Enclosed busbar temperature early warning system and method based on machine vision

Abstract

The invention provides an enclosed busbar temperature early warning system and method based on machine vision. The enclosed busbar temperature early warning system comprises a first camera device used for carrying out image acquisition on a first dial thermometer of an isolated-phase enclosed busbar at a CT position; the second camera shooting device is used for carrying out image acquisition on a plurality of second dial thermometers of the main transformer layer isolated-phase enclosed bus; the first camera device and the second camera device are respectively connected with the image processing system through signal lines; and the image processing system is connected with the temperature prediction system. The system adopts a machine vision monitoring method to clearly capture information of the thermometer on the dial plate of the isolated-phase bus conductor and accurately analyze the temperature information on the dial plate, a traditional manual reading mode is replaced, the accuracy and reliability of temperature monitoring are ensured, meanwhile, a scientific trend early warning analysis method is formulated, and the accuracy and reliability of temperature monitoring are improved. And the targets of intelligent observation, operation and maintenance are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring of generator sets in hydropower stations, and in particular to a closed busbar temperature early warning system and method based on machine vision. Background Art

[0002] The phase-separated enclosed busbar is a large current lead-out device for the power station generator. It is generally connected in a metal shell, and the internal conductor is supported on the shell by a supporting insulator. Since the enclosed busbar adopts a metal shell full connection method, its internal conductor part is not visible. Traditionally, the monitoring of large current enclosed busbar equipment often chooses to arrange a dial thermometer on the busbar shell connected to the main transformer and both sides of the CT, and adopts a manual reading method. The manual reading of the conductor temperature dial is prone to the following results: too many thermometers make the calibration personnel tired and lead to inaccurate calibration results; the subjective factors of the calibration personnel affect the reading of the thermometer and lead to inaccurate calibration results; the data of any thermometer cannot be stored in time, etc., resulting in slow calibration speed. Therefore, it is very necessary to study the automatic calibration of glass liquid thermometers. With the advancement of science and technology, especially the continuous development of intelligence and automation, the automatic calibration of glass thermometers by machine vision has been widely valued by many researchers. Summary of the invention

[0003] In order to solve the current technical problems, the main purpose of the present invention is to provide a closed bus temperature warning system and method based on machine vision. This system adopts a machine vision monitoring method to clearly capture the dial thermometer information of the phase-separated closed bus conductor and accurately analyze the temperature information on the dial, replacing the traditional manual reading method, ensuring the accuracy and reliability of temperature monitoring, and at the same time formulating a scientific trend warning analysis method to achieve the goal of intelligent observation and operation and maintenance.

[0004] In order to achieve the above technical features, the object of the present invention is achieved as follows: A closed bus temperature warning system based on machine vision includes a first camera device for capturing an image of a first dial thermometer of a closed bus at a CT; It also includes a second camera device for collecting images of a plurality of second dial thermometers of the main transformer layer separation phase closed busbar; The first camera device and the second camera device are connected to the image processing system via signal lines respectively; The image processing system is connected to the temperature prediction system.

[0005] The first camera device includes a first camera, the first camera is facing the first dial thermometer, the first camera is fixed to the end of the first support beam through the first column and the first camera mounting plate, and the first support beam is fixed to the bottom end of the busbar support beam through the pressing plate; The first dial thermometer is arranged at the position where the phase-isolated closed busbar casing is located at the CT, and the first main busbar conductor is supported inside the phase-isolated closed busbar casing at the CT through a first insulating support structure.

[0006] The second camera device includes a second camera, the second camera is facing the second dial thermometer, the second camera is fixed to the end of the second support beam through the second column and the second camera mounting plate, the second support beam is inserted and fixed on the busbar support double-slot crossbeam, and the busbar support double-slot crossbeam is fixed to the top of the busbar support column; The second dial thermometer is arranged at the position of the phase-separated closed busbar casing at the main transformer, and the interior of the phase-separated closed busbar casing at the main transformer supports the second main busbar conductor through the second insulating support structure.

[0007] The image processing system is used to process the images collected by the first camera and the second camera to obtain the corresponding temperature data of the first dial thermometer and the second dial thermometer.

[0008] The temperature prediction system collects temperature data processed by the image processing system, pre-processes and analyzes the temperature data, and constructs a busbar temperature change prediction model to provide an early warning of the busbar temperature.

[0009] A closed busbar temperature early warning method based on machine vision, the early warning method is implemented by using the closed busbar temperature early warning system based on machine vision, and comprises the following steps: Step 1: for a first dial thermometer at the shell of the phase-isolated closed busbar at the CT, a first camera device is fixedly installed by using the original support beam of the phase-isolated closed busbar at the CT and adding a new support beam, and an image of the first dial thermometer is collected by the first camera device; Step 2: For the second dial thermometer at the shell of the phase-separated closed busbar of the main transformer layer, use the measurement of the original phase-separated closed busbar of the main transformer layer and fix the double-slot aluminum on the beam, and then fix the second camera device, and use the second camera device to collect images of the second dial thermometer; Step 3, develop an image processing system based on visual detection, the first camera device and the second camera device collect images of the mechanical thermometer in the pipeline through the glass window, and then send the collected images to the processor, perform area division, template matching, and temperature fitting according to the algorithm, and then obtain the temperature data of the first dial thermometer and the second dial thermometer, wherein the detection distance is at least 500mm, and the temperature measurement error is less than or equal to 1.0℃; Step 4: The temperature prediction system collects and analyzes a large amount of temperature data, preprocesses and analyzes the temperature data, and builds a busbar temperature change prediction model. It then uses line graphs, scatter graphs, and bar graphs to display the time changes and trends of the temperature data, and then issues an early warning for the busbar temperature.

[0010] The specific processing process of the acquired image data of the first dial thermometer and the second dial thermometer in step 3 is: The pointer position in the image is located using the edge detection method, and the angle between the pointer and the zero point is calculated to obtain the current dial temperature value. Based on this, the pointer is located by using the template matching algorithm, and the thermometer pointer image on the disk is converted into a pointer image in the Cartesian coordinate system through polar coordinate transformation, thereby realizing efficient positioning and recognition of the dial pointer. The dial pointer is located through the offline calibrated parameters. Based on the position of the dial pointer in the image, and combined with the fact that the camera and thermometer always remain in the same position after installation, the X-coordinate value in the image corresponding to -30℃, -20℃...50℃ can be calibrated offline, and the relationship between the horizontal coordinate X and the temperature value can be fitted by the least squares fitting method to obtain the corresponding temperature data.

[0011] The specific process of preprocessing and analyzing the temperature data in step 4 is as follows: First, the least squares polynomial curve is fitted. According to the given temperature data points, it is not required that the curve passes through these points exactly, but that the fitting curve infinitely approximates the data curve formed by the discrete points. By comparing the error T between the previous temperature estimate and the true value, when T is greater than the preset value T=3, the current temperature value is judged to be abnormal, thereby achieving the purpose of temperature warning; Based on this, it is proposed that when the busbar temperature changes slowly and there is no temperature mutation, it is reliable to use the data of adjacent short time series time nodes to estimate the temperature at the next moment; When the temperature change curve shows large fluctuations, the temperature estimation error curve is used to determine whether the current temperature value is abnormal, thereby achieving the purpose of early warning.

[0012] The present invention has the following beneficial effects: 1. The machine vision industrial camera of the present invention needs to be directly below the glass observation window of the bus conductor dial thermometer to ensure that the temperature information on the dial can be clearly captured and accurately analyzed. Therefore, the present invention proposes to use the original support beams of the original phase-closed busbars of the CT layer, fix the newly added support beams to the original support beams through a customized pressure plate, and use bolts to fix the connection between the newly added support beams and the original support beams on the pressure plate. The camera mounting plate is fixed to the end of the newly added support beam, and finally the camera is fixed to the mounting plate. This installation method will not cause damage to the original support beams and other busbars and auxiliary structures. The newly added support beams, pressure plates and other structures are made of aluminum alloy, and the color of the material is the original color of aluminum alloy. The support materials will be shipped to the site for installation after being manufactured in the factory.

[0013] 2. The machine vision industrial camera of the present invention needs the bus conductor dial thermometer directly below the glass observation window to ensure that the temperature information on the dial can be clearly captured and accurately analyzed. Therefore, the present invention proposes to use the crossbeam of the primary phase-isolated closed busbar of the main transformer layer, fix the double-slot aluminum on the crossbeam, and the end is about 50 mm away from the center of the observation window; then fix the aluminum crossarm on 2 double-slot aluminums, fix the thermometer smart camera mounting plate on the aluminum crossarm, and finally fix the thermometer smart camera on the mounting plate.

[0014] 3. The present invention has developed an image processing system based on visual detection, which collects images of mechanical thermometers in pipelines through glass windows, and then sends the collected images to the processor for area division, template matching, and temperature fitting according to the algorithm. The detection distance is about 500mm, and the temperature measurement error is <=1.0℃.

[0015] 4. The present invention collects and analyzes a large amount of temperature data to preprocess and analyze the data. After constructing a busbar change prediction model, it uses line graphs, scatter graphs, bar graphs and other methods to display the time changes and trends of the data, and proposes an early warning idea. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0017] Figure 1 This is the installation diagram of the dial recognition camera bracket at CT of the present invention.

[0018] Figure 2 For the present invention Figure 1 Center A view.

[0019] Figure 3 It is a side view of the main variable layer camera mounting bracket of the present invention.

[0020] Figure 4 This is a front view of the main variable layer camera mounting bracket of the present invention.

[0021] Figure 5 This is a flow chart of the thermometer identification algorithm of the present invention.

[0022] In the figure: the isolated-phase closed busbar casing 1 at the CT, the first main busbar conductor 2, the first dial thermometer 3, the first camera 4, the first column 5, the first camera mounting plate 6, the first support beam 7, the pressure plate 8, the busbar supporting crossbeam 9, the first insulating support structure 10, the isolated-phase closed busbar casing 11 at the main transformer, the second main busbar conductor 12, the second dial thermometer 13, the second camera 14, the second column 15, the second camera mounting plate 16, the second support beam 17, the busbar supporting double-slot crossbeam 18, and the busbar supporting column 19. DETAILED DESCRIPTION

[0023] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0024] Reference Figure 1-5 A closed bus temperature early warning system based on machine vision is characterized by comprising a first camera device for capturing images of a first dial thermometer 3 of a closed bus separated from the phase at the CT; and a second camera device for capturing images of a plurality of second dial thermometers 13 of a closed bus separated from the phase at the main transformer layer; the first camera device and the second camera device are respectively connected to an image processing system through a signal line; and the image processing system is connected to a temperature prediction system. The present invention collects and analyzes a large amount of temperature data to pre-process and analyze the data, and after constructing a bus change prediction model, uses line graphs, scatter graphs, bar graphs and other methods to display the time changes and trends of the data, and proposes an early warning idea.

[0025] Furthermore, the first camera device includes a first camera 4, the first camera 4 is facing the first dial thermometer 3, the first camera 4 is fixed to the end of the first support beam 7 through the first column 5 and the first camera mounting plate 6, and the first support beam 7 is fixed to the bottom end of the busbar support beam 9 through the pressing plate 8; the first dial thermometer 3 is arranged at the position of the phase-separated closed busbar housing 1 at the CT, and the first main busbar conductor 2 is supported inside the phase-separated closed busbar housing 1 at the CT through the first insulating support structure 10. The dial recognition and reading of the first dial thermometer 3 at the phase-separated closed busbar housing 1 at the CT can be realized through the above-mentioned first camera device.

[0026] Furthermore, the second camera device includes a second camera 14, which is directly facing the second dial thermometer 13. The second camera 14 is fixed to the end of the second support beam 17 through the second column 15 and the second camera mounting plate 16. The second support beam 17 is inserted and fixed on the busbar support double-slot crossbeam 18, and the busbar support double-slot crossbeam 18 is fixed on the top of the busbar support column 19. The dial recognition and reading of the second dial thermometer 13 at the main transformer layer separation phase closed busbar can be realized through the above-mentioned second camera device.

[0027] Furthermore, the second dial thermometer 13 is arranged at the position of the isolated-phase enclosed busbar casing 11 at the main transformer, and the interior of the isolated-phase enclosed busbar casing 11 at the main transformer supports the second main busbar conductor 12 through a second insulating support structure.

[0028] Furthermore, the image processing system is used to process the images captured by the first camera 4 and the second camera 14 to obtain the corresponding temperature data of the first dial thermometer 3 and the second dial thermometer 13 .

[0029] Furthermore, the temperature prediction system collects the temperature data processed by the image processing system, pre-processes and analyzes the temperature data, and constructs a busbar temperature change prediction model, thereby issuing an early warning for the busbar temperature.

[0030] Embodiment 2: A closed busbar temperature early warning method based on machine vision, the early warning method is implemented by using the closed busbar temperature early warning system based on machine vision, and comprises the following steps: Step 1: for the first dial thermometer 3 at the shell of the phase-isolated closed busbar at the CT, the original support beam of the phase-isolated closed busbar at the CT is used and a new support beam is added to fix and install a first camera device, and the first dial thermometer 3 is imaged by the first camera device; Step 2: For the second dial thermometer 13 at the shell of the main transformer layer phase-separated closed busbar, use the measurement of the original phase-separated closed busbar of the main transformer layer and fix the double-slot aluminum on the beam, and then fix the second camera device, and use the second camera device to collect images of the second dial thermometer 13; Step 3, develop an image processing system based on visual detection, the first camera device and the second camera device collect images of the mechanical thermometer in the pipeline through the glass window, and then send the collected images to the processor, perform area division, template matching, and temperature fitting according to the algorithm, and then obtain the temperature data of the first dial thermometer 3 and the second dial thermometer 13, wherein the detection distance is at least 500mm, and the temperature measurement error is <=1.0℃; Step 4: The temperature prediction system collects and analyzes a large amount of temperature data, preprocesses and analyzes the temperature data, and builds a busbar temperature change prediction model. It then uses line graphs, scatter graphs, and bar graphs to display the time changes and trends of the temperature data, and then issues an early warning for the busbar temperature.

[0031] Furthermore, the specific processing process of the acquired image data of the first dial thermometer 3 and the second dial thermometer 13 in step 3 is as follows: The pointer position in the image is located using the edge detection method, and the angle between the pointer and the zero point is calculated to obtain the current dial temperature value. Based on this, the pointer is located by using the template matching algorithm, and the thermometer pointer image on the disk is converted into a pointer image in the Cartesian coordinate system through polar coordinate transformation, thereby realizing efficient positioning and recognition of the dial pointer. The dial pointer is located through the offline calibrated parameters. Based on the position of the dial pointer in the image, and combined with the fact that the camera and thermometer always remain in the same position after installation, the X-coordinate value in the image corresponding to -30℃, -20℃...50℃ can be calibrated offline, and the relationship between the horizontal coordinate X and the temperature value can be fitted by the least squares fitting method to obtain the corresponding temperature data.

[0032] Furthermore, the specific process of preprocessing and analyzing the temperature data in step 4 is as follows: First, the least squares polynomial curve is fitted. According to the given temperature data points, it is not required that the curve passes through these points exactly, but that the fitting curve infinitely approximates the data curve formed by the discrete points. By comparing the error T between the previous temperature estimate and the true value, when T is greater than the preset value T=3, the current temperature value is judged to be abnormal, thereby achieving the purpose of temperature warning; Based on this, it is proposed that when the busbar temperature changes slowly and there is no temperature mutation, it is reliable to use the data of adjacent short time series time nodes to estimate the temperature at the next moment; When the temperature change curve shows large fluctuations, the temperature estimation error curve is used to determine whether the current temperature value is abnormal, thereby achieving the purpose of early warning.

[0033] Embodiment 3: The present invention takes the observation of the thermometer of the shell of the generator main circuit closed busbar at the CT and the main transformer layer in the underground powerhouse of a certain built power station as an example to introduce the implementation of the present invention. The generator main circuit closed busbar is a three-phase horizontal arrangement with a phase-to-phase distance of 2.0m.

[0034] Step 1: For the dial thermometer at the shell of the isolated phase closed busbar at the CT, use the original supporting beam of the original isolated phase closed busbar at the CT layer, fix the newly added supporting beam to the original supporting beam through a customized pressure plate, and use bolts to fix the connection between the newly added supporting beam and the original supporting beam on the pressure plate. Fix the camera mounting plate at the end of the newly added supporting beam, and finally fix the camera to the mounting plate.

[0035] Step 2: For the dial thermometer at the shell of the main transformer layer phase-isolated closed busbar, use the crossbeam of the original phase-isolated closed busbar of the main transformer layer to fix the double-slot aluminum on the crossbeam, with the end about 50 mm away from the center of the observation window; then fix the aluminum crossarm on 2 double-slot aluminums, fix the thermometer smart camera mounting plate on the aluminum crossarm, and finally fix the thermometer smart camera on the mounting plate.

[0036] Step 3: Develop an image processing system based on visual detection, collect images of the mechanical thermometer in the pipeline through the glass window, and then send the collected images to the processor. According to the algorithm, the area division, template matching, and temperature fitting are performed. The detection distance is about 500mm, and the temperature measurement error is <=1.0℃. The edge detection method is directly used to locate the pointer position in the image, and the angle between the pointer and the zero point is calculated to obtain the temperature value of the current dial. Affected by ambient light, thermometer dial material, water vapor interference, camera installation position and viewing angle, the traditional contour segmentation method often causes the failure of thermometer pointer positioning, resulting in the failure of the current task. The dial pointer recognition method based on deep learning can effectively overcome the above environmental interference factors, but it requires a large number of training samples. Since it is difficult to obtain the sample image of the thermometer in this project, the method based on deep learning is not advisable for this project. After careful analysis, we use the following template matching method to realize the positioning of the pointer, and convert the thermometer pointer image on the disk into the pointer image in the Cartesian coordinate system through polar coordinate transformation, so as to realize the efficient positioning and recognition of the dial pointer, and realize the recognition of the dial temperature value through the offline calibrated parameters. The principle of the template matching algorithm mainly involves sliding a template image on the target image and calculating the similarity between the template image and the target image area covering it. By comparing these similarities, the position of the template image that best matches in the target image can be determined. The template matching algorithm based on the correlation coefficient is simple, intuitive, and suitable for finding non-deformed targets. Template matching is very useful in many image processing and computer vision tasks, especially when the size and angle of the object in the target image remain unchanged. Save the metric value: For each position where T covers I, save the metric value to the result image matrix (R). Each position (x, y) in R contains the matching metric value. The whitest position represents the highest match. The position framed by the red ellipse is likely to be the maximum value in the result image matrix, so this area is considered to be a match. In addition, OpenCV provides a variety of template matching methods, including square difference matching (TM_SQDIFF), normalized sum of squared differences (TM_SQDIFF_NORMED), cross-correlation (TM_CCORR), normalized cross-correlation (TM_CCORR_NORMED), correlation coefficient, i.e. zero-mean cross-correlation (TM_CCOEFF), and normalized correlation coefficient, i.e. zero-mean normalized cross-correlation (TM_CCOEFF_NORMED). Each of these methods has its own characteristics and is suitable for different application scenarios and needs. Although template matching has a wide range of applications, it also has its limitations, mainly manifested in that it can only move in parallel. If the matching target in the original image is rotated or changes size, the algorithm is invalid. After the above steps, the position of the thermometer pointer in the image can be obtained, because the camera and thermometer always remain in the same position after installation.Therefore, we can offline calibrate the X coordinate values ​​in the image corresponding to -30℃, -20℃...50℃. The relationship between the horizontal coordinate X and the temperature value can be fitted by the least squares fitting method.

[0037] Step 4: Collect and analyze a large amount of temperature data. After preprocessing and analyzing the data and building a busbar change prediction model, use line graphs, scatter graphs, and bar graphs to display the time changes and trends of the data, and propose early warning ideas. First, use the least squares polynomial curve fitting method. According to the given data points, it is not required that the curve passes through these points exactly, but that the fitting curve infinitely approximates the data curve formed by the discrete points. Assume that the busbar temperature change satisfies the Gaussian distribution, the Gaussian distribution parameter mean is 0, and the standard deviation is 5. Take out a section of temperature data with a total of 10 points y = [39.886239.862339.695339.386838.940438.360637.653736.827035.889034.8493]. Assuming that the busbar changes slowly from adjacent moments t to t + n, use the first 9 data for quadratic curve fitting. , the parameters obtained by solving are (a0, a1, a2) = [39.83290.1403-0.0641]. Then the predicted value of the 10th data predV = 35.9001, and the 10th true value trueV = 34.8493, the absolute error between the predicted value and the true value is: 1.0509. Assuming that the busbar temperature change follows a Gaussian distribution, take out a section of temperature data with a total of 10 points y = [36.526337.391138.138838.761739.253139.608039.822539.894239.822539.6080]. Use the first 9 data to perform quadratic curve fitting , the parameters obtained by solving are (a0, a1, a2) = [35.48051.0958-0.0682]. Then the predicted value of the 10th data is predV = 39.8205, and the 10th true value is trueV = 39.6080. The absolute error between the predicted value and the true value is: 0.2125. Figure 5 It can be seen that when the data points are normal, the temperature estimation error of each point is very small, less than 0.1℃, indicating that the data fit is good. Assuming that the busbar temperature change follows a Gaussian distribution, take out a section of temperature data with a total of 10 points y = [36.526337.391138.138838.761739.253139.608039.822520.000039.822539.6080]. Use the first 9 data to perform quadratic curve fitting , the parameters obtained by solving are (a0, a1, a2) = [35.49071.6015-0.2181]. Then the predicted value of the 10th data predV = 32.2375, and the 10th true value trueV = 39.6080, and the absolute error between the predicted value and the true value is: 7.3705. Before obtaining the 10th measurement value, we cannot know the true value of the temperature. But we can estimate the 10th temperature value based on the temperature curve estimated by the first 9 temperature values. By comparing the error T between the previous temperature estimate and the true value, when T is greater than the preset value T=3, it can be judged that the current temperature value is abnormal, thereby achieving the role of temperature warning. Therefore, the present invention proposes that when the busbar temperature change curve is slow and there is no temperature mutation. Using the data of the adjacent short time series (t = 3-9) time nodes to estimate the temperature at the next moment has good reliability. When the temperature change curve fluctuates greatly, the temperature estimation error curve can be used to judge whether the current temperature value is abnormal, thereby achieving the role of early warning.

Claims

1. A closed busbar temperature early warning system based on machine vision, characterized in that: It comprises a first camera device for collecting images of a first dial thermometer (3) of a phase-isolated closed busbar at a CT; It also includes a second camera device for collecting images of a plurality of second dial thermometers (13) of the main transformer layer separation phase closed busbar; The first camera device and the second camera device are connected to the image processing system via signal lines respectively; The image processing system is connected to the temperature prediction system.

2. According to claim 1, a closed busbar temperature early warning system based on machine vision is characterized in that: The first camera device comprises a first camera (4), the first camera (4) facing the first dial thermometer (3), the first camera (4) being fixed to the end of a first support beam (7) via a first column (5) and a first camera mounting plate (6), and the first support beam (7) being fixed to the bottom end of a busbar support beam (9) via a pressing plate (8); The first dial thermometer (3) is arranged at the location of the phase-isolated enclosed busbar housing (1) at the CT, and the first main busbar conductor (2) is supported inside the phase-isolated enclosed busbar housing (1) at the CT via a first insulating support structure (10).

3. According to claim 2, a closed busbar temperature early warning system based on machine vision is characterized in that: The second camera device comprises a second camera (14), the second camera (14) facing the second dial thermometer (13), the second camera (14) being fixed to the end of a second support beam (17) via a second column (15) and a second camera mounting plate (16), the second support beam (17) being inserted and fixed on a busbar support double-slot crossbeam (18), and the busbar support double-slot crossbeam (18) being fixed to the top of a busbar support column (19); The second dial thermometer (13) is arranged at the location of the phase-separated closed busbar casing (11) at the main transformer, and the interior of the phase-separated closed busbar casing (11) at the main transformer supports a second main busbar conductor (12) via a second insulating support structure.

4. According to claim 3, a closed busbar temperature early warning system based on machine vision is characterized in that: The image processing system is used to process images captured by the first camera (4) and the second camera (14) to obtain corresponding temperature data of the first dial thermometer (3) and the second dial thermometer (13).

5. According to claim 4, a closed busbar temperature early warning system based on machine vision is characterized in that: The temperature prediction system collects temperature data processed by the image processing system, pre-processes and analyzes the temperature data, and constructs a busbar temperature change prediction model to provide an early warning of the busbar temperature.

6. A closed busbar temperature early warning method based on machine vision, characterized in that: The early warning method is implemented by a closed busbar temperature early warning system based on machine vision as described in any one of claims 1 to 5, comprising the following steps: Step 1: for a first dial thermometer (3) at the shell of the phase-isolated closed busbar at the CT, a first camera device is fixedly installed by using the original support beam of the phase-isolated closed busbar at the CT and adding a support beam, and an image of the first dial thermometer (3) is collected by the first camera device; Step 2, for the second dial thermometer (13) at the shell of the main transformer layer phase-separated closed busbar, use the measurement of the original phase-separated closed busbar of the main transformer layer and fix the double-slot aluminum on the beam, and then fix the second camera device, and use the second camera device to collect images of the second dial thermometer (13); Step 3, developing an image processing system based on visual detection, wherein the first camera device and the second camera device collect images of the mechanical thermometer in the pipeline through a glass window, and then send the collected images to a processor, perform area division, template matching, and temperature fitting according to an algorithm, and then obtain temperature data of the first dial thermometer (3) and the second dial thermometer (13), wherein the detection distance is at least 500 mm and the temperature measurement error is less than or equal to 1.0°C; Step 4: The temperature prediction system collects and analyzes a large amount of temperature data, preprocesses and analyzes the temperature data, and builds a busbar temperature change prediction model. It then uses line graphs, scatter graphs, and bar graphs to display the time changes and trends of the temperature data, and then issues an early warning for the busbar temperature.

7. The closed busbar temperature early warning method based on machine vision according to claim 6 is characterized in that: The specific process of processing the acquired image data of the first dial thermometer (3) and the second dial thermometer (13) in step 3 is as follows: The pointer position in the image is located using the edge detection method, and the angle between the pointer and the zero point is calculated to obtain the current dial temperature value. Based on this, the pointer is located by using the template matching algorithm, and the thermometer pointer image on the disk is converted into a pointer image in the Cartesian coordinate system through polar coordinate transformation, thereby realizing efficient positioning and recognition of the dial pointer. The dial pointer is located through the offline calibrated parameters. Based on the position of the dial pointer in the image, and combined with the fact that the camera and thermometer always remain in the same position after installation, the X-coordinate value in the image corresponding to -30℃, -20℃...50℃ can be calibrated offline, and the relationship between the horizontal coordinate X and the temperature value can be fitted by the least squares fitting method to obtain the corresponding temperature data.

8. The closed busbar temperature early warning method based on machine vision according to claim 7 is characterized in that: The specific process of preprocessing and analyzing the temperature data in step 4 is as follows: First, the least squares polynomial curve is fitted. According to the given temperature data points, it is not required that the curve passes through these points exactly, but that the fitting curve infinitely approximates the data curve formed by the discrete points. By comparing the error T between the previous temperature estimate and the true value, when T is greater than the preset value T=3, the current temperature value is judged to be abnormal, thereby achieving the purpose of temperature warning; Based on this, it is proposed that when the busbar temperature changes slowly and there is no temperature mutation, it is reliable to use the data of adjacent short time series time nodes to estimate the temperature at the next moment; When the temperature change curve shows large fluctuations, the temperature estimation error curve is used to determine whether the current temperature value is abnormal, thereby achieving the purpose of early warning.

Images