A method for grid emergency handling of risk assessment

Abstract

The application provides a power grid emergency disposal method for risk assessment, relates to the technical field of image processing, and is used for improving the safe operation ability of a power grid in a mountain environment. The power grid emergency disposal method for risk assessment specifically comprises the following steps: periodically collecting overhead power transmission line video images in a mountain environment, and constructing a power transmission line video set; performing change analysis on the power transmission line video set to obtain an added vegetation coverage area, a vegetation degradation area and a crack propagation area; obtaining the water content change and the area coverage type of each area according to the three areas obtained in step S2; and predicting a mountain landslide risk and generating an emergency disposal scheme according to the water content and the area coverage type of each area. Through the above steps, the application can realize the risk assessment of a mountain landslide, and take early warning and response measures before the risk occurs.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and particularly relates to a power grid emergency disposal method for risk assessment, which is used for improving the safe operation ability of power grid in mountain environment. BACKGROUND

[0002] With the rapid development of tourism, the construction of scenic spot infrastructure is increasingly perfect. As an indispensable important energy network in the scenic spot infrastructure, the power grid plays a crucial role in the daily operation of the scenic spot and the safety experience of tourists. The reliability of the power grid is directly related to the power supply guarantee and operation efficiency of the scenic spot. However, due to the complex terrain and special environment of the mountain scenic spot, the power grid is usually laid in an overhead transmission manner to adapt to the rugged terrain and irregular topography. Compared with the power grid layout in plain areas, the power grid layout in mountain scenic spots has higher cost and greater difficulty in construction, and also faces more technical and economic challenges in operation and maintenance. Although this overhead transmission method solves the limitation of complex terrain on power grid layout to some extent, it also makes the power transmission line more susceptible to external environmental factors, such as natural disasters, external invaders, etc. How to ensure the safe operation of the power grid in the mountain scenic spot has become one of the important problems restricting the sustainable development of the scenic spot.

[0003] In the prior art, the maintenance of the power grid in the mountain scenic spot mainly adopts a post-maintenance mode, that is, the overhead power transmission line is checked and processed by manual inspection. When the power transmission line is affected by external invaders (such as bird nests, garbage randomly discarded by tourists, branches, etc.) or extreme weather (such as ice and snow, strong wind), manual discovery, cleaning and repair are usually relied on. Although this method can solve the problem to some extent, its limitations are obvious. First, manual inspection has a long cycle and cannot discover potential hazards in real time; second, for emergency situations caused by extreme weather or natural disasters (such as heavy rain, snow), the lag of manual maintenance may cause the power grid to be interrupted, and even cause greater economic losses and safety hazards. Especially in the mountain environment, due to the special natural conditions of high altitude, low temperature and strong wind, the power transmission line is exposed to the sun and wind for a long time, and the deterioration speed of the line and the surrounding environment is much higher than that in the plain area. In addition, the complex geological structure of the mountain, such as loose soil and broken rock, easily causes small and large landslides. The impact of such geological disasters on the power grid is often devastating, which may cause the overall collapse of the power transmission line and even large-scale power outages. The existing post-maintenance mode obviously cannot cope with this problem, therefore, a technology and emergency disposal scheme for risk assessment of mountain landslides and other disasters are urgently needed to realize the pre-prediction and prevention of potential risks, so as to ensure the safe operation of the power grid in the mountain scenic spot.

[0004] In recent years, with the rapid development of artificial intelligence technology, especially the wide application of machine vision technology, it has shown mature application value and technical advantages in medical diagnosis, automatic driving, industrial detection and other fields. Through image processing, pattern recognition and deep learning technology, machine vision can efficiently and accurately analyze the dynamic changes in complex scenes, providing important support for the intelligent upgrading of various industries. However, in the field of power grid operation risk assessment in mountainous environment, the application of machine vision technology is still a blank. The power transmission line in the mountain scenic area is often in a steep terrain and harsh environment, and the traditional monitoring and evaluation method cannot meet the requirements of efficiency and accuracy. By collecting image data of mountainous environment and intelligent analysis, the risk of geological disasters such as mountain landslide can be identified and dynamically monitored in early stage, so as to realize the change from 'after-maintenance' to 'pre-prediction'.

[0005] Based on this, the present application provides a power grid emergency disposal method for risk assessment, which evaluates the risk of mountain landslide and takes early warning and response measures before the risk occurs. SUMMARY

[0006] The present application provides a power grid emergency disposal method for risk assessment, which specifically comprises the following steps:

[0007] S1: periodically collecting overhead transmission line video images in mountainous environment, and constructing a transmission line video set;

[0008] S2: performing change analysis on the transmission line video set to obtain new vegetation coverage area, vegetation degradation area and crack extension area;

[0009] S3: obtaining the water content change and area coverage type of each area according to the three areas obtained in step S2;

[0010] S4: predicting the risk of mountain landslide according to the water content and area coverage type of each area and generating an emergency disposal scheme.

[0011] The present application provides a power grid emergency disposal system for risk assessment, which comprises:

[0012] Image acquisition device: the image acquisition device periodically collects overhead transmission line video images in mountainous environment, and constructs a transmission line video set;

[0013] Change analysis module: the change analysis module performs change analysis on the transmission line video set to obtain new vegetation coverage area, vegetation degradation area and crack extension area;

[0014] Water content analysis module: the water content analysis module obtains the water content change and area coverage type of each area according to the three areas obtained.

[0015] The comprehensive judgment module predicts the mountain landslide risk according to the water content of each region and the region coverage type and generates an emergency treatment scheme.

[0016] An electronic device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the power grid emergency treatment method for risk assessment when executing the computer program.

[0017] A computer readable storage medium stores a computer program, and the computer program implements the power grid emergency treatment method for risk assessment when executed by a processor.

[0018] Compared with the prior art, the power grid emergency treatment method for risk assessment identifies the risk area through image detection technology, enhances and weakens the extracted features during the detection process, extracts potential different features, enhances the connection between shallow and deep features, and guides target segmentation; based on the risk area, the type and water content of the risk area are identified, and through the comprehensive judgment of the two key factors, the high-risk area around the power transmission line where the mountain landslide may occur can be identified, and scientific early warning basis can be provided for the relevant departments. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0020] Figure 1 It is a network structure diagram in the present application. DETAILED DESCRIPTION

[0021] The embodiments of the present application will be described in detail below with reference to the drawings.

[0022] Following, the embodiments of the present application are described through specific examples, and other advantages and effects of the present application can be easily understood by those skilled in the art from the disclosure. 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 embodiments, and various modifications or changes can be made to the details in the specification without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0023] It should be noted that the various aspects of the embodiments described below are within the scope of the appended claims. It should be apparent that the aspects described herein can be embodied in a wide variety of forms and that any specific structure and / or function described herein is merely illustrative. Based on the teachings herein one skilled in the art should appreciate that an aspect described herein can be implemented independently of any other aspects and that an aspect can be implemented both as any number of software and / or hardware structures and as any number of combinations of software and / or hardware structures. For example, an apparatus can be implemented as any number and combination of the aspects described herein. Additionally, the features described herein can be implemented as part of one or more separate devices or components.

[0024] In addition, in the following description, specific details are provided to facilitate thorough understanding of examples. However, one skilled in the art will understand that the examples can be practiced without these specific details.

[0025] The embodiments of the present application provide a power grid emergency disposal method for risk assessment, which specifically comprises the following steps:

[0026] S1: periodically collecting video images of overhead transmission lines under mountain environment, and constructing a transmission line video set;

[0027] S2: performing change analysis on the transmission line video set to obtain newly added vegetation coverage area, vegetation degradation area and crack propagation area;

[0028] S3: obtaining water content change and area coverage type of each area according to the three areas obtained in step S2;

[0029] S4: predicting mountain landslide risk and generating an emergency disposal scheme according to the water content and area coverage type of each area.

[0030] In one embodiment, video images of overhead transmission lines under mountain environment are periodically collected, and a transmission line video set is constructed.

[0031] The collected environment and video target are mountain overhead transmission lines. Before video collection, the length of the transmission line area needs to be determined. The transmission line passes through multiple mountains, and some areas may be eroded by strong winds all year round, and the stability of the mountains is poor, especially prone to landslides. In order to determine the risk of landslides in complex environments, the collection period is determined to be once a week. The selection of a weekly cycle is mainly based on the characteristics of geological changes in this mountain environment, avoiding missing the details of geological changes caused by too long data collection period, and taking into account the inspection efficiency and resource utilization. At the same time, each collection task is carried out as much as possible in weather conditions without rain and strong wind to ensure the flight stability of the unmanned aerial vehicle and the image collection quality. In addition, due to frequent strong winds in mountainous areas and high wind speed, weather conditions need to be strictly monitored, and collection tasks are carried out in the morning or afternoon when the wind speed is low.

[0032] The implementation of the collection task is designed with the core idea of segmented inspection of key areas. Specifically, the transmission line is divided into multiple continuous key inspection areas, each area containing multiple transmission towers and mountains along the line. The unmanned aerial vehicle needs to complete the inspection of all areas in the order in the task, and the flight path of each area remains fixed to ensure that the image data at different time periods can correspond to the same geographical location. For the flight path design of each section of the transmission line, the transmission tower is preferred as the node, and the straight flight mode from one transmission tower to the next transmission tower is adopted, while covering the mountain areas that may have landslides above and around the line, such as steep slopes, rock exposed fault zones, etc.

[0033] The unmanned aerial vehicle equipment is preferably an industrial model, such as the Autel EVO Max series, which has strong wind resistance and long endurance, suitable for working in high wind speed environments. The flight height is set according to the location of the transmission network and the height of the vegetation to ensure safety, while avoiding interference from mountain vegetation or sudden changes in wind speed. Combined with the undulating characteristics of the mountain environment, the unmanned aerial vehicle can dynamically adjust the flight height through the built-in laser radar module to maintain an appropriate relative distance from the mountain surface, thereby obtaining clear mountain detail images. Set a fixed flight speed to balance the picture stability and collection efficiency to ensure that a single endurance can complete a collection task in a key area.

[0034] The unmanned aerial vehicle device is also equipped with a multispectral camera and a high-precision inertial navigation system, can simultaneously collect visible light and multispectral image data, and provides a comprehensive analysis basis for mountain environment and vegetation coverage. During flight, the multispectral image is transmitted to the ground monitoring station in real time, and the power grid equipment inspection personnel can check whether the data is complete according to the quality of each spectral image in real time. If the vegetation reflectivity data is missing, the image contrast is poor, or some areas are not covered, the flight route or parameters can be immediately adjusted for supplementary sampling. Since strong winds or air pressure fluctuations often occur along high mountains, the unmanned aerial vehicle needs to have a stable spectral data collection function, automatically correct the image offset caused by wind speed changes, and thus ensure the clarity and accuracy of the multispectral data.

[0035] After the collection is completed, the collected visible light video images are grouped according to the paragraphs of the power transmission line. Finally, a plurality of grouped visible light video images under multiple periods are obtained, and a power transmission line video set is constructed, wherein each sample in the power transmission line video set is a plurality of video clips collected under multiple periods for the same paragraph of the power transmission line, that is, a plurality of visible light video images collected at different times for the same power transmission line.

[0036] In addition, the present application also groups and aligns the simultaneously collected multispectral image data with the corresponding video frames in the visible light video images, to obtain a plurality of grouped multispectral image sets under multiple periods.

[0037] In order to realize the emergency disposal of the power grid, the present application also needs to analyze the changes of the power transmission line video set;

[0038] The change analysis is realized by a change region segmentation network, and the change region segmentation network is composed of an encoding module, a multiscale fusion module and a decoding module.

[0039] The encoding module is composed of three residual networks connected in sequence, and the residual network is composed of a skip connection network and a feature superposition network connected in sequence.

[0040] The skip connection network includes an input layer, an intermediate layer and an output layer, wherein the input layer and the output layer are each provided with a skip connection; the input layer, the intermediate layer and the output layer are defined as follows:

[0041]

[0042] Wherein, input and output of the input layer, represent the output of the intermediate layer and the output layer, respectively, Conv(), Conv 2 represent one and two convolution calculations, respectively, represent two separable convolution calculations;

[0043] The feature superposition network is composed of an enhancement superposition network and a weakening superposition network, wherein the enhancement superposition network is used to enhance the local detail features extracted in the feature map, and the enhanced local detail features are used to guide the segmentation in the decoding module, and the enhancement superposition network is defined as:

[0044]

[0045] wherein, represents the input feature map of the enhancement superposition network, represents the input feature map the first, second, and k regions after extracting k regions in the width direction, and the k region feature maps are constructed into a region vector σ is an activation function, F EN is the output of the enhancement superposition network;

[0046] The weakening superposition is used to weaken the extracted local detail features, so as to prevent the feature map process of the next feature extraction layer from excessively relying on the local detail features extracted by the previous feature extraction layer. In this way, the three residual networks and the subsequent residual networks are not excessively affected by the feature extraction of the previous residual network, which is beneficial to extract more regional detail features. The weakening superposition network is defined as:

[0047]

[0048] wherein, F DE is the output of the weakening superposition network; F DE calculated by each previous residual network is used as the input feature map of the input layer of the skip connection network in the subsequent residual network, and the input feature map of each feature superposition network is the output feature map of the output layer of the skip connection network in the residual network to which the feature superposition network belongs The three outputs of the enhancement superposition network in the three residual networks are used as the input feature maps of the multi-scale fusion module, and are defined as: represents the three outputs of the enhancement superposition network;

[0049] In the process of feature extraction by a neural network, the expression ability of the features gradually transits from shallow texture features to deep semantic features with the deepening of the network layers. The extraction of deep features largely depends on the basis provided by shallow features, and thus the quality of shallow features directly affects the expression of deep features. However, the shallow features are not always the optimal feature basis, and they may interfere with the extraction of deep features due to containing too much redundant information or irrelevant content, thereby limiting the accurate recognition ability of the network to the target. To solve this problem, the present application proposes an innovative feature processing method, which differentiates the extracted features by designing a double-channel structure to optimize the transmission and utilization of the features. Specifically, the method processes the extracted features through two independent channels: in the first channel, the features are enhanced to make them more prominent and clear, so as to provide more accurate feature expression for the subsequent target detection module and improve the accuracy of target detection; while in the second channel, the extracted features are weakened, and the weakened features are superimposed back into the original feature map. In this way, through the processing of the second channel, the interference of shallow features on the extraction of deep features can be effectively suppressed, forcing the deep neural network to mine other potential and more discriminative deep features without relying on shallow features, thereby realizing the capture of more rich semantic information of the target. The core of this double-channel structure design lies in balancing the relevance and independence between shallow features and deep features: on the one hand, the shallow features are enhanced to provide stronger support for the extraction of deep features, and on the other hand, the influence of shallow features is weakened to create conditions for the independent extraction of deep features.

[0050] The multi-scale fusion module includes three parallel fusion networks, wherein the input of the first fusion network is and The input of the second fusion network is and The input of the third fusion network is and

[0051] The fusion process includes:

[0052] Step (1): Calculate the similarity feature map F of each fusion network input feature map and Each element of the similarity feature map F is defined as: s s

[0053]

[0054] wherein, ​​is the pixel value of the first input feature map of the fusion network at (c, u, v), the size of the first input feature map is Channel * W1 * H1, is the pixel value of the second input feature map of the fusion network at (c, x, y), the size of the second input feature map is Channel * W2 * H2, the similarity feature map F s is of size 1 * (W1 * H1) * (W2 * H2);

[0055] Step (2): Based on the similarity feature map F s , the input and is enhanced to obtain the similarity enhanced output and The similarity feature enhancement calculation is:

[0056]

[0057]

[0058] wherein, and represent the enhanced input feature map;

[0059] Step (3): Based on the similarity enhanced output obtained by each fusion network, the fusion is carried out:

[0060]

[0061]

[0062]

[0063]

[0064] wherein, and respectively represent and the output feature enhanced by the first fusion network through the similarity feature, and respectively represent and the output feature enhanced by the second fusion network through the similarity feature, and respectively represent and the output feature enhanced by the third fusion network through the similarity feature;

[0065] In the feature map fusion process, to more efficiently mine and utilize key information from feature maps of different depths, this invention proposes a fusion branch design. This design constructs three branches, each taking three feature maps of different depths as input. Each branch selects two feature maps as input and calculates their similarity. Through similarity calculation, identical feature parts can be extracted from feature maps of different depths. These similar features reflect information that remains stably preserved during multi-layer feature extraction and layer-by-layer feature suppression. This stable presence of similar features indicates their high correlation and importance in the multi-layer representation of the network, and therefore can be considered key features that need to be focused on in object detection tasks. Based on this analysis, this invention further utilizes these similar features to enhance two related feature maps, ensuring that this key information is more prominently expressed. Subsequently, by fusing the enhanced feature maps, the overall expressive power of the features can be significantly improved during the fusion process, providing more effective input support for subsequent object detection.

[0066] The decoding module includes four upsampling stages:

[0067]

[0068] Where i∈{1,2,3}, Conv_De represents deconvolution computation. and F represents the deconvolutional feature map and upsampled feature map of the i-th layer of the decoding module. out This represents the output image that segments out the vegetation cover area, the vegetation degradation area, and the crack extension area.

[0069] Based on the three regions obtained in step S2, obtain the water content changes and regional cover types for each region:

[0070] Based on the segmented vegetation degradation area and vegetation coverage area, set up a vegetation degradation area mask and a vegetation coverage area mask, and use the above masks to crop the visible light image and multispectral image corresponding to the vegetation degradation area, as well as the visible light image and multispectral image corresponding to the vegetation coverage area.

[0071] The soil type of the vegetation degradation area is determined by the visible light image corresponding to the vegetation degradation area, and the water content change of the vegetation degradation area is calculated by the multispectral image corresponding to the vegetation degradation area.

[0072] The vegetation type of the vegetation-covered area is determined by the visible light image corresponding to the vegetation-covered area, and the water content change of the degraded area is calculated by the multispectral image.

[0073] In the process of determining the soil type of the vegetation degradation area based on the visible light image corresponding to the vegetation degradation area, the soil type is sandy soil, loam, and clay. Image features including color and texture features are used, and support vector machine is used as the classification model.

[0074] The water content changes in vegetation-degraded and vegetation-covered areas were calculated based on multispectral images. NDWI and NDVI were calculated from the multispectral images.

[0075]

[0076] Moisture content W T =α·NDWI 2 +β·NDVI+γ;

[0077] Where, ρ green ρ NIR ρ RED The values ​​represent the green light band, near-infrared band, and red light band, respectively. α, β, and γ are weighting parameters, with α ranging from 0.6 to 0.8 and β ranging from 0.2 to 0.4. The NDWI of soil and vegetation is between [-1, 1]. The higher the moisture content, the closer the NDWI is to 1, and vice versa. Considering the saturation effect of NDWI, the sensitivity to moisture decreases when the NDWI is close to 1. In vegetation-covered areas, the higher the NDVI, the healthier the vegetation, which indirectly reflects sufficient soil moisture.

[0078] Calculate the difference in water content between the first and last sampling times in the target area to obtain the change in water content in the target area.

[0079] In the process of determining the vegetation type of the vegetation cover area based on the visible light image corresponding to the vegetation cover area, the soil type is grassland, shrub and tree. The image features used include RGB mean sum method, normalized color index and HIS color features, and random forest is used as the classification model.

[0080] Grassland generally presents a relatively uniform green color with high green channel (G) reflectance, low RGB mean, and consistent hue. Grassland hue is concentrated in the green area, with high saturation and high, uniform brightness. Shrubs have high green reflectance, but due to dense vegetation and complex leaf structure, their RGB mean and variance are larger than grassland. Furthermore, shrubs have higher saturation and slightly lower brightness. Trees exhibit greater color variation (canopy shadows mix with leaves), with a larger RGB variance. Additionally, the brightness distribution of trees is uneven, and canopy shadows cause significant variations in saturation and brightness.

[0081] Predict landslide risks and generate emergency response plans based on the water content and cover type of each region.

[0082] Landslide risk: R_total = w1·R soil +w2·R veg +w3·R crack ;

[0083] w1, w2, and w3 are the weighted scores for the sub-line, respectively; R_total takes a value between 0 and 1.

[0084] For low-risk areas with R total < 0.4, continuously monitor water content and crack dynamics, and conduct weekly inspections of power grid safety and update data.

[0085] For medium-risk areas with 0.4 ≤ R total < 0.8, monitor the water content and crack expansion every three days and notify the power grid operation and maintenance department to start backup circuit dispatch preparation.

[0086] For high-risk areas with R total ≥ 0.8, monitor the moisture content and crack expansion daily, urgently dredge the drainage system to reduce soil moisture content, set up retaining walls or landslide protection nets in crack or landslide warning areas, and immediately activate backup circuits and switch power supply lines to ensure power supply.

[0087]

[0088] ΔW soil W represents the change in water content in areas of vegetation degradation. soil_critical This indicates the critical values ​​for moisture content changes corresponding to different soil types (80% for sandy soil, 70% for loam, and 60% for clay).

[0089]

[0090] ΔW veg W represents the change in water content in the vegetation cover area. veg_critical This represents the critical value for water content change corresponding to the vegetation cover area (the soil type of the vegetation cover area is usually loam or clay, and the average value W is taken). veg_critical =65%),

[0091] S veg Indicates the stability score of vegetation type, grassland: S veg =0.3, Shrubs: S veg =0.6, Trees: S veg =0.9; S max The highest stability score for vegetation type (taken as S) max =0.9 corresponds to trees);

[0092] When vegetation type is unstable (e.g., grassland) and water content varies greatly, R veg→1 (High Risk), when the vegetation type is highly stable (e.g., trees) and the water content changes little, R veg →0 (Low risk);

[0093]

[0094] Where Length and Width are the longest and widest values ​​of the crack propagation region, respectively; L_max and W_max represent the crack propagation warning length and width; w_L and w_W represent the weighting parameters for length and width; and R... crack The value ranges from 0 to 1;

[0095] To accurately predict whether landslides will occur near power transmission lines, this invention fully considers two key natural influencing factors of landslides—mountain type and water content—and assesses landslide risk through phased detection and analysis. First, this invention utilizes visible light video images to monitor the area surrounding the power transmission line, focusing on analyzing changes in vegetation cover, the degree of vegetation degradation, and the distribution characteristics of new cracks over a period of time. By detecting these surface features, potential risk areas can be preliminarily identified, and the landform features of the mountains can be combined to classify their types. This stage of detection not only quickly identifies areas with potential landslide risk but also provides geographical and geological background information for subsequent in-depth analysis.

[0096] After completing the initial risk identification in the first phase, this invention further utilizes multispectral images to detect the water content of the aforementioned potential risk areas. Multispectral images, with their high spectral resolution, can sensitively capture the distribution of moisture content in soil and rock masses. Through reflectance analysis in specific bands, the water content level of the risk areas can be accurately assessed. Water content is a significant contributing factor to landslides; therefore, this invention, through quantitative detection of water content, further refines the assessment of potential landslide risk areas.

[0097] Based on the detection results from the two stages mentioned above, this invention combines mountain type and water content information to construct a landslide risk prediction model. Mountain type determines the geological characteristics and stability of a mountain, while water content reflects the dynamic impact of external environmental changes on mountain stability. By comprehensively judging these two key factors, this invention can identify high-risk areas around power transmission lines where landslides may occur and provide relevant departments with a scientific basis for early warning.

[0098] This invention provides a power grid emergency response system for risk assessment, the system comprising:

[0099] Image acquisition equipment: The image acquisition equipment periodically acquires video images of overhead power transmission lines in the mountainous environment and constructs a video set of the power transmission lines;

[0100] Change Analysis Module: The change analysis module performs change analysis on the transmission line video set to obtain areas of newly added vegetation coverage, areas of vegetation degradation, and areas of crack extension;

[0101] Moisture content analysis module: The moisture content analysis module obtains the moisture content changes and regional coverage types of each of the three regions.

[0102] Comprehensive Judgment Module: The comprehensive judgment module predicts the risk of landslides and generates emergency response plans based on the water content and regional cover type of each area.

[0103] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned power grid emergency response method for risk assessment.

[0104] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned power grid emergency response method for risk assessment.

[0105] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, 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. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0106] In this specification, the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the descriptions of the embodiments described later are relatively simple, and relevant parts can be referred to the descriptions of the foregoing embodiments.

[0107] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A power grid emergency response method for risk assessment, characterized in that, The method specifically includes the following steps: S1: Regularly collect video images of overhead power transmission lines in mountainous environments and construct a video collection of power transmission lines; S2: Perform change analysis on the video set of the transmission line to obtain the newly added vegetation coverage area, vegetation degradation area, and crack extension area; S3: Based on the three regions obtained in step S2, obtain the water content changes and regional coverage types of each region; S4: Predict landslide risks and generate emergency response plans based on the water content and cover type of each region; The soil type of the vegetation degradation area is determined by the visible light image corresponding to the vegetation degradation area, and the water content change of the vegetation degradation area is calculated by the multispectral image corresponding to the vegetation degradation area. The vegetation type of the vegetation cover area is determined by the visible light image corresponding to the vegetation cover area, and the water content change of the degraded area is calculated by the multispectral image. In the process of determining the soil type of the vegetation degradation area based on the visible light image corresponding to the vegetation degradation area, the soil type is sandy soil, loam, and clay. Image features including color and texture features are used, and support vector machine is used as the classification model. The water content changes in vegetation-degraded and vegetation-covered areas were calculated based on multispectral images. NDWI and NDVI were calculated from the multispectral images. Moisture content W T =α·NDWI 2 +β·NDVI+γ; Where, ρ green ρ NIR ρ RED The green light band, near-infrared band, and red light band are represented respectively. α, β, and γ are weighting parameters. The difference in water content in the target area at the first and last collection times is calculated to obtain the change in water content in the target area. In the process of determining the vegetation type of the vegetation coverage area based on the visible light image corresponding to the vegetation coverage area, the vegetation type is grassland, shrub and tree. The image features used include RGB mean sum method, normalized color index and HIS color features, and random forest is used as the classification model. Landslide risk: R_total = w1·R soil +w2·R veg +w3·R crack ; w1, w2, and w3 are risk weight scores, respectively; For low-risk areas where R_totall < 0.4, continuously monitor water content and crack dynamics, and conduct weekly inspections of power grid safety and update data. For medium-risk areas with a water content of 0.4 ≤ R_totall < 0.8, monitor the water content and crack expansion every three days and notify the power grid operation and maintenance department to start backup circuit scheduling preparations. For high-risk areas with R_totall≥0.8, monitor the moisture content and crack expansion daily, urgently dredge the drainage system to reduce soil moisture content, set up retaining walls or landslide protection nets in crack or landslide warning areas, and immediately start the backup circuit and switch the power supply line to ensure power supply. ΔW soil W represents the change in water content in areas of vegetation degradation. soil_critical This represents the critical value for moisture content change corresponding to a soil type. ΔW veg W represents the change in water content in the vegetation cover area. veg_critical S represents the critical value for water content change corresponding to the vegetation cover area; veg S represents the stability score of vegetation type; max The highest stability score representing the vegetation type; Length and Width are the longest and widest values ​​of the crack extension region, respectively. L_max and W_max represent the crack extension warning length and width, and w_L and w_W represent the weighting parameters of length and width.

2. The power grid emergency response method for risk assessment according to claim 1, characterized in that, The change analysis is achieved through a change region segmentation network, which consists of an encoding module, a multi-scale fusion module, and a decoding module.

3. The power grid emergency response method for risk assessment according to claim 2, characterized in that, The encoding module consists of three sequentially connected residual networks. Each residual network is composed of a skip connection network and a feature stacking network connected sequentially. The feature stacking network consists of an enhancement stacking network and a reduction stacking network. The enhancement stacking network is used to enhance the local detail features extracted from the feature map. The enhanced local detail features are used to guide segmentation in the decoding module. The enhancement stacking network is defined as follows: in, This represents the input feature map of the augmentation overlay network. This indicates that the input feature map will be used. After extracting k regions along the width direction, the 1st, 2nd, and kth regions are used to construct a region vector from the feature maps of the k regions. σ is the activation function, F EN To enhance the output of the overlay network.

4. The power grid emergency response method for risk assessment according to claim 3, characterized in that: The weakening superposition network is defined as: Among them, F DE To reduce the output of the superimposed network.

5. A power grid emergency response method for risk assessment according to claim 2, characterized in that: The multi-scale fusion module includes three parallel fusion networks, wherein the input of the first fusion network is... and The input to the second fusion network is and The input to the third fusion network is and 6. A power grid emergency response method for risk assessment according to claim 5, characterized in that: The fusion process includes: Step (1): Calculate the input feature map for each fusion network and Similarity feature map F s The similarity feature map F s Each element is defined as: in, This is the pixel value at (c,u,v) of the first input feature map of the fusion network. The scale of the first input feature map is Channel*W1*H1. This is the pixel value at (c,x,y) of the second input feature map of the fusion network. The scale of the second input feature map is Channel*W2*H2, and the similarity feature map F... s The scale is 1*(W1*H1)*(W2*H2); Step (2): Based on similarity feature map F s , for input and Similarity feature enhancement is performed to obtain similarity-enhanced output. and The similarity feature enhancement calculation is as follows: Step (3): Perform fusion based on the similar enhancement outputs obtained from each fusion network: in, and They represent and The output features are enhanced by the similarity features of the first fusion network. and They represent and The output features are enhanced by the similarity features of the second fusion network. and They represent and The output features are enhanced by similarity features from a third fusion network.

7. A power grid emergency response system for risk assessment, used to execute a power grid emergency response method for risk assessment as described in any one of claims 1-6, characterized in that... The system includes: an image acquisition device: the image acquisition device periodically acquires video images of overhead power transmission lines in the mountain environment and constructs a video set of power transmission lines; a change analysis module: the change analysis module performs change analysis on the video set of power transmission lines to obtain areas of newly added vegetation coverage, areas of vegetation degradation, and areas of crack extension; Moisture content analysis module: The moisture content analysis module obtains the moisture content changes and regional coverage types of each of the three regions. Comprehensive Judgment Module: The comprehensive judgment module predicts the risk of landslides and generates emergency response plans based on the water content and regional cover type of each area.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements a power grid emergency response method for risk assessment as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program that, when executed by a processor, implements a power grid emergency response method for risk assessment as described in any one of claims 1 to 6.

Images