A method, device and medium for monitoring the condition of power transmission lines

Abstract

This application discloses a method, device, and medium for monitoring the condition of transmission lines; relating to the power grid field, it addresses the problems of low efficiency and low reliability of manual inspection of transmission lines. Through sensing terminals installed on transmission towers, it collects conductor contour images, insulator contour images, conductor temperature information, and insulator temperature information on the transmission line. The temperature information is converted into temperature images, and further, edge detection is performed on all physical state images to obtain edge calculation information. The edge calculation information is input into a preset image processing deep learning model for processing to obtain the state type of the transmission line and output the corresponding monitoring results. This comprehensively reflects the state of the transmission line, including external damage or abnormalities, detects potential fault points through temperature changes, unifies all data into image types for easier subsequent processing, and reduces data transmission volume and the computational burden on the central server by performing image edge detection preprocessing at the edge side.

Description

Technical Field

[0001] This application relates to the field of power grids, and in particular to a method, device and medium for monitoring the condition of transmission lines. Background Technology

[0002] In recent years, the scale of power transmission system equipment has continued to expand, and the requirements for power supply reliability from the whole society have also been continuously increasing. As the key to reliable power supply from the power grid, the normal operation of transmission lines is crucial. Therefore, it is urgent to accelerate the construction of intelligent inspection systems for transmission lines to achieve comprehensive perception of the line's operating status.

[0003] Traditional transmission line condition monitoring often relies on manual methods, using human eyes to identify faults. However, this method can only detect some obvious faults, such as insulator bursting, while it is difficult to detect faults such as line corrosion.

[0004] Therefore, how to solve the problems of low efficiency and low reliability of manual inspection of power transmission lines is a technical problem that urgently needs to be solved by people in this field. Summary of the Invention

[0005] The purpose of this application is to provide a method, device and medium for monitoring the condition of power transmission lines, so as to solve the problems of low efficiency and low reliability of manual inspection of power transmission lines.

[0006] To address the aforementioned technical problems, this application provides a method for monitoring the condition of transmission lines, applied to sensing terminals installed on various high-voltage transmission towers, comprising:

[0007] Collect images of conductor outlines, insulator outlines, conductor temperature information, and insulator temperature information on the transmission line.

[0008] The conductor temperature information and the insulator temperature information are converted into conductor temperature images and insulator temperature images;

[0009] Image edge detection is performed on each physical state image to obtain edge calculation information. The physical state images include conductor contour images, insulator contour images, conductor temperature images, and insulator temperature images.

[0010] The edge computing information is input into a preset image processing deep learning model for processing to obtain the state type of the transmission line.

[0011] Output the corresponding monitoring results based on the stated status type.

[0012] As an optional solution, in the above-mentioned transmission line condition monitoring method, the step of converting the conductor temperature information and the insulator temperature information into conductor temperature images and insulator temperature images includes:

[0013] The conductor temperature information and the insulator temperature information are preprocessed.

[0014] The preprocessed conductor temperature information and insulator temperature information are converted into corresponding pixel values ​​using a temperature-pixel mapping algorithm; wherein a mapping relationship is established between the temperature range and the pixel value range.

[0015] Based on the mapped pixel values, generate conductor temperature images and insulator temperature images.

[0016] As an optional solution, in the above-mentioned transmission line condition monitoring method, the step of performing image edge detection on the physical state image to obtain edge calculation information includes:

[0017] The physical state image is subjected to median filtering to remove image noise;

[0018] The gradient information of the filtered physical state image is obtained using the Sobel operator;

[0019] The gradient information is processed by nonmaximum suppression to determine the edge points in the physical state image;

[0020] The edge points are classified using a dual threshold algorithm to determine first-level edge points and second-level edge points, which are used as edge calculation information. The gray values ​​of the first-level edge points and the second-level edge points are different.

[0021] As an optional solution, in the above-mentioned transmission line condition monitoring method, the step of inputting edge computing information into a preset image processing deep learning model for processing to obtain the transmission line condition type includes:

[0022] The edge computing information is input into the embedding layer of a preset improved VisionTransformer model, the embedding layer including image patch embedding, learnable embedding and position embedding;

[0023] The output of the embedding layer is input into the Transformer encoder and features are extracted through the sparse attention module. The Transformer encoder is equipped with a sparse attention module and several coding modules. Each coding module includes a multi-head self-attention layer and a multi-layer perceptron layer.

[0024] The state type of the transmission line is obtained by classifying and predicting the output of the Transformer encoder using a multilayer perceptron classification head.

[0025] As an optional solution, in the above-mentioned transmission line condition monitoring method, the step of inputting the edge computing information into the embedding layer of a preset improved VisionTransformer model includes:

[0026] The input edge calculation information is segmented in the embedding layer to obtain image block tokens of fixed size;

[0027] Add location embedding information to each of the image block tokens;

[0028] A one-dimensional token sequence is obtained by concatenating the image patch token with the learnable embedded category vector;

[0029] The one-dimensional token sequence is used as the output of the embedding layer.

[0030] As an alternative, the above-mentioned transmission line condition monitoring method classifies and predicts the output of the Transformer encoder using a multilayer perceptron classification head to obtain the transmission line condition type, including:

[0031] The feature vector output by the Transformer encoder is subjected to dimensionality reduction processing;

[0032] The probability distribution of transmission line state types is obtained by normalizing the dimension-reduced feature vectors using the Softmax function.

[0033] The state type of the transmission line is determined based on the probability distribution;

[0034] Output the state type of the transmission line.

[0035] As an optional solution, the above-mentioned transmission line condition monitoring method outputs corresponding monitoring results according to the condition type, including:

[0036] The monitoring results are displayed visually on the display device in the monitoring center;

[0037] The monitoring results are used to determine whether the transmission line is in normal condition.

[0038] When the monitoring results indicate an abnormal state, a fault report is generated, including the fault type, location, and severity.

[0039] To address the aforementioned technical problems, this application also provides a transmission line condition monitoring device, applied to sensing terminals installed on various high-voltage transmission towers, comprising:

[0040] The data acquisition module is used to acquire images of conductor outlines, insulator outlines, conductor temperature information, and insulator temperature information on the transmission line.

[0041] The conversion module is used to convert the conductor temperature information and the insulator temperature information into conductor temperature images and insulator temperature images;

[0042] The image edge detection module is used to perform image edge detection on each physical state image to obtain edge calculation information. The physical state images include conductor contour images, insulator contour images, conductor temperature images, and insulator temperature images.

[0043] The analysis and processing module is used to input edge computing information into a preset image processing deep learning model for processing to obtain the state type of the transmission line.

[0044] The output module is used to output the corresponding monitoring results according to the state type.

[0045] To address the aforementioned technical problems, this application also provides a transmission line condition monitoring device, comprising:

[0046] Memory, used to store computer programs;

[0047] A processor is used to execute the computer program to implement the steps of the above-described transmission line condition monitoring method.

[0048] To address the aforementioned technical problems, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the aforementioned power transmission line status monitoring method.

[0049] The transmission line condition monitoring method provided in this application collects conductor contour images, insulator contour images, conductor temperature information, and insulator temperature information on the transmission line through sensing terminals installed on transmission towers. After preprocessing, the temperature information is converted into temperature images. Further edge detection is performed on all physical state images to obtain edge calculation information. Finally, the edge calculation information is input into a preset image processing deep learning model for processing to obtain the transmission line condition type and output the corresponding monitoring results. This application not only collects conductor and insulator contour images but also collects temperature information and converts it into temperature images. Unifying all data into image type facilitates subsequent processing and provides a more comprehensive reflection of the transmission line condition, including not only external damage or abnormalities but also the detection of potential fault points through temperature changes. By performing image edge detection preprocessing at the edge side, the amount of data transmitted and the computational burden on the central server are reduced. This method is applicable to transmission line monitoring under various environmental conditions, improves the efficiency of condition monitoring, and solves the problems of low efficiency and low reliability of manual transmission line inspections.

[0050] In addition, this application also provides an apparatus and medium that correspond to the above-mentioned transmission line condition monitoring method and have the same effect. Attached Figure Description

[0051] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This application provides a flowchart of a method for monitoring the condition of a power transmission line.

[0053] Figure 2 This application provides a schematic diagram of the Vision Transformer model structure as an embodiment;

[0054] Figure 3 This application provides a performance comparison chart of the Vision Transformer model before and after improvement.

[0055] Figure 4 This application provides a performance comparison chart of the Vision Transformer model before and after improvement, as shown in the embodiments of this application.

[0056] Figure 5 A structural diagram of a power transmission line condition monitoring device provided in an embodiment of this application;

[0057] Figure 6 This is a structural diagram of another power transmission line condition monitoring device provided in an embodiment of this application. Detailed Implementation

[0058] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0059] The core of this application is to provide a method, device, and medium for monitoring the condition of power transmission lines.

[0060] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0061] To address the aforementioned problems, this application provides a method for monitoring the status of power transmission lines, applied to sensing terminals installed on various high-voltage transmission towers, such as... Figure 1 As shown, it includes:

[0062] S11: Collect conductor outline images, insulator outline images, conductor temperature information, and insulator temperature information on the transmission line;

[0063] S12: Convert conductor temperature information and insulator temperature information into conductor temperature images and insulator temperature images;

[0064] S13: Perform image edge detection on each physical state image to obtain edge calculation information. The physical state images include conductor contour image, insulator contour image, conductor temperature image, and insulator temperature image.

[0065] S14: Input the edge computing information into the preset image processing deep learning model for processing to obtain the state type of the transmission line;

[0066] S15: Output the corresponding monitoring results based on the status type.

[0067] This embodiment provides a method for monitoring the status of power transmission lines, primarily applicable to real-time status monitoring of high-voltage transmission lines. This method is implemented using sensing terminals installed on each high-voltage transmission tower and is suitable for a wide range of transmission line scenarios, including but not limited to urban power grids, remote mountain power grids, and transmission line monitoring in harsh environments. The sensing terminals are installed on the high-voltage transmission towers and acquire physical status images through image acquisition modules and temperature sensors.

[0068] Step S11, the sensing terminal collects conductor and insulator contour images, providing visual information about the transmission line, while conductor and insulator temperature information reflects the line's operational status. This data collection forms the foundation of the entire monitoring system, ensuring the accuracy and completeness of the data sources for subsequent processing.

[0069] Under certain specific conditions, such as in extreme weather (e.g., heavy rain, heavy snow) or at night, the image acquisition module may need to be waterproof, dustproof, and have night vision capabilities to ensure the accuracy and reliability of image acquisition.

[0070] Step S12 converts the conductor temperature information and insulator temperature information into a temperature image. This process is achieved through a temperature-pixel mapping algorithm. The temperature image can intuitively reflect the temperature distribution of the transmission line, providing a unified format for subsequent image processing and facilitating edge detection and input for deep learning models.

[0071] Temperature sensors, such as thermocouples, thermistors, or infrared thermal imagers, are used to collect temperature information from the conductors and insulators. These sensors acquire the temperature information in real time and convert it into electrical signals. The collected temperature data is then transmitted via wired or wireless communication modules. To improve the reliability of data transmission, verification and error correction can be performed during the transmission process.

[0072] Step S13 performs image edge detection on the physical state image to obtain edge calculation information. The gradient information of the filtered image is calculated using the Sobel operator to initially extract the image's edge features. The Sobel operator is a commonly used edge detection operator capable of calculating the gradient of an image in both horizontal and vertical directions. Non-maximum suppression (NMS) can also be applied to the gradient information to identify edge points in the image. NMS compares gradient magnitudes, retaining edge points with local maxima and removing other non-edge points. A double thresholding algorithm can also be used to classify edge points, distinguishing between strong and weak edge points. The double thresholding algorithm effectively differentiates the intensity of edge points, avoiding misjudgments.

[0073] Under certain specific conditions, such as when the image is noisy or the edge features are not obvious, more advanced edge detection algorithms (Canny operators) can be used to improve the accuracy of edge detection.

[0074] Step S14 inputs the edge computing information into a preset image processing deep learning model for processing to obtain the state type of the transmission line. This step performs deep analysis on the image information after edge detection to identify the state type of the transmission line.

[0075] Edge computing information is input into a pre-defined improved Vision Transformer model. This model includes an embedding layer, a Transformer encoder, and a multilayer perceptron classification head.

[0076] Figure 2 This application provides a schematic diagram of the Vision Transformer model structure, as shown in the embodiment. Figure 2 As shown. In this embodiment, the improved Vision Transformer model consists of an embedding layer, a Transformer encoder, and a multilayer perceptron classification head (MLP Head).

[0077] Figure 3 This application provides a performance comparison chart of the Vision Transformer model before and after improvement, as shown in the embodiments of this application. Figure 4 This application provides a performance comparison chart of the Vision Transformer model before and after improvement, as shown in the embodiments of this application; Figure 3, 4 It can be seen that the improved Vision Transformer model has a smaller loss function value, close to 0.1, and a higher state analysis accuracy, reaching 97%. This is because the improved Vision Transformer model incorporates a sparse attention module, which can retain global classification feature information, further improving the accuracy of transmission line state image analysis and accelerating model convergence.

[0078] Embedding Layer: In the embedding layer, the input edge computation information is segmented into fixed-size image patch tokens, and positional embedding information is added to each image patch (token). A one-dimensional token sequence is obtained by concatenating the image patch tokens with the learnable embedding class vectors.

[0079] Transformer Encoder: The output of the embedding layer is input into the Transformer encoder, which extracts features from the token sequence through a sparse attention module. The sparse attention module can highlight important features and suppress irrelevant features.

[0080] Multilayer perceptron classification head: The multilayer perceptron classification head classifies and predicts the output of the Transformer encoder to obtain the state type of the transmission line.

[0081] Step S15 outputs the applicable environment or conditions of the corresponding monitoring results based on the status type, and displays the monitoring results in a visual manner on the display device in the monitoring center for easy real-time viewing by maintenance personnel. For detected abnormal states, a detailed fault report is generated, including information such as fault type, location, and severity. The monitoring results and fault reports are stored in the cloud database for subsequent data analysis and historical tracing. When an abnormal state is detected, maintenance personnel are promptly notified via SMS, email, or software (APP) push notifications.

[0082] Under certain specific conditions, if further analysis of the monitoring results is required, trend analysis can be performed by combining historical data to identify potential failure trends.

[0083] The transmission line condition monitoring method provided in this application collects conductor contour images, insulator contour images, conductor temperature information, and insulator temperature information on the transmission line through a sensing terminal installed on the transmission tower. After preprocessing, the temperature information is converted into temperature images. Further edge detection is performed on all physical state images to obtain edge calculation information. Finally, the edge calculation information is input into a preset image processing deep learning model for processing to obtain the transmission line condition type and output the corresponding monitoring results. This application not only collects conductor and insulator contour images but also collects temperature information and converts it into temperature images. Unifying all data into image types facilitates subsequent processing and provides a more comprehensive reflection of the transmission line condition, including not only external damage or abnormalities but also the detection of potential fault points through temperature changes. By performing image edge detection preprocessing at the edge side, the amount of data transmitted and the computational burden on the central server are reduced. This method is applicable to transmission line monitoring under various environmental conditions, improves the efficiency of condition monitoring, and solves the problems of low efficiency and low reliability of manual transmission line inspections.

[0084] Furthermore, converting conductor temperature information and insulator temperature information into conductor temperature images and insulator temperature images includes:

[0085] Perform data preprocessing on conductor temperature information and insulator temperature information;

[0086] The preprocessed conductor temperature information and insulator temperature information are converted into corresponding pixel values ​​using a temperature-pixel mapping algorithm; a mapping relationship is established between the temperature range and the pixel value range.

[0087] Based on the mapped pixel values, generate conductor temperature images and insulator temperature images.

[0088] Data preprocessing is the first step in converting conductor and insulator temperature information into temperature images. The main purpose of this step is to eliminate noise and outliers in the raw data, ensuring data quality for subsequent processing. Specifically, data preprocessing may include the following aspects: Data calibration: Calibrating the acquired temperature data to eliminate the influence of sensor errors and environmental factors. For example, by comparing the actual measured values ​​with theoretical values ​​at known temperature points, adjusting the sensor's measurement parameters ensures data accuracy. Data normalization: Normalizing the temperature data to a specific range, such as 0 to 1 or 0 to 255. Normalization facilitates subsequent temperature-pixel mapping, enabling unified processing of data from different temperature ranges.

[0089] Converting preprocessed temperature data into pixel values ​​is a crucial step in generating temperature images. The core of the temperature-pixel mapping algorithm is establishing a mapping relationship between temperature ranges and pixel value ranges. This mapping relationship is determined based on the specific application requirements. For example, the lowest temperature within the range can be mapped to pixel value 0, the highest temperature to pixel value 255, and intermediate temperatures can be mapped to pixel values ​​between 0 and 255 using a linear or non-linear relationship. Based on this mapping, each temperature data point is converted into its corresponding pixel value. This process transforms temperature information into a format recognizable by image processing, providing a foundation for subsequent image analysis.

[0090] Based on the mapped pixel values, conductor temperature images and insulator temperature images are generated. This step converts one-dimensional temperature data into two-dimensional image data, allowing temperature information to be visually represented in space. The mapped pixel values ​​are then filled into the image matrix to generate a complete temperature image. The temperature image can visually display the temperature distribution of the conductor and insulator, facilitating subsequent image analysis and processing.

[0091] Data preprocessing eliminates noise and outliers in the raw data, improving data quality and reliability and providing a solid foundation for subsequent image generation and analysis. Converting temperature data into temperature images allows for a visually intuitive spatial representation of temperature information. This intuitive visualization helps in quickly identifying areas of temperature anomalies, improving monitoring accuracy and efficiency. Temperature images provide rich information for subsequent image analysis, making it possible to use image processing and deep learning techniques for transmission line condition monitoring. For example, image edge detection and deep learning models can more accurately identify faults and abnormal conditions in transmission lines.

[0092] Furthermore, in one specific embodiment, image edge detection is performed on the physical state image to obtain edge calculation information, including:

[0093] Median filtering is applied to the physical state image to remove image noise;

[0094] The gradient information of the filtered physical state image is obtained through the Sobel operator;

[0095] The gradient information is processed by the non-maximum suppression method to determine the edge points in the physical state image;

[0096] Edge points are classified using a dual threshold algorithm to determine first-level and second-level edge points, which serve as edge calculation information. The grayscale values ​​of the first-level and second-level edge points are different.

[0097] Median filtering is applied to physical state images to remove image noise. Median filtering is a non-linear filtering technique that effectively removes salt-and-pepper noise and other random noise by replacing each pixel value in the image with the median value of its neighborhood. Median filtering preserves the image's edge information and avoids blurring.

[0098] Choosing an appropriate filter window size (such as 3×3 or 5×5) is necessary to balance noise removal effectiveness and computational complexity. Smaller windows can preserve more detail but may not completely remove noise; larger windows can remove noise better but may blur the edges of the image.

[0099] The first image information is obtained by performing median filtering. The specific expression is:

[0100] ;

[0101] In the formula, The image contains various status information of the transmission line. x and y represent the coordinates of the first image information, k and l are two-dimensional templates, and med[] represents an algorithm for calculating the difference between two strings.

[0102] The gradient information of the filtered physical state image is obtained through the Sobel operator. The Sobel operator is a commonly used edge detection operator that highlights edge features by calculating the gradient of the image in the horizontal and vertical directions.

[0103] The Sobel operator is used to calculate the image gradient. The Sobel operators in the horizontal and vertical directions are applied to the filtered image respectively to calculate the gradient magnitude and direction of each pixel.

[0104] Edge points in the physical state image are determined by processing gradient information using the non-maximum suppression method. A pixel is considered an edge point if its gradient magnitude is the maximum of the gradient magnitudes of its neighboring pixels along the gradient direction.

[0105] Non-maximum suppression (NMS) compares gradient magnitudes, retaining edge points of local maxima and removing other non-edge points. The specific steps are as follows:

[0106] For each pixel, examine its neighboring pixels along the gradient direction.

[0107] If the gradient magnitude of the current pixel is a local maximum in its gradient direction, then retain the pixel as an edge point; otherwise, set it as a non-edge point.

[0108] Edge points are classified using a dual threshold algorithm to determine first-level and second-level edge points, which serve as edge calculation information. The grayscale values ​​of the first-level and second-level edge points are different.

[0109] A high threshold is used to identify strong edge points. Edge points with grayscale values ​​higher than the high threshold are considered strong edge points, and their grayscale values ​​are typically set to 255 (white).

[0110] A low threshold is used to identify weak edge points. Edge points with grayscale values ​​between the low and high thresholds are considered weak edge points, and their grayscale values ​​are usually preserved or set to a low value (such as 128).

[0111] Points with gray values ​​below a low threshold are considered noise or non-edge points and are usually set to 0 (black).

[0112] The choice between high and low thresholds can be adjusted based on the specific application scenario and image characteristics. Generally, high thresholds are used to detect obvious edges, while low thresholds are used to detect possible weak edges.

[0113] Median filtering effectively removes noise from images while preserving edge information, improving image quality and the efficiency of subsequent processing. Gradient calculation using the Sobel operator accurately extracts edge features, providing reliable data support for subsequent edge detection. Non-maximum suppression precisely locates edge points in the image, avoiding misjudgments and redundant edge points. A dual-threshold algorithm classifies edge points, distinguishing between strong and weak edges, providing richer information for subsequent status monitoring and fault diagnosis. This method is adaptable to different image qualities and environmental conditions; by adjusting parameters such as the filter window size and threshold, it can flexibly address various practical application scenarios.

[0114] Furthermore, in one specific embodiment, edge computing information is input into a preset image processing deep learning model for processing to obtain the state type of the transmission line, including:

[0115] Edge computing information is input into the embedding layer of the pre-defined improved VisionTransformer model. The embedding layer includes image patch embedding, learnable embedding, and location embedding.

[0116] The output of the embedding layer is input into the Transformer encoder and features are extracted through the sparse attention module. The Transformer encoder is equipped with a sparse attention module and several coding modules. Each coding module includes a multi-head self-attention layer and a multi-layer perceptron layer.

[0117] The state type of the transmission line is obtained by classifying and predicting the output of the Transformer encoder using a multilayer perceptron classification head.

[0118] Edge computing information is input into the embedding layer of a pre-defined improved Vision Transformer model. The embedding layer includes image patch embedding, learnable embedding, and location embedding.

[0119] Image patch embedding divides edge computation information into fixed-size image patches (e.g., 16×16 pixels). Each image patch is flattened into a one-dimensional vector and mapped to the model's dimensional space through a linear transformation.

[0120] Learnable embeddings introduce a learnable class token, which is fed into the model along with the image patch embeddings. This class token is continuously learned throughout the training process and is used to store global feature information of the image.

[0121] Since the Transformer architecture itself does not retain spatial information, positional embedding is needed to provide the model with the spatial location information of pixels. Positional embedding can be learned or fixed, such as sinusoidal positional encoding.

[0122] The output of the embedding layer is input into the Transformer encoder and features are extracted through the sparse attention module. The Transformer encoder is equipped with a sparse attention module and several coding modules. Each coding module includes a multi-head self-attention layer and a multi-layer perceptron layer.

[0123] The Transformer encoder consists of multiple identical encoding modules stacked together. Each encoding module comprises two main parts: a multi-head self-attention layer and a multi-layer perceptron layer.

[0124] The sparse attention module introduces sparsity constraints, enabling the model to process image features more efficiently. For example, sparse attention mechanisms can be used to reduce computation while preserving important feature information.

[0125] The learning weights are obtained by the sparse attention module based on the previous layers of the encoding module, which are learned from the last layer of the encoding module. Its expression is as follows:

[0126] ;

[0127] In the formula, m is the first ordinal number. N is the number of encoding module layers. To learn the sub-weights, i is the second ordinal number. , The total number of self-attention heads;

[0128] ;

[0129] In the formula, i is the second ordinal number. , and All are basic learning weights;

[0130] The attention map is globally averaged and pooled into a descriptor using a sparse attention module. A model of the correlation between attention maps is constructed using two connected multi-head self-attention layers and a multilayer perceptron layer. The weight values ​​of the encoding modules (SMs) of each layer are obtained, and the final attention weights are calculated by combining them with the learned weights. Its expression is as follows:

[0131] ;

[0132] In the formula, Let be the weight value of the m-th layer coding module SM;

[0133] By using the weights corresponding to the classification vectors in the final value of the attention weights, the latent features corresponding to the maximum weights are selected from the self-attention head and used as the latent features input to the final encoding module.

[0134] Multi-head self-attention layers: The multi-head self-attention mechanism allows the model to learn information in different representation subspaces, thereby capturing long-range dependencies in the image. Multilayer perceptron layers: Each multi-head self-attention layer is followed by a multilayer perceptron (MLP) for further processing of the feature vectors. The MLP classification head classifies and predicts the state type of the transmission line by classifying the output of the Transformer encoder.

[0135] The improved Vision Transformer model effectively extracts global and local features from images, providing rich information for power transmission line status classification. The introduction of a sparse attention module makes the model more efficient when processing high-resolution images, reducing computational cost and memory usage. The multilayer perceptron classification head accurately predicts the extracted features, and combined with the Softmax function, it outputs the probability distribution for each category, improving classification reliability.

[0136] Furthermore, in one specific embodiment, inputting edge computing information into the embedding layer of a pre-defined improved VisionTransformer model includes:

[0137] In the embedding layer, the edge information of the input is segmented to obtain image block tokens of fixed size;

[0138] Add location embedding information to each image block token;

[0139] A one-dimensional token sequence is obtained by concatenating the image patch token with the class vector of the learnable embedding;

[0140] The one-dimensional token sequence is used as the output of the embedding layer.

[0141] The edge calculation information of the input is segmented in the embedding layer to obtain image block tokens of fixed size.

[0142] The input edge computing information (usually a two-dimensional image) is segmented into fixed-size image patches. For example, a 16×16 pixel image patch size can be selected. Each image patch is flattened into a one-dimensional vector and mapped to the model's dimensional space through a linear transformation.

[0143] The size of the image patch can be adjusted based on the resolution of the input image and the computational power of the model. Larger image patches can reduce computation but may lose some detail; smaller image patches can retain more detail but increase computational complexity.

[0144] Since the Transformer architecture itself does not retain spatial information, location embedding information needs to be added to each image patch token. The location embedding can be learned or fixed.

[0145] Location embeddings provide the model with spatial location information of pixels, enabling the model to better understand spatial relationships within an image. A one-dimensional token sequence is obtained by concatenating image patch tokens with class vectors that can be learned embeddings.

[0146] The category vectors are concatenated element-wise with all image patch tokens to form a one-dimensional token sequence. This concatenated sequence is then used as the output of the embedding layer and fed into the subsequent Transformer encoder. Under certain conditions, such as when handling multi-task learning, multiple category vectors can be introduced to be used for different tasks, thereby improving the model's versatility.

[0147] The one-dimensional token sequence is used as the output of the embedding layer, and the concatenated token sequence is used as the output of the embedding layer, which is then directly input into the Transformer encoder.

[0148] Furthermore, in one specific embodiment, the output of the Transformer encoder is classified and predicted using a multilayer perceptron classification head to obtain the state type of the transmission line, including:

[0149] Dimensionality reduction is performed on the feature vector output by the Transformer encoder;

[0150] The probability distribution of transmission line state types is obtained by normalizing the dimension-reduced feature vectors using the Softmax function.

[0151] Determine the state type of the transmission line based on the probability distribution;

[0152] The status type of the output transmission line.

[0153] Dimensionality reduction is performed on the feature vectors output by the Transformer encoder to reduce the feature vector dimension, thereby reducing computational complexity and avoiding overfitting scenarios, especially when dealing with high-dimensional feature data.

[0154] Multilayer perceptrons (MLPs) are used to reduce the dimensionality of the feature vectors output by the Transformer encoder. MLPs typically contain one or more fully connected layers that perform feature transformations using non-linear activation functions such as ReLU. This maps high-dimensional feature vectors to a low-dimensional space while preserving the information most useful for the classification task. For example, a feature vector can be reduced from 512 dimensions to 128 dimensions.

[0155] By normalizing the dimensionality-reduced feature vectors using the Softmax function, the probability distribution of transmission line state types is obtained, transforming the feature vectors into probability distributions, which is especially useful for multi-class classification tasks.

[0156] The state type of the transmission line is determined based on the probability distribution, and the category with the highest probability is selected as the final classification result. For example, if the probability distribution is [0.1, 0.3, 0.6], then the third category is selected as the state type of the transmission line.

[0157] A confidence threshold can be set, for example, 0.5. If the highest probability is below this threshold, the classification result is considered unreliable and can be marked as "uncertain" or a further check can be requested.

[0158] Output the status type of the transmission line, output the classification results and perform subsequent processing. The output content may include information such as status type, probability value, and timestamp.

[0159] The classification results are displayed in real time on the monitoring center's display devices, allowing maintenance personnel to promptly understand the status of the transmission lines. If the classification result indicates an abnormal state, maintenance personnel can be notified promptly via SMS, email, or app push notifications. Under certain specific conditions, if further analysis of the classification results is required, the results can be stored in a database for subsequent trend analysis and historical tracking.

[0160] Furthermore, in one specific embodiment, the corresponding monitoring results are output according to the state type, including:

[0161] The monitoring results are displayed visually on the display devices in the monitoring center;

[0162] The monitoring results are used to determine whether the transmission line is in normal condition.

[0163] When the monitoring results indicate an abnormal state, a fault report is generated, including the fault type, location, and severity.

[0164] The monitoring results are displayed visually on the monitoring center's display devices, which can be large-screen monitors, monitoring terminals, or mobile devices. These devices need to have high resolution and good color reproduction capabilities to clearly display the monitoring results. The monitoring results can be displayed in various ways, such as: directly displaying images of the physical state of the transmission line, including conductor outline images, insulator outline images, conductor temperature images, and insulator temperature images; using colors or icons to represent the state type of the transmission line, such as green for normal and red for abnormal; and displaying real-time monitoring data, such as temperature values ​​and edge detection results.

[0165] Based on the state type predicted by classification, the system determines whether the transmission line is in a normal state. For example, state types can be categorized as "normal," "minor anomaly," and "serious anomaly." A threshold is set to distinguish between normal and abnormal states. For instance, if the probability of the "normal" category in the state type probability distribution is below a certain threshold (e.g., 0.7), the transmission line is considered to be in an abnormal state. Combining this with other anomaly detection algorithms (such as threshold-based detection, statistical detection, or machine learning-based detection) can improve the accuracy of anomaly detection.

[0166] This embodiment ensures the efficiency and reliability of the transmission line condition monitoring method through detailed design of monitoring result output and fault report generation steps, which can effectively improve the operational safety and maintenance efficiency of transmission lines.

[0167] The above embodiments have described the transmission line condition monitoring method in detail. This application also provides embodiments corresponding to the transmission line condition monitoring device. It should be noted that this application describes the device embodiments from two perspectives: one based on functional modules and the other based on hardware.

[0168] From the perspective of functional modules Figure 5 A structural diagram of a power transmission line condition monitoring device provided in this application embodiment is shown below. Figure 5 As shown, a power transmission line condition monitoring device is applied to sensing terminals installed on various high-voltage transmission towers, comprising:

[0169] The acquisition module 21 is used to acquire the outline images of conductors, the outline images of insulators, the temperature information of conductors, and the temperature information of insulators on the transmission line.

[0170] The conversion module 22 is used to convert conductor temperature information and insulator temperature information into conductor temperature images and insulator temperature images.

[0171] The image edge detection module 23 is used to perform image edge detection on each physical state image to obtain edge calculation information. The physical state images include conductor contour images, insulator contour images, conductor temperature images, and insulator temperature images.

[0172] Analysis and processing module 24 is used to input edge computing information into a preset image processing deep learning model for processing to obtain the state type of the transmission line;

[0173] Output module 25 is used to output the corresponding monitoring results according to the status type.

[0174] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus, which will not be repeated here.

[0175] Figure 6 A structural diagram of another power transmission line condition monitoring device provided in this application embodiment is shown below. Figure 6 As shown, the transmission line condition monitoring device includes: a memory 30 for storing computer programs;

[0176] The processor 31 is used to execute a computer program to implement the steps of the method for obtaining user operation habit information as described in the above embodiment (transmission line condition monitoring method).

[0177] The power transmission line condition monitoring device provided in this embodiment may include, but is not limited to, mobile terminals, personal computers, workstations, etc.

[0178] The processor 31 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 31 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 31 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 31 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 31 may also include an Artificial Intelligence (AI) processor, which handles computational operations related to machine learning.

[0179] The memory 30 may include one or more computer-readable storage media, which may be non-transitory. The memory 30 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 30 is used to store at least the following computer program 301, which, after being loaded and executed by the processor 31, can implement the relevant steps of the transmission line condition monitoring method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 30 may also include an operating system 302 and data 303, and the storage method may be temporary or permanent storage. The operating system 302 may include Windows, Unix, Linux, etc. The data 303 may include, but is not limited to, the data involved in implementing the transmission line condition monitoring method.

[0180] In some embodiments, the power transmission line condition monitoring device may further include a display screen 32, an input / output interface 33, a communication interface 34, a power supply 35, and a communication bus 36.

[0181] Those skilled in the art will understand that Figure 6 The structure shown does not constitute a limitation on the transmission line condition monitoring device and may include more or fewer components than shown.

[0182] The transmission line condition monitoring device provided in this application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the following method: transmission line condition monitoring method.

[0183] Finally, this application also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps described in the above embodiment of the transmission line condition monitoring method.

[0184] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0185] The computer-readable storage medium provided in this embodiment stores a computer program thereon. When the processor executes the program, it can implement the following method: a method for monitoring the status of power transmission lines.

[0186] The transmission line condition monitoring method, apparatus, and medium provided in this application have been described in detail above. The various embodiments in the specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

[0187] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

Claims

1. A method of monitoring the condition of a power transmission line, characterised by, The application is applied to a sensing terminal installed on each high-voltage transmission tower, comprising: Collecting conductor profile images, insulator profile images, conductor temperature information and insulator temperature information on a transmission line; Converting the conductor temperature information and the insulator temperature information into conductor temperature images and insulator temperature images; Performing image edge detection on each physical state image to obtain edge calculation information, wherein the physical state images include the conductor profile images, the insulator profile images, the conductor temperature images and the insulator temperature images; Inputting the edge calculation information into a preset image processing deep learning model for processing to obtain a state type of the transmission line; Outputting a corresponding monitoring result according to the state type; The conversion of the conductor temperature information and the insulator temperature information into the conductor temperature images and the insulator temperature images comprises: Performing data preprocessing on the conductor temperature information and the insulator temperature information; Using a temperature-pixel mapping algorithm to convert temperature data into corresponding pixel values after preprocessing the conductor temperature information and the insulator temperature information, wherein a mapping relationship is established between a temperature range and a pixel value range; Generating the conductor temperature images and the insulator temperature images according to the mapped pixel values; The image edge detection on the physical state images to obtain the edge calculation information comprises: Performing median filter processing on the physical state images to remove image noise; Obtaining gradient information of the filtered physical state images through a Sobel operator; Processing the gradient information through a non-maximum suppression method to determine edge points in the physical state images; specifically, for each pixel point, checking adjacent pixels in its gradient direction; if the gradient amplitude of the current pixel is the local maximum value in its gradient direction, the pixel is retained as an edge point; otherwise, it is set as a non-edge point; Classifying the edge points through a double-threshold algorithm to determine first-level edge points and second-level edge points as edge calculation information, wherein the first-level edge points and the second-level edge points have different gray values; The processing of the edge calculation information into a preset image processing deep learning model to obtain the state type of the transmission line comprises: Inputting the edge calculation information into an embedding layer of a preset improved VisionTransformer model, wherein the embedding layer includes image block embedding, learnable embedding and position embedding; Inputting an output result of the embedding layer into a Transformer encoder and performing feature extraction through a sparse attention module, wherein the Transformer encoder is provided with a sparse attention module and a plurality of layer encoding modules, and each layer encoding module includes a multi-head self-attention layer and a multi-layer perception layer; Classifying and predicting the output of the Transformer encoder through a multi-layer perception classification head to obtain the state type of the transmission line.

2. The power transmission line condition monitoring method according to claim 1, characterized by, The inputting of the edge calculation information into the embedding layer of the preset improved VisionTransformer model comprises: The edge computing information input is segmented in the embedding layer to obtain an image block token of a fixed size; Position embedding information is added to each image block token; A one-dimensional token sequence is obtained by splicing the image block token and the category vector of the learnable embedding; The one-dimensional token sequence is taken as an output result of the embedding layer.

3. The power transmission line condition monitoring method according to claim 1, characterized by, The output of the Transformer encoder is classified and predicted by a multi-layer perception classification head to obtain a state type of the power transmission line, including: The feature vector output by the Transformer encoder is processed by dimension reduction; The feature vector after dimension reduction is normalized by a Softmax function to obtain a probability distribution of the state type of the power transmission line; The state type of the power transmission line is determined according to the probability distribution; The state type of the power transmission line is output.

4. The method of monitoring the state of a power transmission line according to any one of claims 1 to 3, characterized by, According to the state type output, a corresponding monitoring result is output, including: The monitoring result is displayed on a display device of a monitoring center in a visual manner; According to the monitoring result, it is judged whether the state of the power transmission line is normal; When the monitoring result represents an abnormal state, a fault report is generated, including a fault type, a location and a severity.

5. A power line condition monitoring device, characterised in that, The application is applied to a sensing terminal installed on each high-voltage transmission tower, including: An acquisition module is configured to acquire conductor profile images, insulator profile images, conductor temperature information and insulator temperature information on a power transmission line; A conversion module is configured to convert the conductor temperature information and the insulator temperature information into conductor temperature images and insulator temperature images; An image edge detection module is configured to perform image edge detection on each physical state image to obtain edge computing information, wherein the physical state image includes a conductor profile image, an insulator profile image, a conductor temperature image and an insulator temperature image; An analysis processing module is configured to input the edge computing information into a preset image processing deep learning model for processing to obtain a state type of the power transmission line; An output module is configured to output a corresponding monitoring result according to the state type; The conversion of the conductor temperature information and the insulator temperature information into the conductor temperature images and the insulator temperature images includes: Data preprocessing is performed on the conductor temperature information and the insulator temperature information; The preprocessed conductor temperature information and the insulator temperature information are converted into corresponding pixel values by a temperature-pixel mapping algorithm, wherein a mapping relationship between a temperature range and a pixel value range is established; According to the mapped pixel values, conductor temperature images and insulator temperature images are generated; The image edge detection on the physical state image to obtain the edge computing information includes: Median filtering processing is performed on the physical state image to remove image noise; Gradient information of the filtered physical state image is obtained by a Sobel operator; The gradient information is processed by a non-maximum suppression method to determine edge points in the physical state image, specifically including: for each pixel point, checking adjacent pixels in the gradient direction thereof; if the gradient amplitude of the current pixel is a local maximum in the gradient direction thereof, the pixel is reserved as an edge point; otherwise, the pixel is set as a non-edge point; The edge points are classified by a double-threshold algorithm to determine first-level edge points and second-level edge points as edge calculation information, wherein the first-level edge points and the second-level edge points have different gray values; The edge calculation information is input into a preset image processing deep learning model to obtain a state type of the power transmission line, including: The edge calculation information is input into an embedding layer of a preset improved VisionTransformer model, the embedding layer including image block embedding, learnable embedding and position embedding; The output result of the embedding layer is input into a Transformer encoder and feature extraction is performed by a sparse attention module, the Transformer encoder being provided with a sparse attention module and a plurality of layers of encoding modules, each layer of encoding module including a multi-head self-attention layer and a multi-layer perceptron layer; The output of the Transformer encoder is classified and predicted by a multi-layer perceptron classification head to obtain the state type of the power transmission line.

6. A power line condition monitoring device, characterised in that, including: a memory for storing a computer program; a processor for executing the computer program to implement the steps of the power transmission line state monitoring method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer program is stored on the computer readable storage medium and is executed by the processor to implement the steps of the power transmission line state monitoring method according to any one of claims 1 to 4.

Images