A method and system for monitoring electric power equipment based on image processing

Abstract

The present invention provides a method and system for monitoring electric equipment based on image processing, which belongs to the technical field of electric equipment monitoring. The method steps are as follows: a camera, an infrared camera and a monitoring server are arranged at the electric equipment; images of the normal operation and fault status of the electric equipment are collected by the camera and the infrared camera; the monitoring server processes the collected images to form sample data, trains a fault monitoring neural network model, and arranges the model on the monitoring server; the monitoring server receives the real-time images of the electric equipment collected by the camera and the infrared camera, inputs the processed images into the fault monitoring neural network model for fault identification, and sends them to a remote electric power dispatching center. The present invention realizes automatic identification of electric equipment faults through cameras, infrared cameras and monitoring servers, and does not require operation and maintenance personnel to conduct on-site inspections, thereby improving operation and maintenance efficiency and reducing safety accidents.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power equipment monitoring, and in particular relates to an electric power equipment monitoring method and system based on image processing. Background Art

[0002] With the increase in electricity demand, the number of power equipment is increasing. In order to ensure the safe operation of the power grid, power operation and maintenance is very important. However, the increasing number of power equipment will inevitably increase the workload of inspection personnel. The traditional manual inspection method consumes a lot of manpower and time, and the inspection efficiency is extremely low. Secondly, it may cause missed inspections or false inspections due to human factors, resulting in the failure to ensure the accuracy of power equipment fault detection. Third, the inspection personnel directly contact the power equipment and there is a risk of endangering personal safety due to improper operation. On the other hand, the existing power equipment operation and maintenance is often lagging, that is, the inspection of the faulty power equipment is carried out. Once the fault is detected, it is usually the normal operation of the power grid, and even a power outage occurs. In serious cases, safety accidents may occur, affecting the stability of the power grid operation.

[0003] In summary, the increase in the number of power equipment has made power operation and maintenance increasingly important, and the inspection workload has become increasingly heavy. The traditional manual inspection method has problems such as low inspection efficiency, low accuracy, threats to the safety of inspection personnel, and lags in power grid fault identification, resulting in the failure to ensure the stable operation of the power grid.

[0004] This is a shortcoming of the prior art. Therefore, it is very necessary to provide an image processing-based power equipment monitoring method and system to address the above-mentioned defects in the prior art. Summary of the invention

[0005] In view of the fact that the increase in the number of the above-mentioned power equipment in the prior art has led to the increasing importance of power operation and maintenance, and the increasing workload of inspections, the traditional manual inspection method has the problems of low inspection efficiency, low accuracy, threat to the safety of inspection personnel, and lag in power grid fault identification, resulting in the defect that the stable operation of the power grid cannot be guaranteed. The present invention provides a power equipment monitoring method and system based on image processing to solve the above-mentioned technical problems.

[0006] In a first aspect, the present invention provides a method for monitoring electric power equipment based on image processing, comprising the following steps:

[0007] S1. Pre-install cameras, infrared cameras and monitoring servers at the power equipment;

[0008] S2. Pre-collect images of normal operation and failure of power equipment through cameras and infrared cameras, and provide them to the monitoring server;

[0009] S3. After the monitoring server processes the collected images, sample data is formed, and then the fault monitoring neural network model is trained through the sample data, and the trained fault monitoring neural network model is deployed on the monitoring server;

[0010] S4. The monitoring server receives real-time images of power equipment collected by cameras and infrared cameras, processes the collected real-time images and inputs them into the fault monitoring neural network model for fault identification, and sends the identified fault data to the remote power dispatching center.

[0011] Furthermore, the specific steps of step S1 are as follows:

[0012] S11. Deploy a monitoring server in the target inspection area and establish a network connection between the monitoring server and the remote power dispatching center;

[0013] S12. Determine the number and location of inspection points in the target inspection area, and divide the target area into several sub-areas according to the monitoring range of the camera and infrared camera;

[0014] S13. Deploy a camera and an infrared camera in each sub-area;

[0015] S14. Establish a network connection between the cameras and infrared cameras in each sub-area and the monitoring server.

[0016] Furthermore, the specific steps of step S2 are as follows:

[0017] S21. Pre-collect images of equipment during normal operation of each inspection point in the sub-region through a camera, and infrared images of each inspection point in the sub-region during normal operation through an infrared camera;

[0018] S22. Simulate equipment failure at inspection points, collect equipment images at each inspection point failure in the sub-region through a camera, and collect infrared images at each inspection point failure in the sub-region through an infrared camera;

[0019] S23. Set forward labels for the normal operation of the device image and infrared image, and set reverse labels for the device image and infrared image when the device fails, and record the inspection point name;

[0020] S24. Send the device image and infrared image of the inspection point corresponding to the set label to the monitoring server.

[0021] Furthermore, the specific steps of step S3 are as follows:

[0022] S31. The monitoring server analyzes the infrared images received at each inspection point and records the infrared images of the device at normal temperature and the infrared images of the device at fault temperature;

[0023] S32. The monitoring server performs mean filtering on the received device images of each inspection point, divides them into different device parts, and then extracts features from each device part to form sample data;

[0024] S33. Pre-build a convolutional neural network model on the monitoring server, and then train the convolutional neural network model with sample data of forward labels and reverse labels to obtain the corresponding model parameters and generate a fault monitoring neural network model;

[0025] S34. Deploy the infrared temperature image of the device at normal temperature, the infrared image of the device at fault temperature, and the trained fault monitoring neural network model on the monitoring server.

[0026] Furthermore, the specific steps of step S33 are as follows:

[0027] S331. Set up a deep learning framework on the monitoring server, build a convolutional neural network model, initialize model parameters, and set an activation function;

[0028] S332. Select sample data with the same proportion of forward labels and reverse proportions to train the convolutional neural network model, iterate based on the deep learning framework, and calculate the loss of each iteration through a preset loss function until the loss is less than the set threshold, and then end the iteration;

[0029] S333. Substitute the final model parameters into the convolutional neural network model to obtain the fault monitoring neural network model, test the fault monitoring neural network with the remaining sample data, calculate the accuracy, and end the training when the accuracy meets the requirements. If the accuracy does not meet the requirements, return to step S332 to retrain.

[0030] Furthermore, the specific steps of step S4 are as follows:

[0031] S41. The camera collects real-time device images at each inspection point in the sub-region, and collects real-time infrared images of each inspection point in the sub-region through an infrared camera, and provides real-time device images and infrared images to the monitoring server;

[0032] S42. The monitoring server compares the real-time infrared image with the infrared image of the device at normal temperature and the infrared image of the device at fault temperature respectively;

[0033] If the similarity between the real-time infrared image and the infrared image of the device at normal temperature exceeds the threshold, the device temperature at the corresponding inspection point is determined to be normal;

[0034] If the similarity between the real-time infrared image and the infrared image at the equipment failure temperature exceeds the threshold, it is determined that the temperature of the equipment at the corresponding inspection point is abnormal;

[0035] S43. The monitoring server performs mean filtering on the received real-time device image, divides it into different device parts, extracts features from each part and inputs it into the trained fault detection neural network model;

[0036] S44. The fault detection neural network model predicts the status of the equipment at the inspection point;

[0037] S45. Send the prediction results to a remote power dispatching center, and use the prediction results as labels to annotate the real-time equipment image, and then use the features of the annotated real-time equipment image to update the fault detection neural network model.

[0038] Furthermore, in step S42, before comparing the real-time infrared image with the infrared image at the normal temperature of the device and the infrared image at the fault temperature of the device, preprocessing is performed first. The specific steps of preprocessing are as follows:

[0039] The three infrared images are normalized separately;

[0040] The OpenCV image processing library is used to calculate the normalized histograms of the three infrared images, which are used as the real-time temperature histogram, the normal temperature histogram, and the fault temperature histogram respectively;

[0041] Calculate the dot product and norm of the real-time temperature histogram and the normal temperature histogram, and when the calculation result is 1, determine that the real-time infrared image temperature is the same as the infrared image when the device is at normal temperature, and when the calculation result is -1, determine that the real-time infrared image temperature is different from the infrared image when the device is normal;

[0042] The dot product and norm of the real-time temperature value histogram and the fault temperature histogram are calculated, and when the calculation result is 1, it is determined that the real-time infrared image temperature is the same as the infrared image when the equipment fails, and when the calculation result is -1, it is determined that the real-time infrared image temperature is different from the infrared image when the equipment fails.

[0043] In a second aspect, the present invention provides an electric power equipment monitoring system based on image processing, comprising:

[0044] A monitoring equipment setting module is used to pre-install cameras, infrared cameras and monitoring servers at power equipment;

[0045] The sample data acquisition module is used to collect images of the normal operation and fault status of the power equipment through the camera and infrared camera in advance and provide them to the monitoring server;

[0046] The detection model generation and deployment module is used to enable the monitoring server to process the collected images to form sample data, then train the fault monitoring neural network model through the sample data, and deploy the trained fault monitoring neural network model on the monitoring server;

[0047] The equipment monitoring and model updating module is used to enable the monitoring server to receive real-time images of power equipment collected by cameras and infrared cameras, process the collected real-time images, input them into the fault monitoring neural network model for fault identification, and send the identified fault data to the remote power dispatching center.

[0048] Furthermore, the monitoring device setting module includes:

[0049] A monitoring server deployment unit is used to deploy a monitoring server in a target inspection area and establish a network connection between the monitoring server and a remote power dispatching center;

[0050] The inspection area division unit is used to determine the number and location of inspection points in the target inspection area and divide the target area into several sub-areas according to the monitoring range of the camera and infrared camera;

[0051] An image acquisition device deployment unit is used to deploy a camera and an infrared camera in each sub-area;

[0052] A network connection unit is used to establish a network connection between the cameras and infrared cameras of each sub-area and the monitoring server;

[0053] The sample data collection module includes:

[0054] A normal image acquisition unit is used to collect images of equipment when each inspection point in the sub-area is operating normally in advance through a camera, and to collect infrared images when each inspection point in the sub-area is operating normally through an infrared camera;

[0055] A fault image acquisition unit is used to simulate the equipment failure of the inspection point, collect the equipment image of each inspection point in the sub-area when the inspection point fails through the camera, and collect the infrared image of each inspection point in the sub-area when the inspection point fails through the infrared camera;

[0056] An image annotation unit is used to set forward labels for the device image and infrared image during normal operation, and to set reverse labels for the device image and infrared image during failure, and to record the inspection point name;

[0057] The image acquisition and sending unit is used to send the device image and infrared image of the corresponding inspection point with the set label to the monitoring server.

[0058] Furthermore, the detection model generation and deployment module includes:

[0059] The infrared image distinguishing and recording unit is used to enable the monitoring server to analyze the infrared images received from each inspection point and record the infrared images of the equipment at normal temperature and the infrared images of the equipment at fault temperature;

[0060] The sample data generating unit is used to enable the monitoring server to perform mean filtering on the received equipment images of each inspection point, divide them into different equipment parts, and then extract features from each equipment part to form sample data;

[0061] The model building and training unit is used to build a convolutional neural network model on the monitoring server in advance, and then train the convolutional neural network model through sample data of forward labels and reverse labels to generate a fault monitoring neural network model after obtaining corresponding model parameters;

[0062] A model deployment unit is used to deploy infrared temperature images of the equipment at normal temperature, infrared images of the equipment at fault temperature, and a trained fault monitoring neural network model on a monitoring server;

[0063] The equipment monitoring and model updating module includes:

[0064] The equipment real-time image acquisition unit is used to acquire the real-time equipment images of each inspection point in the sub-area through the camera, and to acquire the real-time infrared images of each inspection point in the sub-area through the infrared camera, and to provide the real-time equipment images and infrared images to the monitoring server;

[0065] The temperature detection unit is used to enable the monitoring server to compare the real-time infrared image with the infrared image of the device at normal temperature and the infrared image of the device at fault temperature respectively;

[0066] The device temperature normal determination unit is used to determine that the device temperature at the corresponding inspection point is normal when the similarity between the real-time infrared image and the infrared image of the device at normal temperature exceeds a threshold;

[0067] The device temperature fault determination unit is used to determine that the temperature of the device at the corresponding inspection point is abnormal when the similarity between the real-time infrared image and the infrared image at the device fault temperature exceeds a threshold;

[0068] The real-time image processing and model input unit is used to enable the monitoring server to perform mean filtering on the received real-time equipment image, divide it into different equipment parts, extract features from each part and input them into the trained fault detection neural network model;

[0069] The equipment status prediction unit is used to predict the equipment status of the inspection point through the fault detection neural network model;

[0070] The model updating unit is used to send the prediction results to the remote power dispatching center, and use the prediction results as labels to annotate the real-time equipment images, and then use the features of the annotated real-time equipment images to update the fault detection neural network model.

[0071] The beneficial effects of the present invention are:

[0072] The image processing-based power equipment monitoring method and system provided by the present invention realize automatic identification of power equipment faults through cameras, infrared cameras and monitoring servers, without the need for operation and maintenance personnel to conduct on-site inspections, thereby improving operation and maintenance efficiency and reducing safety accidents. On the other hand, faults can be identified in advance through temperature monitoring, and early intervention before equipment failure can prevent the occurrence of faults and accidents, thereby ensuring the safe operation of the power grid.

[0073] In addition, the invention has a reliable design principle, a simple structure and a very broad application prospect.

[0074] It can be seen that compared with the prior art, the present invention has outstanding substantive features and significant progress, and the beneficial effects of its implementation are also obvious. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0076] Figure 1 It is a flow chart of an embodiment of the method for monitoring electric power equipment based on image processing of the present invention.

[0077] Figure 2 It is a flow chart of another embodiment of the method for monitoring electric power equipment based on image processing of the present invention.

[0078] Figure 3 It is a schematic diagram of the power equipment monitoring system based on image processing of the present invention. DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0080] See also Figure 1The figure shows a method for monitoring electric power equipment based on image processing in a specific implementation, comprising the following steps:

[0081] S1. Pre-install cameras, infrared cameras and monitoring servers at the power equipment;

[0082] S2. Pre-collect images of normal operation and failure of power equipment through cameras and infrared cameras, and provide them to the monitoring server;

[0083] S3. After the monitoring server processes the collected images, sample data is formed, and then the fault monitoring neural network model is trained through the sample data, and the trained fault monitoring neural network model is deployed on the monitoring server;

[0084] S4. The monitoring server receives real-time images of power equipment collected by cameras and infrared cameras, processes the collected real-time images and inputs them into the fault monitoring neural network model for fault identification, and sends the identified fault data to the remote power dispatching center.

[0085] In this embodiment, Figure 2 As shown, the specific steps of step S1 are as follows:

[0086] S11. Deploy a monitoring server in the target inspection area and establish a network connection between the monitoring server and the remote power dispatching center;

[0087] S12. Determine the number and location of inspection points in the target inspection area, and divide the target area into several sub-areas according to the monitoring range of the camera and infrared camera;

[0088] S13. Deploy a camera and an infrared camera in each sub-area;

[0089] S14. Establish network connection between cameras and infrared cameras in each sub-area and monitoring server;

[0090] The specific steps of step S2 are as follows:

[0091] S21. Pre-collect images of equipment during normal operation of each inspection point in the sub-region through a camera, and infrared images of each inspection point in the sub-region during normal operation through an infrared camera;

[0092] S22. Simulate equipment failure at inspection points, collect equipment images at each inspection point failure in the sub-region through a camera, and collect infrared images at each inspection point failure in the sub-region through an infrared camera;

[0093] S23. Set forward labels for the normal operation of the device image and infrared image, and set reverse labels for the device image and infrared image when the device fails, and record the inspection point name;

[0094] S24. Send the device image and infrared image of the corresponding inspection point of the set label to the monitoring server;

[0095] The specific steps of step S3 are as follows:

[0096] S31. The monitoring server analyzes the infrared images received at each inspection point and records the infrared images of the device at normal temperature and the infrared images of the device at fault temperature;

[0097] S32. The monitoring server performs mean filtering on the received device images of each inspection point, divides them into different device parts, and then extracts features from each device part to form sample data;

[0098] S33. Pre-build a convolutional neural network model on the monitoring server, and then train the convolutional neural network model through sample data of forward labels and reverse labels, and generate a fault monitoring neural network model after obtaining corresponding model parameters; the specific steps of step S33 are as follows:

[0099] S331. Set up a deep learning framework on the monitoring server, build a convolutional neural network model, initialize model parameters, and set an activation function;

[0100] S332. Select sample data with the same proportion of forward labels and reverse proportions to train the convolutional neural network model, iterate based on the deep learning framework, and calculate the loss of each iteration through a preset loss function until the loss is less than the set threshold, and then end the iteration;

[0101] S333. Substitute the final model parameters into the convolutional neural network model to obtain a fault monitoring neural network model, test the fault monitoring neural network with the remaining sample data, calculate the accuracy, and end the training when the accuracy meets the requirements, and return to step S332 to retrain when the accuracy does not meet the requirements;

[0102] S34. The infrared temperature image of the device at normal temperature, the infrared image of the device at fault temperature, and the trained fault monitoring neural network model are placed on the monitoring server;

[0103] The specific steps of step S4 are as follows:

[0104] S41. The camera collects real-time device images at each inspection point in the sub-region, and collects real-time infrared images of each inspection point in the sub-region through an infrared camera, and provides real-time device images and infrared images to the monitoring server;

[0105] S42. The monitoring server compares the real-time infrared image with the infrared image of the device at normal temperature and the infrared image of the device at fault temperature respectively;

[0106] If the similarity between the real-time infrared image and the infrared image of the device at normal temperature exceeds the threshold, the device temperature at the corresponding inspection point is determined to be normal;

[0107] If the similarity between the real-time infrared image and the infrared image at the equipment failure temperature exceeds the threshold, it is determined that the temperature of the equipment at the corresponding inspection point is abnormal;

[0108] In step S42, before comparing the real-time infrared image with the infrared image at the normal temperature of the device and the infrared image at the fault temperature of the device, preprocessing is performed. The specific steps of preprocessing are as follows:

[0109] The three infrared images are normalized separately;

[0110] The OpenCV image processing library is used to calculate the normalized histograms of the three infrared images, which are used as the real-time temperature histogram, the normal temperature histogram, and the fault temperature histogram respectively;

[0111] Calculate the dot product and norm of the real-time temperature histogram and the normal temperature histogram, and when the calculation result is 1, determine that the real-time infrared image temperature is the same as the infrared image when the device is at normal temperature, and when the calculation result is -1, determine that the real-time infrared image temperature is different from the infrared image when the device is normal;

[0112] Calculate the dot product and norm of the real-time temperature value histogram and the fault temperature histogram, and when the calculation result is 1, determine that the real-time infrared image temperature is the same as the infrared image when the device fails, and when the calculation result is -1, determine that the real-time infrared image temperature is different from the infrared image when the device fails;

[0113] S43. The monitoring server performs mean filtering on the received real-time device image, divides it into different device parts, extracts features from each part and inputs it into the trained fault detection neural network model;

[0114] S44. The fault detection neural network model predicts the status of the equipment at the inspection point;

[0115] S45. Send the prediction results to a remote power dispatching center, and use the prediction results as labels to annotate the real-time equipment image, and then use the features of the annotated real-time equipment image to update the fault detection neural network model.

[0116] It should be understood that the order of execution of the steps in the above embodiment does not mean that the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0117] The following is an embodiment of the power equipment monitoring system based on image processing provided by the embodiments of the present disclosure. The device and the power equipment monitoring method based on image processing in the above-mentioned embodiments belong to the same inventive concept. For details not described in detail in the embodiment of the power equipment monitoring system based on image processing, please refer to the embodiment of the power equipment monitoring method based on image processing mentioned above.

[0118] like Figure 3 As shown, the system includes:

[0119] A monitoring equipment setting module is used to pre-install cameras, infrared cameras and monitoring servers at power equipment;

[0120] The sample data acquisition module is used to collect images of the normal operation and fault status of the power equipment through the camera and infrared camera in advance and provide them to the monitoring server;

[0121] The detection model generation and deployment module is used to enable the monitoring server to process the collected images to form sample data, then train the fault monitoring neural network model through the sample data, and deploy the trained fault monitoring neural network model on the monitoring server;

[0122] The equipment monitoring and model updating module is used to enable the monitoring server to receive real-time images of power equipment collected by cameras and infrared cameras, process the collected real-time images, input them into the fault monitoring neural network model for fault identification, and send the identified fault data to the remote power dispatching center.

[0123] In this implementation, the monitoring device setting module includes:

[0124] A monitoring server deployment unit is used to deploy a monitoring server in a target inspection area and establish a network connection between the monitoring server and a remote power dispatching center;

[0125] The inspection area division unit is used to determine the number and location of inspection points in the target inspection area and divide the target area into several sub-areas according to the monitoring range of the camera and infrared camera;

[0126] An image acquisition device deployment unit is used to deploy a camera and an infrared camera in each sub-area;

[0127] A network connection unit is used to establish a network connection between the cameras and infrared cameras of each sub-area and the monitoring server;

[0128] The detection model generation and layout module includes:

[0129] A normal image acquisition unit is used to collect images of equipment when each inspection point in the sub-area is operating normally in advance through a camera, and to collect infrared images when each inspection point in the sub-area is operating normally through an infrared camera;

[0130] A fault image acquisition unit is used to simulate the equipment failure of the inspection point, collect the equipment image of each inspection point in the sub-area when the inspection point fails through the camera, and collect the infrared image of each inspection point in the sub-area when the inspection point fails through the infrared camera;

[0131] An image annotation unit is used to set forward labels for the device image and infrared image during normal operation, and to set reverse labels for the device image and infrared image during failure, and to record the inspection point name;

[0132] The image acquisition and sending unit is used to send the device image and infrared image of the corresponding inspection point where the label is set to the monitoring server;

[0133] The equipment monitoring and model updating module includes:

[0134] The infrared image distinguishing and recording unit is used to enable the monitoring server to analyze the infrared images received from each inspection point and record the infrared images of the equipment at normal temperature and the infrared images of the equipment at fault temperature;

[0135] The sample data generating unit is used to enable the monitoring server to perform mean filtering on the received equipment images of each inspection point, divide them into different equipment parts, and then extract features from each equipment part to form sample data;

[0136] The model building and training unit is used to build a convolutional neural network model on the monitoring server in advance, and then train the convolutional neural network model through sample data of forward labels and reverse labels to generate a fault monitoring neural network model after obtaining corresponding model parameters;

[0137] A model deployment unit is used to deploy infrared temperature images of the equipment at normal temperature, infrared images of the equipment at fault temperature, and a trained fault monitoring neural network model on a monitoring server;

[0138] The equipment monitoring and model updating module includes:

[0139] The equipment real-time image acquisition unit is used to acquire the real-time equipment images of each inspection point in the sub-area through the camera, and to acquire the real-time infrared images of each inspection point in the sub-area through the infrared camera, and to provide the real-time equipment images and infrared images to the monitoring server;

[0140] The temperature detection unit is used to enable the monitoring server to compare the real-time infrared image with the infrared image of the device at normal temperature and the infrared image of the device at fault temperature respectively;

[0141] The device temperature normal determination unit is used to determine that the device temperature at the corresponding inspection point is normal when the similarity between the real-time infrared image and the infrared image of the device at normal temperature exceeds a threshold;

[0142] The device temperature fault determination unit is used to determine that the temperature of the device at the corresponding inspection point is abnormal when the similarity between the real-time infrared image and the infrared image at the device fault temperature exceeds a threshold;

[0143] The real-time image processing and model input unit is used to enable the monitoring server to perform mean filtering on the received real-time equipment image, divide it into different equipment parts, extract features from each part and input them into the trained fault detection neural network model;

[0144] The equipment status prediction unit is used to predict the equipment status of the inspection point through the fault detection neural network model;

[0145] The model updating unit is used to send the prediction results to the remote power dispatching center, and use the prediction results as labels to annotate the real-time equipment images, and then use the features of the annotated real-time equipment images to update the fault detection neural network model.

[0146] Although the present invention has been described in detail with reference to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily think of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention.

Claims

1. A method for monitoring electric power equipment based on image processing, characterized in that: The steps include: S1. Pre-install cameras, infrared cameras and monitoring servers at the power equipment; S2. Pre-collect images of the normal operation and fault status of the power equipment through the camera and infrared camera, and provide them to the monitoring server; The camera collects equipment images, and the infrared camera collects infrared images; S3. After the monitoring server processes the collected images, sample data is formed, and then the fault monitoring neural network model is trained through the sample data, and the trained fault monitoring neural network model is deployed on the monitoring server; the specific steps of step S3 are as follows: S31. The monitoring server analyzes the infrared images received at each inspection point and records the infrared images of the device at normal temperature and the infrared images of the device at fault temperature; S32. The monitoring server performs mean filtering on the received device images of each inspection point, divides them into different device parts, and then extracts features from each device part to form sample data; S33. Pre-build a convolutional neural network model on the monitoring server, and then train the convolutional neural network model through sample data of forward labels and reverse labels, and generate a fault monitoring neural network model after obtaining corresponding model parameters; the specific steps of step S33 are as follows: S331. Set up a deep learning framework on the monitoring server, build a convolutional neural network model, initialize model parameters, and set an activation function; S332. Select sample data of forward labels and reverse labels of the same proportion to train the convolutional neural network model, iterate based on the deep learning framework, and calculate the loss of each iteration through a preset loss function until the loss is less than the set threshold, and then end the iteration; wherein, forward labels are set for the device images and infrared images during normal operation, and reverse labels are set for the device images and infrared images during failure; S333. Substitute the final model parameters into the convolutional neural network model to obtain a fault monitoring neural network model, test the fault monitoring neural network with the remaining sample data, calculate the accuracy, and end the training when the accuracy meets the requirements, and return to step S332 to retrain when the accuracy does not meet the requirements; S34. The infrared temperature image of the device at normal temperature, the infrared image of the device at fault temperature, and the trained fault monitoring neural network model are placed on the monitoring server; S4. The monitoring server receives the real-time images of the power equipment collected by the camera and the infrared camera, processes the collected real-time images and inputs them into the fault monitoring neural network model for fault identification, and sends the identified fault data to the remote power dispatching center; the specific steps of step S4 are as follows: S41. The camera collects real-time device images at each inspection point in the sub-region, and collects real-time infrared images of each inspection point in the sub-region through an infrared camera, and provides real-time device images and infrared images to the monitoring server; S42. The monitoring server compares the real-time infrared image with the infrared image of the device at normal temperature and the infrared image of the device at fault temperature respectively; If the similarity between the real-time infrared image and the infrared image of the device at normal temperature exceeds the threshold, the device temperature at the corresponding inspection point is determined to be normal; If the similarity between the real-time infrared image and the infrared image at the equipment failure temperature exceeds the threshold, it is determined that the temperature of the equipment at the corresponding inspection point is abnormal; S43. The monitoring server performs mean filtering on the received real-time device image, divides it into different device parts, extracts features from each part and inputs it into the trained fault detection neural network model; S44. The fault detection neural network model predicts the status of the equipment at the inspection point; S45. Send the prediction results to the remote power dispatching center, and use the prediction results as labels to annotate the real-time equipment image, and then use the features of the annotated real-time equipment image to update the fault detection neural network model; In step S42, before comparing the real-time infrared image with the infrared image at the normal temperature of the device and the infrared image at the fault temperature of the device, preprocessing is performed. The specific steps of preprocessing are as follows: The three infrared images are normalized separately; The OpenCV image processing library is used to calculate the normalized histograms of the three infrared images, which are used as the real-time temperature histogram, the normal temperature histogram, and the fault temperature histogram respectively; Calculate the dot product and norm of the real-time temperature histogram and the normal temperature histogram, and when the calculation result is 1, determine that the real-time infrared image temperature is the same as the infrared image when the device is at normal temperature, and when the calculation result is -1, determine that the real-time infrared image temperature is different from the infrared image when the device is normal; The dot product and norm of the real-time temperature value histogram and the fault temperature histogram are calculated, and when the calculation result is 1, it is determined that the real-time infrared image temperature is the same as the infrared image when the equipment fails, and when the calculation result is -1, it is determined that the real-time infrared image temperature is different from the infrared image when the equipment fails.

2. The method for monitoring electric power equipment based on image processing according to claim 1, characterized in that: The specific steps of step S1 are as follows: S11. Deploy a monitoring server in the target inspection area and establish a network connection between the monitoring server and the remote power dispatching center; S12. Determine the number and location of inspection points in the target inspection area, and divide the target area into several sub-areas according to the monitoring range of the camera and infrared camera; S13. Deploy a camera and an infrared camera in each sub-area; S14. Establish a network connection between the cameras and infrared cameras in each sub-area and the monitoring server.

3. The method for monitoring electric power equipment based on image processing according to claim 2, characterized in that: The specific steps of step S2 are as follows: S21. Pre-collect images of equipment during normal operation of each inspection point in the sub-region through a camera, and infrared images of each inspection point in the sub-region during normal operation through an infrared camera; S22. Simulate equipment failure at inspection points, collect equipment images at each inspection point failure in the sub-region through a camera, and collect infrared images at each inspection point failure in the sub-region through an infrared camera; S23. Set forward labels for the normal operation of the device image and infrared image, and set reverse labels for the device image and infrared image when the device fails, and record the inspection point name; S24. Send the device image and infrared image of the inspection point corresponding to the set label to the monitoring server.

4. An electric power equipment monitoring system based on image processing, characterized in that: The method for monitoring electric power equipment based on image processing according to any one of claims 1 to 3 comprises: A monitoring equipment setting module is used to pre-install cameras, infrared cameras and monitoring servers at power equipment; The sample data acquisition module is used to collect images of the normal operation and fault status of the power equipment through the camera and infrared camera in advance and provide them to the monitoring server; The detection model generation and deployment module is used to enable the monitoring server to process the collected images to form sample data, then train the fault monitoring neural network model through the sample data, and deploy the trained fault monitoring neural network model on the monitoring server; the detection model generation and deployment module includes: The infrared image distinguishing and recording unit is used to enable the monitoring server to analyze the infrared images received from each inspection point and record the infrared images of the equipment at normal temperature and the infrared images of the equipment at fault temperature; The sample data generating unit is used to enable the monitoring server to perform mean filtering on the received equipment images of each inspection point, divide them into different equipment parts, and then extract features from each equipment part to form sample data; The model building and training unit is used to build a convolutional neural network model on the monitoring server in advance, and then train the convolutional neural network model through sample data of forward labels and reverse labels to generate a fault monitoring neural network model after obtaining corresponding model parameters; A model deployment unit is used to deploy infrared temperature images of the equipment at normal temperature, infrared images of the equipment at fault temperature, and a trained fault monitoring neural network model on a monitoring server; The equipment monitoring and model updating module is used to enable the monitoring server to receive real-time images of power equipment collected by cameras and infrared cameras, process the collected real-time images, input them into the fault monitoring neural network model for fault identification, and send the identified fault data to the remote power dispatching center; the equipment monitoring and model updating module includes: The equipment real-time image acquisition unit is used to acquire the real-time equipment images of each inspection point in the sub-area through the camera, and to acquire the real-time infrared images of each inspection point in the sub-area through the infrared camera, and to provide the real-time equipment images and infrared images to the monitoring server; The temperature detection unit is used to enable the monitoring server to compare the real-time infrared image with the infrared image of the device at normal temperature and the infrared image of the device at fault temperature respectively; The device temperature normal determination unit is used to determine that the device temperature at the corresponding inspection point is normal when the similarity between the real-time infrared image and the infrared image of the device at normal temperature exceeds a threshold; The device temperature fault determination unit is used to determine that the temperature of the device at the corresponding inspection point is abnormal when the similarity between the real-time infrared image and the infrared image at the device fault temperature exceeds a threshold; The real-time image processing and model input unit is used to enable the monitoring server to perform mean filtering on the received real-time equipment image, divide it into different equipment parts, extract features from each part and input them into the trained fault detection neural network model; The equipment status prediction unit is used to predict the equipment status of the inspection point through the fault detection neural network model; The model updating unit is used to send the prediction results to the remote power dispatching center, and use the prediction results as labels to annotate the real-time equipment images, and then use the features of the annotated real-time equipment images to update the fault detection neural network model.

5. The power equipment monitoring system based on image processing according to claim 4, characterized in that: The monitoring device setting module includes: A monitoring server deployment unit is used to deploy a monitoring server in a target inspection area and establish a network connection between the monitoring server and a remote power dispatching center; The inspection area division unit is used to determine the number and location of inspection points in the target inspection area and divide the target area into several sub-areas according to the monitoring range of the camera and infrared camera; An image acquisition device deployment unit is used to deploy a camera and an infrared camera in each sub-area; A network connection unit is used to establish a network connection between the cameras and infrared cameras of each sub-area and the monitoring server; The sample data collection module includes: A normal image acquisition unit is used to collect images of equipment when each inspection point in the sub-area is operating normally in advance through a camera, and to collect infrared images when each inspection point in the sub-area is operating normally through an infrared camera; A fault image acquisition unit is used to simulate the equipment failure of the inspection point, collect the equipment image of each inspection point in the sub-area when the inspection point fails through the camera, and collect the infrared image of each inspection point in the sub-area when the inspection point fails through the infrared camera; An image annotation unit is used to set forward labels for the device image and infrared image during normal operation, and to set reverse labels for the device image and infrared image during failure, and to record the inspection point name; The image acquisition and sending unit is used to send the device image and infrared image of the corresponding inspection point with the set label to the monitoring server.

Images