Fault Diagnosis Method for Primary Electrical Equipment in Thermal Power Plants Based on Infrared Image Recognition

Abstract

The present invention discloses a method for diagnosing faults in primary electrical equipment in thermal power plants based on infrared image recognition. The method comprises the following steps: A. classifying equipment by type and capturing standard infrared images of each type of equipment; B. extracting a feature image set from the standard infrared images of each type of equipment; C. establishing a convolutional neural network model and training the convolutional neural network model using the feature image set; D. capturing a real-time infrared image of the target equipment; if an area in the real-time infrared image shows a surface temperature exceeding an alarm threshold, a fault is directly determined and the fault diagnosis step is terminated; otherwise, the method proceeds to step E; E. extracting a feature image from the real-time infrared image; and F. inputting the feature image into the convolutional neural network model trained in step C to obtain a fault diagnosis result. The present invention can overcome the shortcomings of existing technologies and reduce the image capture requirements for infrared image diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of power plant maintenance and repair, in particular to a method for diagnosing faults of primary electrical equipment in a thermal power plant based on infrared image recognition. Background Art

[0002] Primary electrical equipment is critical to power plants. Failures can impact safe operation at best, and even cause personal injury, loss of life, and significant property damage at worst. Infrared diagnostics, a cutting-edge and effective diagnostic technology, can detect thermal defects in operating equipment while it's energized. It requires no contact, no downtime, no sampling, and no disassembly. It comprehensively reflects the temperature distribution on the equipment surface and is simple, safe, and reliable to operate. It plays a crucial role in improving the reliability of electrical equipment. However, existing infrared image diagnostic methods have high requirements for image capture, and fault diagnosis errors often occur due to image capture errors. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method for diagnosing faults of primary electrical equipment in thermal power plants based on infrared image recognition, which can solve the shortcomings of the existing technology and reduce the requirements of infrared image diagnosis on image shooting.

[0004] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.

[0005] A method for diagnosing faults of primary electrical equipment in a thermal power plant based on infrared image recognition comprises the following steps:

[0006] A. Classify the equipment according to its type and take standard infrared images for each type of equipment;

[0007] B. Extract feature image sets of standard infrared images for each type of equipment;

[0008] C. Establish a convolutional neural network model and train it using the feature image set;

[0009] D. Photograph the target device to obtain a real-time infrared image of the target device. If an area where the device surface temperature exceeds the alarm threshold appears in the real-time infrared image, a fault is directly determined and the fault diagnosis step ends. Otherwise, the process proceeds to step E.

[0010] E. Extracting feature images of real-time infrared images;

[0011] F. Input the feature image into the convolutional neural network model trained in step C to obtain the fault diagnosis result.

[0012] Preferably, in step B, extracting the feature image set includes the following steps:

[0013] B1. Calculate the average temperature of the device surface in the standard infrared image;

[0014] B2. Set a temperature deviation threshold and mark the area where the temperature difference from the average temperature is greater than the temperature deviation threshold;

[0015] B3. Perform wavelet decomposition on the marked area to obtain low-frequency features and high-frequency features, then perform low-frequency reconstruction to obtain a low-frequency reconstructed image, and perform binarization processing on the low-frequency reconstructed image to obtain a low-frequency reconstructed binary image.

[0016] Preferably, in step C, the high-frequency features of the marked area are input into the convolutional neural network model for training according to the device category, and a convolutional neural network model training result is obtained for each device category.

[0017] Preferably, in step E, extracting the characteristic image of the real-time infrared image includes the following steps:

[0018] E1. Calculate the average temperature of the device surface in the real-time infrared image;

[0019] E2. According to the temperature deviation threshold set in step B2, mark the area where the temperature difference from the average temperature is greater than the temperature deviation threshold;

[0020] E3. Perform wavelet decomposition on the marked area to obtain low-frequency features and high-frequency features, then perform low-frequency reconstruction to obtain a low-frequency reconstructed image, and perform binarization processing on the low-frequency reconstructed image to obtain a low-frequency reconstructed binary image.

[0021] Preferably, in step F, the low-frequency reconstructed binary image obtained in step E3 is first compared with the low-frequency reconstructed binary image obtained in step B3 to determine a standard infrared image corresponding to the low-frequency reconstructed binary image obtained in step B3, whose similarity with the low-frequency reconstructed binary image obtained in step E3 exceeds a set threshold; then the convolutional neural network model training result of the device category where the determined standard infrared image is located is imported into the convolutional neural network model; finally, the high-frequency features obtained in step E3 are input into the convolutional neural network model to obtain a fault diagnosis result.

[0022] The beneficial effect brought about by adopting the above technical solution is that the present invention uses convolutional neural networks to identify and diagnose infrared images. In order to reduce the interference of image shooting angle and image clarity on diagnostic accuracy, the present invention performs wavelet decomposition on infrared images, uses low-frequency features to compare image categories, and uses high-frequency features to perform classification training of neural networks. When using convolutional neural networks to diagnose faults on infrared images, first use low-frequency reconstructed binary images to determine the device classification corresponding to the real-time infrared image, and then use the corresponding training results to import the convolutional neural network model, and finally use high-frequency features to perform identification and diagnosis. Since high-frequency features are not affected by the original image shooting parameters, they can effectively improve diagnostic accuracy. At the same time, the category range of infrared images can be quickly delineated by comparing low-frequency reconstructed binary images, making it possible to use different categories of high-frequency features to independently train and identify the convolutional neural network model, avoiding recognition errors caused by similar high-frequency features in different categories of images. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic diagram of a specific embodiment of the present invention. DETAILED DESCRIPTION

[0024] Reference Figure 1 , a specific embodiment of the present invention includes the following steps,

[0025] A. Classify the equipment according to its type and take standard infrared images for each type of equipment;

[0026] B. Extracting the feature image set of the standard infrared image of each type of equipment; extracting the feature image set includes the following steps:

[0027] B1. Calculate the average temperature of the device surface in the standard infrared image;

[0028] B2. Set a temperature deviation threshold and mark the area where the temperature difference from the average temperature is greater than the temperature deviation threshold;

[0029] B3. Perform wavelet decomposition on the marked area to obtain low-frequency features and high-frequency features, then perform low-frequency reconstruction to obtain a low-frequency reconstructed image, and perform binarization processing on the low-frequency reconstructed image to obtain a low-frequency reconstructed binary image;

[0030] C. Establish a convolutional neural network model and train it using the feature image set. Input the high-frequency features of the marked area into the convolutional neural network model according to the device category for training. A convolutional neural network model training result is obtained for each device category.

[0031] D. Photograph the target device to obtain a real-time infrared image of the target device. If an area where the device surface temperature exceeds the alarm threshold appears in the real-time infrared image, a fault is directly determined and the fault diagnosis step ends. Otherwise, the process proceeds to step E.

[0032] E. Extracting the feature image of the real-time infrared image; comprising the following steps,

[0033] E1. Calculate the average temperature of the device surface in the real-time infrared image;

[0034] E2. According to the temperature deviation threshold set in step B2, mark the area where the temperature difference from the average temperature is greater than the temperature deviation threshold;

[0035] E3. Perform wavelet decomposition on the marked area to obtain low-frequency features and high-frequency features, then perform low-frequency reconstruction to obtain a low-frequency reconstructed image, and perform binarization on the low-frequency reconstructed image to obtain a low-frequency reconstructed binary image.

[0036] F. Input the feature image into the convolutional neural network model trained in step C to obtain a fault diagnosis result; first, compare the low-frequency reconstructed binary image obtained in step E3 with the low-frequency reconstructed binary image obtained in step B3, and determine the standard infrared image corresponding to the low-frequency reconstructed binary image obtained in step B3 whose similarity with the low-frequency reconstructed binary image obtained in step E3 exceeds a set threshold; then import the convolutional neural network model training result of the device category where the determined standard infrared image belongs into the convolutional neural network model; finally, input the high-frequency features obtained in step E3 into the convolutional neural network model to obtain a fault diagnosis result.

[0037] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0038] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for diagnosing faults of primary electrical equipment in thermal power plants based on infrared image recognition, characterized by: The following steps are included: A. Classify the equipment according to its type and take standard infrared images for each type of equipment; B. Extract the feature image set of the standard infrared image of each type of equipment, The following steps are included: B1. Calculate the average temperature of the device surface in the standard infrared image; B2. Set a temperature deviation threshold and mark the area where the temperature difference from the average temperature is greater than the temperature deviation threshold; B3. Perform wavelet decomposition on the marked area to obtain low-frequency features and high-frequency features, then perform low-frequency reconstruction to obtain a low-frequency reconstructed image, and perform binarization processing on the low-frequency reconstructed image to obtain a low-frequency reconstructed binary image; C. Establish a convolutional neural network model and train it using the feature image set. Input the high-frequency features of the marked area into the convolutional neural network model according to the device category for training. A convolutional neural network model training result is obtained for each device category. D. Photograph the target device to obtain a real-time infrared image of the target device. If an area where the device surface temperature exceeds the alarm threshold appears in the real-time infrared image, a fault is directly determined and the fault diagnosis step ends. Otherwise, the process proceeds to step E. E. Extracting feature images of real-time infrared images; The following steps are included: E1. Calculate the average temperature of the device surface in the real-time infrared image; E2. Mark the area where the temperature difference from the average temperature is greater than the temperature deviation threshold value set in step B2; E3. Perform wavelet decomposition on the marked area to obtain low-frequency features and high-frequency features, then perform low-frequency reconstruction to obtain a low-frequency reconstructed image, and perform binarization processing on the low-frequency reconstructed image to obtain a low-frequency reconstructed binary image; F. Input the feature image into the convolutional neural network model trained in step C to obtain the fault diagnosis result; Specifically, the method includes: first, comparing the low-frequency reconstructed binary image obtained in step E3 with the low-frequency reconstructed binary image obtained in step B3, and determining a standard infrared image corresponding to the low-frequency reconstructed binary image obtained in step B3, whose similarity with the low-frequency reconstructed binary image obtained in step E3 exceeds a set threshold; then, importing the convolutional neural network model training result of the device category where the determined standard infrared image belongs into the convolutional neural network model; finally, inputting the high-frequency features obtained in step E3 into the convolutional neural network model to obtain a fault diagnosis result.

Images