Active bird damage prevention device and method for power transmission line based on artificial intelligence and simulated eagle

Abstract

The invention discloses a power transmission line active bird damage prevention device and method based on artificial intelligence and a simulation eagle, and belongs to the technical field of high-voltage power transmission line safety protection. The lifting platform driven by the pull wire and the lifting motor is innovatively designed, so that the simulation eagle and the ultrasonic transducer can vertically move, the whole space from the ground to the tower top is dynamically covered, and the problem that a fixed device has a protection dead angle is thoroughly solved; a built-in bird image database and an ultrasonic frequency mapping table are adopted, the system can'recognize 'what birds are, the most effective repelling frequency is automatically selected, and the limitation that a single frequency is easy to tolerate is broken through; the three functional modules of sensing, decision making and execution are integrated into a whole. The AI recognition unit, the lifting control unit and the ultrasonic generation unit are cooperatively scheduled by the main controller, so that full-automatic quick response from target discovery to precise expelling is realized; by adopting the design that the lifting motor is matched with the balancing weight, the lifting energy consumption is greatly reduced; solar power supply is suitable for long-term maintenance-free operation.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of high-voltage transmission line safety protection, and particularly relates to a transmission line active bird damage prevention device and method based on artificial intelligence and simulation falcons. BACKGROUND

[0002] Bird damage prevention of a transmission line is a key challenge for safe operation and maintenance of a power grid, especially in dense areas of bird migration channels. The annual equipment maintenance cost caused by bird damage is 64,000 yuan, and 12 hours of power outage is caused by each failure on average. According to the transmission power of 200,000 kilowatts of a 220kV line, the loss of a single accident is more than 160,000 yuan. The existing bird damage prevention methods have the following defects: although the traditional bird prevention thorn has basic physical barrier ability (such as needle length classification of 650 / 550 / 450mm and diameter of 2.0mm), its rigid structure is easy to deform plastically under strong wind conditions, and the static layout cannot adapt to birds with significant wing span differences such as white herons; the sound and light bird repelling device causes rapid formation of bird tolerance (usually invalid within 3-7 days), resulting in a sharp decline in protection efficiency, and the efficiency of artificial inspection mode is only 5km / person·day, which is difficult to cover the total transmission corridor of more than 1,000 kilometers. Although there are existing patent technologies that propose modular bird repeller designs, the core limitation is the lack of prediction ability of bird behavior rules and dynamic optimization mechanism of protection strategies - neither can a risk probability model be constructed based on historical failure data, nor can precise protection be triggered in response to key behavior characteristics such as bird residence time and activity frequency. Therefore, it is urgent to develop an active defense system that integrates environmental perception, intelligent decision-making and adaptive execution to fundamentally break through the technical bottlenecks of existing technologies. SUMMARY

[0003] The technical problem to be solved by the application is to provide a transmission line active bird damage prevention device and method based on artificial intelligence and simulation falcons to solve the technical problem that existing bird damage prevention devices cannot dynamically, accurately and intelligently repel birds.

[0004] The active bird prevention device for power transmission line based on artificial intelligence and simulation hawk comprises a main controller, a fixed base, a lifting motor, a vertical rod, a lifting platform, an ultrasonic transducer, a simulation hawk, a solar panel and an AI recognition camera, the fixed base is fixedly installed on a power transmission line tower, and the fixed base is provided with the main controller, the solar panel and the AI recognition camera; the vertical rod is fixedly installed on the power transmission line tower, and the vertical rod is provided with a lead screw; the lifting motor is connected with the lead screw through a coupling, the lifting motor moves up and down along the lead screw, and the lifting motor is fixedly connected with the lifting platform; the simulation hawk and the ultrasonic transducer are fixedly installed on the lifting platform; the solar panel is connected with a storage battery; the main controller is connected with the storage battery, the lifting motor, the ultrasonic transducer and the AI recognition camera respectively, the main controller is used for AI recognition of birds, control of lifting movement of the lifting motor, selection of ultrasonic frequency of the ultrasonic transducer and data transmission with a background monitoring system of a power system.

[0005] The upper end of the vertical rod is fixedly provided with a fixed pulley; the fixed pulley is sleeved with a traction rope; one end of the traction rope is fixedly connected with the lifting platform, and the other end of the traction rope is fixedly connected with a counterweight, so as to reduce energy consumption of the lifting motor and enhance stability of the lifting platform.

[0006] The lifting platform is further provided with a position sensor, which is used for detecting real-time height of the lifting platform and is in communication connection with the main controller.

[0007] The active bird prevention method for power transmission line based on artificial intelligence and simulation hawk utilizes the active bird prevention device for power transmission line based on artificial intelligence and simulation hawk, and comprises the following steps which are sequentially performed: Step one: continuous monitoring The AI recognition camera performs real-time video monitoring on the surrounding area of the tower; Step two: target recognition and positioning (a) If the AI recognition camera captures birds, the captured bird image is transmitted to the main controller; (b) The main controller analyzes the bird image through a set AI algorithm, identifies the type of the bird, and estimates the relative position of the bird and the tower and the height of the bird from the ground; Step three: intelligent decision and precise driving away (a) According to the estimated height of the bird, a lifting control instruction is generated and transmitted to the lifting motor, the lifting motor is lifted, and the simulation hawk is moved to the same horizontal height as the bird; (b) At the same time, the main controller queries the set most sensitive ultrasonic frequency of the bird from the built-in database of the main controller according to the identified type of the bird, and drives the bird away through the corresponding ultrasonic transducer; Step four: feedback and standby (a) The AI recognition camera continuously monitors the driving effect, and if the AI recognition camera cannot detect birds, the main controller controls the lifting motor to return to the standby position, preparing for the next task; (b) The main controller forms and stores operation logs and bird driving records, which are uploaded to the background monitoring system through the wireless communication module in the main controller.

[0008] The specific method for identifying the species of birds, estimating the relative position of birds and the tower, and estimating the height from the ground in step two is as follows: a. Randomly take a frame of image from the real-time bird image obtained from the AI recognition camera for scaling and normalization preprocessing; b. Use the trained YOLOv5 model for target detection to locate all birds in the image; Wherein, each bird in the image is an element, and the information of the element is represented by [x, y, width, height, confidence, class_id], wherein x is the horizontal coordinate of the center point of the bounding box, y is the vertical coordinate of the center point of the bounding box, width is the width of the bounding box, height is the height of the bounding box, confidence is the confidence that the box contains the target and the class is correct, and class_id is the class number of the target, which is an integer starting from 0, corresponding to the serial number assigned to each class during training; c. Crop the bird image according to the detection box coordinates to obtain the cropped bird candidate region ROI; d. Input the cropped bird candidate region ROI into the trained EfficientNet classification network for fine identification, and output the identified bird species as a probability distribution of each specified category; e. Get the highest probability and its corresponding bird species ID; f. Decision: If the confidence is greater than the set threshold, the identification is valid, the bird species ID is mapped to the bird species name, the bird height from the ground is estimated according to the detection box position, the driving function is called, and the most sensitive ultrasonic frequency corresponding to the bird species name and the driving motor revolutions corresponding to the bird height from the ground are obtained as driving instructions and sent to the ultrasonic transducer and the driving motor respectively; If the confidence is less than or equal to the set threshold, the identification is invalid, and the image is saved for subsequent model optimization.

[0009] The trained YOLOv5 model is trained by using an image dataset containing main harmful birds and using labeled bounding box information to make it learn to locate birds from complex backgrounds; wherein, the birds in each image of the image dataset are labeled by experts, including bounding box and species label.

[0010] The trained EfficientNet classification network is to store the image dataset in folders according to the bird species category, and use the transfer learning technology to retrain the EfficientNet model which has been pre-trained on a large image dataset, so as to focus on distinguishing different species of birds.

[0011] Through the above design scheme, the application can bring the following beneficial effects: 1. Lifting three-dimensional protection mechanism: the lifting platform driven by the traction line and the lifting motor is innovatively designed, so that the simulation falcon and the ultrasonic transducer can move vertically and dynamically cover the entire space from the ground to the tower top, completely solving the problem of dead angle of fixed devices.

[0012] 2. Frequency adaptive strategy based on AI recognition: the built-in bird image database and ultrasonic frequency mapping table enable the system to "recognize" what bird it is and automatically select the most effective repelling frequency, breaking through the limitation of single frequency tolerance.

[0013] 3. Modular integration and collaborative operation: the perception, decision-making and execution modules are integrated into one. The AI recognition unit, lifting control unit and ultrasonic generation unit are collaboratively scheduled by the main controller to realize full-automatic and rapid response from target discovery to accurate repelling.

[0014] 4. Low power consumption and high reliability design: the lifting mechanism adopts a low-power lifting motor combined with a counterweight block design, which greatly reduces the lifting energy consumption. The whole system is powered by solar energy, which is suitable for long-term maintenance-free operation in outdoor tower environment. BRIEF DESCRIPTION OF DRAWINGS

[0015] The application will be further described below in combination with the drawings and specific embodiments: Figure 1 The structure block diagram of the device in the active bird damage prevention device and method for power transmission lines based on artificial intelligence and simulation falcon of the application; Figure 2 The method flowchart in the active bird damage prevention device and method for power transmission lines based on artificial intelligence and simulation falcon of the application.

[0016] In the figure, 1 is a fixed base, 2 is a fixed pulley, 3 is a lifting motor, 4 is a vertical rod, 5 is a lifting platform, 6 is a counterweight block, 7 is an ultrasonic transducer, 8 is a simulation falcon, 9 is a solar panel, 10 is a traction rope, 11 is an AI recognition camera. DETAILED DESCRIPTION

[0017] I. Overall structure As Figure 1The active bird prevention device for power transmission lines based on artificial intelligence and simulation falcons is shown, including a fixed base 1, a lifting assembly, an ultrasonic emission assembly, and a control and energy assembly.

[0018] The fixed base 1 is waterproof and dustproof, and is firmly installed at a suitable position of the power transmission tower by a hoop or a bolt.

[0019] The lifting assembly includes a fixed pulley 2, a lifting motor 3, a vertical rod 4, a lifting platform 5, a counterweight 6, an ultrasonic transducer 7, and a simulation falcon 8. The vertical rod 4 is fixedly installed on the power transmission tower, and a lead screw is installed on the vertical rod 4. The lifting motor 3 is connected with the lead screw through a shaft coupling to realize the lifting and lowering of the lifting motor 3. One end of the lifting motor 3 is fixedly connected with the lifting platform 5, and the ultrasonic transducer 7 and the simulation falcon 8 are fixedly installed on the lifting platform 5. The upper part of the lifting platform 5 is also connected with a traction rope 10. The traction rope 10 passes through the fixed pulley 2 arranged at the top end of the vertical rod 4, and a counterweight 6 is tied at the other end. The counterweight 6 makes the lifting of the lifting motor 3 more labor-saving, thereby reducing energy consumption. Moreover, with the traction rope 10, according to the principle of good stability of a triangle, the stability of the lifting platform 5 is better. A position sensor is also installed on the lifting platform 5 for detecting the real-time height of the simulation falcon, and the position sensor is in communication connection with the main controller.

[0020] The ultrasonic emission assembly includes one or more ultrasonic transducers 7 and their driving circuits. The emission direction can be horizontally outward to ensure effective propagation of sound waves.

[0021] The control and energy assembly includes a main controller, an AI recognition camera 11, a position sensor, a solar panel 9, and a storage battery. The AI recognition camera 11 is installed on the fixed base 1 or at other best viewing positions of the power transmission tower for monitoring birds. The main controller is embedded in the fixed base 1 and is responsible for running the AI recognition algorithm, controlling the lifting of the lifting motor 3, and selecting the ultrasonic frequency. The position sensor can adopt a Hall sensor or an encoder for detecting the real-time height of the lifting platform 5. The solar panel 9 and the storage battery provide energy for the entire system.

[0022] II. Coordinated operation and overall workflow The working process of the device is a complete "perception-decision-execution" closed loop, and the specific steps are as follows: 1. Continuous monitoring: The AI recognition camera 11 performs uninterrupted video monitoring on the surrounding area of the power transmission tower.

[0023] 2. Target identification and positioning: (a) Once the AI recognition camera 11 captures the image of the bird, it is immediately transmitted to the main controller.

[0024] (b) The main controller uses AI algorithms (such as AI algorithms based on models like YOLOv5) to analyze images in real time, identify the species of birds, and estimate their relative position and height above the ground relative to the transmission tower.

[0025] 3. Intelligent decision-making: (a) The main controller queries the built-in database for the most sensitive ultrasonic frequency for the identified bird species (e.g., 22 kHz for crows and 18 kHz for egrets).

[0026] (b) At the same time, based on the estimated height of the bird, a lifting control command is generated, with the goal of moving the lifting platform 5 to the same level as the bird.

[0027] 4. Precise execution: (a) Lifting action: The main controller sends a command to the lifting motor 3, which lifts the lifting motor 3 along the lead screw and drives the lifting platform 5 to move quickly and smoothly to the calculated target height.

[0028] (b) Ultrasonic Repellent: During ascent and descent, or after reaching a designated height, the main controller controls the ultrasonic frequency generator to drive the ultrasonic transducer 7 to emit pre-determined, targeted high-frequency sound pulses to accurately and effectively repel birds. Table 1 shows the recommended ultrasonic frequencies and parameters for repelling major power grid pests.

[0029] Table 1. Recommended ultrasonic frequencies and parameters for repelling major pests in power grids. 5. Feedback and Standby: (a) The AI ​​recognition camera 11 continuously monitors the effect of driving away the birds. After confirming that the birds have flown away, the main controller controls the lifting platform 5 to return to the standby position to save energy and prepare for the next task.

[0030] (b) All operation logs and bird deterrence records can be uploaded to the background monitoring system via the wireless communication module (such as 4G) set on the main controller.

[0031] III. AI Model Processing The specific implementation process of using AI algorithms to identify birds is as follows: Figure 2 As shown, the algorithm includes the following steps: First, the main controller calls the AI ​​recognition camera 11 to acquire images and perform standardized preprocessing; then, the bird region in the image is located through a pre-trained object detection network; next, the cropped bird image is input into a pre-trained image classification network to identify the specific species; finally, the system queries the database based on the recognition results (bird species and altitude) to generate and execute the corresponding repelling strategy. The core logic of the algorithm is as follows.

[0032] 1. Overview of Identification Methods In the device described in the present application, the intelligent recognition of bird species is mainly realized through a target detection and image classification model based on deep learning. The core process includes three core steps: image preprocessing, bird target detection, feature extraction and classification. The system uses a hybrid model architecture that combines the YOLOv5 (You Only Look Once version 5) target detection algorithm and the EfficientNet-based image classification algorithm, which greatly improves the recognition accuracy of different bird species, especially small targets and occluded targets, while ensuring real-time recognition.

[0033] 2. The specific implementation process is as follows: a. Randomly take a frame of image from the bird images obtained in real time by the AI recognition camera 11 for scaling and normalization preprocessing; b. Use the trained YOLOv5 model for target detection to locate all birds in the image; Wherein each bird in the image is an element, and the information of the element is represented by [x, y, width, height, confidence, class_id], wherein x is the horizontal coordinate of the center point of the bounding box, y is the vertical coordinate of the center point of the bounding box, width is the width of the bounding box, height is the height of the bounding box, confidence is the confidence that the box contains the target and the category is correct, and class_id is the category number of the target, which is an integer starting from 0, corresponding to the serial number assigned to each category during training; c. Crop the bird image according to the detection box coordinates to obtain the cropped bird candidate region ROI; d. Input the cropped bird candidate region ROI into the trained EfficientNet classification network for fine recognition, and output the recognized bird species as a probability distribution of each specified category; e. Get the highest probability and its corresponding bird species ID; f. Decision: If the confidence is greater than the set threshold, the recognition is valid, the bird species ID is mapped to the bird species name, the bird height from the ground is estimated according to the detection box position, the drive function is called, and the most sensitive ultrasonic frequency corresponding to the bird species name and the drive motor revolutions corresponding to the bird height from the ground are obtained as the drive instructions and sent to the ultrasonic transducer 7 and the drive motor respectively; If the confidence is less than or equal to the set threshold, the recognition is invalid, and the image is saved for subsequent model optimization.

[0034] 3. Model training and optimization method To enable those skilled in the art to implement, the training method of the model needs to be briefly described: ① Data Preparation: Collect and build an image dataset containing major bird species (e.g., egrets, magpies, crows, etc.). Each bird in the images needs to be labeled by experts, including bounding boxes (Bounding Box) and species labels (Species Label).

[0035] ② Model Training: YOLOv5 Detection Model Training: Use the labeled bounding box information to train the YOLOv5s (small) model, enabling it to locate birds from complex backgrounds.

[0036] EfficientNet Classification Network Model Training: Store the dataset in folders according to bird species, and use transfer learning (Transfer Learning) techniques to fine-tune (Fine-tuning) the EfficientNet model pre-trained on a large image dataset (e.g., ImageNet), focusing on distinguishing different bird species.

[0037] ③ Model Optimization: To adapt to embedded devices, use tools such as TensorRT or OpenVINO to quantize (INT8 precision) and accelerate the trained model, ensuring accuracy while improving inference speed to meet real-time requirements (>15 FPS).

Claims

1. An active bird-proofing device for power transmission lines based on artificial intelligence and simulated falconry, characterized in that: The system includes a main controller, a fixed base (1), a lifting motor (3), a pole (4), a lifting platform (5), an ultrasonic transducer (7), a simulated falcon (8), a solar panel (9), and an AI recognition camera (11). The fixed base (1) is fixedly installed on the power transmission tower, and the main controller, solar panel (9), and AI recognition camera (11) are installed on the fixed base (1). The pole (4) is fixedly installed on the power transmission tower, and a lead screw is installed on the pole (4). The lifting motor (3) is connected to the lead screw through a coupling, and the lifting motor (3) moves along the lead screw. The pole moves up and down, and the lifting motor (3) is fixedly connected to the lifting platform (5); the lifting platform (5) is fixedly installed with a simulated falcon (8) and an ultrasonic transducer (7); the solar panel (9) is connected to the storage battery; the main controller is connected to the storage battery, the lifting motor (3), the ultrasonic transducer (7) and the AI ​​recognition camera (11) respectively. The main controller is used for artificial intelligence AI recognition of birds, for controlling the lifting and lowering movement of the lifting motor (3), for selecting the ultrasonic frequency of the ultrasonic transducer (7) and for transmitting data with the background monitoring system of the power system.

2. The active bird protection device for power transmission lines based on artificial intelligence and simulated falconry as described in claim 1, characterized in that: A fixed pulley (2) is fixedly installed on the upper end of the pole (4); a traction rope (10) is fitted on the fixed pulley (2); one end of the traction rope (10) is fixedly connected to the lifting platform (5), and the other end of the traction rope (10) is fixedly connected to a counterweight (6) to reduce the energy consumption of the lifting motor (3) and enhance the stability of the lifting platform (5).

3. The active bird protection device for power transmission lines based on artificial intelligence and simulated falconry as described in claim 1, characterized in that: The lifting platform (5) is also equipped with a position sensor to detect the real-time height of the lifting platform (5). The position sensor is connected to the main controller.

4. A method for active bird protection of power transmission lines based on artificial intelligence and simulated falconry, utilizing the active bird protection device for power transmission lines based on artificial intelligence and simulated falconry as described in claim 1, characterized in that: Includes the following steps, And the following steps are performed in sequence: Step 1: Continuous monitoring AI-recognized camera (11) performs real-time video monitoring of the area surrounding the tower; Step Two: Target Identification and Localization (1) When the AI-recognized camera (11) captures a bird, it transmits the captured bird image to the main controller; (2) The main controller analyzes bird images using a set AI algorithm, identifies the species of birds, and estimates the relative position of the birds to the tower and their height above the ground. Step 3: Intelligent Decision-Making and Precise Removal (1) Based on the estimated height of the bird, a lifting control command is generated and transmitted to the lifting motor (3). The lifting motor (3) rises and moves the simulated falcon (8) to the same horizontal height as the bird. (2) At the same time, the main controller queries the most sensitive ultrasonic frequency set in the built-in database of the main controller according to the identified bird species, and drives away the birds with corresponding ultrasonic waves through the ultrasonic transducer (7). Step 4: Feedback and Standby (1) The AI ​​recognition camera (11) continuously monitors the effect of driving away birds. If the AI ​​recognition camera (11) cannot detect birds, the main controller controls the lifting motor (3) to return to the standby position and prepare for the next task. (2) The main controller generates and stores operation logs and bird deterrence records, which are then uploaded to the background monitoring system via the wireless communication module set in the main controller.

5. The method for active bird protection of power transmission lines based on artificial intelligence and simulated falconry as described in claim 5, characterized in that: The specific methods for identifying bird species, estimating the relative positions of birds and towers, and their height above the ground in step two are as follows: a. Randomly select a frame from the bird images obtained in real time by the AI ​​recognition camera (11) and perform scaling and normalization preprocessing; b. Use the trained YOLOv5 model to perform object detection and locate all birds in the image. In this context, each bird in the image is represented as an element, and the element's information is represented by [x, y, width, height, confidence, class_id]. Here, x is the horizontal coordinate of the bounding box center point, y is the vertical coordinate of the bounding box center point, width is the width of the bounding box, height is the height of the bounding box, confidence is the confidence that the box contains the target and the class is correct, and class_id is the class number to which the target belongs, which is an integer starting from 0, corresponding to the index assigned to each class during training. c. Crop the bird image based on the detection box coordinates to obtain the cropped bird candidate region (ROI); d. Input the cropped bird candidate region (ROI) into the trained EfficientNet classification network for fine identification, and output the probability distribution of the identified birds for each set category; e. Obtain the highest probability and its corresponding bird species ID; f. Decision-making: If the confidence level is greater than the set threshold, the identification is valid. The bird species ID is mapped to the bird species name, the bird's height off the ground is estimated based on the detection box position, the drive-away function is called, and the most sensitive ultrasonic frequency corresponding to the bird species name and the number of drive motor revolutions corresponding to the bird's height off the ground are used as drive-away instructions and sent to the ultrasonic transducer (7) and the drive motor respectively. If the confidence level is less than or equal to the set threshold, the recognition is invalid, and the image is saved for subsequent model optimization.

6. The method for active bird protection of power transmission lines based on artificial intelligence and simulated falconry as described in claim 6, characterized in that: The trained YOLOv5 model was trained using an image dataset containing major pest birds and labeled bounding box information, enabling it to learn to locate birds from complex backgrounds. In this dataset, each bird in the image was labeled by experts, including bounding boxes and species labels.

7. The method for active bird protection of power transmission lines based on artificial intelligence and simulated falconry as described in claim 6, characterized in that: The trained EfficientNet classification network is created by storing the image dataset in folders according to bird species, and then using transfer learning techniques to retrain an EfficientNet model that has already been pre-trained on a large image dataset, making it focus on distinguishing different species of birds.

Images