Ice hazard early warning method, system and medium based on open neural network exchange algorithm

Abstract

The present invention discloses an ice damage warning method, system, and medium based on an open neural network exchange algorithm, which can be applied to the field of high-voltage power line safety technology. The method of the present invention uses a binocular camera to collect transmission line images in real time and preprocesses the transmission line images to improve subsequent recognition and detection accuracy. The preprocessed transmission line images are then input into a multi-scale target detection network to obtain the ice thickness of the transmission line at the current moment. At the same time, the preprocessed transmission line images are subjected to ice type recognition to obtain the ice type of the transmission line at the current moment. Then, based on micro-meteorological information combined with the current ice thickness and ice type, the ice condition of the transmission line at the next moment is predicted through a memory network. An ice damage warning signal is generated based on the prediction result to realize an automatic ice warning function, improve the timeliness of ice condition detection, and effectively reduce the probability of transmission line accidents.

Description

Technical Field

[0001] The present invention relates to the technical field of high-voltage power line safety, and in particular to an ice hazard early warning method, system and medium based on an open neural network exchange algorithm. Background Art

[0002] In related technologies, when transmission lines are located at the confluence of cold and warm currents, coupled with factors such as the crisscrossing rivers, topography, and climate, a large portion of overhead transmission lines are located in microclimate zones characterized by low temperatures, high humidity, and strong winds, making them susceptible to icing disasters. Current early warning methods for icing disasters rely on manual line inspections, which can lead to delayed detection of icing and increase the probability of transmission line accidents. Summary of the Invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes an ice damage early warning method, system, and medium based on an open neural network exchange algorithm, which can effectively reduce the probability of transmission line accidents.

[0004] In one aspect, an embodiment of the present invention provides an ice disaster early warning method based on an open neural network exchange algorithm, comprising the following steps:

[0005] Obtaining transmission line images captured by binocular cameras in real time;

[0006] Preprocessing the transmission line image;

[0007] Inputting the preprocessed transmission line image into a multi-scale target detection network to obtain the ice thickness of the transmission line at the current moment;

[0008] Identifying the ice type of the pre-processed transmission line image to obtain the ice type of the transmission line at a current moment;

[0009] Obtain micro-meteorological information collected in real time by sensor modules in the area where the transmission line belongs;

[0010] According to the ice thickness at the current moment, the ice type at the current moment, and the micrometeorological information, predicting the ice condition of the transmission line at the next moment through a memory network;

[0011] An ice damage warning signal is generated according to the icing condition of the transmission line at the next moment.

[0012] In some embodiments, preprocessing the transmission line image includes:

[0013] Defogging the transmission line image using an image defogging algorithm;

[0014] The processing of the image defogging algorithm includes:

[0015] The power transmission line image is input into a preset cyclic generative adversarial network, and the power transmission line image is dehazed by color loss and feature loss within the preset cyclic generative adversarial network.

[0016] In some embodiments, the pre-processing of the transmission line image further includes:

[0017] The de-fogging image of the power transmission line is input into a convolutional neural network to obtain ice-covered image features.

[0018] In some embodiments, the activation layer of the convolutional neural network includes a support vector machine; and identifying the ice type of the preprocessed transmission line image includes:

[0019] The ice cover image features are input into the support vector machine to identify the ice cover type.

[0020] In some embodiments, obtaining micro-meteorological information collected in real time by a sensor module in an area to which the power transmission line belongs includes:

[0021] Obtain at least one of the following: ambient temperature, relative humidity, air pressure, wind speed, wind direction, rain, snow, or illumination, which are collected in real time by a sensor module in the area to which the transmission line belongs.

[0022] In some embodiments, the step of predicting the icing condition of the power transmission line at the next moment through a memory network based on the ice thickness at the current moment, the ice type at the current moment, and the micrometeorological information includes:

[0023] analyzing ice formation conditions in real time based on the micro-meteorological information;

[0024] According to the ice formation conditions, the ice thickness at the current moment, and the ice type at the current moment, the ice condition of the transmission line at the next moment is predicted through a memory network.

[0025] In some embodiments, before predicting the transmission line icing condition at the next moment through the memory network, the method further includes:

[0026] Obtain the thickness of the transmission line, the acceleration of the transmission line and the inclination of the tower;

[0027] The weighting coefficient of the memory network is adjusted according to the thickness of the transmission line, the acceleration of the transmission line and the inclination of the tower.

[0028] In another aspect, an embodiment of the present invention provides an ice disaster early warning system based on an open neural network exchange algorithm, comprising:

[0029] The first acquisition module is used to acquire the transmission line image collected by the binocular camera in real time;

[0030] A preprocessing module, configured to preprocess the transmission line image;

[0031] a detection module, configured to input the preprocessed transmission line image into a multi-scale target detection network to obtain the ice thickness of the transmission line at a current moment;

[0032] an identification module, configured to identify the ice type of the transmission line after preprocessing to obtain the ice type of the transmission line at a current moment;

[0033] The second acquisition module is used to obtain micro-meteorological information collected in real time by the sensor module in the area to which the transmission line belongs;

[0034] A prediction module, configured to predict the icing condition of the transmission line at the next moment through a memory network based on the ice thickness at the current moment, the ice type at the current moment, and the micrometeorological information;

[0035] A generating module is used to generate an ice damage warning signal according to the icing condition of the transmission line at the next moment.

[0036] In another aspect, an embodiment of the present invention provides an ice disaster early warning system based on an open neural network exchange algorithm, comprising:

[0037] at least one memory for storing a program;

[0038] At least one processor is configured to load the program to execute the ice hazard early warning method based on the open neural network exchange algorithm.

[0039] On the other hand, an embodiment of the present invention provides a storage medium storing a computer-executable program. When the computer-executable program is executed by a processor, it is used to implement the ice hazard early warning method based on the open neural network exchange algorithm.

[0040] An ice disaster early warning method based on an open neural network exchange algorithm provided by an embodiment of the present invention has the following beneficial effects:

[0041] This embodiment uses a binocular camera to collect transmission line images in real time and preprocesses the transmission line images to improve subsequent recognition and detection accuracy. The preprocessed transmission line images are then input into a multi-scale target detection network to obtain the ice thickness of the transmission line at the current moment. At the same time, the ice type of the preprocessed transmission line images is identified to obtain the ice type of the transmission line at the current moment. Then, based on the real-time micrometeorological information combined with the ice thickness and ice type at the current moment, the icing situation of the transmission line at the next moment is predicted through a memory network. An ice damage warning signal is generated based on the prediction result to realize the automatic icing warning function, improve the timeliness of icing situation detection, and effectively reduce the probability of transmission line accidents.

[0042] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0044] Figure 1 The figure is a flow chart of an ice disaster early warning method based on an open neural network exchange algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0046] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on the present invention.

[0047] In the description of the present invention, "several" means more than one, "plurality" means more than two, "greater than," "less than," and "exceed" are understood to exclude the number itself, while "above," "below," and "within" are understood to include the number itself. The use of "first" and "second" in the description is solely for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, implicitly specifying the number of the indicated technical features, or implicitly specifying the order of the indicated technical features.

[0048] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, and connecting should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0049] In the description of the present invention, reference to terms such as "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the exemplary expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0050] Transmission line icing types can be classified into four categories:

[0051] Rime: When cold water droplets fall onto a cold surface, they form a glassy, ​​smooth, transparent layer of ice. This is called rime. The surface of rime is smooth, and its cross-section is elliptical or wedge-shaped. Its specific structure is closely related to wind direction. Its most notable physical properties are high density, high hardness, and strong adhesion. Rime density is usually greater than 0.8g / cm 2 It can be said that these physical properties determine that the mechanical load caused by rime is far more severe than other forms of icing. Correspondingly, its damage to lines and insulators is the greatest among all types of icing.

[0052] Rime: Rime can be soft or hard. Soft rime, also known as granular rime, often forms in temperatures between -3°C and -8°C. When the surrounding air is foggy or drizzling, the humidity is very high. The abundant water vapor adheres to the conductors and may condense, forming radial crystals. The density of soft rime is 0.1g / cm 2 ~0.5g / cm 2 Its most notable physical properties are weak adhesion and strong granularity. Hard rime, also known as rime or mixed rime, forms when the ambient temperature is between -10°C and -20°C. Supercooled water droplets fall on the windward side of the conductor. The resulting icing becomes a mixture of rime and soft rime, losing its transparency. Hard rime is hard and has strong adhesion.

[0053] Snow accumulation: Snow falls and adheres to the conductor to form snow. Snow accumulation can be divided into dry snow and wet snow. Dry snow is fluffy and dry, and its density is less than 0.1g / cm 3, easy to scatter, not easy to adhere to the wire; the density of wet snow is 0.1g / cm3~0.5g / cm3, and its adhesion is stronger than dry snow, but weaker than rime.

[0054] Hoar frost: When humidity is high, water vapor condenses on surfaces below 0°C. Hoar frost generally increases the corona loss of conductors.

[0055] The current early warning method for icing disasters relies on manual line inspections. However, manual inspections may result in untimely detection and cause accidents. In addition, the areas where high-voltage and ultra-high-voltage transmission lines are distributed have complex terrain and harsh environments, making line inspection work very difficult.

[0056] Based on this, refer to Figure 1 The embodiment of the present invention provides an ice disaster early warning method based on an open neural network exchange algorithm, comprising the following steps:

[0057] Step 110: Acquire the transmission line image captured in real time by the binocular camera. Specifically, the binocular camera can be pre-set in the area of ​​the transmission line to be monitored, and its shooting range includes the transmission line to be monitored.

[0058] Step 120: Preprocess the transmission line image.

[0059] In the embodiment of the present application, due to low visibility in snowy weather, the image quality captured by the binocular camera also degrades. Therefore, to improve the accuracy of subsequent processing steps, the real-time power line images need to be processed. Specifically, an image dehazing algorithm can be used to dehaze the power line images. The image dehazing algorithm includes inputting the power line images into a preset recurrent generative adversarial network, and then dehazing the power line images using the color loss and feature loss within the preset recurrent generative adversarial network. The preset recurrent generative adversarial network does not require paired data for training. During the image processing process, the color loss between the haze-free image generated by the generator and the true haze-free image is optimized, enabling the generator to generate an image with the same color distribution as the haze-free image. Furthermore, the introduction of feature loss solves the image distortion problem that occurs during image conversion in classic recurrent generative adversarial networks, better restoring the detailed features of the original image. The network model used in this embodiment outperforms traditional models in both structural similarity and peak signal-to-noise ratio, achieving excellent dehazing results.

[0060] In this embodiment, after dehazing the real-time transmission line image, the dehazed image can be fed into a convolutional neural network to obtain ice image features. The convolutional neural network used in this step can be a lightweight MobileNetV3 convolutional neural network. During feature extraction, a multi-sensory field of view module can be introduced to increase the model's mapping area for ice images, thereby enhancing its feature extraction capabilities.

[0061] Step 130: Input the pre-processed transmission line image into a multi-scale object detection network to obtain the ice thickness of the transmission line at the current moment.

[0062] In an embodiment of the present application, a multi-scale object detection network includes a single-shot multibox detector (SSD). Ice thickness can be classified into four levels, from severe to mild, namely, I, II, III, and IV. In this embodiment, an early warning is issued when ice thickness reaches Level II. Before the multi-scale object detection network is applied, it can be trained and then tested and verified using ice images perceived in actual scenarios in an edge intelligent device with limited computing resources.

[0063] Step 140: Identify the ice type of the pre-processed transmission line image to obtain the ice type of the transmission line at the current moment.

[0064] In an embodiment of the present application, the ice type can be identified by inputting the ice image features extracted in the above steps into the support vector machine. Specifically, this embodiment first uses a convolutional neural network to extract ice image features, and then uses a support vector machine to replace the softmax layer of the convolutional neural network to achieve the classification of ice and snow cover detection images through the support vector machine. Among them, there are four types of ice: rime, rime, mixed rime and snow. Compared with the convolutional neural network, the hybrid classification model of the convolutional neural network and the support vector machine adopted in this embodiment has better image classification effect, further improves the classification performance of ice and snow power grid detection images, and ensures the reliability and safety of detection.

[0065] Step 150: Acquire micrometeorological information collected in real time by sensor modules within the area of ​​the power transmission line. Specifically, the micrometeorological information includes at least one of ambient temperature, relative humidity, air pressure, wind speed, wind direction, rain, snow, or light intensity. The real-time collected micrometeorological information includes both the current micrometeorological information and the next micrometeorological information.

[0066] Step 160: Based on the ice thickness at the current moment, the ice type at the current moment, and the micro-meteorological information, predict the ice condition of the transmission line at the next moment through a memory network.

[0067] In an embodiment of the present application, at least one item of micro-meteorological information is used to analyze ice formation conditions in real time, and then the ice formation conditions, the ice thickness at the current moment, and the ice type at the current moment are used to predict the ice condition of the transmission line at the next moment through a memory network. The ice condition of the transmission line at the next moment includes the ice type and ice level. In this embodiment, in addition to micro-meteorological information, conductor and tower factors are also important bases for the memory network algorithm prediction. Therefore, this embodiment also obtains the thickness of the transmission line, the acceleration of the transmission line, and the inclination of the tower. Then, before the memory network is applied, the weight coefficient of the memory network is adjusted according to the thickness of the transmission line, the acceleration of the transmission line, and the inclination of the tower. The memory network includes an LSTM network based on a time series. The LSTM (Long Short-Term Memory) network is a long short-term memory network and a time recursive neural network. Compared with the ordinary RNN network, the LSTM network can perform better in longer sequences. Specifically, this embodiment uses past and present ambient temperature, relative humidity, air pressure, wind speed, wind direction, rain, snow, illumination, and future weather forecast information as input, and predicts the type and corresponding level of wire icing as output. Prior to application, the LSTM network is trained, and the model parameters are optimized during the training process. Specifically, different optimizers are compared to minimize error, and the most suitable optimizer is selected. Because the model of this embodiment can utilize continuously acquired real-world data to expand the dataset for model updates, the prediction results can be adjusted in real time, making the predicted values ​​closer to the actual values.

[0068] Step 170: Generate an ice damage warning signal according to the icing condition of the transmission line at the next moment.

[0069] In some application systems, the system can be based on the Atlas 200 multi-sensor module self-organizing network system and use the JETSON AGX XAVIER module to call various sensor data and neural network models. It can use binocular vision cameras to perform zoom sampling and identification of conductors, track and locate ice conditions, monitor conductors and insulators without blind spots, and realize ice cover and ice damage prediction based on edge lightweight computing.

[0070] Specifically, the hardware of the application system can adopt a multi-sensor self-organizing network system. Among them, the multi-sensor includes a micro-meteorological sensor module, a tilt angle sensor module, an acceleration sensor module and an image and sound sensor module. The micro-meteorological sensor module, the tilt angle sensor module and the acceleration sensor module serve as the basic modules, and the sound sensor module serves as the enhanced module. The basic module exchanges data with the enhanced module through 2.4G or 433M local area wireless mode, and the enhanced module exchanges data with the cloud server through NB-IoT (areas with base station signals) or Lora (areas without base station signals). Each module has its own signal acquisition, data processing, data transmission, energy capture and energy storage functions. The enhanced module is equipped with the NVIDIA JETSON AGXXAVIER module, which has functions such as image edge processing, wide area wireless networking, and cloud-edge collaboration.

[0071] Specifically, the system uses a binocular camera to perform zoom sampling of the conductors. The captured images are then transmitted wirelessly over a local area network to a processing terminal or the cloud for ice cover identification and tracking, enabling comprehensive monitoring of conductors and insulators. In this embodiment, due to the cold and high humidity, the binocular camera lenses are easily blurred by fog. Using an improved algorithm based on a recurrent generative adversarial network, images are defogged, restoring detailed image features. The defogged images are then used with a lightweight convolutional neural network, MobileNetV3, to identify ice thickness and issue an alert when ice reaches Level II.

[0072] In this embodiment, to detect icing trends and provide de-icing warnings, micrometeorological and displacement sensing modules collect multiple factors, including ambient temperature, relative humidity, air pressure, wind speed, wind direction, rain and snow, light intensity, inclination angle, and acceleration. A memory neural network algorithm is trained on historical data to predict icing conditions on transmission lines, thereby reconfirming the edge of ice damage. A genetic algorithm and a feedback neural network are combined to form a GA-BP neural network, which predicts conductor galloping. Furthermore, by combining sound signal spectrograms with convolutional neural networks to extract characteristic parameters of sound signals, a number of convolutional neural networks with different structures were constructed. Using the Adam optimization function, the system uses sound recognition to detect tripping sounds, impulse discharge sounds, spherical gap discharge sounds, power frequency discharge sounds, corona discharge sounds, and air noise, thereby monitoring tripping and insulator surface discharge conditions. Continuous binocular vision imaging of conductors and insulators enables real-time ice damage prediction. However, the massive amount of images creates a massive amount of data. The K-Means algorithm, combined with hierarchical clustering and the K-nearest neighbor algorithm, extracts key frames, removes redundant information, reduces data transmission pressure, and reduces the transmission of low-value images, thereby alleviating pressure on communication channels.

[0073] In this application system, the data collection wireless sensor network in the signal-free area adopts Lora networking, and transmits the data to the nearby ice damage early warning aircraft with signal through Lora; in the area with base station signal, NB-IoT is used to interact with the cloud server to upload the data of multiple ice damage early warning aircraft to the cloud in a unified manner.

[0074] An embodiment of the present invention provides an ice disaster early warning system based on an open neural network exchange algorithm, comprising:

[0075] The first acquisition module is used to acquire the transmission line image collected by the binocular camera in real time;

[0076] A preprocessing module, configured to preprocess the transmission line image;

[0077] a detection module, configured to input the preprocessed transmission line image into a multi-scale target detection network to obtain the ice thickness of the transmission line at a current moment;

[0078] an identification module, configured to identify the ice type of the transmission line after preprocessing to obtain the ice type of the transmission line at a current moment;

[0079] The second acquisition module is used to obtain micro-meteorological information collected in real time by the sensor module in the area to which the transmission line belongs;

[0080] A prediction module, configured to predict the icing condition of the transmission line at the next moment through a memory network based on the ice thickness at the current moment, the ice type at the current moment, and the micrometeorological information;

[0081] A generating module is used to generate an ice damage warning signal according to the icing condition of the transmission line at the next moment.

[0082] The contents of the method embodiments of the present invention are all applicable to the system embodiments. The functions specifically implemented by the system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above methods.

[0083] An embodiment of the present invention provides an ice disaster early warning system based on an open neural network exchange algorithm, comprising:

[0084] at least one memory for storing a program;

[0085] At least one processor for loading the program to execute Figure 1 The ice disaster early warning method based on the open neural network exchange algorithm is shown.

[0086] The contents of the method embodiments of the present invention are all applicable to the system embodiments. The functions specifically implemented by the system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above methods.

[0087] An embodiment of the present invention provides a storage medium in which a computer-executable program is stored. When the computer-executable program is executed by a processor, it is used to implement Figure 1 The ice disaster early warning method based on the open neural network exchange algorithm is shown.

[0088] The embodiment of the present invention further provides a computer program product or computer program, the computer program product or computer program including computer instructions, the computer instructions stored in a computer readable storage medium. The processor of the computer device can read the computer instructions from the computer readable storage medium, and the processor executes the computer instructions so that the computer device performs Figure 1 The ice disaster early warning method based on the open neural network exchange algorithm is shown.

[0089] While the embodiments of the present invention have been described in detail above with reference to the accompanying drawings, the present invention is not limited to the embodiments described above. Various modifications may be made within the scope of knowledge possessed by a person skilled in the art without departing from the spirit of the present invention. Furthermore, the embodiments of the present invention and the features thereof may be combined with one another unless there is a conflict.

Claims

1. An ice disaster early warning method based on an open neural network exchange algorithm, characterized in that: The following steps are involved: Obtaining transmission line images captured by binocular cameras in real time; Preprocessing the transmission line image includes: Defogging the transmission line image using an image defogging algorithm; The processing of the image defogging algorithm includes: Inputting the transmission line image into a preset cyclic generative adversarial network, and performing dehazing processing on the transmission line image through color loss and feature loss in the preset cyclic generative adversarial network; Inputting the dehazed transmission line image into a convolutional neural network, performing feature extraction through a multi-sensory field of view module, and obtaining ice-covered image features; Inputting the preprocessed transmission line image into a multi-scale target detection network to obtain the ice thickness of the transmission line at the current moment; Identifying the ice type of the pre-processed transmission line image to obtain the ice type of the transmission line at a current moment; Obtain micro-meteorological information collected in real time by sensor modules in the area where the transmission line belongs; According to the ice thickness at the current moment, the ice type at the current moment, and the micrometeorological information, predicting the ice condition of the transmission line at the next moment through a memory network; generating an ice damage warning signal according to the icing condition of the transmission line at the next moment; Among them, the icing condition of the transmission line is predicted through the micro-meteorological sensor module, displacement sensor module and memory neural network algorithm; the conductor dancing is predicted by combining genetic algorithm and feedback neural network; the tripping and insulator surface discharge conditions are monitored through sound signal spectrogram and convolutional neural network; and the defects of the conductor and the insulator are predicted through binocular vision.

2. The ice disaster early warning method based on the open neural network exchange algorithm according to claim 1 is characterized in that: The activation layer of the convolutional neural network includes a support vector machine; the ice type identification of the pre-processed transmission line image includes: The ice cover image features are input into the support vector machine to identify the ice cover type.

3. The ice disaster early warning method based on the open neural network exchange algorithm according to claim 1 is characterized in that: The step of obtaining micro-meteorological information collected in real time by the sensor module in the area to which the power transmission line belongs includes: Obtain at least one of the following: ambient temperature, relative humidity, air pressure, wind speed, wind direction, rain, snow, or illumination, which are collected in real time by a sensor module in the area to which the transmission line belongs.

4. The ice disaster early warning method based on the open neural network exchange algorithm according to claim 3 is characterized in that: The step of predicting the icing condition of the power transmission line at the next moment through a memory network based on the ice thickness at the current moment, the ice type at the current moment, and the micrometeorological information includes: analyzing ice formation conditions in real time based on the micro-meteorological information; According to the ice formation conditions, the ice thickness at the current moment, and the ice type at the current moment, the ice condition of the transmission line at the next moment is predicted through a memory network.

5. The ice disaster early warning method based on the open neural network exchange algorithm according to claim 4 is characterized in that: Before predicting the icing condition of the power transmission line at the next moment through the memory network, the method further includes: Obtain the thickness of the transmission line, the acceleration of the transmission line and the inclination of the tower; The weighting coefficient of the memory network is adjusted according to the thickness of the transmission line, the acceleration of the transmission line and the inclination of the tower.

6. An ice disaster early warning system based on an open neural network exchange algorithm, characterized in that: include: The first acquisition module is used to acquire the transmission line image collected by the binocular camera in real time; A preprocessing module, configured to preprocess the transmission line image, including: performing defogging processing on the transmission line image using an image defogging algorithm; The image dehazing algorithm comprises the following steps: inputting the power transmission line image into a preset cyclic generative adversarial network, dehazing the power transmission line image through color loss and feature loss within the preset cyclic generative adversarial network; inputting the dehazed power transmission line image into a convolutional neural network, extracting features through a multi-sensory field of view module, and obtaining ice image features; a detection module, configured to input the preprocessed transmission line image into a multi-scale target detection network to obtain the ice thickness of the transmission line at a current moment; an identification module, configured to identify the ice type of the transmission line after preprocessing to obtain the ice type of the transmission line at a current moment; The second acquisition module is used to obtain micro-meteorological information collected in real time by the sensor module in the area to which the transmission line belongs; A prediction module, configured to predict the icing condition of the transmission line at the next moment through a memory network based on the ice thickness at the current moment, the ice type at the current moment, and the micrometeorological information; a generating module, configured to generate an ice damage warning signal according to the icing condition of the transmission line at the next moment; Among them, the icing condition of the transmission line is predicted through the micro-meteorological sensor module, displacement sensor module and memory neural network algorithm; the conductor dancing is predicted by combining genetic algorithm and feedback neural network; the tripping and insulator surface discharge conditions are monitored through sound signal spectrogram and convolutional neural network; and the defects of the conductor and the insulator are predicted through binocular vision.

7. An ice disaster early warning system based on an open neural network exchange algorithm, characterized in that: include: at least one memory for storing a program; At least one processor is configured to load the program to execute the ice hazard early warning method based on the open neural network exchange algorithm according to any one of claims 1 to 5.

8. A storage medium, characterized in that: A computer-executable program is stored therein, and when the computer-executable program is executed by a processor, it is used to implement the ice disaster early warning method based on the open neural network exchange algorithm according to any one of claims 1 to 5.

Images