information disclosed in this Background section is only for enhancement of understanding of the background of the disclosure and therefore it may contain information that does not form the prior art that is already known in the art to a person of ordinary skill in the art.

Contents of the invention

In view of this, the present disclosure provides a method, device, electronic device and computer-readable medium for real-time monitoring of power equipment, which can arrange high-precision and low-complexity calculation models in edge computing equipment, so as to monitor power equipment in real time , To ensure the safe operation of the power grid while reducing the data pressure of the Internet of Things.

Other features and advantages of the present disclosure will become apparent from the following detailed description, or in part, be learned by practice of the present disclosure.

According to one aspect of the present disclosure, a method for real-time monitoring of power equipment is proposed, the method includes: acquiring a real-time image of the power equipment through the Internet of Things; preprocessing the real-time image to generate image data; inputting the image data into the target In the detection model, a target detection result is generated, wherein the target detection model is a deep neural network model based on Newton's method to optimize the loss function for compression; based on the target detection result, the state of the electric equipment is monitored in real time.

In an exemplary embodiment of the present disclosure, it further includes: acquiring a plurality of historical images of electric equipment; performing preprocessing on the plurality of historical images to generate a plurality of historical image data; based on the plurality of historical image data The deep neural network model is trained to generate an initial target detection model; the loss function is optimized based on Newton's method to compress the initial target detection model to generate the target detection model.

In an exemplary embodiment of the present disclosure, performing preprocessing on the multiple historical images to generate multiple historical image data includes: performing data cleaning on the multiple historical images to generate multiple historical image data; And/or performing data conversion on the multiple historical images to generate multiple historical image data; and/or performing data normalization processing on the multiple historical images to generate multiple historical image data.

In an exemplary embodiment of the present disclosure, optimizing the loss function based on Newton's method to compress the initial target detection model to generate the target detection model includes: generating an exponential loss function; generating a compression function; The exponential loss function is optimized and calculated to obtain the minimized exponential loss function and its corresponding model parameters; decoupling the compression function to obtain the weight of the compressed target detection model; based on the model parameters and the weight Generate the object detection model.

In an exemplary embodiment of the present disclosure, generating an exponential loss function includes: obtaining an initialization parameter, the initialization parameter including a compression target, the number of algorithm iterations, and the number of sample iterations; generating the exponential loss function based on the initialization parameter .

In an exemplary embodiment of the present disclosure, generating the compression function includes: acquiring an original weight tensor of the initial target detection model; and generating the compression function based on the original weight tensor.

In an exemplary embodiment of the present disclosure, the exponential loss function is optimized and calculated based on Newton's method to obtain the minimized exponential loss function and its corresponding model parameters, including: determining the iteration initial value; determining the search direction; based on The search direction is iteratively calculated to solve a minimized exponential function; and the model parameters are determined based on the minimized exponential function.

In an exemplary embodiment of the present disclosure, determining the search direction includes: performing linear transformation on the gradient through a Hessian matrix based on the iteration initial value to obtain the search direction.

In an exemplary embodiment of the present disclosure, decoupling the compression function to obtain the weight of the compressed target detection model includes: determining a dual variable; decoupling based on the dual variable and an alternating direction multiplier method The compression function: calculating the decoupled compression function based on iterative calculation to obtain the weight of the target detection model.

According to one aspect of the present disclosure, a real-time monitoring device for electric equipment is proposed, which includes: an image module, used to acquire real-time images of electric equipment through the Internet of Things; a processing module, used for preprocessing the real-time images, and generating Image data; Calculation module, for inputting said image data in target detection model, generate target detection result, wherein, said target detection model is the deep neural network model based on Newton's method optimization loss function to compress; Monitoring module, It is used to monitor the state of the electric equipment in real time based on the target detection result.

In an exemplary embodiment of the present disclosure, it further includes: a model training module, configured to acquire a plurality of historical images of electric equipment; perform preprocessing on the plurality of historical images to generate a plurality of historical image data; based on the The plurality of historical image data are used to train the deep neural network model to generate an initial target detection model; the loss function is optimized based on Newton's method to compress the initial target detection model to generate the target detection model.

According to an aspect of the present disclosure, an electronic device is proposed, which includes: one or more processors; a storage device for storing one or more programs; when one or more programs are executed by one or more processors Execution causes one or more processors to implement the method as above.

According to one aspect of the present disclosure, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above method is implemented.

According to the power equipment real-time monitoring method, device, electronic equipment and computer-readable medium of the present disclosure, the real-time image of the power equipment is obtained through the Internet of Things; the real-time image is preprocessed to generate image data; the image data is input to the target In the detection model, a target detection result is generated, wherein the target detection model is a deep neural network model based on Newton’s method to optimize the loss function for compression; the method of real-time monitoring of the state of the electric equipment based on the target detection result , can arrange high-precision, low-complexity computing models in edge computing devices, so as to monitor power devices in real time, ensure the safe operation of the power grid, and reduce the data pressure of the Internet of Things.

It is to be understood that both the foregoing general description and the following detailed description are exemplary only and are not restrictive of the present disclosure.

Description of drawings

The above and other objects, features and advantages of the present disclosure will become more apparent by describing in detail example embodiments thereof with reference to the accompanying drawings. The drawings described below are only some embodiments of the present disclosure, and those skilled in the art can obtain other drawings according to these drawings without creative efforts.

Fig. 1 is a system block diagram of a method and device for real-time monitoring of electrical equipment according to an exemplary embodiment.

Fig. 2 is a flowchart of a real-time monitoring method for electric equipment according to an exemplary embodiment.

Fig. 3 is a flow chart of a real-time monitoring method for electric equipment according to another exemplary embodiment.

Fig. 4 is a flowchart of a real-time monitoring method for electric equipment according to another exemplary embodiment.

Fig. 5 is a schematic diagram of a real-time monitoring method for electric equipment according to another exemplary embodiment.

Fig. 6 is a schematic diagram of a real-time monitoring method for electric equipment according to another exemplary embodiment.

Fig. 7 is a block diagram of a real-time monitoring device for electric equipment according to an exemplary embodiment.

Fig. 8 is a block diagram of an electronic device according to an exemplary embodiment.

Fig. 9 is a block diagram showing a computer-readable medium according to an exemplary embodiment.

detailed description

Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The same reference numerals denote the same or similar parts in the drawings, and thus their repeated descriptions will be omitted.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided in order to give a thorough understanding of embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced without one or more of the specific details, or other methods, components, means, steps, etc. may be employed. In other instances, well-known methods, apparatus, implementations, or operations have not been shown or described in detail to avoid obscuring aspects of the present disclosure.

The block diagrams shown in the drawings are merely functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor devices and/or microcontroller devices entity.

The flow charts shown in the drawings are only exemplary illustrations, and do not necessarily include all contents and operations/steps, nor must they be performed in the order described. For example, some operations/steps can be decomposed, and some operations/steps can be combined or partly combined, so the actual order of execution may be changed according to the actual situation.

It will be understood that although the terms first, second, third etc. may be used herein to describe various components, these components should not be limited by these terms. These terms are used to distinguish one component from another. Thus, a first component discussed below could be termed a second component without departing from the teachings of the disclosed concepts. As used herein, the term "and/or" includes any one and all combinations of one or more of the associated listed items.

Those skilled in the art can understand that the drawings are only schematic diagrams of exemplary embodiments, and the modules or processes in the drawings are not necessarily necessary for implementing the present disclosure, and thus cannot be used to limit the protection scope of the present disclosure.

Network devices in the Internet of Things environment have accumulated a large amount of data, and cloud computing-related technologies have the potential to be successful in processing these massive amounts of data. However, for most typical monitoring applications, processing and analyzing massive monitoring data is still a challenge. Complex network models have resource limitations when deployed on terminals or edge devices. Research on neural network model compression technology, on the premise of ensuring model accuracy, realizes the transplantation and calculation optimization of neural network models on terminals or edge devices. The key to edge computing complexity.

The inventors of the present application found that in the past few years, various techniques for compressing DNN (Deep Neural Network) models have been proposed. Pruning and quantization are the two most widely used methods in practice. In addition to using weight pruning for model compression, the pruning method also proposes channel (filter/neuron) pruning to remove the entire filter of CNN weights, thus achieving inference acceleration. Besides reducing parameters by pruning, quantization is considered as another direction to compress DNNs. Quantization interval can be uniform or non-uniform, in general, non-uniform quantization can achieve higher compression ratio, while uniform quantization can provide speedup. The quantization bit width can be further reduced by Hoffman coding. Besides scalar quantization, vector quantization can also be applied to DNN model compression.

In order to maximize compression performance, there are some methods that perform training together with pruning and quantization, but in this case there is a problem of layer sparsity and quantization bit width interacting, and these methods rely on setting hyperparameters to Compression layers, increasing the difficulty of manual selection of compression ratios or hyperparameter tuning. In view of the technical bottlenecks in the prior art, the present application provides a real-time monitoring method for electric equipment.

In the real-time monitoring method of power equipment disclosed in the present disclosure, the specific technical description is given by taking the monitoring application of power equipment operation and maintenance as an example. It is understandable that the method disclosed in the present disclosure can also be applied to other fields, and this application does not limit.

More specifically, in the target recognition task of power distribution operation and maintenance, first collect the real-time equipment status images captured by the intelligent patrol equipment used in the patrol business in power distribution operation and maintenance, and select appropriate research samples from them. Afterwards, image enhancement techniques are used to preprocess the data. Secondly, the images are manually annotated to obtain a data set of electric equipment target detection. This image dataset can then be leveraged for edge computing-based object detection tasks. In the automatic DNN compression framework proposed in this paper, it is first necessary to preprocess the image data set, perform data cleaning on the image data, remove dirty data, and then convert the image data into tensors, and then set the mean and variance to perform channel-by-channel The image is normalized; then model training is performed based on the processed image data set, and the optimized automatic DNN compression method is used to compress the trained model, and the optimization method of the loss function is used to speed up the training convergence process of the loss function, and finally Get the compressed neural network model.

Compared with the DNN model in the prior art, the target detection model proposed in the disclosed real-time monitoring method of electric power equipment has a higher compression rate and accuracy, significantly reduces the size of the DNN model, and can effectively reduce the the complexity. The methods in the present disclosure are described below with the help of specific examples.

Fig. 1 is a system block diagram of a method and device for real-time monitoring of electrical equipment according to an exemplary embodiment.

As shown in FIG. 1 , the system architecture 10 may include terminal devices 101 , 102 , 103 , a network 104 and a server 105 . The network 104 is used as a medium for providing communication links between the terminal devices 101 , 102 , 103 and the server 105 . Network 104 may include various connection types, such as wires, wireless communication links, or fiber optic cables, among others.

The terminal devices 101, 102, 103 can interact with the server 105 through the network 104 to receive or send data and the like. Various communication client applications can be installed on the terminal devices 101, 102, 103, such as video surveillance applications, web browser applications, real-time data transmission applications, email clients, and the like.

Terminal devices 101, 102, and 103 may be various electronic devices that have monitoring functions and support data transmission or calculation, including but not limited to power electronic devices, smart cameras, smart monitoring instruments, and the like.

Terminal devices 101, 102, 103 can obtain real-time images of power equipment, for example, through the Internet of Things; terminal devices 101, 102, 103 can, for example, preprocess the real-time images to generate image data; terminal devices 101, 102, 103 can, for example, Input the image data into a target detection model to generate a target detection result, wherein the target detection model is a deep neural network model based on Newton's method to optimize the loss function for compression; terminal devices 101, 102, 103 can, for example, be based on the The state of the electric equipment is monitored in real time based on the detection result of the target.

The server 105 may be a server that provides various services, for example, a background server that provides support for tasks processed by the terminal devices 101 , 102 , and 103 . The background server can analyze and process the received request, and feed back the processing result (compressed monitoring model) to the terminal device.

The server 105 can, for example, acquire a plurality of historical images of electric equipment; the server 105 can, for example, preprocess the plurality of historical images to generate a plurality of historical image data; The network model is trained to generate an initial target detection model; the server 105 may, for example, optimize a loss function based on Newton's method to compress the initial target detection model to generate the target detection model.

The server 105 can be an entity server, and can also be composed of multiple servers, for example. It should be noted that the real-time monitoring method for electric power equipment provided by the embodiment of the present disclosure can be executed by the server 105 and/or the terminal equipment 101, 102, 103 , correspondingly, the apparatus for real-time monitoring of electric power equipment may be set in the server 105 and/or the terminal equipment 101 , 102 , 103 .

Fig. 2 is a flowchart of a real-time monitoring method for electric equipment according to an exemplary embodiment. The method 20 for real-time monitoring of electrical equipment can be applied to edge equipment of the Internet of Things, and at least includes steps S202 to S208.

As shown in Fig. 2, in S202, the real-time image of the electric equipment is obtained through the Internet of Things. Among them, the Internet of Things (IOT) refers to the real-time collection of any object or process that needs to be monitored, connected, and interacted through various devices and technologies such as information sensors, radio frequency identification technology, global positioning system, infrared sensors, and laser scanners. , to collect various required information such as sound, light, heat, electricity, mechanics, chemistry, biology, location, etc., and realize the ubiquitous connection between things and things, things and people through various possible network accesses, and realize the Intelligent perception, identification and management of processes and processes. Edge devices in the Internet of Things can acquire real-time images and process them, where an edge device (edgedevice) is a device that provides an entry point to the core network of an enterprise or service provider.

In S204, the real-time image is preprocessed to generate image data. In order to purify the data set, it is necessary to perform a certain scale screening process on the acquired original power equipment image data, to eliminate blurred images, repeated images, and damaged image files, so as to avoid unnecessary impact on the quality of the training model, and retain Image data with relatively clear targets.

Among them, the data conversion can store the acquired image data set in different folders according to the label, and then load the data, and the gray scale of the image is transformed from [0,255] to [0,1]. between. Randomly crop the given image into different sizes and aspect ratios, then scale the cropped image to the specified size, and randomly flip the given PIL image horizontally with a given probability. Convert the image into a storage format in memory, input the bytes as a stream, convert it into a one-dimensional tensor, reorganize and transpose the tensor, divide each element of the current tensor by 255, and output tensor.

In order to speed up the convergence speed of the model, the image is normalized channel by channel. The calculation formula is shown in formula (1), where mean represents the mean value of each channel, and std represents the standard deviation of each channel. Using this formula, the data can be standardized, and the mean of the standardized result is 0, and the standard deviation is 1.

output=(input-mean)/std(1)

In S206, the image data is input into a target detection model to generate a target detection result, wherein the target detection model is a deep neural network model based on Newton's method to optimize a loss function for compression.

In S208, the state of the electrical equipment is monitored in real time based on the target detection result. A warning message can be generated when detection results contain preset targets. The preset target may be an image indicating damage to electrical equipment.

According to the real-time monitoring method for power equipment of the present disclosure, the real-time image of the power equipment is obtained through the Internet of Things; the real-time image is preprocessed to generate image data; the image data is input into the target detection model to generate a target detection result, wherein , the target detection model is a deep neural network model based on Newton’s method to optimize the loss function for compression; the method of real-time monitoring of the state of the power equipment based on the target detection result can be used to integrate high-precision, low-complexity The computing model is arranged in the edge computing device to monitor the power equipment in real time, to ensure the safe operation of the power grid and reduce the data pressure of the Internet of Things.

It should be clearly understood that this disclosure describes how to make and use specific examples, but that the principles of the disclosure are not limited to any details of these examples. Rather, these principles can be applied to many other embodiments based on the teachings of the present disclosure.

Fig. 3 is a flow chart of a real-time monitoring method for electric equipment according to another exemplary embodiment. The method 30 for real-time monitoring of power equipment can be applied in a server supporting edge computing equipment, and the process 30 shown in FIG. 3 is a supplementary description of the process shown in FIG. 2 .

As shown in FIG. 3 , in S302 , multiple historical images of the electrical equipment are acquired.

In S304, perform preprocessing on the multiple historical images to generate multiple historical image data. Including: performing data cleaning on the multiple historical images to generate multiple historical image data; and/or performing data conversion on the multiple historical images to generate multiple historical image data; and/or performing data conversion on the multiple historical images Images are normalized to generate multiple historical image data.

In S306, the deep neural network model is trained based on the plurality of historical image data to generate an initial target detection model. The image data after data preprocessing is input into the DNN model for training to generate an initial target detection model.

In S308, the loss function is optimized based on Newton's method to compress the initial target detection model to generate the target detection model. For example, generate an exponential loss function; generate a compression function; optimize and calculate the exponential loss function based on Newton's method to obtain a minimized exponential loss function and its corresponding model parameters; decouple the compression function to obtain all compressed The weight of the target detection model; generating the target detection model based on the model parameters and the weight.

In the disclosed real-time monitoring method for electric power equipment, a deep neural network model compression optimization method for intelligent detection of electric power equipment is proposed. Under the premise, by calculating the original weight of the model, initializing parameters, and decoupling the parameters using the Alternate Iterative Multiplier Algorithm (ADMM), this method does not need to use any hyperparameters to manually set the compression rate of each layer, and can simultaneously learn the compression rate and Compress model weights.

In the disclosed real-time monitoring method for electric equipment, a model compression optimization technology based on Newton's method loss function is proposed. Different from traditional classic loss functions such as information entropy and mean square error, this patent uses Newton's method to speed up the training convergence process of loss functions and improve the accuracy of model classification results.

The experimental results show that, compared with the current popular methods, the method proposed in this patent has higher compression rate and accuracy, reduces the size of the DNN model, and can effectively reduce the complexity of the model.

Fig. 4 is a flowchart of a real-time monitoring method for electric equipment according to another exemplary embodiment. The process 40 shown in FIG. 4 is a detailed description of S308 "optimize the loss function based on Newton's method to compress the initial target detection model to generate the target detection model" in the process shown in FIG. 3 .

As shown in Fig. 4, in S402, an exponential loss function is generated. The method includes: acquiring initialization parameters, where the initialization parameters include compression targets, algorithm iteration times, and sample iteration times; generating the exponential loss function based on the initialization parameters.

In S404, a compression function is generated. The method includes: obtaining an original weight tensor of the initial target detection model; and generating the compression function based on the original weight tensor.

In S406, the exponential loss function is optimized and calculated based on Newton's method to obtain a minimized exponential loss function and its corresponding model parameters. The method includes: determining an initial iteration value; determining a search direction; performing iterative calculation based on the search direction to solve a minimized exponential function; determining the model parameters based on the minimized exponential function.

Wherein, determining the search direction includes: performing linear transformation on the gradient through the Hessian matrix based on the iterative initial value to obtain the search direction.

In S408, the compression function is decoupled to obtain compressed weights of the target detection model. The method includes: determining a dual variable; decoupling the compression function based on the dual variable and an alternating direction multiplier method; and calculating the decoupled compression function based on iterative calculation to obtain the weight of the target detection model.

In S410, the target detection model is generated based on the model parameters and the weights.

In one embodiment, the original weight tensor W of the model can be calculated, and the initialization parameters include the target size of model compression, the total number of SGD iterations of the algorithm, and the number of iterations required to train all samples in the training set once. Then define a loss function l, the disclosure uses an exponential loss function (Exponential Loss) to calculate the loss, as shown in formula (2). Among them, n is the number of samples, y is the real value of the sample, and f( _xi ) is the weight of the i-th iteration model.

是拥有L层DNN的权重向量集，b(W⁽ⁱ⁾)是编码W中所有非零元素的最小比特宽，L0范数||W||₀是W中非零元素的个数，Z为模型目标大小，l为损失函数。The general function of DNN compression is shown in Equation (3), which is constrained by the total size of compressed DNN weights, and the goal is to minimize the loss function. in

is the weight vector set with L-layer DNN, b(W ⁽ⁱ⁾ ) is the minimum bit width of all non-zero elements in encoding W, L0 norm ||W|| ₀ is the number of non-zero elements in W, Z is the model target size, and l is the loss function.

并将等式约束吸收到增广拉格朗日函数中，其中λ＞0，是一个超参数。

是DNN权重W的副本。Since b(.) and || _.‖0 are non-differentiable functions, they cannot be solved by normal training algorithms. In the embodiment of this application, the ADMM method can be used to decouple its L0 norm and bit width part, specifically It can be shown in formula (4). By introducing a dual variable

And absorb the equality constraint into the augmented Lagrangian function, where λ>0 is a hyperparameter.

is a copy of the DNN weight W.

Using the ADMM method can enable the model compression method based on joint pruning and quantization to be able to train normally. By iteratively updating the three variables W, V and Y in formula (4), the non-differentiable problem of b(.) and _‖.‖0 can be solved question. Finally, output the weights of the compressed DNN model.

则停止循环。接着开始计算梯度G，如公式(5)所示，其中训练样本总数为n，t＝0,1,...,n，f(x)为损失函数，x为优化的参数对象：In order to optimize the loss function l, in the present disclosure, the idea of Newton's method can be used to perform linear transformation on the gradient of the Hessian matrix to obtain the search direction, thereby speeding up the convergence process of the loss function. First, it is necessary to select the initial value of iteration x0∈Ω, select ε>0, and repeat the following operations. like

then stop the loop. Then start to calculate the gradient G, as shown in formula (5), where the total number of training samples is n, t=0,1,...,n, f(x) is the loss function, and x is the optimized parameter object:

The calculation of the Hessian matrix H is shown in formula (6);

The calculation of the search direction d is shown in formula (7):

The calculation of the last update iteration point is shown in formula (8):

x _t+1 ＝x _t -d _t (8)

As shown above, by continuously iteratively solving the minimum value of the function, the minimized loss function and model parameter values are finally obtained to realize the parameter optimization of the model.

The effect of using the model compression method mentioned in the present disclosure is described below based on a specific experimental result:

Experimental Setup and Datasets

Experiments are carried out using Win10, GPU with GTX1070 and 8G memory and CUDA/CUDNN. The training data set consists of 5,944 state images of electrical equipment collected, including six types of equipment state images such as respirators, meters, insulators, oil leakage, foreign objects, and metal corrosion. The collected data set is divided into standard training/test data, 70% of the entire data set is used for model training as the training set, and 30% of the entire data set is used for model testing as the test set. AlexNet is used as a DNN model for training and testing power image data to test the performance of the disclosed compression method.

The batch size can be set to 128, the momentum SGD is used to optimize the exponential loss function l(W), the initial learning rate is 0.005, the cosine annealing strategy is used to decay the learning rate, and the hyperparameter λ=0.05 is set. In order to make the effect of the comparison between the methods more obvious, the compression budget Z can also be set to a value similar to or smaller than that of the comparison method. Experiments are performed for 120 iterations on a dataset containing 4160 training instances.

Experimental indicators

In order to evaluate the performance of model compression, the compression ratio and accuracy are used as evaluation indicators, as shown in formula (9) and formula (10).

Among them, R _C represents the compression factor, N _original represents the number of original model parameters, and N _compressed represents the number of model parameters after compression.

Among them, A represents the accuracy rate, N _correct represents the number of correctly classified samples, and N _test represents the number of all samples in the test set. In the accuracy rate index, the correct sample classification means that the predicted label takes the largest one in the final probability vector as the prediction result. If the classification with the highest probability in the prediction result is correct, the prediction is correct, and then the frequency of correct classification is calculated. for the accuracy rate.

comparison method

In order to verify the effectiveness of the automatic neural network compression method adopted by the joint pruning and quantization strategy, the present disclosure compares the automatic DNN compression method with the pruning method in deep compression, a currently popular model compression method. Different from the end-to-end framework adopted in this disclosure, this method needs to set the clipping rate as a hyperparameter.

The pruning method is introduced below. The method first learns connections through normal network training. Next, small weight connections are pruned, i.e. all connections with weights below a threshold are removed from the network. Finally, the network is retrained to learn the final weights of the remaining sparse connections. This pruning method reduces the number of parameters of the model.

Experimental Results and Analysis

In this experiment, the method used in this disclosure and the deep compression method used for comparison are used to perform training and testing on the AlexNet model. The training and testing data are all using the power equipment operation inspection image data set to detect equipment in different states, as shown in the figure Figure 5 and Figure 6 show the experimental results of the model compression factor and the classification accuracy of the compressed model for the two methods, respectively.

Figure 5 shows that the compression factor of the AlexNet model using the automatic DNN compression method is significantly higher than that produced by the deep compression method.

After the AlexNet model is compressed by the automatic neural network compression method and the deep compression method, the compressed model is retrained with the training set, and the trained compressed model is tested with the test set. As shown in FIG. 6 , the model compressed by the disclosed method has higher classification accuracy than the deep compression method.

Based on the analysis of Figures 5 and 6, the compression method introduced in this disclosure is more effective than the neural network compression method in the prior art, and has a higher compression factor and accuracy. This method is very suitable for edge computing environments that require processing and in models that analyze large amounts of image data. Therefore, the method adopted in this disclosure is effective in the application of target detection based on power equipment images, especially while achieving high model prediction accuracy, it reduces the size of the DNN model, reduces the model complexity, and further reduces edge computing load.

This disclosure introduces an optimized automatic neural network compression method combined with pruning and quantization strategies, and provides specific implementation measures for using this method in the field of power equipment operation and maintenance. In this disclosure, Newton's method is used to speed up A loss function optimization method for the training convergence process of the loss function. The effectiveness and practicability of this method for neural network compression are verified by a large number of experiments. This method can be used in target detection edge services for processing and analyzing real-time power equipment defect image data, reducing model complexity, reducing the load of edge computing, especially reducing the size of the model while achieving high model prediction accuracy .

Those skilled in the art can understand that all or part of the steps for implementing the above embodiments are implemented as computer programs executed by a CPU. When the computer program is executed by the CPU, the above-mentioned functions defined by the above-mentioned methods provided in the present disclosure are executed. The program can be stored in a computer-readable storage medium, which can be a read-only memory, a magnetic disk or an optical disk, and the like.

In addition, it should be noted that the above-mentioned figures are only schematic illustrations of processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not imply or limit the chronological order of these processes. In addition, it is also easy to understand that these processes may be executed synchronously or asynchronously in multiple modules, for example.

The following are device embodiments of the present disclosure, which can be used to implement the method embodiments of the present disclosure. For details not disclosed in the disclosed device embodiments, please refer to the disclosed method embodiments.

Fig. 7 is a block diagram of a real-time monitoring device for electric equipment according to an exemplary embodiment. As shown in FIG. 7 , the real-time monitoring device 70 for electric equipment includes: an image module 702 , a processing module 704 , a calculation module 706 , a monitoring module 708 , and a model training module 710 .

The image module 702 is used to obtain the real-time image of the power equipment through the Internet of Things;

The processing module 704 is configured to preprocess the real-time image to generate image data;

The calculation module 706 is used to input the image data into the target detection model to generate a target detection result, wherein the target detection model is a deep neural network model based on Newton's method to optimize the loss function for compression;

The monitoring module 708 is configured to monitor the state of the electrical equipment in real time based on the target detection result.

The model training module 710 is used to obtain a plurality of historical images of electric equipment; preprocess the plurality of historical images to generate a plurality of historical image data; train the deep neural network model based on the plurality of historical image data to generate An initial target detection model; optimizing a loss function based on Newton's method to compress the initial target detection model to generate the target detection model.

According to the real-time monitoring device for power equipment of the present disclosure, the real-time image of the power equipment is obtained through the Internet of Things; the real-time image is preprocessed to generate image data; the image data is input into the target detection model to generate a target detection result, wherein , the target detection model is a deep neural network model based on Newton’s method to optimize the loss function for compression; the method of real-time monitoring of the state of the power equipment based on the target detection result can be used to integrate high-precision, low-complexity The computing model is arranged in the edge computing device to monitor the power equipment in real time, to ensure the safe operation of the power grid and reduce the data pressure of the Internet of Things.

Fig. 8 is a block diagram of an electronic device according to an exemplary embodiment.

An electronic device 800 according to this embodiment of the present disclosure is described below with reference to FIG. 8 . The electronic device 800 shown in FIG. 8 is only an example, and should not limit the functions and application scope of the embodiments of the present disclosure.

As shown in FIG. 8, electronic device 800 takes the form of a general-purpose computing device. Components of the electronic device 800 may include, but are not limited to: at least one processing unit 810, at least one storage unit 820, a bus 830 connecting different system components (including the storage unit 820 and the processing unit 810), a display unit 840, and the like.

Wherein, the storage unit stores program codes, and the program codes can be executed by the processing unit 810, so that the processing unit 810 executes the steps described in this specification according to various exemplary embodiments of the present disclosure. For example, the processing unit 810 may execute the steps shown in FIG. 2 , FIG. 3 , and FIG. 4 .

The storage unit 820 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 8201 and/or a cache storage unit 8202 , and may further include a read-only storage unit (ROM) 8203 .

The storage unit 820 may also include a program/utility 8204 having a set (at least one) of program modules 8205, such program modules 8205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include the implementation of the network environment.

Bus 830 may represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local area using any of a variety of bus structures. bus.

The electronic device 800 can also communicate with one or more external devices 800' (such as keyboards, pointing devices, Bluetooth devices, etc.), so that the user can communicate with the devices that the electronic device 800 interacts with, and/or the electronic device 800 can communicate with a Any device (eg, router, modem, etc.) that communicates with one or more other computing devices. Such communication may occur through input/output (I/O) interface 850 . Moreover, the electronic device 800 can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN) and/or a public network such as the Internet) through the network adapter 860 . The network adapter 860 can communicate with other modules of the electronic device 800 through the bus 830 . It should be appreciated that although not shown, other hardware and/or software modules may be used in conjunction with electronic device 800, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives And data backup storage system, etc.

Through the description of the above implementations, those skilled in the art can easily understand that the example implementations described here can be implemented by software, or by combining software with necessary hardware. Therefore, as shown in Figure 9, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, and the software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, or a mobile hard disk). etc.) or on the network, including several instructions to make a computing device (which may be a personal computer, server, or network device, etc.) execute the above method according to the embodiments of the present disclosure.

The software product may utilize any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: electrical connection with one or more conductors, portable disk, hard disk, random access memory (RAM), read only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

The computer readable storage medium may include a data signal carrying readable program code in baseband or as part of a carrier wave traveling as a data signal. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A readable storage medium may also be any readable medium other than a readable storage medium that can send, propagate or transport a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the readable storage medium may be transmitted by any suitable medium, including but not limited to wireless, cable, optical cable, RF, etc., or any suitable combination of the above.

Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages—such as Java, C++, etc., as well as conventional procedural Programming language - such as "C" or similar programming language. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server to execute. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., using an Internet service provider). business to connect via the Internet).

The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by one of the devices, the computer-readable medium realizes the following functions: obtain real-time images of electric equipment through the Internet of Things; The real-time image is preprocessed to generate image data; the image data is input into a target detection model to generate a target detection result, wherein the target detection model is a deep neural network model based on Newton's method to optimize the loss function for compression; The target detection result monitors the state of the electric equipment in real time.

Those skilled in the art can understand that the above-mentioned modules can be distributed in the device according to the description of the embodiment, and corresponding changes can also be made in one or more devices that are only different from the embodiment. The modules in the above embodiments can be combined into one module, and can also be further split into multiple sub-modules.

Through the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described here can be implemented by software, or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of software products, and the software products can be stored in a non-volatile storage medium (which can be CD-ROM, U disk, mobile hard disk, etc.) or on the network , including several instructions to make a computing device (which may be a personal computer, server, mobile terminal, or network device, etc.) execute the method according to the embodiment of the present disclosure.

Exemplary embodiments of the present disclosure have been specifically shown and described above. It should be understood that the disclosure is not limited to the detailed structures, arrangements or methods of implementation described herein; on the contrary, the disclosure is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims (13)

1. A real-time monitoring method for power equipment is characterized by comprising the following steps:

acquiring a real-time image of the power equipment through the Internet of things;

preprocessing the real-time image to generate image data;

inputting the image data into a target detection model to generate a target detection result, wherein the target detection model is a deep neural network model which is based on a Newton method to optimize a loss function for compression;

and monitoring the state of the power equipment in real time based on the target detection result.

2. The real-time monitoring method for the power equipment as claimed in claim 1, further comprising:

acquiring a plurality of historical images of the power equipment;

preprocessing the plurality of history images to generate a plurality of history image data;

training a deep neural network model based on the plurality of historical image data to generate an initial target detection model;

and optimizing a loss function based on a Newton method to compress the initial target detection model to generate the target detection model.

3. The power equipment real-time monitoring method of claim 2, wherein preprocessing the plurality of historical images to generate a plurality of historical image data comprises:

performing data cleansing on the plurality of historical images to generate a plurality of historical image data; and/or

Performing data conversion on the plurality of history images to generate a plurality of history image data; and/or

And performing data normalization processing on the plurality of historical images to generate a plurality of historical image data.

4. The power device real-time monitoring method of claim 2, wherein optimizing a loss function based on newton's method to compress the initial target detection model to generate the target detection model comprises:

generating an exponential loss function;

generating a compression function;

performing optimization calculation on the exponential loss function based on a Newton method to obtain a minimized exponential loss function and corresponding model parameters thereof;

decoupling the compression function to obtain weights of the compressed target detection model;

generating the target detection model based on the model parameters and the weights.

5. The method of real-time monitoring of power equipment of claim 4, wherein generating an exponential loss function comprises:

acquiring initialization parameters, wherein the initialization parameters comprise a compression target, algorithm iteration times and sample iteration times;

generating the exponential-loss function based on the initialization parameter.

6. The power device real-time monitoring method of claim 4, wherein generating a compression function comprises:

acquiring an original weight tensor of the initial target detection model;

generating the compression function based on the original weight tensor.

7. The method for monitoring the power equipment in real time as claimed in claim 4, wherein the optimization calculation of the exponential loss function based on Newton's method to obtain the minimized exponential loss function and the corresponding model parameters thereof comprises:

determining an iteration initial value;

determining a searching direction;

performing iterative calculations based on the search direction to solve a minimized exponential function;

determining the model parameters based on the minimization exponential function.

8. The real-time monitoring method for the power equipment as claimed in claim 7, wherein the determining of the search direction comprises:

and performing linear transformation on the gradient through a Hessian matrix based on the iteration initial value to obtain a search direction.

9. The power equipment real-time monitoring method of claim 4, wherein decoupling the compression function to obtain the weights of the compressed target detection model comprises:

determining a dual variable;

decoupling the compression function based on the dual variable and an alternating direction multiplier method;

calculating the decoupled compression function based on iterative calculation to obtain the weight of the target detection model.

10. The utility model provides an electrical equipment real-time supervision device which characterized in that includes:

the image module is used for acquiring a real-time image of the power equipment through the Internet of things;

the processing module is used for preprocessing the real-time image to generate image data;

the calculation module is used for inputting the image data into a target detection model to generate a target detection result, wherein the target detection model is a deep neural network model which is compressed based on a Newton method optimization loss function;

and the monitoring module is used for monitoring the state of the power equipment in real time based on the target detection result.

11. The real-time monitoring device for electric power equipment according to claim 10, further comprising:

the model training module is used for acquiring a plurality of historical images of the power equipment; preprocessing the plurality of historical images to generate a plurality of historical image data; training a deep neural network model based on the plurality of historical image data to generate an initial target detection model; and optimizing a loss function based on a Newton method to compress the initial target detection model to generate the target detection model.

12. An electronic device, comprising:

one or more processors;

storage means for storing one or more programs;

when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-9.

13. A computer-readable medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to any one of claims 1-9.

Images