Safety intelligent early warning system of power transmission channel

Abstract

The invention relates to the technical field of power transmission channel monitoring, and discloses a safety intelligent early warning system for a power transmission channel, which comprises a power supply unit, a monitoring unit, a processing unit and a deep learning recognition unit, by means of multi-aspect innovation, efficient and reliable safety protection is achieved; in a power supply link, inductive power taking and super capacitor energy storage are combined, so that complexity of an external power supply is avoided, and stable power supply is guaranteed; during monitoring, multiple sensors work cooperatively, data are collected in all directions, and sufficient information is provided for analysis; in the aspect of data processing, a multi-modal fusion technology is matched with a plurality of feature extraction algorithms, so that data values are fully mined, and comprehensive feature vectors are more representative; based on a CNN deep learning model, under the assistance of reasonable data set division and multi-layer structure operation, the hidden danger is accurately judged; once the hidden danger is detected, early warning is quickly performed and information is pushed, so that the response time is effectively shortened, and the safety and reliability of the power system are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power transmission channel monitoring, and in particular to a safety intelligent early warning system for power transmission channels. Background Art

[0002] In the field of power transmission, the safe and stable operation of transmission channels is crucial. As the scale of the power grid continues to expand, the distribution of transmission lines has become increasingly widespread and complex, posing numerous challenges to traditional transmission channel monitoring and maintenance methods. Early power supply methods relied heavily on external power sources, which not only incurred high installation and maintenance costs but also presented wiring difficulties and unstable power supply in remote or complex terrain areas, severely impacting the continued operation of monitoring systems. During monitoring, a single monitoring method struggled to provide a comprehensive, real-time overview of the overall status of transmission channels, leading to frequent missed and false detections.

[0003] During the data processing phase, the lack of effective multimodal data fusion technology results in the isolation of different types of monitoring data, making it impossible to form a comprehensive and accurate description of transmission channel characteristics, which reduces the ability to analyze and identify potential safety hazards. In terms of feature extraction, traditional methods struggle to accurately extract key features from image and sensor data, resulting in unreliable analysis results.

[0004] At the same time, the lack of efficient and intelligent deep learning models for identifying safety hazards makes it impossible to quickly and accurately identify safety hazards in transmission channels in the face of massive amounts of monitoring data, making it difficult to provide early warnings before they occur, posing a significant risk to the safe operation of transmission lines. Therefore, it is urgent to develop a transmission channel safety monitoring system that can provide stable power supply, comprehensive monitoring, accurate analysis, and timely warnings. Summary of the Invention

[0005] The purpose of the present invention is to provide a safe intelligent early warning system for power transmission channels, which solves the technical problems raised in the background technology.

[0006] The purpose of the present invention can be achieved through the following technical solutions: A safety intelligent early warning system for a power transmission channel, comprising: The power supply unit is used to provide power to the monitoring unit, processing unit, and deep learning recognition unit. It uses inductive self-powering technology combined with supercapacitor energy storage. The power supply unit includes: The induction power module uses the changes in the magnetic field on the high-voltage wire to generate electricity; Supercapacitor energy storage module, used to store the generated electrical energy; A monitoring unit is used to perform all-round status monitoring of the power transmission channel and obtain monitoring data of the power transmission channel; A processing unit is used to perform data fusion processing on the monitoring data through multimodal data fusion technology and obtain a comprehensive feature vector of the transmission channel; the processing unit includes: Feature extraction module, used to extract image features from image data, including color, texture and shape features; A preprocessing module is used to normalize the monitoring data obtained by the sensor and obtain the normalized current, surface temperature, ambient temperature and ambient humidity; A data fusion module is used to perform data fusion processing on the normalized current, surface temperature, ambient temperature and ambient humidity; The deep learning recognition unit is used to analyze and identify the fused comprehensive feature vector and extracted image features based on the deep learning model to determine whether there are safety hazards in the transmission channel.

[0007] As a further solution of the present invention: the method of generating and storing electric energy is as follows: First, when the current on the high-voltage wire changes, the strength of the magnetic field around it also changes. The coil in the induction power module generates an induced electromotive force under the change of the magnetic field, thereby generating electrical energy. The output voltage V is calculated as: V out =k×B×A×f; Where: V out represents the output voltage, k is the preset inductance, B represents the pre-monitored magnetic field strength, f is the pre-calculated frequency of magnetic field change, and A is the effective area of ​​the coil; It first passes through the rectifier circuit to convert AC into DC, then passes through the voltage stabilization circuit to stabilize the voltage within the appropriate value range, and then is stored in the supercapacitor; Among them, the energy storage capacity C of the supercapacitor satisfies: C=V g ×Q; Where Q represents the amount of stored charge, V g Represents the operating voltage.

[0008] As a further solution of the present invention: the state monitoring method is as follows: First, image data is acquired by installing two high-definition cameras on two adjacent poles on both sides of the transmission line sag; Among them, the coverage range of the high-definition camera is based on the formula: Calculated; Where D represents the coverage area, h is the camera installation height, and θ is the camera's field of view angle; At the same time, the current of the transmission line is measured by a current sensor provided on the transmission line and marked as L; At the same time, the surface temperature of the transmission line is measured by the temperature sensor installed on the transmission line, and marked as T b ; At the same time, by placing temperature and humidity sensors at a fixed distance from the transmission line, the ambient temperature T of the environment where the transmission line is located is monitored in real time. h and ambient humidity H h .

[0009] As a further solution of the present invention: the color feature extraction method is as follows: The image data is adjusted to an RGB image, and then the pixel values ​​of each pixel in the red, green, and blue channels are extracted and marked as R in turn. k , G k 、B k , where k = 1, 2, ..., e, and e represents the number of pixels in the image data; Then through: ; Calculate the channel mean RP, GP and BP of the image data in the red, green and blue channels respectively; Also through:

[0010] Calculate the channel discrete values ​​RQ, GQ and BQ of the image data in the red, green and blue channels respectively; Among them, the channel mean is used to reflect the overall color bias of the image, and the channel discrete value is used to reflect the degree of color discreteness.

[0011] As a further solution of the present invention: the texture feature extraction method is as follows: Construct a gray-level co-occurrence matrix based on image data, and calculate texture features through the gray-level co-occurrence matrix; Among them, the gray-level co-occurrence matrix describes the spatial distribution relationship of pixels with different gray levels in the image; pass: ; Calculate the contrast DB in the texture feature; Where i and j refer to two different grayscale pixels in the image data, and i and j are variable values, and i≠j; |ij| 2 It is expressed as the square of the difference between two gray levels i and j, reflecting the magnitude of the gray level difference, P i，j It is expressed as the probability value in the gray-level co-occurrence matrix, which refers to the probability that the two gray levels i and j appear at the same time.

[0012] As a further solution of the present invention: the shape feature extraction method is as follows: Used to perform edge detection on image data using the Canny method and ultimately determine the edge image; Then, for the obtained edge image, by traversing all the pixels on the edge image, the number of pixels contained in the contour is counted and used as the contour area of ​​the edge image; at the same time, the distance between adjacent pixels is calculated using the Euclidean distance formula, and then the distances between all adjacent pixels are summed and recorded as the contour perimeter; Then, the circularity Y of the edge image in the relevant image data is calculated using the circularity formula: Y = (4π × S) / C; Where S is the contour area of ​​the edge image, and C is the contour perimeter of the edge image; At the same time, the area-perimeter ratio B of the edge image in the relevant image data is calculated using the formula: B=S / C.

[0013] As a further solution of the present invention: the normalization processing method is as follows: Subtract the lower limit value of the normal operation from the current, surface temperature, ambient temperature or ambient humidity of the transmission line, and divide the difference by the difference between the upper limit value and the lower limit value during normal operation. The obtained value is the normalized current, surface temperature, ambient temperature and ambient humidity. The normalization formula is: ; Where X1 represents the normalized current, surface temperature, ambient temperature and humidity, and X represents the current, surface temperature, ambient temperature and humidity of the transmission line, that is, X={L, T b 、T h 、H h};X min and X max They are respectively the normal operating lower limit value and the normal operating upper limit value of the corresponding monitoring data of the current, surface temperature, ambient temperature and ambient humidity of the transmission line during normal operation.

[0014] As a further solution of the present invention: data fusion processing is to use a linear weighting method to fuse the normalized current, surface temperature, ambient temperature and ambient humidity into a comprehensive feature vector, the calculation formula of which is:

[0015] Where γ1, γ2, γ3, and γ4 are the weight coefficients preset based on the normalized current, surface temperature, ambient temperature, and ambient humidity, respectively; Z is the comprehensive eigenvector; L1, T1 b 、T1 h 、H1 h are the normalized current, surface temperature, ambient temperature, and ambient humidity, respectively.

[0016] As a further solution of the present invention: the deep learning model training process is as follows: Step K1: Collect image data of the power transmission channel in different scenarios and corresponding sensor monitoring data, and divide them into 70% training set, 20% validation set, and 10% test set; Among them, the image data and sensor monitoring data in different scenarios are all provided with annotation information, wherein the annotation information includes whether there are safety hazards and the specific type of hazards; Step K2: Build a CNN model and input the image data and sensor monitoring data into the CNN model. The model extracts and classifies the image data and sensor monitoring data through the convolution layer, pooling layer, and fully connected layer in sequence. Specifically: Step K2.1, in the convolution layer, convolution operation is performed by sliding the convolution kernel on the image data and sensor monitoring data to extract the comprehensive feature vector and image features; Step K2.2, the pooling layer downsamples the features output by the convolutional layer; Downsampling uses maximum pooling, which is calculated as follows: within a pooling window of a specified size, the maximum value of the pixels in the window is taken as the pooling output value; Step K2.3: Input the image feature vector processed by the feature extraction module and the comprehensive feature vector output by the data fusion module into the fully connected layer. The fully connected layer is expressed as Perform linear transformation; Among them, SC is the output value, β r is the weight, SR r is the input value, PZ is the bias value, and v is the input vector dimension; Step K3, then compare the output value with the preset judgment threshold: When the output value is greater than the judgment threshold, the transmission channel has a safety hazard corresponding to the relevant labeled information and the specific type of hazard; then an early warning is triggered, and the relevant labeled information is pushed to the patrol personnel's mobile phone APP and the middle office system via the wireless network; When the output value is less than or equal to the judgment threshold, there is no safety hazard in the transmission channel.

[0017] Beneficial effects of the present invention: Power supply advantages: It adopts the power supply method of inductive power supply technology combined with supercapacitor energy storage, and uses the magnetic field changes on the high-voltage wire to generate electricity. It does not require an additional external power supply, reducing power supply costs and maintenance difficulties. The supercapacitor energy storage module can effectively store electricity, ensuring the continuous and stable power supply of the system.

[0018] All-round monitoring: The monitoring unit conducts all-round status monitoring of the transmission channel through high-definition cameras, current sensors, temperature sensors, temperature and humidity sensors, etc., and obtains rich monitoring data, including image data, transmission line current, surface temperature, ambient temperature and ambient humidity, etc., which can fully reflect the operating status of the transmission channel.

[0019] Multimodal data fusion: The processing unit uses multimodal data fusion technology to fuse different types of monitoring data. The feature extraction module extracts color, texture, and shape features from the image. The preprocessing module normalizes the sensor data. The data fusion module uses a linear weighting method to fuse the normalized sensor data into a comprehensive feature vector. This improves data utilization efficiency and analysis accuracy, and can produce a more comprehensive and accurate comprehensive feature vector for the transmission channel.

[0020] Accurate feature extraction: In terms of color feature extraction, by calculating the mean and discrete values ​​of the image in the red, green, and blue channels, it can accurately reflect the bias and discreteness of the overall color of the image; texture feature extraction calculates the contrast by constructing a grayscale co-occurrence matrix, effectively describing the spatial distribution relationship of pixels of different gray levels in the image; shape feature extraction uses the Canny method for edge detection and calculates parameters such as contour area, perimeter, circularity, and area-to-perimeter ratio, which can accurately extract the shape features of the image. These feature extraction methods provide strong support for subsequent analysis and identification.

[0021] Efficient deep learning recognition: The deep learning recognition unit analyzes and identifies the fused comprehensive feature vectors and extracted image features based on a deep learning model, accurately determining whether there are safety hazards in the transmission channel. The model is trained by rationally dividing the training set, validation set, and test set, and effective processing methods such as convolution operations, maximum pooling, and linear transformations are used in the model's convolutional layer, pooling layer, and fully connected layer to improve the model's recognition accuracy and efficiency. When a safety hazard is detected, it can trigger an early warning and push the relevant annotation information to the inspector's mobile app and the middle-office system, facilitating timely measures to ensure the safe operation of the transmission channel. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The present invention will be further described below with reference to the accompanying drawings.

[0023] Figure 1 This is a system block diagram of a safe intelligent early warning system for power transmission channels of the present invention.

[0024] Figure 2 This is a system block diagram of a monitoring unit in a safe intelligent early warning system for a power transmission channel of the present invention.

[0025] Figure 3It is a system block diagram of a processing unit in a safe intelligent early warning system for a power transmission channel of the present invention. DETAILED DESCRIPTION

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0027] Example 1 See also Figure 1 、 Figure 2 、 Figure 3 As shown, the present invention is a safety intelligent early warning system for power transmission channels, comprising: A monitoring unit is used to perform all-round status monitoring of the power transmission channel and obtain monitoring data of the power transmission channel; The status monitoring method is as follows: First, image data is acquired by installing two high-definition cameras on two adjacent poles on both sides of the transmission line sag; Among them, the coverage range of the high-definition camera is based on the formula: Calculated; Where D represents the coverage area, h is the camera installation height, and θ is the camera's field of view angle; In this embodiment, the higher the high-definition camera is installed, the larger the coverage range D will be when the field of view angle θ remains unchanged; The larger the field of view, the larger the coverage area D; The HD camera uses a HD image sensor with a resolution of no less than 1920×1080 pixels and a frame rate of 30fps, which means the camera can capture 30 clear images per second; A processing unit, used to extract features from the monitoring data and derive image features of the power transmission channel; The processing unit includes: Feature extraction module, used to extract image features from image data, including color, texture and shape features; Color feature extraction: The image data is adjusted to an RGB image, and then the pixel values ​​of each pixel in the red, green, and blue channels are extracted and marked as R in turn. k , G k 、B k , where k = 1, 2, ..., e, and e represents the number of pixels in the image data; Then through: ; Calculate the channel mean RP, GP and BP of the image data in the red, green and blue channels respectively; Also through:

[0028] Calculate the channel discrete values ​​RQ, GQ and BQ of the image data in the red, green and blue channels respectively; Among them, the channel mean is used to reflect the overall color bias of the image, and the channel discrete value is used to reflect the degree of color discreteness; In this embodiment, if the mean value of the blue channel in a certain transmission channel image is high, it may indicate that the image as a whole has a cool tone; a large variance indicates that the color distribution is more dispersed and the image has high color richness; Texture feature extraction: Construct a gray-level co-occurrence matrix based on image data, and calculate texture features through the gray-level co-occurrence matrix; In this embodiment, the construction of the gray-level co-occurrence matrix is ​​a prior art and is therefore not described in detail. Among them, the gray-level co-occurrence matrix describes the spatial distribution relationship of pixels with different gray levels in the image; pass: ; Calculate the contrast DB in the texture feature; Where i and j refer to two different grayscale pixels in the image data, and i and j are variable values, and i≠j; |ij| 2 It is expressed as the square of the difference between two gray levels i and j, reflecting the magnitude of the gray level difference, P i，j It is expressed as the probability value in the gray-level co-occurrence matrix, which refers to the probability of the two gray levels i and j appearing at the same time; In this embodiment, the greater the contrast, the more dramatic the change in grayscale in the image data and the rougher the texture; the smaller the contrast, the finer the image texture. For example, in an image of an industrial power transmission scene, if the texture contrast of the conductor part is small, it means that the surface of the conductor is relatively smooth; while the texture contrast of the surrounding vegetation area is large, indicating that the texture of the vegetation is more complex.

[0029] Shape feature extraction: It is used to perform edge detection on image data using the Canny method and ultimately determine an edge image. The Canny method is an existing technology, so the specific method of edge detection is not described in detail. Then, for the obtained edge image, by traversing all the pixels on the edge image, the number of pixels contained in the contour is counted and used as the contour area of ​​the edge image; at the same time, the distance between adjacent pixels is calculated using the Euclidean distance formula, and then the distances between all adjacent pixels are summed and recorded as the contour perimeter; Then, the circularity Y of the edge image in the relevant image data is calculated using the circularity formula: Y = (4π × S) / C; Where S is the contour area of ​​the edge image, and C is the contour perimeter of the edge image; At the same time, the area-perimeter ratio B of the edge image in the relevant image data is calculated using the formula: B=S / C; A deep learning recognition unit is used to analyze and identify the fused comprehensive feature vector and extracted image features based on a deep learning model to determine whether there are safety hazards in the power transmission channel; The deep learning model training process is as follows: Step K1: Collect image data of the power transmission channel in different scenarios and divide it into 70% training set, 20% validation set, and 10% test set; The image data in different scenarios are provided with annotation information, including whether there are safety hazards and the specific types of hazards, such as trees close to the lines, damaged lines, and foreign objects hanging. Step K2: Build a CNN model and input the image data into the CNN model. The model extracts and classifies the image features through the convolution layer, pooling layer, and fully connected layer in sequence. Step K2.1, in the convolution layer, the convolution kernel is slid on the image data to perform convolution operation and extract local features of the image data, such as edges and textures; Step K2.2, the pooling layer downsamples the features output by the convolutional layer; Downsampling uses maximum pooling, which is calculated as follows: within a pooling window of a specified size, the maximum value of the pixels in the window is taken as the pooling output value; Step K2.3, input the image feature vector processed by the feature extraction module into the fully connected layer; The fully connected layer is expressed as Perform linear transformation; Among them, SC is the output value, β r is the weight, SR r is the input value, PZ is the bias value, and v is the input vector dimension; Step K3, then compare the output value with the preset judgment threshold: When the output value is greater than the judgment threshold, the transmission channel has a safety hazard corresponding to the relevant labeled information and the specific type of hazard; then an early warning is triggered, and the relevant labeled information is pushed to the patrol personnel's mobile phone APP and the middle office system via the wireless network; In this example: When a conductor break is detected, the system immediately issues a red warning signal and notifies the transmission line operation and maintenance personnel via SMS or email. It also provides the specific location of the break and recommends that professional maintenance personnel be dispatched to the site for repairs as soon as possible. If a foreign object is detected hanging on the transmission line, the system will issue an orange or yellow warning based on the size, location and impact of the foreign object on power transmission safety; When the foreign object is small and has little impact on power transmission safety, a yellow warning is issued to remind operation and maintenance personnel to pay close attention to the dynamics of the foreign object; When a foreign object is large and may affect power transmission stability, an orange warning will be issued, along with a recommendation to remove the foreign object as soon as possible and detailed location information of the transmission channel where the foreign object is located. In the event of abnormal parameters such as transmission line current and temperature, different levels of warnings will be issued according to the degree of exceeding the normal range; When the current or temperature slightly exceeds the normal range, a blue warning is issued to prompt operation and maintenance personnel to strengthen monitoring of the transmission channel; When the range is exceeded significantly, a red alert is issued and a decision suggestion is given for taking emergency measures, such as adjusting the transmission load; When the output value is less than or equal to the judgment threshold, there is no safety hazard in the transmission channel; In this embodiment, the output value of the CNN model is a probability value between 0 and 1. For example, an output value of 0.8 indicates that the model believes that there is an 80% probability that the transmission channel has a safety hazard; an output value of 0.2 indicates that there is a 20% probability that the safety hazard exists. This embodiment utilizes high-definition cameras to comprehensively capture transmission channel image data, determining coverage based on a formula based on installation height and field of view. During processing, multiple algorithms are employed to extract color, texture, and shape features from the image, providing accurate data for subsequent analysis. By building and training a CNN model and dividing it into training, validation, and test sets, it is possible to identify safety hazards in transmission channels. Once a hazard is detected, the system can push targeted warnings to inspectors' mobile apps and the central office system based on its type and severity, significantly improving the efficiency of detecting and responding to hazard detection in transmission channels.

[0030] Example 2 See also Figure 1 、 Figure 2 、 Figure 3 As shown, as the second embodiment of the present invention, when the present application is specifically implemented, compared with the first embodiment, the technical solution of this embodiment is different from that of the first embodiment only in that in this embodiment, the state monitoring method is also as follows: The current of the transmission line is measured by a current sensor provided on the transmission line and marked as L; In this embodiment, the current sensor uses the principle of electromagnetic induction to convert the current signal in the transmission wire into a measurable electrical signal. After signal processing and calibration, it accurately outputs the wire current value. For example, when the wire current is 100A, the measured value will be between 99A and 101A. At the same time, the surface temperature of the transmission line is measured by the temperature sensor installed on the transmission line, and marked as T b ; In this embodiment, the temperature sensor adopts a contact or non-contact temperature measurement principle; wherein the contact type is to obtain the temperature by direct contact with the surface of the wire; the non-contact type is such as the infrared temperature measurement principle, which measures the temperature by detecting infrared radiation radiated from the surface of the wire; At the same time, by placing temperature and humidity sensors at a fixed distance from the transmission line, the ambient temperature T of the environment where the transmission line is located is monitored in real time. h and ambient humidity H h ; In this embodiment, the ambient temperature sensor senses ambient temperature changes through a temperature sensing element and converts the temperature changes into electrical signal changes to measure the temperature; Environmental humidity sensors use materials that are sensitive to water molecules to measure humidity based on changes in the physical or chemical properties of the material caused by changes in humidity. The processing unit is further used to perform data fusion processing on the monitoring data through multimodal data fusion technology and obtain a comprehensive feature vector of the transmission channel; The processing unit further includes: A preprocessing module is used to normalize the monitoring data obtained by the sensor; Among them, normalization is used to eliminate the dimensional differences between different physical quantities, so that the current, surface temperature, ambient temperature and ambient humidity of the transmission line can be compared and calculated on the same scale; Normalization is done by the formula: To achieve; Where X1 represents the normalized current, surface temperature, ambient temperature and humidity, and X represents the current, surface temperature, ambient temperature and humidity of the transmission line, that is, X={L, T b 、T h 、H h};X min and X max They are respectively the normal operating lower limit value and the normal operating upper limit value of the corresponding monitoring data of the current, surface temperature, ambient temperature and ambient humidity of the transmission line during normal operation; For example, a set of temperature data has a minimum value of 10°C and a maximum value of 30°C. If the original temperature data X is 20°C, then the normalized data X1 = (20-10) / (30-10) = 0.5; A data fusion module is used to perform data fusion processing on the normalized current, surface temperature, ambient temperature and ambient humidity; The data fusion process uses a linear weighted method to fuse the normalized current, surface temperature, ambient temperature, and ambient humidity into a comprehensive feature vector. The calculation formula is:

[0031] Where γ1, γ2, γ3, and γ4 are the weight coefficients preset based on the normalized current, surface temperature, ambient temperature, and ambient humidity, respectively; Z is the comprehensive eigenvector; L1, T1 b 、T1 h 、H1 h are the normalized current, surface temperature, ambient temperature and ambient humidity respectively; A deep learning recognition unit is used to analyze and identify the fused comprehensive feature vector based on the deep learning model to determine whether there are safety hazards in the power transmission channel; The deep learning model training process is as follows: Step K1: Collect monitoring data from several sensors on the power transmission channel in different scenarios and divide it into 70% training set, 20% validation set, and 10% test set. Among them, sensor monitoring data in different scenarios are provided with annotation information; Step K2: Build a CNN model and input the image data into the CNN model. The model sequentially extracts and classifies the sensor monitoring data through the convolution layer, pooling layer, and fully connected layer. Step K2.1, in the convolution layer, convolution operation is performed by sliding the convolution kernel on the sensor monitoring data to extract the local features of the sensor monitoring data; Step K2.2, the pooling layer downsamples the features output by the convolutional layer; Downsampling uses maximum pooling, which is calculated as follows: within a pooling window of a specified size, the maximum value of the pixels in the window is taken as the pooling output value; Step K2.3, input the comprehensive feature vector output by the data fusion module into the fully connected layer; The fully connected layer is expressed as Perform linear transformation; Among them, SC is the output value, β r is the weight, SR r is the input value, PZ is the bias value, and v is the input vector dimension; Step K3, then compare the output value with the preset judgment threshold: When the output value is greater than the judgment threshold, the transmission channel has a safety hazard corresponding to the relevant labeled information and the specific type of hazard; then an early warning is triggered, and the relevant labeled information is pushed to the patrol personnel's mobile phone APP and the middle office system via the wireless network; Example 2 builds on Example 1 by adding sensors for current, temperature, and humidity to collect multidimensional data. This data is fused through normalization and linear weighting to eliminate dimensional differences and form a more representative comprehensive feature vector. A CNN model is trained based on this fused data to further optimize the identification of safety hazards in power transmission corridors, providing richer information for safety status assessments and improving the comprehensiveness and reliability of system warnings.

[0032] Example 3 See also Figure 1 、 Figure 2 、 Figure 3 As shown, as the third embodiment of the present invention, when the present application is specifically implemented, compared with the first and second embodiments, the technical solution of this embodiment is to combine the solutions of the first and second embodiments. The difference between the technical solution of this embodiment and the first and second embodiments is only in this embodiment; A power supply unit, used to provide power to the monitoring unit, processing unit, and deep learning recognition unit; It uses inductive power supply technology combined with supercapacitor energy storage. The power supply unit includes: The induction power module uses the changes in the magnetic field on the high-voltage wire to generate electricity; Electricity is generated in the following ways: First, when the current on the high-voltage wire changes, the strength of the magnetic field around it also changes. The coil in the induction power module generates an induced electromotive force under the change of the magnetic field, thereby generating electrical energy. The output voltage V is calculated as: V out =k×B×A×f; in: V out Represents the output voltage; k is the inductance, which is related to the number of coil turns and the material properties of the coil. For example, under the same magnetic field change, the more coil turns there are, the greater the inductance k, and the higher the generated voltage V is likely to be. Coils made of different materials also have different inductive abilities to magnetic fields, which also affects the k value. B represents the magnetic field strength. The greater the magnetic field strength, the higher the voltage V generated. f is the frequency of magnetic field change. The higher the frequency, the higher the voltage V. A is the effective area of ​​the coil. The larger the area, the greater the voltage V generated at the same magnetic field change frequency f and magnetic field strength B; Supercapacitor energy storage module, used to store the generated electrical energy; It first passes through the rectifier circuit to convert AC into DC, then passes through the voltage stabilization circuit to stabilize the voltage within the appropriate value range, and then is stored in the supercapacitor; Among them, the energy storage capacity C of the supercapacitor satisfies: C=V g ×Q; Where Q represents the amount of stored charge, V g Represents the operating voltage; For example, at the operating voltage V g Under fixed conditions, the more charge Q is stored, the greater the energy storage capacity C of the supercapacitor. This embodiment adopts a power supply method that combines inductive power self-powering technology with supercapacitor energy storage. Electric energy is generated by the change of the magnetic field of the high-voltage wire, and is stored in the supercapacitor after rectification and voltage stabilization. This power supply method abandons chemical batteries, avoids the need to consider complex issues such as line laying and maintenance when wiring traditional power supplies, and the cost is relatively low; at the same time, it also solves the problem that solar power supply is limited by environmental factors such as weather and light time. Regardless of day or night, sunny or cloudy, as long as there is a current change in the high-voltage wire, power supply can be continued; the power supply unit ensures that the entire early warning system operates stably around the clock, and realizes all-round accurate monitoring and efficient early warning of transmission lines.

[0033] Example 4 See also Figure 1 、 Figure 2 、 Figure 3 As shown, as the fourth embodiment of the present invention, when this application is specifically implemented, compared with the first, second and third embodiments, the technical solution of this embodiment is to combine the solutions of the above-mentioned first, second and third embodiments.

[0034] The fourth embodiment integrates the advantages of the previous three embodiments to construct an intelligent early warning system for power transmission channel safety, integrating multi-dimensional data collection, precise feature extraction, efficient hazard identification, and stable power supply. The system comprehensively collects images and various sensor data, accurately identifies hazards through deep learning models, and provides efficient early warnings under a stable power supply, comprehensively ensuring the safe and stable operation of power transmission channels and significantly reducing the risk of power transmission accidents.

[0035] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0036] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A safety intelligent early warning system for power transmission channels, characterized in that: include: A power supply unit, used to provide power to the monitoring unit, processing unit, and deep learning recognition unit; It uses inductive power self-powering technology combined with supercapacitor energy storage; A monitoring unit is used to perform all-round status monitoring of the power transmission channel and obtain monitoring data of the power transmission channel; A processing unit is used to extract features from the monitoring data and obtain image features of the transmission channel, and to obtain a comprehensive feature vector of the transmission channel by performing data fusion processing on the monitoring data; The deep learning recognition unit is used to analyze and identify the fused comprehensive feature vector and extracted image features based on the deep learning model to determine whether there are safety hazards in the transmission channel.

2. The intelligent safety early warning system for power transmission channels according to claim 1, characterized in that: The power supply unit includes: The induction power module uses the changes in the magnetic field on the high-voltage wire to generate electricity; Supercapacitor energy storage module, used to store the generated electrical energy; The processing unit includes: Feature extraction module, used to extract image features from image data, including color, texture and shape features; A preprocessing module is used to normalize the monitoring data obtained by the sensor and obtain the normalized current, surface temperature, ambient temperature and ambient humidity; The data fusion module is used to perform data fusion processing on the normalized current, surface temperature, ambient temperature and ambient humidity.

3. The intelligent safety warning system for power transmission channels according to claim 2, characterized in that: Electricity is generated and stored in the following ways: First, when the current on the high-voltage wire changes, the strength of the magnetic field around it also changes. The coil in the induction power module generates an induced electromotive force under the change of the magnetic field, thereby generating electrical energy. The output voltage V is calculated as: V out =k×B×A×f; Where: V out represents the output voltage, k is the preset inductance, B represents the pre-monitored magnetic field strength, f is the pre-calculated frequency of magnetic field change, and A is the effective area of ​​the coil; It first passes through the rectifier circuit to convert AC into DC, then passes through the voltage stabilizing circuit to stabilize the voltage, and then is stored in the supercapacitor; Among them, the energy storage capacity C of the supercapacitor satisfies: C=V g ×Q; Where Q represents the amount of stored charge, V g Represents the operating voltage.

4. The intelligent safety warning system for power transmission channels according to claim 1, characterized in that: The status monitoring method is as follows: First, image data is acquired by installing two high-definition cameras on two adjacent poles on both sides of the transmission line sag; Among them, the coverage range of the high-definition camera is based on the formula: Calculated; Where D represents the coverage area, h is the camera installation height, and θ is the camera's field of view angle; At the same time, the current of the transmission line is measured by the current sensor installed on the transmission line, and it is marked as L; the surface temperature of the transmission line is measured by the temperature sensor installed on the transmission line, and it is marked as T b ; By placing temperature and humidity sensors at a fixed distance from the transmission line, and monitoring the ambient temperature T of the environment where the transmission line is located in real time h and ambient humidity H h .

5. The intelligent safety warning system for power transmission channels according to claim 2, characterized in that: Color feature extraction is as follows: The image data is adjusted to an RGB image, and then the pixel values ​​of each pixel in the red, green, and blue channels are extracted, and the channel mean of the image data in the red, green, and blue channels are calculated respectively; At the same time, the standard deviation of the pixel values ​​and channel means of the image data in the red, green and blue channels is calculated to obtain the channel discrete values ​​of the image data in the red, green and blue channels respectively.

6. The intelligent safety warning system for power transmission channels according to claim 2, characterized in that: The texture feature extraction method is as follows: Construct a gray-level co-occurrence matrix based on image data, and calculate texture features through the gray-level co-occurrence matrix; pass: ; Calculate the contrast DB in the texture feature; Where i and j refer to two different grayscale pixels in the image data, and i and j are variable values, and i≠j; |ij| 2 It is expressed as the square of the difference between two gray levels i and j, reflecting the magnitude of the gray level difference, P i，j Represents the probability value in the gray-level co-occurrence matrix, that is, the probability that the two gray levels i and j appear at the same time.

7. The intelligent safety warning system for power transmission channels according to claim 2, characterized in that: The shape feature extraction method is as follows: Used to perform edge detection on image data using the Canny method and ultimately determine the edge image; Then, for the obtained edge image, by traversing all the pixels on the edge image, the number of pixels contained in the contour is counted and used as the contour area of ​​the edge image; at the same time, the distance between adjacent pixels is calculated using the Euclidean distance formula, and then the distances between all adjacent pixels are summed and recorded as the contour perimeter; Then, the contour area of ​​the edge image is multiplied by 4π, and the result is divided by the contour perimeter of the edge image to obtain the circularity Y of the edge image in the relevant image data; At the same time, the area-perimeter ratio of the edge image in the relevant image data is calculated by dividing the contour surface of the edge image by the contour perimeter.

8. The intelligent safety warning system for power transmission channels according to claim 2, characterized in that: The normalization process is as follows: Subtract the lower limit value during normal operation from the current, surface temperature, ambient temperature or ambient humidity of the transmission line, and divide the difference by the difference between the upper and lower limits during normal operation. The resulting value is the normalized current, surface temperature, ambient temperature and ambient humidity.

9. The intelligent safety warning system for power transmission channels according to claim 8, characterized in that: The data fusion process uses a linear weighted method to fuse the normalized current, surface temperature, ambient temperature, and ambient humidity into a comprehensive feature vector. The calculation formula is: ; Where γ1, γ2, γ3, and γ4 are the weight coefficients preset based on the normalized current, surface temperature, ambient temperature, and ambient humidity, respectively; Z is the comprehensive eigenvector; L1, T1 b 、T1 h 、H1 h are the normalized current, surface temperature, ambient temperature, and ambient humidity, respectively.

10. The intelligent safety warning system for power transmission channels according to claim 1, characterized in that: The deep learning model training process is as follows: Step K1: Collect image data of the power transmission channel in different scenarios and corresponding sensor monitoring data, and divide them into 70% training set, 20% validation set, and 10% test set; Among them, the image data and sensor monitoring data in different scenarios are all provided with annotation information, wherein the annotation information includes whether there are safety hazards and the specific type of hazards; Step K2: Build a CNN model and input the image data and sensor monitoring data into the CNN model. The CNN model extracts and classifies the image data and sensor monitoring data through the convolution layer, pooling layer, and fully connected layer in sequence, and obtains the output value of the CNN model. Step K3, then compare the output value with the preset judgment threshold: When the output value is greater than the judgment threshold, it is determined that there are safety hazards corresponding to the relevant labeled information and the specific type of hazard in the transmission channel; then an early warning is triggered, and the relevant labeled information is pushed to the inspection personnel's mobile phone APP and the middle office system via the wireless network; When the output value is less than or equal to the judgment threshold, it is determined that there is no safety hazard in the transmission channel.

Images