Transformer substation unmanned aerial vehicle inspection device and analysis method

Abstract

The invention relates to a transformer substation unmanned aerial vehicle inspection device and an analysis method. The inspection device comprises an unmanned aerial vehicle, the top of a body of the unmanned aerial vehicle is fixedly connected with a positioner, the outer side of the body is fixedly connected with wings in the axial direction of the body, the bottoms of the wings are detachably connected with undercarriages, and an inspection camera mechanism is installed at the bottom of the unmanned aerial vehicle; the inspection device further comprises rear-end equipment, the unmanned aerial vehicle communicates with the rear-end equipment, the unmanned aerial vehicle performs inspection according to an inspection path provided by the rear-end equipment, and the rear-end equipment is used for processing images collected by the inspection camera mechanism; and the inspection camera mechanism comprises an infrared camera. The temperature of the surface of the to-be-inspected equipment is detected through the infrared camera, and the temperature distribution condition of the surface of the equipment is visually and quantitatively detected, so that potential faults possibly existing in the equipment can be found in time, preventive maintenance of the equipment is realized, and accidents such as equipment shutdown possibly occurring are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of substation inspection, in particular to a substation unmanned aerial vehicle inspection device and an analysis method. BACKGROUND

[0002] As a key node of the power system, the long-term stable operation of the internal equipment (such as transformers, circuit breakers, disconnectors, and mutual inductors) of the substation is directly related to the safety and reliability of the entire power grid. These devices may develop potential faults under the influence of factors such as long-term high-load operation, environmental erosion, and material aging. If these faults are not discovered and addressed in a timely manner, minor defects may gradually evolve into serious accidents, such as equipment burnout, even causing fires or widespread power outages, resulting in significant economic losses and social impact. Therefore, regular and efficient inspection of substation equipment to promptly identify potential faults such as thermal defects is a crucial aspect of power system operation and maintenance.

[0003] Traditional substation inspection relies mainly on manual methods, i.e., maintenance personnel carrying tools into the station for inspection. This approach has obvious drawbacks: (1) low safety, maintenance personnel need to be in close proximity to high-voltage equipment, posing a risk to personal safety; (2) low efficiency, the inspection range is large and the equipment is numerous, consuming time and effort, and being easily influenced by subjective factors, resulting in a high rate of missed inspections; (3) insufficient data objectivity, it is difficult to inspect equipment at high altitudes or in hidden locations, and it is difficult to obtain comprehensive and quantifiable data.

[0004] In recent years, unmanned aerial vehicle technology has been introduced into the field of substation inspection due to its high flexibility, maneuverability, and wide perspective. Unmanned aerial vehicle inspection overcomes the above-mentioned shortcomings of manual inspection, enabling rapid and safe access to any angle of the equipment for close observation and image capture, significantly improving the efficiency and coverage of the inspection and reducing the safety risks for maintenance personnel. Unmanned aerial vehicle inspection technology for substations mainly focuses on using visible light cameras for high-definition photography or video recording, and identifying external abnormalities of equipment through image analysis.

[0005] However, there is a significant gap in existing technology: it lacks the ability to provide early warnings for equipment faults. Many electrical equipment faults (such as poor contact and internal wear) will have a sustained heating process before developing into visible physical damage or open flames. Relying solely on visible light inspection cannot capture this temperature change, and only when the fault develops to the middle or late stages, causing substantial damage, can it be discovered, missing the best opportunity for maintenance and repair. Therefore, the existing unmanned aerial vehicle inspection scheme has obvious shortcomings in preventive maintenance and early fault diagnosis, failing to fully exploit the potential value of the unmanned aerial vehicle platform. SUMMARY

[0006] The application provides a power substation unmanned aerial vehicle inspection device and an analysis method to solve the above problems.

[0007] The application adopts the technical scheme of a power substation unmanned aerial vehicle inspection device, which comprises an unmanned aerial vehicle, the top of the body of the unmanned aerial vehicle is fixedly connected with a positioner, the outer side of the body is fixedly connected with a wing along the axial direction, the bottom of the wing is detachably connected with a landing gear, and the bottom of the unmanned aerial vehicle is provided with an inspection camera mechanism.

[0008] The inspection device further comprises a backend device, the unmanned aerial vehicle communicates with the backend device, the unmanned aerial vehicle performs inspection according to an inspection path provided by the backend device, and the backend device is used for processing images collected by the inspection camera mechanism.

[0009] The inspection camera mechanism comprises an external red camera.

[0010] Preferably, the inspection camera mechanism further comprises a camera, and the camera and the infrared camera are respectively located at two ends of the bottom of the unmanned aerial vehicle.

[0011] The bottom of the body of the unmanned aerial vehicle is fixedly connected with a rack, the two ends of the rack are spaced apart from the body, the spacing is referred to as an installation interval, the infrared camera is located at the installation interval, the two ends of the infrared camera are fixedly connected with connecting shafts, the ends of the connecting shafts are rotatably connected to the rack, a micro motor is fixedly connected to the outer side wall of the rack, and the output shaft of the micro motor is fixedly connected to the connecting shafts of the infrared camera through a speed reducer.

[0012] Preferably, the wing is connected to the body through a horizontally arranged connecting rod, the landing gear comprises a horizontally arranged connecting rack, the top of the connecting rack is provided with a connecting groove matched with the bottom of the wing, the bottom of the connecting rack is fixedly connected with two spaced apart adjusting plates, and a landing rod is hingedly connected between the two adjusting plates through a hinge shaft.

[0013] The application further provides a power substation unmanned aerial vehicle inspection analysis method, which is based on the inspection device and comprises the following steps.

[0014] S1: According to the formulated inspection route, the unmanned aerial vehicle flies along the determined inspection route and inspection height, and the camera and the infrared camera carried by the unmanned aerial vehicle are used to take pictures of the equipment to be inspected in the power substation, and the temperature of the equipment to be inspected is detected to obtain equipment temperature data.

[0015] S2: extracting and screening the temperature data of each to-be-inspected equipment collected by the unmanned aerial vehicle;

[0016] S3: analyzing and calculating the temperature data of each equipment extracted in step S2 by using the defect identification system to determine whether the equipment has an abnormality;

[0017] S4: if the equipment has an abnormality in step S3, further analyzing the temperature data of the equipment by using the defect identification system to determine the heating defect level of the equipment.

[0018] Preferably, the analysis method further comprises external red template photo labeling, and the infrared photos are labeled by using a man-machine interactive interface. First, the to-be-detected equipment is classified, and second, the angle, position, and required number of infrared images of each type of to-be-detected equipment are labeled to obtain the infrared template photos. The unmanned aerial vehicle identifies the to-be-detected equipment and determines the equipment type according to the camera carried by the unmanned aerial vehicle, and according to the information labeled in the infrared template photos of the type, the infrared images of different types of to-be-detected equipment are sequentially captured according to a certain distance, pitch angle, and azimuth angle. At the same time, the unmanned aerial vehicle records the position and angle information of the captured photos.

[0019] Preferably, in the image capturing process of the infrared camera of the unmanned aerial vehicle, an image registration algorithm is used to register the infrared image photos captured by the unmanned aerial vehicle with the infrared template photos to determine the key attention areas in the infrared image capturing process.

[0020] The image registration method includes relative registration and absolute registration. The relative registration refers to extracting feature points on the infrared template photos, fitting the feature points of the infrared image photos captured by the infrared camera, fitting the space coordinates of the captured infrared image photos according to the similarity measurement, determining whether the space coordinates are consistent with the infrared template photos, and if the coordinate transformation parameters are fitted, the captured infrared image photos are registered with the infrared template photos.

[0021] Preferably, in step S2, the extraction method specifically comprises:

[0022] A1, converting the infrared image format captured by the infrared camera into a JPG format;

[0023] A2, converting the color image in the JPG format into a grayscale image;

[0024] A3, calculating the temperature of each point on the grayscale image, that is, completing the extraction of the temperature data in the infrared image photos;

[0025] The conversion method of step A2 is to calculate the values of three color channels in the color image by weighting, and the specific calculation formula is:

[0026] Y=ω r Mr +ω g M g +ω b M b ( Formula I )

[0027] In Formula I: ω i is the weight of different color channels, M i is the matrix of the extracted different color channels, r represents the red channel, b represents the blue channel, and g represents the green channel.

[0028] According to the weighted sum of the color components, an equivalent brightness value Y is calculated to convert the color image into a grayscale image.

[0029] Preferably, in step A3, the calculation method of the temperature of each point on the grayscale image is as follows: the parameter value is determined by the method of undetermined coefficients, the temperature value range is read in the infrared imager software, the maximum value Tmax and the minimum value Tmin are taken, which correspond to the maximum value Ymax and Ymin in the grayscale image respectively, and the coefficients of Y=kT+b can be determined according to the following formula.

[0030] k = (T max -T min ) / (Y max -Y min )(Formula II)

[0031] b = T max -kT min (Formula III)

[0032] The obtained linear relationship is used to calculate the temperature of each point on the grayscale image, and the temperature data on the picture can be completely extracted.

[0033] Preferably, the design of the defect identification system includes the following aspects: defect diagnosis classification design, composite insulator defect decision tree model generation, defect diagnosis system design,

[0034] The defect diagnosis classification design includes: pre-analysis by extracting the characteristic parameters of the original infrared image data, reducing the input data dimension of the key equipment area, and in the classifier construction stage, it can be divided into two processes of training and evaluation, using training data for classifier training, using test data for classifier evaluation, correcting the classification algorithm rules through training data, making the classification result consistent with the known data category, and in the use stage, the characteristic parameters of unknown data are extracted as the input data of the classifier for classification.

[0035] Preferably, the composite insulator defect decision tree model generation method is: using the CART algorithm, using the Gini coefficient ratio as the measurement of information gain, processing mixed data, substituting the data into the algorithm environment of Python, obtaining a preliminary shaped decision tree through machine learning, and pruning the decision tree to prevent overfitting;

[0036] The defect diagnosis system design method is: the decision tree analysis file obtained through training is placed in the background, the interactive interface transmits the data to be analyzed to the background analysis, and then feeds back the diagnosis result; after inputting corresponding infrared imaging images, air temperature and the like, the decision tree diagnosis can be performed; the main analysis interface transmits the input result to the trained decision tree.pickle file through the.get() statement, and after diagnosis, the decision tree returns the diagnosis result, which is displayed in the form of text.

[0037] Compared with the prior art, the present application has the following advantages: (1) the temperature of the surface of the equipment to be inspected is detected by the infrared camera, the temperature distribution of the surface of the equipment is directly and quantitatively detected, so that potential faults that may exist in the equipment can be found in time, preventive maintenance of the equipment is realized, and accidents such as equipment shutdown are reduced; (2) the temperature in the collected infrared image is extracted by the system, so that the situation of identifying and analyzing each possible temperature abnormal area and reason in each infrared image by relying on manual work is reduced, and the work efficiency of infrared image photo processing and analysis is improved; (3) the differentiated defect evaluation index is determined according to the category and function of the key equipment of the transformer substation, and the decision tree analysis method is used to realize the state evaluation of the transformer substation equipment. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is a structural schematic view of the unmanned aerial vehicle of the present application;

[0039] Figure 2 is a connection relationship schematic view of the infrared camera and the micro motor of the present application;

[0040] Figure 3 is a defect decision tree model establishment flowchart of the present application;

[0041] Figure 4 is a core flowchart of the decision tree analysis system of the present application. DETAILED DESCRIPTION

[0042] In order to better illustrate the present application, the present application will be further described in conjunction with examples.

[0043] A transformer substation unmanned aerial vehicle inspection device, such as Figures 1-2As shown, the inspection device includes a drone, the top of the body 1 of the drone is fixedly connected with a positioner 2, the position of the drone is accurately positioned through the positioner 2, the positioner 2 can be selected from a Beidou positioner 2 or a radar sensor, the outer side of the body 1 is fixedly connected with a wing 3 along the axial direction, the bottom of the wing 3 is detachably connected with a landing gear 4, a patrol camera mechanism is installed at the bottom of the drone, the patrol camera mechanism includes an infrared camera 5, the temperature of the surface of the equipment to be inspected is detected through the infrared camera 5, the temperature distribution of the surface of the equipment is directly and quantitatively detected, so that potential faults possibly existing in the equipment can be found in time, preventive maintenance of the equipment is realized, and accidents such as equipment shutdown possibly occurring are reduced. An image acquisition module and a wireless communication module (not shown in the figure) are arranged in the body 1 of the drone, in the flight process of the drone, the image acquisition module controls the patrol camera mechanism and inspects the equipment of the transformer substation.

[0044] The inspection device further includes a backend device, the backend device includes a terminal device, a video display and a data processing server, the data processing server is connected with a cloud service platform, the drone is connected with the cloud service platform through the wireless communication module, the drone inspects according to the inspection path provided by the data processor, and the cloud service platform is used for processing the images collected by the patrol camera mechanism.

[0045] In another embodiment, the patrol camera mechanism further includes a camera 6, and the camera and the infrared camera 5 are respectively located at two ends of the bottom of the drone; the camera is used for detecting the appearance of the equipment, and the infrared camera 5 is used for detecting the temperature of the surface of the equipment, so that multi-aspect detection of the equipment is realized and the work efficiency is improved.

[0046] The bottom of the body 1 of the drone is fixedly connected with a rack 7, the two ends of the rack 7 are spaced apart from the body 1, the spacing is called an installation interval 8, the infrared camera 5 is located at the installation interval 8, the two ends of the infrared camera 5 are fixedly connected with connecting shafts 9, the ends of the connecting shafts 9 are rotatably connected to the rack 7, a micro motor 10 is fixedly connected to the outer side wall of the rack 7, and the output shaft of the micro motor 10 is fixedly connected with the connecting shafts 9 on the infrared camera 5 through a speed reducer. In use, the image acquisition module drives the micro motor 10 to rotate by a certain angle according to the acquisition requirements of each equipment input in the image acquisition module, the connecting shafts 9 are driven to rotate through the speed reducer after transmission, so that the angle of the infrared camera 5 is adjusted, the installation interval 8 provides space for the angle adjustment of the infrared camera 5, in addition, the installation of the camera is the same as that of the infrared camera 5, and details are not described here.

[0047] In another embodiment, the wing 3 is connected to the body 1 through a horizontally arranged connecting rod 16, the landing gear 4 comprises a horizontally arranged connecting frame 11, the top of the connecting frame 11 is provided with a connecting groove 12 matched with the bottom of the wing 3, the bottom of the wing 3 is installed in the connecting groove 12 and the connecting frame 11 is connected to the wing 3 through bolts, the bottom of the connecting frame 11 is fixedly connected with two spaced apart adjusting plates 13, the two adjusting plates 13 are hingedly connected with a landing rod 14 through a hinge shaft, the hinge shaft is a damping shaft, the landing rod 14 is in a horizontal or vertical state after rotating along the hinge shaft, and the landing rod 14 in the horizontal state is parallel to the connecting rod 16. By adjusting the landing rod 14 to be in the horizontal state, the landing rod 14 is parallel to the connecting rod 16, so that the unmanned aerial vehicle is conveniently stored and the floor space is reduced; by adjusting the landing rod 14 to be in the vertical state, the unmanned aerial vehicle is conveniently landed through the landing rod 14 contacting the ground. The bottom of the landing rod 14 is fixedly connected with a supporting leg 15.

[0048] Another aspect of the present application protects a kind of substation unmanned aerial vehicle inspection analysis method, which is based on the above-mentioned inspection device, and the method comprises the following steps:

[0049] A kind of substation unmanned aerial vehicle inspection analysis method, which at least includes the following steps:

[0050] S1: according to the formulated inspection route, the unmanned aerial vehicle flies along the established inspection route and the inspection height, and the temperature of the equipment to be inspected in the substation is detected according to the camera carried by the unmanned aerial vehicle to obtain the equipment temperature data;Wherein the inspection route is formulated according to the shooting scheme, each device to be detected is covered, the take-off point, the landing point and each detection point coordinate of the unmanned aerial vehicle inspection route are set, each detection point coordinate is the shooting position of the device to be detected, the detection area position is checked by unmanned aerial vehicle test flight, and the reasonable inspection route is formulated according to the coordinate information, the selected inspection route is the shortest in all paths, so as to save the inspection time and improve the inspection efficiency. By providing the inspection scheme for the substation unmanned aerial vehicle automatic inspection, the unmanned aerial vehicle can perform one-key inspection according to the inspection path. At the same time, the highest height of the inspection route is determined according to the image resolution of the unmanned aerial vehicle and the camera focal length, and the image information returned in the unmanned aerial vehicle test flight.

[0051] S2: the temperature data of each device to be inspected collected by the unmanned aerial vehicle is extracted and screened;

[0052] S3: the temperature data of each device extracted in step S2 is analyzed and calculated by using a defect identification system to determine whether the device has an abnormality;

[0053] S4: If there is an abnormality in the device in step S3, further analyze the temperature data of the device using the defect identification system to determine the level of the heating defect of the device.

[0054] The unmanned aerial vehicle is equipped with an infrared camera, which detects the temperature of the device, and the obtained device temperature data is a colorful infrared image.

[0055] In another embodiment, the analysis method further comprises infrared template photo labeling, which labels the infrared photo using a human-computer interaction interface. First, determine the key area to be detected, enclose the key area by a box, and classify the devices to be detected in the key area. The type of device determines the formulation and selection of the infrared defect identification strategy, which facilitates subsequent identification of various types of infrared defects according to the infrared defect identification standard. The types of devices include bridge, switch, main transformer, and insulating support. Second, according to each type of device to be detected, label the angle, position, and number of infrared images required for the type of device to be detected. The unmanned aerial vehicle identifies the device to be detected and determines its type according to the camera carried by the unmanned aerial vehicle. According to the information labeled in the infrared template photo of the type, the unmanned aerial vehicle photographs and collects infrared images of different types of devices to be detected one by one at a certain distance, pitch angle, and azimuth angle. At the same time, the unmanned aerial vehicle records the position and angle information of the photograph.

[0056] In another embodiment, the infrared camera of the unmanned aerial vehicle uses an image registration algorithm to register the infrared image photo taken by the unmanned aerial vehicle with the infrared template photo during image collection, and then determines the key attention area during infrared image collection, thereby improving the identification accuracy of the key components and parts of the power device. The image registration method can be summarized as relative registration and absolute registration. Relative registration refers to selecting one image as a reference image and registering other related images with it, and its coordinate system is arbitrary. Specifically, using the relative registration principle, feature points are extracted on the infrared template photo, and the feature points of the photographed infrared photo are fitted according to the similarity measure to fit the spatial coordinates of the photographed infrared photo and determine whether it is consistent with the infrared template photo. If the coordinate transformation parameter fitting is fitted, the infrared photo is registered with the infrared template photo, which can be used for defect identification of the key parts of the substation device.

[0057] In step S2, the extraction method specifically includes:

[0058] A1, convert the infrared image format collected by the infrared camera into JPG format; the infrared image can only show the relative distribution of temperature, and cannot obtain the data of temperature. An infrared image is a color photo, and the color of different pixel points represents different temperatures. White to black represents high temperature to low temperature, that is, the brighter the color, the higher the temperature. The image format exported by the infrared imager is *.BMP, which is a bitmap. Converting it into a pixel image is easy to produce noise points, which greatly affects the subsequent temperature extraction. Therefore, the BMP format is first converted into JPG format.

[0059] A2, convert the color image in JPG format into a gray scale image; since the gray scale value of the pixel point has a linear relationship with a certain range of temperature, the color image is converted into a gray scale image. The conversion method is to calculate the values of the three color channels in the color image by weighting. The specific calculation formula is:

[0060] Y=ω r M r +ω g M g +ω b M b (Formula I)

[0061] In formula I: ω i is the weight of different color channels, M i is the matrix of the extracted different color channels, r represents the red channel, b represents the blue channel, and g represents the green channel.

[0062] According to the weighted sum of the color components, the equivalent brightness value Y is calculated to convert the color image into a gray scale image.

[0063] A3, calculate the temperature of each point on the gray scale image, that is, complete the extraction of the temperature data in the infrared image photo. The calculation method of the temperature of each point on the gray scale image is: in order to obtain the linear relationship between the actual temperature value and the image gray scale value, the parameter value is determined by the undetermined coefficient method. Read the temperature value range in the infrared imager software, take the maximum value Tmax and the minimum value Tmin, and correspond to the maximum value Ymax and Ymin in the gray scale image respectively. According to the following formula, the coefficient of Y=kT+b can be determined.

[0064] k=(T max -T min ) / (Y max -Y min )

[0065] b=T max -kT min

[0066] The obtained linear relationship is used to calculate the temperature of each point on the gray scale image, that is, the temperature data on the picture can be completely extracted.

[0067] The aforementioned defect identification system, following a pre-defined infrared defect identification strategy, automatically analyzes and calculates the temperature of key components and parts of power equipment in infrared images, thereby enabling the identification of various types of temperature anomalies in substation equipment, including current-induced heating, voltage-induced heating, and combined heating types. The specific infrared defect identification strategy includes: 1. Checking if the equipment surface temperature exceeds a safety threshold; 2. Comparing temperature differences between equipment of the same type, in the same environment, and with the same function to determine if any anomalies exist; 3. Setting appropriate temperature levels based on the substation equipment at different locations, measuring and extracting the highest temperature of the area, and then calculating the difference between the highest temperature and the equipment's maximum allowable temperature, using this as the basis for determining the level of the heating defect.

[0068] The design of the defect identification system is as follows: Figure 1 As shown, it includes the following aspects: defect diagnosis classification design, composite insulator defect decision tree model generation, and defect diagnosis system design. The defect diagnosis classification design includes: pre-analyzing by extracting feature parameters from raw infrared image data to reduce the dimensionality of input data for key equipment areas; the classifier construction stage can be divided into two processes: training and evaluation; classifier training using training data and evaluation using test data; and correcting the classification algorithm rules using training data to ensure the classification results match known data categories. In the usage stage, feature parameters of unknown data are extracted and used as input data for the classifier. Combining a standard database of infrared template photos of key equipment areas taken within substations, a defect diagnosis algorithm and criteria for key equipment areas are proposed, employing threshold division or intelligent algorithms with low computational time cost to achieve equipment defect diagnosis.

[0069] The method for generating the decision tree model for composite insulator defects is as follows: The CART algorithm is used, the Gini coefficient ratio is used as a measure of information gain, mixed data is processed, the data is substituted into the Python algorithm environment, and a preliminary decision tree is obtained through machine learning. To prevent overfitting, the decision tree is pruned.

[0070] Defect diagnosis system design methods, such as Figure 2 As shown, specifically: the decision tree analysis file obtained through training is placed in the background. The interactive interface will send the data to be analyzed to the background for analysis and then provide feedback on the diagnostic results. After inputting the corresponding infrared image, temperature and other information, the decision tree diagnosis can be performed. The main analysis interface uses the .get() statement to transfer the input results to the pre-trained decision tree .pickle file. After the diagnosis is performed, the decision tree will return the diagnostic results, which will be displayed in text form.

[0071] The above merely describes the preferred embodiments of the present application, and it should be pointed out that, for those skilled in the art, several changes and improvements can be made without departing from the overall concept of the present application, and these should also be considered as the protection scope of the present application.

Claims

1. A substation unmanned aerial vehicle (UAV) inspection device, characterized in that: The inspection device includes a drone, with a locator fixedly connected to the top of the drone's body, wings fixedly connected to the outside of the drone along its axis, landing gear detachably connected to the bottom of the wings, and an inspection camera mechanism installed on the bottom of the drone. The inspection device also includes back-end equipment. The drone communicates with the back-end equipment. The drone performs inspections according to the inspection path provided by the back-end equipment. The back-end equipment is used to process the images collected by the inspection camera. The aforementioned inspection camera device includes an infrared camera.

2. The substation unmanned aerial vehicle (UAV) inspection device according to claim 1, characterized in that: The inspection camera mechanism also includes a camera, with the camera and infrared camera located at opposite ends of the bottom of the drone. A frame is fixedly connected to the bottom of the drone body. The two ends of the frame are spaced apart from the drone body. This space is called the installation space. The infrared camera is located at the installation space. Connecting shafts are fixedly connected to both ends of the infrared camera. The ends of the connecting shafts are rotatably connected to the frame. A micro motor is fixedly connected to the outer wall of the frame. The output shaft of the micro motor is fixedly connected to the connecting shaft on the infrared camera through a reducer.

3. The substation unmanned aerial vehicle (UAV) inspection device according to claim 1, characterized in that: The wing is connected to the fuselage via a horizontally arranged connecting rod. The landing gear includes a horizontally arranged connecting frame. The top of the connecting frame has a connecting groove that matches the bottom of the wing. Two spaced-apart adjusting plates are fixedly connected to the bottom of the connecting frame. A landing bar is hinged between the two adjusting plates via a hinge shaft. The landing bar can be rotated along the hinge shaft to be in a horizontal or vertical state, and the horizontal landing bar is parallel to the connecting rod.

4. The substation unmanned aerial vehicle (UAV) inspection and analysis method according to claim 1, characterized in that: This analysis method is based on the inspection device described in any one of claims 1-3, and includes the following steps: S1: According to the established inspection route, the UAV flies along the predetermined inspection route and at the designated inspection altitude. Using its onboard camera and infrared camera, the UAV takes fixed-point photos of the equipment to be inspected within the substation, and simultaneously detects the temperature of the equipment to be inspected, obtaining equipment temperature data. The external infrared camera detects the equipment temperature, and the obtained equipment temperature data is a color infrared image. The inspection route includes the coordinates of the takeoff point, landing point, and inspection point. A reasonable inspection route is formulated based on the coordinate information provided. S2: Extract and filter the temperature data of each piece of equipment to be inspected collected by the drone; S3: Use the defect identification system to analyze and calculate the temperature data of each device extracted in step S2 to determine whether the device has any abnormalities; S4: If there is an abnormality in the equipment in step S3, the temperature data of the equipment is further analyzed using the defect identification system to determine the level of the heat defect in the equipment.

5. The substation unmanned aerial vehicle (UAV) inspection and analysis method according to claim 4, characterized in that: The analysis method also includes infrared template photo annotation. The infrared photos are annotated using a human-computer interaction interface. First, the devices to be detected are classified. Second, according to the category of devices to be detected, the shooting angle, position, and number of infrared images to be collected are marked to obtain infrared template photos. The UAV identifies the devices to be detected and determines their type based on the camera it carries. According to the information marked in the infrared template photos of that type, infrared images of different types of devices to be detected are captured one by one at a certain distance, pitch angle, and azimuth angle. At the same time, the UAV records the shooting position and angle information.

6. The substation unmanned aerial vehicle (UAV) inspection and analysis method according to claim 5, characterized in that: During the image acquisition process, the infrared camera of the drone uses an image registration algorithm to register the infrared image photo taken by the drone with the infrared template photo to determine the key areas of focus during the infrared image acquisition process. Image registration methods include relative registration and absolute registration. Relative registration refers to extracting feature points from the infrared template photo, fitting the feature points of the infrared image photo acquired by the infrared camera, fitting the spatial coordinates of the captured infrared image photo according to the similarity metric, and determining whether it is consistent with the infrared template photo. If the coordinate transformation parameters are fitted, the captured infrared image photo is registered with the infrared template photo.

7. The substation unmanned aerial vehicle (UAV) inspection and analysis method according to claim 4, characterized in that: In step S2, the extraction method specifically includes: A1 converts the infrared images captured by the infrared camera into JPG format; A2 converts a color image in JPG format to a grayscale image; A3 calculates the temperature of each point on the grayscale image, thus completing the extraction of temperature data from the infrared image. The conversion method in step A2 is to calculate the values ​​of the three color channels in the color image by weighting them. The specific calculation formula is as follows: Y = ω r M r + ω g M g + ω b M b (Equation 1) In Equation 1: ω i M represents the weights of different color channels. i Let r be the matrix of the extracted different color channels, where r represents the red channel, b represents the blue channel, and g represents the green channel. The equivalent luminance value Y is calculated based on the weighted sum of the color components, and the color image is converted into a grayscale image.

8. The substation unmanned aerial vehicle (UAV) inspection and analysis method according to claim 7, characterized in that: In step A3, the method for calculating the temperature of each point on the grayscale image is as follows: determine the parameter value by using the method of undetermined coefficients, read the temperature value range in the infrared imager software, take the maximum value Tmax and the minimum value Tmin, which correspond to the maximum values ​​Ymax and Ymin in the grayscale image respectively, and calculate the coefficient of Y=kT+b according to the following formula. k=(T max -T min ) / (Y max -Y min (Formula 2) b = T max -kT min (Formula 3) The obtained linear relationship can then be used to calculate the temperature of each point on the grayscale image, thus completely extracting the temperature data from the image.

9. The substation unmanned aerial vehicle (UAV) inspection and analysis method according to claim 6, characterized in that: The design of the defect identification system includes the following aspects: defect diagnosis and classification design, generation of composite insulator defect decision tree model, and defect diagnosis system design. The defect diagnosis classification design includes: pre-analyzing the feature parameters of the original infrared image data to reduce the dimensionality of the input data for key equipment areas; the classifier construction stage can be divided into two processes: training and evaluation; training data is used to train the classifier, and test data is used to evaluate the classifier; the classification algorithm rules are corrected using the training data to make the classification results consistent with the known data categories; and in the usage stage, the feature parameters of the unknown data are extracted and used as the input data for the classifier classification.

10. A method for analyzing unmanned aerial vehicle (UAV) inspections of substations according to claim 9, characterized in that: The method for generating the decision tree model for composite insulator defects is as follows: The CART algorithm is used, the Gini coefficient ratio is used as a measure of information gain, mixed data is processed, the data is substituted into the Python algorithm environment, and a preliminary decision tree is obtained through machine learning. To prevent overfitting, the decision tree is pruned. The defect diagnosis system is designed as follows: the decision tree analysis file obtained through training is placed in the background. The interactive interface will send the data to be analyzed to the background for analysis and then return the diagnosis results. After inputting the corresponding infrared image, temperature and other information, the decision tree diagnosis can be performed. The main analysis interface uses the .get() statement to transfer the input results to the pre-trained decision tree .pickle file. After the diagnosis is performed, the decision tree will return the diagnosis results, which will be displayed in text form.

Images