A power equipment status data monitoring system and method based on artificial intelligence

Abstract

The present invention discloses an artificial intelligence-based power equipment status data monitoring system and method, comprising an equipment data acquisition module, a cloud processor, an image data coding feature extraction module, a fault identification module and a fault monitoring center module; the equipment data acquisition module is used to collect real-time image data, operation data, the equipment number and the vegetation distance data around the power equipment; the cloud processor is used to perform corresponding data processing on the real-time image data, the vegetation distance data and the equipment number; the image data coding feature extraction module is used to extract feature parameters of the image data; the equipment fault identification model is used to perform fault analysis on the power equipment; the fault monitoring center receives fault identification results, risk prediction results and comparison results, and can not only diagnose the faults of the power equipment in a timely manner, but also perform risk prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment status data monitoring, and in particular to a power equipment status data monitoring system and method based on artificial intelligence. Background Art

[0002] In recent years, with the development of science and technology, the area of ​​electricity consumption has gradually expanded, and the number of electrical equipment has gradually increased. Once a major failure occurs, it may cause a large-scale power outage, affecting production and life, and even causing major economic losses. The safe operation of power equipment is crucial to ensuring the normal operation of social and economic activities. Therefore, it is very important to monitor the status data of power equipment.

[0003] In the prior art, there are two traditional methods for monitoring power equipment: periodic inspection and fault monitoring. Fault monitoring is based on patrol inspections by operators and maintenance personnel. When abnormal operation of the equipment is detected through sensory organs, the equipment is tested and analyzed to find the cause of the fault. Periodic inspection uses simple small portable detection instruments as a means. Although the above two methods can detect certain problems with power equipment, they cannot detect problems in a timely manner due to manual monitoring, and cannot predict whether the subsequent working status of the power equipment is good. Summary of the invention

[0004] The technical problem solved by the present invention is that the traditional power equipment status data monitoring method cannot timely discover the existence of faults in the power equipment and cannot predict whether the subsequent working status of the power equipment is good.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for monitoring power equipment status data based on artificial intelligence, comprising the following steps:

[0006] Step S1, collecting real-time image data of the real-time operating status of the power equipment, operating data and vegetation distance data around the power equipment, obtaining the device number of the power equipment, and sending the device number, real-time image data, operating data and vegetation distance data around the power equipment to a cloud processor;

[0007] Step S2, the cloud processor reads the device number, performs image denoising on the real-time image data, performs risk prediction on the operation data to obtain a risk prediction result, calculates the distance of vegetation around the power equipment to obtain a comparison result, sends the real-time image data after image denoising to the feature extraction model, and sends the device number, risk prediction result and comparison result after information reading to the fault monitoring center;

[0008] Step S3, retrieving training image data from a device image library and encoding the training image data, constructing a feature extraction model for the encoded training image data, extracting feature parameters of the training image data and sending the feature parameters to a device fault recognition model;

[0009] Step S4, connecting the feature extraction model with the equipment fault identification model, training the equipment fault identification model through the feature parameters corresponding to the SVM classifier, and obtaining a trained equipment fault identification model;

[0010] Step S5, the fault monitoring center receives the fault identification result from the equipment fault identification model, the risk prediction result and the comparison result of the cloud processor, and notifies the maintenance personnel to take corresponding maintenance measures.

[0011] As a preferred solution of the method for monitoring power equipment status data based on artificial intelligence described in the present invention, step S1 specifically includes:

[0012] A monitoring camera is installed on the power equipment to obtain the equipment number of the power equipment. The monitoring camera regularly collects real-time image data of the real-time operation area of ​​the power equipment. The real-time operation area of ​​the power equipment includes the operation environment area, the electrical equipment area, the power component area, the tower area and the distribution network defect area;

[0013] The operating data of the power equipment is collected, wherein the operating data includes voltage and current, the equipment number is a code containing the location data of the power equipment and the initial parameter information of the power equipment, the monitoring camera scans the equipment number, obtains the vegetation distance data through the vegetation distance monitoring unit, and sends the real-time image data, operating data and vegetation distance data to the cloud processor.

[0014] As a preferred solution of the artificial intelligence-based power equipment status data monitoring method described in the present invention, the cloud processor in step S2 performs image denoising on the real-time image data, specifically including:

[0015] The cloud processor performs data segmentation and data enhancement processing on the received image data of the real-time operation of the power equipment to obtain the initially processed image data, deletes abnormal values ​​and supplements missing values ​​in the initially processed image data to obtain the final processed image data, and sends the image data to the feature extraction model;

[0016] The cloud processor calculates and processes the distance of vegetation around the power equipment, specifically including:

[0017] The cloud processor compares the received vegetation distance data with a vegetation distance data threshold pre-stored in the processor, and if the vegetation distance data is greater than the vegetation distance data threshold, sends the vegetation distance data and the location data of the power equipment to a monitoring center.

[0018] As a preferred solution of the artificial intelligence-based power equipment status data monitoring method described in the present invention, the cloud processor in step S2 performs information reading and processing on the equipment number, specifically including:

[0019] The cloud processor receives the device number scanned by the surveillance camera, and reads information of the device number to obtain the power device location data of the device number and the initial parameters of the power device;

[0020] The cloud processor performs prediction processing on the operation data, specifically including:

[0021] Corresponding safe operation data and dangerous condition data are obtained according to the initial parameters of the electric equipment, wherein the safe operation data is the rated current of the electric equipment, and the dangerous condition data is the maximum allowable current of the electric equipment. A risk prediction model is constructed according to the safe operation data and the dangerous condition operation data, and the operation data is input into the risk prediction model to obtain a risk prediction result, and the risk prediction result is sent to the monitoring center, and the location data of the electric equipment is sent to the monitoring center at the same time.

[0022] As a preferred solution of the method for monitoring power equipment status data based on artificial intelligence described in the present invention, step S3 specifically includes:

[0023] Retrieve training image data from a device image library, wherein the training image data includes environment training image data, tower training image data, device training image data, component training image data, and defect training image data, and each of the training image data has a corresponding label. Input the training image data into an encoder for encoding. The encoder has three layers, namely, an input layer, a hidden layer, and an output layer. The training image data is used as the input of the automatic encoder, and the output layer is the feature parameters of the sample image data. The feature parameters are sent to the device fault recognition model.

[0024] As a preferred solution of the method for monitoring power equipment status data based on artificial intelligence described in the present invention, step S4 specifically includes:

[0025] The feature extraction model is connected to the equipment fault identification model, and an SVM classifier is added to the last output layer of the autoencoder. The equipment fault identification model receives the corresponding feature parameters output by the feature extraction model, and inputs the corresponding feature parameters into the corresponding SVM classifier for training to obtain a trained equipment fault identification model. The equipment fault identification model receives the real-time image data from the cloud processor.

[0026] As a preferred solution of the artificial intelligence-based power equipment status data monitoring method described in the present invention, the steps of constructing the risk prediction model are as follows:

[0027] The safe operation data and the dangerous condition data of the power equipment are used as inputs of the risk prediction model, the safe operation data and the dangerous condition data are normalized to obtain risk prediction result parameters, and according to the size of the risk prediction result parameters, the risk prediction model outputs corresponding risk prediction results;

[0028] The safe operation data and the dangerous condition data are normalized, and the mathematical expression thereof is as follows:

[0029] ;

[0030] in, is the dangerous condition data, For the safety operation data, is the operating data, is the risk prediction result parameter;

[0031] When the operation data is less than or equal to the safe operation data, the risk prediction result parameter is less than or equal to 0, and the risk prediction result of the risk prediction model is no risk;

[0032] When the operation data is greater than or equal to the dangerous condition data, the risk prediction result parameter is greater than or equal to 1, and the risk prediction result of the risk prediction model is an emergency risk;

[0033] When the operating data is greater than the safe operating data and less than the dangerous condition data, the risk prediction result parameter is between 0 and 1. When the risk prediction result parameter is between 0 and 0.3, the risk prediction result of the risk prediction model is a first-level risk; when the risk prediction result parameter is between 0.3 and 0.6, the risk prediction result of the risk prediction model is a second-level risk; when the risk prediction result parameter is between 0.6 and 1, the risk prediction result of the risk prediction model is a third-level risk.

[0034] As a preferred solution of the artificial intelligence-based power equipment status data monitoring method described in the present invention, wherein: the dimension of the hidden layer of the autoencoder is smaller than the dimension of the input layer, and a dimensionality reduction operation is performed from the input layer to the hidden layer to describe the real-time image data with a smaller dimension without losing the real-time image data information as much as possible, and the autoencoder is used to compress a large amount of the real-time image data.

[0035] As a preferred solution of the artificial intelligence-based power equipment status data monitoring method described in the present invention, the corresponding SVM classifier includes an operating environment classifier, an electrical equipment classifier, a tower information classifier, a power component classifier and a distribution network defect classifier.

[0036] An artificial intelligence-based power equipment status data monitoring system includes an equipment data acquisition module, a cloud processor, an image data coding feature extraction module, a fault identification module and a fault monitoring center module;

[0037] The equipment data acquisition module is used to collect the real-time image data of the real-time operating status of the power equipment, the operating data, the equipment number and the vegetation distance data around the power equipment, and obtain the equipment number of the power equipment;

[0038] The cloud processor is used to perform corresponding data processing on the real-time image data, the vegetation distance data, and the device number;

[0039] The image data encoding feature extraction module is used to retrieve training image data from a device image library and encode the image data to obtain feature parameters of the image data;

[0040] The equipment fault identification model is used for performing fault analysis on the power equipment;

[0041] The fault monitoring center receives the fault identification result from the equipment fault identification model, the risk prediction result and the comparison result from the cloud processor, and notifies maintenance personnel to take corresponding maintenance measures.

[0042] The beneficial effects of the present invention are as follows: by using the equipment number on the power equipment, the pre-stored safe operation data and dangerous condition data of the power equipment are obtained, and the safe operation data and the dangerous condition data are normalized and calculated to obtain the risk prediction result parameters, so that the risk prediction of the operation data of the power equipment can be performed, and the risk prediction results of the operation data of the power equipment can be obtained. The subsequent operation status of the power equipment is predicted through the risk prediction results of the power equipment, so as to avoid repairing the power equipment only after it fails, causing major economic losses and affecting residents' electricity consumption.

[0043] Real-time image data of the working area of ​​the power equipment is collected by monitoring cameras, and the real-time image data is encoded using an automatic encoder. A large amount of real-time image data and training image data of the equipment image library can be compressed while minimizing data loss. The corresponding SVM classifier is used for fault identification in each working area, which can reduce the storage capacity, thereby greatly improving the data calculation speed of the power equipment status data monitoring system and more accurately identifying the fault type in the working area.

[0044] The vegetation distance data is obtained through the vegetation distance monitoring unit to obtain the distance between the power equipment and the vegetation. If the distance data threshold is exceeded, the vegetation distance data and the location data of the power equipment are sent to the monitoring center to inform the power equipment maintenance personnel of the specific location where the power equipment needs to be cleared of tree obstacles. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A basic flow chart of an artificial intelligence-based power equipment status data monitoring method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0047] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a method for monitoring power equipment status data based on artificial intelligence, comprising:

[0048] Step S1, collecting real-time image data of the real-time operating status of the power equipment, operating data and vegetation distance data around the power equipment, obtaining the device number of the power equipment, and sending the device number, real-time image data, operating data and vegetation distance data around the power equipment to a cloud processor;

[0049] Step S2, the cloud processor reads the device number, performs image denoising on the real-time image data, performs risk prediction on the operation data to obtain a risk prediction result, calculates the distance of vegetation around the power equipment to obtain a comparison result, sends the real-time image data after image denoising to the feature extraction model, and sends the device number, risk prediction result and comparison result after information reading to the fault monitoring center;

[0050] Step S3, retrieving training image data from a device image library and encoding the training image data, constructing a feature extraction model for the encoded training image data, extracting feature parameters of the training image data and sending the feature parameters to a device fault recognition model;

[0051] Step S4, connecting the feature extraction model with the equipment fault identification model, training the equipment fault identification model through the feature parameters corresponding to the SVM classifier, and obtaining a trained equipment fault identification model;

[0052] Step S5, the fault monitoring center receives the fault identification result from the equipment fault identification model, the risk prediction result and the comparison result of the cloud processor, and notifies the maintenance personnel to take corresponding maintenance measures.

[0053] The step S1 specifically includes:

[0054] A monitoring camera is installed on the power equipment to obtain the equipment number of the power equipment. The monitoring camera regularly collects real-time image data of the real-time operation area of ​​the power equipment. The real-time operation area of ​​the power equipment includes the operation environment area, the electrical equipment area, the power component area, the tower area and the distribution network defect area;

[0055] The operating data of the power equipment is collected, wherein the operating data includes voltage and current, the equipment number is a code containing the location data of the power equipment and the initial parameter information of the power equipment, the monitoring camera scans the equipment number, obtains the vegetation distance data through the vegetation distance monitoring unit, and sends the real-time image data, operating data and vegetation distance data to the cloud processor.

[0056] The cloud processor in step S2 performs image denoising on the real-time image data, specifically including:

[0057] The cloud processor performs data segmentation and data enhancement processing on the received real-time image data of the real-time operation of the electric power equipment to obtain the image data after preliminary processing, deletes abnormal values ​​and supplements missing values ​​in the image data after preliminary processing to obtain the final processed image data, and sends the image data to the feature extraction model;

[0058] The cloud processor calculates and processes the distance of vegetation around the power equipment, specifically including:

[0059] The cloud processor compares the received vegetation distance data with a vegetation distance data threshold pre-stored in the processor, and if the vegetation distance data is greater than the vegetation distance data threshold, sends the vegetation distance data and the location data of the power equipment to a monitoring center.

[0060] In this embodiment, the cloud processor performs denoising on the real-time image data, the purpose of which is to remove the noise in the real-time image data, restore the real real-time image data, and minimize the loss of original features in the real-time image data. When the vegetation distance data is greater than the vegetation distance data threshold, the high-voltage line voltage level of the power equipment is different, and the corresponding vegetation distance data threshold is different;

[0061] When the voltage level of the high-voltage line is 35kv, the vegetation distance data threshold is 6.5 meters;

[0062] When the voltage level of the high-voltage line is 110kv, the vegetation distance data threshold is 8.5 meters;

[0063] When the high-voltage line voltage level is 220kv, the vegetation distance data threshold is 11.2 meters.

[0064] The cloud processor in step S2 performs information reading processing on the device number, specifically including:

[0065] The cloud processor receives the device number scanned by the surveillance camera, and reads information of the device number to obtain the power device location data of the device number and the initial parameters of the power device;

[0066] The cloud processor performs prediction processing on the operation data, specifically including:

[0067] Corresponding safe operation data and dangerous condition data are obtained according to the initial parameters of the electric equipment, wherein the safe operation data is the rated current of the electric equipment, and the dangerous condition data is the maximum allowable current of the electric equipment. A risk prediction model is constructed according to the safe operation data and the dangerous condition operation data, and the operation data is input into the risk prediction model to obtain a risk prediction result, and the risk prediction result is sent to the monitoring center, and the location data of the electric equipment is sent to the monitoring center at the same time.

[0068] In this embodiment, the equipment number is interpreted by the cloud processor to obtain the initial parameters and location data information of the power equipment. The cloud processor receives the operating data, inputs the operating data into the risk prediction model to obtain the risk prediction result, and at the same time retrieves the location data of the equipment number and sends it to the monitoring center. The monitoring center can feedback the risk prediction result of the power equipment at a specific location to the staff. When there is a need for immediate exclusion and maintenance by the staff, the specific location and predicted risk result can be known. The staff can adjust the route according to the predicted risk result. If two places need staff to repair at the same time, the predicted risk levels of the two places can be compared, and the staff should be sent first to the place with a higher predicted risk level to repair the power equipment.

[0069] The step S3 specifically includes:

[0070] Retrieve training image data from a device image library, wherein the training image data includes environment training image data, tower training image data, device training image data, component training image data, and defect training image data, and each of the training image data has a corresponding label. Input the training image data into an encoder for encoding. The encoder has three layers, namely, an input layer, a hidden layer, and an output layer. The training image data is used as the input of the automatic encoder, and the output layer is the feature parameters of the sample image data. The feature parameters are sent to the device fault recognition model.

[0071] In this embodiment, the training image data is encoded using an automatic encoder, which can extract the feature parameters of the training image data and avoid the loss of data during the data training process due to the huge amount of training image data in the equipment image library. The huge amount of training image data can ensure better and more comprehensive feature extraction of the training image data of the power equipment.

[0072] The step S4 specifically includes:

[0073] The feature extraction model is connected to the equipment fault identification model, and an SVM classifier is added to the last output layer of the autoencoder. The equipment fault identification model receives the corresponding feature parameters output by the feature extraction model, and inputs the corresponding feature parameters into the corresponding SVM classifier for training to obtain a trained equipment fault identification model. The equipment fault identification model receives the real-time image data from the cloud processor.

[0074] In this embodiment, the last output layer of the automatic encoder outputs feature parameters, and corresponding SVM classifiers are assigned to different working areas of the power equipment. The feature extraction model is trained using the training image data of the equipment image library, and the real-time image data needs to be processed by a cloud processor for image denoising. After the feature extraction model extracts the feature parameters, it can be input into the equipment fault identification model for fault identification, ensuring that the fault identification results are more accurate.

[0075] The steps for constructing the risk prediction model are as follows:

[0076] The safe operation data and the dangerous condition data of the power equipment are used as inputs of the risk prediction model, the safe operation data and the dangerous condition data are normalized to obtain risk prediction result parameters, and according to the size of the risk prediction result parameters, the risk prediction model outputs corresponding risk prediction results;

[0077] The safe operation data and the dangerous condition data are normalized, and the mathematical expression thereof is as follows:

[0078] ;

[0079] in, is the dangerous condition data, For the safety operation data, is the operating data, is the risk prediction result parameter;

[0080] When the operation data is less than or equal to the safe operation data, the risk prediction result parameter is less than or equal to 0, and the risk prediction result of the risk prediction model is no risk;

[0081] When the operation data is greater than or equal to the dangerous condition data, the risk prediction result parameter is greater than or equal to 1, and the risk prediction result of the risk prediction model is an emergency risk;

[0082] When the operating data is greater than the safe operating data and less than the dangerous condition data, the risk prediction result parameter is between 0 and 1. When the risk prediction result parameter is between 0 and 0.3, the risk prediction result of the risk prediction model is a first-level risk; when the risk prediction result parameter is between 0.3 and 0.6, the risk prediction result of the risk prediction model is a second-level risk; when the risk prediction result parameter is between 0.6 and 1, the risk prediction result of the risk prediction model is a third-level risk.

[0083] In this embodiment, the safe operation data is the rated current of the power equipment, and the dangerous condition data is the maximum allowable current of the power equipment. The power equipment can work at the rated current for a long time and can work at the maximum current for a short time. Real-time operation data is collected. The operation data is the magnitude of the current when the power equipment is working. The operation data is input into the risk prediction model, and the rated current and maximum current are used as the minimum and maximum values ​​in the normalization to obtain the risk prediction results of the real-time operation data of the power equipment.

[0084] The dimension of the hidden layer of the autoencoder is smaller than that of the input layer, and a dimensionality reduction operation is performed from the input layer to the hidden layer to describe the real-time image data with a smaller dimension while minimizing the loss of information of the real-time image data. The autoencoder is used to compress a large amount of the real-time image data.

[0085] In this embodiment, when using an automatic encoder to encode real-time image data and training image data, we use an encoder whose hidden layer dimension is smaller than the input layer dimension, which can ensure that the constructed feature extraction model is based on a large amount of training image data, so that the extracted feature parameters can be more accurate, making the recognition accuracy of the subsequent fault recognition model trained using the feature parameters higher.

[0086] The corresponding SVM classifiers include an operating environment classifier, an electrical equipment classifier, a tower information classifier, a power component classifier and a distribution network defect classifier.

[0087] In this embodiment, the fault identification model has a built-in corresponding SVM classifier for different working areas of the power equipment, which improves the fault identification speed of the fault identification model, obtains the fault identification result more quickly, and also improves the accuracy of the fault identification model.

[0088] Embodiment 2 is an embodiment of the present invention, which provides a wireless charging adaptive power regulation system, including a device data acquisition module, a cloud processor, an image data coding feature extraction module, a fault identification module and a fault monitoring center module;

[0089] The equipment data acquisition module is used to collect the real-time image data of the real-time operating status of the power equipment, the operating data, the equipment number and the vegetation distance data around the power equipment, and obtain the equipment number of the power equipment;

[0090] The cloud processor is used to perform corresponding data processing on the real-time image data, the vegetation distance data, and the device number;

[0091] The image data encoding feature extraction module is used to retrieve training image data from a device image library and encode the image data to obtain feature parameters of the image data;

[0092] The equipment fault identification model is used for performing fault analysis on the power equipment;

[0093] The fault monitoring center receives the fault identification result from the equipment fault identification model, the risk prediction result and the comparison result from the cloud processor, and notifies maintenance personnel to take corresponding maintenance measures.

[0094] Through the equipment number on the power equipment, the pre-stored safe operation data and dangerous condition data of the power equipment are obtained, and the safe operation data and dangerous condition data are normalized and calculated to obtain the risk prediction result parameters. The risk prediction of the operation data of the power equipment can be carried out to obtain the risk prediction results of the operation data of the power equipment. The subsequent operation status of the power equipment is predicted through the risk prediction results of the power equipment, so as to avoid repairing the power equipment after it fails, causing major economic losses and affecting residents' electricity consumption.

[0095] Real-time image data of the working area of ​​the power equipment is collected by monitoring cameras, and the real-time image data is encoded using an automatic encoder. A large amount of real-time image data and training image data of the equipment image library can be compressed while minimizing data loss. The corresponding SVM classifier is used for fault identification in each working area, which can reduce the storage capacity, thereby greatly improving the data calculation speed of the power equipment status data monitoring system and more accurately identifying the fault type in the working area.

[0096] The vegetation distance data is obtained through the vegetation distance monitoring unit to obtain the distance between the power equipment and the vegetation. If the distance data threshold is exceeded, the vegetation distance data and the location data of the power equipment are sent to the monitoring center to inform the power equipment maintenance personnel of the specific location where the power equipment needs to be cleared of tree obstacles.

[0097] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. Among them, the storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, magnetic disk or optical disk. These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based power equipment status data monitoring system, characterized in that: It includes equipment data acquisition module, cloud processor, image data coding feature extraction module, fault identification module and fault monitoring center module; The equipment data acquisition module is used to collect real-time image data of the real-time operating status of the power equipment, operating data, equipment number and vegetation distance data around the power equipment, and obtain the equipment number of the power equipment; The cloud processor is used to perform corresponding data processing on the real-time image data, the vegetation distance data, and the device number; The image data encoding feature extraction module is used to retrieve training image data from a device image library and encode the image data to obtain feature parameters of the image data; The equipment fault identification model is used for performing fault analysis on the electric power equipment; The fault monitoring center receives the fault identification result from the equipment fault identification model, the risk prediction result and the comparison result of the cloud processor, and notifies the maintenance personnel to take corresponding maintenance measures; A monitoring camera is installed on the power equipment to obtain the equipment number of the power equipment. The monitoring camera regularly collects real-time image data of the real-time operation area of ​​the power equipment. The real-time operation area of ​​the power equipment includes the operation environment area, the electrical equipment area, the power component area, the tower area and the distribution network defect area; Collecting the operation data of the power equipment, the operation data including voltage and current, the equipment number is a code including the location data of the power equipment and the initial parameter information of the power equipment, the monitoring camera scans the equipment number, obtains the vegetation distance data through the vegetation distance monitoring unit, and sends the real-time image data, operation data and vegetation distance data to the cloud processor; The cloud processor performs image denoising processing on the real-time image data, specifically including: The cloud processor performs data segmentation and data enhancement processing on the received image data of the real-time operation of the power equipment to obtain the initially processed image data, deletes abnormal values ​​and supplements missing values ​​in the initially processed image data to obtain the final processed image data, and sends the image data to the feature extraction model; The cloud processor calculates and processes the distance of vegetation around the power equipment, specifically including: The cloud processor compares the received vegetation distance data with a vegetation distance data threshold pre-stored in the processor, and if the vegetation distance data is greater than the vegetation distance data threshold, sends the vegetation distance data and the location data of the power equipment to a monitoring center; The cloud processor performs information reading processing on the device number specifically including: the cloud processor receives the device number scanned by the surveillance camera, and reads the information of the device number to obtain the power device location data of the device number and the initial parameters of the power device; The cloud processor performs prediction processing on the operation data, specifically including: According to the initial parameters of the electric power equipment, corresponding safe operation data and dangerous condition data are obtained, wherein the safe operation data is the rated current of the electric power equipment, and the dangerous condition data is the maximum allowable current of the electric power equipment; a risk prediction model is constructed according to the safe operation data and the dangerous condition data; the operation data is input into the risk prediction model to obtain a risk prediction result; the risk prediction result is sent to the monitoring center, and the location data of the electric power equipment is sent to the monitoring center at the same time; Retrieving training image data from a device image library, the training image data includes environment training image data, tower training image data, device training image data, component training image data, and defect training image data, and the training image data all have corresponding labels, and inputting the training image data into an encoder for encoding. The encoder has three layers, namely, an input layer, a hidden layer, and an output layer. The training image data is used as the input of the automatic encoder, and the output layer is the feature parameters of the training image data, and the feature parameters are sent to the device fault recognition model; The feature extraction model is connected to the equipment fault recognition model, an SVM classifier is added to the last output layer of the autoencoder, the equipment fault recognition model receives the corresponding feature parameters output by the feature extraction model, inputs the corresponding feature parameters into the corresponding SVM classifier for training, and obtains the trained equipment fault recognition model, and the equipment fault recognition model receives the real-time image data from the cloud processor; The steps for constructing the risk prediction model are as follows: The safe operation data and the dangerous condition data of the power equipment are used as inputs of the risk prediction model, the safe operation data and the dangerous condition data are normalized to obtain risk prediction result parameters, and according to the size of the risk prediction result parameters, the risk prediction model outputs corresponding risk prediction results; The safe operation data and the dangerous condition data are normalized, and the mathematical expression thereof is as follows: ； Among them, x max is the dangerous condition data, x min is the safety operation data, x is the operation data, x new is the risk prediction result parameter; When the operation data is less than or equal to the safe operation data, the risk prediction result parameter is less than or equal to 0, and the risk prediction result of the risk prediction model is no risk; When the operation data is greater than or equal to the dangerous condition data, the risk prediction result parameter is greater than or equal to 1, and the risk prediction result of the risk prediction model is an emergency risk; When the operation data is greater than the safe operation data and less than the dangerous condition data, the risk prediction result parameter is between 0 and 1. When the risk prediction result parameter is between 0 and 0.3, the risk prediction result of the risk prediction model is a first-level risk; when the risk prediction result parameter is between 0.3 and 0.6, the risk prediction result of the risk prediction model is a second-level risk; when the risk prediction result parameter is between 0.6 and 1, the risk prediction result of the risk prediction model is a third-level risk; The dimension of the hidden layer of the autoencoder is smaller than the dimension of the input layer, and a dimensionality reduction operation is performed from the input layer to the hidden layer, so as to describe the real-time image data with a smaller dimension without losing the real-time image data information as much as possible, and the autoencoder is used to compress a large amount of the real-time image data; The corresponding SVM classifiers include an operating environment classifier, an electrical equipment classifier, a tower information classifier, a power component classifier and a distribution network defect classifier.

Images