Method for optimizing operation and maintenance of power grid by adopting machine vision and predictive analysis

Abstract

The invention provides a method for optimizing power grid operation and maintenance by adopting machine vision and predictive analysis, and relates to the technical field of image processing, and the method specifically comprises the steps: collecting a plurality of video images of a power transmission line, and obtaining a video set; performing image splicing on the acquired video images to obtain a corresponding complete panoramic image; inputting the panoramic image into an icing detection model to obtain an icing detection result; through the above steps, the power grid operation maintenance method and system can improve the response capability and the operation efficiency of the power grid, reduce the maintenance cost and optimize the resource configuration.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a method for optimizing power grid operation and maintenance by using machine vision and predictive analysis. Background Art

[0002] Transmission lines often face challenges in extreme weather conditions during operation, especially snow and ice weather. Icing refers to the condensation and freezing of moisture on transmission lines in low temperature environments, which not only increases the weight of the lines, but may also cause line breakage or reduce power transmission efficiency.

[0003] The development of machine vision technology has been widely used in many fields, including manufacturing, medical treatment, transportation, etc. Through advanced image processing and pattern recognition technology, machine vision systems can realize automatic detection and analysis of various objects and scenes. For example, in the automotive industry, machine vision is used to detect defects in the manufacturing process; in the medical field, it is used to analyze medical images to assist in disease diagnosis.

[0004] Although machine vision technology has been widely used in many fields, its application in power grids, especially in extreme climate conditions, is still in its infancy. For example, the power sector often relies on data provided by the meteorological department to assess potential weather risks, but these data are usually provided for larger geographical areas rather than specific local areas of transmission lines. Therefore, these meteorological data often cannot accurately reflect the actual climate conditions at the location of specific transmission lines. In addition, traditional systems often lack the ability to adapt to changing environmental conditions. For example, in the case of strong winds, light ice may fall off naturally, but existing technologies fail to take advantage of this environment and still use standard de-icing procedures, resulting in unnecessary waste of resources.

[0005] In order to improve the grid's ability to respond to extreme weather, reduce maintenance costs, and improve the overall reliability of the grid, this patent proposes a method for optimizing grid operation and maintenance using machine vision and predictive analysis, optimizing grid operation and maintenance under extreme weather conditions, thereby protecting grid assets and ensuring public safety. Summary of the invention

[0006] The present invention provides a method for optimizing power grid operation and maintenance by using machine vision and predictive analysis, the method specifically comprising the following steps:

[0007] S1: Collect multiple video images of the transmission line to obtain a video set;

[0008] S2: stitching the collected video images to obtain a corresponding complete panoramic image;

[0009] S3: inputting the panoramic image into an ice detection model to obtain an ice detection result;

[0010] S4: Generate operation and maintenance recommendations based on the icing detection results.

[0011] The present invention provides a system for optimizing power grid operation and maintenance by using machine vision and predictive analysis, the system comprising:

[0012] Image acquisition device: the image acquisition device acquires multiple video images of the power transmission line to obtain a video set;

[0013] Panoramic image generation module: the panoramic image generation module performs image stitching on the collected video images to obtain a corresponding complete panoramic image;

[0014] Icing detection module: The icing detection module inputs the panoramic image into an icing detection model to obtain an icing detection result;

[0015] Operation and maintenance module: The operation and maintenance module generates operation and maintenance suggestions based on the ice detection results.

[0016] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for optimizing power grid operation and maintenance by using machine vision and predictive analysis is implemented.

[0017] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for optimizing power grid operation and maintenance using machine vision and predictive analysis is implemented.

[0018] Compared with the prior art, the present invention aims to solve the problem of the lack of ability to flexibly adjust maintenance strategies during power grid operation optimization and maintenance in the prior art. The present invention can calculate the curvature of the current line by stitching a panoramic image, and then infer the local wind conditions of the current transmission line. It does not rely on the meteorological data provided by the meteorological department. It obtains more accurate meteorological information for predictive analysis to optimize operating parameters and maintenance plans based on the actual location of the transmission line. In addition, the application also constructs an icing detection model to detect the amount and location information of ice in the current line. The icing detection model integrates multiple layers of features and enhances the features, improves the recognition ability of small targets, and combines the icing detection results with the current wind speed to comprehensively give optimization suggestions for power grid operation parameters and de-icing maintenance plans, thereby improving the reliability of power grid operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 is a flow chart of the optimized operation and maintenance recommendations in this application;

[0021] Figure 2 It is a structural diagram of the ice detection model in this application. DETAILED DESCRIPTION

[0022] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0023] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0024] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present application, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number and aspect described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0025] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the examples can be practiced without these specific details.

[0026] The embodiment of this specification proposes a method for optimizing power grid operation and maintenance using machine vision and predictive analysis, which specifically includes the following steps:

[0027] S1: Collect multiple video images of the transmission line to obtain a video set;

[0028] S2: stitching the collected video images to obtain a corresponding panoramic image;

[0029] S3: inputting the panoramic image into an ice detection model to obtain an ice detection result;

[0030] S4: Generate operation and maintenance recommendations based on the icing detection results.

[0031] The present invention uses drone equipment to shoot video images of overhead power transmission lines. Specifically, the drone equipment can be DJI's Matrice 300RTK, Inspire series drones or other drone equipment with long endurance, stable flight performance, high-quality shooting effects and good environmental adaptability. Preferably, the drone should be equipped with redundant systems, such as dual GPS, dual IMU, etc., to enhance safety and reliability in complex environments.

[0032] Before using drone equipment to shoot video images of overhead power transmission lines, the power grid construction situation in the area to be collected is retrieved in advance, that is, the number and location of transmission towers, the number, height and direction of overhead power transmission lines are retrieved. Based on the obtained power grid construction situation in the area to be collected, a corresponding drone equipment collection plan is formulated, and the collection plan includes setting the flight altitude, speed, flight route of the drone equipment, the take-off and landing position of the drone, and the emergency plan.

[0033] Specifically, for example, the power grid construction situation of a certain overhead transmission area is retrieved in advance to obtain a total transmission line length of about 10 kilometers, covering about 40 transmission towers. Use ArcGIS or QGIS to analyze the geographic data of the transmission line, including the precise location, altitude and surrounding terrain of each transmission tower. Ensure that the flight altitude is at least 20 meters higher than the highest transmission line, and the planned flight altitude is between 120-140 meters to avoid any risk of collision. Set a straight line flight from each transmission tower to the next tower, with an average distance of 250 meters per section, and use the drone's built-in GPS navigation system to maintain an accurate flight path. Set the drone's flight speed to 5 meters per second to balance the stability of the shooting and the efficiency of the mission. Each 250-meter straight line flight takes about 50 seconds, and the entire 10-kilometer line will be divided into 40 sections, with a total flight time of about 33 minutes. Select a drone model based on the flight time and flight distance (such as the DJI Matrice 300 RTK, which has a flight endurance of about 55 minutes). A ground station is set up every 5 kilometers for the drone to monitor its flight status and ensure the smooth progress of the mission. The ground station is equipped with at least two backup batteries to ensure that it can complete the entire mission or respond to any emergency.

[0034] After ensuring that all flight parameters are consistent with the preset parameters, the shooting task is carried out. During the flight, the operator monitors the output of the camera in real time to ensure that the video quality meets the requirements. For each section of the transmission line, when flying from one transmission tower to the next, the operator must control the drone to keep it stable to ensure the continuity and clarity of the transmission line in the video. Every time it arrives at a new transmission tower, mark the start and end points of the video for subsequent processing.

[0035] Preferably, during the filming process, the real-time transmission function of the drone is used to instantly transmit the data back to the ground station. The data is backed up at the ground station to ensure that the data will not be lost even if equipment failure occurs.

[0036] The collected continuous video is cut into pieces according to the transmission tower intervals, and necessary metadata is added to each video, such as the transmission tower number, date and time, flight parameters, etc., to facilitate subsequent analysis and retrieval.

[0037] The image stitching of the collected video images to obtain the corresponding complete panoramic image specifically includes:

[0038] S21: According to the flight speed of the drone, key frames are extracted from the video clips to obtain a key frame image set {D k |k∈[1,n]},D k is the kth key frame image, n is the number of extracted key frames, D1 is preset as the first panoramic image, and the extracted adjacent key frame images have some overlapping areas along the shooting direction;

[0039] S22: Identify and obtain stable SIFT feature points in the t-th panoramic image and the t+1-th frame image, where 1≤t≤n-1;

[0040] S23: Determine the matching feature points in the tth panoramic image and the t+1th frame image according to the k-nearest neighbor algorithm and the Euclidean distance, and obtain the matching feature point set in the tth panoramic image and the set of matching feature points in the t+1th key frame Where N represents the number of matching points. and Respectively represent the i-th matching feature point in the t-th panoramic image and the t+1 key frame image;

[0041] S24: Obtain the t+1th panoramic image I according to the image conversion formula t+1 ;

[0042]

[0043] Among them, tl and tr are the transition area position information, H is the edge position information, γ is the balance weight information, and I t (x, y) is the image information of the t-th panoramic image at coordinate (x, y), D t+1 (x, y) is the image information of the key frame image of the t+1th frame at the coordinate (x, y);

[0044] S25: Iterate steps S22-S24 to obtain the final panoramic image I n-1 As a complete panoramic image I panorama .

[0045] The calculation method of the transition area position information tl and tr specifically includes:

[0046] Matching feature point set and The first conversion matrix T and the second conversion matrix T' are obtained, wherein the first conversion matrix T is used to convert the pixels in the t-th panoramic image to the t+1-th key frame, and the second conversion matrix T' is used to convert the pixels in the t+1-th key frame to the t-th panoramic image; the first boundary B1 of the t-th panoramic image = {(x b ,y b )} Perform the first transformation to obtain the key frame D t+1 The first mapping boundary set E1 in e ,y` e )}, the first conversion calculation formula is:

[0047] (x` e ,y` e )=T(x b ,y b )

[0048] For key frame D t+1 The second boundary B2={(x` b ,y` b )} Perform the second transformation to obtain the key frame D t The second mapping boundary set E2 in e ,y e )}, the second conversion calculation formula is:

[0049] (x e ,y e )=T`(x` b ,y` b )

[0050] Set the first mapping boundary set E1 = {(x` e ,y` e )} in x` eThe maximum value of tr is set to the second mapping boundary set E2 = {(x e ,y e The first boundary is the image boundary corresponding to the flying direction of the UAV, and the second boundary is the image boundary in the direction opposite to the flying direction of the UAV.

[0051] The present invention achieves efficient acquisition of panoramic images of power transmission lines by introducing image stitching technology, and systematically collects image data along the line by controlling the acquisition device to fly in a straight line along the power transmission line in step S1. In this way, the status of the power transmission line can be continuously captured, and these continuous images can be synthesized into a panoramic image using image stitching technology, which not only greatly improves the integrity of the image data, but also ensures the coherence and accuracy of the image information.

[0052] In the panoramic images obtained, the degree of bending of the transmission lines reflects the impact of wind on the transmission lines. Specifically, when wind acts on the transmission lines, it will cause a certain degree of bending or deviation of the transmission lines. This physical deformation can be accurately captured and quantified by analyzing the morphology of the transmission lines in the panoramic images. By conducting in-depth analysis of these image data, it is possible to determine the current wind force that the transmission lines are subjected to, providing reliable weather environment information for subsequent forecasting and analysis.

[0053] The panoramic image is input into the ice detection model to obtain the ice detection result.

[0054] The ice detection model includes a feature extraction module, a feature enhancement module, a feature fusion module, and an attention guidance module; wherein the feature extraction module uses the first five residual modules of Resnet50 as the backbone network for feature extraction; the input complete panoramic image passes through the backbone network and is input into two feature enhancement modules for multi-scale feature fusion, while the attention guidance module is used for feature enhancement, and the multi-scale features are input into the classification module to obtain the ice detection result.

[0055] The feature enhancement module adopts sub-channel parallel enhancement, the number of sub-channels is 4, and the feature enhancement module outputs the feature F of the previous layer. m-1 Perform channel enhancement and use jump connection to splice F m-1 Get the enhanced feature F m :

[0056]

[0057] Among them, Conv represents convolution processing, Act represents activation function, 1*1 and 3*3 represent convolution kernels of size 1 and 3, and d represents expansion size. Represents the output features F of the previous layer respectively m-1subchannel 1, subchannel 2, subchannel 3 and subchannel 4 of the four subchannels; the number of each subchannel is evenly divided according to the total number of channels or divided according to a preset rule, and the feature maps in the subchannels do not overlap with each other;

[0058] The feature fusion module fuses the output features by cross-layer fusion. The cross-layer fusion is specifically to fuse the output features F of the previous layer. m-1 And the next layer output feature F m+1 To perform the fusion:

[0059] F c =Concat(DwnS(C l (F m-1 )), UpS(C h (F m+1 )))

[0060] Among them, F c Output feature F for the previous layer m-1 And the next layer output feature F m+1 Concat represents concatenation by channel, DwnS and UpS represent downsampling and upsampling of feature maps, respectively, to ensure that the scales of feature maps at different layers are consistent. l and C h It means that the low-level features and high-level features are reduced and upgraded by 1*1 to ensure the consistency of feature map channels at different layers;

[0061] Specifically, the present application performs cross-layer fusion on the output F3 of the third residual module in the backbone network and the output F5 of the fifth residual module to obtain a fusion feature The output F4 of the fourth residual module in the backbone network and the output of the first feature enhancement layer Perform cross-layer fusion to obtain fusion features Fusion Features and And the output of the second feature enhancement layer After attention guidance, it is input into the classification module for target detection.

[0062] In addition, the ice detection model also includes an attention guidance module, which uses the output F3 of the third residual module to generate an attention weight to guide the output of the first feature enhancement layer The output of the fourth residual module F4 is used to generate attention weights to guide the output of the second feature enhancement layer

[0063] The calculation method of the attention guidance module is:

[0064]

[0065] in, and is the output feature after attention guidance, σ represents the sigmoid activation function, AVG and MAX represent average pooling and maximum pooling respectively.

[0066] After the panoramic image is input into the ice detection model to obtain the ice detection result, the wind speed is calculated. The wind speed calculation specifically includes:

[0067] S31: grayscale the complete panoramic image obtained in step S2 and perform Gaussian filtering, and use the Canny operator to perform edge detection to obtain an edge image I edge , the edge image I edge including transmission line curves;

[0068] I edge =Canny(GaussianBlur(GrayScale(I panorama )))

[0069] Among them, Canny represents the Canny operator, GaussianBlur represents Gaussian filtering, and GrayScale represents grayscale processing;

[0070] S32: Edge image I edge The endpoints of the transmission line curve are identified to obtain the endpoint coordinates (x1, y1) and (x2, y2), and the midpoint coordinates (x m ,y m );

[0071] S33: Connect the endpoint coordinates (x1, y1) and the midpoint coordinates (x m ,y m ) get the first line segment L1, the endpoint coordinates (x2, y2) and the midpoint coordinates (x m ,y m ) obtain the second line segment L2;

[0072] S34: Calculate the angle θ between the first line segment L1 and the second line segment L2

[0073]

[0074] Where the vector α=(x1-x m , y1-y m ), β=(x2-x m ,y2-y m ), · represents the dot product, |||| represents the vector norm; the wind speed v in the transmission line area is calculated according to the angle value:

[0075]

[0076] Where, v represents wind speed, C d It represents the resistance coefficient of the transmission line, ρ is the air density, A is the diameter of the transmission line, E represents the elastic modulus of the transmission line material, I represents the cross-sectional moment of inertia of the transmission line, and L represents the length of the transmission line.

[0077] Generate operation and maintenance recommendations based on icing detection results.

[0078] The ice detection result includes the number of ice covers T in the complete panoramic image. n and ice position T c , according to the amount of ice T in the complete panoramic image n and the wind speed v in the area of ​​the transmission line determine the number of drones to be dispatched for de-icing;

[0079]

[0080] Where T min The minimum amount of ice that triggers the dispatch of UAVs for de-icing, V maxsafe Indicates the maximum wind speed that guarantees the drone's flight.

[0081] According to the amount of ice T in the complete panoramic image n And the wind speed v in the transmission line area optimizes the grid load parameters:

[0082]

[0083] Among them, S(T n , v) represents the ice coverage amount T n and recommended load parameters for wind speed v in the transmission line area, S max The maximum load capacity of the line, μ1 is the ice influence coefficient, μ2 is the wind speed influence coefficient, T max is the preset ice threshold, V max The preset maximum safe wind speed threshold for the transmission line.

[0084] The present invention provides a system for optimizing power grid operation and maintenance by using machine vision and predictive analysis, the system comprising:

[0085] Image acquisition device: the image acquisition device acquires multiple video images of the power transmission line to obtain a video set;

[0086] Panoramic image generation module: the panoramic image generation module performs image stitching on the collected video images to obtain a corresponding complete panoramic image;

[0087] Icing detection module: The icing detection module inputs the panoramic image into an icing detection model to obtain an icing detection result;

[0088] Operation and maintenance module: The operation and maintenance module generates operation and maintenance suggestions based on the ice detection results.

[0089] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for optimizing power grid operation and maintenance by using machine vision and predictive analysis is implemented.

[0090] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for optimizing power grid operation and maintenance using machine vision and predictive analysis is implemented.

[0091] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0092] In this specification, the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the embodiments described later, the description is relatively simple, and the relevant parts can be referred to the partial description of the previous embodiments.

[0093] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for optimizing power grid operation and maintenance using machine vision and predictive analysis, characterized in that: The method comprises the following steps: S1: Collect multiple video images of the transmission line to obtain a video set; S2: stitching the collected video images to obtain a corresponding complete panoramic image; S3: inputting the panoramic image into an ice detection model to obtain an ice detection result; S4: Generate operation and maintenance recommendations based on icing detection results; The ice detection model includes a feature extraction module, a feature enhancement module, a feature fusion module, an attention guidance module and a classification module; The feature fusion module uses a cross-layer fusion method to fuse the output features. The cross-layer fusion is used to fuse the output features of the previous layer. m-1 And the next layer output feature F m+1 To perform the fusion: F c =Concat(DwnS(C l (F m-1 )),UpS(C h (F m+1 ))) where F c Output feature F for the previous layer m-1 And the next layer output feature F m+1 The fusion output features, m represents the current layer, Concat represents channel concatenation, DwnS and UpS represent downsampling and upsampling of feature maps respectively, C l and C h Indicates 1*1 dimensionality reduction and dimensionality increase of low-level features and high-level features; The feature enhancement module adopts sub-channel parallel enhancement, the number of sub-channels is 4, and the feature enhancement module outputs the feature F of the previous layer. m-1 Perform channel enhancement and use jump connection to splice F m-1 Get the enhanced feature F m : Among them, Conv represents convolution processing, Act represents activation function, 1*1 and 3*3 represent convolution kernels of size 1 and 3, and d represents expansion size. Represent the output features F of the previous layer respectively m-1 Subchannel 1, subchannel 2, subchannel 3 and subchannel 4 among the four subchannels.

2. The method for optimizing power grid operation and maintenance using machine vision and predictive analysis according to claim 1, characterized in that: The image stitching of the collected video images to obtain the corresponding complete panoramic image specifically includes: S21: According to the flight speed of the drone, key frames are extracted from the video clips to obtain a key frame image set {D k |k∈[1,n]},D k is the kth key frame image, n is the number of extracted key frames, and D1 is preset as the first panoramic image; S22: Identify SIFT feature points in the t-th panoramic image and the t+1-th frame image, where 1≤t≤n-1; S23: Determine the matching feature points in the tth panoramic image and the t+1th frame image according to the k-nearest neighbor algorithm and the Euclidean distance, and obtain the matching feature point set in the tth panoramic image and the set of matching feature points in the t+1th key frame Where N represents the number of matching points. and Respectively represent the i-th matching feature point in the t-th panoramic image and the t+1 key frame image; S24: Obtain the t+1th panoramic image I according to the image conversion formula t+1 ; Among them, tl and tr are the transition area position information, H is the edge position information, γ is the balance weight information, and I t (x, y) is the image information of the t-th panoramic image at coordinate (x, y), D t+1 (x, y) is the image information of the key frame image of the t+1th frame at the coordinate (x, y); S25: Iterate steps S22-S24 to obtain the final panoramic image I n-1 As a complete panoramic image I panorama .

3. The method for optimizing power grid operation and maintenance using machine vision and predictive analysis according to claim 2, characterized in that: The calculation method of the transition area position information tl and tr is: According to the matching feature point set and Obtaining a first conversion matrix T and a second conversion matrix T', wherein the first conversion matrix T is used to convert pixels in the t-th panoramic image to the t+1-th key frame, and the second conversion matrix T' is used to convert pixels in the t+1-th key frame to the t-th panoramic image; For the first boundary B1 of the t-th panoramic image = {(x b ,y b )} Perform the first transformation to obtain the key frame D t+1 The first mapping boundary set E1 in e ,y` e )}, the first conversion calculation formula is: (x` e ,y` e )=T(x b ,y b ) For key frame D t+1 The second boundary B2={(x` b ,y` b )} Perform the second transformation to obtain the key frame D t The second mapping boundary set E2 in e ,y e )}, the second conversion calculation formula is: (x e ,y e )=T`(x` b ,y` b ) Set the first mapping boundary set E1 = {(x` e ,y` e )} in x` e The maximum value of tr is set to the second mapping boundary set E2 = {(x e ,y e )} is the minimum value of tl.

4. The method for optimizing power grid operation and maintenance using machine vision and predictive analysis according to claim 1, characterized in that: The calculation method of the attention guidance module is: in, and is the output feature after attention guidance, σ represents the sigmoid activation function, AVG and MAX represent average pooling and maximum pooling respectively.

5. The method for optimizing power grid operation and maintenance using machine vision and predictive analysis according to claim 1, characterized in that: After the panoramic image is input into the ice detection model to obtain the ice detection result, the wind speed is calculated. The wind speed calculation specifically includes: S31: grayscale the complete panoramic image obtained in step S2 and perform Gaussian filtering, and use the Canny operator to perform edge detection to obtain an edge image I edge , the edge image I edge including transmission line curves; I edge =Canny(GaussianBlur(GrayScale(I panorama ))) Among them, Canny represents the Canny operator, GaussianBlur represents Gaussian filtering, and GrayScale represents grayscale processing; S32: Edge image I edge The endpoints of the transmission line curve are identified to obtain the endpoint coordinates (x1, y1) and (x2, y2), and the midpoint coordinates (x m ,y m ); S33: Connect the endpoint coordinates (x1, y1) and the midpoint coordinates (x m ,y m ) get the first line segment L1, the endpoint coordinates (x2, y2) and the midpoint coordinates (x m ,y m ) obtain the second line segment L2; S34: Calculate the angle θ between the first line segment L1 and the second line segment L2 Where the vector α=(x1-x m , y1-y m ), β=(x2-x m ,y2-y m ), · represents the dot product, || || represents the vector norm; the wind speed v in the transmission line area is calculated based on the angle value: Where, v represents wind speed, C d It represents the resistance coefficient of the transmission line, ρ is the air density, A is the diameter of the transmission line, E represents the elastic modulus of the transmission line material, I represents the cross-sectional moment of inertia of the transmission line, and L represents the length of the transmission line.

6. The method for optimizing power grid operation and maintenance using machine vision and predictive analysis according to claim 1, characterized in that: The feature fusion module cross-layer fuses the output F3 of the third residual module in the backbone network and the output F5 of the fifth residual module to obtain a fusion feature The output F4 of the fourth residual module in the backbone network and the output of the first feature enhancement layer Perform cross-layer fusion to obtain fusion features Fusion Features and And the output of the second feature enhancement layer After attention guidance, it is input into the classification module for target detection.

7. The method for optimizing power grid operation and maintenance using machine vision and predictive analysis according to claim 1, characterized in that: Generating operation and maintenance suggestions according to the ice detection results specifically includes: The ice detection result includes the amount of ice in the complete panoramic image T n and ice position T c , according to the amount of ice T in the complete panoramic image n and the wind speed v in the area of ​​the transmission line determine the number of drones to be dispatched for de-icing; Where T min The minimum amount of ice that triggers the dispatch of UAVs for de-icing, V maxsafe Indicates the maximum wind speed that guarantees the flight of the drone. Indicates rounding up.

8. A system for optimizing power grid operation and maintenance using machine vision and predictive analysis, characterized in that The system includes: Image acquisition device: the image acquisition device acquires multiple video images of the power transmission line to obtain a video set; Panoramic image generation module: the panoramic image generation module performs image stitching on the collected video images to obtain a corresponding complete panoramic image; Icing detection module: The icing detection module inputs the panoramic image into an icing detection model to obtain an icing detection result; Operation and maintenance module: The operation and maintenance module generates operation and maintenance suggestions based on the ice detection results. The ice detection model includes a feature extraction module, a feature enhancement module, a feature fusion module, and an attention guidance module; The feature fusion module uses a cross-layer fusion method to fuse the output features. The cross-layer fusion is used to fuse the output features of the previous layer. m-1 And the next layer output feature F m+1 To perform the fusion: F c =Concat(DwnS(C l (F m-1 )),UpS(C h (F m+1 ))) Among them, F c Output feature F for the previous layer m-1 And the next layer output feature F m+1 The fusion output features, m represents the current layer, Concat represents channel concatenation, DwnS and UpS represent downsampling and upsampling of feature maps respectively, C l and C h Indicates 1*1 dimensionality reduction and dimensionality increase of low-level features and high-level features; The feature enhancement module adopts sub-channel parallel enhancement, the number of sub-channels is 4, and the feature enhancement module outputs the feature F of the previous layer. m-1 Perform channel enhancement and use jump connection to splice F m-1 Get the enhanced feature F m : Among them, Conv represents convolution processing, Act represents activation function, 1*1 and 3*3 represent convolution kernels of size 1 and 3, and d represents expansion size. Represents the output features F of the previous layer respectively m-1 Subchannel 1, subchannel 2, subchannel 3 and subchannel 4 among the four subchannels.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for optimizing power grid operation and maintenance by using machine vision and predictive analysis as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method for optimizing power grid operation and maintenance using machine vision and predictive analysis as claimed in any one of claims 1 to 7.

Images