Solar photovoltaic panel bird repelling vehicle

Abstract

According to the solar photovoltaic panel bird repelling vehicle, a vehicle body serves as a carrier, a power supply module, a communication module, a sound wave bird repelling module, a bird repelling execution module and a processor are integrated, all the modules work cooperatively, the power supply module is formed by combining a solar panel, an energy storage battery and a magnetic attraction charging port, and energy self-consistency and endurance are guaranteed; the communication module supports information transmission and emergency communication relay, and expands the cooperative capability of the unmanned aerial vehicle; the bird repelling execution module achieves automatic recognition of bird types, number, activity modes and behavior intentions through an AI visual camera and a deep learning model, multiple bird repelling means are formed in combination with a laser radar, a sound wave generator, a loudspeaker, an illuminating lamp and the like, efficient bird repelling can be achieved in a targeted mode by selecting a single or combined mode, the photovoltaic field area can be actively patrolled, and the bird repelling efficiency is improved. The problem of dead zones of fixed equipment is solved, the bird repelling precision and ecological friendliness are improved, meanwhile, the emergency operation and maintenance capacity of the power station is enhanced, and the method is suitable for bird prevention and control and related auxiliary scenes of various photovoltaic power stations.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of photovoltaic power station supporting equipment, and particularly relates to an intelligent and mobile bird driving device for a solar photovoltaic power station, which is particularly suitable for realizing precise, safe and efficient driving away of birds in the photovoltaic panel area and related operation auxiliary functions through the fusion of artificial intelligence identification, multi-modal bird driving means, new energy power supply and emergency communication cooperation technology. BACKGROUND

[0002] With the vigorous development of the solar photovoltaic power generation industry, large-scale photovoltaic power stations are continuously constructed and popularized worldwide. However, in the operation process of photovoltaic power stations, the negative effects of bird activities are increasingly prominent. The stay, habitat and reproduction of birds in the photovoltaic panel area not only pollute the surface of the photovoltaic panel due to excrement, resulting in a decrease in photoelectric conversion efficiency, but also may cause circuit failure due to bird nesting, and even cause physical damage in the process of flying.

[0003] Traditional bird driving methods, such as simple sound wave bird driving and physical barrier isolation, have significant defects: first, the bird driving method is single and easy to be adapted by birds, with poor long-term effect; second, there is a lack of precise identification and analysis of bird behavior, resulting in energy waste and ecological disturbance due to blind bird driving; third, due to environmental adaptability, it is difficult to work stably in complex terrain and extreme climate, and cannot achieve effective protection in the whole photovoltaic field area and all time periods. In addition, photovoltaic power stations are often built in remote areas, and conventional communication networks are easily interrupted by natural disasters, and lack efficient emergency cooperative inspection means, further increasing the difficulty of power station operation and maintenance. To solve the above problems, the present application designs a solar photovoltaic panel bird driving vehicle, which integrates intelligent identification, multi-element bird driving, emergency communication and unmanned aerial vehicle cooperation functions, adapts to the complex environment of photovoltaic power stations, realizes precise, efficient and intelligent bird driving, and at the same time meets the communication guarantee and inspection cooperation needs in emergency scenarios. SUMMARY

[0004] The present application aims to solve the limitations of existing photovoltaic power station bird driving methods, including single bird driving method leading to easy adaptation of birds, potential harm to birds, difficulty in achieving effective coverage in all areas and all weather conditions, and lack of intelligent decision basis. By integrating intelligent identification, multi-element cooperative bird driving, efficient energy supply and emergency expansion functions, an automatic solution is constructed that can not only precisely drive away birds to ensure the operation efficiency and safety of photovoltaic power stations, but also reduce ecological disturbance to birds, while improving the operation and maintenance capabilities of power stations in complex environments and emergency scenarios, ultimately realizing ecological balance of harmonious coexistence between man and nature and efficient and stable operation of photovoltaic power stations.

[0005] To achieve the above-mentioned purpose, the present application is realized by the following technical solutions:

[0006] The solar photovoltaic panel bird repelling vehicle comprises a vehicle body with a wheel suspension, a plurality of functional modules are arranged on the vehicle body, the functional modules comprise a power supply module, a communication module, a sound wave bird repelling module located at the upper end of the vehicle body, and a bird repelling execution module with bird behavior analysis and prediction functions, and a processor located in the vehicle body, and each module is electrically connected with the processor (central processor);

[0007] The power supply module comprises a battery, a monocrystalline silicon solar panel connected with the battery and a magnetic suction charging port, the monocrystalline silicon solar panel assists in cruising charging during the day, the magnetic suction charging port realizes fast charging of the vehicle body when returning to the base, the communication module comprises a communication antenna for information transmission, the sound wave bird repelling module comprises a sound wave generator, a loudspeaker and a lighting lamp, the sound wave generator executes sound wave bird repelling, the loudspeaker realizes sound externalization for bird repelling, and the lighting lamp is used for light effect bird repelling and night lighting;

[0008] The bird repelling execution module comprises an AI vision camera, a laser radar, a computing device, a storage device and a deep learning model electrically connected, the deep learning model comprises a convolutional neural network, a recurrent neural network and a variant long short-term memory network, the convolutional neural network is used for image feature extraction and bird species identification, the convolutional neural network can automatically extract key features of birds from images collected by the AI vision camera through the combination of convolutional layers, pooling layers and fully connected layers, the key features include feather color and body contour, and the bird species is judged, and the recurrent neural network and the variant long short-term memory network are combined to analyze bird activity patterns and behavior intentions; the AI vision camera is used for discovering birds and collecting bird images, and the storage device is used for storing collected bird image data, trained deep learning model parameters and analyzed bird behavior related data; the related data includes bird species, quantity, activity pattern and behavior intention;

[0009] By using the AI vision camera and the deep learning model, the bird species, quantity, activity pattern and behavior intention are automatically identified, precise bird repelling strategy selection and bird repelling timing judgment are realized, and finally one or more of sound waves, lasers and light effects are triggered to select for efficient bird repelling.

[0010] Preferably, the recurrent neural network is used for processing data with time sequence characteristics and is suitable for analyzing bird activity patterns, the activity patterns include flight trajectories, and the vehicle is positioned and driven for bird repelling; the variant long short-term memory network analyzes changes of bird gathering areas and behavior intentions, and the behavior intentions include a series of action sequences for preparing to land;

[0011] The computing device is a high-performance embedded computing unit carried by a vehicle body, used to run a deep learning model to process and analyze image data collected by a camera in real time; the computing device is an NVIDIA Jetson series module or an edge server.

[0012] Preferably, the data acquisition and labeling method using the bird repelling execution module:

[0013] Image acquisition: using AI vision cameras installed on the bird repelling vehicle, continuously collecting image data of bird activities in different time periods and under different weather conditions in the photovoltaic power station area; during the collection process, various common birds and different bird postures and scenes are covered;

[0014] Data labeling: manually labeling the collected images to mark the species, position (bounding box coordinates) of the birds in the images; for continuously collected image sequences, mark the position changes of the birds in different frames to form flight trajectory data; at the same time, according to the behavior characteristics of the birds, mark whether they are in a state of preparing to land; after labeling is completed, divide the data set into training set, validation set and test set;

[0015] Model training method using bird repelling execution module:

[0016] Model training of convolutional neural network: taking the training set images and their labeled bird species information as input, adjusting the model parameters through back propagation algorithm and stochastic gradient descent optimization algorithm to continuously improve the prediction accuracy of the model for bird species; during the training process, the validation set is used to evaluate the convolutional neural network model to prevent overfitting, and the model hyperparameters, including learning rate and number of convolution kernels, are adjusted according to the evaluation results; when the model achieves satisfactory accuracy on the test set, save the trained convolutional neural network model parameters;

[0017] Model training of recurrent neural network and variant long short-term memory network: for bird activity pattern and behavior intention analysis, the labeled bird position change data and behavior intention annotation information are arranged into data samples in time sequence; for flight trajectory data, the coordinate positions of each bird in different frames are used as input features; for behavior intention analysis, the posture and speed characteristics of the birds are combined to train the recurrent neural network and variant long short-term memory network model using these data samples; the model parameters are adjusted through back propagation and optimization algorithm, the model is evaluated and the hyperparameters are adjusted using the validation set, and finally the recurrent neural network and variant long short-term memory network model that can accurately analyze the bird activity pattern and behavior intention is obtained;

[0018] Real-time recognition and analysis method using bird repelling execution module:

[0019] Image acquisition: During the patrol process, the AI vision camera collects images in the photovoltaic power station area in real time.

[0020] Species identification: input the collected images into the trained convolutional neural network model, and the convolutional neural network model outputs the species and quantity information of the birds in the image.

[0021] Activity pattern and behavior intention analysis: for the continuously collected image sequence, the position information of the birds in each frame of image is extracted to form time series data, which is input into the trained recurrent neural network and long short-term memory network model to analyze the flight trajectory, gathering area change and behavior intention of the birds.

[0022] Use the precise bird repelling strategy selection and timing judgment method of the bird repelling execution module:

[0023] Strategy selection: according to the identified bird species, select the appropriate bird repelling method; if it is a bird sensitive to laser, prefer to choose laser bird repelling; if it is a group of birds, further combine sound waves and / or loudspeakers for sound and light driving; and adjust the patrol route of the bird repelling vehicle according to the gathering area of the birds in real time, focusing on driving away the birds gathering area.

[0024] Timing judgment: when the deep learning model analyzes that the birds have the behavior intention of preparing to land, start the bird repelling measures in advance, drive away the birds before they land on the photovoltaic panel area, improve the bird repelling efficiency, and reduce the interference of the birds to the photovoltaic power station.

[0025] Preferably, the model updating and optimization method: regularly collect new bird image data, retrain and update the existing deep learning model to adapt to the emergence of new bird species and the influence of environmental changes on bird behavior, continuously improve the recognition accuracy and analysis effect of the model.

[0026] Preferably, the laser radar includes a detection and control module, a signal processing module, and a laser emission and scanning module electrically connected, after the AI vision camera of the bird repelling execution module identifies and locates the bird target, the emitter of the high-precision low-power laser radar is activated to emit a specific wavelength and power safe laser beam to drive away the birds.

[0027] Preferably, the sound wave generator includes an induction and triggering module, a signal control and processing module, a sound wave signal generation module, and a sound wave emission module electrically connected, the AI vision camera detects that the birds enter the warning range to trigger the sound wave to interfere with the hearing of the birds, and the sound wave generator can work synchronously with the laser radar, the illuminating lamp and the loudspeaker.

[0028] Preferably, the communication antenna of the communication module transmits information to the remote monitoring center or handheld terminal in real time, including vehicle status, environmental perception, and bird repelling operation log. The communication antenna supports receiving remote control instructions, facilitating remote monitoring, management, and data analysis.

[0029] Preferably, the communication antenna includes a liftable communication mast, which constitutes a small emergency communication relay supporting 4G / 5G, Mesh network, LoRa, and intercom relay. When disasters cause conventional communication interruption, the communication antenna can provide a temporary mobile communication coverage node.

[0030] Preferably, the unmanned aerial vehicle ground control station is integrated in the vehicle body, and the communication link between the unmanned aerial vehicle and the control center or operator is enhanced through the vehicle-mounted communication antenna, realizing real-time control and data relay of the unmanned aerial vehicle, and providing more stable and longer distance unmanned aerial vehicle control and high-definition image and data backhaul.

[0031] Beneficial effects: precise bird repelling to ensure power generation efficiency: through AI recognition and deep learning, the bird species and behavior intention are accurately distinguished, and the bird repelling mode is selected accordingly to reduce invalid interference. Experimental data shows that the incidence of photovoltaic panel hot spot effect caused by bird retention can be reduced by more than 60%, and the power generation efficiency can be improved by 8%-12%.

[0032] Emergency coordination to improve operation and maintenance resilience: emergency communication relay ensures communication in disasters, and unmanned aerial vehicle cooperative inspection extends the monitoring range, enhances the operation and maintenance ability of power station in extreme conditions, and reduces the loss of equipment failure caused by communication interruption and inspection blind area.

[0033] Ecological friendliness and intelligent operation and maintenance: non-injurious laser, simulated enemy sound wave, and other bird repelling means reduce the disturbance to bird ecology; remote monitoring and data analysis function assist operation and maintenance personnel to dynamically optimize bird repelling strategy and reduce artificial inspection cost. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 is a schematic diagram of the overall structure of the present application.

[0035] Figure 2 is a rear view of part of the structure of the present application.

[0036] Figure 3 is a front view of part of the structure of the present application.

[0037] Figure 4 is a principle block diagram of the laser radar of the present application.

[0038] Figure 5 is a principle block diagram of the sound wave generator of the present application. DETAILED DESCRIPTION

[0039] The embodiments of the present application will be described below with reference to the accompanying drawings.Figures 1-5 The technical solutions in the embodiments of the present application are described clearly and completely. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0040] In the description of the application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "setting", "connecting" and the like should be understood broadly, for example, "connecting" can be fixedly connected, can be detachably connected, or integrally connected; can be mechanically connected, or indirectly connected through an intermediate medium. For those skilled in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0041] The vehicle body 1 in the present application preferably adopts a motor-driven four-wheel off-road vehicle. The configuration of the off-road vehicle is not a necessary technical solution of the present application, and therefore will not be described in detail. The vehicle body 1 in the present application can be configured with a GPS module and other commonly used modules.

[0042] Innovative points / principles of the technical solutions of the present application:

[0043] (I) Overall architecture

[0044] The solar photovoltaic panel bird repelling vehicle takes the vehicle body 1 as a carrier, integrates a power supply module, a communication module, a sound wave bird repelling module, a bird repelling execution module, and a built-in processor, and each module cooperates. Through AI visual perception and deep learning model, bird behavior analysis and prediction are realized; relying on multi-element bird repelling means (sound wave, laser, light effect) for precise intervention; with the aid of a liftable communication mast and a UAV ground control station, the emergency communication and inspection cooperation ability is expanded, and a "perception-analysis-decision-execution-cooperation" integrated intelligent bird repelling system is constructed.

[0045] (II) Core module description

[0046] 1. Bird repelling execution module

[0047] Hardware composition: composed of AI visual camera 5, laser radar 6, vehicle-mounted high-performance computing device (NVIDIA Jetson series module or edge server), storage device.

[0048] AI and deep learning models: Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Long Short-Term Memory Network (LSTM variant). CNN extracts image features such as bird feather color and body outline through convolution layers, reduces dimensionality through pooling layers, and classifies through fully connected layers to achieve accurate bird species recognition. RNN and LSTM variant process time series data of bird position changes, analyze flight trajectories and gathering areas, and predict behavior intentions (such as preparing to land).

[0049] Workflow:

[0050] Data collection and annotation: AI vision camera 5 collects images under multiple time periods and weather conditions in the photovoltaic field area, and manually annotates bird species, position (bounding box coordinates), and behavior state (flight / landing, etc.), to build a dataset containing training set, validation set, and test set.

[0051] Model training: Train CNN model with training set data, optimize learning rate and number of convolution kernels, and evaluate validation set to prevent overfitting. Train RNN / LSTM variant model with annotated time series data to learn the correlation between bird activity patterns and behavior intentions.

[0052] Real-time recognition and decision-making: During patrol, the camera takes real-time images, inputs the trained CNN model to output bird species and quantity, extracts consecutive frame bird position information, analyzes flight trajectories, gathering areas, and behavior intentions through RNN / LSTM variant model, and provides basis for bird repelling strategies.

[0053] 2. Multi-element bird repelling coordination module

[0054] Laser bird repelling: Laser radar 6 consists of detection and control module 61, signal processing module 62, and laser emission and scanning module 63. After AI vision camera 5 identifies and locates birds, it triggers laser radar to emit safe laser beams with specific wavelength and power, which stimulate and visually interfere with birds to avoid harm.

[0055] Sound and light coordination bird repelling: Sound wave generator 8 consists of induction and trigger module 81, signal control and processing module 82, sound wave signal generation module 83, and sound wave emission module 84, which can emit specific frequency sound waves such as simulated predator calls and ultrasonic waves; loudspeaker synchronously plays deterrent sound effects, and strobe light 7 flashes warning light. When birds enter the warning range, sound and light coordination triggers to interfere with birds' hearing and vision, enhancing the repelling effect and adapting to medium and close-range group birds.

[0056] Strategy coordination: Based on bird species (such as laser-sensitive birds prioritizing laser repelling), behavior intentions (intervening before preparing to land), and gathering scale (group birds superimposing sound waves / light), intelligently select single or combined bird repelling mode, dynamically adjust patrol route, and focus on bird gathering areas.

[0057] 3. Energy and communication guarantee module

[0058] Power supply module: monocrystalline silicon solar panel 3 assists charging for vehicle body cruising during the day, and stores electric energy to a large capacity battery; magnetic suction charging port 9 supports rapid energy replenishment at the base, realizing multi-element endurance of "solar energy + energy storage + magnetic suction charging", and adapting to the scene of remote power station without power grid.

[0059] Communication and emergency cooperation:

[0060] Normal communication: communication antenna 4 transmits vehicle state (power, equipment working condition), environmental perception (bird identification data), bird repelling log to the remote monitoring center in real time, and supports receiving control instructions (adjusting patrol route, bird repelling parameters).

[0061] Emergency communication relay: after the liftable communication mast is unfolded, it supports 4G / 5G, Mesh self-organizing network, LoRa low-power communication, and intercom relay, and provides temporary mobile coverage nodes when disaster causes normal communication interruption.

[0062] Unmanned aerial vehicle cooperation: the unmanned aerial vehicle ground control station is integrated in the vehicle, the link is enhanced by the vehicle-mounted communication antenna 4, remote control of the unmanned aerial vehicle and high-definition data backhaul are realized, the photovoltaic field inspection range is expanded (such as high-altitude monitoring of large-scale bird migration), and the bird repelling strategy optimization is assisted.

[0063] As shown in Figure 4 , the detection and control module 61: as the core control unit of the laser radar, on the one hand, receives the bird positioning data (including coordinates, distance, motion speed and other parameters) transmitted by the bird repelling execution module, and on the other hand, monitors the working state of the laser emission assembly (such as power output, temperature, light path stability) in real time. When abnormal working condition (such as power exceeding the standard, component overheating) is detected, the protection mechanism is triggered to cut off the laser emission, and at the same time, the fault information is fed back to the processor; in addition, the module can also dynamically adjust the laser emission power according to the distance of the birds (for example, output 5-10 mW power to the target within 10 meters, and increase to 10-20 mW to the target within 10-30 meters), to ensure effective repelling while avoiding energy waste.

[0064] Signal processing module 62: receives the bird contour features and motion trajectory data output by the AI vision camera 5, eliminates environmental interference (such as light and shadow changes, flying insects false touch) through filtering algorithm (such as Kalman filtering), extracts the real-time position coordinates of the birds and converts them into laser aiming parameters (azimuth angle, pitch angle). At the same time, the module can match the preset laser wavelength scheme (for example, 532 nm green laser for raptors, and 650 nm red laser for small birds) combined with the bird species information (provided by the bird repelling execution module), and transmit the processed control instructions to the laser emission and scanning module.

[0065] Laser emission and scanning module 63: contains a rotatable laser emitter and a beam shaping component, which adjusts the emission angle according to the instructions of the signal processing module, and realizes accurate aiming at the bird target. Its scanning range covers 120° horizontally and 60° vertically, with a response speed ≤0.3 seconds, and can track birds moving at a speed ≤10m / s; the diameter of the emitted laser beam is controlled within 5-10cm (adjustable dynamically with distance), and the visual interference effect is enhanced by intermittent flashing (frequency 3-5Hz), and the laser power is strictly limited to Class II safety level (≤5mW continuous output), ensuring no physical harm to birds and humans. When the AI vision camera (5) of the bird repelling execution module identifies and locates the bird target, the detection and control module 61 immediately activates the laser radar 6, and through the coordinated operation of the above three modules, a safe laser beam with specific wavelength and power is emitted to achieve directional repelling.

[0066] As shown in Figure 5 , the induction and triggering module 81: receives the bird intrusion signal of the AI vision camera 5 in real time, and when it detects that a bird enters the preset alert range (a circular area with a radius of 5-20 meters can be set by the processor), it immediately triggers the sound wave bird repelling process; at the same time, it integrates an environmental noise sensor, which automatically increases the sound wave output power (gain adjustment range 0-20dB) when the background noise ≥60dB, ensuring effective coverage; in addition, this module also has a false touch prevention function, which continuously identifies the bird's contour features (excluding interference such as fallen leaves and flying birds), with a trigger delay ≤0.5 seconds and a false trigger rate ≤1%.

[0067] Signal control and processing module 82: receives the bird species, quantity and aggregation density data output by the bird repelling execution module, and generates sound wave control parameters according to the preset strategy. For example, for a single small bird, a high-frequency warning sound (2-5kHz) is called; for a group of birds with more than 10, a composite sound wave mode (superimposed with the call of a natural enemy and a low-frequency oscillation wave) is started; at the same time, the sound wave directivity can be adjusted according to the distance of the birds (omnidirectional diffusion is used for close range, and a 30° fan-shaped beam is focused for long distance). This module can also control the stroboscopic frequency of the lighting lamp 7 (matched with the rhythm of the sound wave, 10-15Hz) and the volume of the loudspeaker (range 60-100dB) simultaneously, realizing sound and light cooperation.

[0068] Sound wave signal generation module 83: built-in audio database containing 50+ bird enemy call sounds (such as hawk scream, falcon cry), 20+ warning sounds (such as explosion, high-frequency pulse), can intelligently match according to bird species (for example, playing shrike call for sparrow, playing vulture call for crow); support custom sound wave combination (update through remote control center), and can simulate non-periodic sound wave sequence (avoid bird adaptation); the generated sound wave signal covers the full frequency band of 20Hz-20kHz, among which the high frequency band of 2-5kHz is used to interfere with bird hearing, and the low frequency band of 100-500Hz produces a slight vibration feeling to enhance the deterrent.

[0069] Sound wave emission module 84: composed of 4 distributed speakers (2 on each side of the vehicle body), covering a horizontal range of 180° and a vertical range of 30°, with an effective propagation distance of 20-50 meters; waterproof and dustproof design (IP65 protection level), suitable for working environment of-30℃ to 60℃; support dynamic adjustment of sound wave intensity (output according to bird distance classification), and have automatic attenuation function (when the bird leaves the warning range, gradually reduce to silence within 3 seconds). When the AI vision camera 5 detects that the bird enters the warning range, the sensing and triggering module 81 starts the sound wave generator 8, and outputs specific sound wave interference to the bird hearing through the cooperation of the above modules; at the same time, the signal control and processing module 82 synchronously activates the laser radar 6 (for long-distance targets), the lighting lamp 7 (at night or in low light environment) and the speaker (to enhance sound coverage), forming a multi-dimensional bird repelling synergy effect.

[0070] A specific embodiment (the parameter setting belongs to a preferred way, should not be understood as a technical limitation):

[0071] (I) Device deployment and initialization

[0072] Hardware installation: AI vision camera 5, laser radar 6 are installed on the top of the vehicle body, ensuring 360° monitoring coverage (360° can be realized by rotating disc); the liftable communication mast is stored in the vehicle body, and the unmanned aerial vehicle ground control station is integrated in the vehicle; monocrystalline silicon solar panel 3 is laid on the roof, and magnetic charging port 9 is reserved for base charging.

[0073] Model training and loading: collect historical bird image data in photovoltaic field area, train CNN, RNN / LSTM variant model after manual annotation, deploy the trained model parameters to the vehicle-mounted computing device, and initialize the data set and log space of the storage device.

[0074] (II) Daily bird repelling operation process

[0075] Patrol and perception: start the bird repelling vehicle, AI vision camera 5 collects images at a frequency of 10-15 frames / second, laser radar 6 scans synchronously, and real-time data transmission to the computing device.

[0076] Bird recognition and analysis: CNN model identifies bird species and quantity, outputs bounding box coordinates; RNN / LSTM variant model analyzes consecutive frame data, plots flight trajectory, and predicts gathering area and behavior intention (e.g., identifies bird wing deceleration and high descent, and determines "preparing to land").

[0077] Bird repelling strategy execution:

[0078] If it is a bird sensitive to laser (e.g., small songbirds), trigger the laser radar 6 to emit a 532nm safe laser beam for directional repelling, with a response delay ≤0.5 seconds;

[0079] If it is a group of birds (e.g., pigeons), simultaneously start the sound wave generator 8 (play eagle calls + 18kHz ultrasonic waves), the loudspeaker (play deterrent sound effects in a loop), and the lighting lamp 7 (10-15Hz stroboscopic warning light), with a sound and light bird repelling cooperation rate ≥95%;

[0080] Based on the analysis of the gathering area, automatically adjust the wheel suspension 2 and motor drive to optimize the patrol route (e.g., move towards the bird gathering area with a deviation ≤±1 meter).

[0081] Data transmission and strategy optimization: The communication antenna 4 uploads the vehicle status (low power warning triggered when the remaining power is 30%) and bird repelling log (bird repelling mode, bird escape rate) to the remote monitoring center every 5 minutes; the operation and maintenance personnel adjust the next day's patrol period and bird repelling parameters (e.g., modify the sound wave frequency and laser power) based on the data.

[0082] (Three) Emergency scene application

[0083] Emergency communication interruption: When conventional communication is interrupted due to disasters such as earthquakes and floods, trigger the liftable communication mast (raise time ≤3 minutes), start Mesh ad hoc network or LoRa mode, cover a radius ≥2 kilometers, and ensure on-site and external emergency communication.

[0084] Unmanned aerial vehicle cooperative patrol: After receiving the cooperative command from the remote control center, the vehicle-mounted ground control station is activated, the communication antenna 4 is enhanced to strengthen the link, and the unmanned aerial vehicle is controlled to take off (response delay ≤1 minute), and high-altitude images of the photovoltaic panel are returned (resolution ≥4K), which helps to identify large-area bird gathering or equipment damage, and expands the patrol range to areas that the bird repelling vehicle cannot reach (e.g., the deep part of the photovoltaic panel array).

[0085] (Four) Model iteration and equipment maintenance

[0086] Model self-updating: Collect new bird image data (≥1000 frames) every month, trigger the model retraining process, and optimize the recognition accuracy (target to improve to 98%); if a new bird species is identified (e.g., migratory birds passing by), manually annotate and supplement the training to adapt to ecological changes.

[0087] Hardware maintenance: check the cleanliness of the laser radar emitter head, the light transmittance of the camera protective glass every quarter; regularly clean the adsorption surface of the magnetic charging port to ensure the charging efficiency; monitor the charge and discharge cycle times of the battery pack, and replace it when the service life is less than 80% to ensure long-term stable operation of the equipment.

[0088] The application provides a full-scene and intelligent bird repelling and operation and maintenance solution for a photovoltaic power station through intelligent identification, multi-element bird repelling and emergency cooperation, balances ecological protection and power generation efficiency improvement, and has remarkable practical value and popularization significance.

[0089] Finally, it should be noted that the present application is not limited to the above embodiments, but can have many variations. All variations that can be directly derived or inferred from the disclosed content by those of ordinary skill in the art should be considered within the scope of the present application.

Claims

1. A solar photovoltaic bird deterrent vehicle, comprising a vehicle body (1) with wheel suspension (2), characterized in that, The vehicle body (1) is provided with multiple functional modules, including a power supply module, a communication module, a sound wave bird deterrence module located at the upper end of the vehicle body (1), and a bird deterrence execution module with bird behavior analysis and prediction functions, as well as a processor located inside the vehicle body (1). Each module is electrically connected to the processor. The power supply module includes a battery and a monocrystalline silicon solar panel (3) connected to the battery, and a magnetic charging port (9). The monocrystalline silicon solar panel (3) assists in charging during daytime cruise, and the magnetic charging port (9) enables the vehicle body (1) to quickly charge upon returning to the base. The communication module includes a communication antenna (4) for information transmission. The sound wave bird deterrent module includes a sound wave generator (8), a loudspeaker, and a lighting lamp (7). The sound wave generator (8) performs sound wave bird deterrent, the loudspeaker enables sound to be played out to deter birds, and the lighting lamp (7) is used for light-effect bird deterrent and nighttime illumination. The bird deterrence execution module includes an electrically connected AI vision camera (5), a lidar (6), a computing device, a storage device, and a deep learning model. The deep learning model includes a convolutional neural network, a recurrent neural network, and a variant long short-term memory network. The convolutional neural network is used for image feature extraction and bird species identification. Through the combination of convolutional layers, pooling layers, and fully connected layers, the convolutional neural network can automatically extract key features of birds from the images collected by the AI ​​vision camera (5). Key features include feather color and body shape outline, thereby determining the species of birds. The recurrent neural network and the variant long short-term memory network are combined to analyze the bird activity patterns and behavioral intentions. The AI ​​vision camera (5) is used to detect birds and collect bird images. The storage device is used to store the collected bird image data, the trained deep learning model parameters, and the analyzed bird behavior-related data. The related data includes the species, number, activity patterns, and behavioral intentions of the birds. By utilizing AI vision cameras (5) and deep learning models, the types, numbers, activity patterns, and behavioral intentions of birds are automatically identified, enabling precise selection of bird deterrence strategies and judgment of bird deterrence timing. Ultimately, one or more of sound waves, lasers, and light effects are selected for targeted bird deterrence.

2. The solar photovoltaic panel bird deterrent vehicle according to claim 1, characterized in that, The recurrent neural network is used to process data with time-series characteristics, which is suitable for analyzing bird activity patterns, including flight trajectories, to facilitate bird deterrence vehicles in locating and driving birds away; the variant long short-term memory network analyzes changes in bird gathering areas and behavioral intentions, including a series of actions in preparation for landing. The computing device is a high-performance embedded computing unit mounted on the vehicle body (1), used to run deep learning models and process and analyze image data collected by the camera in real time; the computing device is an NVIDIA Jetson series module or an edge server.

3. A solar photovoltaic panel bird-repelling vehicle according to claim 1 or 2, characterized in that, Data collection and annotation methods using bird deterrence modules: Image acquisition: Using an AI vision camera (5) installed on a bird deterrent vehicle, image data of bird activities are continuously collected in the photovoltaic power station area at different times and under different weather conditions. During the acquisition process, various common birds and different bird postures and scenes are included. Data annotation: The collected images are manually annotated to mark the species and location of birds in the images; for continuously collected image sequences, the position changes of birds in different frames are annotated to form flight trajectory data; at the same time, based on the behavioral characteristics of birds, it is marked whether they are in the state of preparing to land. After the annotation is completed, the dataset is divided into training set, validation set and test set. Model training method using bird deterrence execution module: Convolutional Neural Network (CNN) Model Training: Using training set images and their labeled bird species information as input, the model parameters are continuously adjusted through backpropagation and stochastic gradient descent optimization algorithms to continuously improve the model's prediction accuracy for bird species. During training, a validation set is used to evaluate the CNN model to prevent overfitting. Based on the evaluation results, the model's hyperparameters, including the learning rate and the number of convolutional kernels, are adjusted. Once the model achieves satisfactory accuracy on the test set, the trained CNN model parameters are saved. Training Recurrent Neural Networks (RNNs) and Variant Long Short-Term Memory Networks (LSTMNs): For bird activity pattern and behavioral intention analysis, labeled bird location change data and behavioral intention annotations are organized into data samples according to time series. For flight trajectory data, the coordinates of each bird in different frames are used as input features. For behavioral intention analysis, bird posture and speed characteristics are combined. These data samples are used to train RNN and Variant LSTMN models. Backpropagation and optimization algorithms are used to adjust model parameters, and a validation set is used for model evaluation and hyperparameter tuning. Finally, a RNN and Variant LSTMN model capable of accurately analyzing bird activity patterns and behavioral intentions is obtained. Real-time identification and analysis methods using bird deterrence execution modules: Image acquisition: During the patrol, the AI ​​vision camera (5) of the bird deterrent vehicle collects images of the photovoltaic power station area in real time; Species recognition: The acquired images are input into a trained convolutional neural network model, which outputs information on the species and number of birds in the images. Activity pattern and behavioral intent analysis: For continuously acquired image sequences, the location information of birds in each frame is extracted to form time series data, which is then input into a trained recurrent neural network and variant long short-term memory network model to analyze the birds' flight trajectory, changes in gathering areas, and whether they are preparing to land. Methods for precise bird deterrence strategy selection and timing determination using bird deterrence execution modules: Strategy Selection: Select the appropriate bird deterrence method based on the identified bird species; if the birds are determined to be more sensitive to lasers, laser bird deterrence will be given priority; if the birds are in groups, sound waves and / or loudspeakers will be used to drive them away with sound and light; and adjust the patrol route of the bird deterrence vehicle in real time according to the area where the birds gather, focusing on driving away the birds from the areas where they gather. Timing judgment: When the deep learning model analyzes that birds are preparing to land, bird deterrence measures are initiated in advance to drive them away before they land on the photovoltaic panel area, thereby improving bird deterrence efficiency and reducing bird interference to the photovoltaic power station.

4. A solar photovoltaic bird-repelling vehicle according to claim 3, characterized in that, Model update and optimization methods: Regularly collect new bird image data, retrain and update existing deep learning models to adapt to newly emerging bird species and the impact of environmental changes on bird behavior, and continuously improve the model's recognition accuracy and analysis effect.

5. A solar photovoltaic bird-repelling vehicle according to claim 1, characterized in that, The lidar (6) includes an electrically connected detection and control module (61), a signal processing module (62), and a laser emission and scanning module (63). After the AI ​​vision camera (5) of the bird deterrence execution module identifies and locates the bird target, it activates the transmitter of the high-precision low-power lidar (6) and emits a safe laser beam of a specific wavelength and power to drive away the birds.

6. The solar photovoltaic panel bird deterrent vehicle according to claim 5, characterized in that, The sound wave generator (8) includes an electrically connected sensing and triggering module (81), a signal control and processing module (82), a sound wave signal generation module (83), and a sound wave emission module (84). When the AI ​​vision camera (5) detects that birds have entered the warning range, it triggers sound waves to interfere with the birds' hearing. The sound wave generator (8) can work synchronously with the lidar (6), the lighting lamp (7), and the speaker.

7. The solar photovoltaic bird-repelling vehicle according to claim 1, characterized in that, The communication antenna (4) of the communication module transmits information to the remote monitoring center or handheld terminal in real time. The information includes vehicle status, environmental perception, and bird control operation log. The communication antenna (4) supports receiving remote control commands, which facilitates remote monitoring, management and data analysis.

8. The solar photovoltaic bird deterrent vehicle according to claim 7, characterized in that, The communication antenna (4) includes a liftable communication mast. The communication antenna (4) constitutes a small emergency communication relay, supporting 4G / 5G, Mesh network, LoRa, and walkie-talkie relay. When a disaster causes a disruption in regular communication, it can provide a temporary mobile communication coverage node.

9. The solar photovoltaic bird-repelling vehicle according to claim 8, characterized in that, The vehicle body (1) integrates a UAV ground control station, which enhances the communication link between the UAV and the control center or operator through the vehicle-mounted communication antenna (4), realizes real-time control and data relay of the UAV, and provides more stable and longer-distance UAV control and high-definition image and data transmission.

Images