A vertical telescopic knife switch closing position identification device and closing state judgment method

Abstract

The present invention discloses a device for identifying whether a vertical telescopic knife switch is closed in place and a method for judging the closing state, which are used to identify whether the vertical telescopic knife switch of a substation is accurately closed in place. The device includes: a camera, a vertical telescopic knife switch identification front-end device, a vertical telescopic knife switch identification background learning and training device, and a data and control bus. The present invention adopts a cloud-edge fusion architecture, uses the YOLO‑V5 algorithm for target identification, and then uses the HED method for edge extraction. On this basis, a method for detecting incomplete closing is proposed, thereby realizing accurate identification of the state of the vertical telescopic knife switch, avoiding serious accidents such as heating and even explosion caused by incomplete closing of the vertical telescopic knife switch, and can replace manual inspections, saving manpower and improving safety. The present invention can also be used to monitor other equipment in substations.

Description

Technical Field

[0001] The present invention relates to a power knife switch state detection technology, in particular to accurate detection of the state of a vertical telescopic knife switch, and specifically to a vertical telescopic knife switch closing position identification device and a closing state judgment method. Background Art

[0002] The knife switch, also known as the high-voltage disconnector, is a major equipment in a substation. The vertical telescopic knife switch described in the present invention refers to a single-phase vertical telescopic knife switch. Each phase of a three-phase vertical telescopic knife switch has a vertical telescopic knife switch, and the three vertical telescopic knife switches are linked. In the present invention, each vertical telescopic knife switch has two vertical telescopic arms.

[0003] During long-term operation, various reasons may cause vertical telescopic knife switches to not close properly. Existing methods for determining whether a switch is closed properly rely on the auxiliary contacts or position sensors built into the vertical telescopic knife switches. Since vertical telescopic knife switches are typically placed outdoors, they are subject to sun and rain, metal corrosion, and wear. Abnormalities in the auxiliary contacts or their transmission components can lead to misjudgment of the closing and opening positions, resulting in incorrect signals being transmitted, threatening equipment safety and lifecycle, and even causing major power accidents, resulting in significant casualties and property losses.

[0004] Traditionally, inspectors use their naked eyes to identify incomplete closing faults. Inspectors are dispatched to the site for visual observation, but since most substations are located in relatively remote areas and each substation has a large number of vertical telescopic knife switches, on-site observation is both time-consuming and laborious. Furthermore, with the continuous development of society, the number of substations and the vertical telescopic knife switches used is also increasing. In order to quickly and simultaneously detect the closing status of numerous vertical telescopic knife switches, more human resources are required than before. This manual double detection method consumes a lot of manpower and has poor real-time performance. It is a shortcoming in the full automation of substations and is also dangerous. Inspectors must observe closely, and when the closing is not in place, the high-voltage disconnector heats up, which may generate discharge sparks or even explosions.

[0005] The development of a dual-verification method to automatically identify incomplete closing faults is extremely necessary. Both the State Grid Corporation of China and the China Southern Power Grid recently issued relevant notices. Current methods for identifying open and closed switch states rely on simple binary logic judgments, but the characteristics of incomplete closing are not readily apparent. Identifying incomplete closing faults is far more challenging than identifying open and closed switches, requiring precise identification devices and methods. Existing image acquisition or video surveillance methods can only determine whether the switch is open or closed, but cannot accurately determine whether it is fully closed. Other methods are also under investigation, such as infrared and gesture sensors. However, infrared methods require a period of power after closing to detect abnormal heating of the vertical telescopic knife switch. A power outage during this period would result in significant economic losses and disrupt the normal operation of the power grid, making them a post-event remedial measure. The gesture sensor requires additional equipment to be installed on the vertical telescopic knife switch, requiring a power outage for installation and maintenance, which also requires a power outage, disrupting grid operation. These methods have not yet achieved the desired results, and the development of better dual-verification methods is urgently needed.

[0006] At present, there are many types and models of knife switches in substations in my country, and their working principles and appearances vary greatly. The double-column vertical telescopic center-break vertical telescopic knife switch, hereinafter referred to as the vertical telescopic knife switch, is one of the most widely used types of knife switches in substations. The present invention studies the recognition of the closed state of this vertical telescopic knife switch. Summary of the Invention

[0007] In order to overcome the above-mentioned defects in the prior art, the present invention provides a vertical telescopic knife switch closing position identification device and a closing state judgment method to solve the above-mentioned problems.

[0008] The present invention provides the following technical solutions:

[0009] A vertical telescopic knife switch closing position recognition device, comprising:

[0010] Camera, vertical telescopic knife switch recognition front-end device, vertical telescopic knife switch recognition back-end learning and training device, data and control bus;

[0011] The camera is used to capture images of the vertical telescopic knife switch. The images are output to the vertical telescopic knife switch identification front-end device via the data and control bus. Three cameras are used to capture images of a three-phase vertical telescopic knife switch, with each camera capturing images of a single-phase vertical telescopic knife switch.

[0012] The vertical telescopic knife gate recognition front-end device sends a control signal to control the shooting action of the camera, and can also control the pan / tilt and rotation of the camera to adjust the shooting direction. The vertical telescopic knife gate recognition front-end device recognizes the image of the vertical telescopic knife gate and detects the closing curvature of the two telescopic arms to determine whether the closing is in place, and sends a signal of whether the closing is in place to the vertical telescopic knife gate recognition background learning and training device. At the same time, the image of the vertical telescopic knife gate is also forwarded to the vertical telescopic knife gate recognition background learning and training device. The recognition parameters of the vertical telescopic knife gate recognition front-end device are provided and refreshed by the vertical telescopic knife gate recognition background learning and training device.

[0013] The vertical telescopic knife switch recognition background learning and training device performs deep learning and training based on the received vertical telescopic knife switch image, sends the learned and trained parameters to the vertical telescopic knife switch recognition front-end device, and at the same time forwards the signal of whether the switch is closed to the dispatching center. Multiple vertical telescopic knife switch recognition front-end devices can share one vertical telescopic knife switch recognition background learning and training device;

[0014] The data and control bus is used to transmit data and control signals between the camera, the vertical telescopic knife gate identification front-end device and the vertical telescopic knife gate identification back-end learning and training device;

[0015] The vertical telescopic knife switches mentioned above all refer to single-phase vertical telescopic knife switches. A three-phase vertical telescopic knife switch contains three linked single-phase vertical telescopic knife switches.

[0016] Preferably, the cameras are arranged as follows: the bases of the left, middle and right cameras are installed in a straight line, the left and right cameras are cameras with fixed shooting angles, and the middle camera is a pan-tilt camera, which are used to shoot each phase of a three-phase vertical telescopic knife switch, and the center of each camera is aligned with the connecting axis of the two telescopic arms of a vertical telescopic knife switch; in addition to shooting the middle knife switch, the middle camera can also be used to shoot other electrical equipment.

[0017] A method for determining the closing state of a vertically retractable knife switch using the aforementioned closing position identification device is characterized in that it comprises the following steps:

[0018] Image acquisition and preprocessing of vertical telescopic knife gate: Acquire an image containing the vertical telescopic knife gate from the camera and perform image preprocessing;

[0019] Vertical telescopic knife switch target recognition: Find the required vertical telescopic knife switch in an image containing background information, filter out the background, extract the vertical telescopic knife switch from the captured image, and select the vertical telescopic knife switch to be tested using a minimum rectangular frame;

[0020] Edge extraction: extracting the contours of the two telescopic arms of the vertical telescopic knife switch within the range selected by the rectangular frame, so as to accurately detect the state of the switch-on by using a geometric method;

[0021] Closing curvature detection: Calculate the linear error of the contour lines of the two telescopic arms of the vertical telescopic knife switch. If the error exceeds a given threshold, a closing failure signal is issued; otherwise, a closing success signal is issued. If any of the three-phase vertical telescopic knife switches fails to close properly, the entire three-phase vertical telescopic knife switch is classified as failing to close properly.

[0022] Preferably, the vertical telescopic knife gate image acquisition and preprocessing includes:

[0023] Data acquisition: issue a shooting command to read the image of the vertical telescopic knife gate taken by the camera;

[0024] Data cropping: Different cameras have different resolutions. To adapt to different cameras and reduce unnecessary calculations, after obtaining photos from the camera, they are uniformly cropped to images with fixed resolution pixels;

[0025] Image noise removal: A filtering algorithm is used to filter the cropped image and perform a weighted average on the entire image. The value of each pixel is obtained by weighted averaging the value of itself and other pixels in its neighborhood. A Gaussian kernel is used to scan each pixel in the image, and then the weighted average grayscale value of the pixels in the area determined by the Gaussian kernel is calculated and used to replace the value of the convolution center pixel. The Gaussian kernel is an odd-sized Gaussian template.

[0026] Preferably, the vertical telescopic knife switch target recognition includes: using YOLO-V5 algorithm to identify the target.

[0027] Preferably, the edge extraction includes:

[0028] Build D network model;

[0029] A multi-scale deep learning algorithm is used on the HED network model.

[0030] Preferably, the closing curvature detection includes:

[0031] Edge cropping: The edge-extracted image is further cropped into an image centered on the edge contour line, saving computing time and speeding up system response;

[0032] Curvature calculation: Take the longest two line segments BC on the straight line outside the lower arm, B and C are the endpoints of the straight line part of the lower arm outer contour line, and point A on the outside of the top of the upper arm, then the curvature

[0033] =(AB+BC) / AC, where AC, AB, and BC are the pixel distances between two points. If the curvature is less than the set threshold, the switch is not closed properly, otherwise it is closed properly.

[0034] Preferably, the pan / tilt control method of the vertical telescopic knife switch closing position recognition device is characterized by adopting a fuzzy control algorithm, comprising the following steps:

[0035] Formula method fuzzy control with recursive adjustment rules;

[0036] Local continuation of the formula method;

[0037] The spherical gap migration global optimization method is used to optimize the local continuation formula method.

[0038] The vertical telescopic knife switch closing position identification device and closing status judgment method provided by the present invention have the following advantages and effects compared with the prior art:

[0039] 1. Can accurately judge whether the knife switch is closed in place;

[0040] 2. Good safety, non-contact optical video detection;

[0041] 3. No need for power outage for installation or maintenance, which will not affect the normal operation of the power system;

[0042] 4. Multi-purpose, in addition to monitoring the status of the knife switch, it can also monitor other power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a schematic diagram of the module structure and principle of the vertical telescopic knife switch closing position recognition device according to an embodiment of the present invention;

[0044] Figure 2 This is a schematic flow chart of a method for determining a closing state according to an embodiment of the present invention;

[0045] Figure 3 Schematic diagram of a vertical telescopic knife switch after target recognition processing according to an embodiment of the present invention;

[0046] Figure 4 Schematic diagram of the vertical telescopic knife gate after edge extraction processing according to an embodiment of the present invention;

[0047] Figure 5 Schematic diagram of curvature calculation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The present invention will be described in further detail below with reference to the embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0049] See also Figure 1As shown, the embodiment of the present invention provides a vertical telescopic knife switch closing position recognition device, which includes: a camera, a vertical telescopic knife switch recognition front-end device, a vertical telescopic knife switch recognition background learning and training device, a data and control bus, and a router;

[0050] Assume that a substation has n vertical telescopic switches. Figure 1 The left, middle and right cameras used to shoot vertical telescopic knife gate 1 and the left, middle and right cameras used to shoot vertical telescopic knife gate n are drawn in the figure. The middle camera has a pan / tilt head that can be tilted and rotated for shooting. Figure 1 A dotted line indicates that there are other vertical telescopic knife switches. The bases of the left, middle, and right cameras are installed in a straight line, shooting each phase of a three-phase vertical telescopic knife switch. The center of each camera is aimed at the connecting axis of the two telescopic arms of a vertical telescopic knife switch in each phase. The image of a vertical telescopic knife switch in a phase of a three-phase vertical telescopic knife switch is as follows: Figure 3 As shown, there are two telescopic arms in the green box, with the connecting axis at the center. A three-phase vertical telescopic knife switch has three such vertical telescopic knife switches.

[0051] Figure 1 In the center, a three-phase vertical telescopic knife switch is equipped with a vertical telescopic knife switch identification front-end device. When the vertical telescopic knife switch identification front-end device sends a shooting signal, the vertical telescopic knife switch images taken by the three cameras on the left, middle and right are output to the vertical telescopic knife switch identification front-end device through the data and control bus; the knife switch identification front-end device can also control the pan / tilt and rotation of the middle camera to adjust the shooting direction.

[0052] The vertical telescopic knife switch recognition front-end device uses a deep learning algorithm to recognize the vertical telescopic knife switch image, then analyzes the closing curvature value of the two telescopic arm contour lines to determine whether the closing is in place, and sends a signal of whether the closing is in place to the vertical telescopic knife switch recognition background learning and training device. At the same time, the image of the vertical telescopic knife switch is also forwarded to the vertical telescopic knife switch recognition background learning and training device. The deep learning network parameters of the vertical telescopic knife switch recognition front-end device are provided and refreshed by the vertical telescopic knife switch recognition background learning and training device. Figure 1 There are n three-phase vertical telescopic knife switches, but only one vertical telescopic knife switch identification background learning and training device is needed, because the deep learning network parameters of these vertical telescopic knife switch identification front-end devices are the same. Furthermore, multiple substations can share one vertical telescopic knife switch identification background learning and training device, so Figure 1 The vertical telescopic knife switch identification front-end device is connected to the data and control bus through a router, which is a currently advanced cloud-edge fusion architecture.

[0053] The vertical telescopic knife switch recognition background learning and training device performs deep learning training based on the received images of the vertical telescopic knife switch, sends the learned and trained parameters to the vertical telescopic knife switch recognition front-end device, and at the same time forwards the signal of whether the closing is in place to the dispatching center.

[0054] Figure 1 The data and control bus can use a common wired Ethernet or wireless Wi-Fi. The cameras are all network cameras, and the vertical telescopic knife gate recognition front-end device and the vertical telescopic knife gate recognition back-end learning and training device all have network interfaces. Data and control signals can be transmitted between any of the devices over the network. The dispatch center can also directly obtain vertical telescopic knife gate images to monitor the vertical telescopic knife gate without going through the vertical telescopic knife gate recognition front-end device.

[0055] Figure 2 This is a flow chart of a method for determining the closing state of a vertical telescopic knife switch closing position identification device.

[0056] The embodiment of the present invention provides a method for determining the closing state of a vertically retractable knife switch using the aforementioned closing position recognition device, which specifically includes the following steps:

[0057] S1. The vertical telescopic knife switch identification front-end device sends a shooting command to the left, center, and right cameras to simultaneously shoot a three-phase vertical telescopic knife switch that has been closed. The cameras send the images of the three vertical telescopic knife switches of the three-phase vertical telescopic knife switch to the vertical telescopic knife switch identification front-end device via the data and control bus.

[0058] S2, the vertical telescopic knife gate recognition front-end device calls the image preprocessing subroutine module to perform data cropping and noise removal on the image;

[0059] S3. Vertical telescopic knife gate identification The front-end device calls the vertical telescopic knife gate target identification subroutine module, finds the required vertical telescopic knife gate in an image containing background information, filters out the background, extracts the vertical telescopic knife gate from the captured image, and selects the vertical telescopic knife gate to be detected with a minimum rectangular frame, such as Figure 3 As shown;

[0060] S4. The vertical telescopic knife gate identification front-end device calls the edge extraction subroutine module to extract the contour lines of the two telescopic arms of the vertical telescopic knife gate within the range selected by the rectangular frame. The present invention adopts the HED network model and adopts a multi-scale deep learning algorithm on the HED network model;

[0061] S5. The vertical telescopic knife switch identification front-end device calls the closing curvature detection subroutine module to calculate the linear error of the contour lines of the two telescopic arms of the vertical telescopic knife switch. If the error exceeds a given threshold, a closing failure signal is issued; otherwise, a closing success signal is issued. If any of the three-phase vertical telescopic knife switches fails to close in place, the entire three-phase vertical telescopic knife switch is classified as failing to close in place.

[0062] S6. The vertical telescopic knife switch identification background learning and training device collects all the closing judgment signals of the vertical telescopic knife switch identification front-end device, presents them through the interface, and forwards them to the dispatching center.

[0063] above Figure 2 In the process, the image preprocessing subroutine module has the following detailed steps:

[0064] Data cropping: Different cameras have different resolutions. To adapt to different cameras and reduce unnecessary calculations, after obtaining photos from the camera, they are uniformly cropped to images with fixed resolution pixels;

[0065] Image noise removal: Use a well-known filtering algorithm to filter the cropped image and perform weighted averaging on the entire image. The value of each pixel is obtained by weighted averaging the value of itself and other pixels in the neighborhood. Use a Gaussian kernel to scan each pixel in the image, and then calculate the weighted average grayscale value of the pixels in the area determined by the Gaussian kernel, and use it to replace the value of the convolution center pixel. The Gaussian kernel is an odd-sized Gaussian template. Here, a 5*5 Gaussian template is used.

[0066] above Figure 2 In the process, the vertical telescopic knife switch target recognition subroutine module adopts the YOLO-V5 algorithm. Figure 3 The green box is a schematic diagram. For more details, see the YOLO-V5 algorithm.

[0067] Figure 4 Obtained by the edge extraction subroutine module, see HED algorithm.

[0068] above Figure 2 In the process, the closing curvature detection subroutine module has the following detailed steps: Edge cropping: the image after edge extraction is further cropped into an image centered on the edge contour line to save calculation time and speed up system response;

[0069] Curvature calculation: Figure 5 In the straight line part on the outside of the lower arm, take the longest two line segments BC, B and C are the endpoints of the straight line part of the outer contour of the lower arm, and point A on the outside of the top of the upper arm, then the curvature

[0070] =(AB+BC) / AC, where AC, AB, and BC are the pixel distances between two points. If the curvature is less than the set threshold, the switch is not closed properly, otherwise it is closed properly.

[0071] The final result example is shown in Figure 5 When the switch is fully closed, the two telescopic arms form a straight line. At this point, AC = AB + BC, and curvature = (AB + BC) / AC = 1. This is the ideal case. In practice, the arms sometimes stop before reaching their full position, a phenomenon known as under-closing. In this case, AC < (AB + BC), and curvature is greater than 1. Sometimes, the arms continue to rotate before reaching their full position, a phenomenon known as over-closing. In this case, AC < (AB + BC), and curvature is greater than 1. In either case, the curvature criterion applies. Of course, curvature = 1 is the ideal case; in practice, there are always errors. Therefore, a threshold can be set. When this threshold is exceeded, a non-position signal is issued.

[0072] Figure 1 The backend learning and training device for vertical telescopic knife gate recognition consists of a deep learning server as hardware, installed with the deep learning software related to vertical telescopic knife gates described in this invention, including vertical telescopic knife gate target recognition and edge extraction. The backend learning and training device performs deep learning training based on received images of the vertical telescopic knife gate. The learned and trained parameters are then transmitted to the frontend device for vertical telescopic knife gate recognition. Without these parameters, the corresponding program in the frontend device cannot run correctly.

[0073] Another function of the vertical telescopic knife switch identification background learning and training device is to summarize the signals of whether the closing is in place sent by multiple vertical telescopic knife switch identification front-end devices, present them through the interface, and forward them to the dispatching center. Multiple vertical telescopic knife switch identification front-end devices in a substation can share one vertical telescopic knife switch identification background learning and training device, and even multiple substations can share one vertical telescopic knife switch identification background learning and training device.

[0074] Figure 1 The two middle cameras can use pan-tilt integrated spherical cameras. The benefits of this solution are as follows:

[0075] 1) Because gun-type cameras do not have a pan / tilt and control mechanism, their price is much lower than that of dome cameras when the optical magnification is the same;

[0076] 2) The gun-type camera has a fixed position and angle, no movement or rotation, and its recognition accuracy is high and the error is small;

[0077] 3) Because the spherical camera needs to rotate and pitch, its positioning accuracy is relatively low, but it can monitor the surrounding area 360 degrees. In addition to shooting knife switches, it can also shoot all other equipment in the substation, such as busbars. By installing a spherical camera at each middle position, the purpose of monitoring other equipment in the substation can be achieved with fewer spherical cameras. Therefore, it can partially replace manual inspections and inspection robots. At the same time, when the left and right cameras are blocked by bird droppings or other faults occur, the middle spherical camera can also temporarily replace them.

[0078] This solution achieves multiple functions while saving costs. The key lies in the control algorithm of the dome camera. The dome camera is a combination of a camera and a pan / tilt head. It can rotate horizontally and tilt up and down, so it can capture a wide range of surrounding scenery. Its field of view is much larger than that of a fixed-angle gun-type camera. However, it also has two disadvantages:

[0079] Its rotation angle is digitally adjusted in steps and cannot be positioned infinitely, so it is difficult to accurately position it at a certain angle; its gimbal is mechanically driven by a motor gear, so there is a gap and usually a certain amount of virtual position error. After shooting the surrounding equipment, it may not be able to accurately return to the position where the knife switch was shot.

[0080] The recognition of the closing state of the knife switch must reach a high-precision positioning level, otherwise it is difficult to accurately identify and judge whether the closing state is in place. For this reason, the present invention adopts a local continuous surface high-precision fuzzy control algorithm to achieve it, see reference 1.

[0081] To achieve precise and rapid control of a dome camera, current mainstream automatic control theory requires developing differential equations and establishing relevant mathematical models. This, however, requires considerable time and experimentation, is difficult and costly, and therefore unsuitable for this project. The high-precision fuzzy control algorithm for locally continuous surfaces eliminates the need to establish a model of the controlled object. It locally continuousizes the control surface using table lookup and formula methods, maintaining high speed while exceeding even the accuracy of rule-based inference methods. Furthermore, the algorithm is simple.

[0082] The high-precision fuzzy control algorithm for local continuous surfaces has many parameters, which have a great impact on the control accuracy and return speed. The present invention uses a spherical gap migration algorithm to optimize the control parameters. For details, see Reference 2.

[0083] References:

[0084] 1) Hu Jinsong, Zheng Qilun, and Wu Jie, "Two-Level Self-Optimizing Fuzzy Controller with Time-Varying Correction Factors," Acta Automatica Sinica, 2002, 6, pp. 1006-1011.

[0085] 2) Jinsong Hu and Qilun Zheng, "Global Optimization Using Sphere Gap Migration Algorithm," Chinese Journal of Computers, Vol. 35, No. 2, February 2012, pp. 193-201.

[0086] The existing vertical telescopic knife switch is an extremely important equipment in the substation. Long-term use will lead to abnormalities in the structure of the vertical telescopic knife switch, resulting in inadequate closing, causing fever and even serious accidents such as explosion. The above embodiment of the present invention focuses on adopting a cloud-edge fusion architecture, using the YOLO-V5 algorithm for target recognition, and then using the HED method for edge extraction. On this basis, a method for detecting inadequate closing is proposed, thereby realizing accurate identification of the state of the vertical telescopic knife switch, which can replace manual inspections, save manpower, and improve safety. Furthermore, the present invention can also be used to monitor other equipment in the substation.

[0087] The above description is merely an exemplary embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A vertical telescopic knife switch closing position recognition device, characterized in that: It includes: Camera, vertical telescopic knife switch recognition front-end device, vertical telescopic knife switch recognition back-end learning and training device, data and control bus; The camera is used to capture images of the vertical telescopic knife switch. The images are output to the vertical telescopic knife switch identification front-end device via the data and control bus. Three cameras are used to capture images of a three-phase vertical telescopic knife switch, with each camera capturing images of a single-phase vertical telescopic knife switch. The vertical telescopic knife gate recognition front-end device sends a control signal to control the shooting action of the camera, and can also control the pan / tilt and rotation of the camera to adjust the shooting direction. The vertical telescopic knife gate recognition front-end device recognizes the image of the vertical telescopic knife gate and detects the closing curvature of the two telescopic arms to determine whether the closing is in place, and sends a signal of whether the closing is in place to the vertical telescopic knife gate recognition background learning and training device. At the same time, the image of the vertical telescopic knife gate is also forwarded to the vertical telescopic knife gate recognition background learning and training device. The recognition parameters of the vertical telescopic knife gate recognition front-end device are provided and refreshed by the vertical telescopic knife gate recognition background learning and training device. The vertical telescopic knife switch recognition background learning and training device performs deep learning and training based on the received vertical telescopic knife switch image, sends the learned and trained parameters to the vertical telescopic knife switch recognition front-end device, and at the same time forwards the signal of whether the switch is closed to the dispatching center. Multiple vertical telescopic knife switch recognition front-end devices can share one vertical telescopic knife switch recognition background learning and training device; The data and control bus is used to transmit data and control signals between the camera, the vertical telescopic knife gate identification front-end device and the vertical telescopic knife gate identification back-end learning and training device; The vertical telescopic knife switches mentioned above refer to single-phase vertical telescopic knife switches. A three-phase vertical telescopic knife switch includes three linked single-phase vertical telescopic knife switches. When extracting the edge, the contour lines of the two telescopic arms of the vertical telescopic knife switch are extracted within the range selected by the rectangular frame, so that the closing state can be accurately detected using geometric methods; During the closing curvature test, the linear error of the contour lines of the two telescopic arms of the vertical telescopic knife switch is calculated. If the error exceeds a given threshold, a closing failure signal is issued; otherwise, a closing success signal is issued. If any of the three-phase vertical telescopic knife switches fails to close properly, the entire three-phase vertical telescopic knife switch is classified as failing to close properly. When detecting the closing curvature, edge clipping is performed: the image after edge extraction is further clipped into an image centered on the edge contour line, saving calculation time and speeding up system response; Then calculate the curvature: take the two longest line segments BC on the straight line portion on the outside of the lower arm, where B and C are the endpoints of the straight line portion of the lower arm's outer contour line, and point A on the outside of the top of the upper arm. The curvature = (AB + BC) / AC, where AC, AB, and BC are the pixel distances between the two points. If the curvature is less than the set threshold, the circuit breaker is not closed properly; otherwise, it is closed properly.

2. The vertical telescopic knife switch closing position identification device according to claim 1 is characterized in that: The cameras are arranged as follows: The bases of the three cameras on the left, middle and right are installed in a straight line. The left and right cameras have fixed shooting angles, and the middle camera is a pan-tilt camera, which is used to shoot each phase of a three-phase vertical telescopic knife switch. The center of each camera is aimed at the connecting axis of the two telescopic arms of a vertical telescopic knife switch; in addition to shooting the middle knife switch, the middle camera can also be used to shoot other power equipment.

3. A method for judging the closing state of a vertical telescopic knife switch using the closing position identification device of claim 1 or 2, characterized in that: It includes the following steps: Image acquisition and preprocessing of vertical telescopic knife gate: Acquire images containing vertical telescopic knife gate from the camera and perform image preprocessing; Vertical telescopic knife switch target recognition: Find the required vertical telescopic knife switch in an image containing background information, filter out the background, extract the vertical telescopic knife switch from the captured image, and select the vertical telescopic knife switch to be tested using a minimum rectangular frame; Edge extraction: extracting the contour lines of the two vertical telescopic arms of the vertical telescopic knife switch within the range selected by the rectangular frame, so as to accurately detect the state of the switch-on by using a geometric method; Closing curvature detection: Calculate the linear error of the contour lines of the two telescopic arms of the vertical telescopic knife switch. If the error exceeds a given threshold, a closing failure signal is issued; otherwise, a closing success signal is issued. If any of the three-phase vertical telescopic knife switches fails to close properly, the entire three-phase vertical telescopic knife switch is classified as failing to close properly.

4. The method for judging the closing state according to claim 3, characterized in that: The vertical telescopic knife gate image acquisition and preprocessing includes the following steps: Data acquisition: issue a shooting command to read the image of the vertical telescopic knife gate taken by the camera; Data cropping: Different cameras have different resolutions. To adapt to different cameras and reduce unnecessary calculations, after obtaining photos from the camera, they are uniformly cropped to images with fixed resolution pixels; Image noise removal: A filtering algorithm is used to filter the cropped image and perform a weighted average on the entire image. The value of each pixel is obtained by weighted averaging the value of itself and other pixels in its neighborhood. A Gaussian kernel is used to scan each pixel in the image, and then the weighted average grayscale value of the pixels in the area determined by the Gaussian kernel is calculated and used to replace the value of the convolution center pixel. The Gaussian kernel is an odd-sized Gaussian template.

5. The method for judging the closing state according to claim 3, characterized in that: The vertical telescopic knife switch target recognition adopts the YOLO-V5 algorithm.

6. The method for judging the closing state according to claim 3, characterized in that: The edge extraction comprises the following steps: Construct HED network model; A multi-scale deep learning algorithm is used on the HED network model.

7. The method for judging the closing state according to claim 3, characterized in that: The closing curvature detection comprises the following steps: Edge cropping: The edge-extracted image is further cropped into an image centered on the edge contour line, saving computing time and speeding up system response; Curvature calculation: Take the longest two line segments BC on the straight line outside the lower arm, where B and C are the endpoints of the straight line portion of the lower arm's outer contour. Point A is the outermost point of the upper arm. Curvature = (AB + BC) / AC, where AC, AB, and BC are the pixel distances between the two points. If the curvature is less than the set threshold, the circuit breaker is not in place; otherwise, it is in place.

8. The method for judging the closing state according to claim 3, wherein: The pan / tilt control method of the vertical telescopic knife switch closing position recognition device adopts a fuzzy control algorithm, including the following steps: Formula method fuzzy control with recursive adjustment rules; Local continuation of the formula method; The spherical gap migration global optimization method is used to optimize the local continuation formula method.

Images