Multi-source big data-based external damage risk grade dynamic evaluation method, equipment and device

Abstract

The invention discloses a multi-source big data-based power transmission channel external rupture risk grade dynamic evaluation method, equipment and device, and the method comprises the steps: extracting external rupture pattern spots of a power transmission channel through a target detection algorithm, and carrying out the recognition and classification of each external rupture type through combining the morphological features and spectral features of each external rupture pattern spot. According to the method, the influence factors of each external breaking type risk are sorted, the external breaking type, the external breaking area, the distance from the external breaking to the tower and the wind power and wind direction factors are combined, the difference of each influence factor under different external breaking types is considered, accurate weighting is carried out, and risk assignment of each external breaking pattern spot is realized through multi-dimensional analysis. And carrying out normalization processing on risk values of various external damage pattern spots, scientifically dividing risk levels, and realizing accurate identification and dynamic monitoring of the external damage risk of the power transmission channel. According to the method, the scientificity and accuracy of risk assessment can be effectively improved, and powerful technical support is provided for safe operation and maintenance of the power transmission line.

Description

Technical Field

[0001] The present invention relates to the technical field of power facility safety monitoring, and in particular to a method, equipment and device for dynamically evaluating external damage risk levels based on multi-source big data. Background Art

[0002] With the deepening development of smart grids, the safe operation of transmission lines, as core infrastructure of the power system, is directly related to the reliability of energy supply. In recent years, the risk of external damage to transmission lines has increased, mainly due to sudden threats such as collisions with construction machinery, floating objects getting caught on power lines, landslides, and fires.

[0003] Traditional risk assessment methods rely primarily on manual inspections and empirical judgment, and suffer from significant technical bottlenecks:

[0004] Current assessment systems often focus on single risk factors (such as wind speed and distance from the tower), lacking differentiated analysis of the specific impacts of external damage types. Wind direction, a core parameter affecting external damage risk, is not systematically incorporated into existing models, weakening the accuracy of assessments. Existing models use fixed weight coefficients, failing to reflect the dynamic correlation between various influencing factors under different external damage types.

[0005] The above-mentioned defects result in the existing assessment methods being insufficiently accurate and limited in applicability, making it difficult to support active prevention and control of transmission channels. Summary of the Invention

[0006] To solve the above-mentioned problems, the present invention provides a dynamic assessment method, equipment and device for external damage risk level based on multi-source big data, which realizes accurate risk assessment and real-time warning through multi-dimensional factor fusion, dynamic weight allocation and normalization processing.

[0007] To achieve the above objectives, the present invention provides a method for dynamically assessing the risk level of power transmission channel external damage based on multi-source big data, the method comprising:

[0008] Acquire remote sensing image data along the transmission channel, identify external damage spots in the image based on a target detection algorithm, and identify the external damage type using a pre-trained external damage type classification model based on the morphological and spectral characteristics of the external damage spots;

[0009] Extracting the mask area of ​​each external broken patch through a semantic segmentation network, calculating the external broken area of ​​each external broken patch based on the mask area, and calculating the Euclidean distance between each external broken patch and the nearest transmission tower based on the geographic coordinates;

[0010] Dynamically obtain real-time wind speed and direction data in the area where the transmission channel is located;

[0011] For each external damage pattern, a preset weight matrix is ​​matched according to the external damage type of each external damage pattern. The weight matrix includes weight distribution values ​​of five dimensions: external damage type, external damage area, Euclidean distance, wind force data and wind direction data;

[0012] Based on the weight matrix, a multi-dimensional risk assessment equation is established to calculate the risk value of each external broken pattern;

[0013] The risk values ​​of all external broken spots are normalized using the range method, and the normalized risk values ​​of each external broken spot are mapped to the preset risk level range to determine the risk level of each external broken spot;

[0014] The external damage location of the transmission channel and the corresponding risk level are displayed in real time through a visual interface.

[0015] Furthermore, the target detection algorithm adopts the YOLOv7 model, the external damage type classification model is based on ResNet50 network training, and the input features include the morphological features and spectral features of the external damage spots. The extraction of the input features includes:

[0016] Morphological feature extraction: The edge detection algorithm is used to obtain the contour information of the outer broken patch area, and the geometric parameters of the outer broken patch contour are calculated as morphological features;

[0017] Spectral feature extraction: The reflectance data of the visible light band and infrared band of the remote sensing image are analyzed for differences, and the normalized vegetation index and thermal radiation anomaly are extracted as spectral features.

[0018] Furthermore, the external damage type is identified by the external damage type classification model, and the specific method includes:

[0019] The extracted morphological features and spectral features are fused to form a fused feature vector;

[0020] Inputting the fused feature vector into the feature extraction layer of the ResNet50 network for feature extraction, and enhancing the deep feature expression through the residual connection layer of the ResNet50 network;

[0021] A fully connected layer and a Softmax classifier are connected to the end of the ResNet50 network to output the probability distribution of the external damage pattern belonging to each preset external damage type; the preset external damage types include at least one or more of construction, greenhouse film, color steel tile house, pond and natural collapse;

[0022] Based on the output probability distribution, determining the external destruction type with the highest probability as the preliminary classification result of the external destruction pattern;

[0023] The preliminary classification results are post-processed by the non-maximum suppression algorithm to eliminate the misjudgment of the external damage type in the overlapping area, and the external damage pattern detection frame with the highest confidence and the corresponding external damage type are retained as the final recognition result.

[0024] Furthermore, the weight matrix is ​​a five-dimensional vector, and each dimension corresponds to the weight distribution value of different influencing factors, specifically including:

[0025] External damage type weight: Determined based on the actual threat level of the external damage type to the transmission channel. The higher the threat level, the greater the corresponding weight value.

[0026] External break area weight: positively correlated with the mask area of ​​the external break pattern. The larger the external break area, the greater the corresponding weight value.

[0027] Euclidean distance weight: Based on the Euclidean distance between the outer broken pattern and the nearest transmission tower, the closer the distance between the outer broken pattern and the transmission tower, the greater the corresponding weight value;

[0028] Wind data weight: positively correlated with the real-time wind value. The larger the real-time wind value, the greater the corresponding weight.

[0029] Wind direction data weight: The value is assigned based on the relative relationship between wind direction and external breaking position. When the wind direction is toward the transmission channel, the corresponding weight value is the largest, and when it is away from the transmission channel, the corresponding weight value is the lowest.

[0030] Furthermore, the multidimensional risk assessment equation is constructed as:

[0031]

[0032] In formula (1), R represents the risk value of the external broken pattern, W i represents the normalized weight value of the i-th dimension in the weight matrix, i = 1, 2, 3, 4, 5 correspond to the type of external damage, the area of ​​external damage, the Euclidean distance, the wind force data and the wind direction data respectively; X i Represents the standardized score value of the i-th dimension.

[0033] Furthermore, the risk level of each external broken pattern spot is determined by the following specific method:

[0034] Based on the statistical distribution characteristics of the normalized risk value, cluster analysis is used to dynamically determine the risk value interval boundaries to form at least three increasing risk level intervals;

[0035] Each risk value interval corresponds to a preset risk level, and the higher the risk value interval, the higher the risk level;

[0036] The normalized risk value of each external broken pattern spot is compared with the preset risk level range to determine the risk level of each external broken pattern spot and associate the corresponding external broken pattern spot position data.

[0037] In addition, to achieve the above-mentioned purpose, the present invention also proposes a device for dynamically evaluating the risk level of power transmission channel external damage based on multi-source big data, the device comprising:

[0038] Multi-source data acquisition module, used to simultaneously acquire remote sensing image data, meteorological monitoring data and geospatial data along the transmission corridor;

[0039] The external damage identification module is in communication with the multi-source data acquisition module and includes:

[0040] Image preprocessing unit, which performs denoising, enhancement and radiation correction preprocessing on the acquired remote sensing images;

[0041] The target detection unit identifies external broken spots in the image based on the target detection algorithm;

[0042] A classification unit is used to identify the type of external damage based on the morphological and spectral characteristics of the external damage pattern using a pre-trained external damage type classification model;

[0043] A feature extraction module, which is in communication with the external damage identification module, includes:

[0044] The mask generation unit is used to extract the mask area of ​​the external broken image patch through the semantic segmentation network;

[0045] An area calculation unit, which calculates the outer broken area based on the mask area;

[0046] A distance calculation unit calculates the Euclidean distance between the outer broken pattern and the nearest transmission tower based on geographic coordinates;

[0047] A weight matching module is connected to the external damage identification module, the feature extraction module and the multi-source data acquisition module respectively, and is used to match a preset weight matrix according to the external damage type. The weight matrix includes weight distribution values ​​of five dimensions: external damage type, external damage area, Euclidean distance, wind speed data and wind direction data;

[0048] a risk assessment module, communicating with the weight matching module, establishing a multi-dimensional risk assessment equation based on the weight matrix, and calculating the risk value of each external broken pattern spot;

[0049] A risk grading module, in communication with the risk assessment module, includes:

[0050] Normalization unit, using the range method to normalize the risk value;

[0051] A risk level mapping unit is used to map the normalized risk value to a preset risk level range and output the risk level;

[0052] The visual warning module is in communication with the risk grading module and is used to display the location of the external damage of the power transmission channel and the corresponding risk level in real time.

[0053] In addition, to achieve the above-mentioned purpose, the present invention also provides a dynamic assessment device for the risk level of external damage of a power transmission channel based on multi-source big data, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement a method for dynamic assessment device for the risk level of external damage of a power transmission channel based on multi-source big data as described above.

[0054] The beneficial effects of the present invention are:

[0055] The present invention constructs a multidimensional risk assessment model by fusing the morphological and spectral features of remote sensing images, combining real-time meteorological data and geographic spatial information, and dynamically matching differentiated weight matrices for different types of external damage, significantly improving the accuracy of risk assessment for different types of external damage.

[0056] The present invention introduces a quantification mechanism for the relative relationship between wind direction and external damage position. By real-time monitoring of the relative orientation of wind direction and external damage position, the contribution of parameters such as wind speed, area, and distance are dynamically adjusted, thereby achieving accurate risk quantification in different scenarios and overcoming the defect of traditional models that ignore wind direction parameters.

[0057] Based on the GIS platform, real-time visualization of the location and risk level of external damage in transmission channels is achieved, supporting simultaneous access by multiple terminals. The area of ​​external damage patches is automatically extracted through the semantic segmentation network. Combined with Euclidean distance calculation and terrain stability analysis, an integrated air-space-ground monitoring network is constructed to meet the needs of refined prevention and control in complex terrain.

[0058] The modular design supports dynamic updates and expansions of algorithm models. For example, when adding new types of external damage, such as ice disasters and fires, only the classification model parameters need to be iterated without reconstructing the entire architecture, meeting the real-time monitoring needs of large-scale transmission networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a flow chart of a method for dynamically assessing the risk level of power transmission channel external damage based on multi-source big data;

[0060] Figure 2 This is a structural diagram of an embodiment of a device for dynamically evaluating the risk level of external damage to a power transmission channel based on multi-source big data. DETAILED DESCRIPTION

[0061] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0062] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0063] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0064] Example 1

[0065] Reference Figure 1 A method for dynamically assessing the risk level of power transmission channel external damage based on multi-source big data, the method comprising:

[0066] Acquire remote sensing image data along the transmission channel, identify external damage spots in the image based on a target detection algorithm, and identify the external damage type using a pre-trained external damage type classification model based on the morphological and spectral characteristics of the external damage spots;

[0067] Extracting the mask area of ​​each external broken patch through a semantic segmentation network, calculating the external broken area of ​​each external broken patch based on the mask area, and calculating the Euclidean distance between each external broken patch and the nearest transmission tower based on the geographic coordinates;

[0068] Dynamically obtain real-time wind speed and direction data in the area where the transmission channel is located;

[0069] For each external damage pattern, a preset weight matrix is ​​matched according to the external damage type of each external damage pattern. The weight matrix includes weight distribution values ​​of five dimensions: external damage type, external damage area, Euclidean distance, wind force data and wind direction data;

[0070] Based on the weight matrix, a multi-dimensional risk assessment equation is established to calculate the risk value of each external broken pattern;

[0071] The risk values ​​of all external broken spots are normalized using the range method, and the normalized risk values ​​of each external broken spot are mapped to the preset risk level range to determine the risk level of each external broken spot;

[0072] The external damage location of the transmission channel and the corresponding risk level are displayed in real time through a visual interface.

[0073] Specifically, remote sensing images along the transmission channel are acquired through the Gaofen-2 satellite, and radiation correction and denoising are performed through the image preprocessing unit. The meteorological bureau's API is dynamically accessed to acquire real-time wind speed and direction data in the area where the transmission channel is located. The geographic coordinates of the transmission towers are extracted from the GIS system.

[0074] As a preferred embodiment of this embodiment, the target detection algorithm adopts the YOLOv7 model, the external damage type classification model is trained based on the ResNet50 network, and the input features include the morphological features and spectral features of the external damage spots. The extraction of the input features includes:

[0075] Morphological feature extraction: The edge detection algorithm is used to obtain the contour information of the outer broken patch area, and the geometric parameters of the outer broken patch contour are calculated as morphological features;

[0076] Spectral feature extraction: The reflectance data of the visible light band and infrared band of the remote sensing image are analyzed for differences, and the normalized vegetation index and thermal radiation anomaly are extracted as spectral features.

[0077] Specifically, the YOLOv7 model is deployed, and Mosaic data enhancement and adaptive anchor box calculation are used to identify external broken patches in images.

[0078] Obtain morphological features of the external broken patches: Use the Canny edge detection algorithm to extract the outline of the external broken patches, and calculate the perimeter-area ratio (PAR) and circularity (Cir) as morphological features;

[0079] Take the spectral characteristics of the external broken spots: According to the reflectivity data of the visible light band and infrared band of the remote sensing image, calculate the NDVI index and thermal radiation anomaly value as the spectral characteristics.

[0080] As a preferred method of this embodiment: identifying the external damage type by the external damage type classification model, the specific method includes:

[0081] The extracted morphological features and spectral features are fused to form a fused feature vector;

[0082] Inputting the fused feature vector into the feature extraction layer of the ResNet50 network for feature extraction, and enhancing the deep feature expression through the residual connection layer of the ResNet50 network;

[0083] A fully connected layer and a Softmax classifier are connected to the end of the ResNet50 network to output the probability distribution of the external damage pattern belonging to each preset external damage type; the preset external damage types include at least one or more of construction, greenhouse film, color steel tile house, pond and natural collapse;

[0084] Based on the probability distribution of the output, the external damage type with the highest probability is determined as the preliminary classification result of the external damage plot;

[0085] The preliminary classification result is post-processed by a non-maximum suppression algorithm to eliminate misjudgments of the external damage type in overlapping areas, and the external damage plot detection box and the corresponding external damage type with the highest confidence are retained as the final recognition result.

[0086] Specifically, the extracted morphological features are normalized and spliced into a 128-dimensional morphological feature vector, the extracted spectral features are normalized and spliced into a 64-dimensional spectral feature vector, and the morphological feature vector and the spectral feature map are merged along the channel dimension to form a 192-dimensional fusion feature vector.

[0087] A channel attention module is inserted in each residual block of ResNet50, the channel weights are dynamically adjusted by global average pooling and a fully connected layer, the key feature responses are enhanced, a feature pyramid network is used to integrate features at different levels, and the small target recognition capability is improved.

[0088] ResNet50 is connected to a fully connected layer and a Softmax classifier at the end, outputs the external damage probability distribution of 5 types such as construction, greenhouse mulch, etc., uses a Soft-NMS algorithm to perform weighted fusion on overlapping detection boxes, retains high confidence targets, effectively reduces the false detection rate, and improves the classification accuracy.

[0089] As a preferred mode of the embodiment: the weight matrix is a five-dimensional vector, each dimension corresponds to the weight distribution value of a different influencing factor, specifically including:

[0090] External damage type weight: determined according to the actual threat level difference of the external damage type to the power transmission channel, the higher the threat level, the greater the weight value of the external damage type;

[0091] Specifically, based on historical accident data and expert evaluation, different weights are given to external damage types such as construction, greenhouse mulch, color steel tile house, pit and pond, and natural collapse, and the higher the threat level, the greater the weight;

[0092] External damage area weight: positively related to the mask area of the external damage plot, the larger the external damage area, the greater the corresponding weight value;

[0093] Specifically, the mask area pixel ratio (area / total monitoring area) is calculated, and the external damage area weight is mapped through a Sigmoid function;

[0094] Euclidean distance weight: based on the Euclidean distance between the external damage plot and the nearest power transmission tower, the closer the distance between the external damage plot and the power transmission tower, the greater the corresponding weight value;

[0095] Specifically, the Euclidean distance between the external damage point and the nearest power tower is calculated, and the Euclidean distance weight adopts an exponential decay function, which decreases with the increase of the distance.

[0096] Wind data weight: positively correlated with real-time wind value, the larger the real-time wind value, the larger the corresponding weight value;

[0097] Specifically, the wind speed is obtained in real time, and is mapped to the wind data weight through linear transformation, and the larger the wind, the larger the weight;

[0098] Wind direction data weight: combined with the relative relationship between the wind direction and the external damage position, when the wind direction is towards the power transmission channel, the corresponding weight value is the largest, and when it deviates from the power transmission channel, the corresponding weight is the lowest.

[0099] Specifically, the angle θ between the wind direction and the axis of the power transmission channel is calculated, and the wind direction data weight adopts a cosine similarity function, when θ≤30°, the weight is 0.5, and when θ>90°, the weight decreases to 0.1;

[0100] According to the seasonal and regional differences of the power transmission channel, the proportion of each weight is dynamically adjusted to enhance the adaptability of the model.

[0101] As a preferred mode of the embodiment: the multi-dimensional risk assessment equation is constructed as:

[0102]

[0103] In formula (1), R represents the risk value of the external damage plot, W i represents the normalized weight value of the i-th dimension in the weight matrix, i=1,2,3,4,5 respectively corresponding to the external damage type, the external damage area, the Euclidean distance, the wind data and the wind direction data; X i represents the standardized score value of the i-th dimension.

[0104] Specifically, the calculation of the standardized score value of each dimension is:

[0105] External damage type score X1: mapped to a numerical value based on the threat level table, the higher the threat level, the larger the value of X1;

[0106] External damage area score X2: calculate the area ratio of the mask Where S max is the maximum area of the evaluation region, and S is the mask area of the external damage plot;

[0107] Euclidean distance score X3: based on the distance D from the nearest tower, adopting an inverse function Where D max is the maximum possible distance of the evaluation region;

[0108] Wind data score X4: based on the normalized real-time wind speed F, the calculation formula is Among them F 阈值 To preset safety threshold;

[0109] Wind direction data score X5: Calculate the angle θ between the wind direction and the axis of the transmission channel, and take the absolute value of the cosine X5 = |cosθ|. When it is in the downwind direction (θ = 0°), X5 = 1; when it is in the sidewind direction (θ = 90°), X5 = 0.

[0110] As a preferred method of this embodiment, the risk level of each external broken pattern spot is determined by:

[0111] Based on the statistical distribution characteristics of the normalized risk value, cluster analysis is used to dynamically determine the risk value interval boundaries to form at least three increasing risk level intervals;

[0112] Each risk value interval corresponds to a preset risk level, and the higher the risk value interval, the higher the risk level;

[0113] The normalized risk value of each external broken pattern spot is compared with the preset risk level range to determine the risk level of each external broken pattern spot and associate the corresponding external broken pattern spot position data.

[0114] Specifically, the range method is used to map the risk value R to the interval [0,1] to eliminate dimensional differences. The normalized data is clustered based on the Gaussian mixture model (GMM), the risk distribution pattern is automatically identified, and the boundaries of each risk interval are dynamically determined. The risk level mapping table is obtained as shown in Table 1:

[0115] Table 1 Risk level mapping table

[0116]

[0117]

[0118] Compare the normalized risk value of each external damage patch with the preset risk level range to determine the risk level of each external damage patch. Associate each external damage patch with the GIS coordinates to generate a risk heat map, which supports statistical risk distribution by administrative area (such as tower section, line section).

[0119] Example 2

[0120] Reference Figure 2 A device for dynamically assessing the risk level of power transmission channel external damage based on multi-source big data, the device comprising:

[0121] Multi-source data acquisition module, used to simultaneously acquire remote sensing image data, meteorological monitoring data and geospatial data along the transmission corridor;

[0122] Specifically, a high-resolution camera is integrated to capture visible light and infrared images of the transmission channel area, supporting all-weather monitoring. A weather station is integrated to collect real-time meteorological parameters such as wind speed, wind direction, temperature and humidity. Geographic coordinate information is synchronized through the Beidou positioning device to form spatiotemporal correlation data. This is combined with the three-dimensional coordinates of the transmission tower and topographic data to build a high-precision Geographic Information System (GIS) basic model.

[0123] The external damage identification module is in communication with the multi-source data acquisition module and includes:

[0124] Image preprocessing unit, which performs denoising, enhancement and radiation correction preprocessing on the acquired remote sensing images;

[0125] Specifically, the CLAHE algorithm is used to improve the contrast of remote sensing images, and the Retinex theory is combined to eliminate the interference of uneven illumination and enhance the accuracy of feature extraction of external broken spots.

[0126] The target detection unit identifies external broken spots in the image based on the target detection algorithm;

[0127] A classification unit is used to identify the type of external damage based on the morphological and spectral characteristics of the external damage pattern using a pre-trained external damage type classification model;

[0128] Specifically, the external damage pattern detection model was trained based on the improved YOLOv7 algorithm, and the pre-trained external damage type classification model was used to identify typical external damage types such as construction machinery and greenhouse ground film.

[0129] A feature extraction module, which is in communication with the external damage identification module, includes:

[0130] The mask generation unit is used to extract the mask area of ​​the external broken image patch through the semantic segmentation network;

[0131] An area calculation unit, which calculates the outer broken area based on the mask area;

[0132] A distance calculation unit calculates the Euclidean distance between the outer broken pattern and the nearest transmission tower based on geographic coordinates;

[0133] Specifically, the mask of the external breach area is extracted through the semantic segmentation network (U-Net architecture), and the area ratio is calculated. Based on the three-dimensional coordinates of the transmission tower and the GPS data of the external breach point, the Haversine formula is used to calculate the Euclidean distance.

[0134] A weight matching module is connected to the external damage identification module, the feature extraction module and the multi-source data acquisition module respectively, and is used to match a preset weight matrix according to the external damage type. The weight matrix includes weight distribution values ​​of five dimensions: external damage type, external damage area, Euclidean distance, wind speed data and wind direction data;

[0135] Specifically, a five-dimensional matrix was constructed based on the external damage type, external damage area, Euclidean distance, real-time wind force and wind direction data, and normalized to the interval [0,1] using the range method;

[0136] a risk assessment module, communicating with the weight matching module, establishing a multi-dimensional risk assessment equation based on the weight matrix, and calculating the risk value of each external broken pattern spot;

[0137] Specifically, the comprehensive risk value of each external broken pattern is calculated through a multi-dimensional risk assessment equation;

[0138] A risk grading module, in communication with the risk assessment module, includes:

[0139] Normalization unit, using the range method to normalize the risk value;

[0140] A risk level mapping unit is used to map the normalized risk value to a preset risk level range and output the risk level;

[0141] Specifically, the range method is used to map the risk value of each external broken spot to the interval [0,1], and the risk level of each external broken spot is obtained according to the preset risk level mapping table;

[0142] The visual warning module is in communication with the risk grading module and is used to display the location of the external damage of the power transmission channel and the corresponding risk level in real time.

[0143] Specifically, based on the GIS platform, risk level color identification is superimposed, and the tower number is associated with the operation and maintenance work order system. When a high-risk area is detected, the drone will be triggered to automatically inspect and push it to the nearest operation and maintenance personnel via SMS / APP.

[0144] This embodiment adopts a modular design to build a collaborative multi-source heterogeneous data collection system, overcoming the limitations of a single data source. A dynamic weight matching mechanism is established, combining multi-dimensional parameters such as damage type, spatial distance, and meteorological conditions to construct an adaptive risk assessment model, effectively addressing the coupling effects of risk variables in different environments. GIS visualization technology enables intuitive presentation of risk levels, and links drone inspections with the operation and maintenance response system to form a closed-loop management chain of "monitoring-assessment-warning-disposal." This embodiment implements a dynamic assessment device for transmission channel damage risk levels based on multi-source big data, effectively improving the timeliness of damage prevention and control.

[0145] Furthermore, the embodiment of the present application also provides a power transmission channel external damage risk level dynamic evaluation device based on multi-source big data, which comprises a memory, a processor, and a power transmission channel external damage risk level dynamic evaluation device program based on multi-source big data stored in the memory and capable of running on the processor. The power transmission channel external damage risk level dynamic evaluation device is configured to implement the power transmission channel external damage risk level dynamic evaluation device method based on multi-source big data as described in the above embodiment one.

[0146] Although the specific embodiments of the present application are described above in combination with the drawings, the description is not a limitation on the protection scope of the present application, and those skilled in the art should understand that various modifications or changes made on the basis of the technical solutions of the present application without creative labor are still within the protection scope of the present application.

Claims

1. A dynamic assessment method for the risk level of power transmission channel external damage based on multi-source big data, characterized by: The method comprises: Acquire remote sensing image data along the transmission channel, identify external damage spots in the image based on a target detection algorithm, and identify the external damage type using a pre-trained external damage type classification model based on the morphological and spectral characteristics of the external damage spots; Extracting the mask area of ​​each external broken patch through a semantic segmentation network, calculating the external broken area of ​​each external broken patch based on the mask area, and calculating the Euclidean distance between each external broken patch and the nearest transmission tower based on the geographic coordinates; Dynamically obtain real-time wind speed and direction data in the area where the transmission channel is located; For each external damage pattern, a preset weight matrix is ​​matched according to the external damage type of each external damage pattern. The weight matrix includes weight distribution values ​​of five dimensions: external damage type, external damage area, Euclidean distance, wind force data and wind direction data; Based on the weight matrix, a multi-dimensional risk assessment equation is established to calculate the risk value of each external broken pattern; The risk values ​​of all external broken spots are normalized using the range method, and the normalized risk values ​​of each external broken spot are mapped to the preset risk level range to determine the risk level of each external broken spot; The external damage location of the transmission channel and the corresponding risk level are displayed in real time through a visual interface.

2. The method for dynamic assessment of transmission channel external damage risk level based on multi-source big data according to claim 1 is characterized by: The target detection algorithm adopts the YOLOv7 model, and the external damage type classification model is based on ResNet50 network training. The input features include the morphological features and spectral features of the external damage spots. The extraction of the input features include: Morphological feature extraction: The edge detection algorithm is used to obtain the contour information of the outer broken patch area, and the geometric parameters of the outer broken patch contour are calculated as morphological features; Spectral feature extraction: The reflectance data of the visible light band and infrared band of the remote sensing image are analyzed for differences, and the normalized vegetation index and thermal radiation anomaly are extracted as spectral features.

3. The method for dynamic assessment of transmission channel external damage risk level based on multi-source big data according to claim 2 is characterized by: The external damage type is identified by the external damage type classification model, and the specific method includes: The extracted morphological features and spectral features are fused to form a fused feature vector; Inputting the fused feature vector into the feature extraction layer of the ResNet50 network for feature extraction, and enhancing the deep feature expression through the residual connection layer of the ResNet50 network; A fully connected layer and a Softmax classifier are connected to the end of the ResNet50 network to output the probability distribution of the external damage pattern belonging to each preset external damage type; the preset external damage types include at least one or more of construction, greenhouse film, color steel tile house, pond and natural collapse; Based on the output probability distribution, determining the external destruction type with the highest probability as the preliminary classification result of the external destruction pattern; The preliminary classification results are post-processed by the non-maximum suppression algorithm to eliminate the misjudgment of the external damage type in the overlapping area, and the external damage pattern detection frame with the highest confidence and the corresponding external damage type are retained as the final recognition result.

4. The method for dynamic assessment of transmission channel external damage risk level based on multi-source big data according to claim 1 is characterized by: The weight matrix is ​​a five-dimensional vector, and each dimension corresponds to the weight distribution value of different influencing factors, specifically including: External damage type weight: Determined based on the actual threat level of the external damage type to the transmission channel. The higher the threat level, the greater the corresponding weight value. External break area weight: positively correlated with the mask area of ​​the external break pattern. The larger the external break area, the greater the corresponding weight value. Euclidean distance weight: Based on the Euclidean distance between the outer broken pattern and the nearest transmission tower, the closer the distance between the outer broken pattern and the transmission tower, the greater the corresponding weight value; Wind data weight: positively correlated with the real-time wind value. The larger the real-time wind value, the greater the corresponding weight. Wind direction data weight: The value is assigned based on the relative relationship between wind direction and external breaking position. When the wind direction is toward the transmission channel, the corresponding weight value is the largest, and when it is away from the transmission channel, the corresponding weight value is the lowest.

5. The method for dynamic assessment of transmission channel external damage risk level based on multi-source big data according to claim 1 is characterized by: The multidimensional risk assessment equation is constructed as follows: In formula (1), R represents the risk value of the external broken pattern, W i represents the normalized weight value of the i-th dimension in the weight matrix, i = 1, 2, 3, 4, 5 correspond to the type of external damage, the area of ​​external damage, the Euclidean distance, the wind force data and the wind direction data respectively; X i Represents the standardized score value of the i-th dimension.

6. The method for dynamic assessment of transmission channel external damage risk level based on multi-source big data according to claim 1 is characterized by: The specific method for determining the risk level of each external broken pattern spot includes: Based on the statistical distribution characteristics of the normalized risk value, cluster analysis is used to dynamically determine the risk value interval boundaries to form at least three increasing risk level intervals; Each risk value interval corresponds to a preset risk level, and the higher the risk value interval, the higher the risk level; The normalized risk value of each external broken pattern spot is compared with the preset risk level range to determine the risk level of each external broken pattern spot and associate the corresponding external broken pattern spot position data.

7. A device for dynamically assessing the risk level of power transmission channel damage based on multi-source big data, characterized by: The device comprises: Multi-source data acquisition module, used to simultaneously acquire remote sensing image data, meteorological monitoring data and geospatial data along the transmission corridor; The external damage identification module is in communication with the multi-source data acquisition module and includes: Image preprocessing unit, which performs denoising, enhancement and radiation correction preprocessing on the acquired remote sensing images; The target detection unit identifies external broken spots in the image based on the target detection algorithm; A classification unit is used to identify the type of external damage based on the morphological and spectral characteristics of the external damage pattern using a pre-trained external damage type classification model; A feature extraction module, which is in communication with the external damage identification module, includes: The mask generation unit is used to extract the mask area of ​​the external broken image patch through the semantic segmentation network; An area calculation unit, which calculates the outer broken area based on the mask area; A distance calculation unit calculates the Euclidean distance between the outer broken pattern and the nearest transmission tower based on geographic coordinates; A weight matching module, connected to the external damage identification module, the feature extraction module and the multi-source data acquisition module, respectively, is used to match a preset weight matrix according to the external damage type, wherein the weight matrix includes weight distribution values ​​of five dimensions: external damage type, external damage area, Euclidean distance, wind force data and wind direction data; a risk assessment module, communicating with the weight matching module, establishing a multi-dimensional risk assessment equation based on the weight matrix, and calculating the risk value of each external broken pattern spot; A risk grading module, in communication with the risk assessment module, includes: Normalization unit, using the range method to normalize the risk value; A risk level mapping unit is used to map the normalized risk value to a preset risk level range and output the risk level; The visual warning module is in communication with the risk grading module and is used to display the location of the external damage of the power transmission channel and the corresponding risk level in real time.

8. A device for dynamically assessing the risk level of power transmission channel damage based on multi-source big data, characterized by: The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement a method for dynamically assessing the risk level of external damage to a power transmission channel based on multi-source big data as described in any one of claims 1 to 6.

Images