Power transmission line galloping detection method based on video

Abstract

The invention relates to a video-based power transmission line galloping detection method, which comprises the following steps of: segmenting a power transmission line by adopting an improved lightweight high-resolution network model, and positioning the specific position of the power transmission line; obtaining the width of the spacer and the coordinate information of the spacer; and calculating the maximum moving distance of the spacer in the x-axis direction and the y-axis direction and the relative maximum distance, and judging whether the power transmission line gallops or not. According to the method, videos are shot through the camera installed on site, the computer vision technology is used for segmenting, detecting and analyzing the motion condition of the spacer, no extra sensor is needed, direct physical intervention on a power transmission line is avoided, and the equipment installation and maintenance cost is greatly reduced; because the acquisition and processing of the video data can realize automation, the power transmission line galloping detection efficiency can be obviously improved, the problem discovery time can be shortened, early warning information can be timely provided, and the stable operation of a power system can be ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of power system detection and monitoring, and in particular to a video-based transmission line galloping detection method. Background Art

[0002] With the rapid development of the electric power industry, the safety and reliability of transmission lines, as an important infrastructure for power transmission, are directly related to the stable operation of the power system. Transmission lines are usually located in areas with complex terrain and changeable meteorological conditions, and are easily affected by environmental factors. Transmission line galloping is a common meteorological disaster phenomenon, which usually occurs under ice, snow or wind conditions. This phenomenon causes the conductors to swing violently and generate mechanical stress, which may cause problems such as conductor contact, wire breakage, and tower damage, seriously affecting the stability and safety of power supply.

[0003] Traditional methods for detecting the galloping of transmission lines mainly rely on on-site manual inspections and mechanical monitoring devices. Although manual inspections can provide intuitive detection results, due to the wide distribution of transmission lines and harsh environments, inspections often require a lot of manpower, material resources and time, and the inspection frequency is low, making it difficult to detect problems in a timely manner. Mechanical monitoring devices such as accelerometers and displacement sensors can provide real-time monitoring data, but they have high installation and maintenance costs, and are limited by the installation location. They cannot fully cover the entire transmission line, and there are monitoring blind spots. In addition, some of these sensors need to be installed on the conductors, which not only adds load to the conductors, but also has high costs. Summary of the invention

[0004] In order to solve the problem that the current galloping detection requires the installation of sensors, which increases the load on the wires and is costly, the purpose of the present invention is to provide a video-based transmission line galloping detection method that greatly reduces the cost of equipment installation and maintenance, significantly improves the efficiency of transmission line galloping detection, provides early warning information in a timely manner, and ensures the stable operation of the power system.

[0005] To achieve the above object, the present invention adopts the following technical solution: a video-based power transmission line galloping detection method, the method comprising the following steps in order:

[0006] (1) Improving the lightweight high-resolution network model to obtain an improved lightweight high-resolution network model, using the improved lightweight high-resolution network model to segment the transmission line and locate the specific location of the transmission line;

[0007] (2) The spacers of the located transmission lines are detected using the YOLOv8 network model, and the BoT-SORT algorithm is used to track the spacers to obtain the width and coordinate information of the spacers;

[0008] (3) According to the width of the spacer bar and the coordinate information of the spacer bar, calculate the maximum distance d that the spacer bar moves in the x-axis and y-axis directions x ,d y , and the relative maximum distance D x , D y At the same time, according to the coordinate information of the spacer bar, the frequency f of the spacer bar dancing is calculated;

[0009] (4) According to the relative maximum distance D x , D y , determine whether the transmission line is dancing.

[0010] In step (1), the improved lightweight high-resolution network model consists of the first unit to the fourth unit, wherein the first unit consists of a convolutional layer and a shuffle module, and the second, third and fourth units are all composed of an SE module and a fusion module; the SE module obtains the global compressed feature vector of the current feature map by performing a global average pooling operation on the feature map, obtains the weight of each channel in the feature map through two fully connected layers, and uses the weighted feature map as the input of the next layer of the network.

[0011] In step (1), the use of an improved lightweight high-resolution network model to segment the transmission line and locate the specific position of the transmission line refers to: inputting a segmentation data set consisting of multiple transmission line images with marked spacers into the improved lightweight high-resolution network model for training to obtain a trained model, and then using the trained model to segment the transmission line to obtain the outline of the transmission line and obtain the specific position of the transmission line.

[0012] The step (2) specifically refers to: downloading multiple images of power transmission lines with spacers, then annotating the spacers in the power transmission line images, the annotated power transmission line images forming a spacer detection data set, inputting the spacer detection data set into the YOLOv8 network model, training the YOLOv8 network model, and obtaining a trained YOLOv8 network model; inputting the to-be-processed power transmission line dancing video with spacers into the trained YOLOv8 network model, the trained YOLOv8 network model processes each frame of the power transmission line dancing video with spacers, and outputs a detection result of the spacers, wherein the detection result of the spacers includes a bounding box and a confidence of each spacer, and inputting the detection result of the spacers into the BoT-SORT algorithm, the BoT-SORT algorithm uses the bounding box of each spacer, a motion prediction model based on a Kalman filter and a Hungarian algorithm to track the motion path of each spacer in the power transmission line dancing video with spacers; calculating the width w of the spacer through the YOLOv8 network model, and recording the coordinate information of the spacer at every moment, i.e., the center point coordinate (x i ,yi ).

[0013] The step (3) specifically refers to: obtaining the maximum distance d of the spacer bar moving in the x-axis and y-axis directions according to the coordinate information of the spacer bar x and d y and the dancing frequency f, the corresponding relative maximum distances in the horizontal and vertical directions are D x =d x / w、D y =d y / w.

[0014] The step (4) specifically refers to: if the relative maximum distance D x >1 or D y >1, it is considered that the transmission line is galloping, otherwise, it is judged that the transmission line is not galloping.

[0015] It can be seen from the above technical scheme that the beneficial effects of the present invention are: first, the present invention uses a camera installed on site to shoot video, and uses computer vision technology, namely segmentation and detection, to analyze the movement of the spacer rods. No additional sensors are required, which avoids direct physical intervention in the transmission line and greatly reduces the cost of equipment installation and maintenance. Second, since the acquisition and processing of video data can be automated, the present invention can significantly improve the efficiency of transmission line dancing detection, shorten the time for problem discovery, provide early warning information in a timely manner, and ensure the stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the structure of the improved lightweight high-resolution network model;

[0017] Figure 2 for Figure 1 Schematic diagram of the shuffle module in ;

[0018] Figure 3 for Figure 1 Schematic diagram of the SE module;

[0019] Figure 4 Visualization of the transmission line segmentation results for the improved lightweight high-resolution network model;

[0020] Figure 5 It is a schematic diagram of the spacer rod detection results;

[0021] Figure 6 Schematic diagram of the change of the amplitude of the spacer bar over time. DETAILED DESCRIPTION

[0022] A method for detecting power line galloping based on video, the method comprising the following steps in order:

[0023] (1) Improving the lightweight high-resolution network model to obtain an improved lightweight high-resolution network model, using the improved lightweight high-resolution network model to segment the transmission line and locate the specific location of the transmission line;

[0024] (2) The spacers of the located transmission lines are detected using the YOLOv8 network model, and the BoT-SORT algorithm is used to track the spacers to obtain the width and coordinate information of the spacers;

[0025] (3) According to the width of the spacer bar and the coordinate information of the spacer bar, calculate the maximum distance d that the spacer bar moves in the x-axis and y-axis directions x ,d y , and the relative maximum distance D x , D y At the same time, according to the coordinate information of the spacer bar, the frequency f of the spacer bar dancing is calculated;

[0026] (4) According to the relative maximum distance D x , D y , determine whether the transmission line is dancing.

[0027] like Figure 1 As shown in Table 1, in step (1), the improved lightweight high-resolution network model consists of the first unit steml to the fourth unit stem4, wherein the first unit steml consists of a convolutional layer and a shuffle module, and the second, third and fourth units stem2, 3, 4 are all composed of SE modules and fusion modules; the SE module obtains the global compressed feature vector of the current feature map by performing a global average pooling operation on the feature map, and obtains the weight of each channel in the feature map through two fully connected layers, and uses the weighted feature map as the input of the next layer of the network.

[0028] Table 1 Composition of the improved lightweight high-resolution network model

[0029]

[0030] like Figure 2 As shown in , the shuffle module first divides the channel into two parts, one part undergoes 1×1 convolution, 3×3 depth convolution and 1×1 convolution, and then concatenates the output with the input and shuffles the concatenated channel. Figure 3As shown in the figure, the SE module is divided into two steps: Squeeze and Excitation. Each feature map is compressed into a single value through global average pooling, and then passes through two fully connected layers with a ReLU layer in the middle of the two fully connected layers. Then a sigmoid activation function is used to output the weight of each channel. Finally, each channel of the original input feature map is multiplied by the corresponding weight to achieve adaptive feature enhancement.

[0031] In step (1), the use of an improved lightweight high-resolution network model to segment the transmission line and locate the specific position of the transmission line refers to: inputting a segmentation data set consisting of multiple transmission line images with marked spacers into the improved lightweight high-resolution network model for training to obtain a trained model, and then using the trained model to segment the transmission line to obtain the outline of the transmission line and obtain the specific position of the transmission line.

[0032] The step (2) specifically refers to: downloading multiple images of power transmission lines with spacers, then annotating the spacers in the power transmission line images, the annotated power transmission line images forming a spacer detection data set, inputting the spacer detection data set into the YOLOv8 network model, training the YOLOv8 network model, and obtaining a trained YOLOv8 network model; inputting the to-be-processed power transmission line dancing video with spacers into the trained YOLOv8 network model, the trained YOLOv8 network model processes each frame of the power transmission line dancing video with spacers, and outputs a detection result of the spacers, wherein the detection result of the spacers includes a bounding box and a confidence of each spacer, and inputting the detection result of the spacers into the BoT-SORT algorithm, the BoT-SORT algorithm uses the bounding box of each spacer, a motion prediction model based on a Kalman filter and a Hungarian algorithm to track the motion path of each spacer in the power transmission line dancing video with spacers; calculating the width w of the spacer through the YOLOv8 network model, and recording the coordinate information of the spacer at every moment, i.e., the center point coordinate (x i ,y i ).

[0033] The step (3) specifically refers to: obtaining the maximum distance d of the spacer bar moving in the x-axis and y-axis directions according to the coordinate information of the spacer bar x and d y and the dancing frequency f, the corresponding relative maximum distances in the horizontal and vertical directions are D x =d x / w、D y =d y / w.

[0034] The step (4) specifically refers to: if the relative maximum distance Dx >1 or D y >1, it is considered that the transmission line is galloping, otherwise, it is judged that the transmission line is not galloping.

[0035] like Figure 4 As shown, the labels on the detection box include id, jgb and floating point numbers, where id represents the number of the tracked spacer bar; jgb represents the name of the identified object, which is composed of the first letters of the three spacer bars; and the floating point number represents the confidence.

[0036] like Figure 5 As shown, Figure 5 The image on the left is the original image; Figure 5 The figure on the right shows the result of segmentation. The white area is the transmission line and the black area is the background. In this way, the position of the transmission line is located.

[0037] like Figure 6 As shown, the horizontal axis is time, and the vertical axis is the relative movement distance in the x direction or the y direction.

[0038] In summary, the present invention uses a camera installed on site to shoot video, and utilizes computer vision technology, namely segmentation and detection, to analyze the movement of the spacer bars. No additional sensors are required, direct physical intervention in the transmission line is avoided, and the cost of equipment installation and maintenance is greatly reduced. Since the acquisition and processing of video data can be automated, the present invention can significantly improve the efficiency of transmission line dancing detection, shorten the time for problem discovery, provide early warning information in a timely manner, and ensure the stable operation of the power system.

Claims

1. A video-based method for detecting power line galloping, characterized in that: The method comprises the following steps in order: (1) Improving the lightweight high-resolution network model to obtain an improved lightweight high-resolution network model, using the improved lightweight high-resolution network model to segment the transmission line and locate the specific location of the transmission line; (2) The spacers of the located transmission lines are detected using the YOLOv8 network model, and the BoT-SORT algorithm is used to track the spacers to obtain the width and coordinate information of the spacers; (3) According to the width of the spacer bar and the coordinate information of the spacer bar, calculate the maximum distance d that the spacer bar moves in the x-axis and y-axis directions x ,d y , and the relative maximum distance D x , D y At the same time, according to the coordinate information of the spacer bar, the frequency f of the spacer bar dancing is calculated; (4) According to the relative maximum distance D x , D y , determine whether the transmission line is dancing.

2. The video-based power transmission line galloping detection method according to claim 1, characterized in that: In step (1), the improved lightweight high-resolution network model consists of the first unit to the fourth unit, wherein the first unit consists of a convolutional layer and a shuffle module, and the second, third and fourth units are all composed of an SE module and a fusion module; the SE module obtains the global compressed feature vector of the current feature map by performing a global average pooling operation on the feature map, obtains the weight of each channel in the feature map through two fully connected layers, and uses the weighted feature map as the input of the next layer of the network.

3. The video-based power transmission line galloping detection method according to claim 1, characterized in that: In step (1), the use of an improved lightweight high-resolution network model to segment the transmission line and locate the specific position of the transmission line refers to: inputting a segmentation data set consisting of multiple transmission line images with marked spacers into the improved lightweight high-resolution network model for training to obtain a trained model, and then using the trained model to segment the transmission line to obtain the outline of the transmission line and obtain the specific position of the transmission line.

4. The video-based power transmission line galloping detection method according to claim 1, characterized in that: The step (2) specifically refers to: downloading multiple images of power transmission lines with spacers, then annotating the spacers in the power transmission line images, the annotated power transmission line images forming a spacer detection data set, inputting the spacer detection data set into the YOLOv8 network model, training the YOLOv8 network model, and obtaining a trained YOLOv8 network model; inputting the to-be-processed power transmission line dancing video with spacers into the trained YOLOv8 network model, the trained YOLOv8 network model processes each frame of the power transmission line dancing video with spacers, and outputs a detection result of the spacers, wherein the detection result of the spacers includes a bounding box and a confidence of each spacer, and inputting the detection result of the spacers into the BoT-SORT algorithm, the BoT-SORT algorithm uses the bounding box of each spacer, a motion prediction model based on a Kalman filter and a Hungarian algorithm to track the motion path of each spacer in the power transmission line dancing video with spacers; calculating the width w of the spacer through the YOLOv8 network model, and recording the coordinate information of the spacer at every moment, i.e., the center point coordinate (x i ,y i ).

5. The video-based power transmission line galloping detection method according to claim 1, characterized in that: The step (3) specifically refers to: obtaining the maximum distance d of the spacer bar moving in the x-axis and y-axis directions according to the coordinate information of the spacer bar x and d y and the dancing frequency f, the corresponding relative maximum distances in the horizontal and vertical directions are D x =d x / w、D y =d y / w.

6. The video-based power transmission line galloping detection method according to claim 1, characterized in that: The step (4) specifically refers to: if the relative maximum distance D x >1 or D y >1, it is considered that the transmission line is galloping, otherwise, it is judged that the transmission line is not galloping.

Images