Intrusion early warning device and method for power transmission line

Abstract

The invention provides an intrusion early warning device and method for a power transmission line, and relates to the technical field of electric power. The intrusion early warning device is installed on a power transmission line tower, and a sound sensor collects sound data around the power transmission line in real time and identifies abnormal sound; the low-power-consumption night vision camera assembly adopts a mechanism combining sound driving with timing acquisition, the timing acquisition is to acquire images according to a preset time interval, the sound driving is to respond to recognized abnormal sound, a camera performs instant image acquisition, and after existence of intrusion hidden danger is confirmed through intrusion hidden danger detection of the images, the low-power-consumption night vision camera assembly is started. Dynamically adjusting the acquisition time interval, performing new timing acquisition and intrusion hidden danger detection, and adjusting to the original time interval until no intrusion hidden danger exists; the acousto-optic early warning device carries out early warning on the intrusion hidden danger according to the result of the intrusion hidden danger detection; the intrusion early warning device meets the requirement of all-weather work, and timely and effectively prevents potential intrusion hazards from threatening the power transmission line.

Description

Technical Field

[0001] The present invention belongs to the field of electric power technology, and in particular relates to an intrusion warning device and method for a power transmission line. Background Art

[0002] The channel conditions, environmental conditions, geological conditions, weather conditions, etc. along the transmission line have a great impact on the safe operation of the line. External force intrusion and damage have exceeded other factors and become the main reason for line tripping. Under the traditional operation and inspection mode, the information acquisition method is traditional and the source is single. The equipment status perception is still mainly based on power outage maintenance and offline testing, and the monitoring and early warning capabilities of external force intrusion hazards are not strong. With the continuous development of the power system, the frequency of engineering operations near transmission lines has increased significantly. It is very easy for engineering vehicles to touch the transmission lines due to improper operation, which may cause damage to the transmission lines and casualties, seriously affecting the normal operation of the power grid. The power grid company adds new transmission lines every year, while the number of operation and inspection personnel is relatively stable, and the average maintenance length per person increases.

[0003] With the development of communication technology and artificial intelligence, the external force intrusion detection of power transmission lines based on visual equipment monitoring has been promoted and applied on a large scale. Compared with the traditional manual inspection method, the inspection efficiency has been greatly improved; however, there are still problems such as untimely monitoring and inability to work normally due to the influence of the external environment, which cannot meet the needs of all-weather work. The specific problems are:

[0004] 1. Existing visualization equipment does not have anti-icing devices. The lens cannot obtain clear images in harsh environments such as fogging and icing. Even if there are devices such as wipers, they can only be controlled manually or turned on at a scheduled time, which cannot effectively ensure the normal operation of the visualization lens.

[0005] 2. Most of the existing visualization equipment works based on images, and most of the monitoring is in timed shooting mode with fixed monitoring intervals, which can easily miss major hidden dangers that occur within fixed intervals. In addition, when there is heavy fog or other bad weather conditions, it is impossible to clearly capture the image of the transmission line channel, resulting in a monitoring blind spot.

[0006] Therefore, although the efficiency of detection and warning of power transmission line intrusion warning devices has been greatly improved, there are still problems such as untimely monitoring and inability to work normally due to the influence of the external environment. It cannot meet the needs of all-weather work and poses a safety hazard. Summary of the invention

[0007] In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides an intrusion warning device and method for a power transmission line, which can meet the needs of working around the clock and timely and effectively prevent intrusion hazards from threatening the power transmission line.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0009] A first aspect of the present invention provides an intrusion warning device for a power transmission line.

[0010] An intrusion warning device for a power transmission line, installed on a power transmission line tower, comprises a low-power night vision camera assembly, a sound sensor and an acousto-optic warning device;

[0011] The sound sensor collects sound data around the power transmission line in real time and identifies abnormal sounds;

[0012] The low-power night vision camera assembly adopts a sound drive combined with a timed acquisition mechanism. The timed acquisition is to acquire images according to a preset time interval. The sound drive is to respond to the recognized abnormal sound. The camera performs real-time image acquisition. After confirming the existence of an intrusion hazard through the image intrusion hazard detection, the acquisition time interval is dynamically adjusted to perform new timed acquisition and intrusion hazard detection until there is no intrusion hazard, and then the time interval is adjusted to the original time interval.

[0013] The sound and light warning device warns of intrusion hazards based on the results of intrusion hazard detection.

[0014] Furthermore, the low-power night vision camera assembly has an adaptive anti-icing function, specifically:

[0015] According to the ambient temperature, humidity and imaging status of the camera, the lens icing status can be predicted in real time, and the opening and closing of the heating and wiper devices can be controlled to ensure clear imaging of the camera.

[0016] Furthermore, the low-power night vision camera assembly is specifically:

[0017] The lens uses a special sapphire glass lens with excellent thermal conductivity, adopts a surround resistance wire heating method, and is equipped with a shield and anti-freeze wiper.

[0018] Furthermore, the predicted lens icing state is based on the constructed icing time series data, and the icing state at the time to be predicted is inferred through the Bi-LSTM bidirectional time series anti-icing prediction model;

[0019] The ice coverage time series data is composed of a group of ice coverage data at consecutive moments, and the ice coverage data includes temperature, humidity, image brightness, and image clarity.

[0020] Furthermore, the abnormal sound recognition is to use a deep learning algorithm to train an abnormal sound recognition model, perform feature extraction and sound classification on the sound data collected in real time, so as to obtain an abnormal sound recognition result.

[0021] Furthermore, the intrusion risk detection is to detect hidden danger targets on images collected in real time through deep learning intelligent detection technology.

[0022] Furthermore, the intrusion warning device has a built-in power supply device, including a solar panel and a rechargeable battery.

[0023] A second aspect of the present invention provides a transmission line intrusion warning method.

[0024] A transmission line intrusion warning method, comprising:

[0025] The sound sensor collects sound data around the power transmission line in real time and identifies abnormal sounds;

[0026] Use low-power night vision camera components to regularly collect images and detect intrusion risks;

[0027] According to the results of the intrusion risk detection, the sound and light warning device will warn of the intrusion risk;

[0028] Among them, the low-power night vision camera component adopts sound drive combined with a timed acquisition mechanism. The timed acquisition is to acquire images according to a preset time interval. The sound drive is in response to the identified abnormal sound. The camera performs real-time image acquisition. After confirming the existence of intrusion risks through intrusion risk detection of the image, the acquisition time interval is dynamically adjusted, and new timed acquisition and intrusion risk detection are performed until there is no intrusion risk, and then it is adjusted to the original time interval.

[0029] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the intrusion warning device for a power transmission line as described in the first aspect of the present invention.

[0030] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the intrusion warning device for a power transmission line as described in the first aspect of the present invention are implemented.

[0031] One or more of the above technical solutions have the following beneficial effects:

[0032] The present invention proposes an intrusion warning device for power transmission lines, which adopts sound drive combined with a timing acquisition mechanism and controlled acquisition of visualized images to ensure timely monitoring and alarm linkage, solves the problems of untimely monitoring and being affected by the external environment, especially the inability to recognize images normally in foggy weather, etc., can meet the needs of all-weather work, and timely and effectively prevent hidden dangers from threatening power transmission lines.

[0033] The present invention collects transmission channel environmental data based on hardware devices such as lens heating devices, anti-freezing wipers, temperature and humidity sensors, etc., establishes an anti-icing prediction model, realizes a dynamic anti-icing mechanism for lenses in harsh environments, accurately predicts the icing state of the lens, and starts the heating and wiping devices, which can timely and effectively ensure clear imaging of the equipment.

[0034] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0036] Figure 1 This is a structural diagram of the intrusion warning device of the first embodiment.

[0037] Figure 2 This is a flow chart of the adaptive anti-icing mechanism of the first embodiment.

[0038] Figure 3 This is a structural diagram of the abnormal sound recognition model of the first embodiment.

[0039] Figure 4 This is a structural diagram of the intrusion risk detection model of the first embodiment.

[0040] Figure 5 This is a flow chart of the linkage mechanism of the first embodiment. DETAILED DESCRIPTION

[0041] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0042] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0043] Embodiment 1

[0044] In one embodiment of the present disclosure, an intrusion warning device with a night vision function, a lens heating device, an anti-freezing wiper and an anti-icing structure is provided. The temperature and humidity sensors are used to collect temperature and humidity data and combine them with the imaging status to realize an adaptive dynamic lens anti-icing mechanism, automatically control the activation of the wiper and heating functions, and timely and accurately prevent the lens from being disturbed by the external environment; and a sound sensor is used to collect sound information around the line in real time, and a sound-image linkage mechanism is designed. A lightweight abnormal sound recognition model is used to realize real-time sound analysis, and image capture and intrusion hazard detection are linked in time to form an effective warning mechanism, improve the all-weather intrusion warning capability, and timely and effectively prevent intrusion hazards from threatening the transmission line.

[0045] The following is a detailed explanation of the hardware system, adaptive anti-icing mechanism, abnormal sound recognition, intrusion risk detection, and linkage mechanism.

[0046] 1. Hardware system

[0047] like Figure 1 As shown, the intrusion warning device is composed of a temperature and humidity sensor, a low-power night vision camera, a sound sensor, an acoustic and light warning device, a heating device, an anti-freezing wiper, a solar panel, a rechargeable battery, etc., and is controlled through an intelligent prediction and control module.

[0048] 1) The low-power night vision camera component is used, which has the ability to automatically adjust the image and adjust the exposure and white balance parameters according to the ambient light conditions, so as to capture clear images at different times of the day and night. The lens uses an industrial-grade CMOS sensor with a target size of 1 / 1.8 inches and an F2.0 aperture to increase the amount of light exposure to ensure the quality of nighttime imaging.

[0049] 2) The lens heating component is designed to heat the lens at low temperatures to avoid fogging and icing, meeting the requirements for collecting clear images around the clock. This device uses a special sapphire glass lens with excellent thermal conductivity and a surrounding resistance wire heating method, which ensures heating efficiency while not affecting light-transmitting imaging; it has multiple heating modes and supports remote start and automatic start.

[0050] 3) The lens is equipped with a shield to prevent rain from falling; the lens uses a high thermal conductivity mirror, and the mirror is sprayed with a high molecular hydrophobic layer using a vacuum evaporation process, which greatly reduces the adhesion of ice on the lens surface and effectively removes ice.

[0051] 4) The lens is equipped with an anti-freezing wiper and used with the lens heating function. After the ice layer in contact with the mirror surface melts, the adhesion of the ice will be significantly reduced. Combined with the wiper, the ice layer on the lens surface can be easily removed, effectively ensuring that the lens can capture clear images in low temperature environments.

[0052] 5) Set up a sound sensor collection module to collect sounds around the device in real time.

[0053] 6) The intelligent prediction and control module has powerful computing power, can perform front-end intelligent analysis, identify abnormal sounds and image potential danger targets, and can perform anti-icing prediction in real time, realize data analysis and dynamic adaptive control functions, and realize routine intrusion potential danger inspections and line inspection tasks after icing. The processing results are transmitted to the control center via 4G / 5G.

[0054] 7) The rechargeable battery is charged and the device is powered by solar panels, which has over-current, over-voltage and lightning protection functions.

[0055] 8) The temperature and humidity sensor realizes the function of collecting temperature and humidity in the environment, provides environmental data in a timely manner for accurate prediction, and prevents the lens from fogging and icing.

[0056] 2. Adaptive anti-icing mechanism

[0057] The adaptive anti-icing mechanism based on lens heating device, anti-freezing wiper, temperature and humidity sensor collects environmental temperature and humidity, image information, records the collection time, and uses this information to establish a Bi-LSTM bidirectional time series anti-icing prediction model to infer the pre-icing state of the lens, start the heating and wiper device, and ensure clear imaging of the equipment in a timely and effective manner, thus realizing a dynamic anti-icing mechanism for the lens in harsh environments.

[0058] Specifically, a Bi-LSTM bidirectional time series anti-icing prediction model is constructed based on the sample data collected by sampling, and a prediction is made based on the real-time sampling data to determine whether the anti-icing module is currently turned on. When the prediction detects that the surrounding temperature, humidity, imaging status, etc. are about to freeze or affect the lens, the anti-icing module is started, and the resistance wire is automatically powered on and heated, and continues to heat to prevent ice from condensing on the lens surface to achieve the best anti-icing effect; during the process, the temperature, humidity and imaging status are monitored in real time. When the clear imaging requirements are met, the anti-icing module stops working to ensure energy saving in low temperatures, such as Figure 2 As shown, the main steps are:

[0059] 1) Collect raw sampling data

[0060] Three months of data of the tested transmission line were selected as the data set for test verification. The data information included: time, temperature, humidity, image brightness, image clarity, and lens icing status.

[0061] The image brightness adopts the mean brightness of all pixels, and the image clarity can be calculated based on the Laplacian gradient function. The Laplacian operator is defined as follows:

[0062]

[0063] Definition of clarity:

[0064] D(f)=∑ y ∑ x |G(x,y)|(G(x,y)>T)

[0065] Among them, G(x, y) is the convolution of the Laplacian operator at the pixel point (x, y).

[0066] 2) Data preprocessing

[0067] Convert the original data into a data format suitable for processing time series prediction problems, that is, time series data, specifically:

[0068] Feature normalization processing, data format: [samples, timesteps, features], where samples: the number of samples, here is 35038; timesteps: the time step, is 1; features: attribute feature dimension, here are the six attributes of time, temperature, humidity, image brightness, image clarity, and lens icing status.

[0069] After removing outliers, the data is divided into training set and test set.

[0070] 3) Establish a Bi-LSTM bidirectional time series anti-icing prediction model

[0071] Bi-LSTM learns the bidirectional long-term dependencies between time series data based on the entire time series, which can extract more representative time series features and help improve prediction accuracy.

[0072] Bi-LSTM adds an independent LSTM network layer with negative time input to the LSTM network structure with positive time input. The timing information is input into the first layer of LSTM at positive time to obtain the sequence information h at the current moment. t R , input the negative time into the second layer LSTM to obtain the reverse information h of the current moment t L The two output information are combined to form the hidden output h at the current moment t ; Therefore, the output at the current moment includes both past and future time series information, which is context-related. The Bi-LSTM structure hides the output, which can be expressed as:

[0073]

[0074] Among them, the current input x t and the previous hidden output h t-1, with R for positive time sequence, with L for negative time sequence (reverse sequence). W1-6 are weight coefficients, h t R Represents the order information of the current moment, h t L represents the reverse order information at the current moment, and ht represents the output information of the current moment that combines the forward order and reverse order.

[0075] The Bi-LSTM model was built. The loss function used the mean square error (MSE). The optimization algorithm used was Adam. The model used 100 epochs and the size of each batch was 36. The formula for the mean square error was:

[0076]

[0077] Where N is the total number of samples, is the predicted value, r i is the true value.

[0078] 4) Analysis and prediction

[0079] As needed, multiple data spans can be set for prediction; all feature data at times T-3, T-2, and T-1 can be used to predict the ice cover status value at time T.

[0080] If the predicted value is a pre-icing state, the anti-icing module is started and the judgment is continued during the process, otherwise it ends.

[0081] 3. Abnormal sound recognition

[0082] Collect real-time sound data, and intercept abnormal sound fragments from the ambient sound of the power transmission line in real time through index parameters such as threshold value and signal-to-noise ratio. Use deep learning algorithm to train the abnormal sound recognition model, automatically extract features and classify the intercepted abnormal sound fragments, and judge different types of sounds such as various types of construction machinery, explosions, thunder, birds, etc. The specific steps are as follows:

[0083] (1) Record and collect sound samples in real time.

[0084] (2) Capture sound clips.

[0085] (3) After preprocessing, sample labeling is performed to mark the sound type of each sound clip.

[0086] (4) The audio Mel-spectrogram features of the sound clip are obtained and used as the input of the abnormal sound recognition model. The sound type of the sound clip is used as the output to construct a sound sample for model training.

[0087] (5) After tuning, an abnormal sound recognition model is generated.

[0088] (6) Filter the sound data collected in real time by using sound energy and signal-to-noise ratio thresholds.

[0089] (7) Start abnormal sound recognition and output sound target.

[0090] This paper provides an implementation method of an abnormal sound recognition model based on a lightweight backbone network, the efficient convolutional neural network EfficientNet. The model structure is as follows: Figure 3 As shown:

[0091] First, the 10-second waveform (i.e., the audio Mel spectrum features of the sound clip) is converted into a 1056×128 logarithmic Mel filter bank (fbank) feature vector and input into the efficient convolutional neural network EfficientNet. The output of the second layer of the network is a 33×4×1408 tensor.

[0092] Then, frequency average pooling is applied to generate a new representation, which is fed into a 4-head attention pooling module; in each head, the CNN output is converted to a corresponding tensor by a set of 1×1 convolutional layers with parallel attention and classification branches; the outputs of each branch element are multiplied and temporal average pooling (implemented by summation) is applied.

[0093] Finally, the weighted outputs of each attention head, scaled by learnable weights, are summed and produce the final predictions for all categories.

[0094] The loss function of the abnormal sound recognition model uses multi-label classification cross entropy to calculate the loss, and the training strategy adopts a 5-fold cross validation method. In order to maximize the role of real collected data, when training the model, the network pre-training strategy in transfer learning is combined, and the pre-training model based on the large-scale classification data AudioSet is optimized. The specific optimization method is:

[0095] The first few layers of the pre-trained model are frozen to increase the model's prior knowledge, and then the unfrozen network layers of the model are iteratively optimized using real-world scenario data. Finally, the model with the best performance is obtained by trying different freezing strategies. The model can achieve multi-classification of sounds such as construction work and bird calls in power transmission channel scenarios.

[0096] 4. Intrusion risk detection

[0097] Design a front-end intelligent and lightweight intrusion hazard detection algorithm, collect sample image data of transmission channel hazard, and use deep learning intelligent detection technology to realize the front-end intelligent detection of channel hazards such as wildfires, smoke, foreign objects in wires, and construction machinery (tower cranes, cranes, excavators, dump trucks, bulldozers, and cement pump trucks). The specific steps are as follows:

[0098] 1) Collect transmission line image samples.

[0099] 2) Mark sample hidden danger targets and indicate the type and location of hidden dangers.

[0100] 3) Design a lightweight intrusion risk detection model.

[0101] The intrusion risk detection model is as follows: Figure 4 As shown, an efficient and computationally light neural network (lightweight feature extraction network) is used to extract key features from input data; the feature extraction network mainly includes convolutional layers, pooling layers, feature connection layers, etc.; the convolutional layer is used to extract local features of the image, the pooling layer is used to reduce the size and computational complexity of the feature map, and the feature connection layer may be used to combine features of different layers.

[0102] In order to ensure the effectiveness of feature expression, a feature cascade method is adopted to fully reuse the feature map output by each previous layer before the input of each layer, that is, the input of each layer contains the output feature maps of all previous layers; this design not only reduces the loss of information, but also promotes the flow and reuse of features between different levels, effectively ensuring the feature expression ability of the network.

[0103] The feature cascade network adopts a two-way dense layer design to obtain fields of view of different scales; one way uses a smaller 3×3 convolution kernel, which can better capture small-scale targets because it can focus on more detailed features; the other way uses two consecutive 3×3 convolution kernels (equivalent to a 5×5 convolution kernel but with fewer parameters and more efficient calculation). This design helps to learn the visual features of large-scale targets.

[0104] To ensure the effectiveness of feature expression, the structure performs three-layer downsampling feature extraction on the input image; downsampling is usually achieved by using a pooling layer or convolution layer with a stride greater than 1, which helps reduce the size of the feature map while retaining important feature information.

[0105] A lightweight Resblock residual block is added before the detection layer of each detection branch; Resblock is a basic module in the residual network (ResNet). It solves the gradient vanishing and gradient exploding problems in deep neural networks by introducing skip connections, ensuring that the feature expression ability does not weaken with the increase of network depth, thereby ensuring the feature expression ability.

[0106] Through iterative training (i.e., repeatedly adjusting model parameters to minimize the loss function), the model can eventually output the target type (Class) and its location box (box); the location box is usually a four-tuple that represents the position and size of the target in the image.

[0107] In summary, this intrusion risk detection model describes a lightweight feature extraction network customized for the transmission line intrusion risk detection task, which combines feature cascading, lightweight residual blocks, and two-way dense sampling methods to improve feature expression capabilities and detection performance while maintaining computational efficiency.

[0108] 4) Training with hidden danger image samples.

[0109] 5) Optimize and generate the hidden danger target identification model.

[0110] V. Linkage Mechanism

[0111] A timely and effective alarm linkage strategy based on a lightweight abnormal sound recognition model and an intrusion risk detection model uses sound drive combined with a timed acquisition mechanism and controlled acquisition of visual images to ensure timely monitoring and alarm linkage; Figure 5 As shown in the figure, an abnormal sound recognition model is designed to intelligently analyze the sound data collected by the sound sensor in real time. After identifying abnormal sounds such as construction machinery, the camera is started to capture images in time, and the intrusion hidden danger detection model is used to timely identify hidden danger targets in the image. Subsequently, the monitoring frequency is dynamically adjusted according to whether there are hidden danger targets in the image.

[0112] The sound is linked to the camera to take pictures, realizing the full perception of the sound and image on the scene. A timely and effective alarm linkage strategy based on a lightweight abnormal sound recognition model and an intrusion hidden danger detection model is formulated to solve the problems of untimely monitoring and filming, being affected by the external environment, especially the inability to recognize images normally in foggy weather.

[0113] Embodiment 2

[0114] In one embodiment of the present disclosure, a method for early warning of intrusion of a power transmission line is provided, comprising:

[0115] Step S1: collecting sound data around the power transmission line in real time through a sound sensor and identifying abnormal sounds;

[0116] Step S2: using a low-power night vision camera component to regularly collect images and perform intrusion risk detection;

[0117] Step S3: Based on the result of the intrusion risk detection, the sound and light warning device warns of the intrusion risk;

[0118] Among them, the low-power night vision camera component adopts sound drive combined with a timed acquisition mechanism. The timed acquisition is to acquire images according to a preset time interval. The sound drive is in response to the identified abnormal sound. The camera performs real-time image acquisition. After confirming the existence of intrusion risks through intrusion risk detection of the image, the acquisition time interval is dynamically adjusted, and new timed acquisition and intrusion risk detection are performed until there is no intrusion risk, and then it is adjusted to the original time interval.

[0119] Embodiment 3

[0120] The purpose of this embodiment is to provide a computer-readable storage medium.

[0121] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of an intrusion warning device for a power transmission line as described in the first embodiment of the present disclosure.

[0122] Embodiment 4

[0123] The purpose of this embodiment is to provide an electronic device.

[0124] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the intrusion warning device for a power transmission line as described in the first embodiment of the present disclosure are implemented.

[0125] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A transmission line intrusion warning device, characterized in that: Installed on transmission line towers, it includes low-power night vision camera components, sound sensors, and sound and light warning devices; The sound sensor collects sound data around the power transmission line in real time and identifies abnormal sounds; The low-power night vision camera assembly adopts a sound drive combined with a timed acquisition mechanism. The timed acquisition is to acquire images according to a preset time interval. The sound drive is to respond to the recognized abnormal sound. The camera performs real-time image acquisition. After confirming the existence of an intrusion hazard through the image intrusion hazard detection, the acquisition time interval is dynamically adjusted to perform new timed acquisition and intrusion hazard detection until there is no intrusion hazard, and then the time interval is adjusted to the original time interval. The sound and light warning device warns of intrusion hazards based on the results of intrusion hazard detection.

2. The intrusion warning device for a power transmission line according to claim 1, characterized in that: The low-power night vision camera assembly has an adaptive anti-icing function, specifically: According to the ambient temperature, humidity and imaging status of the camera, the lens icing status can be predicted in real time, and the opening and closing of the heating and wiper devices can be controlled to ensure clear imaging of the camera.

3. The intrusion warning device for a power transmission line according to claim 2, characterized in that: The low-power night vision camera assembly is specifically: The lens uses a special sapphire glass lens with excellent thermal conductivity, adopts a surround resistance wire heating method, and is equipped with a shield and anti-freeze wiper.

4. The intrusion warning device for a power transmission line according to claim 2, characterized in that: The predicted lens icing state is based on the constructed icing time series data, and the icing state at the predicted time is inferred through the Bi-LSTM bidirectional time series anti-icing prediction model; The ice coverage time series data is composed of a group of ice coverage data at consecutive moments, and the ice coverage data includes temperature, humidity, image brightness, and image clarity.

5. The intrusion warning device for a power transmission line according to claim 1, characterized in that: The abnormal sound recognition is to use a deep learning algorithm to train an abnormal sound recognition model, perform feature extraction and sound classification on the sound data collected in real time, so as to obtain an abnormal sound recognition result.

6. The intrusion warning device for a power transmission line according to claim 1, characterized in that: The intrusion hidden danger detection is to detect hidden danger targets on images collected in real time through deep learning intelligent detection technology.

7. The intrusion warning device for a power transmission line according to claim 1, characterized in that: It also includes that the intrusion warning device has a built-in power supply device, including a solar panel and a rechargeable battery.

8. A method for early warning of intrusion of a power transmission line, characterized in that: include: The sound sensor collects sound data around the power transmission line in real time and identifies abnormal sounds; Use low-power night vision camera components to regularly collect images and detect intrusion risks; According to the results of the intrusion risk detection, the sound and light warning device will warn of the intrusion risk; Among them, the low-power night vision camera component adopts sound drive combined with a timed acquisition mechanism. The timed acquisition is to acquire images according to a preset time interval. The sound drive is in response to the identified abnormal sound. The camera performs real-time image acquisition. After confirming the existence of intrusion risks through intrusion risk detection of the image, the acquisition time interval is dynamically adjusted, and new timed acquisition and intrusion risk detection are performed until there is no intrusion risk, and then it is adjusted to the original time interval.

9. An electronic device, comprising: a memory for non-transitory storage of computer readable instructions; a processor for executing the computer-readable instructions; Wherein, when the computer-readable instructions are executed by the processor, the method of claim 8 is performed.

10. A storage medium, characterized in that: The computer readable instructions are non-transitorily stored, wherein when the computer readable instructions are executed by a computer, the method of claim 8 is performed.

Images