CN116091027A - Electric power equipment maintenance method, device and system - Google Patents

Abstract

The application discloses a power equipment overhaul method, device and system, and relates to the technical field of intelligent management of power equipment, wherein the method comprises the following steps: acquiring a 3D image of the target power equipment; inputting the 3D image into a target fault recognition model, and performing fault recognition on the target power equipment to obtain a recognition result; the target fault recognition model is constructed based on a 3D U-NET full convolution neural network algorithm; generating a fault removal strategy according to the identification result and the electronic file of the target power equipment; and sending a fault removal operation guide to the equipment maintenance system according to the fault removal strategy. The scheme of the utility model realizes the intellectualization of the elimination of the power equipment faults, and improves the accuracy and efficiency of the elimination and maintenance of the power equipment faults.

Description

Power equipment maintenance method, device and system

technical field

The present application relates to the technical field of intelligent management of electric equipment, in particular to a method, device and system for overhauling electric equipment.

Background technique

The development of power equipment maintenance system is roughly divided into three stages: accident maintenance, planned maintenance and condition maintenance. Troubleshooting, also called troubleshooting, refers to repairing equipment when it breaks down and cannot continue to operate. Planned maintenance, also known as regular maintenance, is a pre-set maintenance work content and cycle, and the equipment is repaired after a period of operation. Condition-based maintenance is a maintenance system based on the actual operating state of the equipment, and a maintenance method based on computer technology, detection technology, power technology, diagnostic technology, and prediction technology.

The existing power equipment maintenance field mainly uses the Internet of Things, infrared imaging technology, RFID technology, etc. to monitor and analyze the status of power equipment, and finally realizes the maintenance of power equipment to ensure the safe and efficient operation of power equipment.

However, in the above scheme, the state of the power equipment can only be represented by the collected information including equipment simulation information, red heat imaging, electronic label information, etc., which cannot show the real status of the power equipment very intuitively and accurately, which is not conducive to experts , technicians, etc. have an understanding of the failure and operation of power equipment, which affects the effect and efficiency of power equipment maintenance.

Contents of the invention

The purpose of the embodiments of the present application is to provide a method, device, and system for overhauling electric equipment, so as to solve the problems of low efficiency and poor overhaul effect for electric equipment in the prior art.

In order to achieve the above purpose, an embodiment of the present application provides a method for overhauling electric equipment, including:

Obtain a 3D image of the target electrical equipment;

The 3D image is input to the target fault recognition model, and the fault recognition is carried out to the target electrical equipment to obtain the recognition result; wherein, the target fault recognition model is constructed based on the 3D U-NET full convolutional neural network algorithm;

generating a troubleshooting strategy according to the identification result and the electronic file of the target electric device;

According to the troubleshooting strategy, the troubleshooting operation guide is sent to the equipment maintenance system.

Optionally, the method also includes:

In the case of failure of the target electric device, displaying the 3D image through a remote terminal device;

receiving an analysis result determined from the 3D image displayed on the remote terminal device;

According to the troubleshooting strategy, send troubleshooting operation instructions to the equipment maintenance system, including:

generating the troubleshooting operation guide according to the troubleshooting strategy and the analysis results;

Sending the troubleshooting operation guide to the equipment maintenance system.

Optionally, generating a troubleshooting strategy according to the identification result and the electronic file of the target electric device, including:

According to the 3D image, retrieve the electronic file of the target electric device from the device electronic file library;

generating the troubleshooting strategy according to the electronic file and the identification result;

Wherein, the electronic archive includes at least one of parameter information, historical operation data, historical troubleshooting data, historical 3D images, troubleshooting operation guide, maintenance video information and/or maintenance record information of the target electronic device.

Optionally, inputting the 3D image into the target fault identification model, performing fault identification on the target electrical equipment, and obtaining identification results, including:

In the case that the target electric equipment is predicted to be faulty according to the monitored operating state data of the target electric equipment, or when an indication that the target electric equipment is faulty is received, the 3D image is input to the target electric equipment. The target fault identification model is used to perform fault identification on the target electric equipment to obtain identification results.

Optionally, the method also includes:

receiving maintenance video information and/or maintenance record information of the target electrical equipment;

saving the maintenance video information and/or the maintenance record information in the electronic file of the target electrical equipment.

Optionally, the method also includes:

Receiving and storing updated troubleshooting instructions during troubleshooting and repair of the target electrical device.

Optionally, the method also includes:

saving the 3D image to an electronic file of the target electric device.

Optionally, the method further includes: training the fault identification model based on the following steps to construct the target fault identification model:

Based on the 3D U-NET full convolutional neural network and the troubleshooting data in the electronic archives of the target power equipment, construct a fault identification model corresponding to the type of the target power equipment;

Input the collected 3D image sample set into the fault identification model, and adjust the parameters of the fault identification model;

When the prediction result of the fault identification model does not meet the expected condition, return to the step of adjusting the parameters of the fault identification model until the prediction result meets the expected condition, and obtain the target fault identification model.

The embodiment of the present application also provides a power equipment maintenance device, including:

The first acquisition module is used to acquire the 3D image of the target electrical equipment;

An identification module, configured to input the 3D image into a target fault identification model, perform fault identification on the target electrical equipment, and obtain identification results; wherein, the fault identification model is constructed based on a 3D U-NET full convolutional neural network algorithm ;

A generating module, configured to generate a troubleshooting strategy according to the identification result and the electronic file of the target electric device;

The sending module is configured to send the troubleshooting operation guide to the equipment maintenance system according to the troubleshooting strategy.

The embodiment of the present application also provides a power equipment maintenance system, including: a processor, a memory, and a program stored on the memory and operable on the processor. When the program is executed by the processor, the above The steps of the electric equipment maintenance method.

An embodiment of the present application further provides a readable storage medium, wherein a program is stored on the readable storage medium, and when the program is executed by a processor, the steps of the above electric equipment maintenance method are realized.

The above technical solution of the present application has at least the following beneficial effects:

In the power equipment maintenance method of the embodiment of the present application, firstly, the 3D image of the target power equipment is obtained to accurately display the real situation of the power equipment; secondly, the 3D image is input into the target fault identification model, and the fault of the target power equipment is detected. Identify and obtain identification results, wherein, the target fault identification model is constructed based on the 3D U-NET full convolution neural network algorithm; the intelligent identification of power equipment faults is realized, and the dependence on human factors is reduced; again according to the identification Based on the results and the electronic files of the target power equipment, a troubleshooting strategy is generated; the troubleshooting strategy is automatically generated based on historical data; finally, according to the troubleshooting strategy, a troubleshooting operation guide is sent to the equipment maintenance system. In this way, the problems of low maintenance efficiency and poor maintenance effect of electric equipment in the prior art are solved.

Description of drawings

FIG. 1 is a schematic flow diagram of a power equipment maintenance method according to an embodiment of the present application;

Fig. 2 is a schematic structural diagram of a power equipment maintenance device according to an embodiment of the present application;

FIG. 3 is a schematic structural diagram of a power equipment maintenance system according to an embodiment of the present application.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of them. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of this application.

The terms "first", "second" and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific sequence or sequence. It is to be understood that the data so used can be interchanged under appropriate circumstances such that the embodiments of the application can be practiced in sequences other than those illustrated or described herein. In addition, "and/or" in the specification and claims means at least one of the connected objects, and the character "/" generally means that the related objects are an "or" relationship.

Before describing the power equipment maintenance method of the embodiment of the application, the technical content related to it will be described first:

The principle of 3D U-Net fully convolutional neural network algorithm:

Net is improved based on Fully Convolution Network (FCN), and uses data enhancement to train a relatively small number of samples. U-Net is a variant of the fully convolutional neural network. It is named U-Net because of its structural form letter U. The entire neural network is mainly composed of two parts: the search path (contracting path) and the expanding path (expanding path). The search path is mainly used to capture the context information (context information) in the picture, and the corresponding extended path is to accurately locate (localize) the part that needs to be segmented in the picture.

The model of 3D U-Net is changed based on the model of U-Net, which includes an encoding part and a decoding part. The encoding part is used to analyze the whole picture and perform feature extraction and analysis, while the decoding part is to generate a segmented block image.

3D holographic projection principle:

3D holographic projection mainly uses the principle of interference and diffraction to record and reproduce the real three-dimensional image of the device. Firstly, the light wave information of the object is recorded by using the principle of interference. The power equipment forms a diffuse object beam under laser irradiation; another part of the laser light is used as a reference beam to irradiate on the holographic film and interfere with the object beam superposition, and convert the phase and amplitude of each point on the object light wave into a spatially varying Intensity, so as to use the contrast and spacing between the interference fringes to record all the information of the light wave of the object. After the film with interference fringes is processed through developing, fixing and other processing procedures, it becomes a hologram, or a hologram. Then use the principle of diffraction to reproduce the light wave information of the device, so that the device can be displayed in a three-dimensional image.

The power equipment maintenance method, device, and system provided by the embodiments of the present application will be described in detail below through specific embodiments and application scenarios with reference to the accompanying drawings.

As shown in Figure 1, it is a schematic flow chart of the power equipment maintenance method of the embodiment of the present application, and the method includes:

Step 101, acquiring a 3D image of the target electrical equipment;

In this step, the 3D image of the target power equipment can be obtained by 3D modeling of the images collected by the video capture equipment deployed around the power equipment, or the power equipment maintenance system directly acquires the 3D image of the target power equipment from the video capture equipment image, wherein the video capture device can be a camera, specifically, the video capture device is a device with special functions, for example, the video capture device can have a certain penetrating power, such as in the geological survey field or the medical field, so that it can gain insight into The internal condition of the target electrical equipment.

Compared with the current detection of electrical equipment is generally in two-dimensional or one-dimensional level, the 3D images collected in this step can effectively avoid the influence of some adverse environmental factors, so as to have a more intuitive understanding of the actual situation of the equipment and improve the Accuracy of fault identification.

Step 102, inputting the 3D image into the target fault identification model, performing fault identification on the target electrical equipment, and obtaining the identification result; wherein, the target fault identification model is constructed based on the 3D U-NET full convolutional neural network algorithm;

In this step, the 3D image is input into the target fault identification model to realize the identification of the fault of the target power equipment, realize the automatic identification of the target power fault, and reduce the dependence on human factors. The specific identification process will be in the follow-up Be explained.

Step 103, generating a troubleshooting strategy according to the identification result and the electronic file of the target electric device;

Here, it should be noted that the electronic files are constantly updated with the use of the target electric equipment. The original electronic files should include the equipment model, manufacturer, production date, operation manual, installation video, original 3D image, etc. of the target electric equipment. .

Step 104, according to the troubleshooting strategy, send the troubleshooting operation guide to the equipment maintenance system.

In this step, the troubleshooting operation guide is sent to the equipment maintenance system, so that the equipment maintenance system can troubleshoot and repair the target electrical equipment based on the troubleshooting operation guide, thereby realizing the use of the troubleshooting operation guide to guide the failure of the target electrical equipment Troubleshoot and repair.

In the power equipment maintenance method of the embodiment of the present application, firstly, the 3D image of the target power equipment is obtained to accurately display the real situation of the power equipment; secondly, the 3D image is input into the target fault identification model, and the fault of the target power equipment is detected. Identify and obtain identification results, wherein, the target fault identification model is constructed based on the 3D U-NET full convolution neural network algorithm; the intelligent identification of power equipment faults is realized, and the dependence on human factors is reduced; again according to the identification Based on the results and the electronic files of the target power equipment, a troubleshooting strategy is generated; the troubleshooting strategy is automatically generated based on historical data; finally, according to the troubleshooting strategy, a troubleshooting operation guide is sent to the equipment maintenance system. In this way, the problems of low maintenance efficiency and poor maintenance effect of electric equipment in the prior art are solved.

Further, as an optional implementation, the method also includes:

In the case of failure of the target electric device, displaying the 3D image through a remote terminal device;

Here, it should be noted that the process of judging whether the power equipment is faulty can be: the power equipment maintenance system performs prediction and judgment based on the real-time 3D image of the power equipment, or, according to the operation and maintenance personnel confirming that the target power equipment is faulty, send the power equipment maintenance The system directly or indirectly inputs the information that characterizes the fault of the target power equipment to judge.

receiving an analysis result determined from the 3D image displayed on the remote terminal device;

Here, it should be noted that the analysis result can be the result of the technical experts remotely viewing the actual situation of the target power equipment according to the 3D image displayed on the remote terminal device, and analyzing the faults. In this way, the technical experts can participate in fault identification. process, improving the accuracy of fault identification. Among them, technical experts can check the 3D image of the target power equipment through virtual reality (Virtual Reality, VR) technology or 3D projection.

On this basis, step 104, according to the troubleshooting strategy, send the troubleshooting operation guide to the equipment maintenance system, including:

generating the troubleshooting operation guide according to the troubleshooting strategy and the analysis results;

Sending the troubleshooting operation guide to the equipment maintenance system.

In this optional implementation method, based on the identification results of the target fault identification model and the automatic generation of troubleshooting strategies from the electronic files of the target power equipment, as well as the remote analysis results of technical experts, a troubleshooting operation guide is generated. Experts give guidance based on experience. This optional implementation method makes full use of big data analysis, realizes intelligent analysis of equipment failures, reduces the dependence on human factors, and at the same time avoids the identification results generated only by intelligent analysis. Troubleshooting operation guide has the problem of giving wrong troubleshooting operation guide due to inaccurate recognition results caused by intelligent analysis errors, which improves the accuracy of troubleshooting operation guide.

As an optional implementation, in step 103, generate a troubleshooting strategy according to the identification result and the electronic file of the target electric device, including:

According to the 3D image, retrieve the electronic file of the target electric device from the device electronic file library;

Here, it should be noted that the 3D image will carry information related to the device, such as the serial number pasted on the target electric device, etc. Therefore, this step can automatically identify the device information based on the 3D image, so that the device information can be obtained from the device electronically based on the device information. The electronic archives of the target electrical equipment are retrieved from the archives.

generating the troubleshooting strategy according to the electronic file and the identification result;

Wherein, the electronic archive includes at least one of parameter information, historical operation data, historical troubleshooting data, historical 3D images, troubleshooting operation guide, maintenance video information and/or maintenance record information of the target electronic device.

Here, it should be noted that the information in the electronic file is continuously updated along with the use of the equipment. For example, if the equipment breaks down and the maintenance personnel repair the fault, all the data in the entire processing process will be saved in the electronic file. , in order to continuously increase the data information used for fault identification and processing, and finally realize the automatic identification and maintenance of faults.

In other words, the process of this optional implementation is: firstly, automatically identify the equipment information in the 3D image, secondly, obtain the electronic file of the equipment based on the equipment information, and thirdly, determine the fault according to the identification result and the information in the electronic file exclusion strategy.

As an optional implementation, step 102, input the 3D image into the target fault identification model, perform fault identification on the target electrical equipment, and obtain identification results, including:

In the case that the target electric equipment is predicted to be faulty according to the monitored operating state data of the target electric equipment, or when an indication that the target electric equipment is faulty is received, the 3D image is input to the target electric equipment. The target fault identification model is used to perform fault identification on the target electric equipment to obtain identification results.

That is to say, in this optional implementation mode, when it is determined that the target power equipment has a fault, the 3D image will be input into the target fault identification model for fault identification, wherein the power equipment maintenance system can be based on the target power The operating state of the equipment combines the parameter information and historical operation data of the target electric equipment to predict whether there is a fault in the target electric equipment, or the electric equipment maintenance system can determine whether there is a fault in the electric equipment based on the received indication information, wherein the indication information can be Information entered directly or indirectly by the user.

The following describes the specific implementation process of this optional implementation mode:

(1) Down-sampling process: 3D video image data is input to the target fault recognition model, after passing through two 3D convolutional layers, extracting different features, continuously passing through the convolutional layers and using the activation function Relu function, and passing through the 2×2×2 maximum The pooling layer continuously performs downsampling to extract features.

(2) Upsampling process: Deconvolution operation is performed to restore and decode the "abstract" fault features to the size of the original data.

(3) Connection process: Concatenate (connection) is a very important part of the network model structure design. The features are combined, the features extracted by multiple convolutional feature extraction frameworks are fused, or the information of the output layer is integrated. Through the 3D U-Net fully convolutional neural network model training, combined with training at different levels, captures fault features at different levels, so as to achieve accurate description of power equipment faults.

(4) The output layer is a 1×1×1 3D convolutional layer, using the sigmoid activation function to map the power equipment failure probability value to the interval [0,1]. The size of the output data and the input data are the same, both are 128×128×128, and the faults of the electric equipment can be identified.

Here, it also needs to be explained that the target fault identification model can be verified before the fault identification of the electric equipment. If the verification fails, the target fault identification model can be trained again to improve the accuracy of fault identification. The training process of the model will be described in detail later.

Further, as an optional implementation, the method also includes:

receiving maintenance video information and/or maintenance record information of the target electrical equipment;

saving the maintenance video information and/or the maintenance record information in the electronic file of the target electrical equipment.

Here, it needs to be explained that during the maintenance process of the target electrical equipment, the maintenance personnel will wear a real-time video camera to record the whole process of on-site maintenance. After the maintenance personnel complete the maintenance, an actual maintenance report will be generated. Record the information, so that the video recording and maintenance records are sent to the electric equipment maintenance system to update the corresponding information in the electronic file.

Here, it also needs to be explained that technical experts can also remotely guide the maintenance process based on the real-time video camera worn by the maintenance personnel. If other conditions occur on site, they can guide maintenance based on the actual situation on site until the maintenance of the target power equipment is completed. .

Furthermore, as an optional implementation, the method also includes:

Receiving and storing updated troubleshooting instructions during troubleshooting and repair of the target electrical device.

Here, what needs to be explained is that this optional implementation method is aimed at the situation of emergencies in the maintenance process. Technical experts have given maintenance suggestions for emergencies. After the maintenance is completed, the maintenance personnel need to submit a A copy of the actual operation record, to correct the errors in the troubleshooting operation guide generated before the maintenance in a timely manner, to ensure that the information stored in the electronic file is true and reliable, so as to improve the accuracy of the basis for subsequent troubleshooting and maintenance.

As an optional implementation, the method also includes:

saving the 3D image to an electronic file of the target electric device.

In this optional implementation mode, after the 3D image collected by the video capture device is obtained, the 3D image is saved in the electronic file of the target electric device, so as to increase the information in the electronic file and increase the judgment basis for subsequent fault targeting and maintenance , and finally realize intelligent fault identification and maintenance of power equipment based on big data technology, reducing dependence on human factors.

As an optional implementation, the method further includes: training the fault identification model based on the following steps, so as to construct the target fault identification model:

Based on the 3D U-NET full convolutional neural network and the troubleshooting data in the electronic archives of the target power equipment, construct a fault identification model corresponding to the type of the target power equipment;

Input the collected 3D image sample set into the fault identification model, and adjust the parameters of the fault identification model;

When the prediction result of the fault identification model does not meet the expected condition, return to the step of adjusting the parameters of the fault identification model until the prediction result meets the expected condition, and obtain the target fault identification model.

That is to say, the training process of this optional implementation is as follows:

Step 1: Collect 3D video image data, which can also be troubleshooting data in electronic files;

Step 2: Build a 3D U-Net fully convolutional neural network model;

The third step: parameter selection and optimization analysis;

The optimization function uses the Adam algorithm commonly used in deep learning, the loss function uses the binary cross entropy loss function (BCE Loss), the activation function uses sigmoid, and the callback function uses the Dice function to implement.

The Adam algorithm uses the first-order moment estimation and second-order moment estimation of the gradient to dynamically adjust the learning rate of each parameter, so that the learning rate has a fixed-range step size every time it is updated, so that the parameters remain stable when updated. Calculated as follows:

ε＝0.001，σ＝10^-7。Among them, generally,

ε=0.001, σ=10 ⁻⁷ .

The calculation formula of the binary cross entropy loss function (BCE Loss):

BCELoss(y,y')=-y×log(y')-(1-y)×log(1-y')

In the above formula, y is the label value, that is, the real value, and y' is the predicted value calculated by the output layer of the convolutional neural network.

The Dice function is a set similarity measurement function, the formula is:

In the formula, N is the total number of sampling points of the collection device, w is the set weight parameter, μ=1.0 prevents the denominator from being 0 and reduces overfitting, y is the label value, that is, the real value, and y' is the calculated predicted value. When the value of the loss function decreases during training, it means that the error between the predicted result and the actual marked fault point and non-fault point has been shrinking until the loss function DLoss(y, y') converges to 0.001.

Step 4: Input the video image data into the 3D U-Net fully convolutional neural network model for model testing.

Step 5: Judge the test effect of the model. If it meets the expectation, the model training is completed. If it does not meet the expectation, return to the third step.

Here, it should be noted that the initial implementation process of the power equipment maintenance method in the embodiment of the present application is: by collecting video of the power equipment, building a three-dimensional model of the power equipment, and transmitting the collected video data of the power equipment to experts , and displayed in the form of VR or three-dimensional projection. Experts judge faults based on experience and give suggestions. The system uses big data, artificial intelligence algorithms and other technologies to identify faults and give corresponding troubleshooting instructions to guide maintenance. Workers perform repairs on electrical equipment. In the case of using the power equipment maintenance method of the embodiment of the present application to carry out multiple maintenances to accumulate enough data, the 3D U-Net full convolution neural network algorithm is used to intelligently carry out a large amount of troubleshooting data entered in the database. Through chemical analysis, build a target fault identification model, identify equipment faults, and automatically formulate troubleshooting strategies. At the same time, by monitoring the operating status of the equipment, the operating status data of the equipment can be obtained, including the environmental data of the equipment, the color change of the equipment, the temperature of the equipment, the location of the equipment failure, etc., combined with some expert opinions, and constantly corrected Model parameters are used to predict equipment, discover equipment failure points in time, and eliminate them as soon as possible. In order to achieve the final realization: video and image acquisition equipment can directly collect the situation of power equipment, and the situation of the equipment can be transmitted to the maintenance system of power equipment, so that the operating status of the equipment and the degree of aging of the equipment can be judged. When the equipment fails, the system automatically analyzes the cause of the failure, forms a failure report, formulates a troubleshooting strategy, and gives instructions to the robot to repair the equipment and eliminate the failure. At the same time, it can predict the occurrence of equipment failure and replace the equipment in time.

In the power equipment maintenance method of the embodiment of the present application, in the early stage, the technical experts participate in the intelligent judgment of the power equipment maintenance system to overhaul the power equipment, and save the data used in the entire identification and maintenance process in the electronic file, so as to Realize the continuous improvement and learning of the power equipment maintenance system. With the increase in the amount of data collected on equipment failures, the Suzaku degree of the power equipment maintenance system will continue to increase, which will provide assistance for the intelligent and digital transformation of future power equipment maintenance.

Below, in conjunction with specific examples, the process of the power equipment maintenance method in the embodiment of the present application will be described:

Install video and other image acquisition equipment in places with a lot of power equipment such as hydropower stations, substations, and power distribution rooms. Create an electronic file for each power equipment, which includes the 3D model of the power equipment, equipment model, manufacturer, production date, operation manual, installation video, etc. Construct a power equipment database, and store all equipment files in the database, which is convenient for system retrieval and update.

First of all, the image acquisition equipment monitors the power equipment at all times and captures the images of the power equipment at any time. When the power equipment fails, the image acquisition equipment scans the 3D image of the equipment and quickly transmits the formed 3D image to experts in other places , displayed in front of experts in the form of 3D projection, and at the same time compare the 3D image of the equipment with the equipment in the power equipment database to find the electronic files of the same type of power equipment. Experts can retrieve electronic files in the power database through remote terminal equipment, and analyze them with 3D projections, videos, etc. of the equipment. At the same time, the fault intelligent analysis system (power equipment maintenance system) uses the convolutional neural network model to analyze equipment faults. Combined with the advice of experts, a strategy for equipment troubleshooting is formulated, and the strategy lists as much as possible a step-by-step process operation. Pass the troubleshooting strategy to the equipment maintenance platform, and the equipment maintenance platform will form a work order and dispatch it to the corresponding maintenance worker. The maintenance worker's handheld terminal will receive the work order information. Process operation, maintenance of power equipment, in the maintenance project, maintenance workers wear video acquisition equipment, record the entire maintenance process, and upload the video to the electronic archive of power equipment, and form a record in the system.

Through 3D modeling, the fault situation of electric equipment is compared and analyzed with the data in the electronic archives, and the convolutional neural network is used to diagnose and analyze the faults of electric equipment, and the characteristic points of equipment faults are extracted, and experts analyze this Fault analysis of faults, equipment troubleshooting videos and other data are collected. When the collected power equipment data is sufficient, using the fault intelligent analysis system based on the convolutional neural network algorithm, it will be possible to quickly analyze the faults of the power equipment and arrange maintenance workers to deal with them. With the development of robot technology, it will be possible to directly Assigning to robots for maintenance can realize unmanned and intelligent in substations, hydropower stations and other places.

The specific implementation process is as follows:

(1) Use cameras and other video and image acquisition terminals to collect the conditions of power equipment in substations, hydropower stations and other places.

(2) Classify and store the collected video, image and other data into the equipment electronic archives, and at the same time, the 3D modeling system constructs the 3D model of the equipment from the collected equipment information.

Among them, the main construction process of the 3D model is as follows. First, through the collection terminal equipment equipped with 3D scanning technology, it is deployed around the power equipment, so that the whole picture of the equipment can be collected, and various points on the equipment can be collected. to the coordinate data (x, y, z) of the equipment; then, different coordinates are registered and spliced, and the data in different coordinate systems are transformed into the same coordinate system; then the transformed coordinate system data are trimmed to retain the established The data required by the model, and then encapsulate the data, use 3D modeling software to generate a 3D model, and repair and simplify the model; finally, render and process the model, and output the model.

(3) When the power equipment fails, the system will automatically identify the collected equipment information, enter the electronic archives of the equipment to retrieve the electronic archives of the equipment, compare the collected information with the archived information of the equipment, and find out the factory information of the equipment. The current data situation and the historical operation data of the equipment. At the same time, experts can check the condition of the equipment through the remote terminal equipment, through VR technology or 3D projection.

(4) Input the collected video and image data into the intelligent fault identification system (or power equipment maintenance system). The intelligent fault identification system builds a fault identification model based on the 3D U-Net full convolution neural network algorithm, and the fault identification;

Among them, the specific process of the model is:

First, down-sampling is performed, and the collected 3D video image data is input into the model. After passing through two 3D convolutional layers, different features are extracted, and the data is then passed through the convolutional layer. The convolutional layer has 8 3×3×3 The convolution kernel, and use the activation function Relu function, then go through the 2×2×2 maximum pooling layer downsampling, and then go through two 3D convolution layers with 16 3×3×3 convolution kernels, and then go through 2 ×2×2 maximum pooling layer downsampling, then two 3D convolutional layers with 32 3×3×3 convolution kernels, then 2×2×2 maximum pooling layer downsampling, and then Two 3D convolution layers with 256 3×3×3 convolution kernels, and then down-sampled by 2×2×2 maximum pooling layer;

Secondly, the bilinear interpolation method is used for the upsampling process, that is, the four real pixel values around the target point in the original image are used to jointly determine a pixel value in the target image, and the original image is restored to the original image by deconvolution step by step. the size of the image;

Again, the connection process is performed, the features are combined, the features extracted by multiple convolutional feature extraction frameworks are fused, or the information of the output layer is integrated, through the training of the 3D U-Net full convolutional neural network model, combined with different Hierarchical training captures fault features at different levels, so as to realize accurate identification of power equipment faults.

Finally, the output layer is a 1×1×1 3D convolutional layer, using the sigmoid activation function to map the power equipment failure probability value to the interval [0,1]. The output data is the same size as the input data, both are 128×128×128. Identify faults in electrical equipment.

At the same time, according to the equipment information viewed, combined with the results given by the fault intelligent analysis system, the technical experts make comprehensive considerations, put forward analysis opinions, formulate troubleshooting strategies, and upload the troubleshooting strategies to the system to form a complete set of troubleshooting Operation guide.

(5) The system automatically transmits equipment maintenance instructions and troubleshooting operation guidelines to the equipment maintenance system module. The equipment maintenance system module will communicate with the worker system, notify the workers to perform maintenance, and perform maintenance operations according to the maintenance operation guidelines during the maintenance process. During the maintenance process, workers wear cameras with real-time video recordings. Experts can guide them remotely at any time, and make video recordings of the entire on-site maintenance process. If other situations occur on site, maintenance will be carried out according to the actual situation on site until the power equipment maintenance is completed. .

(6) After the worker repairs, the worker refers to the expert's advice, resubmits an actual operation record, and saves the repair video. The operation guide and operation records should be continuously updated, and any errors in the operation guide should be corrected in time.

In the power equipment maintenance method of the embodiment of the present application, on the one hand, in the existing technical solutions, the detection of the power equipment is generally in the two-dimensional or one-dimensional level, while the present application analyzes the three-dimensional status of the power equipment, It can more accurately show the real situation of the equipment, avoid the influence of some unfavorable environmental factors, and have a more intuitive understanding of the actual situation of the equipment; both, for experts and technicians who cannot understand the situation on site, there is a more visual part This way of busy line is beneficial to formulate equipment maintenance strategies and guide maintenance personnel to carry out maintenance, and can record the actual operation of maintenance, which is convenient for future study, and also provides reference materials for robot operation; the third, through the use of 3D U- Net full convolutional neural network algorithm performs target detection and fault identification on the faults of power equipment, which can judge the fault situation more accurately, facilitate technical experts to understand the situation of power equipment remotely, and guide on-site maintenance personnel to work, which can help enterprises to digitize and wirelessly Human transformation. Fourth, in the existing technical solutions, only the faulty equipment is generally detected, and no electronic archives are established for the history of equipment failures. Only experienced experts and technical personnel are used to judge equipment failures. Apply to make full use of the failure cases of power equipment, and use emerging technologies such as big data analysis to realize intelligent analysis of power equipment failures. In the early stage, it can provide some references for experts and technicians. Intelligently identify the faults that occur, formulate a feasible maintenance plan, and finally arrange the robot to repair the equipment.

As shown in Figure 2, the embodiment of the present application also provides a power equipment maintenance device, including:

The first acquisition module 201 is configured to acquire a 3D image of the target electric device;

The identification module 202 is used to input the 3D image into the target fault identification model, perform fault identification on the target electrical equipment, and obtain identification results; wherein, the target fault identification model is based on 3D U-NET full convolutional neural network Algorithm construction;

A generation module 203, configured to generate a troubleshooting strategy according to the identification result and the electronic file of the target electric device;

The sending module 204 is configured to send the troubleshooting operation guide to the equipment maintenance system according to the troubleshooting strategy.

In the power equipment maintenance device of the embodiment of the present application, firstly, the first acquisition module 201 acquires the 3D image of the target power equipment, realizing the accurate display of the real situation of the power equipment; secondly, the recognition module 202 inputs the 3D image into the target fault recognition model, Perform fault identification on the target electric equipment to obtain identification results, wherein the target fault identification model is constructed based on the 3D U-NET full convolution neural network algorithm; the intelligent identification of electric equipment faults is realized, and human factors are reduced Dependence; re-generating module 203 generates a troubleshooting strategy according to the identification result and the electronic file of the target electric device; realizes automatically generating a troubleshooting strategy based on historical data; finally sending module 204 sends a troubleshooting strategy based on the troubleshooting strategy The equipment maintenance system sends troubleshooting operation guide. In this way, the problems of low maintenance efficiency and poor maintenance effect of electric equipment in the prior art are solved.

Optionally, the device also includes:

A display module, configured to display the 3D image through a remote terminal device when the target electric device fails;

A first receiving module, configured to receive an analysis result determined according to the 3D image displayed on the remote terminal device;

The sending module 204 includes:

A first generating submodule, configured to generate the troubleshooting operation guide according to the troubleshooting strategy and the analysis result;

The sending submodule is configured to send the troubleshooting operation guide to the equipment maintenance system.

Optionally, the generating module 203 includes:

The calling submodule is used to call the electronic file of the target electric device in the device electronic file library according to the 3D image;

The second generation submodule is used to generate the troubleshooting strategy according to the electronic file and the identification result;

Wherein, the electronic archive includes at least one of parameter information, historical operation data, historical troubleshooting data, historical 3D images, troubleshooting operation guide, maintenance video information and/or maintenance record information of the target electronic device.

Optionally, the identification module 202 is specifically configured to:

In the case that the target electric equipment is predicted to be faulty according to the monitored operating state data of the target electric equipment, or when an indication that the target electric equipment is faulty is received, the 3D image is input to the target electric equipment. The target fault identification model is used to perform fault identification on the target electric equipment to obtain identification results.

Optionally, the device also includes:

The second receiving module is configured to receive maintenance video information and/or maintenance record information of the target electrical equipment;

The first saving module is configured to save the maintenance video information and/or the maintenance record information into the electronic file of the target electric equipment.

Optionally, the device also includes:

The third receiving module is configured to receive and store the updated troubleshooting operation guide during troubleshooting and maintenance of the target electrical equipment.

Optionally, the device also includes:

The second saving module is used to save the 3D image into the electronic file of the target electric device.

The device also includes a training module for training the fault identification model based on the following steps to construct the target fault identification model:

Based on the 3D U-NET full convolutional neural network and the troubleshooting data in the electronic archives of the target power equipment, construct a fault identification model corresponding to the type of the target power equipment;

Input the collected 3D image sample set into the fault identification model, and adjust the parameters of the fault identification model;

When the prediction result of the fault identification model does not meet the expected condition, return to the step of adjusting the parameters of the fault identification model until the prediction result meets the expected condition, and obtain the target fault identification model.

As shown in Fig. 3, the embodiment of the present application also provides a power equipment maintenance system, including: a processor 300, a memory 320 and a program stored in the memory 320 and operable on the processor 300, the When the processor 300 executes the program, it implements the various processes of the above-mentioned embodiments of the electric equipment maintenance method, and can achieve the same technical effect. In order to avoid repetition, details are not repeated here.

The transceiver 310 is configured to receive and send data under the control of the processor 300 .

Wherein, in FIG. 3 , the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by the processor 300 and various circuits of the memory represented by the memory 320 are linked together. The bus architecture can also link together various other circuits such as peripherals, voltage regulators, and power management circuits, etc., which are well known in the art and therefore will not be further described herein. The bus interface provides the interface. Transceiver 310 may be a plurality of elements, including a transmitter and a receiver, providing a means for communicating with various other devices over a transmission medium. For different user devices, the user interface 330 may also be an interface capable of connecting externally and internally to required devices, and the connected devices include but not limited to keypads, displays, speakers, microphones, joysticks, and the like.

The processor 300 is responsible for managing the bus architecture and general processing, and the memory 320 can store data used by the processor 300 when performing operations.

The embodiment of the present application also provides a readable storage medium, on which a program is stored, and when the program is executed by the processor, each process of the above-mentioned embodiment of the power equipment maintenance method is realized, and can achieve The same technical effects are not repeated here to avoid repetition. Wherein, the readable storage medium is, for example, a read-only memory (Read-Only Memory, ROM for short), a random access memory (Random Access Memory, RAM for short), a magnetic disk or an optical disk, and the like.

Finally, it should also be noted that in this text, relational terms such as first and second etc. are only used to distinguish one entity or operation from another, and do not necessarily require or imply that these entities or operations, any such actual relationship or order exists. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article, or terminal equipment comprising a set of elements includes not only those elements, but also includes elements not expressly listed. Other elements mentioned above, or also include elements inherent in such process, method, article or equipment. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

The above description is a preferred embodiment of the present application, and it should be pointed out that for those of ordinary skill in the art, some improvements and modifications can also be made without departing from the principles described in the application. These improvements and modifications It should also be regarded as the protection scope of the present application.

Claims (11)

1. A method for overhauling electrical equipment, comprising: Obtain a 3D image of the target electrical equipment; The 3D image is input to the target fault recognition model, and the fault recognition is carried out to the target electrical equipment to obtain the recognition result; wherein, the target fault recognition model is constructed based on the 3D U-NET full convolutional neural network algorithm; generating a troubleshooting strategy according to the identification result and the electronic file of the target electric device; According to the troubleshooting strategy, the troubleshooting operation guide is sent to the equipment maintenance system. 2. The method according to claim 1, characterized in that the method further comprises: In the case of failure of the target electric device, displaying the 3D image through a remote terminal device; receiving an analysis result determined from the 3D image displayed on the remote terminal device; According to the troubleshooting strategy, send troubleshooting operation instructions to the equipment maintenance system, including: generating the troubleshooting operation guide according to the troubleshooting strategy and the analysis results; Sending the troubleshooting operation guide to the equipment maintenance system. 3. The method according to claim 1, wherein generating a troubleshooting strategy according to the identification result and the electronic file of the target electric device includes: According to the 3D image, retrieve the electronic file of the target electric device from the device electronic file library; generating the troubleshooting strategy according to the electronic file and the identification result; Wherein, the electronic archive includes at least one of parameter information, historical operation data, historical troubleshooting data, historical 3D images, troubleshooting operation guide, maintenance video information and/or maintenance record information of the target electronic device. 4. The method according to claim 1, characterized in that, inputting the 3D image into the target fault identification model, performing fault identification on the target electrical equipment, and obtaining identification results, comprising: In the case that the target electric equipment is predicted to be faulty according to the monitored operating state data of the target electric equipment, or when an indication that the target electric equipment is faulty is received, the 3D image is input to the target electric equipment. The target fault identification model is used to perform fault identification on the target electric equipment to obtain identification results. 5. The method according to claim 1, wherein the method further comprises: receiving maintenance video information and/or maintenance record information of the target electrical equipment; saving the maintenance video information and/or the maintenance record information in the electronic file of the target electrical equipment. 6. The method according to claim 5, further comprising: Receiving and storing updated troubleshooting instructions during troubleshooting and repair of the target electrical device. 7. The method according to claim 1, further comprising: saving the 3D image to an electronic file of the target electric device. 8. The method according to claim 1, further comprising: training the fault identification model based on the following steps to build the target fault identification model: Based on the 3D U-NET full convolutional neural network and the troubleshooting data in the electronic archives of the target power equipment, construct a fault identification model corresponding to the type of the target power equipment; Input the collected 3D image sample set into the fault identification model, and adjust the parameters of the fault identification model; When the prediction result of the fault identification model does not meet the expected condition, return to the step of adjusting the parameters of the fault identification model until the prediction result meets the expected condition, and obtain the target fault identification model. 9. A power equipment maintenance device, characterized in that it comprises: The first acquisition module is used to acquire the 3D image of the target electrical equipment; An identification module, configured to input the 3D image into a target fault identification model, perform fault identification on the target electrical equipment, and obtain identification results; wherein, the fault identification model is constructed based on a 3D U-NET full convolutional neural network algorithm ; A generating module, configured to generate a troubleshooting strategy according to the identification result and the electronic file of the target electric device; The sending module is configured to send the troubleshooting operation guide to the equipment maintenance system according to the troubleshooting strategy. 10. A power equipment maintenance system, characterized in that it comprises: a processor, a memory and a program stored on the memory and operable on the processor, when the program is executed by the processor, the The steps of the electric equipment maintenance method described in any one of claims 1 to 8. 11. A readable storage medium, characterized in that a program is stored on the readable storage medium, and when the program is executed by a processor, the power equipment maintenance method according to any one of claims 1 to 8 is realized A step of.

Images